Historical Semantics - Historical Word-Formation 9783110850178, 9783110104677

Table of contents :
Preface
List of conference participants
The ordering of nominal compounds in Irish
The semantic development of will
Semantic shifts in Korean honoriflcation
Prefixal negation of English adjectives: psycholinguistic dimensions of productivity
Cross-linguistic comparison and the development of grammatical meaning
Derivation, inflection, and semantic change in the development of the Chukchi verb paradigm
Suppletion in word-formation
Etymology and the lexical semantics of the Old English preverb be-
Cognitive restrictions on the structure of semantic change
On the origin of the suffix -ly
Gender systems and semanticity: two case histories from Bantu
Segmental phonology and word-formation: agency and abstraction in the history of Irish
Deverbal nouns in Old and Modern English: from stem-formation to word-formation
Where do concessives come from? On the development of concessive connectives
The influence of semantic fields on semantic change
On semantic change in a dynamic model of language
Contact-induced semantic change and innovation
Inferential features in historical semantics
Diachronic irreversibility in word-formation and semantics
Semantic development of borrowings
Etymology and semantics. Theoretical considerations apropos of an analysis of the etymological problem of Spanish mañero, mañeria
Lexical and syntactic semantics in historical aspect
The synchrony-diachrony division in word-formation
A historical perspective on the productivity of the suffixes -ness and -ity
Word-formation, learned vocabulary and linguistic maturation
LaƷamon’s compound nouns and their morphology
On the functional development of repetition in Antiguan Creole morphology, syntax, and discourse
Modality and semantic change
The lexicological analysis of older stages of languages
‘Left’ or ‘right’?
Index of authors
Index of languages

Citation preview

Historical Semantics Historical Word-Formation

Trends in Linguistics Studies and Monographs 29

Editor

Werner Winter

Mouton Publishers Berlin · New York · Amsterdam

Historical Semantics Historical Word-Formation edited by

Jacek Fisiak

Mouton Publishers Berlin · New York · Amsterdam

Professor Jacek Fisiak Institute of English Adam Mickiewicz University Marchlewskiego 124/126 PL-61-874 Poznan Poland

Library of Congress Cataloging in Publication Data Main entry under title: Historical semantics • Historical word-formation. (Trends in linguistics. Studies and monographs; 29) Includes indexes. 1. Lexicology — Addresses, essays, lectures. 2. Historical linguistics — Addresses, essays, lectures. I. Fisiak, Jacek. II. Series. Ρ326. Η55 1985 412 85-18764 ISBN 0-89925-115-3

CIP-Kurztitelaufnähme der Deutschen Bibliothek Historical semantics - Historical word-formation / ed. by Jacek Fisiak. — Berlin ; New York ; Amsterdam : Mouton, 1985. (Trends in linguistics : Studies and monographs ; 29) ISBN 3-11-010467-9 NE: Fisiak, Jacek [Hrsg.]; Trends in Linguistics / Studies and monographs 29

Printed on acid free paper © Copyright 1985 by Walter de Gruyter & Co., Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form — by photoprint, microfilm, or any other means — nor transmitted nor translated into a machine language without written permission from Mouton Publishers, Division of Walter de Gruyter & Co., Berlin. Typesetting: Arthur Collignon, Berlin. — Printing: Druckerei Gerike, Berlin. — Binding: Lüderitz & Bauer Buchgewerbe GmbH, Berlin. Printed in Germany

Preface

The present volume contains a selection of papers prepared for the International Conference on Historical Semantics and Historical Word-Formation held at Blazejewko near PoznaA from March 28 to 31, 1984. The papers discuss numerous problems in the vast areas of historical semantics and historical word-formation. Some of them are more central for the field, other could be considered more peripheral, perhaps more important for the analysis of particular languages than as contributions to the formation of general theories. The material contained in the volume and the discussion at the conference seem to reflect adequately the present state of the disciplines in question, i.e. a lot remains still to be done both as regards a general semantic theory as well as in historical semantics, and the area of historical word-formation requires more investigations in particular languages before some crucial issues could be explained more definitely. The theoretical bias of the conference contributions was not restricted in any way and representatives of various linguistic persuasions have expressed their ideas. The languages analyzed include Korean, Bantu, and Chukchi next to Indo-European ones. It is hoped that the conference and the present volume have contributed in some sense at least to the development of historical semantics and word-formation, if not by solving any issues, at least by posing numerous questions which have to be answered if any progress in the field is to be made. Finally it is our duty to express our sincere thanks to all conference members for their active participation and in particular to all those whose contributions have been included in this volume. Special words of thanks go to the Vice Rector of Adam Mickiewicz University, Professor Stefan Paszyc, for providing the necessary funding. Last but not least thanks are due to the Conference Secretary, Ms. Barbara Ptocinska whose devotion to the cause and

VI

Preface

skillful handling of administrative matters during and long before the conference has contributed to its success. Poznaii, August 1984

Jacek Fisiak

Table of contents

Preface

V

List of conference participants

XI

Anders Ahlqvist The ordering of nominal compounds in Irish

1

Karin Aijmer The semantic development of will

11

Eung-Jin Baek Semantic shifts in Korean honorification

23

Philip Baldi, Victor Broderick and David S. Palermo Prefixal negation of English adjectives: psycholinguistic dimensions of productivity

33

Joan L. Bybee and William Pagliuca Cross-linguistic comparison and grammatical meaning

59

the

development

of

Bernard Comrie Derivation, inflection, and semantic change in the development of the Chukchi verb paradigm

85

Wolfgang U. Dressier Suppletion in word-formation

97

Thomas Fraser Etymology and the lexical semantics of the Old English preverb be113 Dirk Geeraerst Cognitive restrictions on the structure of semantic change . 127 Claude Guimier On the origin of the suffix -ly

155

VIII

Contents

Robert Κ. Herbert Gender systems and semanticity: two case histories from Bantu 171 Raymond Hickey Segmental phonology and word-formation: agency and abstraction in the history of Irish 199 Dieter Kastovsky Deverbal nouns in Old and Modern English: from stemformation to word-formation 221 Ekkehard König Where do concessives come from? On the development of concessive connectives 263 Adrienne Lehrer The influence of semantic fields on semantic change

283

Barbara Lewandowska-Tomaszczyk On semantic change in a dynamic model of language

297

Charles N. Li Contact-induced semantic change and innovation

325

Leonhard Lipka Inferential features in historical semantics

339

Helmut Lüdtke Diachronic irreversibility in word-formation and semantics 355 Witold Manczak Semantic development of borrowings

367

Francisco Marcos-Marin Etymology and semantics. Theoretical considerations apropos of an analysis of the etymological problem of Spanish manero, maneria 377 Anatolij M. Mukhin Lexical and syntactic semantics in historical aspect

397

Herbert Pilch The synchrony-diachrony division in word-formation

407

Elizabeth M. Riddle A historical perspective on the productivity of the suffixes -ness and -ity 435

Contents

IX

William C. Ritchie Word-formation, learned vocabulary and linguistic maturation 463 Hans Sauer La3amon's compound nouns and their morphology

483

Susan Shepherd On the functional development of repetition in Antiguan Creole morphology, syntax, and discourse 533 Carmen Silva-Corvalan Modality and semantic change

547

Jürgen Strauss The lexicological analysis of older stages of languages . . . . 573 Werner Winter 'Left' or 'right'?

583

Index of authors

597

Index of languages

605

List of participants at the International Conference on Historical Semantics and Historical Word-Formation held at Blazejewko, Poland, March 28 — 31, 1984

Director Professor Jacek Fisiak

Adam Mickiewicz University, Poznan

Conference Secretary Ms. Barbara Plocmska, M. A.

Participants Dr. Anders Ahlqvist

Adam Mickiewicz University, Poznah

National University of Ireland, Galway Docent Karin Aijmer University of Lund Professor Frank Anshen State University of New York, Stony Brook Docent Wiestaw Awedyk Adam Mickiewicz University, Poznah Dr. Eung-Jin Baek University of Toronto Professor Philip Baldi Pennsylvania State University Professor Gero Bauer University of Vienna Professor Peter Bierbaumer University of Graz Professor Joan L. Bybee State University of New York, Buffalo Professor Bernard Comrie University of Southern California, Los Angeles Professor Jan Cygan University of Wroclaw Dr. Barbara Dancygier University of Warsaw Docent Bernhard Diensberg University of Bonn Professor Klaus Dietz Free University, West Berlin Mrs. Katarzyna Dziubalska- Adam Mickiewicz University, Poznah Kolaczyk Dr. Thomas Fraser University of Lille III

XII

List of participants

Professor Udo Fries Dr. Dirk Geeraerts Mrs. Marinel Gerritsen Professor Johannes Gerritsen Dr. Heinz. J. Giegerich Dr. Antonina Grybosiowa Mr. Claude Guimier Docent Edmund Gussmann Mr. Camiel Hamans Professor Robert K. Herbert Dr. Raymond Hickey Professor Ernst H. Jahr Docent Roman Kalisz Professor Dieter Kastovsky Docent Göran Kjellmer Docent Veronika Kniezsa Dr. Roman Kopytko Dr. Günter Kotzor Mr. Jerzy Krzyszpien Dr. Barbara LewandowskaTomaszczyk Professor Charles N. Li Professor Leonhard Lipka Professor Magnus Ljung Ms. Ingegard Lohmander Professor Helmut Lüdtke Professor Witold Mariczak Professor Francisco Marcos-Marin Professor Manfred Markus Mr. Arthur Mettinger, M. A. Dr. Michael Miller Professor Anatolij M. Mukhin Professor Ruta Nagucka Mrs. Barbara Nykiel-Herbert

University of University of Royal Dutch University of

Zürich Leiden Academy, Amsterdam Groningen

University of Edinburgh University of Silesia, Katowice University of Caen Catholic University of Lublin University of Leiden Adam Mickiewicz University, Poznah University of Bonn University of Troms0 University of Gdahsk University of Vienna University of Göteborg University of Budapest Adam Mickiewicz University, Ρ ο ζ ^ ή University of Munich Jagiellonian University of Cracow University of Lodz University of California, Santa Barbara University of Munich University of Stockholm University of Göteborg University of Kiel Jagiellonian University of Cracow University of Madrid University of Innsbruck University of Vienna Jagiellonian University of Cracow Academy of Sciences of the USSR, Leningrad Jagiellonian University of Cracow Poznan

List of participants

Professor William Pagliuca Professor Herbert Pilch Dr. Michal Post Professor Elizabeth Riddle Professor William C. Ritchie Docent Jerzy Rubach Professor Kari Sajavaara Dr. Hans Sauer Professor Charles T. Scott Dr. Susan C. Shepherd Professor Carmen SilvaCorvalän Professor Jürgen Strauss Docent Aleksander Szwedek Dr. Bogdan Szymanek Ms. Malgorzata Teclaw, M. A. Mr. Jerzy Tomaszczyk, M. A. Dr. Alicja Wegner Dr. Jerzy Welna Professor Werner Winter Mr. Adam Wojcicki Mrs. Anna Zbirska-Sawala, M. A.

XIII

State University of New York, Buffalo University of Freiburg, Freiburg i. Br. University of Wroclaw Ball State University, Muncie, Ind. Syracuse University University of Warsaw University of Jyväskylä University of Munich University of Wisconsin, Madison Free University, West Berlin University of Southern California, Los Angeles University of Trier Pedagogical University, Bydgoszcz Catholic University of Lublin University of Gdansk University of Lodz Adam Mickiewicz University, PoznaA University of Warsaw University of Kiel University of Warsaw Adam Mickiewicz University, Poznah

ANDERS AHLQVIST

The ordering of nominal compounds in Irish

In a contribution to another series of conferences devoted to historical linguistics, I have (1980) discussed what I believe to be a fundamental change in the early history of Irish. This is the shift from a situation in which the verb had its unmarked position at the end of the sentence to one where it moved to the very beginning thereof. Since the former pattern is attested in the oldest literature, dating from ca. 700 AD., it seems legitimate1 to assume that the change-over took place over a period of time during the couple of centuries before that. Among other features changing at the same time, we may note that genitives in Modern Irish always and in Classical Old Irish mostly come after the noun they qualify, whereas in the oldest material, there are quite a few examples2 of preposed genitives, a feature that is of course consistent with the occurence of verbs in sentence final position (Greenberg 1963: 62). Regarding adjectives, on the other hand, we may note that they usually come after the noun, but that some fairly common ones normally precede, forming a sort of a compound, as indicated by the spelling of 1) deagh-shompla 'good example' (de Bhaldraithe 1953: 241). If one uses another adjective with an identical meaning, the order changes to la) sompla maith 'good example'. As I have noted elsewhere (1977: 270) this feature has parallels in other languages. Turning now to compounds, in which both parts are nouns, we find that the order of elements is that of determinans + determinatum: starting with Old Irish, we find: 2) ban-liaig 'woman physician' {DIL Β, 70.55-6), which is echoed by the modern form 2a) ban-dochtiiir 'woman doctor' {EID, 198 col. 1.12), for which the same source (col. 1.11) supplies the alternative form 2b) dochtuir mnä 'woman doctor'. 3

2

Anders

Ahlqvist

There is an obvious parallelism here, between 1 and la on one hand and 2a and 2b on the other, both as regards the ordering of determinans and determinatum and in respect of the status of the two constructions in spoken Modern Irish. The number of adjectives that may be used in constructions like that of 1 is finite,4 whereas such is not the case of the type exemplified in la. The pattern found in 2a may by now be considered bookish: 2b is now normal in spoken Modern Irish. For the sake of convenience, I shall henceforth refer to these two types as hysterocentric compounds and proterocentric noun phrases, respectively. Although given as part of the language in standard grammars of Modern Irish, 5 hysterocentric compounds are no longer productive. De Bhaldraithe 6 has this to say about the matter, in introducing exemples of various ways of forming compounds: "it is not a living manner to form compounds in any of the following ways, except in the case of putting two adjectives together," adding in a footnote (2541) that the "living manner that exists in the case of nouns is to form a phrasal noun, and for the second noun to be in the genitive, e. g., bean tighe" 'housewife'. 7 The obvious implication of all this is that hysterocentric compounds (with the order determinans -I- determinatum) are a category that is on the way out, something that would appear to be consistent (Miller 1975: 32) with the VSO type of language that Irish is now and has been for more than a thousand years. However, some questions remain open, mainly of chronology, in that it is difficult to fix the time when hysterocentric compounds ceased to be productive. Here I can add nothing to de Bhaldraithe's statement, except to point out that proterocentric NPs have been in existence for some time.8 Also, I have from Professor William Gillies of Edinburgh the following piece of information about their status in Scottish Gaelic: "The construction is relatively unusual in Scottish Gaelic, but fairly widespread as to the sort of sources it appears in. It tends to be limited to set phrases, which may suggest it once had greater currency than now." In origin, it seems to me very likely that this construction is derived from the genitive of apposition of the earlier language. As Thurneysen states in the Grammar of Old Irish (1946: 158): "certain uses of the qualificatory genitive are more common in Irish than in other languages."

The ordering of nominal compounds

in Irish

3

At this point, it may be useful to review briefly what is known about the formation of compounds in Old Irish. The standard source information is of course Thurneysen, who states: "The capacity to form compounds of various kinds with substantival stems survives in Old Irish on much the same scale as in Greek and Germanic. Even dvandva-compounds 9 are not unknown; cp. sall-c[h]arna 'bacon and fresh meat', üacht-gorta 'cold and hunger'. In verse, as might be expected, compounds are formed more freely than in prose. For the lenition of the initial of the second element, even after stems which originally ended in a consonant, see § 235 E. A few nouns become neuter io-stems in composition; e. g. finguine 'kin-murder' (guin), leth-gille 'half-pledge' (gell). Feminine personal nouns may be formed from the corresponding masculine nouns be prefixing ban-, composition form of ben 'woman'; e.g. ban-näm(a)e 'female enemy', ban-dälem 'female cup-bearer, spencer', ban-dea 'goddess'; similarly ban-chii 'bitch'." Most of the not very numerous examples he gives are ones with ban-, as in the material presented above. About their morphology, he notes (Thurneysen 1946: § 235 E): "The second element of a compound is lenited: When the first is noun, adjective, or numeral, even if it is a consonantal stem; e. g. rig-suide 'royal seat' (stem rig-; teglach 'household' from teg- fastem) and slog 'troop' with silent s. In such cases a composition vowel had been early inserted; cp. Gaul. Rig-o-magus 'royal field', Cinget-o-rix 'king of heroes'. Traces of the older method of composition are still furnished by numerals ending in a nasal: ηόη-bur, in spirto secht n-delbichsin 'of that septiform spirit' (and the placename Noindruimm), beside the later formations deichthriub 'ten tribes', noidecde 'cycle of nineteen years'." This gives an outline of the facts, but does not supply as many examples as do Zeuss (1971: 854-856) and Pedersen (1913:2-5); some further information is given by Vendryes (1908: 162) and Dottin (1913: 104-105) for the Old and Middle Irish periods, respectively. However, with one exception, none of these mention any alternatives to the internal ordering of elements in compounds, which in all examples cited seems to be determinans +

4

Anders

Ahlqvist

determinatum. 10 The exception is Zeuss, who mentions (1871: 854) that: "Praemittitur item, quod postponitur in aliis linguis, subst. dun (castrum, oppidum) in nominibus urbium vel monitum munitorum: 'ecclesia in cacumine montis, qui appellatur Dundeuenel. in cacumine montis Dunbreten ... quinta vero ecclesia Dunedene, quod anglica lingua dicta Edenburg ...",11 adding a footnote to mention the expression fius seel 'tidings', literally 'knowledge of stories', as follows: "Prorsus seiungenda est formula fius seel, fis seel (nuntius), ..." This is by no means the only fixed noun phrase with the second part in the genitive, but it is interesting to note that Zeuss saw fit to deal with it in the context of compounds, just like de Bhaldraithe and Sjcestedt-Jonval (see footnotes 6 and 7) had done in respect of the modern language. Semantically, the connection between the two types, i. e., hysterocentric compounds and proterocentric noun phrases is obvious, but formally, one might object that levels are getting mixed here, in that the former belongs to that part of morphology which deals with word-formation and the second to syntax. However, it seems to me that we may follow Benveniste in this instance (1967: 30): "C'est que l'impulsion qui a produit les composes n'est pas venue de la morphologie, ou aucune necessite ne les appelait; eile est issue des constructions syntaxiques avec leurs varietes de predication. C'est le modele syntaxique qui cree la possibilite du compose morphologique et qui le produit par transformation. La proposition, en ses differents types, emerge ainsi dans la zone nominale." In any case, the most important question to arise out of the Irish material mentioned in this paper is that of the typological status of the two constructions. As already mentioned, the order determinans + determinatum does not seem consistent with a VSO type of language, whereas the opposite one of course does. Thus, language universale typology would appear to cover the disappearance of hysterocentric compounds reasonably well, except for one matter of particular interest to historical linguists, which is the very considerable time lag involved. As we have seen, VSO became became the unmarked order of Irish sentences not later than 700 AD., whereas hysterocentric compounds are only now

The ordering of nominal compounds

in Irish

5

giving way to the forms to be expected from a typological point of view. At first sight, this may seem difficult to account for, but parallels appear to exist. Givon (1971: 401-405) has hinted at a similar lag in Germanic. On the other hand, Lehmann (who has written much on languages typology) prefers (1969: 2) not to see in compounds survivals of earlier syntactic patterns. Dealing with the work of Hermann Jacobi (1897), he states: "By Jacobi's point of view, compounds would be relics of earlier morphological categories and syntactic patterns. Jacobi's view was not generally accepted even in his day. Some compounds, to be sure, like members of any linguistic class, may reflect earlier patterns; yet from analysis of languages spoken today we have no grounds for assuming that compounds are particularly archaic. Rather, the analyses made for contemporary languages suggest that most compounds of a given language are best interpreted by comparison with other syntactic patterns contemporaneous with them." However, I find it impossible to relate the occurence in Modern Irish of hysterocentric nominal compounds to any contemporaneous syntactic pattern other than that of the closed class of prenominal adjectives (example 1 above). Furthermore, this too relates better to a SOV than to a VSO pattern, so that Irish would seem to be a language supplying a fairly likely counter-example to Lehmann's general statement, linking compounds primarily with contemporaneous syntactic patterns. Rather, we may here see yet another manifestation of the well-known important principle, according to which "today's morphology is yesterday's syntax" (Givon 1971:413). Also, it seems to me relevant to observe that the change of ordering that we have observed here is one that seems to have a parallel in modern Western European languages. As Fehling (1980: 376-377) has observed very perspicaciously, simple juxtaposition, in set noun phrases, of determinatum and determinans, is at present gaining ground in these languages, in a way rather reminiscent of the Semitic status constructus. 12 But whether this may be related to the fact that for instance French is now proceeding towards being a verb-initial language (Harris 1978: 7) is a matter best left to specialists in those languages. At this stage, the main points of this brief paper having been covered, an even shorter recapitulation may be useful. In Irish,

6

Anders Ahlqvist

natural speech nowadays seems to shun hysterocentric compounds in favour of proterocentric noun phrases. Although a considerable time lag seems to be involved, this appears to be related to the word order change (SOV-VSO) that may be observed in archaic Irish material. As the present title indicates, I have mainly dealt with compounds and phrases involving noun + noun. However, certain cases (other than those exemplified by 1 and la) of compounds involving noun + adjective may have a bearing 13 on the matter. Thus, Schmidt has made a useful collection 14 of a type he has described thus: das "umgekehrte Bahuvrihi". As he points out (1953: 80), this happens sporadically in English, where we may find arm-strong and strongarmed. A typical example from Irish is gob-reamhar 'fat-beaked', where gob means 'beak' and reamhar 'fat'. However, there are semantic restrictions 15 on the occurence of this type as well as doubts 15 whether it can really be regarded as an "Umkehrung" of the usual type of bahuvrihi. Also, I am personally not sure that these compounds may be described as being proterocentric in quite as clear-cut a way as phrases like 2b, so that I should, at any rate for the moment, prefer not to assume that they need be part of the word order related processes described in the main part of my paper. Finally, a purely practical point may be raised. In Modern Irish, hysterocentric compounds are often coined by people making up neologisms. In the spoken language, on the other hand, they are slow to gain acceptance. If this is, as suggested above, in any way related to typological trends inherent in the language perhaps language planners might do well to take note of the findings of typological linguistics before formulating their policies, so as to give these a better chance of being accepted by actual speakers of the languages involved.

Notes 1. To me, at any rate; however, cf. Wagner's rather different and by no means negligible conclusions (1967: 312-314). 2. Wagner (1967: 296-297) gives a useful list; I have commented on it elsewhere (1977: 271-272). 3. The declension of Old Irish ben, Modern Irish bean 'woman' is not regular (see my notes of 1980 a). For our purposes, it may be notes that ban- is the

The ordering of nominal compounds in Irish

4. 5. 6. 7.

8. 9.

10.

11.

12.

13.

14. 15.

1

composition form, whereas the genitive singular is mnä and the corresponding plural is ban, as for instance in dochtuir ban 'gynaecologist' (FGB, 97 col. 2.54). See de Bhaldraithe (1953: 240-244) for a list of those that occur in the Cois Fhairrge dialect of Connemara. See for instance the Christian Brothers (1960: 303); this is probably the most comprehensive of modern reference grammars. 1953: 254. I myself am responsible for these translations from Modern Irish. Literally, this means 'woman of house'. Similar information is supplied by Sjoestedt-Jonval (1938: 50, cf. 18 and 107 — 108), who states that compounds are "de caractere livresque et artificiel, la langue parlee preferant ou emprunter le mot anglais, ou le traduire par un groupe substantif + genitif". Thus one finds examples closely like 2b at least as early as the Middle Irish period, as in dalta mnä 'foster-daughter' ( D I L Β, 68.43-44). (1946: §254). This does not enlarge much on his earlier treatment (1909: 161). It must be noted that dvandva-compounds are in fact rather more numerous than this statement would seem to allow; cf. Binchy (1972: 38-41) and note (as Liam Breatnach informs me) that some of these may be classified as "merismatic" (Huld 1983) ones. However, as the internal ordering in dvandvas is relatively unimportant, they need not concern us here further. For a good definition, see especially Pedersen (1913: 3) "Mit Bezug auf die Bedeutung ist das erste Kompositionsglied Bestimmungswort, das letzte Glied Ausdruck für den Hauptbegriff. Ein Substantiv als erstes Kompositionsglied stimmt in der Funktion einigermaßen mit den mannigfachen Verwendungen des Genitivs überein; zur genauen Angabe der Beziehung wären in den verschiedenen Fällen sehr verschiedene Umschreibungen nötig ..." These are now usually spelt Dun Eidean (in modern Scottish Gaelic) and Edinburgh (in English). For an example where dim occurs according to the order determinans + determinatum, cf. the Gaulish place-name Lug(u)dunum 'Lyons, Laon etc.' (Evans 1967: 221). To this one might add that Welsh has a rather similar way of dealing with this pattern, in that genitive inflection disappeared very early, so that the only formal link between one noun in the nominative and another in the genitive is provided by word-order: see Jones-Thomas (1977: 192-193) for further details for the situation in Modern Welsh and Pedersen (1913:82) for an historical perspective. As Professor Pilch pointed out in his interesting contribution to the discussion after the paper had been presented at the conference, Modern Welsh is rather like Modern Irish in that the old way of forming compounds is still very popular among inventors of new technical terminology, particularly civil servants. I am grateful to Ruairi 0 hUiginn and Jürgen Uhlich for useful comments "an Bonner Kaminfeuern" on this and other matters connected with this paper, even if our opinions naturally enough may differ on certain points. See also de Bhaldraithe (1953: 256) for some further examples and note that he would not (254) accept this as a living formation. For the former, see Schmidt (1953: 88) and for the latter 81.

References Ahlqvist, Anders 1977 "Typological notes on Irish word-order", Studies in descriptive and historical linguistics: Festschrift for Winfred P. Lehmann, edited by P. J. Hopper (Amsterdam: Benjamins), 267-281.

8

Anders Ahlqvist 1980

"On word order in Irish", Papers from the Fourth International Conference on Historical Linguistics, edited by Elizabeth Closs Traugott (Amsterdam: Benjamins), 107-113. 1980 a "Old Irish be, ben ' w o m a n ' E r i u 31: 156-163. Benveniste, Emile 1967 "Fondements syntaxiques de la composition nominale", Bulletin de la Societe de Linguistique de Paris 62: 16-31. De Bhaldraithe, Tomas 1953 Gaeilge Chois Fhairrge: An Deilbhiocht [The Irish of Cois Fhairrge: morphology], (Dublin Institute for Advanced Studies). Binchy, Daniel A. 1972 "Substantival dvandva-compounds in Irish", Indo-Celtica: Gedächtnisschrift für Alf Sommerfelt, edited by Η. Pilch and J. Thurow (Munich: Hueber), 3 8 - 4 1 . Christian Brothers 1960 Graimear Gaeilge na mBräithre Criostai [Christian Brothers' Irish grammar], (Dublin: Gill & Son). DIL 1913-76 Dictionary of the Irish language (Dublin: Royal Irish Academy). Dottin, Georges 1913 Manuel d'irlandais moyen: I. Grammaire (Paris: Champion). EID 1959 English-Irish dictionary, edited by T. de Bhaldraithe (Dublin: Stationery Office). Evans, D. Ellis 1967 Gaulish personal names (Oxford: University Press). Fehling, Detlev 1980 "The origins of European syntax", Folia Linguistica Historica 1/2: 353-387. FGB 1977 Focloir Gaeilge-Bearla [Irish-English dictionary], edited by N. 0 Donaill (Dublin: Stationery Office). Givon, Talmy 1971 "Historical syntax and synchronic morphology: an archaeologist's field trip", Papers from the 7th Regional Meeting (Chicago: Linguistic Society), 39Φ-415. Greenberg, Joseph H. 1966 "Some universale of grammar with particular references to the order of meaningful elements", Universals of language, edited by J. Greenberg (Cambridge, Mass.: MIT Press), 58-90. Harris, Martin 1978 The evolution of French syntax. A comparative approach (London: Longman). Huld, Martin E. 1983 "Birds, beasts and Indo-European merismatic compounds in Albanian", KZ 96, 1: 152-158. Jacobi, Hermann 1897 Compositum und Nebensatz (Bonn: Friedrich Cohen). Jones, Morris — Thomas, Alan R. 1977 The Welsh language: studies in its syntax and semantics (Cardiff: University of Wales Press). Lehmann, Winfred P. 1969 "Proto-Indo-European compounds in relation to other Proto-IndoEuropean syntactic patterns", Acta Linguistica Hafniensia 12: 1-20.

The ordering of nominal compounds in Irish

9

Miller, D. Gary 1975 "Indo-European: VSO, SOV, SVO, or all three?", Lingua 37: 31-52. Pedersen, Holger 1913 Vergleichende Grammatik der keltischen Sprachen II (Göttingen: Vandenhoeck & Ruprecht). Schmidt, Karl Horst 1957 "Die Komposition in gallischen Personennamen", ZCPh 26: 33-301. Sjoestedt-Jonval, Marie-Louise 1938 Description d'un parier irlandais de Kerry (Paris: Champion). Thurneysen, Rudolf 1909 Handbuch des Alt-Irischen: I. Teil: Grammatik (Heidelberg: Winter). Thurneysen, Rudolf 1946 A Grammar of Old Irish (Dublin: Institute for Advanced Studies). Vendryes, Joseph 1908 Grammaire du vieil-irlandais (Paris: Guilmoto). Wagner, Heinrich 1967 "Zur unregelmäßigen Wortstellung in der altirischen Alliterationsdichtung", Beiträge zur Indogermanistik und Keltologie Julius Pokorny zum 80. Geburtstag gewidmet, edited by W. Meid (Innsbruck: Sprachwissenschaftliches Institut), 298-322. Zeuss, Johann Caspar 1871 Grammatica Celtica (2nd ed.) (Berlin: Weidmann).

KARIN AIJMER

The semantic development of will

A characteristic feature of the modal auxiliaries is their propensity to semantic change. The change from one meaning to another is not random but takes place according to certain principles. These principles predicate future changes and explain changes which have taken place. My presentation will deal with the semantic changes of will, especially the change from the modal notion Volition to a temporal expression with future meaning. The traditional view of the modal auxiliaries is that they can be identified with discrete, non-overlapping categories with meanings like Obligation, Volition, Necessity, etc. There is a fixed number of such categories related to each other as members in a system. This view of the modal category as a discrete unit in opposition to other categories is however difficult to combine with the vagueness and non-distinctness of the modal categories in language. A 'weaker' definition of a (modal) category better reflecting the situation in natural language is to identify it with a prototype. The notion of prototype was first used by the psycholinguist E. Rosch and her associates to analyse natural kind categories like 'fruit' or 'bird'. Rosch assumes that there is a taxonomy consisting of a 'basic level' category (e. g., bird) and its hyponyms (sparrow, finch, eagle, penguin, etc). The idea is now that for instance 'sparrow' is a better or more prototypical member of the category 'bird' than 'hen' or 'penguin' (Rosch — Mervis 1975). Other instantiations of the category are 'matched' with the prototype on the basis of similarity. The non-prototypical members themselves may not have anything in common but 'family resemblance' and their similarity to the prototype. The idea that a category is based on a prototype can also be used to explain the relationship between the various senses of will both synchronically and diachronically. By looking at the synchronic prototypical structure of the modal auxiliaries we can reconstruct what the changes must have been.

12

Karin

Aijmer

The prototype functions as a focus from which the other meanings can be derived by extension. The extensions from the focus coincide with possible semantic changes. The prototype theory thus explains and constrains the notion 'semantic change'. How the change has developed can only be reconstructed from historical texts however. Fortunately there exists a substantial amount of research within a philological tradition documenting the development of future will (and shall) at various stages of the language. 1 In Old English, will (OE willan) does not express modality. In its earliest meaning the verb is used about a physical rather than a mental activity (Björkstam 1919: 171). The subject has a need, craving, desire directed to a certain goal. This meaning is strongest when will is found with a nominal object as in: (1)

Tostenc öiode da jefeht willad 'Destroy the people who want fight' Vesp. Ps. 67, 31 (Visser 1969: 1677).

The physical meaning is weakened when will occurs with a thatclause (2) or an infinitival object (3 and 4). (2)

Hwcet wylt öu öcet ic öe do 'What "will" you that I do to you' Mk (the Gospel of Mark) 10, 51 (Standop 1957: 137).

(3)

Ic wille mid flode folc acwellan Ί "will" with flood kill people' Genesis 1296. (Visser 1969: 1677).

(4)

Beowulf is min nama. Wille ic asecgan sunu Healfdene ... gif he us geunnan wile, pcet we hine ... gretan moton: 'B is my name. I "will" tell H's son if he "will" allow us that we greet him' Beowulf 343 (Visser 1969: 1677).

(3) and (4) are syntactically distinct from (2) which consists of a complex sentence.2 (3) and (4) are single sentences in which the infinitive refers to an activity attributed to the subject rather than to the goal itself. In its earliest meaning will requires a human subject but the more abstract notion of Volition (illustrated by will in examples 3 and 4) can be extended to non-human subjects. The properties possessed by the instances of will in (3) and (4) include:

The semantic development

of will

13

(a) the desire is 'active' (the speaker strives towards a desired goal); (b) the subject is human; (c) the main verb denotes an activity. These properties correspond to a prototype schema from which sentences in which will has a different use can be derived. The new uses become associated with different 'definitions' depending on which properties or combinations of properties are selected from the schema. In English and many other languages the modal category Volition (Desiderative) undergoes the semantic change to Future. 3 In modern English this is almost the only meaning of will. In older stages of the language there is however considerable disagreement about when will has the meaning (pure) Future, when it means Volition, and when it has both meanings. The 'rule' Volition Future states the starting-point and the end-point of the change. It does not tell us anything about how the change starts, what the intermediate stages of the change are, or whether the change is still in operation. The linguistic context in which the word which undergoes a change is used is furthermore important. Thus depending on whether the change away from prototypical Volition starts out in the first or the third person of will the results of the change are different. With a first person subject the condition that the desire must be 'active' can be weakened. This gives rise to a different meaning of will. The extension starts in contexts such as: (5)

I wol gladly yelden hire my place Chaucer, c. 1386 (Visser 1969: 1678).

Ί wol' is interpreted in terms of'external initiation'. The will to act 'is prompted by the situation and is a response to a request, offer, invitation, etc, from the interlocutor' (Haegeman 1983: 105). First person will can be paraphrased by the 'bilateral' verb willing or agree. If the speaker is willing to do something, it follows conversationally that he intends to do it and that the future action will take place. In the case that the change starts in the first person, future meaning always depends on the volitional interpretation. The semantic development of will into a 'pure' or nonvolitional future is initiated instead in the third person where it is caused by the extension from human to nonhuman subjects. In the original

14

Karin

Aijmer

(ie prototypical) use will requires a human subject. This (selectional) restriction can be relaxed so that will also collocates with a nonhuman subject. The change is mediated by the existence of vague cases characterized by the fact that the subject is neither human nor non-human. The vague area can be placed between the endpoints on a gradient scale (cf. Leech — Coates 1980). Human (noun) represents one end-point on the scale; non-Human (noun) represents the other. Points on the scale correspond to selectional classes of noun which are neither clearly Human nor clearly nonHuman. If the subject of will is not a human noun, the hearer tries to assimilate it to the human noun of the prototype. If this is not possible will loses its volitional meaning and a development to a future marker can take place. If the subject is non-Human but animate (eg animal or plant) will is less prototypical. (6)

pe smale j seil, fish j he j scil. the whale j wile öus beswiken 'the small /fish/ he /the whale/ "will" thus deceive' c. 1250 Bestiary 400 (Visser 1969: 1682).

(7)

Reed that boweth down from every blast Ful lightly, cesse wynd, it wol aryse Chaucer, Troil. 2, 1388 (Visser 1969: 1682).

An animate subject can be derived by weakening the requirement that the subject should be human in order to allow animate subjects. The subject ('the whale') is treated as if it were a human being with will. When the subject is the name of a plant /as in (7) / will loses the meaning of Volition and refers to the concrete manifestations of Volition only. If the noun is non-Human but abstract it can be a personification. (8)

Malencolie ...An hundred times in an houre Wol as an angri beste loure c. 1390 Gower, C. A. 3, 29 (Visser 1969: 1682).

A personified subject can be derived by weakening the requirement in the prototype schema of will that the subject must be human so that sentences where the subject can be an inanimate noun with human properties are allowed.

The semantic

development

of will

15

Another deviation from the prototype is provided by examples in which the subject of will is an inanimate noun with a human-like capacity. Although the subject is inanimate, there is no reference to the future. Machines (means of transportation, tools, medicines, etc) can be thought of as being able to do things by themselves and as having Volition but in a less direct way than human beings. Thus the interpretation of will in a sentence with such a subject can be compared with and assimilated to the prototypical interpretation of will with a human (agentive) subject. The following example is from modern English. (9)

Tell them you had an automobile accident and your car won't operate

We think of the car as resisting our efforts if it does not do what it is expected to do. Will is typically negated as in the following example. (10)

This screw won't turn

What the speaker actually means is 7 can't move this screw'. Close (1977) from whom I have taken the last example speaks about 'shifting responsibility for human incapacity to an inanimate agent'. If the effect desired by the speaker in a situation involving the inanimate noun does not take place, the responsibility is attributed to the inanimate noun. We can interpret the sentence by assimilating it to the prototypical instantiation with a human noun. The difference from the prototype may show up as a slight difference of meaning. Thus will may convey the nuance 'ability' in these examples. If will becomes too remote from the prototype its meaning is weakened. In the following example from Old English with an abstract inanimate subject Visser interprets will as a future marker without any volitional element. (11)

hwider hrepra gehygd hweorfan wille 'whither the thoughts of his mind will turn' Wanderer (Visser 1969: 1692).

We have seen that with a non-human subject the transition to Future can be established; if the subject is human, on the other

16

Karin

Aijmer

hand, will retains the meaning Volition although future meaning may be implied. There are however also indeterminate cases in which will can be interpreted both as Volition (including Willingness, Intention) and as Future. F o r example: (12)

he ponne wile deman ... anra gehwilcum 'he then "will" judge each' Dream of the Rood, 103 (Visser p. 1693).

It is not always possible to determine in a process of semantic change which of the two meanings is present in a particular example. It is impossible to know if will judge means 'wishes to judge' or 'will (future) judge'. On the one hand, the subject is human which suggests at this stage that will denotes Volition. On the other hand, the complement-prediction can be interpreted as a state which is a sign that will is non-volitional (ie a pure future). The communicative effect is the same with both interpretations of will (Stern 1931: 355). It does not matter what the speaker's intention has been. These vague cases in which will has two possible definitions which are combined are a sign of transition from one meaning to another. The direction of the change is furthermore clear. There will be more and more examples in which will must be interpreted in the new way as the future. In the last stage of the change Future emerges as the new meaning of will in (nearly) all contexts. (Stern 1931: 351 'Permutation'). The development which has been outlined is usually seen as a gradual development from Old English to older New English. There is however uncertainty about when the change was concluded (Weida 1975: 324). When will is disassociated from the subject in the second person it is either a future auxiliary (associated with epistemic modality) or it expresses deontic modality (demanding somebody to d o something). (13)

You will light my fire at 7 o'clock 1905, Onions, An Adv Eng Syntax, 136 1697).

(Visser 1969:

If will is deontic we must assume a further extension Future ->· Deontic. The deontic interpretation is however indistinguishable from the future one if the person who predicts something is in the position to bring about that the action takes place (Lyons 1977: 846).

The semantic

development

of will

17

We have seen that Future is derived by extension from the prototype Volition. It can itself function as a focus (another prototype) and develop new meanings. The development Future -> Modality is the 'reverse' of the earlier change from Modality (Volition) Future (cf. Fleischmann 1982). Will starts out as a predicate in which the modality (Volition) is attributed to the subject — it ends up as expressing epistemic modality, a kind of modality in which the speaker expresses his attitude to the whole proposition. The latter development in this chain is explained by the link between Future and (epistemic) modality. 'There is a demonstrable historical connection between reference to future and non-factivity in too many languages for it to be regarded as a matter of accident that languages should rarely, if ever, distinguish systematically between statements of fact about the future and subjectively modalized propositions' (Lyons 1977: 817). It follows that a sentence with a future form is similar to a sentence with an epistemic modal auxiliary. The modal meaning associated with will at this stage of the change is to express a high degree of certainty. Thus a vW//-sentence referring to the future could be qualified by Ί am convinced that'. When will refers to the future, modality and temporal elements of meaning are in balance. Once will is used to indicate modality, the proportions between temporal and modal meanings can change however. At the final stage of the change all temporal meaning is lost. Additional changes have the effect that fluctuations of meaning within the range of possible subjective epistemic meanings may occur. Thus the purely modal will expresses various qualifications such as probability, inference, likelihood, tentativeness besides certainty. Epistemic will is found in Old English (14) and then not until the fourteenth century. With the exception of dialects it plays no role until the beginning of the twentieth century (Weida 1975: 444). The change from Future is first seen in non-factive environments as illustrated by the following examples:

(14)

I am aferd there wylle be somethyng amyss c. 1450, Coventry Myst; Assumption 349 (Visser 1969: 1701).

(15)

Hwcet wile pis wesan? Blickl Horn, 239, 29 Mid.

(Visser 1969: 1701).

18

Karin Aijmer

In the first sentence will is found in a subordinate clause embedded in a non-factive matrix; the second example is a question of deliberation or doubt rather than a declarative sentence. The sentence-qualifying function is illustrated by: (16)

Ye wil peraduenture meruel of thys c. 1522 St ThMore (Wks. 1557) (Visser 1969: 1701).

Will presents a state of affairs (in the present or the past) with the qualification that it is only tentative. It is often difficult to distinguish between a diachronic change which is a gradual process in time and an extension of the prototype which is a synchronic process and does not cause a change but a new definition of the prototype. If the reference is to a class of events rather than a single event, will develops the meaning disposition or characteristic property. Sentence (17) has a generic subject and therefore refers to a class of events. (17)

yn drunkenes men wyl rage 1303 Rob of Brynne. Handl. Synne 9713 1681).

(Visser 1969:

In (17) neither subject nor predicate are associated with prototypical Volition. Consider also: (18)

Tapres of a pound, to brenne fromme Ye, Even forth as longe as they woll dure 1428 Early Eng. Wills 81, 17 (Visser 1969: 1684).

The verb 'dure' describes a permanent (inherent) state. It is more like a class of events than a single event. The subject is furthermore inanimate. As a result of these deviations from the prototype will loses the meaning Volition. The new meaning resulting from the extension is called 'capacity' (Haegeman 1983), 'power' (Palmer 1979) or 'potential' (Lawler 1972). The meaning Habit is slightly different from the meanings above. It arises if the activity can be understood as repeated (habitual). (19)

Hwcenne wylle je synjan of en oppe nihtsanjc? 'When "will" you sing the evening or night song?' JElfric, Colloq 276 (Visser 1969: 1681).

The semantic development of will

19

Diagrammatically Future (and Epistemic Modality) and the 'omnitemporal' meanings Habit, Disposition and Capacity are related to each other and to the prototype as follows. The prototype is placed centrally in the extension of the term. The non-prototypical meanings are closer to the periphery of the extension. Habit, Capacity and Disposition are more similar to each other than to the future. Epistemic modality

Volition

Summary The meaning of a word is 'dynamic, only partially determined' and 'open and susceptible to contextual modifications' (Linell 1982: 93). The most stable part of the meaning of a word which is most likely to be intersubjectively shared by members in the linguistic community is the prototype. Both diachronically and synchronically the meaning of will (and other modal auxiliaries) can be looked upon as being organized around a prototype. Some shifts of meaning caused by the prototypical structure of the word are synchronic, others start off in the same way as extensions from a prototype and gradually spread to new contexts. The relationship between the various meanings is that of a focus and its extensions. Extension from the prototype explains both generalization or broadening of meaning and contextual shift.

20

Karin Aijmer

A category organized around a prototype contains vague areas (cf. Karlsson 1983). Leech — Coates 1977) discuss two types of vagueness, gradience and merger which are found in the modal auxiliaries. Both types of vagueness are found in will and play a role in its semantic development. In gradience the vague cases can be interrelated as steps on a gradient scale relating the old and the new use of a lexical item along some dimension. In a merger two originally distinct meanings may under certain circumstances be close enough to function in the same way. The change Volition -> Future is initiated by the extension along the dimension Human -> non-Human (noun) when will is used with a third person subject. The generalisation of future meaning to contexts where it cannot be explained by extension from the prototype is, on the other hand, the result of a merger between Future and Volition (Willingness and Intention). The old and the new meaning come so close to each other in the course of the semantic development that they can occasionally refer to the same event. This convergence results in the emergence of future as a new meaning.

Notes 1. Cf. Weida (1975), Standop (1957) and Tellier (1962). The examples have generally been quoted from Visser (1969). 2. The constructions in (1) and (2) disappear in Middle English. 3. The emergence of the Future also involves the development of will to a grammatical element (a modal auxiliary) by a process of grammaticization. This makes it possible that there is a semantic explanation for the syntactic change (or vice versa). Cf. Lightfoot (1979: 100) for a different view.

References Björkstam, Harald 1919 De modala hjälpverben i svenskan I: Tör, lär, mon, mä, matte och vill (Lund: Hakan Ohlssons boktryckeri). Close, R. A. 1977 "Some observations on the meaning and function of verb phrases having future time reference", Studies in English usage, edited by WolfDietrich Bald and Robert Ilson. (Frankfurt/M.: Lang), 125-156.

The semantic development of will

21

Fleischman, Suzanne 1982 The future in thought and language (= Cambridge studies in linguistics 36) (Cambridge: Cambridge University Press). Haegeman, Liliane 1983 The semantics of will in present-day British English. A unified account (= Verhandelingen van de koningklijke Academie voor wetenschappen, letteren en schone kunsten van Belgie). Karlsson, Fred 1983 "Prototypes as models for linguistic structure", Papers from the seventh Scandinavian conference of linguistics II, edited by Fred Karlsson (= University of Helsinki, department of general linguistics, publications 10), 583-605. Lawler, John 1972 "Generic to a fault", PCLS 8, edited by Paul M. Peranteau et al, 247-259. Leech, Geoffrey — Jennifer Coates 1980 "Semantic indeterminacy and the modals", Studies in English linguistics for Randolph Quirk, edited by Sidney Greenbaum et al (London: Longman), 79-90. Lightfoot, David 1979 Principles of diachronic syntax (Cambridge: Cambridge University Press). Linell, Per 1982 The written language bias in linguistics (= Studies in communication 2, University of Linköping). Lyons, John 1977 Semantics 2 (Cambridge: Cambridge University Press). Palmer, Frank Robert 1979 Modality and the English modals (London: Longman). Rosch, Eleonor — Mervis, C. B. 1975 "Family resemblances: studies in the internal structure of categories", Cognitive psychology 7: 573-605. Standop, Ewald 1957 Syntax and Semantik der modalen Hilfsverben im Altenglischen magan, motan, sculan, willan ( = Beiträge zur englischen Philologie 38). (Bochum-Langendeer: Pöppinghaus). Stern, Gustaf 1931 Meaning and change of meaning with special reference to the English language (= Göteborgs högskolas ärsskrift XXXVIII) (Göteborg: Elanders boktryckeri aktiebolag). Tellier, Andre 1962 Les verbes perfecto-presents et les auxiliaires de mode en anglais ancien (VIII e S. - XVI e S.) (Paris: C. Klincksieck). Weida, Gudrun 1975 Der Gebrauch von shall/should und will/would in englischer Prosa am Ende des 16. Jahrhunderts (Band 1 u. 2) (Augsburg: Blasditch GmbH). Visser, F. Th. 1969 An historical syntax of the English language III (Leiden: E.J. Brill) [Visser].

E U N G - J I N BAEK

Semantic shifts in Korean honorification

Honorification is one of the salient features of the Korean language. The Koreans are forced to make a choice of speech levels, before they can utter a sentence: polite level or non-polite level. The forms used at the polite level are called honorifics. These honorifics are used to show deference. In this paper, I will examine historical developments of the honorifics on the basis of Prof. Lipka's descriptive framework of obligatory inherent and inferential semantic features (Lipka, 1984). I will show that there were semantic changes in the history of Korean honorifics: a) that the verb honorifics in Modern Korean were the markers of social superior class in Middle Korean. b) that the non-polite personal pronouns in Modern Korean were neutral or unmarked forms in the fifteenth century, that is, they were used in talking to superiors as well as inferiors. I will also argue that semantic change is intimately related to change in social structures. First of all, I will describe the system of honorifics in Modern Korean in Section 1 and the speech levels in Middle Korean in Section 2, and then semantic changes in Korean honorification will be discussed in Section 3.

1. Modern Korean honorifics The choice of speech levels in Modern Korean depends on the speaker's evaluation of his relationship with the person he is addressing (addressee), and with the person about whom he is talking (referent). The speaker's evaluation is based on his social status. Its determining factors are age, position (occupation), sex and intimacy. Here the key notion is that of social superiority, which is determined on the basis of the four factors. The speaker-addressee relationship

24

Eung-Jin

Baek

and the speaker-referent relationship may be explained in terms of the axis of address and the axis of reference, respectively.

1.1 Axis of address (addressee honorification) In Korean, verb endings indicate the speaker-addressee relationship. Style Style Style Style

A: B: C: D:

apeci cipey issapeci cipey issapeci cipey issapeci cipey issfather home stay 'Is Father at home?'

upnikka? eyo? na? e? verb endings

All these four sentences seem to be synonymous. However, the speaker has to choose one of the verb endings according to his attitude toward the addressee. These verb endings show the speaker's relationship with the addressee: whether the speaker is the superior or inferior, or whether the addressee and the speaker have an equal social status. The verb endings in Style A and Style Β are used to show the speaker's deference to the addressee. The obligatory inherent semantic feature associated with the verb endings in Style A and Style Β is [honorific]. These endings are called addressee honorifics. The verb endings in Style C and Style D do not show any respect to the addressee at all.

1.2 Axis of reference The referent on the axis of reference may be the subject, the direct object or the indirect object.

1.21 Subject honorification In order to show deference to the subject of the sentence, we may use the honorific verb suffix -si-, which is called subject honorific, as in (2.b), honorific verb stems, as in (2.c), honorific nouns, as in (2.d) and (2.e), or the honorific subject marker -kkeyse, as in (2.f).

Semantic shifts in Korean honorification

(2)

a. b. c. d. e. f.

apeci pap apeci pap pap apeci apeci cinci cinci apeci- nim cinci apeci-nimkkeyse father meal 'Did Father eat?'

mekmekcapswucapswucapswucapswueat

usisisisisi-

25

essessessessessess-

e? e? e? e? e? e?

Past

Ending

1.22 Direct object honorification A speaker may use honorific verb stems, as in (3.b), or honorific nouns, as in (3.c), to show respect to the direct object. (3)

-lul a. apeci b. apeci -lul c. apeci- nim -ul father Accusative Ί will see Father.'

popoyppoypsee

keysskeysskeyssProspective

ta ta ta Ending

1.23 Indirect object honorification In order to show respect to the indirect object, we may use honorific verb stems, as in (4.b), honorific nouns, as in (4.c), or the honorific dative -kkey, as in (4.d). (4)

a. b. c. d.

apeci -eykey apeci -eykey apeci-nim -eykey apeci-nim -kkey father Dative

semwul semwul semwul semwul present

-ul -ul -ul -ul Accusative

cwutulitulituligive

essessessessPast

ta. ta. ta. ta. Ending

Ί gave Father a present.' 1.24 Honorific personal pronouns Modern Korean has developed the humble forms of the first-person pronouns and the polite forms of the second-person pronouns, as in (5). These humble forms and the polite forms are called honorific

26

Eung-Jin

Baek

pronouns. The inherent semantic feature associated with these pronouns is [honorific]. (5) First-Person: Second-Person:

Humble Non-Humble Humble Non-Humble

Singular ce na sensaynghim (male) ne

Plural cehuy wuli

nehuy

2. Speech levels in Middle Korean Traditional grammars of Middle Korean all agree that Middle Korean honorifics are classified into three categories, as in (6).2 (6)

a. Addressee Honorifics: b. Subject Honorifics: c. Object Honorifics:

-ηΡ -si-sap-sava

-Rap-Rava

-tsap-tsava-

According to traditional grammarians, the addressee honorific occurs when the speaker wants to show his deference to the addressee. The subject honorific occurs, when the subject denotes a person to whom the speaker wants to show his respect. The object honorific occurs, when the object denotes a person (or an action, a state, words or children of a person) to whom the speaker wants to show his deference. There is a consensus that the obligatory inherent semantic feature associated with these honorifics is [honorific] or [deference]. However, according to my analysis of Middle Korean, as reflected in the Sekpo Sangcel (SS, 1447 AD.) and the Penyek Nokeltay (PN, 1517 AD.), the term honorific or deference seems to be misleading. The morpheme -yi, which is usually called addressee honorific, was used when the speaker did not want to show his deference to the addressee. Let us examine the dialogue (SS 24:18) between the king and the prison guard who was appointed by the king. (7)

King:

na -lal tsuk- yulye ha- nan- ta. I Accusa- die Causa- Inten- be Pres- Endtive tive tive ent ing 'Do you intend to kill me?'

Semantic shifts in Korean honorification

11

Prison kali ho- lye ha- norji- ta. guard: that do Inten- be PresEndtive ent ing Here, the prison guard tries to kill the king. We can not say that the prison guard wants to show his deference to the king. It seems to me that the morpheme -qi was obligatorily used to the king, since the social status of the king was superior to that of the prison guard. Then, we may conclude that in Middle Korean the morpheme -ljiwas used as a social class marker, and that the meaning of the morpheme was 'social superiority.' The obligatory inherent semantic feature associated with -yi- was [social superior]. On the basis of the inherent meaning of social superiority, I would like to devise a new terminology which captures the semantic features of the Middle Korean morphemes which correspond to the honorifics in Modern Korean. I shall refer to the three categories in (6), respectively, 'superior addressee', 'superior subject,' and 'superior non-subject.'

2.1. Axis of address This axis shows us the relationship between the speaker and the addressee. In Middle Korean, the speaker — addressee relationship was based on social superiority or inferiority. The inferior had to use the morpheme -yi, whether or not the speaker wanted to show his deference to the addressee. This morpheme indicates that the addressee is socially superior to the speaker. The inherent semantic feature associated with this morpheme is [superior addressee], as shown in (7). 2.2 Axis of reference This axis shows us the relationship between the speaker and the referent. The referent may be the subject or the non-subject. 2.21 Superior subject The subject may denote a person whose social status is superior to the speaker. In Middle Korean the morpheme -si- indicates that the

28

Eung-Jin Baek

subject is a person superior to the speaker. This morpheme was used as the superior subject marker. The inherent semantic feature associated with this morpheme is [superior subject], as in (8). Prison guard:

(8)

wang-i mot king Nomina- not tive 'The King shall not

na-si lita. move Prospec- Endout tive ing go out.'

2.22 Superior non-subject The morpheme Superior Non-Subject indicates that a non-subject is superior to the subject or the speaker, or that the subject is inferior to a non-subject. The non-subject may be the object, the indirect object, any other noun phrase other than the subject or someone in discourse. The inherent semantic feature associated with this morpheme seems to be [superior non-subject], as in (9). (9)

a.

Superior object ... putyehi matsnaBuddha Accusa- meet tive ' . . . met the Buddha and . . .

b. Superior indirect object ... putyeskdy patBuddha Dative give

-ni Circumstantial Converbial

-/tsekdy Prospec- time tive

'When . . . gave it to the Buddha c. Superior noun phrase with dative ... putye· -skdy oBuddha Dative come Infinitival ' . . . came to the Buddha . d. Superior in discourse ... thstsa- i kissava- a prince Nomina- happy Infinitival tive ' . . . the prince was happy ..

Semantic shifts in Korean honorification

29

2.23 Neutral personal pronouns In Middle Korean, personal pronouns were neutral as to the speech levels of the constructions in which they were employed. There were no special forms used in talking to superiors or inferiors. The same personal pronouns in (10) were used both to superiors and inferiors. (10) First person Second person

Singular na ne

Plural uli nehay

3. Semantic changes 3.1 Semantic shift As shown above, the verb honorifics of Modern Korean were used as social superior markers in Middle Korean. The inherent semantic feature [superior] of Middle Korean corresponds to the semantic feature [honorific] in Modern Korean, which may have been the inferential feature derived from the inherent feature [superior] in Middle Korean. In Modern Korean, the semantic feature [superior] is regarded as the inferential feature derived from the inherent semantic feature [honorific]. We may conclude that a semantic shift took place in the history of honorifics: 4 from

superior (honorific)

to

honorific (superior)

3.2 Semantic restriction Personal pronouns in Middle Korean were neutral as to the social class status, superior or inferior. Thus, the same personal pronouns were used in talking to superiors and inferiors. However, in Modern Korean, all the personal pronouns of Middle Korean are confined to reference to a person to whom the speaker does not need to show his deference at all. These can not be used in talking to superiors in order to show respect. The new honorific personal pronouns, which are used to show deference, have been transplanted

30

Eung-Jin Baek

from somewhere else. We may conclude that another semantic change took place in the history of honorifics: from unmarked personal pronouns to non-honorific personal pronouns. This semantic change is regarded as a case of semantic restriction. In connection with the semantic feature [social superior] it is interesting to note that in Middle Korean, as in the SS, the three morphemes with the semantic feature [superior] were used often in talking to gods, Buddhas, kings, and the heaven. On the other hand, these morphemes were hardly used in dialogues between common people, as in the PN. It seems to me that the origin of the honorifics in Korean has something to do with gods, Buddhas, religions, and kings. We may assume that the Korean honorifics originated with gods or religions. We may also speculate that a history of Korean honorifics is divided into three stages: First stage:

The morphemes corresponding to the honorifics in Modern Korean were used in talking to gods and kings. Second stage: These morphemes were used in talking to superiors: kings, upper class government officials, older people, or husbands. This stage was influenced by Confucian teachings. Third stage: These morphemes are used to inferiors and superiors to show deference. The history of Korean honorifics seems to reflect the history of social class structure in the history of Korea. Accordingly, and historical study of Korean honorifics seems to require a thorough examination of the historical developments of social class hierarchies and their relation to linguistic features. It seems to me that semantic change, particularly semantic change of Korean honorifics, is intimately related to change in social structures and hierarchies.

Notes 1. 2. 3. 4.

The Yale Romanization System is used for Modern Korean. See He, Wung (1975) and Ahn, Pyung-hi (1982). A phonemic system is used for Middle Korean. Inferential semantic features are enclosed in parentheses.

Semantic shifts in Korean honorification

31

References Ahn, Pyung-hi 1982

"Humble forms in Middle Korean revisited" [in Korean], Journal of Korean Linguistics 11: 1-24.

He, Wung 1975 Middle Korean grammar [in Korean] (Seoul: Saym Mwunhwa). Lipka, Leonhard this volume "Inferential features in historical semantics".

PHILIP BALDI, VICTOR BRODERICK A N D DAVID S. PALERMO

Prefixal negation of English adjectives: psycholinguistic dimensions of productivity* I. Introduction The study of negative prefixing in English has long been of concern to historical linguists and to those interested in the morphology of the language. Jespersen (1942) is a standard resource, and the data are presented in admirable detail in Marchand (1969). Zimmer's (1964) monograph, though primarily concerned with affixal negation, is a groundbreaking study of morphological productivity and should be consulted as a fine example of early attempts to determine the productivity of morphological constructs. Negative prefixing is a fertile testing ground for theories of morphological productivity. The general issue of productivity of word-formation rules has attracted a great deal of attention recently, much of it prompted by Aronoff (1976). The productivity issue is now an important locus of discussion in both synchronic and diachronic studies of word-formation. 1 Our purpose in this paper is to discuss the results of two psycholinguistic experiments which were designed to test the productivity of certain negative prefixes in English adjectives, and to offer some proposals concerning the principles which underlie the form and distribution of these prefixes. We are specifically concerned in this paper with the negative prefixes un-, IN- (in-, im-, il-, ir-)2 and, to a lesser extent, non- and dis-. As is well-known, un- is the only one of these prefixes which is of native English origin (though un-, IN-, and Gk. a- (an-) are all from the same PIE source, *«).3 According to Marchand (1969: 168), the Latin negative prefix IN- can be considered an independent formative in English by about 1500 AD. This development, of course, was brought about by the massive number of borrowings into English from Latin and French which was systematically carried out from the fourteenth to the nineteenth

34

Philip Baldi, Victor Broderick

and David

Palermo

centuries, but dates to the earliest periods of Old English. Marchand makes the claim that un- has always been 'stronger' than IN- and is 'ousting it more and more'. In Early Modern English, for example, IN- could be prefixed to almost any adjective with a Latin or French basis, and always according to the Latin rules governing assimilation, viz., im- before labials (m, b, ρ), il- before /, ir- before r, and in- elsewhere. Thus such words as inceremonious, incertain, inchangeable, ingrateful, incomfor table, impleasing and impopular were common at this time, though one can readily see that these words have all given way to un- preflxation in Modern English (cf. uncomfortable, ungrateful, unpopular, etc.). There are also cases where Latinate words formerly attested with the un- prefix have given way to their earlier etymological negative with IN-, e. g., uncredible, undubitable, uneffable, unexpert, unfirm, unformal, unglorious, unperfect, unpiteous, unpossible. All these words and others like them are now prefixed with IN- (cf. incredible, informal, imperfect, impossible). It is significant, in our opinion, that such a change took place only with adjectives which were etymologically Latin/French. It is also significant that there are no examples of etymologically native English words which are prefixed with IN-. Thus we never find such forms as *imborn 'not born', *irrocky 'not rocky', *illoved 'not loved'. If IN- is indeed an 'independent formative', as Marchand claims, why is it restricted to Latinate roots? 4 In approaching the matter of morphological productivity and prefixal negation of English adjectives we focus on three central questions: 1) Is it possible to demonstrate experimentally which negative prefix or prefixes are the most productive? What prefix or prefixes are speakers most likely to use to negate newly introduced words when there is no preexisting lexical specification of a negative prefix? 2) Are native speakers of English sensitive to the etymological distribution of negative prefixes and root words, or are the Latin origins of IN-, dis- and non- and the native origins of un- opaque to modern speakers? Do speakers tend to use Latinate prefixes with Latinate roots, or is Latin etymology no longer a factor which influences negative prefix production in modern English?5 3) Do speakers of English observe the historical phonological constraints on the distribution of IN- when negating newly introduced words? If speakers do not observe the historical phonological constraints, this might mean that some or all of negatively

Prefixal negation of English adjectives:

dimensions of productivity

35

prefixed adjectives like illogical, irreverent, impossible, intolerant, etc., might simply contain relics of Latin morphophonology and are not creatively generated by English speakers (cf. fn. 4). If, on the other hand, speakers do observe the historical phonological patterns, what type of model best accounts for the form and distribution of the negative prefixes? There are at least three distinct models which could account for the phonological distribution of IN- as in-, im-, il-, or ir-. Briefly, these are:

A. Classical assimilation model This model posits a single underlying form /in-/ which is phonologically transformed in appropriate environments by a set of assimilation rules which change η to m before labials, η to I before / and η to r before r.

B. Item-and-arrangement (matching) model This model posits four separate (unrelated) lexical entries, viz. in-, im-, il-, and ir-. The use of each entry is governed by euphonic constraints; for example, im- cannot occur before a non-labial, ilcannot occur before any segment other than /, etc.

C. Analogical construction model This model assumes four separate lexical entries, as does the itemand-arrangement model. However, there are no abstract euphonic rules that can operate purely phonologically (without referencing particular lexical items to illustrate the phonology in question). Rather, there is an analogy-making procedure that seeks out instances from the lexicon which resemble the current problem. In attempting to negate sempient, for instance, insentient and insensitive might be considered. This model assumes a search procedure, and a similarity metric, as well as an evaluative mechanism for deciding among competing sets of analogies. In order to account for the phonological constraints on the distribution of negative prefixes,

36

Philip Baldi, Victor Broderick and David

Palermo

the similarity metric presupposed by this model must specify that root initial segments are the important locus of similarity in analogies.

II. Method In approaching the three questions on which we focus in this paper, we present two experiments which were designed to test native speakers' intuitions about negative prefixes. One of the experiments is a production task (fill-in-the-blank), and the other is a preference task (multiple choice).

Experiment #1: Production test (fill-in-the-blank) Subjects: Seventy-two introductory psychology and introductory linguistics students of the Pennsylvania State University participated for credit. Five were eliminated for not following instructions, leaving sixty-seven subjects whose data were used. Materials: A list of sixty pseudo-adjectives was prepared to test the productivity of negative prefixation with novel lexical items (see appendix for the complete list). These sixty words contained thirty pairs of adjectives. While the pairs were matched for initial sound, the first member of each pair was a Latinate pseudo-word, the second was a native pseudo-word (e. g. Lat. limoral: native lurmy). The Latinate/native distinction was created both through native speaker brainstorming and criticism, and by contrasting the pairs on three variables: 1) Number of syllables. Latin words tend to be polysyllabic, whereas native words tend to have fewer syllables. Accordingly, Latinate words were made one syllable longer than their corresponding native words, e. g., Lat. limoral (3 syllables); native lurmy (2 syllables). 2) Suffixes. Latin adjectival suffixes are generally longer and have a more complicated internal structure than their native counterparts. Such suffixes are also quite plentiful in English, and are easily attached to pseudo-roots (e. g., -al, -ive, -ant, -ous). The native suffixes -ed, -Jul, and -ish and -y were affixed to the native words. 3) Spelling and consonant peculiarities were incorporated where feasible. For example, c was used for Latinate words whereas k or

Prefixal negation of English adjectives: dimensions of productivity

37

ck was used for native words. Latinate words tended to have fewer and more restricted consonant clusters, whereas native words were invested with consonant clusters as much as could be realistically done. Six pseudo-words for each of five phonological environments (conditions) were created in view of the historical phonological constraints on IN-. The initial segment of each pseudo-root was assigned to one of the following conditions: a) Vowels (half of the vowel-initial pairs of words began with the letter a, half with e). b) "Neutral" consonants, i. e., those which do not trigger assimilation (two of the pairs of words began with the letter n, the remaining with the letters c(k), j\ t, and 5). c) Labials (half of the words began with ρ, half with m). d) / e) r According to the historical phonological constraints on Latin words, vowels and neutral consonants should receive in- prefixes, and the remaining phonological conditions should receive im-, il-, and />-, respectively. Six pairs of words (six Latinate and six native) were assigned to each phonological condition; examples of these pseudoadjectives follow. Phonological condition

Latinate

Native

vowel neutral consonant labials

exilar nexibular profinous lanurous retulous

ertful neshorly prafilled lattished ruckful

Procedure: Subjects were given a sheet with instructions; attached to the instruction sheet was a list of the Latinate pseudo-adjectives and a list of the native pseudo-adjectives, counterbalanced across subjects for order of presentation. N o indication was given to the subjects that the lists differed in any way. The instructions directed subjects to write a negative prefix before each pseudo-word to form its opposite. This was the fill-in-the-blank (production) version of the task.

38

Philip Baldi, Victor Broderick and David

Palermo

Experiment #2: Preference test (multiple choice) Subjects: Forty-one introductory psychology students of the Pennsylvania State University participated for credit. Materials: A list of 60 pseudo-adjectives and 90 real adjectives was constructed. The pseudo-adjectives were identical to those used in the production task of experiment #1. The real adjectives consisted of 30 real native words whose negatives regularly take the prefix un-, 30 real Latinate words regularly taking the native prefix un-, and 30 real Latinate words taking the Latinate prefix IN-. These three real word lists were matched with their corresponding pseudowords for number of syllables and phonological condition. Examples of these real words are given below. Phonological condition

Real Un-Native

Real Un-Latin

Real IN-Latin

PseudoLatin

PseudoNative

vowel neutral consonant labial 1 r

aware needed proven lucky ripened

affected natural paralleled lyrical romantic

accurate nocuous probable logical reverent

addorsent neculous profinous locanous robulous

arnful nellful prafilled loofish ruckful

(See appendix for the full list)

The resulting list of 150 real and pseudo-words was divided into two lists of 75 words: half of the six words within each of the 25 conditions (5 phonological conditions crossed by 5 word type conditions) was randomly selected for "List 1" and the other half for "List 2". Multiple choice response forms were constructed from these lists. First the 75 root words from the lists were randomized. Then each root word was followed by a list of six possible negations of that root. The negations were constructed by affixing un-, non-, in-, im-, il-, and ir- to the root word. These prefixes were selected because they were the most common responses given in the production task of experiment #1. Each word in each form used an independent random ordering of negative choices, so that no two subjects received equivalent orderings of choices or of root words. An example of a multiple choice item follows: nexibular: ilnexibular innexibular

Prefixal negation of English adjectives:

dimensions of productivity

39

nonnexibular unnexibular irnexibular imnexibular Procedure: Subjects were tested in groups, each receiving an instruction sheet and a multiple choice response form. Subjects were instructed to decide which of the six alternatives was the best and second best negation of the root word, and to indicate their judgments by writing a 1 and a 2 by their first and second choices, respectively. After completing the multiple choice task, 24 of the subjects were given a list of the 150 words and asked to judge whether each was a native word or a borrowed foreign word.

III. Results A) Real words We will first present analyses of the responses to the real words of the preference task in experiment #2. We present these data first because the real word data provide an important backdrop against which responses to the pseudo-word data can be evaluated. Each real word in the preference task is lexically associated with a particular negative prefix, which we assume must be explicitly specified in each speaker's mental lexicon. In fact, subjects selected the expected prefix as first choice 84% of the time (see table 1). However, un- and IN- did not appear to have equal status: IN- was selected 83% of the time when appropriate for Latinate roots, whereas un- was selected 92% of the time for native roots. (We will discuss un- for Latinate roots below.) Examining the intrusions we find that un- was selected 12% of the time when IN- was the appropriate response, while IN- intrusions occurred only 3% of the time for native English roots, which require un-. Hence, although lexical specification seems to be the basis for subjects' responses to real words, the tendency to give un- responses was stronger than the tendency to give IN- responses, both in terms of correct responses and in terms of intrusions. Non- was generally a rare first choice, but for the real native and Latinate un- words, non- was the most popular second choice. On the other hand, IN- was less frequent as a second choice for these

40

Philip Baldi,

Victor Broderick

and David

Palermo

Table 1 Percentage of first (and second) choices of each prefix type for real words in the preference task Prefix type appropriate

native INun-

non-

intrusions incorrect IN-

un-Latin

83 (7) 92 (6) 3(62)

phonologically correct INun-

IN-Latin

77 (18) 1(30)

3(21)

14 (46)

10(25) 12(53)

1(12)

4(11)

1 (10)

Total 84 (10)* 6(46)

8 (33)** 2(11)

Note: Some columns do not add up to exactly 100% because of rounding of individual cells. * 84 (10) is based on the correct responses in rows 1 and 2 combined. ** 8 (33) is based on the combination of the two types of intrusions.

words (62% and 46% for non- vs. 21% and 25% for IN-). The reluctance to choose IN- when the lexicon specifies un- stands in contrast to the relative freedom with which un- was chosen (53%) when the lexicon specified IN-. Once again, un- appears to be the most productive prefix. The inclusion of Latinate roots that take the native prefix unpermits us a brief look at the nativization or assimilation of borrowed words. This is in itself a difficult issue, as there are no convenient formulae on which to base claims about 'degree of nativization.' Generally speaking, words are considered nativized when they conform entirely to the native phonological and morphological patterns of the recipient language. For example, the Latin/French word reasonable takes stress on the first syllable, is suffixed with the native morpheme -ness in nominalization (reasonableness), and it takes the prefix un- in negation {unreasonableness). It is totally nativized, whereas the Latin/French word sensitive conforms to the Latin/French pattern in forms like insensitive and insensitivity (see Bynon 1983: 217 for discussion). Correctly used un- proved to be a less likely response for Latinate roots (77%) than for native roots (92%). IN- intrusions were fairly common for the Latinate un- words (10%). Moreover, the lexical specification of un- for Latinate words did not preempt competing

Preßxal

negation of English adjectives: dimensions of productivity

41

prefixes as well as the other real words did. In particular, non- was much more likely to be the first choice for Latinate un- words than for the other real words. In light of these patterns it seems that speakers were less certain as to the correct response for real words of Latin origin which take un-. Nativization or assimilation of these words into English seems to be only partially complete, at least by the measure used in this research. The twenty-four subjects who were asked to give native vs. borrowed judgments on the full set of words demonstrated fairly accurate intuitions about the distinction. Table 2 summarizes their judgments. Subjects were able to judge real native words as native about as well as they judged real Latin words as Latinate. Subjects were also able to judge pseudo-native and pseudo-Latin words in the appropriate class.

Table 2 Judgments of nativeness vs. Latinness for each word type Word type

Percent judgments 'Latin' (number of words out of 30 judged 'Latin' by a majority)

Percent judgments 'Native' (number of words out of 30 judged 'Native' by a majority)

Real Un-Native Real Un-Latin Real IN-Latin Pseudo-Latin Pseudo-Native

26 (3) 72 (28) 69 (24) 69 (26) 32(3)

74 (27) 28 (2) 31 (6) 31(4) 68 (27)

Furthermore, subjects judged pseudo-Latin words to be just as Latin as real Latin words, and pseudo-native words to be just as native as real native words. These data are reassuring with respect to our attempted manipulation of Latinness vs. nativeness in creating the list of pseudo-words. 6 One might have expected Latin un- words to have been judged as more native than Latin IN- words because of the native prefix un- on the former. Subjects did not, however, indicate any difference in the nativeness of these words. Thus it appears that, even though speakers do seem to have information about the etymology of native/foreign lexical items, such knowledge cannot account for the historical shift of IN- to un- in English, nor does it allow us to

42

Philip Baldi, Victor Broderick and David

Palermo

predict which lexical items are most likely to give up IN- in favor of un- in the future. Nativization may be a random lexical diffusion process.

B) Pseudo-words While only preference responses were obtained for the real words, we have the production responses of experiment #1 as well as the preference responses of experiment #2 for the pseudo-words. The two types of responses to pseudo-words permit us to examine the productivity of the various prefixes independent of lexical specifications, since there would, of course, be no a priori "correct" specification for a pseudo-word. Table 3 shows the mean percentage of each response for the production task and the preference task. Table 3 Percentage of prefix selection for pseudo-words in the production and preference tasks

un-

non-

correct IN-

incorrect IN-

dis-

other

Pseudo-Native Pseudo-Latin

48 35

12 19

14 23

1.5 2.3

13 10

11 10

Total

42

16

18

1.9

12

10

Preference 1st (2nd) Choice Pseudo-Native Pseudo-Latin Total

48 (31) 32 (37) 40 (34)

Production

9(33) 9(28) 9(30)

35(16) 50(17) 42 (17)

8(20) 9(20) 9(20)

As may be seen in the table, the general trends of the data are similar but there are some differences depending upon whether the subject had to produce the negative prefix or merely select the negative prefix. The first thing to note is that un- and IN- are the most likely responses in both tasks. The likelihood of selecting unin production and preference is about the same for both tasks. Note too that in both cases un- is more likely for pseudo-native words while IN- is more likely for pseudo-Latin words. IN- responses, while the second-most likely response in the production task, are

Preßxal

negation of English adjectives:

dimensions

of productivity

43

much more often selected in the preference task, where the possible responses are offered as alternatives, than in the production task, where the person must come up with the response on his/her own. The reverse of this trend is found with non- responses, where nonis more likely to be produced and less likely to be selected (except as a second choice). Looking now at the incorrect IN- responses, it may be seen that few such errors are made productively, but as many as nine percent of such errors are made in the preference test. Finally, since there are no constraints on responding in the production task, dis- and a variety of other negative prefixes were chosen. Those responses could not be made in the preference task because those alternatives were not available. In light of the latter methodological consideration, the marked similarity in the relative frequency of the responses available in both tasks in quite remarkable. Pseudo-root etymology had a definite effect on the choice of prefix in both tasks. Pseudo-native words tended to attract a disproportionate number of un- responses, whereas pseudo-Latinate words attracted a disproportionate number of IN- responses. For the production task, non- appears to have gravitated somewhat toward the pseudo-Latinate words, but this was not the case in the preference task. The distribution of dis- appears to be unaffected by root etymology, showing that it is at least neutral with respect to nativization. As noted above, the gravitation of IN- toward Latinate pseudoroots is more pronounced in the preference task than in the production task, and this actually results in more IN- responses overall than un- responses. This difference must be an artifact of the multiple choice design: four choices for IN- (in-, im-, il-, ir-) but only one un- choice are given for each item, thereby artifactually inflating the likelihood that subjects will choose a variant of IN(recall that dis- and other responses were not available in this task). However, if we assume that there is a single underlying form for all IN- variants, in the production task only that one underlying form is competing for production. To the extent that the difference between the two tasks demands such an explanation, a model positing a single underlying form (viz., /in-/) is supported. Table 4 shows the percentage each prefix was chosen within each of the five phonological conditions. Comparing the relative productivity of in-, im-, il- and ir- to each other, the most common phonologically appropriate choice was not in- (33%), which cor-

44

Philip Baldi, Victor Broderick and David

Palermo

responds to the presumed based form /in-/, but ir- (42%): see the second last row of table 4. The distribution of in-, im-, il-, and ir- is almost completely predictable from the phonological condition, the pattern being consistent with the historical phonological constraints. On the other hand, the frequency of un-, non- and dis- is largely independent of the phonological condition (see table 3). The rate of phonological errors is quite low. We will briefly examine the error data for real words, followed by an extensive analysis of the pseudo-word error data. Responses to real words in the preference task admit two distinct types of errors, namely, intrusions and phonological errors. First choice errors were mostly intrusions rather than phonological errors: only 21%, 3% and 19% of the errors for Latinate in-, Latinate un-, and native un- roots, respectively, were phonological mismatches. Of the 36 phonological errors made on real words, 19 were in- errors and 16 were il- errors. The remaining error was an ir- error. As will be seen in the following section, this preponderance of in- and il- errors is consistent with the error data for pseudo-words. Most (22) of the 36 IN- errors were made on the Latin IN- roots, which makes sense since IN- responses in general were more common for Latinate IN- roots. Table 4 Percentage of first choice prefixes within each phonological condition

vowel

neutral consonant

/I/

labial

Μ

Total

unnoninilimir-

53.0 8.6 34.0 1.8 0.0 2.3

59.0 4.7 33.0 2.3 0.8 0.3

48.0 7.5 2.1 41.0 0.3 0.3

54.0 9.4 5.0 1.5 30.0 0.5

47.0 4.9 2.1 3.6 0.7 42.0

52.0 7.0 15.4 10.1 6.3 9.0

correct INerror IN-

34.0 4.1

33.0 2.6

41.0 2.8

30.0 7.0

42.0 6.3

36.0 4.7

Condition Prefix

Error predictions The issue of which model to choose in providing an account of the prefixation data is a central concern of this paper. Each of the three

Prefixal

negation

of English adjectives:

dimensions

of productivity

45

models mentioned in the introduction (i.e., the classical assimilation model, the item-and-arrangement model and the analogical construction model) predicts a phonologically appropriate distribution of the four variants of IN- (in-, im-, Π-, ir-). How, then, can a choice be made as to the best model of the three? We believe that evidence about the advantages of one model over the others is to be found in the patterns of errors which subjects made in the two tasks. Thus, while it may be true that each model predicts the same correct distribution of IN-, they differ in the predictions which they make about the relative likelihood of various sorts of deviations from the phonological ideal, i. e., errors. We restrict ourselves in this discussion to errors on pseudo-roots, since the overall number of errors with real words was too small to be significant.

A) Assimilation model The assimilation model assumes a single underlying form /in-/ which must undergo a phonological transformation in specified environments. 7 Specifically, /in-/ will become im- before labials, irbefore r, and //- before /; it remains in- elsewhere. Assimilatory rules reduce the phonological distance between participating segments, in this case adjacent segments. In this model there are three conceivable types of phonological errors: failure to assimilate, unmotivated assimilation, and countermotivated assimilation (dissimilation). Failure to assimilate occurs whenever in- appears in an environment in which assimilation is obligatory, i. e., before labials, r, or /; cf. inmonarial, inrepulient, inlogenic. Unmotivated assimilation occurs when a phonological change in the surface form is made in the absence of a motivating phonological environment, for example, imabfunious, ilexilar, or iraddorsent. Countermotivated assimilation (dissimilation) occurs when phonological distance is increased (e. g., irnexibular, imneculous, ilfissular), or when a phonologically leveling change is usurped (e. g., irlanuous, imrobulous, ilmonarial). There is a definite ranking of these three error types in the assimilation model. Failure to assimilate is most likely to occur, because it involves the simple omission of a rule. Unmotivated assimilation and dissimilation would be less likely, since im-, il- and ir- are not available as underlying forms (only in- is), and must be

46

Philip Baldi, Victor Broderick and David

Palermo

introduced by the misexecution of the assimilation rule. Moreover, the misexecution could not be assimilatory since assimilatory rules are leveling rules. A leveling process cannot account for the change of η to m, I or r, before vowels in particular, which are neutral with respect to these rules.8 Counter-motivated changes (e. g. irnexibular) would be even less likely than unmotivated changes (e. g. ilexilar). Even if part of the phonological distribution of IN- is attributable to deletion rather than assimilation (cf. fn. 8), the same line of argument holds.

B) Item-and-arrangement (matching) model Though we will not propose an explicit item-and-arrangement model, we feel that several predictions are warranted. An arrangement model would predict a high compatability between vowelinitial roots and all four forms of IN-. Vowels go well with any consonant, consequently this model would predict that errors would be just as common as correct responses before vowels. That is, since the item-and-arrangement model has four separate, unrelated lexical entries in-, im-, il- and ir- which are matched up with roots according to euphonic constraints, there is no reason to suggest that inaddorsent, for example, should be any more likely to occur than imaddorsent, iladdorsent, or iraddorsent. Before consonants, this model would predict more resistance to errors. Prefixation before consonants results in consonant clusters, whose euphonic status is inferior to consonant vowel combinations. Moreover, a subset of errors before consonants should be particularly rare, namely, those preceding n, m, I, or r. The correct responses in these environments yield identical consonant pairs (innexibular, immonarial, illogenic, irresular), whereas erroneous responses result in mixed clusters (imnexibular, ilmonarial, irlogenic, inresular). The euphonic value of identical consonant pairs exceeds that of mixed clusters, so error responses competing with identical consonant pair responses would be at a disadvantage. Errors before other consonants (viz. c(k), f , p, t, s) would not suffer such a disadvantage and would be expected to be the most common type of error before consonants.

Prefixal negation of English adjectives:

dimensions of productivity

47

C) Analogical model The analogical model depends on the analogical availability of lexical items in the language which are similar in phonological shape to the stimulus item. At present there exists no precise theory of analogical availability; however, a rough index of availability might be the number of comparable lexical items, weighted for familiarity and degree of similarity. For each response this index would be contrasted with the availability index of competing responses. Thus, under this model the phonologically correct form of IN- would be highly available because of the many instances of correct IN- in the lexicon of English. Analogies for erroneous responses are much less common, and invariably involve an inappropriate analysis of the lexical items involved. An analogy for il- before consonants other than / might be made (inappropriately) using words constructed with the prefix(?) ill-. The possible basis for such analogies is the following exhaustive list of words: ill-advised, ill-afford, ill-at-ease ill-conceived, ill-considered, ill-conditioned ill-fated, ill-favored, ill-fame, ill-founded ill-kempt ill-mannered ill-natured, (illness) ill-suited, ill-starred, ill-sorted ill-tempered, ill-treated, ill-timed ill-received, ill-repute

il/a il/c il/f il/k il/m il/n il/s il/t il/r

Thus for the environments represented by the letters c(k),f, m, n, s, t, r, and vowels there is some analogical support for il- errors. However, the analogy requires an erroneous equation of the prefix il- and the prefix(?) ill-. Ill- is such a productive quasi-negative formative that it may be confused with il-, whether or not analogy is at work. Thus, although il- before a non-/ consonant would be the most likely error type under the analogy model, the il-jillconfusion is an alternative source for that particular error under any model. In- errors, earlier referred to as "failure to assimilate" could occur on analogy to the following words:

48

Philip Baldi, Victor Broderick

and David

Palermo

inlaid, inlay, in-law, inlet, inland inmate, inmost inpatient, input inroad, inrush

in/1 in/m in/p in/r

These analogies should not, however, have high availability because the vocabulary is low in similarity to the task items: 1) they do not involve negation; 2) they are not adjectives; 3) the in- is stressed; 4) except for inpatient, the roots are monosyllabic, native-like words. Errors before vowels could be generated on analogy to misparsed words such as imm-obile, ill-ogical and irr-elevant. There is no a priori basis for predicting how frequent such analogies might be relative to il- and in- before consonants, but they are most certainly more likely than the remaining two error types. II- errors before vowels should be especially common, since they receive analogical support both from misparsing such as this, and from il-jill- confusions (ill-advised, ill-afford). Unlike the above three error types, phonological errors with imand ir- have essentially no analogical support. The only example of ir- followed by a non-r consonant is irksome. There are no examples of words having im- followed by a non-labial consonant. Ir- and im- analogies should be non-existent before consonants. Because of the different predictions generated by the three models, it was anticipated that an error analysis would provide evidence favoring one of them as superior. Table 5 presents the error predictions for each model ordered in terms of probability of occurrence and the actual error data obtained in both the pseudoword production conditions of experiment #1 and the first and second preference conditions of experiment #2. It should be noted that errors in this case are confined to allomorphs of IN-, and those errors were more frequent in the multiple-choice preference task than in the production task. Apparently errors are more likely when they are explicitly presented as possible choices. Table 5 is divided into three sections, one for each model. The table is constructed so that within each section, the values within each column should decrease from top to bottom if the model is viable. Also, values within a single row should be more comparable to each other than to values in other rows. The last column summarizes the values in each row in one of two ways. For the arrangement model, the analogy model and for the preference items within the assimilation model, the mean of each row is calculated, since

Prefixal negation of English adjectives: dimensions of productivity c rt ω (Ν

2 > 3 singular > 3 plural, similar to the inverse forms of Algonquian languages (Comrie 1980). The following sets of forms fall outside this general pattern, and provide the data base for the discussion of the body of this paper. First, several verb forms for transitive verbs with first person singular objects, and in Chukchi only for some with first person plural objects, do not take an object suffix, but instead use the prefix ine-jena- (first person singular) or the suffix -tku/-tko (first person plural) in the morphological position usually reserved for derivational affixes (i. e. adjacent to the verb stem); the person-andnumber suffixes then show agreement with the subject, not the object (section 1). Second, Chukchi has a future tense with the tense prefix re-jra- (and often a suffix -q), with similar forms in the Koryak dialects — the existence of a specific tense prefix in a system which otherwise eschews morphological oppositions of tense requires an explanation (section 3). Third, Chukchi and Koryak have a perfect with a morphological composition completely different from that of the other verb forms discussed so far: there is an invariable prefix ge-/ga-, and a series of suffixes that encode primarily the subject of intransitive verbs and the object of transitive verbs (with the result that subjects of transitive verbs are normally not encoded morphologically). Finally, Chukchi, and also apparently the more northerly dialects of Koryak (Alutor and Palana, but not Chavchuven) have an imperfect, formed with the invariable prefix n- and suffixes encoding sometimes the subject, sometimes the object. The analyses of these four exceptional sets of forms

Derivation,

inflection and semantic change in the Chukchi verb paradigm

87

presented below are based on the assumption that these are all later additions to the core system as outlined above.

1. Antipassive and first person objects In Chukchi and Koryak, first person singular direct objects, where the subject is second person or third person singular, are marked not by person-and-number suffixes but rather by the prefix ine-/ ena-, which immediately precedes the verb stem. In Chukchi (except for the Khatyr dialect), but not in Koryak, first person plural objects are likewise encoded by the suffix -tkuj-tko, immediately following the verb stem, when the subject is second person. Otherwise, the morphology of the verb in such cases is constructed exactly as if the verb were intransitive, i. e., the person-and-number suffix encodes the subject rather than the object. Below, some forms of the transitive verb pela- 'leave' are compared with forms of the intransitive verb viri- 'descend': ena-pela-g9e pela-tko-g9e ena-pela-tdk pela-tko-tdk

'you/he left me' 'you left us' 'you-all left me' 'you-all left us'

viri-g?i 'you/he descended' viri-tdk 'you-all descended'

Elsewhere in Chukchi, the affixes ine- and -tku are used as derivational affixes, serving to derive intransitive partners of transitive verbs, the subject of the new intransitive derivative corresponding to the subject (agent) of the transitive verb. If the direct object of the transitive verb remains overtly expressed with the derived intransitive, then it appears as a nondirect object (in the instrumental case). These affixes are thus antipassive markers. As is set out in greater detail in Comrie (1979), the unusual first person object forms set out above can be given a diachronic explanation in terms of reanalysis of the antipassive markers as first person object markers (although it remains to be stipulated arbitrarily that ineis reinterpreted as a first person singular object marker, and -tku as a first person plural object marker). This then accounts for all aspects of the morphology of these forms. Diachronically, they are derived intransitives, therefore naturally both prefix and suffix person-and-number slots encode agreement with the subject, and the object remains uncoded by the normal person-and-number agreement mechanism. The regular subject person-and-number pre-

88

Bernard

Comrie

fix can incidentally be seen in the imperative, where the second person singular has an overt prefix q-, e. g. q-ena-pela-ge 'leave me' (cf. qd-viri-gi 'descend'). Synchronically, however, the first person object forms displayed above cannot be treated as intransitive, not even as intransitive derivatives of a transitive verb. It is quite possible, although unusual except for purposes of emphasis, to include overt pronouns in addition to the person-and-number affixes on the verb, and if one of these verb forms takes an overt pronoun referring to its object, that pronoun stands in the absolutive case while the subject, if expressed, stands in the ergative case, just as with a transitive verb, and quite differently from the case array with a synchronic antipassive (subject in the absolutive, object in the instrumental), as in gdnan gam ena-pela-g?e 'you left me'. The conclusion is thus that the forms cited above for first person objects are explainable diachronically as being derived antipassives, but they have been reanalyzed synchronically with the interpretation of first person object forms, even though the formal device of person-and-number marking here is completely different from the found elsewhere for objects.

2. Desiderative and future Chukchi has a future tense with the prefix re- and, in a number of forms, the suffix -η. Similar formations, taking into account regular phonetic correspondences, are found in the Koryak dialects. In the context of the rest of the Chukchi verb system, it is unusual to find tense marked in this way. Even if one attempted to analyse the future as a mood rather than a tense, the combined use of prefix and suffix would be unusual. An explanation of the anomaly can be found if we note that Chukchi also has a desiderative derivation, utilizing the prefix re- and the suffix -η. The development of future tenses from desideratives is attested in a wide range of languages, including English (will). The basis of our analysis is thus that the desiderative has been reanalyzed as a future. The question now arises as to whether there is, in fact, any justification for saying that Chukchi has distinct desiderative and future, given that the same markers are used for both. Crucially, however, despite the obvious similarity of desiderative and future,

Derivation, inflection and semantic change in the Chukchi verb paradigm

89

and despite their clear common etymological origin, in the modern language for certain forms the desiderative and the future are formally distinct. One distinction is that the desiderative derivative can be conjugated fully as a verb in all moods and both aspects, whereas the future is restricted to two forms, traditionally called Future I and II, distinguished by the continuous aspect suffix -rk(an) on the Future II. There are two more significant distinctions. First, while the desiderative invariably has the suffix η (except for changes due to regular morphophonemic processes), the future lacks this suffix in a wide range of forms. In Chukchi, the nasal suffix is absent in the Future I for intransitive verbs with a singular subject or with a first person plural subject (e. g. ta-reviri-g^e Ί will descend'), and for transitive verbs with the first person singular object suffix -gam, second person singular object suffix -gat, first person plural object suffix -mak, second/third person singular suffix -g^e (this last being possible of course only in forms with the prefix ine- or the suffix -tku) (e. g. ta-ra-pela-gat Ί will leave you'); in the Future II, the nasal suffix is invariably absent with transitive verbs, and is present for intransitive verbs only in the third person plural (e. g. re-viri-rkani-rj-at 'they will descend'). The distribution of the nasal suffix is slightly different in the Koryak dialects, although the same phenomenon of differentiation of desiderative and future is found through partial loss of the nasal suffix in the future. Thus, in the Chavchuven dialect, the nasal suffix is found throughout the Future I of intransitive verbs (though, since its presence leads to loss of some of the person-and-number suffixes, the forms are still sometimes different from those of the desiderative) and in the Future I of transitive verbs except where the object suffix is first person singular -gam, second person singular -ge, or first person plural -mak; it is absent throughout the Future II of both intransitive and transitive verbs. The second distinction concerns the order of the desiderative suffix -η and the aspectual suffix rk(an) (since Koryak does not use the nasal suffix in the Future II, this point obviously applies only to Chukchi). In the desiderative, the desiderative suffix, like other derivational suffixes, comes closer to the verb stem than does the aspect suffix, as in re-viri-ija-rk-at 'they want to descend'. In the corresponding (nondesiderative) future form, however, the nasal suffix follows the aspect suffix, in keeping with the general crosslinguistic tendency for aspect affixes to come closer to the verb than do tense affixes, thus giving re-virirkani-y-at 'they will descend'. (Because of the restrictions on the

90

Bernard

Comrie

occurrence of the nasal suffix in future forms in Chukchi, this is in fact the only minimal pair of this kind.) Thus, we can see the clear origin of the future in the desiderative. However, the two can no longer be identified synchronically. What has happened diachronically is that a single set of forms, the desiderative, has split into two by semantic reanalysis, with distinction of desiderative and future, the basic evidence for this split being the formal divergence between the two sets of forms. A further small piece of evidence for the semantic reanalysis of desiderative as future tense is provided by the relative order of the nasal suffix and aspect suffix in Chukchi, where reanalysis in the direction of tense has led to placement of the suffix after the aspect suffix. The diachronic origin of the future as a desiderative accounts for the anomaly of tense marking, otherwise absent from the Chukchi and Koryak verb paradigms.

3. Possessive and perfect Both Chukchi and Koryak have a perfect (traditionally: Past II), formed with the prefix ge-jga- and a suffix indicating person-andnumber. In both languages, intransitive verbs show agreement with their subject by means of this suffix. In Koryak, transitive verbs show agreement with their object by means of this suffix (and thus show no agreement with their subject). 2 In Chukchi, the same holds true in general, except that for those verb forms given in section 1 as having the prefix ine- or the suffix -tku as marker of first person singular or plural object, the object is encoded in this way in the perfect too, the person-and-number suffix slot then being used to encode the person-and-number of the subject. The normal pattern is thus illustrated by ga-pela-ygdt Ί/he/we/they have left you' (or: 'you have been left'), the exceptional pattern by g-ena-pela-ygdt 'you have left me', ga-pela-tko-ygdt 'you have left us', cf. intransitive geviri-ygdt 'you have descended'. Note that Koryak lacks these forms with ine- in the perfect even though, for the forms in ine- discussed in section 1, it evinces the same pattern as Chukchi. Presumably, the occurrence of these ine- and -tku forms in the Chukchi perfect is an extension from the more typical aspect-moods discussed in section 1, i. e. is to be treated as an analogical innovation. In formation, this perfect is identical to the possessive. A noun marked with the prefix ge- and a person-and-number suffix is the

Derivation,

inflection and semantic change in the Chukchi verb paradigm

91

normal way of encoding the fact that the referent of the noun in question belongs to the referent of the suffix, e. g. ga-qaa-ygdt 'you have reindeer'. One feature of this construction is that it has only one person-and-number affixal slot, namely the suffix. When this morphological construction was extended to indicating the perfect, there was therefore no way in which, for transitive verbs, both subject and object could be indicated in the regular way, nor more generally any way in which the person-and-number of a participant could be indicated by means of a prefix. The solution was to use the suffix to encode the person-and-number of one participant only, in fact the subject of intransitive verbs and the object of transitive verbs, thus giving rise to an ergative system of agreement. There are two possible explanations for the choice of an ergative system, rather than, for instance, consistent agreement with the subject, perhaps both of them intervening to produce the observed configuration. First, given the general rule that verb suffixes in Chukchi encode intransitive subjects but transitive objects, this pattern with the perfect could simply be an extension of the pattern found in the more central parts of the system. Second, crosslinguistically there is a strong tendency for perfect to correlate with ergativity, and the Chukchi pattern could therefore be a language-specific manifestation of this universal (itself based in the semantic properties of the perfect; see Comrie 1981). The anomaly of the perfect as regards verb agreement is thus that this form was not originally verbal in formation, and retains a morphological structure reminiscent of its origin rather than of its present function within the verb system. The extension of the ine- and -tku forms in Chukchi can be seen as a step in the direction of integrating the perfect into the more core parts of the verb system.

4. Adjective and imperfect Chukchi has one further anomalous set of verb forms, the imperfect, formed with the prefix n- and person-and-number affixes. (Parallel verb forms are found in the Palana and Alutor dialects of Koryak, although I have not been able to investigate them in detail; they are not found in the Chavchuven dialect, for which see below.) As with the perfect, agreement is possible with only one participant even for transitive verbs. With intransitive verbs, this person-and-

92

Bernard Comrie

number suffix invariably marks agreement with the subject. For transitive verbs, the distribution of agreement is complex, although the principle is quite straightforward: agreement may be with the subject, in which case either the prefix ine- or the suffix -tku is required, or it may be with the object, in which case neither of these affixes occurs. The forms for a transitive verb are set out below: n-ena-pela-ygdm n-ena-pela-ygdt nd-pela-tko-ygdt n-ena-pela-qen n-ena-pela-qenat nd-pela-ygdt nd-pela-more nd-pela-tore n-ena-pela-more n-ena-pela-tore nd-pela-tko-tore nd-pela-ygdm nd-pela-ygdt nd-pela-qen nd-pela-more nd-pela-tore nd-pela-qenat

Ί leave you/him/you-all/them' 'you leave me/him/them' 'you leave us' 'he leaves me/him' 'he leaves them' 'he leaves you' 'he leaves us' 'he leaves you-alF 'we leave you/him/you-all/them' 'you leave me/him/them' 'you leave us' 'they leave me' 'they leave you' 'they leave him' 'they leave us' 'they leave you-all' 'they leave them'

As was noted in Comrie (1979), this system can be regarded as the interaction of two different principles, as far as the occurrence or nonoccurrence of ine-l-tku is concerned. First, every combination of subject and object which requires ine- or -tku by the discussion of section 1 will also take ine- or -tku here. Second, and subordinate to this first principle, the verb must agree with the participant highest on the person-and-number hierarchy 1/2 > 3 singular > 3 plural; where both subject and object are first or second person, agreement is with the subject. Where agreement is with the subject, the prefix ine- is required. In a sense, then, agreement in the imperfect is on an ergative basis, as with suffixal agreement in Chukchi in general, although the manifestation of object agreement is blocked in a number of instances by use of the prefix ine- which, as in the discussion of section 1, gives rise to subject agreement, as if the verb form were intransitive. Just as in section 1, although these verb forms look intransitive, they behave syntactically as transitive, as can be seen by inserting overt pronouns, e. g. gdmnan

Derivation,

inflection and semantic change in the Chukchi verb paradigm

93

gat n-ena-pela-ygdm Ί leave you', where the subject is in the ergative and the object in the absolutive. Summarizing agreement in the imperfect, especially in so far as the second principle noted above is relevant, we can say that agreement is with the participant highest on the person-and-number hierarchy; if this participant is the subject, then the ine- prefix must be used to generate a morphological form (intransitive) which permits suffixal subject agreement. The behavior of the third person is, incidentally, precisely the same as is found with inverse forms: third person singular is higher than third person plural, so for third person singular acting on third person plural we have direct forms in the core paradigm and agreement with the subject in the imperfect; for third person plural acting on third person singular we have inverse forms in the core paradigm and agreement with the object in the imperfect; third person singular acting on third person singular is treated as direct in the core paradigm and triggers subject agreement in the imperfect (seen in the use of ine-); third person plural acting on third person plural is treated as inverse in the core paradigm and triggers object agreement in the imperfect (no ine-). These last observations extend the discussion in Comrie (1979), where this parallelism between imperfect and inverse was not noted. In form, the imperfect is identical to the basic adjectival formation in Chukchi, which has an invariable prefix n- and a suffix agreeing in person-and-number with the entity having the property referred to by the stem, e. g. n-erme-ygdm Ί am strong', n-erme-qin 'he is strong'. The unusual nature of the agreement in the imperfect, as compared with the core paradigm, is again (as with the perfect) explainable diachronically as a residue of the agreement pattern of the adjective, which provides only a single, suffixal slot for personand-number agreement. Unlike the perfect, with the imperfect there is no reason other than the general pattern of suffixal agreement in Chukchi to expect an ergative pattern of agreement, and the widespread use of ine- to permit, in effect, agreement with transitive subjects reflects this — there is no a priori reason for subject or object agreement to be preferred, and in fact agreement is basically with the participant highest on the person-and-number hierarchy. Although the Chavchuven dialect of Koryak does not have an imperfect formed in as systematic a way as that of Chukchi, it does have deverbal adjectives with the prefix n- and person-and-number suffixes (the unmarked one being third person -qin (et)), as discussed by Zukova (1972: 147-151). For intransitive verbs, the resulting

94

Bernard Comrie

derivative always attributes a property to the subject, e. g. m-yiijeqin 'flying'. For transitive verbs, however, the picture is rather more complex: some such forms attribute a property to the subject, e. g. m-vaqla-qen 'demanding', while others attribute a property to the object, e. g. m-nvat-qen 'foldable' (or 'folding', as in 'folding chair', since the English present participle evinces a similar fluidity). This is the kind of lack of systematicity that so often characterizes derivational morphology. The most probable relation between the Koryak and Chukchi uses of these forms is thus as follows: The Koryak system is more archaic, and presents a derivational process for deriving adjectives from verbs; the syntactic relation between the subject of the adjective and the participants of the corresponding transitive verb is fluid, presumably each individual lexical item being given the meaning that is most useful for that lexical item. In Chukchi, this formation has shifted from the domain of derivational morphology to that of inflectional morphology, with concomitant tightening up of the syntactic relations involved, the precise mechanism being that outlined above, whereby the verb encodes either subject or object in its affixal morphology, primarily in accordance with the person-and-number hierarchy, although the argument not encoded in the verb morphology can be expressed as an independent noun phrase. This is the kind of systematicity that one expects to find in an inflectional system.

5. Conclusions In this paper, a number of apparently anomalous sets of verb forms in Chukchi have been examined. All cases involve the shift of a derivational process into the inflectional morphology, leading to forms that are anomalous given the rest of the inflectional morphology (though they can readily be explained in terms of their derivational origin), but which nonetheless show clear signs of having been integrated syntactically and semantically, and even to a limited extent morphologically, into the verb system as a whole.

Notes 1. The following sources of verb paradigms have been used: Chukchi: Moll — Inenlikej (1957), Skorik (1977); Koryak: Chavchuven: Moll (1960), Zukova

Derivation, inflection and semantic change in the Chukchi verb paradigm

95

(1972); Palana: Zukova (1980); Alutor: Mel'cuk (1977). Note that Alutor is sometimes considered a separate language. Morpheme alternants separated by / are vowel-harmony variants. 2. Since a case can be made for analyzing these forms as passive, a more accurate statement would be that the perfect of (lexically) transitive verbs shows agreement with the Ρ (patient), not the A (agent).

References Comrie, Bernard 1979 "Degrees of ergativity: some Chukchee evidence", Ergativity: towards a theory of grammatical relations, edited by Frans Plank (London: Academic Press), 219-240. 1980 "Inverse verb forms in Siberia: evidence from Chukchee, Koryak, and Kamchadal", Folia Linguistica Historica 1: 61-74. 1981 "Aspect and voice: some reflections on perfect and passive", Tense and aspect ( = Syntax and Semantics 14), edited by Philip Tedeschi and Annie Zaenen (New York: Academic Press), 65-78. Mel'cuk, I. A. 1973 Model' sprjazenija ν aljutorskom jazyke, 2 volumes ( = Problemnaja gruppa po eksperimental'noj i prikladnoj lingvistike, Predvaritel'nye publikacii, vypusk 45-46 (Moscow: Institut russkogo jazyka). Moll, T. A. 1960 Korjaksko-russkijslovar', edited by I. S. Vdovin (Leningrad: Ucpedgiz). Moll, T. A. and Inenlikej, P. I. 1957 Cukotsko-russkijslovar', edited by P. Ja. Shorik (Leningrad: Ucpedgiz). Skorik, P. Ja. 1961-1977 Grammatika cukotskogo jazyka, 2 volumes (Moscow-Leningrad: Izdvo AN SSSR, and Leningrad: "Nauka"). Zukova, A. N. 1972 Grammatika korjakskogo jazyka: fonetika, morfologija (Leningrad: "Nauka"). 1980 Jazyk palanskix korjakov (Leningrad: "Nauka").

WOLFGANG U. DRESSLER

Suppletion in word-formation

§ 1 A model of Natural Morphology as I espouse it1 consists of 3 main parts: 1) a theory of universal morphological naturalness (or markedness) from which universal scales are deduced for each parameter of morphological naturalness; 2) a typological theory where I elaborate on Skalicka (1979); 3) a theory of language specific system adequacy,2 supplemented by a diachronic investigation of "historical accidents" which resulted in morphologically unnatural phenomena. In my opinion, one of the strategic aims of Natural Morphology in its present stage of development must consist in tackling the "most unnatural" morphological phenomena which seem to present scandalous counter-examples against any theory of morphological naturalness. One such scandal is suppletion,3 a morphological technique which is most or very unnatural on two parameters (and scales) of morphological naturalness, viz., the parameters/scales of morphotactic transparency and of morphological diagrammaticity. § 2 Morphotactic transparency is more natural/ less marked than morphotactic opacity. This follows from the semiotic4 principles of semiotic transparency (cf. Koj 1979) and has found empirical support in psycholinguistic experiments (cf. MacKay 1978). On the following eight-point-scale of morphotactic transparency3 I is most natural, VIII is least natural. On this scale on morphotactic transparency suppletion, especially strong suppletion is most unnatural, i. e. most opaque. Suppletion is defined by Mel'cuk (1976: 50) as "the relationship between any linguistic units A and Β which meet the following condition: the semiotic distinction between A and Β is regular, while the formal distinction between them is not regular". Formal regularity (including subregularity) is expressed, in my model, either by 1) spellout MRs and PRs as in E excite -{-merit (I of the scale of morphotactic transparency) where PRs differentiate word-final and syllable-final /

98

Wolfgang U.

Dressier

I

II

III

IV

Intrinsic allophonic PRs

PRs intervene e. g. resyllabification

neutralizing PRs e. g. flapping

MPRs (no fusion) velar softening

excite* + merit excite

exis^t + ence exist

rid + er (American) ride

electric + ity electric

V

VI

VII

VIII

MPRs with fusion

M R s intervene e. g. G V S

weak suppletion

strong suppletion (no rules!)

conclusion conclude

decision decide

childr + en child

be, am, are, is, was

t/ very slightly, or as in exist+ ence,W where a PR of resyllabification intervenes, or in rid-er^m\ where in some English dialects a neutralizing PR of flapping intervenes in addition, 2) or by a spell-out MR and MPRs such as velar softening in electric-ity (IV on the scale), in addition to a PR of resyllabification or as in conclusion (V) where the final /d/ of conclude and the initial glide of the suffix are fused. 3) or by spell-out MRs and allomorphic MRs such as the allomorphic MR of monophthongization which relates dec[i]sion to dec[ai]de, in addition to the MPRs and PRs referred to above. The relation (VII) of ch[ai]ld to ch[i]ld[r] -en exemplifies weak suppletion (alternation of single segments), the relation (VIII on the scale) of be W (am) W are W was W (were) W is strong suppletion (alternation of whole stems). It can be easily shown that the distinction between weak and strong suppletion is gradual, (cf. § 9) e. g. in childr-en one phoneme is added 6 ; one phoneme is replaced (irregularly) in science scient-ist and in has vs. have, whereas more change is involved in Glasgow -* Glaswegian·, was and is have the inflectional ending -s in common (although in highly irregular ways), was and were show an affinity in their "stems" whereas am and is do not etc. To differentiate this continuum here would fall out of the scope of this symposium (cf. Dressier, in preparation). § 3 The scale of morphological diagrammaticity derives from the concept of inconicity.7 One form of this scale is: I) diagrammatic suffixation: e.g. exite + ment, exist + ence, rid+er, electric + ity.

Suppletion

in word-formation

99

III) partially diagrammatic modification via directional allomorphic MRs, e. g. sing -*• song, foot feet IV) adiagrammaticity: IVa) conversion ("zero affixation"), e. g. to cut a cut IVb) strong suppletion where only a semantic, but no morphotactic direction can be establisched, e. g. are am. In are were one could argue that were has one phoneme more than are, but this is not much diagrammaticity. V) antidiagrammaticity in subtraction as in German dialectal hond 'dog', pi. hon. Now we can supplement II, a hybrid between I and III. II) suffixation plus modification as in IIa) suffixation and opaque (fusing) MPRs, M R s as in conclusion, decision lib) weak suppletion as in childr + en II c) subtypes of strong suppletion as in is where the suffix -s could be identified. There is an essential distinction between I, and II—V: In I the stem of the base form can be easily identified in the derived form, and this is the basis for comparing addition of meaning with addition fo form. Increasing morphotactic opacitiy threatens to destroy the morphological basis for the comparison. Thus the scales of diagrammaticity and morphotactic transparency partially converge — which is to be expected, because the principle of semiotic transparency is connected with the principle of iconicity (cf. Mayerthaler 1981). Nevertheless suppletion is rather or very unnatural on the scale of diagrammaticity as well. § 4 One prediction of Natural Morphology is that the more unnatural a morphological technique is on a scale of naturalness, the rarer it should be in the languages of the world. The most unnatural techniques (e. g. suppletion) should be non-existent or peripheral/ marginal. This is clearly borne out for suppletion. But why does suppletion occur at all? Why do not languages "select" only the most natural solutions of each parameter/scale? The main answer is that many parameters of naturalness are in conflict with one another so that all languages must contain both rather/very natural techniques of some parameters and rather/very unnatural tech-

100

Wolfgang U.

Dressler

niques from others. We will discuss these parameters which conflict with transparency (§ 2) and diagrammaticity (§ 3) in § lOff. § 5 In my view a language type is constituted by a specific choice of (very) natural techniques from some parameters and unnatural ones from others. Certain parameters are higher valued in one language type than others, and it is predicted that from highly valued parameters (nearly) only (very) natural techniques 8 are selected. E. g. the agglutinating type highly values diagrammaticity, semantic and morphotactic transparency whereas the inflecting type does so much less. Therefore we can predict that agglutinating languages show very little suppletion if at all, at least much less than inflecting languages. This prediction is spectacularily borne out. Within inflecting languages word formation is often more agglutinating than inflection (cf. Skalicka 1979). Thus we predict that there should be less suppletion in word formation than in inflection. 9 §6 A corollary of this prediction is the following diachronic prediction: If a language changes from the agglutinating to the inflecting type, the number of suppletions is expected to increase. This seems to be borne out as well: E. g., among Baltic-Finnic languages Estonian is generally considered to have become much more inflecting than its cognate Suomi-Finnish. And, as predicted there are much more suppletions in Estonian than in Suomi-Finnish. § 7 These explanations of suppletion are general ones. A more specific explanation of suppletion in word formation is diachronic explanation in the sense of "histoire des mots (et formes)", i. e. by tracing the continuous developments of particular suppletions (suppletive alternations) from their historic origins. This type of explanation falls under the notion of "genetic explanation" in science theory 10 and is absolutely necessary within the frame-work of Natural Morphology, viz. in its third part, the analysis of language specific naturalness. It has been deduced from the naturalness scales of transparency and diagrammaticity (§ 2-3) that supppletion should occur on the periphery of morphology. But periphery 11 must not only be characterized in terms of frequency of type, not of token) and exceptionality (irregularity), but also in terms of structural marginal!ty; and it must be subjected to genetic explanation.

Suppletion

in word-formation

101

§ 8 Typical domains of suppletive word formation are adjectival derivations from place names, especially inhabitants of cities and countries: Let us first exclude names of foreign countries and cities, since suppletions which occur there, are subject to all sorts of loan phenomena and political factors, and tell us relatively little about the language in question, e. g. weak suppletion in Fr. Londres -* adj.: London-ien = It. Londra -*• Londin-ese\ Fr. Madrid —• Madrilene = It. Madrid Madril-eno; stronger suppletion in Fr. Madagascar —> Malgache, Made-gasse = G. Madagaskar —*• Madegass-e, adj. Madegass-isch, = E. Madagascar -*• Mal(a)gasy, (weaker) Madegasc-art; Hung. Ausztria -> Oszträk 'Austrian' etc.; strong suppletion in It. Ceylon singal-ese = Fr. Ceylan -> Cingalais = G. Ceylon/Sri Lanka —• adj. (regular) Ceylon-esisch, singalesisch = Hung. Ceylon —• szingaliz, (regular) ceylon-i, Pol. Wggry Madziar, (regular base form Wggier) etc. E. g., Hungarian Oszträk has been borrowed from older German Starrichi (> G. Österreich), whereas Ausztria is a humanist, latinate form. Notice that even here agglutinating languages (e. g. Hungarian, SuomiFinnish) seem to tolerate less suppletion than inflecting languages. For Suomi-Finnish A. Ahlqvist and K. Sajavaara (personal communication) pointed only to the Swedish name Göteborg 'Gothenburg' and its fennicized derivative Göötepori-lainen 'Gothenburger', an example of weak suppletion. N o such examples occur in Turkish. § 9 As to indigenous geograhic suppletions the agglutinating languages Hungarian, Suomi-Finnish, Turkish do not have any suppletion. For some inflecting languages take weak suppletion in Fr. Reims [Res] Rem-ois [Remwa], Blois [blwa] -> Blis + ois [blezwa], Epinal [epinal] Spinal-ien etc. or in It. Arezzo -*• Aret-ino, Rieti Reatino, (regular) Riet-ino, Nepi —• Neves-ino, (regular) Nep-ese etc. Stronger suppletion in Fr. Besanfon [bczasö] -> Bisont-in [bizöte], Arras [aRa] —• Artes-ien [aRtezie], Saint-Ιό —• Laudin-ien [lodinie] etc. or in It. Rovigo Rodig-ino, (regular) Rovig-otto, Susa Segus-ino, Gubbio -> Eugub-ino, (regular) Gubb-ino etc. or in E. Glasgow —• Glasweg-ian, Liverpool Liverpudl-ian. Strong suppletion in Fr. Le Puy ->· Anicien, Ponot, Saint-Etienne ->• Stephan-ois. Fontainebleau Bellifontain etc. It. otranto Idrunt-ino, (regular) Otrant-ino, Monterotondo -> Eret-ino, (weaker) Monterond-ese, chieti Teat-ino, (regular) Chiet-ino etc. More

102

Wolfgang U. Dressler

suppletive Italian forms can be easily found in Cappello — Tagiavini 1981. Again it must be noted (cf. § 2) that there is no ready way of grading (not to speak of quantifying) strength of suppletion. Obviously it is inverse to the degree of similarity between the stemforms of the base and of the derivative. The number of phonemes added/subtracted/replaced plays a role in establishing the degree of similarity between the suppletive forms, but also the respective similarity of phonemes replacing one another and the position of replaced/added/deleted phonemes within the stem, because its beginning is more salient than its end. §10 Some of these suppletions are due to the interaction of morphology and phonology, i. e. natural phonological process are applied, but the resulting PRs were then morphologized and lexicalized, e. g. the suppletion of It. Arezzo -> Aret-ino goes back to assibilation before glide in Late Latin as applied to lat. Arretium Arretinus. §11 But most of the suppletions are due to learned Latinate substitutions of popular forms, e. g. It. Chiet-ino was replaced by Teat-ino, the revived form of Lat. Teat-inus adj. cordiale 'cordial'/ cardiaco 'cardiac' (whereas the Old Italian regular derivation corale is now extinct); Fr. chien 'dog' —• adj. canicul-aire (but chiennerie 'avarice'), mer 'sea' -> adj. maritime, coeur 'heart' adj. cord-ial,13 As an example of "non-Latin" repristinations I want to cite Modern Greek examples, such as psömi [psomi] 'bread' -> (regular) psömίζ-ο Ί give bread', psömo-zetes 'beggar', psöm-äs 'baker', psöm-ädiko 'bakery', but suppletive arto-pöles « arto-poios 'baker', artoροιέιοη Ä arto-pöleion 'bakery' etc. derived from learned ärtos 'bread' ( = Ancient Greek ärtos). Now there are, presumably, very many who never use ärtos, psömädiko and even psömäs, but only psömi, artopoiös, artopoieiön etc.; so for these speakers there exists a real suppletion; cf. krasi 'wine' (kraso-p6les 'wine-seller', krasopöleio 'wine-shop'), (oinos 'wine') oino-pdles, oino-pöleion, sc. [ino-polis] etc. from AGk. oinos. All these types of suupletion (but not necessarily all tokens) go back to periods of diglossia. This holds also for the well-known pairs E. sheep — mutton, calf — veal, pig — pork. Only a very wide conception of suppletion includes also these examples. First of all, mutton, veal, pork are affixless — but so is Oszträk, szingalez (§ 8); however these two forms are members of a large class (derivational paradigm) where all other names of inhabitants have suffixes. §13 If we say that F. mer is the suppletive base of adj. maritime and marine 'navy', can we say with the same right that E. maritime and navy are suppletive derivations from E. sea? In other words where does suppletion cross over into non-suppletive semantic relations of near-synonymy, membership in the same semantic field, identity of certain semantic components etc. (cf. Lyons 1977:1 § 8 f.)? Going back to Mel'cuk's (1976: 50) definition of suppletion which necessitates a 'regular semantic distinction between' the supposed base and its putative semantic derivation 14 we can see that F. mer -*• maritime etc. seems to belong to a derivational paradigm 15 whose core is constituted by regular morphological derivations such as F. matin 'morning' -> adj. matin-al, lune 'moon' -» adj. lun-aire.

104

Wolfgang V. Dressler

Moreover F. coeur, mer etc. have no, chien etc. very few regular derivations, but have putative suppletive ones, they cannot occur in compounds, whereas E. sea is productively used in compound formation which substitutes for adjective formation. Finally we must differentiate non-suppletive E. maritime from suppletive Ε. paternal·, patern-al, matern-al, fratern-al are learned doublets of father-ly, mother-ly, brother-ly with a parallel, i. e. regular meaning difference both in relation to father etc. and to fatherly etc. And there is regular suffixation with E. -al. In other words French cord-ial, canicul-aire, and to a lesser extent, maritime on the one hand E."patern-al, matern-al, fratern-al on the other hand, belong to a derivational paradigm, whereas E. maritime does not — unless we want to weaken the concept of paradigm 16 to such an extent that it can mean mere semantical relatedness. Since the paradigm and paradigm distinctions play a smaller role in word formation than in inflection, and in agglutinating languages than in inflecting ones, we find another reason for the smaller frequency of suppletion in word formation and in agglutinating languages (cf. § 5). §14 Next, following Plank (1981: 96 ff.) I want to discuss suppletion in feminine motion: In English and German (and many other languages) we find the following derivational paradigm which betrays some male chauvinism: lion tiger wolf bear

-> lion-ess -*• tigr-ess -* (wolf-ess), she-wolf she-bear

Löwe Tiger Wolf Bär

-*• Löw-in Tiger-in -> Wölf-in -> Bär-in

There are more derivations in German than in English because in German not only grammatical gender is more important than in English, but also the derivational expression of sex. In contrast we have the suppletive set: cock bull stallion ram

hen cow mare ewe

Hahn Stier Hengst Widder

Henne Kuh Stute/Mähre Schaf

This set may be called suppletive, because it belongs to a larger set where feminine motion is expressed via suffixation (cf. § 12 f.). This holds at least for German where suffixation with -in is wide-

Suppletion

in word-formation

105

spread; suffixation with -ess is much rarer in English; if more words like wolf-ess died out, we would have the problem of deciding at which point we should consider examples such as bull — cow not as suppletive, but as of the sheep — mutton type (§ 12). One semiotic function of suppletion is that it signals difference. 17 In contrasting the two sets we find suppletion precisely with animals where sex difference is important for man, for peasant economy, and where the two sexes have quite different economic functions and manners of life. In other words, suppletion signals metaphorically a pragmatic difference, and this is pretty stable in diachrony (cf. Plank 1981: 96 ff.). §15 So far our explanation of strong suppletion within the framework of Natural Morphology has been: 1) Metaphorical signalling of pragmatic and social difference (§11,14). 2) Interaction with phonology (application of sound laws) — rare (§10). 3) Historical accidents due to the influence of another language, be it a more prestigious foreign living or dead language (§ 8, 9, 11, 12, 13). But the indication of more prestige is again a metaphorical social signal. 4) Loans from a closely cognate language, with the same functions as strong suppletion (see 2) above). Note that often more than one motivation holds simultaneously for the same example. Such convergent motivation or multicausality is typical for functional explanation (cf. Dressier, to appear). §16 The last and very general area of suppletion is constituted by the derivation of ordinal 18 numbers from cardinal numbers: Some languages have agglutinative ordinal formation throughout, e.g. Turkish bir Ί ' -*• bir-inci, iki '2' -> ik-inci etc., or Vogul, Mongolian, Maori, Aztek. For suppletion, the following universale seem to hold: A) If a language has only one strong suppletive derivation, it is the derivation offirst from one as in German, Dutch (een eerste) Bask, Vogul (as a variant of agglutination), Mongolian (the same), Arabic, Old Egyptian, Igbo, French (un -*• premier, deux-ieme), Koryak. B) If a language has two suppletive derivations, it is one -* first, two second as in most Indo-European languages, Hungarian,

106

Wolfgang U. Dressler

Estonian, Finnish, Zyryen, Accadian, Aramaic, Ethiopian, Algerian Arabic. Then the three generalizations are possible: C) The etymology of first can be traced to words for 'first, before' as in most Indo-European languages, Hungarian, Zyryen, Mongolian (optional terigiin) or to words for 'beginning, end' as in Vogul (optional), Mongolian (optional angxa-dugar), Koryak, Haussa. D) The etymology of second can be traced to words for 'other' as in Polish drugi and in other Slavic languages, Lithuanian, Old Church drugi and in other Slavic languages, Lithuanian, Old Church Slavonic vütorü, Avestan, Vedic (optional), Swedish, Danish, most Celtic languages, Hungarian, Zyryen, Finnish, Estonian, the archaic German variant anderer. Or to words for 'following' such as Latin secundus, AGk deiiteros or for 'nearest' as Cornish nessa. Most of the three etymologies go back to prehistoric times and may betray a very concrete way of number conception (cf. Gonda 1953: 5; Thurnwald 1929), but most of the times the way how semantic affinity turned into suppletion lies in the dark. Antoine Meillet (as cited by Slawski 1956: 169) traced the replacement of Pol. wtory 'second' in the 14th century by drugi 'other' back to expressions such as jeden drugiego (gen.), jeden drugiemu (dative) 'one another'; of course strong suppletion may also be due to loaning, e. g. the Cornish variant secund < E. second. < F. second, Zyryen variant pervoj 'first' < russ. pervyj. E) Weak suppletion is usually restricted to the lowest numbers, and is due to the interaction with phonology (cf. above § 10).19 Thus weak suppletion may become stronger in time as in Latin 20 sex sextus > It. sei -*• sesto, but suppletion also may decrease such as in Latin > French, Germanic languages, Breton. §17 But the question remains why in diachrony suppletion between ordinal and cardinal numbers has not been completely eliminated in favour of transparent and diagrammatic derivations. One reason is — as with most types of suppletion (cf. Dressier, in prep.) — that suppletion occurs within a relatively small class (i. e. numbers 1-10) where regular alternations exert a weaker analogical force on irregular (incl. suppletive) alternations than in larger classes. But the main reason seems to lie in the particular way children learn number names: They don't learn one-first, two-second, three-third etc., but they learn to count the series of cardinals and ordinals

Suppletion

in word-formation

107

separately. Therefore in diachronic change much more analogies occur in adjacent numbers of the same series (see Winter 1969, cf. Szemerenyi 1960) than between a cardinal and its respective ordinal number. Moreover the earlier numbers are learned earlier and are used more frequently, therefore they are better stored in the brain so that it may be more economical to retrieve them directly from the storage than to compute derived numbers via MRs. 2 1 Finally — in contrast to the other ordinal numbers — "second" and especially "first" are often used outside of counting (this is a synchronic motivation: signalling difference, cf. § 14). These are the bases of universals A, Β and generalization E. §18 What 1 have tried to show in this contribution is 1) that and why suppletion is marginal (peripheral) in word formation. The peripheral character of the unnatural class of suppletion fits well to a "natural" approach to morphology, and one might be content with the fact that a theory of Natural Morphology explains the core of more morphotactically transparent and diagrammatic derivations, but not the periphery of opaque and undiagrammatic suppletive derivations. But I have tried to show more than this: 2) that most tokens of suppletions have a synchronic semiotic motivation such as metaphoricity, indexicality, economy (of storage vs. computation), i. e., whereas the core of non-suppletive derivations is explained by the principles of semiotic transparency and iconicity (diagrammaticity), the periphery is explained by less important semiotic concepts: a) metaphors are less natural icons than diagrams, b) indices are less natural signs than icons. Therefore if there is a conflict between diagrammaticity and metaphoricity or between iconicity and indexicality, the more natural principles are expected to be more powerful, i. e. more, if not all, and more essential phenomena should be derived from the more basic principles of naturalness. This is the case with suppletion. Principles of language acquisition (§ 17) come in as intervening variables for explaining the relative diachronic stability of suppletion in certain categories throughout the languages of the world. So much for the explanatory power of the universal subtheory of morphological Naturalness/Markedness. The results of morphological Naturalness/Markedness. The results of this theory can be further refined by the filter of the typological subtheory of

108

Wolfgang U. Dressler

morphological naturalness. In this connection we have 3) predicted the higher occurrence of suppletion in inflecting languages and in inflectional morphology than in agglutinating languages and in derivational morphology (§ 5, 6,13). 4) As to purely language-specific factors, I have tried to show that genetic explanation, i. e., diachronic investigation becomes imperative for the proper understanding of the actual occurrence of suppletive alternations in languages. This may seem as trivially true for diachronic linguistics in general. But what I specifically mean here is to trace the genesis of specific instances of suppletion and fall outside of the scope of universal and typological properties of language. Of course, these properties always explain only classes, not instances of phenomena, but in the case of suppletion we have seen the relevance of quite peculiar historic situations such as diglossia (§ 12).

Notes 1. Dressier (1977; 1982 a; b; 1983 in press); Dressier — Acson (in press); KilaniSchoch (to appear); Dressier (in prep.); cf. Mayerthaler (1981); Wurzel (1980 in press); Dressier — Mayerthaler — Panagl — Wurzel (in prep.). 2. For inflectional morphology see particularly Wurzel (in press). 3. Mel'cuk (1976a, b; 1982: llOff.); Mayerthaler (1977: 138ff„ 179; 1981: 39, 141 ff„ 182); Lyons (1977: II § 10.1., 13.4.); Rudes (1980); Plank (1981); Werner (1977); Schwarze (1970); Markarjan (1970: 320 ff.); Pohl (to appear); Dressier (to appear, in prep.). 4. The basis of semiotics as used here is Peirce (1965). 5. PR = phonological rule, MR = morphological rule (inflectional or derivational), MPR = morphonological rule. 6. Notice that the "monophthongization" in ch[ijldr-en is irregular, i. e., suppletive, whereas an allomorphic MR of monophthongization can be established for wordpairs with English 'Great Vowel Shift' such as decide — decision', line — linear, crime — criminal, bile — bilious, biliary, whereas similar "monophthongizations" in non-Latinate derivations are more irregular, cf. wild — wilderness, wide — width, five — fifth, fifteen, fifty, but nine — ninth, nineteen, ninety. On the differentiation of PRs, MPRs, and MRs see Dressier (1981, to appear). 7. Cf. Mayerthaler (1980; 1981); Wurzel (1980); Dressler (1982 a; b; in press a; b); Dressier — Acson (in press); Kilani-Schoch — Dressier (to appear); Dressier — Mayerthaler — Panagl — Wurzel (in prep.). 8. In the construct of an ideal language type "only very natural techniques" are selected, but since natural languages do not coincide with an ideal construct, "nearly only natural techniques" is a more realistic formulation. 9. Provided that word formation is homogeneous within the respective language; English word formation is not: Germanic derivational morphology is agglutinating, thus nearly without suppletion (e. g., Glasgow — Glaswegian), Latinate

Suppletion in word-formation

10. 11.

12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

109

derivational morphology is of the inflecting type, thus suppletion is to be expected, cf. § 13. See Stegmüller (1969: 117 ff.); cf. Stegmüller II (1979: 134ff.). Cf. similar notions in the Prague School of Structuralism (e. g., volume 2 (1966) of the Travaux Linguistiques de Prague is devoted to this subject) and in generative core grammar (cf. Chomsky 1981; and the contributions of D. Lightfoot and L. White in the same volume). (1895: 62 ff.); cf. Schwarze's (1970: 29 ff.) term "Übersetzungssuppletion". Baudouin de Courtenay (1895: 63); Lightner (1972; 1974); Schwarze (1970: 23); Beard (1981: 76, 278 f.); Dell - Selkirk (1978). Cf. Beard (1981, in press); for directionality cf. Matthews (1974: 132). Cf. Zemskaja (1979); Guilbert (1975). Seiler (1966); Kastovsky (1982: 21 ff., 73 ff.); Dressier - Mayerthaler - Panagl — Wurzel (to appear). Cf. Anttila (1975 a; b); Mayerthaler (1981: 39). Less so in other number series, e. g., fraction numbers like two -» half. Cf. the larger (near-!) universal 39 in Greenberg (1978: III: 281): "The degree of morphological fusion varies inversely with the size of the numerical value". Lat. quinque -* quintus > It. cinque -»· quinto shows an increase in suppletion due to the latinizing form quinto. Cf. Mayerthaler (1977: 139ff.); for a computation model cf. Koskenniemi (1983). As to diachrony, frequent words resist analogy better than less frequent ones, cf. Tauli (1958: 24).

References Anttila, Raimo 1975 a "The indexical element in morphology", Innsbrucker Beiträge zur Sprachwissenschaft, Vorträge 12. 1975 b "The metamorphosis of allomorphs", Lacus 2: 238-249. Beard, Robert 1981 The Indo-European lexicon (Amsterdam: North Holland). in press "Generative Lexicalism", Quaderni di Semantica, volume "Word Formation", edited by J. Alinei. Baudouin de Courtenay, Jan 1895 Versuch einer Theorie phonetischer Allernationen (Strassburg: Trübner). Bily, I. 1979 "Zu einigen Besonderheiten bei der Ableitung von Adjektiven und Bewohnernamen aus russischen Ortsnamen mit adjektivischer Endung", Linguistische Arbeitsbereiche 22: 129-134. Cappello, Teresa — Tagliavani, Carlo 1981 Dizinario degli etnici e dei toponimi taliani (Bolgona: Patron). Chomsky, N o a m 1981 "Principles and parameters in syntactic theory", Explanation in linguistics, edited by Ν. Hornstein & D. Lightfoot (London: Longman), 32-75. Dell, Francois — Selkirk, Elisabeth O. 1978 "On a morphologically governed vowel alternation in French", Recent transformational studies in European languages, edited by J. Keyser (Cambridge: MIT Press), 1-51.

110

Wolfgang U. Dressler

Dressler, Wolfgang 1977 "Elements of a polycentristic theory of word formation", Wiener Linguistische Gazette 15: 15-32. 1981 "Outlines of a model of morphonology", Phonologica 1980, edited by W. Dressler et al. (Innsbruck: Inst. f. Sprachwissenschaft), 113-122. 1982 a "On word formation in Natural Morphology", Wiener Linguistische Gazette 26: 3-14 ( = Reprints of the 13th International Congress of Linguists, Tokyo, 149-155). 1982 b "Zur semiotischen Begründung einer natürlichen Wortbildungslehre", Klagenfurter Beiträge zur Sprachwissenschaft 8: 72-87. 1983 "The predictiveness of Natural Morphology". Henry Sweet Lecture at the LAGB meeting (Sept. 22nd, 1983, Newcastle on Tyne). Written version in prep, for Journal of Linguistics. in press "Subtraction in word formation and its place within a theory of Natural Morphology", Quaderni di Semantica, volume "Word Formation", edited by M. Alinei. in press "Subtraction in a polycentristic theory of Natural Morphology", Rules and the lexicon. Studies in word formation, edited by E. Gussmann (Lublin: Cath. Univ.). to appear Morphonology (Ann Arbor: Karoma Press), in prep. Studies in Natural Morphology (Tübingen: Narr). Dressier, Wolfgang — Acson, Veneeta in press "On the diachrony of subtractive operations: evidence for semiotically based models of Natural Phonology and Natural Morphology from Northern and Anatolian Greek", Proceedings of the VHth International Conference on Historical Linguistics, edited by J. Fisiak (Amsterdam: Benjamins). Dressler, W. U. - Mayerthaler, W. - Panagl, O. - Wurzel, W. U. in prep. Leitmotifs in Natural Morphology (Amsterdam: Benjamins). Gonda, Jan 1953 Reflections on the numerals "one" and "two" in Ancient Indo-European languages (Utrecht: Oosthoeck). Greenberg, Joseph 1978 "Generalizations about numeral systems", Universals of human language, edited by J. Greenberg (Stanford: Univ. Press), III: 249-295. Guilbert, Louis 1975 La creativite lexicale (Paris: Larousse). Haiman, John 1983 "Iconic and economic motivation", Language 59: 781-819. Kastovsky, Dieter 1982 Wortbildung und Semantik (Düsseldorf: Schwan-Bagel). Kilani-Schoch, Marianne — Dressier, Wolfgang U. to appear "Natural Morphology and Classical vs. Tunesian Arabic", Wiener Linguistische Gazette and Studia Grammatyczne. Koj, Leon 1979 "The principle of transparency and semiotic antinomies", Semiotics in Poland (edited by J. Pelc) (Dordrecht: Reidel) 376-406. Koskenniemi, Kimmo 1983 Two-level Morphology. A general computational model for word-form recognition and production. (PhD thesis, Univ. of Helsinki). Lightner, Theodore M. 1972 "Some remarks on exceptions and on coexistent systems in phonology", The Slavic word, edited by D. S. Worth (The Hague: Mouton), 426-436.

Suppletion in word-formation 1974

111

"Preliminary remarks on derivational morphology of French", Actes du I. colloque franco-allemand de grammaire transformationeile, edited by C. Rohrer & N. Ruwet (Tübingen: Niemeyer), 142-164.

Lyons, John 1977 Semantics I, II (Cambridge: Univ. Press). MacKay, Donald G. 1978 "Derivational rules and the internal lexicon", Journal of Verbal Learning and Verbal Behavior 17: 61-71. Markarjan, Ranus A. 1970 Tipy semanticeskogo protivodejstvija ν sfere formoobrazovanija i slovoobrazovanija (Erevan: Izd. Erevanskogo Universiteta). Matthews, Peter H. 1974 Morphology (Cambridge: Univ. Press). Mayerthaler, Willi 1977 Studien zur theoretischen und französischen Morphologie (Tübingen: Niemeyer). 1980 "Morphologischer Ikonismus", Zeitschrift für Semiotikl: 19-37. 1981 Morphologische Natürlichkeit (Wiesbaden: Athenaion). Mel'cuk, Igor A. 1976 a "On suppletion", Linguistics 170: 45-90. 1976 b "Suppletiologische Etüden", Wiener Linguistische Gazette 12: 57-65. 1982 Towards a language of linguistics (München: Fink). Peirce, Charles 1965 Collected papers, edited by Ch. Hartshorne & P. Weiss (Cambridge Univ. Press). Plank, Frans 1981 Morphologische (Ir-) Regularitäten (Tübingen: Narr). Pohl, Alex to appear "Das Personalpronomen im Polnischen und Deutschen", Studio Grammatyczne. Rüdes, Blair A. 1980 "On the nature of verbal suppletion", Linguistics 18: 655-676. Schwarze, Christoph 1970 "Suppletion und Alternanz im Französischen", Linguistische Berichte 6: 21-34. Seiler, Hansjakob 1966 "Das Paradigma in alter und neuer Sicht", Kratylos 11: 190-205. Skalicka, Vladimir 1979 Typologische Studien (Braunschweig: Vieweg). Slawski, Franciszek 1956 Slownik etymologiczny jpzyka polskiego (Krakow: Towarzystwo. Milosniköw Jfzyka Polskiego). Stegmüller, Wolfgang 1979 Hauptströmungen der Gegenwartsphilosophie II (Stuttgart: Kröner). 1969 Das ABC der modernen Logik und Semantik. Der Begriff der Erklärung und seine Spielarten (Berlin: Springer). Szemerenyi, Oswald 1960 Studies in the Indo-European system of numerals (Heidelberg: Winter). Tauli, Valter 1958 The structural tendencies of languages (Helsinki: Suomalainen Tiedeakatemia). Thurnwald, R. 1929 "Zählen", Reallexikon der Vorgeschichte, edited by Max Ebert (Berlin: de Gruyter), 14: 459-479.

112

Wolfgang U. Dressler

Werner, Otmar 1977 "Suppletivwesen durch Lautwandel", Salzburger Beiträge zur Linguistik 3: 269-283. Winter, Werner 1969 "Analogischer Sprachwandel und semantische Struktur", Folia Unguistica 3: 29-45. Wurzel, Wolfgang U. 1980 "Some remarks on the relations between naturalness and typology", Travaux du Cercle Linguistique de Copenhague 20: 103-113. in press "Flexionsmorphologie und Natürlichkeit", Studio Grammatica. Zemskaja, Elena A. 1979 "Unites fundamentales du systeme synchronique de la formation des mots", PICL 12, edited by W. Dressler & W. Meid (Innsbruck: Institut für Sprachwissenschaft), 78-83.

THOMAS FRÄSER

Etymology and the lexical semantics of the Old English preverb be-

1. Most handbooks of the older Germanic languages, following on a tradition which goes back at least to Grimm, have a chapter on word-formation in which the authors provide alphabetical lists of prefixes, each one being attributed a variety of meanings. So the prefix be-, which will be examined here, is ascribed three meanings/ functions by Mitchell (1968: 55): "1) In some words be- is the same as the preposition 'about'. 2) Sometimes it is a privative. 3) It can make an intransitive verb transitive." The same scanty information is provided by the dictionaries of Old English. Thus, for example, Clark Hall and Meritt describe be- first of all as "specializing the meaning of the transitive verb" then as "not altering the meaning". This lexicographical method is followed by authors of monographs, the most complete of which on this prefix is by Lenze (1909). Lenze distinguishes, for the particle, a spatial meaning — or rather a series of spatial meanings (wohin, auf/über, um ... herum, an ... vorbei) — then four main functions: it can transitivize an intransitive verb; it can intensify the meaning of the verb; it has a privative meaning; it forms denominative verbs. Lenze's conclusions, however, are limited, for he makes no attempt to point to relationships between the various meanings or functions and his study is in fact no more than a classification based on the meanings of the simple verbs. In a more recent study Katz (1958), taking as his starting point the observation that the preverb has no obvious independent meaning, bases his analysis on the meanings of the compound verbs, and therefore tells us very little about the particle itself. Only Marchand makes any attempt to explain the roles of be- by a semantic shift undergone by the particle itself. Concerning verbs of the Modern English type besmear, beweep, he writes (Marchand 1969: 146): "It is also possible that the sense development [of the prefix] was from 'around' to 'all around' to 'all over, thoroughly, completely'." On

114

Thomas Fräser

the one hand, therefore, we have those who consider that the prefix is a meaningful element and can extend its meaning; on the other hand, be- must mean nothing, since it does not alter the simplex. Between the two extreme viewpoints there are those who consider that be- has been grammaticalized. From the simple point of view of observation of the facts, all three statements are correct and need not be contradictory, since it is a constant feature of languages to present cases of signs which cover a number of different significates 1 (see, for example, Bybee and Pagliuca, and Guimier, this volume). One of the aims of this paper will be to show how such apparent contradictions, noted but rarely accounted for by lexicographers, can be resolved. 2. Before going on to examine the different meanings of be-, I would like here to enter into some etymological considerations which will highlight the relationships between be- and another Old English particle: ymb-. The original spatial meaning of be("movement up to a point") is to be found in various Indo-European cognates: thus, for example, Latin ob- (cf. movere = "to move, shake" vs. obmovere = "to approach"); or again the Sanskrit preposition/adverb abhi (movement towards); and for Germanic, we can take Gothic bi-. These particles probably derive from a Proto-Indo-European root, *obhi: *obhi

Skt. abhi

Lat. ob-

Ο. Ε. be-

Got. bi-

Figurc 1

Now, Gothic occupies a particular position in the Germanic languages in that a good third of its compounds with bi- express a circular movement (movement around a point). According to Delbrück (1893: 684) this can be explained by the fact that Gothic bi- has a dual origin — on the one hand *obhi, on the other *mbhi. Now the latter particle is also the etymon for Latin ambi-, Greek amphi, and in the Germanic languages Old Norse umb-, Old Saxon umbi-, and Old English ymb-:

Etymology and the lexical semantics of the Old English preverb be-

115

*mbhi

Got. bi-

Ο. E. ymb-

Ο. N. umb-

Lat. ambi

Figure 2

Formally, of course, there is a relationship between *obhi and *mbhi, which has led Meid, for example (1967: 38) to suggest that IndoEuropean formed a root *bhi by aphaeresis from *rribhi\ similarly, Priebsch and Collinson (1966: 253) point out that German be- "goes back to PIE *ipbhi 'around', a word which has given in Germanic both *umbi and, with full stress on the last syllable, *bi. Thus um-, be-, bei- are originally the same word." There are certain problems in accepting some aspects of Meid's and Priebsch and Collinson's point of view, at least as far as the development of meaning is concerned, but there is, I believe, sufficient etymological evidence to allow a comparison between be- and ymb- in Old English. 3. Before examining the relationships between be- and ymb-, a number of preliminary remarks should be made concerning ymb-. First of all it should be noted that this prefix disappeared in the course of the Middle English period, whereas the compound verbs with be- on the other hand had resisted the general disappearance of Germanic preverbs. Yet it might have been expected that the existence of Old Norse compounds with umb(i)- would have at least maintained the ymb- compounds in English. A second point to note is that in Old English ymb- was less productive than beand that it was used mainly with verbs expressing actions that could be directed "around" the direct object. Moreover, the role of ymbwas purely lexical, and unlike be-, it could not be used either to transitivize verbs or to form denominative verbs. If we look at instances of the two prefixes with verbs of movement, it will be seen that whereas ymb- clearly expresses movement around something, be- on the other hand can express movement towards a point. Let us compare, for example, ymbfaran, as in: (1) "He het J^aet faeste land utan ymbfaran, J^aet him mon sceolde an ma healfa on feohtan Jsonne on an." Orosius, 80,26. (He

116

Thomas Fräser

had the fortified land surrounded on all sides, so that they could fight on more than one front.) with befaran in the following instance from the Chronicle 2 : (2) "Jja landesmenn hine beforon innan faere burh and hine ofslogen." Chron. 1068 D. (The citizens came upon him in the town and killed him.) Similarly, ymblcedan in the Vespasian Hymns translates the Latin circumducere: (3) "ymblaedde hine and gelaerde hine and heold hine swe swe sian egan." Hymns 7,19. (He led him round, gave him instruction, considering him as the apple of his eye.) whereas in the Rule of St. Benedict (MS. C. C. D. Ill) we find inducere translated by belcedan: (4) "t>u belaeddest us on grin." (You led us into a trap.) Diagrammatically we can represent the two preverbs as follows: L

L

beFigure 3

As we have in fact two successive interceptions of the same movement, it is not surprising to find cases where be- takes on the meaning of ymb-, as for example: bedelfan and beweorpan in the following: (5) "oJd ic hyne Q)one fictreow) bedelfe and ic hyne beweorpe myd meoxe." St. Luke 13,8. (Till I dig around the fig tree and spread dung around it.) This semantic development 3 seems to have taken place in classical or late Old English. For example, whereas the Canterbury Psalter translates the Latin circumdare by ymbtrymian, in later texts — as in the 11th century translations of the Gospels, — we find betrymian:

Etymology

and the lexical semantics

of the Old English preverb be-

117

(6) "t>yne fynd J?e betrymiaj)." St. Luke 19,43. (Your enemies will cast a trench about you.) Here, however, a not unusual phenomenon is to be observed: this shift of meaning undergone by the prefix be- takes place at the same time as loss of meaning, or dematerialization. I will come back to this phenomenon below with respect to the roles and functions of the be- prefix (see § 4). However, one of the consequences of such weakening of the spatial senses of be- is that compound verbs are occasionally reinforced by an adverb or by a second prefix. There is, for example, a case in the Chronicle where be- with the meaning "around" is reinforced by abutan: (7) "swylce hig woldon J>aes cynges scipa abutan betrymman." Chron. 1052 C. (as if they were about to surround the king's ships.) When be- is weakened from the meaning "towards", on the other hand, the reinforcing particle can be to, in or on. The following is an example from a late 9th century entry (ca. 892) in the Chronicle: (8) "£>a gatu him to belocen haefdon." Chron. 755 A/E. (They had locked the doors on themselves.) It is of course possible here to interpret to as a postposed particle ("him to"); in the following example, however, there is no doubt as to the preverbal character of in-: (9) "t>a £>a duru inbeleac aefter him." Blick. Horn. 217,26. (Then he shut the doors behind them.) Similarly, in Aelfric's Grammar, Latin verbs with the in- prefix are frequently translated by a be- verb reinforced by on: (10) "ic on besette : insero" (166,3) "ic on beslea : incutio" (169,9) "ic on beslea : illido" (171,4) "ic on besende : inmitto" (172,11) On the other hand, in the same text, ymb- apparently retains its full meaning, and we find: (11) "icembedo : circumdo" (139,13) In this last example it is worth drawing attention to the vowel /e/ which appears before the base verb 4 , and which is in fact present

118

Thomas Fräser

sporadically in 10th and l l t h century texts. The Chronicle has one example: (12) "Jjaet land on |?a sae healfe mid scipum ymbe laeg."5 Chron. 1072 E. (With his ships he set up a naval blocade around that land.) Particular attention should be paid to the meaning of the verb in this context: whereas in earlier Old English ymblicgan meant "to surround" in what might be called a "static" sense (i. e. "to lie round"), and belicgan was used in the sense "to surround in order to lay siege to", here we find ymbe licgan with exactly this meaning. Verbs with an ymbe-/embe-/umbe- prefix make their appearance in English up until the 14th century; in Sir Gawain and the Green Knight, for example, the forms umbefolde, umbeweved and umbeteye are to be found, and while there are no corresponding Old English compounds with ymb-, we do have befealdan, bewefan and beteon. Since such forms begin to appear at a time when be- was regularly being reinforced by in, on and to, it would be my contention that they are the result of a fusion between the two prefixes, ymb- and be-, a fusion which could take place all the more easily as the two preverbs had become synonymous 6 . 4. I would like here to go on to look at those cases of be- in which there is no obvious spatial meaning. To explain this I will base my arguments on Gustave Guillaume's theory of internal subduction 7 according to which the notional content of a word can be reduced to a greater or lesser degree; a complete cancellation of this notional content will produce a form unable to exist on its own and which needs to make up for the lexical matter lost during subduction. In the case under consideration, we can take as a starting point the adverb big, which, like all adverbs which have developed into prefixes, expresses a particular idea of movement or spatial position, as in: (13) "f)a eode he {?aer rihte big on sume stowe." Blick. Horn. 221,23. (Then he went and stood immediately alongside in a certain place.) This particular idea is the lexical matter (M) or notional content, which is then provided with the formal characteristics of the adverb (F) before the word is ready to be used at sentence level. In order

Etymology

and the lexical semantics

of the Old English preverb be-

119

for a word to exist, then, the two component parts must be balanced, corresponding to the formula Μ (matter) + F (form) = I (integral) As long as it corresponds to this formula, the adverb, like other predicative parts of speech (noun, adjective, verb) will continue to exist as an independent word. If, however, the word undergoes a certain amount of dematerialization, while remaining formally complete — corresponding to the new formula: M —q + F = I —q (where q represents the quantity of matter lost) — the result will be a word which cannot lead an independent existence and which must must make up for the matter lost by associating with another part of speech, in most cases a noun 8 . In our example, we move from the adverb big to the preposition be, in which the idea of movement is reduced to a somewhat vaguer notion of spatial or temporal relationship. Since the process of subduction is a factor of semantic instability, it is not surprising that dictionaries of Old English should find it difficult to give a modern equivalent of the preposition be. Here, for example, is the entry in Clark Hall and Meritt's dictionary: "(of place) near, in, on, upon, with, along, at, to; (of time) in, about, by, before, while, during; (notion) for, because of, in consideration of, by, by means of, through, in conformity with or imitation of, in comparison with, about, concerning, in reference to." The preposition, however, although dematerialized, continues to exist as a word and is recognizable as such. This is because, in Guillaume's terms, subduction remains immanent to the stematic vector which carries the word in the course of its dematerialization. Moving down the stematic vector, the word will lose more and more of its notional content, but without the latter disappearing entirely. If, however, subduction goes beyond the first vector, it will continue along the astematic vector, which cannot carry the word as such, since its internal structure has been modified. This can be represented by the following figure. What we are left with, once subduction has transcended the stematic vector, is an affix, that is a form, in the position of matter which, in a way, takes over from the matter 9 . In the case under con-

120

Thomas Fräser

F Μ

immanence (stematic vector)

L (limit) Μ

transcendence (astematic vector)

F Figure 4

sideration the prefix be-, the form, "becomes" a movement, and as such is capable of propelling, or pushing the verb to which it is attached. This propulsive action of the preverb (the term is used by Gustave Guillaume in two unpublished lectures given in May 1940 and December 1943, but he never expands on it) can operate both lexically and grammatically. I will concentrate here on the lexical role of be-, but the two main grammatical roles ascribed to it can be mentioned briefly: 1) it forms denominative verbs; this can be explained by the fact that be- propels the noun out of the spatial universe and into the temporal universe, the two universes being conceived in that order (see Joly 1975: 24-27 on this point); 2) secondly, be- forms transitive verbs; without expanding on this difficult problem here, it can be said that under the propulsive action of the preverb the verbal notion is carried to its goal, i. e. the direct object. Frorrt a lexical point of view, the propulsive action of the preverb can also operate in two main ways. First of all, the verb can be "pushed" to the end-limit of its own lexical field. In a figure: L

field 1

(operation) Figure 5

Etymology and the lexical semantics of the Old English preverb be-

121

The result will be an intensification of the notion expressed by the simple verb, or an idea of completion, as in notian as opposed to benotian in the following two examples: (14) "... amang J?am feo t>e we on Jjysum dagum notiaj)." Aelfric, Lives 23,706. (... among the coins we use today.) (15) "hie haefdon ... hiora mete benotodne." Chron. 894 B. (they had used up all their provisions.) With a similar meaning we can oppose brucan ("to use") and bebrucan ("to use to a good purpose"). Intensification can also be observed when be- is associated with verbs expressing visual perception. Thus, simple perception can be expressed by hawian, as in: (16) "... sona swa hi waeron swa gehende J>aet aeg^er heora on ojDer hawede." Chron. 1003 E. (... as soon as they were near enough for each [army] to notice the other.) Behawian, on the other hand, expresses perfect perception: (17) "La J)u liccetere ado aerest ut J)one beam of t>inum agenum eagan and behawa fjonne ^aet |DU ut ado J?aet mot of Jsines broker eagan." St. Matt. 7,5. (You hypocrite. First cast out the beam from your own eye, and then you will see clearly to cast out the mote from your brother's eye.)10 In the same way we can compare sceawian ("to watch") with besceawian, which implies not only the physical activity, but also the mental activity which accompanies the act of watching: (18) "£>u ne besceawest nanes mannes had." St. Matt. 22,16 (You do not take the condition of men into consideration.) It should be noted, however, that if be- insists on the fact that the verb's event is carried to completion, it is by no means a necessary element with verbs of this type since their semantic content implies that the event is seen as complete. If now the propulsive action of the preverb is continued, the verb will be propelled out of its own semantic field and it will then express a notion situated resultatively beyond the first field, as represented by the following figure:

122

Thomas Fräser

(operation)

field 2 (result)

Figure 6

This accounts for the shift in meaning undergone by delfan ("to dig") when be- is prefixed to it: (19) "and hine J?ar let ofslean and deop bedelfan" Chron. 1049 C. (He had him killed and buried deep.) In the same way, whereas sierwan expresses the operation of plotting, the compound expresses a possible result of such an activity: (20) "t>ises geares eac t>a Scottas heora cyng Duncan besyredon and ofslogen." Chron. 1094 E. (In this year also the Scots set a trap for their king Duncan and murdered him.) A third example is the verb wcegan which expresses the notion "to deceive", that is, an active and deliberate operation on the part of the subject. Opposite the simple verb we have bewcegan which expresses a consequence of deception, i. e. disappointment: (21) "swor dryhten sojjfestnisse and ne biwaegde hine ..." Vesp. Ps. 131,11. (The Lord has sworn it [to David] and he will not disappoint him.) (Lat. frustrabitur) Finally there is the case where the result phase is in fact a reversal of the initial operation expressed by the simple verb. So we have, for example, dalan ("to distribute") opposite bedcelan ("to deprive") and rcedan, one meaning of which is "to possess", opposite bercedan ("to dispossess"). There are of course privative verbs in which the particle simply intervenes as an intensifying prefix, as for example bereafian ("to rpb") or bescieran ("to cut off'); in such cases, we have an interception of the movement at L^ whereas in the case of bedcelan the interception is beyond the limit of the verb's semantic field. The majority of such verbs expressing privation are in fact denominative verbs, and here a curious phenomenon should be

Etymology

and the lexical semantics of the Old English preverb be-

123

pointed out: many such denominative verbs can express either the idea of "covering with X" or "providing with X" (where X is the notion expressed by the noun), or else the idea of "depriving of X". The two notions may appear to be contradictory, but in fact they are complementary and are related in a notional chronology, for, in order to deprive someone of something, he must previously have been provided with that thing. In other words, "provision" is a notion which extends from non-possession ( —) to possession ( + ), whereas "privation" is a movement in the opposite direction, from ( + ) to ( —). In a figure: provision

privation +

+

Figure 7

The same prefix can therefore quite easily express the two successive notions, and so we have: (22) "bedicodon s y ^ o n t»a burh uton J)get nan man ne mihte ne inn ne ut." Chron. 1016 E. (They built fortifications in front of the town, so that no one could go in or out.) where the denominative verb expresses provision, as opposed to privation in the following: (23) "Se cynincg het forceorfan his tungan and hine behaettian and his handa forceorfan and eac befotian." Aelfric, Lives 25,115. (The king had his tongue cut out and had him scalped, and had his hands and also his feet cut off.) Such verbs are normally to be found with parts of the body — where provision is presupposed — or else they exist alongside denominative simplex verbs which express the notion of provision, as for example horsian ("to provide with horses") and behorsian ("to take away the horse"). 5. All too often in linguistic studies grammatical morphemes are excluded from semantic analysis because they are considered as meaningless. Such is the case for certain determiners, and even more so for auxiliaries (cf. Joos' "meaningless do") and affixes (as for example in the voluminous literature on ge~). The problems encountered by dictionary-makers in defining such items show just

124

Thomas Fräser

how difficult it is to analyse their semantic contents, and this has resulted, amongst other things, in the often contradictory meanings put forward for preverbs. Such contradictions, or such apparent meaninglessness, exist however only at the surface level of discours, and it should be one of the tasks of the linguist to go beyond this and analyse the underlying meaning in langue.

Notes 1. I have used sign in the Guillaumean — and traditional pre-Saussurean — sense to refer to the physical aspect of the word, and significate, which translates the French signifie, to refer to the mental representation, which may be either lexical or grammatical. 2. Earle and Plummer, in their edition of the Chronicle, give only "to surround" as the meaning of befaran. The Laud Manuscript (E) clearly shows, however, that the movement is towards something: "{ia comon {5a landes menn togeanes him and hine ofslogen." 3. The Latin prefix ob- seems to have followed the same development, and so comes to occupy the same semantic field as amb(i)-. Thus, the verb obire, which in Old Latin and early Classical Latin meant "to approach", takes on the meaning of "to surround" in Ovid: "chlamydem limbus obibat aureus". 4. In the case of embedo the extra syllable could be explained by considering /e/ as a linking vowel between the two occlusives, but not all cases can be explained in this way, as for example the preposition embeutan. 5. Manuscript D of the Chronicle, which is earlier than the Laud MS, has ymb laig, without the intervening vowel. 6. It should be added that the same type of phenomenon is to be observed in Latin, where we find circumobruere = "to cover all over with earth", and in Old High German, as in the following: "den se ein wait hat umbebewahsen" (a wood has grown up all around the lake). 7. Part of this section was first presented in French in Fraser (1977). Guillaume's theory is applied to auxiliaries in "Theorie des auxiliaires et examen des faits connexes" in Guillaume (1964). See also Guimier (this volume) for the same theory applied to the adverbial suffix. 8. Adverbial phrases, or discourse adverbs, are formed in this way in Old English; so we have be nihte (at night), be dale (partly), be fullan (entirely), etc. The second element in each case represents a quantitative equivalent (q) of the matter lost during subduction. 9. Not all Old English prefixes are the result of a transcendent subduction; some of them, such as ofer-, under-, up-, for])- are adverbial in character and are scarcely dematerialized; others, such as cet-, for(e)-, wip-, to- can be either stematic or astematic; a third group is made up of "pure" prefixes, ge-, a-, onand be-. The dematerialization described here has already been observed by some linguists, of course, but what is not always realized is that the loss of matter is the cause of the loss of autonomy of prefixes. For Lindemann (1970: 27), for example, the cause becomes a consequence: "Whereas the preposition generally retained the single concrete meaning of the original adverb, the preverb was more influenced by its environment, yielded some of its concreteness in being fused onto the verb, and consequently became more abstract." (My italics).

Etymology

and the lexical semantics of the Old English preverb be-

125

10. It is interesting to note that in the Greek version of the New Testament the verb is intensified by an adverb — "tote diablepseis". The Vulgate, on which the Old English version is based, has no intensive marker ("... et tunc videbis eiicere festucam de oculo fratris tui."), but the translator seems to have been aware of the value to be given to the verb.

References I. Corpus Earle, John & Plummer, Charles, eds. 1892/1965 Two of the Saxon Chronicles Parallel (Oxford: Oxford University Press). Grünberg, Μ., ed. 1967 The West-Saxon Gospels (Amsterdam: Scheltema & Holkema Ν v.). Logeman, Henri, ed. 1888 The Rule of St. Benet ( = Early English Text Society, Original Series 90). Morris, Richard, ed. 1974-1880 The Blickling Homilies ( = Early English Text Society, Original Series 58, 63, 73). Skeat, Walter W., ed. 1881/1966 Aelfric's Lives of Saints ( = Early English Text Society, Original Series 76, 82, 94, 114). Sweet, Henry, ed. 1883 King Alfred's Orosius ( = Early English Text Society, Original Series 79). 1885/1966 The Oldest English Texts (including the Vespasian Psalter and Hymns) ( = Early English Text Society, Original Series 83). Zupitza, Julius, ed. 1880 Aelfrics Grammatik und Glossar (Berlin). II. Works

consulted

Clark Hall, John R. - Meritt, Herbert D. 1894/1966 A Concise Anglo-Saxon Dictionary (Cambridge: Cambridge University Press). Delbrück, Berthold 1893 Vergleichende Syntax der Indogermanischen Sprachen (Strasbourg: Trübner) Fräser, Thomas 1977 "Les roles du preverbe en vieil-anglais", Actes du Colloque "Linguistique et philologie" edited by Danielle Buschinger (Amiens: Centre d'Etudes Medievales, distribue par Paris: Champion). Guillaume, Gustave 1964/1973 Langage et science du langage (Paris: Nizet, Quebec: Presses de l'Universite Laval). Joly, Andre 1975 La negation verbale en anglais moderne (unpublished thesis). Katz, R. 1958 "Razvitie proizvodnyx glagolov s prefiksom be- ν anglijskom jazkye", Issledovanija po anglijskoj filologii I: 99-121.

126

Thomas Fräser

Lenze, Josef 1909 Das Präfix bi- in der altenglischen Nominal- und Verbalkomposition, mit gelegentlicher Berücksichtigung der anderen germanischen Dialekte (Dissertation) (Kiel: R. Cordes). Lindemann, J. W. Richard 1970 Old English preverbal Ge-: its meaning (Charlottesville: University Press of Virginia). Marchand, Hans 1969 The categories and types of present-day English word formation: a synchronic-diachronic approach (München: C. H. Beck). Meid, Wolfgang 1967 Germanische Sprachwissenschaft III: Wortbildungslehre (Berlin: Walter de Gruyter). Mitchell, Bruce 1968 A guide to Old English (2nd edt.) (Oxford: Basil Blackwell). Priebsch, R. — Collinson, William Edward 1966 The German language (6th edt.) (London: Faber & Faber).

DIRK GEERAERTS

Cognitive restrictions on the structure of semantic change

1. Introduction Recent work in linguistic semantics suggests that the gulf between cognitive psychology and the study of natural language is less unbridgeable than it seems in transformational grammar, which simply proclaims itself an independent cognitive discipline and thenceforth feels itself entitled to ignore psycholinguistics as a whole. A number of studies by linguists such as Lakofif (1977, 1982), Langacker (1982, 1983), Fillmore (1977, 1982), Lindner (1981) and Brugman (1981) argue that the prototypical view of categorization developed in psycholinguistics by Eleonor Rosch and her colleagues (Rosch 1975, 1977, 1978; Rosch - Mervis 1975, 1981) presents an adequate framework for the description of natural language. These studies being restricted to the synchronic aspects of linguistic structure, I would like to show that by and large prototype theory also adequately characterizes the diachronic structure of lexical times. This conclusion, which will be reached in paragraph 3 on the basis of a case study carried out in paragraph 2, will be deepened in two ways. First, a theoretical explanation of the descriptive correspondence between the diachronic data and the synchronic conception of categorization will be given in paragraph 4, using functional requirements on cognition as a basis. The theoretical import of this explanation is one of the basic aspects of this paper: entirely in accordance with prototype theory's rejection of absolute dichotomies, diachrony and synchrony will appear to be much more intimately entwined than is usually assumed by (neo)structuralist conceptions of language; moreover, this entanglement will be explained as an immediate consequence of the cognitive function of language. Second, paragraph 5 contains a defence of the prototypical freedom to maximize polysemy against Goossens' structuralist claim (1963) that languages tend to minimize polysemy.

128

Dirk

Geeraerts

To round off this introduction, I would like to mention that the case study presented in paragraph 2 has been published in Dutch with slightly different theoretical emphasis (1984). An additional example of a prototypical semantic development is given in Geeraerts (1983a). My (1983b) article discusses the importance of prototype theory for historical lexicography, and (1983 c) explores the possibility of incorporating prototype theory into a functional classification of the mechanisms of semantic change. Within this article, paragraphs 3 and 4 contain the gist of the argument; readers who are less interested in the empirical details of the example discussed may skip paragraph 2.

2. A case study: Dutch type Dutch type is a loan word taken over from French; French type goes back to Latin typus, meaning 'model, picture, symbol', and hence to Greek typos, which is derived from typtö, 'to hit', and which has the meaning '(impression produced by a) blow; sign, picture, model'. The first quotation of Dutch type to be found in the sources of the Woordenboek der Nederlandsche Taal1 dates to 1816; in 1824 it occurs for the first time in a dictionary (Weiland). It gradually replaces the Latin loan word typus, for which the W. Ν. T. has a first quotation of 1658 and a final one from 1866. These etymological characteristics of type mean that we will have to allow for semantic borrowing as a factor determining the senses of type. In particular, it should be noted that the older Dutch typus as well as French type exhibit two figuratively related meanings: on the one hand, a material object used for printing or pressing images on or into paper, wax, metal and so on (and the impression created by that object); on the other hand, an abstract characteristic such as may be produced by an imaginary printing-form or pressingform. I will therefore assume that type, at the time when it is borrowed from French and when it begins to replace typus, already carries with it both semantic aspects. In other words, I will not consider the transition from the material sense to the figurative, abstract sense (which are both to be found in the semantic structure of type) as an autonomous development of the lexical item type, but I will assume that it is borrowed from the semantic structures of Dutch typus and French type. As such, I will restrict the attention to the internal developments of both kernel meanings.

Cognitive

restrictions

on the structure

of semantic change

129

As far as the material, 'printing-form/pressing-form' sense is concerned, the following chronologically ordered quotations are relevant 2 . (1) Dat ... zoo spoedig mogelijk zal worden aangemunt een nationale Indische Standpenning, hebbende den naam, en voerende de type van den gewonen Nederlandschen gulden, met dit onderscheid, dat nevens het standbeeld zullen geplaatst worden de letters Ν. Ο., Stbl. ν. Ν.-I. 1816, p. 213. (2) De langzame druk van het XVIIde Deel (der Verhandelingen) (is) te wijten aan den daarbij gebruikten vorm van tijpen ..., waarvan slechts een zoo geringe voorraad bij's Lands Drukkerij voorhanden is, dat telkens niet meer dan een zetter daarmee onledig kan worden gehouden, Verh. Batav. Gen. 17, vi, 18394. (3) Een oude houtsnede plaat waarvan de type op het stedelijk museum aanwezig is, De Lange v. Wijng.-Scheltema, Gouda 3, 110, 1879s. (4) De letters of teekens van allerlei soort (typen) waaruit elk drukwerk gevormd wordt, staan in omgekeerde orde, als Spiegelschrift, op langwerpige Vierkante metalen staafjes, V. d. Meulen, Boekh. en Bibl 107, 18836. (5) Na elk woord, det van een volgend gescheiden dient te zijn, wordt een iets lagere type (het wit) gezet. Daar deze bij het drukken niet door de inktrol aangeraakt kan worden, blijft die plaats open, V. d. Meulen, Het boek 29, 18927. Each of these examples presents the first occurrence of a particular sense or nuance of type. In (1), type is used numismatically with regard to the effigy of a coin; the concept 'printing-form' is metonymically connected with the result of the action of imprinting. This application can also be found in a quotation from 1886, but it does not reappear afterwards. In (2), the concept 'printing-form' is specified in a typographical sense; it refers to typographical types in their material form, i. e. to small blocks of wood or metal that bear the image of a character. This meaning still exists; it is also very common in English (cfr. words such as typewriter). In (3), the concept 'printing-form' is used with regard to a wood-cut (the material object in wood, not the printed result); this nuance has only been found once. (4) and (5) are direct developments from (2). In (5), type still refers to a small bar in wood or metal used for

130

Dirk

Geeraerts

printing, only this time it does not bear the image of a character: it denotes the small blocks that are used for obtaining blank spaces between the words on the printed page. In (4), the notion 'typographical type' is extended from the material objects used for printing to the abstract kinds of characters that are printed; it indicates a fount, a set of characters with similar characteristics. (4) is still used today; (5) has only been found once. Senses (2), (4) and (5) are more closely connected with each other than with (1) and (3); in a dictionary of the average kind, they would probably be represented as a single meaning. Here, it is shown how that meaning can exhibit nuances. In this respect, (5) is particularly important because it shows how the original motivation for calling the things mentioned under (2) types (viz. the fact of bearing the image of a character) can be lost (the thing mentioned under (5) does not bear such an image). Also, it should be mentioned that the distinction between the material interpretation (2) and the abstract interpretation (4) is not always easy to make. (In sentences of the sort This book is set in Linotype Baskerville, both are equally possible.) As far as the abstract, figurative sense of type is concerned, it is convenient to make a distinction between nuances indicating an abstract characteristic or a kind (of people or things) characterized by such a characteristic, and nuances indicating characteristic individuals. With regard to the former, the following quotations have to be taken into account. (6) (Op Marken) heeft zieh ... het waarachtige type onzer oudste voorouders bewaard, in die mannen van meer dan zes voet, met schouders als Atlassen en goudgele lokken, Beets, C. O. 339, 18418. (7) Men (zoude) ... gevaar loopen zieh van de wijze, waarop de natuur bij de volmaking der dieren is te werk gegaan, een geheel verkeerd denkbeeld te vormen, indien men niet twee ... bijzonderheden in aanmerking nam ... Ik bedoel de onderlinge onafhankelijkheid van de volmaking der verschillende stelsels en organen — en de verscheidenheid der grondvormen of typen, Alb. d. Nat. 1854, p. 3769. (8) Möge de koorts ook geen constante type vertoonen, steeds is zij aanwezig, V. Seiten, Acute militair tuberculose 53, 187910. (9) Tegen zijn fors geteekend type (dat van Jamieson) komt het gelaat van Callenbach, predikant te Nijkerk, eigenaardig uit, Pierson, Ο. Tijdgen. 112, 188311.

Cognitive

restrictions

on the structure of semantic change

131

(10) De typen, zooals die zijn vastgesteld door de commissie voor de herziening der ongebouwde eigendommen, die dit jaar hebben voorgelegen, zijn bepaald als volgt: f90, f80, f 7 0 ... (enz.) per H. A. huurwaarde, Onderz. Landb. 1890, XXII, 22 12 .

(11) Ten einde de dynamo direct met de stoommachine te kunnen koppelen, moet het vereischte aantal omwentelingen niet te groot zijn, hetgeen leidt tot ankers van groote middellijn ... Daaruit heeft zieh het vrijwel algemeene huidige type ontwikkeld, V. Capelle, Electr. 122, 190813. The quotation in (6) is a straightforward example of the meaning 'abstract characteristic; set of characteristics and properties that a particular set of people, things etc. have in common; essential form'. (7) is metonymical with regard to (6); it denotes a set of entities that have such a characteristic in common. (6) and (7), which can both be found in present-day Dutch, have the same mutual relationship as intensions and extensions: (6) refers to an abstract characteristic, (7) to the set of entities determined by such a characteristic. The boundary between (6) and (7) is not sharp: at least in (7) one might read 'generic essence, specific characteristic' just as well as 'genus, species' 14 . Unambiguous examples of the intensional reading can be found fairly easily, but this is much more difficult with regard to the extensional reading 15 . The same situation holds for the different senses that develop on the basis of (6) and (7). On the one hand, (8) and (9) are connected with (6); on the other, (10) and (11) can be connected with (6) as well as with (7). In (8), the properties of a fever are described. This kind of usage, which can also be found in the semantic range of typus and which is also mentioned by Van Dale, the standard dictionary of contemporary Dutch, shows how the general meaning of (6) can be specified in a particular medical context. Moreover, contemporary medical dictionaries show that it is not restricted to fevers but that it can refer to different kinds of clinical pictures or syndromes. (9) indicates a particular kind of specific property, viz. the face. Note that this nuance (which can be found unambiguously only in one quotation) does not denote the characteristics of a group of people or a set of things, but characterizes a single individual on the basis of his outward appearance 16 . (10) and (11) have in common that they do not refer to natural categories and categorial properties (as is the case in most instances of (6) and (7)), but rather to artificial

132

Dirk

Geeraerts

categories, viz. to industrial artifacts and products in (11) and to norms that are imposed or agreed upon by people in (10). The meaning illustrated by (10) occurs mostly in a commercial sense, denoting certain fixed norms that traded goods have to conform with, but occurs in a juridical sense in (10), denoting certain norms for imposing taxes. It is found at least until 1921, whereas (11) is still very much alive nowadays, particularly with regard to the different models in which industrial goods are produced (such as different kinds of cars). Finally, the following quotations have to be mentioned with regard to the meaning 'characteristic person or thing'. (12) Houd de twee blinden voor een oogenblik eens voor typen van het toenmalig onderwijs hier te lande, en toets er zijn strekking aan; onder het vrijgeleide der vroomheid, kwam de kunst het huis in — was het wonder dat de eene als de ander zieh doorgaans in den uitslag harer pogingen zag teleurgesteld? Potgieter 1, 111, 1844 17 . (13) De echte rooker! de type van een' rooker! Oct. Kwarto 80, 1853 18 . (14) Romeinsche en Grieksche schoonen boeien als typen, weergaloos en grootsch; Waar Syries jonge maagden bloeien, Daar blijft alle andere schoonheid doodsch, V. Ackere, Winteröl. 26, 1867 19 . (15) Temidden van een wereld vol zelfbehagen, zelfgenoegzaamheid en eigenbaat, en om haar van deze gruwelen te zuiveren en te verlossen, is hij het type, het toonbeeld der zelfverloochening geweest; hij, de Zoon van God, mensch geworden om dit te zijn, Beets, St. Uren 8, 232, 1875 20 . (16) Vreede was een soort van type aan de Academie. Een man, van wien anecdotes te verhalen waren, Quack, Stud. 218, 1881 21 . (17) Dat blijft de eerste bezorgdheid van den Engeischen letterkundige, — dichter, drama- of romanschrijver — het leven nagaan, doorgronden, weergeven, levende beeiden scheppen, typen in 't leven roepen, die voor alle geslachten, als naastbestaanden, zullen meeleven, Belpaire, Landlev. 90, 1902 22 . (18) Ieder mensch heeft wel een tiep, Dat de klap somsbij 'm liep, Speenhoff, Liedjes 1, 35, 190323.

Cognitive

restrictions

on the structure of semantic change

133

(19) Daar de mensch tot een sociaal wezen is gemaakt; daar de verhouding waarin hij tot zijn moeder, die hem teeder opkweekt, Staat, de type en uitdrukking vormt zijner afhankelijke betrekking tot alle medemenschen, poogt hij een samenleven te vestigen op de harmonie van alien, Quack, Soc. 1, 303, 191124. In (12), (13) and (15), we find the sense 'person who is characteristic for something, who embodies something'. What this something can be becomes apparent by comparing those three cases. In (12) a situation, an aspect of state of society is involved. In (13) a group of people with particular characteristics is represented by a specific member of that group. And in (15) a human property, a way of conduct is embodied in one man. Further examples of all these cases have been found; in each of them, the relevant situation, property and so on is mentioned in the context of the sentence. That is not the case in example (14), which is therefore not easy to interpret. I think we should read that Roman and Grecian beauties do not draw out attention primarily as individual personalities, but rather as embodying an abstract ideal; the W. Ν. T. defines type in this example as 'ideal, ideal image'. Because what is (literally) embodied here is still an abstract property, (14) and (15) can be put together. Now, the sense 'person embodying something' gives rise to two different developments. The minor one is illustrated by (19): instead of a person, a thing or a situation is mentioned as characteristic for something. It is not easy to make out whether (19) links up with (12), (13) and (14)—(15). Because there is a thing and not a person that is characterized, (12) exerts a certain amount of influence. Because the thing characterized is of the same nature as the characteristic thing, one might say that the latter characterizes a set of human relations, which points in the direction of (13). Finally, because human relations and modes of conduct are involved, (14)—(15) also have a bearing on (19). Instead of choosing among these possibilities, I will assume that (12), (13) and (14)—(15) alike have to be taken into account to explain the occurrence of (19)25. In (16), (17) and (18) on the other hand, type is used with regard to persons, just as in (12)—(15). In (16), a person that is characteristic as such is named, instead of a person that is characteristic for something: a type in (16) is a remarkable, picturesque, striking, sometimes even odd or strange person. On the one hand, (16) gives rise to (18), where the meaning of type is

134

Dirk Geeraerts

further weakened to the nuance 'person, individual as such' 26 . On the other hand, (16) leads to the sense 'personage in a work of literary art' in the artistic context of (17), albeit with the connotation 'personage that is notable for his specific nature'. Because it is not entirely clear whether a purely individual characteristicness is involved, or whether the personage in question is characteristic for something or other, a direct link of (17) with (12)—(15) cannot be excluded. In figure 1, the semantic development of type is schematically represented. The Arabic numerals correspond with the nuances and senses distinguished in the examples given above; short indications of what these senses amount to are given in the legend (these indications are by no means full-fledged definitions). Along the horizontal axis, the historical continuum is cut up into synchronic slices in order not to overload the picture. Rather pragmatically, these slices comprise ten years. This division does not imply that each decade is a synchronic period in the theoretical, structuralist sense of the term; it is merely a device to make the diagram simpler. At each synchronic point, a new meaning is presented as a development from the situation at the previous point. This yields the oblique lines in the picture; these lines indicate the existing senses out of which a new kind of usage develops. Horizontal lines mean that a particular sense subsists over the period indicated by the line. The dotted lines in the figure indicate relations that are less clear or less outspoken. The Roman numerals indicate the basic meanings of Dutch typus and French type which may have influenced the development of type. In this respect, it may be useful to point out that the figure contains a number of idealizations. In the first place, it is not always easy to determine whether a particular semantic development is an independent process or whether it takes place under the influence of French or Latin. As I mentioned before, I will assume that Dutch type develops its meanings independently, except for its first senses. Such a point of view is not taken without reason. On the one hand, certain developments of Dutch type do not have a clear counterpart in French type or the earlier Dutch typus. (This is e. g. the case with (3), which is not to be found in any of the historical dictionaries of 19th century French). On the other hand, even if certain developments correspond with French, this would probably only mean that the development of French type would exhibit a structure that is similar to the one presented here. (Unfortunately, the materials given in

Cognitive

restrictions

on the structure

of semantic change

135

!\

1810-1820: 1820-1830:

/

1830-1840: 1840-1850: 1850-1860: 1860-1870: 1870-1880:

/

1880-1890: 1890-1900: 1900-1910:

\/ C > CC •

ς ν ς c ν ς C VC CYC

(saibhir) ifochell) (uasal) (deacair)

There is no phonological process, such as assimilation, progressive or regressive, which would account for the feature values of the resolved clusters in (13). Note that to postulate that the lexically stored forms are the unsyncopated ones and to have syncope introduced later, as was considered above, is not a viable solution either as one could not say what value for [palatal] the resulting syncopated clusters would have. For example C V C results in both C C (13)b and C C (13)d. As it stands one must accept that the syncopated forms are stored with additional information on how the clusters resolve. It has been maintained that syncope applies before suffixation. The initial argumentation against the derivation in (9) was that if syncope applies after suffixation there is no way of telling what value for [palatal] the syncopated cluster will have. It might be countered now that my postulation that the lexically stored forms contain information on both the syncopated and the nonsyncopated forms would render my objection to the derivation in (9) invalid. But further considerations show that the later standpoint which I have evolved is correct. These concern the form of the suffix used. The assumed form of the agency suffix is /sdo:r/ (see (4) above). It would be expected then that the base form /foxl/ should have this suffix added as in (14). (14) a b

/foxl/ + /odoir/ */foxlado:r/ (i) by suffixation

Now if one compares Modern Irish productive forms with /odo:r/ one can arrive at a generalization which accounts for the unacceptability of (14)b. Consider (15) a b c d e f

croch crochadoir glic gliceadoir dli dliodoir

/krAx/ /krAxadoir/

mi

/glikadoir/ ΙΦΙ

/41i:do:r/

'gallows' 'hangman' 'clever' 'trickster' 'law' 'lawyer'

Segmental phonology and word-formation in Irish

205

The generalization to be drawn from (15)a-f is that the form /adoir/ is only added to stems which are (i) vocalic or (ii) end in a single segment. But if syncope applies after suffixation as suggested in (9) then (14)b should be the correct output as the non-syncopated /foxil/ of (9) ends in a single segment and so should accept the unreduced suffix /adoir/. As it patently does not, syncope must have occurred before suffixation. One of the conditions for the process illustrated by (10), suffix reduction, is a stem-final cluster. This gives the most radical reduction. An intermediate type of reduction is seen in the form /to:r/. To illustrate this and for the ensuing analysis I use Early Modern Irish forms. Consider the appearence of the intermediate suffix. (16) a b

coill coillteoir

/kill/ /ki:lto:r/

'wood' 'woodsman'

According to the remarks above the expected form would be (17)

*/ki:bdo:r/

As there is a stem-final single segment. Now while this form is not attested similar forms are, along with those with the /to:r/ suffix. (18) a feann b feanntoir c feannnadoir

/fcxin/ /fa:nto:r/ /fainadorf/

'to skin' 'skinner' 'skinner'

What in fact has occurred in (16)b and (18)b, which represent by far the most common forms, is syncope. If one recalls the formulation of syncope given in (8) then one sees that (18)c and (17) both match the structural description for it: an unstressed vowel is lost if the consonants which come together as a result of it represent a permissible phonotactic sequence. Here one must distinguish between tautosyllabic and heterosyllabic clusters. Thus while /xl/ (in (10)) can only occur syllable initially and /nt/ only syllable finally both can occur with a syllable boundary running through the cluster as for syncope. This accounts for the impossibility of (14)b as no syllabification would make the cluster acceptable in terms of Irish phonotactics. 5 (19) a b

*/-x$lt-/ */-xl$t-/

206

Raymond

Hickey

It can be seen that on syncope (d/ changes to /t/. This is part of a general rule of the form (20)

Voice dissimilation rule + son C [avoice] / — avoice

which in effect means devoicing as Irish has no phonemically voiceless sonorants. It can be seen to effect any input which meets its structural description and has operated throughout the whole history of Irish. In fact its form was somewhat more complicated in Old Irish. An extended version to account for the situation in Old Irish would be (21)

Voice and manner dissimilation rule — voice + voice / [ + son] — continuant + continuant

(21) now accounts for the following set of forms. (22) a

fäilid

/failiö/

'joyous'

b

fäilte

/failts/

'welcome' (fäilte

/fcdtex/

'joyous'

c

fäilteach

(fäili

(fäilteach

'affable' (Modi)) 'welcome' (Modi)) 'welcoming' (Modi))

That the form in (18)b is really a case of syncope can be seen from (22) b and its derivation. (23)

fäilid

/failiö/ /fail φ / /fail /fa:lt 9 /

(i) (ii) (iii)

by abstract suffixation by syncope by voice and manner dissimilation

It might be asked why for the base form above I have not postulated /fa:lö/ and given desyncope in the derivation of /fa:li§/. This could be done of course. As it is, the actual surface form acts as a good starting point as then the various steps on the way to /failja/ can be seen clearly. A second query might be: is not the very

Segmental phonology

and word-formation

in Irish

207

existence of the surface form /failiö/ a denial of the validity of the syncope hypothesis? In view of optionally syncopated forms like (18)b it is not. Desyncope is obligatory in Irish but syncope is an optional but strongly favoured rule. The non-existence of fäilt /fa:lt'/ along with fäilid may be an accident of attestation. Before leaving this examination of the agency suffix two conditions on it must be considered. Firstly for the syncope which I have been discussing to operate the segments which cluster must be homorganic. As the first segment is a sonorant and as there are three positions for sonorants in Irish, labial, dental and velar and as the velar nasal only occurs with a following homorganic stop only two positions, namely labial and dental, would have to be specified. The labial position must be excluded being heterorganic to the /d/ of the suffix. This is evidenced by forms like (24) a b

gliomach gliomadoir

'lobster' 'lobster-fisher'

Because of this only the dental position needs to be specified in the voice and manner dissimilation rule. This can be done with the feature [coronal]. (25)

~ -1- voice" + cont - + cor -

~ — voice" — cont - + cor -

An automatic palatal agreement rule would prohibit sequences such as jn\j, /It/, etc. The features [high] and [back] in the relevant combination of values can be used to formulate this rule and it can be added at the end of the derivation to bring clusters which might disagree among their segments for the feature [palatal] into line. Although the palatal agreement rule is simple in its operation and responsible for converting (26)a into (26)b (26) a b

/ki:lto:r/ /ki:lto:r/

it is complex inasmuch as it must have access to morphological information. In Irish assimilation is usually progressive so that (26)a should result in (27)

*/ki:lto:r/

208

Raymond

Hickey

but from this and other forms it is clear that the cluster arising from syncope on suffixation agrees in terms of palatality with the final consonant of the stem although this runs counter to normal phonetic assimilation in Irish. Ignoring this morphological fact would result in (27). The second condition on agency suffixation also concerns the final consonant of the stem. Consider the Modern Irish forms (28) a b

bad bädoir

/ba:d/ /ba:do:r/

'boat' 'boatsman'

As there is just a single stem-final consonant in the stem (28)a one would expect /ado:r/ but instead the form for a stem-final cluster is found. This has a simple explanation. No two consecutive syllables can begin with a /d/ so the /d/ of the suffix is deleted as is the unstressed vowel before it.6 Truncation may also feed the rule of second /d/ deletion. Consider the derivation of Modern Irish cladoir 'sea-shore dweller'. (29)

cladach

/kladax/ /klad/ /kladadoir/ /kladsoir/ /kladoir/

'sea-shore' (i) by truncation (ii) by agency suffixation (iii) by second /d/ deletion (iv) by pre-vocalic shwa deletion

Because of truncation above a complicated derivation has ensued. In view of the fact that it has this result one would wonder why it should take place at all. Its motivation would seem to be to render a monosyllable of the stem before suffixation for as will be seen from the next case to be considered agency suffixation always implies a monosyllabic stem. The second major agency suffix in Irish is also a borrowing (from English). It derives from the suffix '-er' and can be seen in obvious loan words like (30)

siuineir

'carpenter'

( < English joiner)

This and similar loans must be quite early as they show affricate simplification (/d.3/ -» /%/) and sibilant devoicing (/3/ -> /J/) in accordance with the strict loan-word phonology of Irish up to the

Segmental phonology

and word-formation

in Irish

209

later early modern period whereby no affricates or voiced sibilants were tolerated. In the present-day language affricates are tolerated (Hickey, 1982: 150-151; de Bhaldraithe, 1953 b: 15-17) but sibilants are still devoiced. As the result of a process of using the genitive for the nominative and of having English loan-words frequently end in -a /a/, all forms with original -eir are now found in spoken Connemara Irish (de Bhaldraithe, 1953 a: 248) with the ending -eara1. (31) a b

bäiceara tinceara

'baker' 'tinker'

(English loan word) (English loan word)

For the issue at hand however the use of the genitive for the nominative is irrelevant. What is important is that this ending involves a change in the final consonant of the stem. (32) a b c d

sgläta sgläitiara guna guineara

/sklaits/ /sklarteira/ /g\n a/ /gineira/

'slate' 'slater' 'gun' 'gunner'

On suffixation the stem-final consonant is palatalized. For the determination of the lexically stored form of the suffix there are then two possibilities. The first is that along with /e:r/ additional information that the consonant preceding the suffix is palatalized is given much as with syncope where the resolution of a cluster is lexically specified as involving palatal or non-palatal consonants or a combination of these (see (13) above). This process I call the EXPLICIT PALATALIZATION RULE. The second possibility is that on suffixation the stem-final consonant adopts the inherent value for [palatal] which the initial vowel of the suffix has in accordance with the normal progressive assimilation of Irish. In this case it is [-(-palatal] as /e:/ is a front vowel. This would be an example of a NATURAL PALATALIZATION RULE and would suffice in this case as all the stem-final consonants of bases which take /e:ra/ are either palatal already or palatalize. But in fact it must be regarded as an explicit palatalization rule which happens to coincide with a natural palatalization rule. It is explicit as it alters the form of the base to which it is added and natural as the change coincides with the inherent value of the suffix vowel. However a suffix with a back vowel which also palatalizes would be explicit

210

Raymond

Hickey

and unnatural, just as a front vowel which depalatizes would be. The remaining fourth possibility, a back vowel which depalatalizes, is explicit and natural. For the analysis below I will characterize the rules which alter the value of the final consonants of stems as all being explicit and those which do not as being natural leaving aside the question of whether the alternation is inkeeping with or contrary to the inherent value for [palatal] of the suffix vowel. Now compare the above situation with that of the /ado:r/ suffix. With this the non-palatal stems remain non-palatal whereas the palatal stems (all ending in a sonorant) palatalize the initial consonant of the suffix, see (16) above where /t/ -»· /t///!/ —. One can see that for the /adoir/ suffix (in its post-sonorant form /to:r/) a natural palatalization rule applies as the suffix assumes the value of [palatal] which the stem-final consonant shows. On the contrary the palatalization effect of /e:ra/ must be lexically specified as it does not allow a formation such as */gAne:ra/ for the form in (32)d. In order to arrive at a stem for suffixation, truncation may be applied to a given form, The truncated element is usually a verbal suffix 8 as in (33)

bäcäil 'bake'

/baikail/ /ba:k/ /bcnkeira/

(i) (ii)

by truncation by suffixation

resulting in a monosyllabic stem input to suffixation. Apparent exceptions to the monosyllable condition on suffixation are found in (34) a feilmeara b peaisineara c pinsineara

/felimeira/

'farmer' 'passenger' 'pensioner'

The first of these has an epenthetic vowel between the stem-final consonant sonorant cluster which is added after all word-formation processes have applied by scanning the output of the word-formation rules for acceptability with regard to phonotactics. The second and third forms in (34) while having disyllabic sources in English tend to lose the second vowel by syncope in the spoken language. (35) a b

peaisineara pinsineara

—• [paeisneus] -»· [pinsneua]

Segmental phonology

and word-formation

in Irish

211

2. Abstraction Abstract nouns in Irish have been and still are produced with a variety of suffixes. From Old Irish (Thurneysen, 1946: 167-168) onwards one suffix dominates above all others. It had and has the form /sxt/ with some variation in its realization which will be the subject of remarks below. It is added to various types of stems, the interaction of stem and suffix forming the object of interest here. Consider to begin with a set of Early Modern Irish forms. 9 (36) a b c d

bän bäineacht dall doilleacht

/ba:n/ /bainaxt/ /da:l/ /debxt /

'white' 'whiteness' 'blind' 'blindness'

It is evident from these that palatalization occurs on suffixation. 10 It can be ruled out as natural palatalization not only because the stem adapts itself to the suffix but also because the central vowel /a/ is non-palatal of its nature anyway. Thus an explicit palatalization rule must be lexically specified. Not only that but the information for the suffix must have access to at least one item of information on the stem, namely that it is monosyllabic, for with disyllabic inputs there is no palatalization (see below). Note that the forms in (36) are neither truncated nor syncopated. But a large number of adjectives which syncopate add /axt/ as well. Consider (37) a b

c d

socair socracht mi Iis milseacht

/sAkir/ /sAkraxt/ /mills/ /mil§axt/

'calm' 'calm(ness)' 'sweet' 'sweetness'

What is obvious from these is that the value for [palatal] which the syncopated clusters show in general (e. g. with agency suffixation considered above) is retained on abstraction suffixation. The condition that /axt/ palatalizes the stem-final consonant of monosyllabic inputs runs into difficulty here because forms like those in (37) a and c are assumed to be underlyingly monosyllabic as the argumentation given above in connection with agency suffixation has shown. This would imply that the suffix /axt/ must have access not only to the fact that a syllable is monosyllabic but also that it is syncopated as in the latter case the explicit palatalization rule which is operated

212

Raymond

Hickey

by /axt/ on monosyllables is overriden. In accordance too with the condition that epenthesis applies after word formation rules one has /axt/ after clusters with epenthesis. As these are underlyingly non-epenthetic, i. e. monosyllabic, suffixation as with normal monosyllables occurs. (38) a b

garbh gairbheacht

/garav/ /genyaxt/

'rough' 'roughness'

It should be noted in connection with these suffixes that explicit palatalization rules while involving the adoption of a certain value for [palatal] by a stem in its final consonant are never, in any morphological operation in Irish, inversion rules of the type (39)

[apalatal]

[ - apalatal]

but always of the type in (40) a or b. (40) a b

[apalatal] - + [ - p a l a t a l ] [apalatal] ->· [ + palatal]

An example of the former type is found when one proceeds with an examination of abstraction suffixation. Consider (41) a fearuil b fearülacht

/faru:]/ /farurbxt/

'manly' 'manliness'

As the form in (41) a is itself suffixed ( = /far/ + /u:l/) one could view both /u:l/ and /uitaxt/ as lexicalised and so do away with a combersome explicit depalatalization rule for /axt/ when added to /u:l/. This solution seems to miss a generalization which can ostensibly be reached when one views further forms. (42) a b

galänta galäntacht

/galamta/ /galamtaxt/

'stylish' 'stylishness'

Now if /u:l/ and /u:taxt/ are regarded as lexicalised then one requires an additional rule which stops the generation of (43)

/galamtsxt/

Segmental phonology

and word-formation

in Irish

213

with palatalization. Such a rule would specify that explicit depalatalization occurs with all polysyllables. It would account for both (41) b and (42) b. However this single rule runs into difficulties of its own with forms like (44) a b

siüräilte siüräilteacht

/su:ra:lte/ /su:ra:ltaxt/

'certain' 'certainty'

An examination of further examples of abstract nouns shows that a natural palatalization rule applies to all stems which are not monosyllables. The suffix does not change the value for [palatal] of the stem-final consonant. Thus the seeming generalization linking (41) b and (42) b turns out to be spurious and /u:l/ and /u:taxt/ must after all be lexicalised or a unique rule given which depalatalizes /u:l/ on suffixation of /axt/ which is the same in terms of complexity. Truncation has a role to play in this suffixation as well. Disyllabic forms which end in /a/ have this truncated and /axt/ added after truncation with the explicit palatalization rule operating. The procedure is thus similar to that for deriving (32)b and d from (32)a and c above. (45)

fada

/fada/ /fad/ /fadaxt/ /fi^axt/

(i) (ϋ) (iii)

by truncation by suffixation by explicit palataliza tion rule

But there is a further manifestation of the /axt/ suffix which is found with a number of monosyllables such as fada when truncated. The alternative abstraction suffixation is then. (46)

fada

/fada/ /fad/ /fadiixt/

(i) (ii)

by truncation by suffixation

Given the existence of the form fadaiocht 'length' it might seem that there are two suffixes. However they represent two manifestations of the same suffix. This is given in (47). (47) (i)

Suffixation of monosyllables add suffix /axt/

214

Raymond

(ii) a b

Hickey

palatalize stem-final consonant or add palatal vowel after it.

The exclusiveness of the alternative suffixation rules in (ii) accounts for the non-existence of either (48)a or b. (48) a b

*/fi4i:xt/ »/fadaxt/

In both (45) and (46) I have argued that truncation has taken place. Arguments could be brought forward however to show that suffixation is not preceded by truncation at all and alternative derivations can be supplied for forms with final shwa and without it, such as fada and trom. a

b

fada

trom

/fad / /fadaxt/ /fi^axt/

'long' (i) (ϋ)

/truim/ /truimxt/ /truimaxt / /trimsxt/

'heavy' by /xt/ suffixation (i) (ii) by shwa insertion (iii) by explicit palataliza tion rule

by /xt/ suffixation by explicit palataliza tion rule

But while a shwa insertion rule 11 will deal with the derivation of truimeacht 'heaviness' a truncation rule is still required to derive those monosyllabic forms which add /i:/ instead of palatalizing the stem-final consonant. If no truncation rule is available then the derived form in (46) would be /fadaiixt/ which is phonotactically impermissible in Irish (apart from not being attested) as a shwa cannot precede a long vowel anywhere. Palatal vowel suffixation as in (46) is frequently the only kind possible with a given stem. This has a clear diachronic explanation. Consider the Old Irish forms (Thurneysen, 1946: 167). (50) a fili b filed c filedacht

/fib/ /filiO/ /fili^axt/

'poet-NOM' 'poet-GEN' 1 2 'poetry'

Originally there was only /sxt/ as a suffix but the vocalization of the sequence /i§/ to /i:/ lead to it being used as an alternative for

Segmental phonology

and word-formation

in Irish

215

abstraction suffixation with forms which did not have /ιφ/ to begin with. In the present-day language the alternatives for suffixation given in (47) (ii) are reduced to a single choice when the stem already ends in a palatal consonant. Here only palatal vowel suffixation is found. Thus the modern equivalent of (50)c is (51)

filiocht

/filiixt/

'poetry'

which is derived as is fadaiocht in (46) (with truncation of wordfinal /a/). It can be seen that this type of derivation has been reached by j\ij achieving suffix status when one examines forms which in Old Irish did not have a genitive in /iö/. One such instance is Old Irish delb 'shape, form' which had the genitive delbe ~ delbae (Thurneysen, 1946: 100) and has the following abstract noun form in Modern Irish. (52)

deilbhiocht

/^elviixt/

'morphology'

In connection with this a second situation needs to be considered. Quite a number of forms which act as the input for suffixation already have long high vowels in their input forms (although these may be morphologically complex in themselves, see (53)c) where there is no such process operating like that in the derivations of (51) and (52). (53) a b c d

bulai bulaiocht comrädai comradaiocht

/Mi:/ /bAliixt / /kAmra:di:/ /kAmra:di:xt/

'bully' 'bullying' 'comrades' 'comradeship'

It is obvious that in these cases the abstraction suffix has been reduced from /axt/ to /xt/. Now there are many polysyllabic forms which end in a final shwa and given the possibility (for the forms in (53) and their like) that the suffix be reduced to /xt/ it might be supposed that in a set of forms like (54) a b

forasta forastacht

/fArasta/ /fArastaxt/

'stable' 'stability'

/xt/ is added to the entire stem of (54)a with shwa truncation reserved for the special case of underlying monosyllablic forms.

216

Raymond Hickey

Whether this view is correct can be ascertained by examining a further set of derivations. Consider the group of adjectives in /ax/ which also form abstract nouns. (55) a b

gionach gionacht

/gAnax/ /gAnaxt/

'voracious' 'voraciousness'

Here I assume that there is no truncation either. For abstraction suffixation truncation is reserved for underlyinigly monosyllabic forms ending in /a/. For if one had truncation of /ax/ as in agency suffixation then the input to abstraction suffixation in (55) would be n/. But this would automatically meet the structural description of the explicit palatalization rule which would give the form /ginaxt/ after suffixation which is incorrect. Because of this the abstraction suffixation in (55) has been further reduced to /t/ but only because of the following condition. (56)

/xt/ -> / t / / / x /

-

which arises as /xx/ is not permissible phonotactically. That truncation is not involved in forms with final /ax/ can be seen from contrasting sets which have palatalization applied or not, depending on their base form. (57) a

gealach

/galax/ /galaxxt/ /galaxt/ >

(57) b

gealacht geall

/galaxt/ /gal/ /galaxt/ /gilaxt/

gileacht

/gilaxt/

'brightness, moonlight' (i) by abstraction suffixation (ii) by geminate simpliflcation 'brightening' 'bright' (i) by abstraction suffixation (ii) by explicit palatalization rule 'whiteness, brightness'

Lastly one further manifestation of the abstraction suffix calls for comment. In a number of forms the suffix appears to begin consonantally. Consider the Modern Irish forms

Segmental phonology

(58) a b

meirin meirinteacht

and word-formation

in Irish

217

'small finger' 'act of fiddling with something' (de Bhaldraithe, 1953 a: 253)

An examination of these and similar forms shows that the /t/ which is found suffix-initially only occurs after /n/. If one considers Early Modern Irish forms then one finds /t/ after /l/ also as in (59) a fearüil b fearultacht

/faruil/ /faruiltsxt/

'manly' 'manliness'

This is not something which is morphologically conditioned. Rather in Irish there is a tendency to develop a voiceless alveolar stop after certain continuant segments. This gives /a'riist/ for aris 'again' with an epenthetic /t/ for example. Just such a 7 1 3 is present in (58)b and (59)b, which is supported by the existence of Τ after the sonorants L and Ν in countless other forms, not least among agency nouns (see (16) above for example). The presence of suffix-initial Τ after the sonorants L and Ν in the derivational forms for agency nouns may have exerted pressure towards a similar situation developing with abstraction suffixation. Thus the voiceless homorganic stops which arose through syncope in the one area would have been introduced by epenthesis in the other. With the diminishing force of analogy the stops were deleted after the lateral sonorant but are still to be found after the nasal. In conclusion I should add that in the discussion of agency and abstraction above I have continually referred to 'forms' which are underlying, act as inputs, etc. This very general term is deliberate and reflects the fact that, for all the processes considered, affiliation to a specific lexical class plays no role. The input for agency derivation may come from the lexical class noun or verb. Equally for abstraction derivation both nouns and adjectives serve as inputs, although for the majority of my examples I have taken adjectives. In fact the input for abstraction derivation can itself be the output of agency derivation, something which can be seen from the following set of forms. (60) a feilmeara b feilmearacht

'farmer' 'farming'

218

Raymond

Hickey

Notes 1. Though its distribution in Modern Irish is vastly restricted in comparison with Old Irish. 2. The arguments for and against deletion in Irish are basically similar to those used for English (see Kastovsky, 1980: 219): greater generalization, economy of derivations, etc. 3. This is a decisive difference between it and epenthesis which always separates segments of a phonotactically impermissible cluster. But the two phenomena are closely related not least in the conditions which govern their appearence. On these and the interrelationship of epenthesis and syncope, see Hickey (1985). 4. Seen from the point of view of distinctive feature theory this is an abbreviation for combinations of the two features [high] and [back]. In Irish however where palatalization or velarization has a strictly morphological function there is justification in having a separate feature, viz. [palatal], apart from the phonological ones, which characterizes the strictly morphological role which palatalization and velarization play in Irish. 5. Note that I refer globally to Irish phonotactics, syncope, truncation, etc. when there is no discernible difference between their manifestation in Old Irish and that in Modern Irish. If there is, the particular statement is specified for a certain stage of the language. 6. That this condition is correct can be seen from dissimilations of voiced alveolar stops which have occurred throughout the history of Irish. Early Modern Irish dadamh gives present-day Irish /tada/ with devoicing of the first /d/. 7. MacEoin (1974: 63) comments on this and a variety of other forms where the genitive and nominative interchange. He rightly remarks that with the collapse of the older case system and the loss of the accusative and dative only the nominative and genitive (and the strictly non-grammatical vocative) were left and these then interacted. This interaction was furthered by the fact that the form of the genitive in one nominal class was frequently the same as that of the nominative in another class. 8. Though by no means always. Many English loans are disyllabic although the English source was monosyllabic. Thus gun guna in Irish. This /a/ is added in many places and not always those where, for the Middle English forms which may have been the source (MacEoin, 1974: 63-64), the unstressed shwa, later lost in English, can still be assumed to have been pronounced. 9. Early Modern Irish is indicated by the spelling (which has been since reformed, though with little or no effect on the forms quoted below). The pronunciation is based on that of Connemara Irish today. Note that there are vowel alternations (back to front vowel) which cooccur with palatalization in the forms given. Whether this vocalic gradation (Hickey, forthcoming: VI. 1.2.) was already present in Early Modern Irish is difficult to say as the spelling does not indicate it and no external evidence points to vocalic gradation having had occurred by some particular point in time. The upshot of this is that the pronunciation offered in this section is that of contemporary (Western) Irish and the orthographical forms are Early Modern Irish (with the proviso that there are similar to those of Modern Irish) although the processes being dealt with, such as syncope, reach back into Old Irish and the stages of the derivations which I postulate, if they are correct for the modern language, were certainly correct for previous stages of it as well. 10. Palatalization also occurs when adjectives such as those in (36) a and c are altered from the positive to the comparative degree. But it would not be an explanation of palatalization here to remark that it occurs because the abstract nouns derive from the comparative form of adjectives. Firstly allowing such a

Segmental phonology and word-formation in Irish

219

derivational history for abstracts would just shift the problem of the conditions for palatalization to adjectives and secondly and more seriously it can be objected that the comparative form of adjectives and abstract nouns are parallel and do not stand in a derivational relationship. To prove that they do one would have to show conclusively that native speakers when forming abstract nouns first change the adjective into the comparative degree and then add a further suffix to give the abstract noun. While such pathways through a language's morphology may be useful didactically it is more than doubtful that speech production occurs on similar lines. 11. A further argument against an insertion rule at this point is that there is already a deletion rule for agency noun formation. Having an insertion rule for abstraction formation would mean that the grammar would be mixed in a manner which would run counter to the aim of maintaining it as simple as possible. 12. The genitive case was used to provide the base for abstraction suffixation already in Old Irish, see Thurneysen (loc. cit.). 13. This is a convention adopted from Hickey (forthcoming) where capital Italic letters are cover symbols for palatal and non-palatal versions of a segment. . Thus Τ = m + ft/, £ = /!/ + /Ι/, Ν = /η/ + etc.

References Aronoff, Mark 1976 Word-formation in generative grammar (Linguistic Inquiry Monograph 1) Cambridge, Mass.: Μ. I. T. Press). Chomsky, Noam — Morris, Halle 1968 The sound pattern of English (New York: Harper and Row). De Bhaldraithe, T. 1953 a Gaeilge Cois Fhairrge. An deilbhiocht (Dublin Institute for Advanced Studies). 1953 b "Nua-iasachtai i nGaeilge Chois Fhairrge", Eigse 7: 1-34. Hickey, Raymond 1982 "The phonology of English loan-words in Inis Meäin Irish", Eriu 33: 137-156. (forthc.) The phonology of Modern Irish. (1985) "The interrelationship of epenthesis and syncope, evidence from Irish and Dutch", Lingua 65: 239-259. Kastovsky, Dieter 1980 "Zero in morphology: a means of making up for phonological losses", Historical morphology, edited by J. Fisiak (The Hague: Mouton), 213-250. MacEoin, Gearöid S. 1974 "Genitive forms as nominatives in Irish", Zeitschrift fur celtische Philologie 33: 58-65. Thurneysen, Rudolf 1946 A grammar of Old Irish (Dublin: Dublin Institute for Advanced Studies).

DIETER KASTOVSKY

Deverbal nouns in Old and Modern English: from stem-formation to word-formation1

1.1. A comprehensive history of English word-formation has still to be written. Despite considerable differences in their theoretical orientation, the classical handbooks by Jespersen (1942), Koziol (11937, 21972) and Marchand (21969) are basically descriptions of M o d e r n English, because they all take the present state of the language as a starting point, i. e., they describe the present patterns, and then add some remarks about their history. The only major difference between Marchand and these earlier handbooks is that Marchand on the whole includes only those patterns that are clearly productive in Modern English or have at least led to some sporadic new formations during the last centuries 2 , while Koziol and Jespersen also list definitely unproductive patterns, which are merely of historical interest from a Modern English point of view. Even more important is the fact, however, that Koziol and Jespersen also very often include completely demotivated formations (e. g., defile, batch, glazier) or words which from a Modern English point of view are completely unanalysable (e. g., pioneer, volunteer, treasure, dairy, chivalry, etc.), while Marchand only deals with analysable syntagmas. And, after all, it is only analysable syntagmas and not monemes that are the subject matter of word-formation. Now, even despite their wealth of historical information, such descriptions of individual patterns do not add up to a comprehensive history of English word-formation, however. They do not do this, because they do not ask in what way the overall system of English word-formation has changed in the course of time. Nor could such a question be answered, of course, from a point of view which merely takes Modern English as a starting-point. In Middle English, many Old English patterns (prefixes, suffixes) were lost, and many new ones were introduced from French and (Neo-)Latin, which, together with some other developments discussed in this paper,

222

Dieter

Kastovsky

resulted in a considerable restructuring of the whole system. This fact was by no means overlooked by the above-mentioned handbooks, of course. But this kind of restructuring cannot be described properly without appropriate, comprehensive synchronic analyses of the Old English and Middle English patterns, and, unfortunately, such analyses are not available for all periods and all patterns in the required detailed form. We will therefore very often have to fall back on the fairly global remarks in the various Old English and Middle English grammars 3 as a complement for the existing monographs dealing with individual morphological and/ or semantic patterns. The following comparison of the derivational patterns of deverbal nouns in Old English and Modern English will therefore be no more than a first attempt at a more comprehensive view of the history of English word-formation. It will consequently focus more on general tendencies characterizing the development of the system as a whole than on individual patterns and idiosyncratic details such as lexicalizations, etc. This will also be necessary in view of the data that were readily available. 1.2. The description of Modern English is based on Marchand (21969), still the most comprehensive empirical study in existence. His analyses had to be reinterpreted somewhat, however, because of a slightly different theoretical framework adopted in the present paper. For the analysis of compounds and deverbal nouns, Marchand basically uses the syntactic-transformational approach to nominalizations developed in Lees (1960, 1966), i. e., the suffixes of deverbal nouns are related to syntactic functions like Subject, Object, Adverbial Complements, Predicate. The present analysis, on the other hand, is based on a case grammar framework, and the resulting categories come much closer to those of traditional, nontransformational, semantically oriented word-formation studies. Thus, the suffixes of deverbal nouns are interpreted as representing either underlying semantic-syntactic case relations, such as Agentive (employ I er, letter-writ I er*, Objective (bed-spread/0), Experiencer (ihear/er), Benefactive (grant/ee), Locative (bus-stop/0), Instrumental (atomizjer), Temporal (spring/0), or abstract head-nouns of complement sentences such as Fact, Act, Action, Process, Event, State, etc. (employ/merit, jump/0, runn/ing)5. The description of Old English is based on a list of all deverbal nouns (including compounds) recorded in Bosworth — Toller (1898/

Deverbal nouns in Old and Modern English: from stem- to word-formation

223

1921) and Clark Hall (41960), which I had collected quite some time ago 6 . I have not included a similarly detailed description of Middle English, partly, because I did not have available a comparatively extensive data base, partly, because the developments during this period, notably the loss of certain suffixes and the introduction of new ones, merit a separate study. Moreover, this was not the primary objective of this investigation, which I intend simply to be a fairly general survey of how the overall system of deverbal nouns changed between Old English and Modern English, in particular also from a typological point of view. 2.0. For reasons of easier exposition, I will begin with a brief description of the Modern English system. According to Marchand (21969), the following suffixes are involved in the derivation of Modern English deverbal nouns: -age, -al, -ancej-ence, -ard, -ation, -ee, -er, -ery, -ing, -lej-el, -ling, -ment, -ster, -th, -ure and -0 (zero morpheme). Their functions can be described as follows. 2.1.1. -age: Borrowed from French, it forms both denominal and deverbal nouns; within the deverbal group, Action nouns 7 seem to be the most productive group, cf. coinage, leakage, steerage; but other types occur as well: Objective (floatage, luggage, package), Factitive/Result (sweepage, breakage, wreckage), Locative (anchorage, storage). Many derivatives represent several groups, as is typical of deverbal nouns, i. e., this also applies to the other types and also to the Old English material. The suffix does not, however, seem to form Agentive, Experiencer or Instrumental nouns. It seems to be fairly productive, and is not restricted to the non-Germanic vocabulary. 2.1.2. -al: Borrowed from French, it forms more or less exclusively Action and Factitive nouns from French or Latin verbs, cf. arrival, acquittal, reversal, trial, revival (Action), proposal, surmisal (Factitive), but occasionally also Agent nouns (arrival) 8 . 2.1.3. -ancej-ence: Borrowed from French, it derives primarily Action nouns (acceptance, admittance, dependence, guidance), but also Factitive nouns (utterance), Objective nouns (inheritance, allowance) and Locative nouns (entrance). The type is not restricted to loans. 2.1.4. -antj-ent: Borrowed from French, it derives primarily on a Neo-Latin basis of coining: Agentives (defendant, adherent, servant, claimant), personal Objective nouns (decedent, dependent) and In-

224

Dieter

Kastovsky

strumentals {solvent, coolant, digestant). Most of these are technical terms, but the type seems to be fairly productive. 2.1.5. -ard: Borrowed from French, it has led to just a few analogical formations used as epithets, like dotard, blinkard, laggard, stinkard (coined after bastard, buzzard, coward), which are probably best interpreted as representing Agentive and personal Objective nouns. Its productivity is practically nil. 2.1.6. -ation: Borrowed from French, it derives only Action (identification, christianization, de-Maofication) and Factitive nouns (creation, translation, formalization). It is the only loan suffix that derives Action nouns on a large scale, since it is the only option for verbs ending in -ize, -ate and -ify. 2.1.7. -ee: Borrowed from French, it was originally restricted to the derivation of personal Goal/Benefactive nouns as legal terms (grantee, payee, transferee), but now also derives Objective nouns (drawee, draftee, trainee, murderee) and Agent nouns (escapee, standee). 2.1.8. -er. Of Old English origin, it derives primarily Agent nouns (personal: baker, writer, reader,; non-personal: pointer, squeaker), but occurs in most other deverbal types as well: Experiencer (hearer), Objective (drawer, tier), Instrumental (blotter, atomizer), Locative (diner, sleeper, slipper, boiler), Action nouns (disclaimer, breather). Only the types Factitive, Goal and Temporal seem to be excluded. 2.1.9. -ery: Borrowed from French, it mainly functions as a denominal suffix (pottery, swannery)·, as a deverbal suffix, it primarily derives Locative nouns (bakery, printery, brewery), which, however, can also be related to the respective profession-denoting nouns. The same holds for the Action nouns robbery, jugglery, while thievery is clearly deverbal 9 . 2.1.10. -ing: Of Old English origin, it derives primarily Action nouns (drinking, driving, beating, ending) but occurs also in most other deverbal types, i. e., Objective nouns (stuffing, lending, offering; being (personal)); Factitive nouns (painting, building, clearing, cutting)·, Locative nouns (opening, crossing, dwelling), Instrumental nouns (clothing, binding, coating, lacing), Agentive nouns (following (collective)). The only types that are excluded are Benefactive, Experiencer and Temporal nouns.

Deverbal

nouns in Old and Modern English: from stem- to word-formation

225

2.1.11. -lej-el: Of Old English origin, it has formed a few Instrumental {prickle, hackle, stopple, treadle) and Objective nouns {spittle), but is no longer productive (acc. to Marchand dead since about 1400). 2.1.12. -ling·. Of Old English origin, mainly denominal; it has formed a few Objective {shearling, suckling, eanling, weanling, starveling) and Agent {gruntling) nouns, all denoting animals, and all probably coined primarily as names rather than as nominalizations proper 10 . Its productivity seems to be rather limited. 2.1.13. -ment: Borrowed from French, it derives primarily Action nouns {admonishment, amusement, bewilderment, embezzlement, development), but occurs also in most other groups, i. e., Objective nouns {shipment), Factitive nouns {achievement, assignment), Instrumental nouns {reinforcement, equipment), Locative nouns {encampment, settlement), and even Agent nouns {embarrassment). 2.1.14. -ster: Of Old English origin, originally a female suffix, now deriving sex-indifferent Agentive nouns, but of restricted productivity {brewster, dryster); probably mainly operating on the basis of direct analogy. 2.1.15. -th: Of Old English origin, with just a few formations, viz. Action (growth, spilth) and Objective nouns {spilth); practically unproductive. 2.1.16. -ure: Borrowed from French, it is restricted to verbs ending in -t or -s, and derives mainly Action nouns {closure; pressure, exposure, departure) and a few Objective nouns {enclosure). Its productivity is questionable. 2.1.17. 0: Of Old English origin, it derives within almost all types of deverbal nouns with the exception of the Benefactive and Experiencer types, cf.: Action nouns {look, hunt, hit, love, feel, alarm, hate), Agentive nouns {cheat, flirt, sneak, spy, coach, cook, guide), collective Agentive nouns {hunt, help), Objective nouns {discard, convert, flunk; deposit, handout), Factitive nouns {design, brew, hash), Instrumental nouns {whistle, rattle, pull), Locative nouns {dump, lounge, stop, hide-out), Temporal nouns {spring, fall).

226

Dieter

Kastovsky

2.2.1. This brief survey can be summarized in the following table: (1) Action Fact -age

+ +

-al

+ 4-

-ance/ -ence

+ +

Agentive

Experiencer

Object- Factiive tive +

+ + + +

+

-ard

(+)

(+)

+ +

-ee +

-ery

( + ?)

-ing

+ +

+ +

(+ ) collect.

-le/-el

-ment

+ +

-ster

Temporal

+

+

Productivity + +

+

+

+

+

+

+ + +

•

+ +

+

+

+

+

+ +

+

+

+

+ + +

+

-ling

Locative

+ +

-er

Instrumental

+

-ant / -ent

-ation

Goal/ Benefactive

+

+

+ +

-1- +

+

+

+ +

+

-

+ +

+

+

+

-th

+

+

-ure

4- +

+

0

+ +

+

+

-

(+) +

+ +

+

+

+ +

+ + = d o m i n a n t semantic type within the suffix ( ) = borderline or d o u b t f u l cases

As this table demonstrates, the number of basic semantic-syntactic types in this area is rather limited, i. e., roughly nine, if one adopts fairly general case definitions. This set could even be reduced to eight categories, if one disregards Temporal nouns. These are only instanced by the zero derivatives spring and fall, which, moreover, seem to be highly lexicalized. These global categories may of course contain various semantic subgroups; thus, Action nouns occur in the basic variants of Fact nouns on the one hand and Action, Process, State, Act, etc., nouns on the other; Agent nouns may be [ + / —human], Objective and Factitive nouns may be concrete or abstract, etc. Such distinctions depend partly on the underlying verb, partly on the immediate syntactic context, and partly they are due to various degrees of lexicalization. They are disregarded here.

Deverbal nouns in Old and Modern English: from stem- to word-formation

227

As has already been observed, most of the suffixes derive several semantic types, but usually, although not always, one or two types dominate with a given suffix. Thus, -age, -al, -ance, -ing, -ment, -ure, -0 are primarily Action noun formatives, -ant and -er derive primarily Agent and Instrumental nouns, -ee Benefactive and Objective nouns, etc. Only very few suffixes are limited to one or two types, as, e.g., -ation (Action, Factitive), -ery (Locative), -ster (Agentive). Deverbal noun-forming suffixes are thus typically poly functional. 2.2.2. Of the 17 suffixes listed above, only 7, i. e. 41%, are of native origin; the others are loans from French. And if one disregards -lej-el, -ling, -ster, -th, which are more or less unproductive in Modern English, we are left with only 3 native formatives ( = 17,6%), viz. -er, -ing and -0. On the other hand, these are definitely the most productive ones, since the Romance suffixes also manifest various restrictions of their productivity. Thus, -ard is practically dead, -ure is restricted to verbs ending in -t and -s and therefore hardly productive; -age, -al, -ancej-ence, -ant/-ent, -ee, -ery and -ment seem to be reasonably productive, as is indicated by their extension to native bases, but their productivity is by no means unrestricted; only -ation is hardly restricted within its domain, which is verbs ending in -ate, -ize, -ify, for which it is the only possible Action noun suffix. 2.2.3. One general remark as to the morphological aspect seems to be necessary in view of the following. Derivation always operates by adding the suffix to the verbal base form, which is identical with the infinitive and the quotation form of the verb in the dictionary. In other words, the basis is always a word in the sense of "free form", and so is the derivative. We can thus quite legitimately speak of "word-formation". Moreover, with the overwhelming majority of these suffixes, no morphophonemic alternations arise; exceptions are partly -antj-ent 11 , -ätion (stress shift) 12 , -ee (stress shift), -ure (consonant alternation). 3.1.1. Let me now turn to a more extensive survey of the Old English patterns, which is based on an exhaustive collection of all deverbal nouns quoted in Bosworth—Toller (1898/1921) and Clark Hall (41960). For reasons of space, this material can of course not be quoted in full, but will have to be reduced to a representative

228

Dieter

Kastovsky

sample illustrating the morphological and semantic properties of the respective patterns. When we try to establish the word-formation patterns of a dead language or of the earlier stages of a living language — i. e., of what is also called "corpus languages" — we are usually confronted with delimitation problems to a much greater extent than when we are dealing with a living language. The relevant criteria in both cases would seem to be analysability, existence of a morphologicalsemantic pattern, and productivity, if one adopts Marchand's definition of word-formation: Word-formation is that branch of the science of language which studies the patterns on which a language forms new lexical units, i. e. words. Word-formation can only be concerned with composites which are analysable both formally and semantically ... a composite is a syntagma consisting of a determinant and a determinatum ... (Marchand 1969: 2-3). Obviously, these criteria constitute an implicational hierarchy, i. e., productivity implies the existence of a pattern, and a pattern implies the analysability of its members. The opposite direction does not hold, of course: the analysability of a single lexical item, e. g., laugh/ter, does not necessarily imply the existence of a pattern, and the existence of a pattern, e. g., brass ~ brazier, glass ~ glazier, grass ~ grazier, does not necessarily imply its productivity. For Marchand and many other specialists in the field, the decisive criterion is the productivity of a given pattern, i. e., they would not include patently unproductive patterns in a synchronic description. This is certainly justified, although it is well known that it is not always easy to establish whether a given pattern is in fact productive or not. Moreover, the notion of productivity itself is all but welldefined, since it is a relative rather than an absolute concept. And Marchand himself has a fairly broad notion of productivity in view of the inclusion of suffixes like -ard, -le or -th, for which only very few formations are attested, and, what is more, for which it would hardly be possible to predict any additional, new formations. 3.1.2. For a corpus language, productivity as the decisive criterion as to whether a given pattern should be included in the description or not, is somewhat problematic, since the major criteria for the establishment of productivity, viz. introspection, elicitation and

Deverbal nouns in Old and Modern English: from stem- to word-formation

229

acceptability judgments of neologisms, are not available. But, as has been pointed out repeatedly, word-formation has two aspects, an analytic and a synthetic one (Kastovsky 1982 a: 16-17, Hansen et al. 1982: 32-33). The synthetic aspect will focus on the production of word-formation syntagmas and will consequently only deal with productive patterns. The analytic aspect, on the other hand, which is a prerequisite for the synthetic point of view, need not necessarily be as restrictive; it will in fact have to establish quite generally which part of the vocabulary is complex and will describe its morphological and semantic structure. Here, analysability (motivation) and the existence of a pattern are the primary criteria. Such an approach will not only include those formations that are due to synchronically productive patterns, but also those lexical items which are morphologically and semantically analysable, but which were formed on the basis of patterns that have ceased to be productive, provided they are not isolated instances but fit into a recognizable pattern. It seems to me that this kind of approach is preferable, at least in the initial stages of an investigation, when dealing with a corpus language such as Old English. This does not mean, of course, that productivity judgments should not be included, only that productivity does not play the same critical role as in the synthetic approach, among others because it can only be established more or less tentatively by circumstantial evidence. In this connection, Panagl (1982) discusses four useful criteria, which were at least informally also used in the following analysis of the Old English material, viz. 1) quantity of new formations in a given period, 2) quality of the formations, i. e., semantic regularity and absence of lexicalization, 3) grammaticalization (shift of a pattern towards inflection), 4) typology, i.e., extrapolation from typological regularities found in living languages. To these, the following considerations can be added. If a certain pattern already existed in Old English and is still productive in Modern English, it can be assumed that it was also productive in Old English. A further indication is the degree of morphological transparency, i. e., the amount and type of morphophonemic alternation, in particular whether such an alternation occurs elsewhere, e. g., in the inflectional system, or whether it is isolated. 3.1.3. On the basis of these criteria, and with the analytic aspect as a theoretical background in mind, the following suffixes have been included: -d (alternants -ed, -opj-ap, -p, -1), -el (alternants -I, -of),

230

Dieter

Kastovsky

-els, -en, -end, -er(e),

-estr(e),

-ett, -icg(e),

-ing\ (alternant -ung),

-ing2, -ling, -ness, -0, i. e., 14 suffixes (Modern English 17). There are a number of formations containing elements that could be identified as suffixes and that historically speaking indeed are suffixes, but that exhibit semantic irregularities, e. g., a high degree of lexicalization, or an isolated morphophonemic alternation; to these belong: -dorj-porj-tor. gealdor 'incantation, spell' : galan 'to sing', beorpor 'childbirth, offspring' : beran 'to bear, bring forth', hleahtor 'laughter' : hlihhan 'to laugh'; -s: Ices 'letting of blood' : lätan 'to let, let out', rces 'counsel, deliberation' : rcedan 'to advise', bices 'blowing, blast' : bläwan 'to blow', häs 'command, bidding' : hätan 'to command', äs 'food' : eta η 'to eat'. -m: cwealm 'death, destruction' : cwelan 'to die', wielm 'fount, stream, inflammation' : weallan 'to well, boil, be h o t \ ß e ä m 'flight' : fleön 'to flee'; -eld: fcereld 'going, journey, way, road': far an 'to go, travel'. Such formations probably had the same status as, e. g., laughter (: laugh) has in Modern English; there is a semantic connection between the verb and the noun, but no longer any genuine derivational relationship. Such cases have been excluded on the basis of the above-mentioned criteria, since the respective patterns are clearly no longer productive. 3.1.4. At this point, a brief remark about certain morphological aspects of Old English word-formation seems to be indicated; more will follow after the presentation of the Old English data. Old English is an inflectional language with grammatical gender (masculine, feminine, neuter), which is thus not predictable on the basis of semantic properties. Moreover, there are various inflectional classes cutting across the gender distinctions, which are usually referred to as strong (type cyning, cyninges) and weak (type guma, guman) declensions. Originally, the affiliation to a given declension was determined by a stem-formative, but this in most instances has coalesced with the case/number exponents and is recoverable synchronically only to a limited extent. Recoverability is easy only when the stem-formative has caused mutation of the stem vowel, cf. drenc m. 'drink' < *drank-i-z : drincan. As will be seen in the following, only very few suffixes have a unique affiliation to one gender and/or one inflectional class, because the Old English pat-

Deverbal nouns in Old and Modern English: from stem- to word-formation

231

terns are at least partly the result of the coalescence of several morphologically different, but closely related word-formational processes. This also explains why in certain patterns the suffix has caused -/-mutation in some of the derivatives but not in all. Thus, there are some patterns completely without -/-mutation, but no pattern where every item capable of being affected by -/-mutation has in fact undergone this change; it is always only some part of the derivatives that has. Finally, there are two types of verbs functioning as bases; strong verbs, which form their preterites and past participles by vowel change (gradation, ablaut), and weak verbs, which normally have no vowel alternation and form their preterites and past participles by means of a dental suffix. The former but not the latter, thus have various stem forms or root alternants, and all of them are not only theoretically, but also actually available as bases for derivation. The situation is thus quite similar to German binden and der Band, das Band, die Binde, das Bündel, das Bündnis, das Gebinde, der Bund, das Bändel, die Bande, etc. I will return to these aspects after a survey of the Old English data. These will contain examples for each semantic type existing within a given suffix arranged according to morphological differences (strong, weak verbs, stem form, gender, inflectional class, etc.). 3.2.1.1. -d (alternants -ed; -op/-ap/-p; -1) Action nouns: flöd m., n. 'flood' : flöwan 'to flow', han-cred m. 'cook-crow' : cräwan 'to crow', äp-swyrd η. 'oath' : swerian 'to swear', cwild m., f. 'death, destruction' : cwelan 'to die' // äled m. 'fire' : älan 'to burn' // cebrecp f. 'sacrilege' : brecan 'to break', gewilcp n. 'rolling' : (ge)wealcan 'to roll', fyrmp f. 'washing' : (ge)feormian 'to wash, clean', gepingp f. 'intercession' : pingian 'to beg, intercede' // sweolop(a) m. 'heat, burning' : swelan 'to burn', spiwepa m. 'vomiting': spiwan 'to vomit', fiscop m. 'fishing' : fiscian 'to fish', hergap m. 'harrying': hergian 'to make war, harry', hceletop m. 'greeting' : hälettan 'to greet' // gift f., m. 'giving' : giefan 'to give', ymb-hwyrft m. 'circuit, bend, turn' : ymb-hweorfan 'to turn round, revolve', ceht f. 'possession, power' : ägan 'to own', slieht m. 'striking' : sleän 'to strike', Tht f. 'increase' : eäcan/ican 'to increase', wyrht f. 'work' : wyrcan 'to work', nceft f. 'need, want' : nabban 'to have not, to lack'. Agentive nouns: fierd f. 'national levy or army' : faran 'to travel, go' // wcel-slihta m. 'murderer' : sleän 'to slay', serift m. 'one who shrives, confessor': scrifan 'to shrive', wyrhta m. 'workman', cys-

232

Dieter

Kastovsky

wyrhte f. 'cheesemaker' : wyrcan 'to work' // folgap m. 'train, retinue' : folgian 'to follow'. Objective nouns: säd η. 'seed' : säwan 'to sow' // piefpjpeöfpj peöft f. 'stolen goods' : peöfian 'to steal' / peöf 'thief, äfyrmpa f. 'sweepings, rubbish' : feormian 'to scour, clean' // gifepe n. 'what is granted by fate' : giefan 'to give' // äht f. 'possessions, serf : ägan 'to own', slieht m. 'animals for slaughter' : sleän 'to slay'. Factitive nouns: geblced m. 'blister': bläwan 'to blow', bled f. 'flower, blossom': blöwan 'to blossom' // migopa m. 'urine': rriigan 'to make water', sceafopa m. 'chip' : sceafan 'to shave', spiwepa m. 'what is vomited, vomit' : splwan 'to vomit', huntop m. 'what is hunted, venison' : huntian 'to hunt' // cyst f., m. 'what is chosen' : ceösan 'to choose', (ge)sceaft m., f., η. 'what is created' : scieppan 'to create', scrift m. 'penalty' : scrifan 'to shrive'. Instrumental nouns: weft m./wift i.jwefta m. 'weft' : wefan 'to weave', wist f. 'food, sustenance' : wesan 'to be, exist'. Locative nouns: flöde f. 'a place where anything flows, channel' : flöwan 'to flow' // gepingp f. 'court where claims are settled' : (ge)pingian 'to intercede, settle', nirwp f. 'prison' : nirwan 'to constrain' // fiscop m. 'a place for fishing' : fiscian 'to fish'. Thus, this group of suffix alternants derives primarily Action nouns, other groups are much weaker. Often more than one possibility exists for one and the same morphological form, sometimes differentiated by gender and/or inflectional class, cf. flod m., n. 'flood' vs. flöde f. 'channel', (ge)wyrht f. 'work' vs. wyrhta m. 'worker' vs. cys-wyrhte f. 'cheese-worker', sometimes without such a differentiation, cf. byrd f. (: beran 'to carry') 'birth' and 'burden' (Action vs. Objective noun), säd η. 'seed' and 'sowing', gift f., η. 'gift' and 'giving'. 3.2.1.2. The decision to regard -d, -ed, -op, -ap, -p and -t as alternants of a common suffix -d rather than as independent suffixes requires an explanation. Pilch (1970: 111, 113) apparently assumes two completely different suffixes -d (deverbal) and -p (deadjectival; -d according to him appears as -t after stem-final fricatives and as -0 after stem-final -t or consonant + -d (cf. wyrd 'fate' = *wyrdd12); and -p also has an alternant -t after fricatives. This analysis is questionable, however, since there are quite a few deverbal derivatives with -p, which Pilch does not mention. Moreover, this latter form is in fact the one that survived, cf. spilth, growth. We will thus have to include the fricative forms in this family. Basically,

Deverbal nouns in Old and Modern English: from stem- to word-formation

233

however, the reasons for treating these formatives as alternants of one and the same dental suffix are semantic and historical. There does not seem to exist any clear correlation between a certain semantic type and any of the alternants, or between gender and/or inflectional class and any of the alternants. This is due to the history of this morphological pattern, which also accounts for the fact that these alternants exhibit at least to a certain extent complementary distribution. The pattern goes back to a Proto-Indo-European suffix family clustering around the formative -t- (cf. Krähe — Meid 1967: 139 ff.), to which various vocalic extensions functioning as stem formatives were added (-to-, -tä-, -tio-, -tiä-, -ti-, -tu-, etc.). Whether these stem formatives originally correlated with certain semantic patterns is not clear and is irrelevant in this connection, since at least for Old English no such correlation can be established. But these stem formatives determined inflectional class affiliation and gender. Thus, while these stem formatives were still clearly distinct, the various patterns were kept apart at least morphologically. But one of the characteristic features of the Germanic languages is the levelling of unstressed vowels including the stem-formatives, which furthermore partly coalesced with the case/number endings. This resulted in the loss or merger of various originally separate inflectional classes. The same apparently happened to the various originally separate -iF-formatives. The consonantal alternation, on the other hand, is due to the fact that part of the members of this suffix family were stressed in the pre-Germanic period and were thus affected by Verner's Law. This and the various sound changes and assimilation processes in the wake of the Germanic sound shift produced the variants -d, -p, and -t, which probably were in complementary distribution, at least originally. This is still indicated by certain distributional tendencies in Old English. Thus, -d occurs mainly after roots ending in a semivowel, liquid or nasal (ßöd, bled, cwild, fierd), -p is mostly found after stops (brecp(a), rtpp, gicpa), and -t usually follows a fricative (friht, gift, äht, graft, wist). But there are exceptions (gesiht/gesihp) 'sight': seön, pifp/peöft, wcefpj waft), which can only be explained as analogical reformations, because the factors causing the complementary distribution were no longer transparent. 3.2.1.3. All three genders are represented, but there is a marked predominance of the feminines, while neuters are fairly rare. The same holds true of the inflectional classes, i. e., all are represented,

234

Dieter

Kastovsky

but the strong declensions dominate. Since the suffix alternants not only contained stem formatives, but also could be preceded by socalled linking vowels ("Bindevokale") 13 , the root vowel could be affected by -i- mutation, if the stem formative or the linking vowel were -/-. Thus we have mutated bled, bleed, cred, gesiht, gewilep, scyrft, cyst besides unmutated flöd, peöft, hwearft, brecp(a). Derivation is made both from strong and weak verbs. With weak verbs, the base is always the infinitive stem, cf. ic-an — ih-t, giccan — gic-pa, til-ian — til-p(e). With strong verbs, usually the stem form containing the weak grade, which also occurs in most past participles and preterite plurals, is selected, cf. cyst, gedryht, flyht, ärist, scrift. With verbs of the 5th, 6th and 7th class, however, past participle and infinitive stem are identical, cf. flöd, gift, graft, slieht, and from there analogical formations with the present tense stem as base are to be expected also in the other classes, cf. brecpa, ripp, äp-sweord, cwild, etc. In view of the fact that -th has led to a few new formation later on, we may assume that the fricative alternants were also productive in Old English. This is corroborated by such derivatives as fiscop, häletop based on weak verbs which themselves are derived from nouns by productive Old English word-formative processes. The stop alternants, on the other hand, have not led to new formations later on. Moreover, they occur more frequently with strong than with weak verbs, and the weak verbs they occur with can not be regarded as derivatives in Old English. This suggests that the attested formations are already old, and that the stop alternants were no longer active in Old English. 3.2.2.1. -el (alternants -/, -ol) Action nouns: bindele f. 'tying': bindan 'to bind', hwyrfel m. 'circuit, whirlpool': hweorfan 'to turn, revolve', swingel(e) f. 'stroke, blow' : swingan 'to beat, strike', pweäl η. 'washing' : pweän 'to wash', onsetl n. 'a sitting on' : onsittan 'to sit on', andetla m. 'confession' : andettan 'to confess', scendle f. 'reproach' : scendan 'to blame', preäl f. 'reproof : preäganjpreäwian 'to rebuke' // begangol m. 'worship' : begangan 'to worship'. Agentive nouns: after-gengel m. 'successor' : gangan 'to go', by de I m. 'herald, beadle' : beödan 'to command, summon', forridel m. 'forerider' : rldan 'to ride' // begangol m. 'worshipper' : begangan 'to worship', andfangol m. 'susceptor': onfön 'to receive, take', bcecslitol m. 'backbiter' : sfitan 'to slit, bite'.

Deverbal nouns in Old and Modern English: from stem- to word-formation

235

Objective nouns: scytel m. 'dart, missile': sceötan 'to shoot', hangelle f. 'an implement that hangs' : hönjhangian 'to hang', stricel m. 'teat' : strlcan 'to stroke' // bitol n. 'bridle' : bltan 'to bite'. Factitive nouns: scitel m. 'excrement' : scltan 'to shit', windel m. ' b a s k e t ' : windan 'to wind', hring-gewindla m. 'something rolled into circular form' : windan, fyndeI m. 'invention' : findan 'to find'. Instrumental nouns: sceacel m. 'shackle' : sceacan 'to shake', spinel f. 'spindle' : spinnan 'to spin', swingel(e) f. 'a scourge, whip' : swingan 'to beat, strike', tredel m. 'sole of the foot' : tredan 'to tread', perscel m. 'threshel, flail': perscan 'to thrash', pweäl η. 'what is used in washing, ointment' : pweän 'to wash'; hlcedel m. 'an instrument for drawing water, ladle': hladan 'to draw water', pwirel m. 'whisk' : pweran 'to stir, churn', gnidel m. 'pestle' : griidan 'to rub, grind together', slegel m. 'plectrum' : sleän 'to strike', stricel m. 'strickle' : strlcan 'to stroke', gyrdel m. 'girdle' : gyrdan 'to gird', sticel m. 'sting, goad' : stician 'to prick'. Locative nouns: smygel m. 'burrow, retreat' : smügan 'to creep', setl n. 'seat' : sittan 'to sit', stigel m. 'stile' : stigan 'to mount'. 3.2.2.2. Obviously, the strongest groups are Action and Instrumental nouns, the other types are much weaker. As in the case of the preceding suffix, it is difficult to tell whether -(e)l and -ol should be treated as different suffixes or as suffix alternants. Phonological similarity and lack of a clear-cut functional differentiation seem to point towards the latter solution. Historically, we have a similar situation as with the preceding suffix. The original Proto-Indo-European formative is -/-, which was augmented by stem-formatives, and could be preceded by linking vowels. Whether these secondary formatives differed semantically is difficult to assess; for Old English, no consistent difference exists. Since the various stem formatives, which kept the subtypes apart at least morphologically, no longer appear overtly in Old English, there does not seem to exist any cogent reason for treating these formatives as different suffixes rather than as suffix alternants. Again, all genders are represented, but there is a definite preponderance of the feminines among the Action nouns, and of the masculines among the Agent and Instrumental nouns; neuter nouns are rare. Masculines and neuters mainly belong to the strong declension, feminines mainly follow the weak declension. Many derivatives exhibit -/-mutation, cf. hwyrfel, cefter-gengel, bydel, pwirel, but others do not, cf. pweäl, andetla, scendle, begangol,

236

Dieter

Kastovsky

hangelle, tredel, perscel, etc. The overwhelming majority of the derivatives are based on strong verbs, partly on the full grade of the present (bindele, swingel, spinel, pwirel), partly on the reduced grade (rynel, smygel, gnidel, stigel, stricel, bydel), but there are also some derivatives from weak verbs (andetla, geoehtle, scendele, preäl, sticel). The preponderance of strong verbs as bases suggests that many formations are probably already quite old, since the strong verbs belong to the older stratum of the language. On the other hand, there are also some formations based on weak verbs. Since these latter tend to be younger than the strong verbs, and in many cases can be regarded as synchronic derivatives, it is quite likely that suffixes added to weak verbs were still productive in Old English. In the present case, this is corroborated by the fact that at least a few derivatives were formed by this suffix in the later stages of the language (Marchand 1969: 324), so that it can safely be assumed to have been at least mildly productive in Old English. 3.2.3.1. -els Action nouns: rädels m. jrädelse f. 'counsel, consideration' : rädan 'to advise', blgels m. 'inclination' : bigan 'to bend'. Objective nouns: brcedels m. 'carpet' : brcedan 'to spread, stretch o u t \ f e d e l s m. 'fading': fedan 'to feed'. Instrumental nouns: spennels m. 'clasp' : spannan 'to join fasten', scyttel(s) m. 'bar, bolt' : scyttan 'to bar, bolt', wrigels m. 'covering' : wreön 'to cover', wripels m. 'bandage, band' : wripan 'to twist, wrap, bind', cnyttels m. 'thread, sinew' : cnyttan 'to fasten', gyrdels m. 'girdle' : gyrdan 'to gird', smirels m. 'ointment' : smierwan 'to smear, anoint', sticels m. 'goad, stimulus' : stician 'to prick, goad'. Locative nouns: byrgels m. 'tomb' : byrgan 'to bury', hydels m. 'place of concealment' : hydan 'to hide'. 3.2.3.2. Obviously, the Instrumental group is the strongest, others are weak. The majority of the derivatives are strong masculines and are based on weak verbs; those derived from strong verbs {spennels, wrigels) have -/-mutation whenever this is possible. The suffix also belongs to the family of PIE -/- and represents Germanic *-isla~, *-islan- with subsequent metathesis. In view of the consonantal -s-extension, I have treated it as a separate suffix and not as an alternant of -elj-ol, although it is closely related to the latter. Its productivity does not seem to have been very strong.

Deverbal nouns in Old and Modern English: from stem- to word-formation

237

3.2.4.1. -en Action nouns: strong verbs, full grade: gehealden(n) f. 'observance' : gehealdan 'to observe', sie η f. 'sight' : seön 'to see', pigen f. 'taking of food' : picgan 'to take, receive', fillen f. 'falling, dropping' : feallan 'to fall', rndenn f. 'rule, government, estimation' : rädan 'to rule, deliberate', hengen f. 'hanging': hön 'to hang', wacen f. 'watching, vigil' : wacan 'to awake' / wacian 'be awake', swefen n. 'sleep, dream' : swefan 'to sleep' // reduced grade: ägniden f. 'rubbing off : ägmdan 'to rubb o f f , lygen f. 'lie' : leögan 'to lie', onflygen n. 'infectious disease' : fleögan 'to fly', druncen n. 'drunkenness' : drincan 'to drink' // weak verbs: bycgen f. 'buying': bycgan 'to buy', byrgen f., η. 'burying' : byrgan 'to bury', glemen f. 'care, heed' : gieman 'to care for', hierdenn f. 'hardening' : hierdan 'to harden', söcen f. 'search' : secan 'to search', fasten n. 'fasting' : fastan 'to fast', and quite a few others. Objective nouns: strong verbs, full grade: pigen f. 'what is taken as food' : picgan 'to take' // weak verbs: seienjseilen f. 'gift' : sellan 'to give', hcefen f. 'possessions' : habban 'to have', fasten f. 'fortress' : fastan 'to fasten, make fast'. Factitive nouns: strong verbs, full grade: rädenn f. 'reckoning, estimation' : rädan 'to reckon' // weak verbs: streöwen f. 'couch, bed' : streöwian 'to strew, scatter', sagen/segen f. 'saying' : secgan 'to say, speak'. Instrumental nouns: strong verbs, full grade: sien f. 'eye' : seön 'to see', hengen f. 'rack, cross' : hön 'to hang', hlceden m. 'bucket' : hladan 'to draw water' // weak verbs: lifen f. 'sustenance' : libban 'to l i \ e \ fasten n. 'fastener' : fastan 'to fasten', mcesten n. 'pasture' : mcestan 'to feed with mast'. Locative nouns: strong verbs, full grade: hengen f. 'rack, cross' : hön 'to hang' // weak verbs: byrgen f., η. 'grave' : byrgan 'to bury'. 3.2.4.2. Obviously, the strongest group are Action nouns, other types are again comparatively weak. The majority of the derivatives are either strong feminines or strong neuters, partly with, partly without mutation; masculines are the exception. There are quite a few derivatives from weak verbs based on the infinitive stem, but also some with "backmutation", i. e., based on the original root without mutation caused by the verbal stem formative (söcn, sagen). With strong verbs, the preferred stem form seems to be the 2nd participle, i. e., reduced grade for classes 1-3, and full grade as in the infinitive for classes 5-7. The type must still have been productive in

238

Dieter

Kastovsky

Old English in view of the sizeable number of attested derivatives, in particular Action nouns with the function of syntactic recategorization, which seems to be a clear indication of the productivity of a given pattern.

3.2.5.1. -end Action nouns: strong verbs: md-nimend f. 'rapine' : niman 'to take', blinnende f. : 'rest, ceasing' : blinnan 'to cease' // weak verbs: landbüend f. 'colony' : büan 'to dwell, live', peöfend f. 'theft' : peöfian 'to steal', leöht-sceäwigend f. 'light-seeing' : sceäwian 'to see'. Agentive nouns, personal: strong verbs: berend m. 'carrier' : beran 'to carry', biddend m. 'petitioner' : biddan 'to bid', bläwend m. 'inspirer' : bläwan 'to blow', galend m. 'enchanter' : galan 'so sing', ridend m. 'horseman, knight' : rldan 'to ride', swelgend m. 'glutton, drunkard' : swelgan 'to swallow', hearm-cwepend m. 'slanderer' : cwepan 'to speak', teönd m. 'accuser': teön 'to accuse' // weak verbs: andettend m. 'confessor' : andettan 'to confess', bodigend m. 'one who announces, preacher' : bodian 'to announce, preach', länend m. 'creditor, lender' : Icenan 'to lend', leerend m. 'teacher' : Iceran 'to teach', and many others. Agent nouns, animate: strong verbs: feper-berend m. 'feathered creature' : beran 'to bear', creöpend m. 'reptile' : creöpan 'to creep', lyft-fleögend m. 'bird' : fleögan 'to fly', slincend m. 'reptile' : slincan 'to creep, crawl', lagu-swimmend m. 'fish' : swimman 'to swim' // weak verbs: sun-folgend m. 'solisequia' : folgian 'to follow'. Agent nouns, inanimate: strong verbs: geötend m. 'artery' : geötan 'to pour', fore-ceorfend m. 'fore-cutter, front-tooth' : ceorfan 'to cut', dcelnimend m. 'participle' : dcelniman 'to take part', swelgend m., f., η. 'whirlpool, vortex' : swelgan 'to swallow, devour', slgend m. 'wave' : sTgan 'to sink, move, advance'. Experiencer nouns: strong verbs: on-cnäwend 'one who knows' : oncnäwan 'to know' // weak verbs: lufigend m. 'lover' : lufian 'to love'. Objective nouns: strong verbs: belifend m. 'surviver' : belTfan 'to remain, be left' // weak verbs: pearfend m. 'needy person' : pearfian 'to be in need'; gehceftend m. 'prisoner' : hceftan 'to arrest, detain'. Benefactive nouns: strong verbs: ägend m. 'possessor' : ägan 'to possess', andfengendjonfangendjonfönd m. 'receiver' : onfön 'to receive', begietend m. 'one who gets' : begietan 'to get', hleötend 'one who obtains' : hleötan 'to obtain by lot' // weak verbs: ähniend m.

Deverbal nouns in Old and Modern English: from stem- to word-formation

239

'possessor' : ähnian 'to possess', hcebbend m. 'one who has' : habban 'to have'. Instrumental nouns: weak verbs: (ge)bicnigend 'forefinger' : bicnian 'to beckon, make a sign', hälettend m. 'middle-finger, by which the sign of greeting is made' : hälettan 'to greet', eär-clänsend m. 'little finger, ear-cleaner' : clänsian 'to clean', tcecnend m. 'index finger' : tcecnan 'to make a sign'. 3.2.5.2. This is one of the most productive Old English suffixes, deriving primarily Agent nouns, but also, though less frequently, other types. All derivatives are strong masculines, with the exception of the few Action nouns and some occurrences of swelgend, which are feminine. The suffix does not cause -/-mutation and is always added to the infinitive stem. 3.2.6. -ere14 Action nouns: strong verbs: geligere, dirne-geligere n. 'adultery' : dirn-licgan 'to commit adultery'. Agent nouns, personal: strong verbs: bcecere m. 'baker' : bacan 'to bake', beätere m. 'beater, boxer' : beätan 'to beat', wceter-berere m. 'water-bearer' : beran 'to bear', biddere m. 'petitioner' : biddan 'to pray, ask', creöpere m. 'cripple' : creöpan 'to creep', freiere m. 'glutton' : fretan 'to eat up, devour', galere m. 'enchanter' : galan 'to sing', bTgengere m. 'worker': begangan 'to work', grafere/grafere m. 'engraver' : grafan 'to grave, engrave', leögere m. 'liar' : leögan 'to lie', swelgere m. 'glutton' : swelgan 'to swallow' / ätere m. 'glutton' : etan 'to eat' // weak verbs: andet(t)ere m. 'confessor' : andettan 'to confess', clänsere m. 'cleanser, purifier' : clänsian 'to clean', cwellere m. 'killer': cwellan 'to kill', fiscere m. 'fisher': fiscian 'to fish', folgere m. 'follower' : folgian 'to follow', leornere m. 'learner, disciple' : leornian 'to learn', and many others. Agent nouns, animate: weak verbs: fiscere m. 'kingfisher' : fiscian 'to fish'. Object nouns: strong verbs: sceäp-ätere m. 'carcas of a sheep' : etan 'to eat' 15 // weak verbs: sceäwere m. 'buffoon; mirror' : sceäwian 'to look at, see'. Factitive nouns: weak verbs: stTpere m. 'prop, pillar' : stipan 'to raise, build high'. Benefactive nouns: weak verbs: ägnere m. 'owner, possessor': ägnian 'to own'.

240

Dieter

Kastovsky

Instrumental nouns: weak verbs: pünere m. 'pestle' : pünian 'to pound'. Locative nouns: weak verbs: word-samnere m. 'catalogue' : samnian 'to collect', sceäwere m. 'watchtower': sceäwian 'to look at, observe, reconnoitre'. 3.2.6.2. The most productive group are again Agentive nouns, other groups are even less productive than with -end. All derivatives are strong masculines, with the exception of the Action nouns {dime-) geligere, which are strong neuters. The suffix does not cause -imutation, and the deriving base is the infinitive stem. Deviations from this pattern like bigengere, ätere, hcelgere are either denominal formations, or at least influenced by parallel nouns like bigeng, begenge, hcelu, hcelend. Alternations like grafan/grcefere, bacanj bcecere were probably still phonologically conditioned and thus do not affect the basic regularity of this extremely productive pattern. 3.2.7.1. -estre Agent nouns: strong verbs: gefestre i., forgifestre f. 'female giver', giefan, forgiefan 'to give', hleäpestre f. 'dancer' : hleäpan 'to jump, dance', wcescestre f. 'washer' : wascan 'to wash', bcecestre f. 'baker' : bacan 'to bake', wyrt-gcelstre f. 'a woman who uses herbs for charms' : galan 'to sing', rädestre f. 'female reader' : rcedan 'to read' // weak verbs: cennestre f. 'mother' : cennan 'to bring forth', hcelestre f. 'saviour' : hälan 'to heal, cure', hoppestre f. 'female dancer' : hoppian 'to leap, dance', Icerestre f. 'female teacher' : Iceran 'to teach', toeppestre f. 'female tavern-keeper' : toeppan 'to open a cask', wregestre 'female accuser' : wregan 'to accuse', fultumgestre f. 'female helper' : fultumian 'to help', huntigestre f. 'huntress' : huntian 'to hunt'. Experiencer nouns: weak verbs: lufestre f. 'female lover' : lufian 'to love'. All nouns are weak feminines and denote females; the suffix does not cause -/'-mutation itself, and the base is the infinitive stem. The suffix seems to have been reasonably productive, which is corroborated by the fact that it is still active in Modern English, although in a slightly different function. 3.2.8.1. -ett Action nouns: strong verbs, full grade: forgrindet n. 'grinding, pounding': forgrindan 'to grind', rewett n. 'rowing': röwan 'to row', hiwett n. 'hewing, cutting' : heäwan 'to hew, cut', mced-mcewett n.

Deverbal

nouns in Old and Modern English: from stem- to word-formation

241

'mowing of a meadow' : mäwan 'to mow', säwet η. 'sowing' : säwan 'to sow' // weak verbs: bcernett n. 'burning' : bcernan 'to burn', nirwett n. 'narrowness' : nirwan 'to constrain, repress, confine', rymett n. 'space, extension, clearance' : ryman 'to clear, widen'. Objective nouns: strong verbs: rewett n. 'ship' : röwan 'to row'. Factitive nouns: strong verbs: grafet(t) n. 'excavation' : grafan 'to dig'. Locative nouns: weak verbs: nirwett n. 'place of confinement, prison' : nirwan 'to constrain, confine'. 3.2.8.2. The suffix derives mainly Action nouns from both strong and weak verbs and causes -/-mutation in most cases. The basis is the infinitive stem. All derivatives are stronjg neuters. Since the listing above is almost exhaustive, the suffix was probably no longer really productive in Old English. This is corroborated by the fact that no class 2 weak verbs act as bases, and that the pattern has not led to any later formations. 3.2.9. -icge Agent nouns: weak verbs: sealticge f. 'dancer' : sealtian 'to dance', byrdicge f. 'embroideress' : byrdan 'to embroider', bearn-cennicge f. 'mother' : cennan 'to bring forth', hunticge f. 'huntress' : huntian 'to hunt'. The suffix is also denominal and a rival of -estre, i. e., it also derives weak feminine nouns denoting females. There may have been a dialect difference in the distribution of these two suffixes (cf. von Lindheim 1958; 1969). -icge does not cause -/-mutation and derives from the infinitive stem. 3.2.10.1. -ing\\ (alternant -ung) Action nouns: strong verbs/infinitive stem: binding f. 'binding' : bindan 'to bind', bläwung f. 'blowing, blast' : bläwan 'to blow', ceöwung f. 'chewing' : ceöwan 'to chew', delfing f. 'digging' : delfan 'to dig', hlöwung f. 'lowing' : hlöwan 'to low', sliting/slftung f. 'tearing' : slltan 'to tear', etc., etc. / non-infinitive stem: naming f. 'acceptance, bargain' : niman 'to take'; wiper-cwedung 'contradiction' : wijper-cwepan 'to contradict' // weak verbs: äcsung f . 'inquiry' : äcsian 'to ask', bcerning f. 'burning' : bcernan 'to burn', bäsnung f. 'expectation : bäsnian 'to expect', bisceopung f. 'confirmation' : bisceopian 'to confirm', brastlung f. 'rustling' : brastlian 'to rustle', eriung f. 'ploughing' : erian 'to plough', huntung f. 'hunting' : huntian 'to hunt' and many others.

242

Dieter

Kastovsky

Agentive nouns (all semantically collectives) : weak verbs: fyrdung f. 'militia, t r o o p s ' : fyrdian 'to march, go on an expedition', gaderung f. 'gathering, assembly': gaderian 'to gather', (ge)meting f. 'meeting, assembly' : (ge)metan 'to meet', (ge)mitting f. 'meeting, assembly' : gemittan 'to meet', (ge)samnung f. 'assembly, council' : (ge)samnian 'to gather', setlung f., 'an assembly of persons sitting' : setlan 'to settle, sit', pegnung f. 'retinue' : pegnian 'to serve'. Objective nouns: strong verbs: forgifung f. 'gift, d o n a t i o n ' : forgiefan 'to give', beorning f. 'incense' : beornan 'to burn' // weak verbs: ägnung f. 'possessions' : ägnian 'to own', breeding f. 'roast meat' : brcedan 'to roast', birning f. 'incense' : birnan 'to burn', clysing f. 'a closed place, clause' : clysan 'to close', huntung f. 'game' : hintian 'to hunt', gelapung f. 'assembly, congregation': (ge)lapian 'to invite, summon', scotung f. 'missile' : scotian 'to shoot', wilnung f. 'desired object' : wilnian 'to desire', and others. Factitive nouns: strong verbs: finding f. 'device, invention' : findan 'to find', snlpung f. 'cut, wound' : snipan 'to cut', onwriting f. 'inscription' : onwritan 'to write on', etc. // weak verbs: äesung f. 'question' : äesian 'to ask', bodung f. 'message' : bodian 'to tell, announce', bytlingjbylding f. 'building' : bytlian 'to build', gyddung f. 'poem, saying' : giddian 'to sing, recite', holung f. 'excavation' : holian 'to hollow out', screädung f. 'shred, paring' : screädian 'to shred, pare', timbrung f. 'building' : timbrian 'to build' and many others. Instrumental nouns: weak verbs: fering f. 'vehicle': ferian 'to convey, carry', froet(e)wung f. ' o r n a m e n t ' : frcettewian 'to adorn', gegaderung f. 'bond' : gaderian 'to gather, join together', hceftung f. 'fetter' : hceftan 'to bind, arrest', läcnung f. 'medicine': läcnian 'to heal, cure', wering f. 'dam' : werian 'to hinder, check' and others. Locative nouns: strong verbs: biding f. 'abode' : bidan 'to abide' // weak verbs: bying f. 'habitation' : büan 'to dwell', cyping f. 'market place' : ciepan 'to buy, sell', celing f. 'a cool place' : celan 'to be, become cool', ear ding jgeear dung f. 'habitation, dwelling' : (ge)eardian 'to dwell', lending!landing f. 'landing-place' : lendanj Icendan 'to land', manung f. 'place where toll is demanded' : manian 'to claim, demand', orpung f. 'pore' : orpian 'to breathe', wunung f. 'dwelling' : wunian 'to dwell', and a few others. 3.2.10.2. Thus, almost all typical deverbal groups are well represented, but Action, Objective and Factitive nouns clearly dominate. Often, more than one possibility exists for one and the same verb.

Deverbal nouns in Old and Modern English: from stem- to word-formation

243

The two forms, -ing and -ung, have to be regarded more or less as free variants in Old English, although originally they seem to have been in complementary distribution, -ing occurring with weak verbs of the 1st class and strong verbs, and -ung occurring with weak verbs of the 2nd class, cf. älan : celing vs. äcsian : äcsung. All derivatives are strong feminines. The suffix derives from both weak and strong verbs, usually without causing -/-mutation (exception: bying, perhaps denominal), and is added to the infinitive stem (exceptions: naming, wipercwedung, forsceapung 'mishap', perhaps also nominal influence?). It is extremely productive and functions like Modern English -ing deriving "mixed forms" of the type the shooting of the hunters and derived nominals like newspaper cutting, opening, etc. 3.2.11.1. -ing2 Agentive nouns: weak verbs: leasing m. 'false person' : leäsian 'to lie'. Objective nouns: weak verbs: flyming m. 'fugitive': flyman 'to cause to flee, banish', föstringm. 'foster-child': föstrian 'to foster, nourish', Using m. 'a free man' : Usan 'to release', räping-weard m. 'a keeper of prisoners' : räpan 'to bind, make captive', wcedling m. 'needy person' : wädlian 'be poor'. 3.2.11.2. The suffix thus derives personal nouns, primarily of the Objective type. The derivatives are all strong masculines, and as bases only weak verbs seem to occur. The suffix does not cause -imutation. On account of its functional and morphological properties — it derives exclusively masculine personal nouns — it has been treated not as a variant of feminine -ing/-ung, but rather as an independent suffix. Originally a denominal suffix (in this function it was much more productive in Old English), it was extended to deverbal derivation via nouns like leasing, flyming, where there was a verb (leäsian, flieman), which was in turn derived from a noun {leas, fleam), thus allowing a dual connection. In this way, a deverbal interpretation became possible, but the number of clearly deverbal derivatives has remained relatively small. 3.2.12.1. -ling Objective nouns: weak nouns: hceftling m. 'prisoner' : hoeftan 'to arrest, bind', hyrling m. 'hireling' : hyr(i)an 'to hire'/hyr 'hire', mcestling m. 'fatling' : mcestan 'to fatten', nldling m. 'slave, one who serves of necessity' : nidan 'to force, compel', rcepling m. 'prisoner'

244

Dieter

Kastovsky

: rcepan 'to bind', hwirßing m. 'that which turns, orb' : hwirfan 'to turn, revolve'. 3.2.12.2. The suffix thus derives only Objective nouns from weak verbs. It originated as a variant of masc. -ing2, and is thus also primarily denominal, being extended to deverbal derivation in the same manner as the preceding. All derivatives are strong masculines. The suffix does not cause -i-mutation itself. Productivity as a deverbal formative is rather weak. 3.2.13.1. -ness Action nouns: strong verbs/infinitive stem: äbläwness f. 'inflation' : äbläwan 'to blow up, breathe', blinness f. 'cessation' : blinnan 'to cease', brecness f. 'breach': brecan 'to break', gebregdness f. 'sudden movement' : gebregdan 'to move suddenly', ymbceorfness f. 'circumcision' : ymbceorfan 'to circumcise', scmness f. 'brightness, radiance' : sclnan 'to shine', etc., etc. // present participle: äblinnendness f. 'cessation': äblinnan 'to cease', ästandendness f. 'continuance' : ästandan 'to stand up, persist, continue' // second participle: (ge)corenness f. 'election, choice' : (ge)ceösan 'to choose', wipcwedenness f. 'contradiction' : wipcwepan 'to contradict', onfundenness f. 'experience, discovery' : onfindan 'to find out, discover', andfengness/onfangenness f. 'receiving' : onfön 'to receive', ägotenness f. 'effusion' : ägeötan 'to pour out', hrinenness f. 'touching, contact' : hrman 'to touch', üpärisenness f. 'resurrection' : üpärlsan 'to rise up', ofslegenness f. 'killing' : sleän 'to slay', etc. // weak verbs/infinitive stem: ambehtness f. 'service' : ambehtan 'to serve', andetness f. 'confession' : andettan 'to confess', gebeorhtness f. 'brightness, splendour': gebeorhtian 'to be, become bright, shine', bryrdness f. 'pricking' : bryrdan 'to prick', cosiness f. 'temptation' : costian 'to tempt', cwipness f. 'lamentation' : cwipan 'to lament', gicceness f. 'itch' : giccan 'to itch', ärärness f. 'raising, exaltation' : ärceran 'to rear, exalt', etc., etc. // present participle: gebetendness f. 'emendation, correction' : gebetan 'to improve', gTtsiendness f. 'avarice' : gitsian 'to covet, desire' // second participle: cirredness f. 'turning' : cirrart 'to turn', äcwellednessjäcwealdness 'killing, slaughter' : äcwellan 'to kill', fordemedness f. 'condemnation' : fordeman 'to condemn' gedreccednessjgedrehtness f. 'affliction, contrition' : gedreccan 'to vex, afflict', behcefedness f. 'restraint' : behabban 'to restrain' and many others. Objective nouns: strong verbs/present participle: ägendness f. 'prop-

Deverbal nouns in Old and Modern English: from stem- to word-formation

245

erty' : ägan 'to own' // second participle : ongeän-hworfenness f. 'what lies in the way, obstacle' : hweorfan 'to throw' // weak verbs/ infinitive stem: onbcernness f. 'incense': onbcernan 'to burn' // second participle : älegedness f. 'interposition, interjection': älecgan 'to put, lay down', foreset(ed)ness f. 'preposition': foresettan 'to put before'. Factitive nouns: strong verbs/second participle: äfundenness f. 'device, invention' : äfindan 'to find out' // weak verbs/infinitive stem: setness f. 'ordinance, regulation' : settan 'to set, ordain' (also: äsetness, insetness), streöwness f. 'bed, mattress' : streöwian 'to strew, scatter', tyderness f. 'branch' : tydrian 'to bring forth', tynness f. 'enclosed place' : tynan 'to enclose', ontynness 'aperture' : ontynan 'to open', gemangnessjgemengness f. 'mixture, spice': gemangian 'to mix' // second participle: gegaderedness f. 'gathering of diseased matter, abscess' : gegaderian 'to gather', tömearcodness f. 'description, enumeration' : tömearcian 'to describe', forsoegedness f. 'preface' : forsecgan 'to say beforehand, foretell', strecedness f. 'couch, spreading' : streccan 'to stretch, spread out', gefigedness f. 'figure, shape' : gefegan 'to construct, join, put together', gemetedness f. 'invention' : gemetan 'to find out, discover', gereccedness f. 'history, report' : gereccan 'to relate, tell', and others. Instrumental nouns: strong verbs/infinitive stem: anfengness f. 'receptacle' : onfön 'to receive' // weak verbs/infinitive stem : fidness f. 'nourishment' : ßdan 'to feed', gereordness f. 'food' : gereordian 'to take food, to refresh (oneself)', smireness f. 'ointment' : smirian 'to smear' // second participle: geßgedness f. 'conjunction' : geßgan 'to join'. Locative nouns: strong verbs/second participle: purh-farenness f. 'inner chamber' : purhfaran 'to penetrate' // weak verbs/infinitive stem: land-bünessjgebüness f. 'habitation' : büan 'to dwell', wuneness f. 'dwelling' : wunian 'to dwell' // second participle: behydedness f. 'secret place' : behydan 'to hide, conceal'. 3.2.13.2. The suffix thus derives primarily Action nouns, other types are relatively weak. All derivatives are strong feminines. As bases we find not only strong and weak verbs, but also adjectives and nouns. In a few cases the suffix seems to have caused -/-mutation, cf. bigengness 'cultivation': begangan, byrneness f. 'hard, fiery trial' : beornan 'to burn', behef(ed)nes f. 'need, utility' : behößan 'to have need to, be necessary', although this might also be explained by assuming derivation from a nominal base itself already exhibiting -/-mutation.

246

Dieter

Kastovsky

This suffix presents some morphological problems, which are interesting also from a more general point of view. The above examples have already indicated that the suffix derives from infinitive stems as well as from participles, and often we find doublets. There is, however, no functional difference between these derivatives, i. e., the meanings of the participial derivatives do not in any way differ from those of the derivatives based on an infinitive stem. This is most obvious if we look at some of the doublets, cf. töcnäwnessjtöcnäwenness 'knowledge, understanding', forgifnessj forgifenness 'forgiveness', onginnessjongunnenness 'undertaking', ongitnessjongitenness 'understanding', fortlcetnessjforlcetenness 'leaving, remission', dälnimeness/dälnimendness 'participation', töbrcednessjtobrädedness 'extent', gecignessjgecigedness 'call, summons', cwemnessjcwemednessisa.tisfaction\gedrefnessjgedrefedness 'trouble, disturbance', geßgnessjgefegedness 'conjunction', leörnessjleörendnesslleöredness 'departure, passing away', ällsnessjällsendness/älTsedness 'redemption', äwendendness/äwendedness 'change'. Clearly, all of these exhibit typical Action noun meanings, and not the meanings of 'passive state' which one would associate with a second participle, if this was also the deriving base from a semantic point of view (cf. Modern English -ness). Obviously, semantically speaking the verbal stem as such acts as basis, regardless of whether this is represented morphologically by the infinitive stem or one of the participles. This is even more obvious with the following group, and I will return to this problem below. 3.2.14.1. Zero derivatives 16 Derivation without an explicit suffix is already extremely common in Old English and is thus by no means a characteristic feature only of Modern English, as some handbooks claim. Historically, the following Old English patterns arose from the addition of nominal stem formatives to (verbal) roots. These formatives probably had some derivative-semantic function just like the consonantal suffixes, but primarily they determined the inflectional class of the lexical item in question. In the various Germanic languages, these stem formatives gradually coalesced with the case/number exponents or were lost altogether, i. e., they were no longer recoverable synchronically except if they had caused some phonological change in the root such as -/-mutation, and this loss resulted in a restructuring of the inflectional system. This is why in Old English words like NSg gum-α, GSg gum-an 'man' as well as in derivatives like wig-α, bor-

Deverbal nouns in Old and Modem

English: from stem- to word-formation

247

a, etc., -a, -an have to be regarded as inflectional (case/number) suffixes and not as derivational suffixes; the derivational element in these cases is not represented overtly, i. e., it is zero. We will thus have to postulate the following morphological analysis: wig/0-α 'fighter' : wig-an 'to fight'. For a more extensive justification of this analysis, cf. Kastovsky (1968: 81 ff.). 3.2.14.2. Action nouns Strong masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: drepe 'slaying' : drepan 'to strike', drinc 'drink' : drincan 'to drink', ece 'ache' : acan 'to ache', cyme 'coming' : cuman 'to c o m f i e l d 'fold, crease' : fealdan 'to fold', gangjgeng 'going' : gänj gangan 'to go', heöf 'lamentation' : heöfan 'to lament', fanglfeng 'grasp' : fön 'to take, grasp' // 2nd part., 2nd part. = pret. pi. stem : fröre 'frost' : freösan 'to freeze', cwide 'speech' : cwepan 'to speak', bite 'bite' : bit an 'to bite', drync 'drink' : drincan 'to drink' gripe 'grip, grasp' : gripan 'to seize', ryne 'run, flow' : rinnan 'to run, flow', stride 'stride' : strldan 'to stride', byge 'bending' : bügan 'to bend', cyrf 'cutting' : ceorfan 'to cut', eyre 'choice' : ceösan 'to choose', hryre 'fall' : hreösan 'to fall', wyrp 'throw' : weorpan 'to throw', and others // lengthened grade (pret. pi.): cet 'eating' : etan 'to eat' // full grade — o-ablaut (pret. sg. stem): drenc 'drink' : drincan 'to drink', hlemm 'sound, crash' : hlimman 'to sound, resound', sang 'singing' : singan 'to sing', stenc 'stench' : stincan 'to emit a smell', swenc 'trouble, tribulation' : swincan 'to labour, work' // weak verbs: bale 'belch' : bcelcan 'to belch', cid 'strife' : cidan 'to quarrel', drenc 'drowning' : drencan 'to drown', slice 'prick, stab' : stician 'to prick', wysc 'choice' : wyscan 'to wish'/ backmutation: gepanc 'thought, thinking' : pencan 'to think', etc. Weak masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: steorfa 'mortality' : steorfan 'to die', sceapa 'harm' : sceppan 'to injure' // reduced grade (2nd part., pret. pi. stem): fnora 'sneezing' : fneösan 'to sneeze', sopa 'sup, draught' : süpan 'to swallow, sup' // lengthened grade: sceta 'a holding of land' : sittan 'to sit' II weak verbs: hopa 'hope, expectation' : hopian 'to hope', plega 'play, fight' : plegan 'to play, fight', wena 'supposition': wenan 'to hope, imagine' / backmutation: sala 'sale' : sellan 'to sell'. Strong feminines: strong verbs/infinitive stem, infinitive = 2nd part, stem: hreöw 'sorrow' : hreöwan 'to distress, be sorry', faru 'journey, going' : faran 'to go', giefufgiefe 'gift, favour' : giefan 'to give', wcesc 'washing' : wascan 'to wash' // 2nd part., 2nd part. =

248

Dieter

Kastovsky

pret. pi. stem: bedu 'request' : biddan 'to ask', beäh-pegu 'ring-receiving' : picgan 'to receive', geswicu 'cessation': geswican 'to cease', gebygu 'bend' : gebügan 'to bend' // lengthened grade (pret. pi.): bräc 'breach': brecan 'to break', sprcec 'speech': sprecan 'to speak' // pret. sg. stem (full grade, o-ablaut): bäd 'expectation' : bldan 'to wait, expect', dräf/drcef a driving o u t ' : drlfan 'to drive', gräp 'grasp' : gripan 'to grasp', stalu 'theft' : stelan 'to steal' // weak verbs: äl 'burning' : celan 'to burn', hlyst 'hearing' : hlystan 'to hear', hunt 'hunt': huntian 'to hunt', lufu 'love': lufian 'to love', edwend 'change' : edwendan 'to change' / backmutation: cwalu 'murder, killing' : cwellan 'to kill', sagu 'say, speech' : secgan 'to say, speak'. Weak feminines: strong verb/infinitive stem, infinitive = 2nd part, stem: birce 'barking' : beorcan 'to b a r k \ f e o h t e 'fight' : feohtan 'to fight', swinge/swyrtge 'strike, blow': swingan 'to strike', blod-läte 'blood-letting': lätan 'to let' // weak verbs: äsce 'search, inquisition' : äcsian 'to ask', gicce 'itch' : giccan 'to itch', scearpe 'scarification' : scearpian 'to scarify', secge 'speaking' : secgan 'to speak', wcecce 'watch' : waccan 'to watch'. Strong neuters: strong verbs/infinitive stem, infinitive = 2nd part, stem: beorc 'bark(ing)' : beorcan 'to bark', geberst 'eruption' : geberstan 'to burst', delf 'digging' : delfan 'to dig', feoht 'fight' : feohtan 'to fight', rip 'harvest' : rlpan 'to reap', winn 'strife' : winnan 'to strive', gebcec 'baking' : bacan 'to bake', blöt 'sacrifice' : blötan 'to sacrifice', far 'journey' : faran 'to travel' // 2nd part, stem, 2nd part. = pret. pi. stem: bod 'preaching' : beödan 'to command, proclaim', bed 'supplication' : biddan 'to pray', broc 'break' : brecan 'to break', gehror 'fall' : gehreösan 'to fall', flit 'strife' : flitan 'to strive', swic 'deception' : swican 'to deceive' // lengthened grade (pret. pi.) : gewäge 'weighing' : gewegan 'to weigh' // pret. sg. stem: gebrcec 'breaking' : brecan 'to break', wrote 'suffering, punishment' : wrecan 'to avenge, punish' // weak verbs: gebirg 'taste' : gebirgan 'to taste', gecld 'strife' : gecidan 'to quarrel', reed 'reading lesson' : rädan 'to read', sicett 'sigh' : sicettan 'to sigh', gestyr 'movement' : gestyrian 'to move' / backmutation: weorc 'work': wyrcan 'to work'. 3.2.14.3. Agent nouns Strong masculines: weak verbs; läwe 'betrayer' : Icewan 'to betray', weard 'guard' : weardian 'to guard'. Weak masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: äw-breca 'law breaker' : brecan 'to break', winna 'opponent' : winnan 'to fight', cräwa 'crow' : cräwan 'to crow', cuma

Deverbal

nouns in Old and Modern English: from stem- to word-formation

249

'guest' : cuman 'to come', mere-fara 'seafarer' : faran 'to travel', nihtegala 'nightingale' : galan 'to sing', gifa 'giver' : giefan 'to give', spreca 'speaker' : sprecan 'to speak' // 2nd part., 2nd part = pret. pi. stem: bora 'one who bears' : beran 'bo bear', glida 'kite, vulture' : glTdan 'to glide' // lengthened grade: hläf-ceta 'domestic servant' : etan 'to eat' // pret. sing, stem: gestala 'one who steals with another' : stelan 'to steal' 11 weak verbs: andetta 'confessor' : andettan 'to confess', bytla 'builder' : bytlan 'to build', ciepa 'merchant' : cTepan 'to buy, sell', dema 'judge' : deman 'to judge', drefla 'driveller' : dreflian 'to drivel', hunta 'hunter' : huntian 'to hunt', sella 'giver' : sellan 'to sell', and others. Strong feminines (only collectives): strong verbs: faru 'troops, train' : faran 'to go, travel' // weak verbs: weard 'watch, body of men keeping watch' : weardian 'to watch'. Weak feminines: strong verbs/infinitive stem: fleöge 'fly' : βeögan 'to fly', cräwe 'crow' : cräwan 'to c r o w \ f e l d e - f a r e 'field-fare, thrush' : faran 'to go', nihtegale 'nightingale' : galan 'to sing', bi-genge 'a female worshipper' : begangan 'to worship' // 2nd part.: loc-bore 'one wearing long hair' : beran 'to bear' // weak verbs: hafenbläte 'haven-bleater, sea-gull' : blätan 'to bleat', cild-föstre 'nurse' : föstrian 'to nourish', gcers-hoppe 'grasshopper' : hoppian 'to hop'. 3.2.14.4. Objective nouns Strong masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: fleöt 'ship, raft' : fleötan 'to float', bring 'offering' : bringan 'to bring, offer', drinc 'drink' : drincan 'to drink' // 2nd part, stem: fang/feng 'what is taken, booty', fön 'to take' // lengthened grade: ät 'food' : etan 'to eat' // pret. sg. stem (o-ablaut): drenc 'drink' : drincan 'to drink', sang 'song' : singan 'to sing' // weak verbs: brcede 'roasted meat' : brädan 'to roast', gehaft 'captive' : hceftan 'to arrest' / backmutation: sand 'messenger' : sendan 'to send'. Weak masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: hweorfa 'spindle' : hweorfan 'to turn', drinca 'drink' : drincan 'to drink', unna 'grant' : unnan 'to grant', freö-läta 'one made free' : Icetan 'to let' 11 2nd part.: loca 'locker' : lücan 'to look', strong feminines: strong verbs/infinitive stem, infinitive = 2nd part, stem: unn 'grant, what is given' : unnan 'to grant', giefu 'gift' : giefan 'to give' // pret. sg. (o-ablaut) : stalu 'what is stolen' : stelan 'to steal' // weak verbs: lufujlufe 'object of desire' : lufian 'to love', send 'present' : sendan 'to send' / backmutation: läf 'what is left' :

250

Dieter

Kastovsky

läfan 'to leave', sand 'message' : sendan 'to send', talu 'tale' : tellan 'to tell, narrate'. Weak feminines: strong verbs/infinitive stem, infinitive = 2nd part, stem: fleöte 'raft, ship' : fleötan 'to float', unne 'grant' : unnan 'to grant', drcege 'dragnet' : dragan 'to draw, drag'. Strong neuters: strong verbs/infinitive stem, infinitive = 2nd part, stem: gield 'payment': gieldan 'to give, pay', rip 'what is reaped in' : ripan 'to reap', gebcec 'anything baked' : bacan 'to bake' // 2nd part, stem: gecow 'what is chewed, food' : ceöwan 'to chew', gescot 'weapon shot or hurled': sceötan 'to shoot' // weak verbs: gewrixl(e) 'what is given in exchange, payment' : (ge)wrixlan 'to change, barter' / backmutation: on-äl 'incense' : oncelan 'to burn'. 3.2.14.5. Factitive nouns Strong masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: hliet 'lot, fate': hleötan 'to cast lots, get by lot' // pret. pi., 2nd part, stem: bryce 'fragment' : brecan 'to break', cwide 'sentence, decree' : cwepan 'to speak', cyrf 'what is cut off : ceorfan 'to carve' 11 weak verbs: hyge 'thought': hycgan 'to think', soc 'soakings' : socian 'to soak'. Weak masculines: strong verbs: bita 'bit, morsel' : bltan 'to bite' // weak verbs: mearca 'space marked out' : mearcian 'to mark' / backmutation: hräca 'expectoration' : hräcan 'to reach, bring up', saga 'saying, story' : secgan 'to speak'. Strong feminines: strong verbs/infinitive stem, infinitive = 2nd part, stem: hläf-gebrecu 'fragment of bread': brecan 'to break', grafu 'cave' : grafan 'to dig' // 2nd part, stem: bund 'bundle' : bindan 'to bind' // pret. pi. stem: spräc 'what is said, speech' : sprecan 'to speak' // pret. sg. stem: snäd 'slice' : snlpan 'to cut' // weak verbs: screäd 'piece cut off : screädian 'to shred, peel' / backmutation: sagu 'tale' : secgan 'to speak', talu 'tale' : tellan 'to tell, narrate'. Strong neuters: strong verbs/infinitive stem, infinitive = 2nd part, stem: delf 'anything dug out' : delfan 'to dig', gesweorf 'filings' : (ge)sweorfan 'to file, polish', geweorp 'what is thrown up, heap' : geweorpan 'to throw', gewring 'what is pressed out, drink, liquor' : (ge)wringan 'to wring', bann 'command, edict' : bannan 'to command', hat 'promise, vow': hätan 'to promise' / 2nd part, stem: bod 'command' : beödan 'to command', bed 'prayer' : biddan 'to pray', broc, 'fragment': brecan 'to break', gescrif'what is prescribed' : scrlfan 'to prescribe', snid 'cut' : snlpan 'to cut', writ 'writing' : wrltan 'to write' 11 weak verbs: gediht 'composition' : (ge)dihtan 'to

Deverbal nouns in Old and Modern English: from stem- to word-formation

251

compose, write', gepeöde 'translation' : (ge)peödan 'to translate' / backmutation: weorc 'work, fortress' : wyrcan 'to work'. 3.2.14.6. Benefactive nouns Weak masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: äga 'owner' : ägan 'to own', med-gilda 'one who receives pay' : gildan 'to give, pay' // weak verbs: tipa 'receiver, grantee' : tipan 'to give, bestow'. 3.2.14.7. Instrumental nouns Strong masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: heard-heäw 'chisel' : heäwan 'to hew', weald 'bridle' : wealdan 'to have power over, control', wröt 'snout, trunk' : wrötan 'to root up' 11 2nd part., pret. pi. = 2nd part, stem: slege/slcege 'plectrum' : sleän 'to beat', snide 'saw' : snfpan 'to cut' // pret. pi. stem: cyrf 'cutting instrument' : ceorfan 'to carve, cut' // pret. sg. stem: scear 'plough-share' : seer an 'to cleave, cut' // weak verbs: mcest 'mast' : mcestan 'to fatten, mast', sweg 'instrument' : swegan 'to sound', wer(e) 'weir, dam' : werian 'to hinder, check'. Weak masculines: strong verbs/infinitive, infinitive = 2nd part, stem: heard-heäwa 'chisel' : heäwan 'to hew', sceafa 'plane' : sceafan 'to shave, scrape' // 2nd part, stem: ceäc-bora 'yoke for buckets' : beran 'to carry' // weak verbs: plega 'implement for a game' : plegan 'to play'. Strong feminines: strong verbs/infinitive stem, infinitive = 2nd part, stem: bän-beorg 'greave' : beorgan 'to protect', slä 'instrument used to beat up the weft' : sleän 'to beat', gehild 'defence from attack' : gehealdan 'to hold, guard, protect', wegu 'carriage' : wegan 'to move, transport' // lengthened grade (pret. pi. stem): beer 'bier, pallet' : beran 'to carry', sceär, 'shears' : scieran 'to cut, shear', wäg 'implement for weighing' : wegan 'to weigh' // weak verbs: feorhneru 'sustenance, food' : nerian 'to save, preserve', steörjstyr 'rule, regulation, punishment' : steöranjstyran 'to steer, guide' / backmutation: läd 'carriage' : lädan 'to carry, transport, mylen-waru 'milldam' : werian 'to hinder, check'. Weak feminines: strong verbs/infinitive stem, infinitive = 2nd part, stem: binde 'band' : bindan 'to bind', gearn-winde 'yarn-winder, reel' : windan 'to wind', wringe 'press' : wringan 'to press', seohhe 'strainer, sieve': seön/slhan 'to strain' // pret. pi. stem: pwäre 'pestle' : pweran 'to stir, churn', wäge 'scales' : wegan 'to weigh' 11 pret. sg. stem: band/bend m., f. 'band, bond' : bindan 'to tie' // weak verbs: handle 'handle' : handlian 'to handle', hirste 'frying-pan' :

252

Dieter

Kastovsky

hirstan 'to fry, roast', hwistle 'pipe' : hwistlian 'to make a hissing noise, whistle' // backmutation: pll-stampe 'pestle' : stempan 'to stamp, pound'. Strong neuters: strong verbs/infinitive stem, infinitive = 2nd part, stem: gebeorg 'that which protects' : (ge)beorgan 'to protect', fcer 'carriage, vessel' : faran 'to travel', gemet 'instrument for measuring' : (ge)metan 'to measure', gespann 'yoke, clasp' : (ge)spannan 'to join together' 11 2nd part, stem: hoi 'cover' : helan 'to cover, hide', loc 'lock, bolt' : lücan 'to lock' // 2nd part. = pret. pi. stem: lid 'ship' : llpan 'to go by sea', scrid 'carriage, litter' : scripan 'to go, move', gewrid 'covering, husk' : gewripan 'to wrap up, cover, bind' // weak verbs: gecnycc 'bond' : gecnyccan 'to bind together' 11 backmutation: tödäljtödal 'dividing point' : tödälan 'to divide'. 3.2.14.8. Locative nouns Strong masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: fleötlfllet 'place where vessels float, bay, estuary': fleötan 'to float', hel 'shelter': helan 'to conceal, cover', gangjgeng 'passage, drain, privy' : gangan 'to go', ingang 'entrance' : ingangan 'to enter', onheäw 'a block to hew on' : heäwan 'to hew', stepejstcepe 'step' : steppan 'to step' 11 pret. pi. stem: ryne 'path, course' : rinnan 'to run', syde 'decoction, water in which anything has been boiled' : seöpan 'to boil'. Weak masculines: strong verbs/infinitive stem, infinitive = 2nd part, stem: stiga 'path' : stlgan 'to go' sceäda 'parting of the hair' : sceädan 'to separate' // 2nd part. = pret. pi. stem: brand-rida 'firegrate' : ridan 'to ride' // weak verbs: llc-hanga 'gibbet' : hangian 'to hang'. Strong feminines: strong verbs/infinitive stem, infinitive = 2nd part, stem: here-beorg 'shelter, lodgings' : beorgan 'to shelter', ädfaru 'way to the funeral' : faran 'to go' // pret. sing, stem: dräfjdrcef 'drove, road' : drifan 'to drive', for 'way, approach' : faran 'to go' // weak verbs: ferh-neru 'refuge' : nerian 'to save, rescue', rest 'restingplace' : restan 'to rest', sät 'ambush, lurking-place' : sätan/sätian 'to lie in wait'. Weak feminines: strong verbs/infinitive stem, infinitive = 2nd part, stem: hunig-süce 'plant from which honey may be sucked, clover': sücan 'to suck', genge 'privy, drain': gangan 'to go', ütgenge 'exit' : ütgangan 'to go out', hllpe 'a place to leap from, jump over' : hleäpan 'to leap', sceäp-wcesce 'place for washing sheep' : wascan 'to wash' 11 2nd part, stem: gelege 'lair, bed' : gelicgan 'to lie'.

Deverbal nouns in Old and Modern English: from stem- to word-formation

253

Strong neuters: strong verbs/infinitive stem, infinitive = 2nd part, stem: gedelf 'quarry' : delfan 'to dig', gedrlf'dr'wc' : drifan 'to drive', far 'thoroughfare, road' : faran 'to go', Isen-grcef 'ironmine' : grafan 'to dig', weed 'ford' : wadan 'to go, wade' // 2nd part, stem: hlot 'the urn in which the lots are placed' : hleötan 'to cast lots', set 'seat, place where the sun sets, etc.' : sittan 'to sit', onrid 'riding horse' : ridan 'to ride' // weak verbs: büjby 'dwelling' : büan 'to dwell', purh-Jere 'secret chamber' : purh-jeran 'to penetrate into', gener 'asylum, refuge' : generian 'to save, preserve', gerest 'restingplace, couch' : (ge)restart 'to rest', gewraxl 'wrestling place, gymnasium': wraxlian 'to wrestle' // backmutation: weg-gedäl 'place where a road divides' : gedälan 'to divide, part'. 3.2.14.9. All semantic types are represented with the exception of Experiencer nouns, but this might also be due to insufficient data. Clearly, Action and Agentive nouns dominate — the selected examples roughly mirror the relative strength of the types —, but the other types are not weak, either. Derivation is made both from strong 17 and weak verbs, with and without -/-mutation, and in the case of strong verbs, practically all ablaut grades are used as bases, although with a preference for infinitive and past participle stems. All genders are represented in all types, but personal Agentive nouns are more or less restricted to masculines and feminines. 3.3. The overall system of the Old English deverbal nouns can now be summarized as in table (2) — see p. 254. 3.4. What are the most conspicuous features of this system, especially when we compare it with the Modern English system summarized in (1)? First of all, there is a striking continuity of the semantic patterns, which have not changed at all, with the exception, perhaps, of the more or less abortive attempt to introduce the category of temporal nouns. But this is not really surprising, since the possible major semantic-syntactic categories 18 of deverbal nouns are predetermined by the number of possible cases and head nouns of complement clauses. And it seems that only a limited subset of these tend to figure in deverbal derivation, which has proved fairly stable during the history of English. Thus the addition of a further category, viz. temporal nouns, has not proved successful. The only more noticeable change concerns the category of Benefactive nouns. These were an extremely weak type in Old English with just a

c^·

+1 +1 +1 +1 +

+ + + + c- •

+1 +1 +1 +1

feminine masculine

+

Locative

+

+

+

+ +

+ + +

+ +

+ +

+ +

+

+

+

τ

+ +

+ + +

Goal/ Benefactive

+ +

Instrumental

neuter

1

+

+ Φ + +

+ + +

1

I

+1

+ +

+ +

+ + + +

+ +

+

+

+1 +1

I

I

I

+ϊ +1

I

+ +

Τ

+ +

+ +

+ + + +

+ +

+ +

+ +

Τ

+

+

+ +

+

+

+

+

+

+

+

Experiencer

+

+

+

+

+

+

Agentive

+ +

+ +

+ +

τ

"ο 1 jo (Λ 'S" ^ "u t I

+ +

+ +

+ +

e ache vb., field > fold : fold vb., etc.). 3.5.3. Stem-allomorphy due to -/-mutation was, at least originally a purely phonological phenomenon and had nothing to do with the

Deverbal nouns in Old and Modern English: from stem- to word-formation

257

morphological system of the language as such. This is not true of stem-allomorphy caused by gradation, since gradation functioned as an exponent of morphological categories in the verbal paradigm, and in fact still does so in Modern English, although less regularly than it did in Old English. Why, then, did English lose almost all deverbal nouns characterized by ablaut, and why did those that were preserved (song, writ, frost, etc.) lose their derivational connection with the respective verbs, while German more or less preserved the status that also obtained in Old English? The answer to this question is that in English a more radical typological reorientation from stem-inflection and stem-formation to word-inflection and word-formation has taken place than in German. In order to realize what has happened here, a brief look at the inflectional system of Old English may be helpful. In the verbal paradigm, there was no form that was without an inflectional marker indicating person and number and/or tense, with the exception of the imperative singular, which was in fact characterized exactly by the absence of such a marker. Thus, if we subtract the person/number and tense markers, we are left with a stem that could not function on its own as a free form, i. e., as a word in a sentence. A stem became a word only when it was combined with an inflectional ending. The same is on the whole also true of many nouns, at least if we analyse end-e, wit-e, tal-u, gum-α, etc., as containing an inflectional ending signalling the nominative singular, which is probably the most appropriate interpretation. It is therefore not quite correct to speak of word-formation in Old English, because the results of the various processes creating new lexical items are stems, not words. Old English is thus characterized, typologically speaking, by stem-formation and stem-inflection, rather than wordformation and word-inflection. The situation is radically different in Modern English. There, we have an unmarked base form without any inflectional ending, which functions as infinitive and present tense form, or unmarked singular, and also as quotation form in the dictionary, and which is at the same time a word, a "free form". Inflectional endings are added to this base form in an almost agglutinative manner (cf. ox-en-s, Gen. PI.), so that Modern English is characterized by word-inflection and word-formation in the strict sense of the term. The change from the Old English to the Modern English type is due to the radical reduction of inflectional endings 20 , whose details need not concern us here. Its result, however, is extremely important for the functioning of the word-formation

258

Dieter

Kastovsky

patterns in Old English and Modern English, and is apparently the direct cause of the loss of the ablaut nouns. Another, indirect cause is, of course, the disintegration of the verbal ablaut patterns themselves on account of phonological developments (Great Vowel Shift), which in many cases disrupted the morphophonemic transparency of the noun-verb correlation, and the change of many ablaut verbs into the weak class. 3.5.4. Since in Old English no verbal form was without an inflectional ending, no form within the paradigm had a dominating role in the same way as the unmarked base form has in Modern English. Person/number exponents were added to the verbal stem, and in the case of the strong verbs, there were several stems, which were mainly correlated with tense distinctions — but not exclusively so, since the opposition ic wrät : we writon, Ί wrote' : 'we wrote' does not mark a tense distinction. In such a situation it would not be unnatural to regard all morphophonemic representations of the verb as more or less equally eligible bases for the derivation of deverbal nouns — as is still the case in German from a purely descriptive point of view, although new formations now probably are restricted to the infinitive stem. This, of course, is a synchronic description, since historically the ablaut differences are due to a number of different factors. But this would explain why all ablaut forms still occur as bases in Old English deverbal nouns. Additional support for this analysis comes from the behaviour of -ness in Old English, which, as we have seen, is added to infinitive stems as well as to first and second participles without any difference in the meaning of the resulting derivatives. In Modern English, on the other hand, where -ness no longer derives from verbal stems, the derivatives from second participles have lost their Action noun meaning and only denote a passive state, cf. the meanings of Modern English cursedness, devotedness etc. with the Old English examples above. If the various verb stems hat not been regarded as more or less equivalent in this respect, this Old English alternation could not be explained. On the other hand, there was of course a tendency to increasingly derive from the infinitive stem, which was due to the existence of weak verbs, because there the stem normally did not vary, or at least less so than with strong verbs 21 . It is thus not surprising that the typological change in the inflectional system, which created an unmarked base form not only in the nominal system — this happened in German, too — but also in the verbal

Deverbal nouns in Old and Modern English: from stem- to word-formation

259

paradigms — and this did n o t happen in German — did have major consequences for word-formation as well. It established the new base form ( = infinitive) as the only natural starting-point for deverbal derivatives, which automatically resulted in a dissociation of the ablaut nouns from their verbal bases and corresponding lexicalizations or analogical re-formations. This process was of course accelerated by the general decay of the Old English ablaut system in connection with the Great Vowel Shift, a process which would merit a separate, detailed investigation. As a result, stemallomorphy was reduced to a minimum, which is why those French and Neo-Latin patterns that involve alternations are clearly marked as non-native. Thus, changes in the inflectional system of the language had quite far-reaching consequences also for the area of word-formation. And in German, where the verbal inflectional system did not undergo such a restructuring, and where there is no verbal unmarked base form, the ablaut nouns still form a large, although perhaps no longer productive subset of deverbal nouns.

Notes 1. I should like to thank the participants of the conference, and in particular Jacek Fisiak, Bernhard Diensberg, Klaus Dietz, Johannes Gerritsen, Göran Kjellmer, Herbert Pilch, Jürgen Strauß and Anna Zbierska for many helpful comments and suggestions. 2. He is not completely consistent, however, cf. the suffixes -ard, -le or -th below. 3. Cf. the comprehensive bibliography on English word-formation by Stein (1973), and Quirk-Wrenn (21257: 104-119), Pilch (1970: 109-113; 117-119; 126-135); Wright (31925: 307-333) for Old English, Fisiak (1965; 1968: 111-120) for Middle English. 4. A single line marks the determinatum, which is correlated with the case roles just mentioned. 5. For details, cf. also Kastovsky (1982 a: 206 ff., 231-246). 6. Part of this material is documented in Kastovsky (1968, 1971). 7. The term "Action noun" is used here as a cover term for the various subtypes of nouns derived from the heads of complement sentences, cf. Kastovsky (1982 a: 189, 199 ff.). 8. As in The later arrivals would have quite a distance to walk. 9. This example is due to Johannes Gerritsen. 10. For the distinction between "naming" and "syntactic recategorization" (here nominalization) as the two basic functions of word-formation and the role of analogy in the former, cf. Kastovsky (1982 b), Mötsch (1979). 11. If the derivation is not made on a native basis of coining, as in reside ~ resident, oppose ~ opponent, vs. cool ~ coolant, pollute ~ pollutant. 12. From an historical point of view, such an analysis is probably justified; synchronically, however, such cases are probably better analysed as zero derivatives.

260

Dieter Kastovsky

13. Ε. g., G -φο- in OE äled 'fire' : älan 'burn', G -apa(n)- in OE migopa m., spiwepa m., huntojj m., etc. 14. For a detailed analysis of -er(e)-nouns, cf. Kastovsky (1971). 15. The status of this derivative, as well as of ätere above, is questionable, since it would be derived from a base with a lengthened grade instead of from the stem with the full grade (infinitive), which would be exceptional. Probably, it has to be analysed as denominal, i. e., as derived from ä t m., f., η. 'food', cf. Kastovsky (1971:294). 16. For a comprehensive treatment and documentation of this type, cf. Kastovsky (1968). 17. It might be argued that the inclusion of derivatives from strong verbs based on non-infinitival stems is unjustified, because, as is commonly assumed, this pattern was no longer productive in the individual Germanic languages. This can be countered by the observation that the ablaut derivations fit in perfectly into the overall semantic patterns, and that on account of their number it would seem unjustified to exclude them from a description of the Old English wordformation patterns that focuses on the analytic aspect. Moreover, it is by no means settled that this type was really completely dead, since there are quite a few synthetic compounds, where the second part does not occur on its own. But such formations typically occur with productive patterns. Nevertheless, it has to be admitted that the really productive type in Old English was of course the derivation from the infinitive/present stem. 18. Not, of course, minor, more specialized subgroups such as the type sit-in, lovein, etc., or anchorage 'fee for lying at anchor', etc., which are basically analogical formations due to the naming function of word-formation. 19. -end probably disappeared in the wake of the change of the present participle -ende to -ing (avoidance of "homonymie genante"?), with a concomitant extension of the range of -er, which took over the functions of -end. 20. Cf. Kastovsky (forthcoming) for a more detailed analysis of the restructuring processes involved. Note also that grammatical gender was replaced by natural gender. 21. Incidentally, the cases of backmutation listed above in connection with zero derivatives also contributed to the feeling that verbal stems in deverbal derivation were variable.

References Bosworth, Joseph — Toller, T. Northcote 1898 An Anglo-Saxon dictionary (Oxford: Oxford University Press). 1921 Supplement (Oxford: University Press). Clark Hall, John R. 1960 A concise Anglo-Saxon dictionary (4th ed.) (Cambridge: Cambridge University Press). Fisiak, Jacek 1965 Morphemic structure of Chaucer's English (Alabama: University of Alabama Press). 1968 A short grammar of Middle English. Parti. Graphemics, phonemics, and Morphemics (Warsaw: Panstwowe Wydawnictwo Naukowe). Hansen, Barbara - Hansen, Klaus - Neubert, Albrecht — Schentke, Manfred 1982 Englische Lexikologie. Eine Einführung in Wortbildung und lexikalische Semantik (Leipzig: Enzyklopädie).

Deverbal nouns in Old and Modern English: from stem- to word-formation

261

Kastovsky, Dieter 1968 Old English de verbal substantives derived by means of a zero morpheme (Esslingen: Langer). 1971 "The old English suffix -er(e)", Anglia 89: 285-325. 1982 a Wortbildung und Semantik (= Studienreihe Englisch 14) (Düsseldorf: Schwann & Bern: Francke). 1982b "Word-formation: A functional view", Folia Linguistica 16: 181-198. forthc. "Typological changes in the nominal inflectional system of English and German", Workshop on Natural Morphology, SLE Meeting, Poznan 1983. Koziol, Herbert 1937 Handbuch der englischen Wortbildungslehre (2. neubearbeitete Auflage 1972) (Heidelberg: Winter). Krähe, Hans — Meid, Wolfgang 1967 Germanische Sprachwissenschaft III. Wortbildungslehre (Berlin: de Gruyter). Lees, Robert B. 1960 The grammar of English nominalizations (Bloomington: Indiana University; repr. 1962: The Hague: Mouton). 1966 "On a transformational analysis of compounds: A reply to Hans Marchand", Indogermanische Forschungen 71: 1-13. Lindheim, Bogislav von 1958 "Die weiblichen Genussuffixe im Altenglischen", Anglia 76: 479-504. 1969 "Die weiblichen Genussuffixe im Altenglischen. Korrekturen und Nachträge", Anglia 87: 64-65. Marchand, Hans 1969 The Categories and types of present-day English word-formation (2nd ed.) (München: Beck). Mötsch, Wolfgang 1979 "Zum Status von Wortbildungsregularitäten", DRLAV Papier No. 20. Paris: 1^*0. Panagl, Oswald 1982 "Produktivität in der Wortbildung von Corpussprachen: Möglichkeiten und Grenzen der Heuristik", Folia Linguistica 16: 225-239. Pilch, Herbert 1970 Altenglische Grammatik (München: Hueber). Quirk, Randolph — Wrenn, C. L. 1957 An Old English grammar (2nd ed.) (London: Methuen). Stein, Gabriele 1973 English word-formation over two centuries. In honour of Hans Marchand on the occasion of his sixty-fifth birthday. (Tübingen: Narr). Wright, Joseph 1925 Old English grammar (3rd ed.) (Oxford: Oxford University Press).

E K K E H A R D KÖNIG

Where do concessives come from? On the development of concessive connectives 1. Introduction Although a popular topic during the first half of this century, concessive sentences have received relatively little attention in the last three decades, both from a historical and from a synchronic point of view. Those earlier historical studies of concessives in various European languages (Mensing, 1891; Burnham, 1911; Lerch, 1925; Quirk, 1954; Kuhlmann, 1891; etc.), however, are based on inadequate analyses of the synchronic facts, the semantic properties of concessive sentences and their relatedness to other adverbial clauses, notably conditionals. The present paper attempts to give a more general and more comprehensive account of the development of concessive sentences on the basis of historical developments observable in English, German and French, as well as on the basis of a synchronic study of concessive sentences in about fifty languages taken from language families all over the world. In contrast to other markers of adverbial relations, concessive connectives have a fairly transparent etymology. Regardless of whether such markers are realized by free morphemes (prepositions: despite·, conjunctions: although; conjuncts/pro-adverbials: nevertheless) or bound morphemes, they are, in most cases, complex in nature. Moreover, their components are easy to identify in both form and meaning and can easily be related to another, original or at least earlier, meaning. In European languages, at least, a wide variety of concessive markers is available and newcomers are constantly added to this class. Most of these connectives seem to have developed relatively late in the history of a language. There were very few, if any, clearly concessive markers in Old English (cf. Burnham, 1911; Quirk, 1954), Old High German (cf. Mensing, 1891) or Old French (cf. Lerch, 1925). In language acquisition, too, concessive constructions come in fairly late, by far later than

264

Ekkehard König

conditionals, for instance, which are acquired fairly late themselves (cf. M. Bowerman, 1983).

2. Conditionals, concessive conditionals and concessives Synchronically, a strict distinction has to be drawn between conditionals and concessives, on the one hand, and concessive or irrelevance conditionals and concessives proper, on the other (cf. Quirk et al., 1972: 749 ff.). Like most distinctions drawn in the area of adverbial clauses, these distinctions can best be made on the basis of semantic criteria, i. e., by looking at the semantic relations existing between the proposition expressed by the complex sentence and those expressed by the component clauses. Conditionals and concessives are easy to distinguish semantically: the former entail neither their antecedent nor their consequent, whereas the latter entail both of their component clauses. Irrelevance (concessive) conditionals like (1) have properties in common with both conditionals and concessives and thus are sometimes grouped together with the former and sometimes with the latter: (1)

a. Whether he is right or not, we must support him. b. However much advice you give him, he does exactly what he wants to do. c. Even if you drink just a little, your boss will fire you.

In contrast to simple conditionals, irrelevance conditionals relate a series of antecedent conditions to a consequent. This series can be specified by a disjunction as in (l)a., by a universal or 'free choice' quantifier as in (l)b. or by a scalar expression as in (l)c. A characteristic property of such 'free choice' quantifiers is that they signal free choice in the substitution of values for a variable in the antecedent. In other words, we do not find the usual restriction to a universe of discourse which is otherwise typical of universal quantifiers. Concessive conditionals of type (l)b. are related to Whquestions in the same way in which those of type (l)a. are related to Yes-No questions. In (l)c. a series of antecedents is specified by asserting a conditional relationship for an extreme (unlikely) value on a scale. By implication, this relationship can also be assumed to obtain for other values on the same scale. The consequent is asserted to hold under any of the conditions specified by the antecedent.

On the development

of concessive connectives

265

In most cases, concessive conditionals will therefore entail their consequent and, to simplify matters slightly, we will assume that this is always the case. The three constructions discussed so far can thus be distinguished on the basis of semantic criteria along the following lines: (2)

Conditionals typical form: if p, then q entailments: —

(3)

Concessive conditionals typical form: a. Whether ρ or not-p, q b. (Vx) (if p x , q) c. Even if p, q entailments: q

(4)

Concessives typical form: although p, q entailments: p, q

The semantic properties just mentioned do not exhaust the contribution that concessive connectives make to the meaning of a sentence. The use of such connectives implies that, by normal standards, there is an incompatibility or conflict between the facts described by ρ and q: 'p' and 'q' do not normally go together. This implication can roughly be described as follows: (5)

Normally (if p, then not-q)

Because of the abnormal projection behaviour of this implication — it survives embedding into negative, interrogative and conditional contexts (cf. (6)) — and because of its cancellability in reductio arguments (cf. (7)), we will regard it as a presupposition: (6)

a. It is not the case that Fred wants to go out for a walk even though it is raining. b. Does Fred want to go out for a walk even though it is raining? c. If Fred wants to go out for a walk even though it is raining, he must be crazy.

266

Ekkehard König

(7)

Even though I put this chemical into the water, the water does not change its colour. This shows that this chemical does not affect the colour of water in any way.

One reason why concessive conditionals have so often been grouped together with concessives is the fact that they, too, may carry an implication or suggestion of incompatibility between two situations. Given the fact that such conditionals relate a set or series of antecedents to a consequent, one of those conditions (either ρ or not-p or one substitution instance for the variable in px) will often be regarded as being in conflict with the consequent. Indeed, irrelevance conditionals can be expressed by paratactic constructions in which an extreme or unlikely value on a scale of possible values is specified: (8)

a.

You may drink a drop of alcohol and your boss will fire you. b. I can drink a whole bottle of alcohol and my boss won't fire me.

Even though concessive (irrelevance) conditionals and concessive sentences are grouped together in the description of many languages, a clear distinction between these two types of sentences can be made in most European languages. Sentences introduced by even though or although always entail their component clauses, sentences introduced by even if never entail their antecedents, although they may assume a factual character in certain contexts. We will return to this problem below.

3. Semantic domains and concessive connectives As has already been mentioned, concessive connectives in a wide variety of unrelated languages are composite in nature and have a fairly transparent etymology. In most cases other meanings and functions of these components can easily be identified and concessivity can thus be related to other semantic domains. By investigating the affinity between concessivity and other semantic domains, we can in fact identify various historical sources for concessive connectives. On the basis of such affinities, concessive

On the development

of concessive connectives

267

connectives can be divided into five groups (cf. König — Eisenberg, 1984): First, there is a close relationship between concessivity and universal quantification. Concessive connectives in many languages contain a component which is also used as a universal quantifier. The relevant affinity becomes even more obvious if in addition to genuine universal quantifiers we also include those connectives which directly express the notion of free choice like Latin licet. (i) Universal quantification (free choice) E. although, albeit, for all ..., all the same, however, G. bei all ..., allerdings', Lat. quamquam (cf. quisquis, 'whoever'), licet', F. toutefois, tout ... que, (com)bienque; Fin. vaikka 'although' (cf. vaikka kuka, 'whoever'); Hung, habär 'although' {ha ' i f , bärki 'whoever', ki 'who'); Russ. vse-taki, chotja; Pol. chociaz; Turk, her ne kadar {her 'every, all', kadar 'quantity'); Chin, swei rän ... döujye ('although' ... 'all'/'also') The examples listed above include expressions with straight-forward universal quantifiers as one component (e. g., E. although', G. allerdings', Russ. vse-taki; Turk, her ne kadar) as well as expressions that can also be used as 'indefinite elements' (free-choice quantifiers) in concessive conditionals (e. g., Finn, vaikka; Latin quamquam). A third group of connectives derives from verbs of volition (Russ. chotja; chotet' 'wish') or permission (Lat. licet 'it is permitted'). These notions of 'volition' or 'permission' play an important role in the development and make-up of so-called 'indefinite elements' in general (cf. Haiman, 1974; Coyaud — Hamou, 1976). Concessive connectives in many languages are composed of an originally conditional (e. g., E. i f ) or temporal (e. g., E. when) connective and/or an additive focus particle or emphatic particle like E. even, also, too. The particle invariably follows the conditional marker whenever these elements are added as clitics to a verb: (ii) Focus particle/emphatic particle and/or conditional/temporal conjunction E. even though, even so; G. ob-gleich, ob-wohl, ob-schon, wenn-gleich, wenn ... auch; Lat. et-si; Gk. ei-kai; Dutch: ofschoon; F. quand meme, lorsque meme; Finn, joskin, sittenkin; Serb.-Croat, iako; Jap. keredomo, mo; Malayalam: -enkil-um ('if-even/also'); Ewe: ne-ha ('if-and'); Turk, cond + de 'also'; Abkhaz: Cond + g 9 'even'; Skt.

268

Ekkehard König

yadiapi 'if-also/even'; Cambodian: dae(r) 'too, also, even so'; Susu: hali ...to 'even when/as'; Tibetan: Gerund + na-yan 'if also'; Sotho: le ha/ho 'even/also if Members of the third group all imply remarkable co-occurrence or co-existence of two facts as part of their literal meaning. This implication may be expressed in different ways: a) by asserting that one fact 'p' does not prevent another fact 'q'· b) by asserting that a fact 'q', or rather the agent responsible for it, does not pay any attention to another fact 'p'. c) by asserting simultaneity of 'p' and 'q' or the unhindered continuation of 'q' given 'p': (iii) Remarkable co-occurrence/co-existence E. nevertheless, notwithstanding, just the same, regardless, unimpressed by, still, yet\ G. gleichwohl, dennoch, ungeachtet, unbeschadet', Lat. nihilο minus; F. tout de meme, cependant, η 'empeche que; Sp. aunque; Turk, bununla beraber ('together with this'); Hung. migis ('still + too', 'nevertheless'); Sp. con todo, todavia; Chin, häi; Arabic: ma'a dalika 'together with that'; Jap. nägarä 'while/although' Members of the fourth group directly lexicalize notions of conflict, obstinacy, dissonance or our reactions to such situations: (iv) 'Obstinacy, conflict, disorder' E. in spite of, despite; F. en depit de, au mepris de; G. trotz, trotzdem; Sp. a pesar de (pesar 'sorrow, regret'); Chin, däo ('invert, pour, on the contrary, yet, but'); Dutch: ondanks (ondank 'ingratitude'); Arabic: ragman 'nevertheless' (ragama 'compel'); Turkish: ragmen Members of the following list can be grouped together because of their relatedness to causal connectives or because they mark in some other way the factual character of the clauses linked by them: (v) 'factual' G. zwar ('it is true'); Jap. no ni 'fact + essive', cf. no de 'since, because'; Cambodian: dbyt 'because/although', dbyt-tae

On the development

of concessive connectives

269

(-'only/but') 'although'; Mundari: - te - ο 'because + even'; Nahuatl: intlanel ('is-if-in fact'), immanel ('is-let be-in fact'); Indonesian: sementang 'just because', semantang pun 'although'

4. The development of concessive connectives In what follows, we will concentrate exclusively on the development of connectives listed in our first three groups. The two remaining groups are less important cross-linguistically. Moreover, the development of connectives in these two lists is clearly based on a process of 'bleaching' the notions of 'conflict', 'obstinacy' or 'truth'. Members of the first three groups provide the more interesting challenge for a theory of semantic change, specifically for a theory of grammaticalization. Our account of the historical development of concessive connectives will partly be a rational reconstruction of the connection between the literal, original, meaning of the elements in (i)-(iii) and their derived concessive meaning. This reconstruction will be supplemented by a look at the actual historical facts observable in the development of concessive constructions in English, German or French.

4.1. Irrelevance conditionals as source The formal make-up of members of our first two groups of connectives clearly points to irrelevance conditionals as historical source of concessive sentences. This should be obvious as far as our first group is concerned: in this case, the connectives in question are identical or nearly identical to those used as markers of irrelevance conditionals, i. e. as expressions of free choice. The elements of our second list point into the same direction. If we abstract from idiosyncratic developments within a specific language, members of this list can be said to have the following general form: (9)

even/also + if/when

That either conditional or temporal markers may be components of such connectives is not surprising in view of the fact that many languages do not differentiate between temporal and conditional

270

Ekkehard König

clauses in non-past tenses (cf. G. wenn, Dutch als, Japanese -tara, etc.). In fact, conditional connectives frequently derive from temporal connectives (cf. Traugott, 1983). If an additive focus particle like E. even, G. selbst, auch or French meme is added to conditionals and focuses on a part of the antecedent or the antecedent itself, the simple conditional is changed into a concessive conditional. Let us assume that the general form of even-if conditionals can be represented as follows (cf. König, 1981): (10)

(Even, ρ, (λ, χ (if χ, then q)))

The contribution that even and related particles make to the contribution of a conditional can now be described as follows:1 (a) they presuppose that there is an alternative to 'p' which satisfies the variable in the open conditional sentence; (b) they presuppose that the value given, i. e., ' p \ is the least likely and therefore most surprising value for the open sentence in question of all values under consideration in a certain context. Note that this contribution of even corresponds exactly to two of the ingredients which we found to be essentially involved in the meaning of concessive conditionals, namely a series of antecedent conditions one of which is in conflict with the consequent and therefore an unlikely candidate for the conditional relationship. The other, remaining ingredient of irrelevance conditionals is also a property of, at least certain, even-if conditionals. As pointed out in Bennett (1982), such conditionals entail (or rather presuppose) their consequent, whenever even focuses on the whole antecedent rather than a subpart of it. Thus, only the first example presupposes the consequent: (11) a. Even if he is a little slow, he is actually very intelligent. b. Even if you drank just a drop of alcohol, your boss would fire you. The assumption that concessives derive from conditionals, notably from concessive (irrelevance) conditionals, can be based on the formal similarity exhibited by connectives in the two constructions (cf. (i), (ii)) as well as the following historical and synchronic facts: (a) There is sufficient semantic similarity between irrelevance conditionals and concessive sentences to make the assumption of a common origin and a later split a plausible one. Concessives pre-

On the development of concessive connectives

271

suppose an incompatibility between the facts expressed by the component clauses. Irrelevance conditionals suggest that there is a conflict between one condition of the series denoted by the antecedent and the situation described in the consequent. The only thing that has to change if an irrelevance conditional is to develop into a concessive construction is the relationship of the antecedent to the real world. (b) In Old English and Middle English, as well as in Old High German, conditionals and concessives were not clearly distinguished. In Middle High German and even Early Modern German, both conditional and concessive clauses were introduced by either the conjunction wenn or ob. A concessive relationship was signalled by additional markers: adversative conjuncts or proadverbials in the main clause (e. g., doch) or focus particles {auch, gleich, schon, noch, joch) in the subordinate clause (cf. Mensing, 1891:69-70). But these expressive devices only differentiated between conditionals, on the one hand, and irrelevance conditionals and concessive sentences, on the other. The following examples taken from Behagel (1928: 51-52) can be interpreted either as irrelevance conditionals ( = even if p, q) or als concessive sentences ( = even though p, q), depending on whether the antecedent is considered as factual or not: (12) a.

Wenn ewer sünde gleich blutrot ist, soll sie doch schneeweis werden. b. Ob auch das Roß sich grauend bäumt, nicht rast ich bis sie sich gewöhnen.

In Modern German, the former conditional conjunction ob and the present conditional and temporal conjunction wenn have combined with particles like gleich, schon, wohl, zwar to form genuine concessive conjunctions: wenngleich, wenn auch, obwohl, obschon, obgleich, obzwar. The following examples show that though φαj, peah) was still used in the sense of 'even if in concessive conditionals in both Middle English and Early Modern English:2 (13) a.

Com je pere, je be kylled, may pe knyjt rede Trawe je me pat trewely, paj je had twenty lyues to spende.

Ill

Ekkehard König

'If you go there, ... you will be killed ... even if you had twenty lives to lose.' (Gawain 2111-12) b. I'll speak to it though hell itself should gape and bid me hold my peace. (Shakespeare, Hamlet, I.II.) This use of peah is also well documented for Old English (cf. Burnham, 1911; Quirk, 1954). Here, too, it was a process of 'strengthening' (i. e., adding the quantifier al or the particle even) that led to the development of genuine concessive conjunctions: although and even though (cf. Visser, 1972: 905 ff.). Further evidence for the assumption that though (peah, paj, etc.) was originally used as a conditional conjunction is provided by the fact that if and though are still interchangeable in combination with as (as if = as though). Even though irrelevance conditionals and concessive constructions are clearly differentiated in both Modern English and Modern German, the former can still be used in the sense of the latter under certain contextual conditions. It will be shown below that irrelevance conditionals introduced by even if in English or auch wenn in German may assume a factual character in certain contexts, even though they do not entail their antecedents. (c) Irrelevance conditionals provide a constant source for the development of concessive sentences. In addition to the elements listed in our first two groups, the French connectives malgre and malgre que also developed out of a marker of'irrelevance': bon gre, mal gre ('willy-nilly'). In English, regardless and anyway, clearly markers of irrelevance (concessive) conditionals in sentences like (14), have developed a concessive use in the recent history of English: (14) a. He went out for walks, regardless of the weather. b. You can give me your letter. I have to go to the post office anyway. Note that anyway or regardless only have a concessive interpretation ( = 'nevertheless') whenever the two sentences linked by these conjuncts express facts known to be normally incompatible: (15) a. I knew it was useless. I did it anyway. b. I had not invited him. He came regardless.

On the development

of concessive connectives

273

(d) Given that conditionals (both simple conditionals and irrelevance conditionals) do not entail their subordinate clauses and given that concessive sentences do, the former can only change to the latter if they lose their hypothetical character. Now it is a wellknown fact that the (generalized) clausal implicatures (Pp, Ρ—,ρ) (Levinson 1983: 137) carried by a conditional utterance of the form if ρ, then q can be cancelled in a context where 'p' is given3: (16) A: I was in France for a year. B: If you were in France for a year, your French must be excellent. In contexts such as these conditionals lose their hypothetical character and assume a factual character. Exactly the same phenomenon may be observed in connection with even-if conditionals: (17) a. He looked at me kindly, if somewhat sceptically. b. It was the loneliness of the neighbourhood, they supposed, that kept the house next to theirs empty ... The house stood two hundred yards from the Bartleby's and A. liked looking out of the window now and then and seeing it, even if it was empty. (Highsmith, Suspension of Mercy, 6) c. ' / believe you're looking at my freckles' ' Yes, I was. I like them And then I made up my mind that even if I did like her freckles, I was going to find out where I was at ... (Caine, Double Indemnity, 16) It is precisely in such contexts that evew-z/conditionals (or concessive conditionals in general) and concessive sentences are practically indistinguishable. So we may assume that in earlier stages of English and German concessives and irrelevance conditionals were kept apart purely on the basis of contextual factors. At a later stage concessive sentences were distinguished as a separate type of adverbial clause. Another contextual condition that favours the development of concessive constructions out of concessive conditionals is illustrated by the following example: (18)

However (no matter how) hard he tried, he could not lift the stone.

IIA

Ekkehard König

In cases such as these, we know in fact that only serious attempts (i. e. trying very hard) will lead to success. In other words, it is only values at the upper end of the scale that play a role. So the meaning of such sentences is not much different from that of genuine concessive sentences like the following: (19) a. Even though he tried very hard, he couldn't lift the stone. b. He tried very hard. However, he couldn 't lift the stone. Given this similarity, it should not come as a surprise that it is especially indefinite elements of manner (i. e. free choice quantification over manner) like E. however, Dutch hoewel, Lat. quamquam, F. (com)bienque, G. wiewohl that have developed into concessive connectives.

4.2. Conversational implicatures become conventionalized The problem of explaining the development of the connectives listed in our third group practically amounts to reconstructing the path that leads from the literal meaning of those connectives ('simultaneity', 'concomitance', 'unhindered continuation') to their derived concessive meaning. In the case of these connectives, it is plausible to assume, however, that the concessive presupposition (or conventional implicature) originated as a conversational implicature which later became part of the conventional meaning of the connective. In order to reconstruct this development, it seems useful to begin by considering expressions with an identical, or at least similar, literal meaning which only have a concessive reading in certain contexts. Among such expressions or constructions we find the coordinating conjunction and, the construction as Adj as NP be, the conjunctions when and while, the adverbial at the same time in English, the construction tout en W-ant in French and the so-called adverbial participle or gerund in Russian or Italian, to give only a few examples. The following pairs of sentences show that these expressions and constructions, which all express concomitance as part of their literal meaning, may have a concessive reading in some contexts, but lack such a reading in other contexts: (20) a. I have to do all this work and you are watching TV. b. Fred is reading and Mary is watching TV.

On the development

of concessive connectives

275

(21) a. Poor as he is, he spends a lot of money on horses. b. Rich as he is, he spends a lot of money on horses. (22) a.

It is not easy to find examples of services that are of general social benefit, and at the same time, not costly. b. This is exactly what people expect us to do and, at the same time, it is very sensible policy in general.

(23) a.

There was a funny smile on Dickie's face, as if Dickie were pulling his leg by pretending to fall in with his plan when he hadn't the least intention of falling in with it. (Highsmith, The talented Mr. Ripley, 75). b. This road can be very slippery when it is wet.

(24) a.

II etait d'accord, tout en gardant ses objections damentales. b. II chantait, tout en travaillant.

fon-

It is quite plausible to assume that examples such as (20)-(24) show us a great deal about the historical origin of the concessive implications which are now part of the conventional meaning of the connectives listed in group (iii). In certain contexts, so it seems, the literal meaning of expressions from certain semantic domains ('simultaneity', 'concomitance', etc.) is augmented by concessive inferences. The universal character of this process suggests that this augmentation is based on conversational maxims and the implications in question are conversational implicatures. But which maxims exactly are involved and how is it possible to construct an argument that would lead from those maxims and the literal meaning of utterances like (20)-(24)a. to the concessive implicature? It is quite clear that the inferences in question are not based on violations of conversational maxims and do not require specific situational contexts except that two facts are known to be normally in conflict. It follows that, if anything, they can only be generalized conversational implicatures. None of the Gricean maxims (Quality, Quantity, Relevance, Manner), however, seems to play any role in the calculation of these inferences 4 . In this respect they are similar to the causal implications sometimes carried by utterances expressing a temporal sequence between two events (post hoc ergo propter hoc), which have also come to be conventionally associated with originally temporal connectives in many languages (e. g., E. since). For such and similar inferences, Atlas — Levinson (1981) have proposed an

276

Ekkehard König

additional principle or maxim, the principle of informativeness, which in some cases allows us to read into an utterance more information than it actually contains. It seems that this principle, which roughly states that the best interpretation of an utterance is the most informative proposition among competing interpretations that is consistent with the common ground, is crucially involved in the concessive interpretation of (20)-(24)a. and also in the development of the concessive connectives listed in (iii). Note that the concessive inference (if p, then normally not-q) is only added to the literal meaning of the examples given above whenever it is consistent with background (i. e., world) knowledge about what may or may not go together. In addition to this principle of informativeness, the pragmatics of negation seem to play an important role in the development of the concessive connectives under discussion. It is a well-known fact that negatives are only uttered in a context where corresponding affirmatives have already been discussed, or else where the speaker assumes the hearer's belief in — and thus familiarity with — the corresponding affirmative (cf. Givon, 1978: 179). To what kind of inference does this restriction on the use of negatives, which seems to be based on the maxim of Quantity, lead in the case of those connectives in (iii) which contain a negative element? Consider the literal meaning of the following French connective: (25)

p. N'empeche que q.

It seems quite plausible to assume that an utterance of such a form is only informative if its literal meaning is augmented by the inference 'if p, then normally not-q'. Analogous additions can be assumed to have led to the concessive meaning of nevertheless, nonetheless, notwithstanding and their counterparts in other languages. Both the principle of informativeness and the pragmatics of negation seemed to have played a role in the process that led to a concessive reading of E. stillIyet and related elements in other languages. Note that the literal meaning of sentences like the following can always be augmented by a concessive implicature if this is consistent with the background knowledge of speaker and hearer: (26)

It is midnight and he is still working.

On the development

of concessive connectives

111

Such an augmentation is even more likely if the second clause contains a negation: (27)

It is midnight and he still has not stopped

working.

The constraint on the use of the negation mentioned above requires that the corresponding affirmative be under consideration in this case. A broadening out of the notion 'continuation' implied by still in sentences like (26) and (27) could have given additional support to this development of a concessive reading. Originally, still was only used as a verb phrase operator in the sense of 'quietly' 5 . In the course of Middle English this word came to be used as a sentential adverb with the meaning 'continuation (of a state or process) u p to an imminent boundary'. This overall meaning is the joint effect of several contributions made by still to the meaning of a sentence: still does not affect the truth conditions of a sentence in any way, but presupposes that the state or process denoted by the rest of the sentence also obtained before the interval (I 0 ) determined by the tense of the sentence (i. e., ρ before Io) and that there is an imminent change of state (i. e. not-p after I 0 ): (28) a. b

He is still sleeping ( = still p) I, < I 0 < h ρ ρ not-p

This implication of temporal continuation is widened to something like 'continues to be true in several possible worlds' if still is used in the consequent of a conditional: (29) a.

(Even) if they offer me a lot of money, I am still not going to do it.

If we let 'w p ' stand for the antecedent world in which the consequent ('q') is asserted to be true, we can represent the contribution made by still to the meaning of (29)a. as follows: (29) b.

wt < w p < w2 q q not-q

Conditionals like (29)a. differ from the examples discussed before insofar that the concessive implication is not due to a simple augmentation of a basic temporal meaning.

278

Ekkehard König

To summarize, the concessive reading of Ε. still/yet, G. dennoch, F. encore que, etc., can be assumed to be the result of both an augmentation of a temporal meaning by implicatures based on conversational maxims and on the pragmatics of negation and a widening of temporal notions. The question concerning the exact sequence of these two processes and interaction between them is beyond the scope of the present paper.

5. Four uses of anyway Finally, I would like to return to a conjunct briefly discussed above, its various uses in Present-Day English and the semantic changes involved in the development of those uses. Anyway, the conjunct in question, is a member of a group of related elements containing a universal or 'free choice' quantifier as one of their components and a general, non-specific, noun or pronoun as the other: anyhow, at any rate, in any case, in any event, at all events, however. Expressions of this type can have at least four different uses in Present-Day English: (A) concessive-conditional use ( = whether ρ or not-p) (30)

I don't want to invite Fred. He won't I wouldn't come anyway.

(B) concessive use ( = nevertheless) (31)

I had not invited him. (But) he came anyway.

(C) restrictive use ( = at least) (32)

Natural language expressions tend to have simple, stable and unitary senses (in many cases anyway).

(D) change of topic/return to previous topic (33)

A. (answering the phone) Well, I'm afraid you're out of luck this time ... I imagine she's gone out somewhere. Anyway, she isn't here.

Not all of the expressions listed above have all of these uses, but every expression has at least two or three of these uses. Moreover,

On the development

of concessive connectives

279

expressions of the same formal make-up and the same etymology show a similar range of uses in other languages. Examples are auf jeden Fall, in jedem Fall, auf alle Fälle, jedenfalls, wie auch immer in German, en tout cas, de toute fagon in French, in ieder (elk) geval in Dutch, 'ala hull hal in Egyptian Arabic, herhalde in Turkish, etc. Of the four uses under discussion, the use of anyway as a marker of a concessive-conditional relationship is clearly the primary one. This use of anyway can be observed whenever ρ and q, the sentences linked by this conjunct, denote eventualities that typically go together, as in (30) or in the following example: (34)

You can give me your letter. I have to go to the post-office anyway.

The contextual conditions relevant for the development of a concessive use of anyway are the very opposite of the conditions just mentioned. P, Q anyway has a concessive interpretation whenever ρ and q denote facts and the facts are known to be normally in conflict. Another way of looking at the difference between (30) and (31) is to say that in (30) alternative conditions ('p' and 'not-p') are under consideration which are both irrelevant for the consequent q, whereas in (31) the unfavourable condition is asserted as a fact together with q. The restrictive reading, too, derives from a basic concessiveconditional reading. In this and the fourth use, anyway relates to epistemological rather than the causal notions. Asserting that a fact expressed by a preceding utterance is irrelevant for a fact mentioned subsequently amounts to emphatically asserting the truth of the second statement. The contextual condition responsible for the restriction or retraction expressed by anyway in such cases is the specific relationship between 'p' and 'q', the two sentences or clauses linked by anyway. In (32), the first clause ( = p) expresses a much stronger claim than the emphatically asserted (reduced) second clause ( — ρ in many cases). So, what the speaker does in using anyway is in fact retracting a stronger claim and replacing it by a weaker one in accordance of the Gricean maxim of Quality. The fourth use of anyway is based on a transfer of the notion 'irrelevance' from a conditional relationship between propsitions to the idea of irrelevance of parts of a conversation to a general topic.

280

Ekkehard König

6. Conclusion The development of concessive connectives is based on (i) a process of bleaching involving such motions as 'spite', 'ingratitude' or 'conflict', (ii) the conventionalization of conversational implicatures carried by assertions of co-occurrence and co-existence of two facts and (iii) the factual use of irrelevance conditionals in certain contexts as well as the meaning of additive focus particles like E. even. 'Concessivity' is the result of semantic changes involving any of these notions, constructions and contexts in a wide variety of languages. The fact that these different sources led to the same result is one reason for the wide variety of concessive connectives available in many languages. By revealing the essentially derived nature of concessivity, the present paper offers an explanation for the relatively transparent etymology and the late acquisition of concessive connectives. The view that the development of grammatical meaning is motivated principally by unfulfilled communicative needs does not get any support from the facts discussed in this paper. Connectives with a concessive meaning or use are constantly added to a language, even if a large number of them is already available. What seems to play an important role in the development of such connectives is a well-known principle of creative language use, namely the constant attempt to express the same meaning in other words.

Notes

1. 2. 3. 4.

5.

The research reported on here was initially undertaken at the Netherlands Institute for Advanced Study in the Humanities and Social Sciences (Ν. I. A. S.), Wassenaar. I wish to express my gratitude for this fellowship and for help of the Institute staff. Cf. Karttunen — Peters (1979) and Bennett (1982) for a detailed discussion of the meaning of even. We may also note that Old English f>eah is related to the conditional connective pauh (pau 'in that case' + -h 'and/also') in Gothic (cf. OED, s. v. though). For a detailed discussion of such conditionals, cf. Akatsuka (1983). The only maxim that could be argued to play a role is the maxim of Quantity. There are many things going on simultaneously with a given state, process or event 'p' which are hardly worth mentioning. To point out these facts and their simultaneity can be totally uninformative and perhaps also irrelevant. For a discussion of the semantic development of E. still and yet, cf. König — Traugott (1982).

On the development of concessive connectives

281

References Akatsuka, Noriko 1983 "Conditionals are context-bound", On conditionals, edited by E. C. Traugott et al. Atlas, Jay David — Levinson, Stephen C. 1981 "It-clefts, informativeness, and logical form: radical pragmatics", Radical pragmatics, edited by P. Cole (New York: Academic Press), 1-61. Behagel, Otto 1928 Deutsche Syntax (Heidelberg: Germanistische Bibliothek). Bennett, Jonathan 1982 "Even i f ' , Linguistics and Philosophy 5: 403^18. Bowerman, Melissa 1983 "First steps in acquiring conditionals: some cognitive, semantic, formal, and pragmatic considerations", to appear in On Conditionals, edited by E. C. Traugott et al. Burnham, Joan M. 1911 Concessive constructions in OE prose (New York: Holt). Coyaud, Maurice — Khaled Ait-Hamon 1976 "Indefinis et interrogatifs", Semantikos 1.3: 83-88. Givon, Talmy 1978 "Negation in language: pragmatics, functions, ontology", Syntax and Semantics 9, edited by P. Cole (New York: Academic Press), 62-112. Haiman, John 1974 "Concessives, conditionals and verbs of volition", Foundations of Language 11: 341-360. Karttunen, Lauri — Peters, Stanley 1979 "Conventional implicature", Syntax and Semantics 11: Presupposition, edited by Ch. Oh — D. A. Dinneen (New York: Academic Press), 1-56. König, Ekkehard 1981 "The meaning of scalar particles in German", Words, worlds, and contexts, edited by Eickmeyer, H. J. — H. Rieser (Berlin: de Gruyter), 107-132. König, Ekkehard — Traugott, Elizabeth C. 1982 "Divergence and apparent convergence in the development of yet and still", BLS 8: 170-180. König, Ekkehard — Eisenberg, Peter 1984 "Zur Pragmatik von Konzessivsätzen", Pragmatik in der Grammatik, edited by G. Stickel (Düsseldorf: Schwann), 313-332. Kuhlmann, Heinrich 1891 Die Konzessivsätze im Niebelungenliede und in der Gudrun (Unpublished Doctoral Dissertation: Universität Kiel). Lerch, Eugen 1929 Historische Französische Syntax II (Leipzig: Reisland). Levinson, Stephen C. 1983 Pragmatics (Cambridge: Cambridge University Press). Mensing, Otto 1983 Untersuchungen über die Syntax der Concessivsätze im Alt- und Mittelhochdeutschen (Unpublished Doctoral Dissertation: Universität Kiel). Quirk, Randolph 1954 The concessive relation in Ο Ε poetry (New Haven: Yale University Press).

282

Ekkehard König

Quirk, Randolph et al. 1972 A grammar of contemporary English (London: Longman). Traugott, Elizabeth C. 1983 "Conditional markers", to appear in Papers from the Symposium on Iconicity, edited by J. Haiman. Visser, F. Th. 1972 An historical syntax of the English Language II (Leiden: Brill).

ADRIENNE LEHRER

The influence of semantic fields on semantic change* Introduction The theoretical activities and insights of the last two decades in linguistics have not spilled over into etymology and the study of semantic change, even though there has been much important work in both historical linguistics and in semantics. One reason for this neglect of semantic change is that the changes themselves seem to be sporadic. Every word has its own history. About the best we have come to hope for is a taxonomy, or classification schema, as found in Ullmann (1957), Stern (1931, 1968), or Williams (1975). These categories of semantic change summarize the tendencies or possibilities, which may in fact have opposite effects, such as narrowing vs. broadening. Our current state of knowledge does not allow us to make interesting, falsifiable statements concerning the lexicon as a whole. In this paper I shall show that some insights into the principles of semantic change can be found by looking, not at the whole lexicon, but at words which belong to a single semantic field. A semantic field is a set of lexemes which cover a certain conceptual domain and which bear certain specifiable relations to one another. An example of a simple semantic field would be the conceptual domain of cooking, which in English is divided up into the lexemes boil, bake, fry, roast, etc. A basic premise of semantic field theory is that to understand lexical meaning it is necessary to look at sets of semantically related words, not simply at each word in isolation. By 'semantically related' I refer to relationships between lexical items such as synonymy, as in big and large, antonymy, as in big and small, hyponymy, as in rose and flower, converseness as in buy and sell, incompatibility, as in cat, dog, cow, horse, pig, etc. A list of such lexical relationships and their entailments can be found in Lyons (1977) or Lehrer (1974). I hope to show that our understanding of semantic change can be enriched by looking at the histories of semantically related words.

284

Adrienne

Lehrer

Semantic fields and semantic change The connection between semantic change and semantic fields was stressed by Trier (1931), one of the earliest linguists to develop semantic field theory. Trier showed that semantic change affects the structure of semantic fields. His famous example of the meaning of words dealing with knowledge in medieval German shows that both the inventory and the semantic structure changed between 1200 and 1300. The words that remained in the language had different meanings. 1200

1300

Wisheit Kunst

Wisheit

Kunst

Wizzen

List

Figure 1

Figure 2

Around 1200 Wisheit was a general term covering the whole field of knowledge. A century later the meaning narrowed to 'religious or mystical knowledge'. Kunst at the earlier time referred to courtly knowledge and later to mundane skills and knowledge. List, which was used for technical skills, moved out of the field, and Wizzen, relating to art, moved in. These two figures show that the "semantic space" was divided into two parts in 1200, with a superordinate word for the whole, whereas in 1300 the space was divided into three parts. 1 Trier concludes that a change in the meaning of one word in a field requires changes in the meaning of other items because of two assumptions he makes: (1) There are no overlaps of meaning in a field. (2) There are no gaps. Both of these assumptions are false, however. The existence of partial synonymy — a very common phenomenon — shows that the first assumption is false, and the discussion of lexical gaps in Lehrer (1974, chapter 5) provides many counterexamples to Trier's second assumption. This study will take a different direction. It will show how semantically related words show parallel semantic changes. One earlier

The influence of semantic fields on semantic change

285

study which demonstrates this phenomenon is by Stern (1931,1968), who traces the semantic changes of a set of words in English from the meaning of 'rapidly' at Stage I to both 'rapidly' and 'immediately' at Stage II to 'immediately' at Stage III. (See table 1). The chronological discrepancies in the table are explained by Stern as due to the

Hrcedlice Hrat»e (Ra{)e) Ardlice Lungre Ofstlice Sneome SwiJ)e Swiftly Caflice Swift Georne Hi3endliche

Quickly Smartly Snelle Quick Belife Nimbly Rapely Skete Tite Wight Wightly

Sense I 'Rapidly'

Sense II 'Rapidly, Immediately'

OE OE OE OE OE OE OE OE OE OE OE 1200 1200 1290 1300 1300 1200 1430 1225 1300 1300 1300 1350

OE OE OE OE OE OE OE OE OE 1360 1290 1200 1200 1300 1275 1290 1200 1470 1300 1300 1350 1360 1350

Sense III 'Immediately' OE OE OE OE OE OE 1175 1200 1370 1300? 1400? 1300 1200 1200 1300 1300 1300 1200 1400 1325 1200 1300 14th cent. 1300

scantiness of the OE and ME texts... It is only about 1300 that there is a satisfactory supply of texts to illustrate the state of the language... In other cases, a meaning may have arisen in colloquial language, which is scarcely represented in our texts. Also, in some cases, the words are comparatively rare. In these circumstances, it is evident that a meaning may be much older than the earliest preserved record ... Stern (1968: 187)

286

Adrienne

Lehrer

The semantic change of any of the words is quite ordinary. It can be viewed in either of two ways: (1) Stage I

-> broadening Stage II

- • narrowing Stage III

or (2) Single meaning Sense χ

addition Ambiguous loss of of a sense -* a sense Senses x, y

Single meaning Sense y

The first model views Stage II 'rapidly/immediately' as a case of vagueness; the second views it as ambiguity (which apparently is how Stern treats the matter). What is striking about the data is that the same change occurred thoughout this relatively large set of synonymous or partially synonymous words. Words meaning 'rapidly' entering English after 1400 did not undergo this semantic change (Stern 1968: 189). Given the uncertainty of dating the earliest instances of these words and their meanings, Stern's arguments and conclusions about this particular case are not particularly convincing. However, he has proposed an important hypothesis — one that is worth further investigation — which is that semantically related words may undergo parallel semantic changes. In fact, I wish to propose an even stronger hypothesis: semantically related words are more likely to undergo parallel semantic changes than semantically unrelated ones precisely because of their semantic relationships. Semantic relationships tend to remain constant, so that if one word changes meaning, it will drag along other words in the domain.

Animal Metaphors The first set of semantic fields I will examine is a set of conventionalized animal metaphors, based on an analysis by Battan (1978). The words basically denote animals {wolf, mouse, turkey), but they can be applied to people, as in John is a wolf {mouse, turkey). These extended meanings are conventionalized. (Definitions are taken from the American Heritage Dictionary and the Oxford English Dictionary.) Therefore, the mature speaker-hearer has learned the extended meaning of the words and need not figure out the

The influence of semantic fields on semantic change

287

interpretation as he or she would have to for a truly novel metaphor. These animal words as applied to humans have originated as metaphors, of course, and because the literal meaning — that is, the meaning which denotes the animal — remains in the language, the relationship between the animal and the transferred meaning is salient for the speaker. The main point here is that since the transferred meaning is conventional, the hearer does not need to compute the meaning. Animal metaphors are also interesting to look at because they are offered as paradigm cases of metaphors that involve some similarity between the objects which are implicitly compared. (Black (1954) points out that what is important here is our stereotypes, which could be false.) The semantic field of animal words provides a good case for testing the effects of semantic change of one word on other items in the field, since the meaning of each word — at least in its literal sense — is influenced and limited by its denotation. We should not expect the meaning of tiger to be affected by the existence or non-existence of co-hyponyms: lion, cheetah, panther, leopard. If cheetahs were to become extinct and if the word cheetah were to disappear from the active vocabularies of speakers, the meaning of tiger should not be affected. And if a new species of felines were discovered and named, then that fact also should not affect the meaning of tiger. What would change the semantic structure would be a reclassification by taxonomic zoologists, e. g., a decision that lions and tigers are really the same thing. However, given the current state of such taxonomies and the words for the classes of animals, one would expect there to be little influence of the meaning of one word on another. Fourteen words have been selected, which can be grouped into classes corresponding to zoological classes and exhibiting a similarity of semantic extension. Current meanings are taken from the American Heritage Dictionary and earlier meanings from the Oxford English Dictionary. The diagrams are to be interpreted as follows: (straight line) shows the period of time for which the word has been recorded in the language with its zoological meaning. (string of dots) shows the period of time that a word has existed with a metaphorical meaning that is either obsolete, archaic, or less salient than the current (and most salient) metaphorical meaning.

288

Adrienne

Lehrer

xxxxxxxxx (string of x's) shows the period of time that the word has been in the language with its current and most salient metaphoric meaning. The numbers at the top provide the time scale. Let us look first at figure 3 which shows the semantic transfer for three primate names. These words are arranged chronologically in terms of their length of time in the language. Figure 3 shows that the metaphorical meaning for ape 'a fool' developed early and the meaning for baboon 'a general term of abuse' developed somewhat later. The meaning 'a brutish person' appeared for all of these words more recently — first for gorilla,2 then for ape and baboon. The meaning 'fool' for ape has become archaic, according to the OED.

Animal category: Primate Semantic transfer: Brutishness

1000

1200

1400

1600

1800

current

ape (700)

'a fool'

(1330)

xxxxxxxxxx

baboon

(i4oö)

;

"

general

(1500)

term of abuse'

xxxxxx gorilla (1799) 2 (1847)

xxxxxxx (1884) Figure 3

This is a pattern which will recur in other diagrams as well: the animal word to enter the domain most recently (in this case gorilla) was the first to acquire the extended meaning of 'brute, brutish person'. At a later time, two other primate names acquired this meaning, even those that already had different metaphorical meanings. In the case of ape the earlier metaphorical meaning of 'fool' became obsolete. There appears to be a triggering mechanism.

The influence of semantic fields on semantic change

289

When a member of an animal word set acquires a metaphorical meaning, it faciliates a comparable change in other members of the set.

Animal category: Bird Semantic transfer: Foolishness 1000

1200

1400

1600

1800

current

goose (1000)

xxxxxxxxxxxxxxxxxxxxxxxx (1547)

cuckoo (1240)

xxxxxxxxxxxxxxxxxxxxxx (1580)

pigeon (1440) 'coward' (1586) xxxxxxxxxxxxxxxxxxxxx (1593) coot (1382) (1430)

. 'general term of abuse' xxxxxxxxxxxx

turkey (1555) xxxxxx Figure 4

Figure 4 contains five words denoting birds, all of which have a current metaphorical meaning of 'foolish'. Notice that three terms, goose, cuckoo, and pigeon acquired the same metaphorical meaning around the same time, although the three words had been in the language as animal words for 300 to 500 years. This could have been a coincidence, of course, but a better explanation is that the semantic field properties facilitated these changes. The OED gives no precise definition for coot as a general term of abuse; rather there are proverbial phrases such as as bald (mad, stupid) as a coot. This older, general meaning has become obsolete and replaced by a narrower meaning: 'foolish person, simpleton', in accord with

290

Adrienne

Lehrer

other bird names. An earlier meaning for pigeon, 'coward', also became obsolete.

Animal category: Snake Semantic transfer: Treachery 1000

1200

1400

1600

1800

current

snake (1000)

XXXXXXXXXXXXXXXXXXXXXX (1591)

viper (1526) xxxxxxxxxxxxxxxxxxxxxxxx (1526) Figure 5

Figure 5 consists of only two words, both with the same metaphorical meaning. Viper was first noted in English in both its zoological and metaphorical sense in Tindale's 1526 translation of the Bible from Old French. Snake, which had been in the language centuries before viper was introduced, soon acquired the same metaphorical meaning that viper had. This change can be explained by assuming that viper served as a trigger. Since viper and snake belong to the same animal set, the semantic transfer of snake was facilitated and triggered by that of viper.

Animal category: Scavenger Bird Semantic transfer: Greed 1000

1200

1400

1600

1800

current

buzzard (130°)

'worthless person' xxxxxxxxx

vulture (1374) xxxxxxxxxxxxxxxxxxxxx (1602) Figure 6

The influence of semantic fields on semantic change

291

Figure 6 shows the semantic development for two bird names. Vulture was recorded slightly later than buzzard, and its metaphorical meaning triggered a change in buzzard. Buzzard has lost its earlier meaning of 'worthless person'. Figure 7 seems to show a widening of metaphoric meaning. Mule and donkey had acquired the meaning of 'silly' or 'stupid person' (as had ass and jackass). However, later donkey acquired the meaning of 'obstinate' as well, which triggered or facilitated a similar change in mule to include 'obstinate'. The current metaphoric meaning of mule probably focuses on obstinacy. 3

Animal category: Mule Semantic transfer: Obstinacy 1000

1200

1400

1600

1800

current

mule (1000)

'silly person' (1470)

xxxxxxxxx

donkey (1785) (1840) 'silly person' xxxxxx Figure 7

These examples show the following pattern: there is a domain of animal names in which there are subclasses corresponding to zoological subclasses. One of the words acquires a metaphorical meaning, which in turn triggers or facilitates corresponding transfers among other lexemes in the subclass. In the examples above, the newer word to be recorded was usually the one whose metaphorical meaning determined that of the words for the subclass, often displacing other metaphorical meanings which older words had. Whether this last observation is a coincidence (that is, an accidental feature of the sample) or whether it reveals something important about processes of semantic change remains to be determined.

292

Adrienne Lehrer

Gambling terms A similar phenomenon can be seen in a set of terms from gambling vocabulary (Teske 1983). Most of the terms belong to the slang register or are jargon to the relevant semantic field (see figure 8). All four words have undergone parallel historical development. In each case a specific gambling sense is transferred to an auctioneering meaning. Each transfer involves a common semantic change (widening), but the fact that the whole set is affected in the same way is striking.

Gambling Terms 1800

1900

current

setter

'a sharper's decoy' (1591) xxxxxxxxxxxxxxxxxxxxx (1840)

booster

'runner-up of prices at auctions' 'gambler's decoy'

(1906) xxxxxxxxxxxxx (1915) capper

'auctioneer's confederate' 'sharper's decoy'

(1853) xxxxxxxxxxxxx (1875)

'auctioneer's confederate' 'gambling confederate'

shill (1898) xxxxxxxxxxxxx

'runner-up of prices at auctions'

Figure 8

Metaphorical transfer and semantic change Many scholars have pointed out the relationship between metaphorical transfer and semantic change. A speaker or writer (or several speakers and writers) uses a word metaphorically, and if this usage spreads to other speakers in the community, the word has acquired a new sense. (The old meaning might remain or it might disappear.)

The influence of semantic fields on semantic change

293

One particularly interesting case has been among words denoting sensation and perception. Williams (1976) has proposed a law of semantic change among such adjectives where transfers are unidirectional:

touch

-»-taste

-»-smell

dimension. »-

color | sound

Lehrer (1978) has proposed an alternative analysis for these data. I will not review the specific disagreements, but I will draw on part of Williams' research because it shows quite convincingly the ways in which lexical sets may undergo parallel semantic development. If there is a set of words that have semantic relationships in a semantic field, and if one or more items pattern in another semantic field, then the other items in the first field are available for extension to the second field. The semantic relationships will remain the same: synonyms will remain synonyms, antonyms will remain antonyms, etc. Lehrer(1978: 96f

I will illustrate the process by considering dimension words which were transferred to the taste domain, primarily for beverages, and more particularly for wine (Lehrer 1983). Figure 9, based on Williams (1976:475-6), gives the dimension words with their earliest recorded dates, shown by a straight line, and then the appearance of these words in the taste domain, shown by x's. Many of the words with the taste meaning can readily be found in books and articles about wine. Therefore, I have added this meaning with the label 'wine' under the heading for current. Since these meanings have not yet been recorded in the major dictionaries (but are found in wine glossaries), I cannot date their first uses with the taste meanings. My guess is that their use is fairly recent (that is, within the last century). s I have tried to show that semantic field theory — the view that our lexicon is organized into semantic fields — can contribute to our understanding of semantic change. The phenomenon I discuss does not replace other explanations of why words change meaning in the first place or why they change from A to Β instead of from A to C. But I hope to have shown that the histories of words are not completely independent. Sometimes semantically related words share a historical development.

294

Adrienne Lehrer

Semantic Field: Dimension 6 Transfer: Taste 1000

1200

1400

1600

1800

current

small (725)

xxxxxxxxxxxxxxxxxx

(wine)

(1676) high (825)

xxxxxxxxxxxxxxxxxx (1430)

deep (854)

XXX

(wine)

XXX

(wine)

thick (888) fat (893)

xxxxxxxxxxxxxxxxxxx

(wine)

(1609) even (893)

XXX

(wine)

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

(wine)

thin (900)

(1377) empty (971)

XXX

(wine)

(1000)

XXX

(wine)

(1000)

XXX

(wine)

full little hollow (1250)

(wine)

big (1386)

χχχ

(wine)

flat (1400)

XXXXXXXXXXXXXXXXXXX

(wine)

(1609) acute

XXX

(wine)

Figure 9

Notes * This paper is an expanded version of Lehrer and Battan (1983). I with to thank Richard Janda and Bruce Mannheim for their comments on an earlier draft.

The influence of semantic fields on semantic change

295

1. The accuracy of Trier's description has been challenged by Scheidweiler (ZDA 79. 1942: 249-268) and Spence (1961), but that does not concern the theoretical points in this paper. 2. Gorilla was first applied to the name for a wild hairy person in a Greek tale of the 5th or 6th century B.C. It was first recorded in English in 1799 as a term for a group of hairy savages and only later applied to the animal. However, I hypothesize that the later transferred meaning sprang from the animal name, not the earliest Greek or English use. 3. The OED reports that in Greek fables and proverbs, the ass was a figure of clumsiness, ignorance, and stupidity. In the Bible and the earliest English texts, ass and mule did not have this connotation, however, donkey began to replace ass in the 18th century. 4. Kittay and Lehrer (1981) provide further examples of the process where whole fields of words transfer metaphorically. 5. Some taste meanings are listed by Williams as obsolete. However, if they are used as wine descriptors, I have extended the x's to the column for 'current', although the current use may indeed have been a new transfer from dimension to taste. 6. Williams also lists three words which were first used in the touch domain before transferring to taste and dimension.

sharp crisp smart

touch

taste

dimension

825 900 1023

1000 wine 1648

1537 1398 1668

References Battan, P. L. 1978 A study of standardized animal metaphors: a search for structure in the lexicon (Tucson: ms.). Black, Max 1954 "Metaphor", Proceedings from the Aristotelian Society 55: 273-94 [reprinted in M. Black 1962. Models and metaphors, Ithaca: Cornell University]. Kittay, Eva — Lehrer, Adrienne 1981 "Semantic fields and the structure of metaphor", Studies in Language 12: 35-46. Lehrer, Adrienne 1974 Semantic fields and lexical structure (Amsterdam: North Holland). 1978 "Structure of the lexicon and transfer of meaning", Lingua 45: 95-123. 1983 Wine and conversation (Bloomington: University of Indiana Press). Lehrer, Adrienne — Battan, P. L. 1983 "Semantic fields and semantic change", Coyote Papers 4, edited by E. Martin-Callejo (Tucson). Lyons, John 1977 Semantics (Cambridge: Cambridge Uniersity Press). Spence, Nicol C. W. 1961 "Linguistic fields, conceptual systems, and the Weltbild', Transactions of the Philological Society, 88-106.

296

Adrienne Lehrer

Stern, G. 1931 Teske, M. 1982 Trier, Jost 1931

Meaning and change of meaning (Bloomington: University of Indiana Press) [reprinted 1968]. Gambling terms and semantic change (Tucson: ms.).

Der Deutsche Wortschatz im Sinnbezirk des Verstandes (Heidelberg: Winter). Ullmann, Stephen 1957 The principles of semantics [2nd ed.] (Glasgow Univ. Publ. 84) (Glasgow: Jackson, Son & Co.). Williams, Joseph M. 1975 Origins of the English language (New York: Free Press). 1976 "Synaesthetic adjectives: a possible law of semantic change", Language 52: 461-78.

BARBARA LEWANDOWSKA-TOMASZCZYK

On semantic change in a dynamic model of language

The starting point for the present paper is the assumption that diachronic language change is intimately related to synchronic variability. One of the assumptions behind the dynamic model of language proposed here (cf. also Lewandowska-Tomaszczyk 1983 c) is that the processes which take place in human verbal interaction cause systemic language change when projected onto a large scale. It is hoped that the model adequately accounts for the types of construction such as the ones given below from Polish which exemplify either a completed semantic shift or change in progress. Language change in time can be associated with a whole spectrum of language users' behaviour pattern, starting from deliberate goaloriented activity, through partly purposeful types of behaviour, up to actions which lack any explicitly motivated character (for typology cf. Andersen 1973 and 1980). Intentional neologisms occur frequently in the language of literature where they may be either referentially empty and serve first of all certain affective function (alansacja (Tuwim) 'alansation', murbie w sosie agamelinowym (Witkiewicz) 'murbs in agammelin sauce'), or, as in some science fiction stories, they introduce items from a reality proposed by the author (sepulki, 'sepules', defryzon 'defrostee' (Lem)). As is known, such constructions do not form a dichotomous set with intra-systemic norm-keeping novelties in language, but exhibit degrees of peripherality with respect to the norm. Thus they cannot be evaluated from the point of view of any absolute criteria. They can be judged instead on a relational scale of interpretability. Some of them seem to be more widely accepted, for instance such literary neologisms as gdybac 'to i f , lopatologia 'shovellology' — "scientific" method of teaching slow learners; term based on a current colloquial Polish expression 'put knowledge into somebody's head with a shovel', mniemanologia 'considerology' —

298

Barbara

Lewandowska-Tomaszczyk

a term ridiculing pseudo-science (pseudoscientists being noted for the overuse of mniemam, affected for 'think'), are currently making some career in a semi-serious style, while similar ktorzyc 'to which' or wstecznologia 'regressology', remain only nonce-formations. Deliberate creations formed according to language productive rules are, in some conversational styles, devised to break the monotony of the language convention, e. g., padlodajnia 'carrion-house' in analogy to jadlodajnia 'eating-house' (some of the above examples are from Buttler 1974). Most frequently, however, they are formed in order to name new concepts expressing novelties in the outside world which require labels (zaopatrzeniowiec 'in a factory person responsible for arranging supplies of raw materials, spare parts, etc.; supplier', malomiasteczkowosc 'provincialism; smalltowniness'. Similar motives for concept formation and labelling can be observed in connection with new technological discoveries, inventions, etc., that come to be known by language users via non-native forms. Such is the case with the word computer, which exhibits varying degrees of phonetic assimilation with the Polish system. Its referential synonym maszyna liczaca 'computing machine' seems to have a different interactional status. Other loan-words enter the native system and become established in the nativized form parallel to their closest native equivalents as a consequence of covering partly different conceptual domains, e. g. farma — most frequently 'English, American, Canadian farm' vs gospodarstwo rolne, zagroda — 'farmstead', 'croft'. Both widening and narrowing of meaning can be observed in such cases as akcelerator 'accelerator' — more general than pedal gazu 'gas pedal', or ferma 'farma', from French, tending to be narrowed to ferma zwierz^ca 'stock farm'. Lexical creativity is also a domain of technocrats in Poland, who, besides introducing new names for new products, tend to replace standard words denoting most frequently used objects (lampa 'lamp', sliniaczek 'bib', scierka do kurzu 'dusting cloth') with new technical labels (zwis 'hang' /noun/, podgardle dzieciqce 'children's throat belt', skurzawka 'duster', respectively). Some of the above cause unexpected funny effects {podgardle is used primarily in the sense of 'dewlap'), some others are felt pseudoscientific, while the whole idea — difficult to accept for an average Polish speaker. Another process connected with the interpersonal aspect of meaning can be observed in cases of hypercorrection, that can lead to observable phonological, syntactic, and semantic variation and

On semantic change in a dynamic model of language

299

change, e. g. a well-known case of sabla (now only dialectal) 'sabre', changed to szabla around XVI c. or contemporary non-standard obczasy from obcasy 'shoe heels', both being the case of false demasurisme; also cases of semantic extension, e.g. profesja 'profession' may be synonymous to zawod 'occupation' or zajgcie 'job'. The latter may also refer to 'any activity', hence the use attested by Pisarek (1978: 92): klopot w nielatwej profesji robienia zakupow 'trouble in the uneasy profession of doing shopping'. This type of shift seems to result from false inferences and, similarly as other instances of hypercorrection, it is aimed at increasing the evaluation of overt markers of the speaker's social status. It takes place first of all in the cases of limited exposure to varying contexts in which a given linguistic unit can be used, which may result in forming wrong hypotheses about its meaning (cf. also alternatywa 'alternative', which, in spite of its original meaning, tends to be used as: wybor ζ dwoch lub wiecej mozliwosci 'selection out of two or more possibilities', as in: mam dwie alternatywy, kilka alternatyw Ί have two alternatives', 'a few alternatives' with the intended sense: mam alternatywe Ί have an alternative' and kilka mozliwosci 'a few possibilities'). The adaptation of one's speech to that of another person may be also based, as is known, on the speaker's idea about the interlocutor's norm. Such an accomodation of the speaker's forms and meaning based on the premises of what he believes, correctly or incorrectly, to be the listener's norm, may also lead to semantic shifts (cf. Householder 1983: 8 and further in this paper). Another case refers to shift of properties expressing the foregrounding of different denotational and/or connotational characteristics of a lexical item. Certain salient perceptual, functional, emotional, etc., properties of one object get to be associated with observable characteristics of another object by metonymy, metaphor, etc., thus leading to semantic shifts (tchorz, dzierlatka, first used as names of animals 'polecat', 'crested lark', then extended to human beings — 'coward' and 'immature lively girl' respectively; however, while the former has an uncontroversial status of a generic term, the latter is one of the many descriptive emotive options, also considered dated by some informants; cf. also gluchy 'deaf and glupi 'stupid' and possibly glaupy 'young' in some Slavic languages. Shifts of certain characteristics of real world objects can condition a new sense evolving from the older one with slight or with no modification of the form, e. g.,: wladca 'ruler' from wladykajwlodyka

300

Barbara

Lewandowska-Tomaszczyk

'lord', wlodarz 'steward, supervisor'; suknia — originally 'a woman's garment made only of raw woolen cloth' called sukno, currently — 'a woman's dress made of any type of cloth'; Sejm 'Polish Parliament'. Conditions for decreasing frequency and eventual elimination of an item may evolve at the moment when a certain object or phenomenon is getting out of use or starts being considered old-fashioned narzeczona 'fiancee', sympatia 'girlfriend' — being substituted for the more general and much less comitting dziewczyna 'girl', preceded by a possessive moja 'my', twoja 'your', jego 'his'. The form itself may be frequently felt to be worn out, which refers first of all to occasional and slang formations, most mobile elements in lexis e. g., bombowy koszul 'smashing shirt'; where koszul /Masculine/ was formed from koszula /Feminine/, was current among young people in the early sixties. Euphemisms referring to various taboo-phenomena show, as is widely known, a tendency to have a relatively short lifespan as well (cf. klozet, wygodka, wychodek, ustep, ubikacja, toaleta, etc., for 'lavatory'), with newer forms replacing the older ones. What has been discussed so far is the tendency to semantic shifts which stem from 1) the change of the outside reality, 2) specific perception of reality by language users, and 3) hearers' abduction (Andersen 1973) concerning language patterns and their semantic load, which does not conform to those intended and actually used by speakers. Such circumstances motivate 1) deliberate, conscious activity of language users geared towards devising new or modifying established patterns, 2) total or partial borrowing from other codes, and 3) gradual filling of new or modified content into conventional word-forms and syntax. The shifts, however, can be signalled and spread only via verbal interactions among language users. Meaning of words and utterances in such situations has a starting point in the conventional given, but in the course of ongoing interaction meaning is negotiated, i. e., jointly and collaboratively constructed by sequential contributions of participants (Lewandowska-Tomaszczyk and Machova, in press). This is the setting of semantic variability and change. Out of the available semantic theories, some are better —, some others — not so well-suited to account for language and meaning conceived of as above. A componential analysis, utilizing a conjunction of semantic markers as employed by Katz (1972), for

On semantic change in a dynamic model of language

301

instance cannot account for the gradual shift in meaning that is observed in suknia, alternatywa, or in some novel lexical constructs of native or foreign origin. First of all, it appeared that denotational attributes of some objects are neither necessary nor sufficient conditions for their categorial membership (Rosch 1973, Labov 1973). Objects of natural kind are considered by language users to occupy a space either more towards the centre of a categorial domain, e. g., dog in ANIMALS, chair in FURNITURE, in the category of BIRDS robin for English (Rosch 1973), and wrobel 'sparrow' — for Polish data (Miodunka 1980), thus being considered the most typical class representatives, and such category members which approach its peripheries (e. g. penguin and ostrich for BIRDS), the properties of which are not identical with those in the centre. Placement of abstract objects and their concepts may be even more diverse. Secondly, from a diachronic point of view semantic features such as those postulated by Katz would have to gradually undergo such drastic changes that, at least in some cases, the result could be their negation at some point. Katz's semantic apparatus does not allow for such a possibility; to quote an example of the semantic shift of ceorl from Dahlgren (1978: 61): ceorl 'man' -> ceorl 'free landholder' /690 AD./ ceorl 'semiserf /1050 AD./ cheorl 'serf /1100 AD./ cherl 'rude, lowbred fellow' /1386 AD./ Dahlgren stresses in this connection the unavailability of a proper analytic sentence of the type All ceorls are free, which would justify the introduction of the semantic marker FREE, underlying the original meaning of ceorl, as "the moment the first ceorl gave up his land to a lord toward feudalism, this sentence became false, much less analytic" (Dahlgren 1978: 61). The decreasing scale of analyticity of the original proposition then is what underlies the changing meaning of ceorl. A similar observation can be made with respect to such vague terms as 'free' or 'freedom' themselves, with semantic markers whose values may change with individual ideology. Thus change has to be correlated with the ever-changing world outside as well as with our 'private' worlds in which both properties of objects as well as stereotypical patterns of reality are reconsidered and new features added either by readjustment to the context or to the interlocutor, as its part. For that reason I think it necessary to

302

Barbara

Lewandowska-Tomaszczyk

incorporate into the model of language change — global patterns of knowledge (cf. Schank — Abelson 1977), which are prototypical schemata for the categorization of socially accepted reality and knowledge of typical events, processes, etc., which serve as a conventional point of reference to variability and mutations. That such knowledge is relevant both to language production and processing has been convincingly demonstrated in numerous publications (cf., e. g., Lakoff 1982). In the language user's mental store there can be present conflicting knowledge, which is used according to his/ her psychological needs (cf. Lewandowska-Tomaszczyk 1983 b and 1983 c) and is responsible for individual semantic variations, e. g.: (1)

Homo sapiens is a representative of mammals in animals

(2)

He is behaving like an animal

(3)

"Take cars ... / don't even like old cars ... I'd rather have a goddam horse. A horse is at least human ..," 1

All three sentences can be present in the repertoire of the same individual, although each of them assumes a different, incompatible with the other two, categorization of reality. Example (1) represents a scientific classification where MAN is subordinate to ANIMAL. In sentence (2) we discover a typical categorization with its roots in the Biblical "have dominion over the fish of the sea, and over the fowl of the air, and over every living thing that moveth upon the earth", where man is opposed to and superior to animal. Animality then equals degradation for man. Example (3) entails a metaphorically extended categorization with the element H U M A N understood as NATURAL in opposition to TECHNICAL, ARTIFICIAL. What I would like to propose is a framework for a dynamic model of language that would account for the interplay of dynamics and stasis that characterizes it. Although resembling models conceived of as cybernetic feedback control systems with homeostatic self-regulation to some extent, language never, in fact, seems to achieve the state of ideal equilibrium. The model here is an alternative to such as the one proposed for instance in Nöth (1983) in reference to Laszlo (1969), symetrically modelling linguistic morphogenesis in a helical shape with the following four phases: (1) equilibrium, (2) disturbance, (3) disequilibrium, (4) compensation phase, replaced by a new equilibrium (!'). Besides the states of ideal

On semantic change in a dynamic model of language

303

equilibria, implied here is the presence of an absolute parallelism, or one-to-one correspondence between the new and the old system, although on different levels. As was implied in my discussion above, there is no possibility to guarantee such an idealistic equivalence in comprehending a concept at a different place, time, or setting. Hence the concepts underlying some semantic criteria may be sensitive to interpretation as well. For that reason the second variable has been introduced into the model, signalling the gradual displacement processes and possible reinterpretation of semantic criteria. The model has a form of an Archimedean (displaced) spiral with an additional dimension as in Fig. 1.

Η

X Fig. 1. Archimedean spiral of language change

The model represents language change conditioned by two variables: the system itself and the projection of reality onto the language user's mind. The system is recoverable due to certain regularities of change such as analogy, compensation, etc. The perception of reality is treated as a partly relative process (cf. also Beaugrande 1983: 87) permitting individual self-reflexion and change. What it implies,

304

Barbara

Lewandowska-Tomaszczyk

however, is that a number of (not entirely overlapping) Archimedean spirals should be proposed here to reflect individual language variation and change (cf. Bickerton's (1972) "polylectal" grammars as well as Romaine's (1982) critical discussion of the Labovian "language community" concept). That the spirals would partly overlap can be confirmed by consistently interpretable linguistic and nonlinguistic outcomes obtained from individuals experiencing comparable stretches of reality (Beaugrande 1983). As stated above, it is assumed that linguistic units get determined in an interactional context. The process of meaning negotiation seems to be a mini-replica of the Archimedean spiral, subject to two sorts of conflicting principles referring to interactants' individual actions on the one hand and their interaction on the other. The first set of principles refers to mutually exclusive interests of the speaker and the hearer. Productive and perceptual signal simplicity (Langacker 1977: 105 calls them "signal simplicity" and "perceptual optimality" respectively) are both a variation of the "least effort principle". The former minimizes the speaker's efforts. The latter, in tending towards one-to-one optimality between the form and the content, makes the message easier to process by the listener. Productive signal simplicity can be observed not only on the phonetic (articulatory) level. It is expressed in lexical selection, condensation and complexity of syntactic patterns, or excessively laconic conversational contributions of the speaker. It is in the listener's interest, on the other hand, for the utterance to be easy to perceive and for the messages to be unambiguous. The other extreme is expressed by the second set of principles, namely, Gricean "principle of co-operation" (1975). They may level out possible results of the former ones. Furthermore, both signal and signal-content complexity cannot be determined as absolute values — they should be considered instead to be context-bound and be assigned gradual complexity scales from maximum signal simplicity (minimum complexity) to minimum simplicity (maximum complexity). Both extremes of the scale are marked by blocking the processing, as a result of which no communication can take place. All cases in between correspond to different context-bound options, where the speaker adapts his/her language behaviour to all parameters of the communicative act: hearer, setting, topic, social roles, etc., and tunes his/her language accordingly. In the case of procedural problems (Lewandowska-Tomaszczyk and Machova, in press) the interactant

On semantic change in a dynamic model of language

305

who signals them calls for repair, and he/she is either satisfied by the speaker's rephrasing the utterance, or else the negotiation is continued. Whether those vague, indeterminate, or novel constructions used by the speaker, which are further elaborated, defined or explained in an indirect way, enter the active vocabulary of the interlocutor, to what extent and in what shape, should be experimentally tested, especially in reference to semantic data. It can be only conjectured at present, that the tendencies Langacker (1977) mentions in his study are regularly weakened in those interactions which are conventionally co-operative, while in some other cases (e. g., in the language of literature) they may be totally reversed. There the speaker (author) may intentionally make the signal more complex to possibly evoke in the hearer (reader) alternative world-models, not necessarily corresponding to the actual model of reality. In such a case the principle of perceptual optimality is ignored not only by the author, a situation that could be considered maximally unco-operative, if it were not motivated, but it is also remitted by the addressee. This is merely a consequence of his/her expectations in reference ot the type of text he/she is going to read. A detailed analysis of the complexities of verbal encounters on contemporary material can bring about some new insights to diachronic linguistics. Both increased meta-language awareness of the speaker or lack of sufficient "self-feedback" (Greenberg et al. 1965: 153-5) on some other occasions, as well as the "social feedback" (Nöth 1983: 120) of the audience, who either accept or reject linguistic shifts in their intended or misrepresented versions, are vehicles of language change. Naturally, the meaning of a linguistic sign cannot be unambiguously "established" in a single negotiation. This can be achieved rather in a recurrent process where the semantics of linguistic units is open both on individual and social levels, and where a consensus is likely to be reached on a conventional level of practice. Language, as understood here, expresses partly conventional categorization of the socially accepted reality and an open potential for capturing alternative realities, which is asymmetrically constrained on different linguistic levels (in the sense of Jakobson 1971: 242), by the properties of linguistic convention and psychophysical abilities of its users. On each occasion meanings of linguistic units referring to these realities are individually perceived and negotiated among the participants of the communicative act.

306

Barbara

Lewandowska-Tomaszczyk

In order to account for those properties the model of language has to be embedded into a cognitive model of communicating minds (for similar assumptions cf. Fawcett 1980). Fig. 2 presents a proposed cognitive model of communicative parameters in linguistic action and interaction. The main components of this model are affective states and psychological and physical needs of the speaker (context 3 ), properties of the interacting addressee, such as his social status, needs, expectations, etc., that are perceived prior to and during the interaction (context 2 ), the linguistic and paralinguistic response of the addressee to speaker' activities (context 2 ), and features of actual reality (contextj) other than the interactant, such as actual properties or objects, processes, etc. Knowledge of the universe, that is projected on the actual or alternative reality, is a data-base containing the encyclopaedic information referring to the world, enriched by the personal, sociocultural, as well as the interactional experience of the language user. All the individual and contextual information updates the prototypical structures on the semantic levels in the language component, which occupies some space in the compartment of action and interaction. Semantic structures are transduced (translated) onto syntactic and morphonological structures, which feed relevant "analysability reflexes" (cf. Langacker 1982) into corresponding semantic structures. This implies a bi-directional rule movement as well as access to language code at any time of signal/content production and processing. Potential realizations of the language code are constrained by speech acts and discourse organization schemata (conversational maxims, turn taking, etc.), which undergo control under suitable conditions and are surveyed in the problem solver where all the relevant information is retrieved so that the final linguistic outcome can be decided on. Recurrent readjustments of the linguistic performance of the alternating speakers to the linguistic outputs of their interlocutors, are oriented towards reaching the familiar state of homeostatic balance between them, i. e., total comprehension. Although the state of optimal understanding may oscillate around a maximum, its highest value cannot be achieved by definition: since there is always a degree of individualization (cf. experiential component of meaning below, actualizations of different components of the context, different selections of syntactic, morphological, and phonetic options)

On semantic change in a dynamic model of language

307

[PROFE] occupation professed (Latin)

+

[SJA]

nominal suffix to foreign stems (steady action)

Speaker 2 : plan: impress Hearer 2 method: substitute foreign for native

Stage 1 CC: F (Activities (steady/occasional)) Interactional/Affective: profesja (dated) (jocular) (foreign) zajgcie\

φ

zajgciei

profesja

—

zaj§cie\

profesja

φ

zajgciei

On semantic change in a dynamic model of language

317

Stage 2 Hearerj CC: F (Activities) Interactional/Affective: profesja (foreign) profesja = zaj§cieu 2,... η Speaker 2 CC: Plan (Impress Hearer 2 ) Substitute: foreign/native = profesja/zaj^cie Cx: conventional (abduction —•) In stage 2 Hearer! makes wrong inferences concerning the lexical item profesja by extending its meaning to any type of zajgcie while narrowing it in the Interactional and Affective components to that of (foreign), which causes the formation of the reconfigurated scheme for profesja. In the role of a speaker, he/she employs a concept of Plan, which is another, besides Frame and Script, prototypical pattern of knowledge utilized by language users. The component Substitute: foreign/native in the above Plan, which is only a drastically abbreviated version of a more elaborate schema, confirms the necessity of the socio-cultural layer in the meaning component of the language since, although such a subgoal in Impress-Hearer-Plan can always be subject to individual variation, it seems to be more prototypical for a native speaker of Polish than, for instance, for a native speaker of English. In both cases it is constrained or weakened in the experiential and actual layers of meaning. The respective average can be experimentally established in each case. Stage 3 is developed in diagram (16). It shows the possible reactions of Hearer 2 to Speaker's modified use of profesja, indicating at the same time the (spiral) effect of Hearer 2 /Speaker 3 's response upon the establishment of the alternative sense of the word with Speaker 2 . (16)

a = profesja a' = profesja' (product of wrong inferences)

Interpersonal operations active in the processes diagrammed above are subject to context 2 and Affective constraints, which results in using different possible strategies: Pass, Announce procedural problem, Fail.

318

Barbara

Lewandowska-Tomaszczyk

H2:

i Disconfirms a'

The paradigm of Announcing procedural problem; Negotiating shows the following pattern: (17)

(numbers in brackets indicate the optional turns of the interactants) 1. Si: Introductory turn 2. S2: Procedural problem announcing (recursive) Cause: a) f missing information b) [disagreement 3. S}: Attempt at repair 4. S2: 2 5. Sj: 3 6. S2: Acknowledgment 7. S2: Response to 1

In the case of Fail in (16), turn (6) in (17) is: Evaluation negative, which constitutes a disconfirming evidence for Speaker!. In the framework of Heringer's (1978) semantics of "Action" and "Interaction" the announcing and negotitating procedural problem scheme may be represented as follows:

On semantic change in a dynamic model of language

319

(18) (Assertion, Question, etc. convey the message χ by Speaker's uttering y) Ass

[ |

S, ·< Ask > (x)

Utterance fAsk

S;

Procedural problem announcing

(-Disagree !

ι S,

!

i

; J S: •--recursive-J

Repair (y) —> U t t e r a n c e

1) Acknowledge:

ipl

2) Negate: s q > _ S:

„ , Respond to χ

rfP [r

Ι

—• < C o m m e n t ! · —• U t t e r a n c e (z) J

ΓΡ

U t t e r a n c e ( m )

φ y ^ χ —* U t t e r a n c e (n)

Break

(Utterance (u)) (a)

The above model of recurrent (verbal) actions can be understood as a non-deterministic pushdown automaton (cf. Skvoretz — Fararo 1980), where no bound is given on the number of successive actions. One utterance establishes (a) condition(s) (T-test) that should be satisfied before a particular sequence of utterances (actions) can occur. The success or failure of the test is evaluated by the problem solver. Thus the interaction entails the following: (19)

Interactants: Speaker^ Speaker 2 UTTerance n by Speaker n establishes Testn UTT n + 1 by S n + 1 satisfies T n and establishes T n + j

51 utt,

Tj

52 υττ2 ~^ T , + + T 2 Si υττ3 ~ > T 2 + + T 3 The meaning content of the linguistic unit proposed in the present paper comprises, as has been shown, both its conventional sense as well as the intended message conveyed indirectly by the speaker. The meaning structures proposed above are taken to represent the T-tests put forward by the speaker to be satisfied by the interactant. If one of the tests is not satisfied (T n ~), i. e., the intended meaning structure does not correspond to the one reconstructed 4 by the interlocutor, the failure in communication may occur. It is signalled

320

Barbara

Lewandowska-Tomaszczyk

either by its termination (20) or by some procedural problems discussed above. (20)

S, UTT, T, $2 UTT2 Ti~ + T2 Si utt, ~ T 2 - #

In natural language interactions the conditions for Τ satisfaction are not always unambiguously stated (intended vagueness of a unit, indeterminate (non-contrastive) context, unmarked salience, ambiguities and inherent vagueness, etc.)· In such cases, if the procedural problems are not signalled, the approximative Τ satisfaction can take place, which creates especially favourable conditions for false inferences. In addition, the addressee may be faced with such T-tests whose conditions are explicit but not conventionally shared (e. g., alternatywa , profesja'). The announcing of procedural problems may not take place there at all (cf. (16)) due to a special social conditioning for instance. It will be promoted mainly by favourable context 2 conditions. The situations of that type sanction, so to say, the semantic change by withholding the disconfirming reactions towards this sort of shift in meaning. The dynamic framework proposed in this paper is an attempt to show language change processes as shaped not only by systemic but also by cognitive, social, and emotional parameters. Therefore, a type of a dynamic componential language paradigm has been proposed, in which all the processes are pertinent to language change. Also the feature analysis employed cannot be limited to definitional, criterial markers, as it is frequently connotational or non-criterial properties that are involved in shifts in natural language semantics.

Notes 1. J. D. Salinger. The Catcher in the Rye. I owe this example to R. Tissot. 2. The problems connected with metaphorical extensions and establishing analogical dimensions between different cognitive domains have been discussed in Lewandowska-Tomaszczyk (1984). 3. Analysability principles in Langacker (1982) have the same function, although they are incorporated into a different framework. I am using the symbolism proposed by Langacker for rules of Analysability.

On semantic change in a dynamic model of language

321

4. I am trying to avoid the term "decoding" as it seems to imply a reduced share of the addressee's own receptive creativity in language understanding.

References Andersen, Henning 1973 "Abductive and deductive change", Language 49: 765-793. 1980 "Morphological change: towards a typology", Historical morphology. Trends in Linguistics. Studies and Monographs, edited by J. Fisiak (The Hague: Mouton), 1 — 50. Bailey, Charles-James — Shuy, Roger W., eds. 1973 New ways of analyzing variation in English (Washington, D. C.: Georgetown University Press). Beaugrande, R. de 1983 "Surprised by syncretism. Cognition and literary criticism exemplified by E. D. Hirsch, Stanley Fish, and J. Hillis Miller", Poetics 12: 83-137. Bickerton, Derek 1972 "The structure of polylectal grammar", Sociolinguistics: Current trends and prospects. Proceedings of the Twenty-Third Annual Round Table, Monograph Series on Languages and Linguistics. No 25, edited by R. W. Shuy (Washington, D. C.: Georgetown University Press). Buttler, Danuta 1974 Polski dowcip jgzykowy (Warszawa: PWN). Cole, Peter — Morgan, Jerry L., eds. 1975 Syntax and semantics, vol. 3: Speech acts (New York: Seminar Press). Dahlgren, Kathleen 1978 "The nature of linguistic stereotypes", Papers from the parasession on the lexicon, edited by D. Farkas et al. (Chicago: Chicago Linguistic Society, University of Chicago Press), 58-70. Davenport, M. — Hansan, E. — Nielsen, H. F., eds. 1983 Current topics in English historical linguistics. Odense University Studies in English, vol. 4 (Odense: Odense University Press). Edelson, M. — Janicka-Swiderska, I. — Nal?cz-Wojtczak, J. — Sturgess, Ph. — Styczynska, Α., eds. 1984 Studies in English and American literature. In honour of Witold Ostrowski (Warszawa: PWN). Farkas, Donka — Jackobsen, Wesley M. — Todrys, Karol W., eds. 1978 Papers from the parasession on the lexicon (Chicago: Chicago Linguistic Society, University of Chicago Press). Fawcett, Robin P. 1980 Cognitive linguistics and social interaction (Exeter Linguistic Studies 3) (Heidelberg: Julius Groos Verlag and Exeter University). Fisiak, Jacek, ed. 1980 Historical morphology. Trends in Linguistics. Studies and Monographs (The Hague: Mouton). 1981 Contrastive linguistics: prospects and problems (The Hague: Mouton). Greenberg, Joseph H. — Osgood, C. E. — Saporta, S. 1965 "Language change", Psycholinguistics, edited by C.E.Osgood (Bloomington: Indiana University Press), 146-163. Grice, H. Paul 1975 "Logic and conversation", Syntax and semantics, vol. 3: Speechacts, edited by P. Cole and J. L. Morgan (New York: Seminar Press), 41-58.

322

Barbara

Lewandowska-Tomaszczyk

Hajicova, Eva, ed. 1982 CO LING 82. Abstracts. (Praha: Univerzita Karlova). Hartmann, Reinhard Rudolf Karl, ed. 1983 Lexicography: principles and practice (London: Academic Press). Heringer, Hans Jürgen 1978 Practical semantics (Trends in linguistics 3) (The Hague: Mouton). Householder, Fred W. 1983 "Kyriolexia and language change", Language 59: 1-17. Jakobson, Roman 1971 "Two aspects of language and two types of aphasic disturbances", Selected writings, vol. II: Word and language (The Hague: Mouton), 239-259. Katz, Jerrold J. 1972 Semantic theory (New York: Harper and Row). Labov, William 1970 The study of non-standard English (Urbana: National Council of Teachers of English). 1973 "The boundaries of words and their meanings", New ways of analyzing variation in English, edited by Ch.-J. Bailey and R. Shuy (Washington, D. C.: Georgetown University Press), 340—373. Lakoff, George 1982 Categories: an essay in cognitive linguistics. Linguistics Department and Cognitive Science Program. (University of California at Berkeley.) Langacker, Ronald W. 1976 "Semantic representations and the linguistic relativity hypothesis", Foundations of language 14: 307-357. 1977 "Syntactic reanalysis", Mechanisms of syntactic change, edited by Ch. N. Li (Austin: University of Texas Press), 57-140. 1982 "Space grammar, analysability, and the English passive", Language 58: 22-80.

Laszlo, E. 1969 System, structure, and experience (New York: Gordon and Brea). Lewandowska, Barbara 1982 "Meaning negotiation in dialogue", CO LING 82. Abstracts, edited by E. Hajicova (Praha: Univerzita Karlova), 182-185. Lewandowska-Tomaszczyk, Barbara 1983 a "Convention and creativity in natural language", Studia Anglica Posnaniensia XVI: 97-105. 1983 b "Some psychological aspects of contrastive pragmatics or in search of a cognitive contrastive analysis of social interaction (CCASI)", Crosslanguage analysis and second language acquisition 1. Jyväskylä CrossLanguage Studies No 9., edited by K. Sajavaara (Jyväskylä: Department of English, Univesity of Jyväskylä), 157-169. 1983 c "The cognitive basis for dynamic semantics", in press. 1984 "Metaphor as a mental phenomenon", Studies in English and American literature, edited by M. Edelson et al. (Warszawa: PWN), 95-100. Lewandowska-Tomaszczyk, Barbara — Machova, S. 1984 "Meaning negotiation in man-man and man-machine dialogue", in press. Li, Charles N., ed. 1977 Mechanisms of syntactic change (Austin: University of Texas Press). Miodunka, Wladyslaw 1980 Teoria pöl jgzykowych (Warszawa: PWN).

On semantic change in a dynamic model of language

323

Nöth, Winfried 1983 "Systems theoretical principles of the evolution of the English language and literature", Current topics in English historical linguistics, edited by M. Davenport et al. (Odense: Odense University Press), 103-122. Osgood, Charles E., ed. 1965 Psycholinguistics (Bloomington: Indiana University Press). Pisarek, Walery 1978 Slownik j§zyka niby-polskiego, czyli bl§dy jgzykowe w prasie (Wroclaw: Zaklad Narodowy im. Ossolinskich). Rieger, Burghard B., ed. 1981 Empirical semantics: new approaches in the field, vol. I (Bochum: Brockmeyer). Romaine, Suzanne 1982 Socio-historical linguistics: its status and methodology (Cambridge Studies in linguistics 34) (Cambridge: University Press). Rosch, Eleonor H. 1973 "Natural categories", Cognitive Psychology 4: 328-350. Sajavaara, Kari 1984 "Psycholinguistic models, second language acquisition, and contrastive analysis", Contrastive linguistics: prospects and problems, edited by J. Fisiak (Berlin, New York: Mouton), 379^08. Sajavaara, Kari, ed. 1983 Cross-language analysis and second language acquisition 1, (Jyväskylä Cross-Language Studies no. 9) (Jyväskylä: Department of English, University of Jyväskylä). Schänk, Roger — Abelson, R. 1977 Scripts, plans, goals, and understanding (Hillsdale, N. J.: Erlbaum). Shuy, Roger W., ed. 1972 Sociolinguistics: Current trends and prospects, Proceedings of the Twenty-Third Annual Round Table. Monograph Series on Languages and Linguistics No. 25 (Washington, D. C.: Georgetown University Press). Skvoretz, J. — Fararo, Th. J. 1980 "Languages and grammars of action and interaction: a contribution to the formal theory of action", Behavioral Science 25: 9-22. Steels, L. 1981 "Frames and descriptions", Empirical semantics: new approaches in the field, vol. I, edited by Β. B. Rieger (Bochum: Brockmeyer), 192-235. Tomaszczyk, J. 1983 "On bilingual dictionaries", Lexicography: principles and practice, edited by R. R. K. Hartmann (London: Academic Press), 41-51. Wierzbicka, Anna 1972 Semantic primitives, translated by A. Wierzbicka and J. Besmares (Frankfurt: Athenäum). Dictionaries Brückner, Alexander 1970 Slownik etymologiczny jgzyka polskiego (Warszawa: Wiedza Po[1927] wszechn^). Doroszewski, Witold, ed. 1958-68 Slownik jgzyka polskiego, vols 1-10 (Warszawa: PWN).

CHARLES Ν. LI

Contact-induced semantic change and innovation* 0. Introduction This paper represents a departure from the traditional investigation of semantic change in two ways: First, it is not restricted to semantic change of lexical elements; secondly, it elucidates the path of semantic change which, as a consequence of languages in contact, involves either an unusual exploitation of homophony or an unexpected scrambling of syntactic structures. I will begin by introducing in Section I the three language of Central Asia from which data will be drawn in later sections. In Section II, I will present two cases of contact-induced semantic change at the lexical level. In Section III, a case of morphological innovation resulting from a semantic change is presented, and finally, Section IV discusses the semantic change of a complex sentence type as a result of borrowing.

1. Three languages of Central Asia The linguistic data in this paper is from the part of Central Asia which is located in China. The area covers the upper reaches of the Yellow River, not far south of the western terminal point of the Great Wall. To the north and northeast of this area is the homeland of the Mongols; to the northwest are the Turkic tribes of the Uighurs and Khazakhs; to the south and southwest lies the domain of the Tibetan culture; and to the east is the old civilization of the Han Chinese. It is, therefore, not surprising that this area should be a center where different ethnic groups, languages, and cultures come in contact with each other. Today aside from the Han Chinese, most of whom arrived as immigrants following the establishment of the establishment of the People's Republic of China, seven other ethnic groups make their home in this region: Amdo Tibetan, (Anduo Zangzu 1 ), Santa (Dongxiangzu), Salar (Salazu), Baonan (Baoanzu),

326

Charles Ν. Li

Monguor (Tuzu), Hui (Huizu), und Wutun. 2 Contact between these languages and the resultant linguistic borrowing and interference in this area is intense. The three languages, from which the data for this paper is extracted, are Baonan, Wutun and Hui. Baonan is a language of the Mongolian family. It is significantly different from Khalkha Mongolian, the standard language of the Mongolian People's Republic. The genetic distance between the two languages is comparable to that between German and English. There are two distinct dialects of Baonan, one representing the speech of approximately 5000 Moslems occupying five villages in Jishishan County in southwestern Gansu province, the other being the speech of approximately 2000-3000 Lama Buddhists residing in Tongren County in southeastern Qinghai province. The major difference between the two dialects is that the language of the Lama Buddhists has been strongly influenced by Amdo Tibetan, whereas the language of the Moslems shows a great deal of borrowing from the Chinese. The Baonan data in this paper is taken from the Moslems of Ganhetan village in Jishishan County. The Wutun language belongs to an ethnic group which immigrated from the eastern seaboard of China to their present home in southeastern Qinghai province several centuries ago. According to the criteria of lexicalstatistics, Wutun is a member of the Chinese family. However, the lexicon as well as the phonology, morphology and syntax of the language bear little resemblance to other Chinese languages. For instance, it has lost its tones — an imporant phonologic feature of Chinese; it has a full-fledged case system and complex suffixal verb morphology; its syntax is that of a typical polysynthetic, verb-final language; and finally, its lexicon includes a large number of Amdo Tibetan loan words. Over the centuries, Amdo Tibetan has had a major impact on Wutun, but Boanan and Monguor have also left their mark on the language. The Hui language of Linxia in southwestern Gansu province is a Chinese dialect spoken by the Moslems. Although the lexicon of Hui is overwhelmingly Chinese, its grammar displays a wide array of features commonly associated with verb-final languages. It is pointed out in Li (1948 a) that the verb-final features of Hui are probably the result of substratum interference, that is, interference of the native language of an immigrant people during their acquisition of the language of their new home, rather than the result of borrowing, that is, the incorporation of features from other

Contact-induced semantic change and innovation

327

languages into the native language of a people. 3 In this paper Hui is discussed only in the light of its role as a language in contact with Baonan. 4

II. Lexicon Consider the Baonan words for 'one thousand' to 'nine thousand:' (1)

/nagudar/ /iruardar/ /Kuraqdar/ /deraqdar/ /ttAiqdar/ /dzirauqdar/ /daluqdar/ /nimaqdar/ /jesuqdar/

'one thousand' 'two thousand' 'three thousand' 'four thousand' 'five thousand' 'six thousand' 'seven thousand' 'eight thousand' 'nine thousand'

Contrasting (1) with the Baonan numerals from one to nine; (2)

/naga/ /iruar/ /iruraq/ /deräq/ /ta^üq/ /dziriruq/ /dalüq/ /nimäq/ /jesiiq/

'one' 'two' 'three' 'four' 'five' 'six' 'seven' 'eight' 'nine'

One can easily postulate the morpheme for 'thousand' in Baonan as /-dar/. This analysis is straightforward except for /nagudar/ 'one thousand'. Contrasting /nagudar/ 'one thousand' with /naga/ 'one', one cannot avoid questioning the source of the vowel /u/ in /nagudar/. Two pieces of evidence are relevant to the question pertaining to the vowel /u/ in /nagudar/: First, Baonan did not have a morpheme denoting 'thousand' before it came into contact with Chinese. O n e thousand' was simply expressed as /harwäq ndzyn/ 'ten hundred'.

328

Charles Ν. Li

Secondly, the word /nagudar/ in modern Baonan has two meanings: 'one day' and 'one thousand'. The first piece of evidence strongly suggests that the semantic notion of 'thousand' is borrowed from another language which, in this case, must be Hui of Linxia, the language that is most closely in contact with Baonan. Let us, now, examine the second piece of evidence: The word for 'day' in Baonan is /ud6r/. /nagudar/ 'one day' is obtained from /naga/ 'one' + /udar/ 'day' after the application of a Baonan morphophonemic rule which deletes the vowel (a/ when it is concatenated with any other vowel:

In order to account for the afore-mentioned facts. The following hypothesis may be postulated: (3)

Through contact with Hui, the Baonan word /nagudar/ 'one day' became polysemous (i. e., underwent a semantic change) and acquired a new meaning 'one thousand'. Subsequently a re-analysis of the word /nagudar/ 'one thousand' from /nags/ 4- /udar/ into /nagu-dar/ results in the birth of a new morpheme /-dar/ 'thousand'.

The data from Hui bears out the validity of this hypothesis. In Hui, the word for 'day' and the word for 'thousand' are homophonous. Thus, /tqan/ means 'day/thousand' and /ji tean/ means 'one day/one thousand'. We can now reconstruct this contact-indeed diachronic scenario as follows: (4)

Through contact with Hui, the Baonan people saw the need for the semantic notion, 'thousand'. Instead of borrowing the Hui morpheme, /tean/, for 'thousand', the Baonan people copied the homophonous nature of the Hui expressions for 'one day' and 'one thousand' by assigning the same two readings to their indigenous word /nagudar/. Subsequent re-analysis of the word

Contact-induced semantic change and innovation

/nagudar/

yields

the

new

329

phonologically-indigenous

morpheme /-dar/ 'thousand'. A similar, but less complicated case o f contact-induced semantic change is also found in Baonan. Consider the Baonan w o r d for 'ten-thousand': /ajK0/

(5)

/naga ajifa/

'one ten-thousand = ten thousand'

/Kuar ajKs/

' t w o ten-thousand = twenty thousand'

/ifuraq ajifa/ 'three ten-thousand = thirty thousand' /deraq ajifa/ 'four ten-thousand = forty thousand'

/ajKa/ in Baonan also means 'bowl'. Thus, /nags ajifa jamύ/ one bowl noodle one bowl o f noodles. As with the case o f 'thousand', Baonan formerly did not have a morpheme for the semantic notion, 'ten-thousand'. However, its language o f contact, Hui, does have a morpheme, /wan/, f o r the notion 'ten-thousand'. In Hui, the morpheme f o r 'ten-thousand' is homophonous with the morpheme f o r 'bowl', i. e., /wan/ has t w o meanings, ' b o w l ' and 'ten-thousand'. Instead o f borrowing /wan/ f r o m Hui, the Baonan people adopted the homophony between ' b o w l ' and 'ten-thousand' in Hui, and added the meaning 'tenthousand' to their indigenous w o r d f o r bowl, /ajifa/. The above t w o cases o f contact-induced semantic change represent an interesting and unusual path o f semantic association. T h e two semantic notions, ' b o w l ' and 'ten-thousand', become associated with one phonologic f o r m in Baonan because they are homophonous in Hui. T h e same holds true f o r 'one day' and 'one thousand'. A

somewhat similar, but nevertheless distinct path o f semantic

association was observed by Weinreich (1968). H e noted that a morpheme X

in

language A

may

take

on

the

meaning

of

a

morpheme Y in language Β if X and Y are seen as identical or similar. F o r example, because Yiddish speakers in America identify the Yiddish morpheme /op/ ' o f f , d o w n ' with the English morpheme up, /op/ has taken on the meaning 'up' as in /op-rufn/ 'to call

330

Charles Ν. Li

up'. Weinreich calls this type of contact-induced semantic change 'Replica functions for equivalent morphemes' (cf. Weinreich 1968: 39). The keystone of this kind of change is clearly the equivalence of the forms of the two morphemes, according to the speaker's perception. In the Baonan cases, however, there is no equivalence of forms between X and Y, but only the identity of the original meaning of X with one of the meanings of Y. Buck (1949), in his study of synonymous Indo-European words, observed that the paths of semantic association and semantic change are so complex that one can only conclude that each word had its own semantic history. Despite resemblance to Weinreich's generalization, the peculiar path of semantic change in Baonan described here bears witness to Buck's conclusion.

III. Morphology It is well known that the most salient typological characteristics of the Chinese language family are that it is isolating and verb-medial: it does not exhibit any inflection or agreement, and it has few morphological or grammatical markers. But Wutun, which belongs to the Chinese language family according to lexicalstatistic criteria, is different. It is a full-fledged polysynthetic language and it is rigidly verb-final. The change from being isolating and verb-medial to being polysynthetic and verb-final is the result of several centuries of influence by Amdo Tibetan, Monguor and Baonan. However, hardly any of the grammatical morphemes marking cases, numbers, tenses and aspects in Wutun are loan words. A typical grammatical morpheme in Wutun is a Chinese word which was formerly a content word. Its grammatical function was acquired as the language gradually became polysynthetic and verbal-final. In this section, I will describe the emergence of the Wutun instrumental case marker through the contact-induced semantic change of a content word. The instrumental case suffix in Wutun is /-liaqgs/. The following two sentences illustrate its occurrence: (6)

rti magd - liarjga fui hoqdi jy you (Nom.) what - Instrument water spray Verb-Particle What do you use to spray water?

Contact-induced

semantic change and innovation

331

ηO toy - liaqga QUI hoydi jy I pail - Instrument water spray Verb-Particle I use a pail to spray water.

(7)

This instrumental case suffix is homophonous with the word /liaqga/ 'two-Classifier = two'. (8) and (9) illustrate the use of /liag-ga/: (8)

(9)

nihy Uaq-gd daughter two-Classifier Two daughters. a.

Question nia nihy djigd jy? you (Dat.) daughter how many have How many daughters do you have?

b. Answer liaq-ga two-Classifier Two. Since case markers developed in Wutun after the people immigrated to their present home, one may reasonably conjecture that the instrumental case meaning is a new addition to the meaning of the word /Iiaqgs/ 'two'. The evidence in support of this conjecture comes from Baonan, one of the three verb-final languages which have been in close contact with Wutun. First the instrumental case suffix in Baonan is /-Kata/ as shown in (10): (10)

Habib bi - if a Id teidzir - m Habib pen-Instrument flower-Accusative xuagd - dzb draw-Imperfective Habib was drawing the flower with a pen.

Secondly, each of the Baonan numerals from one to nineteen has a corresponding pronominal form denoting animate beings: (11)

/nagula/ /Kuala/ /Kurula/ /derula/

'one (human/animal)' 'two (humans/animals)' 'three (humans/animals)' 'four (humans/animals)'

332

Charles Ν. Li

Sentences (12) and (13) illustrate the use of the pronominal forms of Baonan numerals. (12)

äda

agy

maid

dzänjiq

ndza-dzo

father daughter two movie see-Imperfective Father and daughter, the two are watching a movie. (13)

Kuald

vestiy

xachh-dzo

two grass cut-Imperfective Those/these two are cutting the grass. Consider the Baonan instrumental suffix /-καΐό/ and the pronominal form of the numeral 'two' /kuo16/. The phonological similarity between the two morphemes, especially to the ears of the Wutun people who are not native speakers of Baonan, is striking. Since the meaning of /καΐό/ in Baonan and the meaning of /liaij-ga/ in Wutun are nearly identical, it is most likely that the Wutun speakers then used /liaqga/ as an instrumental suffix on the basis of the phonological similarity, or perhaps what they erroneously perceived as the phonological identity, between /Kuala/ 'two' and /-καΐό/ 'instrumental case' in Baonan. Historically, the semantic change of the word /liaqgs/ in Wutun results in the acquisition of a grammatical function of marking the instrumental case. The significance of this semantic change is that it represents a reconstructed step in the contact-induced drift of Wutun from a verb-medial, isolating language to a poly synthetic, verb-final language. This diachronic drift — a major typological change — developed under the influence of unrelated as well as very distantly related languages. The step of the drift reconstructed here is an interesting example of categorial diffusion between unrelated languages. 5

IV. Syntax The case of semantic change discussed in this section involves a complex sentence type which may be considered to be the serial verb construction of Baonan. Like serial verb construction in other languages, the two verbs in the Baonan construction share a common subject. Another important feature of the Baonan serial verb construction is that one of the two verbs is marked with a subordinator /-dzi/. Sentences (14) and (15) are illustrations:

Contact-induced

(14)

Habib

eidzay-dd

Habib Tibet-Locative

semantic change and innovation

dunei

333

atei-dzi

goods transport-Subordinator

ei-dzi

go-Perfective Habib went transporting goods to Tibet. (15)

dzay

dzay-nd

bangd-dzi

moroq-nä

s/he oar-Accusative pull-Subordinator river-Accusative guogd-dzi

cross-Perfective S/he crossed the river by rowing (a boat). As examples (14) and (15) show, the typical semantic role of the subordinate clause in the Baonan serial verb constructions is to describe the circumstances under which the event in the main clause occurs. The subordinate clause in the Baonan serial verb construction cannot function as a purposive clause or an object complement of the main verb. However, a new semantic function of the subordinate clause in the Baonan serial verb construction has emerged, apparently in recent decades. This new semantic function is to signify the cause of the event denoted by the main predicate. Sentences (16) and (17) are illustrations: (16)

bu

Habib-na

§gd-dzi

fajynga-dzi

I Habib-Accusative hit-Subordinator faint-Perfective I hit Habib with the result that he fainted. (17)

dzat1

difu-na

ur-dzi

S/he doctor-Accusative call-Subordinator rd-dzi

come-Perfective S/he called the doctor with the result that the doctor came. In (16), Ί hit Habib' is the cause and '(Habib) fainted', the result, whereas in (17), 'She called the doctor' is the cause and 'The doctor came', the result. The force behind this semantic change of the Baonan serial verb construction is the influence of the Chinese resultative compound. Since the resultative compound is structurally and semantically identical in all members of the Chinese language family, I will

334

Charles Ν. Li

describe it with examples from the widely-known Mandarin Chinese rather than the relatively obscure Hui, although Hui is the language in contact with Baonan. A typical resultative compound in Chinese is composed of two elements, with the second element signaling some result of the action or process conveyed by the first element. 6 For example, (18)

jiäo-läi call-come call (someone) with the result that s/he comes.

(19)

dä-yün hit-faint hit (someone) with the result that s/he faints.

Resultative compounds are verbs with a high degree of transitivity.7 The morphological process of creating resultative compounds in Chinese is a productive one. Resultative compounds occur frequently in all genres of modern Chinese discourse. Sentences (20) and (21) show examples of resultative compounds serving as transitive verbs in Mandarin Chinese: (20)

wo bä xiäoli da - yün - le 1st. sg. Xiaoli hit-faint-Perfective I hit Xiaoli with the result that s/he fainted.

(21)

tä bä däifu jiäo-läi-/e 3rd. sg. doctor call-come-Perfective S/he called the doctor with the result that the doctor came.

Returning to the Baonan serial verb sentences (16) and (17), we can clearly see the semantic influence of the Chinese construction, as exemplified in (20) and (21): The causative relationship between the two elements of the Chinese resultative compound is transplanted into the relationship between the two verbs in the Baonan serial verb construction. This transplantation has occurred inspite of the fact that Baonan has a productive affixal causative construction. Thus, the Baonan equivalent of the Mandarin Chinese sentences (20) and (21) should be the affixal causative sentences (22) and (23) rather than (16) and (17):

Contact-induced

semantic change and innovation

(22)

bu Habib - ηέ §gd - dzi I Habib-Accusative hit-Subordinator fajyngd -Kd- dzi faint-cause-Perfective I caused Habib to faint by hitting him.

(23)

dzaq difu-nd ur-dzi S/he doctor-Accusative call-Subordinator \d- lid-dzi come-cause-Perfective S/he caused the doctor to come by calling him/her.

335

Indeed, (22) and (23) are perfectly well-formed sentences in Baonan. They are unambiguously causative. Sentences (16) and (17), however, are ambiguous. Their causative readings are newly acquired as a result of contact with Chinese. Their indigenous meanings are the circumstantial readings given below as (16') and (17'): (16')

bu Habib-m §gd-dzi faj'ngd-dzi I Habib-Accusative hit-Subordinator faint-Perfective Hitting Habib, I fainted.

(17')

dzaq defu - m ur-dzi S/he doctor-Accusative call-Subordinator rd-dzi come-Perfective Calling the doctor, s/he came.

With the new causative meaning, the Baonan serial verb construction as exemplified by (16) and (17), exhibit an unusual syntactic phenomenon. Consider (16) for example. The main predicate of the sentence is /fajynga-d^i/ 'fainted'. Its subject which is co-referential with the object of the subordinate verb is absent. The absence of the subject of the matrix verb requires that it must be inferred on condition of co-referentiality with the object of a subordinate verb. Such a requirement is rarely found in languages of the world. The contact-induced semantic change of the Baonan serial verb contruction demonstrated above shows the strength of the on-going sinicization of Baonan. The traditional source of Chinese linguistic influence on Baonan is Hui. Now, as the use of Mandarin Chinese

336

Charles Ν. Li

spreads throughout the People's Republic of China, it has added its influence through education and the news media, intensifying the sinicization of Baonan. It will be interesting to see whether or not the new causative reading of the serial verb construction eventually results in the displacement of the traditional affixal causative in Baonan.

Notes * Preparation of this article was supported by National Science Foundation Grant SBNS 83-08220. The data in this article was collected during a 1982-83 field trip supported by a grant from CSCPRC, National Academy of Sciences and grant #BNS 79-20944 from the National Science Foundation. 1. The term in parenthesis is the Chinese term written in Pinyin, the official Chinese romanization system. 2. The languages of these seven ethnic groups belong to four language families: Tibetan, Turkic, Mongolian and Chinese. Amdo Tibetan, a member of the Tibeto-Burman family, is significantly different from Lhasa Tibetan; Santa, Baonan and Monguor are three distinct Mongolian languages; Salar is Turkic; Wutun and Hui belong to the Chinese family. 3. For an in-depth analysis of substratum interference vis-a-vis borrowing, see Thomason (1982). 4. A more detailed description of Hui, Wutun, and Baonan is available in Li (1984 b). 5. Categorial diffusion between unrelated languages was assumed to be impossible until relatively recently, e.g., Meillet (1921-87). One of the most interesting counterexamples to this traditional assumption, regarding the use of tenseaspect and pronominal categories in Proto-Chinookan, was put forth by Silverstein (1977). 6. For a detailed description of resultative compounds, see Li — Thompson (1982). 7. On the notion of transitivity, see Hopper — Thompson (1980).

References Buck, Carl Darling 1949 A dictionary of selected synonyms in the principal Indo-European languages: A contribution to the history of ideas (Chicago: University of Chicago Press). Hopper, Paul — Thompson, Sandra A. 1980 "Transitivity in grammar and discourse", Language 56.2: 251-299. Li, Charles N. 1981 Mandarin Chinese: A functional reference grammar (Berkeley and Los Angeles: University of California Press). 1984 a "From verb-medial, analytic language to verb-final synthetic language: A case of typological change", Proceedings of the Tenth Annual Meeting of the Berkeley Linguistics Society.

Contact-induced semantic change and innovation 1984 b

337

"Languages in contact in Western China", Papers in Asian languages, 31-51. Meillet, Antoine 1921 Linguistique historique et linguistique generale (Paris: Champion). Silverstein, Michael 1977 "Person, number, gender in Chinook: Syntactic rule and morphological analogy", Proceedings of the Third Annual Meeting of the Berkeley Linguistics Society, edited by K. Whistler et al., 143-56. Thomason, Sarah Grey 1982 "Are there linguistic pre-requisites for contactinduced language changes?", MS. Weinreich, Uriel 1968 Languages in contact (The Hague: Mouton).

LEONHARD LIPKA

Inferential features in historical semantics

1.1. After a time of uncritical adoption and use of the concept of feature — during the heyday of Transformational Grammar — semantic features and componential analysis have recently come under attack, as witnessed in publications by Lyons (1977: 317-335), Leech (21981: 117-122), and summarized by Sprengel (1980). In the following I shall try to show that, in spite of all remaining problems and difficulties, a subclass of features, viz. inferential semantic features, may be applied profitably and successfully to historical semantics. For this purpose a number of examples from the history of English will be discussed. 1.2.1. Before coming to the actual illustration of my points, it is necessary to clarify a few basic issues. First of all I should like to point out that more than ten years ago I investigated the use and status of semantic features and the related question of semantic tests in my book on Semantic structure and word-formation (1972). Moreover, on the basis of this theoretical discussion, I carried out extensive empirical research in this monograph, establishing semantic features inherent in all contemporary verb-particle constructions with out and up. In a later article on semantic components of English verbs and nouns (Lipka 1979) the field of application was further extended, and the justification of metalinguistic elements was again treated explicitly. I considered three general methods of establishing underlying semantic elements: 1. morphological evidence from complex lexemes, i.e. compounds, prefixal and suffixal derivatives; 2. the extraction of semantic components on the basis of paraphrase relationships; and 3. the use of semantic tests including logical relations such as implication, tautology, and contradiction. These procedures, however, are not applicable to the so-called inferential features.

340

Leonhard Lipka

1.2.2. In the article just mentioned I have set up a taxonomy of seven types of semantic features: denotative, connotative, inferential, relational, transfer, deictic, and distinctive features. I cannot go into details here. Suffice it to say that except for the very important class of inferential features, all others function as distinctive features. Only some of them are binary. Like connotative features such as [ + ARCHAIC] in steed and smite, inferential features are supplementary in nature. They differ, however, from the former in being optional, not obligatory and inherent, and they usually depend on context. In my definition the class of inferential features covers not only properties usually associated with a referent — such as slyness with a fox, clumsiness with an ox etc. — but also the influence of co-text and extralinguistic context. 1.2.3. In dictionaries inferential features in my use of the term are normally marked by labels such as "especially" or "usually", or simply brackets. Thus beat is often defined as 'hit (especially with a stick)' or 'not hard'. In the following I will use braces as a notational device for marking such optional features, a convention taken over from Lehrer (1974). There is a two-fold advantage in the recognition of optional semantic elements as opposed to strictly inherent features, based on yes/no-decisions and the principle of the all-or-none. First, they can be used to capture the fuzziness of meaning and linguistic variation in synchrony. Secondly — and more important here — they open a door for describing, formalizing, and explaining semantic change in historical linguistics. 2.1. A good example of the parallelism of regional and stylistic synchronic variation and language change is the case of starve discussed in Lyons (1968:452), Pyles (21971: 348), and Görlach (21982: 124). All three authors also mention the genetically related German sterben, but do not explicitly draw any possible contrastive conclusions1. Lyons makes the point that frequent syntagmatic modification (such as starve of hunger) may lead to incorporation of the sense 'hunger' into starve, and that in the English spoken in some areas of Northern England starve may incorporate 'with cold'. Görlach explains the semantic change in Norghern England and Middle English by restriction to certain contexts and the beginning opposition to the superordinate die. He uses a simplified feature notation which will be adapted in schema (1) together with Lyons' remarks and information from the LDCE and the COD 6:

Inferential features

in historical semantics

341

(1) OE NE ModE

steorfan [BECOME NOT ALIVE] > starve [BECOME NOT ALIVE] + {of COLD} 2 starve [BECOME NOT ALIVE] 4- {of HUNGER} 3

(OE = Old English; N E = Northern English; ModE = Modern English)

Since the Old English period the inferential features {of COLD} and {of HUNGER} have been added in Northern English and standard Modern English respectively, denoting the cause of the process. It was unspecified in Old English, as it still is in German sterben, but the result (death) seems to be an obligatory inherent feature in both languages. The Modern English collocation starve to death would have been tautological in Old English. This demonstrates that a complex semantic component [DIE] is not today a necessary element of starve as we will discuss presently. We could therefore postulate an inferential feature {to DEATH} in Modern English in some contexts. That the inferential feature {of HUNGER} has not been obligatory for a long time is proved by the possibility of syntagmatic modification in the formerly existing compound verb hunger-starv^. According to Lyons (1968: 452), the collocation starve with cold is still possible in Northern England, so that — with an incorporated inferential feature {of COLD} — I'm starving is roughly equivalent to Standard English I'm freezing. Obviously, in the standard expression I'm starving, meaning 'I'm very hungry', the result to death is neither obligatory nor even possible. This also holds for other contexts where, in addition, the inferential feature {of HUNGER} is missing, e. g., She is starving for companionship, The engine was starved of petrol, He's completely sex-starved. They might be accounted for on the basis of a metaphorical relationship and explained with the help of transfer features. Leaving aside this problem here, I should like to illustrate some aspects of semantic change involving metaphorical shift from Old to Modern English in the following diagram:

(2) OE steorfan [DIE] + {SUFFER} > ModE starve (met.)

1 0

1 [SUFFER] [from LACK] + {of FOOD} {of LOVE}

(OE = Old English, ModE = Modern English)

342

Leonhard Lipka

The component [DIE] has disappeared completely in these polysemous uses of starve. For Modern English die, the semantic equivalent of Old English steorfan, none of the standard dictionaries mention suffer in their definitions. It can therefore not be established as an inherent feature of die on the basis of paraphrase relationships, as discussed at the beginning. Suffer is defined in LDCE as 'experience pain'. This is certainly what happens normally when someone dies. We are consequently justified in postulating an inferential feature {SUFFER} for Old English steorfan. This changes into an obligatory inherent feature in all metaphorical uses of the polysemous Modern English verb starve, where the component [DIE] has disappeared. Further inferential features are added from cotext or situational context, such as {of FOOD} etc. The new inherent component [from LACK] might perhaps be related to the complex element [DIE], paraphrasable as 'stop living, no longer have life, lack life'. However, this is rather speculative. 2.2.1. Let us now turn to a more pleasant subject, viz., the example of the semantic change of holiday, which Leech (1974: 123-124) discusses in the first edition of this book on semantics only, in connection with the problem of the fuzziness of meaning. He uses the following diagram (3) for explaining the gradual transition from state Ά', where the expression had the meaning 'holy day, viz. Sunday or religious feast', to state 'D', the Modern English meaning 'a period when one is not required to work'. (3)

holiday: A 'holy day, viz. a Sunday or religious feast' > D 'a period when one is not required to work'

His main point is a distinction between what he calls "criterial components", i. e. obligatory semantic features, and "non-criterial" or "optional features", which are enclosed in brackets. These are identical with my inferential features, viz., {NO WORK}, {of a DAY}, and {HOLY}. They are intended to account for the fact that

Inferential features

in historical semantics

343

the same word, at a given time, can have "two or more overlapping definitions" (Leech 1974: 123). Such inferential features may either become obligatory or disappear completely. In combination, the result of such a "step-by-step progression" may be "a complete shift in the reference of the expression" (Leech 1974: 124). According to Leech the inferential feature {NO WORK} was optional in the original meaning, corresponding to the Modern English collocation holy day.5 This would not be an adequate paraphrase for the lexicalized compound holiday today. Since the optional feature was frequently associated with the expression, it became gradually obligatory. On the other hand, the obligatory components 'of a day' and 'holy' are lost in Modern English, by transition through the states 'B' and 'C' in which they had become optional inferential features. Such a extension of meaning, can be termed 'semantic shift' (cf. Görlach 21982: 119; Ullmann 1972:227; Berndt 1982: 81,87). 2.2.2. Leech's schematic representation looks very convincing. Nevertheless, it is tempting to follow up the actual linguistic development with the help of the OED. First of all it is not surprising to find that the dictionary registers considerable variation, both formally and semantically. Let us first consider the formal side (neglecting phonological developments) as summarized in simplified form in schema (4): (4)

Old English (a) haligdceg (b) halig dceg dat. pi.: dat. pi.: haligdagum halgum dagum from 15th century: hallidai, halliday (and in northern dialects); from 14th century: holidai, holidaie, holiday.

The OED makes a distinction between a "combined form" (4a) and an "uncombined form" (4b), in modern terminology 'compound' vs. 'syntactic group' or 'collocation'. The basis is a purely formal criterion, viz., inflection of the first constituent or its lack. Various spellings are given. As is well known, the Northern English form halliday survives in personal names. Basically three different meanings are distinguished, as quoted in abbreviated form in (5), of which the first is the oldest and the other two have earliest recordings around 1300:

344

(5)

Leonhard Lipka

1. A c o n s e c r a t e d day, a religious festival — c. 950 2. A d a y on which o r d i n a r y occupations ... are suspended; of exemption or c e s s a t i o n from w o r k ; of festivity, r e c r e a t i o n , or amusement - a 1300 ... 3. A t i m e or p e r i o d of cessation from work ... (= collect, pi. or sing.) — 13

I have marked relevant semantic material by spacing. This shows that Leech's analysis is basically correct, and that meanings with or without the inferential feature {of a DAY} coexist from about 1300 onwards. State Ά ' in diagram (3) is further justified by the remark in the dictionary that meaning 2 in (5) is "in early use not separable from 1". It is interesting to note that, according to the OED, form (4b) became more frequent "as the distinction in signification between sense 1 and sense 2 became more marked". In spite of some difficulties the following quotations (6) may illustrate how the inferential feature {NOW W O R K } came in and {of a DAY} and {HOLY} went out: (6)

For meaning 2\ a 1300 Iesus went him for to plai Wit childir on a halidai 1478 One for the halydays ... and a no thy r for the workyng days 1601 Hence: home you idle Creatures, get you home: Is this a Holiday? (Shakespeare, Julius Caesar) For meaning 3: 13.. Er pe halidayez holly were halet out of toun 1667 The Christmas holidays giving more leave and licence to all kinds of people.

2.2.3. The OED also mentions two further subsidiary meanings of holiday which I shall now discuss. One is contained in the obsolete or extinct phrase to speak holiday defined as 'to use choice language, different from that of o r d i n a r y life'. This could lead us to introduce an additional inferential feature {EXTRAORDINARY} (cf. also meaning 2 in (5)), related to the feature {HOLY}. Such an inferential feature could be supposed to have been present already in state Ά '

Inferential features

in historical semantics

345

of diagram (3), while all the other features of that stage must have disappeared in the phrase. The second use of holiday is characterized as colloquial and nautical by the dictionary and defined as 'a spot carelessly left uncoated in tarring or painting'. The following diagram (7) may explain the semantic change and again illustrates the usefulness of an inferential feature {EXTRAORDINARY}:

TIME {of a DAY} {HOLY} {EXTRAORDINARY} NO WORK

(b)

PLACE

EXTRAORDINARY NO WORK

Holiday in this specific nautical register can be defined as 'extraordinary place where no work has been done', which is crudely formalized in (7b). The starting-point of the semantic change, viz., (7a), can be compared to state Έ ' in (3). I have replaced 'period' by [TIME] and added the inferential feature {EXTRAORDINARY}. This is converted into an obligatory component. The other inferential features have disappeared, while the feature [NO WORK] is retained in (7b). The most radical semantic change from [TIME] to [PLACE] cannot be interpreted with the help of inferential features. In my opinion it must be regarded as an abrupt conversion or replacement rather than a gradual transition, or step-by-step progression for which Leech set up his model. Obviously, we here reach the limits of inferential features. Furthermore, the actual process of the shift of meaning of holiday may have been quite different from the reconstruction given in (7). Perhaps the situation of coining the new meaning was something like an officer saying: "Is this were you took a holiday?" to a lazy sailor, pointing to the spot in question. It is well known that not all shifts of meaning are gradual, but some are rather sudden, accidental, and anecdotal. Possibly the best examples of unique changes are those of the history of ModE mint and money and F croissant, denoting a milk-roll, as explained by Ullmann (1972: 197). The four fundamental "types" of change: metaphor, metonymy, popular etymology, and ellipsis may all involve cases of sudden shift of meaning. On the other hand it is quite

346

Leonhard

Lipka

normal that a word gradually acquires a new sense, which then coexists with the old one for some time or even indefinitely. 2.2.4. If we now look at holiday from a strictly synchronic point of view, we can state that some inferential features are relevant for Modern English as well, not only for a historical approach. Thus the inferential feature {of a DAY}, postulated for the transition from meaning 2 to 3 in schema (5), is not at all obligatory in Modern English, but optional. This is evident in the following definition of holiday from the LDCE: 'time of rest from work, a day ... or longer' (cf. also (7a)). Another possible inferential feature is {RE-CREATION}, derived from sense 2 and 3 in (5) and the following definition in COD 6: 'day of festivity or recreation, when no work is done ... period of this'. Its optionality is supported by the fact that recreation is not mentioned in the LDCE. 2.3.1. Turning away from our holiday to serious work, it may not seem out of place, in the context of a conference on historical semantics and word-formation, to make a few remarks on lexicalization (cf. Lipka 1977; 1981). As I understand this term (cf. Lipka 1981: 120 f.) lexicalization is a multi-layered historical phenomenon, in which a complex lexical item, through frequent use, gradually loses the character of a syntagma and formally and semantically tends to become a single, specific, lexical unit. This process may involve graphemic and phonological changes, sometimes referred to as demotivation, but also morphological and syntactic alterations, and especially semantic modifications. The latter may be largely captured by the theoretical construct of loss or addition of semantic features, which describe the gradual process of idiomatization. In holiday both processes can be found, as well as phonological and morphological changes that isolate the lexicalized compound from its constituents and a parallel syntactic group (cf. 2.2.2. and Bloomfield 1933: 434). In other cases such as blackboard or watchmaker, which are perfectly analysable today, demotivation has been caused by developments in the extralinguistic world, something I have referred to (Lipka 1981: 124) as 'referential change' ("Referenzwandel"). 2.3.2. I distinguish lexicalization from what may be called 'instantaneous' or 'individual coining', a phenomenon I once termed "Einzelprägung" (Lipka 1981: 122). This is tied up with the naming function of simple and complex lexical items and must be considered

Inferential

features

in historical

semantics

347

as a singular act, in which new concepts or concrete extralinguistic referents are given a name. This may lead to instanteneous idiomaticity — a fact already noted by Hermann Paul (cf. Lipka 1981: 122) — but does not necessarily do so. Examples for the former are streaker 'person running naked across a public place' and G Geisterfahrer 'car going in wrong direction on a motorway'. Examples of unidiomatic unique coinings may be pedestrianization or G Windabweiser 'part of car that keeps wind away'. 3.1. I will now turn — in less detail — to some other examples of semantic change that have been repeatedly used in the literature (cf. Bloomfield 1933: 426, 431^32, 440-441; Pyles 21971: 347 ff.; Ullmann 1982: 227 ff.; Görlach 21982: 118-119; Berndt 1982: 76 ff.; Nöth 1979: 27 ff.). As Pyles (21971: 347) points out, many of them can already be found in a book by Greenough and Kittredge published in 1901, but have been adopted since "they make their point better than less familiar ones would do". I will here concentrate on two fundamental "categories" (cf. Ullmann 1972: 227) of semantic change, to which a feature approach is most amenable, and which are based on the result of change and the range of words, viz., extension and restriction of meaning as illustrated in schema (8): (8)

(a) Restriction (features added): deer, fowl, hound, liquor, starve; (b) Extension (features subtracted): barn, bird, dog, meat, mill, tail.

In the following, I shall distinguish two groups of authors who have dealt with these examples and will try to review the most important points they make. With all of them the analyses of specific items do not differ much. 3.2.1. Bloomfield, Pyles, and Ullmann belong together in that they do not draw on the concept of feature. Bloomfield surveys previous research, notably Paul, with his distinction between general and occasional meaning and Sperber, with his stress on the context of new meanings. In this connection Bloomfield uses the term "extension of meaning"; otherwise the speaks of "narrowing" and "widening". For him (Bloomfield 1933: 440 f.) finding the context or situation in which a linguistic form may be used with both the old

348

Leonhard Lipka

and new meanings is the key for explaining semantic extension. At the same time the paradigmatic and syntagmatic co-text — in modern terminology — must be considered, e. g., the competition of meat and flesh, and possible unfavourable connotations 6 . 3.2.2. Pyles distinguishes between "specialization" and "generalization" of meaning, a classification "based on scope". He treats some examples that are not discussed in the other books, viz., barn, mill, tail, and liquor, and also makes reference to regional semantic variation between British English and American English. Thus barn was originally a compound of OE bere 'barley' and am 'house'. We may therefore postulate an inferential feature {BARLEY} for the specific kind of cereal or grain that disappeared in the development to modern British English barn 'storehouse for grain'. In American English and some other varieties of English even the feature [GRAIN] is no longer obligatory, since barn may be defined as a 'building for storing hay, livestock, vehicles, etc.' (cf. COD 6). Finally, the feature [for STORING] may also be absent in some other uses of barn as in the definition of barn3 in the LDCE as 'a big bare plain building'. On the synchronic level we can therefore postulate two inferential features {GRAIN} and {for STORING}. Pyles (21971: 347) further points out that a mill was formerly a place for making things by grinding, viz., meal (etymologically related), and that it is now only "a place for making things", because "the grinding has been eliminated" 7 . As evidence he cites the syntagmatic modifications in "woolen mill" (sic!), steel mill, and even gin mill. The situation is, however, more complicated and we might be induced to postulate inferential features such as {GRINDING}, {BUILDING}, and {GRAIN} both for diachronic and synchronic purposes. The third item considered by Pyles alone is tail, from OE teegl, which he defines as probably having meant 'hairy caudal appendage, as of a horse'. According to him the "hairiness" and "horsiness" have been eliminated in the development to ModE tail, which could be represented by the inferential features {HAIR} and {HORSE}, or {EQUINE}, the latter found as inherent features in ModE horse, stallion, mare, gelding, foal, filly, colt etc. The semantics of tail are far more complex if we include metaphorical and metonymic extensions, as a glance at the dictionaries will show. Thus, for example, we probably need an inherent feature [BACK] to account for tail in the sense of 'reverse of coin'.

Inferential features

in historical semantics

349

I therefore turn to the last item, liquor, which is also special in being a case of semantic restriction. Pyles (21971: 348) claims that originally it simply meant 'fluid', but that "we have added 'alcohol' ". This is not borne out if we look up liquor in the COD 6, where alcohol plays a very small role. On the other hand the LDCE is much more alcoholic, since the first definition is plainly 'alcoholic drink', and the second one, labelled American English, even reads 'strong alcoholic drink'. It is therefore probably not unwise to postulate an inferential feature {ALCOHOL} both synchronically and diachronically. 3.2.3. Ullmann (1972: 195) draws attention to the fact that many semantic changes arise in "ambiguous contexts" and notes (Ullmann 1972:229) that "extension of meaning" is apparently "a less common process than restriction". He gives some very interesting examples for extension (Ullmann 1972: 231), viz. F pigeon, dindon 'turkey-cock', hetre 'beech', and ModE bird from OE brid 'young bird'. In all these cases, which denote whole species, the original meaning was 'young animal or plant' (cf. also G Schößling 'young plant'). We can therefore postulate an inferential feature {YOUNG} that has disappeared. It was formerly an inherent distinctive feature, and still is in a number of Modern English lexical items, such as boy, girl, foal and its hyponyms, etc.8. The reverse process, viz. restriction of meaning, is illustrated by Ullmann (1972: 229) with the help of the examples deer, earlier meaning 'beast', hound, formerly 'dog', fowl, once denoting 'bird' in general, and starve. He notes that G Tier, Hund, Vogel, and sterben have "retained the wider meaning". We might postulate the additional inferential feature {family CERVIDAE} 9 added to the sense of OE deor, {for HUNTING} in hound, and {family GALLUS} incorporated in the sense of earlier OE fugol. It would seem that German always retains the original sense and is thus less susceptible to semantic change. However, Pyles (21971: 350, fn. 10), giving the further items G Knabe, selig, Knecht (to which we might add Korn and Mühle), had already warned against jumping to this conclusion, and argued that the impression would not be the same with a different choice of examples. 10 To finish the review of this group of linguists, let me state that all three, viz. Bloomfield, Pyles, and Ullmann, draw on relevant German material and do not diverge much in their opinion on particular lexical items.

350

Leonhard Lipka

3.3.1. Görlach (1974: 118 f.) was to my knowledge the first to use a simplified feature notation for the explanation of semantic change in the history of English, if we disregard Leech's isolated schema for holiday. He illustrates extension of meaning by an inherent feature [ + jung], becoming optional in ModE bird, which corresponds to our inferential feature {YOUNG}. For restriction of sense Görlach uses [ + zur Jagd] and [+ fleischl.] in ModE hound and meat respectively, which were missing in OE hund and mete. He gives detailed chronological tables (Görlach 21982: 123) for the semantic development of ModE sad, silly, nice, and stout, where the overlap of particular senses at certain times is clearly shown. This could also be captured with the notion of inferential feature, something Görlach does not attempt to do. Finally, his approach is exceptional in that he does not confine himself to the investigation of single lexical items in isolation, but tries to analyse a whole wordfield in its semantic development (Görlach 21982: 126 f.). Thus, the interdependence between ModE farm, hamlet, village, town, city and their earlier equivalents are studied on the basis of a text corpus. This field of 'a collection of dwellings', for which an archilexeme is missing, also includes, e. g., OE wie, häm, castel, ceaster, and burg, for which semantic continuations in Modern English do not exist. 3.3.2. Berndt (1982: 81-87) also takes into account paradigmatic lexical relationships in his discussion of semantic change, but in addition includes syntagmatic context by quoting extensive syntagmas or full sentences. He argues that "changes in the meaning of a word ... have to be seen from the point of view of their effects upon the relation of this word to other words in the same semantic field or the same subsystem" (Berndt 1982: 80), and therefore jointly investigates the history of fowl and bird (81-82), deer, beast, and animal (82-83), meat, flesh, and fodder (83-84), hound and dog (86), and the co-occurrence of ME sterven, swelten, and dien (85-86). In many cases he demonstrates overlap and even synonymy at some time during the Middle English period. He explicitly deals with meaning differences between "cognate words in English and German (partly) due to narrowing of the range of reference in the history of the English items" (86-87) Berndt (1982: 81) generally explains "specialization" and "generalization" as due to the " a d d i t i o n of certain structural components" and the " s u p p r e s i o n or loss of certain structural components" 11 , with substitution combin-

Inferential features

in historical semantics

351

ing the two processes (81). However, he never makes an attempt to isolate or formalize these components with a feature notation. 3.3.3. Our last author, Nöth, is exceptional because he applies such a notation to both diachronic and contrastive semantics at the same time. Using many of the English and German examples already treated here, and a considerable number of others, his aim is to find out the affinities and differences between the two branches of semantics. He argues for a more dynamic view of the interrelation between the two systems, which is quite compatible with my conception of inferential features. 4. This brings us back to my own proposal and the conclusions one can draw from it. I believe that the notion of the inferential feature is both necessary and inevitable, if one accepts the idea of semantic components at all. It is furthermore extremely useful for descriptive adequacy and possible generalization, as well as for descriptive and didactic purposes. In synchrony, it can account for fuzziness of meaning, for polysemy, and for regional, stylistic, and other variation (cf. 2.2.4.). On the diachronic scale, it can capture semantic restriction, extension, and shift and possibly other changes of meaning. A number of specific features have been established in the course of this paper, as summarized in (9), with (9a) and (9b) following the same order as (8a) and (8b), and (9c) containing the inferential features mentioned alsewhere in the order of appearance in the text: (9)

(a) {family CERVIDAE}, {family GALLUS}, {for HUNTING}, {ALCOHOL}, {of HUNGER} (b) {BARLEY}, {GRAIN}, {for STORING}; {YOUNG}; {particular BREED}; {FLESH}; {GRINDING}, {BUILDING}, {GRAIN}; {HAIR}, {HORSE} = {EQUINE} 12 (c) {of COLD}, {CAUSE}, {of HUNGER}, {to DEATH}, {SUFFER}, {of FOOD}, {of LOVE}, {NO WORK}, {of a DAY}, {HOLY}, {EXTRAORDINARY}, {RECREATION}.

Obviously, the items discussed in section 3 and listed in schema (8) have not been treated in the same detailed way that was reserved for our crucial example holiday. Nevertheless, I am convinced that

352

Leonhard Lipka

if they were followed up in OED in the same manner, this would further support my argument and prove the value of inferential features for synchronic and historical semantics.

Notes * I should like to thank Elspeth Davidson and Helmut Gneuss for very helpful comments on an earlier version of this paper. The term "inferential feature" ultimately derives from Nida (cf. Lipka 1979) and denotes non-obligatory meaning components which may be "inferred" from the use of an expression. 1. Berndt (1982:86), however, states: "The semantic differences between ModE starve and its modern German cognates, H G sterben and Low German jstaivnj, arose as a result of addition of further structural components to the underlying structure of the earlier meaning shared with ME dien." Cf. Lipka (1980:102-103) for hyponyms of sterben and die, contrastive equivalents, and the problem of choosing between analytical paraphrases. 2. Cold and hunger may be further analysed as 'absence or lack of heat/food', cf. LDCE starvei and starve3 'to (cause to) suffer from not having come stated thing'. For possible approaches to the optional causativity in all three meanings of starve cf. Lipka (1982). We might postulate an inferential feature {CAUSE}. 3. ME sterve(n), according to Görlach (21982: 124), is already characterized by an obligatory feature [ + durch Hunger], However, the six occurences of the verb in Chaucer's works (cf. Skeat 1912: 381, 431, 435, 479, 557, 653) all clearly show the lack of this feature. Although the editor in his glossary (Skeat 1912:105) in one instance explicitly defines sterve as 'die of famine', a look at the passage immediately shows that this is not correct, since we find: sholde hir children sterve for famyne. We have a clear case of syntagmatic modification, which is the best test for non-incorporation of the meaning 'of famine' in sterve. Of the 50 entries in the Chaucer Concordance under starve not a single one justifies [of HUNGER], The earliest date for the sense 'to die of hunger' in the OED is 1578. 4. Cf. Pyles (21971: 348). The OED gives 1390 and 1879 as the earliest and latest dates respectively for this verb. For the general problem of the interrelation between syntagmatic modification and paradigmatic lexical structuring cf. Lipka (1981). 5. If we follow Liebermann (1903-1916; 1960: 399, 656 f., s. v. Feiertag 5), Sonntag 3) and 4)) this is not correct, since there were strict laws that prohibited work on a holiday in the Anglo-Saxon world. 6. In another connection Bloomfield (1933: 432, 440) mentions meat and drink, sweetmeats, and counted one's bedes (more currently tell one's beads). The coexistence of the former cases with ModE meat 'flesh-food' could be accounted for by an inferential feature {FLESH} cf. also the Modern English saying one man's meat is another man's poison. In the expression tell one's beads the noun bead 'prayer' is synchronically unrelated to bead 'small object' and there is only an etymological connection. 7. Cf. the definitions in SDCE: 1. '(a building containing) a machine for crushing corn or grain into flour'; 2. 'a factory or workshop', and in COD 6: 1. 'Building fitted with machinery for grinding corn'; 2. 'Any mechanical apparatus for grinding corn ... any solid substance ...'; 3.'Any machine, or building fitted with machinery, for manufacturing-processes etc.'. Nöth's (1979: 33) treatment

Inferential features in historical semantics

8.

9.

10. 11.

12.

353

of the "metaphorical extension" of the meaning of mill is apparently based on the COD or OED. Cf. also the syntagmatic modification in: a) flourmill, watermill, windmill, b) coffee-mill, pepper-mill, c) cotton-mill, paper-mill as a test. As a binary feature it is also relevant for the system of address in many languages, e. g., as one of the factors involved in the distinction between tu, du (for children) and vous, Sie (for adults). Cf. COD 6 s. v. deer; Nöth (1979: 31); Pyles (21971: 348), who points out that the feature was not yet present in Shakespeare's "Mice, and Rats, and such small Deare". Cf. also Berndt (1982: 92-95) for boor and G Bauer; boy, knave, child and G Knabe; silly and G selig; and knight and G Knecht. Pejoration and (a)melioration are accounted for by "gain or loss of evaluative meaning components" (Berndt 1982: 93 ff.). Thus e. g. knave, silly, and G Knecht are said to have acquired "negative evaluative components", while knight is a case of additional "positive evaluative components". For a semantic feature Negative Evaluation [ + NegEv] in G stinken and participial adjectives like choked up, snowed up, mixed up, glued up, inked up cf. Lipka (1972:136-138, 208). For ME dogge 'dog of a particular breed' cf. Bloomfield (1933:426) and also G Dogge. This leads to an inferential feature {particular BREED}. For meat cf. Bloomfield (1933: 425-426, 431, 440-441), Pyles (21971: 348).

References Berndt, Rolf 1982 A history of the English language (Leipzig: VEB Verlag Enzyklopädie). Bloomfield, Leonard 1933; 1935 Language (London etc.: George Allen & Unwin). Brekle, Herbert E. — Kastovsky, Dieter (eds.) 1977 Perspektiven der Wortbildungsforschung (Bonn: Bouvier). Görlach, Manfred 1974 Einfuhrung in die englische Sprachgeschichte (Heidelberg: Quelle & Meyer). 1982 Einfuhrung in die englische Sprachgeschichte, 2nd edition (Heidelberg: Quelle & Meyer). Kastovsky, Dieter, ed. 1980 Perspektiven der lexikalischen Semantik. Beiträge zum Wuppertaler Semantikkolloquium vom 2. bis 3. Dezember 1977 (Bonn: Bouvier). Leech, Geoffrey 1981 Semantics. The study of meaning, 2nd edition (Harmondsworth: Penguin). Lehrer, Adrienne 1974 Semantic fields and lexical structure (= North-Holland Linguistic Series 11) (Amsterdam etc.: North-Holland). Liebermann, Felix, ed. 1903-1916; Die Gesetze der Angelsachsen, Vol. II, (Berlin; Aalen: Scientia). 1960 Lipka, Leonhard 1972 Semantic structure and word-formation. Verb-particle constructions in contemporary English (= International Library of General Linguistics 17) (München: Fink).

354 1977

1979

1980

1981

1982

Leonhard Lipka "Lexikalisierung, Idiomatisierung und Hypostasierung als Problem einer synchronischen Wortbildungslehre", Perspektiven der Wortbildungsforschung, edited by Herbert Ε. Brekle & Dieter Kastovsky (Bonn: Bouvier), 155-164. "Semantic components of English nouns and verbs and their justification", Angol Filolögiai Tanulmänyok 12 / Hungarian Studies in English 12: 187-202. "Methodology and representation in the study of lexical fields", Perspektiven der lexikalischen Semantik, edited by Dieter Kastovsky (Bonn: Bouvier), 93-114. "Zur Lexikalisierung im Deutschen und Englischen", Wortbildung, edited by Leonhard Lipka & Hartmut Günther (Darmstadt: Wiss. Buchgesellschaft), 119-132. "Causatives and inchoatives in English and their treatment in recent lexicographic practice", Studia Anglica Posnaniensia 14: 3-16.

Lyons, John 1968 Introduction to theoretical linguistics (Cambridge: Cambridge University Press). 1977 Semantics (Cambridge etc.: Cambridge University Press). Nöth, Winfried 1979 "Contrastive semantics in the light of the theory of semantic change", Anglistik & Englischunterricht 8: 25-39. Pylcs. Thomas 1971 The origins and development of the English language, 2nd edition (New York etc.: Harcourt —Brace —Jovanovich). Skeat, Walter W., ed. 1912 The complete works of Geoffrey Chaucer (London etc.: Oxford University Press). Sprengel, Konrad 1980 "Über semantische Merkmale", Perspektiven der lexikalischen Semantik, edited by Dieter Kastovsky (Bonn: Bouvier), 145-177. Tatlock, John S. P. — Kennedy, Arthur G. 1927 A concordance to the complete works of Geoffrey Chaucer and to the Romaunt of the Rose (Washington: Carnegie Institution). Ullmann, Stephen 1972 Semantics. An introduction to the science of meaning (Oxford: Basil Blackwell). Dictionaries COD 6 1976 The Concise Oxford dictionary of current English, 6th edition, edited by J. B. Sykes (Oxford: Oxford University Press). LDCE 1978 Longman dictionary of contemporary English, edited by Paul Proctor (London: Longman). OED 1933 The Oxford English Dictionary, edited by James A. H. Murray, Henry Bradley, William A. Craigie & Charles T. Onions (Oxford: Oxford University Press).

HELMUT LÜDTKE

Diachronic irreversibility in word-formation and semantics 1. Epistemological remarks on language change The causes of language change (if the word "cause" is appropriate) are no longer unknown: there is a preliminary theory which explains observed, and prognosticates additional, phenomena of language change (cf. Lüdtke 1980 a (ed.)), on a universal basis.

1.1. Universals of language change Contrary to what most linguists have hitherto held to be the case, language (i. e. the normal speaker-hearer's production and perception of speech utterances, together with the psychological processes that give rise to them) is neither a natural nor a cultural phenomenon. This Aristotelian dichotomy is unsuitable for the understanding and describing of a class of events which some late eighteenth century Scotch economists labelled "invisible-hand processes". Such processes result from man's activity (and therefore cannot be accounted for with physical laws alone) — but not from his intent and purpose (for historical details, see v. Hayek 1969: 99-102, Keller 1982: 5-14). Let me give two examples, the first concerning inflation of a currency. Under conditions of market economy, although viewed as undesirable by everybody in the community, inflation may well progress for the simple reason that everybody is pursuing his own economic aims. Take as a second example a column of motorcars moving, say, at 100 km per hour. If a driver, for one reason or other, suddenly slows down to 50 km, the following driver will perhaps slow down to 45 (in order to avoid an accident), the next one to 40 etc. ... until there is a brief halt in the flow of traffic which, of course, none of the drivers involved had intended to provoke (see Keller 1982: 8-9).

356

Helmut

Lüdtke

Language change, too, is an invisible-hand process. Those taking part in it are the community of speaker-hearers, while the relevant parameters are: communication goal, time, and articulatory effort. The rules of the game (i. e. the language system) allow for onesided deviations within the limits of two parameters. There is a phonetological maximum (corresponding to the pronunciation indicated in teaching books) which is a mark not to be overshot but which the normal speaker, in ordinary performance, will fall more or less short of (in the present writer's German, a four-syllable utterance like ich habe es [^'haiba^es] easily shrinks to monosyllabic [gaps])1. The semantactic parameter, on the other hand, consists of minimum structures, e. g., he Xs it, where he may be expanded into a noun, a noun plus adjective(s), plus a relative clause, etc., X into a verb, a verb plus an adverb, etc. Any lexeme may also be replaced by a longer lexeme or a compound one. Owing to thousands and thousands of occurrences, generations of speaker-hearers grow more and more accustomed to shortening and expansion, and less and less familiar with the phonetological maxima and the semantactic minima. The moment that a whole generation of speaker-hearers fails to acquire a given semantactic minimum this option is irretrievably lost. (This happened in France shortly before 1600, i. e., before the language fell prey to grammarians' whims, when children failed to learn the option of using inflected verb forms without subject pronouns, which had been current throughout the 16th century.) The same holds true — the other way round — for any phonetological maximum: once it is lost accidentally, it is lost forever, a slightly shorter pronunciation taking its place. Although not a single accident can be predicted at any given moment, the direction it will take is known beforehand. A speaker can neither randomly add sounds nor drop obligatory semantactic items, if he wishes to reach his communication goal. Therefore, given a statistically relevant number of speech events (which implies a time span of not much less than 1000 years), the stochastic processes of both slur and prolixity will inevitably result in phonetological shrinking and semantactic accretion. And since there is no fixed limit either to slur or to prolixity, both chains of processes are bound to continue ad infinitum. Through accumulation of chance events chance turns into necessity (cf. Monod 1970; see also 2.1.).

Diachronic

irreversibility

in word-formation

and semantics

357

1.2. Accessory language change Beside those ineluctable phenomena described above there are instances of language change brought about mainly in two different ways, viz., — language planning, i. e., all sorts of conscious metalinguistic activity aiming at putting into use or disuse some lexical item, at changing the pronunciation of single words (e. g., "pronouncing formerly mute consonants", cf. Ε often, Fr. sculpteur, etc.), and other instances of purposeful manipulation; — interference due to the speaker/hearers' consciousness of conflicting linguistic norms within their community, and their consequent desire to behave according to the more prestigious norm 2 . These phenomena are accessory in that they are not universal, i. e., bound to happen sooner or later, but contingent upon some special conditions; and they do not pursue an irreversible course. Nor are they, for both these reasons, in any way predictable.

1.3. Change and variation If a language change is understood as an instance of some linguistic item coming into existence that has not been there before or as some process happening that has not previously taken place, then language changes hardly occur. Only a few marginal phenomena like interference or linguistic borrowing may be said to bring about something new within a language system. True internal innovations (e. g., saying lui il vient for the first time in French where formerly only il vient had been usual), on the other hand, start as sporadic occurrences which may or may not give rise to a new pattern. Taken in a narrow sense, language change — as stated in 1.1. — is the opposite of innovation, viz., obsolescence and disappearance, from a linguistic system, of that which was formerly current practice. Language change in a broader sense, i. e., embracing any shift whatever in speech performance taking place along the time axis, must be viewed as the diachronic aspect of language variation.

358

Helmut

Lüdtke

The subsumption of change under variation allows us to behold its phenomena statically, which is analogous to Einstein's static view of motion (cf. Einstein — Infeld 1971: 205-208). According to this concept change is always there, just as variation is. No actuation problem arises since innovations enter casually, in minimal steps, while the seemingly important processes, i. e., collective abandoning of maxima and minima (see 1.1.), are but changes in frequency, from a near zero to zero. A language may be regarded as a system-time-space continuum (analogous to Einstein's four-dimensional time-space, etc., continua — cf. Einstein — Infeld 1971: 204-205), with variation being as universal a property as heterozygosis is in the make-up of human populations.

1.4. Universal laws of language change Variation consists of both quantitative and qualitative processes; only the former are — over long stretches of time — irreversible and hence susceptible to some sort of prediction which may be formulated as follows (cf. Lüdtke 1980 a: 14): lPh(t)

^

EPV)

Σ ι«.»




Stated verbally: statistically speaking, the phonetologic representation of a semantactic item given at a time t will have shrunk by t', while the sum of items needed to convey a given message will increase during the same time span.

2. Diachrony of word-formation 2.1. A geometrical model Taking place in accordance with the universal laws of language change plus their first corollary, viz. merger (or fusion) of syntactically adjacent items (cf. Lüdtke 1980 a: 15), the quantitative aspect of word-formation shift occurring between two distant points of time may be represented by the following model:

Diachronic

item(s)

irreversibility

time arrow •

in word-formation

item(s)

and semantics

359

process SHRINKING ACCRETION MERGER

This set of processes is recursive, insofar as the output of merger is also the input of the other two processes. Functioning and changing at the same time, language is a perpetuum variabile.

2.2. Economy balance From a synchronic point of view, variation — according to the above model — appears as a triple of reversible processes, of which — shrinking takes place along a gradient scale between a given boundary and zero (b je became obligatory, and EGO FECI > je fis was ousted, in popular speech, by EGO HABEO FACTUM j'ai fait /3efe/. As regards economy, the outcome could not have been more balanced: two syllables / four phonemes in each case. So far, we have looked at word-formation processes and their impact on the lexicon followed across time. What if we take the opposite approach? If the triple {shrinking; accretion; merger} incessantly keeps working on every living linguistic system, we cannot but conclude that the lexical stock of any language is made up of items that were either derivatives or compounds in the recent or distant past. So we may search for words that are no longer transparent (cf. Gauger 1971: 7-15, Rettig 1981) but were so in past stages, whether these are attested or have to be reconstructed through comparison. It is not a hopeless endeavour, at least in a certain number of cases. Starting, for example, with Fr. usine, via its latin antecedent officina we arrive at the two roots present in OP(US) and FAC(ERE). In a similar way, a good many German

362

Helmut

Liidtke

words, e. g., Adler, albern, beide, bieder, Demut, Drittel, elend, Efeu ... 3 appear as obscured compounds, bleiben, glauben as derivatives. For English, we may cite barley, barn, bridal, breakfast, garlic, gospel, husband, lady, lord, sheriff, steward, stirrup, woman among many others (cf. Faiß 1978, Fill 1980). In some cases, like Christmas, cupboard, forecastle, forehead, shepherd, etc., only orthography tells us that the words in question are former compounds. This is particularly true for Chinese where quite a few lexical items end in -dz('i) or -r. They would be unanalysable if it were not for the ideograms transcribed ζϊ and er in pinyin, both originally meaning "son, child". Let us imagine that German -t in Gruft, Kluft, Last, Schrift, Tracht, Trift, etc., had a lexical ideogram as its normal graphic rendering: then, of course, everybody could grasp the diachronic fact of derivation. In the absence of any longer tradition of writing, comparison with cognate languages may yield similar results. Thus Maori ra "sail" and uho "umbilical cord" could hardly be suspected of being compounds; and yet, a comparison with Samoan and with the more conservative Bahasa Indonesia:

Indonesian Samoan Maori

"umbilical cord"

"sail"

pusat ("navel") uso uho

lajar lä ra

reveals the fact that both Maori words are indeed obscured compounds, since Indonesian la- and pu-, which recur in several other words, can be assigned particular meanings (cf. Lüdtke 1980 a: 215-216).

2.5. The universal fate of lexical items Since linguistic diachrony consists of infinite processes that take place in accordance with universal laws, we may — after having answered the question of where the lexical stock of any language stems from — ask what will become of single lexical items existing in a given language at a given moment. This answer will be threefold:

Diachronic

irreversibility

in word-formation

and semantics

363

during a time span t -1' the items existing at t will divide into three parts so that (a) some of them remain, (b) some of them get eliminated (become obsolete), (c) some of them merge with syntactically contiguous items. The set of the remaining items (a) further splits into two subsets, viz. such as remain as lexemes and others that grammaticalize (cf. Lüdtke 1980 a: 234^238, Lehmann 1982). In the long run, however, grammatical items, too, are either eliminated from the repertoire or merge with others. Every item is doomed, for — given a sufficiently long time span — it will eventually lose its identity either through elimination or through merger.

3. Semantic diachrony Starting from the premise that an overwhelming number of instances of semantic shift will have to be accounted for without recourse to any general principle other than a classificatory one (cf. Lüdtke 1980 a: 234^238, Wellander 1968), and taking for granted that semantics is far more closely tied to social settings than anything else in language, we could still argue that there might nevertheless be some domain where — short of laws — at least quantitative, and perhaps irreversible, processes can be detected.

3.1. Meaning, frequency, and change Apparently, the frequency with which a lexical item occurs, has no bearing upon its meaning. And yet, there is a relationship — although a more subtle one — between the two parameters. Any change of meaning due to some external event may entail a rise or fall in frequency. And since frequency is statistically linked to (phonetological) size which, in turn, is bound to shrink in the long run — there is some, however remote, possibility that an item may get a somewhat longer respite from its inexorable fate of loss of identity (cf. Lüdtke 1980 b: 273-281; see 2.2.) by acquiring either a different or an additional meaning, if this process goes along with higher frequency of occurrence.

364

Helmut

Lüdtke

A well known class of examples is at hand: words denoting spacial relations tend to acquire temporal meanings (cf. German-English etymological couples like (be) vor / (be)fore; achter / after; nah, nach j nigh; cf. Lehmann 1959: 289, Pottier 1968: ch. XIII), while denotation of temporality tends to widen into cause, manner etc. (cf. weil / while; für / for; Fr. puisque, Sp. antes, It. anzi). Other long-term survival candidates in the field of lexical items have acquired such general meanings as "thing" (cf. CAUSA > chose; G. Ding, E. thing (itself)) from where they may shift even further to "what?" (cf. It. cosal, Welsh betW), or "be" (cf. IE. *bhü-, originally "grow", as in Greek (phü-), "go" (cf. AMBULARE > aller). These semantic shifts are certainly akin to what comes under the heading of "grammaticalization" (cf. Lüdtke 1980 a: 234^238, Lehmann 1982), the tertium comparationis being a noticeable rise in frequency which accounts for longer survival. One might ask whether the opposite will not happen just as easily; but when it does, the combination of drop in frequency and shrinking of phonetological size will hasten the process of obsolescence, so that to the linguist's observant eye these cases seem rarer.

3.2. From clause to item The two main uses speaker-hearers make of language are naming (of things, processes, states, properties, etc.) and statement (of "what is the case"). While a statement ordinarily takes the shape of a clause, naming is performed either by a label (which is the shortest way) or by a description. In this latter respect, languages behave differently; an extreme case is that of Cahuilla, where the speakerhearers resort to description to a far greater extent than we are familiar with from European (and other) languages (cf. Seiler 1975). The question may be raised as to what diachronic relationship there is between describing and labelling. Some examples from European languages, in fact, suggest that many words used today as labels were previously descriptions. ADVOCATUS once meant "(someone who is) called to"; Ε child, G Kind are cognate with Latin genitum "(what is) begot", so that their -d is a product of accretion plus merger; the corresponding Scandinavian word barn (also OE beam, Scotch bairn) originally meant "(what is) born".

Diachronic irreversibility

in word-formation

and semantics

365

Such phenomena are not difficult to explain: since phonetological shrinking is present everywhere in language diachrony, evolution cannot but lead irreversibly from clause towards item, from description towards labelling. It is a question of speech economy to have labels for the most frequently occurring notions. If there are too many of them in the repertoire, the mean phonetological size of labels will exceed the norms set by speech production economy, if homonymy is to be avoided (cf. the numerous instances of synonymic compounding in Modern Chinese). Therefore — statistically speaking — there will be a balance between both procedures of naming, label and description, in the long run.

4. Bio-linguistic implications Since variation in language is biologically given (unlike young bees, children have to learn in order to be able to communicate), the question may be raised whether there is any sense (from the point of view of biological evolution) or — in other words — any survival value attached to variability and learnability. I would venture the hypothesis that lexical renewal — though seemingly only a byproduct of a triple of other processes (see 1.4.-2.3.) — keeps man's mind open for appropriate reactions to changes in the outside world. This seems to have been of paramount importance for a tropical ape who was going to extend his habitat even beyond the arctic circle.

Notes 1. For slur in Standard German, see the detailed description by Kohler (1977: 207-230). 2. These phenomena have been thoroughly studied by Labov (1978). They should, however, not be confused with change within a linguistic system, which is a distinct class of phenomena. 3. See Kluge — Mitzka (1960: 915), s.v. "unkenntlich gewordene Zusammensetzungen".

366

Helmut

Lüdtke

References Chao, Yuen Ren 1968 Language and symbolic systems. (London: Cambridge Univ. Press). Einstein, Albert — Infeld, Leopold 1971 The evolution of physics. Faiß, Klaus 1978 Verdunkelte Compounds im Englischen: ein Beitrag zu Theorie und Praxis der Wortbildung (TBL 10) (Tübingen: Narr). Fill, Alwin 1980 Wortdurchsichtigkeit im Englischen. (Innsbruck). Gauger, Hans-Martin 1971 Durchsichtige Wörter. Zur Theorie der Wortbildung. (Heidelberg: Winter), v. Hayek, F. A. 1969 Freiburger Studien. Gesammelte Aufsätze. (Tübingen). Keller, Rudi 1982 "Zur Theorie sprachlichen Wandels", ZGL 10. 1982: 1-27. Kluge, Friedrich — Mitzka, Walther 1960 Etymologisches Wörterbuch der deutschen Sprache. 18th ed. (Berlin: de Gruyter). Kohler, Klaus 1977 Einführung in die Phonetik des Deutschen (Grundlagen der Germanistik 20) (Berlin: E. Schmidt). Labov, William 1978 Sociolinguistic patterns (Oxford). Lehmann, Christian 1982 Thoughts on grammaticalisation. A programmatic sketch I. (AKUP: Arbeiten des Kölner Universalien-Projekts 48). Leumann, Manu 1959 Kleine Schriften (Zürich/Stuttgart: Artemis). Lüdtke, Helmut (ed.) 1980 a Kommunikationstheoretische Grundlagen des Sprachwandels (Berlin: de Gruyter). 1980 b "The place of morphology in a universal cybernetic theory of language change". In: J. Fisiak (ed.), Historical Morphology (Trends in Linguistics, Studies and Monographs 17) (The Hague: Mouton), 273-281. Monod, J. 1970 Le hasard et la necessite (Paris). Pottier, Bernard 1968 Lingüistica moderna y filologia hispänica (Madrid: Gredos). Rettig, Wolfgang 1981 Sprachliche Motivation. Zeichenrelationen von Lautform and Bedeutung am Beispiel französischer Lexikoneinheiten (Studia Romanica et Lingüistica 12) (Frankfurt/M.: Lang). Seiler, Hansjakob 1975 "Die Prinzipien der deskriptiven und der etikettierenden Benennung". In: H.Seiler (ed.): Linguistic Workshop III. (Structura vol. 9) (München: Fink), 5-57.

WITOLD MANCZAK

Semantic development of borrowings

It is very easy to show that the number of meanings of a word borrowed from one language into another diminishes. E. g., English borrowings in German have fewer meanings in German than in English. Here are some examples from Muret-Sanders 1932 (everything that does not refer to meaning has been omitted): budget 1. Lederbeutel, Ranzen, Sack, Tasche; Schlauch, Lederflasche; to open one's budget sein Herz öffnen, ausschütten. — 2. Vorrat, Bündel; a budget of news ein Sack voll Neuigkeiten (oft Zeitungsname). — 3. Staatshaushalt, Budget, Voranschlag; to open the budget das Budget vorlegen. — 4. Karabiner-Halfter, -Tasche. — 5. s. mumbudget.

Budget budget; das Budget vorlegen to introduce (propose) the budget; aufs Budget bezüglich budgetary; Vorleger des Budgets proposer of the budget, budgeteer.

club 1. Keule, schwerer Knüttel. — 2. Tölpel; roher Bursche (= clown 1). — 3. Ball-stock, -kelle, Schlägel (bei verschiedenen Ballspielen). — 4. Gewehrkolben. — 5. a) keulenförmiger Fühler; b) keulenförmiges Organ (der Keulenpilze). — 6. Klumpen, Knoten. — 7. Haarknoten. — 8. wulstiger Auswuchs an Rüben (durch Insekten verursacht). — 9. clubs

Klub club, association; gemütlicher Klub in einem Wirtshaus fee-and-easy\ fähig oder geneigt einem Klub beizutreten club(b)able; sich zu einem Klub vereinigen to club; Vereinigung zu einem Klub clubbing.

368

Wit old Manczak

Treff, Kreuz, Eicheln (beim Kartenspiel). — 10. a) Vereinigung zu einer Masse; Masse; b) gemeinsame Beiträge (zu einem Essen, Vergnügen); c) Versammlung im Gasthause zur Unterhaltung; d) geheime (politische) Versammlung, Clique. — 11. a) Klub, Verein, geschlossene Gesellschaft, Kasino, Ressource; The (Literary ) Club Name einer von Dr. Johnson 1763 gegründeten Gesellschaft; b) = benefit club (s. benefit 1); to be on the Club aus der Vereinskasse unterstützt werden. — 12. = club-house. — 13. Art Bootsspiere. — 14. clubs s. clump 3. dog 1. Hund; Prairiehund. — 2. Männchen (einiger Tiere, ζ. B. dog-fox). — 3. Hund; Hunde-; dog horse Schindmähre. — 4. Mensch, Kerl, Bursche, Schlingel. — 5. = bulldog 2.-6. = dogfish. — 7. Dog a) (großer bzw. kleiner) Hund (s. canis II); b) Hundsstern, Sirius. — 8. Name verschiedener Werkzeuge: a) Feuerbock; b) = doghead; c) Klammer, Klammerhaken; d) Klemmschraube; e) Sperrhaken; f) Teufelsklaue, Schlittenständer; g) Schlosserei: Dorn; h) Drahtzieherei: Ziehzange; i) Brechstange. — 9. Fischerei: Art Boje aus Tierblase. — 10. Laufkarren, Hund. — 11. dogs Aktien der Newfoundland Land Co.

Dogge dog, bulldog; englische Dogge mastiff; dänische {deutsche, Ulmer, dalmatische) Dogge great Dane, Danish dog.

Semantic development of borrowings

Sport

369

sport 12. Belustigung, Kurzweil, Zeitvertreib; Spiel, Scherz, Spaß; in sport, for sport, for sport's sake zum Spaß, zum Zeitvertreib; sport of nature Naturspiel; sport of words Wortgeklingel, Spiel mit Worten; that is (a) sport to him das macht er spielend, das ist Spaß für ihn; to be in sport scherzen; to give sport belustigen; we had a capital sport wir haben uns königlich amüsiert; wir haben eine vortreffliche Jagd gehabt; to make a sport scherzen, spielen, sich belustigen, Spaß treiben; to make (a) sport with (oder of, at) a person jemanden zum besten haben, sich über jemanden lustig machen, seinen Scherz mit jemandem treiben; to spoil a person 's sport jemandem den Spaß verderben, jemandem einen Strich durch die Rechnung machen. — 13. Sport, Vergnügung im Freien (wie Jagd, Wettrennen, Boxen). — 14. Wettkampf. — 15. Spott, Hohn (at, with über); Zielscheibe des Spottes. — 16. Mutwille; Liebeslust. — 17. a) Sportliebhaber; b) Spieler, Wetter, Gauner. - 18. Spielart. — 19. to show sport sich kampflustig zeigen.

Sport sport; sporting.

treibend

strike 1. Streichen, Streich. — 2. Streichen (der Schichten). — 3. Streich-holz, -maß (für Gemäße); to sell by the strike nach gestrichenem Maß ver-

Streik strike, stoppage (or suspension) of work, turn-out.

370

Witold Manczak

kaufen. — 4. Streichholz (der Töpfer); Abstreichlineal (der Gießer). — 5. Rühr-stange, -krücke; Stoßstange, Steinmeißel (beim Gießen). — 6. Schippe (der Bäcker). — 7. Trockenmaß = 2 bushels (schwankt aber von 1/2 bis 4 bushels). — 8. Ablaßquantum (Sirup); auf einmal gekochtes Zuckerquantum. — 9. Fehlschlag; Ball, den der Schläger hätte schlagen sollen. — 10. Kegelspiel: Umwerfen aller Kegel; to make a strike alle neun (bzw. zehn) werfen. — 11. Streik, Strike, Arbeitseinstellung, Ausstand; to be on strike streiken, im Streik stehen, die Arbeit eingestellt haben. — 12. gewissenloser Versuch, um (auf dem Wege der Abfindung, wofür dann gewisse politische Schritte unterlassen werden) Geld oder Vorteile zu erpressen, politische Erpressung. — 13. volles oder gestrichenes Maß; Brauerei: gestrichenes Maß Malz; ale of the first strike Bier der besten Sorte oder Qualität. — 14. Gesamtheit der auf einmal geschlagenen Münzen, Prägungsbetrag. — 15. Schriftgießer: nicht justierte Matrize. — 16. Schließ-blech, -platte (an der Tür). — 17. kristallinisches Aussehen (harter Seife); richtige Marmorierung. — 18. Anzahl von fünfundzwanzig Aalen. - 19. Stoß ins Horn. 20. Pfund Sterling. - 21. eiser-

Semantic development

of borrowings

371

ner Stab (an Palisaden). — 22. strike of day Tagesanbruch. — 23. strike of flax Handvoll Flachs (die man auf einmal hecheln kann).

I do not know whether anybody has ever tried to account for the fact that as a rule borrowings have fewer meanings in the new language than in the old one. The purpose of the present paper is to explain this fact by a law which I formulated in an article published in 1966. The law in question is very simple and reads as follows: the more frequently used linguistic elements are more differentiated than those which are used less often. By linguistic elements I mean letters, phonemes, morphemes, words, and groups of words. Here are some examples illustrating this law and referring to spelling, phonetics, word formation, inflection, syntax, and vocabulary. Spelling: The small letters are more often used than the capital letters and, if there is a difference between small letters and capital letters, the former are more differentiated than the latter. E. g., in Greek, there are two signs for the small sigma (σ, ς) but only one for the capital sigma (Σ). In German, there is a difference between ss and ß, e. g., in Gasse and Straße, but there is no difference while using the capital letters: GASSE and STRASSE. In different European languages, there are two ways of writing the small r, while there is only one possibility of writing the capital R. In French, the capital letters can be used without diacritic signs, so that Ε may correspond to five small letters: e, έ, e, e, e. The Arabic figures are used more often than the Roman, and this accounts for the fact that in Spanish there is a difference between 5 = cinco and 5.° = quinto, while V means both 'cinco' and 'quinto'. Phonetics: The voiceless consonants are more often used than the voiced and if the number of voiced consonants is not equal to that of the voiceless, the number of the latter is as a rule larger. In German there is a voiceless ζ (e. g., in zu), but a corresponding voiced consonant is lacking. In Indo-European languages, dentals are more often used than labials or velars, and this explains why Span, d has three allophones (as in dar, piedra, and cantado), whereas b and g have only two variants (as in boca, haber and grande, rogar). Oral vowels are more frequently used than nasal

372

Witold

Manczak

vowels, and therefore oral vowels are more differentiated; in other words, they are more numerous than nasal vowels (e. g., in Polish, French, or Portuguese). Word-formation: There are more derivatives and compound words from a frequently used word like mother than from a rarely used word like olive. In many languages, quite a different word serves to form the feminine in the case of the most frequently used words, e. g., brother — sister, a suffix is used with less frequently used nouns, e. g., master — mistress, while the feminine is not formed at all from words which are seldom used, such as witness. The same applies to animal names. In the case of the most frequently used animal names, there are two different words, e. g., bull and cow. With less frequently used words, the difference is less, cf. tiger — tigress. Finally, the feminine is not formed at all from words which are seldom used, such as hyena. Inflection: In Latin, from the point of view of frequency of use, the indicative takes the first place, the subjunctive the second, and the imperative the last; that is why the indicative has six tenses, the subjunctive four, and the imperative only two. The present is more often used than other tenses, and this explains why in the present of the verb write there are two forms: write, writes, but only one in the preterite: wrote. From the point of view of frequency of use the order of the numbers is as follows: singular, plural, dual. This explains why in Old Church Slavic the singular has seven different case forms, the plural only six (there is no special form for the vocative), while the dual shows only three forms: one for the nom.acc.-voc., one for the gen.-loc. and one for the dat.-instr. The passive voice is less frequently used than the active voice, and this explains why in Latin and Gothic the synthetic forms of the passive are less numerous than those of the active. It is important to point out that suppletion is a particular case of the general law concerning the connection between differentiation and frequency of use of linguistic elements. This allows us to understand the noteworthy fact that in different languages suppletion occurs in words of more or less the same meaning. It is mostly a question of the following categories of words: the the the the

pronoun of the first person: I, me, we, us, Fr. je, me, nous·, pronoun of the second person: thou, you, Fr. tu, vous; pronoun of the third person: he, they, It. egli, essi; word meaning 'every, all': Germ, jeder, alle, Fr. chaque, tous;

Semantic development

of borrowings

373

the noun meaning 'man, people': Germ. Mann, Leute, Fr. homme, gens; the adjective meaning 'good, better': Germ, gut, besser, Fr. bon, meilleur; the adjective meaning: 'bad, worse': Fr. mauvais, pire, Russ. plochoj, chuze; the adjective meaning: 'large, larger': It. grande, maggiore, Russ. velikij, bol'sij; the adjective meaning 'small, smaller': Fr. petit, moindre, Pol. maly, mniejszy; the adverb meaning 'much, more': Germ, viel, mehr, Fr. beaucoup, plus; the adverb meaning 'little, less': Germ, wenig, minder, Fr. peu, moins; the numeral meaning 'one, first': Germ, eins, erster, Fr. un, premier, the numeral meaning 'two, second': Fr. deux, second, Russ. dva, vtoroj; the verb meaning 'be': Germ, bin, ist, war, Fr. est,fut, ete\ the verb meaning 'go': Fr. va, aller, ira, Pol. isc, szedl. That suppletion occurs in different languages in more or less parallel ways is very easy to account for: the suppletion occurs only in the most frequently used words and these are everywhere practically the same. If somebody doubted that there were a connection between suppletion and frequency of use, other facts could be cited in explanation. The inflection of the French verb aller shows four roots: va, allait, ira, and fut (used sometimes in the meaning of 'alia'), but the only tense where two roots occur is the present indicative (vais, vas, va, allons, etc.), which is the most frequently used tense. The same applies to the German verb sein, cf. on the one hand bin, bist, ist, etc., and on the other hand all other tenses, which are less often used and show no suppletion. The same applies to the English verb be, cf. on the one hand the pres. ind. am, are, etc., and on the other hand the remaining tenses. It happens in different languages that pronouns show suppletion in the singular and no suppletion in other, less frequently used numbers, cf. Fr. je, me (sing.) but only nous (plur.). In all Indo-European languages, the verb 'be', which practically always is suppletive, is the most frequently used verb. In the Indo-European languages, the suppletion occurs in the numerals meaning 'one, first' always, in

374

Wit old

Manczak

the numerals meaning 'two, second' sometimes, and in the higher numerals never, and one knows that 'one' is used more frequently than 'two', 'two' more frequently than 'three', and so on (see Mariczak 1970, 140-179). It is necessary to insist on the fact that suppletion is not a heritage from Proto-Indo-European, but comes into being in all periods of the history of languages. Suffice to say that suppletion occurs in Romance languages in some words that had a regular inflection in Latin. There was no suppletion in the Latin pronoun nos, but in Italian one says noi, ci 'we, us', and ci < Lat. ecce hie. There was no suppletion in the Latin verbs vadere, ire, or ambulare, while the inflection of Fr. aller, Span, ir, or It. andare is suppletive. The Latin adjective magnus, major, maximus did not show suppletion, whereas Italian replaced magnus, major with grande, maggiore. Syntax: The personal pronouns, which are more often used than the substantives, can be combined with verbal forms in all three persons, while the nouns can be combined only with the forms of the third person. The indicative, which is more often used than the imperative, can be combined both with a nominative and an accusative, while the imperative cannot be used with a nominative: he helps her, but help her. The Latin preposition in can be combined with two cases, while the less frequently used preposition inter can be combined with only one case. The definite article is more frequently used than the indefinite one, and that is why the use of the former is more differentiated than that of the latter. Therefore, in any syntax of English, the chapter on the definite article is longer than that on the indefinite one. W. C. Ritchie mentions that one can say both John sent the package to Mary and John sent Mary the package, but only John missent the package to Mary (*John missent Mary the package being impossible). That the syntactic use of send is more differentiated than that of missend is to be accounted for by the fact that send is more frequently used than missend. Vocabulary: A noun like hand has many more meanings than a rarely used substantive like thorn. In other words, the meaning of a frequently used word is more differentiated than that of a word which is seldom used. Having illustrated the law concerning the connection between differentiation and frequency of linguistic elements with so many examples, it is easy to answer the question why borrowings have fewer meanings in the new language than in the old one: simply because they mostly are less frequently used in the new language

Semantic development of borrowings

375

than in the old one. The frequency dictionary by Eaton (1961) contains, among others, 6000 English and 6000 German words. According to Eaton, the places of budget, club, etc., and their German equivalents on frequency lists are as follows: budget club dog sport strike

6000-7000 1500-2000 500-1000 1000-1500 500-1000

Budget Klub Dogge Sport Streik

absent 5500-6000 absent 3500-4000 3500-4000

It follows from these data that Germ. Budget, Klub, etc., are less frequently used than Engl, budget, club, etc. Finally, it is important to stress the fact that the law concerning the connection between differentiation and frequency of linguistic elements throws a new light on many problems, e. g., it shows that the theory of natural morphology is false (see Mariczak 1982).

References Eaton, Helen 1961

An English-French-German-Spanish word frequency dictionary (New York: Dover Publications), new ed. Mariczak, Witold 1966 "La nature du suppletivisme", Linguistics 28: 82-89. 1970 Ζ zagadnien jgzykoznawstwa ogölnego (Wroclaw: Ossolineum). 1982 Review of Morphologische Natürlichkeit by Mayerthaler, Studies in Language 6: 146-152. Muret, Eduard — Sanders, Daniel 1932 Enzyklopädisches englisch-deutsches and deutsch-englisches Wörterbuch (5th ed.) (Berlin-Schöneberg: Langenscheidt).

FRANCISCO MARCOS-MARIN

Etymology and semantics. Theoretical considerations apropos of an analysis of the etymological problem of Spanish MANERO, MANERIA

The origin of the study of Diachronic Semantics in Spanish can be implicitly found in the pages of Origenes del Espanol, where Menendez Pidal (1956) takes up Lexicography and, particularly, Lexical Geography. In these pages, the neo-grammarian concern with isolated words which characterized the first part of the volume gives way to the discussion of related words. In § 82, which deals with Mozarabic "dictionaries", quite a few interrelated lexical items are studied as well as some complete conceptual fields, such as the names of the horse in the Glossary at the Leyden Library (Seybold: 1900). In § 85 5 on derivatives of prunum, nisum and cereola, not only do we find geographic interpretation, but also several semantic and lexicological nuances are identified by including, e. g., some derivatives. And later, § 84 bis not only explains the geographic distribution of names assigned to the mustela in Hispania and other Romanic areas (cf. et. Schot: 1935), but also studies the causes for changes in the denomination of this animal. Finally, § 85 is, from the start, a clear example of onomasiological semantics: "Topographical names which designate an isolated rise of land of less height than the mountain". In all of these cases, we are speaking of the onomasiological and etymological semantics of words. But, as has been indicated above, there are also interesting aspects for the study of derivational procedures or for the formation of locutions. This concept of Semantics has more to do with pragmatic implications — including literary considerations (Marcos —Marin: 1979) — than with those of either Morphology or Syntax. Therefore, it constitutes but one aspect of many existing possibilities. (For the diachronic interrelationship of Syntax, Semantics and Pragmatics, see Lapesa: 1970).

378

Francisco

Marcos-Marin

Given these preliminary observations and the inherent limitations involved, the time has come to present the following contribution. Here it is attempted to develop an etymological semantic study which is not only onomasiological, but also partially semasiological since the evolutionary data will indicate the appearance and development of certain characteristics of variation in the history of the terms. We shall attempt to put some order into the scattered notes on the etymology of mafieria and manero, deal with the chronological question and, after a critical review of the possibilities and rejection of the impossibilities, we shall suggest which of those possibilities still available to us seems to be the most reasonable. Mafieria and manero have come down to Modern Spanish as medieval judicial terms (Garcia Gonzalez: 1955; Gonzalez Olle: 1976, 9.65), referring to a legal tribute or obligation and to the person obligated by the tribute. The tribute of mafieria is the judicial form substituting the obligation of a colonist manero to restitute his lands to his lord at the manero's death. The term manero is not only applied to a colonist subjected to the tribute of mafieria which allowed those lands to be bequeathed to other heirs in the case there were no legitimate sons; it also refers to a person who, at his death, left no legitimate succession, and whose state reverted to the sovereign, that is, the king in the case of the nobility, or the lord in the case of the colonist peasants. We insist on the concepts of 'legitimacy' or 'legitimate' sons: the word manero, applied to persons, does not necessarily mean 'sterile' as it does when applied to animals, even though this meaning does seem to lie at the origin of its semantic evolution. This semantic subtlety can be confirmed by the examination of the history of the word itself in Spanish. (The Latin and pre-Latin forms will be taken up later.) The earliest Castilian example to be quoted corresponds to the form manero and is found in a document of a. 1030 of Clunia published by Menendez Pidal (1956, p. 36): DeTorre deGisando, domno Gisando et don Kintla et don Gutierre et don Monnio fuerunt bassallos de illo comite Garcia Fernandiz, et mattarunt illos mauros jnZeruera, et fuerunt manneros: pro jnde jntrarunt earunt [sic] diuisas jncometato, et mandarunt illas suos merinos qui Clunia tenuerunt; et post obitum dedomno Sanciocomite, partirunt illas jnfanciones.

Etymology

and semantics:

apropos Spanish manero, maneria

379

A possible antecedent of maneria may be the form mannaria as documented in the Fuero de Mendavia of 1274 (Lacarra: 1934): Et que su merino nin su sayon, non les tome ninguna cosa contra su uoluntad nin ayan sobressi ningun fuero malo de sayonia ni de nuda, nin de mannaria ... In Castilian, the first meaning of the lexeme which corresponds to these lexical items is that of'sterile' or 'sterility'. With the passage of time, this 'sterility' becomes relative and one may speak of 'legal sterility' or 'judicial sterility'. Since, in ancient times it was believed that the woman was at fault when a couple did not have children, we should not be surprised that the feminine form, manera, is more frequent than the masculine manero. However, because of the greater influence of the man, the judicial term is the masculine for the individual. In the Alphonsine works of the thirteenth century, the meaning of the word is 'sterile' and refers, for example, to animals ('sterile female') in the General Estoria (1930: 13,a,12; 569,a,9). Here it is also used to refer to land (80,a,9; 120,a,13) and to the water in areas in which nothing grows nor can be grown (133,b,49 1 ). It seems to be clear in several passages of Castigos e documentos del rey don Sancho (1292-93, copy after 1351) that the word refers to sterility and not impotence 2 . Cap. LXXVI: Otrosi, si en el comentimiento hobiere tal condicion que procuremos maneria e non hayamos fijos, ... alii non hay casamiento ninguno. Cap. LXXVII: Est a manera pone Valerio Maximo en el segundo libro de los Reys e fechos maravillosos, do dize que despues que fue fecha la cibdat de Roma fasta ciento e cincuenta anos, nunca hobo y home que repoyase [repudiated] a su muger, e el primero que la repoyo por amor que era manera [because she was 'manera'], fue Carbino Fornisino, el cualfue movido por razon de su esterilidat; empero fue mucho reprehendido por ello. (...) La segunda razon se toma de parte de los fijos; ca maguera el casamiento sea manero, e non se deba departir por la maneria, segund dicho es, empero, si hobiere fijo ο fijos [if there were sons], es bien comun del home e de la muger. Cap. LXXIX: Ε si una muger fuese casada con muchos varones, todos es tos bienes cuatro se embargarian. (...) Lo tercero, que se embargaria la generacion de los fijos; ca fazerse-y-an las mugeres

380

Francisco

Marcos-Marin

casadas maneras, onde las malas mugeres que se dan a muchos homes por esta razon son maneras, segund que dize el filosofo e lo probamos de fecho. Onde porque las mugeres casadas non pier dan la generacion de los fijos, nin sean maneras, nunca deben ser ayuntadas en uno con muchos maridos3. It is therefore evident that the condition of 'mafiera' may be acquired by a woman of her own will (contraception), and has nothing to do with 'impotence'. At the very beginning of La estoria del rey Anemur e de Josaphat e de Barlaarrf, it is clearly shown that manero or the derivative manereza j maheresa refers to sterility and never to impotence. Manuscriptes Ρ and G tell us of king Avenir ( = Anemur) who was surrounded by great comforts and pleasures. He lacked only one thing for his happiness to be complete, "he was manero, because he couldn't have children" (Keller — Linker 1979: 5; we quote from Ρ although there are some minor variants which are not relevant for our present purposes): e aquella [that thing] atormentava la su anima de fiera guysa de muchos pensamientos: e esto era porque era manero, ca non podia aver ningund fijo. Ε por ende vyvia en muy grand cuydado [worry] e trabajavase mucho de commo podiese ser suelto de tal enlazamiento, e fuese llamado padre de fijos, que es cosa de muchos codiciada. In text S, somewhat briefer, the king is called Anemur and we can read (Keller - Linker 1979: 365): Ε commo [although] visquiese en muchos deleytes, avia un mal de manereza, el qual menguava la su gloria e atormentava el su coragon, ca non podia aver fijos. Those who are familiar with the story know that king Avenir or Anemur will have a son somewhat later {P and G, on page 20, S on page 367), called Josafat, who will protagonize the subsequent story. There is no doubt that one cannot speak of sterility in this case, but rather of dilatoriness since the king's paternity is unquestioned in the work. Therefore, the condition of manero can refer to temporary accident. Some nuances will be added to this concept later, in the Spanish world. Thus as concerns the second meaning, 'judicial sterility' I daresay, a manero for the Sephardim is any man who

Etymology

and semantics: apropos Spanish maiiero, maneria

381

dies without succession or with no direct heirs, even though he may have had children 5 . At the beginning of the sixteenth century in the Farsa del Sacramento de Moselina, a manera is a "woman who has had two children but who, because of her age, is no longer apt for generation". (Β. AA. EE. LVIII, l i b , cited in RFE, XXV, 247). Here it is obviously a matter of secondary sterility and not of sterility itself nor of impotence 6 . With these observations behind us, let us now examine the formal problems of the word's evolution. Both maiiero and maneria appear as derivatives, in -ero and -eria, respectively, of a radical which — if we recall the form mannaria registered above — may have been MANN — us. This form is quite well documented, with its derivative mannulus as recorded in the Thesaurus1: mannus, -i m. [vocab. illyricum, cf. messap. (Iupiter) Menzana. alban. mes, mas 'pullus equinus vel asininus', gall. Epo-manduodurum. J.B.H.] de orig.: CONSENT, gramm. V 364, 9 Gallorum, -i. SCHOL. Hor. carm. 3, 27, 7 (add. V2) -i ideo dicti, quod mansuetudine familiarius manum sequantur. de signif.: v. vol. II 2250, 72 sqq. s. v. buricus (PORPH., SCHOL. HOR., ISID., GLOSS.). GLOSS, equus parvus. i. q. equus parvulus (cf. Schuster, RE XIV 1228 sqq.): LVCR. 3, 1063 currit agens -os ad villam praecipiter. HOR. epod. 4, 14 Appiam -is terit. carm. 3, 27, 7 serpens terruit -os. epist. 1, 7, 77 impositus -is. PROP. 4, 8, 15 detonsis avecta est Cynthia -is (Beroald., ab annis codd.). OV. am. 2, 16, 49 rapientibus esseda -is. SEN. epist. 87, 11 v. vol. II982, 70. SER. SAMM. 804 fimus -i. AVSON. 397, 7 vel celerem -um vel ruptum terga veraedum conscendas. HIER, epist. 66, 8 ferventes -os. deriv.: mannulus. It should be added here that Nebrija (1492) records mannus, -i as a 'dwarf horse' as well as mannulus -i which the grammarian explains with 'por aquello mismo' (for that very reason), i. e., because of the smallness. Above, we promised to come back to the term mannus as documented in the Fuero de Mendavia (1274) in our etymological arguments. In fact, one finds this passage in a Lusitanian inscription from Mertola, dated March 17 of era 532 (Hispanic), i. e., the year 494 AD. This tombstone is reproduced by Hübner (1891, rep. 1975). It may be useful to us later to caution that Maria Lourdes Albertos (1966) does not document Mannaria as a proper noun because she

382

Francisco

Marcos-Marin

only deals with Tarraconensis and Baetica. Because it is Latin and of a rather late period, the term also cannot appear in Palomar Lapesa (1957), although this fact might as well indicate that, if the word is to be considered an anthroponym, it would not be preLatin. Let us see the text: 309 (...) Mannaria famula Chr[ist]i vixit annos novem menses quattuor, requievit in pace d[ie] XVI Aja/[endas] Apriles era DXXXII. First, one might ask whether this Mannaria is an adjective or a proper noun. Were it a proper noun, it would be the only example known. This fact, of course, would support those who consider it an adjective. However, this is not an unsurmountable obstacle, given the scarcity and dispersion of anthroponymic remains. Syntactically, as an adjective, mannaria is placed before the noun it modifies, according to the order in Classical Latin, where the adjective normally precedes the noun unless: the adjective is a monosyllabic noun modifier; it is derived from a proper noun, or; it forms part of a set expression. On the other hand, it is common knowledge that both the Satiricon and the Peregrinatio offer many cases of the noun-adjective order which are not explained by significant or expressive intensification. Thus, the above arguments are weakened by the end of the fifth century. We cannot even refer to the conservative nature of funereal inscriptions because Mannaria famula appears only once in an inscription and it is precisely here. Since it does not seem likely, considering the material characteristics of this tablet, that another anthroponym could come first, the fact that the inscription is headed by the deceased person's name would seem to corroborate the hypothesis of mannaria as a proper noun. Nothing conclusive can be obtained from either possibility. There can be no doubt that there is an example of the word-form mannaria in the fifth century and that the radical mann- is by no means unknown to us. This fact invalidates those hypotheses which would propose a later etymon such as that of an Arabic loan word, rejected elsewhere (Marcos — Marin, 1982). The words manero, maneria and other cited above not only appear in the world of Romance lexic, but are also evidence of a primitive lexeme of wide-spread use: Portuguese maninha 'sterile woman', Galician maniha 'a cow which is not sterile but does not always conceive', Castilian manera (to follow some sort of geographical order), Gascon mano 'sterile cow' and mane 'sterile woman', Be-

Etymology

and semantics:

apropos Spanish manero, maneria

383

arnese mane 'sterile ewe or animal', Bearnese, Aranese mänou, -o, Catalan mana 'sterile cow', the meridional Galloromanic dialect of Aveyron, mono, in Cantal, 'sterile cow', in Nice mana 'sterile animal', Occitan magneiro 'an ewe whose lamb has died and receives another to nurse in its place' which in the valley of the Rhone is maniero and in Nice maniera, Italian manzo 'steer' translated as 'buey' in Corominas' DECH (the Ochse of Meyer-Lübke's REW 5289). Sardinian manzu 'steer', Corsican mandzonu 'nickname for the steer', Engadinish manz 'young bull' 8 , Rumanian minz 'colt, donkey', fem. minza, minzati 'a cow without offspring', minzare 'milking ewe', minzoc 'young horse'. To the North we have the Wallon term mos 'sterile cow' 9 . With these data in mind, we can establish a tentative geographic distribution of the term and draw the borders of its semantic equivalences. In the Romance languages, the word encompasses the Mediterranean area to the West of meridian 26° E, it spills over the 9° W meridian and penetrates North to parallel 49°. Semantically, its meanings are 'young animal', 'sterile animal', ovine, bovine or equine. Within these geographic borders, some non-Romance languages, mainly two Indo-European and one pre-Indo-European, offer examples of our lexeme. The Indo-European languages are Albanian and Germanic and the pre-Indo-European language is Basque. In Albanian we have mes, mezi 'colt, young ass', or 'young mule', mezat 'young bull' for Cioranescu (s. v., minz deriving from Rumanian minzat), in High Middle German menz, in Tirolese manz 'sterile'. This evidence fills in the geographic area we have defined above, with two exceptions, and we have extended the semantic fan to include the sterile animal par excellence, the mule. The exceptions can be found in Slavic languages and within the Hellenic linguistic space. At this point, we are ready to talk about chronology. In the Hellenic world, Kretschmer (Glotta, XVI, 1928: 182, quoted by Tovar, 1949: 157), documents dialectal mod. Gr. Cretan μα(ν)ζέττα 'cow'. In clas. Gr. we have μαζός, Doric μασδός, 'breast', 'teat' 'mamma' and the adjective μαζούσιος, 'in the form of a breast', 'protruding like a teat', μαστός or μασθός, 'teat', 'mamma', 'breast'. Tovar (1949: 158) documents some of these forms and relates them to Albanian ment, 'she nurses', mendese 'wet-nurse', and to the Indo-European roots *mand- and *mnd-. The only information I have concerning the Slavic languages comes from Cioranescu and refers to Bulgarian mandzara, which comes

384

Francisco

Marcos-Marin

from Rumanian minzare. But Albanian and Greek are not alone among the Indo-European languages, as we shall see. Nevertheless, first we must fulfill our promise to present our chronological discussion for which we need only agree on a few initial, but incomplete, conclusions: 1. The widespread incidence of the lexeme within all the Romance languages would clearly indicate that the antecedent form of manero and maneria, that is, the basic form MANN-, already existed in Latin. 2. However, due to the late date of documentation, one might think it to be a regional or rustic form or that it came into Latin from another language. 3. In spite of this manifestation in Latin, the lexical base is documented in all the Indo-European languages which were slowly implanted in the Northern Mediterranean basin towards the tenth century BC. 4. The form mand- in Basque would suggest that this language did not receive the term from the source documented in Latin. The word came into Basque through or from another language. 5. Some important semantic contrasts have been noted: on the one hand, there is the meaning 'teat', 'mamma' and, on the other, 'young animal' and 'sterile or sterilized animal'. This would lead us to think that there already existed in the Indo-European languages a lexical base (or radical, if one prefers) MA(N)D~, a thousand years before our era. Let us mention now the Indo-European people who invaded the Northern Mediterranean basin at the beginning of the last millenium BC., and dispersed over the whole area: the Celts. This IndoEuropean people poses many problems; we have very little knowledge of the Hispanic Celtic languages or of Gallic, although we do know more about several other Celtic languages, specially the insular ones such as Welsh, Gaelic, Irish and Breton (insular Celt of late emigration to the continent). However, there are several scholars, such as Pokorny, Tovar and, of course, Menendez Pidal (1943), who accept the existence of pre-Celts or para-Celts called Illyrians, Ligurians, Illyrian-Ligurians, Ambrons or Ambroillyrians. Menendez Pidal (1943, 1968: 174-5) synthesizes one of the most apparently obvious aspects of these peoples: the Ambrons, who gave the name to the island of Amrun in Southeast Denmark, were amber merchants in the North Sea. 'Amber', in Greek, was called precisely λιγύριον i.e., 'ligurian'. (Pauly —Wissowa 3.300. Here-

Etymology and semantics: apropos Spanish manero, mafieria

385

after we quote from the second edition of Adolfo Schulten's, 1955, see also Krahe, 1955, I: 103). Tovar (1977) points out that the reference to amber seems to indicate an etymology a posteriori and he comes to the conclusion that it is an interpolation. Nevertheless, he does not attack "en su conjunto la construction levantada por Schulten sobre ideas de G. Schöning y K. Mullenhoff'. The Celts gradually pushed this group to the South towards Liguria, according to Rufus Festus Avienus (end of the fourth century AD.), a later author in whose Massalliot periplus from ca. 520 BC. in book I of his Ora Maritima we can find in lines 129-136 the following: ... siquis dehinc ab insulis Oestrymnicis lembum audeat urgere in undas, axe qua Lycaonis rigescit aethra, caespitem Ligur[g]um subit cassum incolarum. namque Celtarum manu crebrisque dudum proeliis vacua arva sunt Liguresque pulsi, ut saepe fors aliquos agit, uenere in ista ... Schulten (1955: 97) confirms that the Ligurian formerly lived on the coast of Frisia. He based this theory on several factors: the fable of the Frisian king Cicno; the name of amber, as we have just mentioned, and the common origin of the Ligurians and Ambrons according to Plutarch ( L i f e of Marius, 19). Marius took Ligurians from Italy in his army which fought against Teutons (Germanic), Cimbrians (Celts) and "Ambrons" in the battle of Aquae Sextiae. Marius' Ligurians were surprised to learn that the Ambrons shouted "Ambrons!" to identify themselves in battle, just as the Ligurians did themselves because, as Menendez Pidal translates, "The Ligurians call themselves and their race Ambrons". Of course, it is not intended here to review D'Arbois de Jubainville's hypothesis of a cohesive Ligurian civilization, nor do I intend to pass judgement on pre-Celtic problems. For the purposes of this paper it is to be understood that, according to those who have traditionally postulated two invasions, the Celts push, and carry with them, these people even less known than themselves, and it is precisely these groups which have been related to our etymology of mafieria and manero. Ligurians and Illyrians would have occupied the Alpine area now contained in Switzerland, Austria and Italy and probably spread throughout the Northern Mediterranean basin

386

Francisco

Marcos-Marin

before or at the same time as the Celts. Therefore they should not be ignored in our analysis. On quoting the Thesaurus above, it was noticed that its writers assign an Illyric etymology to Latin mannus. This thesis is supported both by Alpine forms and Rumanian as much as by the cult of Jupiter Menzana, to whom "the Messapians made offerings of a horse" in Southern Italy. (REW; Tovar, 1949: 157-8, remembers that this cult is known through a text of Paulus Festus.) The Celtic hypothesis is supported by the testimony of the grammarian Consentius (fifth century AD.) partially reproduced by Tovar, who based his hypothesis on it {Gramm. Lat., V, ed. H. Keil, Leipzig: 364; Tovar, 1949: 157): Omne peregrirtum nomen, siquidem id iam receptum est, ut Gallorum manni, Medorum acinacis vel gaza, Poenorum tubur, dirigi ad eius latini sermonis similitudinem debet, cui proximus est. Accepting, after Tovar, the authenticity of this testimony, some aspects should be differentiated. Scholars do not neccessarily agree on the genuiness of the paragraph quoted. It seems convenient to say that Tovar was primarily interested in the etymology of the name of the famous Illergetean chieftain Mandonio (from Ilerda, Cast. Lerida, Cat. Lleida, Ν. E. of Spain). The form Mando- can be seen as the root of that noun, joined to a well-known Celtic suffix, -onios, which Illyrian also attests. This suffix is to be related to the Urnenfelder culture, the first wave of the Celts, ca. 900 BC., which soon merged with the Iberians. It is an acceptable etymology, believed to be "totally demostrated" by Maria Lourdes Albertos (1966: 146). Nevertheless, it occupies a marginal place in this discussion, which will not enter into its specific difficulties. It is fair to recognise, however, the basic help provided by Tovar's studies in the development of this part of the present paper. Thus, I must dispense with the doubtful forms some authors seem to reduce mainly to *mendo, such as Mid. Ir. menn, Welsh mynn, Cornish min, Bret, men 'kid goat'. On the other hand, Maria Lourdes Albertos cites many anthroponyms and toponyms with the base mando-, such as: Mandalonius (CIL XIII4130) in Belgium; Mandatus in the Illyrian region, Gaul and Germania; Mandelana and Mandilo (CIL V 6803, 5001) in Cisalpine; Mando (from which Mandonius is directly derived), a type of pottery in Gaul; Mandorinus (CIL IX 1404) in Italy, etc. In composition, see in Hispania Melmandi and its variants

Etymology and semantics: apropos Spanish manero, mafieria

387

and Lica-manda. (...) However, Mandia (CIL XIII 7893) in Low Germania seems to come from the base *mandio. In toponyms, see Illyrian μανδύριον, Manduria', Gallic Epomanduo-durum, etc. Evans (1967: 197-199; 222-3) has discussed the etymology of the compound name Epomanduodurum. Referring to the second element, mandu-, he writes: "Names in mand(u)- ... are doubtless of multiple origin. A number of them are probably to be related to the form mannus 'a pony' or 'a little draught horse' first attested in Latin by Lucretius 3. 1063 (dimin. mannulus Plin. ep. 4.2.3) and claimed as Celtic by Consentius". He continues (1967: 223): "However, d'Arbois de Jubainville (NG 127 f.) claimed that Gaulish mandu- (in names such as Mandubracius, Mandubilus, and Manduessedum) should be related to the root *mendh- seen in Gk. μδνθηρη φροντίς, μανθανω Ί learn', etc. (see W. - P. 2. 270 f., IEW1W), and that the form pointed to an 'adjectif preceltique mndhii-s "qui reflechit ä", "qui se preoccupe de", en gallois mandti-s"'. Besides, it would seem correct to suppose that the anthroponym Mandullus or Mandulius of Barcelona ( C I L II 6147 = 4516), for which there are variants Manduilus in Aquitaine {DAG 322) and Manduillos, Manduilla in Cisalpine, has the same radical and is involved in this very discussion. Maria Lourdes Albertos proposes, following Schmidt (1957: 236), the radical *mand-, D'Arbois de Jubainville, however, and Holder later have related it to *mendh'to think', as we have just remembered, quoting Evans. But their etymological proposal for Manduessedum is worth discussion: Maria Lourdes Albertos maintains that it can be related to mando-, although Tovar accepts it with reticence. Manduessedum was a Roman garrison in Britania, today Mancetter, in Warwich County, England (21 kilometers north of Coventry). It appears in the form of Manduessedo ( I A 470, 3). Holder, s. v., records the interpretations offered by Pictet, πολυάρματον and D'Arbois de Jubainville: *Mandu-essedo-s 'celui qui veille sur les chars de guerre' or 'qui s'occupe des chars de guerre' and refers to Tarv-essedum. However, earlier (col. 404), in a discussion of the radical mandu-, according to d'Arbois de Jubainville, Holder relates the term to pre-Celtic mndh-ü-s 'qui reflechit a' 'qui se preoccupe de', which is related to the Greek radical mendh, as we saw. On the contrary, Tovar wisely points out (p. 155) that if the parallel form is taruessedum, meaning 'ox-cart', there is no reason why manduessedum cannot be analyzed as mandu-essedum

388

Francisco

Marcos-Marin

and "mean simply 'mule cart' or some such thing". Here, he relies on Dottin {La langue gauloise: 358) and, although he does not explicitly say so, he takes into consideration the meaning of mannus in Latin, "small draught horse". Birkhan (1970: 402, n. 1033) adds a note on the name of Mandubracius, the chief of the Trinovantes. It is again a compound noun of questionable etymology. In his own words: "Die zuletzt vorgeschlagene Etymologie SCHMIDTS 'Steiß eines jungen Pferdes habend' (154, 237) ist ganz auf der Voraussetzung aufgebaut, daß *brac- wirklich 'Hinterer' bedeutet habe, was zwar für das Germ., nicht aber für das Kelt, zutrifft, da im Kelt, ja *brag- (ohne L[aut]Verschiebung]!) zu erwarten wäre. Man wird doch nicht annehmen wollen, daß das Kelt, das Wort für 'Hinterer' aus dem Germ, entlehnt habe! — Ich halte es daher für notwendig, entweder anzunehmen, daß das Zweitglied zu gall, bräca 'Hose' gehört oder mit D'ARBOIS DE JUBAINVILLE (...1891: 131 f.) an das Malz(Getreide-) Wort gall, bracis anzuschließen ist. Die Bedeutung des Zweitgliedes hängt natürlich wesentlich auch an der des Erstgliedes, die selbst wieder sehr umstritten ist (...). Daß sich allgemein hinter dem N-Element *mandu- die Bedeutung 'junges Pferd, Pony' verbergen soll, wie SCHMIDT (237) meint, leuchtet mir nicht ein, zumal die lat. Formen mannus, manulus [sic] (W. — H. 3 II, 29 f.) kein -d- enthalten. Es scheint nur in bask, mando 'mulus' auf. Zum "illyr." Etymon s. H. KRÄHE, Die illyrische Namensgebung. Würzburger JBB. 1 (1946) 189, 191, 216 (zu Iuppiter Menzanas). Zu *mandu- s. reiche Lit.-Angaben bei EVANS 222 f. Wenn im Zweitglied das Hosenwort steckt, so darf man annehmen, daß dieses Kleidungsstück des Reiters als vornehm galt und ein ähnlich hohes Wortethos hatte wie bei uns die Reitsporen." It would seem preferable to mention in passing another of the etymologies which has been pondered and which came to be quite accepted. In ZRPh, XI: 256, corresponding to "Miscellanea Etymologica" and under the heading "Spanisch manera", W. Meyer — Lübke writes: For Sp. manera 'sterile woman', Galician maneira 'barren female', maneiro: 'sterile', Port., Gall, maninho 'sterile' for animals, manero 'he who dies leaving no legitimate heirs', one finds in Gascon only one simple word, mane. The information available would suggest a relation between the Portuguese and Gascon forms and a basic form *manna, and not *manya. The meaning inmediately

Etymology and semantics: apropos Spanish manero, mafieria

389

brings to mind Italian meno dealt with in Caix Studi 46 and Schuchardt Litteraturblatt 1885, in particular: 114. However, the sounds are against it. Whatever the etymon of menno is, an e in Italian cannot correspond to an ä in the Iberian Peninsula. Since manna does not appear in Latin and there is no corresponding form in Arabic, only Basque and Germanic are left. The former has against it that the word also appears in Portuguese, and in the latter we have manna 'the man' whose feminine manna, with the meaning given above, would behave similarly to Latin taura, Port, toura 'sterile cow' as related to Lat. taurus, Port, tauro 'bull'. Although the information we possess today allows us to reject this etymology (as did the DECH earlier), it should be mentioned here, if only for the sake of exhaustiveness since it is accepted in the review in Romania, XVII: 149-150 (— 'brehaigne' < Ger. mann) and by Americo Castro in RFE, XVI: 413. It was corrected in Briich (1919), where the relation to mannus was established, the Germanic form Menz documented and where Basque mando was incorrectly thought to be a Romanic loan according to Baist, Ζ RPh, XIV: 183. Five provisional conclusions were proposed above which allowed us to postulate a radical MA(N)D-. Now it seems that something can be added to them: 6. Three types of radicals have been encountered: MAND-, which seems to be basic, MANN- by assimilation of the dental sound to the preceding nasal, and ΜAND + yodh (palatal glide) which would result in the palatalization and assibilation of the dental sound. According to our information, the basic Indo-European type could be *mand- and *mnd-. (See Tovar, 1949: 158. For a critique of J. Loewenthals' statement, also mentioned in the DECH, see Wörter und Sachen, IX, 1924/26: 188 ff., and ibidem: 158, 159, and note 1 on this last page.) 7. I totally agree with Tovar (1949: 159): "Semantically, this data can only be reduced to a unity when one starts with the idea of 'udder'". From that point, one progresses to the concept of 'weaness', then to 'an animal too young yet to reproduce' and, finally, to the concept of 'sterile'. The intermediate step is well documented for animals and people by the Rumanian minz meaning 'colt' or 'boy'. One could add another observation a propos of 'young animal': this meaning, applied to people, would correspond to 'boy', 'girl'. Perhaps it would be plausible to agree with J. Corominas'

390

Francisco

Marcos-Marin

and J. A. Pascual's opinion in the DECH (although taking into account the reservations they point out there) and look there for the etymology of Sp. mano (cf. s. v.) and mafia, maybe from mannus, a word with an initial meaning of 'young', 'one who has not yet had children' also in Valencian zones where Catalan is spoken. This does not exclude the simultaneous development in Spanish of a haplological form of germanus, with the parallel hermano / mano in Mexico and other Central American areas. An Arabic etymology has also been ventured, in this case, but does not offer the necessary methodological rigor nor consistency (a critique in Marcos —Marin, 1982). 8. From 'sterile animal' stems, naturally, on the one hand, steer, mule and, on the other, maneria and manero, semantically in that order. 9. Basque mando means 'mule' according to the Diccionario vasco-espanol-frances by R. Maria de Azkue. But it also corresponds to 'sterile' for female animals and, pejoratively, for women. Other meanings are 'large, bulky' as opposed to zorri 'small' (originally 'louse' which may indicate the metaphorical origin of the semantic change: large as a mule, small as a louse). The affective form, with palatalization of the d, is used for 'male mule', 'ass', 'small sterile woman' and 'a bird hybrided of two species'. 10. It was stated too that Albanian and Greek are not alone among the Indo-European languages. Also in the Germanic languages, in Old High German, manzon 'ubera' is documented by Trautmann and quoted by Tovar (1949: 158, referring to ZvS, XLV, 1913: 252). Rudolf Schützeichel (1974) records it as manzo meaning 'Brust' ('breast' or 'teat'), a weak masculine noun documented in Tatianus, Cod. 56: 25-342 of the St. Gallen conventual library. 11. Those languages which have lost the primary meaning of 'udder', such as Celtic, Basque, Latin and the Romance languages, present the form without palatal glide. 12. Those languages which have retained the primary meaning 'udder' (Germanic, Albanian, Southern Italian dialects) tend to use the form with palatal glide for the meanings 'young animal' or 'sterile woman'. This, of course, may be a morphological indication of this form's secondary lexical-grammatical nature. These fragmentary conclusions, although not all of the same validity, may be considered sufficient to venture a synthesis. Manero, -a and maneria are derived from an Indo-European base MANN-, the result of MAND- by assimilation of the group -nd-,

Etymology and semantics: apropos Spanish manero, maneria

391

and present in the Celtic-Latin lexical form mannus. Semantically, our two words represent the last step in the evolutionary process, that is the reference to sterility, which occured at a relatively modern period since the meaning 'young animal' is still frequent in the Romance languages. By logic and from a semantical point of view, the first term must be manera, 'sterile woman' from which maneria would be formed, at first for a vital situation, and later for judicial reference. Finally, manero would be formed fundamentally as a judicial term. This contrast in the meaning of the derivatives in -ero, -a, depending on the gender, is not rare in Spanish, although it is usually ordered in relation to the primitive gender. Thus, masc. bombon, fem. bombonera, fem. gallina, masc. gallinero (Garcia: 1970).

These observations would be sufficient for a Romanist (and should also be sufficient for the author) since they have reached the limit of etymological speculation. For the Indo-Europeanist, there is still one more path to follow, a path somewhat more removed chronologically: the possible Illyrian origin of the radical and its pre-Celtic nature. This aspect may also be of interest to Hispanists as one more datum on the contacts sustained between Basques and Iberians in what was not yet called Hispania, nor Iberia, but perhaps was beginning to be known as Oestrymnis or Ofiusa.

Notes 1. I greatly appreciate having been allowed to consult the files of the Seminario de Lexicografia de la Real Academia Espafiola since the Diccionario Historico is still quite distant from M. It is a pleasure to acknowledge how much I am indebted to Prof. Dr. K. Dietz for his comments on an earlier redaction of this paper. Prof. Dr. H. Lüdtke also jotted down some valuable remarks. I fear, however, they may not recognize their generous help in the results. 2. I quote from the edition in the Biblioteca de Autores Espaiioles (BAAEE) since the variants which may exist due to the lack of editorial rigor, are not relevant to the purpose of this paper. See the modern edition by Agapito Rey, Indiana University, 1952. However, the punctuation is slightly corrected and the accentuation is suppressed. 3. The text of chapter LXXIX is partially quoted by Julio Cejador, 1929. Almost the exact same text appears in Juan Garcia de Castrojeriz' Regimiento de los principes, a translation of the Latin work by Frier Gil de Roma, Gil de Columna or Egidius Romanus, ca. 1345, published in Sevilla, 1494, by Ungut and Polono, fol. LXXXV: "Lo terzero se embargaria la generacion de los fijos. Ca fazer se yan las muegeres [sic] casadas por esta razon mafieras". A Regimiento de principes was also written by Gomez Manrique (XV c.). 4. Lauchert's edition of manuscript S of Strassburg in Romanische Forschungen 7, 1893: 331-402. Manuscripts Ρ, G and S have been recently edited by J. E. Keller and R. W. Linker, 1979. These editors read manereza in S: 365.

392

Francisco

Marcos-Marin

5. Quote from a cedula of the Real Academia Espanola which erroneously remits to Yahuda, RFE, 1914: 348. The DECH remits to Bol. R. Α. Ε., XV: 219. 6. Although the coincidence is probably due to popular etymology and a cross I won't go into now, it might be mentioned here that mannera is the name of the 'argemone' in the anonymous Hispanic-Mussulman botanist of ca. 1100 (M. Asin Palacios, 1943, 324: 166). This plant is from the papaveraceous family and is considered an antidote or counterpoison. It would be interesting to know whether it is also an abortive. 7. Also recorded there is: MANNULUS, -I m. a mannus dentin. GLOSS L - Corp. Μ 57 -us: caballus, buricus. MART. 12, 24, 8 nusquam est mulio: -i tacebunt. PLIN. epist. 4, 2, 3 habebat puer -os multos et iunctos et solutos. 8. The move from 'sterile' to 'impotent' is apparently possible in one case: Landes dialect mane "sterile, impuissant (en parlant du betail)" (Metivier, Agric. des Landes, 732), cited in the DECH, s. v. manero. One should keep in mind, however, that in modern Italian, manzo is a 'castrated bovine animal'. M. Alinei (1980: 19-20) has proposed an etymon mansus in agreement with Diez' proposal, without considering the great extension of the lexical family we are studying and its presence in non-Latin Indo-European languages. In view of our data, one may suppose that the influence of manso on manzo could be explained by a popular etymology. 9. The FEW gives us new forms under the pre-Roman headings *mand- and *mandiä, which it explains after the Indo-European radical MEND-. It agrees with Tovar on the semantic development, and discusses its relation to MANNUS (p. 147): "Nach Stolz - Leumann 169 ist es zweifelhaft, ob lt. MANNUS zu den vereinzelten Wörtern gehört, die -nd- zu -nn werden lassen (wie etwa ablt. grundio > grunnio; dispennite, zu pandere, neben distendite; tennitur). Es erhebt sich deshalb die Frage, ob dieser Lautwandel bei MANNUS nicht schon sehr alt ist, zwar nicht im Lt., aber im Gallischen, wovon die heute aufs Gask. beschränkte Lauterscheinung nur noch ein Relikt wäre;... Darauf könnten auch apr. hyperkorrekte Bildungen wie coronda für corona, colonna, ferner endir < HINNIRE, endilhar neben enilhar < HINNITULARE (...) deuten, sowie die Tatsache, daß die mit. Belege für MANNUS aus Gallien und Spanien stammen". I don't believe so obvious the connection with the noun manna and the verb mannar or mannear discussed by Georg Sachs in the Miscelänea of RFE, 23, 1936 (183-4). The proposed etymology, "Iberian *manna 'sterile' ( R E W , 5307)" is, of course, wrong. The noun manna appears in 1283, in the Alphonsine Libro de Ajedrez (fol. 10b) in the phrase juego mana, Sp. 'tables', 'chess game drawn by reason of inability of either player to checkmate'. The verb mannar appears in fol. 135b. A verbal adjective manneada which refers to a verb mannear is recorded in the Fuero de Cuenca (Urefia ed., Madrid, 1935: 533). The meaning 'to equal' can only be remotely related to the concept of 'sterility', although such a remote connection cannot be denied.

References Adas 1976

F. Jorda, J. de Hoz and L. Michelena, Actas del I Coloquio sobre lenguas y culturas prerromanas de la Peninsula Iberica (Salamanca, 27-31 de mayo de 1974), (Salamanca: Universidad).

AH DE Anuario de Historia del Derecho Espahol. Madrid.

Etymology and semantics: apropos Spanish manero, maneria

393

Albertos Firmat, Maria Lourdes 1966 La onomästica personal primitiva de Hispania. Tarraconense y Betica. (Salamanca: C. S. I. C. and Colegio Trilingüe de la Universidad). Alinei, Mario 1980 "Questioni di metodo e di fatto nella ricerca etimologica romanza", Romanica Europaea et Americana, Festschrift Harri Meier (Bonn: Bouvier), 19-20. Arbois de Jubainville, H. d' 1891 Les Noms gaulois chez Cesar et Hirtius De Bello Gallico, Premiere Serie, Les composes dont RIX est le dernier terme (Paris). Asin Palacios, Miguel (ed.) 1943 Glosario de voces romances (Madrid-Granada: Escuela de Estudios Arabes). Birkhan, Helmut 1970 Germanen und Kelten bis zum Ausgang der Römerzeit (Wien: Österreichische Akademie der Wissenschaften). Bol. R.A.E. Boletin de la Real Academia Espanola, Madrid. Brüch, Josef 1919 "Zu Meyer —Lübkes etymologischem Wörterbuch", ZRPh 39: 200-211. C. I. L. 1863 et ss. Corpus Inscriptionum Latinarum (Berlin: consilio et auctoritate Academiae litterarum regiae Borussicae editum). C. S . I . C . Consejo Superior de Investigaciones Cientificas. Spain. Cejador y Frauca, Julio 1929 Vocabulario medieval castellano. (Madrid: Hernando). Cioranescu, Alejandro 1960 Diccionario etimologico rumano (La Laguna: Universidad). Corominas, Joan and Pascual, Jose Antonio see DECH. DECH 1976 and ff. J. Corominas and J. A. Pascual, Diccionario etimologico castellano e hispänico (Madrid: Gredos). Evans, D. Ellis 1967 Gaulish personal names. Α study of some continental Celtic formations (Oxford: Clarendon Press). FEW 1928 and ff. Französisches Etymologisches Wörterbuch. Eine Darstellung des galloromanischen Sprachschatzes von Walther von Wartburg (Basel: R. G. Zbinden). Fuero de Mendavia see Lacarra: 1934. Garcia, Erica C. 1970 "Gender switch in Spanish derivation ...", RoPhil 24: 39-54, with a post-script by Y[akov] M[alkiel]: 55-57. Garcia Gonzalez, J. 1955 "La maneria", AHDE 12: 224-299. General Estoria 1930 ed. A. Solalinde (Madrid: Centro de Estudios Historicos). Holder, Alfred 1904 Alt-Celtischer Sprachschatz (Leipzig: Teubner. Rep. 1962, Graz: Akademische Druckanstalt).

394

Francisco

Marcos-Marin

Hübner, Ernst Willibald Emil 1871 Inscriptiones Hispaniae Christianae (Berolini apud Georgium Reimerum. Rep. 1975, Hildesheim New York: Georg Olms). Hubschmid, Johannes 1954 Pyrenäenwörter vorromanischen Ursprung und das vorromanische Substrat der Alpen (Salamanca: Universidad). I. E. W. see Pokorny: 1959. I. H. C. see Hübner: 1871. Keller, John Esten, and Linker, Robert W. (ed.) 1979 Barlaam y Josafat (Madrid: C. S. I. C.). Krähe, Hans 1955 Die Sprache der Illyrier. Erster Teil: Die Quellen (Wiesbaden: O. Harrassowitz). Lacarra, Jose Maria (ed.) 1934 "Fuero dado a Mendavia por Sancho el Sabio y confirmado por el Infante don Fernando de Castilla el 19 de Noviembre de 1274", AH DE 11: 1157/492. Lapesa, Rafael 1970 "Personas gramaticales y tratamientos en espanol", Revista de la Universidad de Madrid. Homenaje a R. Menendez Pidal 19: 141-167. Malkiel, Yakov 1956 "Studies on Spanish and Portuguese animal names (Conclusion). IV. Verbs derived from animal names", HR 26: 207-231. Marcos-Marin, Francisco 1977 "Notas de literatura medieval (Alejandro, Mainete, Marco Polo ...) desde la investigation lexica de 'brahman' y sus variantes", Vox Rom. 36: 121-161. 1982 "Eunnas no es 'eunnas'". (Notas lexicas sobre los textos del privilegio de Valderejo)", DICENDA, Cuadernos de Filologia Hispänica, Universidad Complutense, 1: 169-175. Menendez Pidal, Ramon 1943 "Ligures ο ambroilirios en Portugal", Revista da Faculdade de Letras de Lisboa, X, included in Toponimia Prerromänica Hispänica (Madrid: Gredos, rep. 1968: 159-178). 1956 Origenes del espanol (4th rev. ed.) (Madrid: Espasa Calpe). Meyer-Lübke, Wilhelm see REW Nebrija, Elio Antonio de 1492 Diccionario Latino-espanol (Salamanca. Facsimil ed., estudio preliminar de German Colon y Amadeu —J. Soberans, Barcelona: Puvill, 1979). NG see Axbois de Jubainville, H. d' Palomar Lapesa, Manuel 1957 La onomästica personal pre-latina de la antigua Lusitania (Salamanca: C. S. I. C. and Colegio Trilingüe de la Universidad). Pauly — Wissowa 1839 Real-Encyclopädie der classischen Altertumwissenschaft, by August von Pauly; new edition by Georg Wissowa, 1893 (Stuttgart). Pokorny, Julius 1959 Indogermanisches etymologisches Wörterbuch, I (Bern: Francke).

Etymology and semantics: apropos Spanish manero, maneria REW 1972

395

W. Meyer — Lübke, Romanisches etymologisches Wörterbuch (5th ed.) (Heidelberg: Winter). Schmidt, Karl Horst 1957 "Die Komposition in gallischen Personennamen", ZCPh 26: 26-32. Schot, Elsbet 1935 Das Wiesel in Sprache und Volksglauben der Romanen (Dissertation Tübingen). Schulten, Adolfo (ed.) 1955 Avieno. Ora maritima, in Fontes Hispaniae Antiquae, I (Barcelona: C. S. I. C.). Schützeichel, Rudolf 1969 Althochdeutsches Wörterbuch (Quoted after the 2nd rev. and comp, ed. 1974) (Tübingen: Max Niemeyer). Seybold, Chr. Fred, (ed.) 1900 Glossarium latino-arabicum ex unico qui exstat codice Leidensi, undecimo saeculo in Hispania conscripto (Berlin). Thesaurus 1900 et ss. Thesaurus Linguae Latinae (Leipzig and München). Tovar, Antonio 1949 Estudios sobre las primitivas lenguas hispänicas (Buenos Aires: Instituto de Filologia). 1977 "El nombre de celtas en Hispania", Revista de la Universidad Complutense (Horn. Garcia Bellido 3) 26: 165-166. Urquijo e Ibarra, Julio 1949-51 Homenaje a D. - (San Sebastian). Walde — Hofmann 1910 A. Walde, Lateinisches etymologisches Wörterbuch (2nd ed.) (Heidelberg: Winter) (1930 ff. 3rd ed. by J. B. Hofmann). Walde — Pokorny see Pokorny: 1959.

ANATOLIJ Μ. M U K H I N

Lexical and syntactic semantics in historical aspect

The task of the present paper is to discuss from the perspective of the history of the English language the possibilities of systemic analysis of lexical and syntactic units, that is lexemes and syntaxemes. In pursuing this purpose I can turn to my own experience in analysing systemic relations of those units in Modern English and Russian, which is reflected to some extent in my books (Mukhin 1964: 248-278; 1976: 99-229; 1980: 99-133). 1. Lexical semantics constitutes the content of the elementary lexical units, which are distinguished primarily by the part of speech they represent. So we recognize substantival, verbal, adjectival, adverbial, pronominal lexemes, or in other terms, substantives (nouns), verbs, adjectives, adverbs, pronouns. An important formal characteristic of lexemes belonging to a definite part of speech is their capacity to combine with other lexemes within lexical constructions, called phrases. Such lexical constructions are made up of lexemes, such as a verb and a substantive, or a verb and a pronoun, etc., as distinct from syntactic constructions, comprising elementary syntactic units, named syntaxemes, which will be discussed later on (3). Here I am primarily concerned with verbal phrases based on government, which is manifested by prepositions or their absence. According to the presence or absence of government verbs are characterized as transitive or intransitive. The same distinction is observed among the lexemes of other parts of speech — nouns, adjectives and adverbs. Hence, in English, Russian, and other IndoEuropean languages, transitive lexemes, including transitive verbs may be defined as lexical units capable of prepositional and nonprepositional government. As for government itself it is revealed by experiments and may be shown in symbolic models. For example, the impossibility of replacing or omitting the preposition for follow-

398

Anatolij Μ. Mukhin

ing the verb starve, which means "long for, be in great need o f ' (to starve for knowledge, etc.), testifies to the fact that it is the lexical semantics of this transitive verb that calls forth the use of the preposition. In other words, the preposition for serves as an indication of the government of the verb starve. On the basis of this prepositional government a lexico-semantic group of transitive verbs including ache, die, languish, starve, wish, etc., is formed in Modern English. Having the same prepositional government (V tr + for + C), all these verbs show identical or similar lexical semantics, and form similar combinations with lexemes serving as complements (complementary lexemes designated by the letter C in the model). Cf.: to ache for a rest, to die for a drink, to languish for relieve, to starve for friendship, to wish for peace, to wish for happiness, etc. But most verbs of this group including crave, gape, gasp, hanker, hunger, long, lust, pant, strive and some others may show variable government, which is accomplished by prepositions for and after (V tr + for/after + C): to gape for knowledge — to gape after knowledge, to gape for profit — to gape after profit, to lust for power — to lust after power, to pant for money — to pant after money, etc. Therefore, transitive verbs having the meaning of ardent desire, craving for something form two different subgroups according to whether they have variable or non-variable government. The two criteria — identity or similarity of lexical semantics and similarity of combinatorial capacity (due to the same government) are obligatory for any group of transitive verbs. Furthermore, it is to be emphasized that one should always take into account the capacity of transitive verbs to combine with other lexemes, which are represented in the above phrases by nouns denoting, as a rule, abstract notions. The fact that the meaning of transitive verbs is determined with due regard for their capacity to combine with lexical (and not syntactic) units shows that it is lexical semantics that is instrumental in this case and it differs considerably from syntactic semantics. Moreover, while finding out lexico-semantic groups of transitive verbs it is necessary to consider some other facts besides their ability to combine with substantival or pronominal lexemes on the basis of appropriate government. It is possible for most of the above monotransitive verbs having the meaning of ardent desire or craving for sth to combine with the infinitive used instead of a noun with the preposition, as in the phrases: to ache to see somebody, to die to learn a secret, to long to go away, to strive to do something, etc. Although these phrases fail

Lexical

and syntactic

semantics

in historical

aspect

399

to demonstrate prepositional government (by means of prepositions for or after), the capacity of the above verbs to combine with the infinitive can not be disregarded as it singles out this group of monotransitive verbs from other lexico-semantic groups of monotransitive verbs that lack such a combinatorial characteristic. Similarly, several groups of ditransitive verbs permit the infinitive to be used instead of a noun with the preposition or without it, which distinguishes them from many other lexico-semantic groups of ditransitive verbs. In Modern English the same distinctive function is performed by the ability to combine with gerund which is characteristic of the group of ditransitive verbs with the meaning "stop or keep from doing or happening" and the government V tr + C 1 + from + C 2 : to preclude somebody from performing this task, to prohibit somebody from coming, to prevent somebody from getting married, etc. Lexico-semantic groups of transitive verbs often correlate in their meaning and combinatorial capacity with the groups of transitive lexemes of other parts of speech, especially those of transitive substantival and adjectival lexemes that may be established on the basis of appropriate government. The existence of such systemic relations between transitive lexemes of different parts of speech enables a researcher to carry out experiments which include transformations of verbal phrases into substantival or adjectival ones. Thus, for example, phrases with monotransitive verbs denoting emotions or emotional attitude — adore, dislike, hate, like, love, pity, respect, etc., may be transformed into substantival phrases with the preposition for (Vtr + C -*• S tr + for + C): to adore a woman adoration for a woman, to dislike a boy —• dislike for a boy, to hate a person -*• hatred for a person, etc. Phrases with some other monotransitive verbs having the same non-prepositional government may be transformed into substantival phrases with the prepositions at (Vtr + C -* S tr + at + C) or on (Vtr + C S tr + on + C), cf.: 1. to attempt a task —• an attempt at a task, to essay a problem -> an essay at a problem, to try a new detergent —• a try at a new detergent', 2. to impress somebody —> impression on somebody, to influence somebody's decision —> an influence on somebody's decision, to sway somebody's opinion —• a sway on somebody's opinion, etc. The experiments of such type which are made possible by the existence of systemic relations between transitive lexemes of different parts of speech may serve a purpose of establishing a border line between the studied lexico-semantic groups of transitive

400

Anatolij Μ. Mukhin

verbs, in particular those having the same non-prepositional government. Similar experiments can be carried out on the material of a dead language, such as Old English, for example. In this case, however, they should be supported by the evidence of the texts which might demonstrate a range of combinatorial potentiality of the lexemes under study, while for the experiments performed on the material of any modern language (English included) it is quite sufficient to be approved by native speakers. It should be pointed out that systemic analysis of transitive verbs which takes into account the fact of their belonging to certain lexico-semantic groups is of great importance for the history of language. In this case the historical study does not concern isolated linguistic facts, but systems of elementary lexical units, liable to comparison in different historical periods of language. It becomes possible because of the fact that transitive verbs have quite definite meanings (lexical semantics) as well as formal combinatorial characteristics, which are relevant for any historical period of the language. As a starting point for the analysis, one can take the groups of transitive verbs in Old English which combine with substantives and pronouns by means of a preposition or without it and trace their development up to modern English. For example, among Old English transitive verbs governing the dative case (Vtr + C d ), we can single out a lexico-semantic group of verbs with the meaning of service, obedience or submission to somebody or something: bugan, gebugan, onbugan, underbugan, folgian, hyran, peowian, underlicgan, mishieran and others. Cf. peowian {folgian, hyran) hlaforde "to serve, to obey one's lord", etc. Sometimes the government of the dative case by means of the preposition to and the accusative case by means of the preposition on may be observed along with non-prepositional government. Variable government is characteristic of many transitive verbs in Old English (Safarov 1975: 253-258). When tracing the development of groups of transitive verbs from Old English to Middle English and further on, one may see the differences and similarities in their systemic relations throughout different historical periods of the language. This especially concerns the use of prepositions as means of government against the background of the existence or absence of case distinctions, as well as the extent to which variable prepositional and non-prepositional government is represented. Besides, one should give special attention to quantitative composition of groups of transitive verbs in different

Lexical and syntactic

semantics

in historical aspect

401

historical periods of the language. As an outcome, the analysis may show that one group of transitive verbs in Old English may not necessarily correspond to only one group in Middle or Modern English. The study of systemic relations of transitive verbs in different historical periods of the language would also have to take into account some syntactic and derivational phenomena. The latter may result from whether these transitive verbs have prefixes or suffixes in their derivational structure or not. Thus, the study of systemic relations of transitive verbs based on government in Old English or any other historical period of the English language requires the analysis of a wide range of language phenomena and has a direct bearing on the problems of a comparative-typological investigation of languages which is also based on systemic relations of linguistic units. 2. Unlike transitive verbs, which are subdivided into monotransitive and ditransitive due to their government, intransitive verbs are not capable of government. This can be shown by the experiments similar to those mentioned. We have already seen that in the phrase to starve for knowledge the preposition for can neither be replaced by any other preposition nor omitted, because it is conditioned by the lexical semantics of the transitive verb starve ("long for, be in great need of')· But in the phrase with the intransitive verb to lie on the table (or in the sentence The book is lying on the table) the preposition on can be replaced by the prepositions in, under, near, besides, above and others. None of these prepositions is conditioned by the lexical semantics of the verb lie ("to be resting flat on sth") which falls under the category of intransitive verbs. The fact that these prepositions do not serve as means of government in the phrases with the verb lie is confirmed by the possibility of their omission when the adverbs there or here are used instead of a noun: to lie there, to lie here. But the absence of government does not necessarily mean that intransitive verbs cannot form lexico-semantic groups when they have the same or similar meaning and similar combinatorial capacity. The latter is the formal (distributional) feature which characterizes all the elementary lexical units, including transitive and intransitive verbs. It should be emphasized that lexemes can only combine with lexemes — the same as syntaxemes can only combine with syntaxemes, morphemes with morphemes, and phonemes with phonemes. For example, the intransitive verbs with the meaning of motion or movement (travel, walk, wander and

402

Anatolij Μ.

Mukhin

others) are characterized by their ability to combine with substantives designating space or territory: to travel round the world, to walk along the road, to wander about the world, etc. It seems likely therefore that these intransitive verbs form a certain lexicosemantic group. It results from their close meaning and similar combinatorial capacity, which in the given phrases is brought about by different prepositions. These intransitive verbs correlate with some transitive verbs (travel, walk, wander and others) which have non-prepositional government (Vtr + C) and should therefore be regarded as their homonyms: to travel the whole world, to travel a country, to walk a street, to wander the world, to wander the forest, etc. These transitive verbs make it possible for a researcher to construct a symbolic model of government (Vtr + C), whereas the above mentioned intransitive verbs lack such a possibility (cf. different prepositions preceding substantives, designated by the letter S : V + round + S; V + along + S; V + about + S, etc.). It should be also mentioned that both groups are characterized by different systemic relations with the substantives of the same root: in Modern English there are corresponding intransitive substantives which are used with the same prepositions (round, along, about, and others), but there are no substantives correlating with the above mentioned transitive verbs. Of special importance for the historical study is the problem of the mutual relation between intransitive verbs and transitive (monoand ditransitive) verbs, which are close by their lexical semantics but have different systemic relations in the language, that is they enter into different lexico-semantic groups and correlate with different groups of lexemes belonging to other parts of speech (substantives, adjectives and adverbs). This sphere of lexical analysis covers numerous pairs of homonymous monotransitive and intransitive verbs in Modern English, such as are often discussed in relation to the problem of the so-called middle voice: openλ "to spread out, expand" — open2 "to become open" (/ opened the door — The door opened)·, bumx "to destroy by fire" — burn2 "to be on fire" (/ burnt the paper — The paper burnt), boil\ "to heat to the boiling point" — boil2 "to bubble up and vaporize by being heated" (/ boiled the water — The water boiled), resume\ "to continue after interruption" — resume2 "to proceed after interruption" (We resumed the conference — The conference resumed), sell\ "to exchange sth for money" — sell2 "to be sold" (I sell the book — The book sells well), etc.

Lexical and syntactic

semantics

in historical aspect

403

No less importance should be attached to the historical study of the correlation between homonymous monotransitive and ditransitive verbs which, however close they may be by their lexical semantics, form different lexico-semantic groups, as, for example, in Modern English: run\ "to stumble upon sth" — run2 "to drive (an object) into or against something", hit χ and striket "to come against something with force" — hit2 and strike2 "to bump something against something", etc. Such correlative verbs may combine with the same substantives or pronouns, designated by the letters C and C 2 in the models of government, cf.: Vtr + against + C to to to to

run against a rock run against a wall hit against a wall strike against a stone

Vlr + C 1 + against + C 2 — to — to — to — to

run one's foot against a rock run one's head against a wall hit one's head against a wall strike one's foot against a stone.

The above mentioned considerations show that of primary importance is the study of the homonimy of transitive and intransitive verbs and the extent to which it has spread during different historical periods. It can only be made on systemic grounds, that is, taking into account the belonging of verbs to certain lexico-semantic groups and their correlation with appropriate groups of lexemes of different parts of speech, if any. 3. While studying lexical semantics of transitive verbs within lexical constructions the preposition is only regarded as a means of government, for it does not form a unit either with the governing lexeme (transitive verb) or with the governed one (complement). Quite a different situation is observed if we study prepositions in sentences where they form an indissoluble syntactic unit with the substantive or the pronoun. Such combinations with the prepositions following transitive verbs are used as means of expressing various objective or indirect-objective syntaxemes. For example, in the following sentences an objective syntaxeme is represented by two positional variants, one of which (a prepositional variant "after S", where S denotes a substantive) functions in the dependent position, the other (a discontinuous prepositional variant "S ... after") — in the position of the subject: Four doctors are looking after the children. The children are being looked after by four doctors. Similarly, an agentive syntaxeme is also represented in the above

404

Anatolij Μ.

Mukhin

sentences by two prepositional variants, one (a prepositional variant "by S") being used in the dependent position (The children are being looked after by four doctors), the other (non-prepositional variant "S") occupying the position of the subject {Four doctors are looking after the children). In case of intransitive verbs combinations with prepositions may also represent a number of syntaxemes but no objective or indirectobjective syntaxemes can be found among them. For example, in the sentence (The boys travelled round the world) the combination of the preposition with the substantive (round S) serves as a means of expressing a locative syntaxeme, while in the sentence (The boys travelled on Sunday) the combination with the preposition (on S) represents a temporal syntaxeme. Each of these syntaxemes has its own system of variants, expressed by different means, such as the combinations with different prepositions conveying identical syntactic semantics or, in other words, having identical syntacticosemantic content. Thus, besides the above mentioned prepositional variant "on S" the temporal syntaxeme has a number of prepositional variants: "in S" {He arrived in May), "during S" {He was invited to visit them during the summer), as well as a non-prepositional one "S" {He said he'd be home Wednesday), etc. Syntaxemes differ in their formal distributional features, such as the ability to combine with other syntaxemes, their distance from other syntaxemes as well as the range of their syntactic positions in sentences. Hence, a syntaxeme can be defined as an elementary syntactic unit (invariant), represented in the language by a system of its variants which may be expressed by both individual lexemes and syntactically indissoluble combinations of lexemes with auxiliary elements, for example, prepositions. The content of a syntaxeme is formed by its syntactic semantics, or, to be exact, the totality of different syntactico-semantic features which manifest themselves by the distributional characteristics of the syntaxeme, as well as its specific system of variants. Determination of syntaxemes and their variants within syntactic constructions in different periods of the history of the language enables a linguist to study syntactic semantics of language units on the basis of their systemic functional relations no matter what morphological or lexical means are used to express them. The situation changes when we aim at investigating any particular morphological means of expressing certain syntactic semantics, taking

Lexical and syntactic semantics in historical aspect

405

no account of its systemic functional relations to other elements, which may be quite different outwardly. When, for example, we are studying the case forms of substantives or pronouns conveying syntactic semantics of some kind, we have to limit ourselves to those historical periods when such case distinctions can be recognized. Thus, a linguist investigating dative, accusative or nominative cases of Old English cannot correlate them with the appropriate forms of Middle English and Modern English where they are missing. The above considerations show that a historical study of syntactic units should be based on their syntactic semantics while paying due attention to all the other means of conveying the same syntactico-semantic content. In this case the task of investigating syntax in its historical aspect (in terms of syntaxemes and their variants) becomes closely connected with the task of the comparative-typological analysis of syntactic units in different languages.

References Muxin, Anatolij M. 1964 Funkcional'nyj analiz sintaksiceskix elementov (Ν a materiale drevneanglijskogo jazyka) [Functional analysis of syntactical elements (on the basis of material from Old English)] (Moskva-Leningrad: Nauka). 1976 Lingvisticeskij analiz. Teoreticeskie i metodologiceskie problemy [Linguistic analysis. Theoretical and methodological problems] (Leningrad: Nauka). 1980 Sintaksemnyj analiz i problema urovnej jazyka [Syntaxemic analysis and the problem of language levels] (Leningrad: Nauka). Safarov, S. 1975 "Upravlenie i sistemnost' leksiceskogo urovnja jazyka (Na materiale drevneanglijskich glagolov)", Lingvisticeskie issledovanija 1975. Voprosy stroja indoevropeiskix jazykov. Cast' 2 (Moskva), 246-262.

HERBERT PILCH

The synchrony-diachrony division in wordformation

I. Empirical motivation 1. The problem. Synchrony and diachrony provide different, but equally valid points of view in linguistics. This has been generally recognized for a long time.1 Still, the two points of view have been traditionally interpreted in such a way that synchrony means 'at a given point of time', diachrony 'during a certain period of time'. This interpretation is surely inadequate, as any amount of data takes 'a certain period of time' to collect. Consequently, the synchronic point of view is physically impossible to practise, and the diachronic point of view absurdly focuses on the time it takes to collect the data. The more adequate alternative interpretation takes synchrony to mean 'considering speech as linguistic function', diachrony 'considering change of linguistic function as correlated with time'. The difference is in the researcher's point of view. It is not in the presence or absence of linguistic change as such (cf. Hammarström 1966: 60-61; Lüdtke 1980: 183-187). In other words, linguistic variety is always there. To the synchronic view, it appears as variability within a given language. To the diachronic view, it appears as successive stages of the same language. As far as word formation is concerned, the traditional interpretation has given rise to a certain amount of confusion. For instance, word formation has been confused with the etymology of affixes. The protocolization of the data has been confused with the chronology of first records. Polysemy has been confused with meaning change. Such confusion implies inconsistency or (to use the philosophical term) metabasis, producing a host of insoluble problems. Let us briefly review some of those perennial problems which, like the proverbial bad penny, always return into the discussion of word formation:

408

Herbert

Pilch

(i) The English suffix -ic "places stress on the immediately preceding syllable" ( C h o m s k y - H a l l e 1967: 88), witness: hero : sulphur : base :

hero-ic sulphur-ic bas-ic

However, certain unexplained exceptions to the rule arise, such as politic, catholic, Arabic. The exceptions arise, as the synchronic affix -ic (as in bas-ic) is being confused with the etymological affix -ic (as in politic < lat. polit-icum). On top of this, certain formations in -ic are correlated with abstract nouns in -icity, entailing an unmotivated alternation of /k/ in -ic, js/ in -icity, witness: electric phonemic

: :

electricity phonemicity2

In fact, the motivation is supplied by the synchronic fact of NeoLatin coinage (cf. section IV below), but this is being confused with the diachronic identity of the affix -ic in Latin and English. (ii) The English suffix -able involves no stress shift, i. e., "the stress pattern is that of the underlying form" (Chomsky — Halle 1967: 86), as in: read return present

: : :

readable returnable presintable

Again, certain unexplained exceptions arise which do involve a stress shift and even segmental changes, such as: admire despise neglect

: : :

ädmirable dispicable negligible

The problem arises, as the etymological identity of the affix -able is believed to guarantee the synchronic identity of the paradigm. (iii) Certain pairs of bisyllabic verbs and nouns do involve a stress shift, witness: permit : conflict : invite :

permit conflict invite

The synchrony-diachrony

division in word-formation

409

However, many others do not, witness dismäy, program, and there is a lot of dialectical variation, witness the nouns Br.E. research, address, dispute, exploit, Am.E. ally, but Am.E. research, (postal) address, Sc.Ε. dispute, Hb.E. exploit, Br.E. ally. The irregularity arises, as the synchronic existence of the paradigm is being confused with the chronology of first records (cf. section V.3 below). The direction of the derivation is less than obvious, for that matter. Presumably, the noun invite is derived from the verb invite. And obviously, the verb input (as in "we inpüt data") is derived from the noun input. (iv) The English prefix re- is one 'occurring in many words borrowed from Latin and French" (Quirk 1972: 990), witness reclaim. Yet it also occurs in many native words, e. g., re-write, reread, re-do. On top of this, the prefix re- (allegedly depending on the date of the formation) varies in pronunciation, Br.E. /n/ in return 'come back', but /re/ in recommend, /ri/ in re-consider: "Note the contrast between unstressed jnj found in borrowed words, and the /ri/ ... of a newer formation" (Quirk 1972: 990). The problem arises, as the (diachronic) chronology of first records is being confused with the (synchronic) productivity of the paradigm (cf. V.3 below). (v) The meaning of tatpurusa compounds with the suffix -er unpredictably fluctuates between 'agent as person' and 'agent as tool', witness: pay tax own land go (to) church mow lawn record (on) tape guzzle gas

: taxpayer (person) : landowner : churchgoer : lawnmower (tool) : taperecorder : gas guzzler

On top of this, the paradigm seems to involve erratic meanings, as in shoemaker 'agent who does NOT make shoes, but repairs them', chain smoker 'agent who smokes NOT chains, but cigarettes', skyscraper 'very high building' (neither agent nor tool). Prediction fails, as polysemy is being treated as a diachronic rather than a synchronic fact (cf. section V.4 below). Certain predicates, for that matter, derive tatpurusa without the suffix -er, such as the bellhop, the chimney sweep, the toothpick. How come two derivational paradigms co-exist with the same meaning? Should we, perhaps, test a population of undergraduates to see

410

Herbert

Pilch

whether they will accept invented tatpurusa like *bellhopper or *taxpayl Conversely, there are agent nouns in -er which do not correspond to an existing predicate, e. g. * say sooth * sing folk * write type

: : :

soothsayer folksinger typewriter

Both these problems arise, as productive derivation is being confused with historical derivation (cf. section V.l-2 below). (vi) The suffix -ness derives abstract nouns from adjectives in English, witness: kind useful arbitrary

: : :

kindness usefulness arbitrariness

However, what about richness, poorness? To the extent that these two words exist, they are not simply the abstract nouns corresponding to rich and poor. The difficulty arises as polysemy is being disregarded as a synchronic process (cf. section III.3 below). And what about tangiblenessl Or is it tangibility? The problem arises as the derivational paradigm is seen apart from the appropriate lexical paradigm. It can be solved neither by the appeal to first records nor by the administration of freshman tests (cf. section III.2 below). For that matter, what about forgiveness, wilderness? The former is correlated not with an adjective, but with the verb forgive. The latter is correlated with wild. Or is it?3 The problem arises, as productive and residual paradigms are being confused (cf. V.l below). 2. Requirements for a solution. In order to solve such problems, we propose a model of word formation which will enable us to: (i) Distinguish between hero-ic as a derivative of hero on the one hand, politic as a primary (non-derivative) on the other hand. This will basically be Hans Marchand's model which is pivoted on the focal concept of "derivative relevancy" alias "productivity" (Marchand 1969: 2-5). True, the latter notion is widely used in word formation, but not with adequate precision (see appendix below). (ii) Integrate word formation into lexicology, so we can view the two pairs kind : kindness and poor : poverty as the same lexical paradigm within which the productive derivation in -ness forms a

The synchrony-diachrony

division in word-formation

411

special morphological class. This will also enable us to apply to morphological derivatives the lexicological notion of polysemy, so we can account for the polysemy of such paradigms as the tatpurusa. (iii) Distinguish between monolingual and bilingual derivation. This is a generalization on Hans Marchand's division between the "English basis of coining" (as in read-able) and the "Neo-Latin basis of coining" (as in electricity). In this context, we allow for borrowing in such a way that a word can be monomorphemic in language lj (the borrowing language), even though it is polymorphemic in language 12 (the lending language). For instance, the Russian word sputnik contains the three morphemes s-put-nik. Borrowed into English, it is monomorphemic, even though it has triggered the English -mTc-paradigm of beatnik, nudenik (see appendix below). (iv) Redraw the dividing line between synchrony and diachrony. This division is usually understood in the sense of an earlier or later date at which a particular derived word is first recorded. However, we need a dividing line which will explain soothsayer diachronically as derived from the now extinct predicate say sooth, but typewriter synchronically as an extension of the tatpurusa paradigm, shoemaker as a conventional interpretation ("lexicalization") of the meaning regardless of chronology.

II. A model of morphological analysis 1. An exemplary paradigm. Let us first look at a set of English signs (words), taking it for granted they have been properly protocolized in the lexicon, including the pronouncing dictionary: write start consider do read re-write

re-write re-stärt re-consider re-do re-read re-rewrite

Our selection of precisely these signs and their present arrangement (rather than something else) is initially motivated by a hunch. It is ultimately borne out by its adequacy to the solution of our problem. This set is ordered in such a way that 4 :

412

Herbert

Pilch

(i) It is binary, i. e., it is divided into two subsets arranged in two columns. (ii) There is one-one-correspondence between the items in each given row. (iii) Each pair is similar in phonological form, grammatical function and meaning, and the differences involved are the same in all pairs. Specifically: In terms of phonological form, column II adds some phonemic material, namely /ri/, to the word in column I, and the whole word in column II has level stress. In terms of grammatical function, both members of each pair belong to the same syntactic subclass of verbs, i. e. they fill the same positions in the same constructions in each instance. In terms of meaning, column II adds the semantic feature 'duplication' to column I, such as re-read 'read again' alias 'duplicate the reading'. (iv) The two classes of words are open, i. e., they have an unlimited (not an infinite!) number of members. Consequently, no dictionary could possibly list all the members and only the members of the paradigm (see section (1) of appendix below). (v) There is linear arrangement of the added re- /ri/ and the original phonemic material. In such a situation, we segment the words in column II into the two morphemes (or monemes) re- + read and so on. Specifically: The morpheme /ri/ remains constant throughout the paradigm. By definition, the constant is an affix. The other morphemes write, consider etc. are variable in every row. They are stems (equally by definition). The range of the variable is, in this instance, the class of all English verbs — except the intransitive verbs of motion (such as fall, go), the stative verbs (such as stand, know, love), and the auxiliaries (such as have, can). However, those "exceptional" verb classes do enter the paradigm when their meaning is transferred metaphorically as in re-run (for election). 2. Generalization. We can now generalize on the above paradigm to present a model of word formation. The general model is a binary paradigm with a uniform relation of phonological form, grammatical function and meaning for each pair. Every particular paradigm of word formation is characterized by a particular relationship of this kind. For instance, write : re-write belongs to one particular paradigm, baby sitting : babysit belongs to another. For that matter, every particular paradigm belongs to a particular language. For instance the write : rewrite paradigm belongs to

The synchrony-diachrony

division in word-formation

413

English (not to Latin or French), even though some of the morphemes involved, such as re-, may be borrowed from Latin. There are several special features to the exemplary paradigm above. These features do not apply to all paradigms of word formation: (i) Above, each binary couple is ordered. In such cases, the first member of each pair is called primary, the second derivative. The derivative is said to be derived from the primary. The derivation is productive, when the paradigm concerned is open. However, not all binary paradigms are ordered, witness invite : invite, input : input above (section 1.1 (iii)). How can we tell the direction of the derivation? We can on the strength of the more specific sense of the derivative (cf. III.3 below). (ii) The paradigm write : rewrite involves an affix. However, not all derivational paradigms involve affixes, for instance compounds do not, nor does invite : invite. Remember that we have defined affix by paradigm, not inversely, so the rewrite paradigm owes its identity not to the prefix re-, but inversely the prefix re- owes its identity to the paradigm. The paradigm write : rewrite is recursive, as the derivative (e. g., rewrite) reenters the paradigm as a primary (witness re-rewrite in the last row). However, not all derivation is recursive. Recursiveness is tied to the necessary condition that the primary and the derivative have the same (or an overlapping) grammatical range. This theorem follows from the definitions above. We say that the morphological analysis of any given word (e. g., re-write) consists in assigning it to its particular paradigm of word formation and inflection (if applicable). Otherwise, it is monomorphemic, i. e., not analyzable morphologically from the synchronic point of view, at least not within the present model. 5 This is (we claim) a sufficient definition of morphological analysis. Consider, as an example, the derivative coastal navigate, as used in the advertisement: learn to sail and coastal navigate in the sheltered waters of the Solent (The Yachting World, April 1980: 266b). For the purposes of morphological analysis we assign this expression to the paradigm: Action Noun

Verb

dinghysailing babysitting coastal navigation

dinghysail babysit coastal navigate

414

Herbert

Pilch

It is usually assumed that all derivatives within a given paradigm belong not only to the same grammatical class, but also to the same morphological class, i. e., to the same gender, inflectional class, stem class, stress pattern, etc. (to the extent that such classes exist in the language concerned). This assumption is borne out notably by the Germanic and Slavic languages. For instance, all German diminutives in -chen, -lein, Swiss German -i are neuter regardless of natural gender, witness das Mädchen, Swiss German 's Susi (girl's name). Russian deverbative substantives with zero-suffix are feminine /-stems stressed on the initial syllable, witness propoved', zivopis', podpis'. The assumption seems to be fairly safe, so we are inclined to incorporate it into our model. 3. Solution of empirical problems. Some of those perennial problems of English morphology are solved by our model: (i) The element re- of verbs such as recognize, recommend, return, reform, is not an affix in English. We have defined affix by paradigm, and these verbs are not derivatives in the paradigm write : rewrite (nor in any other English paradigm). Consequently, they do not contain the prefix re- of the rewrite paradigm in the synchronic sense, witness the fact that the verb recognize has no correspondent *cognize and that the verb recommend differs from commend in a different way than in the paradigm both phonologically and semantically. Similarly, the words return, reform start with /ri/ (not /ri/), and they do not mean 'turn, form again'. All these words are adequately dealt with diachronically as borrowings from French or Latin. The etymological identity of the elements re- in recognize and rewrite still leaves them different synchronically (cf. Marchand 1954: 301-302). On the other hand, the derivatives re-commend, re-turn- re-form (with initial /ri/ and level stress and the meaning 'commend again' etc.) do exist in English, and they do belong to the paradigm above. However, they are similar to recommend 'advise', return 'come back', reform 'change' in spelling only. True, the three latter verbs are borrowed from French. Their morphological status in French is, however, not the same thing as their morphological status in English. This follows from our stipulation that every paradigm "belongs to a particular language". (ii) The element -ness of wilderness, forgiveness is not a suffix in Modern English. The latter two nouns are not derivatives in the paradigm kind : kindness, but monomorphemic words synchron-

The synchrony-diachrony

division in word-formation

415

ically, witness the fact that the grammatical relationship between them and their presumed primaries * wilder, forgive is not the same as in the kind : kindness paradigm. In fact, * wilder (pronounced as in wilderness) is not an English word at all, forgive is a verb (not an adjective). As we have defined affix by paradigm, the homophony of the final /n 3 s/ of forgiveness and kindness is not a sufficient condition of their morphological identity. Otherwise, why not also accept the homophonous final /n 3 s/ of bonus as "the suffix -ness,,r! Surely, the reason cannot be in the difference of spelling. (iii) The element -ible of negligible, despicable, legible is not a suffix in English, witness the fact that the phonemic relationship with their presumed primaries neglect, despise, *lect is not the same as in the read : readable paradigm. The words negligible, despicable, legible are adequately dealt with diachronically as borrowings, not synchronically by assigning them to the read : readable paradigm. True, neglect and negligible do involve the same Latin stem morpheme etymologically. Their morphological analysis in English is, however, not the same thing as their etymology.

III. Morphological analysis in lexicology 1. The lexical paradigm. From the vocabulary of English, we can assemble paradigms such as: Toponym

Inhabitant

(a) London New York New Zealand

London-er New York-er New Zealand-er

(b) Seattle Ann Arbor Michigan

Seattl-ite Ann Arbor-ite Michigan-ite

(c) Halifax Aberdeen Wales

Haligonian Aberdonian Welshman

(d) Birmingham Indiana North Carolina

Brummy Hoosier Tarheel

416

Herbert

Pilch

This paradigm fits the model of the derivational paradigm (as set out above) in all respects but one, namely the paired items differ from each other in the same way in grammatical function and meaning over the whole paradigm, but not in phonological form. Such a paradigm we call not derivational, but lexical. In fact, the paradigm of word formation is, by definition, a special case of the lexical paradigm, namely a lexical paradigm with uniform phonological differences within the pairs and, consequently, one subject to morphological analysis. The subsections (a) and (b) of the paradigm above are, in fact, subject to morphological analysis, yielding the suffixes -er, -ite, but the lexical paradigm as a whole is not. Either of these subsections can therefore be taken as one morphological paradigm, and both these morphological paradigms are integrated into the overall lexical paradigm. 6 The lexical paradigm is, in addition, characterized by satellite words 7 and by polysemy. It follows from our definition that these notions must also apply to the morphological paradigm (which is just a special case of a lexical paradigm).

2. Satellite words. The satellite word is a synonym which co-exists in the paradigm with "the established word", as Kiwi co-exists with New Zealander, Michigander with Michiganite, Brummagem with Birmingham (England). The satellite word usually carries special meanings, connotations and such like. It is therefore easy, in many cases, to decide which of those synonyms is the established word. Now this situation equally applies to the derivational sub-paradigms, when we view them as integrated into overall lexical paradigms. For instance, the derivative explainable can be taken as a satellite word of the established word explicable in the lexical paradigm concerned, similarly admirable of ädmirable, chimney sweeper of chimney sweep, fruitfull-ity of fruitfulness, poorness of poverty, etc. The exact connotations of those satellite words are somewhat erratic. Many of them are just noncewords, such as fruitfullity (see appendix below), and speakers (and linguists) widely disagree as to the exact "shade of meaning" involved. One particular satellite word situation notably concerns us in word formation. This is the derived word which co-exists with the established word of the overall lexical paradigm. Consider poor-ness. This word does exist by definition, as the appropriate derivational

The synchrony-diachrony

division in word-formation

417

paradigm exists (see section V.3 below). However, it co-exists with the established word poverty in the lexical paradigm. It is a typical satellite word of the latter, being mainly heard (if it is heard) in a metaphorical sense, such as the poorness 'poor performance' (of a student) in opposition to his poverty 'lack of wealth (even spiritual wealth)', but speakers (including linguists) characteristically disagree on what exactly poorness does mean (if it is accepted as a word at all). Generalizing on the case of poorness, we claim that co-existence with an established synonym in the lexical paradigm reduces the derived word to satellite status (so much so that the satellite word may not even be recorded anywhere and be rejected by native speaker informants). Diachronically, however, the status may be reversed, so that the former satellite word acquires established status and conversely. For instance, Brummagem was at one time the established name of Birmingham (England). This consideration again solves a number of empirical problems: (i) The widespread deprecation of the derivatives admirable, compärable, referable in the read : readable paradigm is due to their status as satellite words, competing with their established synonyms ädmirable, comparable, referable in the lexical paradigm. The fact that the latter are, in turn, historical derivatives (see section V.l [ii] below) is not relevant in this context. (ii) The state nouns corresponding to rich and poor are wealth and poverty respectively in the lexical paradigm. The derivatives richness, poorness are just satellite words in the lexical paradigm with their special meanings and connotations. (iii) Competing pairs such as the nouns Am.E. (postal) äddress, (presidential) address are historically interpreted in such a way that the noun äddress is a member of the permit : permit paradigm, so are the nouns research, dispute, exploit and the verbs input, Am.E. (Southern) influence, ally, Hb.E. commint. Conversely, the nouns address, research, etc., are not derivatives, but pre-established synonyms of äddress, research in the lexical paradigm. In fact, the permit : permit paradigm did not yet exist in the seventeenth century (judging from Shakespeare's and Milton's metre (Kökeritz 1953: 335)), but it was established by the eighteenth century (judging from Alexander Pope's verse (Günthner 1978)). The fact that derivative nouns such as dismay have, as far as I know, not yet been observed is due to its co-existence with the preestablished synonym dismäy in the lexical paradigm.

418

Herbert

Pilch

It is, consequently, misleading to claim that a word like äddress has shifted its stress as compared to the earlier form address. In fact, äddress is a new word. It is derived in the permit : permit paradigm. As far as the lexical paradigm is concerned, it co-exists with the established (non-derived) word address, and various meaning differences, regional differences etc. have been observed for the two words. 3. Polysemy. Polysemy is a fact of linguistic synchrony. We recognize at least four different types of polysemy: (i)

whole : part (Pilch 1976: 95-96), e. g., America (1) 'the whole Western hemisphere', (2) 'the United States'.

(ii) primary : metaphorical, e. g., lemon (1) 'kind of citrus fruit', (2) 'ramshackle car'. (iii) primary : metonymical, e. g., council (1) 'advice', (2) 'person(s) who give(s) advice'. (iv) primary : name of established concept, e. g., trade wind (1) 'wind connected with trade', (2) 'the steady winds blowing between the 10th and 30th degrees of latitude' (cf. footnote 8). Polysemous interpretation is a fact of language use. Whenever a word is used, it refers to a specific denotatum (real or fictitious), while its linguistic meaning is tantamount to a class of denotata (Hervey 1980: ch. 3). The latter cannot be a closed class, as many occurrences involve new denotata. Consequently, it is a synchronic fact that a word is made to refer to an open class of denotata. In this sense, the four types of polysemy described above describe the modes in which a given word is re-interpreted as referring to different subsets of denotata. As morphological derivatives figure in lexical paradigms, we must allow for polysemy when interpreting those uniform differences of meaning envisaged by our model. For instance, the uniform difference of meaning as postulated for the paradigm envisages the paradigmatic meaning, i. e., 'agent' (animate or inanimate) for the tatpurusa like taxpayer, 'subclass of b connected with a' for the determinative compounds ab such as street car, where street (a) determines car (b). But then, the determinative compound trade wind

The synchrony-diachrony

division in word-formation

419

is conventionally re-interpreted to have not just its paradigmatic meaning as a determinative compound 'wind connected with trade', but to specifically denote those winds blowing between the 10th and 30th degrees of northern and southern latitude. 8 Similarly, the waiter is conventionally interpreted not just as any person who waits, but specifically as one who waits on patrons in a restaurant, the street car not as 'any car in the street', but specifically as a 'public service vehicle on rails ...'. It has been argued that, as a result of such specific interpretations, the derivatives change their meaning and that this is a diachronic change (Mulder — Hervey 1980: 141; Hervey 1980: ch. 3). However, those specific interpretations are very often not fully conventionalized. In fact, as long as the derivational paradigm exists, the paradigmatic meaning can always be recovered and be re-interpreted in a new way. For instance, a recent article on the prospects of American foreign trade in "The Christian Science Monitor'''' was headlined American Trade Winds. In this instance, the term was reinterpreted metaphorically as 'trends of American foreign trade'. Similarly, we could pun on the polysemy of waiter (1) 'who waits' (e. g., for someone failing to arrive on time), (2) 'who serves patrons in a restaurant'. In fact, it is (we believe) precisely by specific polysemous interpretation that a derivative becomes established as the name of a conventional concept. For instance, the German determinative compound das Waldsterben has recently been given a conventional interpretation as the 'dying away of forests as recently observed in Central Europe', and a heated public debate is going on about its causes, remedies, etc. The word is now part of the German lexicon in this specific sense. Does this mean the word did not exist before? It certainly did exist as a member of a derivative paradigm, even though it did not designate a specific concept. It only had the paradigmatic meaning of 'action noun' as given by its derivational paradigm, as in: Intransitive Predicate

Action Noun

Kinder turnen Pferde rennen Blumen welken

das Kinderturnen das Pferderennen das Blumenwelken

Accepting the polysemous interpretation of paradigmatic meaning as a synchronic fact, we solve several further problems:

420

Herbert

Pilch

(i) The fact that some tatpurusa conventionally designate persons, others specific tools is a matter of polysemous interpretation. It is by such an interpretation that the word church goer has been established to denote not just 'a person who goes to church', but 'one that goes more often than I think he should'. By the same token, a gas guzzler is not 'any person or object that guzzles gasoline', but the established designation of 'an automobile with a low gas mileage'. Similarly, the skyscraper is usually interpreted metaphorically either as 'a specific type of topsail' or, alternatively, 'an exceptionally high-rise building', the chainsmoker metaphorically as 'one who smokes unceasingly'. Yea, the recognized interpretation may contradict the paradigmatic meaning, as in the case of the shoemaker who does not make shoes; the British public school which is, in fact, a private school; those papers that I shipped to Blazejewko, while, in fact, they were carried not by ship, but by train. (ii) By the same token, an invite is conventionally interpreted as an informal social function to which guests have been invited, a pervert as a person who has been perverted, a permit as a piece of paper documenting that something has been permitted, an intern not as a person who has been interned. The paradigmatic meaning of the derivatives is much less specific. It could be paraphrased as 'something connected with the action of the verb'.

IV. Bilingual derivation So far we have considered monolingual paradigms in the sense that the primary and derivative range over the vocabulary of the same language. However, this is not always the case. Consider, for instance, the paradigm of inflammatory diseases in English (and other European languages). Its primaries range over the GrecoLatin (not the English) vocabulary of European (not Classical) Latin: Greek Greek Greek Greek Greek Latin

laryngg-os (gen.) brogch-os (gen.) gastr-os (gen.) hepat-os (gen.) ot-os (gen.) periton-eum

Engl, Engl, Engl, Engl, Engl, Engl,

laryng-itis bronch-itis gastr-itis hepat-itis ο t- it is periton-itis

The synchrony-diachrony

Latin Latin Latin

appendic-is tonsill-ae sinus

division in word-formation

421

Engl, appendic-itis Engl, tonsill-itis Engl, sinus-it is

The paradigm is sometimes extended to the native vocabulary, as in golf-itis, headlin-itis, but those derivatives still have the status of noncewords or satellite words compared to the established modes of expression. Similarly, German has a deverbative paradigm of nouns in -in, such as: teach love drink

: teach-in : love-in : drink-in

The primaries of this paradigm range over the class of English (not German) verbs. The derivatives are, however, German (not English). They have conventional meanings in German (not in English). Their stress pattern is not the same as of the corresponding English words, witness Gm. teach-in, Engl, tiach-in. Even so, the derivatives have the status of noncewords and satellite words. In one type of diachronic change, such paradigms are extended to the native vocabulary not just for nonce-formations, but generally. This has apparently happened to the write : re-write paradigm (Pilch 1976: 156). At one time, its range was limited to the Latinate vocabulary, as is still claimed by Quirk (Quirk 1972: 990). Today, however, it ranges routinely over the native vocabulary. By the same token, the -itis paradigm has become established by extension, in Colloquial German, with many German verbs and corresponding derivative nouns such as Heuler-itis, Heirater-itis. Similarly, it is often claimed that the permit: permit paradigm ranges just over Franco-Latin prefixed verbs. This is certainly no longer the case (if ever it was), witness the nouns torment, ferment, the verb input. The model of bilingual derivation solves the problem of the etectric : electricity paradigm. The primaries of this paradigm range just over the Latinate vocabulary, and it uses not any "deep structure" to convert the final /k/ of -ic into the /s/ of -icity, but just the conventional English school pronunciation of Latin. Consequently, we find in the paradigm derivatives such as analyticity, (certainly not a word borrowed from Latin), even basicity as a

422

Herbert

Pilch

derivative from the chemical term basic (antonym of acid), but not *basicity as a derivative from the derived word basic 'fundamental' of the general vocabulary.

V. Diachronic morphology Reconsidering morphological analysis from the point of view of diachrony, we infer from our definitions the following theorems: 1. Closure of paradigm. Every particular paradigm exists as long as it is productive. This is both a necessary and a sufficient condition of its existence. Once it ceases to exist, however, those of its members which have become established as designating a specific concept (cf. section III.3 above) survive as individual words. For instance, the English verb set survives to this day. Once it was a derivative of sit in a paradigm of causative verbs (Pilch 1970: 130). Such a survivor is called a historical derivative. In fact, the verb set belongs to a closed set of historical derivatives, all of them derived from a ProtoGermanic paradigm of causative verbs which has been defunct for a long time. This closed set also contains the pairs lie : lay, rise : rear, drink : drench, sing : singe. We call such a set a residual paradigm. This theorem solves some of our perennial problems: (i) The words forgiveness, witness are historical derivatives in a deverbative paradigm of Proto-Germanic 9 . Cognate residual paradigms survive to this day in German and Dutch, witness: Gm. versäumen argern vorkommen Du. begraven ontvangen

:

Versäum-nis Ärger-nis Vorkomm-nis begrafe-nis ontvange-nis

(ii) The word admirable is a historical derivative of the Early Modern English paradigm accept : äcceptable, which involves a stress shift (Kökeritz 1953: 337). It belongs to this residual paradigm today together with referable, preferable, comparable, ädorable, reputable, revocable. (iii) The tatpurusa with zero-suffix such as chimney sweep, presumably, belong to a residual paradigm which was productive in Old English, forming both masculine and feminine «-stems (Pilch 1970: 112):

The synchrony-diachrony

division in word-formation

423

OE. beagas gifan : beag gifa (Bwf 1102) run witan : run wita (Bwf 1325) freoöe wefan : freoöu webbe (Bwf 1942) 2. Loss of primary. Suppose a particular primary word becomes extinct (while the paradigm continues to exist). This does not necessarily imply that the corresponding derivative should also become extinct. On the contrary, once established (cf. section III.3 above), the latter survives, conserving all the characteristics of its derivative status apart from the non-existence of the former primary. This theorem solves the problem of soothsayer. Its primary say sooth did exist up till Early Modern English, but has since become extinct. 3. Chronology. Once a set of primaries exists in a particular paradigm, the corresponding set of derivatives exists by definition. This follows from the definition of existence above (section V.l). Consequently, whether or not a particular derivative (say the verb input) exists and since when it exists, does not hinge on the chronology of records, but on the range of the paradigm. As we have seen (section II above), the very existence of an affix is guaranteed by its occurrence in a particular paradigm, not by its etymology. On this score, I should take issue with Professor Kastovsky. His paper in the present volume is pivoted on "affixes" as defined by etymology rather than by derivational paradigm — even to the point where one affix simultaneously belongs to different inflectional paradigms and stem classes. This is the same criticism as leveled by Hans Marchand against Herbert KozioPs "Englische Wortbildungslehre", and it appears to apply not only to Herbert Koziol. 4. "Lexicalization". Once a derivative paradigm ceases to exist, the paradigmatic meaning of the historical derivatives can no longer be recovered and re-interpreted polysemously. For instance, the paradigmatic meaning of spinster 'female who spins' is forgotten today. The word is just used with its established metonymical meaning of 'unmarried woman'. It has become a taboo word, for that matter, being replaced, in respectable usage, by its gloss (or should we take it as a satellite?) 'unmarried woman'. Still, it remains a historical derivative in a residual paradigm. This applies equally to historical derivatives whose corresponding primary has dropped out of the paradigm (cf. section V.2 above). For instance, the

424

Herbert

Pilch

paradigmatic meaning of ladybird 'flying insect associated with Our Lady' is forgotten today. What survives is just the established interpretation as a particular species of beetle. The same is true of trade wind (cf. footnote 8). Established derivatives sometimes "establish" not only special semantic interpretations, but also special morphophonemic alternants, so that even the phonological relation between the primary and the derivative is no longer the same as in the paradigm. Wellknown instances are the former determinative compounds shepherd, vinyard, boatswain, hussy, for which the enclitic pronunciation has become established today. The so-called karmadharaya compounds are attributive groups which have both an established semantic interpretation and a special stress pattern, namely forestress, witness the (metaphorical) diimb waiter, white lie, the (specific) White House, blackbird, First Street. The blackbird (and so on for the other examples) is, we take it, a special case of the paradigmatic black bird — by the same token as the topical Waldsterben is a special case of 'dying forest'. In fact, even the paradigmatic black bird occurs with either stress pattern (as in a bläck bird, not a green one), and some black birds are blackbirds. True, "albino blackbirds" (we understand) exist (Mulder — Hervey 1980: 130). The apparent contradiction is part of the recognized interpretation (cf. section III.3 above). Should we therefore treat blackbird as a compound, black bird as an attributive group? This is not necessary — by the same token as the topical Waldsterben is still the same word as the paradigmatic Waldsterben. On top of it, such treatment reduces us to the absurd claim that morphological status is defined by a phonological property (namely stress) — as is done even by Marchand (1969: 22) — and, consequently, that the forestress group First Street is a compound, but the level stress group First Avenue is an attributive group. Is there any morphological difference between the two? There may be excellent reasons to treat both expressions as compounds or neither, but there can be none to treat them differently. In fact, what is more adequate to both than the category of compound is Andre Martinet's category of syntheme. Both expressions qualify as synthemes (Martinet 1979: 19, 233). What we propose is to take the first street (on our right) and the street name First Street not as two different morphological entities, but as phonological alternants of the same morphological entity, even though with different syntactic properties. The forestress

The synchrony-diachrony

division in word-formation

425

(sometimes, not always) provides a phonological cue (Pilch 1984: 201) to a specific polysemous interpretation which, syntactically, turns the expression into a syntheme. This view easily accounts for the fact that either stress pattern can be observed for either meaning (if the truth were known), as in "take the first street on the right, not the second". This view can also be reconciled with the fact that not every particular interpretation of attributive groups is cued by forestress, but just the odd one, for instance the social worker, the bläck märket, but not social insurance, the Black Maria, either, phonological alternant is commonly heard for the fast woman. All these expressions are synthemes when they have their specific established meanings. It is for this reason that we have put the word "lexicalization" in inverted commas above. What is involved is not the establishment of a new word or lexeme, but the conventionalization of a specific polysemous interpretation and, in some instances, of a specific allomorph. 5. Range extension. Paradigms of word formation can extend their range. This is a fact of linguistic synchrony, as long as it produces just nonce-formations. A case in point is the verb coastal navigate noted above. The standard range of the paradigm is the tatpurusa nouns in -er, -ing, as in baby sitter : babysit. The extension to the attributive group coastal navigation : coastal navigate (as recorded) is just a nonce-matter. It often provokes adverse comments from native speakers. However, the extension becomes a fact of linguistic diachrony once the range is extended to a new class of primaries (rather than just the odd new instance). For instance, the tatpurusa paradigm has also been extended (presumably, fairly recently) to the new class of co-ordinate predicates, witness: manufacture cars and bicycles : car and bicycle manufacturer build and repair bicycles : bicycle builder and repäirer Has the word wilderness arisen through a similar productive extension of the kind : kindness paradigm? This is a doubtful proposition, as the extension involves just one instance, not a class of instances. In fact, wilderness very specifically involves the morphological re-interpretation of the Old English word wildern 'wilderness' as in wilder-ness — just as Fr. groseille has been re-interpreted as goose-berry in English. Those may have originated as nonceformations and later have become "lexicalized".

426

Herbert

Pilch

6. Revival of paradigm. Occasionally, attempts have been made to revive the productivity of a residual paradigm. For instance, new coinages have been made time and again since the sixteenth century in the residual paradigm of de-adjectival abstract nouns in -th (as deep : depth). The coinage was of deverbative abstract nouns such as tilth 'tilled land' (Jespersen 1961: 441-442). The only survivor today is growth. Usually, no more is involved in such revivals than just a few items and/or nonce-formations. Probably, the bellhop (cf. section I.l(v) above) is one of them. Theoretically, however, the closure of a paradigm does not rule out its re-opening later. An observed instance is the recent revival of the vme-paradigm (as in language-wise, time-wise, education-wise) which earlier was limited to clockwise and its antonym counterclockwise10. This revival is not limited to the odd new coinage, but it has a very wide range indeed, including the class of all substantives. Is this an entirely new paradigm or the revived paradigm of clockwise? This boils down to a question of mere terminology. 7. Reconstruction. In summary, we recommend discrimination between synchrony and diachrony in such a way that only members of productive paradigms are morphologically analyzable in the synchronic domain, e. g., kind-ness as a member of its paradigm, but not forgiveness. Diachronic analysis, on the other hand, applies to residual paradigms and historical derivatives, such as forgive : forgiveness. They can be reconstructed into earlier open paradigms, in the present instance into the same deverbative paradigm as Gm. verloben : Verlöbnis. The Present-Day English word forgiveness contains the suffix -ness strictly in the diachronic sense of being reconstructible into the deverbative -«ess-paradigm of Proto-Germanic. By analogy, the same analysis applies to inflexions 11 . The segmentation of child-r-en into two or three morphemes is acceptable strictly as a diachronic reconstruction first into the productive -en plural paradigm of Middle English, then into the thematic -rparadigm of Proto-Germanic. The form is not segmentable synchronically in English. Similarly, the English word humanity contains the suffix -ity etymologically, not synchronically. More precisely, its Latin etymon human-itatem (or is it French humanitel) contains the element -itat. The latter has also been separately borrowed into English. In English, this element figures as a suffix in the electric : electric-ity

The synchrony-diachrony

division in word-formation

All

paradigm. So this is a diachronic relationship via Latin morphology. The counterhypothesis to the effect that the English pair human : humanity belongs to the same paradigm as electric : electric-ity is weakened by the unique semantic relation of the former pair. While electric-ity means 'the state of being electric' (and so on for the other members of the paradigm), humanity means not 'the state of being human', but 'all men collectively'. True, this is a metonymical meaning of Latin humanitas, yet it is the primary meaning of the English word humanity. By the same token, both decide and decision are loanwords from French. They are not members of any productive English paradigm. Consequently, all those deep structure rules which derive decision from decide (and so on for the other members of the alleged paradigm) are simply superfluous. Similar diachronic analysis applies to exist : existence, neglect : negligible, Arab : Arabic, etc. All those spurious derivational paradigms and the complicated deep structure rules which they conjure up stem from the simple failure to adequately distinguish between synchrony and diachrony.

Appendix: Criteria of productivity 1. Limit on paradigm membership. How can we tell whether a given paradigm is open (i. e., productive) or closed (i. e., unproductive and, presumably, residual)? On principle, the criterion is implicit in the definition: A paradigm is open, if and only if it has unlimited membership (see section II.1 (iv) above). Otherwise it is closed. The notion of limit is, in this context, the usual settheoretical notion which implies a lower limit (cf. Hockett 1965: 48). The lower limit on the members of the paradigm implies that these members can be enumerated, i. e., listed. Suppose we wish to check on the hypothesis proposed above that the write : rewrite paradigm is open. As we know from epistemology, a hypothesis can never be definitely "proved", it can only be weakened (invalidated) by counterevidence (cf. Mulder — Hervey 1980: 30). The counterevidence to the effect that the paradigm is closed would, by definition, have to consist in an enumeration of all the members (and only the members) of the paradigm. Our productivity hypothesis affirms they cannot be enumerated. Should it turn out that they can, the hypothesis would be invalidated.

428

Herbert

Pilch

Our productivity hypothesis about the English re- verbs affirms, in its most general version, that the left-hand column of the paradigm ranges over the (open) class of English verbs. We have successively corrected this hypothesis, as our opponents suggested specific subclasses of verbs which, as far as we know, do not enter the paradigm, and we have mentioned those subclasses above (section II.1), thus narrowing the range of the paradigm. Conceivably, further modification of this sort will be required. This does not, however, affect the openness of the paradigm as such. Conversely, suppose we wish to invalidate a hypothesis to the effect that a particular paradigm is closed, for instance the deverbative paradigm forgive : forgiveness. Suppose, it should turn out that one more pair exists, even two or three. The hypothesis would thereby not be invalidated, as two is still a limited number of members, and so is three and four. All that would have to be corrected is the list of members. What we would have to specify, in order to weaken the closure hypothesis, would be an open range of the paradigm such as "the class of English transitive verbs" (as in section II. 1 (v) above), and this range would have to be plausible as an inductive generalization on observed instances (see section (3) below). "Enumerate all the members and only the members". The latter stipulation is implicit in the notion of lower limit. It is to forestall the interpretation of limit as 'upper limit'. Otherwise, my imaginary opponent might propose an unrealistically high limit on the paradigm, say two million members (or more), then propose an unrealistically high limit on the length of English verbs as written (say 100 letters maximum). He could then instruct a computer to enumerate all possible letter sequences of the Latin alphabet starting with re as the constant first two letters and a minimum of three letters, maximum of 100 letters. The printout would, presumably, contain all the English re-verbs which we know, and in that sense list them completely, but it would contain not only those re-verbs. Consequently, such a list is not acceptable as counterevidence. 2. Degrees of productivity. An open paradigm is not necessarily one with very many known members. Consider the agent nouns in -nik such as beatnik, nudenik, peacenik. The paradigm arose after the sputnik scare of 1957 by morphological re-interpretation. We would hypothesize that it is open and that it ranges over all monosyllabic nouns, adjectives and verbs of English. Sure, I could

The synchrony-diachrony

division in word-formation

429

not list even (say) 100 members, but this is one of those irrelevant upper limits. What is relevant in this context is my failure to specify a lower limit. A closed paradigm is not necessarily one with very few members. Consider the tatpurusa paradigm chimney sweep as described above. It contains quite a few more members, namely bootblack, shoeblack, bellhop, toothpick (Marchand 1969: 76-77). Their number is (we believe) limited. We would hesitate to accept into this paradigm agent nouns like pastry cook, industrial claims judge, life guard, but propose to analyse the latter as determinative compounds, semantically as 'a particular kind of cook, judge, guard'. If those were tatpurusa (rather than determinative compounds), it would be inexplicable that they should be productive only with verbs which are accompanied by homophonous agent nouns (such as cook), but not by others (such as write). On the principle of Occam's razor, we therefore interpret the primary cook of pastry cook etc., as the agent noun, not as the homophonous verb. What, then, about "degrees of productivity"? Our definition rules them out. A paradigm can, by definition, only be either open or closed, not more or less open or closed. Those "degrees of productivity" (as the term is commonly understood) envisage not the closure of the paradigm, but the frequency with which members of the paradigm are observed. For instance, we would propose the hypothesis that the diminutive paradigm pig : piglet is productive in English today. This hypothesis could be invalidated by a complete list of members such as kinglet, princelet, starlet, frontlet etc. To my knowledge, such a complete list has not been presented. On the other hand, we have certainly observed fewer diminutives in -let than duplicate re- verbs. We do not object to speaking of different degrees of productivity in this sense, nor does this sense conflict with our theorem that paradigms can be only either closed or open. 3. Evaluation of evidence. Many linguists use just the conventional parts of speech to specify the range of paradigms, for instance subclassifying compounds as Ν + N, Adj. + N, Pron. + N, etc. *This is inadequate in many cases. For instance, the specification Pron. + Ν is insufficient for Layamon's compound al-drihten (cf. Dr. Sauer's paper in this volume), as there is no evidence (apart from this single instance, which may easily be a loan translation of Greek pantokrätör via Lt. omnipotens) that such a paradigm of compounds exists in Middle English, i. e., one ranging over the class

430

Herbert

Pilch

of Middle English pronouns. The hapax legomenon cel-drihten is surely insufficient evidence to inductively infer the existence of such a paradigm. Disagreement often arises whether or not certain counterevidence should be recognized as such. Take the word tangibleness which I would claim is a member of the -ness paradigm by definition, as this paradigm ranges over the class of English adjectives (cf. section I.l(vi) above). Professor Anshen has told this conference the word does not exist. He bases his counterclaim on the responses of native speaker test populations. I do not question his expertise as far as those tests are concerned. I could, of course, invoke the fact that tangibleness is an entry in Webster's International Dictionary and doubtless in other dictionaries. In fact, however, I did not even check on this entry before proposing the range of my paradigm. What would induce me to narrow this range excluding tangibleness, would not be disagreement over specific single words, but the recognition of specific subclasses which are not members of the paradigm (as in the case of the re- verbs above). In fact, we have concluded above (section V.l) that any derived word exists, if the appropriate derivational paradigm exists — regardless of dictionary entries, test populations and the like. We cannot, of course, propose a general algorithm which will relieve the linguist of the burden of his professional discretion. Nobody except the linguist himself can decide what sort of evidence he will or will not accept and how he will interpret this evidence. His choices will be motivated and arguable (if informed by his professional discretion) (Pilch 1968: 23-24). This situation is the same as in all other branches of science. For instance, even when operating with those "objective" oscillograms and spectrograms, the researcher has to discriminate between good evidence and artefacts. There is, on the one hand, a sizable body of evidence on which we probably all agree, say the write : rewrite paradigm above or the non-existence of the word *cold-th as an analogue of warmth. On the other hand, there are many less clearcut, arguable cases. Ultimately, the investigating linguist has to use his discretionary judgment. He will base it (hopefully) on extensive professional experience. It is an illusion to believe we could be "truly objective" by just administering tests or relying on dictionaries (rather than use our discretionary judgment). Those are modes of "objectivity" which, in effect, mean that we are relying on somebody else's discretion (which may be even less "objective" than ours).

The synchrony-diachrony

division in word-formation

431

What about the word fruitfull-ity'} I would claim the word does not exist, as the primaries of the appropriate paradigm range over the vocabulary of Latin adjectives (see section IV above), and fruitful is not one of them. True, I have heard one member of this conference pronounce fruitfull-ity, and he even was a native speaker of English. I would interpret this as a nonce-word. By definition, nonce-words are not part of the vocabulary. At the same time, the speaker concerned was extending the range of the paradigm exactly in the sense in which we have used the term 'extension' above (section V.5). Such extension is, of course, a common phenomenon which we can observe all over the place notably in punning, in joking, in erroneous usage. Such observations are (we take it) not sufficient to establish productivity. Surely, nobody would seriously argue that fruitfull-ity is now an English word in the same sense as (say) tangibil-ity. Consequently, the hapax legomenon is like the proverbial one swallow that does not make a summer. True, what starts as a nonce-extension may snowball into a general extension, so the earlier range of the paradigm becomes extended at a later stage (cf. Gm. Heuler-itis, section IV above), but this is a diachronic, not a synchronic perspective. Synchronically, a nonce-word does not necessarily qualify as part of the lexicon.

Notes 1. The classic statement is by Ferdinand de Saussure 1949: 117. 2. The problem has been widely discussed in generative and natural phonology and has been solved, within this framework, by the "velar softening rule". 3. Even Chomsky — Halle (1967: 181) have reservations about this derivation. 4. These statements differ in theoretical status: (i) and (ii) arrange the protocolized data, (iii) empirically motivates the arrangement, (iv) is an inductive generalization, (v) presents definitions for morphological analysis. 5. An expanded model to analyse thematic formatives and amredita compounds has been presented by this writer (Pilch 1967: 160-178). 6. Professor Dressier, in his paper in this volume, applies the terms "suppletion" and "derivation" to all pairs of the lexical paradigm (not just to the morphological sub-paradigms). 7. I take this notion from Walter von Wartburg (1962: 145). 8. Dr. Jerzy Welna (Warsaw) calls my attention to the fact that the primary trade of trade wind has the now defunct meaning of 'steady track, course'. This makes trade wind 'steady wind on the transatlantic East-West route' a historical compound (cf. section V.2 below). However, the polysemous pun on American Trade Winds as here reported shows (I take it) that the word can synchronically be re-interpreted as 'wind connected with trade', i. e., that the paradigmatic meaning has been re-interpreted to fit the productive paradigm of determinative compounds.

432

Herbert Pilch

9. Derivation from the past participle forgiven is doubtfully suggested by Marchand 1969: 336. Dr. Günter Kotzer (Munich) calls my attention to the co-existence of two -new-paradigms in Old English, one apparently derived from the present stem (as forgife-ness), the other from the past participle (as forgifen-ness). Presumably, these two paradigms belong to different dialects of Old English. Our reconstruction involves only the former paradigm. The word drunkenness is, presumably, a survivor of the latter, but forgiveness of the former. 10. Formations like anywise, edgewise are still listed in dictionaries. In actual usage, they were replaced a long time ago by similar formations in -ways. 11. My attention has been called to this problem by Professor Claude Guimier (Caen).

References Chomsky, Noam — Halle, Morris 1967 The sound pattern of English (New York: Harper). Günthner, Cornelia 1978 Konversion und Akzentverschiebung im Englischen des 18. Jahrhunderts (Freiburg i. Br.: Staatsexamensarbeit). Hammarström, Göran 1966 Linguistische Einheiten in der modernen Sprachwissenschaft (Berlin: Springer). Hervey, Sandor 1980 Axiomatic semantics (Edinburgh: Scottish Academy Press). Hockett, Charles 1966 The state of the art (Janua Linguarum, series minor 73) (The Hague: Mouton). Jespersen, Otto 1961 Modern English grammar (London: Allan & Unwin, reprint). Kökeritz, Helge 1953 Shakespeare's pronunciation (New Haven: Yale University Press). Lüdtke, Helmut 1980 Kommunikationstheoretische Grundlagen des Sprachwandels (Berlin: de Gruyter). Marchand, Hans 1954 "Notes on English prefixation", Neuphilologische Mitteilungen 55: 294-304. 1969 The types and categories of present-day English word formation (2nd edition) (Munich: Beck). Martinet, Andre 1979 Grammaire fonctionnelle du franfais (Paris: Presse universitäre de France). Mulder, Jan — Hervey, Sandor 1980 The strategy of linguistics (Edinburgh: Scottish Academy Press). Pilch, Herbert 1967 "Modelle der englischen Wortbildung", Festschrift Hans Marchand, (The Hague: Mouton), 160-178. 1968 "Linguistische Existenzurteile", Folia Linguistica 3: 13-28. 1970 Altenglische Grammatik (Munich: Hueber). 1976 Empirical linguistics (Munich: Francke). 1984 "Intonation als kommunikative Funktion", Festschrift F. Link, ed. U. Halfmann et al. (Berlin: Duncker and Humblot), 195-210.

The synchrony-diachrony division in word-formation

433

Quirk, Randolph et. al. 1972 A grammar of contemporary English (London: Longman). Saussure, Ferdinand 1949 Cours de linguistique generale (4th edition) (Paris: Payot). v. Wartburg, Walter 1962 Einführung in die Problematik und Methodik der Sprachwissenschaft (2nd edition) (Tübingen: Niemeyer).

ELIZABETH Μ. RIDDLE

A historical perspective on the productivity of the suffixes -ness and -ity

The notion of morphological productivity has been the focal point of a number of recent studies of English lexical structure. One pair of suffixes, -ness and -ity, have enjoyed the limelight as ostensibly synonymous affixes performing an identical function but differing with regard to their potential for affixation to particular bases. The factors responsible for the choice between the two suffixes for any one base have thus been a topic of special interest. Aronoff (1976: 51-52) posits an abstract feature [ ± latinate] to distinguish the bases eligible for -ity affixation from those allowing only -ness. Monomorphemic words of latinate historical origin are considered [— latinate] since they normally do not allow -ity affixation. He states (1976: 26) that the productivity of an affix must be evaluated with respect to particular bases. For example, -ity is said to be more productive with bases ending in -ic and -He than -ness is, and -ness is more productive with -ive and -ous bases than -ity is, although the other suffix is not impossible in either case. Aronoff — Schvaneveldt (1978) claim that the greater productivity of -ness on -ive bases reflects a preference for word boundary affixes (# ness) over formative boundary affixes (4- ity). Cutler (1980, 81) argues that preservation of the phonological transparency of the base is the primary factor determining the choice between the two suffixes, but Anshen — Aronoff (1981) appear to supply counterexamples. Randall (1980) concludes from tests in which subjects rated nonce -ness and -ity formations that native speakers of English distinguish between classical and native morphemes, although the judgments of her subjects may have been affected by familiarity with some of the bases of the nonce words as well.

436

Elizabeth

Μ.

Riddle

Aronoff (1982) distinguishes between actual and potential forms, claiming that the more productive a pattern is, the less likely a speaker is to be able to distinguish potential from actual words formed in that pattern. Words formed on the model of less productive word formation patterns, on the other hand, can be judged more easily, and such words often have special arbitrary meanings. Moreover, they may be more frequent as actually occurring individual words than words of less marked form. Aronoff claims that such less productive patterns are often used for the purpose of foregrounding. Anshen — Aronoff (1984) compare the greater frequencies of -ity on -ible bases and -ness on -ive bases and conclude that the number of existing words of a given type, their frequency in discourse, and their date of introduction into the language affect the productivity of a particular suffix for a given base. Each of the works mentioned above discusses the productivity of -ness and -ity under the assumption that the two suffixes are synonymous. Randall, in fact, ruled out meaning as a possible factor in the choice between the two suffixes by instructing her subjects to judge words purely on the basis of phonological acceptability and not on their meanings. Authors of more general works discussing English morphology, such as Maetzner (1874: 449-450, 456), Nesfield (1922: 386, 404), Koziol (1967: 57-58; 1972: 203-204, 258-259), Jespersen (1942: 449) and Marchand (1969: 227-228), also treat -ness and -ity as synonymous. Marchand recognizes that in general, suffixes appearing to be synonymous only partially overlap and that "each one suffix has a different totality of semantic features." (1969: 227). By this he means that "no two combine alike formally or with the same intellectual or emotional connotation" (Marchand 1969: 227) and "any one sign is determined by the totality of combinations in which it may occur and which cannot be the same as that of any other sign" (1969: 228). He claims near interchangeability for certain pairs, and specifically pairs of the form -ablenessI-ability, however, and proposes no specific (and no denotative) distinction between -ness and -ity per se. Aronoff (1976: 56) remarks that pairs such as humanness and humanity are not synonymous and that the current meaning of the latter is the result of drift. He also states (1976: 38-39) that the semantics of certain sets of base + affix (e. g., Xousness) is more coherent than that of others (e. g., Xity). In other words, the

A historical perspective

on the productivity

of the suffixes -ness and -ity

437

meanings of Xousness words are more predictable than those of Xity words which are related to -ous words. No one, however, has considered the choice between -ness and -ity for a base to be a function of a difference in meaning between the suffixes themselves. In this paper, I will argue that -ness and -ity are semantically distinct for a large portion of the lexicon. That is, the suffixes themselves have different meanings when occurring on many bases, but the distinction is not realized on all bases. I will suggest that this is the result of an unfinished historical change gradually affecting the lexicon, and that cases of words which do not exhibit the distinction can be explained on other grounds. In addition to the phonological transparency factor noted above, pragmatic, social and historical factors sometimes take precedence over the semantic factor, thus causing the generalization to be but a partial one. In addition, words which have not yet been affected by the change also appear as exceptions. I conclude that the productivity of lexical elements cannot be satisfactorily discussed without reference to meaning, context and history. First I will argue for the semantic distinction between -ness and -ity and then proceed to the historical discussion. Basically, the distinction between -ness and -ity is as follows: -ness tends to denote an embodied attribute or trait, while -ity tends to denote an abstract or concrete entity. Examples of what I consider to be abstract entities are the names of concepts and situations and of characteristics in the generic sense. For example, an -ity word may refer to a characteristic, in the generic sense, while there is a tendency for the corresponding -ness word formed on the same base to describe an embodied attribute 1 . A discussion of several examples will clarify the distinction. First I will consider minimal pairs. In examples 1-12, the -ness words denote an embodied trait or attribute, while the -ity words denote an abstract or concrete entity. Consider the examples in 1, taken from a single newspaper article: (1)

a. "However, don't call this third-grader a picky eater. She's a selective one, a Feingold diet subscriber, whose hyperactiveness has decreased, her mother says, since she began the program four years ago." (Crider 1982: El) b. "But to date there is no evidence that this type of dietary regime will have any effect on hyperactivity in children." (Crider 1982: El)

438

Elizabeth Μ. Riddle

Hyperactiveness in a. denotes an embodied attribute of a particular child, while hyperactivity in b. names the condition. That is, it denotes an abstract entity. A similar contrast can be seen in (2): (2)

"Instead of asking what reality is, he gave matters a subjective phenomenological twist, italicizing the following question: Under what circumstances do we think things are real? The important thing about reality, he implied, is our sense of its realness in contrast to our feeling that some things· lack this quality." (Goffman 1974: 2)

Reality in (2) is clearly the name of an abstract entity, while realness describes an embodied trait of that entity. Further examples of the entity-trait contrast are provided in (3M5): (3)

Her ethnicity was not a factor in the hiring decision. We are an equal opportunity employer.

(4)

Her ethnicness was certainly a big factor in the director's decision. He wanted someone who personified his conception of the prototypical Greek to play the part.

(5)

The decorations made the restaurant seem very ethnic. In fact, its j * ' ??ethnkTity} s e e m e d

exa

Sgerated·2

In (3), ethnicity refers to nationality or race, an abstract entity, while in (4), ethnicness refers to an embodied trait involving personal physical characteristics. Ethnicity is unacceptable in (5), as ethnic refers to the individual trait rather than a nationality per se. In (6), either ethnicity or ethnicness is acceptable, but the meaning of the sentence differs, depending on which word is used. (6)

The lanterns demonstrated the < f ' ^ j 1 1 1 ! 0 ^ I 0 f the b. ethnicness J restaurant. ^

In a., the lanterns enabled the speaker to identify to which ethnic group the restaurant belonged, while in b. the lanterns evidenced the ethnic (as opposed to non-ethnic) nature of the restaurant.

A historical perspective

ort the productivity

of the suffixes -ness and -ity

439

Notice that the latter form and interpretation are strongly favored in (7): (7)

The lanterns contributed to the restaurant.

! S [ of the >

It is possible to reinforce an embodied trait, but not an independently existing abstract entity such as a nationality. The entity-trait distinction may also be observed in the contrast between brutality and brutalness below: (8)

I can't stand any kind of j ^ ' rft^utalness}' w ^ e t ^ e r ^ comes from rednecks or radicals.

Since it denotes a type of behavior or act (cf. comes from) rather than a personal attribute, brutality is the preferred word in (8). In example (9), in contrast, the occurrence of the phrase he's got provides the appropriate context for the form used to describe personality traits: (9)

I don't want to have anything to do with Fred Martin. He's got a brutalness the likes of which I've never seen.

In (10), either word is possible, but the resulting sentences have different senses: (10)

The {*·

remarks shocked us.

(10) a. focuses on the brutal nature of the remarks themselves, while b. focuses on their utterance as being brutal. In (11), tangibility is preferred because the topic is a concept in the abstract: (11) A: We need tangible evidence. B: Yes, j ^ ' < ^ ^ | g | J | j J n e s s |

be

an

important factor.

When a trait of some particular evidence is described, however, the -ness form is preferred:

440

Elizabeth

Μ.

Riddle

(12) A: Her description of the danger really swayed the jury. B: Its tangibleness was what was so convincing. You could practically taste her fear. Example (13) provides an interesting contrast: (13) Here, tangibility refers to existence itself (an entity) and tangibleness to degree of concreteness in an individual (a trait). The words very and doubt make it clear that existence is the question, and thus, the -ity form is preferred. Finally, consider the pair of phrases in (14): (14) a. "... A Study in Pronominality" b. A Study in Pronominalness

(Campbell 1969)

(14) a. is part of the title of a book on the forms and use of pronouns in a particular language. The hypothetical title in (14)b., on the other hand, promises a (perhaps squishy) treatment of what it means to be a pronoun. To summarize, the examples in (1)—(12) above exhibit a consistent semantic distinction between -ness and -ity words, despite the morphology of the base forms. This poses problems for analyses such as those of Marchand and Aronoff (see also Reichl 1982: 157-239) which emphasize the form of the base as the most important factor in the choice between the two suffixes. Further support for the claim that -ness and -ity display a tendency to semantic differentiation may be found in the fact that there exist semantic groups of words which allow only one and not the other suffix. For example, color words, which describe inherent traits, only take -ness, no matter what their historical origin or morphological or phonological form. Thus, not only do all native monomorphemic color words take -ness (e. g., redness, greenness, goldness), but also all color words of Romance origin. Compare the pairs in (15)—(17). The c. examples are included to show that the b. examples should not be ruled out on phonological or morphological grounds. (15) a. maroonness b. *maroonity

vs.

c.

immunity jejunity

A historical perspective

on the productivity

of the suffixes -ness and -ity

(16) a. ceriseness b. *cerisity

vs.

c.

obesity

(17) a. lavenderness b. *lavenderity

vs.

c.

dexterity

441

Out of a list of some fifty (one-word) color words I have compiled, the vast majority of which are of Romance origin, none allow -ity in contemporary English. One color-related word ending in -ity, i. e., viridity, meaning 'greenness' or 'verdancy' is now either rare or obsolete. Moreover, in a number of the examples cited in the Oxford English Dictionary (henceforth OED), it is not used simply to denote greenness, but rather to describe the flourishing of green plants, i. e., verdancy. Four other attested -ity words, namely vividity, clarity, lividity and pallidity, have or had meanings closely related to the description of a color. However, vividness seems to be currently preferred to vividity, at least in American usage, and the relevant meaning of clarity, i. e., 'brightness', has been lost. Lividness, meaning 'leaden blueness', seems to be preferred to lividity, and pallidness and pallor seem to have replaced pallidity to a great extent. To summarize, color words in contemporary English only take -ness, a suffix which I claim denotes an embodied attribute. Another such semantic group is ethnic names, which describe inherent traits. When characterizing the ethnic nature of an entity, -ness is almost always used, despite the fact that a particular word might have a phonological form which, according to most analyses, calls for -ity, e. g., -ic as in electricity. For example, consider the sentences in (18) and (19):

(18) braids and beribboned headpiece. (19) feeling right about the pill, even though she admitted to its desirability.3 There are a few attested ethnic or language names with -ity, for example latinity. However, this word denotes Latin linguistic construction or style, or else Roman culture or civilization rather

442

Elizabeth

Μ.

Riddle

than the embodiment of an attribute in an individual person or thing. Consider next example (20), which was uttered at the conference in Poland where the papers in this volume were presented, under the following circumstances. An announcement was made by the conference organizer that a reporter was present and would like to interview the three Scandinavian participants. A (non-Scandinavian) member of the audience then said: (20)

"Is he interested in their Nordicity?"

The -ity form was appropriate here because the question was whether the reporter was interested in these people simply because they were Nordic, and not whether he was interested in their personal Nordic characters. The last semantic group I will discuss is count nouns. For the most part, -ness words do not belong to this group, which is to be expected if -ness denotes an attribute. Only a handful of -ness words regularly pluralize, and many of these were well established as count nouns before -ity was borrowed. Some examples are: kindnesses, likenesses, weaknesses, illnesses and sicknesses. Such very common words are often resistant to historical change. I have collected a few unusual plural -ness words from conversation, e. g., carelessnesses, but they seemed to be used self-consciously and at least partly in jest. -Ity words, on the other hand, frequently occur as count nouns, an unsurprising fact if they commonly denote abstract or concrete entities. Some examples are: oddities, rarities, impurities, sensibilities, festivities. Notice that the bases of each of these words can take -ness to denote an attribute. The next type of data I will discuss as evidence for the semantic distinction concerns the impossibility of occurrence of a particular form with -ness or -ity. Some words can take only one of the two suffixes, and this does not depend on either the origin or phonological form of these base, but rather on pragmatic factors. That is, some potential words seem to be ruled out because they would describe strange or rare situations. For example, furtiveness and abortiveness would be used rather than *furtivity and *abortivity. I believe that this is because there is normally no practical use for an individual word denoting a furtive or abortive characteristic in a generic sense. These non-occurring words contrast with captivity,

A historical perspective

on the productivity

of the suffixes -ness and -ity

443

which denotes a conventional concept, as it were. Compare, e. g., in captivity vs. the captiveness of the audience. Some potential -ity words do not actually occur on particular bases because there already exists another word which expresses the meaning that would have been associated with the potential -ity word. This phenomenon has been called "blocking" (Aronoff 1976: 43) and "pre-emption" (Clark - Clark 1979: 798). Some examples of this are given below: (21) a. aggressiveness b. conservativeness c. excessiveness

*aggressivity * conservativity *excessivity

aggression conservatism excess

This concludes my argument that -ness and -ity are semantically distinct with respect to a considerable portion of the lexicon. I will now discuss some apparent problems for my claim that semantics plays a significant role in the choice between -ness and -ity and show that they do not invalidate my analysis. To this end, I will review in more detail some of the works cited above. One factor in the formation of -ness and -ity words mentioned above, and which has long been recognized, is the phonological factor. Affixation of -ity sometimes changes the stress pattern of the base (e. g., sensible, sensibility), while the affixation of -ness does not (e. g., sensible, sensibleness). In addition, affixation of -ity may cause velar softening (toxic -*• toxicity) and trisyllabic laxing (grave gravity). Aronoff (1976: 40) attributes the difference in stress behavior between -ness and -ity words to the purported fact that -ness follows a word boundary and -ity a morpheme boundary. Cutler (1980) argues on the basis of results from four investigations designed to test the productivity of -ness and -ity in the formation of nonce-words that the crucial factor is not the existence of a morpheme or word boundary, but preservation of the phonological transparency of the base. This factor may cause some latinate bases to prefer -ness to -ity, since -ity can trigger trisyllabic laxing and velar softening. In an experimental study where both the -ness and -ity words being tested preserved the transparency of the base (e. g., jejuneness, jejunity), Cutler found no significant preference on the part of her subjects for word-boundary versus morpheme-boundery words. In response, Anshen — Aronoff (1981) argue that phonological transparency cannot be the primary factor in the choice between the two suffixes, since, they claim, words in -ible prefer -ity despite

444

Elizabeth

Μ.

Riddle

the fact that its affixation results in a change in stress (e. g., sensible, sensibleness, sensibility). Cutler (1981) counters that transparency is a gradable concept; it is not a matter of preserving the entire base word intact, but rather just enough of it to make possible correct access of the base word's lexical entry. For example, suppressibility retains the relative syllable prominence and segmental values of the base word suppress up to the recognition point, i. e., where it can be distinguished from other words such as supremacy, but this is not the case with a form like suppressivity. The prediction is thus that subjects judging actual words, possible words and non-words on a lexical decision task will prefer a word like suppressiveness, which does retain the relevant values to the recognition point, to suppressivity, which does not, but will show no preference between pairs such as suppressibility and suppressibleness. This conflicts, however, with Anshen and Aronoffs experimental findings that -ity is preferred about 75% of the time to -ness for -ible words. Despite the -ibility counterexample to Cutler's claim, Aronoff correctly recognizes that phonological transparency is relevant to word formation, and I agree. It plays a role, but is not the prime factor in determining suffix choice. It seems worth commenting, however, on why Cutler's (1980) study yielded the results it did, since it appears to show that semantic factors are not relevant. The problem is simply that Cutler's tests were not designed in such a way as to allow for the possibility of a semantic distinction. In other words, her contexts were insufficient to force a choice on the basis of meaning. Another potential problem for my analysis is the fact that there are apparent preferences on the part of -ness and -ity for particular bases. This would seem to show that meaning is not a factor. For example, as noted above, Aronoff (1976: 36) claims that some bases, such as those ending in -ic and -ile, are more likely to take -ity than -ness, and that other bases, such as those ending in -ive and -ous, are more likely to take -ness, although the other suffix is not impossible in either case. He attributes this to greater productivity on the part of a particular suffix for a particular base. It is thus a purely morphological fact in his view. However, this treatment does not take into consideration the fact that -ic, -ile, -ive and -ous might themselves have semantic significance which contributes to the meaning of a particular word and which might rule out either -ness or -ity for a particular word, depending on the meaning of the base and on the context.

A historical perspective

on the productivity

of the suffixes -ness and -ity

445

A detailed study of suffixes such as -ic and -ile must await another time. I would like, nevertheless, to discuss a pair of -ile words to show that the meaning of the bases in question determines whether -ness or -ity is affixed for use in a particular context. Consider the contrasts in (22) and (23): (22)

Her |gg|||]^y e S S e S |

kept her from following the con-

versation. (23)

Her Ι ? ? ^ ^ ^ ^ !

causes people not to take her

seriously. Notice that senility describes a condition of health, i. e., an abstract entity, while senileness denotes a personal attribute. The corresponding contrast does not occur for the base juvenile, however. I suggest that this is because we have no condition or state in life that we would wish to term 'juvenility'. Except in the worlds of advertising (juvenile furniture) and law (juvenile delinquent), the adjective juvenile means 'childish', and it does not seem culturally relevant to need a special word to denote 'the condition of being childish'. This is not an institutionalized notion, as it were. Another important factor in judging the productivity of the two suffixes with respect to -ile words is the historical factor. For the majority of -ile words listed in two reverse dictionaries, Lehnert (1971) and Walker (1936), the OED lists the noun in -ity as being borrowed as such from French. In a number of cases, the -ity form is represented by citations which are considerably earlier than the adjective in -ile. It is possible that these -ity words became fixed in usage before they were associated with the corresponding adjectives and were thus less open to replacement by -ness forms later on. It is interesting to note here that the transparency criterion would predict a preference for -ness rather than -ity for -ile words, contrary to fact. The semantics and history of all the suffixes in question require further study, but it seems clear that they cannot be ignored. Social factors may also affect suffix choice. Words of Latin and French origin have long been considered to be educated words in English,4 and science in the nineteenth and twentieth centuries in particular has turned to the classical languages to form a large part of its vocabulary. This high prestige seems to extend to the suffix

446

Elizabeth

Μ.

Riddle

-ity, and I suggest that it can take precedence over other factors involved in word formation. In short, I have argued that -ness and -ity are semantically distinct when occurring on many, but not all bases. Preservation of phonological transparency, blocking, lack of need within society for a word with a particular meaning, the circumstances of borrowing, and perceived linguistic prestige also play a role in determining the choice between the two suffixes for any one base. The existence of so many conflicting factors partly explains why speakers may disagree when asked for judgments on any individual items. As will be shown below, other historical factors are also partly responsible for the failure of the semantic distinction to predominate in many cases. I will now offer evidence for the existence of a historical change whereby -ness and -ity have begun and are continuing to diverge, and will suggest that this is a case of lexical diffusion. The first type of evidence concerns the fact that many Old English and Middle English -ness words which denoted abstract or concrete entities have become obsolete. Some have been replaced by words formed with other native suffixes, or else another native suffix which was used simultaneously finally displaced -ness. Examples of these are given below.5·6 (24) a. b. c. d. e. f. g· h. i. j·

wlsness wisdom forseonnes —• foresight forspyllednes -> spilling (waste) gröwenness —• growth wldness —• width strangnyss -*• strength deopnis depth, abyss hethenesse —• heathendom midnes —• middle, midst besceawodnes —• sight

Other -ness words denoting entities dropped the -ness and did not replace it with another suffix, or else an alternate form without a suffix and identical to the adjective base displaced the -ness form. In some cases this amounted to returning to an earlier form. (25) a. truthness -*• truth b. anguishness —> anguish

A historical perspective

c. d. e. f. g. h. i.

on the productivity

of the suffixes -ness and -ity

447

byrdness —• birth nydnyss need täcnyss —• token egeniss -*• awe aferednesse —• fear ägylines guilt brysednes —• bruise

Still other -Hess words denoting entities were replaced by -ity words formed on an etymologically related base: (26) a. b. c. c. e. f. g.

prinnes7 trinity cristness ->· Christianity ännes —• wm'/y xtnes —> edibility asprenesse —• asperity medemlicnes —• mediocrity sxdnes satiety

The examples in (27) are -«ess words denoting abstract entities eventually replaced by entirely different words. These are only a few of the huge number of such examples. Note that in some cases the base word was also lost, while in other cases the base word was retained. (27) a. b. c. d. e. f. g. h. i.

mildhertnesse —• compassion, inbrydnisse —>· humility lufsumnes -> amiability änrednisse —>• constancy golnesse -> lechery häelnes salvation godnesse -> deity oferetolnes gluttony lessness —*• inferiority

mercy

In a number of cases, a -ness word which earlier had a sense denoting an entity lost this sense, but kept another sense denoting a trait, or else changed its meaning to denote a trait. Elsewhere a word was lost and an equivalent form was later recoined with a new trait sense. Examples of these are given in (28).

448

Elizabeth

(28) a. b. c. d. e. f. g. h. i. j. k. 1. m.

Μ.

Riddle

dearness — 'an expression or token of affection' vileness — 'foul matter' särness — 'pain' (cf. ModE soreness) fairness — 'beauty' (rather archaic in Modern English) hefigness — 'melancholy', 'sadness'; 'weight', 'burden' (cf. ModE heaviness) greatness — 'size', 'dimension' grimnis — 'anger', 'fury' menniscness — 'incarnation', 'human condition' (cf. ModE mannishness) mödignes — 'pride' (cf. ModE moodiness) hälignes — 'religion' (cf. ModE holiness) fulness — 'decay', 'corruption' (cf. ModE foulness) smiriness — 'oil' (cf. ModE smeariness) cleannes — 'chastity'

Some similar changes occurred with -ity words as well. For a time, a number of minimal pairs of -ness and -ity words existed, apparently with the same sense. (This cannot be stated with certainty, however, without a detailed analysis of the items in context.) Subsequently, the -ity word denoting a trait or embodied characteristic was lost, e. g.: (29) a. b. c. d.

bareinesse/bareinite8 barrenness debonairness I debonarity debonair eness; gentleness 9 vastness/vastity -> vastness dullness jdullity -*• dullness

Other -ity words denoting traits were replaced by a different -ness word or a -ness word on the same base. In some cases, the -ity word was an "inkhorn" term, i. e., one consciously coined by a writer or dictionary maker but which never took hold. In fact, such words were often "branded" or condemned by other dictionary makers (Osselton 1958: 21-108). (30) a. b. c. d. e. f.

lenity —• mildness, gentleness, mercifulness terreity -> earthiness salinity ->smoothness scabredity roughness anility —• dotage, foolishness vility —• vileness

A historical perspective

g. h. i. j.

on the productivity

of the suffixes -ness and -ity

449

acrity —> keenness, sharpness raucity —• harshness, roughness sordidity -*• sordidness scant ity -> scantness, scarceness

A few -ity words possessing a sense denoting a trait lost that sense and now only denote an abstract or concrete entity. These are given in (31). (31) a. cavity — 'hollowness' b. levity — 'smoothness' c. surety — 'sureness' It is significant that all of these different changes have contributed to the same net result: except for a very few old -ness words denoting entities, such as witness, and fastness, -ness is ceasing to be used to form words denoting anything other than a trait, while a great number of -ity words now denote an abstract or concrete entity. Other contemporary -nessf-ity pairs denoting traits seem preferable, at least for some speakers, in their -ness forms: (32) a. b. c. c.

collectivity ->• collectiveness oddity —>• oddness suavity —* suaveness verticality —• verticalness

The above facts suggest that the changes occurring, which have had the net result of semantically distinguishing -ness and -ity to a considerable extent, have affected the lexicon on a nearly word-byword basis over time. For this reason, I believe that this change can be called a case of lexical diffusion. This lexical diffusion is significant in several respects. First, it is possible that the incipient change could spread only by lexical diffusion since there were/are so many conflicting factors involved. The change only affects words for which some other factor does not take precedence. Second, the fact that the change has been spreading gradually causes items which have not been affected yet to appear to be counterexamples at any given synchronic stage (cf. the work of Wang (1969) and Chen (1972) in phonology in this regard). It is also possible that the unaffected words cause confusion regarding the generalization for English speakers and will thus eventually hinder the spread of the change. Finally, it should be

450

Elizabeth

Μ.

Riddle

expected that individual items will exhibit the change in the lexicons of some speakers, but not others. There are other historical factors which I believe may have contributed to obscuring the new generalization concerning -ness. The first is that there was apparently an earlier stage of competition between -ness and several other suffixes forming nouns from adjectives, i.e., -dom, -hed (-hede) ( = ModE -hood), and -scipe (-shepe), (-ship).10 Although according to Thiele (cited in Gadde 1910: 8), -ness was by far the most frequently used suffix for this purpose even in Old English, and although the suffixes -dom, -hed, and -scipe had non-overlapping senses as well, there was a good deal of competition between all the different suffixes as well as the Romance suffixes at various points in time (Maetzner 1874: 437; Kington-Oliphant 1891: 438, 500; and Jespersen 1942: 316). It appears that during the Old and Early Middle English periods, the above suffixes were somewhat interchangeable when forming nouns from adjectives. (These would have to be studied in context to ascertain whether there was in fact no sense distinction, but it is entirely possible that during a period of flux, alternate forms with the same meanings existed. It is also possible that dialectal variation was involved.) Some examples of these are given below: (33) a. b. c. d. e. f. g. h. i. j.

raunesse, rauhede — 'rawness' gladness, gladscip — 'gladness' fainnes, fainhed — 'gladness' wisness, wlshede, wisdom — 'wisdom' arjnesse, arhede, aerscipe — 'cowardice', 'sloth' weorpscipe, worthness — 'honor', 'worthiness' blmhede, blindness — 'blindness' dronkenesse, drunkenhede, drunkeshepe — 'drunkenness' derkness, deorkhede — 'darkness' boldhede, boldness — 'boldness'

-Ness eventually consolidated its position as the primary suffix forming nouns from adjectives, and in Modern English -dom, -hood and -ship are not used productively in this function. -Ness words also began to covary with nouns without a suffix and identical to the adjective base such as the following: 11 (34) a. glaed — gladness b. deork — darkness

A historical perspective

c. d. e. f. g. h.

on the productivity

of the suffixes -ness and -ity

451

göd — goodness birhtu — brightness blake — blackness bittere — bitterness dear — dearness wurd — worthiness

Thus, just around the time that -ity began to make inroads on the position of -ness in English word formation (i. e., the fourteenth century (Serjeantson 1935: 136)), the -ness affixation process was strengthening its own position vis-ä-vis other native abstract nounforming processes. Just as Wang (1969) has argued with respect to phonological change, it is possible that these conflicting tendencies were responsible for a residue of examples which do not fit in with the new generalization concerning the function of -ness in contemporary English. Another relevant historical factor, noted by Aronoff (1976: 56), is the tendency to semantic drift or metaphorical extension. It is common for words to take on new senses, and typical such changes are from concrete to abstract and literal to metaphorical senses. From a sense denoting an abstract entity to a sense denoting a trait would be a similar type of change. This strong tendency to sense extension probably also blurs the general semantic distinction between -ness and -ity, which after all is a very subtle one. The motivation for the historical change remains to be discussed. An immediate cause would be the commonly encountered tendency for synonymous linguistic elements to diversify. But languages occasionally tolerate such duplications, and sometimes there is only a register difference between otherwise synonymous elements. Why should -ness and -ity have changed in this particular direction? My hypothesis is that this was a result of the circumstances under which -ity words and other abstract substantives were introduced into English. I believe that the area of religion was of great significance in providing the impetus for the historical change. After the conquest, most English abbots and bishops were replaced by Norman and Central French speakers, which naturally had a great effect on ecclesiastical vocabulary in the monasteries and upper echelons of the church (Shelly 1921: 32; Vising 1923: 11; Jespersen 1955: 91-92). Even after the Normans had assimilated to the point where they were no longer seen as a separate ethnic group, French continued,

452

Elizabeth

Μ.

Riddle

along with Latin, to be important in monastic life, since religious figures looked to the continent as a source of culture and learning. For example, the important book of monastic rule, the Ancrene Riwle of circa 1225 was written in English, but very much influenced by French tradition, to the point where many synonymous pairs of English and French words, such as cherite and luve, pacience and polemodnesse, were used side by side (Jespersen 1942: 100). Large numbers of words pertaining to moral ideas and religion were borrowed and the Old English terms were dropped. As KingtonOliphant (1891: 452) notes, although the law, learning, fashion and chivalry were areas of great French influence, they were restricted to the upper classes. Religion went to all people, however. When virtually all the Old English theological terms were replaced with French and Latin loans, it made a huge difference to the language of the common people. According to Kington-Oliphant (448-452), the Franciscan movement was probably significant in the spread of French and Latin loans, since the friars, who usually knew these languages, had contact with all people and not just the nobility. By the fourteenth century, French had even crept into the Lord's Prayer, and Wycliffe and his collaborators introduced more than 1000 new Latin terms into English in translating the Bible (Mosse 1947: 103). Many of these French/Latin religious terms denoted abstract entities such as the names of virtues, vices, the seven deadly sins, good works, etc. It is my hypothesis that this wholesale borrowing of terms in entity senses caused the native suffix -ness to lose its entity sense. It appears that at first, French and Latin borrowings with a variety of suffixes, such as (in their Modern English forms) -ancej -ence, -(t)ion, -ment, -ity, etc., replaced the -ness words. It was not until well into the Middle English period that -ity words denoting entities began to be borrowed in large numbers, primarily from Latin, and directly competed with -ness. Thus, the first stage in the change was the loss of a coherent class of -ness words, i. e., those denoting abstract or concrete entities. Although much work remains to be done, some suggestive figures may be offered here. Of some 2370 entries under the letter A in Borden's (1982) A Comprehensive Old English Dictionary (henceforth COED), 121 or 5.1% are -ness words. Of these, 96 have entity senses. In contrast, under the letter A in the Middle English Dictionary (henceforth MED) edited by Kurath and Kuhn, of

A historical perspective

on the productivity

of the suffixes -ness and -ity

453

approximately 3100 entries, only 44 or 1.4% are -ness words. I could find only 22 -ness words that are listed in the COED in the MED as well (taking into account phonological and spelling changes). It is worth noting that 63% of the bases on which the COED entity -ness words not occurring in the MED are formed are fairly transparently represented in other lexical items in the MED. Thus, while the -ness form denoting an entity was lost, the corresponding adjective, verb or adverb is still represented. This shows that the loss of such -ness words was not due merely to the loss of particular bases. Some examples of -ness entity words from Old English replaced by French/Latin loans in Middle English culled from the A volume are given below: (35) a. b. c. d. e. f. g. h. i. j.

xfterweardnes -* posterity xwicnes -> eternity ärednes -*• degree, condition, covenant äszegdnes mystery, sacrifice astandennes -> perseverance, existence, subsistence ästignes -> ascent ätendnes incentive äwecenes -*• incitement awegawörpnes -> abortion äwordennes degeneration

Notice that all of the above words are relevant to areas of concern in medieval religion and morality. Just as striking are the examples of -ness words in Middle English (starting with A) denoting entities, again from the sphere of religion, which do not occur in contemporary English. Some of these are: (36) a. b. c. d. e. f. g. h. i. j.

admodnesse -» humility alwyttynes -*• omniscience andetnes -*• confession of sins, profession of faith arghnesse -*• cowardice, timidity arisnesse -* the resurrection armhertnesse -» contrition asetnesse appointed order, institute ateowednisse —• manifestation attemprenesse -> temperance (the Christian virtue) arwurönesse -> dignity, reverence

454

Elizabeth

Μ.

Riddle

To summarize, it appears that a significant proportion of the -ness words lost from Old and Middle English were terms pertaining to religion and morality, or else general words commonly occurring in these contexts. I hypothesize that these contexts favored the entity rather than the trait senses of the French and Latin replacements and that the loss of so many entity -ness words from English led to a loss of that sense for -ness in general. This resulted in a gap in English word formation processes (because -ness was no longer productive in the entity sense) which was to be eventually filled by foreign suffixes, notably -ity. This analysis may provide a reason for why, according to Aronoff (1976: 38-39), the set of contemporary -ness words exhibits greater semantic coherence than that of the -ity words. The trend initiated by the encroachment of religious loans upon the domain of -ness was reinforced by the importance of Latin and French in general medieval and renaissance scholarship. Many latinate English words, and in particular, -ity words, were borrowed in the process of translation of classical works of philosophy, literature and history, mainly from Latin, but sometimes via the French version of a Greek original. I hypothesize that because of the subject matter being translated, Latin -itas and French -ite words, from which English -ity words were borrowed, tended to occur in abstract and concrete entity rather than embodied trait senses. This led to a reanalysis of the Latin and French originals as denoting entity senses in English and reinforced the loss of these senses by -ness words due to the common tendency for duplicates to diversify. According to Baugh (1957: 259), English was viewed as lacking abstract words capable of conveying the richness of Latin and French originals, and this led to a great deal of borrowing. He states: "the scholarly monopoly of Latin throughout the Middle Ages had left the vernaculars undeveloped along certain lines." It is probable that the loss of -ness as a productive suffix for the formation of words with entity senses was also significant in this regard. As Baugh observes, however, there was a fine line between needed loans and borrowing for ornamental purposes. Interest in stylistic variation and elegance was great during certain periods (Shipley 1955: X-XI, Schlauch 1959: 66-80). One reflex of the interest in rhetorical elegance which may have facilitated the replacement of -ness words denoting abstract entities was the use of synonymous word pairs mentioned above. 12 Caxton used many

A historical perspective

on the productivity

of the suffixes -ness and -ity

455

pairs of native and French words (Jespersen 1955: 101),-and Koskenniemi (cited in Robinson 1972: 166) found in a study of word pairs in Old and Middle English prose that the majority denoted abstract ideas. One possible reason for why -ity words in particular became so popular is that there was great concern among English writers, especially during certain periods of the sixteenth through the eighteenth centuries, about the so-called monosyllabic nature of English (Kington-Oliphant 1891: 505; Mosse 1947: 131, 202; Shipley 1955: X; Tucker 1961: 57-120). Although not all English writers felt this way (e. g., George Jeffreys, Elizabeth Elstob and Archibald Campbell), many agreed with Dryden's judgment that the native monosyllables with which English was said to abound were inelegant. part of the dislike of monosyllables centered on the fact that several consonants occurred in the same syllable, which according to Lord Monboddo, made English sound harsh (Tucker 1961: 119). In comparing English to Italian, Dryden states in the Dedication of his translation of the works of Virgil (1721: 411): "... 'tis all we can do to give sufficient Sweetness to our language: We must not only chuse our Words for Elegance, but for Sound. To perform which, a Mastery in the Language is requir'd; the Poet must have a Magazin of Words, and have the Art to manage his few Vowels to the best Advantage, and that they may go farther." 1 3 There had to be considerable contact between the -ity words borrowed during the late Middle Ages and Renaissance through scholarly translation and the remaining -ness words in the everyday language for the two groups to continue to diverge in the directions indicated. Otherwise one would expect merely a register difference to have arisen. This was probably facilitated by the renaissance goal of spreading classical ideals to the general population and the desire of the newly literate and rising middle class to educate itself. According to Schlauch (1959: 67), the English reformation removed education from the exclusive control of the Latin-speaking clergy and transferred it to larger, secular groups. Although this weakened the position of Latin as the only language of instruction, it probably contributed a great deal to the dissemination of Latin and French loans into ordinary people's vocabulary. As noted above, translation of the Bible into English also played a role, especially since the Bible was very popular reading at the time. In fact Schlauch (1959:

456

Elizabeth Μ. Riddle

68) cites an anonymous author of 1528 who claimed that reading the Bible in English replaced an earlier taste for romances. Finally, dictionaries of "hard words" began to be compiled, and these abounded in -ity words. A notable example is Cawdrey's 1604 dictionary (containing 112 -ity words), which he described on the title page as "A Table Alphabetical!, conteyning and teaching the true writing, and understanding of hard usual English wordes, borrowed from the Hebrew, Greeke, Latine, or French, &c. With the interpretation thereof by plaine English words, gathered for the benefit & helpe of Ladies, Gentlewomen, or any other unskillfull persons. Whereby they may the more easilie and better understand many hard English wordes, which they shall hear or read in Scriptures, Sermons, or elsewhere, and also be made able to use the same aptly themselves." Thus, there was sufficient intermingling of -ness and -ity words to supply continued pressure toward semantic divergence rather than allowing them to remain distinct merely on the stylistic level as [ ± learned]. At this point I would like to say a few more words about the historical change discussed in this paper being a case of lexical diffusion. Even the small sampling of terms beginning with the letter A from Old to Middle to contemporary English shows that the change took place gradually on an item-by-item basis. Although I do not yet have exhaustive counts of the items and senses occurring during all the relevant periods, it is clear that in Old English, -ness words had both embodied trait and entity meanings, that there were fewer -ness words with entity senses in Middle English, and still fewer in contemporary English. Conversely, -itas and -ite words in Latin and French appear to have had both trait and entity senses, while -ity words in contemporary English predominate in the entity sense.14 As noted earlier, a lexical diffusion analysis provides a partial explanation for many apparent counterexamples to the semantic generalization in that they can be viewed as the as-yetunaffected residue of items; in other words, the existence of this group of items is seen as a partial result of the fact that lexical diffusion has occurred. In any case, I believe that the existence of so many conflicting factors will probably prevent the extension of the change to the entire English lexicon.

A historical perspective

on the productivity

of the suffixes -ness and -ity

457

To summarize, the productivity of the suffixes -ness and -ity cannot be accounted for in any simple way. A number of factors, including meaning, phonological transparency, blocking, conditions of borrowing, the lack of practical need for a word with a particular meaning, register distinctions, and historical changes (including metaphorical extensions) must all be considered. Although the analysis presented here does not appear to have the formal neatness of a feature analysis nor of an analysis taking "productivity" as a primitive, nor of an analysis based purely on phonological transparency or the morphological form of the base, it is nevertheless much more explanatory than any of them. This more complex and less elegant analysis relates to more relevant data and attempts an explanation for why certain groups of words mean what they do, which an analysis based on a primitive notion of productivity cannot do. This analysis also casts grave doubt on the validity of experiments that do not consider the possibility of a semantic distinction for -ness and -ity words. The experiments described above purporting to show that either preservation of phonological transparency or the morphological form of the base is the crucial criterion must be questioned, as the tests did not allow for all the variables. In fact, this is probably why Cutler on the one hand, and Aronoff — Schvaneveldt and Anshen — Aronoff on the other, offer conflicting conclusions. In the absence of sufficient contexts to force particular readings, experimental subjects must fall back on other criteria to make the relevant choices. To conclude, this paper does more than offer an elucidation of the use of the English suffixes -ness and -ity: it calls into question an entire approach to morphological study.

Notes * I am deeply indebted to Paul Neubauer for many hours of discussion and encouragement. I would also like to thank Merle Fifield for her helpful comments. Earlier versions of this paper were presented at the Linguistic Society of America 1980 Summer Meeting, the 14th Annual CAES [Committee for the Advancement of Early Studies] Conference, 1983, and the 4th International Conference on Historical Linguistics: Historical Semantics/Historical WordFormation, 1984, and I thank the audiences for their comments and discussion. This research was partially supported by a Faculty Research Grant from Ball State University, 1983-84. I am grateful to the American Council of Learned Societies and Ball State University for providing the funds for my attendance at the conference where the papers in this volume were presented.

458

Elizabeth Μ. Riddle

1. It appears to be somewhat similar to the distinctions between type and token, and generic vs. specific (I owe the latter observation to Ross Clark), but does not correspond to them exactly. 2. Henceforth I will use double question marks to indicate unacceptability in a given context. 3. Catholicism is also possible here, but refers more to a structured code of beliefs than to an embodied trait. I think any tendency for Catholicism to be preferred in this context is attributable to the fact that it is already a well-established lexical item, whereas Catholicness is not. 4. Nesfield (1922: 257) notes that many of the French borrowings between 1350 and about 1600 are learned words originally borrowed by the French from Latin books and "put into a French dress." For example, ferocity is from literary rather than popular French. 5. I have made no attempt to regularize the spelling of the words cited. The words are reproduced here as I found them in the various dictionaries, grammars, and literary texts from which they were taken. (Sources used to compile the data base for this study are listed in the References, but not cited individually in the text.) Typographical conventions, however, were regularized. The following are all spelling variants of -ness: -nes, -nesse, -nis, -nisse, -nys, -nysse. Sometimes the η belonging to the suffix is dropped if the base ends in an n, e. g., gröness. 6. The arrow (—•) is merely meant to indicate that the particular lexical item in question dropped from the language and a corresponding word came into use. It is not meant to imply that the item to the left actually evolved into the item on the right, although this may be the case for some examples. The histories of individual lexical items in this sense are not crucial to this study. 7. prinnes and cristness were originally loan translations. 8. -ite is an early spelling of -ity. -Ty is a variant of -ity but was never a productive form (Marchand 1969: 315). 9. The -ity words in c.-e. never had much currency. 10. The suffix -th was already ceasing to be productive in Old English. 11. Kington-Oliphant (1891: 33) and Mosse (1968: 91) note the use of adjectives as substantives at various periods in the history of English. 12. This is not the only reason for the existence of such pairs. See Robinson (1972: 166-168) for discussion. 13. Although the volume from which this particular quote was taken was a translation of the Aeneid in verse, the concern with monosyllabicity and plethora of consonants extended to prose as well. 14. My assumption that -itas and -ite in Latin and French respectively cover both the trait and entity senses is based upon a consultation of the following works treating the morphology of these languages: De Chevallet (1858: 373), Brugmann (1891: 472-473), and Nyrop (1936: 168). (Joan Bybee brought this consideration to my attention.) This question requires further investigation of the items in question in context.

References Anshen, Frank — Aronoff, Mark 1981 "Morphological productivity and phonological transparency", Canadian Journal of Linguistics 26 (1): 63-72. 1984 "Morphological productivity in historical perspective", read at the 4th International Conference on Historical Linguistics: Historical Semantics/Historical Word-Formation, Blazejewko, Poland, March 1984.

A historical perspective on the productivity of the suffixes -ness and -ity

459

Aronoff, Mark 1976 Word formation in generative grammar (Cambridge, Mass.: The MIT Press). 1982 "Potential words, actual words, productivity and frequency", read at the XIII International Congress of Linguists, Tokyo. Aronoff, Mark — Schvaneveldt, Roger 1978 "Testing morphological productivity", Annals of the New York Academy of Sciences 318: 106-114. Bailey, Richard W., ed. 1978 Early Modern English — Additions and antedatings to the record of English vocabulary 1475-1700 (Hildesheim: Georg Olm Verlag). Baugh, Albert C. 1957 A history of the English language (2nd ed.) (New York: AppletonCentury-Crofts, Inc). Borden, Arthur R., Jr. 1982 A comprehensive Old-English dictionary (Washington, DC: University Press of America). Bosworth, Joseph 1898 An Anglo-Saxon dictionary (Edited by T. Northcote Toller) (Oxford: The Clarendon Press). Bright, James W. 1917 An Anglo-Saxon reader (4th ed.) (New York: Henry Holt and Company). Brown, A. F., ed. 1963 Normal and reverse English word list (Vol. 7) (Philadelphia: University of Pennsylvania/Air Force Office of Scientific Research). Brugmann, Karl 1891 A comparative grammar of the Indo-germanic languages (Vol. II, Part 1. Translated by R. S. Conway and W. H. D. Rouse) (New York: Westermann & Co). Campbell, Russell N. 1969 Noun substitutes in Modern Thai — A study in pronominality (The Hague: Mouton). Cawdrey, Robert 1604 A table alphabeticalI of hard usual English words (Facsimile reproduction with introduction by Robert A. Peters, 1966.) (Gainesville, Fl: Scolar Facsimiles & Reprints). Chen, Matthew 1972 "The time dimension: contribution toward a theory of sound change", Foundations of Language 8: 459-498. Chevallet, A. de 1858 Origine et formation de la langue franfaise (Paris: L'Imprimerie Imperiale). Clark, Eve V. - Herbert H. Clark 1979 "When nouns surface as verbs", Language 55: 767-811. Crider, Kitty 1982 "Feingold diet revolutionizes habits of eating", Austin AmericanStatesman, Thursday, April 22, 1982, Ε 1. Cutler, Anne 1980 "Productivity in word formation", Papers from the 16th Regional Meeting — Chicago Linguistic Society. (Chicago: Chicago Linguistic Society). 1981 "Degrees of transparency in word formation", Canadian Journal of Linguistics 26 (1): 72-77.

460

Elizabeth Μ. Riddle

Dryden, John 1721 The works of Virgil translated into English verse (Vol. 2) (London: Jacob Tonson). Fisiak, Jacek 1968 A short grammar of Middle English ( P a r t i ) (Warsaw: Panstwowe Wydawnictwo Naukowe). 1972 An Early Middle English reader (Warsaw: Panstwowe Wydawnictwo Naukowe). Funk & Wagnall's New standard dictionary of the English language 1963 (New York: Funk & Wagnall's). Gadde, Frederick 1910 On the history and use of the suffixes -ery, (-ry), -age, and -ment in English (Lund: Gleerupska Univ. Bokhandeln). Goffman, Erving 1974 Frame analysis (New York: Harper Torchbooks, Harper & Row, Publishers). Jespersen, Otto, with Paul Christopherson, Niels Haislund and Knud Schibsbye 1942 A modern English Grammar on historical principles. Part 6, Morphology (Copenhagen: Ejnar Munksgaard). Jespersen, Otto 1955 [1948] Growth and structure of the English language (Garden City, NY: Doubleday & Company, Inc). Kington-Oliphant, Thomas L. 1891 The Old and Middle English (2nd ed.) (London: MacMillan and Co). 1968 Handbook of Middle English (Translated by James A. Walker. 5th printing, corrected, augmented.) (Baltimore: The Johns Hopkins Press). Koziol, Herbert 1972 [1937] Handbuch der englischen Wortbildungslehre (Revised edition) (Heidelberg: Carl Winters Universitätsverlag —Buchhandlung). 1967 Grundzüge der Geschichte der englischen Sprache (Darmstadt: Wissenschaftliche Buchgesellschaft). Kurath, Hans — Kuhn, Sherman, eds. 1956Middle English dictionary (Ann Arbor: University of Michigan Press). Lehnert, Martin 1971 Reverse dictionary of Present-day English (Leipzig: VEB Verlag Enzyklopädie). Maetzner, Eduard 1874 An English grammar: methodical, analytical and historical (vol. 1. Translated by Clair James Grede.) (London: John Murray). Marchand, Hans 1969 The categories and types of Present-day English word formation — synchronic-diachronic approach (2nd ed.) (Wiesbaden: Otto Harrassowitz). Mayhew, A. L. - Skeat, Walter W. 1888 A concise dictionary of Middle Engish (Oxford: At the Clarendon Press). Mosse, Fernand 1947 Esquisse d'une histoire de la langue anglaise (Lyon: Edition I. A. C.). Nagucka, Ruta 1977 A Late Middle English reader (Warsaw: Panstwowe Wydawnictwo Naukowe). Nesfield, J. C. 1922 English grammar past and present (London: MacMillan and Co., Limited).

A historical perspective on the productivity of the suffixes -ness and -ity

461

Nyrop, Kristoffer 1936 Grammaire historique de la langue frangaise (Vol. 3) (Copenhagen: Gyldendalske Boghandel Nordisk Forlag). Osselton, Ν. Ε. 1958 Branded words in English dictionaries before Johnson (Groningen: J. B. Wolters). The Oxford English Dictionary 1933 (Reprinted 1970.) (Oxford: The Clarendon Press). Randall, Janet 1980 "-ity: A study in word formation restrictions", Journal of Psycholinguistic Research 9 (6): 523-534. Reichl, Karl 1982 Categorial grammar and wordformation: the de-adjectival abstract noun in English (Tübingen: Max Niemayer Verlag). Robinson, Fred 1972 "Review of repetitive word pairs in Old and Early Middle English prose, by Inna Koskenniemi", Anglia 90: 166-168. Schlauch, Margaret 1959 The English language in modern times (Since 1400) (Warsaw: PWN — Polish Scientific Publishers). Seijeantson, Mary S. 1935 A history of foreign words in English (London: Kegan Paul, Trench, Trubner & Co., Ltd). Shelly, Percy Van Dyke 1921 English and French in England 1066-1100 (Ph. D. dissertation, University of Pennsylvania, Philadelphia). Shipley, Joseph T. 1955 Dictionary of Early English (New York: Philosophical Library). Stratmann, Francis Henry 1891 A Middle-English dictionary (Revised by Henry Bradley) (Oxford: The Clarendon Press). Tucker, Susie I. 1961 English examined — two centuries of comment on the mother tongue (Cambridge: The University Press). Vising, Johan 1923 Anglo-Norman language and literature (London: Oxford University Press). Walker, John 1936 [1775] The rhyming dictionary of the English language (Revised and enlarged by Lawrence H. Dawson) (New York: E. P. Dutton and Co). Wang, William 1969 "Competing changes as a cause of residue", Language 45: 9-25. Wardale, Ε. E. 1937 An introduction to Middle English (London: Kegan Paul, Trench, Trubner & Co., Ltd.). Webster's Third New International Dictionary 1971 (Springfield, Mass.: G. & C . Merriam Company, Publishers). Wright, Joseph 1905 The English dialect dictionary (Oxford: Henry Frowde). Wright, Joseph — Wright, Elizabeth Mary 1923 An elementary Middle English grammar (London: Humphrey Milford). 1925 Old English grammar (3rd ed.) (London: Humphrey Milford. Reprinted 1934, Oxford University Press).

WILLIAM C. RITCHIE

Word-formation, learned vocabulary and linguistic maturation Introduction The so-called Dative Alternation in English, as exemplified in (1) and (2) below, has been analyzed in a number of different ways in recent work within the tradition of generative grammar. (1)

a. John sent the package to Mary. b. John sent Mary the package.

(2)

a. John boiled an egg for Mary. b. John boiled Mary an egg.

In general, within this tradition, the (a) and (b) examples have been treated as systematically related in some way — either through a transformational rule of Dative Movement in the theory of the mid-60's (see, e. g., such standard texts in syntax as Burt, 1971; Akmajian and Heny, 1975), through some sort of lexical redundancy rule (e. g., Oehrle, 1976), or as a consequence of Case Theory within the Government-Binding framework (see Czepluch, 1982/83). In any case, all approaches (that is, those within which the question is raised at all) treat the (a) structures (henceforth, prepositional Indirect Object — structures) as unmarked in the sense of Markedness Theory whereas the (b) structures (henceforth, prepositionless Indirect Object — structures) are treated as marked. Specifically, within the GB framework, both Chomsky (1981: 171) and Czepluch (1982/83) derive the prepositional Indirect Object structures quite directly from plausible principles of core grammar — particularly, Case Theory. However, the (b) structures (prepositionless Indirect Object) present a number of theoretical difficulties that are solved, within these analyses, by the introduction of certain exceptional features rendering them marked. (One more point of terminology — we will henceforth refer to Verbs that

464

William C.

Ritchie

accept either the prepositional Indirect Object structure alone or the prepositionless Indirect Object structure alone or both as Indirect Object Verbs.) Two classes of Indirect Object Verbs that are exceptions to the Dative Alternation, in the sense that they accept only Prepositional Indirect Verbs, have also received attention. In the context of a discussion of the subcategorization frames of derived lexical items, Carlson — Roeper (1981) note exceptions like (3) and (4). (3)

a. John mis-sent the package to Mary. b. ?John mis-sent Mary the package.

(4)

a. John re-boiled an egg for Mary. b. ?John re-boiled Mary an egg.

They propose a general constraint on the subcategorization of derived Verbs, again, in terms of Case Theory. In a rather different context, Baker (1979) notes the problematic character of "exceptions" to traditionally optional transformations for the logical problem of language acquisition given that only positive evidence is available to the language learner. Among the structures he discusses are exceptions to Dative Movement of the sort given in (5) and (6). (5)

a. John transmitted the package to Mary. b. *John transmitted Mary the package.

(6)

a. John prepared an egg for Mary. b. *John prepared Mary an egg.

We return below to a detailed discussion of the problems that these examples present for a theory of language acquisition. Oehrle (1976: 121-125, 137-138) also discusses these exceptions to the Dative Alternation in a more traditional way — regarding this task as (at least) the description of the phenomena in terms of the mature grammar of English alone rather than treating it in terms of learnability — and proposes a constraint that links the exceptionality of Verbs like transmit and prepare to their internal morphological structure. Specifically, Oehrle proposes (1976: 124) the constraint in (7).

Word-formation,

(7)

learned vocabulary

and linguistic maturation

465

( = Oehrle's (47)) The dative alternation does not apply if the verb in question has the internal structure [ X ] p f i x = [ Y J s t e m re

where ' = ' is a morphological boundary specific to the latinate stratum of the lexicon. This constraint simply has the effect of excluding latinate Verbs from the prepositionless Indirect Object structure while allowing Verbs in the germanic stratum to exhibit both structures. Though Carlson — Roeper's explanation of the exceptionality of derived Verbs may find an explanation of the usual sort (that is, through Universal Grammar) in the form of some sort of linkage between the "derivational complexity" of lexical items and their subcategorizations, neither Oehrle's traditional analysis nor Baker's learnability account addresses the question of why it is that, in the overwhelming number of cases, the non-derived exceptions to the Dative Alternation are latinate Verbs — that is, verbs from the learned vocabulary of English. Oehrle's account may be descriptively correct but, by his own statement, lacks explanatory force. Baker's account supports Oehrle's lexical analysis (see below) but does not address the question of why it is specifically learned vocabulary that resists the marked prepositionless Indirect Object structure. In addition to the considerations discussed above, we may note that the class of Verbs allowing the prepositionless Indirect Object structure has diminished radically through history since the Old English period (see, e. g., Visser, 1963, vol. I, pp. 621-635). A proposal by Czepluch (1982/83) suggests indirectly that this is one consequence of a typological change in the structure of English with respect to type of Case-assignment — a shift that renders the prepositionless Indirect Object structure highly marked in Modern English, though less so in Old English. The Dative Alternation is thus a rich area of investigation for a number of subproblems in the study of syntactic, morphological, and lexical structure and their interaction — both synchronically and diachronically though existing accounts of the exceptions to the Dative Alternation remain sketchy and non-explanatory. The purpose of this paper is to bring the Dative Alternation facts to bear on yet another theoretical question — that of how Chomsky's idealization to instantaneous language acquisition (see, e. g., Chomsky, 1965: 202, footnote 19) is to be relaxed to give a

466

William C.

Ritchie

more realistic account of the process of language acquisition. Briefly, I will propose a non-instantaneous view of language acquisition incorporating a fairly straightforward notion of "critical period" (see Lenneberg, 1967) within which the mature grammar of a particular language is seen not as the consequence of the choice of one from a class of mature grammars available to the organism in its "initial state" — as the process is often characterized (see, e. g., Chomsky, 1975: 137-8) — but rather as the product of the interaction of a genetically determined maturational schedule with specific linguistic experience more in keeping with Chomsky's discussions of language acquisition as the "growth" of a "mental organ" (e.g., Chomsky, 1975: 10-11). Under this view, the exceptions to the Dative Alternation exemplified in (3)-(6) above receive an account which is both quite natural and explanatory in a significant sense. Briefly, it is suggested that these exceptions are to be accounted for as the product of a maturational sequence that determines the opening and closing of critical periods for the acquisition of rules of certain sorts. The paper has three Sections. Section 1 sketches three general approaches to the explanation of features of the mature grammars of particular languages. Section 2 presents and argues for a particular type of explanation for the latinate exceptions to the Dative Alternation. Section 3 is a conclusion.

1.0. Three types of linguistic explanation The task of this section is to set the general problem of explanation of properties of mature grammars in more detail in preparation for an explanation of the specific case of exceptions to the Dative Alternation. Since the current approaches to the explanation of both the mature grammars of particular languages and the process of acquisition of a mature grammar are based on the notion of Universal Grammar, it will be important to bring certain features of the notion of Universal Grammar to light before we relate it to claims about the process of acquisition. Likewise, there appear to be two major formulations of the relationship between Universal Grammar and the process of acquisition — one noted above under which Universal Grammar is part of the "initial state" of the organism and the process of

Word-formation,

learned vocabulary

and linguistic maturation

467

acquisition is "stationary" (in the sense of Pinker, 1981) and the other under which the process is taken to be "non-stationary". A stationary view of acquisition would take the acquisitional capacity of the learner to be constant through time (even though the learner's capacity to use the language, including his grammar for the language, would differ from stage to stage in the acquisition process) whereas a non-stationary view would allow for changes in the learner's acquisitional capacity over the period of acquisition. These alternatives will be discussed in sections 1.2 and 1.3, respectively.

1.1. Explanation in terms of Universal Grammar The characterization of universal properties of linguistic structure (as opposed to the idiosyncratic properties of the grammars of particular languages) has been the central concern of work in generative grammar since its first beginnings in the mid 1950's. Recent work in this tradition — particularly that in syntax — has reached new levels of depth and specificity in its claims through the formulation of Universal Grammar as a highly restrictive schematism characterizing those properties that are shared (or claimed to be shared) by the grammars of all languages as a matter of biological necessity, with well-specified points ("parameters") where structured variation is possible, differences among particular languages being accounted for by the "fixing" of these parameters in different "values." For example, whereas English places Direct Objects after Main Verbs (e. g., see the cat) sentential complements after their head Nouns (the fact that John left), the complements of a Preposition after the head Preposition (in the house), and the complements of an Adjective after the head Adjective (fond of Harry), Japanese exhibits precisely the opposite ordering. Given a characterization of the syntactic part of Universal Grammar as a schematism in which all phrases in all languages have heads and complements, and a parameter that, in the expected case, allows two variants of the order of heads and their complements — head-complement and complement-head — word order in the grammar of English is obtained by fixing this parameter in one value (head-complement) and word order in Japanese is obtained by fixing the parameter in its other value (complement-head). The full set of facts about word

468

William C.

Ritchie

order in English and Japanese is thus accounted for with a difference in value for a single parameter. Within this framework, then, features of the grammars of particular languages are thus considered to be explained (in a fairly straightforward sense) when they are attributable to Universal Grammar and the fixing of parameters in Universal Grammar. Among the values of a given parameter, one may be "more expected" or "more usual" than the other(s). For example — with respect to facts at hand — given that Universal Grammar provides for both prepositional and prepositionless Indirect Object structures as expressions of a particular semantic structure, and given that English assigns Case in the way it does (see, e. g., Czepluch, 1982/ 83, for discussion), the prepositionless Indirect Object structure is less expected for English than the prepositional Indirect Object structure. Phenomena of this sort are accommodated within the theory by assigning prepositionless Indirect Object the status of a "marked" structure and prepositional Indirect Object the status of an "unmarked" structure in the sense of traditional Markedness Theory. Interpreting Universal Grammar in the usual way as a hypothesis about one aspect of the language faculty, the grammar that would result from choosing the unmarked values of all parameters is termed the core grammar; marked structures are part of what is termed the periphery of the grammar of a particular language and require positive evidence from the environment for their implementation. This approach allows a more restrictive formulation of Universal Grammar than previous approaches (such as that of Chomsky, 1965), a feature which is highly desirable on both methodological and empirical grounds. Methodologically, a more restrictive version of Universal Grammar is, all things being equal, to be preferred for the same reasons that any more restrictive explanatory hypothesis is to be preferred over less restrictive ones: it has greater empirical content (equivalently, is more highly falsifiable, more sensitive to the phenomena, etc.). From an empirical point of view, since Universal Grammar is to be interpreted as an empirical hypothesis about the class of grammars available tacitly to the child as hypotheses about the structure of the language of his speech community — specifically, the class that is genetically specified for the species — then the more restrictive a particular hypothesized Universal Grammar is, all things being equal, the closer we come to an explanation of the speed with which the child acquires the language of his community

Word-formation,

learned vocabulary

and linguistic maturation

469

and an explanation of the uniformity of the mature grammars that result from the acquisition process. Since Universal Grammar, whatever we hypothesize it to be on empirical grounds, is genetically determined, a corollary of the preference for a highly restrictive account of Universal Grammar over less restrictive accounts is that linguists' hypotheses about the structure of a particular language which attribute features of that structure to a genetically determined, biologically necessary property contributed by the organism's genetic program — as opposed to accidental and fragmentary experience — are to be preferred to hypotheses which view the same structural features as an idiosyncratic property of the language in question. By attributing properties of the grammar of a language to genetically determined structures wherever possible, the linguist renders those properties more widely testable (falsifiable) since they are thus claimed to be properties not only of the grammar of the language under investigation but of the grammars of all human languages. One final point concerning explanation in terms of Universal Grammar: Universal Grammar is specifically a theory of mature grammars. Conventionally, arguments for properties of Universal Grammar — whether shared across languages or specified as variables across languages — are based on mature grammars. For example, given the facts of word order discussed above for the mature grammars of English and Japanese, an adequate account of Universal Grammar would have to allow for both possible word orders. In the case of exceptions to the Dative Alternation, an adequate account would have to allow for both prepositional and prepositionless Indirect Object. As will be argued below, this raises certain problems when attempts are made to incorporate Universal Grammar into theories of acquisition.

1.2. The process of acquisition: a stationary model In a number works published over the last ten years or so (Chomsky, 1975: 137-138; 1980: 187-188; 1981: 7-9), Chomsky has characterized the process of language acquisition as a series of cognitive states as follows (Chomsky, 1980: 187-188): ... we may suppose that there is a fixed, genetically determined initial state of the mind, common to the species with at most

470

William C.

Ritchie

minor variation apart from pathology. The mind passes through a sequence of states under the boundary conditions set by experience, achieving finally a "steady state" at a relatively fixed age, a state that then changes only in marginal ways. The basic property of this initial state is that, given experience, it develops to the steady state. Correspondingly, the initial state of the mind might be regarded as a function, characteristic of the species, that maps experience into the steady state. Universal grammar is a partial characterization of this function, of the initial state. The grammar of a language that has grown in the mind is a partial characterization of the steady state attained. In one quite natural interpretation, this characterization of acquisition claims that the class of mature grammars (specified by Universal Grammar) is tacitly available to the child in its "initial state" — perhaps at birth. The process of acquisition then consists in the choice of one of these grammars (equivalently, the fixing of the parameters of Universal Grammar) on the basis of experience. In addition to Universal Grammar ("... a partial characterization ...of the initial state ...") one may also assume as part of the initial state a "... strategy for finding hypotheses ..." (Chomsky, 1965: 203, footnote 22). Under the interpretation adopted here, Universal Grammar and a single, constant learning strategy taken together would then constitute the capacity for language acquisition throughout the acquisition process, changes in the learner's internalized grammar over the career of the learner being determined entirely by the interaction of these stationary mechanisms with experience.1 Note that on this view, the course of acquisition is, in fact, determined wholly by the interaction of learning strategies and the character of experience, the class of hypothesizable grammars available to the learner (that is, Universal Grammar) remaining constant throughout. Since, empirically, the course of acquisition runs, in general terms, from unmarked structures to marked, we might impute to the learning strategy the property that its first "finds" unmarked hypotheses from the class of those available and resorts to more marked hypotheses (available at birth) only at later stages of acquisition as they are required by positive evidence from the environment. Two recent lines of investigation bear quite directly on these issues — those reported in Baker (1979) and in Berwick (1982).

Word-formation,

learned vocabulary

and linguistic maturation

471

Baker (1979) notes, as mentioned above, that exceptions to optional rules present severe difficulties for a theory of learnability which can rely only on positive evidence from the environment — that is, which relies on the actual occurrence of utterances relevant to the adequacy of the child's developing grammar as opposed to negative evidence in the form of the information that particular utterances are ungrammatical. Reliance on positive evidence is crucial because of the fact that systematic investigation (Braine, 1971; Brown - Cazden - Bellugi, 1969; Brown - Hanlon, 1970) indicates that children receive little or no negative evidence regarding the grammaticality of utterances, whether their own or others'. How, given the traditional analysis of the Dative Alternation as mediated by an optional rule of Dative Movement, could the child arrive at the information that (3)b, (4)b, (5)b, and (6)b are ungrammatical given that all of (l)a, (l)b, (2)a, and (2)b are grammatical and that (3)a, (4)a, (5)a, and (6)a are grammatical without negative evidence from the environment — that is, the specific evidence that (4)b and (5)b are ungrammatical? In short, why doesn't the child "overgeneralize" the rule of Dative Movement to make the b examples of (3)-(6) grammatical? Baker's solution is as follows. First of all, he replaces the traditional transformational analysis for the mature grammar of English with the proposal that the two members of the Dative Alternation (prepositionless and prepositional Indirect Object) are lexicallydetermined subcategorizations of the individual Verbs in question so that send and boil are (lexically) subcategorized for both prepositional and prepositionless Indirect Object whereas mis-send, reboil, transmit, and prepare are (also lexically) subcategorized only for prepositional Indirect Object. Secondly, the acquisition of these structures is then hypothesized as proceeding in accordance with the following assumption (Baker, 1979: 548): ... verbs are not assumed to occur in the [ NP NP] environment [i. e., the prepositionless Indirect Object structure] unless the primary data specifically include examples of such occurrences Given the view of acquisition under discussion, this (learner's) assumption would be built into an account of the learning strategy in some way. Baker's proposal has major consequences for the present discussion. First, it shows — perhaps for the first time in detail — how

472

William C. Ritchie

consideration of the character of the process of acquisition bears on hypotheses about the mature grammar: if a given feature of a linguist's hypothesized grammar is not learnable, given plausible assumptions about the nature of acquisition, then it is discredited as a hypothesis about the mature grammar. Second, it helps to establish the view, extended by Berwick (see below), that the major mechanism of language acquisition is transition from an unmarked grammar to a marked one on the basis of positive evidence from the environment that disconfirms the unmarked grammar but does not disconfirm the marked grammar. Significantly, Oehrle (1976) argues convincingly on quite independent grounds against a transformational treatment of the Dative Alternation and in favor of a lexical treatment like Baker's. The Baker — Oehrle analysis is highly compatible with Chomsky's Projection Principle (Chomsky, 1981) which derives the complements of phrases by "projection" from their heads. Specifically, in essence (Chomsky, 1981: 29): (8)

The Projection Principle (informal) Representations at each syntactic level ... are projected from the lexicon, in that they observe the subcategorization properties of lexical items.

In the present case, Indirect Object Verbs would be claimed to project either prepositional Indirect Object alone (transmit, prepare), prepositionless Indirect Object alone {ask, envy), or both prepositional and prepositionless Indirect Object (send, boil). An important implication of the Projection Principle would appear to be that lexical items are acquired along with their syntax in a single "package". Hence the kinds of structures with which a given lexical item is acquired are specific to that item. This consideration will enter into the explanation to be proposed below for exceptions to the Dative Alternations. Returning to the main theme, Baker's proposal for the Dative Alternation (as well as many other proposals concerning the principles that underlie the process of language acquisition) are subsumable under a certain property of the succession of grammars that characterize the process of acquisition discovered by Berwick (1982: 8-9) and termed the Subset Property. The Subset Property consists in an ordering of the grammars that characterize the process of acquisition according to the "size" of the languages generated

Word-formation,

learned vocabulary

and linguistic maturation

473

by those grammars. Earlier grammars are "more restrictive" in the sense that they exclude structures generated by later grammars. In the structures under discussion, the Subset Property predicts that the child's early grammars will include only one of the two subcategorizations in the Dative Alternation. Given that prepositional Indirect Object is unmarked and the additional plausible assumption that when one member of an alternation of this sort appears in the language it will be the unmarked member rather than the marked, the Subset Property subsumes Baker's solution as a particular case, since it claims, essentially, that the unmarked member of the Dative Alternation — prepositional Indirect Object — will be acquired to the exclusion of the marked member — in this case, prepositionless Indirect Object — unless there is positive environmental evidence for the marked member as well. Again, within the stationary framework under discussion, the Subset Property would be made a consequence of the learning strategy of the learner rather than a property of the class of hypothesizable grammars available to the learner since the latter — Universal Grammar — is assumed to be constant across the acquisition process.2 Let us consider now some consequences of this stationary view. Notice that it claims (apparently) that all that is required to attain the mature grammar is experience with the language. Since there are no changes in the capacity for acquisition — both major elements of the mechanism (the source of hypotheses, Universal Grammar, and the learning strategy) are constant — the stationary model claims that if, by some tragic accident, a child were frozen at the level of cognitive maturation of a two-year-old (or, for that matter, of a newborn), then sufficient experience would lead to attainment of a mature grammar anyway, since (under the present model) experience is all that is required for full acquisition. Though, obviously, the relevant experiment cannot be done, this seems dubious as a likely outcome if it could be. Clearly, a more realistic approach to a theory of acquisition is required.

1.3. The process of acquisition: a non-stationary model Apparently, what is needed in approaching first language acquisition more realistically is a view of the process which takes into account

474

William C.

Ritchie

maturational changes in the child's capacity for language acquisition — in short, what is needed is a non-stationary view in which the capacity for acquisition, as well as the grammar being acquired, changes over the career of the language learner. This view seems rather more in accordance with Chomsky's view of language acquisition (mentioned above) as the "growth of a mental organ" analogous to the growth of a physical organ of the body (1975: 8-11). ... it is taken for granted that the physical structure of the organism is genetically determined, though of course variation along such dimensions as size, rate of development, and so forth, will depend in part on external factors. From embryo to mature organism, a certain pattern of development is predetermined, with certain stages, such as the onset of puberty or the termination of growth, delayed by many years. Variety within these fixed patterns may be of great importance for human life, but the basic questions of scientific interest have to do with the fundamental, genetically determined scheme of growth and development that is characteristic of the species and that gives rise to structures of marvelous intricacy ... [H]uman cognitive systems, when seriously investigated, prove to be no less marvelous and intricate than the physical structures that develop in the life of the organism ... The idea of regarding the growth of language as analogous to the development of a bodily organ is thus quite natural and plausible. and (Chomsky, 1975: 145): ... Notice, incidentally, that biologically determined systems might begin to function only at a particular level of maturation, or after appropriate triggering experience. Thus the theory of faces might be innate, though fully functional only at a particular stage of development. Clearly, a realistic account of language acquisition — one that takes maturational factors into account — must allow for changes in the capacity for language acquisition itself. How is this to be done? Consider once again Berwick's Subset Property. Suppose this property is interpreted not as a claim about the property of some learning strategy as discussed in Section 1.2 but as a condition on successive classes of grammars available to the organism as a matter of biological necessity. In short, the Subset Property is not merely

Word-formation,

learned vocabulary

and linguistic maturation

475

a property of the grammars attained by the learner but is, in fact, a property of the grammars available to the learner at each of a succession of maturational stages with later grammars excluded from those available at earlier stages. Rather than a condition on the order of consideration of grammars made available by a constant source of grammars (that is, Universal Grammar) it is a condition on the succession of sources of grammars themselves with the principles of markedness translated into principles determining the biologically necessary ordering of maturational stages. 3 Consider the Indirect Object structures under discussion here. Under the present, maturational view of acquisition, the unmarked structure — prepositional Indirect Object — would be available to the organism at some fairly early stage of maturation — call it S; — as a matter of genetic programming. With this hypothesis available, the child is able tacitly to characterize (that is, assign a structural description to) sentences exhibiting these structures and to incorporate them into his grammar. On the other hand, the child is not able to characterize the prepositionless Indirect Object structures properly because the required structures are not available and they therefore resist acquisition in spite, perhaps, of rich positive evidence from the environment to the contrary. Let us now consider a possible later stage of maturation — call it Sj — at which the principles required to characterize the marked prepositionless Indirect Object structure become available to the child — again, as a matter of genetic determination. At that point, he will characterize the structure — assuming he encounters evidence for it in the environment — and incorporate it into his grammar. Consider now an additional possibility within a non-stationary view, a still later period in the child's (or, perhaps, the adult's) maturation — call it Sk — at which, once again, the principles required to characterize prepositionless Indirect Object are unavailable. Again, at this stage, the child/adult will be unable to acquire the structure regardless of how rich the environment is in positive evidence for their exsitence. Under this view, the period Sj would qualify as a critical period for the acquisition of the marked prepositionless Indirect Object structure. The essential point, however, is that the class of available grammars is, under the non-stationary view, allowed to vary across the career of the language learner.4 This approach has a number of immediately attractive consequences. First, it makes clear in what sense the course of language

476

William C.

Ritchie

acquisition is genetically determined through the specification of a sequence of sources of hypothesizable grammars. Secondly, it provides an empirically meaningful interpretation for notions of markedness — marked structures are those that are only available during a specified critical period. Finally — and crucially for the present discussion — it allows the natural formulation of questions not only about transition from a maturational stage at which a particular hypothesis is not available to the learner (e. g., S, above) to a stage at which it is available, it also allows the natural formulation of questions about transition from stages at which a given hypothesis is available to a stage at which it is not (as form Sj to Sk). That is, this framework allows the formulation of significant hypotheses about genetically-determined critical or sensitive periods for the acquisition of particular structures. In fact, it is in this way that the study of second language acquisition bears on theoretical questions. We return to these issues below. Notice that under this view, the notion of "Universal Grammar as a universal theory of mature grammars" is replaced by a notion of "sequence of theories of grammars available to the learner at particular stages of maturation." In order to differentiate these two notions, I have elsewhere suggested the term "Operational Universal Grammar at maturational stage S " or " O U G " (Ritchie, 1983). Within the present maturational framework, then, Universal Grammar is seen, correctly it seems, as an "epiphenomenon" of less immediate empirical significance than the sequence of Operational Universal Grammars that characterize the genetically determined maturational schedule in the cognitive domain of language and from which the properties of Universal Grammar are more or less derivable. The mature grammar of a given language, then, constitutes the consequence of the interaction of this maturational schedule and linguistic experience and explanation of features of the mature grammar in terms of such interaction are fully as valid — if not more so in view of the higher degree of falsifiability of an Operational Universal Grammar sequence over a Universal Grammar, the former being open to falsification by data about the course of language acquisition and the latter not — as explanation in terms of Universal Grammar. It should be noted that explanation in terms of maturational schedule and specific experience is not, strictly speaking, opposed to explanation in terms of Universal Grammar, the latter serving within a general research program in linguistic theory as an essential

Word-formation,

learned vocabulary

and linguistic maturation

All

first approximation for the former. Perhaps an example from biology will serve to illustrate the point. One may approach the investigation of some physical organ — say the human heart — in two ways relevant to the present discussion. In the context of comparative anatomy, one can characterize the mature human heart as having certain structural characteristics that distinguish it from, say, the heart of the mature chimpanzee noting those characteristics of all mature human hearts that distinguish them non-accidentally from all mature chimpanzee hearts. The biologist might then, quite reasonably, assume that these specific differences are due to genetic factors of some sort. The second approach would be a developmental one with the purpose of determining how the genetic program "unfolds" through a series of changing potentialities and realizations that eventuate (given suitable environmental conditions) in the mature heart, attending particularly to the intricate timing mechanisms that distinguish, e. g., a healthy mature heart from one that is diseased or defective either genetically or through environmental conditions. In this case explanation would take a rather different form. Surely, the study of comparative anatomy provides a first approximation for the developmental study though a developmental account is more detailed and falsifiable. However, it should be clear that these are explanations of different orders and there is no need to consider them to be "rivals" of any sort. 5

2.0. The Dative Alternation and linguistic maturation Returning to the main task at hand — the explanation of the exceptions to the Dative Alternation — recall that the data are essentially as indicated in (1)—(6), repeated here for convenience. (1)

a. John sent the package to Mary b. John sent Mary the package

(2)

a. John boiled an egg for Mary b. John boiled Mary an egg

(3)

a. John mis-sent the package to Mary b. ?John mis-sent Mary the package

(4)

a. John re-boiled an egg for Mary b. ?John re-boiled Mary an egg

478

William C.

Ritchie

(5)

a. John transmitted the package to Mary b. *John transmitted Mary the package

(6)

a. John prepared an egg for Mary b. *John prepared Mary an egg

Recall that these examples are typical in that derived verbs are generally resistant to the prepositionless Indirect Object structure and latinate Verbs, in the vast majority of cases, reject them entirely. In keeping with earlier discussion, let us turn immediately to the problem of the latinate Verbs. Recall that Oerhle's constraint (7) has little explanatory value — that is, there seems little or no reason to believe that resistance to prepositionless Indirect Object and internal morphological structure can be linked by some genetically given principle characterizable in terms of Universal Grammar — that is, in terms of a theory of mature grammars. Consider now the following maturational account of these exceptions. First of all, assume (a) Chomsky's Projection Principle and its apparent consequences for acquisition (p. 472 above), (b) that marked structures are available to the language learner only during specified, genetically-determined critical periods in the maturation of the organism, (c) that the prepositionless Indirect Object structure is marked and the prepositional Indirect Object structure is unmarked, and (d) that the germanic and latinate strata of the lexicon are acquired at relatively disjoint stages of acquisition, the germanic stratum appearing early and the latinate stratum, representing learned vocabulary, appearing relatively late. Assume, further, that there is, in fact, a genetically-determined critical period for the acquisition of prepositionless Indirect Object (whatever the principled basis of its markedness may be), that germanic vocabulary is acquired during that ciritcal period and that latinate vocabulary is acquired after its close. Then the exceptions to the Dative Alternation are explained on grounds of genetic factors — specifically, the sequencing of the critical period for the acquisition of prepositionless Indirect Object and the input of latinate lexicon from the environment. Let us consider the parts of this account one-by-one. Chomsky's Projection Principle has already been discussed. The general assumptions about markedness and critical periods is quite natural and is supported, first, by the earlier discussion of the Subset Property with respect to the beginning of critical periods and, second,

Word-formation,

learned vocabulary

and linguistic maturation

479

(somewhat more crudely) by the generally unmarked character of languages acquired by adults as second languages (e. g., pidgin languages) with respect to the ends of critical periods. The assumptions of the markedness of prepositionless Indirect Object and the unmarkedness of prepositional Indirect Object fall out of current work in the Government-Binding framework (particularly, as noted above, Case Theory — see the analyses of Chomsky, 1981, and Czepluch, 1982/83, for details; detailed discussion is beyond the scope of this paper). The early acquisition of germanic lexicon and later acquisition of learned, latinate lexicon seems indisputable in general though it is, of course, open to empirical test. (See Rinsland, 1945, for — admittedly crude — empirical support for the maturational explanation.) In addition to the fact that it is based on a genetic account, the hypothesis has a number of immediate conceptual advantages over non-maturational accounts. Note that it implies that if a native speaker of a language other than English attempts to acquire the prepositionless Indirect Object structures as they are represented in the grammar of the native speaker of English after the age at which native speakers of English begin to acquire latinate lexicon, the hypothesis predicts that, as a matter of biological necessity, the learner will be unable to do so regardless of input from the evironment since the prepositionless Indirect Object structures will no longer be available to the learner within his tacit notion of "possible human language". (This, in fact, is a paradigm case of the relevance of second language data to linguistic theory.) From a historical point of view, if this hypothesis is correct, and, in addition, Czepluch (1982/83) is correct in attributing the markedness of the prepositionless Indirect Object structure to a typological difference between languages which, like Modern English, mark Case structurally only, and languages which, like Modern German, mark Case inherently or lexically as well, then the presence in Old English (a language sharing crucial typological features with Modern German in contradistinction to Modern English) of a large class of Verbs accepting prepositionless Indirect Object and the gradual loss of this structure over time (as mentioned above, see Visser, 1963) correlated with its acquisition only during specific periods in the maturational process is explained. As another possibility, the case of exceptional derived Verbs might also be explained as a consequence of the sequencing of the period during which the morphological rules that generate derived forms

480

William C. Ritchie

and the period during which prepositionless Indirect Object is available with the former following the latter. In all of these cases the mature grammar of English is regarded as the consequence of the interaction of maturation — conceived of as a fairly delicately timed sequence of critical periods — and experience. Under these circumstances the mature grammar should be expected to exhibit the residue of structures acquired early and those acquired late with no necessary, systematic relation among them within a theory of mature grammars such as Universal Grammar, as is the case with the exceptional behavior of many lexical items and their subcategorizations, the case of the Dative Alternation among them.

3.0. Conclusion In conclusion, I have suggested an explanation for certain recalcitrant features of the Dative Alternation as a conception of the linguistic genetic endowment of the human species not in terms of a single universal theory of mature linguistic structure — Universal Grammar — but rather as a program determining a sequence of such theories each of which characterizes the class of grammars available to the learner at a specific maturational stage.

Notes 1. That this is, in Chomsky's view, only a first approximation is clear from other discussions of his — in fact, discussions in the works from which these quotes are taken. The purpose for including this view here is expositional — to contrast it with the nonstationary view discussed later in the text. 2. It is not at all clear that Berwick intends this interpretation. See note 3 below for discussion. 3. It should be noted that Berwick himself does not appear to take a definite position with respect to stationarity. However, since the operation of a learning strategy presupposes a certain class of hypotheses, the placement of heavy restrictions on the class of hypotheses would appear to constrain the account of acquisition more heavily, hence to constitute a stronger claim. The reasoning is the same as that which has motivated the research strategy pursued in generative grammar over the last twenty years, of accounting for the speed of language acquisition by placing heavy restrictions on Universal Grammar rather than rendering learning strategies more powerful. 4. See Chomsky (1975: 118-122) and Lightfoot (1982: 172-184) for some relevant discussion.

Word-formation, learned vocabulary and linguistic maturation

481

5. Note the unlikelihood of the biologist's constructing a theory of mature human hearts — parallel to Universal Grammar — and then attributing representation of this theory to the "initial state of the organism" and characterizing the development to maturity as passage from that state — in which the full class of mature hearts is somehow represented in the organism — through a series of physical states in which only one of that class is so represented, the process determined by some "growing strategy" parallel to the learning strategy in the stationary model sketched in the text. It is not at all clear how such a model would contribute to an understanding of the process of development in the biological case nor is it clear how its analog — the stationary model of the text — contributes to an understanding of the "growth" of language.

References Akmajian, Adrian — Heny, Frank 1975 An introduction to the principles of transformational grammar (Cambridge, Mass.: Μ. I. T. Press). Baker, C. L. 1979 "Syntactic theory and the projection problem," Linguistic Inquiry 10: 533-581. Berwick, Robert C. 1982 Locality principles and the acquisition of syntactic knowledge (Unpublished Ph. D. Dissertation, Μ. I. T.). Braine, Martin D. S. 1971 "On two types of models of the internalization of grammars," The ontogenesis of grammar: a theoretical symposium, edited by D. I. Slotin (New York: Academic Press), 153-187. Brown, Roger — Cazden, C. B. — Bellugi, U. 1969 "The child's grammar from I to III," Minnesota Symposium on Child Psychology, vol. 2, edited by S. P. Hill (Minneapolis: U. of Minnesota Press), 28-73. Burt, Marina K. 1971 From deep to surface structure (New York: Harper & Row). Carlson, Greg — Roeper, T. 1980 "Morphology and subcategorization: case and the unmarked complex verb," Lexical Grammar, edited by T. Hoekstra et al. (Dordrecht: Foris Publications), 123-164. Chomsky, Noam 1965 Aspects of the theory of syntax (Cambridge, Mass.: Μ. I. T. Press). 1975 Reflections on language (New York: Pantheon Books). 1980 Rules and representations (New York: Columbia U. Press). 1981 Lectures on government and binding (Dordrecht: Foris Publications). Czepluch, Hartmut 1982/83 "Case theory and the dative construction," The Linguistic Review 2: 1-38. Lenneberg, Eric H. 1967 Biological foundations of language (New York: John Wiley & Sons). Lightfoot, David 1982 The language lottery: toward a biology of grammars (Cambridge, Mass.: Μ. I. T. Press).

482

William C. Ritchie

Oehrle, Richard T. 1976 The grammatical status of the English dative alternation (Unpublished Ph. D. Dissertation, Μ. I. T.). Pinker, Steven 1981 Comments on K. Wexler, "Some issues in the theory of learnability," logical problem of language acquisition, edited by C. L. Baker and J. McCarthy (Cambridge, Mass.: Μ. I. T. Press), 53-63. Rinsland, H. D. 1945 Basic vocabulary of elementary school children (New York: Macmillan). Ritchie, William C. 1983 "Universal grammar and second language acquisition," The acquisition of symbolic skills, edited by D. Rogers and J. Sloboda (New York: Plenum), 471-483. Visser, F. Th. 1963 An historical syntax of the English language (Leiden: E. J. Brill).

H A N S SAUER

La3amon's compound nouns and their morphology 1. Introduction 1.1. The text and previous research La3amon's Brut ( L a j B ) is an Early Middle English alliterative poem of more than 16000 long lines, which was written around 1200.1 It survives in two manuscripts from the second half of the 13th century. Manuscript C (London, British Library, Cotton Caligula Α.IX) is nearer to the original; manuscript Ο (British Library, Cotton Otho C.XIII) shows a conscious effort at modernization where, among other changes, compounds are often, but by no means always, replaced or omitted. Much more rarely, Ο introduces new compounds. 2 A list of nominal compounds in La3amon, based on manuscript C, is given by Oakden (1935) in his survey of Middle English alliterative poetry. 3 Oakden usually omits compounds that occur in manuscript Ο only, whereas Ο is also taken into account here. Oakden (1935: II, 130) states that there are 411 nominal compounds in LajB. His list is therefore impressively long, but it is neither complete nor entirely consistent and also contains some items which have no place here. For instance, Oakden omits /Eldrihten and alwaldinde; he has bermannen but not bercnihtes; he gives Brutlond, but not Brutleoden, Brutpringen and Brutware.4 He largely excludes combinations with suffixoids, namely -dom, -fast, -ful, -had, -les, -rceden, but he includes the prefixoid kine- (cf. chapter 2.13 on prefixoids and suffixoids in LajB). He also includes a few words which are clearly formed with prefixes, e. g., cer-wene, cer-witte, elreordi and el-peod, som-rune, step-moder. These inconsistencies are possibly due to the fact that Oakden was not interested in the morphological aspect of nominal compounding but rather in the distinction between poetic and prose compounds and the relation of the Middle English poetic compounds to the Old English poetic tradition. This distinction does not, however, seem to bear much,

484

Hans Sauer

if any, relation to the morphological patterns of word-formation. Moreover, as far as La3amon's Brut is concerned, Oakden's results are probably not completely reliable due to his omissions and wrong inclusions.5 I shall not pursue the question of prose and poetic compounds here but shall instead concentrate on the morphological description of the compounds in La3amon's Brut. I shall also confine myself to compound nouns, although nominal compounds strictly speaking comprise both compound nouns and compound adjectives. Compound nouns are far more numerous in LajB than compound adjectives. 6 I have tried to collect all compound nouns from the general vocabulary, whereas for place names and personal names a representative selection only is given, with no claim to completeness. I hope that this analysis will not only illuminate some aspects of La3amon's use of nominal composition, but will also point out problems in and suggest solutions for the description of compounds in other Middle English texts. As the critical remarks about Oakden have shown, such an undertaking is not possible without discussing a few theoretical points first. It must also be stressed that a morphological analysis alone cannot cover all the aspects of nominal compounds; it has to be supplemented by the description of their syntactic, semantic and other structures (cf. for example Marchand 1969: 54-59; Warren 1978: 5-52; Kastovsky 1982: 168-215). A systematic analysis of these is not possible here, but even in a morphological description they have to be taken into account for certain points (see for example 2.1. below).

1.2. The definition of nominal compounds and their separation from syntactic groups Like other linguistic concepts, the concept of nominal compound is widely used but difficult to define in a way that is at the same time short, simple and satisfactory. 7 Nominal compounds can be defined as nouns and adjectives consisting of two immediate constituents, both of which are words (or potential words). But there is not only a relatively clear-cut central area, e. g., noun/noun, adj/ noun, etc., but also a marginal area shading off into other patterns of word-formation. In this area we must include combinations containing locative particles, e. g., ModE insight, outlaw, synthetic compounds, which result from a combination of compounding and

Lajamon's

compound nouns and their morphology

485

suffixation, e. g., ModE theatre-goer, house-keeping; and the socalled exocentric compounds, e. g., ModE greenhorn, cutpurse. All of these are in one way or another problematic with respect to the definition given above, 8 but are nevertheless considered to be nominal compounds here because they are nouns or adjectives and contain two words (or potential words). Formations which must be excluded from nominal compounds are derivations from compounds, for example Old English god-spellere 'evangelist', which is derived from god-spell 'gospel', and combinations of two words which are used as adverbs only, for example lim-mele 'limb by limb' {LajB CO 12 785), which is wrongly included by Oakden. 9 Even more difficult than defining the nominal compound is the attempt to draw a dividing line between compounds and parallel syntactic groups. 10 It is usually assumed that the compound is in some way isolated from the syntactic group, but the criteria available are not sufficient to decide all cases. The word-division of the manuscripts is unreliable. The criterion of accent is clear in theory for Old English and Middle English: compounds have a main accent on their first element. 11 In practice this is, however, of little value because in prose texts accent often cannot be determined, whereas in poetry the ordinary accent pattern can be blurred by the pattern of verse accent. The criterion of morphology and morphosyntax covers a variety of phenomena. 1 2 One of these is that pronouns and articles refer to the compound as a whole, i. e., not just to the first element, but also to the second. Because in La3amon traces of inflection are still preserved, a decision is possible in some cases, but sometimes there is ambiguity or even conflicting evidence; moreover La3amon does not always reflect Old English usage correctly. 13 Let me give some examples: pere lond-sorje (manuscript C) must be a compound, because lond was neuter and sorje feminine and pere refers to sorje. be pere sce-brimme 'by the sea-coast' (manuscript C) is more difficult. It looks like a syntactic group, because see was feminine and brimme neuter, and pere should refer to see. In manuscript O, however, the form bi pan see-brimme certainly is a compound. A decision is impossible with the construction in pare halle-dure, because both halle and dure were originally feminine and the article therefore could refer to either. The combination mid rede golde stauen ( C I O 555) seems semantically to be a syntactic group meaning 'letters of red gold', and not 'red gold-letters', cf. ring of rede golde(C 15 376). But morphologically golde stauen should be a compound because

486

Hans Sauer

the genitive of gold was usually goldes.14 Another difficult case is anes bule hude (manuscript C, 7080): Anes seems to refer to bule, but if bule were a genitive one would expect the form bules.15 The reading of manuscript O, a bole hude, certainly represents a compound. The criterion of semantic isolation is problematic, too. Three conclusions can be drawn from these considerations. Firstly, when dealing with Middle English compounds, a number of formations often remains whose status is doubtful. Secondly, lists of compounds will therefore frequently contain a number of uncertain examples and thirdly, the figures given for the compounds of a text or for certain types of compounds often cannot therefore be taken at face value, but just as rough approximations. This applies to Oakden as well as to the figures given below.

1.3. The morphological analysis of nominal compounds According to Hans Marchand, the morphological analysis of compounds has to proceed on the levels of morphological shape and morphological structure (Marchand 1969: 54; cf. Kastovsky 1982: 214). Marchand's approach has been questioned or simply ignored by some recent authors (for instance Adams 1973; Pennanen 1982: 241-261; Bauer 1983), but in spite of the difficulties it raises, it still seems useful. On the level of morphological shape, the word-class and the morphological status of the elements making up a compound are described, for instance noun + noun for boc-fell 'parchment', noun + s + noun for cunnes-mon 'kinsman', noun + adjective for snaw-white, etc. Problems arise at this level when the word-class of one element is ambiguous; on this question see 3.7. below. The level of morphological structure is somewhat more sophisticated. It is based on the assumption that the immediate constituents of a compound form a syntagm and stand in the relation of determinant and determinatum. In English, the principle is for the determinant to precede the determinatum. 16 The determinant modifies the determinatum and restricts its reference; thus the whole compound usually becomes a hyponym of its determinatum. The determinatum in its turn determines the word-class and the lexicosemantic category, for example animate vs. inanimate. This relationship is obvious with formations like cnaue-child 'boy' and meidenchild 'girl', where child is the determinatum and cnaue or meiden

Lajamon's

compound nouns and their morphology

487

the determinant; it is less obvious at first sight in formations which have traditionally been called exocentries, e. g., ModE redcoat and cutpurse, where it leads to the postulation of a zero morpheme as determinatum (redcoat/Φ; cutpurse/0).17 In other cases, especially with synthetic compounds, 18 the question of a dual analysis of the morphological structure arises. If we start from a syntactic paraphrase and compare ModE theatre-goer 'someone who goes to the theatre' with ModE redcoat 'someone who has a red coat', we are forced to take 'someone' as the determinatum, which is represented by a zero-morpheme in the case of redcoat and by the suffix -er in the case of theatre-goer. If, on the other hand, we compare theatre-goer to compounds like ship-owner and car-dealer, where owner and dealer also occur as independent words, and to teamerchant, where the determinatum merchant cannot be further divided morphologically, we are justified in taking the deverbal formation goer as the determinatum of theatre-goer, which in this and similar cases seems the more natural analysis from the morphological point of view.19

1.4. Establishing morphological types of nominal compounds Taken together, the concepts of morphological shape and morphological structure yield the morphological types of compounds as they are more or less completely listed in the handbooks (cf. Marchand 1969: 60-95; Koziol 1972: 51-82; Hansen et al. 1982: 52-63; Bauer 1983: 201-212). But setting up these types is not as easy as it might seem. A glance at the handbooks reveals considerable variation in classification. Some authors arrange their material into relatively few and accordingly fairly broad types, whereas others have a far more detailed classification. 20 A middle way certainly has to be found in establishing morphological types. If the classification is too broad, important patterns will be lost or obscured. On the other hand, not every morphologically distinct combination of words and morphemes that occurs constitutes a type of composition. Some formations have to be classified as subgroups of a particular type, and a few are exceptions which do not represent a productive type of composition at all. Examples for the former will be given below (see e. g., 2.1. and 3.1.), an example of the latter is Old English and Middle English wel-ded 'good deed' (also used by La3amon), which can be described as consisting of adverb/noun,

488

Hans Sauer

but since it is more or less the only example of such a combination in the history of English, the combination adverb/noun can hardly be called a productive type of nominal composition in English. 21 In the following survey of types the emphasis is put on the determinatum and the arrangement is in the order of: primary determinatum, i. e., determinatum which is not deverbal; deverbal determinatum; zero determinatum; other cases. Within this order, the arrangement is according to the word-class of the determinant in this sequence: noun, adjective, numeral, pronoun, particle, verb. This arrangement certainly does not solve all the problems, but perhaps it results in a certain amount of coherence and consistency in the presentation. 22

2. The types of compound nouns in La3amon's Brut23 2.1. Noun/noun The type noun/noun is by far the most frequent of all types in LajB; this is probably true not only for Early Middle English but also for the entire history of English. 24 I have counted 268 different noun/noun compounds in LajB, excluding place and personal names; a selection of place and personal names has yielded 66 more noun/noun compounds. Out of the 268, 129 have been inherited from Old English and 139, more than half, are new formations in Middle English. Morphologically noun/noun compounds can be divided into determinative compounds and dvandvas (often called copulative compounds). The former can be analysed according to the pattern of determinant/determinatum, the latter have no determinant/determinatum structure. A major division is therefore sometimes made between determinative compounds and dvandvas. 25 It seems, however, preferable to follow Marchand in dividing the noun/noun compounds syntactically into copula compounds, where the verb 'to be' is inserted in a paraphrase, e. g., cnaue-child 'the child is a cnaue', and into rectional compounds, where another verb has to be inserted, e. g., craft-mon 'the man practises a craft', gold-wir 'the wire consists of gold'. In this classification, the dvandvas form a subgroup of the copula compounds. The rectional noun/noun compounds are a much larger group than the copula compounds. Morphologically they are not isolated

Laj am oncompound

nouns and their morphology

489

from syntactic groups in Early Middle English, because some formations certainly go back to genitive groups, others possibly so, and in some cases the genitival character might still have been felt, e. g., Mone-dcei, halle-dure, leode-celder (see 3.3. below). The copula compounds are morphologically isolated; 26 they can be subdivided on semantic grounds into additive, attributive, subsumptive and tautological compounds, although there are borderline-cases which are difficult to classify. This subdivision is founded on morphology only to a minor extent, nevertheless I believe it is useful to mention it here. Additive copula compounds is another term for dvandvas. They designate two entities (denotata) that together form a higher unit, as in the Old English pattern adum-swerian 'father in law and son in law' and the Modern English pattern Austria-Hungary21. They do not figure in La3amon and are very rare in Early Middle English generally. The attributive compounds show two aspects of the same entity and in contrast to the dvandvas the second element can be taken as determinatum and the first as determinant. 28 A few groups can be distinguished. One denotes sex: cnaue-child, cniht-bcern, maidenchild, wap-men29 and wif-men (wimmen) — the latter two occur several times as a contrasting pair; another status: alder-man, munecchild,30 wine-mceies; a third nations, countries and towns: Brutleoden, Brut-pringen, Sax-leode, Scot-leode, Scot-peode; Flaundrelond, Gric-lond, Puille-lond 'Apulia', Lundene-burh, Rome-burh. Eitlond 'island', hoker-leop, hoker-word, hux-word literally 'insult-song', 'insult-word' are probably also attributive. 31 In the subsumptive copula compounds, the determinant is the hyponym and the determinatum its hypernym (i. e., all the semantic features of the determinatum are contained in the determinant), whereas in most other compounds, the whole compound is a hyponym of its determinatum. Subsumptive compounds occur in biological and mineralogical classification (cf. Carr 1939: 325-326); LajB has examples for the latter only: jim-ston, marbre-ston, marman-ston. Copulative designations for time should probably also be classified among the subsumptive compounds: ceuen-time and jeol-dcei.32 Names for peoples and countries, however, should probably be included among the attributive copula compounds. The subsumptive copula compounds and some of the attributive copula compounds are pleonastic (cf. Carr 1939: 330 for terminology). The determinatum could be omitted because its semantic elements are

490

Hans Sauer

contained in the determinant (e. g., Brut-leoden, cnaue-child, jimston), but it is not synonymous with the determinatum. In the tautological compounds, our fourth group, both elements are synonymous. 33 It is, of course, often difficult to know exactly to what extent words are synonymous, especially in the older stages of the language, but the following formations probably or possibly belong to this subgroup (cf. Carr 1939: 330-339; see also Strauss 1980: 305-315): driht-folk 'people, army, host', gleo-drem 'mirth, music', leod-folk 'people', lond-riche 'country, state', 34 mor-uenne 'moorland fen, marsh', pmt-wrenche 'trick', rake-teje 'chain, fetter', peod-folk 'people', prym-ferde 'army'. Run-stauen, writ-runen and a few others are problematic. 35 The reasons for the formation of tautological and pleonastic compounds vary. With some, the determinant was a loanword (jim-, marman-, marbre-) or obsolescent (drihtmarman-) and needed some semantic support to motivate it 36 better; with others, the determinant perhaps serves as a kind of intensifying element (possibly driht-, leod-, peod- and prym-); on these, see 2.13.4. below. In poetry, metrical considerations could have played a role, too. Apparently, most of these formations were not felt to be tautological or pleonastic by the speakers. 37

2.2. Noun + s/noun In this type, which can be named genitive compound, the determinant is a noun in the s-genitive, and the determinatum is a noun. The type is morphologically not isolated from syntactic genitive groups, especially since in Old and Middle English a noun in the genitive could much more easily premodify another noun than is possible in Modern English. According to some handbooks, genitive compounds to not exist, at least not in Modern English 38 . In Early Middle English, the type certainly existed, but its definition for this stage of the language is problematic for another reason: Old English had a variety of genitive morphemes (-e, -a, -an, -s, -0 for the genitive sg.). Most of these, apart from zero- and 5-genitives, were levelled under -e in Early Middle English. An -e at the end of the determinant can, however, sometimes be interpreted differently: it can also represent an Old English nominative (in -a, -e, -o, -u), it can have been generalized from an inflected form or it can be intrusive. 39 Because it is sometimes difficult to decide which of these possibilities applies in a given case, formations where the

Lajamon's

compound nouns and their morphology

491

determinant ends in -e are included here under the type noun/ noun. 40 In LajB I have counted 30 different noun + s/noun combinations (C has 23, Ο has 20) plus nine place-names. About half of the 30 go back to Old English (16) and half are new formations in Middle English (14). A few groups can be distinguished in which the compound status of the members is largely certain. One of these are place-names with the determinant in the genitive, e. g., Amberes-bury, Beljces-jate or Bellinges-jate, Cristes-chirche, Cuninges-burh or Conanes-borh, Ludes-jate, Teynis-wich. * Amber- and *Lud- do not occur as independent words (cf. Blenner —Hassett (1950) for details), which makes the compound status of the combinations in which they occur certain. Genitive place-names which might be semantically isolated syntactic groups are Agnetes Hull and Michaeies Munt. A number of genitival combinations with -tun as second element are probably not compounds, but syntactic groups, namely Lundenes tun, Tintageles tun, Verolames tun, Winchcestres tun. Their semantic structure, however, is like that of the attributive copula compounds and the second element is semantically superfluous, because the semantic feature 'town' is already implied in Lunden, Tintagel, etc. The days of the week form a second group. Of these, Sateresdai, Tis-dcei, punres-dm (see 3.6. below) and poris-dai have an sgenitive in Early Middle English. Their compound status is certain because most of them were probably obscured by La3amon's time (see 3.6. below). A third group whose compound status is fairly certain has -man as determinatum: There are seven such formations in LajB: cunnesmon, erendes-mon, Icedes-mon, sondes-mon, steores-mon, peines-men, wardes-men41. With most other combinations it is difficult to decide whether they are compounds or not. Examples from La3amon are dcejes liht, dceöes kare, Domes-dai, heruestes dcei, hinges istreon, midewinteres doei, twines pred, welles-stram, whiles-water, whalesbone. In many of these, the genitive is classifying (as opposed to specifying), i. e., it does not refer to a particular instance of the first element, but to the first element in general. Whales-bone, for example, does not refer to the bones of a particular whale, but to the bones of whales in general. It is, however, doubtful whether the existence of a classifying genitive alone justifies regarding a combination as a compound. There are moreover sometimes ambiguous cases. Kinges istreon could mean 'the offspring of a particular king'

492

Hans Sauer

or 'the offspring of kings, royal offspring'. Domes-dai is specifying, nevertheless it is a compound in Modern English and probably was one in Early Middle English, too; 42 midewinteres dai, on the other hand, is doubtful. Noun/noun compounds and genitive combinations were often interchangeable in Old English and Early Middle English; apparently, these formations were semantically equivalent (cf. Strauss 1980: 313-314). In LajB, manuscript Ο several times has a genitive combination where manuscript C has a compound (cf. Begholm 1944:41), but in some instances noun/noun-compounds and genitive combinations co-exist in manuscript C, too. The examples are: 43 manuscript cerchebiscop-stol biscop-stol castel-jeat dai-liht leod-folk lif-griö mod-sorje mon-drem munec-claöes sond-mon stcen-cun steor-men welle-water weorld-scome

C C CO CO CO

C C c c c c

CO

C c

manuscript

0 Ο c

archebissopes stol bissopes stol castles jcet dceies liht, daises liht leodes-folk lifues grid modes sorje marines drem monekes elopes sondes-man stones cun steores-man, steres-mon welles water weorldes scome

ο c

Teynis-wich

0

CO

c

CO

c 0

CO CO

0

CO

and among the place-names: Teinne-wic

c

Genitive compounds are usually rectional but, apart from the pattern Lundenes tun, which probably represents a syntactic group, there are a few which might be copulative, peines-men 'the men are feines' is fairly certainly copulative, whereas sandes-mon and wardes-mert are ambiguous. Sonde can mean 'message' or 'messenger', ward can mean 'guard' or 'guardian'. They would be rectional, if the first meaning was intended but copulative, if the second was intended. 44

Lajamon's

compound nouns and their morphology

493

2.3. Adj/noun Like the genitive compounds, the type adj/noun is morphologically not isolated from the corresponding syntactic group adj + noun in Middle and Modern English. 45 It is therefore impossible to give exact numbers for this type. In LajB I have counted 37 formations from the general vocabulary which are or could possibly be compounds, and five names of people and places. 19 of the 37 are Old English formations, 18 Middle English ones. Godd-spell, leof-man, midden-erd and mid-niht are certainly compounds: leof-man and probably also godd-spell are on their way to obscuration, while midden- does not occur on its own. Other formations might also be syntactic groups in LajB, for example: Englisce-men, Frensce-men (but both are compounds in Modern English), freo-king, hcejhe-dcei, lift-hond and riht-hond (rihthand was a compound in Old English), luper-craftes (cf. mid luperen heore craften, C 14467), middel-broper (cf. pe midieste broöer, C 6442), soö-cwide (which was a compound in Old English, but cf. of sode mine Spellen, C 10972), priue-men, weste-pades. Heh-king existed as a compound in Old English, but since LajB has pes hejes kinges (C 23) and for hehne king (C 14476), the form heh king (C 1022) would rather seem to be a syntactic group. Oakden lists heje men (C 14831) as a compound, but a construction like was swa hah mon (C 1284) makes this unlikely: the adverb swa can only refer to the adjective hah and not to the noun mon (cf. OED s. v. So.). Occasionally, a combination occurs as a syntactic group and as a compound: ne beo he na swa leof mon (C 6919) is a syntactic group, while the frequent leof-mon 'lover, beloved' is a compound, which is shown by its semantic isolation plus the fact that in manuscript Ο it often appears in its obscured form lemman. Combinations like ludere stefne, soöere worden, to wraöere hele 'with a loud voice', 'with true words', 'to one's misfortune' are interesting because the ending of the adjective goes back to the strong dat. sing, fem., but certainly in soöere worden and probably in the others, too, it is a frozen form no longer used in its original function: word was neuter and worden reflects a dat. pi.; it certainly could not originally take a strong dat. sg. fem. adj. Oakden lists wraöerehele as a compound, but not the other formations. Very likely none of them is a nominal compound in LajB; apparently they were syntactic groups used as adverbial forms. 46 This is also

494

Hans Sauer

clear from forms like to godere pire hale (C 1796), to wroper heore hele (C 247), mid ludere stefne (C 716).

2.4. Numeral/noun Compounds of the form numeral/noun could be listed among the type adj/noun, but are here counted as a separate type. 47 There are not many of these formations. LajB has seven which are or could be compounds. Four of them are Middle English formations: duszepers, a loanword from French, hcelue-suster, haluen-del, seouen-dajes (soue-dajes), which possibly was a syntactic group in LajB. One formation is Old English: pridden-dale. Seoue-niht existed as a bahuvrihi adj. in Old English, but is used as a noun in Middle English. 48 Feowertene niht (manuscript C 12 814, 14093, etc.) was probably still a syntactic group, but fourti-niht (manuscript Ο 12 814, 14123) has coalesced into an (obscured) compound (see Faiß 1978: 167-168).

2.5. Pronoun/noun The type pronoun/noun is very weak, not only in LajB but in Early Middle English generally. LajB just has the Middle English coinage ΑΪΙ-drihten, a hapax legomenon 49 Nevertheless the type is listed here because it existed in Old English and is also productive in Modern English, and because the compound /El-drihten shows that the type was not dead in Middle English. /El-drihten is morphologically isolated, because all can only refer to a plural or to a collective noun. 50 This compound is used in LajB in the combination ALldrihten God. The MED therefore classifies al-drihten not as a noun, but as an adjective. 51 This classification is not justified: ^El-drihten in LajB is probably a noun, put in apposition to God, similar to the construction Lauerd Drihten Crist (CO 12 760).52

2.6. Particle/noun Here the determinant is a particle, the determinatum is a noun. Without aiming at completeness for this type, I have counted 19 formations; 8 of these are Old English and 11 are Middle English. Particles are words that are used independently as prepositions,

Laj am on's compound nouns and their morphology

495

adverbs or both. 53 The type is morphologically isolated, but in composition the particles approach the status of prefixes which makes the type particle/noun somewhat marginal among the types of nominal composition. The determinatum is often a deverbal noun because these compounds are frequently derived from particle/ verb or verb/particle-constructions, e. g., ingan, gan in are nominalized as in-jeong. Examples from LajB are fore-ward, in-siht, sunder-rune, to-nome, under-king, ut-lajen,54 wiöer-craftes, wiöerdeden and several other compounds with wider- as determinant. 2.7. Vb/noun The type vb/noun is not very frequent but productive in Early Middle English. There are 9 certain or possible examples in LajB, four of which are Old English and five are Middle English formations. The type is morphologically isolated, because there is no corresponding syntactic group of verbal stem + noun. Nevertheless a few of these formations pose morphological problems. These are linked with the history of the type. 55 It is assumed that the type arose from such noun/noun-compounds where the determinant was a deverbal noun: in these cases the determinant could come to be directly associated with the underlying verb. There appear to be two or three examples for this development in LajB: The forms plaje-iueren and spcec-hus in manuscript C are still noun/ noun-compounds (OE plega and sp(r)cec were deverbal nouns), 56 but the corresponding forms plei-uere and spec-hus in manuscript Ο can also be interpreted as verb/noun-compounds. 57 The same development could have taken place in the determinant of wallestream 'spring'. 58 In the majority of formations, however, the determinant is certainly the verb because there is no corresponding deverbal noun: ber-cnihtes and ber-men 'porters', beor-time (manuscript O) 'time of birth', hcere-word 'word of praise', til-aht and welde-king, probably 'king who wields [sc. his rice]'.59

2.8. Noun/vb + 0, i. e., noun/zeroderived deverbal noun This type is relatively strong in LajB. I have counted 37 formations, 20 of which are Old English and 17 Middle English, plus the name Dorchestre-seten. In Old and Middle English this type formed partly agent nouns, e. g., eorö-tilie, dure-ward, partly action nouns, e. g.,

496

Hans Sauer

balu-feht, blod-gute, mon-lure.60 By definition the type forms synthetic compounds, i. e., the formations are nominalisations of sentences and the determinatum does not necessarily occur as an independent word, to give a Modern English example: bellhop 'hotel porter' literally is 'someone who hops at the call of the bell'. 61 In as much as the determinatum does not occur as an independent word, the type is morphologically isolated. There are, however, not many such cases in LajB: the determinatum is not an independent word (neither in Old English nor in Middle English) in Dorchestreseten, erend-rake, mon-sware, here-toje, grist-bat (erend-rake was probably, here-toje possibly obscured). Apart from -bat all these determinata go back to Old English. But the majority of the determinata occurred also as independent words in Old English, e. g., ward, which is frequent in LajB (bac-ward, bur-ward, dureward, jcete-ward, etc.), also gute {blod-gute), feht, fiht (balu-feht, stcel-feht) and so on. 62 Moreover it is doubtful whether all independent nouns which are etymologically deverbal were still felt as deverbal synchronically, i. e., in Early Middle English. Synchronic and diachronic analysis do not always tally here and the same determinatum can also be felt to be deverbal in one compound and primary in another. 63 A number of these formations have therefore been listed here under the type noun/noun because their deverbal character was possibly no longer evident, e. g., herre-feng 'booty', originally 'what has been caught by the here [i. e., army]', 64 and some of the compounds listed under noun/vb + 0 might possibly be better placed in the noun/noun group. It is for example doubtful whether ward was still regarded as deverbal ('someone who guards something'), nevertheless it is included as a zero-derived deverbal noun here. 65 Fortunately there are a few examples in LajB showing that at least in some formations the determinatum was still felt to be connected with the corresponding verb and that the type as such is therefore justified. (1) Feond-rces: to him reasde mid his feondreases ( O i l 956; C has a different verb: fusden); (2) blod-gute: this compound occurs several times in manuscript C in the clause par was muchel blod-gute, e. g., 9552, but it is regularly converted back into a sentence in manuscript O: pere was mochel blod igote. (3) eorde-itilie: pat lond ... nes hit neuere itiledpurh ncenne eoröe-itilie, C 4999-5000. (4) leod-swike: swike occurs independently and in a genitive phrase (Malgod, swike he wes C 2198; leodene swike C 378), but also in connection with the verb: pa pohten he to swiken ... 7 pus leh pa leod-swike ... C 6457-59. 66

Lajamon's

compound nouns and their morphology

497

2.9. Noun/vb+ /, i. e., noun/deverbal action noun with suffix -t This type is somewhat doubtful and is not mentioned in the handbooks. Like the type noun/vb + 0, it forms synthetic compounds by definition; but in all the examples from LajB the determinatum exists as an independent word, too; a classification of these formations as noun/noun cannot therefore be excluded. Nevertheless I list them as a separate type because they are morphologically distinct and the deverbal connection still seems obvious in some cases.67 LajB has six examples, three of Old English, three of Middle English origin; but the five which are not obscured have just two different determinata: slceht in feond-slcehtes, mon-slceht, wcel-slaht and wcen-slahtes; seht, siht in heh-seht 'eye-sight' (and cf. in-siht).68

2.10. Noun/vb + ere The type noun/vb + ere, i.e., noun/deverbal agent noun with the suffix -ere is represented in LajB by just one formation, namely wcei-witere 'guides', literally 'those who know the way', a Middle English formation which is a hapax legomenon and occurs in manuscript C only — manuscript Ο replaces it by woei-wittie.69 The scarcity of this type in LajB may seem surprising at first sight because the type existed in Old English and in the course of Middle English largely ousted its competitors in the formation of compound agent nouns, viz. the types noun/vb + 0 and noun/vb + ende.10 As we have just seen, the type noun/vb + 0 is still much stronger in LajB. This is, however, generally true for Early Middle English texts. Moreover, the virtual absence of the type noun/vb + ere from LajB is perhaps one of the features that connect it with the Old English poetic tradition, because according to Carr the type noun/ vb + ere existed in Old English prose but was not used in Old English poetry. 71

2.11. Noun/vb + ing, i.e., noun/deverbal action noun with the suffix -ing This type is also very weakly represented in LajB. There are just two examples, grist-batinge 'gnashing of teeth', which goes back to

498

Hans

Sauer

Old English, 72 and heortne-graning 'groaning of the heart', which is Middle English but might be a syntactic group, because granung occurs as an independent word in Old and Middle English and heortne could be a reflex of the Old English genitive heortan.

2.12. Exocentric compounds: bahuvrihi-nouns (noun + noun/0) Exocentric compound nouns are extremely rare in LajB. The only example is Draken-hefd, which belongs to the type of bahuvrihinouns. The bahuvrihi-nouns can be described morphologically as noun + noun/0, i. e., they consist of a composite determinant and a determinatum represented by a zero morpheme. They can be described more precisely with a syntactic paraphrase: 'Someone who has what is indicated by the complex determinant'. This paraphrase, however, has to be extended for Draken-hefd, which refers to Uther Pendragon and is explained in the text. It is not an independent formation but a translation from Welsh Pendragon (cf. Ackerman 1952: 75, 190), and it does not mean that Uther Pendragon has a dragon's head, but that he has a dragon's head on his banner. The types of exocentric nouns are a new development in Middle English; their scarcity in LajB is therefore not surprising. 73

2.13. Formations with affixoids Affixoids, i. e., preflxoids and suffixoids are elements which have a status in between full words and pure affixes. Each affixoid has its own history. As they cannot be discussed in detail here, a few examples will have to suffice.

2.13.1. -dorn. In Modern English, the noun doom jduimj and the suffix -dom jddmj, which has developed from the noun, are phonologically distinct. This difference cannot be proved for Early Middle English yet, but a difference in meaning existed even in Early Middle English. As a suffixoid, dom often meant 'status of (being (a)) ...', e. g , , f r e ( o ) - d o m , peow(e)-dom, frral-dom, wis-dom, sometimes 'territory of ...' (eorl-dom), sometimes it had other meanings (cf.

Lajamon's

compound nouns and their morphology

499

riche-domes 'collectivity of ...', swike-domes 'acts of treachery'), whereas the independent word dom had either lost these meanings in Middle English or had never possessed them. A morphological sign for the suffix-status of -dom is that combinations with it can be prefixed with un-: unwisdom (cf. Marchand 1969: 356, with reference to ModE -like). On the other hand, wisdom could still be analysed into its elements: peo quene ...of hire wisdome sprong pat word wide, pat heo wes swide wis of wordliche dome. (C 3139-42, similarly O).

2.13.2. -had: ModE -hood is a suffix because it does not exist as an independent word any longer. But it was still used independently in Early Middle English (though not in LajB) and therefore cannot be classified as a suffix for this period yet, especially since the meaning of independent word and suffix is basically the same, namely 'status o f . LajB has child-had.

2.13.3. kine-: kine- never occurs as an independent word; yet it should not be classified as a prefix, 74 but rather as a morphologically conditioned allomorph of king. It has the same meaning as king and is used as a determinant of compounds instead of king, whereas king never occurs as a determinant of compounds in La^B.15 This distribution goes back to Old English and is lost in later Middle English, when kine- gradually dies out; 76 it is strictly observed in La^B in both manuscripts. There are 10 compound nouns formed with kine-, some of which occur very frequently (kine-lond, over 50x in C; kineriche c. 31x in C; kine-dom, c. 24x in C; kine-louerd, c. 13x in C). 77

2.13.4. Other cases Other elements to be treated under the heading of pre- and suffixoids are -em, -ful (as noun-forming suffixoid), -lac, -reden, twi-, -ware. Furthermore it could be argued that words which serve more or less regularly as intensifying elements when used as determinants

500

Hans

Sauer

of compounds should be included among the prefixoids, too. 78 But these words also still occur more or less regularly in their original meaning as determinants. The main example from LajB is leod(cf. also 2.1. above and 5.1.1. below). It probably has its original meaning 'people, nation' in leod-quide 'national language', leodöeaw 'national custom' and possibly leod-scop 'national bard', whereas it seems intensifying in leod-cnihtes, leod-ferde, leod-kempen, leod-runen (cf. MED s. v. led{e n. (2); Carr 1939: 353). But because a decision is difficult in some cases and in order not to tear apart the examples, compounds with leod- as determinant have here been counted as belonging to the type noun/noun. Compounds with heued- as determinant have been treated similarly. In hcefd-bon and hafued-wunde, hcefd has its literal meaning, whereas in hcefd-men it is used in a transferred meaning 'main, principal'.

2.14. Inversion compounds Inversion compounds have the determinatum preceding the determinant. This order is un-English and due to foreign influence. In LajB this type occurs in a number of place-names of Celtic origin, e. g., Kaer Leir, Kaer Liun, Kaer Lud, Kcer Usch, Port Lud19 and in the Celtic nickname Pendragon. It is not productive with English elements in LajB and can therefore not be called an English type of composition. Its structure and at least to a certain degree the meaning of the elements were, however, understood because for some of these compounds their English equivalents are given by Lajamon, e. g., Kaer Leir is translated as Leirchestre (ModE Leicester), Kaer Lud as Ludesgate, Pendragon as Drakenhefd, and so on. Muntgiu 'Alps' (Latin Möns Iovis) is an inversion compound which ultimately goes back to Latin. 80 Its variant spellings in manuscript O, Montagu and Munt-ioye, show that it was obscured in Early Middle English and that several different words had fallen together here. 81

3. Morphological variants and peculiarities Apart from stating the morphological types of composition, a morphological description of compounds also has to take into account the morphological variants that occur within each type as well as

Lajamon's

compound nouns and their morphology

501

morphologically problematical cases. Under this heading we must consider compounds containing more than the minimum number of words or morphemes; elliptical compounds; compounds with the determinant in the plural or genitive; compounds with intrusive elements in the composition joint; compounds with constituents that do not occur as independent words; obscured compounds; compounds where the word-class of one element is ambiguous; interesting to note are also determinants and determinata which occur very frequently. Loan influence can be treated here, too. Some of these peculiarities have been pointed out in the preceding sections; they will be discussed somewhat more systematically now.

3.1. Compounds morphemes

with more than

the minimum

number

of

Compound types can be described as consisting either of two free morphemes, e. g., noun/noun, adj/noun, etc., or as consisting of two free and one bound morpheme; to this category belong the synthetic compounds (e.g., noun/vb + ere), the exocentrics, and noun + s/noun. Formations where either the determinant or the determinatum is a compound itself are rare in LajB and in Early Middle English generally; compounds consisting of more than three words do not occur at all in Early Middle English. 82 Within the type noun/noun LajB has Cantware-buri and Noröhumber-lond (adj + noun/noun); 83 within the type noun + s/noun there is midewinteres dcei (adj + noun + 5/noun), whose compound status, however, is not certain. Whit-suneday belongs to the type adj/noun, whereas Whitesune-tide and Witson-time are somewhat different. In Whitsuneday, whit is the determinant and suneday the determinatum, but there must have soon been a change in the morphological structure of this compound reflecting its lexicalization and obscuration, because in Whitesunetide and Witson-time the determinant is Whitesune and the determinatum is tide or time.84 Other compounds with three words are Dorchestre-seten, Est-Engelond, Suö-hamton and wapmon-kun, but in these the compound constituent was already obscured. The determinant is a prefix-formation in cerchebiscop-stol and a suffix-formation in alder-mon, aldere-men.S5 The determinant has a frozen case-ending (an original gen., dat. or acc. ending which has been taken over into the nominative) in: midderniht (cf. German

502

Hans Sauer

Mitternacht), haluen-del, mesten-dcel, pridden-dale and possibly also middenerd (all adj/noun or num/noun). 86 In manuscript C, the determinatum has the prefix i- ( < OE ge-) in eoröe-itilie, hird-iferen, plaje-iferen, witene-imot and wiöer-iwinnen.87 The determinatum is suffixed in uore-genglen (particle/noun).

3.2. Elliptical compounds The middle element has been dropped in Dorsete, which comes from Dorchestre-sete (cf. CODEPN, s. v. Dorset:; Blenner—Hassett 1950: 32), and possibly also in Brutware, which may have to be derived from a hypothetical *Brutlondes ware.68

3.3. Compounds with the determinant in the plural or genitive Compounds, especially noun/noun-compounds, with the determinant in the plural are fairly frequent in Modern English. They were rare in Old and Middle English; there are three examples in LajB: weien-lceten (gen. pi.), the loan word incubii demones and the place-name Sparewen-chestre (see C 14643-45). 89 Compounds where the determinant ends in an ^-genitive are here regarded as constituting a type of their own (see 2.2. above). Compounds where the determinant possibly or certainly goes back to another form of the genitive are included here among the noun/noun compounds. Clear examples of original genitive combinations are the days of the week: Mone-dcei, Fri-day, Sone-day (they all go back to West Germanic). Other examples are halle-dure, leode-celder, leoden-king, witene-imot, etc., and among the place-names Dene-marke, Oxeneford. In other formations, the -e might or might not represent a genitive, for instance in chirche-griö, hefne-riche, Rome-walles, sunne-liht, etc. (cf. 2.2.; footnotes 38-39).

3.4. Compounds with connecting elements in the composition-joint Whereas connecting elements in the composition-joint are fairly common in Modern German (e. g., Ankunftszeit), they are very rare in Modern English and they were rare in Middle English, apart from the -e at the end of the determinant, which, however, can

Landmen's

compound nouns and their morphology

503

often be interpreted in several ways and is therefore excluded here (cf. 2.2. above). One possible example for a connecting element from LajB is loft-song (manuscript C 36, etc.; manuscript Ο 36 has loue-sang), which goes back to OE lof-sang 'song of praise'. But the -t- may not be simply intrusive, it may represent an attempt to give the compound a new motivation and to re-form it as loft-song 'song of heaven' (cf. MED s. v. Wf n. (1); löf-söng; loft n. 2(c).). The reading of manuscript O, loue-sang, seems to indicate an interpretation as 'love-song'.

3.5. Compounds where one element does not occur as an independent word These fall into three categories: (1) The determinatum does not have to occur independently in the types of synthetic compounds which were analysed above (2.8.-2.11.). 90 (2) Bound elements which occur frequently as determinant or determinatum in place of the corresponding words have to be classified as allomorphs of these words. There were not many such pre- and suffixoids in Middle English; the examples from LajB are kine-, twi- and -ware (cf. 2.13. above). 91 (3) If bound elements occur in just one or two compounds and do not belong to the categories just mentioned they have to be classified as blocked morphemes. Compounds containing them are only partly motivated and are therefore partly obscured. Often the blocked morpheme is a word that occurred independently in Old English or earlier, but then died out in independent use. The determinant does not occur independently any longer in dweomer-crceft, dweomer-lak, jursten-dcei, i-lond, (but illond, yllond is motivated), 92 lei-uen, stiward; the determinatum no longer existed as an independent word in hale-wei 'healing liquid, balm', mer(e)-minnen 'mermaid' (OE mennen 'female servant'), fore-gengel 'predecessor' (-gengel apparently never existed as an independent word). Many place names have a blocked morpheme as determinant, e. g., Barbe-fleot, Bellasgate, Leir-chestre, or as determinatum, e. g., Stanhenge.93

3.6. Obscured compounds 94 Obscuration is not rare among the Early Middle English compounds. Among its main reasons are (1) the shortening of the

504

Hans

Sauer

determinant, e. g., before consonant clusters and in trisyllabic words, (2) the assimilation in the composition-joint, (3) the loss of accent on the determinatum, and (3), as stated above, the loss of a word in independent use. These factors can occur in combination. There are also degrees of obscuration: the result can be anything between slight and total obscuration. Accordingly, obscuration appears in several forms. — The determinant has been shortened in Enge-lond, wap-men, rake-tege and mcer-jeue. The final -n of the determinant has been dropped in eue-time, fourti-niht (manuscript O), more-liht, maidechild, marme-stan, seoue-niht, but since this happened more or less regularly also in the independent use of many of these determinants (ModE eve, morrow, maid, ME seoue) only a part of the compounds can be classed as obscured (marme-stan, fourti-niht). — The determinant has been shortened due to loss of the middle consonant in a cluster of three consonants in feon-dewces (< feonddewes), lon-gauel (< lond-gauel, both forms in C), sul-trome (manuscript O; < sceld-trome manuscript C), wor-man manuscript Ο (weorc-mon manuscript C). — The determinant has been shortened and changed in Cantelburi, Ο (< Cantware-buri, C); it has been changed for no recognizable reason in cantel-cape, originally canter-cape 'robe of the choir-master' (cf. MED s. v. canter-cöpe). In both cases, r has been changed to I. — The determinant has been changed due to regressive assimilation in lemmon (manuscript O), womman; assimilation has also taken place in Poillond, Ο (< Puille-lond, C). — Cases where the determinant or the determinatum does not exist independently any longer have been mentioned above (see 3.5.). — The determinatum has been weakened in likame; the determinatum has been changed in cerend-rake.95 — The compound has been entirely obscured, it has become a moneme, in: lauerd, leuedi.fulluht, orchard (manuscript C; horechard manuscript O), weofed. Obscuration does not occur all of a sudden; often the original and the obscured form coexist for some time. Thus manuscript C has wifmen and wimmen, manuscript Ο has wifman, wimmen and wommen. Manuscript C has Englene-lond, Engle-lond and /Engelond, manuscript Ο has Englene lond and Engelond', manuscript C has leue-mon, leoue-mon only, whereas manuscript Ο often has lemmon, but leof-man occasionally. The days of the week are particularly

Lajamon's

compound nouns and (heir morphology

505

interesting. La3amon devotes a passage to explaining their etymology, usually correctly. Nevertheless many of them were certainly obscured by his time, because in most cases the determinants of the compounds are spelled differently than in independent use:96 Manuscript Ο

Manuscript C Sunne Mone Tidea Woden punre Freo(n) Saturnus

Sonedcei Monedcei Tisdcei punre s dee i Fridcei Scetterdcei

Mone Tydea Woden ponre Frea Saturnus

Moneday Tisdei Wendesdai porisdai Friday Sateresdai.

3.7. Elements with ambiguous word-class Sometimes it is difficult to assign a particular compound to a certain type because the word class of one of its elements is ambiguous. Examples for the difficulty of deciding between verb and deverbal noun have been given with the type vb/noun (2.7.), examples for the problem of distinguishing between deverbal noun and (synchronically) primary noun have been given for the type noun/vb + 0 (2.8.). Throughout the history of the English language, adjectives have often been morphologically identical with the corresponding nouns (cf. for example Koziol 1972: 48, 281 ff.; Mustanoja 1960: 642-643). Whereas this distinction is usually clear in a sentence, it is sometimes difficult to make in a compound. Est-, norö-, suöand west- have here been included among noun/noun, but they could also be included among adj/noun. 97 The same applies for sibbe-men {sib-men) 'relatives'. Middel-niht on the other hand has here been included among adj/noun, but it could possibly be listed as a noun/noun-compound. 98 The adjectival character of middel seems, however, clear in middel-broper (cf. pe midleste broper, C 6442) and middel-erpe. Godd-spel was originally an adj/nouncompound 'good message', but was apparently re-interpreted as a noun/noun-compound 'God's message' in Middle English, as is indicated by manuscript C's spelling.99 Other cases where the determinant could be an adj. or a noun are frim-dcej.frum-ras, westepaöes, weste-side; they have been included among adj/noun here.

506

Hans Sauer

3.8. Frequent determinants and determinata Frequent determinants in the type noun/noun are leod- (13 x ) , sce(11 χ ) , burh-, Rome- (9 χ ) , here- (8 χ ) , hired-, wode- (6 χ ), weorld(5 χ ) , in the type particle/noun wider- (10 χ ) , among the prefixoids kine- (12 χ). The most frequent determinatum is -man, which is probably true throughout the history of English. 100 LajB has -man, -mon 33 times (including personal names): 19 χ in the type noun/noun, 7 χ in the type noun + s/noun and 7 χ in the type adj/noun. Other relatively frequent determinata are, for example, -day (23 χ ), cniht (9 χ ) , king (8 x ) and -ward (5 χ ) .

3.9. Loan influence 101 Loan influence among the compounds appears in different forms, in LajB especially in the form of loan compounds and in the form of hybrid compounds, i. e., compounds which are English formations, but contain a loan-word. 102 Loan-compounds can be further differentiated according to whether they were obscured or motivated in English; the latter is usually the case if their elements had been taken over as independent words, too. Compounds taken over from Old Norse were often assimilated, for example ON -maör as determinatum was regularly replaced by English -man. These cases ought probably still to be classed among the loancompounds and need not be classed among the hybrid compounds. With some compounds it is not certain whether they are loans or native formations, or whether they are loans or hybrid formations.

3.9.1. Loan compounds — A Latin loan is incubii demones, whose compound status is weak (because the determinant is in the plural). — A loan compound from French is dusze-pers (manuscript C), dosse-peres (manuscript O) which is marked as a foreign expression by La3amon (pe Freinsce heo cleopeden dusze pers), but became a common word in Middle English (cf. MED s. v. dousse-per). — Loan compounds from Old Norse are hus-bond (noun/noun; Late Old English); in-siht (particle/noun); ut-lajen (particle/noun;

Landman's

compound nouns and their morphology

507

Late Old English). 103 Gcersum, gersume 'treasure', husting 'council, assembly' (frequent in manuscript C; in most instances replaced by manuscript O), hcer-berge 'lodging', wandreöe 'misery', and windewen were probably obscured.

3.9.2. Hybrid compounds 1 0 4 — Both elements are Latin loan-words in cantel-cape 'robe of the choir-master' ( < canter-cape).10s — Possible or certain hybrid compounds with French loan-words as determinants are forrest-ende (manuscript O), illond/yllond (manuscript O) (see footnote 92), marbre-ston (manuscript O), priuemen. The determinatum is a French loan-word in jat-essel. — Hybrid compounds with Old Norse elements: The determinant is a loan-word in bleo-men; brunie-hode (manuscript C); 106 bulehude; scaöe-ded; scaöe-werk; Tis-dai ( < O N tysdagr, the Old English form was tiwes-dceg);107 jyoris-day (manuscript O; manuscript C's punres-dcei is the Old English form); prel-werkes. The determinatum is a loan-word in herd-swein, here-dring, mcein-clubbe, wil-tiöende, wider-heppen (wider-happes).

4. Conclusions The exact number of compounds in La3amon's Brut, as well as in other Early Middle English texts, is difficult to ascertain because in a number of cases compounds cannot be clearly distinguished from the corresponding syntactic groups. The attempt to give numbers for certain types of compounds is furthermore complicated by the fact that the word-class of a constituent is sometimes ambiguous (e. g., noun or adjective; noun or verb; primary noun or deverbal noun). It is also not easy to establish the exact number of compounding types in LajB or in other texts, because certain combinations can be treated either as a distinct type or as a subgroup of another type; for example numeral/noun could possibly be treated as part of adj/noun, but is regarded as a separate type here. Our survey has yielded 14 types of compound nouns in La3amon's Brut, the importance of which varies considerably. Even if the exact number of compounds in each type is not certain, the types can

508

Hans

Sauer

clearly be grouped into four categories with regard to their importance and productivity. (1) By far the strongest is the type noun/noun with 268 formations from the general vocabulary alone. (2) The next most frequently found types are noun/vb+ 0 with 37, adj/noun with 37, and noun + s/noun with 30 formations. The type particle/noun is represented by 19 formations and compounds with affixoids with over 30. The latter two types are, however, somewhat marginal in the system of nominal compounding and have not been collected exhaustively here. (3) The types vb/noun with 9 formations, num/noun with 7, and noun/vb-K with 6 are weaker. The inversion compounds have a marginal status, too: I have found eight formations on a selective count, but they occur in place names (and nicknames) only and are almost exclusively loan-words. They are not an English type of word-formation and are not productive in LajB. (4) The remaining types are represented with just one or two examples each: pron/ noun, noun/vb + ere, noun/vb + ing, noun + noun/0, i. e., bahuvrihinouns. (5) A few types of compound nouns are not represented at all in LajB, namely vb + ing/noun, e. g., ModE writing table, dancing girl, although this type existed in Old English and has lived on into Modern English. There are neither imperative compounds (of the structure vb +noun/0, e.g., ModE pickpocket) nor reduplicative nouns (e. g., ModE tick-tick, tick-tack, etc.),108 but because both types as well as the bahuvrihi-nouns are new in Middle English, the fact that they are missing in LajB is perhaps less surprising. La3amon is thus fairly conservative in his use of compounding types as far as nouns are concerned: he hardly uses types that are new in Middle English; he also hardly uses those synthetic and verbal-nexus types that go back to Old English and have become very productive in Modern English (vb + wg/noun; noun/vb + ere; noun/vb + ing), and the type pronoun/noun is also still very weak. None of these types is, however, particularly strong in other Early Middle English texts; although the Ancrene Riwle, for example, has a few formations of the type vb + wg/noun (clensing fur, swiing wike), vb +noun/0, i.e., imperative compounds {kealche-cuppe, weschedisch), and of reduplicative nouns (baban, giue-gauen) and is thus slightly more progressive (and also more colloqiual) than La3amon.109 Among the morphological variants, a number of compounds with complex constituents, of obscured compounds, and of loan compounds and hybrid compounds are worthy of note. It is often said that the productivity of compounds declined sharply in

Lajamon 's compound nouns and their morphology

509

Middle English; 110 but it certainly was not dead in Early Middle English, as is shown by L a j a m o n ' s many new formations.

5. List of compounds in La3amon's Brut The compounds are arranged according to the morphological types discussed in chapter 2. In order to save space, usually just one linereference is given for each manuscript and one spelling for each compound (the spelling is usually from manuscript C), unless the variants are important for the morphological description. 111 Worddivision has usually been regularized in this list: a dash has been inserted between the two elements (unless the compound is clearly obscured). The line numbers refer to the edition of Brook — Leslie. If a compound occurs more than once, this is indicated by 'etc'. Names of places and people are given after the general vocabulary in each section, but with no claim to completeness. Formations listed by Oakden which are not compounds are given at the end (5.3.).

5.1. Compound nouns 5.1.1. Noun/noun 1 alder(e)-mon C 712 etc 2 almes-mon C 9800 etc 3 archebiscop-stol C 6316 etc (archebissopes stol Ο 6316 etc) 4 bale-siö C 285 etc 5 bend-hus Ο 525 6 biscop-stol C 9091 etc (bissopes stol Ο 9091 etc) 7 blod(e)-stram C 13 328 etc 8 boc-fell C 26 9 boc-runen C 2244 10 boc-spell C 8728 etc; CO 9691 11 boc-stauen C 3807 12 breost-fronk C 969 13 bule-hude CO 7080 etc 14 bur-cnihtes C 8863 etc bour-cniht Ο 9461 etc burh-cnihtes C 8854 etc 15 bur-öein C 7663 etc burh-frein C 6844 etc

OE OE OE

16 17 18 19

OE ME OE

20

ME OE ME ME OE OE ME OE ME OE

21 22 23 24 25 26 27 28 29 30 31 32 33

bur-lutlen C 15 368 burh-cnauen CO 7762 burh-folc C 4866 etc burh-jate C 4713; CO 8818 burh-mon CO 6208 etc bur-men C 6798 burh-walles CO 11 024 burh-weren C 14159 etc burne-hod C 8242 etc brunie-hod 0 8242 etc candel-liht CO 11 853 cantel-cape C 14847 castel-buri C 3346 castel-jat CO 9308 etc (castles j a t C 9482 etc) chap-men C 5665; CO 6645 etc chireche-grid CO 11 138 chiric-lond C 7413 cnaue-child 0 7748 cniht-barn C 7748 cniht-weorede C 13 359

ME ME ME OE OE OE OE ME OE OE ME ME OE OE OE ME ME ME

510 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80

Hans Sauer ME craft-mon C 14444 dai-ende C 13 385 ME dcei-liht CO 2881 etc ME dad-siö C 3164 etc ME deap-si}} 0 3164 etc deor-friö CO 720 ME driht-folk C 57 etc OE driht-men C 7343 OE/ME dujeöe-cnihtes C 5070 ME dujeöe-king C 12954 ME dujeöe-men C 7019 ME ME dweomer-crceft C 15 291 ast-cende C 8903 etc OE east-ende Ο 10877 etc cest-half CO 11 902 OE ME ast-side C 10877 OE JEstere-dcei C 12048f. eiY-Zonc/ C 561; Ο 3659 etc ME OE i/ond 0 569 OE/ME Ο 3616 etc OE ende-dcei C 1858 etc eord-hus CO 1181 etc OE ceuen-time C 6417 etc OE eue-time Ο 6417 etc ME far(e)-cost CO 747 etc feon-öewces C 291 ME OE feper-beddes 0 8706 feöer-home CO 1436 etc OE fohjel-cun CO 4044 OE folk(e)-king C 4542 etc OE ME forrest-ende Ο 10410 OE uoster-moder CO 12926 fur-b(u)rondes ME CO 12780 Fri-dai CO 6952 OE ME 3cet-essel CO 9477 jeol-dai C 11 346 OE jim-ston C 10 550 OE glees-fat C 8845 etc OE OE gleo-crceft C 3495 OE gle(o)-drem CO 913 gleo-men C 3491; OE Ο 2548 etc ME Godd-cunde CO 12456 ME gold-ring CO 2252 etc ME golde-stauen C 10 555 ME gold-wir CO 3513 ME gras-bed C 11 722 etc ME guö-strenede C 799 ME hale-wei CO 11 513 etc ME halle-dure CO 10478 etc ME hal-imot C 15 973 ME 0halle wah C 12920) hile-woh Ο 12920

81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125

hcefd-bon CO 735 etc hcefd-men C 8058 hcefued-wunde CO 3796 herd-swein C 2824 here-burne CO 11 959 here-cniht C 15 367 herre-feng C 5845 hcere-gumen C 7254 etc hcere-kemppen C 10 287 etc here-marke C 9089 etc hcere-scrud C 2528 here-öringes C 2580 etc here-dring CO 4290 etc heorte-blod CO 7907 heouen-king C 16 059 hefne-riche C 14788 etc hired-childeren C 8261 hird-cnaue C 2825 etc hired-cnihtes C 1779 etc; 0 7488 hird-folc C 3222 hird-iferen C 3306 hired-gume C 6123; Ο 9563 hired-men C 1126 etc; CO 1177 etc hoker-loö C 14408 hoker(e)-word CO 9777 etc hond-scex C 3227 hors-leden C 11 483 hors-men CO 13 285 hus-bonde CO 15 953 husting CO 430 etc hux-wordC 10 819 incubii demones CO 7876 lache-craft C 3797 etc ley-uen CO 11 395 lar-spel CO 5068 etc leer-stow C 8466 etc leode-celder C 684 leod-cnihtes C 3718 leod(e)-ferde C 421 etc leod(e)-folk C 1025 etc; 0 1019 (leodes folk C 4728) leod-kempen C 3004 etc leod-king CO 436 etc (leoden king C 249) leod-quide C 1456 leod-runen C 7264 etc leod-scome C 13 124 leod-spelle CO 7863 etc

OE OE OE ME OE ME ME ME OE ME ME ME ME OE OE ME ME OE ME ME ME OE ME ME OE ME ME OE OE OE ME OE ME OE OE ME ME ME ME ME OE ME ME ME ME

Lajamon's compound nouns and their morphology OE 126 leod-öeaw C 1031 ME 127 leod-peines C 3327 ME 128 leod(e)-wise CO 3544 etc OE 129 likame C 2509; CO 3814 etc OE 130 lich-rast C 8597 OE 131 lif-daj C 5413; Ο 485 etc ME 132 lif-griö C 4418 (lifues grid CO 4400 etc) ME 133 lif-time Ο 503 134 lod-cniht CO 12 841112 ME OE/ME 135 loft-song C 36 etc loue-song 0 36 ME 136 lond-cnihtes C 9963 etc; CO 12 829 OE 137 lond-folc C 11 114 etc; Ο 15439 OE 138 lond-gauel C 3884 lon-gauel C 3721 OE 139 lond-riche CO 12 774 etc ME 140 lond-sorje C 11 696 141 luf-ping C 86 ME 142 meiden-child C 1613 OE maide-child C 7175 etc 143 mcein-clubbe C 7630 ME 144 marme-stcen C 571 etc OE 145 marbre-ston 0 571 etc ME OE 146 masse-preostes C 14908 147 merfeJ-minnen CO 663 OE 148 mete-bur des CO 1817 ME ME 149 mate-cun C 473 150 mod-kare CO 1556 etc OE 151 mod-sorje C 4335 etc OE (modes sorje C 12 603) 152 mon-kun CO 220 etc OE 153 man-kunde 0 7424 ME 154 mon-drem C 11 949 OE (mannes drem O i l 949) 155 mon-ueorde C 2835 etc; ME 0 5360 etc 156 mon-qualm CO 1950 etc OE 157 Mone-dcei C 6954 OE 158 mor-uenne CO 10062 ME 159 morjen-liht C 8955 OE more-liht 0 8955 160 morö-gomen C 11431 ME 161 morö-spell C 9807 ME morpre spell 0 9807 162 munec-child C 6614 OE 163 munec-claÖes C 6480 etc ME (monekes elopes Ο 6542)

164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187

188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207

(monekene elopes 0 8814) munecclif C 14831 etc nail-sax CO 15 263 niö-craft CO 3547 norö-ende CO 1389 etc norö-ard C 5816 etc norö-jeet C 3799 norö-half CO 7953 norö-side CO 12234 nute-scalen C 14605 orchard C 6465 horechard 0 6465 pic-forc CO 10 777 prat-wrench C 2644 etc quale-hus CO 365 etc quale-siö C 15 924 rake-teje CO 8360 etc run-stauen C 4967 Satter-dai C 6953 (Sateres-dai 0 6952-3) scaöe-ded C 14761 scaöe-werc C 774 sceld-trome C 4714 etc sultrome Ο 8216 etc scid-wal CO 5164 scip-ferde C 1078 etc; 0 3607 etc scip-gumen C 2276 scip-men C 1115 etc; Ο 669 etc {scipen-monnen C 6884) sce-brim CO 3183 etc sa-cliua CO 933 etc sa-fisc CO 11 252 sa-flod CO 518 etc sa-grund CO 2012 etc sa-men CO 584 etc sce-oure CO 4281 sa-rime CO 3098 sa-side CO 4068 etc sa-stremes CO 165 etc sce-strond C 1076 etc; Ο 744 etc seil-claö CO 2270 sail-rap C 8682 sibbe-freond C 1222 sibe-laje C 208 sibb(e)-men CO 682 size-craft C 7735 Sone-dai C 6953 sorh-siöes C 5541 spere-scceft CO 7362

511

OE OE ME OE ME ME OE ME ME OE ME ME ME ME OE ME OE ME ME OE OE OE ME OE

ME OE OE OE OE OE ME OE ME OE OE ME ME ME ME ME ME OE ME OE

512

Hans Sauer

208 stcen-cun C 1422 209 210 211 212 213 214 215

(stones-cun Ο 1422) stan-gr äffen C 15 915 ston-wall C 96; Ο 866 etc stel-boje C 11926 sunne-liht C 8913 suö-ende C O 1684 etc suö-3cet C O 13 941 suö-halue C 10 878;

OE ME? OE ME ME OE ME OE

0 12232 etc 216 217 218 219 220 221 222 223 224 225 226 227

suö-lond C O 1057 etc

swerd-broper C 2068 etc tceuel-brede C 4056 treo-wrekes C 11 427 treo-workes Ο 11 427 peod-folk C 13 223 prel-werkes C O 229 prum-ferde C 680

wa-dai C 4364 wei-siö C 12899 etc wcel-kempe C 284 etc

wal-spere C 14260

wan-sides C 1542 etc wen-sides C 57 etc wap-men C O 172 etc wapmon-kun C 251

228 229 230 wari-treo C 2850 weri-treo Ο 2850 231 wed-broper C O 7221 232 weien-laten CO 7739 233 wei-wittie Ο 6418 234 welle-water C 2255 etc

OE ME ME OE ME ME ME ME ME ME OE ME OE ME OE OE OE ME OE

254 255 256 257 258 259 260 261

worle-ping 0 14040 west-ande CO 9378 etc west-halue C O 11 901 etc west-lond C O 8119 wi-cex C 785 etc

wif-men C 186 etc wimmen Ο 186 etc

wif-man Ο 4786 etc womman Ο 1119 etc

wode-burj C 1084 wode-lond C O 851 wode-rim C 372 wode-rot C O 235 wude-scaje C 10 759 etc wode-saye 0 13 659 wored-strencöe C 256

Place-names and names o f people (selective)

3 Brut-lond C O 1097 etc

Ο 10193 OE OE OE OE OE OE OE ME ME

CO 12692 246 wimmon C 78 etc

etc wun-sele C 7836 etc113 wine-mceies C 2908 etc witene-imot C 5759

OE

(weorldes scome C 9279)

241 242 243 244 245

wifnjdewen C 15 385 win-tunnen C O 15 312

1 Barbe-fleot C O 12 707 etc 2 Brut-leode(n) C 4180 etc;

worle-kin[g] 0 3661

237 weorld-lif C O 16012 238 weorld-mon C 14040 239 woruld-riche C 3493 etc worle-riche Ο 3493 etc 240 weorld-scome C 4151

wil-gomen C 10450 will-spall C 8803 etc wil-tidende C 8529

262 263 worö-munt C 9407 264 writ-runen C 2868 265 w( ojnder-craftes CO 576 etc 266 wonder-ping Ο 8548; CO 8584 267 wunder-weorc 0 8673 268 wn-folk C 694

(·wor-mafn] 0 5160)

236 weoreld-king C 3154 etc

wif-öing C 2218 etc wil-dajes C 901

OE

{welles water 0 2255 etc)

235 weorc-mon C 5160 etc

247 248 249 250 251 252 253

OE

4 Brut-pringen C 6429 5 Camel-ford CO 14239 Cantware-buri C 1409 etc Cantel-buri Ο 1409 etc 7 Dene-marke C O 2281 etc 6

8 Dertene-mud C 8046 etc Derte-moup Ο 8046 9 Deuene-scire C 14928 etc 10 s£st-/Engle C 6114 etc 11 Est-Engelond 0 6114 12 Alst-lond C O 15 440 etc

13 ALst-scex CO 7662 etc; Ο 7678 14 Englene-lond C O 9 etc Engle-lond C 3148 etc /Engelond C 15 578 Engelond Ο 7 etc 15 Eover-wic C O 1335 etc

OE ME ME OE ME ME OE OE OE OE OE OE OE ME ME ME OE ME OE ME OE ME

Lajamon's compound nouns and their morphology 16 Flaundre-lond C 3610 Flandres lond 0 3610 17 Gloi-chestre C 4801 Glou-cestre 0 4801 18 Gric-lond CO 166 etc 19 Gut-lond CO 11 259 etc 20 Had-feld C 15 581 Apel-feld 0 15 581 21 Hamton CO 4677 etc 22 Houen-feld C 15666 23 Humber-lond CO 2225 24 lr-lond CO 3118 etc 25 Is-lond CO 11212 26 Lajamon C 1 etc Laweman 0 1 etc 27 Lane-castel CO 7108 28 Leir-chestre C 1456 etc; CO 1859 29 Lundene-burh C 2138 30 Maidene Castel CO 1340 31 Nord-humbre CO 9223 etc 32 Noröhumber-lond CO 3185 etc 33 Nord- Wales C 12142; Ο 3712 etc 34 Oxene-uord CO 13097 etc 35 Port-chcestre CO 4606 36 Puille-lond C 6086 Poillond Ο 6086 37 Rome-burh CO 2667 etc 38 Rom-cnihtes C 4624; Ο 12 399 39 Rom-folk C 3953 40 Rom-leode(n) CO 2818 etc 41 Rom-lond C 4840 Rome-lond CO 5596 etc 42 Rom-monen C 2950 43 Rom(e)-peode CO 4512 etc 44 Rom(e)-walles CO 2952 etc 45 Rom-wisen C 5576 (wise of Rome C 5576) 46 Scot-dale C 2412 47 Scot-ferde CO 2050 48 Scot-leode CO 10003 49 Scot-lond CO 1065 etc 50 Scot-peode C 10187 51 Sex-leode C 7552 etc 52 Sax(e)-londC 7002; 0 7149 etc 53 Sax-peode C 7234 54 Sparewen-chestre C 14645 55 Stan-henge CO 8732 etc 56 Sud-hamton CO 2413 etc 57 Sud-saxe C 7668 58 Suö-Wales CO 14935 etc

513

59 bwong-chastre CO 7104 etc 60 Wal-bruttes C 5361 Walse-Bruttes Ο 5361 61 Wale-broc CO 5407 62 West-mering C 4972 ff. Westenering 0 4972 ff. 63 West-muster C 15 330 64 West-sax CO 7678-80 65 West-Wales C 5361 66 Wine(n) t-lond CO 11 285 etc

5.1.2. Noun + s/noun 1 archebissopes stol OE Ο 6316 etc 2 bissopes stol 0 9091 etc OE 3 castles j a t C 9482 etc ME 4 cunnes-mon CO 1363 etc ME 5 dajes liht CO 7423 etc ME 6 daöes kare CO 7755 ME 7 Domes-dai CO 7873 etc OE 8 erendes-mon C 12407 ME ME 9 heruestes dai C 12 678 ME 10 kinges (i)streon CO 11 276 11 lifes grid CO 4440 ME 12 lades-men C 3113 OE lodes-men 0 3113 OE? 13 modes sorje C 12603 14 mannes drem 0 11 949 OE 15 midewinteres dai OE CO 11 430 16 monekes elopes C 8814; ME Ο 6542 17 Sateres-dai 0 6952-3 OE 18 sondes-mon CO 6785 etc; ME 0 5239 (sond-mon C 6361) 19 steores-man CO 677 etc OE (,ster-men C 3911; CO 14193) OE/ME 20 stones cun Ο 1422 21 Tis-dai CO 6954 OE 22 twines-prad C 7096 ME (,twined pred Ο 7096) 23 peines-men C 7460 ME 24 punres-dai C 6951 OE poris-dai Ο 6950-1 ME? 25 wardes-men C 9633 ME 26 welles-stram C 9858 etc OE/ME 27 welles-water 0 9858 OE/ME

514

Hans Sauer

28 Wendes-dei Ο 6950-1 29 weorldes scome C 9279 30 whales-bone CO 1182

OE OE/ME ME

Place-names 1 Angnetes munt C 1339 2 Agnetes hull Ο 1339 3 Amberes-buri C 8177; Ο 8248 etc 4 Beljces-jate C 3021 etc Belynes-jate Ο 3021 Bellinges-jate Ο 7519 5 Cuninges-burh C 5814 etc Conanes-borh Ο 5814 etc 6 Cristes-chirche C 1254 7 Ludes-gate C 3559; Ο 3560 8 Mihceles Munt CO 9937 etc 9 Teynis-wich Ο 10 783 (Teinne-wic C 10783)

Syntactic groups are probably: 10 Lundenes tun C 9231 etc; Ο 9317 11 Tintageoles tun CO 9580 12 Verolames tun CO 9768 13 Winchcestres tun CΟ 9806 (ιWinchastre tun C 9795)

5.1.3. Adj/noun 1 ced-mede C 3316; CO 4993 2 aide-fader C 1694 etc holde-fader Ο 1694 3 cer-dced CO 4361 4 bleo-men CO 12 666 5 feie-siö C 154 etc; CO 1863 etc 6 freo-king C 1595 etc 7 freo-men C 3833; CO 9675 8 frim-daj C 15947 9 frum-rces C 4322 10 furn-daj CO 13 536 etc 11 god-dede CO 10 514 12 godd-spel C 14726 etc 13 jursten-dai CO 8516 etc 14 hali-rift C 14211 15 hea-der C 159

16 hcehje-dai C 5341 etc 17 hah-sattele C 8307; CO 9245 18 laö-spcel C 10382 19 leue-mon C 97 etc leof-man 0 2323 etc lemmon 0 2239 etc 20 lift-hond C 8392 etc 21 luper-craftes 0 12435 etc 22 mesten-dcel C 3921 23 mid-festen C 11106 24 mid-niht C 843 etc; 0 7956 etc 25 middel-broper Ο 6442 26 middel-(e)arde C 3592 etc middil-erpe Ο 3658 etc 27 middel-niht C 10 282 28 midden-erde C 12365 29 midder-niht C 7956 etc 30 priue-men Ο 3428 31 riht-hond CO 8327 etc 32 soö-quides C 4749 etc 33 weste-paöes C 8650 34 weste-side C 10 880 35 White-sune-dai C 8725 etc (a Whiten-sunen-dceie C 8727 etc) Witesone-dais Ο 8725 etc Witsonday 0 12135 36 Whitesune-tide C 15 736 37 Witson-time 0 15 736

ME OE OE ME

ME ME ME OE OE ME ME

OE OE OE ME OE OE ME ME ME

ME ME

OE OE OE ME ME ME OE ME? ME OE OE OE OE OE OE

The following are probably adverbial combinations: 1 ludere stauene C 8204; CO 10 391 etc 2 soöere worden C 11 240 etc sopere wordes Ο 11 240 3 to wradere hele CO 14750

ME ME?

OE

Place names and personal names 1 Dense-men C 6649 2 Englisce-men CO 12 164; Ο 988

Lajamon's compound nouns and their morphology 15 16 17 18 19

3 Frensce-men CO 3819 4 Middel-saex CO 7680 5 Wit-sond CO 4855 etc

5.1.4. Numeral/noun dusze-pers CO 813 ME (feowertene niht C 12 814 OE etc) fourti-niht Ο 12 814 etc ME halue-suster C 4195 ME haluen-del 0 1668; ME CO 3536 seouen dajes C 4869 ME soue dawes Ο 4869 seoue(n)-nihte CO 818 OE/ME etc seoue-niht CO 703 etc />ridden-dale CO 7822 OE

wiöer-lajen C 5471 wifjer-reases Ο 12 524 wiöer-saka CO 906 wider-side CO 5973 wiöer-winna Ο 391 wiöer-iwinnen C 2097 etc wiper-iwinnes Ο 2851 etc

515 ME ME OE ME OE

5.1.7. Verb/noun ber-cnihtes C 9952 ber-men CO 1657 etc be or-time 0 6115 hare-word CO 5946 etc plaje-iueren C 7800 plei-ueres Ο 7800 spac-hus C 6506 etc spec-hus Ο 6545 etc til-ahö C 8413 walle-stream C 1423 welde-(king) Ο 1597

ME OE ME OE ME OE ME OE ME

5.1.5. Pron/noun 1 /.El-drihten C 14077

ME

5.1.8. Noun/vb + 0

5.1.6. Particle/noun (selective) 1 uore-genglen C 12 517 2 forp-fare 0 1259; CO 2997 etc 3 for(e)-ward C 549; 0 548 etc 4 in-jeong C 14160 in-gonde{\) 0 14160 5 in-siht CO 15222 6 sunder-rune C 15 681 7 to-nome C 6511 8 under-kinge C 15 644 9 vt-lajen CO 563 etc 10 wiöer-craftes C 138 etc; CO 9794 11 wiöer-deden C 8143; CO 10521 12 wiöer-ueht C 14306 13 wiöer-gome CO 12 326 14 wider-heppes CO 204

OE ME OE

5 6 7 8

OE ME ME OE OE OE ME ME ME ME ME

9 10 11 12 13 14 15 16 17 18 19 20

bac-ward(e) CO 11 884 balu-fehte C 2964 balu-ras C 5154 etc; Ο 12944 blod-gute CO 317 etc (WOJ mochel blod ijote Ο 9552 etc) bur-ward CO 9569 dure-ward CO 8819 eah-sene CO 1544 etc eorde-itilie C 5000 etc erpe-tilie Ο 5000 etc arend-rake CO 332 etc feond-ras CO 11 956 feond-scaöe CO 12996 jate-ward CO 9480 gram-res Ο 2593 grist-bat C 2586 her-beria C Ml5 har-berje CO 3773 etc here-jeong C 9079 her(e)-toje C 2932 etc hired-plaie C 7227 horse-bere CO 9695 lauerdC 136, Ο 333 etc

ME ME ME

OE ME

OE ME

OE OE OE OE OE ME ME

OE OE OE ME

OE OE

516

Hans Sauer

21 lauerd-swike CO 11 047 etc 22 leod-scopes C 11 465 etc 23 leod-swike C 6459 etc; 0 378 etc (leodene swike C 378) 24 leuedi CO 597 etc 25 lond-tilien C 7409 26 man-lure 0 321 (his monne lure C 321) 27 mon-sware C 2070; CO 11048 28 mcer-3eue C 7183 etc 29 rad-jiuen C 5794 etc; 0 7857 etc 30 sce-flot C 2261 31 stal-fiht CO 315 32 sti-ward C 111, CO 739 etc 33 wceld-scceöe C 3213 etc 34 water-scenc CO 9827 35 win-scench C 1762; 0 4833 etc 36 wind-rces C 4611 37 wode-scape 0 12906

OE ME ME

OE ME ME OE OE OE OE ME OE ME ME ME OE ME

Personal names 1 Dorchestre-seten C 14 780 Dor(e)-sete CO 10 717

5.1.9. Noun/vb + t feond-slcehtes C 8212 fulluht CO 18 etc heh-seht 0 4104 mon-slceht C 13 889; CO 13 891 5 wal-slaht C 686 6 wan-slcehtes C 4747

1 2 3 4

ME OE ME OE OE ME

5.1.11. Noun/vb + ing 1 grist-batinge C 945 2 heortne-graning C 8880

OE ME

5.1.12. Noun + noun/0 (Bahuvrihi-nouns) Personal names 1 Draken-hefd C 9097 Drake-heued Ο 9097

ME

5.1.13. Combinations with affixoids -dom 1 Cristin-dom CO 5040 etc 2 eorl-domes CO 3573 etc 3 fre(o)-dom(es) CO 10427 etc 4 hali-dom(es) CO 7655 etc 5 martir-dom CO 5047 6 riche-dom(es) Ο 1186; CO 1662 etc 7 swike-dom(es) CO 2058 etc 8 peowe-dom CO 183 etc 9 pral-dom C 14 550 etc 10 wis-dom CO 223 etc (un-wisdom CO 1690 etc) 11 wrake-dom C 40

OE OE OE OE OE OE ME OE ME OE (OE) ME

-em 1 quarc-ern CO 365 etc quart-ern C 1882 etc

ME

-ful (noun) 1 scip-ful CO 11 824

5.1.10. Noun/vb + ere 1 wcei-witere C 6418 (wei-wittie Ο 6418)

ME

ME

-had 1 child-had CO 10134

OE

Lajamon's compound nouns and their morphology kine1 kine-cerde C 9696 kine-erf>e Ο 9696 2 kine-bearn C 101 etc; Ο 626 3 kine-bench C 4833 4 kine-dom CO 1400 etc 5 kine-helm C 3372; CO 4033 etc 6 kine-lond C 93 etc; Ο 110 etc 7 kine-Iouerd C 1251 etc; Ο 3304 etc 8 kine-riche Ο 299; CO 1447 9 kine-stol CO 2254 10 kine-peode C 1472 etc; CO 11 128

M E

2 Cant-uare C 14853 Kent-war re C 4158 3 Rom-ware C 3957 etc

OE ME

5.1.14. Inversion-compounds

OE Place- and personal names ME

OE 0 E

Qg

Personal names 1 Kine-bord CO 15 325 2 Kine-marck Ο 1956 (king Marke C 1956)

1 2 3 4 5

Kaer Leir C 1455 Kaer Liun C 3012 Kaer Lud C 3540 Kaer Usch C 2995 Munt-giu C 2814 etc Mun-giu C 2621 etc Montagu Ο 2621 Munt-ioye Ο 11 480 6 Munt Senis CO 2622 7 Pen-dragon CO 9096 8 Port Lud C 3559 Ροφ-lud Ο 3559

5.1.15. Adverb/noun

-lac, -lock 1 dweomer-lac C 137 etc 2 fair-lac C 5806 fair-lock Ο 5806 3 raf-lak C 2014 etc 4 wed-lac CO 199

517

ME ME

1 wel-ded C 1578; Ο 1491 etc

OE

OE 0E

5.2. Compound adjectives 5.2.1. Noun/adj

-reden 1 fceredene C 2029; CO 3002 2 hired C 1169 etc 3 mon-redne C 212 etc; Ο 2394 etc

OE OE 0 E

twi1 twi-read Ο 9688 (itwi-rced adj. C 9688) -ware

Personal names 1 Brut-ware C 6333

0 E

1 feond-strong C 11 939 2 Jeue-custi C 2427 3 gold-faje C 4880 etc; Ο 13 329 4 grund-hat C 2839 5 mete-custi CO 177 etc 6 milc-whit CO 7954 7 mils-liöe C 7409 8 sce-werie CO 2306 9 snau-wh[i]te C 12 237 10 stale-worjje Ο 1903 etc stille-worpe Ο 6405 (not in C) 11 weorld-seliC 5508 12 weorlde-wis C 6847 etc worle-wise Ο 6847 etc 13 wod-sturne C 11 236 14 wunder-creftie CO 579

ME ME OE ME ME OE ME OE OE OE

ME OE ME ME

518

Hans Sauer

5.2.8. Pron/part pres

Names 1 Gos-whit C I O 552 (used as noun) 2 Woest-Walsce C 7719; CO 8799

1 al-waldinde C 9754

OE

5.2.9. Noun/part pret 5.2.2. Adj/adj 1 dere-worpe dure-wuröe 2 jere-witele war-wit(t)y

Ο 1017 etc C 1778 etc C 1512 etc Ο 7256 etc

OE OE

5.2.3. Num/adj 1 hcelf-jaru CO 4318

ME

5.2.4. Pron/adj 1 al-miten C 30 etc al-mihti C 8377 etc; Ο 10923

OE

5.2.5. Particle/adj 1 ouer-soh C 1732 2 purh-costned CO 12696

ME ME

1 2 3 4 5

boc-ilcered C 7298 etc cheues-boren CO 2163 gold-ileired C 4630 grund-fulled C 546 grund-lade(n) CO 555 etc 6 win-drunken C 4052 (dronge of wine Ο 4052)

1 bezst iboren C 8163 2 bezst itojen C 12299 3 hah-iboren C 4980 etc hep-bore Ο 170 etc 4 loh-iboren C 10999 5 neow-cumen C 4270 6 wel-ibore CO 341 7 wil-cume CO 2447 etc wel-come Ο 6679 etc 8 wel-idon CO 9363 etc

ME? ME ME? ME ME? OE OE ME?

5.2.11. Combinations with affixoids -fast

OE

5.2.7. Adj/part pres Adv/part pres 1 wel-donde Ο 2929

OE

5.2.10. Adj/part pret Adv/part pret (possibly most examples are syntactic groups)

5.2.6. Noun/part pres 1 sce-liöende C 3900 etc

ME OE ME ME ME

1 2 3 4

blceö-fest C 3482 etc soö-fest CO 32 etc stadel-uast C 4896 swome-fest CO 1554

OE OE OE OE

-fold

OE

1 feole-feld C 2120 etc; Ο 13 541 2 many-fold Ο 13 379

OE OE

Lajamon's compound nouns and their morphology -ful 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

19 20 21 22 23 24

25 26

aijes-ful C 8968 balu-fulle C 7522 etc blis-ful CO 3015 etc brade-fulle C 16087 car(e)-ful CO 141 etc deol-ful 0 3440 etc der-ful C 9886 god-ful C 1700 etc gomen-fulC 10693 grid-ful CO 4370 guö-fui C 1476 etc hoh-ful C 7034 laf-ful C 5414; CO 1515 lid-ful C 633 mod-ful C 12046 etc neod-ful Ο 11 848 niö-ful C 5096 redes-ful C 66 rad-ful C 1453 etc; Ο 2639 sorh-ful CO 165 etc swic-ful CO 4000 etc teon-ful C 2289 tir-ful C 1445 etc peu-ful C 900 wize-ful C 1439 wiht-ful C 602 wijele-ful C 10 548 wit-ful C 4536; 0 900 etc wiöer-ful C 5273; CO 10738 etc

-wende OE OE ME ME OE ME ME ME? ME ME ME? OE OE ME? OE OE OE ME?

OE OE OE ME ME ME?

ME? ME

kine1 kine-wuröe C 2982 etc; 0 6680 etc 2 kine-boren C 170; CO 5036 etc

OE OE

-les 1 2 3 4

aie-les C 9685 barn-les C 4484 care-les CO 6226 etc fader-les C 10927; 0 11 788

OE OE OE OE

seld1 sel-cuö C 142; CO 245 etc

519

OE

1 hal-wende C 1424 -wis

OE

1 riht-wis CO 1046 etc

OE

5.2.12. Bahuvrihi-adjectives (noun + noun/0; adj + noun/0, etc) 1 ad-mod C 11 188 etc; Ο 11605 2 aöel-mod C 11605 3 bar-fot C 4410 4 bliöe-mod C 14823 5 mild-heorte C 8390 6 run-hende C 3259 7 sori-mod 0 12 640 sari-mod C 14776 etc 8 twi-rad C 9688 9 pole-mod CO 1569 10 wa-med C 3174 (eds.: to-wamed probably wrongly)

OE ME OE OE OE ME OE OE ME OE?

5.2.13. Extended bahuvrihiadjectives (adj + noun/ ed, etc) 1 an-lapi C 6668 alpi(e) CO 1747 etc 2 feower-noked CO 10 978 3 heor-lockede C 12 899 hor ilocket Ο 12899

OE ME ME

The following are probably syntactic groups (but are listed by Oakden): 4 hard iheorted C 5982 5 stiö biwaledC 15 322 (•wel bi-walled Ο 15 322) 6 stiö imained CO 12 886 7 stiö imoded C 9890 etc 8 wel iöawed CO 3258 9 wel iwepned CO 13 577

ME ME ME OE ME? ME?

520

Hans Sauer

5.3. Items from Oakden's list which are not compounds 1 2 3 4 5 6 7 8 9 10 11

aie-leste (suffixation) cer-wene (prefixation) cer-witte (prefixation) drinc-hail (phrase) el-reordi (prefixation) el-peod (prefixation) god-cunde, adj (suffixation) heje men (syntactic group) hinder craft (syntactic group) ifurnjer (syntactic group) larspelundre (wrong reading; should be: larspel undernom) 12 liöendemon (syntactic group) 13 meterum (wrong reading; should be: a mete rumhende)

14 15 16 17 18 19 20 21 22 23 24 25

of-longed (prefixation) on-freondes (prefixation) som-rune prefixation) step-moder (prefixation) step-sune (prefixation) wikenares (suffixation) wunderbli^e (synt.gr.: adv + adj) wunderfeole {synt.gr.: adv + adj) wundergod (synt.gr.: adv + adj) wunderkene (synt.gr.: adv + adj) wundermere (synt.gr.: adv + adj) wundermuchel (synt.gr.: adv + adj) 26 wunderriche (synt.gr.: adv + adj) 27 wunderstrong (synt.gr.: adv + adj) [On the cpd-adverb limmele see 1.2. above; on the adverbial group wraöerehele see 2.3. above]

Notes 1. Bibliographical information can be found in NCBEL, I, 460-463 and in R e n w i c k - O r t o n - W a k e l i n (1966: 348-351). I follow Brook-Leslie (19631978) for quotations and line-numbering. I am indebted to Prof. H. Gneuss and Dr. K. Toth for reading a draft of this paper and making valuable comments; my thanks are also due to Prof. L. Lipka for his help and support, and to Mrs. S. Bollinger for casting a critical eye over my English. 2. It is usually assumed that manuscript C was written about half a century earlier than manuscript O, but the dates given for the two manuscripts vary considerably. If Ker's dating of manuscript C is accepted, both manuscripts are of approximately the same age, although manuscript Ο is linguistically more progressive than C. See for example Wyld (1930: 1): manuscript C 'written about 1200', manuscript Ο 'written about fifty years later'; these dates are repeated by Strang (1970: 258, 265), but the date for C is certainly too early. MED, Plan and Bibliography: 51, dates manuscript C 'a. 1225'; manuscript Ο 'c. 1300'. R e n w i c k - O r t o n - W a k e l i n (1966: 348) date manuscript C 'c. 1250', manuscript Ο 'third quarter of the thirteenth century'; Ker (1963: ix, xvi, xvii) dates manuscript C in 'the second half of the thirteenth century' [authoritative]. 3. Oakden (1935: II, 130-165). The place-names are dealt with by Blenner-Hassett (1950). A number of La3amon's compounds are also mentioned by Wyld (1930-37), but Wyld is interested in the use and meaning of La3amon's words, not in morphological questions or composition patterns. 4. MED, s. v. Brit, has the same omission: it gives Britlond, but not the other compounds. 5. There are other doubtful points about Oakden's method. He apparently takes distribution as his criterion: according to him, poetic compounds do not occur 'in ordinary prose usage' (Oakden (1935: II, 115)). But it is not quite clear what he means by 'ordinary'. Moreover, Oakden contradicts his own principle when he remarks on gold-wire (II, 142) Ά rare Μ. E. word confined to poetry but in no sense a poetic word'. Here obviously another concept of poetic

Lajamon's compound nouns and their morphology

6. 7.

8.

9. 10.

11.

12.

13. 14. 15.

16.

17.

18.

521

compound comes into play, namely some poetic quality of the compound in question. Such a poetic quality is certainly not always easy to pinpoint, but obviously it cannot be ignored. Why, for example, should burh-^ate be a poetic compound, whereas castel-jeat is a prose compound, and castel-buri 'castleboundary' is poetic again? Oakden's classification of compounds as Old English or Middle English is not always correct, either. It must, however, be admitted that it is at times difficult to find out whether a compound was formed in the Old English or in the Middle English period. This is probably true throughout the history of English. For comparison, the compound-adjectives from LajB are listed in chapter 5.2. Levi (1978) tries to circumvent the question by abandoning the concept of nominal compound in favour of the term 'complex nominal'. This does not, however, solve our problem, because it covers constructions that are certainly not compounds, especially combinations with nonpredicating adjectives such as criminal lawyer, linguistic scholar, polar bear. Locative particles are independent words but show some affinity to prefixes when used in compounds; the second member of a synthetic compound does not always occur as an independent word (but this can be taken care of by the notion of 'potential word'; cf. Hansen et al. (1982: 49-50); Kastovsky (1982: 179, 210)); the exocentric formations contain two words, but their immediate constituents are noun + noun (or adj + noun, etc.) on the one hand and a zero morpheme on the other. See MED s. v. limmel(e). Another example is aneoweste (< OE on nea-wiste), e. g., C 8698, cf. MED s. v. aneweste. See further 2.3. on ludere stefne, etc. Most authors dealing with compounds also give a survey of criteria; see for example Bloomfield (1933: 227-233); Marchand (1969: 20-30); Matthews (1974: 188-195). In Modern English the analysis of accent in compounds is complicated through the existence of level stress in some compounds. Level stress in compounds probably did not exist yet in Old English and Middle English. What is called morphological isolation by Marchand (1969: 21 ff.), should be called phonological isolation. — Another indication of morphological isolation is if the order of the elements in a compound is not possible in the corresponding syntactic group, for example ModE grass-green (noun/adj compound), but as green as grass (syntactic group). See for example sodere worden, chapter 2.3., and cf. Bogholm 1944: 47-48, 51-52. One might even think that the combination rede golde-stauen forms a compound, but this is probably not the case. Therefore Hall (1920: II, 475) assumes that manuscript C's bule-hude is nevertheless a compound. Possibly gold and bule in the combinations mentioned are to be regarded as anaphoric peninsulas; on this concept see Kastovsky (1982: 218). The order of determinant/determinatum is typical for the Germanic languages, whereas the order determinatum/determinant is not rare in the Celtic and Romance languages, see 2.14. Because redcoat does not mean 'a red coat', but 'someone wearing a red coat', and cutpurse does not mean 'a purse which has been cut (off)', but 'someone who cuts other people's purses ofF, cf. OED s. v. Cutpurse (in contrast to mincemeat, which originally meant 'meat that has been minced', see OED s. v. Mincemeat). The animate determinatum 'someone' is represented by the zero morpheme, whereas redcoat and cutpurse form the composite determinant. Sometimes also called polysynthetic compounds, in German Zusammenbildungen (see for example Kastovsky (1982: 170, 179, 210)).

522

Hans Sauer

19. The analysis in terms of the underlying sentence and the analysis in purely morphological terms lead to different results in the case of synthetic compounds. See Kastovsky (1968: 3-4, 7 ff., 27), and Kastovsky (1982:190-191, 209-214) on this question. 20. An example for a classification into fairly few basis types is Carr (1939: xxixff.). Carr does not separate primary determinatum and clearly deverbal determinatum in his basic classification, nor does he separate pronouns and adverbs from adjectives, if they occur as determinants. Examples for a more detailed classification are Marchand (1969); Koziol (1972); see also Jönsjö (1979: 30-35), who has 28 different morphological patterns, not all of which seem to occur in the general vocabulary. Moreover, Jönsjö and Koziol take only morphological shape, but not morphological structure into account, nor does Bauer (1983). 21. OE wel-dced is probably a loan-translation after Latin beneficium. — Jönsjö (1979: 31) and Bauer (1983: 206) have one example each for an adv/nounformation, namely ME Wid-fare and ModE now generation (Wid-fare [ < ON viö + fari] strictly speaking should be classified as adv/vb + 0). — Adverbs which are locative particles do not belong here; see 2.6. 22. The handbooks do not always point out the fact that their classification is not the only possible one. It is, moreover, doubtful whether an entirely satisfactory classification can be achieved; cf. Bauer (1983: 201-202). This shows once more that language is not only structured and patterned, but at the same time also a continuum which is sometimes difficult to break up into clear-cut segments. 23. For literature on each type see footnote 20 above. 24. For Modern English cf. Marchand (1969: 60); Bauer (1983: 202). There are no reliable figures for Old English; cf. Gardner (1968). 25. For instance by Carr (1939: xxix); Hansen et al. (1982: 4 3 ^ 4 , 52-53); cf. Bauer (1983: 30). — The determinatives comprise not only most of the noun/nouncompounds, but also all the other types of compounds that have a determinant/ determinatum structure, that is all except the dvandvas and the exocentric compounds; the latter are listed as the third basic category by Carr (1939: xxix). 26. They can be paraphrased by a sentence with the verb 'to be', but they can usually neither be paraphrased by a genitive phrase (cnaue-child φ 'cnaue's child') nor by a prepositional phrase (cnaue-child φ 'child for a cnaue', etc). 27. Compounds where two aspects of the same unit are equally prominent are usually also taken to be dvandvas, e. g., ModE poet-painter, conductorcomposer, fighter-bomber, but these are sometimes difficult to distinguish from attributive copula compounds. 28. Sometimes they are also called appositional compounds; see e. g., Carr (1939: 327-329); Bauer (1983: 30). 29. wap-man is ambiguous: it is an obscured compound and can either be derived from OE wcepned-man 'the man is a wsepned' [i. e., 'male'] or from wcep(e)nman 'the man has a weapon' (cf. BT). It is copulative in the former case, but rectional in the latter. 30. munec-child is perhaps not copulative ('the child is a monk'), but rectional, see the definition given by BT s. v. munuc-cild Ά boy that is being brought up to be a monk'. 31. Probably less convincing are paraphrases like 'the song [leojj] constitutes an insult [hoker]\ etc. 32. Of course not all designations for time are copulative, e. g., not lif-dag. 33. Nevertheless the first element can probably be seen as the determinant and the second as the determinatum.

Lajamon's

compound nouns and their morphology

523

34. There is no opposition to sea-riche implied. 35. In these and some other compounds, a constituent is polysemantic and therefore both constituents are synonymous in one meaning, but not in others. It is not easy to decide which meaning has been selected in the compound. Run, staf and writ can mean 'letters' and the compounds run-stauen and writ-runen probably simply mean 'letters'; see the glossary in Madden's edition. A meaning 'runic letters' for run-stauen and 'secret letters' for writ-runen can, however, not be excluded. Cf. Wyld (1930: 15), who translated leod-runen as 'mysteries', but (1937: 32) as 'discourse, counsel'. On run, cf. also Wyld (1937: 35-38); on the meanings of gleo- cf. (1937: 51-52). Other difficult cases are herd-swein, wode-burje, wode-scage, and the compounds with -kin as dm (mon-kin, metecun). 36. marma(n) - does not occur independently in Middle English any longer; marmestan is last attested in LajB, see MED s. v.; it is superseded by marble-ston, see MED s. v. For driht cf. MED. In manuscript Ο driht is often replaced or omitted. 37. There are also a number of rectional compounds in LajB, where either the determinant or the determinatum seems superfluous, e. g., dujeöe-cnihtes, herekempe, wan-sipes\ these cannot be investigated here (on wan cf. Wyld (1933: 187)). 38. The genitive-compounds are not mentioned by Quirk-Greenbaum (1973: Appendix I) (and cf. §4.70) and are excluded by Bauer (1983: 240-241). They are, however, described by Marchand (1969: 65-69); Koziol (1972: 60-61); Hansen et al. (1982: 54); see also Smith (1956: I, xxi). 39. An intrusive vowel between the constituents of a compound was not rare in late Old English, see Campbell (1959: § 367) (and cf. §§341, 348-350, 359, 204.1 for Old English usage with respect to connecting vowels). On the function of the -e in Middle English cf. also Bergsten (1911: 43-47). 40. Another reason is that genitives other than in -5 (and other than the o/-genitive) are receding in Middle English. Cf. Mustanoja (1960: 70 ff.). 41. Almes-mon does not belong here. It was a noun/noun compound in Old English and Middle English, because the -s of almes was part of the stem ( < lat. eleemosyna); in Modern English, however, the form was re-interpreted and alms is now treated as a plural, cf., e. g., ODEE, s. v. alms. 42. Originally it was a syntactic group, which then coalesced into a compound (in German: Zusammenrückung). 43. Cf. also Lundene-burh and Lundenes tun\ Bellinges-jat, but Leir-chestre; leodene swike C 378 and leod-swike Ο 378. Furthermore La$B has craft-mon, while it is craftsman in Modern English; LajB has lades-men, lodes-men, whereas it had been lad-men in Old English (and it is lod-cniht in LajB). The distribution of dai-liht and dajes lihte is according to metre and formula in LajB: the latter is frequently and mainly used in the formula drihte(n) fre scop jyes dajes lihte(n), e.g., CO 7423, 9274, etc; it could however be a syntactic group in this formula. 44. The syntactic paraphrase for sandes-mon is either 'The man delivers a message' or 'The man is a messenger'. L a j Ä uses sonde often in the meaning 'messenger', e. g., CO 213, 1561, 1574, etc. 45. In Old English, the distinction between compound and syntactic group was at least clear in the inflected cases, but adjectival inflexion was lost in Middle English. There are a few traces of it left in La^B (see Begholm (1944: 51-52); cf. Mustanoja (1960: 275-277)), but they are not always reliable (cf. soöere worden). 46. This seems to be true for most (if not all) of their occurrences in Early Middle English. Wroöere hele is syntactically restricted and occurs mainly after to, see

524

47.

48.

49.

50. 51.

52. 53. 54.

55.

56. 57.

58.

59.

60.

Hans Sauer the instances given by OED s. v. Wrother-heal, and cf. OED s. v. Goder-heal·, MED s. v. goder hele; d'Ardenne (1936: 219 (§ 82)). For ludere stefne see MED s. v. loud adj.; cf. Wyld (1937: 29-30); d'Ardenne - Dobson (1981: 210, 270-271, 321). Grammars often devote a separate section to numerals, but attributive numerals are syntactically largely equivalent to attributive adjectives. Koziol (1972: 63) treats num/noun as a distinct type; Marchand (1969) does not distinguish it from adj/noun. Seoue-niht possibly vacillates between syntactic group and compound in LajB, cf., e. g., Seoue niht 7 cenne dcei ..., C 703, and pat he wolde beon seoueniht pcere, C 1195. Cf. also OED s. v. Sennight, and Faiß (1978: 167-171). The classification of all varies; see for example MED s. v. al: 'lim. adj. & n.'; Campbell (1959: 261 (§638) and 294 (§725)): adj. or pronoun; Quirk/ Greenbaum (1973: 108-109): pronoun. Both Marchand (1969: 74) and Koziol (1972: 64) regard all as a pronoun. — Unlike the adj/noun-compounds, Aildrihten cannot be regarded as having coalesced from a corresponding syntactic group (cf. footnote 50 below). — Na-ping is classified as pron. and as adv. by MED s. v. no-thing, but as sb. and adv. by OED s. v. Nothing. A syntactic paraphrase of Aildrihten would be 'drihten rules all' (and not * 'drihten is all'). Because dryhten (and indryhteri) are occasionally inflected as adjs. in Old English. On the formula drihtenum Gode (instead of drihtene Gode) see Β TS s. v. dryhten adj. (?); cf. Toth (1984: 35). Cf. also Modern English constructions of the type Lord God, where Lord certainly is not regarded as an adjective. Some in addition also occur as adjs. or nouns, cf. OE wider. ut-lajen is a loan word from Old Norse. Its morphological analysis in English is complicated, because a zero morpheme has to be postulated as determinatum: out-law means 'someone who is outside the law'. However it cannot be classed among the bahuvrihi-compounds (see 2.12.), because these are defined as 'someone who has [sc. what is indicated by the complex determinant]'. See Carr (1939: 175 ff.). The type seems to be relatively young, not older than West-Germanic. — An element can also appear as a noun in one compound and as a verb in another, cf. ModE playgoer ('someone who goes to plays') with playboy ('boy who plays' [lexicalized]). See Kastovsky (1968: 247-248 and 170-172). On OE sp(r)cec cf. also Seebold (1970: 456). In a syntactic paraphrase it would probably be easier to use the vb: 'House where one can speak'; 'The iueres play with each other'. But OED s. v. playfellow analyses as sb/sb. luere (OE gefera) originally was a deverbal noun, too (cf. Kastovsky (1968: 243)), but was probably no longer felt as such. For the deverbal character of welle see Kastovsky (1968: 110-111) and Seebold (1970: 538). But OED gives Wellstream as sb/sb, and the compounds wellesstrcem and wcelles-water (see 2.2. above) could also point to an analysis of walle-strem as noun/noun-compound. — Bend-hus, quale-hus, scaöe-ded, scaöewerc have here been classified as noun/noun-compounds. The latter seems somewhat strange: it appears to be pleonastic (in the sense that the determinant is superfluous) and might be a corrupted form. Madden (1847: III, 460) suggests holde king or weorld-king as the original form. — The MED apparently does not list bercnihtes and beortime. For Old English, the type has been extensively treated by Kastovsky (1968). For the deverbal character of those determinata that are derived from strong verbs see also Seebold (1970) (e. g., -feht, -gute, -lure). On tilie as an agent noun derived from tilian, cf. OED s. v. Tilie. — In Germanic, these formations

Lajamon's

61. 62.

63.

64.

65. 66. 72.

73.

74.

compound nouns and their morphology

525

had no zero suffix yet, but were derived with an explicit suffix. Because the suffix was subsequently lost, they can be regarded as zero-derivations from Old English onwards, cf., e. g., Germanic * gut-i-z > OE gyte > ME gute. [On the description of deverbal nouns see now also the paper by D. Kastovsky in this volume.] According to Marchand (1969: 76) the type forms verbal-nexus compounds, but it should more specifically be described as forming synthetic compounds. Determinata that occurred as independent words in Old English are: -bere, -feht (-fiht), -flot, -geong, -jiue, -gute, -kempen (cf. Kastovsky (1968: 241)), -lure, -plaie, -ras, -scaöe, -scop, -scenc, -sene, -swike, -(i)tilie, -ward. Marchand (1969: 38-39, 76-78) has the same problem for Modern English, although he makes its solution seem easier than it actually is. Thus, he analyses bus stop, sheep walk, etc., as n o u n / v b + 0 compounds, but inkstand, sidewalk, etc., as noun/noun compounds. For this problem in Old English, cf. Kastovsky (1968: 93 ff.), and passim. Other determinata which are here regarded as primary nouns for Early Middle English are -gomen, -iueren, -kempe, -sid, -stram, -spell, -ponk, -work. — Kastovsky (1968) starts basically from the semantic side of the problem and classifies many nouns that can be regarded as the nominalization of a verbal action (i. e., that can be paraphrased as 'someone who does ...' or 'something which is done ...') as deverbal. But he is aware of the fact that the morphological derivation did not always work in the same direction in Germanic and Old English (cf. Kastovsky (1968: 93 ff.)). Weak verbs especially are often derived from nouns (and not the other way round) or both go back to the same root. For example: Kastovsky (1968: 241-242) lists cempa 'fighter, warrior' as deverbal, although he admits himself that it is a denominal formation ( < OE camp) and not derived from the verb OE campian (which is itself derived from camp). Old English weorc is ultimately derived from the Germanic root *werk-, but it is not synchronically derived from the Old English verb wyrcan (weorc is treated by Kastovsky (1968: 283-289)). Sid 'journey', etc., can be taken to mean 'act of journeying', but morphologically the verb sidian is derived from the noun sid. An example of a deverbal noun which was no longer felt as such is Old English feond: even in Old English, it was no longer connected with the verb feo(ga)n 'to hate'; see Gardner (1968: 43 ff.). The compound wardes-men C 9633 also points toward a synchronically primary status of ward. The form pohten is 3rd person sg. (for regular jjohte). There is quite a variety of forms in Old English (cf. BT): gristbite, gristbitian, gristbitung; grist-batian, grist-batung. — Gristbat, which occurs in LajB (see 5.1.8.), seems to be Middle English, cf. MED s. v. grisbaten ff. and grist-bat. It is often maintained that the -ing combinations are not nominal compounds, but derivations from the compound-verb gristbitian (grist-batian) 'to grind or to gnash one's teeth', see for instance MED s. v. grisbating. It is, however, more probable that the compound-verb was originally formed from the nominal compound grist-bitung, see Schrack (1966: 56, 98); cf. Marchand (1969: 100 ff.) for general considerations. They are still generally rare in Early Middle English literary texts, whereas they occur more frequently in Early Middle English personal names; see for instance Selten (1969) and (1975). As do MED s. v. kine- and Smith (1956: I, 123); but cf. OED s. v. kine-. A morphological indication that kine- is not a prefix is furnished by kinelich C7051: assuming that -lich was an adjectival suffix even in Old English (although it originated from the noun lie 'body'), kine- cannot be a prefix, because there is no English word-formation pattern *prefix/suffix (cf. Bauer

526

67.

68.

69.

70.

71.

75.

76.

77.

78.

79. 80.

81.

82.

83.

Hans Sauer (1983: 213-214)). Another indication that kine- is not a prefix is the formation kinelouerd, which seems to be a copula compound 'the lord is a king'. [On the history of the suffix -lich > ModE -ly, see now the paper by C. Guimier in this volume.] Cf. for example Warren (1978: 62), who regards ModE birthday, nickel content as compounds whose determinant or determinatum can synchronically be regarded as deverbal. To this group also belong historically fulluht, an obscured compound which had probably become a moneme ( < OE full-wiht), and neo-west, which however is used as adv. and prep, only (in the form aneoweste, see footnote 9 above). Slceht was probably no longer felt to be deverbal in all combinations, because the elements of wal-slaht are (at least in one sense) tautological: both can mean 'slaughter' (on this problem cf. footnotes 29-30 above). -witere does not occur as an independent word. It is probably derived from witan 'to know'. Wittie is the Middle English form of Old English witega\ see OED s. v. Wide. The type noun/vb + ende, i. e., noun/present participle or noun/substantivized present participle, is rare in LajB. The only example is adjectival: sce-lidende men (cf. also wel-donde — adv/present participle; al-waldinde (God) — pronoun/present participle). Literature on this type: Kärre (1915); Carr (1939: 209-215); Visser (1966: II, 1069 ff.). — Uncompounded agent nouns in -ere are very rare in Lajfi, too. One example is Hamun pe ridere, CO 4633. Carr (1939: 229), Oakden (1935: II, 158) nevertheless classifies wcei-witere as a poetic compound, which again shows that the distinction between prose and poetic compounds, as he makes it, is not satisfactory. The only exception is Cuninges-burh (manuscript C), Conanesborh (manuscript O), which, however, shows Old Norse influence and is formally distinct from king. Cf. Blenner-Hassett (1950: 31); Smith (1956: I, 123-124; II, 7). Most formations did not survive the Early Middle English period; only the frequently used formations kinedom and kineriche survived till the later Middle English period; cf. OED, s. v. kine-, and MED, the entries following kine-. Moreover, there are two personal names: manuscript C has king Marke, which manuscript Ο transforms into Kine-marck; both C and Ο have the name Kinebord. Cf. Carr (1939: 344-354), who has a section on intensifying elements: jyeod is treated on p. 349-350; leod on p. 353. See also Ingersoll (1978: 85 ff.), especially 137-139 (leod) and 140-141 (peod)\ for affixoids in Modern German, see Wellmann (1975: 98 ff.) (passim). On these place names see Blenner-Hassett (1950); on Kaer cf. also Smith (1956: I, 76) s. v. cair. Originally it refers to the Great St Bernard pass, and then (in English) by extension to the Alps; see Blenner-Hassett (1950: 49). According to Blenner-Hassett (1950: 71) it is a French loanword; but because it already occurs in Old English (e. g., yElfric), it is probably derived from Latin and represents an Old English (hybrid) loan-translation. For Old English instances, see also Bately (1980: 285-286, 420). Munt-ioye 'mount of joy' goes back to a montem gaudia, see Tobler/ Lommatzsch s. v. monjoie (and cf. MED s. v. mount η. (1) (a) and (b)). Montagu points to a Möns acutus. Cf. Marchand (1960: 411^416). Marchand denies that combinations of more than three words are compounds in English; according to him they are generally syntactic groups. Noröhumbre means 'dwellers north of the Humber', but also 'country north of the Humber'. The name of a people is often also used as the name of the

Lajamon's

84.

85.

86. 87.

88. 89. 90. 91.

92.

93.

94.

95. 96.

compound nouns and their morphology

527

country they inhabit in LajB, e. g., AZst-Aingle, ALst-sax. Cf. Modern German Bayern, Franken, Hessen, Schwaben, Sachsen, and see CODEPN, xiii-xiv. Whitesuneday has coalesced from the syntactic group Old English on hwitan sunnandag(e), which is still represented in manuscript C as a Whiten-sunendceie\ it originally means 'white Sunday', not *'day of the white sun'. Its obscuration and change in semantic structure (i. e., lexicalization) is also reflected in word-division: In LajB it is sometimes spelt as three words and sometimes as Witesone daije. Cf. Ο DEE s. v. Whit Sunday, Faiß (1978: 184). A morphological analysis as cerche-biscopstol is less likely, although there is also the compound biscop-stol in LajB. Old English ealdor was a de-adjectival noun formed from the adj (e) aid + suffix -or. Cf. MED s. v. halven-del(e) n., adj., and adv.; most adj. sup. & n. la (c) and 5; OED s. v. Thirdendeal. Manuscript Ο has only wiöer-iwinnes, but erpe-tilie and plei-ueres. The Old English form was eorö-tilie, which shows that the i- in eoröe-itilie did not have a real function any longer. The most recent investigation of the prefix Middle English i- ( < OE ge-) and its function is Stanley (1982: 25-66). Alternatively it would have to be explained as an attributive copula compound 'the Bruttes are ware', which is not convincing. For Old English and Middle English, see Bergsten (1911: 74-78); for Modern English, see Johansson (1980). As was pointed out above (footnote 62), the determinatum of synthetic compounds in fact often occurs as an independent word, too. kine- and ware- died out in Middle English; twi- is no longer productive. Morphologically conditioned allomorphs in Modern English are of the NeoLatin type, e. g., Anglo-French 'English-French', politico-economical; see Marchand (1969: 89-90) and Bauer (1983: 213-216). The determinant of i-lond ( < OE ieg-land) was later associated with ile < O F isle 'island' and thus newly motivated. — The determinant of sti-ward was not felt to be connected with stig > ModE sty(e), see ODEE s. v. steward. For Stonehenge, see CODEPN. — La3amon often explains how places were given their names; his explanations, however, are often fanciful and do not represent the true etymology. On obscured compounds see for example: Bergsten (1911: 9-24); Götz (1971); Faiß (1978). Obscuration refers to morphological changes in a compound. These morphological changes are often accompanied by loss of motivation and changes in the semantic structure. Obscuration can be regarded as a form of morphological lexicalization. On lexicalization, see, e.g., Lipka (1981: 119-132); Kastovsky (1982: 164-168); Bauer (1983: 42-61). Originally, the determinatum ended in -wreca; according to AEW s. v. arendrace, it was reformed in analogy to Old English racu 'narrative'. In 6935-46 La3amon enumerates a number of gods and in 6947-6954 he relates them to the days of the week; but not all of the gods of 6935-46 appear among the days of the week, and not all of the gods after whom the days of the week have been named are mentioned in 6935-46. Cf. also Hall (1920: II, 471). Manuscript C 6937 Phebus, Saturnus 6938 Woden 6939 Iupiter 6940 Mercurius 6941 Appollin 6942 Teruagant 6945 Frxx

6947 6951 6952 6953

Wodenfyunre -fyunresdai Freo(n)-Frida>i Saturnus-&mm/cF/ sunne-Sonedcei 6954 Mone(n)-Monedcei Tidea-Tisdcei

528

Hans Sauer Manuscript Ο 6937 Phebus, Saturnus 6938 Woden 6939 lubiter 6940 Merchurius 6941 Appolin 6942 Teruagant 6945 Frea

6954

Μ one- Μ one day Tydea- Tisdei 6950-51 Woden- Wendesdei fronre- porisdai 6952-53 Frea-Friday Saturnus- Sateresdai

97. The MED is inconsistent here: it classifies est as a noun only, but north as a noun, an adj and an adv. Est-del, for example, is listed under est n. 5a(e), but north-del under north adj. 1(a). 98. On sib, see BT and BTS, s. v. gesib and sib(b); OED s. v. Sib sb1 and Sib a. and sb 2 . For middel see MED s. v. middel n. and adj. 99. Cf. OED s.v. Gospel and Faiß (1978: 123-124) - In certain compounds, confusion between two words seems to have arisen, although this does not affect word-class. This is the case with bur 'room, chamber' and burh 'castle, town', cf. bur-cniht and burh-cniht\ bur-men and burh-men; bur-folc, C 14178, instead of burh-folc. It is also the case with firing 'crowd' ( < OE gepring) and dreng 'man, warrior, free tenant' ( < ON), cf. here-dring and here-öringes. Cf. MED s. v. dreng; OED s. v. dreng·, thring sb1 and sb 2 . 100. Cf. Marchand (1969: 60-61). — In LajB -mon occurs as determinatum in noun/noun: alder(e)-, almes-, burh-, chep-, craft-, driht-, dujeöe-, gleo-, hafd-, hired-, hors-, scip-, see- (see-), sibbe-, weorc-, weorld-, wif- (wimmon), furthermore in Löja- (Lawe-) and Rom-, -mon as determinatum in noun + i/noun: cunnes-, erendes-, lades-, sondes-, steores-, peines-, wardes-, -mon as determinatum in adj/noun (possibly not all of these combinations are compounds): bleo-, Dense-, Englisce-, Frensce-, freo-, leue-, priue-, 101. Cf. Luhmann (1906: 183-193). 102. The imperative compounds and the inversion compounds are loan compounding types. As pointed out above (2.12. and 2.14.), the former do not occur in LajB, while the latter were experienced as a foreign pattern that was not productive in LajB. — Wolff (1975: 51) says that the number of hybrid compounds in Middle English is overrated, but he gives neither evidence for this statement nor references to the literature he attacks. — There are also compounds which are loan-translations, cf. the section on inversion compounds above (2.14.), but the question to what extent they are represented in LajB could only be answered by a systematic comparison of LajB with its Old French model. 103. For in-siht, see Ο DEE s. v. insight; for ut-lajen, see Ο DEE s. v. outlaw and footnote 54 above. Fare-coste 'behaviour, circumstances' is given as a loancompound by some, but not by MED. 104. Castel-, which was taken over in Late Old English, is not taken into account here. 105. Cantel-cape probably consists of two Latin elements, see MED s.v. cantercäpe\ cf. cöpe n. and OED s. v. Cope sb 1 ; but cf. OED s. v. Cantel-cape. 106. bleo as well as brunie have been taken over from Old Norse and Old French, see MED s. v. bio adj.; brinie. The determinant of burne-hode, manuscript O, is, however, the native word. 107. On Tisdai and porisdati see Sunden (1941/42: 290-292). 108. wceilaucei is an interjection. 109. See Tolkien (1962) [manuscript A]: clensing-fur, f. 62r/4; swiing-wike, f. 17v/7; kealche-cuppe, f. 58v/23; baban, f. 63r/23; giue-gauen, f. 52r/6; wesche-disch is in manuscript C only: see Dobson (1972: f. I76v/I1).

Lajamon 's compound nouns and their morphology

529

110. See for instance Bergsten (1911: 65-73); B a u g h - C a b l e (1978: §49, 140, 217-218). — The frequency of nominal compounds in Old English poetry is exceptional and cannot be taken as a yardstick for compounds in other texts. 111. The list therefore cannot be used for an investigation of the spelling and the inflection of manuscripts C and O. 112. Brook —Leslie emend manuscript C (but not manuscript O) to lofnjd-cniht, which is unnecessary. 113. Probably OE win-sele·, but the interpretation as wynn-sele (ME) 'hall of joy' cannot be excluded; cf. Oakden (1935: II, 164).

Abbreviations and signs AEW BT, BTS

C CODEPN EM Ε LajB ME MED ModE NCBEL Ο ODEE ΟΕ OED OF ON / 0

Ferdinand Holthausen, Altenglisches etymologisches Wörterbuch (Heidelberg: Carl Winter, 1934; 3. unveränderte Auflage 1974). J. Bosworth — Toller, Τ. N., An Anglo-Saxon dictionary (Oxford: University Press, 1898); Supplement, by Toller, Τ. N. (Oxford: University Press, 1921). Manuscript C: London, British Library, Cotton Caligula Α. IX. Ekwall, Ellert, The Concise Oxford Dictionary of English Place-Names, 4th edition (Oxford: University Press, 1960). Early Middle English. La3amon's Brut, edited by Brook —Leslie 1963-1978 (see references). Middle English. Middle English Dictionary, cited by H. Kurath et al. (Ann Arbor: University of Michigan Press, 1952 ff.). Modern English. The New Cambridge Bibliography of English Literature, edited by George Watson, vol. I (Cambridge: University Press, 1974). Manuscript O: London, British Library, Cotton Otho C. XIII. The Oxford Dictionary of English Etymology, edited by C. T. Onions et al. (Oxford: University Press, 1966). Old English. The Oxford English Dictionary, edited by J. A. H. Murray et al., 13 vols. (Oxford: University Press, 1933). Old French. Old Norse. connects determinant and determinatum. zero morpheme.

References Ackerman, Robert W. 1952 An index of the Arthurian names in Middle English ( = Stanford University Publications, University Series, Language and Literature X) (repr. New York: A M S Press, 1967). Adams, Valerie 1973 An introduction to Modern English word-formation ( = English Language Series 7) (London: Longman).

530

Hans Sauer

Bately, Janet, ed. 1980 The Old English Orosius ( = EETS S. S. 6) (Oxford: University Press). Bauer, Laurie 1983 English word-formation ( = Cambridge Textbooks in Linguistics) (Cambridge: University Press). Baugh, Albert C. — Cable, Thomas 19783 A history of the English language (London: Routledge & Kegan Paul). Bergsten, Nils 1911 A study on compound substantives in English ( P h . D . dissertation, Uppsala: Almqvist & Wiksell). Blenner-Hassett, Roland 1950 A study of the place-names in Lawman's Brut ( = Stanford University Publications, University Series, Language and Literature IX. 1) (repr. New York: AMS Press, 1967). Bloomfield, Leonard 1933 Language (New York; repr. London: George Allen & Unwin, 1935). B0gholm, N. 1944 The Layamon texts. A linguistical investigation ( = Travaux du Cercle linguistique de Copenhague 3) (Copenhagen: Einar Munksgaard). Brook, George Leslie — Leslie, R. F., ed. 1963-78 Layamon: Brut, 2 vols. ( = EETS 250 and 277) (Oxford: University Press). Campbell, A. 1959 Old English grammar (Oxford: University Press). Carr, Charles T. 1939 Nominal compounds in Germanic ( = St Andrews University Publications 41) (London: Humphrey Milford). d'Ardenne, S. R. T. O., ed. 1936 pe liflade ant te passim of Seinte Iuliene (Liege); repr. ( = EETS 248) (Oxford: University Press, 1961). d'Ardenne, S. R. T. O. - Dobson, E. J., ed. 1981 Seinte Katerine ( = EETS S. S. 7) (Oxford: University Press). Dobson, E. J., ed. 1972 The English text of the Ancrene Riwle ed. from B.M. Cotton MS. Cleopatra C. VI ( = EETS 267) (Oxford: University Press) [manuscript C]. Faiß, Klaus 1978 Verdunkelte Compounds im Englischen ( = TBL 104) (Tübingen: Gunter Narr). Gardner, Thomas J. 1968 Semantic patterns in Old English substantival compounds (Ph. D. dissertation, University of Heidelberg). Götz, Dieter 1971 Studien zu den verdunkelten Komposita im Englischen ( = Erlanger Beiträge zur Sprach- und Kunstwissenschaft 40) (Nürnberg: Hans Carl). Hall, Joseph 1920 Selections from Early Middle English 1130-1250, 2 vols. (Oxford: University Press). Hansen, Barbara — Hansen, Klaus, et al. 1982 Englische Lexikologie. Einführung in Wortbildung und lexikalische Semantik (Leipzig: VEB Verlag Enzyklopädie). Ingersoll, Sheila Most 1978 Intensive and restrictive modification in Old English ( = Anglistische Forschungen 124) (Heidelberg: Carl Winter).

Lajamon's compound nouns and their morphology Jönsjö, Jan 1979

531

Studies on Middle English nicknames. I. Compounds ( = Lund Studies in English 55) (Lund: CWK Gleerup). Johansson, Stig 1980 Plural attributive nouns in present-day English ( = Lund Studies in English 59) (Lund: CWK Gleerup). Kärre, Karl 1915 Nomina agentis in Old English ( = Uppsala Universitets Ärsskrift 1 ...) (Ph. D. dissertation, University of Uppsala: A.-B. Akademiska Bokhandeln). Kastovsky, Dieter 1968 Old English deverbal substantives derived by means of a zero morpheme (Ph. D. dissertation, University of Tübingen). 1982 Wortbildung und Semantik ( = Studienreihe Englisch 14) (Düsseldorf: Schwann-Bagel; Bern und München: Francke). Ker, N. R. 1963 The Owl and the Nightingale, reproduced in facsimile ... ( = EETS 251) (Oxford: University Press). Koziol, Herbert 1972 Handbuch der englischen Wortbildungslehre (2nd edition) (Heidelberg: Carl Winter). Levi, Judith N. 1978 The syntax and semantics of complex nominals (New York: Academic Press). Lipka, Leonhard 1981 "Zur Lexikalisierung im Deutschen und Englischen", in Wortbildung, edited by L. Lipka — H.Günther ( = Wege der Forschung 564) (Darmstadt: Wissenschaftliche Buchgesellschaft), 119-132. Luhmann, Adolf 1906 Die Überlieferung von Lajamon's Brut ... ( = Studien zur englischen Philologie 22) (Halle/Saale: Max Niemeyer). Madden, Frederic, ed. 1847 Lajamons Brut, or Chronicle of Britain, 3 vols. (London: Society of Antiquaries). Marchand, Hans 1960 "Die Länge englischer Komposita und die entsprechenden Verhältnisse im Deutschen", Anglia 78: 411-416. 19692 The categories and types of present-day English word-formation (Munich: C. H. Beck). Matthews, P. H. 1974 Morphology ( = Cambridge Textbooks in Linguistics) (Cambridge: University Press). Mustanoja, Tauno F. 1960 A Middle English syntax. I. Parts of speech ( = Memoires de la Societe Neophilologique de Helsinki 23) (Helsinki: Societe Neophilologique). Oakden, J. P. 1935 Alliterative poetry in Middle English [II.] A survey of the traditions, with assistance from Elizabeth R. Innes ( = Publications of the University of Manchester 236; English Series 22) (Manchester: University Press). Pennanen, Esko V. 1982 "Remarks on syntagma and word-formation", Folia Linguistica 16: 241-261. Quirk, Randolph — Greenbaum, Sidney 1973 A university grammar of English (London: Longman).

532

Hans Sauer

Renwick, W. L. - Orton, H. - Wakelin, M. F. 19663 The beginnings of English literature to Skelton 1509 (London: The Cresset Press). Schrack, Dieter 1966 Ne. 'to spotlight' und Ίο stagemanage'. Studien zur Entwicklung und Struktur der verbalen Pseudo-Komposita im Englischen ... (Ph. D. dissertation, University of Tübingen). Seebold, Elmar 1970 Vergleichendes und etymologisches Wörterbuch der germanischen starken Verben ( = Janua Linguarum, series practica 85) (The Hague: Mouton). Selten, Bo 1969 Early East-Anglian nicknames: 'Shakespeare' names (= Scripta Minora Reg. Soc. Hum. Litt. Lundensis 1968-69: 3) (Lund: CWK Gleerup). 1975 Early East-Anglian nicknames: bahuvrihi names (Scripta Minora Reg. Soc. Hum. Litt. Lundensis 1974-75: 3) (Lund: CWK Gleerup). Smith, A. H. 1956 English place-name elements, 2 parts ( = English Place-Name Society 25-26) (Cambridge: University Press). Stanley, E. G. 1982 "The prenominal prefix ge- in late Old English and early Middle English", Transactions of the Philological Society (Oxford: Basil Blackwell), 25-66. Strang, Barbara Mary Hope 1970 A history of English (London: Methuen & Co). Strauss, Jürgen W. 1980 "Compounding in Old English poetry", Folia Linguistica Historica 1: 305-315. Sunden, Κ. F. 1941/42 "Notes on the vocabulary of Layamon's Brut", Studia Neophilologica 14: 290-292. Tobler- Lommatzsch 1915 ff. Altfranzösisches Wörterbuch (Berlin; later: Wiesbaden: Franz Steiner). Tolkien, John Ronald R., ed. 1962 The English text of the Ancrene Riwle: Ancrene Wisse, edited from MS. CCCC 402 ( = EETS 249) (Oxford: University Press) [manuscript A]. Toth, Karl 1984 "Altenglische Interlinearglossen zu Prospers Epigrammata und Versus ad coniugem", Anglia 102: 1-36. Visser, F. Th. 1966 An historical syntax of the English language, Part II (Leiden: E.J. Brill). Warren, Beatrice 1978 Semantic patterns of noun-noun compounds ( = Gothenburg Studies in English 41) (Göteborg: Acta Universitatis Gothoburgensis). Wellmann, Hans 1975 Deutsche Wortbildung, Typen und Tendenzen ... 2. Hauptteil: Das Substantiv (Sprache der Gegenwart 32) (Düsseldorf: Schwann). Wolff, Dieter 1975 Grundzüge der diachronischen Morphologie des Englischen ( = Anglistische Arbeitshefte 7) (Tübingen: Max Niemeyer). Wyld, Henry Cecil 1930-37 "Studies in the diction of La3amon's Brut", Language 6 (1930): 1-24; 9 (1933): 47-71 and 171-191; 10 (1934): 149-201; 13 (1937): 29-59.

SUSAN SHEPHERD

On the functional development of repetition in Antiguan Creole morphology, syntax, and discourse

People often comment on the musicality and distinctive rhythm of Creole languages. One factor contributing to this impression is the use speakers make of repetition. What I refer to as repetition exists in the speech of an individual speaker on grammatical and pragmatic levels, and includes such devices as reduplication, formulaic repetition of a clause which occurred earlier in the utterance, and repetition of a particular key theme in the structuring of discourse and narratives. Little attempt has been made to show the relationship between these various forms of repetition, yet the functions they serve are remarkably similar. Emphasis, intensification, and humor are the most common functions, as well as the most widespread in terms of level. There is a connection between these functions. Enough repetition goes beyond emphasizing a point and exaggerates it, leading to a comic interpretation. The boundaries here are not clearly drawn, and the meanings conveyed may be ambiguous. A listener may choose to interpret a speaker's use of repetition as humorous when in fact intense anger was being expressed. Both speakers and listeners use the potential ambiguity to their own advantage. By drawing upon data from Antiguan Creole, an English-based Creole spoken in the West Indies, and child language, I would like to propose several ways in which the uses Creoles make of repetition have been encouraged and reinforced historically. Mühlhäusler (1980: 33) has suggested that "... hardly any word or construction in a pidgin or Creole can be traced back to a single origin." Certainly that is the case here. Reduplication is used extensively in Creole languages, and its functions have frequently been discussed both for Creoles and other languages (Moravcsik (1978), Hancock (1980), Alleyne (1980)).

534

Susan Shepherd

Hancock has stressed the role of reduplication in lexical expansion. Four primary uses listed by Hancock and Alleyne are relevant here. i) ii) iii) iv)

intensive — emphatic iterative attributive semantic extension

In Antiguan Creole the intensive — emphatic use is by far the most pervasive. One finds reduplicated adjectives, adverbs, and verbs, as in the following examples. (1)

Me swim fas-fas an far-far. "I swam very fast and very far."

(2)

De girl crazy-crazy. "The girl is really crazy."

(3)

Dey don come dong ya know. Dey ease-dong ease-dong. "They don't come down openly you know. They creep up very stealthily."

(4)

De dress red-red. "The dress (she had on) was bright red."

Less frequently reduplication marks iteration or habitual action. This occurs with predicate adjectives and verbs. (5)

De pikny sicky-sicky. "The child is always getting sick." (Contrast with De pikny sick-sick, "The child is very sick.")

(6)

She a wok-wok. "She is always working."

Obviously these two functions are not unrelated. The iterative examples could be paraphrased as "Someone does something (or is something) constantly, or extremely often." Reduplication with an attributive function is less clearly related. It is distinguished from emphatic meanings by the intonation pattern. Thus, (4) uttered with falling intonation and a relatively low tone would mean, "The dress was a reddish color (but not true red)." This use of reduplication

Repetition

in Antiguan Creole morphology,

syntax,

and discourse

535

is not found extensively. One could not say, for example, "Me go far-far," and mean with that utterance, "I went sort-of far." On the syntactic level, repetition is used in topicalization when the focus is placed on predicate adjectives and verbs. This is not true for topicalized elements in other word classes. Patterns of topicalization are discussed in more detail in Shepherd (1981) and will be briefly outlined here. Topicalized predicate adjectives occur both in the fronted position, following the topicalizer a, and in the normal sentence position. Thus, the examples under (7) below are the topicalized forms of the sentences given in (8). (7)

a. A emty de cup emty. "The cup is empty]" b. A sick Clave lie sick. "Clavelle is really sick." (She's not just skipping work.)

(8)

a. De cup empty. "The cup is empty." b. Clavelle sick. "Clavelle is sick."

Topicalized verbs follow a similar pattern. They occur twice in the sentence — in fronted position after a and in normal sentence position. (9)

A tiif me tiif? A borrow me borrow. "I stole it? (No,) I borrowed it."

(10)

A mosa lef she mosa lef. "She must have left (by now)."

(11)

A rongo me rongo. Me na min tap long. "I hurried! I didn't stay long."

A second common syntactic structure involves repetition of a main clause at the end of the sentence. This use is related to, and sometimes hard to distinguish from, certain discourse uses of repetition which will be discussed later. In such sentences, the repeated section may also contain an expansion. More detailed information may be added.

536

Susan Shepherd

(12)

But dey long, dey long up in de air. "But they're tall, tall up in the air." (describing a mocka jumby, a character who walks on stilts)

(13)

Dat time drum play, ya know drum drum play. "At that time drums played, you know drums played." (The speaker may have repeated drum to indicate that drums were continually playing, or the repetition may simply have been a hesitation phenomenon.)

(14)

i tek smell tek smell, i tek smell. "He kept sniffing, he sniffed."

The openings of examples (16), (17), and (18) also involve this type of structure. Such repetition is a form of focusing attention on an important or central point. As with reduplication, the major function of repetition in such structures is intensive — emphatic. When we look at how discourse is structured, the functions of repetition branch out. A speaker often makes a point or tells a story which is then summarized by repetition of a key phrase or clause which contains the main point. (15)

De fire dat me a play wit done burn me aready. Kyaan burn again. Kyaan burn again. Me done get burn. "The fire that I am playing with (a man) has already burned me. It can't burn (me) again. It can't burn (me) again. I already got burned." (I've learned my lesson.)

(16)

Muh see Rut, muh see Rut. She say she might come but she na know if she a come ... dat wha she say. She might come. "I saw Ruth. I saw Ruth. She said she might come but she doesn't know if she's coming ... that's what she said. She might come."

(17)

i godfather say so, i godfather say so aready ... ya bu de godfather say so in de early. "The godfather said so, the godfather said so already ... Yes, the godfather said so in the early (part of the gathering)." (from Reisman 1964: 180)

Repetition in Antiguan Creole morphology, syntax, and discourse

(18)

537

Dey don even serve notice, dey don even serve. Come now masquerade, don even serve come roun now an play. "They don't even notice, they don't even notice. Now when the masqueraders come, they don't even notice when they come around now and play."

In each of the above utterances a clause or phrase which is central to what is being told, is repeated at the end of the story. This device occurs frequently in natural conversations, and is also used extensively in narratives. In the examples given the function is once again emphasis. It also occurs (as in (16) and (17)) when speakers feel they have not been believed or paid attention to. The examples discussed thus far involve repetition within a single utterance or turn taken by a speaker. Repetition also play a significant role in conversations and arguments. Reisman has described Antiguan conversations as contrapuntal in nature (1964: 1974). There are a set of themes ... which alternately or simultaneously fade in and out, but never seem to be lost. When a new theme enters, or tries to, the ongoing rush of voices crowds it out. It knocks again, and eventually after several attempts it takes a lead and the other voices join themselves to it, until the crowd of voices is again assaulted by an old theme wishing reentry. (1964: 223) Repetition serves to emphasize a point of view or suggested topic and to hold ground for an individual speaker. When used effectively it determines the course a conversation takes. The uses of repetition can be ambiguous. For example, the two (or more) speakers taking part in an argument may each simply make the same statement over and over, without seeming to listen to the other(s). According to Reisman, "... Antiguan argument is distinguished by the lack of attempts at 'rational' persuasion." Group feeling generally determines which point is allowed to dominate. An argument quickly becomes heated, and a speaker's anger is often apparent in the force with which he repeats his stand in the matter. However, a second function of repetition often results in such arguments ending with no hard feelings, sometimes with the participants all laughing. Repetition is central to humor. The more often something is repeated the funnier it becomes. An outsider often has trouble determining when an argument is occurring in earnest, and when in jest. The ambiguity and potentially "light" interpretation gives a

538

Susan

Shepherd

speaker who is angry about something a means of expressing the anger without necessarily antagonizing the others involved. He can always claim to have been joking. The following utterances constitute one five year old child's contributions in a conversation that lasted approximately five minutes. She was angry because the other two participants (also age five) did not choose to play the game she had suggested. Instead these two children continued to play and look at books. Their utterances, which have been left out, concern their activities.

Κ: I don wan noise in muh head

Me na wan na noise on muh head ... me na wan na noise on muh head Na wan na noise on muh — an you not readin nuttin. You dus turn de page to look an / chupsj paper ya wan? me na wan na noise in muh head naa a wonder if ya — wonder if muh s- ya- ya know muh serious

I don't want any noise in my head. (I don't want to hear any noise.) I don't want any noise in my head. ... I don't want any noise in my head. I don't want any noise in my — you're not reading anything. You just turn the pages to look /chups is an expression of disgust/ You want paper? I don't want any noise in my head you know. I wonder if you — wonder if I — you know I'm serious.

muh say muh na wan no noise in muh head

I said I do-'t want any noise in my head.

a ya mosa gon play wi de sudn. muh na ...

All of you have to go play with the thing. I don't ...

wonder if ay a na know me serious wonder if aya na know me serious serious me serious serious

I wonder if all you don't know that I'm serious. I wonder if all you don't know I'm very serious. I'm very serious.

Repetition in Antiguan Creole morphology, syntax, and discourse

539

me serious serious serious

I'm very very serious.

muh serious me na wan na noise in muh head muh serious

I'm serious. I don't want any noise in my head. I'm serious.

At this point one of the children she is angry at draws her back into their game. Having expressed her anger to her own satisfaction, she rejoins the other children with no sign of her former discontent. In the next passage Ρ has complained because the food she has been offered at a friend's house contains pork, which she does not eat. In order to lighten the force of her initial comments, she continues to repeat her negative statements even after the other participants have moved on to a new topic. They eventually recognize the humor in what she is doing and realize that she is telling them not to take her too seriously. No one is insulted. In the section up to the line, all of P's utterances are given. The section under the line, before her last comment about pork, consists of miscellaneous conversation lasting three and a half minutes. All participants, including P, are talking about subjects other than food.

P: M: L: P: M: P: L: P: M: P:

Me na eat dem subm ya. Wey- a pig gut she got? I tink I tink cattle. Pig an cattle no no no. dis a cattle? de whitey one hav pig No. Now wey de cattle? De whitey one is swine. No me na wan notn out in deah today. a pig meat na. Pig meat yech.

P: me na go eat notn na, cos metoo much a swine ya

Ρ: M: L: P: M: P: L: P: M: P: P:

I don't eat that thing there. Wait- is it pig gut she got? I think cattle. Pig and cattle, no, no, no. Is this cattle? The whitish ones are pig. No. Now where's the cattle? The whitish one is swine. No, I don't want anything out of there today. It's pork you know. Pork, yuck. I'm not going to eat anything you know, because Itoo much swine there

540

Susan

Shepherd

Me na eat dat deah. Too mucha pork.

I'm not going to eat that. Too much pork.

ay a aya tell she fuh sen pork cos ya know me a come today? M: mmmhmmm. /laugh/

Did all you tell her to send pork because you knew I was coming today? M: Yes.

Repetition plays a role in the structuring of narratives. It occurs frequently in both natural, spontaneous narratives and in more structured storytelling. Repetition of a significant phrase or clause may occur several times within a given narrative, and gives cohesion to the story. It often indicates what the teller views as the theme, and may also be used for clarification. While carrying out fieldwork in Antigua, one method of data collection used was a joint picture book "reading" activity. Stories based on a book with pictures (and no words) that tell a complex story were elicited. The methodology is based on work carried out by Bamberg (1983; 1984) with Germanspeaking children. The Antiguan children whose narratives have been analyzed used an average of five phrasal repetitions of the type discussed earlier per story. Some examples from their stories follow. (21)

De deer trow off de boy in un river an de dog an all drop ova in dey too. When de boy drop ova in de river headway, de dog an all drop in dey too. "The deer threw the boy off (his head) into a river and the dog fell in there too. When the boy fell in the river headfirst, the dog fell in there too."

(22)

I na see de krapa — de krapa. Dus he an de clos. Na see de krapa. "He didn't see the frog. Just he and the clothes (are in the room). He didn't see the frog."

(23)

A wey i dey? say de boy. De dog put i head in de bokl. De boy a call out "Frog". De dog-de dog is- de dog put i head in de bokl. "Where is he? said the boy. The dog put his head in the bottle. The boy is calling out 'Frog'. The dog put his head in the bottle."

Repetition

in Antiguan

Creole morphology,

syntax,

and discourse

541

Repetition as a narrative technique has been discussed with respect to standard English and Black English in America as well (Halliday - Hasan (1976), Kernan (1977), Heath (1983)). The examples given above could be viewed as the result of a common narrative practice rather than as something characteristic of the Creole. For the sake of comparison, comparable data from 8 and 9 year old black and white children from working class backgrounds in the United States were also collected. The five black American children all had one to two such repetitions per story. The five white American children used repetition of the type described once or not at all in their stories. This is significantly lower than the number (five) used by the Antiguan children. The numbers given for these two groups should probably be even lower because some of the counted repetitions seem to be qualitatively different from those given above. (24)

It was a reindeer or somethin /Int: Yeah/ a reindeer. It was a reindeer.

(25)

An the bees started chasin after the dog and the beehive fell down an all the bees started chasin after the dog. The boy went up in a tree an started lookin in a hole in the tree. An then the owl pushed the boy off the tree an then the bees started chasin after the dog.

In (24) the speaker looks to the interviewer for confirmation, which is given. Upon receiving it, he repeats the name of the animal, and then his entire previous utterance. He is probably using the repetition to get back to his story after a digression. The repetition in (25) results at least in part from the nature of the task. The children paged through the book as they told the story. In this case, more than one picture show the bees chasing the dog. It is possible that this story-teller was simply telling what she saw on each page. The Antiguan Creole speakers' repetitions cannot be similarly accounted for. This was determined by comparing the point at which a page was turned to the point at which the repetition occurred. It has been claimed here that cohesion is one function of repetition in narratives and discourse. This should be explained in more detail as it may not be immediately apparent why this is so. A particular event is opened and focused on, and the repeated initial segment provides its closing, often by calling attention to an idea, activity, or state of being that will be important at a later point in the story

542

Susan Shepherd

or conversation. (22) is a good example of this. The picture book tells the story of a boy and a dog who go searching for a frog that has disappeared from a jar in the boy's bedroom. They look in many places and have several "adventures" before they finally find the frog. The repeated element, "na see de krapais central to the story. The repetition in (23), "de dog put i head in de bokl," does not signal a key element of the story as a whole, but foregrounds what is to occur in the next pictures. The dog walks to the window with the jar still on his head, falls out the window, and the jar breaks. The cohesiveness supplied by repetition can apply to the story as a whole, as in (22), or simply to one section, as in (23). It is interesting that young children speaking standard English also make use of repetition as they learn the language. Most studies which mention repetition in child language refer to utterances made by another person and repeated by the child (Keenan, 1977; Heath, 1983). A look at detailed transcripts, however, reveals that most children repeat their own utterances as well. Unfortunately many researchers have left this type of repetition out of their transcripts. Ochs (1979), Ochs - Schieffelin - Piatt (1979), and Scollon (1979) discuss self-repetition as well as other-repetition. According to Ochs (1979: 60) self-repetition is used as an attention-getting device, to emphasize a point, and to clarify or correct a previous utterance. It also seems to be a form of verbal play, and occurs in what might be called language practice when the child is playing alone. The creole-speaking children I have worked with make use of these types of repetition before they learn the specific uses of repetition used in the adult language. By around age four children speaking standard American English use repetition for these functions less than younger children, having learned other devices to signal these meanings. Creole-speaking children at this age continue to use a great deal of repetition, but in new functions conforming to the adult model. The similar functions of the types of repetition discussed here probably have their origins in several sources. Reduplication, due to its extensive use in some of the African substrate languages involved in the development of the English-based Caribbean Creoles, provided a model for the use of repetition for semantic extension and to convey emphasis-intensification, iteration, and attribution. The prevalence of reduplication in Creole languages in general and other languages indicates that universal strategies were also at work here.

Repetition

in Antiguan Creole morphology,

syntax,

and discourse

543

Consideration of morphological/grammatical structures alone does not adequately explain the extent to which repetition is used. It merely provides a possible source for the particular device chosen to serve the functions mentioned. When the sociohistorical background to Caribbean Creoles is taken into consideration, we see that this grammatical structure probably received reinforcement from discourse processes. Repetition is often used for clarification when a speaker has not been understood by a listener. It probably occurred frequently in the speech of slave trader, slave owner, overseer, and slave in the early stages of pidginization and creolization. Repetition for humorous effect may have arisen out of this. As Reisman (1970) and Rickford — Rickford (1980) show, linguistic and cultural ambiguity is often in evidence in Creoles. "The existence of public and more 'acceptable' interpretations is exploited by Blacks for the communication of more private or 'unacceptable' meanings," (Rickford — Rickford 1980: 362). A slave could thus have made fun of an overseer by repeatedly asking for clarification, or by continuing to repeat a Creole phrase which has not been and is unlikely to be understood. A similar phenomenon can be observed today in interactions between Antiguans and tourists. An Antiguan may answer a tourist's question in Creole, and when not understood continue to repeat the same Creole answer several times although capable of speaking standard English. Observers find this quite amusing, and an 'acceptable' explanation (from the point of view of the tourist) can be offered for such behavior. The speaker can always claim not to speak standard English, or to have been unaware that there was a language problem involved in the tourist's inability to understand the utterance. Further reinforcement for the uses creole speakers make of repetition comes from child language and the role children play in creolization. Children use repetition for functions related to the functions in the adult language — emphasis, attention-getting, and clarification. In addition, standard English speakers sometimes repeat words or phrases that they wish to emphasize or focus attention on, although their use of this device has not become as formulaic as that of creole speakers. I propose that the above-mentioned factors in combination have led to the development of the system described for Antiguan Creole. The existence of one of these factors in isolation would probably not have led to the pervasive use of repetition that we find in the creole, and it would be impossible to trace this use to a single source.

544

Susan

Shepherd

An interesting point to be made in conclusion is that for many West Indians repetition has become a basic pragmatic device not restricted to the Creole. It occurs in their speech regardless of whether they happen to be speaking standard English or a more creolized form. That is, a linguistic characteristic that originated in the Creole has become part of standard West Indian English, and distinguishes it from other standard varieties. Non-creole speakers may be inclined to devalue this, claiming that frequent repetition sounds childlike or simplistic. They evaluate the language according to their own pragmatic system and set of presuppositions, failing to recognize what to a West Indian might be an extremely witty joke or complex play on words.

References Alleyne, Mervyn 1980 Comparative Afro-American (Ann Arbor: Karoma). Bamberg, Michael 1983 "Temporality and backgrounding — a developmental study of narrative construction" (Paper presented at Linguistics Colloquium Series, University of California, Berkeley). 1984 "Anaphorische Pronomen in Erzählungen von 4-9jährigen Kindern" (Paper presented at the Sixth Jahrestagung der Deutsche Gesellschaft für Soziologie, Bielefeld). Bickerton, Derek 1975 Dynamics of a creole system (Cambridge: Cambridge University Press). 1981 Roots of language (Ann Arbor: Karoma). Hancock, Ian 1980 "Lexical expansion in creole languages", Theoretical orientations in creole studies, edited by Albert Valdman and Arnold Highfield (New York: Academic Press), 63-88. Halliday, Michael Alexander Kirkwood — Hasan, Ruqaiya 1976 Cohesion in English (London: Longman). Heath, Shirley Brice 1983 Ways with words: language, life, and work in communities and classrooms (Cambridge: Cambridge University Press). Keenan, Elinor Ochs 1977 "Making it last: Repetition in children's discourse", Child discourse, edited by Susan Ervin-Tripp and Claudia Mitchell-Kernan (New York: Academic Press), 125-138. Kernan, Keith T. 1977 "Semantic and expressive elaboration in children's narratives", Child discourse, edited by Susan Ervin-Tripp and Claudia Mitchell-Kernan (New York: Academic Press), 91-102. Moravcsik, Edith A. 1978 "Reduplicative constructions", Universals of human language, Volume 3, edited by Charles A. Ferguson and Edith A. Moravcsik (Stanford: Stanford University Press), 297-334.

Repetition in Antiguan Creole morphology, syntax, and discourse

545

Mühlhäusler, Peter 1980 "Structural expansion and the process of creolization", Theoretical orientations in creole studies, edited by Albert Valdman and Arnold Highfield (New York: Academic Press), 19-55. Ochs, Elinor 1979 "Planned and unplanned discourse", Syntax and Semantics Volume 12: Discourse and semantics, edited by Talmy Givon (New York: Academic Press), 51-80). Ochs, Elinor — Schieffelin, Bambi B. — Piatt, Martha L. 1979 "Propositions across utterances and speakers", Developmental pragmatics, edited by Elinor Ochs and Bambi B. Schieffelin (New York: Academic Press), 251-268. Reisman, Karl 1964 "The isle is full of noises": a study of Creole in the speech patterns of Antigua, West Indies (unpublished Ph. D. dissertation, Harvard University). 1970 "Cultural and linguistic ambiguity in a West Indian village", AfroAmerican anthropology, edited by Norman Whitten and John Szwed (New York: The Free Press), 129-144. 1974 "Contrapuntal conversations in an Antiguan village", Explorations in the ethnography of speaking, edited by Richard Bauman and Joel Sherzer (Cambridge: Cambridge University Press), 110-124. Rickford, John — Rickford, Angela 1980 "Cut-eye and suck-teeth: African words and gestures in New World guise", Perspectives on American English, edited by J. L. Dillard (The Hague: Mouton), 347-365. Scollon, Ronald 1979 "A real early stage: an unzippered condensation of a dissertation on child language", Developmental pragmatics, edited by Elinor Ochs and Bambi B. Schieffelin (New York: Academic Press), 215-227. Shepherd, Susan 1981 Modals in Antiguan Creole, child language acquisition, and history (unpublished Ph. D. dissertation, Stanford University). in prep. "The impact of the child's world on pairing form and function in Antiguan Creole and English", Contrastive pragmatics, edited by W. Oleksy (Amsterdam: John Benjamins).

CARMEN SILVA-CORVALAN

Modality and semantic change

The variable use of the conditional (C) or the imperfect (I) indicative form in contexts where standard varieties of Spanish use the imperfect subjunctive (S) characterizes some varieties of Spanish spoken in Old Castile and the Basque country. Examples 1 to 3 illustrate the alternation between these three verbal forms: (1)

ιDönde vas con cincuenta afios? Unicamente si los hijos emigraran (S) y harian (C) alii for tuna, vamos, entonces si harian (C) for tuna, se llevarian (C) a los padres cuando seriamos (C) mäs viejos. (V, CV 25)1 'Where do you go when you are fifty? Only chance is if the children emigrated and made a fortune, well then, if they made a fortune they might take their parents when we were older.' Standard forms: emigraran (S); hicieran (S); llevarian ( C); fuiramos (S)

(2)

iSabe lo que haciamos para que no se caerian (C) de la cama? (G, CV 1) 'You know what we did so that they didn't (wouldn't fall off the bed? Standard form: cayeran (S)

(3)

Pues lo mismo si era (I) del Esta'o, a lo mejor pone pinos ο que se yo. (A, CV 11) 'Well just the same if it belonged to the State, may be it'd plant pine trees or who knows what.' Standard form: fuera (S)

The variants include the two forms of the imperfect subjunctive (e. g. cantara and cantase, to which I at times refer as "the -ra variants"), the conditional (e. g., cantaria, "the -ria variant"), and

548

Carmen

Silva-Corvalän

the imperfect indicative (e. g., cantaba, "the -ba variant"), which alternate in various syntactic contexts in the speech of a given speaker, as examples 4 and 5 illustrate: (4)

Es como si quedaria (C) muerto, ino?, pero no importa porque ya no me vuelve a molestar. (T, CV 36) 'It's as if it became dead, right?, but it doesn't matter because it doesn't bother me again.' Standard form: quedara (S)

(5)

Es como si te lo quisieras (S) llevar todo para ti, ino?; lo quieres llevar a pedacitos. (T, CV 36) 'It's as if you wanted to take it all for you, right?; you want to take it in small pieces.' Standard form: quisieras (S)

It is the thesis of this paper that the synchronic variability attested in Covarrubias represents a stage in a relatively slow but continuous chain of changes in the verbal system of Spanish, which involves a process of semantic weakening of the verbal forms which may be used to convey modality. I further suggest that one of the factors which triggers the loss of certain linguistic forms is the egocentric and concrete nature of everyday human interaction and propose a principle of distance to account for some types of linguistic change. The alternate use of the conditional and the subjunctive appears to be the consequence of a change which probably started before the sixteenth century, as Lavandera (1975) notes in a study of this phenomenon in Buenos Aires Spanish. This change has involved an extension of the meaning of the conditional form from a stage in which it served as an indicator of tense ('past of the future', example 6), to a stage in which it also served to convey modality, as in examples 7 and 8, and to the present stage, illustrated in examples 1, 2, and 4, in which the conditional has taken over syntactic and semantic features of the imperfect subjunctive. 2 (6)

Juan decidio que iriamos (C) manana. 'John decided that we would go tomorrow.'

(7)

Si viniese, lo veria (C), 'If he came, he would see it.'

(8)

Entonces tendria (C) unos doce anos. Ί was probably twelve then.'

Modality and semantic change

549

On the other hand, the use of the imperfect in the apodosis of both past real and non-past unreal conditions, in alternation with -ra and -ria, is attested as early as the sixteenth century by Keniston (1937), but its occurrence in the protasis of past and non-past unreal conditions appears to be restricted even today. Thus, examples of the type of 3, with an imperfect in the protasis of an unreal condition, are rare. Table 1 shows the relative frequencies of the variants in the protasis and apodosis of conditional sentences, and in all other possible contexts in the data from Covarrubias. 3 Table 1. Relative frequency of the variants in a number of linguistic contexts.

Total Ν Protasis Apodosis All other contexts

-ra\-se

ria

72 33

25

35%

46 64% 19 58%

169

36

21%

133 79%

-ba 1 1% 14 42%

In this paper, I assume as correct the analysis which proposes that each verbal form conveys a basic grammatical meaning with respect to the categories of tense, mood and aspect (Alarcos 1970; Bello 1977; Cartagena 1976-77).4 This basic meaning may be extended in certain contexts in such a manner that the meaning differences between two or more forms may become neutralized. These possibilities indicate that the verb system constitutes a potential locus of semantic ambiguity and, consequently in all likelihood a source of change. In fact, the historical development of the verb systems in the Romance languages bears out this hypothesis since, as it is well known, they have all undergone a number of changes including the loss and addition of forms, and the reduction or extension of the meaning of certain existing forms. In what follows, I restrict my attention to the modal meanings conveyed by the three verbal forms under study (-ba, -ria, -ra). Mood has been defined (Gili Gaya 1971; Lyons 1977; Real Academia Espanola 1973) as the verbal category which specifies the predication of the clause with respect to the possibility or factuality of the event (i. e., epistemic modality) and/or with respect to the necessity of the event (i. e., deontic modality). Frequently, -ba, -ria, and -ra forms are used to communicate subjective epistemic and

550

Carmen

Silva-Corvalän

deontic modality, a possibility derived from their irrealis meaning (Cartagena 1976-77). The fact that these verbal forms are used to express subjective judgments of events makes them good candidates for reinterpretation given the difficulty for the language user of recovering different subtle modal meanings very precisely. I proceed to present the contexts where the substitution of -ria for -ra occurs in Covarrubias Spanish. 1. The alternation in conditional sentences. One important question posed by the alternation is whether it is associated with any meaning differences at the present stage of the language, a possibility which I investigate below. There is general agreement that in contemporary Spanish the difference between real and unreal conditions is grammaticalized in the form of the verb used in the protasis: indicative forms express real conditions, i. e., those where the condition and the consequence represent a natural relation (example 9), or those whose relations is judged by the speaker to be factual (example 10); subjunctive forms express unreal or contrafactive conditions. (9) (10)

Si llueve se moja el suelo. 'If it rains the ground gets wet.' Si hace buen tiempo iremos al campo. 'If the weather is good we'll go to the country.'

I will have nothing to say about real conditions here because Covarrubias Spanish appears to be like all other varieties of Spanish in this respect. It seems, therefore, that it is in the sphere of irrealis, non-factual events that subtle meaning differentiations are more likely to be made, thus paving the way to variation and change. Unreal or contrafactive conditions are further subdivided into non-past and past, a tense difference which correlates with the probability that the condition may or may not be fulfilled. 1.1 Non-past conditions. The verb forms which may occur in this type of conditional sentence are presented in 11: (11)

Non-past conditions Standard Spanish Covarrubias Spanish s Protasis (-ba), -raj-se -ba, -raj-se, -ria Apodosis (-ba), -ra, -ria -ba, -ria

Modality

and semantic change

551

Non-past unreal conditions are counterfactual inasmuch as they present a situation which is not a fact in the real world, but they may be judged by the speaker as more or less likely to be a fact in the future. Thus, these conditions may be marked by the features [ +contrary to fact], [ ± possible in a future world]. This semanticpragmatic difference is not grammaticalized in Spanish, however, as shown by examples 12 and 13: (12)

Si viviera (S) en Australia, seria diferente. [ +possible] 'If I lived in Australia, it would be different.'

(13)

Si viviera (S) Juanita, seria diferente. [—possible] 'If Jane lived, it would be different.'

If the indicative mood is the mood of factivity and assertion, it appears possible to hypothesize that the conditional form, considered to be part of the indicative mood in Spanish, is used in the protasis to indicate that the situation is [ +possible] 6 so that we would predict its occurrence in sentences of the type 12 but not of 13. This is not so, however, since -ria and -ra do not contrast semantically in the protasis of conditional sentences, as illustrated in 14 to 17:

(14)

(15)

[ + possible] El si saldria (C) otra cosa, se marchaba (I), pero estä tan malo Espana, estä muchisimo malo. (A, CV 11) 'He, if something came up, would leave, but things are bad in Spain, real bad.' Standard form: saliera (S); marchaba (I) iDönde vas con cincuenta anos? Unicamente si los hijos emigraran (S) y harian (C) alii for tuna, vamos, entonces si harian (C) fortuna, se llevarian a los padres cuando seriamos mäs viejos. (V, CV 25) 'Where do you go when you are fifty? Only chance is if the children emigrated and made a fortune, well then, if they made a fortune they might take their parents when we were older.' Standard forms: emigraran (S); hicieran (S); hicieran (S) [ — possible]

552

Carmen

Silva-Corvalän

(16)

Ahora que si yo me volveria (C) joven, me dirian (C) "Te vas a volver joven, Joaquina, y no vas a cobrar una perra de pension," yo mejor queria (I) la juventud. (J, CV 19) 'Now, if I became young, if they told me "You're going to become young again, but you're not going to get your pension." I would prefer youth. Standard forms: volviera (S); dijeran (S); querria (C)

(17)

Y deciamos, "Si ahora despertaria (C) y se viese (S) con su trajecito del Corazon de Jesus."1 (G, CV 2) 'And we said, "If she woke up now and saw herself in her Sacred Heart of Jesus dress." ' 7 Standard form: despertara (S), viese (S)

Examples 14 to 17 show that conditional and imperfect subjunctive forms do not contrast in the protasis of non-past conditional sentences, i. e., linguistically, they are in free variation since they both may occur in [ +possible] and [ — possible] contexts. Examples 15 and 17, where -ria and -ra occur in conjoined protases, offer good support for the lack of meaning differences. In example 15, to emigrate (coded in the subjunctive) and to make a fortune (coded in the conditional) are possible events, while in example 17, to wake up (coded in the conditional) and to see oneself (coded in the subjunctive) are not possible events. I have said above that the I also occurs in unreal conditions in alternation with C and S, but that its occurrence in the protasis appears to be restricted. Thus, examples of the type 3 (repeated below with number 18 for ease of reference) are rare (1%, see table 1), while examples of the types 14, 16, and 19, with I in the apodosis, are as frequent as those with C (see table 1). (18)

Pues lo mismo si era (I) del Esta'o, a lo mejor pone pinos ο que se yo. (A, CV11) 'Well just the same if it belonged to the State, maybe it would plant pine trees or who knows what.'

(19)

Si hiciesen el pantano, se podia (I) hacer zona de regadios.(JL, CV 22) 'If they built the dam, they could build a watering system.'

Here again the question arises of whether the imperfect indicative contrasts semantically with the C and the S in the protasis of unreal

Modality and semantic change

553

conditionals, as in 18, and whether it contrasts with the C in the apodosis, as in 19. In the standard language, the alternation between imperfect, conditional, and subjunctive appears to be meaningful in the context of epistemic and deontic modality. The three forms convey a decreasing degree of certainty and a decreasing degree of obligation. 8 Thus, 19 would convey the message that the "watering system" is more likely to be built, but in 20 the consequence appears less likely, and in 21 it is the least likely: (20)

Si hiciesen el pantano, se podria (C) hacer zona de regadios.

(21)

Si hiciesen el pantano, se pudiera (S) hacer zona de regadios. 'If they built the dam, they could build a watering system.'

On the other hand, the deontic necessity in 22, 23, and 24 gradually decreases from obligation to weak obligation to advice, as I have tried to convey in the translations: (22)

Debias (I) ir al simposio. 'You should go to the symposium.'

(23)

Deberias (C) ir al simposio. Ί think that you should go to the symposium.

(24)

Debieras (S) ir al simposio. 'It would be advisable for you to go to the symposium.'

It is interesting that in Covarrubias Spanish the imperfect subjunctive never occurs in the apodosis of non-past unreal conditions, so the opposition in this context is now between I and C, while in the protasis of these conditions the I is practically absent. It seems to me that if the status of factuality and the time frame of the unreal condition are created in the protasis (often also in combination with extralinguistic factors, though), then the absence of the I in this context appears to be a good indication that this form does contrast semantically with the C. Indeed, the C is in all contexts future oriented and irrealis; the I, on the other hand, may be past oriented and realis. This difference would account for the occurrence of the C and the non-occurrence of the I in a context which requires the creation of an unreal, future-oriented frame.

554

Carmen

Silva-Corvalän

It is not difficult to see, however, that the semantic differences which appear to be associated with the forms under scrutiny are subtle, and since they are related to the speaker's subjective judgment of the situation, explicit supporting contexts are almost impossible to find. I will try to show that there seems to be a difference, though, by comparing examples 25, 26, and 27 with 28 and 29: (25)

jUy! Si me darias (C) mil pesetas, en casita me estaba (I). (A, CV11) 'Wow! if you gave me 1000 pesetas, I'd stay home.'

(26)

Si me pagarian (C) a mi del Esta'o, ra'iba (I) contigo por ahi. (FP, CV 17) 'If they paid me retirement pension, Yd go around with you.'

(27)

Si ella estaria (C) sola, se arreglaba (I) con ese diner ο. (J, CV 20) 'If she were alone, she'd have enough with that money.'

(28)

Si tendria (C) mucho dinero, compraria (C) unos terrenos, ya te dije, entre Retuerta y Contreras. (T, CV 36) 'If I had a lot of money, Yd buy, I told you, some land between Retuerta and Contreras.'

(29)

Seria (C) feliz si no tendria (C) este dolor. Ύά be happy if I didn't have this pain.'

(T, CV 36)

The conditional statements in 25 to 29 are counterfactual and, apparently, they all seem to be possible in a future world. However, those in 25-27 with an imperfect form in the apodosis, seem more probable than those in 28 and 29, which have a conditional verb form. In fact, the speaker who produces these two examples is very poor and has talked to me on several occasions about his longstanding dream of buying land and raising cattle, but the speaker knows that the dream will most likely remain a dream; furthermore, happiness is a tenuous concept, and "being happy" may be a very unlikely state to reach given that the speaker has suffered from a recurrent pain for about seven years and has lost any hope that a doctor will be able to cure him. The situations in 25-27 seem more probable, on the other hand. Example 25 responds to my asking the speaker to imagine that I would give her 1000 pesetas a day (about U.S. $10) and to tell me if she would continue working

Modality

and semantic change

555

under those circumstances, so it is not difficult for the speaker to view the condition and especially its consequence as almost real. In 26, the speaker, who is already 61 years of age, may in fact soon start receiving pension, in which case he would not have to work and he would be able to go around to different places with his interlocutor. Finally, in example 27 ella 'she' refers to a friend of the speaker's who is having financial problems because she is having to support her unemployed husband; clearly, the speaker presents the consequence, if the condition is fulfilled, as almost a fact. In sum, then, the system of non-past counterfactual conditionals in Covarrubias Spanish may be represented as in 30: (30)

protasis -raj-se, -ria, (-ba) + contrary to fact + possible in the future

apodosis -ba + contrary to fact + [stronger possibility]

-ria + contrary to fact + [weaker possibility]

1.2 Past unreal conditions. The verb forms which may occur in this type of conditional sentence are presented in 31, and illustrated for Covarrubias Spanish in 32 to 36: (31)

Standard Spanish Protasis:

{hubiera 1 \hubiese J + -do

i

hubiera

hubiese V + -do (habria μ hubiera 1 hubiese > + -do (habia) J (-ba) (present indie.) Si hubiera (S) estado vivo mi marido, aquel si que le hubiera contado. (B, CV 4) 'If my husband had been alive, for sure he would have told you.' Apodosis: (hubiera \ habria J»· + -do l(habia) J

(32)

Covarrubias Spanish

{

556

Carmen Silva-Corvalän

(33)

Si no hubiese (S) apedrea'o, me hubiese gustado que hubiese ido por ahi, por ese llano. (M, CV 24) 'If it hadn't hailed, I would've liked you to have gone there, along that valley.'

(34)

Si se habria (C) muerto antes habia (I) sido mejor. (Ca., Obs.) 'If he had died before it would've been better.' Standard forms: hubiera (S) muerto; habria (C) sido

(35)

Yo no he tenido hermanas asi tan gor das, perο si hubieran (S) esta'o asi gordas, gordas, no las dejo (pres. indie.) ir. (J, CV 20) Ί haven't had sisters that fat, but if they had been fat, fat, I don't let them go.' Standard form: habria (C) dejado

(36)

Si mi hermana no hubiera (S) esta'o asi, si se casaba (I). (J, CV 20) 'If my sister hadn't been like that, for sure she would've gotten married.'' Standard form: habria (C) casado

The absence of the prescribed sequence of tense rules in 35 and 36 indicates that the expression of modality takes precedence over grammaticalized constraints of tense agreement, and that speakers may view the consequence of a condition somewhat independently of the condition itself. Once the status of factuality and the time frame have been created by either the tense of the protasis or the extended context, the speaker is free to be more or less assertive in his judgment of the factivity and/or necessity of the consequence. The protasis of conditional sentences is the only past context in which the conditional may be starting to substitute for the subjunctive in Covarrubias Spanish. In all other contexts, the pluperfect subjunctive is the form used, just as in standard Spanish, and it is the only one used even in the apodosis, where the conditional perfect is allowed in the standard. The proposal that in the history of the Spanish verb system there has been a repeated process of semantic weakening accounts rather nicely for the maintenance of the pluperfect subjunctive in the context of past unreal conditions. Observe that the negation of the possibility of actualization, that is to say the feature [—possible], is an absolute value which is not susceptible to gradation. Accordingly, zero possibility is the most

Modality

and semantic change

557

certain context; it does not correspond to a probable fact. Indeed, a protasis coded in the pluperfect subjunctive in the affirmative form refers to a non-fact, an absolute irreality, and in the negative form it refers to a fact, a reality, as I show in examples 37 and 38, to be compared with 39: (37)

Si hubieras ido a Espana ... = no fuiste a Espana 'If you had gone to Spain ...' = you did not go to Spain, [-fact]

(38)

Si no hubieras ido a Espana ... = fuiste a Espana 'If you had not gone to Spain ...' = you went to Spain. [ + fact]

(39)

Si fueras (S) a Espana ... = podrias ir un dia a Espana. 'If you went to Spain ...' = you might go to Spain one day. [ +possible fact]

If we accept that the negation of the possibility of actualization is the most certain context in that it corresponds to a fact or a nonfact and is, therefore, definite and unambiguous with respect to modality, then my hypothesis about semantic change predicts that in the process of diffusion, past unreal conditions will not be a favorable context for the extension of the C. No subtle epistemic meaning differences are possible here to favor this extension. My hypothesis assumes as correct, and indeed the data provide evidence in support of, the statement that language innovations spread gradually. In the case of the changes studied here, the spread to various contexts has depended on semantic-pragmatic factors. 2. Adverbial and adjectival clauses. In clauses of this type, the occurrence of an indicative or a subjunctive verb form correlates with the features [ ± fact] in the real world, and the choice of subjunctive form with the speaker's judgment of the possibility of something being so. Indicative forms do not occur in purpose clauses; in the latter ones, only the imperfect subjunctive may occur, and in purpose and time clauses the tense is somewhat controlled by that of the main clause. Examples 40-44 illustrate all these facts for standard Spanish: (40)

Me acuerdo de eso como si fuera (S) hoy. Ί remember that as if it were today.

558

Carmen

(41)

Silva-Corvalän

(dio "] [asista (pres. subj.)} Le i f dinero para que) asistiera (S) J al simv-dar J fasista (pres. subj.)1 posio. a j*asistiera (S) j 'She [ g w % e g i v e ] him money to attend the symposium.'

(42)

_ , fva (pres. indie.) ] [ + fact] Lo cuido hasta que se { . ,' > , / [vaya (pres. subj.)J [r —fact, + stronger possibility] Ί take care of him until he leaves.'

(43)

Lo cuidaba (I)/cuidaria (C) hasta que se fuera (S). [—fact, + weaker possibility] Ύά take care of him until he left.''

(44)

Lo habria (C) cuidado hasta que se hubiera (S) ido. [ — fact, —possible] Ί would've taken care of him until he had left.'

In contexts of the type illustrated by 40, 41, and 43, Covarrubias Spanish substitutes the conditional for the imperfect subjunctive. Furthermore, both present and imperfect indicative may substitute for the subjunctive in hypothetical manner clauses. Examples 45-52 illustrate:

(45)

Hypothetical manner clauses Digamos que yo estoy, como se dice, como si estaria (C) reprimido todo el dia. (T, CV 36) 'See, I am, as they say, as if I were repressed all day long.'

(46)

C: l Y te acuerdas de todo eso? FP: jMe cago en diez! Me acuerdo como si es (pres. indie.) hoy. (FP, CV 18) C: 'And you remember all that?' FP: 'Shit! I remember it as if it were today.'

(47)

Si que me sirve igual, vamos, como si estaban (I) granadas. (T, CV 9) 'They surely are useful, see, as if they were ripe.'

Modality

(48)

(49)

(50)

(51)

(52)

and semantic change

559

Purpose clause Y Pedro, como comia tanto, pues siempre habia que apartarle la cabeza pa' que se entretendria (C) con los huesos. (A, CV 26) 'And Peter, as he ate so much, well we always had to leave him the head so that he entertained himself with the bones.' Time clause La maldicibn era que cuando naceria (C) su hijo, porque su mujer estaba embarazada, pues que se la llevaria a su castillo. (C, CV 6) 'The curse was that when his son was born, because his wife was pregnant, well that she would take her to her castle.' Place clause Pues teniamos que ir a donde nos mandaria el. (FJ, CV 17) 'Well we had to go where he sent us.'

(C)

Concessive clauses Estaba mala, pero como la llamara cualquier persona, aunque seria (C) α las dos de la noche, volaba. (D, CV 13) 'She might be sick, but if anyone called her, even if it was at two in the morning, she flew there.' Relative clause Queria una que seria (C) rica, y aquellas pues no le querian. (T, CV 9) 'He wanted one who would be rich, and those didn't want him.'

The substitution of indicative for subjunctive forms in adverbial and adjectival clauses is not categorical, however, as example 53 shows (see table 1 for frequencies in "other contexts"): (53)

Haciamos un agujerito asi en el, por la puerta para que saliesen (S) y entrasen (S) cuando las parecia a ellas. (G, CV 2) 'We made a hole like this, in the door, so that they went out and came in whenever they pleased.'

560

Carmen

Silva-Corvalän

Furthermore, as in the case of conditional constructions, the substitution does not take place in those contexts where the possibility of actualization of the situation is zero, i. e., [—fact, —possible], as illustrated in examples 54 and 55: (54)

Lo que tenian que haber hecho son fäbricas para que se hubiera (S) empleado la juventud. (M, CV 6) 'What they should have built is factories so that the youth would've been employed.''

(55)

[Habria trabajado con el] aunque no hubiese (S) ganado diner ο. (Τ, CV 36) '[I would've worked for him] even if I hadn 't made any money.'

3. Nominal clauses. Standard Spanish grammaticalizes the difference between reporting mands or statements in the mood of the subordinate verb: indicative for statements, subjunctive for mands. Thus, verbs of saying or communicating take a complement in the indicative, and verbs of ordering, requesting or demanding take a complement in the subjunctive, as shown in 56 and 57: (56)

Comunico que irias (C) manana. 'He communicated that you would go tomorrow.'

(57)

Pidio que fueras (S) manana. 'He requested that you go tomorrow.'

In 56 and 57, the relationship between the semantics of the main verb and the choice of mood is transparent, and even the use of a conditional form, for example in 57, would not cause ambiguity. However, the most frequently used reporting verb, decir 'to tell/ say', may report both a mand or a statement and it is the mood of the subordinate verb what disambiguates the meaning, as illustrated in 58: (58)

On the basis of a principle of inconicity of form and function we would expect that sentences like 58 would not be a favorable context

Modality

and semantic change

561

for the use of the conditional instead of the subjunctive, but this is not the case. Indeed, the occurrence of the C is very frequent in indirect reports of mands, as examples 59 to 63 illustrate. In Covarrubias Spanish, these types of constructions are ambiguous if they are cited in isolation, that is to say, without any further information they may be interpreted as reports of either mands or statements. My interpretations are based on my knowledge of the extended context in which 59-63 occurred. (59)

Ya ve lo que ha dicho el de la caja, que la llevaria (C), que iria (C) uste' conmigo. (J, CV 19) 'You see what the bank employee said, that I should take you, that you should go with me.'

(60)

Pues dicen que le han avisa'o de que vendria (C) a ver a T. que estä muy mal. (G, CV 29) 'Well they say they told him to come to see T. who is very sick.'

(61)

Entonces yo le decia que me compraria (C) novelas del oeste y tal. (T, CV 33) 'So I told him to buy me cowboy novels and things.'

(62)

Y la dije redondamente que iria (C) con quien la diera la gana. (A, CV 5) 'And I simply told her to go with whom she pleased.'

(63)

La dijo a su mujer que se saltaria (C) a la huerta de la bruja. (C, CV 6) 'He told his wife that he would jump into the witch's orchard.'

The ambiguity of examples of the type of 59 to 63 is resolved in real life communication, however, where utterances occur in a context, which generally allows the addressee to recover the appropriate meaning. Either the preceding or following verbal information, or the non-verbal information, makes it possible for the addressee to infer what the speaker wishes to communicate. Thus, it is only the wider context of the conversation what has allowed me to infer that 59 to 62 are reports of mands, and that 63 reports a statement. Even further, in the case of 63 judgment had to be suspended until later information disambiguated the referent of saltar, since when 63 was uttered it was unclear to me whether it

562

Carmen

Silva-Corvalan

was the speaker or his wife the one who would do the jumping. The overt expression of the subject in the noun clauses would have helped to disambiguate the meaning. However, perhaps somewhat optimistically, speakers rely on their interlocutors' inference capabilities and withhold the expression of disambiguating information. It is interesting, then, that the conditional has proceeded with neogrammarian regularity to replace the imperfect subjunctive even in contrastive contexts where the neutralization is a source of ambiguity. It is not surprising, then, that this phenomenon occurs in all other noun clauses which require the subjunctive in the standard language, i. e., where indicative and subjunctive do not contrast. Examples 64 and 65 illustrate: (64)

De nada te servia que irias (C) en buenas condiciones ni en malas. (A, CV 5) 'Whether you went in good or bad condition it was the same.'

(65)

La otra, la mayor, dice que queria que estaria (C) con ellos. (F, CV 27) 'The other one, the oldest, says that she wanted me to be with them.'

In noun clauses, the epistemic modality of the conditional and the subjunctive is said to be irrelevant because both forms may be [±fact] and [±possible]. However, notice that the events encoded in the noun clauses illustrated are [ + possible] in a time posterior to the time reference of the main verb. The generalization that I propose, therefore, is that in Covarrubias Spanish the conditional form has taken over all the modal and syntactic functions of the imperfect subjunctive except in [ — possible] situations in the past. We may further ask ourselves if the extension of the modal functions of the conditional has affected the basic values of this form. The answer appears to be positive, as I show below. 4. The changing moods of the conditional. As we have seen, in Covarrubias Spanish the conditional has kept its epistemic modality meaning and its original meaning of past of future. We have also seen that in conditional sentences and in adjectival and adverbial clauses the imperfect indicative may also be used in addition to the conditional. It seems, then, possible that a larger shift is in progress,

Modality

and semantic change

563

involving -ra, -ria, and -ba forms. An investigation of the data shows that, in fact, the imperfect indicative has replaced the conditional in all deontic utterances, as examples 66 to 68 show: (66)

Entonces pues ya, usted tenia (I) que ir a visitar la colegiata. (G, CV 1) 'Then, you should visit the church.'

(67)

Si fuesen ellos unas personas honestas y unas personas buenas, debian (I) decir "iFueraΓ (L, CV 24) 'If they were honest people and good people, they should say "Out!"'

(68)

Yo soy partidario de que los curas, los curas se tenian (I) que casar; serian mäs, mäs puros en todo. (T, CV 13) Ί believe that priests should get married·, they'd be purer, purer in every way.'

It seems to me quite possible that the conditional no longer cooccurs with verbs of obligation and that the three way distinction illustrated in examples 22, 23, and 24 has been lost, so that only one form remains: the imperfect indicative. This development would be in agreement with Bybee — Pagliuca's (this volume, p. 66) observation that "the semantic development of modals points to the unidirectional evolution of subject-oriented modalities 9 into epistemic modalities," even though in the case of the conditional only one of the subject-oriented modalities has been affected, obligation. The loss of the possibility for the conditional to cooccur with the notion of obligation strongly supports the hypothesis that the variation observed in Covarrubias reflects a process of semantic weakening of the forms used to express modality. This process may be observed throughout the history of Spanish in the following attested changes (cf. Marcos Marin 1980: 402^12): (a) the weakening of the pluperfect subjunctive form in -se (cantase 'that he had sung') from a meaning of past fact or non-fact in Medieval Spanish to its present meaning of [ ± possible] in the non-past; (b) the weakening of the Medieval Spanish pluperfect indicative form in -ra (eantara 'he had sung') to its primary present meaning of remote possibility as a form of the imperfect subjunctive; (c) in the variety of Spanish under study,-the weakening of -ria. In its early stages of development, the conditional, a Romance creation, existed side by side with the non-bound form Infinitive + haber (e. g., cantar habia,

564

Carmen

Silva-Corvalän

lit.: sing had to). It seems to me quite possible that the non-bound form continued to be used with modal meanings while the new form (cantaria) was at first only a marker of tense. At a later stage the bound form may have started to cooccur with subject-oriented modalities and subsequently (or concurrently) developed epistemic meanings. If Bybee and Pagliuca's (this volume) predictions are correct, by the time habia became fused as a bound morpheme to the infinitive of the main verb, however, its meaning may have been generalized enough to have lost some of its subject-oriented properties, specifically in the case of Covarrubias Spanish, and quite likely in most of Spain, the notion of obligation. What is happening in Covarrubias Spanish, or has happened if we are willing to disregard a small amount of variation, is, more or less simplified, what I have shown in table 2: Table 2. Modality shift in Covarrubias Spanish

Older Stage'

-bei

Fctmnnpr

nnceiKilitvll"

-ria -ra

-raj-se + past participle [zero possibility] Present stage:

-ba

[stronger possibility] [strong/weak obligation] -ria [weaker/remote possibility] -raj-se + past participle [zero possibility]

5. A tentative explanation. I have proposed that the changes discussed here reflect a process of semantic weakening, without questioning why semantic weakening may occur. This is not an easy question to answer, but at least two explanations appear to be possible. One has been proposed by Bybee and Pagliuca to account for the addition of an epistemic meaning. They suggest that the semantic weakening of modals is the consequence of the metaphorical uses of subject-oriented modalities. Metaphorical extension leads to the loss of certain specificities or restrictions involved in

Modality and semantic change

565

the meaning of the verb; this loss in turn results in the addition of an epistemic function. The increasingly frequent use of the epistemic function leads to the weakening of the specific notion of obligation. On the other hand, the issue that I am concerned with is different because it involves changes in the meaning of bound verbal morphemes. In this respect, I would like to propose that semantic weakening is the consequence of the combined effect of both internal and external factors. Specifically, I want to propose that something like a principle of distance plays an important role. According to this principle, if a language system has several closely related forms in the same syntactico-semantic sphere, the form which is farthest away from the speaker, in the sense that it refers to objects or events which are the farthest from him in his objective (e. g., actual distance) or subjective (e. g., possibility of actualization) world, will be lost. 11 I further propose that one of the factors which triggers this loss is interactional, i. e., sociolinguistic. Speech is egocentric (Givon 1976), i. e., speakers tend to speak about themselves and their immediate world, about events that have occurred or will occur in their lives, about the various ways in which they can manipulate their environment in order to bring about concrete results.12 The frequent use of forms which convey the meaning of stronger possibility and obligation empties out their modal meanings and adds to them weaker notions of possibility and obligation. 13 At this point, there is little need for the forms which are farther away from the speaker's immediate world. Since the system offers other forms with closely related meanings, the most peripheral form, which is used with a much lower frequency, tends to disappear. The consequence is the development of a paradigm which is maximally differentiated in form but not in meaning, since the more peripheral form now includes the meaning of the form lost. Schema 69 tries to represent a dynamic pattern of change, where 1 stands for the form which is closest to the speaker's objective or subjective world, and the numbers to the right represent the various forms ordered in an increasing degree of distance from the speaker. (69)

1 V

~

2 ^

3 ^

4 ^

V

5 > 0 form meaning

Schema 69 may be interpreted as follows: given a number of closely related forms in a syntactico-semantic sphere, if one of them

566

Carmen

Silva-Corvalän

is lost it will be the one which refers to objects, states, events, etc. which are farthest away from the speaker's immediate objective and subjective reality; the meaning of the form lost is incorporated into the form to the left. Outside the verbal paradigm, supporting evidence for the principle of distance may be offered by the recent changes in the paradigm of the Spanish demonstrative deictics este 'this', ese 'that', aquel 'that over there'. This system has been reduced to two in spoken Latin American Spanish and, predictably according to the principle proposed here, the form lost has been aquel. The development of complex pasts and futures in Romance, as discussed by Fleischman (1983: 183-214), could also be relevant evidence for the principle of distance. Fleischman observes that in both past and future systems the complex formation which began as a marker of present relevance later acquired an additional temporal meaning. Thus, "the complex tenses can now refer to situations completely detached from 'now', situations which at one time could be described using only the simplex forms" (Fleischman 1983: 201). The semantic change has been coupled with a steady increase of the frequency of occurrence of the complex forms, that is, the forms which primarily refer to the speaker's more immediate world. It is of course necessary to define explicitly what is meant by "closely related meaning," a question which I will answer only intuitively at this point. It seems to me that the linguist knows that two forms have closely related meanings when it becomes difficult to differentiate them and define their differences, that is to say, the forms may be interchangeable in some contexts, such that the linguist will have to state that the forms have become "neutralized." One may further wonder why subtle semantic differences may trigger change in only some varieties of a language. In other words, how do we account for the geographic distribution of a particular innovation? In order to answer this question, I would like to propose the concurrent effect of an external factor: language contact. Notice that the use of the conditional for subjunctive has been reported to be very frequent in the Spanish spoken by Basques 14 and Lavandera (1975) reports a strong correlation between conditional for subjunctive and Italian background in Buenos Aires. This sociogeographic distribution of the phenomenon led me to propose in an earlier study (Silva-Corvalän 1982: 87-96) that language contact was an external cause which contributed to the actuation of the change. Indeed, it is well-known that in situations of extended bilingualism

Modality

and semantic change

567

the two (or more) languages may converge grammatically (Gumperz — Wilson 1971) and/or that the structure of the second language may be modified as a result of incomplete learning due to limited exposure to all the registers, styles, and modes available to educated native speakers. Both factors (convergence and incomplete learning) may have played a role in Buenos Aires and in the Basque country, thus paving the way to and accelerating the spread of the innovation first in bilingual areas and later in adjacent monolingual areas. Both places appear to illustrate substratal influence, albeit of a different sociolinguistic kind. Basque and Spanish have been in contact for centuries, and in Basque speaking communities there is no conscious effort from whithin to suppress Basque in favor of Spanish. The situation may be described as one of relatively stable bilingualism, while in Argentina, where Italian and Spanish have come in contact very recently, mostly as a result of massive immigration from Italy at the beginning of the twentieth century, bilingualism appears to be a transient phenomenon (Whinnom 1971). Furthermore, while the subjunctive is very much alive in Italian (at least in the standard language), it is reported to be rare or non-existent in spoken Basque (Eys 1883; N'Diaye 1970). Thus, the form used in contexts where a past subjunctive would be expected in standard Basque is the past of the future, i. e., a form which corresponds to the Spanish conditional, as example 70 illustrates: 15 (70)

Parise-a Paris-to

jun-go go-future particle

ba-nitxa-ke if-aux-KE

gausa thing

asko many

neu-kez aux-KE

erosi-ko buy-fut. part.

'If I would go to Paris I would buy many things.' Constructions of the type 70 may have constituted the source of the substratal influence. In Italian, on the other hand, present and future counterfactual conditions have the same tense sequence as in standard Spanish: imperfect subjunctive in the protasis, conditional in the apodosis (see example 71). Future conditions, however, have the same verbal form, the future, in both clauses (see example 72), and this construction 16 might be the source of the possible substratal influence suggested here. (71)

Se fossimo (S) in ritardo prenderemmo (C) 1'autobus. 'If we were late we would take the bus.'

568

Carmen

(72)

Silva-Corvalän

Se saremo (fut.) in ritardo prenderemo (fut.) l'autobus. 'If we are late, well take the bus.'

6. Conclusion. To conclude, then, I have proposed that the existence of forms with closely related meanings may be a source of linguistic change, and that this change is governed by a principle of distance which has a sociolinguistic/interactional motivation. In the specific instances analyzed in this paper, the substitution of the conditional for the subjunctive and the encroachment of the imperfect indicative into the realm of epistemic and deontic modality, I have shown that the changes constitute a process of continuous semantic weaking, that the diffusion of the change has been gradual, i. e., it appears to have spread to the various contexts in a stepwise manner, and that it has been spurred by language contact. In addition, the results of this study bear on the question of typologies of language change. Malkiel (1978: 91), for instance, assumes a typology of internal processes of language change which includes, fundamentally, three categories: increase of economy, sharpening of clarity, enhancement of expressivity. It is clear that the changes discussed here do not sharpen clarity nor do they enhance expressivity, the two categories which seem to be more semantically motivated. On the other hand, the loss of -ra does result in an increase of economy of forms, but it does not increase semantic economy. This indicates that Malkiel's typology would have to be modified in order to accomodate changes which affect form and meaning in different ways or, at least, that increase of economy should not be part of the same typology of change as sharpening of clarity and enhancement of expressivity since they refer to different levels of analysis: form and meaning, respectively.

Notes * I would like to thank the people of Covarrubias, whose friendship and cooperation made this study possible. I would also like to thank B. Comrie, Y. Malkiel, and J. Wonder for valuable comments, and two of my graduate students, Jose Ignacio Hualde and Fernando Martinez-Gil, for their help in collecting some of the data for this paper. I am particularly indebted to Joan Bybee and Erica Garcia for their criticism of earlier versions of this paper and for the very helpful discussions we had about change, meaning, modality and explanation.

Modality and semantic change

569

1. Within the parenthesis, the first letter identifies the speaker, CV plus a number identifies the cassette tape. The data were collected in the Summers of 1981 and 1982 in Covarrubias, a small town of about 1000 people, located in the province of Burgos (Old Castile) in Spain. 2. In comparison with other Romance languages, it is interesting to note that the use of the conditional form in many of the contexts where it is considered nonstandard in Spanish is in fact the standard norm, such as in French and in Rumanian. Furthermore, though still nonstandard in French, the use of the conditional in the protasis of conditional sentences is reported to occur in certain varieties of spoken French. Outside the Romance languages family, it appears that English, at least American English, is moving in a similar direction in conditional clauses, as one frequently hears examples like the following: If they would tell him he would do it; If they would've told him he would've done it. Is this a case of "linguistic preference" (Venneman 1982)? The substitution of the conditional for the subjunctive should certainly occupy a prominent place in a "catalog of possible language change." 3. The figures presented in table 1 are based on the occurrences of the variants in the samples of speech from all the speakers recorded: 14 women and 12 men, 8 under 30 years of age, and 18 over 30. Other possible syntactic contexts are discussed later in the paper. 4. Grammars (e. g., Real Academia Espanola, 1973; Marcos Marin, 1980) propose the existence of the following verbal forms for present day Spanish: Non-finite forms: Infinitive: Present participle: Past participle:

Simple hablar hablando hablado

'to speak' 'speaking' 'spoken'

Infinitive: Present participle:

Compound haber hablado habiendo hablado

'to have spoken' 'having spoken'

Finite forms: Indicative Mood Present: Preterit: Future: Conditional: Pres. perfect: Past perfect: Pluperfect: Fut. perfect: Cond, perfect:

habla hablo hablarä hablaria ha hablado hubo hablado [almost obsolete] habia hablado habra hablado habria hablado

'speaks' 'spoke' 'will speak' 'would speak' 'have spoken' 'had spoken'

Subjunctive Mood Present: Imperfect: Future: Pres. perfect: Pluperfect: Future perfect:

hable hablara/hablase hablare [obsolete] haya hablado hubierajhubiese hablado hubiere hablado [obsolete]

Imperative Mood Present:

habla

'had spoken' 'will have spoken' 'would have spoken'

'speak'

570

Carmen Silva-Corvalän

5. I have written the imperfect indicative morpheme -ba in parentheses in the Standard Spanish column because normative grammars state that it is "rare in the literary language though frequent in conversation" (Gili Gaya 1971: 162, my translation). 6. This analysis has in fact been proposed by Lavandera (1975), who hypothesizes that the use of the conditional in the protasis of conditional clauses was motivated by the need to refer unambiguously to [ +possible] situations once the future subjunctive had been lost. Unfortunately, my data are not large enough to allow a reliable statistical analysis to investigate the correlation between [ ± possible] and the choice of -ria or -ra. 7. The quote in example 17 refers to how the speaker's daughter was dressed after she died. 8. The Real Academia Espanola (1973) associates different degrees of politeness with imperfect and conditional forms, derived from the fact that these two forms are said to convey stronger and weaker assertion respectively; the imperfect subjunctive and the conditional are considered to be interchangeable, on the other hand: Queria (I) hablar contigo Ί wanted to talk to you', Querria (C)/ quisiera (S) hablar contigo Ί would like to talk to you'. 9. Bybee and Pagliuca divide modality notions into two groups: subject-oriented modalities, which include the notions of ability, obligation, desire and intention, and epistemic modalities, which include the notions of possibility, probability and prediction or future. 10. Obviously, it is the verb itself the one that expresses the meaning of'obligation'. The bound morpheme {-ba, -ria, -ra) simply modifies the strength of the obligation. 11. It is reassuring to note that this principle makes explicit an idea implied in Bolinger (1956). I read Bolinger after having written this article and was glad to find out that his notes on the alternation between imperfect subjunctive forms, imperfect indicative and conditional confirm most of my own analysis and conclusions. With respect to the loss of -se forms, for example, Bolinger (1956: 346) notes that 'In the 20th-century Peninsular dramatic works studied by Francis B. Lemon, -ra forms outnumbered -se forms by 172 to 26, indicating a preference for -ra in everyday speech which is just what one would expect if -ra signals meanings that are warmer and more immediate.' 12. As a matter of fact, one often hears among sociolinguists interested in syntax and semantics how difficult it is to obtain samples of speech containing counterfactual conditional sentences or other hypothetical statements/speculations. This appears to be an indication of the speakers' preference to discuss the world in terms of more concrete possibilities. 13. This type of change is well attested in the development of formulaic expressions, greetings and forms of address (e. g., adios 'goodbye' from a Dios seas 'may God be with you', usted 'you-formal' from vuestra merced 'your lordship'). 14. This statement is based both on my own observation of Basque Spanish and on reports from numerous individuals familiar with this variety of Spanish. 15. Example 70 is taken from Hualde (1983). It was produced by a Basque speaker from Ondarroa, Vizcaya. 16. This is based on the assumption that the Italian variety spoken by immigrants was like standard Italian in this respect.

Modality and semantic change

571

References Alarcos Llorach, Emilio 1970 Estudios de gramätica funcional del espanol (Madrid: Gredos). Bello, Andres — Cuervo, R. 1977 Gramätica de la lengua castellana (Buenos Aires: Sopena). Bolinger, Dwight 1956 "Subjunctive -ra and -se: free variation?", Hispania 39: 345-349. Bybee, Joan L. — Pagliuca, William 1985 "Cross-linguistic comparison and the development of grammatical meaning.", this volume, 59 — 83. Cartagena, Nelson 1976-77 "Estructura y funciön de los tiempos del modo indicativo en el sistema verbal del espanol", R. L. A. 14-15: 5-44. Eys, W. J. 1883 Outline of Basque grammar (London: Trubner). Fleischman, Suzanne 1983 "From pragmatics to grammar", Lingua 60: 183-214. Gili Gaya, Samuel 1979 Curso superior de sintaxis espahola (Barcelona: Biblograf). Givon, Talmy 1976 "Topic, pronoun, and grammatical agreement", Subject and topic, edited by Charles N. Li. (New York: Academic Press), 149-188. Gumperz, John J. — Wilson, Robert 1971 "Convergence and creolization: a case from the Indo-Aryan/Dravidian border", Pidginization and creolization of languages, edited by Dell Hymes. (Cambridge: Cambridge University Press), 151-167. Hualde, Jose Ignacio 1983 "Expresion de la condition en vasco y condicional por subjuntivo en castellano." (MS, University of Southern California). Keniston, Hayward 1937 The syntax of Castilian prose: the sixteenth century (Chicago: University of Chicago Press). Lavandera, Beatriz 1975 Linguistic structure and sociolinguistic conditioning in the use of verbal endings in si-clauses (Ph. D. dissertation, University of Pennsylvania) (Ann Arbor: University Microfilms). Lyons, John 1977 Semantics, Vols. 1 & 2 (Cambridge: Cambridge University Press). Malkiel, Yakov 1978 "Hypercharacterization of pronominal gender in Romance", Language, meaning, and style: essays in memory of Stephen Ullmann, edited by Τ. E. Hope et al. (Leeds: Leeds University Press), 91-107. Marcos Marin, Francisco 1980 Curso de gramätica espanola (Madrid: Cincel-Kapelusz). N'Diaye, Genevieve 1970 Structure de Dialecte Basque de Maya (The Hague: Mouton). Real Academia Espanola 1973 Esbozo de una nueva gramätica de la lengua espanola (Madrid: EspasaCalpe). Silva-Corvalan, Carmen 1982 "Conditional for subjunctive in Old Castile", Proceedings of the Eighth Annual Meeting of the Berkeley Linguistics Society, 87-96.

572

Carmen Silva-Corvalän

Vennemann, Theo 1982 "Theories of linguistic preferences as a basis for linguistic explanations." Paper presented at the Conference Causality in Language Change, U. C. L. Α., May 1982. Whinnom, Keith 1971 "Linguistic hybridization and the 'special case' of pidgins and Creoles", Pidginization and creolization of languages, edited by Dell Hymes, 91-115. (Cambridge: Cambridge University Press).

JÜRGEN STRAUSS

The lexicological analysis of older stages of languages 1. Introduction Dictionary definitions dealing with older stages of languages, particularly Old English, are often unsatisfactory in many respects. Therefore for a long time now there have been attempts to analyse selected areas of vocabulary more precisely. Many publications on Old and Middle English word fields have been written in Germany and elsewhere, above all in the context of the theoretical discussion about the nature of semantic fields. Although the methodological and semantic premises are often basically different, a meaningful evaluation and summary of the findings — such as with the framework of a revision of the wellknown Old English dictionaries (Grein; Bosworth-Toller) — may seem fruitful. Moreover, the discussion of semantic theories in the last few years has offered so many interesting and promising new possibilities of investigating meanings that renewed considerations about the desirability of the revision of dictionaries seem appropriate. Since in addition linguistic data processing nowadays can facilitate many processes of searching and sorting in lexicology and lexicography, the effort involved in such analyses seems to have become less. In the following some examples will be given to demonstrate such possibilities.

2. Meaning, metaphors, and fields The theoretical basis for our considerations is provided by Wittgenstein (1960) and others at the level of language philosophy, and the semantic insights above all by Leisi (1961), Lyons (1963, 1968, 1977), Fillmore (1977) et al. on the linguistic level. They all take as their assumption that the meaning of the word — in the

574

Jürgen Strauss

words of Wittgenstein (1960: 311) — 'is its use in language'. The Swiss linguist Ernst Leisi speaks in this context of the semantic 'conditions which determine the use of the particular word' which must be known to a user of the language in order for him to use the word correctly (Leisi 1961: 23). A knowledge of the 'conditions of usage' is gained by the individual speaker through experience: if he wants to use a particular word in a new situation he compares the present conditions of usage with those that he has already experienced, and then decides whether the conditions are sufficiently similar to justify the particular word. It is probable that the speaker has in his mind a 'prototypical scene' as advanced by Fillmore (1977). Applied to Labov's (1973) famous 'cup'-experiment, this means that a speaker describes a drinking vessel as cup whenever he recognizes the conditions of usage of 'proper cuphood' (Fillmore 1977: 56). If, on the other hand, the depicted drinking vessel differs from his idealized model in conditions essential for the observer, i. e., prototypical conditions, he looks for another word {bowl, glass), or rejects the object as 'unknown'. The prototypical scenes propounded by Fillmore can be compared — obviously without Fillmore's theoretical foundation — with long familiar concepts in semantics such as 'lexical meaning' (Stern 1931: 68 ff.; Schmidt 1961:28 et passim), 'usual meaning' ('usuelle Bedeutung'), 'core meaning', 'central meaning', 'eigentliche Bedeutung' (basic meaning), etc. The attractiveness of Fillmore's conception lies purely in the fact that the explanatory strength and the simplicity of his model are directly convincing. In addition, the definition of meaning by means of typical conditions of usage or scenes seems to be applicable, too, if we deal with figurative meanings like metonymies, metaphors, kenningar, etc. As has been depicted by Lakoff and others (Lakoff — Johnson 1980; most recently Lakoff — Kövecses 1983), metaphorical paraphrases also have typical conceptual aspects or scenes such as 'love is war', 'love is a journey', 'love is madness', etc., which can be grouped into language-specific systems. In Old English poetry, in particular, the art of finding metaphorical paraphrases which stress certain prototypical scenes is perhaps the most important characteristic of the genre. A sword, for example, is thought of as 'the helper in the battle' (hilde-frofor, mcegenfultum), 'the battle-light' (beado-leoma, hilde-leoma), 'the leavings of the hammers or the files' (homera laf, fela laf).

The lexicological

analysis of older stages of languages

575

Typical metaphorical scenes for the sea are: 'the sea is the home of a sea animal' (fisczs epel, hweeles epel, etc.), 'the sea is the road of a sea animal' (swan-rad, hron-rad, hwcelweg, etc.), 'the sea is the bath of a sea animal' (ganotes bceö,fisces bceö). Characteristic scenes for the lord are among others: 'the lord is the delight/joy of the retainers' (cepelinga wynn, duguöa wyn, etc.), 'the lord is the shelter/protection of the retainers' (cepelinga hleo, heriga helm, etc.). A semantic conception which orientates itself by typical conditions of use or prototypical scenes also has the advantage that in addition to its immediate powers of explanation, it helps to answer the question of the delimitation of sections of vocabulary (fields). The problem of delimitation of semantic fields has been much discussed and is generally considered as an insoluble problem involving, for instance, metaphors etc. Above all, componential analysis has attempted with very painstaking lines of inquiry and 'components' nevertheless to divide up such border areas. However, against this kind of field analysis can be raised the same objections as Fillmore brings against the 'checklist-theory of meanings'. The core of a semantic field is comparable with an idealized cognitive model which we can describe as the prototypical centres of the field. About this core are grouped in a continuous succession words and word groups (metonymies, metaphors, kenningar, etc.) moving outwards in circles of increasing vagueness. The more uncertain the membership of a word to a particular field becomes — for example, the more 'daring' a metaphor becomes — the more the language user will ask himself whether the relevant linguistic unit still corresponds to his idealized cognitive model of the core, whether the unit corresponds more to another prototypical field core, or whether a decision is impossible (possibly because the metaphor is too 'bold' and the listener is no longer in a position to make a decision). Further investigations in this direction will perhaps show that such a definition of fields is better able to answer the question of the 'synonymy' of members of the same field. Diachronic changes within the fields and between related fields can probably be explained as changes of the prototypical centres of the fields: thus the social changes in the eleventh century changed the relationship between the rulers and the subjects to such an extent that the word dryhten with its underlying conception (idea of personal relationship between the chief and his retainers: Germ. 'Gefolgschaftsgedanke')

576

Jürgen Strauss

was replaced by others since the principle of 'Gefolgschaft' had become meaningless as a result of foreign domination.

3. The analysis of the conditions of usage of Old English words in poetry For the lexicologist concerned with the older stages of a language the problem arises of how to make deductions from the existing occurrences of words in the extant texts about the prototypical conditions of usage, in other words, how to move from the tokens to the (proto)types without the possibility of introspection or asking informants. This can only happen if the actually occurring uses are carefully analysed and checked for the underlying conditions of usage. Since we are dealing with restricted corpora, if possible all the instances of a word should be investigated in this way.1 This is the only way of building up a maximum of different conditions of usage which can subsequently be reduced to the essential or 'prototypical' conditions. Today, there is more point than ever in such quantitative analyses since the large amount of information about the conditions of usage can be processed much more economically with the help of data processing machines than was possible in the days of Grein, the author of the Angelsächsischer Sprachschatz. For example, Grein restricts himself in his lexicography of the Old English poetic language to listing the lemmata with, in each case, only a few — mostly Latin — glosses, and then cites a selection of occurrences of a lexical item in the respective contexts from which he gains his intuitive insights. The lexicographic compilation of the Old English word material in the most extensive dictionary by Bosworth-Toller is constructed similarly. Such semantic intuitions, however, valuable they may be as points of orientation, are basically only a 'scaffolding' (Lyons 1963: 23) which must be completed by references to the various prototypical conditions of usage. These prototypical conditions can only be worked out by careful analyses on the referential level and made of use for the lexicological determination of a word. This simply means that all the occurrences of a word in the extant texts must be gathered and investigated. Such a semantic investigation can orientate itself by prototypical scenes which, for example, in the case of the Old English words for 'sword' can be:

The lexicological

analysis of older stages of languages

577

— the sword is presented as a (valuable) gift — the sword is depicted in a concrete battle scene — the sword fails in its duty, etc. If we take the Old English words for 'blood' (cf. Kühlwein 1968), typical 'scenes' might be the following: — is it human blood — is the blood clotted — is the blood flowing in the body of a human being or an animal — is the blood pouring out — is it flowing onto the ground — is a battle scene involved — is it the blood of Christ, etc. Yet another field, that of the Old English words for 'lord and master', must orientate itself by prototypical scenes such as: — how does the lord behave towards his subjects: — punishing — distributing gifts — protecting, etc. — how does the subject behave towards his lord: — serving — praising — rebelling, etc. — in what situation is the word used: — (secular) fight — the idea of personal relationship between a chief and his retainers ('Gefolgschaftsgedanke') is implied (such as Jesus as a retainers' chief) — God as the avenging deity, etc. These catalogues are formed operationally, that is to say they must be constantly modified in the course of an investigation, certain questions are deleted, others are added. At any rate, in all cases definite conditions of usage which can be regarded as typical crystallize out relatively quickly. A lexicographical description which includes such typical conditions of usage can give very much more information than the standard dictionaries of Old English. So we

578

Jürgen

Strauss

find in Hall, Grein, or Bosworth-Toller for the Old English words blod, dreor, swat, and heolfor only the glosses 'blood, gore', or the corresponding Latin terms sanguis and cruor. This glossing can be complemented by essential conditions of usage. For example, heolfor designates particularly clotted blood, whereas swat is hardly ever used for blood of this kind. Heolfor, on the other hand, is almost never used for human blood or blood flowing in the body. If the blood is 'pouring out', dreor or swat is used, swat particularly when the condition 'fight' is involved. Also in the case of the elements of the word field for the 'lord and master' striking uses for 'prototypical conditions' of usage can be noted. An essential prototypical scene for dryhten is the situation of the 'Gefolgschaft': if the lord is represented as 'the personal leader of a certain individual', as a 'distributor of gifts or honours' or 'helping, saving, protecting', the word dryhten is used most often in relation to the other words of this field. In the religious sphere the idea of salvation, which can also be understood as a part of the function of the idea of the 'Gefolgschaft' (i. e., the lord supporting his retainers), is prototypical of dryhten: again in comparison to the other members of the field, dryhten is most often used for the situations 'rising again/redeeming' or 'God/ Christ at the Day of Judgment'. If, on the other hand, a situation is represented in which the lord is depicted as 'punishing' or 'commanding' or 'having glory and power', the word cyning is the one which corresponded most closely to this prototypical conception, again in comparison to all other words denoting 'lord or master'. Other typical uses also underline expressly the rex-function which the word possessed. A fact that has been largely overlooked so far is that cyning also seemed particularly appropriate as a word for the lord in battle situations or as a leader of a host of warriors. The semantically difficult word peoden, used only in poetry, is translated in the dictionaries by 'king, chief of a theod, lord, dominus'. The analysis of the conditions of usage shows that the idea of 'Gefolgschaft' is implied, as with dryhten, but that the close emotional link between the lord and his subject, as with dryhten, is missing. Instead, peoden is used above all in 'solemn' situations whereby the marked decorative function of this word becomes clear: surely peoden already had in poetry a rigid semantic structure. This is born out by the frequent occurrences in the context of a variation (52.8% of all occurrences in the ASPR are in the context of a

The lexicological

analysis of older stages of languages

579

variation!), its frequent use in the vocative (which also applies to frea), and the restriction of peoden to very few modifying adjectives (mainly mare). Another typical use of peoden is when the subject in the given situation already holds a high position in the aristocracy or similar — this, too, is possibly an indication of the decorative function of the word. If we take the prototypical scenes as a departure point, interesting oppositions result. Whereas the scene 'praising, in general' is associated with cyning, the scene 'praising for a specific reason (or giving thanks) 1 , on the other hand, is associated with dryhten, which implies the close relationship between the lord and his subject once again. When in religious texts the people or an individual rebels against God, the word most frequently chosen is god or wealdend, whereas if the subjects are represented as 'begging for help', the Old English poet chose dryhten, and if a subject is 'mourning for his lord', then the appropriate word is hlaford which also involves the close personal relationship between lord and subject. It has already become clear that other semantic aspects of the Old English words resulting from specific peculiarities of use can be shown. So, for example, the stylistic contraints imposed upon the words by the stylistic means of variation can be taken into account (cf.peoden). Also references to morphological-semantic connections such as arise in Old English poetry because of the necessity for alliteration can be integrated in a lexicographic revision of the Old English poetic vocabulary by basing the description upon usage (Strauß 1980). It must be emphasized once again that the quantitative analysis must be accompanied by the individual analysis of typical examples of use which, as far as possible, take into account intuition. The quantitative analysis gives us information about the 'degree of centrality' (Lyons 1968: 479) of certain norms of usage and is objectively 'correct' as a method. It is nevertheless clear that the extant texts convey only a section of the applications of a word. The supplementary interpretation must definitely also be provided — with all the possibilities of error which this involves — in order to give an overall picture of the supposed meaning; this is all the more necessary, the fewer occurrences we find in the Old English texts.

580

Jürgen Strauss

4. Steps in the analysis

The lexicological analysis of older stages of languages

581

Notes 1. Of course, such an individual semantic analysis of all the occurrences of a word assumes a thorough acquaintance with the old documents. How difficult the correct emendation of a manuscript reading can be has been convincingly described in the doctoral thesis of Westphalen (1967). Other examples of the difficulties involved in the semantic description of particular instances are the contributions by Klegraf (1971) on the Beowulf-crwx Bwf 769 ealuscerwen, or the question of the assignation of elements in composita, which is very often not clear from the Old English manuscripts: the combining of lexical items to form compounds was often undertaken at some stage by an editor and then handed down uncritically by generations of editors (cf. Strauß 1982: vii). Of course, differences of opinion about the reading of particular instances will always remain, and a quantitative analysis of this kind must live with the fact that the statistical objectivity of the results is valid only insofar as the results reflect the investigator's point of view on the readings.

References Fillmore, Charles J. 1977 "Scenes-and-frames semantics", Linguistic structures processing, edited by A. Zampolli (Amsterdam: North Holland), 55-81. Klegraf, Josef 1971 "Beowulf 769: ealuscerwen.", Archiv 208: 108-112. Kühlwein, Wolfgang 1968 Modell einer operationalen lexikologischen Analyse: Altenglisch 'Blut' (Heidelberg: Winter). Labov, William 1973 "The boundaries of words and their meanings", New Ways of Analyzing Variation in English, edited by C.-J. N. Bailey and R. Shuy (Washington, D. C.: Georgetown University Press), 340-373. Lakoff, George — Johnson, Mark 1980 Metaphors we live by (Chicago: University Press). Lakoff, George — Kövecses, Ζ. 1983 The cognitive model of Anger inherent in American English (Trier: L. A. U. T.). Leisi, Ernst 1961 Der Wortinhalt. Seine Struktur im Deutschen und Englischen (Heidelberg: Quelle und Meyer). Lyons, John 1963 Structural semantics. An analysis of part of the vocabulary of Plato (Oxford: Blackwell). 1968 Introduction to theoretical linguistics (Cambridge: University Press). 1977 Semantics (Cambridge: University Press). Stern, Gustaf 1931 Meaning and change of meaning. With special reference to the English language. (Göteborg: Elanders). Strauss, Jürgen W. 1974 Eine Komponentenanalyse im verbal- und situationskontextuellen Bereich (Heidelberg: Winter).

582 1980

Jürgen

Strauss

"Compounding in Old English poetry", Folia Linguistica Historica 1: 305-316. 1982 Beowulf. Konkordanz und Glossar, Beowulf und die kleineren Denkmäler der altenglischen Heldensage Waldere und Finnsburg, Teil 3, edited by G. Nickel (Heidelberg: Winter). Westphalen, Tilman 1967 Beowulf 3150-55: Textkritik und Editionsgeschichte (München: Fink). Wittgenstein, Ludwig 1960 Schriften I. Philosophische Untersuchungen (Frankfurt/M.: Suhrkamp).

WERNER WINTER

'Left' or 'right'? Some thirty years ago, we lived in faculty quarters of the University of Kansas. We were housed in converted Navy barracks, in an apartment on the second floor. Little by little, we added to the furniture we had been able to rent with the apartment. One day, movers brought a huge green sofa which we had bought secondhand. Two people had to carry it, the one in front going up the stairs backwards, the one at the lower end, facing upwards. He was the one who gave instructions to his partner: Ά little further east — a bit further west', until they had reached the top of the stairs and could turn into the apartment. For newcomers from Europe, still unaccustomed to the checkerboard layout of a mid-American town, this way of giving directions came as a surprise; but on second thought, it was only too obvious that under the circumstances this was a much better solution than what would have resulted from recourse to what we had expected, namely, instructions referring to 'right' and 'left' — for, whose right and whose left would have been meant if the man below had told the other mover where to go in these terms? 'East' and 'west' were the same for both men; 'right' and 'left' differed in reference to the actual situation depending on whether the person giving instructions was speaking in his own terms or in those of his partner facing him. The problem encountered — and solved — by our furniture movers was of course not a new one. Other answers had been found before. Typically, a convention was established that a point of reference should remain stable. Thus, the left-hand side of a boat was always determined in relation to its bow and called larboard or port, with starboard is its counterpart. 'Left' and 'right' in a house of parliament have been fixed as seen from the Chair. Under everyday conditions, however, in particular when no hierarchy immediately suggests itself, we time and again feel compelled to ask: 'Do you mean on your right or on mine?', and only after clarification has been obtained will it be possible to take proper action. I would like to discuss here a problem of text philology from two languages with badly damaged, and not very voluminous, corpora

584

Werner

Winter

of texts, viz., Tocharian A and Β. I will give a complete inventory of all published passages containing the forms in question, and will take stock of the various arguments that can and must be adduced to answer the question which of two competing terms denotes 'left' and which one 'right'. As long as we just look at handbooks, there seems to be no problem at all. In the glossary included in Thomas — Krause (1964), we find: Α päci 'rechts', sälyi 'links', Β saiwai 'rechts', swälyai 'links' without so much as a question mark, which did appear in Sieg — Siegling — Schulze (1931:17 η. 1). Later on in the same book (p. 69), the authors seem to remain noncommittal to a discriminating reader, suggesting, yet at the same time not explicitly affirming, that Α päci meant 'right' and Α sälyi, 'left'. Somewhat later, Sieg (1944: 16) no longer used qualifying terms concerning these translations, probably under the influence of the argumentation by Konow (1942: 207-208); see Sieg (1944: 28 η. 10). Van Windekens (1976), who much earlier (1938-1939: 182-184; 1941: 87, 134) had proposed etymologies for Α päci and Β swälyai, Α sälyi on the strength of the hints given by Sieg — Siegling — Schulze, echoes his earlier views as well as those of Thomas — Krause (1964). Thus, all seems to be clear to everyone, except that the etymologies proposed by Van Windekens (1941; 1976) leave much to be desired. In itself, this would not rule out correctness of the meaning identifications; after all, in all languages we incur words without good etymologies. Still, one begins to wonder, and one wants to find out what exactly the information is that can be derived from the texts. We can proceed from the assumption that one pair of the set A sälyi, Β swälyai: Α päci, Β saiwai means 'left', the other one, 'right'. So as not to influence the decision, I shall 'translate' the first pair of forms by 'X', the second, by Ύ ' . Our data are: Α sälyi: A 12 b 4 kliso päccäs possäsä to lap sälyim kolyeyac 'resting on the Y side, having put its head on the X xxx' (said of a lion; A kolye is not likely to mean 'tail' as proposed by Sieg 1944: 16),

A 132 b 4 // tmäs sälyim tsar-yo' // 'then with the X hand ...'; A 146 a 5 // kuli täs sälyäs po((ssä) )sä mäcri käts se(m yämuräs)

'Left'or

'right?

585

— ' i f . . . was a woman, having made, on the X side, ... mother's belly ... shelter'; A 150 b 4 // iycäk mäskatär sälyäs po(ssäsä) — 'is xxx; on the X side...'; A 158 a 1 // näktennäy sälyim peyä // '... at the X foot of the goddess ...'; A 199 b 2 — m crank(s)äm klop säl. / / ' . . . told(?) them(?); sorrow A 342 a 5 tmäs (amäsäh) lantas ke kälpos karesa saseyus sälyis po'(ssäsä) II 'then (the ministers(?)), with the king's leave(?), leaning on their swords, (standing(?)) on the X side(?) ...' (A possäh 'side' is clearly a feminine in A 150 b 4; only with reluctance will one consider changing sälyis into a regular form of the masculine accusative plural, viz., sälyi (nä)s — but it should not be overlooked that po'(ssäsä), first proposed by Couvreur 1955-1956: 78, is merely conjectural); A 378.2 // sälyäs päcyäs lalänkuntim s(a)m täkenc: tsem säkwis kusram tmassäl täkmäm tili sälyi pä'(ci) // 'hanging down to X and Y there will be a pair of braids(?) of blue hair; comparable to that your X and Y ...'. Β swälyai: Β 74 b 4 II (ta)karske kaunäkte laktse äntse-ne tan swälyai Ikätär yaitu yaltse swaficain-tsa: 'clear and shining, the Sun is seen on your X shoulder, adorned with a thousand rays'; Β 74 b 5: swälyai märkwat-sa okpokai Visn(u) 'on your X thigh (is) the eight-armed Visnu'; Β 371 a. 2 II su tüyäkne-sa swälyai paiyyehhe mok(ocintse äntene) 11 '... thus (at the tip) of the X big toe'; Β 567 a 1. hake lämat swälyai älyi-ne Ime swä' // 'now you will sit, in the X palm the awareness ...'; Β 567 a 2 // tsaksentär swälyai älyi-ne naumyesse ksätre l(k)äs(ty 11 '(they) shine; in the X palm you see a jewelled umbrella ...'; Β 580 b 2 swälyai (paijyyenne mokocintse änte-ne tuciyai tanäkkai Ikässäm 'at the tip of the X big toe he sees a yellow mark'. Α päd: A 12 b 4 (see above under Α sälyi); A 296 b 2 II (smi)mäm akmal-yo bodhisattu päcim tsar-yo n(ä)ktes kunti tsitoräs // 'with a smiling face, the Bodhisattva, having touched the god's pot with his Y hand, ...';

586

Werner

Winter

A 3 1 4 a 7 / / ram päcim tsar to ses cincär: 'he had put his lovely Y hand into ...'; A 320 a 3 //päccäspossäsä yaiwu (swä)ncen-yo: '... having entered on the Y side ... with rays'; A 328 a 2 // päcim pe-yo sik toräs tä // 'having made a step with the Y foot ...'; A 358 a 2-3: pämcyä(s käjlytärci Vajrapä(n)i — (sä)lyäs pracar Änant sämam wsokone-yo Ikehcy akmlac: 'to your Y Vajrapäni is Standing, to your X your brother Änanda, the monk; they are looking at you with joy'; A 378.2 (see above under Α sälyi); A 379 a 4 //' päcim tsar-yo kakärnäm: 'with the Y hand he(?) hit him(?)\ Β saiwai: Β 56 a 2 // ssäm saiwai no so j j '... while on the Y ...'; Β 74 b 4: saiwai äntse-sa yaitu — '// 'near the Y shoulder adorned ... , Β 74 b 5 saiwai-sa no Mahisvare märkwarc-tsa tan kauursapkai 'to the Y, however, on your thigh, (there is) the bull-armed Mahesvara (or: Mahlsvara)'; Β 74 b 6 // onkolma tan Ikätär saiwai ckäckai-ne: 'a she-elephant is seen on your Y shank'; Β 580 b 3 arkwaniia tanäkko saiwai (paiy)y(e'nne mokocintse änte-ne) // 'a white spot at (the tip of the) Y (big toe) ...'. Α päci can be considered an adjective based on a noun Α pät* which can be extracted from an extension by a prefix Α ä- (umlauted to ä- from a-* before -ä- of the following syllable before the latter vowel was reduced to -a- by syncope), viz., Α äpat 'on the Y side'. This word is attested in A 24 a 6 Brähadyutis kapsahi äpat twantam yatsi osät 'the body of Brhaddyuti began, on the Y side, to pay respect'; A 213 a 3 som-som yokasi spartu tatämsu äpat sä // 'each lock of hair, having been brought forth separately, on the Y side ...'. What looks like a case form of Α äpat, viz., Α äptä (another instrumental/perlative of a prefixed form is found in Α opärkä 'in the east'), occurs in contexts which do not contribute immediately to endeavors to determine the precise meaning of Α äpat and A päci, but can at best be subjected to further interpretation once this

'Left' or 'right'? 587 meaning has been established. We will return to some of these passages later on. Α äpat in A 24 a 6 at first glance seems to be of crucial importance for the question of what Α pät* and Α päci can be said to mean. Reference is made, as clearly seen by, e. g., Sieg (1944: 28 η. 10), to the practice common to Hindus and Buddhists of having the person or object to be honored on one's right side. The action involved is called pradaksinam kr-, and it was therefore only natural to believe that Α äpat and päci should be equated with Skt. daksina- 'right'. However, a seemingly minor point should not be overlooked: Sieg (1944: 28) translates Α äpat by 'nach rechts'. Closer inspection of forms with a prefix Β e-, A a- (and its umlauted derivatives) shows that a directional interpretation is called for only when an accompanying verb implies directionality; otherwise, the forms seem to express adjacency and thus something close to the instrumental in its adessive function. One may list here: A oklop 'in danger', oseni 'at night'; Β amäskai 'with difficulty', apläc 'in conversation', epinkte 'between' (cf. Winter 1983: 321-322), eraitwe 'in connection with', esatkai 'in excess, very', etsuwai 'in combination with', eweta 'in conflict with', aiksnar 'around', orkäntai 'back and forth', ossale (osle) 'in the evening'. The verb A yatsi 'make' in A 24 a 6 cannot be said to imply directionality; on the other hand, the phrase A twantam yatsi 'to pay homage' would leave this interpretation at least a possibility. As respect is expressed by having the person honored toward one's right, it is tempting to accept it. Still, is there a compelling need for such a decision, or is it possible to retain an adessive reading for Α äpafl It does indeed seem so. The same act can be described equally adequately by saying that the person honoring another one does so by taking a place to this person's left. If so, we are free to align A äpat with the prefixed items interpreted as adessives, changing, however, the proposed meaning of Α äpat radically: There is nothing wrong, as far as the content is concerned, with rendering A 24 a 6 by 'Brhaddyuti's body, being placed on the left side (of the Buddha), began to show his respect'. Clearly, this is a possibility, and apparently not a weak one. But we would like to have further confirmation for our assumption. For the moment, I want to disregard etymology, but limit myself to an inspection of the texts. In spite of their fragmentary state of preservation, a few observations can be made.

588

Werner

Winter

First, there is the order in which Α sälyiand päci and Β swälyai and saiwai occur. Let us call Α sälyi, Β swälyai 'X' again and A päci, Β saiwai, Ύ \ The order 'X'-'Y' then is found in A 378.2 (twice), Β 74 b 4, 74 b 5, 74 b 5-6, 580 b 2-3; all these passages show close internal parallelism. The reverse order Ύ'-'Χ' occurs in A 12 b 4 and 358 a 2-3; of these, only the latter passage involves a parallelism. In Sanskrit tradition, the order of 'left' and 'right' appear to be fixed; witness the presence of a compound daksinottara- 'right (and) left', but not of one with the opposite sequence, °uttaradaksina- 'left (and) right'. It is perhaps simplest to associate the preferred order with the correlation that Indie tradition established between 'right' and 'left' and cardinal points. First among the directions of the compass was the point of the sunrise, the east; the natural direction to follow from there was the course of the sun. Facing east, absolute 'south' and person-related 'right' referred to the same reality; the same applied, as long as the basic orientation was retained, to 'north' and 'left'. To go from right to left in a natural way was to move clockwise; even in our western tradition, the same pattern is observed to this day — to move withershins, that is, counterclockwise, is considered unlucky by some. The ratio six-to-one in favor of 'right' before 'left' in our texts reflects the natural order of things, which some of us tend to call unmarked; the one instance of 'left' before 'right' requires an explanation. Two of the instances of Α päci at first sight seem to present difficulties when translated by 'left'. In A 296 b 2 we would find that the Bodhisattva touches the god's pot with his left hand; in A 379 a 2 someone seems to hit someone else with his left hand. The second case may resolve itself quite easily — the hostile act of hitting may very well be performed with the hand reserved for activities that are not 'right'. In the first instance, we would, however, consider it natural that the touching should be done with the right hand. Consider, though, Β 85 b 4-5 tu-mem walo seme sarsa U(tta)'rem mneuskem e(nku wace sar-sa äyo)rs(s)e w(a)r enku 'thereupon the king, having gotten hold with one hand of Prince Uttara, with the other, of the gift water'; here, 'the second hand' can hardly be anything but the left hand, to which the less important action is entrusted, and this action involves precisely the handling of a vessel. Therefore, A 296 b 2 offers no conclusive argument against an identification of Α päci as 'left'.

•Left' or -right'?

589

A 314 a 7, not discussed so far, may be used in support of the equation Α päci = 'left', if indeed the two cola (seven plus four syllables) in the beginning of the line are to be reconstructed as // (sälyim tsa)ram päcim tsar to ses cincär. 'he had put his lovely left hand into his right one', with 'right' given the naturally preferred first position. It remains to have another look at A 358 a 2-3. Here, if Α päci is correctly identified as 'left', we not only find 'right' mentioned after 'left', but Vajrapäni is placed to the left and Änanda to the right, and if Vajrapäni is a boddhisattva, one would expect to find him in the place of honor, that is, to the right of the Buddha. (For alternative identifications of Vajrapäni see Edgerton 1953: 467.) The problem might be resolved by taking into account the nature of the text A 358: it is a poem addressing the Buddha; the speaker sees before him the Buddha surrounded by a host of supranatural beings; he describes what he sees, and so for him what is 'right' for the Buddha of course is 'left', and vice-versa — yet he makes reference first to the one honored by the Buddha, even if this upsets the speaker's 'natural' order. In a way then, the relative, and not absolute, nature of the terms 'right' and 'left' asserts itself, with the speaker not being able to cope with the problem quite as successfully as the furniture movers in Kansas. Our findings so far may be summed up as follows: Internal evidence of the texts does not contradict and indeed, as far as the preferred order is concerned, strongly supports an identification of Α sälyi, Β swälyai as 'right' and of Α päci, Β saiwai 'as 'left'. The time has come now to turn to questions of etymology. The morphological explanations given in Krause — Thomas (1960: 146) are largely unobjectionable: Α sälyi, Β swälyai and A päci are part of a class of denominal adjectives characterized by a suffix Β -(i)ye, A -i (both with palatalization of the preceding consonants — there are no nominative forms °siski 'lion's', °wärti 'forest-', °wasti 'house-' as given by Krause — Thomas, but only such in Α -im; Β kewiye 'cow's' requires a special explanation). A sälyäs and päccäs are properly identified as originating from A sälyäs (päccäs) possäsä; it may be possible to derive the feminine accusative singular form Β swälyai from an earlier collocation swälyai kalymi-sa 'at the right side (lit.: direction)'. Taken by itself, Β swälyai admits of a very simple etymology. Β swäl 'bait', Α swäl 'meat' show different specializations of an earlier meaning '*food; something to be eaten', regular beside the common

590

Werner

Winter

verb stem Β Α swä- 'eat'. Β swäl + -(i)ye thus meant 'pertaining to food'. Considering the fact that in the Indie tradition it is the right hand, and only the right hand, that is to be used in eating, the choice of the term is straightforward; if further support were needed, we may point to the parallel found in Khotanese Saka where hvarandaa- 'right; southern' goes with hvar- 'eat' (cf. also Sogdian yw'r'nt 'southern'). Α sälyi cannot be derived from Α swäl — the loss of the -wcannot be explained. However, evidence from Β can be adduced to eliminate the problem. In this language, we have two verb stems side-by-side, one, Β swä- (/swa-/), the other, Β sawä- (/sawa-/). This alternation does not reflect a productive process; as it cannot be due to borrowing, it seems legitimate to project it back into Common Tocharian. The regular match of Β /sawa-/ would be Α /säwa-/ (with reduction of /ä/ to /a/ in the second syllable due to syncope). The sequence Α /äwa/ is further reduced to /ä/ in forms like A sämäm 'living' based on a thematic verb stem Α säw- 'live' (cf. Winter 1965: 205). A verbal noun derived from a base */säwa-/ would then be Α säl*, the form which is the nucleus of the derived adjective Α sälyi. The explanation offered here for the set Β swälyai, Α sälyi has the advantage of requiring recourse only to morphological structures found elsewhere in Β and A and to sound changes known from both languages. It eliminates the need for etymologies of the type offered by Van Windekens (1976) which are poorly based semantically and frought with the necessity of having to operate with sound changes unsupported by any parallels whatsoever ('dans le dialecte A gw- a ete simplifie en p-'). For Β saiwai, now identifiable as 'left', the connection once suggested by Couvreur (1948: 128) may now be reconsidered: Couvreur proposed to compare Β saiwai with Skt. savya-, Av. haoya OCS suji 'left', referring these forms to a reconstructed PIE *sewyoand assuming a metathesis *-wy- > -yw- in B. This latter assumption is necessary if one wants to analyze Β saiwai, in analogy to Β swälyai, as a feminine accusative singular form, as this case ends in Β -yai throughout (cf. Krause — Thomas 1960: 144-157 for a convenient overview). The metathesis may recur in Β raiwe if indeed this word means something like 'sluggish' (an interpretation which, however, would be difficult to maintain for Β 497 a 4): rather than reconstruct an otherwise unattested PIE *rey-, as does Van Windekens (1976: 400), it would seem possible to connect Β raiwe

'Left' or 'right?

591

with *röwä as found in O H G ruowa 'rest, quiet' as a *-yo- derivative (considerable difficulties remain — the meaning of Β raiwe is far from certain, and a Sievers-Edgerton variant *-wiyo- rather than *-wyo- would be expected). It seems better to have the case of Β saiwai: Skt. savya-, etc., rest on its own merits; even here, however, it should not be overlooked that the Β form cannot be derived from PIE *sewyo-, but seems to require an underlying *sowyo-. Thus, the proposed etymology should be considered a possibility, hardly more. This leaves us with Α päci and äpat. A basic difficulty one has to face in dealing with terms denoting 'left' in Indo-European languages is that here euphemism abounds — see Buck (1949: 866). This leads us to pursue a possible avenue of explanation: Α äpat and päci both point toward an underlying simple noun Α pät(V)*. Disregarding meaning for the moment, we note that such a noun is found in the second part of the compound Α arämpät, Β erepate, which serves as a translation of Skt. rüpa-. Van Windekens (1976: 176), taking up, and slightly modifying, a proposal by Pisani (1942-1943: 28), derives Α p ä t , Βpäte (read: pate*) from PIE *bhätoor *bhato- without clarifying what precisely the reconstructed noun is to mean; therefore, little is gained by the proposal. However, in very general terms it may be said that Β päte*, Α pät in Β erepate, Α arämpät seems to signal a positive quality; hence, the term could have become the basis of euphemistic expressions particularly if, as our texts seem to indicate, the original use became limited to a compound which could easily acquire a partial opacity. Arguments of the type just used are rather unsatisfactory as they can hardly be raised beyond the level of mere possibility or, at best, plausibility. I have alluded to the possibility that Α päci and Α pät in arämpät may be related only because here, as in the case of Β swälyai, Α sälyi, there seems to be a chance that an intra-Tocharian explanation can be offered which may not be as interesting from an Indo-Europeanist point of view as that proposed for Β saiwai, but by its very nature offers fewer problems. But I would readily admit that the argument in the case of Α päci is very much weaker than that for 'right' — not surprisingly so as a recourse to euphemism in essence amounts to invoking the lucus-a-non-lucendo principle. We thus get a scale of probabilities: No formal and no semantic problems with swälyai, Α sälyi, over and above the fact that strong and well-understood ties with other Tocharian forms exist; I would not hesitate to say that the proposed etymology has a good chance

592

Werner

Winter

to be the correct one. Β saiwai is isolated in Tocharian; this may not be decisive as the same applies to the supposedly related forms in Indo-Iranian and Slavic. More perturbing is the fact that no certain parallels can be offered for the metathesis assumed (the reference by Couvreur (1948: 128 to §40 of his still unpublished Vergleichende tocharische Grammatik is of no help; Van Windekens 1976: 95 claims that in Β -w- was deleted, not metathesized, but his only plausible example, Β saiyye 'sheep': Α säyu (name of an animal), involves palatalized *-w- ( > -y-) and can therefore not be used in support of his claim). I would rank the etymology offered here for Β saiwai second: semantically perfect, insufficient support, though no counterevidence on, the side of the form. Finally, the connection suggested between Α päci, äpat, and -pat in Α arämpät, while unobjectionable in terms of form, suffers from very major shortcomings on the side of meaning: we do not know what precisely is denoted by -pat in Α arämpät — whether it adds to, or merely duplicates, the meaning of Α aräm; we do not know whether invoking the principle of euphemism is warranted, or whether pät* in Α äpat and pät* in Α arämpät just happen to be homonyms. I am inclined to leave it at that; a possibility, and in many ways a weak one at that, has been pointed out — I do not think that more can, and should, be done. I stated earlier that the passages containing Α äptä, which might well be considered a case form of Α äpat, could not be used to help elucidate the meaning of Α äpat, but could only be interpreted after this form had been properly identified. This agrees with the procedure adopted by Kölver (1965: 37-38). This author takes over, as does the entire Sieg — Thomas school, the argument put forward by Konow (1942: 208) that Α äpat twantam should be considered the precise equivalent of Khotanese Saka hvaramcmä tvandana'turning toward the right', classifies Α äpat as a noun used adverbially, and takes Α äptä to be the 'perlative' of this noun. His translations are not consistent; they do not always agree with his own definition of the functions of the 'perlative'. Kölver (1965: 30) states explicitly that depending on whether the governing verb denotes rest or motion, the 'perlative' is used to signal location or direction; yet he translates A 305 b 2 ni ... äptä mä katkar by 'they did not pass to my right' and A 332 b 5 mä skam äptä nenc by 'they are not turned toward the right'. The references made above to the function of the prefix Β e-, Α α-, of course, continue to apply; Kölver's interpretations can therefore not be accepted. It is not

'Left'or

'right'?

593

uninteresting that Kölver's mentor, in his own dissertation (Thomas 1952: 44), renders A 436 a 2 täne äptä yämäl sem tmäs kospremnassi kälpäl sem by '(If) I had acted differently(?) then, how much would I have obtained?', while Kölver proposes 'If I had acted in the right way then, how much of it would I have obtained?'. It is obvious that there exists at least a third possibility, namely, to translate A äptä in accordance with our findings about Α äpat: If Α äpat meant 'on the left side', that is, the side which is not 'right' in both senses of the word, then Α äptä would mean 'in the manner of what is on the wrong side', 'in the wrong way', 'wrongly'. All passages cited so far appear to permit such an interpretation; moreover, it seems to be strongly indicated for a bit of text which Kölver was unable to handle: In A 115 a 5 we find, after a major text break, mä näni snesrum täprem rohcam (kätäkl) // 'not without cause (arose) in me (feminine!) jealousy to such an extent ...'; in a 6 the text continues, after a lacuna, // (mäl) // kulyi äklus täs temi äptäprastam ... roncäm kätä(nkäs) // 'if a woman is (not) taught (properly), in her there arises, at the wrong time, jealousy'. As the result is quite negative, the time cannot be auspicious; therefore a translation along Kölver's lines by 'at the right time' is quite impossible. Hence, scant as the evidence for Α äptä is, it, too, seems to support the interpretations offered here for Α äpat and Α päci. The time has come for a few concluding remarks. In dealing with dead languages, we are constantly faced with the problem of being unable to determine directly the meaning of forms we encounter. We have to rely essentially on one or more of the following methods: If translations are available that may be considered reliable, we can adopt the interpretations given in them as points of reference. Lacking translations, we can try to develop working hypotheses based on collocations with forms with known meanings; if a study of texts other than those we used to find our hypotheses bears them out, we can retain them. Finally, we can introduce evidence from related languages as a basis for what can at best be intelligent guesses about what a hitherto obscure word might mean in the texts we study. Traditionally, the preferred method has been that of identification through a close investigation of context. Its success, however, depends to a large extent on the quality and the transparence of the context; the more broken or opaque it is, the less can we rely on the contextual method. Recourse to translations will lead to acceptable results if a satisfactory degree of bilingualism on the part of the people involved in the translation process can be assumed

594

Werner

Winter

and as long as there are no drastic changes necessary to make a message idiomatic in the target language. If possible, a combination of methods should be used; this applies in particular to the etymological one — it should not be applied as the primary approach if that can be avoided —; on the other hand, it can be eminently useful to confirm or disconfirm hypotheses arrived at by a study of context or of translation. All three methods have been utilized in the present paper. Context was studied to determine the meaning of Β swälyai, Α sälyi; etymological analysis, performed here within the bounds of two languages, was employed to find support for the hypothesis offered. Evidence from parallel texts in other languages was adduced and, in this case, rejected as it was contradicted by contextual and morphological data that were thought to be of higher rank. Had the texts been more complete, the task at hand would have been easier, but also less interesting — a state of affairs very common in historical work. Perhaps that is one reason why we enjoy it so much.

References Buck, Carl Darling 1949

A dictionary of selected synonyms in the principal Indo-European languages (Chicago: University Press). Couvreur, Walter 1950 'B-Tocharische Etymologien', Archiv Orientälni 18: 126-130. 1955-56 'Bemerkungen zu Pavel Pouchas Thesaurus linguae tocharicae dialecti A \ La Nouvelle Clio 7-8: 67-98. Edgerton, Franklin 1953 Buddhist Hybrid Sanskrit grammar and dictionary 2: Dictionary (New Haven, Connecticut: Yale University Press). Kölver, Bernhard 1965 Der Gebrauch der sekundären Kasus im Tocharischen (Frankfurt am Main). Konow, Sten 1942 'Notes concerning Khotanese', NTS 13.199-224. Krause, Wolfgang — Thomas, Werner 1960 Tocharisches Elementarbuch 1: Grammatik (Heidelberg: Winter). Pisani, Vittore 1942-43 'Glottica Parerga 5: Etimologie tocariche', Rendiconti, Reale Istituto Lombardo di Scienze e Lettere, Classe di Lettere 76: 2: 21-34. Sieg, Emil 1944 Übersetzungen aus dem Tocharischen 1. Abhandlungen, Preußische Akademie der Wissenschaften 1943. Phil.-hist. Klasse 16. (Berlin: Akademie der Wissenschaften).

'Left'or

'right'?

595

Sieg, Emil — Siegling, Wilhelm — Schulze, Wilhelm 1931 Tocharische Grammatik (Göttingen: Vandenhoeck & Ruprecht). Thomas, Werner 1952 Die tocharischen Verbaladjektive auf -1. Eine syntaktische Untersuchung. Institut für Orientforschung, Deutsche Akademie der Wissenschaften, Veröffentlichung 9. Thomas, Werner — Krause, Wolfgang 1964 Tocharisches Elementarbuch 2: Texte und Glossar (Heidelberg: Winter). Van Windekens, Albert Joris 1938-39 'Linguistische Studien V: De vertaling van Tocharisch Α päci en ςälyC, Philologische Studien 10: 183-188. 1941 Lexique etymologique des dialectes tokhariens. Bibliotheque du Museon 11. (Louvain: Museon). 1976 Le tokharien confronte avec les autres langues indo-europeennes 1: La phonetique et le vocabulaire. Travaux, Centre International de Dialectologie Generale, Universite catholique neerlandaise de Louvain, 11. (Louvain: C I D G ) . Winter, Werner 1965 'Zur Vorgeschichte einiger Verbformen in Tocharisch A', KZ 79: 203-210. 1983 Ά tentative English-Walapai dictionary and Etymologicum Tocharicum: Unterschiede und Gemeinsamkeiten', in: Alfred Bammesberger, Das etymologische Wörterbuch (Regensburg: Pustet), 313-327.

Index of authors

Abelson, R. 302 Acson, V. 108 η. 1, n.7 Akmaijan, A. 463 Akmaijan, A. — Heny, F. 463 Ahlqvist, A. 1, 101 Ahn, P. 30 n. 2 Aijmer, K. 148 Akatsuka, N. 280 n. 3 Alarcos Llorach, E. 549 Albertos Firmat, M. L. 381, 386, 387 Alexandre, P. 186-188, 195 n. 15 Alinei, M. 392 n. 8 Alleyne, M. 533, 534 Andersen, H. 297, 300 Anshen, F. 430, 435, 436, 443, 444, 457 Anshen, F. — Aronoff, M. 435, 436, 443, 444, 457 Anttila, R. 109 n. 17 Aronoff, M. 33, 55 n . l , 201, 435, 436, 440, 443, 444, 451, 454, 457 Aronoff, M. — Schvaneveldt, R. 55 η. 1, 435, 457 Arbois de Jubainville, H. d' 385, 387, 388 Ard, J. 67 Asin Palacios, M. 392 Atlas, J. D. 275 Atlas, J. D. — Levinson, S. C. 275 Baist 389 Baker, C. L. 464, 465, 470 - 473 Baker, W. J. 55 η. 1 Bamberg, M. 540 Barney, S. A. 168 n. 2 Baron, N. S. 171, 192, 193, 194 η. 1 Bately, J. 526 n. 80 Battan, P. L. 286, 294 Bauer L. 486, 487, 522 n. 20-22, n.24, n. 25, n. 28, 523 n.38, 525-526 n.74, 527 n. 91, n.94 Baudouin de Courtenay, J. 109 n. 13 Baugh, A. C. 454, 529 n. 110

Baugh, A. C. - Cable, T. 529 n. 110 Beard, R. 109 n.13, n. 14 Beaugrande, R. de 303, 304 Behagel, O. 271 Bello, A. 549 Bellugi, U. 471 Bennett, J. 270, 280 n. 2 Benveniste, E. 4 Bergsten N. 523 n. 39, 527 n. 89, n. 94, 529 n. 110 Berndt, R. 343, 347, 350, 352 n . l , 353 n. 10, n. 11 Berwick, R. C„ 470, 472, 474, 480 n. 3, n. 4 Bhaldraithe, T. de 1, 2, 4, 7 η. 4, η. 14, 209, 217 Bickerton, D. 304 Bily, I. 102 Binchy, D. A. 7 n. 9 Birkhan, H. 388 Björkstam, Η. 12 Black, Μ. 287 Blenner-Hassett, R. 491, 509, 526 η. 75, η. 79, η. 80 Bloomfield, L. 346, 347, 349, 352 η. 6, 353 η. 12, 521 η. 10 Bagholm, Ν. 492, 523 η. 45 Bokamba, Ε. Β. 188, 189 Bolinger, D. 570 η. 11 Borden, Α. R. 452 Bosworth, J. 222, 223, 227, 573, 578 Bosworth, J. - Toller, Τ. Ν. 222, 223, 227, 573, 578 Bowerman, Μ. 264 Braine, Μ. D. S. 471 Brook, G. L. 520 η. 1, 529 η. 112 Brook, G. L. - Leslie, R. F. 520 n . l , 529 n . l 12 Brown, R. 471 Brown, R. — Cazden, C. B. — Bellugi, U. 471 Brown — Hanlon 471

598

Index of authors

Brüch, J. 389 Brugman, C. M. 127 Brugmann, K. 458 n. 14 Bruner 142 Buck, C. D. 330, 591 Burnham, J. M. 263, 272 Burt, Μ. K. 463 Buttler, D. 298 Bwantsa — Kafungu, S. P. 189 Bynon, T. 40 Bybee, J. 59, 76, 76 n.4, 114, 151 n.29, 458 n. 14, 563, 564, 570 n.9 Bybee, J. - Moder, C. L. 151 n.29 Bybee, J. - Pagliuca, W. 114, 563, 564, 570 n.9 Cable, T. 529 n.UO Caix 389 Campbell, A. 523 n. 39, 524 n.49 Campbell, R. N. 440 Cappello, T. 102 Cappello, T. — Tagliavini, C. 102 Carlson, G. 464, 465 Carlson, G. — Roeper, T. 464, 465 Carr, C. T. 489, 490, 521 n. 13, 522 n. 20, n. 25, n. 28, 524 n. 55, 526 n. 70, n. 71, n.78 Cartagena, N. 549, 550 Castro, A. 389 Cawdrey, R. 456 Cazden, C. B. 471 Cejador, J. 391 n. 3 Chao, Y. R. 359 Chen, M. 449 Chevallet, A. de 458 η. 14 Chomsky, Ν. 109 η. 11, 142, 201, 408, 431 η. 3, 463, 466, 4 6 8 - 4 7 0 , 472, 474, 478, 479, 480 n . l , n.4 Chomsky, Ν. - Halle, Μ. 201, 408, 431 η. 3 Christian Brothers 7 η. 5 Clark, Ε. V. 443 Clark, Ε. V. - Clark, Η. Η. 443 Clark, Η. Η. 443 Clark, R. 458 Clark Hall, J. R. 113, 119, 223, 227 Close, R. A. 15 Coates, J. 14, 20, 64, 65, 73, 74, 77 n.6, n.8, n.9 Collinson, W. E. 115 Comrie, B. 60, 61,76 η. 1, 86,87,91 - 9 3 , 191, 194, 195 n.17 Corominas, J. 383, 389, 390 Corominas, J. — Pascual, J. A. 389, 390

Couvreur, W. 585, 590, 592 Coyaud, M. 267 Coyaud, M. — Hamou, K. 267 Creider, C. A. 177, 178 Crider, K. 437 Cutler, A. 435, 443, 444, 457 Czepluch, H. 463, 465, 468, 479 Dahlgren, K. 301 d'Ardenne, S. R. T. O. 524 n. 26 d'Ardenne S.R.T.O. - Dobson, E. J. 524 n. 6 Delbrück, Β. 114 Dell, F. 109 η. 13 Dell, F. - Selkirk, Ε. Ο. 109 η. 13 Derwing, Β. 55 η. 1 Derwing, Β. - Baker, W. J. 55 η. 1 Denny, J. P. 177, 178 Denny, J. P. - Creider, C.A. 177, 178 Diez 392 Dishington, J. 168 n. 3 Dobson, E. J. 524 n. 6, 528 n. 19, 542 n. 46 Doke, C . M . 175, 176, 179, 182, 191 Dottin, G. 3, 388 Dressier, W. 105, 106, 108 η. 1, n. 2, n. 6, n.7, 109 n. 16, 431 n.6 Dressler, W. - Acson, V. 108 η. 1, n.7 Dressier, W. — Mayerthaler, W. — Panagl, O. - Wurzel, W.U. 108 η. 1, n.7, 109 η. 16 Dryden, J. 455 Earle, J. 124 n. 2 Earle, J. - Plummer, C. 124 n. 2 Eaton, H. 375 Edgerton, F. 589 Einstein, A. 358 Einstein, A. — Infeld, L. 358 Eisenberg, P. 267 Evans, D.E. 7 n . l 1, 387, 388 Eys, W.J. 567 Faiß, Κ. 362, 494, 524 η. 48, 527 η. 84, η. 94 Fararo, T.J. 319 Fawcett, R. P. 306 Fehling, D. 5 Fill, A. 362 Fillmore, C.J. 127, 5 7 3 - 5 7 5 Fisiak, J. 259 n. 3 Fleischmann, S. 17, 60, 67, 73, 78 n. 13, n. 14, 566 Fräser, T. 124 n. 7

Index of authors Gadde, F. 450 Gaffmann 438 Garcia E. C. 391 Garcia Gonzalez, J. 378 Gardner, T.J. 522 n.24, 525 n.64 Gauger, H.-M. 361 Geeraerts, D. 128, 142, 143 Gerritsen, J. 55 n. 6 Gili Gaya, S. 549, 570 n. 5 Gillies, W. 2 Givon, T. 5, 60, 176, 177, 194 n.8, n. 12, 276, 565 Görlach, M. 340, 343, 347, 350, 352 n. 3 Götz, D. 527 n. 94 Gonda, J. 106 Gonzales Olle 378 Goossens, J. 127, 143, 145, 146, 148 Greenbaum, S. 523 n.38, 524 n.49 Greenberg, J. Η. 1, 62, 76 n. 3, 109 n. 19, 171, 182, 194 n. 3, n.4, 305 Greenberg, J. H. et al. 305 Greenough 347 Greenough — Kittredge 347 Grein 573, 576, 578 Grice, H.P. 275, 304 Grimm 113 Günthner, C. 417 Guilbert, L. 109 n. 15 Guillaume, G. 118-120, 124 n . l , n.7, 156, 158-160, 163, 165, 168 n.9 Guimier, C. 114, 124 n.7, 162, 168 n.7, 432 n. 11, 526 n.66 Gumperz, J. J. 567 Gumperz, J.J. - Wilson, R. 567 Guthrie, M. 175 Haegeman, L. 13, 18 Haiman, J. 102, 267 Hall, J. 521 n.15, 527 n.96, 578 Halle, M. 201, 408, 431 n.3 Halliday, Μ. Α. K. 541 Halliday, M . A . K . - Hasan, R. 541 Hammarström, G. 407 Hamou, K. 267 Hancock, I. 533, 534 Hansen, B. 229, 487, 521 n. 8, 522 n. 25, 523 n.38 Hansen, B. et al. 229, 487, 521 n. 8, 522 n.25, 523 n.38 Harris, M. 5 Hasan, R. 541 Hayek, F. A. v. 355 Heath, S.B. 541, 542 He, W. 30 n. 2

599

Heny, F. 463 Herbert, R. K. 194 n. 7 Heringer, H.J. 318 Hervey, S. 418, 419, 424, 427 Hewson, J. 162, 163 Hickey, R. 201, 209, 218 n.3, n.9 Hinnebusch, T.J. 179, 182 Hirtle, W.H. 160 Hockett, C. 427 Holder, A. 387 Hopper, P. 336 n.7 Hopper, R — Thompson, S. A. 336 n. 7 Horn, L. 64, 65 Householder, F. W. 299 Hualde, J.I. 570 n.15 Hübner, E.W.E. 381 Huld, M . E . 7 n.9 Inenlikej, R I . 94 η. 1 Infeld, L. 358 Ingersoll, S. M. 526 n. 78 Jacobi, H. 5 Jakobson, R. 62, 305 Jespersen, O. 33, 221, 426, 436, 4 5 0 - 4 5 2 , 455 Jönsjö, J. 522 n. 20, n. 21 Johansson, S. 527 n. 89 Johnson, M. 574 Joly, A. 120 Jones, M. 7 n. 12 Jones, M. — Thomas, A. R. 7 n. 12 Joos 123 Kärre, K. 526 n. 70 Karlsson, F. 20 Karttunen, L. 280 η. 1 Karttunen, L. - Peters, S. 280 η. 1 Kastovsky, D. 109 n. 16, 218 n.2, 229, 247, 259 n . 5 - 7 , n. 10, 260 n. 1 4 - 1 6 , n.20, 423, 484, 486, 521 n.8, n.15, n. 18, 522 n. 19, 524 n. 5 6 - 5 8 , n.60, 525 n. 60, n. 6 2 - 6 4 , 527 n.94 Katz, J.J. 300, 301 Katz, R. 113 Keenan, E. O. 542 Keil, H. 386 Keller, R. 355 Keller, J.E. 380, 391 n.4 Keller, J.E. - Linker, R.W. 380, 391 n.4 Keniston, H. 549 Ker, N . R . 520 n.2 Kernan, K.T. 541

600

Index of authors

Kilani-Schoch, M. 108 η. 1, n.7 Kilani-Schoch, M. - Dressler, W. 108 n.7 Kington-Oliphant, T.L. 450, 452, 455, 458 n. 11 Kittay, E. 295 n. 4 Kittay, E. - Lehrer, A. 295 n. 4 Kittredge 347 Klegraf, J. 581 η. 1 Kluge, F. 156 Kluge, F. 365 η. 3 Kluge, F. - Mitzka, W. 365 n. 3 Kökeritz, H. 417, 422 Kölver, B. 592, 593 König, E. 267, 270, 280 n. 5 König, E. — Eisenberg, P. 267 König, E. - Traugott, E. C. 280 n. 5 Kövecses, Z. 574 Köhler, K. 365 η. 1 Koj, L. 97 Konow, S. 584, 592 Koskenniemi, K. 109 n. 21 Kotzer, G. 432 n. 9 Koziol, H. 221, 423, 436, 487, 505, 522 n. 20, 524 n. 47, n.49 Krähe, H. 233 Krähe, H. - Meid, W. 233 Krause, W. 584, 589, 590 Krause, W. - Thomas, W. 589, 590 Kretschmer 383 Kühlwein, W. 577 Kuhlmann, H. 263 Kuhn, S. 452 Kunene, E. C. L. 179, 185, 195 n. 13 Kurath, H. 452 Kurath, H. - Kuhn, S. 452 Labov, W. 301, 365 n.2, 574 Lacarra, J. M. 379 Lakoff, G. 127, 302, 574 Lakoff, G. — Johnson, M. 574 Lakoff, G. — Kövecses, Z. 574 Lambermont, H. 149 n.2 Langacker, R.W. 127, 304-306, 308, 320 n. 3 Lanham, L.W. 180 Lapesa, R. 377 Laszlo, E. 302 Lauchert 391 n.4 Lavandera, B. 548, 566 Lawler, J. 18 Leech, G. 14, 20, 339, 342-345 Leech, G. - Coates, J. 14, 20 Lees, R. B. 222

Lehmann, C. 363, 364 Lehmann, W. P. 5 Lehnert, M. 445 Lehrer, A. 283, 284, 293, 294, 295 n. 4, 340 Lehrer, A. - Battan, P. L. 294 Leisi, E. 573, 574 Lenneberg, Ε. H. 466 Lenze, J. 113 Lerch, E. 263 Leslie, R. F. 520 η. 1, 529 n. 112 Levi, J.N. 521 n.7 Levinson, S. C. 273, 275 Lewandowska-Tomaszcyk, B. 297, 300, 302, 304, 320 n. 2 Lewandowska-Tomaszcyk, B. — Machova, S. 300, 304 Li, C.N. 326, 336 n.6 Li, C. N. — Thompson, S. A. 336 n. 6 Liebermann, F. 352 n. 5 Lightfoot, D. 20 n.3, 109 n. 11, 480 n.4 Lightner, Τ. M. 55 η. 1, 109 n. 13 Lindemann, J.W.R. 124 n. 9 Lindheim, B. v. 241 Lindner, S. 127 Linell, P. 19 Linker, R.W. 380, 391 n.4 Lipka, L. 23, 339, 346, 347, 352 n . l , n.2, 353 n. 11, 527 n.94 Loewenthal, J. 389 Lüdtke, H. 355, 358, 362-364, 407 Luhmann, A. 528 n. 101 Lyons, J. 16, 17, 64, 65, 103, 108 n.3, 283, 339-341, 549, 573, 576, 579 MacEoin, G.S. 218 n.7, n.8 Machova, S. 300, 304 MacKay, D. G. 97 Madden, F. 523 n. 35, 524 n. 59 Maetzner, E. 436 Malkiel, Y. 568 Manczak, W. 374, 375 Marchand, H. 33, 34, 113, 165, 168 η. 1, 221 - 223,228,410,411,414,423,424, 429, 432 n. 9, 436, 440, 484, 486, 487, 499, 521 n. 10, n.12, 523 n.38, 524 n. 47, n. 49, 525 n. 61, n. 63, n. 72, 526 n. 82, 527 n.91, 528 n.100 Marchese, L. 61, 67, 78 n. 15 Marcos-Marin, F. 377, 382, 390, 563, 569 n.4 Maria de Azkue, R. 390 Markarjan, R.A. 108 n.3 Martinet, Α. 424

Index of authors

601

Matsch, W. 259 n. 10 Matthews, P. H. 109 n. 14 Mayerthaler, W. 99, 108 η. 1, n.3, n.7, 109 n. 16, n. 17, n.21 Meid, W. 115 Meillet, A. 336 n. 5 Meinhof, C. 175 Mel'cuk, I. A. 95 η. 1, 97, 103, 108 n. 3 Menendez Pidal, R. 377, 378, 384, 385 Mensing, O. 263, 271 Meritt, H.D. 113, 119 Mervis, C.B. 11, 127 Metivier 392 Meyer-Lübke, W. 383, 388 Miller, D.G. 2, 193 Miodunka, W. 301 Miranda, R.V. 193 Mitchell, B. 113 Mitzka, W. 365 n. 3 Moignet, G. 166 Moll, T. A. 94 η. 1 Moll, T. A. - Inenlikej, P. I. 94 η. 1 Monod, J. 356 Moravcsik, E. A. 173, 191, 193, 533 Mosse, F. 157, 452, 455, 458 n. 11 Mustanoja, T. F. 505, 523 n. 40, n. 45 Mühlhäusler, P. 533 Mulder, J. 419, 424, 427 Mulder, J. - Hervey, S. 419, 424, 427 Mullenhof, K. 385 Muret, E. 367 Muret, E. — Sanders, D. 367 Muxin, A. M. 397

Pascual, J. A. 389, 390 Paul, H. 347 Pauly, A. v. — Wissowa, G. 384 Pedersen, H. 3, 7 n. 10, n. 12 Peirce, C. 108 n. 4 Pennanen, Ε. V. 486 Perkins, R. 59, 77 n. 5, n. 12 Peters, S. 280 η. 1 Piaget 142 Pictet 387 Pilch, H. 7 n. 12, 232, 259 n.3, 418, 421, 422, 425, 430, 431 n. 5 Pinker, S. 467 Pisarek, W. 299 Pisani, V. 591 Plank, F. 104, 105, 108 n.3 Platt, M. L. 542 Plummer, C. 124 n.2 Pottier, B. 364 Priebsch, R. 115 Priebsch, R. - Collinson, W.E. 115 Pohl, A. 108 n. 3 Pokorny, J. 158, 384 Pyles, T. 340, 347-349, 352 n.4, 353 n.9, n. 12

N'Diaye, G. 567 Nesfield, J. C. 436, 458 n. 4 Nöth, W. 302, 305, 347, 351, 352 n.7, 353 n.9 Nyrop, K. 458 n. 14

Randall, J. 435, 436 Real Academia Espanola 392 n. 5, 549, 569 n.4, 570 n.6 Reichl, K. 440 Reisman, K. 537, 543 Renwick, W. L. 520 η. 1, n. 2 Renwick, W. L. - Orton, H. Wakelin, M.F. 520 n . l , n.2 Rettig, W. 361 Rey, A. 391 n. 2 Richardson, I. 181, 184, 185, 190, 194 n.7 Rickford, A. 543 Rickford, J. 543 Rickford, A. - Rickford, J. 543 Rinsland, H. D. 479 Ritchie, W. C. 374, 476 Robinson, F. 455, 458 n. 12 Roeper, T. 464, 465 Romaine, S. 304

Oakden, J.P. 483-485, 509, 520 n.3, n. 5, 526 n.71, 529 n.113 Ochs, E. 542 Ochs, E. - Schiefielin, Β. B. - Piatt, M. L. 542 Oehrle, R.T. 463, 465, 472, 478 Okhotina, N.V. 185 Orton, H. 520 η. 1, n.2 Osselton, Ν. Ε. 448 Pagliuca, W. 76, 114, 563, 564, 570 n.9 Palmer, F. R. 18 Palomar Lapesa, M. 382 Panagl, O. 108 η. 1, n. 7, 109 n. 16, 229

Quirk, R. 259 n. 3, 263, 264, 272, 409, 421, 523 n. 38, 524 n.49 Quirk, R. et al. 264 Quirk, R. - Greenbaum, S. 523 n. 38, 524 n. 49 Quirk, R. - Wrenn, C. L. 259 n. 3

602

Index of authors

Rosch, E. 11, 127, 136, 137, 139-141, 301 Rosch, E. - Mervis, C.B. 11, 127 Rudes, B. A. 108 n. 3

Sachs, G. 392 Safarov, S. 400 Sajavaara, K. 101, 309 Sanders, D. 367 Sauer, Η. 429 Saussure, F. de 124 η. 1, 431 η. 1 Schänk, R. 302 Schänk, R. - Abelson, R. 302 Schapera, I. 179 Schapera, I. — van der Merwe, D. F. 179 Scheidweiler 295 η. 1 SchiefTelin, Β. B. 542 Schlauch, M. 454, 455 Schmidt 574 Schmidt, Κ. H. 6, 7 n. 15, 387, 388 Schöning, G. 385 Schot, E. 377 Schrack, D. 525 Schuchardt 389 Schützeichel, R. 390 Schulten, A. 385 Schulze, W. 584 Schvaneveldt, R. 55 η. 1, 435, 457 Schwarze, C. 108 n. 3, 109 n. 12, n. 13 Scollon, R. 542 Seebold, E. 524 n. 56 n. 58, n. 60 Seiler, H. 109 n. 16, 364 Selkirk, Ε. Ο. 55 η. 1, 109 η. 13 Selten, Β. 525 η. 73 Seijeantson, Μ. S. 451 Seybold, C. F. 377 Shelly, P.V.D. 451 Shepherd, S. 535 Shipley, J. T. 454, 455 Shook, L.K. 168 n. 8 Sieg, Ε. 584, 587, 592 Sieg, Ε. - Siegling, W. - Schulze, W. 584 Siegling, W. 584 Silva-Corvalan, C. 566 Silverstein, M. 336 n. 5 Sjcestedt-Jonval, M.L. 4, 7 n.7 Skalicka, V. 97, 100 Skeat, W. W. 352 n. 3 Skorik, P. J. 94 η. 1 Skvoretz, J. 319 Skvoretz, J. — Fararo, T.J. 319 Slawski, F. 106

Smith, A.H. 523 n.38, 525 n.74, 526 n. 75, n. 79 Spence, N.C.W. 295 η. 1 Sperber 347 Sprengel, K. 339 Standop, E. 12, 20 η. 1 Stanley, E.G. 527 n.87 Steele, S. 64, 65 Steels, L. 310, 313 Stegmüller, W. 109 n. 10 Stemberger, J. 56 n. 8 Stern, G. 16, 259, 283, 285, 574 Stolz-Leumann 392 Strang, Β. Μ. H. 520 n. 2 Strauß, J. 148, 579, 581 η. 1 Strauss, J. W. 490, 492 Sunden, K. F. 528 n. 107 Suzman, S.M. 185, 190 Sweet, H. 165 Szemerenyi, Ο. 107 Tagliavini, C. 102 Tauli, V. 109 n. 21 Tellier, A. 20 η. 1 Teske, M. 292 Thiele 450 Thomas, A.R. 7 n. 12 Thomas, W. 584, 589, 590, 592, 593 Thomas, W. - Krause, W. 584 Thomason, S. G. 336 n. 3 Thompson, S. A. 336 n. 6, n. 7 Thurneysen, R. 2, 3, 199, 200, 211, 214, 215, 219 n. 12 Thurnwald, R. 106 Tindale 290 Tobler — Lommatzsch 526 n. 81 Tolkien, J. R. R. 528 n. 109 Toller, Τ. N. 222, 223, 227, 573, 578 Tomaszczyk, J. 308 Toth, K. 524 n. 51 Tovar, A. 383-387, 389, 390, 392 Traugott, E.C. 193, 270, 280 n.5 Trier, J. 284, 295 Tucker, S. I. 455 Uhler, K. 163 Ulimann, S. 140, 143, 283, 343, 345, 347, 349 Ultan, R. 67, 70 Valente, J. F. 182, 186 Van Dale 131 Van der Mewe, D.F. 179 Van Windekens, A.J. 584, 590-592

Index of authors Vendryes, J. 3 Venneman, T. 569 n. 2 Verner 233 Vising, J. 451 Visser, F.T. 12-18, 272, 465, 479, 526 n. 70 Vries, M. de 149 η. 1 Walker, J. 156-158, 164, 168 n . l , n.2, 445 Wakelin, M. F. 520 n . l , n.2 Wang, W. 449, 451 Warren, B. 484, 526 n. 67 Wartburg, W. v. 431 n.7 Weida, G. 16, 17, 20 η. 1 Weiland 128 Weinreich, U. 329, 330 Wellander 363 Wellmann, Η. 526 η. 78 Weimers, W.E. 173, 175, 181, 194 η. 6 Welna, J. 431 η. 8 Werner, Ο. 108 η. 3 Westphalen, Τ. 581 η. 1

603

Whinnom, Κ. 567 White, L. 109 η. 11 Whiteley, W.H. 177 Wierzbicka, Α. 311 Willemyns, R. 146 Williams, J. M. 283, 293, 295 Wilson, R. 567 Winter, W. 107, 587, 590 Wittgenstein, L. 138, 573, 574 Wolf, P.P. de 173 WolfT, D. 528 n. 102 Wrenn, C. L. 259 n. 3 Wright, J. 259 n. 3 Wurzel, W.U. 108 n . l , n.2, n.7, 109 n. 16 Wyld, H.C. 520 n.2, n.3, 523 n.35, n. 37, 524 n. 46 Zemskaja, Ε. A. 109 n. 15 Zeuss, J. C. 3, 4 Ziervogel, D. et al. 191 Zimmer, Κ. E. 33 Zukova, Α. N. 93, 94 η. 1, 95 η. 1

Index of languages

Abkhaz 267 Accadian 106 Acoma 70 Ainu 64, 77 Albanian 381, 383, 384, 390 Algerian 106 Alutor 85, 86, 91, 95 Arabic 67, 105, 106, 181, 268, 389, 390 Arabic, Egyptian 279 Aramaic 106 Aranese 383 Avestan 106, 590 Aveyron 383 Aztec 105 Balto-Slavic 168 Bantu 172, 174-176, 179-181, 184, 185, 187, 189-195 Bantu, Proto- 175-178, 183, 184, 192, 194 Bantu, Southern 179, 180, 185 Baonan 326 — 336 Basque 61, 64, 65, 77, 105, 383, 384, 388-390, 566, 567, 570 Bearnese 383 Breton 106, 384, 386 Bulgarian 383 Burushaski 61, 62, 77 Cahuilla 364 Cambodian (Khmer) 64, 65, 77, 268 Cantal 383 Car 64, 77 Castillian 378, 386 Catalan 383, 386 Celtic 106, 388, 390, 391, 500, 521 Chavchuven 85, 86, 89, 91, 93 Chinese 267, 268, 334, 335 Chinookan, Proto- 336 Chukchi 8 5 - 9 4 Cornish 106, 386 Corsican 383

Creole, Antiguan 533 — 544 Cretan 383 Danish 106 Diegueno 64, 70, 77 Doric 383 Dutch 77, 102, 105, 128-134, 144, 149-151, 268, 274, 279, 422 Egyptian, Old 105 Engadinish 383 English 6, 7, 13, 15, 16, 33-43, 47, 48, 53-57, 61, 65-67, 72-77, 88, 9 7 - 9 9 , 101, 103, 104, 108, 113, 129, 155, 161, 163, 164, 168, 171, 173, 181, 193, 194, 199, 208-210, 218, 221-223, 227, 229, 230, 240, 246, 253, 255-259, 263-268, 270-280, 283, 285-292, 294, 301, 302, 308, 309, 315, 326, 339-343, 345-353, 357, 362, 364, 367-371, 373, 374, 397 - 399, 401-404, 408 - 432, 435-457, 463-465, 467-469, 471, 472, 477-480, 484-490, 492, 493, 496, 498-500, 502, 504, 506, 508, 521-526, 569 English, Black 541 English, Middle 20, 115, 155, 193, 218, 221-223, 271, 277, 285, 340, 344, 350, 352, 353, 405, 426, 429, 430, 446-448, 450-456, 458, 483-529, 573 English, Northern 340, 341, 343 English, Old 12-18, 34, 113-119, 121-125, 155, 156, 158, 162-165, 168, 192, 193, 221-225, 227, 229-253, 255-260, 263, 271, 272, 280, 285, 301, 341-343, 348-350, 364,400, 401,405,422,423,425, 429, 430, 432, 446-448, 451-456, 458, 465, 479, 483, 485, 487-500, 502, 503, 506-529, 573-579, 581

606

Index of languages

Estonian 105 Ethiopian 106 Ewe 267 Ewondo 192 Ewondo populaire 187, 188 French 33, 34, 55, 66, 67, 101, 103, 104, 106,124,128,134,135,151,160,221, 223-225, 256, 259, 263, 267, 268, 270,272,274,276,278,279, 345, 349, 357-361, 371-373, 409, 413, 414, 425-427, 451-455, 458, 506, 507, 569 French, Old 263, 290, 361, 527, 528 Gaelic, Scottish 2, 7, 60, 62, 384 Galician 382, 388 Gallic 381, 387, 388, 392 Garo 69-71, 74, 77 Gascon 382, 388, 392 Georgian 61, 64, 77 German 101, 102, 104-106, 115, 151, 171, 231, 256, 258-260, 263, 267, 268, 270-273, 278, 279, 284, 326, 340, 347, 349-353, 356, 362, 364, 365, 367-373, 375, 383, 389, 414, 419,421,422,424,426,431, 479, 502, 521, 523, 526, 527 German, Low 352 German, Middle High 271 German, Old High 124, 263, 271, 390, 591 German, Swiss 414 Germanic 3, 5, 67, 77, 106, 108, 113, 115, 155-157, 168, 236, 246, 383, 389, 390, 414, 521, 525 Gothic 114, 115, 156, 157, 168, 280 Greek 3, 33, 55, 67, 102, 106, 114, 125, 128, 156, 157, 172, 295, 364, 371, 383-385, 387, 390, 420, 429, 454 Haitian 6 4 - 6 7 , 77 Haussa 106 Hui 326-329, 334-336 Hungarian 101, 106, 267, 268 Hopi 308 Iatmul 61, 62, 64, 77 Igbo 61, 105 Illyrian 387 Indo-European, Proto- 33, 114, 115, 158, 193, 233, 235, 236, 590 Indonesian 269, 362 Indonesian, Bahasa 362

Irish 1 - 6 , 199, 200, 202-218, 384 Irish, Connemara 209, 218 Irish, Middle 3, 6, 386 Irish, Old 1 - 3 , 6, 200-202, 206, 214, 218, 219 Italian 101-103, 274, 364, 372-374, 383, 388, 392, 455, 567 Japanese 267, 268, 467, 469 Khatyr 87 Khoisan 180 Khotanese Saka 592 Kiwai 61, 62, 64, 70, 74, 77 Korean, Modern 23-27, 29, 30, 70 Korean, Middle 23, 2 6 - 3 0 Koryak 85-91, 93, 94, 105, 106 Kru 61, 66, 67, 78 Kutenai 77 Kuwaa 61 Latin 33-44, 54-56,102, 103,106,108, 109, 114-117, 124, 128, 134, 168, 200, 221, 223, 256, 259, 267, 268, 274, 356, 361, 364, 372, 374, 381, 382, 384-386, 388-392, 409, 413-415, 420,421,426,427,429,431, 452,454, 458, 506, 507, 522, 526, 528 Lingala 188-190, 192 Lithuanian 106 Logbara 61 Luiseno 64, 65, 77 Luvale 186, 187 Maasai 61, 64, 77 Malayalam 6 9 - 7 1 , 77, 267 Mandarin 64, 77, 334 Maori 105, 362 Mongolian 105, 106 Mongolian, Khalkha 326 Monguor 330, 336 Mundari 269 Nahuatl 61, 62, 77, 269 Navaho 64, 77 Norse, Old 114, 115, 155, 156, 506, 507, 526, 528 Occitan 383 Ojibwa 70 Palana 85, 86, 91 Palaung 64, 77 Pawnee 61, 62, 6 9 - 7 1

Index of languages Polish 101, 102, 106, 172, 173, 194, 267, 2 9 7 - 3 0 1 , 308, 309, 313-317, 320, 372, 373 Portuguese 372, 382, 388, 389 Romance 67, 74, 76, 78 Rumanian 67, 383, 384, 389, 569 Russian 102, 267, 274, 373, 397, 411, 414 Salar 336 Samoan 362 Sanskrit 114, 268, 590, 591 Santa 336 Sardinian 383 Saxon, Old 114 Scotch 364 Semitic 5 Senoi 61, 62 Serbian 61, 62, 64, 66, 67, 75, 77 Serbo-Croatian 267 Shona 182 Sierra Miwok 61, 62, 67, 69, 71 Slavic 106, 414 Slavic, Old Church 67, 102, 106 Somali 67 Sotho 268 Sotho, Northern 179 Sotho, Southern 179 Spanish 62, 66, 173, 268, 364, 371, 374-386, 388, 3 9 0 - 3 9 2 , 547-565, 567-570 Spanish, Corravubias 548, 550 — 552, 555, 556, 558, 5 6 1 - 5 6 4 Spanish, Medieval 563 Subia 179 Suomi-Finnish 101, 106, 267 Susu 268 Swahili 181, 188 Swazi 195 Swedish 101, 106

607

Tagalog 67 Tarascan 61, 70 Temne 191 Thai 64, 77 Tibetan 268 Tibetan, Amdo 326, 330, 336 Tirolese 383 Tiwi 69, 71 Tocharian A 5 8 4 - 5 9 4 Tocharian Β 5 8 4 - 5 9 4 Touareg 61, 62 Turkish 69, 71, 77, 101, 105, 267, 268, 279 Tsonga 179 Tswa 172, 173, 176 Ukrainian 67 Umbundu 182, 186, 187 Vedic 106 Venda 191, 192 Vietnamese 77 Vogul 105, 106 Wallon 383 Welsh 7, 61, 364, 384, 386, 498 Wutun 326, 330-332, 336 Xhosa 180 Yanomana 61 Yao 177 Yiddish 329 Yoruba 61 Yukaghir 61, 70 Zapotec 61, 70, 77 Zulu 74, 180, 191 Zyryen 106