Dutch Morphology: A Study of Word Formation in Generative Grammar [Reprint 2020 ed.] 9783112327708, 9783112327692

178 94 11MB

Dutch; Flemish Pages 190 [196] Year 1977

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Dutch Morphology: A Study of Word Formation in Generative Grammar [Reprint 2020 ed.]
 9783112327708, 9783112327692

Citation preview

Studies in Generative Grammar

Editors'

introduction

The goal of Studies in Generative Grammar is to publish those texts that are representative of recent advances in the theory of formal grammar. Too many studies do not reach the public they deserve because of the depth and detail that make them unsuitable for publications in article form. We hope that the present series will make these studies available to a wider audience than has been hitherto possible.

Jan Köster Henk van Riemsdijk

This book has been originally by the Peter de Ridder Press.

¥

FORIS PUBLICATIONS DORDRECHT-HOLLAND

published

G. E. BOOIJ

DUTCH MORPHOLOGY A STUDY OF WORD FORMATION IN GENERATIVE GRAMMAR

LISSE

THE PETER DE RIDDER PRESS 1977

© Copyright reserved No part of this book may be translated or reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the author. Proefschrift Universiteit van Amsterdam 1977 ISBN 90 316 0150 0 Printed in The Netherlands

voor

üerry

ACKNOWLEDGEMENTS

This thesis was written under the supervision of Prof. Dr. W.G.Klooster, University of Amsterdam. His encouragement was of great importance to me and his criticisms of earlier versions of this thesis led to many improvements. To Prof. Dr. H.Schultink, University of Utrecht, who whas so kind to accept the function of co-referent, I am much indebted. His critical comments stimulated me to reformulate several parts of my thesis, and to come to a better exposition of my arguments. I would also like to express my gratitude to my colleague Jaap van Marie with whom I had many lenghty and fruitful discussions on the subject matter of this thesis, and who also gave valuable comment on earlier drafts. I also thank my colleague Sies de Haan for his remarks on Chapter 1. My colleagues of the Department of Dutch linguistics ("Vakgroep Nederlandse Taalkunde"), University of Amsterdam were so kind to grant me a sabbatical leave for half a year, which helped me considerably in finishing this study.

Finally I thank Mrs. Susan Kok for checking and correcting my English.

Amsterdam, August 1977

iv

SAMENVATTING

Deze Studie behandelt de vraag, hoe woordvorming, in het bijzonder die van het Nederlands, beschreven moet worden binnen het kader van de generatieve grammatica. In hoofdstuk 1 wordt uiteengezet, dat woordvormingsprocessen niet verantwoord dienen te worden door middel van syntactische transformaties. Het lexicon van een generatieve grammatica moet naast een beschrijving van alle bestaande woorden ook een verzameling woordvormingsregels bevatten. Deze regels voorspellen enerzijds welke informatie met betrekking tot bestaande gelede woorden redundant is, en geven anderzijds aan, hoe de verzameling gelede woorden van een taal uitgebreid kan worden. Verder wordt de hypothese verdedigd, dat woordvormingsregels niet-transformationele, structuur-bouwende regels zijn. Ook flexie moet beregeld worden door regels in het lexicon. Hoofdstuk 2 behandelt twee aspecten van de structuur van gelede woorden. In de eerste plaats wordt betoogd, dat er twee typen interne grenssymbolen moeten worden aangenomen voor gelede woorden, het morfeemgrenssymbool + en het woordgrenssymbool §. Suffixen worden voorafgegaan door + of

#, prefixen

worden gevolgd door #. De woordvormingsregels specificeren welk grenssymbool is geassocieerd met een affix. Deze hypothese met betrekking tot interne grenssymbolen wordt fonologisch gemotiveerd door een analyse van de accent- en syllabificatiepatronen van gelede woorden en door een analyse van het verdwijnen en ingevoegd worden van schwa's in gelede woorden. Uit de analyse van accent- en syllabificatiepatronen blijkt, dat ook een zekere ordening van accent- en woordvormingsregels noodzakelijk is. In de tweede plaats wordt betoogd, dat gelede woorden interne benoemde haken kunnen bevatten. Dit zijn de haken van de woorden waarvan ze zijn afgeleid. Deze interne haken speien een cruciale roi bij de cyclische toepassing van accentregels. Een analyse van klinker-reductie ondersteunt deze cyclische toepassing van accentregels, en zo ook de aanname van interne haken v

voor gelede woorden en de hypothese dat woordvormingsregels structuurbouwende regels zijn. In hoofdstuk 3 tenslotte, wordt onderzocht door welke typen condities op woordvormingsregels de productivitéit van deze regels wordt beperkt. Er kunnen zowel fonologische, morfologische als syntactische eisen gesteld worden aan de woorden die als basis voor een bepaald woordvormingsproces kunnen functioneren. Sommige van deze restricties zijn regel-specifiek, andere regel-onafhankelijk. Daarnaast bevat de woordvormingscoraponent ook algemene negatieve condities op de toepassing van woordvormingsregels. Beide typen restricties beperken de productiviteit van de individuele woordvormingsregels.

vi

CONTENTS

Acknowledgements

iv

Samenvatting

v

1. The nature of word format-ion rules 1.1. Some preliminary principles of morphology

1 1

1.1.1. Lexical competence

2

1.1.2. The notion 'complex word'

3

1.1.3. Productivity

4

1.2. Transformationalist theories of word formation 1.2.1. Compounds

5 5

1.2.1.1. Lexical deletion

7

1.2.1.2. Output conditions on transformations

8

1.2.1.3. Botha's theory 1.2.2. Derived nominals

9 10

1.2.3. Sentence adverbs

14

1.2.4. Bresnan's Nuclear Stress Rule Hypothesis

16

1.2.5. Summary of conclusions

17

1.3. Lexicalist theories of word formation 1.3.1. Halle's theory of word formation 1.3.2. Some criticism of Halle's theory 1.3.2.1. Some additional properties of WF-rules

17 17 19 20

1.3.2.2. Are word formation rules rules with a global condition?

20

1.3.2.3. The position of the affixes

22

1.3.2.4. The filter

23

1.3.2.5. Summary of conclusions

24

1.3.3. Jackendoff's theory of word formation

24

1.3.4. Constraints on WF-rules: Aronoff's theory

31

1.3.4.1. The 'one affix a rule' hypothesis

32 vii

1.3.4.2. Are all regular w o r d formation processes word-based?

32

1.3.4.3. Morpheme-deleting WF-rules or Truncation Rules? ... 39 1.3.4.4. The bases of WF-rules

44

1.3.4.5. Concluding remarks on Aronoff's theory

45

1.3.5. Summary of conclusions 1.4. W o r d formation and flexion

45 46

1.4.1. The separation of derivation and flexion

46

1.4.2. Halle's theory of flexion rules

49

1.4.3. Flexion rules in Jackendoff's theory

50

1.4.4. Lexical insertion a n d Bresnan's Nuclear Stress Rule

51

1.4.5. Psychological objections?

52

1.4.6. Summary of conclusions

53

1.5. Summary of conclusions of Chapter 1

53

2. Phonological evidence for non-phonological properties of complex words . 55 2.1. The stress patterns of simplex words a n d compounds

56

2.1.1. Stress a n d stress differentiation

57

2.1.2. The Main Stress Rule

58

2.1.3. Exceptions a n d morphological features

62

2.1.4. The Alternating Stress Rule and the problem of minor rules . 67 2.1.5. The stress patterns of compounds 2.2. Affixes and their boundaries 2.2.1. Suffixes and their boundaries

70 72 72

2.2.1.1. Syllabification rules

75

2.2.1.2. Schwa-deletion

81

2.2.1.3. A special class of stress-shifting adjectival suffixes

82

2.2.1.4. Schwa-insertion in complex words

85

2.2.1.5. Boundary weakening

87

2.2.1.6. A first summary and some residual problems concerning the classification of derivational suffixes 2.2.1.7. The introduction of boundary symbols

91

2.2.1.8. The boundary of inflectional suffixes

93

2.2.1.9. Summary of conclusions 2.2.2. Prefixes and their boundaries

97 97

2.2.2.1. Underlying stress in affixes

104

2.2.2.2. The ordering of prefixation and stress rules

104

2.2.3. Summary of conclusions viii

90

107

2.3. Vowel reduction and the structure of complex words

108

2 . 3 . 1 . Vowel reduction in simplex words

110

2 . 3 . 2 . Vowel reduction in complex words

113

2 . 3 . 3 . Brame's Natural Bracketing Hypothesis

114

2.3.4. Vowel reduction and s t y l e d i f f e r e n c e s

117

2.3.5. Summary of conclusions

118

2.4. Summary of Chapter 2 3. Competence restrictions

on the productivity

119 of word formation

rules

3.1. Conditions on the bases of WF-rules

.. 120 122

3 . 1 . 1 . Conditions on the phonological form of bases

122

3.1.2. Morphological conditions on bases

127

3 . 1 . 2 . 1 . Conditions with respect to the presence of s p e c i f i c morphemes i n bases

127

3 . 1 . 2 . 2 . Conditions with respect t o the complexity of bases

128

3 . 1 . 2 . 3 . Conditions with respect to s t r a t a l features of morphemes in bases 3.1.3. Syntactic conditions on bases 3 . 1 . 4 . Summary of conclusions

131 140 -142

3.2. Some other general p r i n c i p l e s of the Word Formation Component . . . 143 3 . 2 . 1 . E x t r i n s i c ordering of WF-rules

143

3 . 2 . 2 . 1 . The organization o f the English Word Formation Component

146

3 . 2 . 1 . 2 . The organization of the Dutch Word Formation Component 3 . 2 . 2 . Recursive a p p l i c a t i o n of WF-rules 3.3. Concluding remarks Notes

147 153 155 157

Bibliography

169

Index

177

ix

CHAPTER 1

THE NATURE OF WORD FORMATION RULES

This chapter gives a general discussion of the nature of word formation rules (henceforth WF-rules or WFR's). In section 1.1. I present some preliminary principles which form the basis of this study of word formation. In section 1.2. I argue that the theory that word formation processes must be accounted for by means of syntactic tranformations must be rejected. Section 1.3. discusses two recently proposed theories of word formation in a lexicalist framework, the theories outlined in Halle(1973a) and Jackendoff (1975). Arguments in favour of Jackendoff's theory will be given. This section also discusses the kind of operations which a WF-rule can perform and supports the hypothesis put forward in Aronoff(1976) that WF-rules can only add one affix to a word. Section 1.4. discusses the delation between derivation and flexion. Finally, section 1.5. gives a summary of the conclusions of this chapter. These conclusions specify the lines along which further morphological research should be conducted.

1.1. Some preliminary principles of morphology In this section I will present three preliminary principles for the study of word formation: (i)

word formation rules form part of the linguistic competence

(ii)

the notion 'complex word' should not be defined beforehand

(iii) productivity is not a matter of degree, but determined by various interacting competence restrictions. 1

1.1.1. Lexiaal competence The aim of a generative grammar of Dutch is to describe and explain the competence which native speakers of Dutch have of their language. This competence comprises the (mostly unconscious) knowledge of the semantic, syntactic, morphological and phonological rules of Dutch, and knowledge of the vocabulary of Dutch. If we try to explain the competence of native speakers of Dutch, this implies that we should relate the structural principles of that competence as much as possible to the general principles which are valid for every language and which are specified in the general theory of language. On the other hand,, the investigation

of Dutch can contribute to the formulation of a more

adequate general theory of language. Therefore, I shall try to relate hypotheses concerning the structure of Dutch as much as possible to hypotheses concerning the general properties of language, especially to that version of transformational generative grammar which has come to be known as the Extended Standard Theory (Chomsky 1970, 1971, 1973) and the standard version of generative phonology as outlined in Chomsky and Halle's The Sound Pattern of English(1968). In this way I will try to contribute to a generative grammar of Dutch which is not only descriptively but also explanatorily adequate. Part of the competence of the native speakers of Dutch is knowledge of the vocabulary of Dutch, i.e. knowledge of all the existing Dutch words and their semantic, syntactic, morphological and phonological properties. I will refer to a linguistic description of this knowledge by means of the term 'the lexicon of Dutch'. Of course no Dutchman knows all the existing words of his language. But linguistics abstracts from the memory-, experience- and other restrictions of the individual language user and describes the lexical competence of the 'ideal native speaker'. In this way it is possible to describe the competence (including the vocabulary) as the common possession of the members of a language community. In the vocabulary of a language an important difference can be observed between two types of words: simplex or mono-morphematic words and complex or poly-morphematic words. Mono-morphematic words show an arbitrary relation between form and meaning. In certain words one can recognize more than one morpheme and therefore

these words have a certain morphological structure.

Traditionally we distinguish between compounds and derived words. Compounds can be considered to have been formed by combining two (simplex or complex) words, e.g. 2

huisdeur (lit. house door) from huis (house) and deur (door), or

wonderstamppot

(lit. miracle hotchpotch) from wonder

stamppot

(miracle) and

(hotchpotch). Derived words can be considered to have been formed by adding em affix to a word or a stem. An example is groenig

(greeny) from groen

(green)

and the suffix -ig. There exists a third type of complex word, the so-called 'synthetic compounds' (Bloomfield 1935:231). These words are the products of both compounding and derivation. For instance, the word tweedehands hand) consists of the words tweede

(second

(second) and hand (hand) and the suffix -s.

The vocabulary of a language is not a closed set of simplex and complex words. The set can be extended by making new simplex words (e.g.

stalf,

prats)

which are in accordance with the morpheme structure conditions of that language, in this case Dutch, and by making new complex words. we can form blauwte

(blueness) from

operatorship) from

medewerker

blauw(blue),

For instance,

medewerkerssohap

(co-

(co-operator).

Because of the word-forming creativity which the speakers of a language possess, an adequate theory of Dutch should not only specify the existing words of Dutch but also all the possible words of Dutch. This study deals with the question of how the knowledge of the existing and possible words of Dutch should be accounted for.

1.1.2. The notion

'complex

word'

We might try to define complex words as words, the meaning of which can be more or less deduced from the meanings of the parts of those words. It is sometimes said that in this way the arbitrariness of the linguistic sign is reduced. However, such a definition is purely semantic, and therefore insufficient, as we shall see. I will not try to give a complete theoretical definition of the notion 'complex word' beforehand. Only after an investigation of complex words in different languages can the general theory of language specify such a notion. Of course we must have some preliminary ideas about what a complex word is, but these ideas can be modified and more articulated in the course of investigation. My working hypothesis about complex words is, that these are words, the meaning of which can be more or less deduced from the meanings of parts of these words, or words to which this semantical description does not apply, but which nevertheless behave in similar ways with respect to phonological and morphological rules. The same idea about the incorrectness of such a theoretical definition beforehand was put forward by Ljung(1975). He points out that we should not define the notion 'morpheme' as 'a mimimal meaning-bearing unit of the 3

language'. A morpheme is "any minimal morphological unit required by the rules set up for the description of language". A theory of the morphological structure of words should not only be able to account for the semantic, but also for the phonological, morphological and syntactic behaviour of words. The verb beveet (to command) for instance does not show any semantic complexity but nevertheless this verb must be assumed to consist of a prefix be- and a morpheme veet because the past participle is bevolen and not gebevoten as would be the case if the word were simplex. The rule for past participles is, that simplex verbs get the prefix ge-, whereas complex verbs with an unstressed prefix do not (Schultink 1973). Another example is the difference between verveel (to bore) and vermis (to varnish). Verveel must be considered to be complex, whereas

vermis is a simplex word since the past participles are

verveeld and gevernist respectively. The same point is made by Aronoff(1976: 14-15).

1.1.3. Productivity An important aspect of the vocabulary of a language is that it can be extended by making new words according to the principles by means of which the existing complex words have been formed. Schultink(1961:113) defines the notion 'productivity' as follows: "Productivity as morphological phenomenon is the possibility which language users have to form an in principle uncountable number of new words unintentionally, by means of a morphological process which is the basis of the form-meaning correspondence of some words they know" [my translation, GEB].1 By using the concept of unintentionality in his definition of the notion 'productivity' Schultink was able to distinguish what later on was called 'rule-governed creativity' versus 'rule-changing creativity' by Chomsky(1965). Word formation rules have to account for that part of rule-governed creativity which manifests itself in the derivation of new complex words. An instance of rule-changing creativity would be the formation of the word auratie from the word curator (curator). Normally words on -or can be derived from words on -ie (agressie+agressor, kwalifiaatie+kwalifiaator).if we reverse the direction of the rule (an act of rule-changing creativity) we could get auratie from curator} gladiatie from gladiator. For instance, the curator of a university could say: "I have the curatie of this university". But to say this, is felt to be intentional: a case of rule-changing creativity. This view of word formation is opposed to that of De Groot's who considered 4

the formation of each new word a diachronic phenomenon (De Groot 1966:126). Two types of morphological processes must be distinguished: productive processes and unproductive ones. An example of an unproductive process is the formation of feminine nouns by adding the suffix -egge to a masculine noun. Dutch has only one word of this type: dievegge (female thief) *• dief (thief). Although there is a clear correlation between the semantic and the morphological complexity of this word, the process is not productive. So we have to specify this WF-rule as unproductive. It only has the function of explaining the structure of the word dievegge. On the other hand, the formation of new compounds for instance is unrestrictedly productive. Schultink(1961) points out that it makes no sense to speak about

degrees

of productivity if we assign this notion a statistical interpretation. It is the task of morphology to investigate by which kinds of conditions on the WF-rules and by which other principles of the word formation component of the grammar the productivity of WF-rules is restricted. The same position is taken by Aronoff(1976:37). In this way, the degree of productivity of a WF-rule can be seen as inversely proportional to the amount of competence restrictions on that WF-rule. This does not amount to the claim that performance factors do not play any role in the explanation of the quantitative productivity of certain WF-rules. But we cannot say anything specific about the role of performance factors before we have investigated

which competence factors restrict the

productivity of a WF-rule.

1.2. Transformationalist theories of word formation In this section

I will discuss three types of word formation for which a

transformational account has been proposed: compounds, derived nominals and sentence adverbs. It will be shown that these theories give rise to unsurmountable problems due to the unpredictable idiosyncratic properties of complex words, the very powerful transformations which are needed to account for word formation and the position of the Nuclear Stress Rule in the cycle of syntactic transformations.

1.2.1. Compounds The first theories of word formation in the framework of generative grammar tried to account for word formation processes by means of syntactic

transformations. Lees(1963) e.g. proposed to derive the compound car thief from the deep structure of The thief steals a car2. Botha(1968:43-47) gives a summary of Lees' theory of compounding, adapting this theory, originally written in the framework of Syntactic Structures (Chomsky 1957) to the Aspects-theory (Chomsky 1965). The sentence John needs a girl friend e.g. has to be derived from the following deep structure (Botha 1968:44): (1)

S

Botha gives the following description of the transformational derivation of the sentence John needs a girl friend-. "Firstly, a relative transformation applies to this S', replacing its leftmost NP (the friend) by Who and erasing the sentence boundary markers, ' tttt' between who and the rightmost NP (a friend) of the highest S. The resulting transform is

#John pres need a friend who pres be a girl#.

The Wh-deletion transformation then applies to the latter transform, yielding the transform #John pres need a friend a girl#. To this transform the Noun Shift transformation applies, deleting the DET (a) occurring immediately to the left of girl, and then shifting girl (actually the N) to the position immediately to the left of friend (actually the rightmost N of the highest S). The resulting transform is #John pres need a girl friend#. Next, a transformation shifts pres to the position immediately to the right of need (i.e., shifts AUX to the right of V), resulting in the transform #John need pres a girl friend#. Finally, two transformations apply, inserting boundary markers into the 6

appropriate positions. Inter-word boundaries are inserted first to yield the transform #John # need pres If a it girl friend # . The second of these transformations insert[s] the appropriate intra-word boundary markers '-', yielding the transform # John # need-pres § a # girl-friend #."(p. 45) However, Botha does not give any independent evidence for the different rules necessary for the derivation of the compound girl-friend. Lees gives the following semantic arguments for a transformational treatment of compounds (I follow Botha's summary here): (i)

by means of this derivation it is specified that friend and girl in girl-friend

(ii)

stand in a subject-predicate relation,

if a compound is semantically ambiguous, this can be explained by assuming more than one deep structure for that compound,

(iii) the difference between e.g. wind-mill and flour-mill can be explained by assuming different deep structures for these compounds: Wind powers the mill (mill is object) versus The mill grinds flour (flour is subject). So all the arguments of Lees' concern the task of grammar to specify the meaning of a compound (partially) on the basis of the meaning of its parts. But these arguments are not very compelling seen in the light of the present state of the theory. Semantic regularities can also be expressed by other types of rules, e.g. rules in the lexicon. Moreover, Lees' theory gives rise to huge problems: the abundant deletion of lexical material, and the necessity of very powerful conditions on transformations, as will be shown in the next subsections.

1.2.1.1. Lexical deletion Lees' theory of compounding needs quite a lot of transformations in the syntactic component of the grammar which are allowed to delete all kinds of lexical material. In deriving the word

oar thief for instance, the verb

steal has to be deleted, in deriving the word wind mill the verb

power , etc..

Note that Botha only gave an example of a relatively simple derivation, using the transformation Relative Clause Reduction which has perhaps independent syntactic motivation. These kinds of deletion transformations were also proposed by Meys(1975) who wants to derive e.g. sea breeze from breeze from the sea by deleting from the and reversing the nouns. 7

However, serious objections can be raised against the abundant deletion of lexical material. It is of great importance to constrain the power of transformations in order to arrive at an adequate characterization of the notion 'natural language', the principal aim of a general theory of language (Bach 1971). One of the constraints on the power of transformations is the condition on the recoverability of deletions (Katz and Postal 1964) which says that terminal elements of a structure can only be deleted if they are identical with another terminal element of that structure, if they are PRO-elements, or if they are specified in the structural description of the transformation. So in principle Katz and Postal admit the deletion of terminal elements. They refer to the condition that deletable elements must be specified in the structural description of the rule as condition But this condition

(i) (Katz and Postal 1964:81).

(i) admits far too many deletions, as Den Besten(1974:

12) points out convincingly: "It can be seriously questioned whether condition

(i) of Katz and Postal

is really sufficient for stating which deletions are permitted, apart from deletions under identity and deletions of PRO-elements. In this way the theory permits that each constituent disappears from a sentence structure, if only the requirement that the terminal elements of that constituent are mentioned in the structural description of the relevant transformations has been met.[...] In extremis, the theory predicts that every word and therefore every sentence can be deleted, which implies that every sentence could be realized by silence, instead of being spoken. We cannot accept such a prediction." 3 [my translation, GEB] Den Besten(1975) is an attempt to constrain this type of deletion. So the first problem in a transformational treatment of word formation is that it blocks the possibility of further constraining the notion 'transformation1,and consequently the notion 'natural language'.

1.2.1.2. Output conditions on transformations The second serious difficulty is the fact that we would make all kinds of wrong compounds by means of the transformations suggested above. The base component could generate for instance the structure [the friend

[ the friend

steal a girl ]] which might result in the compound girl friend through the application of the steal-deletion

transformation. But now the wrong meaning

'friend who steals a girl' is assigned to the compound girl friend. 8

How can this be prevented? By specifying all the existing compounds in the lexicon, by specifying by means of which transformations they have been derived and by formulating a general condition on all compounding transformations which says that a transformation is blocked, if there is a compound in the output of the transformation which is not positively specified for that transformation. This is a very strong type of condition which we would like to eliminate from the grammar.

1.2.1.3. Botha's theory Botha, who accepts Lees' transformational theory

of compounding, modifies

Lees' theory in certain respects. He concludes (Botha 1968, Chapter 5) that compounds in Afrikaans, a language spoken in South Africa and historically related to Dutch, have four types of idiosyncratic properties which are not predictable by rule: (i)

they may have an unpredictable internal link phoneme

(ii)

they may have a deviating stress pattern

(iii) they may have idiosyncratic meaning aspects (iv)

they may have a metaphorical meaning.

As for (i), Botha proposes the assumption of a phonological dictionary, in which all the existing compounds with a link phoneme must be listed. He does not make a proposal concerning the account of the three other idiosyncratic aspects of compounds. Before the application of the phonological rules, the syntactic surface structure passes the phonological dictionary. The dictionary adds the correct link phoneme to every compound listed in the dictionary, if it appears in a syntactic surface structure. Apart from the fact that Botha's proposal does not account for the idiosyncracies (ii)-(iv), it fails because the 'regular' compounds without a link phoneme are not specified in the dictionary. This implies that the general condition mentioned above on the output of compounding transformations cannot work properly because it needs information about all the existing compounds, not only about those with a link phoneme. For compounds with an idiomatic meaning it is even impossible to be listed in the dictionary, because due to its position between the syntactic and the phonological component the dictionary would change the meaning of sentences in that case, and this conflicts with the principle of the Extended Standard

Theory that the functional meaning of a sentence is determined by

deep structure (cf. Jackendoff 1972). Therefore, Botha's analysis of compounds

inevitably leads to the conclusion that all the existing compounds have to be specified in the lexicon. But even if we would do this, the objection remains valid that we would allow very powerful

kinds of transformations by which the generative power of

the grammar increases enormously. Therefore, seen in the light of the linguistic discussions in the last few years about the power of transformations (e.g. Peters and Ritchie 1969;1973, Bach 1971) it is rather strange that Botha still appears to prefer a transformational theory of compounding, as he explicitly states in Botha(1974:9): "Although it had definite limitations, the only reasonably well-outlined and partially justified theory of compounding is an extended version of Lees' transformationalist theory."1* 1 conclude therefore that a transformational theory of compounding is inadequate.5

1.2.2. Derived nominals A second type of word formation for which a transformational treatment has been proposed are the so-called derived nominals, nouns derived from verbs. The word weigering (refusal) in Ve dokter zijn weigering van het aanhod (The doctor's refusal of the offer) must, according to this proposal be derived from the structure [

[ de dokter weiger het aanbod ]- s/ —

y

As predicted by convention (28), rule (30) will only apply before the suffix +y (democrat-democracy; president-presidency) which is stress-shifting, and not before the stress-neutral suffix #y (ch5colat-chócotaty;bràt-bvàtty) .But this independent evidence is rather questionable because dictionaries of English unanimously give the pronunciation presidency. Furthermore, there are also i/s-alternations in pairs such as adequate-adequacy; consistentconsistency where nevertheless the stress does not shift. 74

For Dutch independent evidence for the assumption of two suffix boundaries is provided by the phenomenon of devoicing of obstruents within words. This well-known rule is often (31) [-son] ->- [-voice] /

formulated as: #

Since this rule applies before a word boundary, it also applies before Class II-suffixes within complex words, as can be seen in the phonetic forms of words such as voodaehtig, dwaasheid, hoedloos and halfling listed in (27). However, it has been claimed by several linguists that final devoicing of obstruents does not apply before word boundaries but before syllable boundaries, i.e. devoicing is conditioned by a phonological syllable boundary symbol, not by a grammatical word boundary symbol. This has been argued for German by Andersen(1972), Hooper(1972) and Vennemann(1972), and for Dutch by Kooij(1976) and Gilijamse(1977). Gilijamse points out that in the Dutch pronunciation of loanwords such as Sydney and Cambodja where no internal word boundaries are present, nevertheless devoicing is applied to the internal [d]: [sitni], [kambotja] because these words have the syllable structures Syd$ney and Cam$bod$ja respectively (where $ is the symbol for the syllable boundary). The claim that final devoicing is conditioned by the syllable boundary instead of the word boundary seems to imply that devoicing of obstruents is no longer an independent argument for the assumption of two boundaries for suffixes (cf. Schultink 1977) . But now we face a new problem: how can the syllable boundaries within a word be predicted correctly? In the next section it will be shown that for a correct account of the syllable structure of complex words, the distinction between two types of boundaries is indispensable.

2.2.1.1. Syllabification rules Hooper(1972), Vennemann(1972) and Gilijamse(1977) argue that the notion 'syllable boundary' is necessary for a correct account of certain phonological processes. Therefore, this notion must be defined formally in generative phonology. Hooper(1972) proposes to define the notion 'syllable boundary' by means of a set of syllable boundary insertion rules which insert syllable boundaries between segments and which apply persistently throughout the phonological component, changing the position of the syllable boundaries if the sequences of segments change. According to her, these rules are partly 75

universal, partly language-specific. She proposes the following set of disjunctively ordered rules of syllable boundary insertion: (32) a

0

$ / [+syll] —

[+syll]

b

0

$ / [+syll] —

[-syll] [+syll]

c

0

$/

d

0

$ / [+syll] [-syll] 0 —

[+syll] [-syll] Q —

[-son]

[+syll]

[+cons] [-cons]0[+syll]

She claims that these four rules represent "the universal definition of the phonological syllable, and that languages may or may not have additional late rules that define a language-specific, phonetic syllable" (Hooper 1972:536). The rules (32a,b) indeed make correct predictions concerning the syllabification patterns of Dutch simplex words and words with Class I suffixes. Rule (32a) inserts a $ between two vowels, e.g. in aha$OS} Bo$az. Rule (32b) inserts a $ before the consonant if the vowels of the two syllables are separated by only one consonant: va$der, la$ohen, hui$ahel, ba$ker3 ba$(k)ker

etc.. Rule (32c) asks for more discussion. This rule states that

the syllable boundary is inserted before two non-vowels at most followed by a vowel. But then the second non-vowel must be a [j], [w] , [l] or [r], and the first an obstruent. Otherwise the syllable boundary is inserted before the sequence non-V,V. For many Dutch words, this rule predicts correct syllabification patterns, e.g. for the following words:

(33)

moef$ti, ast$ma, korn$werd, tar$zan, ti$flis, na$bloes, ta$bleau,a$pril, ko$pra, neu$tron, a$trix

But in some cases the syllable boundary must be inserted in a position different from that predicted by rule (32c). In the first place, a syllable boundary must appear before the consonant cluster st, e.g. in pa$stoor, pa$stei, ek$ster. The possibility of sucti exceptions was noted by Hooper (1972:535). She proposed to add a set of possible language-specific, exceptional rules to the set of universal syllabification rules. Each language may have one or more of these exceptional rules. A second class of exceptions in Dutch is formed by the combinations of a non-sonorant and a glide [j] or [w]. In such cases, the syllable boundary must be inserted between the non-sonorant and the glide, at least if this cluster is preceded by a lax vowel:

76

(34) cam$bod$ja, to$rad$ja, at$jar I will not present a complete formal account here of these exceptional cases since this does not concern the main argument in this section, i.e. the role of grammatical boundaries in syllabification rules. The fourth rule, rule (32d), has to account for all those cases not accounted for by the previous rules. Since the rules (32) are disjunctively ordered and rule (32c) "takes care of any clusters that might contain obstruents, we are left only with clusters of nasals, liquids and glides. Here we find the following: (a) If there are two nasals, two liquids, liquid plus nasal, or nasal plus liquid (i.e. any two consonantal segments), then the $-boundary separates them [...] (b) If there is one consonantal segment, it begins the syllable [...]" (Hooper 1972:536). The statement sub (a) is correct for Dutch: Am$ram3 Her$man, Hel$ma etc., but the statement sub (b) is invalid for Dutch: Ar$jan, not A$rjan, ran$ja, not ra$nja} Hol$werd, not Ho$lwerd. So the fourth rule must be corrected by a specific rule for Dutch, since Hooper's rules are claimed to be universal. Alternatively, we might restate rule (32d) as follows: (32d)1 0

$ / [+syll] [-syll]0 —

t-syll] [+syll]

That is, we could claim that the fact that e.g. in Spanish the syllable boundary is located before Zj as in a$lyent$tO is language-specific, and must be accounted for by an additional rule. Finally a rule is needed for the insertion of syllable boundaries at the beginning and end of words. Hooper proposes the following rule: (35)

0

-

$ / {" [+ segment]} L+ segment] fttt

i.e., syllabification is sensitive to grammatical word boundaries. It is not my aim here to present a complete account of all the syllabification details of Dutch. The rules discussed so far suffice to make it possible to discuss the role of grammatical boundaries in the syllabification of complex words. Since only the word-internal syllable boundaries play a crucial role in the argument of this section, I will omit the syllable boundaries on the left and the right side of words represented with their syllabification patterns. Let us now consider in more detail the syllabification patterns of 77

complex words. As far as words with Class I-suffixes are concerned, their way of syllabification is identical with that of simplex words. If we assume that Class I-suffixes are preceded by the formative boundary + , this is predicted by convention (28), the convention that phonological rules are not blocked by +. Consequently, no syllable boundary will appear before the suffix, but a $ will be inserted before one or two non-vowels before the suffix, since all Class I-suffixes are vowel-initial. Therefore, a final obstruent of a basis word will never be devoiced before a Class I-suffix, since it is not immediately followed by a syllable boundary. In (36) I list some words with their syllable boundaries inserted (in Dutch spelling, consonant symbols are geminated to indicate the lax nature of the preceding vowel): (36)

sy$no$daal, mo$ha$(m)med$daan, mil$jar$dair, am$ba$(s)sa$deur

Compounds on the other hand behave quite differently with respect to syllabification. As predicted by the presence of an internal word boundary, the rules syllabify the parts of the compounds independently of each other since they do not apply across word boundaries. Consequently, we get such minimal pairs as: (37) bal#kanker - bal$kan$ker balk#anker - balk$an$ker

(testicle cancer) (beam brace)

vrouwelnaaier - vrou$we$naa$ier (woman fucker) vrouwen#aaier - vrou$wen$aa$ier (women caresser) (examples from Zaalberg 1976) Such a minimal pair

Bye-trait.

is also given by Stanley(1973:185): night-rate

versus

It is clear from these examples that syllabification is blocked

by internal word boundaries, and that rule (38), a modified version of rule (36) must be assumed: (38)

*

-

$/{

#

t+ se ment 9 ]} [+ segment] tt

Rule (36) is modified for two reasons. In the first place, there is no reason to assume that compounds have two internal boundaries (cf. section 2.2.1.7.). Furthermore, according to the SPff-conventions for boundary insertion, it is not true that every word is flanked by two word boundaries (cf. section 2.2.1.7.). 78

It appears now that complex words with Class II-suffixes such as -aohtig, -loos and -Zing

behave exactly like compounds with respect to syllabification,

i.e. syllabification is blocked by the suffix boundary and syllable boundaries must be inserted between the basis word and the suffix: (39)

rood$ach$tig

[rotaxtSx]

ge$dach$tig

[ySdaxtax]

rib$loos

[riplos]

ta$bleau

[tablo]

half$ling

[halflin]

ti$flis

[tiflis]

compare:

This difference between the two classes of suffixes with respect to syllabification is also pointed out by BasbfSll (1975:127) : "The syllable boundary between two vowels (with intermediate consonants) belonging to different morphemes always occurs at the morpheme boundary if it is

but not (necessarily) if it is + (in that case the location

of $ depends on the sequence of segments)." Speakers of Dutch might claim that they can syllabify words such as graafloos and half ling as follows: graa$floos, hal$fling. But this is no counterevidence: at the level where the rule of devoicing of obstruents applies, the syllable boundary must appear where the suffix boundary is; otherwise we cannot explain why final devoicing has been applied in these words. This situation must be interpreted as follows: the syllabification rules given above define the notion 'phonological syllable boundary-. But in casual speech the strenght of the boundaries can be weakened and consequently the syllabification becomes later on like that of simplex words: rule (32c) predicts a $ before the sequence fl. So in casual speech the phonological syllabification of a word might differ from its phonetic syllabification. The other Class II-suffixes (those mentioned sub (b) in (27)) are indeterminate in this respect since according to the syllabification rules a syllable boundary can e.g. never appear before a cluster of a consonant and the [d] of -dom

or the [h] of -heid and therefore automatically appears at

the boundary between basis word and suffix. That is, it cannot be decided for these suffixes by means of syllabification patterns whether they must be assumed to be preceded by # or by +. But since one of the two must be chosen, and since complex words with these suffixes show the same stress-behaviour as those with the three Class II-suffixes mentioned above (i.e. they are X I 1 1 stress-neutral: beznvloed-beznvloedbaar, bedaehtzaam-bedaohtzaamheid, arbeider-arbeiderdom, dominee-dbmineetje, dbminee-dbmineeschap, arbeid79

avbeidster ,. we cannot but classify them as #-suffixes. This is a perfectly reasonable method: it is clear that we have to divide suffixes into two classes with respect to their stress-behaviour. This classification is performed by means of the assumption of two types of boundaries. For some Class II-suffixes this assumption is independently motivated by means of syllabification phenomena. Therefore, we assign without any extra cost (since one of the two boundaries must be mentioned) the word boundary # to the other Class Il-suffixes which are indeterminate with respect to choice of boundary as far as syllabification is concerned. In this way their stress-neutrality is predicted without invoking any other formal mechanism. Moreover, it will be shown in Chapter 3 that, given the classification of suffixes proposed here, it is possible to formulate a general principle concerning the ordering of suffixes in multiply complex words. The fact that e.g. the word wendbaar (manoeuvrable) has the phonetic form [wendba:r] instead of Iwentbair] is no problem, neither for a theory of word-final devoicing nor for a theory of syllable-final devoicing: the [d] of the basis word Wend has been devoiced but has been voiced again later on by a rule of voice assimilation.5 For the suffixes -sel and

-te neither syllabification nor stress pattern

arguments are available to determine which boundary must be assigned to them. There are no examples of complex words with these suffixes where the stress could possibly shift. Dutch has the word aanhangsel (appendix) but we cannot say that this word is derived from a compound aanhang (compare note 4), and so we cannot conclude that -sel is stress-neutral. But there are more indirect arguments for assiging the boundary # to these two suffixes: this assumption makes correct predictions about the possible suffix-sequences in which they can

occur (cf. section 3.2.1.2.) and it can be maintained now that all

+-suffixes are vowel-initial, i.e. the MS-condition for +-suffixes proposed in section 1.3.2.3. can be maintained. Another suffix of Dutch, -evd ([3rt]) was not mentioned so far. It must be assumed to be preceded by +, as can be concluded from the syllabification patterns of words with this suffix, e.g. vies—oie$zevd (dirty person). Nevertheless this suffix does not bear stress because its vowel is the schwa. No examples of words where the stress of the basis word could have shifted rightward are available. Furthermore, this suffix has an allomorph +aar>d, e.g. geestig-geestigaard, wveed-wreedaard. This allomorph can be derived by means of a rule which changes the schwa into an [a] (cf. Smith 1976).6

80

2.2.1.2.

Schwa-deletion

If a complex word is derived from a basis word by means of a Class I-suffix, and the basis word has a final schwa, this schwa is always deleted (the word-final e in Dutch spelling always indicates a schwa): (40)

+aal

synode

synodaal

+aan

Syracuse

Syracusaan

+air

elite

elitair

+eer

code

codeer

+ein

Rome

Romein

+esk

ballade

balladesk

+eur

ambassade

ambassadeur

+eus

rancune

rancuneus

+iek

muze

muziek

+ieus

mode

modieus

+iseer

periode

periodiseer

+ist

falange

falangist

+iteit

stupi de

stupiditeit

On the other hand, the schwa never disappears before a Class II-suffix, even not if the suffix is vowel-initial: (41)

#achtig

zijde

zijdeachtig

jute

juteachtig

groente

groenteachtig

The rule of schwa-deletion should therefore be formulated as follows: (42)

[3]

0 /

+ [+ syll]

The condition that the first segment after the formative boundary must be a vowel is not strictly necessary given these data, since all the derivational +-suffixes are vowel-initial. But in section 2.2.1.8. it will be shown that the schwa also disappears before vowel-initial inflectional suffixes, but not before inflectional suffixes with an initial consonant. Therefore, the vowel in the context of (42) is indispensable. The difference in behaviour with respect to schwa-deletion between Class I- and Class II-suffixes can thus be explained by their different 81

boundaries. Therefore, schwa-deletion supports the assumption of two types of boundaries for suffixes. The only other description of the restrictions on schwa-deletion that might seem to be possible would be the addition of the condition that the vowel of the suffix must bear [l stress] (-aahtig bears [2 stress]). But such a condition would not explain why prefixes with a schwa such as be- and gedo not lose this vowel in words such as beacon and geacht

where the vowel of

the next syllable bears primary stress. If schwa-deletion is made dependent on the presence of a boundary + on the other hand, this fact can be explained since prefixes are always followed by the word boundary # as will be shown in section 2.2.2.. Moreover, the schwa is also deleted before vowel-initial inflectional suffixes which bear no stress at all. This leads to the conclusion that the rule given above is the correct one.

2.2.1.3. A special class of stress-shifting adjectival suffixes In section 2.2.1. I mentioned the stress-neutral adjectival suffixes -achtig, -baar and -loos

and the stress-shifting adjectival suffixes -aal, -air, -esk,

-ief and -(i)eus). Apart from these two classes of adjectival suffixes there is a third class of stress-shifting adjectival suffixes: they form complex adjectives with main stress on the final vowel (except the schwa) before the suffix. This class comprises the suffixes -ig, -isch, -zaam, -(e)lijk and -eldos. Note first that their stress-shifting behaviour cannot be explained 2 1 by assigning them the boundary +. For instance, the adjective godsdienstig 1 2 (religious), derived from the compound godsdvenst (religion) would get an incorrect stress pattern since the internal word boundary blocks the application of the MSR on the final cycle: (43)

[[[gods]N#[dienst]N]N+ig]A l

St

cycle, MSR

2 n d cycle, CR 3

rd

cycle, MSR CR

1 1

2

blocks 1

Nevertheless, the suffixes -ig ([9x1) and

3 -isch ([is]) must be assigned the

boundary + since the syllable boundary is not located at the boundary between basis word and suffix: goe$dig, lo$gisch etc.. On the other hand, assumption of the boundary + for the suffix -lijk ([ 13k]) would fail to explain why the 82

syllable boundary is located at the suffix boundary: graafilijk, lief$lijk etc.. This position of the syllable boundary must be assumed because syllable-final devoicing has devoiced the [v] of /graav/ and /liev/. Therefore, the suffix -lijk must

be assigned the boundary

This boundary also makes

the correct prediction that this suffix can be preceded by the link phoneme [8] (cf. section 2.2.1.4.). The suffix -eloos is a variant of the suffix -loos, the variant with the link phoneme. Therefore, this suffix must be assigned the structure effloos. Finally, it must

be explained why the suffix -isah. does not bear stress

although it is a strong cluster, and preceded by +. It is clear, therefore, that apart from a classification of suffixes by means of two types of boundaries

another classification is necessary in order

to account for the behaviour of these adjectival suffixes. I present here

some data

concerning the stress-behaviour of these

adjectival suffixes. Note first that the adjectival suffixes -aohtig3 -baar7 and -loos are stress-neutral: (44)

afgod

afgodachtig

sleutelbloem

sleutelbloemachtig

beantwoord

beantwoordbaar

beoordeel

beoordeelbaar

godsdienst

godsdienstloos

uitzicht

uitzichtloos

In the following examples the suffixes shift the main stress of the basis word to the final strong cluster before the suffix, whether the basis word contains internal word boundaries or not: (45)

1 2 hoofdzaak 1 2 hartstocht

2 1 hoofdzakelijk 2 1 hartstochtelijk

aartsvader

aartsvaderlijk

voorbeeld 1 2 hartstocht

voorbeeldeloos hartstochteloos

1 v, v,2 herberg

2 2 u vl herbergzaam

2

2

1

2

1

2

83

1 2 arbeid

2 1 2 arbeidzaam

afgqd 1 proza

afgodisch 2 1. . prozaisch

1 2 dienstplicht 1

2

aandacht

2 1 dienstplichtig

2

1

aandachtig

From the data in (44) and (45) it must be concluded that a special stress rule is necessary for words containing one of the suffixes mentioned in (45). This rule must be ordered disjunctively before the MSR and the CR: (46)

V

-»• [1 stress] / [ x

c

0

^

C

0

)

Z

^A

Conditions: X may contain § Z e { #zam, O)#18k,3#loz, + 3x, +is } If we assign the suffixes -eloos and -zaam

[l stress] in their underlying

form, the following correct derivations are possible: (47)

[[arbeid]„#zaaml V

1

S

cycle, MSR ASR

A

1 1

2

2n-

I- stress] / [c. U

C § C V U Ur, . 1 11 s t r e s s ]

Condition:

CQVCQff e {

X

1 V

:>nd3r#, o v 8 r # , her# }

The boundary # o f these p r e f i x e s w i l l be argued f o r b e l o w . I assume u n d e r l y i n g [ l s t r e s s ] and n o t . [ - s t r e s s ] only occur i n v e r b s .

s i n c e the s t r e s s a l t e r n a t i o n s i n these p r e f i x e s

In nouns, the p r e f i x her-

o f allomorphy has been c a l l e d 352):

i s always s t r e s s e d . This type

'morpholexical v a r i a t i o n '

by B l o o m f i e l d ( 1 9 3 9 :

" d i f f e r e n t words o f t e n c o n t a i n m o r p h o l o g i c a l elements o f

i d e n t i c a l meaning but d i f f e r i n g somewhat i n f o r m " . I w i l l a rule a morpholexical r u l e ,

The n e g a t i v e p r e f i x on-

following

otherwise

t h e r e f o r e c a l l such

Anderson(1974).

can be added t o both nouns and a d j e c t i v e s .

If

it 101

is attached to nouns, it bears primary stress. In adjectives, however, the basis word often bears main stress: (83)

,

1 2 (nonsense) onzin

zin (sense) genoegen (pleasure) mens (human)

ongenoegen (displeasure) 1 2 onmens (brute)

aardig (kind) 1

onaardig (unkind)

gewoon ( c oiranon ) heus (polite)

, ^

1

ongewoon (uncommon) onheus (impolite)

With respect to adjectives, things are rather complicated, as Van Den Berg (1970) shows: the prefix on- does not bear primary stress in adjectives in predicative position or in metalinguistic use (Bi¿ is onaardig. He is unkind. Bet woord 'onaardig', The word 'onaardig'). But in certain syntactic positions the main stress can also be on the prefix, e.g. in adnominal position: een ongewone jongen (an uncommon boy), een onbekwame burgemeester (an incompetent mayor). A complete analysis of the prefix on- therefore depends on a correct syntactic analysis of the different constructions where the stress can shift. Since this does not affect the hypothesis that prefixes are always followed by a word boundary, I will not go deeper

into this matter here. I simply assume

underlying [l stress] for this prefix. In certain syntactic positions, it must be destressed in adjectives. Now the question must be answered which boundary or boundaries must be associated with prefixes. My hypothesis is that they all have the boundary #. This was also proposed for Danish by Basbszill (1975) . If the prefixes in (76b) are specified with [l stress] in their underlying form, we can derive the correct stress patterns as follows: (84)

[be#[dijk]N]v . st 1 cycle, MSR 2 n d cycle, MSR CR

[aarts#[vader]N]N

1 blocks blocks

1 blocks 1

2

There is independent evidence for the assumption of # as the prefix boundary: 102

the syllabification patterns of complex words with prefixes

where the final

segment of the prefix is a consonant and the first segment of the basis word a vowel. In C85) some examples of the syllabification patterns of such words are listed:

examp le:

syllabification pattern:

veras

ver$as

ontaard

ont$aard

aartsachterlijk

aarts$ach$ter$li jk

oeroud

oer$oud

wanorde

wan$or$de

herassureer

her$a$ (s)su$reer

onaardig

on$aar$dig

hyperemie

hy$per$e$mie$

superaardig

su$per$aar$dig

omarm

om$arm

doorader

door$a$der

overeet

o$ver $eet

In all these examples the syllabification rules are correctly blocked by the prefix boundary since otherwise these rules would assign the final consonant of the prefix to the first syllable of the basis word. It might be that in very fast speech

speakers of Dutch syllabify a word

such as ontaard as on$taavd. This can be explained by the assumption that in casual speech the grammatical boundary tt is weakened to a + or erased, since in such a style of speech the syllabification rules can even apply across the boundaries between words, e.g. hij$he$pe$mal (hij heb em al). Another apparent counterexample is the verb herinner (to remind, lit. to re-internalize). This verb must be considered complex, because its past participle is herinnerd and not *geherinnerd (cf. Schultink 1973). But the syllabification pattern is always her$i$(n)ner. We see here again a case of boundary weakening: the word boundary has been weakened to a morpheme boundary due to a loss of semantic transparency of the complex word. Boundary loss can be observed in the word abortus (abortion). Most speakers of Dutch consider this word simplex and conseqently syllabify it as follows: a$bor$tus. Only those Dutch speakers who are well aware of the original (Latin) structure of this word, syllabify it as ab$or$tus. More independent evidence for the prefix boundary # comes from 103

schwa-deletion: in the words beaam, bearbeid, geacht, geaard with the prefixes b3- and gd-

respectively the rule of schwa-deletion does not apply to the

final segment of the prefix which occurs before a vowel. This is predicted by the boundary # since, as we saw in section 2.2.1.2., schwa-deletion requires a + in its context.

2.2.2.1. Underlying stress in affixes With respect to some suffixes and prefixes I made use of the mechanism of underlying stress in order to predict the correct stress patterns of complex words derived by means of these affixes. The reason for this assumption is that otherwise we would be forced to arbitrarily assign these affixes , and to extend

labeled brackets to

the domain of application of the MSR to affixes

with labeled brackets. But this extension of the use of labeled brackets for non-lexical morphemes is not motivated otherwise. Therefore, it is preferable to assign underlying stress to these affixes. Furthermore, the assumption of labeled brackets for affixes would conflict with the Natural Bracketing Hypothesis put forward by Brame(1974). Brame's hypothesis will be more extensively discussed

in section 2.3.2.3.. It says that only those parts

of complex words that can also

occur as independent words with the same

meaning as when occurring in complex words, can be bracketed. The fact that some affixes get underlying primary stress reflects the fact that native speakers of Dutch have to learn that certain #-affixes bear stress, whereas others do not. The prefix ont- e.g. does not bear stress whereas ondoes, although both of them are strong clusters. Only in those cases where the affix-vowel is a schwa, it is predictable that the affix is unstressed because a schwa never bears stress.

2.2.2.2. The ordering of pre fixation and stress rules The prefixes listed in (77) impose a general condition on the phonological form of their basis words: the first syllable of the basis word must bear primary stress: e.g. doorloop, doorseharrel, doorkevkelijk3 dooreoho but *doorcomponeer. This condition can also be inferred from the data given by De Vries(1975). Note, incidentally, that there is a conspiracy of this stress condition on prefixation and rule (80): in both cases the effect is that an unstressed prefix is followed by a stressed syllable. The fact that the WF-rules for these prefixes need information about the 104

stress patterns of basis words implies that the word stress rules (MSR, ASR, rule (46) and the French and Latin stress rules) must apply before prefixation but the Compound Rule afterwards. Perhaps one might be inclined to think that mentioning a stress condition in these WF-rules could be avoided by stating a requirement on the segmental structure of their basis words, e.g. 9C C

x =

q V C q ( O ( # ) C ) 0 } ) , where x stands for the sequence of segments and non-

segments of the basis verb. But this would be an unnecessary complication of the WF-rules since this segmental structure is a duplicate of the context of the MSR. Moreover, this condition could not explain the correctness of e.g. 1 2 . 12. doorarbeid derived from the verb arbeid. The stress pattern of this verb is derived by means of the MSR and the ASR. The fact that arbeid is a correct basis for the prefix door-

would be expressed very clumsily, if we would

mention segmental structure in the WF-rules. This leads to the following hypothesis concerning the linear ordering of rules in the lexicon: (86)

Suffixation Rules (except a few including the flexion rules) Word Stress Rules Prefixation Rules Compound Rule

These rules are linearly ordered, but apply cyclically. As far as stress rules are concerned, this will be motivated extensively in section 2.3.. Cyclic application of .WF-rules correctly predicts that suffixation can apply after

prefixation, e.g. mens -*• oermens •+• oermenselijk. The ordering hypothesis proposed here embodies the claim that each application of a suffixation rule creates a new cycle for either the word stress rules or the Compound Rule, whereas prefixation rules only create a new cycle for the Compound Rule!" Note that the WF-rules are optional, i.e., a word can pass a block of WF-rules without undergoing affix-attachment. The word doorspzegelbaar e.g. will now be derived as follows:11 (87)

1 s t cycle:

[spiegel]^ MSR Prefixation CR

nC

2 ^ cycle .-Suffixation MSR CR

.1 [door#[spiegel] ] (blocks because no # follows the vowel with [ [door# [spiegel] v]v#baar]ft

1

stress)

(blocks) 1

2 105

Another argument for this ordering hypothesis can be found in the stress patterns of complex adjectives, e.g. adjectives with the prefixes aavts- and oev~, two intensifying prefixes which can be attached to adjectives: (88)

1 2 1 2 1 2 aavts: aartsdom (very stupid), aartsgierig (very avaricious), aartsgoed 1 2 (very good), aartsnieuwsgierig (very curious) 1 2 1 2 1 2 oev: oergezellig (very cosy), oergezond (very healthy), oerdom (very 1 2 stupid), oerleuk (very funny)

Both prefixes can also be attached to nouns but then they sometimes have idiosyncratic meanings. (89)

1 2 1 2 aartsbisschop (archbishop), aartsvader (patriarch) 1 2 1 2 oermens (primitive man) , oerbegm (the beginning of the world)

Due to the fact that aavts- and oev- can be attached to both nouns and adjectives, the following derivations of the adjectives aavtsvadevlijk and oermenselijk are possible: (90)

vader (father) ->- aartsvader (patriarch) •+ aartsvaderlijk (patriarchal) vader (father) mens (man)

vaderlijk (fatherly) •*• aartsvaderli jk (very fatherly)

oermens (primitive man)

oermenselijk (typical of primitive men)

mens (man) -*• menselijk (human) -+• oermenseli jk (very human) As a consequence of these two possible ways of derivation, two words aavtsvadevlzjk and two words oevmenselijk arise, both with different meanings. The difference in morphological structure is reflected by the stress contours: if the adjectives have the meaning component 'very', the main stress is on the prefix, but otherwise it is on the final strong cluster before the suffix. But if we would assume that prefixation precedes the word stress rules, these differences with respect to stress between the two interpretations of these adjectives could not be predicted: both would be assigned the stress contour 2

1

2

1

21: aavtsvadevlijk, oevmenselijk:

106

[[âarts#[vader] ] #lijk] (patriarchal) N N A 1_

(91)

1 s t cycle, „nd 2 cycle, 3

1

cycle, (46)

[aartsl[[vader] #lijk] ] (very fatherly) N A A

st 1 2

2

cycle, MSR nd

cycle, CR

3 r d cycle, (46) Since, as I argued in section 2.2.1.3., rule (46) must be disjunctively ordered before the CR, this rule will apply on the third cycle, giving the wrong stress contour 21 for the adjective with the meaning 'very fatherly 1 . The problem is solved by the ordering hypothesis proposed above. We then get the following correct derivation of the stress pattern 12:

(92)

l S t cycle:

[vader] N 1

MSR 2

cycle: suffixation

[[vader] N #lijk] A

rule (46) prefixation

1 [aarts# [ [ vader j^lijk]^]^

CR

1

2

Therefore, the stress patterns of adjectives oer-

with such prefixes as aarts- and

support the hypothesis that prefixation rules must be

ordered after the

stress rules.

2.2.3. Summary of conclusions In this section (section 2.2.) I tried to give an answer to the question: which kind of morphological boundaries must be assumed for Dutch? I adduced several kinds of evidence which led to the conclusion that word-internal boundaries are indeed indispensable for an adequate account of the phonological properties of complex words. Suffixes must be assumed to be preceded by + or # and all prefixes to be followed by #. The phonological evidence in favour of these claims concerned (i) the stress patterns of complex words, (ii) the syllabification patterns of complex words, and (iii) the restrictions on schwa-deletion in complex words. 107

I came to the conclusion that four classes of suffixes must be distinguished: (i)

stress-neutral suffixes (preceded by #)

(ii)

stress-shifting suffixes (preceded by +)

(iii) adjectival suffixes which shift the main stress to the final strong cluster before the suffix (preceded by + or #) (iv)

suffixes which are attached after the application of the stress rules (including the inflectional suffixes)

Prefixes

are always followed by # but prefixation rules must apply after the

word stress rules. One of the consequences of the theory proposed here is that the WF-rules specify whether an affix has underlying stress or not. (If nothing is specified, then it is unstressed in virtue of a general rule (SPE, p.66)). That is, it must be learned for affixes whether they have stress or not, whereas for all the words of lexical categories the stress patterns are predicted by means of a system of stress rules.

2.3. Vowel reduction and the structure of complex words In the introduction to this chapter I pointed out that one of the assumptions about the structure of words is, that complex words may have internal labeled brackets. In this section I will argue that these internal brackets which were necessary for morphology, are also necessary for phonology, i.e. for a correct account of the stress patterns of, and the vowel reduction possibilities in complex words. The central issue involved here is that the presence of internal brackets makes it possible to let the stress rules cyclically apply to complex words. The cyclic application of stress rules predicts certain stress contours for complex words. The correctness of the stress contours obtained in this way can be checked by testing whether the correct vowel reduction possibilities are predicted. Of course such a test presupposes an independently motivated rule of Vowel Reduction. Therefore, the first subsection proposes a rule of Vowel Reduction based on an analysis of vowel reduction in simplex words. The method used here to investigate the correctness of certain stress contours obtained by cyclic application of stress rules is identical with that used by SPE to justify the hypothesis of cyclic application of stress rules: SPE too uses vowel reduction to support the hypothesis that stress rules must be applied cyclically. The rule of Vowel Reduction in SPE reduces unstressed lax vowels only. The vowels are reduced to a schwa. SPfip.llG) notes e.g. the 108

fine distinction between compensation

and condensation derived from the verbs

compensate and condense respectively. "The cycle applies first to the 1 1 2 eirtbedded verb (yielding condense and compensate) , and then to the whole noun, thus giving a result in which both end in -ation but only the former has a stress on its second syllable"(McCawley 1975:159). This results in the stress patterns compensation versus condensation. Consequently, only the vowel in the second syllable of compensation

can be reduced to a schwa.

same difference is noted by Halle and Keyser(1971:153) for the pair 3 4 The 1 1 3 1 1 *^ attestation (-»-attest) - devastation (-^devastate) •. "the assumption that rules apply in cyclical order is absolutely crucial to obtaining the correct results. Any theory that wishes to operate without cyclical application of rules would have to account for the same facts in an equally motivated manner. Though not impossible in principle, it is quite unlikely that this challenge will be easily met. And unless and until it is met, the assumption that cyclical rules apply below the word level must be accepted." Cyclic application of stress rules is also defended by Brame(1974). On the other hand, it has been questioned, at least for English, by Ross(1972), Schane(1972) and McCawley(1975:158-59). The implication of non-cyclic stress assignment has been formulated by Schane(1972:251) as follows: "If there is no cycle then the set of rules must apply immediately to the whole word and the stresses will have to be assigned by running through the rules exactly once. Second, there is no need for nested structures. Of course we still require morphological information, such as whether we are dealing with a noun or a verb, and also we will need to know where morpheme boundaries are located. It is just hierarchical structure which is no longer necessary." Schane's conclusion concerning the consequences of a rejection of a cyclical application of stress rules is somewhat too strong: hierarchical structure will no longer be necessary for phonology but it still remains necessary for morphology, for the redundancy rule function of WF-rules. A decision with respect to the hypothesis of cyclic application of stress rules and its universal validity is very hard to make. Therefore, my modest aim for the next subsections on vowel reduction will be to show how cyclic application of stress rules combined with a rule of Vowel Reduction makes correct predictions concerning the vowel reduction possibilities of complex words in a relatively simple way. I would not know how to account for these 109

facts in an equally simple way without the mechanism of cyclic application of stress rules. Therefore, I consider the analysis of vowel reduction presented here a provisional corroboration of the theory of cyclic application and consequently of the morphological theory of nested structure for complex words. The analysis will proceed as follows: first I propose a rule of Vowel Reduction based on an analysis of vowel reduction in simplex words. Secondly, vowel reduction in complex words is treated. Finally, the consequences for Brame1s Natural Bracketing Hypothesis will be discussed. The analysis of vowel reduction presented here only predicts vowel reduction possibilities. The actual application of the rule of Vowel Reduction not only depends on whether the conditions of Vowel Reduction are met, but also on factors such as the style of speech (Vowel reduction normally only applies in casual speech), the frequency of a word (Fidelholtz 1975) and the nature of the word (learned words are often only used in non-casual speech). These factors are also mentioned by Dressier(1975) and Zwicky(1972) for German and English respectively. Furthermore, non-native affixes generally appear to oppose

vowel reduction. For instance, the vowel [i] in the suffixes

-iseer and -iaan is not very frequently reduced although it occurs before a stressed syllable.

2.3.1. Vowel reduction in simplex words In Dutch unstressed vowels can be reduced to a schwa in the casual style of speech. The words in (93) can get the phonetic forms indicated in the third column:13

(93)

i metaal

[metal]

[m3tal]

banaan

[banan]

[bänan]

muziek

[myzik]

[mäzik]

persoon

[person]

[person]

lokaal

[lokal]

[l3kal]

In all these examples the MSR assigns primary stress to the vowel of the second syllable. The vowel of the first syllable is unstressed and can reduce. However, not every unstressed vowel can reduce. Unstressed vowels at the end of such words as ega, olie, soldo, amerika, pagina do not reduce. Unstressed vowels in the first syllable of a word can only be reduced if the next syllable

110

bears stress. Compare:11* (94) kabouter

1

[k3bDut3r]

fiducie

[f3dysi]

parochie

[p3r:>xi]

So a vowel can only reduce if followed by a stressed vowel. This principle 1 2 correctly predicts that in such a word as almanak with secondary stress on the final syllable, the second vowel can reduce: [alm3nctk] . Another constraint on vowel reduction can be inferred from the fact that the unstressed vowel of the first syllable of a word cannot be reduced if no consonant precedes it, as can be concluded from (95): (95)

anaal, *[3nal]; egaal, *[3yal]} elite, *[31it3]; ellips, *[31ips]

In order to account for these facts concerning vowel reduction I propose the following phonological rule: (96) Vowel Reduction (optional) V

[3] / # X C

CQ [- stress]

V [+ stress]

Y #

Since diphtongs can be considered combinations of two vowels, it is predicted that unstressed diphtongs do not reduce: (97)

houweel,-[h3wel]

laurier,

*[l3ri:r]

Rule (96) reduces both tense and lax vowels if they are followed by one or more consonants. The following examples support this claim (the vowel in italics reduces): (98) tense vowels: frekwentie, relikwie, jwstitie, persiflage, per-iscoop; lax vowels:

carambool, colporteur, documentatie, compensatie

But reduction of a lax vowel followed by more than one consonant is only possible in the first syllable, if the vowel is an [e] followed by [r] plus C or if both consonants belong to the second syllable: (99)

per$soon, per$cent, pa$stoor, pa$stei

but: 111

rem$bours, *[r3mbu:rs]; pen$dant, *[p3ndant]; flamlbouw, *[fl3mb:>uw] kar$ton, *[k3rt}n]j por$tier, *[p3rti:r]j kwar$tier, *[kw3rti:r]

In this connection it is an interesting fact that words such as karton kwartier can have the following phonetic forms respectively:

and

[k3ton],

[kw3ti:r]. In these forms, the [r] has been deleted, so the vowel is not followed by a consonant in the same syllable anymore. This makes the application of Vowel Reduction possible: deletion of the [r] feeds Vowel Reduction. A second point to be noted is that this restriction on vowel reduction in the first syllable implies that syllable boundaries can occur in the context of phonological rules, and that therefore these syllable boundaries must be present in the phonological form of the words.

This conclusion was

independently arrived at in section 2.2.1.1.. The boundaries

are introduced

by syllabification rules given in section 2.2.1.1.. We add the following condition to rule

(96) to account for the restriction discussed here:

(96)'

/ i X ct

V

{3]

CQ [- stress]

Condition:

if X = 0, then

V Y if [+ stress] V C Q = (i) stress] .... (n)

er C Q ; .„ V $ CQ

or

This rule accounts for all the cases of vowel reduction discussed above. As Martin (1968:167) points out, a preceding the following words: humaan3

hevaut, heroisah.

[h] blocks reduction, e.g. in This is predicted by rule (96)'

because the Ih] has the feature [-cons] and therefore blocks reduction. For 1 2 the same reason, reduction of the second vowel in sanhedrtn if

is blocked; but

the [h] is deleted, reduction is possible and the correct phonetic form

[sanSdrm] is derived. Apart

from rule (96)' there is another rule of Vowel Reduction which

reduces unstressed lax vowels in the final syllable if the preceding syllable bears stress and (100)

112

koning

the word has at least one final consonant, e.g.: with reduction:

[kon3n]

kennis

[ken3s]

penis

[pen3s]

motor

[mot3r]

avond

[av3nt]

Reduction does n o t apply i f

the vowel i s not f o l l o w e d by a consonant. The

word eega e . g . does n o t admit r e d u c t i o n o f the f i n a l v o w e l .

Furthermore,

r e d u c t i o n i s not p o s s i b l e i n n o n - f i n a l s y l l a b l e s under these c o n d i t i o n s . word

Amerika

e.g.

cannot be reduced t o

The

[amer3ka]. So t h i s r u l e must have the

f o l l o w i n g form:

(101) Vowel Reduction I I V + I- tense]

[3] /

(optional)

v c1 1 [+ s t r e s s ] F

c

i

#

stress]

This r u l e a l s o c o r r e c t l y p r e d i c t s t h a t vowel r e d u c t i o n does not take p l a c e

if

the f i n a l u n s t r e s s e d vowel i s preceded by another unstressed v o w e l , as in 1 1 begrafenis, vergadering where the vowel o f the f i n a l s y l l a b l e does not reduce.

2.3.2. Vowel reduction

in complex words

I n s e c t i o n 1 . 4 . I proposed the h y p o t h e s i s t h a t WF-rules are s t r u c t u r e rules.

T h i s i m p l i e s t h a t WF-rules produce i n t e r n a l b r a c k e t s i n t h e i r

building output

words. I f we would admit r u l e s o f the f o l l o w i n g type which form words by means of

l i n e a r c o n c a t e n a t i o n o f morphemes:

(102)

[ x ]

-*•

[x

+ y]^

where X and Y a r e l e x i c a l

categories

the i n t e r n a l c a t e g o r y X would be erased and consequently the WF-rules would have t r a n s f o r m a t i o n a l power. Given the assumptions t h a t complex words c o n t a i n i n t e r n a l l a b e l e d b r a c k e t s and t h a t s t r e s s r u l e s apply c y c l i c a l l y ,

the f o l l o w i n g s t r e s s p a t t e r n s and

vowel r e d u c t i o n p o s s i b i l i t i e s are p r e d i c t e d f o r words c o n t a i n i n g the s u f f i x e s

+aal, +iteit. (103) a

+isme. +ist,

+eur> and +iseer

1 grammatika

2 . 1 grammatz-kaal

1

2]

paradox

paradoxaal

[[grammatika] 1

cycle,

L a t i n SR

2 n d c y c l e , MSR

b

debiel crimineel

respectively:

N

+aal]

A

1

J.

1

2 •1

debiliteit . 2 . 1 criminal!teit 113

c

d

.2 . 1 tri^altteit 1 boeddhìsme 2 1 katholicisme 3, 1 alcoholisme

triviaal 1 boeddha

2 1 propagandist

propaganda 1

reservist

reserve

monopolie

ambassadeur .2 . 11 pertodiseer , 2-, • 1 kanaltseer 1 monopoliseer

vitamine

vitaminiseer

e

ambassàde

f

periode kanaal

The reduction possibilities for the vowels in italics prove the correctness of the stress patterns derived by cyclic application of the MSR, and so indirectly support the hypothesis that these complex words contain internal brackets. I conclude therefore that the hypothesis put forward by Aronoff (1976:25) that WF-rules produce internal brackets for the basis parts of their output words is supported by Vowel Reduction in Dutch. Correct predictions are also made for complex words with prefixes: the vowel of Ver- in Verloop

can reduce since condition (i) of rule (96)' is met.

But reduction in other prefixes is impossible because they do not meet the requirements of rule (96)'.

2.3.3. Brame's Natural Bracketing Hypothesis Now the problem must be discussed whether the assumption of internal brackets proposed in the preceding section is in accordance with the Strong Natural Bracketing Hypothesis proposed in Brame(1974). Brame proposes that only under a certain condition can internal brackets be assumed. This condition is necessary but not sufficient for the assumption of internal brackets. Brame's hypothesis is given in (104): (104)

Strong Natural Bracketing Hypothesis For a substring if of a string to be bracketed, i|j must be equipotent to a string a, and the meaning of (J) must be a compositional function of the meaning of a and zbrer|9n] [lantsxiii : r 3 n / l a n s x ^ : r 3n]

It]-deletion in (b) is only possible after application of the readjustment rule for inter-word boundaries, because the rule mentions only one word boundary in its context. This accounts for the fact that [t]-deletion in (b) is only possible in casual speech, whereas the phonetic forms listed in (a) are also feasible in a more formal style of speech. Selkirk ( 1972) calls the three styles of speech distinguished by her for French: I Conversation familière II Conversation soignee III Lecture ou discours These three styles of speech are discrete subdialects of French, defined in a purely linguistic way: certain rules only apply in certain styles of speech. This does not imply that in actual performance only one style is used. Actual performance might show a mixture of linguistic variants. Labov(1972) e.g. has given

a characterization of performance-styles with respect to [t]-deletion

in word-final consonant-clusters, with the degree of attention paid to speech as a parameter :"There are a great many styles and stylistic dimensions that can be isolated by analytical means. But we find that all such styles can be ranged along a single dimension measured by the amount of attention paid to speech"(Labov 1972:208). 162

In Labov's analysis of [t]-deletion there are only two linguistic variants: a variant with application of [t]-deletion, and a variant without [t]-deletion. The percentage of application of each variant is determined by the degree of attention paid to speech and the socio-economic class of the speaker. The innovation proposed here is that with respect to certain rules there are three variants: (i) the rule always applies; (ii) the rule only applies within words and between words separated by one word boundary at most; (iii) the rule never applies. In this way it can be explained why in more careful speech the percentage of [t]-deletion within words is much greater than the percentage of [t]-deletion at the borders between words of lexical categories. Moreover, the hypothesis that in casual speech the strength of the boundaries between words of lexical categories is weakened also predicts that other rules with one #, e.g. nasal assimilation, also apply more frequently within words than at the border between words of lexical categories in more careful speech. That is, we can correlate the percentages of application of the different rules. The frequency of the readjustment rule itself may then be determined by the sociolinguistic variables of style (= degree of attention) and socio-economic class. Selkirk(1972:184) also points out that certain phonological processes operate more easily between the two parts of a compound than between two words of lexical categories. She concludes that "one is inclined to consider the strength of the boundary internal to compounds to be less than the ## specified in the CSR [= Compound Stress Rule]". Systematic investigation of the possible difference in strength of the boundary in compounds and that between words of lexical categories has not yet been executed for Dutch. There is only one quantitative investigation of It]-deletion (De Vries e.a. 1974) but in their paper the difference between inter-word and intra-word deletion is not investigated (cf. Booij 1977 for comment on this paper). The same remark applies to the quantitative investigations of voice assimilation in Dutch by Rijnbach and Cramer(1939) and Demeulemeester(1962). The suggestion made in this footnote therefore only defines a research program with respect to style-dependent phonological rules. 10. Since the WF-rules for stress-neutral #-suffixes only create new cycles for the Compound Rule, they could be ordered between the Word Stress Rules and the Compound Rule as well. I have no empirical arguments to make a choice, but simply order all the suffixation rules in one block. 11. The derivation of doorspiegelbaav can also be performed in three cycles, but without any empirical difference: 163

[spiegel]v 1 s t cycle, MSR

1

nt

2 ^ cycle, prefixation

[door#[spiegel]^]^

CR

(blocks, because no § follows the vowel with [l stress]

rd 3 cycle, suffixation

1 [[[door#[spiegel]v]v#baar]

CR

1

2

12. Note, however, that the accentuation of devastate given by dictionaries of 12 British English, e.g. Jones' English Pronouncing Dictionary (1963 ) is devastate. 13. Some data concerning vowel reduction in Dutch can be found in Martin (1968). But he mixes synchronic cases of vowel reduction with synchronic cases. The same objection must be raised against De Schutter(1975). Data concerning regional variation in vowel reduction are provided by Stroop(1974). 14. Apart from Vowel Reduction Dutch has another rule for unstressed vowels which laxes non-high tense vowels in the first syllable of a word or morpheme: [a]

[a]

banaan

[banan]

[banan]

tabel

[tabel]

abonnee

[abane]

[abane]

Napoleon

[ñapóle jon]-»-[napolej 3n]

cachet

[kasje] •*• [kasje]

irrationeel [irasjonel]-»-[irasjonel]

[o] -*• [d] kolossaal

[kolasal]

[kolosal]

[e]

telefoon

[telSfon]

[til3fon]

televisie

[tel3visi]

[tilSvisi]

[l]

-*-[tabel]

ge^reformeerd Iy9ref3rme :rt] ->-[y3rif 3rme :rt] The following rule accounts for this laxing process: V -»• [- tense] / L [- stress] " '

"

t"

^

15. This notation of the derivation of the stress pattern of grammatikaal is an abbreviation of the following notation: [grammatika] st 1 cycle, Latin stress rule 2

nd

cycle, WF-rule for +aal MSR

164

1 [ [grammatik>(]N+aal] 2

1

Notes to Chapter 3 1. This p o s i t i o n with r e s p e c t t o the i n f l u e n c e o f the e x i s t e n c e o f

complex

words on the d e r i v a t i o n o f o t h e r , synonymous complex words from the same b a s i s word i m p l i e s the r e j e c t i o n o f s o - c a l l e d

'blocking principles'

i n the Word

Formation Component. " B l o c k i n g i s the nonoccurrence o f one form due t o the simple e x i s t e n c e o f a n o t h e r "

(Aronoff

1976:43). The assumption o f a b l o c k i n g

p r i n c i p l e i n the Word Formation Component was proposed by A r o n o f f ( 1 9 7 6 : 4 5 ) : " f o r each stem t h e r e cannot be more than one i t e m in each meaning s l o t " ,

i.e.

we cannot l i s t two c o m p l e t e l y synonymous words d e r i v e d from the same b a s i s word i n the l e x i c o n . A r o n o f f ' s f o r m u l a t i o n o f

this blocking

principle

presupposes t h a t o n l y those complex words t h a t have a t l e a s t one p r o p e r t y a r e l i s t e d i n the l e x i c o n . The d e r i v a t i o n o f e . g . (distillation)

destillering

from the verb destilleer

b l o c k e d although destillatie

(distillation)

idiosyncratic

the noun

(to d i s t i l l a t e )

already e x i s t s ,

is

not

because

i s c o m p l e t e l y r e g u l a r and needs not be l i s t e d i n the l e x i c o n

destillering

for

that reason. Note furthermore t h a t A r o n o f f makes h i s b l o c k i n g p r i n c i p l e n e a r l y vacuous because a c c o r d i n g t o him those r u l e s which d e r i v e new words normally complex words which are c o m p l e t e l y r e g u l a r :

"As f a r as I can t e l l ,

d i r e c t l i n k between semantic coherence and p r o d u c t i v i t y "

(Aronoff

2. Not everybody w i l l perhaps a g r e e t h a t an a b s o l u t e d i s t i n c t i o n

derive

there i s a 1976:39).

between

p r o d u c t i v e and unproductive r u l e s should be made because i t i s c l e a r t h a t from a s t a t i s t i c a l p o i n t o f v i e w some p r o d u c t i v e WF-rules are more o f t e n than o t h e r s . The s u f f i x -he-id

e.g.

i s p r e f e r r e d t o the s u f f i x -te

speakers o f Dutch f o r the d e r i v a t i o n o f nouns from a d j e c t i v e s .

applied

by n a t i v e

Compare:

N stijf

(rigid)

stijfheid

stijfte

(rigidness)

blauw

(blue)

blauwheid

blauwte

(blueness)

slank

(slim)

slankheid

slankte

(slimness)

geel gek

(yellow) (mad)

geelheid

geelte

(yellowness)

gekheid

gekte

(madness)

Some o f my informants c o n s i d e r blauute l i s t e d i n Van D a l e ( 1 9 7 6 ) .

Geelte

a c c o r d i n g t o my i n t u i t i o n s . a b s o l u t e l y unproductive.

r a t h e r awkward, even though i t

and gekte

is

are not l i s t e d but c o r r e c t

I t would be t o o f a r f e t c h e d t o c l a i m t h a t -te

But s i n c e -heid

i s more o f t e n used than -te,

might be i n c l i n e d t o s p e c i f y f o r each WF-rule i t s p r o b a b i l i t y o f

is

one

application. 165

The rule for -heid would have then a far higher probability than the rule for -te. But if these rules would have a different probability of application this does not imply that the set of possible words defined by the rule with a lower probability of application is smaller than the set of possible words defined by the rule with a higher probability of application. It is only the set of actual words derived by the first rule that is smaller. One might invoke the well-known distinction between grammaticality and acceptability here. The Word Formation Component defines the well-formed (i.e. the grammatical) words of a language, some of which are less acceptable than others. For instance, I do not consider the word geelte impossible, but less acceptable than geelheid. The degree of acceptability might correlate then with the probability of application of a WF-rule. That is, the notion 'probability of application' could be incorporated in a performance theory of word formation which predicts the degree of acceptability of words. 3. The formulation of the part of rule (2) to the right of the arrow is an abbreviation of I[x V ([+ c o n s ]) [- v o c ] 1 # 3 1 . This kind of + son A t N abbreviation will be used throughout the rest of this chapter. 4. The word Friese (woman from Friesland) is no counterexample since this word can be assumed to be derived from the adjective Fries (Frisian) by adding a schwa. That is, there is no need for the assumption of a suffix -se for the derivation of this word. If we would assume a suffix -se, this suffix would apparently be attached to nouns (Haarlem-Haarlemse) but then we would expect FriesZandse derived from the noun Friesland. Dutch has an unproductive suffix -se in such words as generaalse (wife of a generaal), domineese (wife of a dominee), kosterse (wife of a koster) . But this is clearly another suffix, which does conform to the phonological condition discussed here. 5. This set of conditions on the formation of -or-nouns is probably not sufficient. Semantic conditions might also be involved, e.g. the condition that the basis noun mentions an action or process. For instance, one cannot derive *nator from natie (nation) because this word is not an action/process noun. This would imply that there are also semantic conditions on bases. 6. Possible exceptions to this condition are ondiepte (shallow) and goedkoopte (cheapness). Ondiepte can be assumed to be derived from either the adjective [oni[diep] ](shallow) or from the noun [[diep]^ite]^. If the latter possibility is the correct one, it is no exception at all. Goedkoopte is a exception if the adjective goedkoop (cheap) has the morphological structure [ [goed] ^""P^v/N^A*

these words are not counterexamples,

because the class of exceptions cannot be extended (cf. section 3.1.2.3. for 166

this criterion). 7. Note that this does not imply that these prefixes always bear primary stress in the surface forms of the complex words in which they occur, because their stress might be lowered due to the application of stress rules. 8. Saciuk(1969) claims that stratal features are always properties of morphemes, because according to him complex words might consist of two morphemes from different strata, where each morpheme undergoes the specific rules of its own stratum. An evaluation of Saciuk's arguments for this claim depends on an evaluation of the phonological and morphological analyses of several languages used by him, and will therefore not be given here. As far as Dutch is concerned we could also assume that stratal features are properties of lexical items and of affixes. Given this assumption, no special distribution convention for the non-phonological properties of morphemes is necessary. 9. This has certain consequences for the evaluation measure: we would like to relate e.g. the word [[strijk]^#je]

to the WF-rule for diminutives, although

its morphological structure does not completely match the morphological structure of the output of the WF-rule for diminutives. We want to relate this word to the WF-rule, because the fact that strijkje is a noun with the meaning aspect 'little' should not be counted as independent information. I will not pursue here the exact formulation of such an additional principle in the evaluation measure since that requires a more detailed analysis of the notion 'exception' in linguistic theory. 10. If we do assume two separate WF-rules for the prefix belt we cannot express that the attachment of this prefix always results in obligatorily transitive verbs. This generalization can be expressed if we collapse the rules, but this would imply that we have to admit negative syntactic conditions on the bases of WF-rules. The collapsed rule would have the following form:

[ x ]x

-

\h&§ I x ] x ] v [+ transitive]

Conditions: X * A if X = N, then V=> TO PROVIDE WITH X if X = V, then V=TO X ON

Aronoff (1976:62) however proposes to exclude both disjunctive (the unitary base hypothesis) and negative conditions from WF-rules. That is, if we prefer to constrain the formal nature of WF-rules as much as possible, the similarities in the behaviour of the two be-'s cannot be expressed. The 167

negative condition 'X ^ A' is necessary although there are a few verbs with adjectival bases, e.g. bedroef (to make sad). But such verbs must be considered cases of affix-generalization since they are exceptional. 11. Aronoff (1976 :69) who supports Siegel's position states that "boundaries encode the place in the phonological derivation of the base at

which the

operation of a particular WFR takes place: + is prephonological, # is post-cyclic (word level)...". This rather vague statement seems to reduce the role of boundaries in phonology: they are no longer theoretical entities, i.e. theoretical constructs which function in several types of phonological rules, e.g. stress and syllabification rules, but only indicate whether a WF-rule applies before or after the word-level stress rules. 12. Block 5 in schema (53) mentions the WF-rules for +er and +ing. The suffix +er is ambiguous: it can also be used to derive nouns from geographic names with the meaning 'inhabitant of N', and adjectives from geographic names with the meaning 'originating from N'. That is. block 5 has to contain three different rules for +er. The two suffixes mentioned here must also be assumed to be preceded by + because of the syllabification patterns of the complex words in which they occur, and because they cause schwa-deletion: Haar$lem

Haar$le$(m)mer

Bellingwolde

Bellingwolder

Hoo$ge$veen

Hoo$ge$vee$ner

Vlagtwedde

Vlagtwedder

13. See Leben and Robinson(1977) for the suggestion that boundaries can be dispensed with by means of rule ordering. They do not support this suggestion with well-motivated and detailed phonological analyses.

168

BIBLIOGRAPHY

Abbreviations: FdL

= Forum der

Letteren

IULC - Indiana University

Linguistics

LB

= Leuvense Bijdvagen

Lg

- Language

LI

- Linguistic

Ntg

- De Nieuwe Taalgids

Sp

= Spektator

Club

Inquiry

Andersen, H. 1972

'Diphtongization',

Anderson, 1974

Lg 48,

11-50

S.R.

'On the t y p o l o g y of p h o n o l o g i c a l r u l e s '

Lagaly(eds-) Papers from the Parasession Chicago,

i n : A.Bruck, R.A.Fox and M.W.

on Natural

Phonology,

1-12

A r o n o f f , M.

1976

Grammar, Cambridge Mass.

Word Formation in Generative

Bach, E. 1971

'Syntax s i n c e A s p e c t s '

in:

Developments of the Sixties

R.J.O'Brien S . J . ( e d . )

- Viewpoints for

Linguistics.

the

Seventies,

Georgetown, 1-17

1974 Basbjrfll,

1975

Syntactic

Theory, New York

H.

'Grammatical boundaries in phonology', Annual Reports of the of Phonetics

Berg,

B.

van

of the University

den

3

1964 1970

Institute

of Copenhagen 9, 109-35

Foniek van het Nederlands,

6

Den Haag (1964 , 1972 )

'Het woordaccent van a f l e i d i n g e n met h e t p r e f i x -on',

Ntg,

Van

Haeringen-nummer, 1-18 169

Besten, H. den 1974

'Terugvindbaarheid: de condities op de deleerbaarheid en substitueerbaarheid van lexicale elementen' (unpublished, Institute of General Linguistics, University of Amsterdam)

1975

'Het kiezen van lexicale delenda1, Sp 5, 415-32

Blancquaert, E. 1962

Praktische uitspraakleer van de Nederland.se taal, Antwerpen, 6 t h ed.

Bloomfield, L. 1933

'The structure of learned words' in: Ch. Hockett(ed-) A Leonard Bloomfield Anthology, Bloomington Ind. 1970, 252-56

1935

Language (British edition)

1939

'Menomini morphophonemics' in: Ch. Hockett(ed.) A Leonard Bloomfield Anthology, Bloomington Ind. 1970, 351-62

Booij, G.E. 1974

'Zinsbepalingen in het Nederlands', Sp 3, 619-46

1977

'Vragen.bij een Leids onderzoek', FdL 18, 15-22

Botha, R.P. 1968

The Function of the Lexicon in Transformational-generative Grammar, The Hague-Paris

1974

'Problematic aspects of the exception filter in Professor Halle's "Morphology"', IULC, june

Brame, M. K. 1974

'The cycle in phonology: stress in Palestinian, Maltese and Spanish', LI 5, 39-60

Brasington, R.W.P 1972

'The treatment of exceptions in phonology', Lingua 29, 101-19

Bresnan, J.W. 1971

'Sentence stress and syntactic transformations', Lg 47, 257-87

Chapin, P.G. 1967 1970

On the Syntax of Word Derivation in English (unpublished MIT-thesis) 'On affixation in English' in: M.Bierwisch and K.E.Heidolph(eds.) Progress in Linguistics, The Hague-Paris, 51-63

Chomsky, N. 1957

Syntactic Structures, The Hague-Paris

1965

Aspects of the Theory of Syntax, Cambridge Mass.

1970

'Remarks on nominalization1 in: R.A.Jacobs and P.S.Rosenbaum(eds.) Readings in English Transformational Grammar, Waltham Mass., 184-221

1971 170

'Deep structure, surface structure and semantic interpretation' in:

D . D . S t e i n b e r g and L . A . J a k o b o v i t s ( e d s . ) Reader

in Philosophy,

Linguistics

Semantics.

An

Interdisciplinary

Cambridge Mass.,

and Psychology,

62-

118 1973

' C o n d i t i o n s on t r a n s f o r m a t i o n s '

i n : S . R . A n d e r s o n and P . K i p a r s k y ( e d s . )

A Festschrift

New York e t c . ,

for

Morris

Halle,

232-86

Chomsky, N. and M . H a l l e 1968

The Sound Pattern

of

New York

English,

Cohen, A. 1958

' H e t Nederlands d i m i n u t i e f - s u f f i x , een m o r f o n o l o g i s c h e p r o e v e 1 ,

Ntg

51, 40-45 Cohen, A. 1961

e.a.

Fonologie

van het

Nederlands

en het

Fries,

s'Gravenhage

D a l e , van 1976

Groot

woordenboek

Demeulemeester, 1962

der Nederlandse

s"Gravenhage

taal,

(10th

printing)

F.

' A s s i m i l a t i e van s t e m l o z e e k s p l o s i e v e n en f r i k a t i e v e n v o o r b en d ' , en Tongval

Taal

14, 20-36

D r e s s l e r , W. 1975

'Methodisches zu A l l e g r o - r e g e l n ' Phonologica

1972,

i n : W.Dressier and F . M a r e s ( e d s . )

München, 219-34

Emonds, J . E . 1976

A Transformational preserving,

Evers, 1975

Approach

and Local

to English

Transformations,

Syntax.

Root,

New York

Structure-

etc.

A. The Transformational Utrecht,

Cycle

reproduced by

in Dutch

and German ( d i s s . U n i v e r s i t y

of

IULC)

E v e r s , A. and M.A.C.Huybregts 1975

"The c y c l e i n Dutch: i t s r e l e v a n c e t o the t h e o r y o f grammar' A.Kraak(ed.)

Fidelholtz, 1975

Linguistics

in

the Netherlands

in:

Assen,

1972-1973,

25-33

J.

'Word frequency and vowel r e d u c t i o n ' Vance ( e d s . ) Papers Linguistic

Society,

from

the

Chicago,

ll^1

i n : R.E.Grosman, L.J.San and T . J .

Regional

Meeting

of

the

Chicago

200-13

G i l i j a m s e , H. 1977

'Over h e t l e t t e r g r e e p g r e n s s y m b o o l ' ,

l e c t u r e d e l i v e r e d a t the annual

meeting o f the Algemene V e r e n i g i n g v o o r Taalwetenschap,

january

Groot, A.W. de 1966

' S t r u c t u r e l e v e r s c h i l l e n tussen f l e c t i e en d e r i v a t i e '

in:

id., 171

Betekenis en betekenisstructuur, Groningen, 75-100 Haeringen, C.B. van 1947

'De meervoudsvorming in het Nederlands' in: id. Neerlandica, Den Haag 1949, 186-209 and in: J.Hoogteyling(ed.) Taalkunde in artikelen, Groningen 1969, 87-108

1949

'Participia praeverbalia', Ntg 42, 187-91

Halle, M. 1973a 'Prolegomena to a theory of word formation', LI 4, 3-16 1973b 'Stress rules in English: a new version', LI 4, 451-64 Halle, M. and S.J.Keyser 1971

English Stress: Its Form, Its Growth and Its Role in Verse, New York

Hetzron, R. 1975

'Where the grammar fails', Lg 51, 859-72

Hoard, J.E. 1975

'The new phonological paradigm' in: D.L.Goyvaerts and G.K.Pullum(eds.) Essays on The Sound Pattern of English, Ghent, 21-62

Hooper, J.B. 1972

'The syllable in phonological theory', Lg 48, 525-40

Hoppenbrouwers, C.A.J. 1977

'Het meervoud in het Nederlands', Taalkundig Bulletin van het Nederlands Instituut der Eijksuniversiteit Groningen 7, 23-27

Hyman, L. 1975

Phonology : Theory and Analysis, New York etc.

Jackendoff, R.S. 1972 1975

Semantic Interpretation in Generative Grammar, Cambridge Mass. 'Semantic and morphological regularities in the lexicon', Lg 51, 63971

Jakobson, R. 1929

'Remarques sur l'évolution phonologique du russe comparée a celle des autres langues slaves' in: Selected Writings Vol 1, The Hague 1971"2

1949

'The phonemic and grammatical aspects of language in their interrelations' in: Essais de Iniguistique générale, Paris 1963

1965

'A la recherche de l'essence du langage'. Diogene 51, 22-38

Jones, D. 1956

An Outline of English Phonetics, Cambridge ( 8 ^ impression)

Katwijk, A. van 1974

Accentuation in Dutch (diss. Univ. of Utrecht)

Katz, J.J. and P.M.Postal 172

1964

An Integrated Theory of Linguistic Description, Cambridge Mass.

Kenyon, J.S. and T.A.Knott Ì953

A Pronouncing Dictionary of American English, Springfield Mass.

Kiparsky, P. 1973a 'Phonological representations' in: O.Fujimura(ed.) Three Dimensions of Linguistic Theory, Tokyo, 1-136 1973b '"Elsewhere" in phonology' in: S.R.Anderson and P.Kiparsky(eds.) A Festschrift for Morris Halle, New York etc., 93-106 1973c 'Productivity in phonology' in: M.J.Kenstowicz and Ch.Kisseberth(eds.) Issues in Phonological Theory, The Hague-Paris, 169-76 Kisseberth, Ch. 1972

'Cyclical rules in Klamath phonology', LI 3, 13-34

Kooij, J.G. 1976

'Schwa insertion in Dutch: phonology or morphology?', to appear in W. Dressier (ed. ) Proceedings of the 3rc^ International Phonology Meeting Vienna 1976, Innsbrück

Koster, J. 1975

'Dutch as an SOV language', Linguistic Analysis 1, 111-36

Labov, W. 1972

3 Socvolinguistic Patterns, Philadelphia (1975 )

Ladefoged, P. 1971

2 Preliminaries to Linguistic Phonetics, Chicago (1973 )

Lakoff, G. 1970

Irregularity in Syntax, New York etc.

Leben, W.R. and O.W.Robinson 1977

'"Upside-down" phonology', Lg 53, 1-20

Lees, R.B. 1963

The Grammar of English Nominalizations, Indiana Univ. Bloomington (5*"*1 printing Bloomington/The Hague 1968)

Ljung, M. 1975

'Review of H.Marchand, Studies in Syntax and Word Formation ed. by D.Kastovsky, München 1974', Foundations of Language 13, 475-80

Martin, W. 1968

'De verdoffing van gedekte en ongedekte e in niet-hoofdtonige positie bij Romaanse leenwoorden in het Nederlands', Ntg 61, 162-81

McCawley, J.D 1973

'A review of Noam A.Chomsky, Studies on Semantics in Generative Grammar', IULC 173

1975

'Review of The Sound Pattern of English' in: D.L.Goyvaerts and G.K. Pullum(eds.) Essays on The Sound Pattern of English, Ghent, 145-97

Meinsma, G.L. 1958

'Assimilatie en assimilatie-onderzoek, Taal en Tongval 10, 107-14

Meys, W.J. 1975

Compound Adjectives in English and the Ideal Speaker-Listener, Amsterdam/New York

Peters, P.S. and P.W.Ritchie 1969

'A note on the Universal Base Hypothesis', Journal of Linguistics 5, 150-52

1973

'On the generative power of transformational grammar', Information Sciences 6, 49-83

Postal, P.M. 1968

Aspects of Phonological Theory, New York etc.

Reece Allen, M. 1975

'Vowel mutation and word stress in Welsh', LI 6, 181-201

Riemsdijk, H.C. van 1974

'De relatie tussen postposities en partikels', Sp 3, 447-62

Rijnbach, A. en W.F.Cramer 1939

'De assimilatie van stem en plaats in het Nederlandsch bij het lezen van losse zinnen', Onze Taaltuin 8, 97-107

Rijpma, E. en F.G.Schuringa 1968

Nederlandse Spraakkunst, Groningen (21th ed.)

Ross, J.R. 1969

'A proposed rule of tree-pruning' in: D.Reibel and S.A.Schane(eds.) Modern Studies in English, New Jersey, 288-99

Saciuk, B. 1969

'The stratal division of the lexicon', Papers in Linguistics 1, 464532

Sassen, A. 1968

'Problems of pre-, post- and co-existence and the word class of past participles in Dutch', Lingua 21, 400-16

Schane, S.A. 1968

French Phonology and Morphology, Cambridge Mass.

1972

'Non-cyclic English word stress', in: D.L.Goyvaerts and G.K.Pullum (eds.) Essays on The Sound Pattern of English, Ghent 1975, 249-59

Schreiber, P.A. 1972 174

'Some constraints on the formation of English sentence adverbs', LI 3,

183-201 Schultink, H. 1961

'Productiviteit als morfologisch fenomeen', FdL 2, 110-25

1962

De morfologische Nederlands,

1964

valentie

adjectief

in

modern

Den Haag

'De bouw van nieuwvormingen met her-', Taal- en Letterkunde

1973

van het ongelede

Tijdschrift

voor

Nederlandse

80, 151-84

'Het prefix ge- in Nederlandse

(en Duitse) verleden deelwoorden', Ntg

66, 409-18 1974a 'Plaats en aard van morfologische regels in een transformationeelgeneratief taalmodel', FdL

15, 23-39

1974b 'Dubbele diminutieven in het Afrikaans en transformationeelgeneratieve taalbeschrijving' in: Taalkunde: aan prof. 1975

dr. W.Kempen,

'Output conditions in word formation?' in: W.Abraham(ed.) Ut Contributions

to an Understanding

on the Occasion 1977

'n lews, studies

opgedra

Kaapstad-Johannesburg

of his 70^

of Linguistics.

Birthday,

For Pieter

Videam. Verbürg

Lisse, 263-72

'Affixen en hun boundaries', Sp 6, 472-76

Schutter, G. de 1975

'De plaats van de 3 in een fonologische beschrijving van het Nederlands', LB 64, 173-202

Selkirk, E.O. 1972

The Phrase

Phonology

of English

and French

(unpublished MIT-thesis)

1974

'French liaison and the X-notation', LI 5, 573-90

Siegel, D. 1974

Topics

in English

Morphology

(unpublished MIT-thesis)

Smith, N.S.H. 1976

'-Aar', LB 65, 485-96

Stanley, R.S. 1967 1973

'Redundancy rules in phonology', Lg 43, 393-436 'Boundaries in phonology' in: S.R.Anderson and P.Kiparsky(eds.) A Festschrift

for Morris

Halle,

New York etc., 185-206

Stroop, J.P.A. 1974

'lets over de uitspraak van de protonische vokaal in Romaanse leenwoorden', Ntg 67, 314-40

Toorn, M.C. van den 1973

Nederlandse

grammatica,

Groningen

Tops, G.A.J. 175

1974

'Assimilation of voice in Dutch. A aenerative approach', LB 63, 143-51

Vago, R.M. 1976

"Theoretical implications of Hungarian vowel harmony, LI 7, 243-63

Vennemann, Th. 1972 1976

'On the theory of syllabic phonology', Linguistische

Berichte

18, 1-18

'Vowel alternations in English, German, and Gothic: remarks on realisti in phonology' in: M.A.Jazayery e.a.(eds.) Linguistic Studies

in Honor

of Archibald

A. Hill

Volume

and

Literary

1, Lisse, 337-59

Vries, J.W. de 1975

Lexicale

morfologie

van het werkwoord

in modern

Nederlands,

Leiden

Vries, J.W. de e.a. 1974

'De slot-t in consonantclusters te Leiden: een sociolinguistisch onderzoek', FdL 15, 235-49

Walker, D.C. 1975

'Word stress in French', Lg 51, 887-900

Welschen, A.J. 1970

'De morfologische valentie van -esk',

Ntg 63, 441-52

Wissing, D.P. 1971

Fonologie

en morfologie

Afrikaans;

van die simplekse

selstandige

'n transformationeel-generatiewe

naarmooord in

beschri.jving,

Amsterdam

Wurzel, W.U. 1970

Studien

zur deutschen

Lautstruktur,

Berlin

Zaalberg, C.A. 1976

'Afbreuk', Meta

10, no.6, 16

Zwaardemaker, H. en L.P.H.Eijkman 1928

Leerboek

der phonetiek,

Haarlem

Zwarts, F. 1975

'-Aar, -arij, -se- en -te', Taalkundig Instituut

der Rijksuniversiteit

Bulletin

Groningen

van het

Nederlands

6, 9-23

Zwicky, A.M. 1972

'On casual speech' in: P.M.Peranteau, J.N.Levi, and G.C.Phares(eds.) Papers Society,

176

from

the 8^

Regional

Chicago, 607-15

Meeting

of the Chicago

Linguistic

INDEX

acceptability 166

Booij, G.E. 14,163

Adverbialization 14,15

Botha, R.P. 6,7,9,10,86

affix

boundary

ambiguous — 140

grammatical — 75,77

discontinuous — 32,45

syllable - 75-79,82,112

learned - 132

word-internal - 18,22,27,40,55,70-97, 127,129,131,168

loan — 23 native - 23,131-39,143

— insertion 78,91-93

non-native - 110,131-39,143

— loss 103

normal — 132

— reduction 161

— generalization 136,140

-weakening 79,87-89,103,163

Agent Postposing 12 Andersen, H. 7 5

Brame, M.K. 29,104,109,110,114-16,119, 145

Anderson, S.R. 101

Brasington, R.W.P. 69

Aronoff, M. 1,4,5,31-35,37-45,114,

Bresnan, J.W. 16,51,53,92

115, 133, 140,142,146, 165,168 Chapin, P.G. 12,153

assimilation voice — 80,94,159,160

Chomsky, N. 2,4,6,10-14, 58

nasal - 160,161,163

Chomsky-adjunction 41 Cohen, A. 86,161

Bach, E. 8,10,41

comparative 47,48,50,94

back formation 27

competence 2

Basbrfll, H.

79,102,161,162

lexical —2,17

Berg, B. v.d. 102,161

Complementizer Placement 14

Besten, H. den 8,158

compounds 2,5-10,14,23,44-48,56,70,71,

Blancquaert, E. 61

78,79,86,87,93,99,100,157,160-63

blocking principle 165

adjectival — 159

Bloomfield, L. 3,64,101,132

synthetic - 3,44,45,160 177

Compound Rule 56,70,71,74,88,99,105

Fidelholtz, J. 110

144,147,152, 163

filter 18,20,23-25,30,49

Minor —

final devoicing 75,80,83

56,71,88

conspiracy 126 Cramer, W.F. 163 creativity

flexion 46-54 - r u l e s 46-54,133,151,152 Fuckin-rule 42

word-forming — 3 rule-changing — 4,138,149

gerunds 10,11

rule-governed — 4,137

Gilijamse, H. 75 grammaticality 166

deletion 7,35,36,43

Groot, A.W. de 4,47

lexical — 7,8 morpheme — 39

Haeringen, C.B. van 28,95

recoverability of — 8

Halle, M. 1,2,17-20,22-26,45-50,52,53,

— rules 36,44,45

57,58,70,109,157,159

Demeulemeester, F. 163

Hetzron, R. 52

derived nominals 10-14

Hoard, J.E. 69

diminutives 136,153,158

Hooper, J.B. 75-77

diphtongs 60,111

Hoppenbrouwers, C.A.J. 95

Dressier, W. 110,117

Huybregts, M.A.C. 16,58 Hyman, L. 22,58

Eijkman, L.P.H. 57 Emonds, J.E. 13

implicit transposition 32,42

evaluation measure 25,26,62,167

internal brackets 42,55,108,113-16, 118,

Evers, A. 16,58,117

119

exception 62,63,66,136, 137, 140,150 166-68

Extended Standard Theory 2,9 Extraposition .14

Jackendoff, R.S. 1,9,12,17,19,24-27,30, 31,37,41,45,46,50,51 Jakobson, R. 158 Jones, D. 57,164

features diacritic — 6 3,69

Katwijk, A. van 57

exception — 43,44,49,58,62,69

Katz, J.J. 8,157

grammatical — 46,51

Kenyon, J.S. 57

negative rule — 68

Keyser, S.J. 57,58,70,109

non-phonological — 56

Kiparsky, P. 66.158

positive rule — 68

Kisseberth, Ch. 42

stratal - 63-65,131-39,167

Knott, T.A. 57

178

Kooij, J.G. 75,87

Nuclear Stress Rule 16,46,51,52,92

Labov, W. 162,163

Object Raising 12,13

Ladefoged, P. 57

output condi tion 8

Lakoff, G. 68,69 Lees, R.B. 6,7,9,157 lexical insertion 41,42,51 lexicalist theory of word formation 12-15,17,45,93

paradigm 46-52 defective — 49 idiosyncratic — 50 parenthesis notation 25,40

lexical representation 58

particle 99,100,159

liaison 92,161

past participle

link phoneme 9,48,83,86,160

Peters, P.S. 10

Ljung, M. 3

phonological dictionary 9

4,29,50,103,133,159

Postal, P.M. 8,63,157 major rule 117

prefixes 82,97-108,114

Marie, J. van 158

Class I - 144

Martin, W. 112,164

Class II - 144

McCawley, J.D. 12,109

Greek - 132

Meys, W.J. 7

Latin - 132

minor rules 67-69,117

native — 132

morpheme 3,4,18,19,22,127,129,131 grammatical — 46,51

non-native — 132 productivity 4,5,120,128,142,143

inflectional — 48

degrees of — 5

lexical — 22,64,65

potential - 120,121,127,155

native - 131-39

qualitative — 120,121

non-lexical — 65,116

quantitative — 5,120

non-native — 141-39

pruning principle 116-18

zero — 32 - list 18,30

readjustment rule 46,51,91,92,94,97

—structure conditions 20,22-24,31, Reece Allen, M. 48,96,97 64,80,85,131,132 morpholexical rule 101

Relative Clause Reduction 6 Riemsdijk, H.C. van 99 Rijnbach, A. 163

Natural Bracketing Hypothesis 29,104, Rijpma, E. 48 114-18 nested structure 27,28,109,110

Ritchie, P.W. 10 roots

neutralization process 56,158

learned — 132

Nominalization 11,12,16

normal — 132 179

Ross. J . R . rule

informal -

109,117

non-casual — 110

application

cyclic -

56,70,105,108-10,117-19

147,148,152,155

governed — 86

n e g a t i v e c o n d i t i o n s on — 148,153,

R.S.

24,78

degrees o f — 57,58 d i s t i n c t i v e — 58 u n d e r l y i n g — 74,104 — l o w e r i n g convention 67,159

155

stress

r e c u r s i v e — 153 rule

Stanley, stress

conventions f o r — 68,73,155 lexically

117,118,161

rule

Alternating -

ordering

disjunctive -

62,66,76,84,107,155

56,62,67-69,105

French — 63,65,66,105

e x t r i n s i c — 142,143,145

L a t i n — 67,105

i n t r i n s i c — 142

Main -

56,58-62,66,67,69,74,104,105,

110

l i n e a r — 142

Primary — 144 Saciuk, B. 64,65,13.1, 167

Stroop, J . P . A .

Sassen, A.

subcategorization r e s t r i c t i o n s

28,29

Schane, S.A.

Subject Raising

96,109,159

4,5,21,29,36,47,48,

12,13

Class I -

72-74,78,81,86,144

75,89,95,100,101, 103, 121,124, 132,

Class I I -

133, 136, 149, 150, 15 3, 157, 158

inflectional -

72-74,79-81,86,87,144

Schuringa, F.G. 48

l a t i n a te — 133

Schutter,

native -

G. de 164

schwa-deletion

72,81,82,94,96,97,

131-139,150

non-native -

61,131-39,150

108,151

85,86

segment s t r u c t u r e c o n d i t i o n s 64, 131,132

s t r e s s - s h i f t i n g — 72-74,82,108 — o r d e r 80,153

selection restrictions E.O.

D.

simplicity

10,13,19

49,91,92,96,161-63

sentence adverbs Siegel,

81,82,93-97,108,151

s t r e s s - n e u t r a l — 72-74,79,80,90,95,

104,107 schwa-insertion

Selkirk,

14,15

39,40,49,72,93,97,143-47 25,26

Smith, N.S.H.

80,149

superlative

47,50,94

syllabification

75-80,94,96,103,107,149,

168

p h o n o l o g i c a l — 79 p h o n e t i c — 79 — rules

76-78,103,112

syllable

speech casual -

79,103,110,161-63

p h o n o l o g i c a l — 76

formal -

117,161-63

p h o n e t i c — 76

180

19

suffix

S c h r e i b e r , P . A . 14 S c h u l t i n k , H.

164

146,165

Toorn, M.C. van den 157 Tops, G.A.J. 159

suffix-replacing — 39-41

Tough Movement 13

structure-building — 41,42,113

transformational theory of w o r d

transformational — 42,43,113 unproductive -

formation 5-16,92 Truncation rule 33,36,39,40,42-45

165,166 Wurzel, W . U .

116,118,128,146

5,30,31,39,89,121

153

Umlaut-rule 68,69

X-bar convention

13,14,19

Vago, R.M. 63

Zaalberg, C.A. 78

Vennemann, Th. 69,75

Zwaardemaker, H. 57

vowel copy rule 42,43

Zwarts, F.

vowel laxing 164

Zwicky, A. 110

122,123,149

vowel mutation 97 vowel reduction

56,57,88,108-19,152,

164 Vries, J.W. de

24,99,100,163

Walker, D.C. 63,96 We Ischen, A.J.

134,137

Wissing, D.P. 46 word,

passim

foreign — 64 French - 60,127 Greek -

127

Latin -

127

learned - 64,110 loan - 61,75,132 native - 60,61,132 non-native — 60-64 normal —

64

W o r d Formation Component

18,20,23,46,49,

51, 120-22, 126, 127, 132, 143, 151,155, 156, 165 Word Formation rules, bases of —

passim

18,44,45,48

productive -

5,29,30,31,33,38,121,127 181