Old English Weak Verbs: A Diachronic and Synchronic Analysis (Linguistische Arbeiten) (German Edition) [Reprint 2011 ed.] 3484301120, 9783484301122

Table of contents :
INTRODUCTION
CHAPTER ONE: THE DATA
1. The Germanic verb system
2. The Old English verb system
2.1 Major characteristics
2.2 Weak verb classes
3. Variations in the OE system
3.1 Early Old English versus Late Old English
3.2 Dialectal variation in the OE period
4. Summary
CHAPTER TWO: THE DIACHRONIC DEVELOPMENT OF THE OLD ENGLISH WEAK VERBS
1. Introduction
2. Diachronic rules
2.1 Glide Vowel Coalescence
2.2 Vocalization
2.3 West Germanic Gemination
2.4 Aufhellung
2.5 Breaking
2.6 Palatalization
2.7 i-umlaut
2.8 j-deletion
2.9 Apocope
2.10 Syncope
2.11 OE Vowel Reduction
3. Diachronic development of the OE weak verbs
3.1 Class I
3.2 The weak verbs of classes II and III
4. Theoretical issues
4.1 Phonological change in standard generative phonology
4.2 Diachronic evidence and generative phonological theory
4.3 Recoverability and productivity
CHAPTER THREE: A SYNCHRONIC ANALYSIS OF THE OLD ENGLISH WEAK VERBS
1. Introduction
2. Restructuring of underlying lexical representations
3. The j formative
4. Degemination
5. Schwa-insertion
6. Synchronic derivations of the OE weak verbs
6.1 Class I
6.2 The weak verbs of classes II and III
CHAPTER FOUR: THE TRANSFER OF OE WEAK VERBS OF CLASS I TO CLASS II
1. Introduction
2. Stages of transfer
2.1 Loss of the j formative
2.2 A two-paradigm system
3. Reasons for the transfer
4. Analogy
5. Conclusion
APPENDIX
BIBLIOGRAPHY

Citation preview

Linguistische Arbeiten

112

Herausgegeben von Herbert E. Brekle, Hans Jürgen Heringer, Christian Rohrer, Heinz Vater und Otmar Werner

Detief Stark

The Old English weak verbs A diachronic and synchronic analysis

Max Niemeyer Verlag Tübingen 1982

CIP-Kurztitelaufnahme der Deutschen Bibliothek Stark, Detlef: The Old English weak verbs : a diachron. and synchron, analysis / Detlef Stark. Tübingen : Niemeyer, 1982. (Linguistische Arbeiten; 112) NE:GT ISBN 3-484-30112-0

ISSN 0344-6727

©Max Niemeyer Verlag Tübingen 1982 Alle Rechte vorbehalten. Ohne ausdrückliche Genehmigung des Verlages ist es nicht gestattet, dieses Buch oder Teile daraus auf photomechanischem Wege zu vervielfältigen. Printed in Germany. Druck: fotokop Wilhelm Weihert KG, Darmstadt.

ACKNOWLEDGEMENTS

I am grateful to my teachers at the University of Wisconsin-Madison, Charles T. Scott and Peter A. Schreiber, for their help and support in the preparation of this book. I would also like to thank my colleague Jim Dishington, Harvard University, for his advice in the preliminary stages. But most of all I am indebted to Martin D. Pam, Gesamthochschule Wuppertal, whose knowledge of linguistics and generosity proved to be crucial many times. To my wife, Ruth Francisca, and my mother, Margarete Stark, this book is dedicated.

Hannover July 1981

D. S,

VII

TABLE OF CONTENTS

Page

INTRODUCTION

1

CHAPTER ONE: THE DATA

5

1. The Germanic verb system 2. The Old English verb system 2.1 Major characteristics 2 . 1 . 1 Strong and weak verbs 2 . 1 . 2 The -j- formative 2.2 Weak verb classes 2 . 2 . 1 Class I 2 . 2 . 2 Class II 2 . 2 . 3 Class III 3. Variations in the OE system 3.1 Early Old English versus Late Old English 3.1.1 EOE versus LOE variations in class I 3.1.2 EOE versus LOE variations in class II 3.1.3 EOE versus LOE variations in class III 3.2 Dialectal variation in the OE period 3.2.1 Class I 3.2.2 Class II 3.2.3 Class III 4. Summary

31 32 32 33 33 34

CHAPTER TWO: THE DIACHRONIC DEVELOPMENT OF THE OLD ENGLISH WEAK VERBS

37

1. Introduction 2. Diachronie rules 2.1 Glide Vowel Coalescence 2.2 Vocalization 2 . 3 West Germanic Gemination 2.4 Aufhellung 2.5 Breaking 2.6 Palatalization 2.7 i-umlaut 2.8 j-deletion 2 . 9 Apocope 2 . 1 0 Syncope 2.11 OE Vowel Reduction 3. Diachronie development of the OE weak verbs 3.1 Class I

5 8 8 8 9 11 11 16 22 27 27 28 30

37 39 40 42 43 46 49 52 55 58 60 60 61 63 63

VIII

Page 3.1.1 bycgan, sellan 3.2 The weak verbs of classes II and III 3.2.1 Class II 3 . 2 . 2 Class III 4. Theoretical issues 4 . 1 Phonological change in standard generative phonology 4 . 2 Diachronie evidence and generative phonological theory 4 . 3 Recoverability and productivity CHAPTER THREE: A SYNCHRONIC ANALYSIS OF THE OLD ENGLISH WEAK VERBS

1. Introduction 2 . Restructuring of underlying lexical representations 3. The j formative 4. Degemination 5. Schwa-insertion 6. Synchronic derivations of the OE weak verbs 6.1 Class I 6.1.1 bycgan, sellan 6.2 The weak verbs of classes II and III 6.2.1 Class II 6 . 2 . 2 Class III CHAPTER FOUR: THE TRANSFER OF OE WEAK VERBS OF CLASS I TO CLASS II

1. Introduction 2. Stages of transfer 2.1 Loss of the j formative 2.2 A two-paradigm system 3. Reasons for the transfer 4. Analogy 5. Conclusion

69 73 73 80 83 83 88 91 95

95 99 102 104 107 109 109 114 116 116 121 122

122 124 124 125 126 130 132

APPENDIX

135

BIBLIOGRAPHY.

144

INTRODUCTION

In the present study the system of the Old English (OE) weak verbs will be discussed. This includes a treatment of the diachronic, phonological development of the weak verbs from a Germanic (Gmc.) to an early OE stage, and a synchronic analysis of the weak verb system as

it

presented itself at the beginning of the OE period. Weak verbs, although morphologically much more regular than the strong verbs, appear to be in a state of transition during the OE period. They no longer comprise four classes, as for instance the Gothic weak verbs, and they are still a long way from the two classes of Middle English ( M E ) . In f a c t , class I with its three subclasses (exemplified by fremman, nerian, deman) and two important subgroups (bycgan, sellan), class II with its uniform conjugational pattern

basically

( l u f i a n ) , and the four

verbs of class III (habban, libban, secgan, hycgan) present a varied and often inconclusive picture. Also, beginning in early OE times, the originally short-stem verbs fremman, nerian, show a remarkable tendency to transfer to class II and to re-form their stems and endings according to class II (fremman vs. f r e m i a n ) . In recent years there have been several attempts to deal with the OE weak verbs from the point of view of generative phonology, the theoretical framework put forward most comprehensively by Noam Chomsky and Morris Halle in their The sound pattern of English ( 1 9 6 8 ) . Wagner ( 1 9 6 9 ) , Keyser ( 1 9 7 5 ) , Lass/Anderson ( 1 9 7 5 ) , and Kiparsky/O 1 Neil ( 1 9 7 6 ) are the most notable examples. The generative model in its general outlines will also provide the theoretical basis in the present study, although, unlike in the papers quoted above, various positions will

be modified. The fundamental difference is that a distinction will be maintained throughout the paper between the recoverability of certain phonological processes and the productivity of the same processes in a synchronic grammar of OE. This will lead, in the diachronic part, to a detailed analysis of the historical processes which appear to have played a role in the development of the weak verbs and which can be recovered from OE or, occasionally, other Gmc. dialects. In the synchronic part, many such processes as West Gmc. gemination, breaking, i-umlaut, etc., will be shown to be no longer productive and therefore no longer part of the synchronic rule inventory of OE. A synchronic analysis of the OE weak verbs will be presented which is far more concrete than the treatments by Wagner, Keyser, Lass/Anderson, and Kiparsky/O'Neil. This is to say that, based on the transparency of the surface structure, underlying representations will become more concrete. At the same time, restructurings of underlying representations will occur more frequently than in standard generative phonology because it will no longer be a basic tenet of the theory that all morphological alternations must be accounted for by rule if these alternations can be recovered by the specialist. In other words, if we intend to write a grammar not of the structural constitution of the language (utilizing all the resources we have for explaining all the various elements in the language), but a grammar of an actual speaker of the language, we are bound to arrive at a more concrete description than those in standard generative phonology. Chapter one contains a detailed survey of the data available on the OE weak verb system. The major characteristics, especially the function of the j_ formative which follows the verb stem, will be considered. Paradigms which are representative of all three classes will be discussed along with the variations that occur during the OE period. This will allow a comparison of the early OE verb forms with those of late OE. Similary, the dialectal

forms often provide interesting insights into the development of a language. They are therefore included in chapter one. In chapter two, phonological alternations and variations within attested OE and comparative evidence from other Gmc.dialects will help establish a set of major rules and proto-forms which reflect the historical (diachronic) development of the weak verbs from Gmc. to OE times. The weak verbs will be shown to undergo these diachronic rules and f u l l derivations will be given for five representative forms of each paradigm. The last section of this chapter contains a discussion of phonological change and of the crucial difference between recoverability and productivity already alluded to above. Chapter three will be concerned with a synchronic analysis of the OE weak verbs. Leading up to the synchronic derivations of the verbs, which will involve only three rules, is a treatment of the role of re-structuring in the present study and a discussion of the rules basic to this synchronic analysis, degemination and schwa-insertion. An important change, if compared to the diachronic derivations, involves the original j_ formative which has become part of the inflectional ending in OE und thus contributes to the phonetic similarity between class I nerian and class II lufian and ultimately to the transfer of class I to class II. The transfer of class I weak verbs to class II is treated in chapter four. Reasons for this transfer can be found in the high degree of formal similarity between class I nerian (in later OE frequently nerigan, nerigean) and class II lufian, the regular and simple morphological structure of class II verbs, the overwhelmingly large number of class II verbs, and a general tendency to regularize verb stems and endings ( c f . the two-class weak verb system of ME) . Verbs of the frenunan type transfer to class II as a result of analogy, which is discussed and defended as a viable concept in section 4. In general,

it is the goal of chapter four to show how the transfer from class I to class II supports the claim that the system underlying the classical OE weak verb paradigms had undergone considerable restructuring since its ProtoGmc.stage.

CHAPTER ONE: THE DATA

1.

The Germanic verb system

The Germanic verb system d i f f e r s sharply from that of the other branches of Indo-European. It d i f f e r s with respect to both conjugational complexity and the categories to which these conjugations refer. For example, the Romance languages, while sharing with Germanic a temporally-oriented system, typically have forms for expressing, both in the indicative and the subjunctive moods, a variety of tenses ( e . g . , present, simple past, continuous past, future, e t c . ) . On the other hand, the Slavic languages, while sharing with Germanic a basically binary conjugational system, make a primary distinction between completed and non-completed action rather than the past and non-past temporal distinction made in Germanic. In trying to reconstruct the history of the verb forms in the individual older Germanic dialects, scholars sometimes assume that we know more about the Proto-IndoEuropean verb system than is actually the case. But Prokosch, for example, warns (1938: 1 4 5 ) : "It would be wrong to ascribe to Indo-European the complicated tense system of Sanskrit, Greek, or Latin. A good deal of this is secondary innovation." Moreover, even in those cases where we can confidently assign a given category to ProtoIndo-European, it is not necessarily to be assumed that there has been a regular and deducible development to the daughter languages. Quoting Prokosch again ( 1 4 8 ) : "The present system of the Gmc. verb has been greatly standardized." This is one reason why we will pay very little attention to Indo-European here. Another, and far more important, reason is that Indo-European lacked that

6

feature which is most characteristic

of the Germanic

verb system. This is the division of verbs into two major classes — the so called 'strong' verbs and 'weak' verbs — according to the way in which the past tense is marked. Every Germanic grammar ( e . g . , Prokosch 1938, Krahe/Meid 1 9 6 9 ) and indeed every grammar dealing with an older Germanic dialect ( e . g . , Wright 1 9 1 0 , Braune/Eggers 1975, Campbell 1 9 5 9 , Brunner 1 9 6 5 ) present seven classes of verbs whose past systems are defined by a stem vowel (or vowels) different from that of the present tense, often with

accompanying consonant changes resulting from the

application of V e r n e r ' s Law. These 'strong 1 verb classes represent what Germanic has retained of the original Indo-European system. The 'weak' verbs, on the other hand, with their socalled 'dental preterite', are a Germanic innovation. Originally, there seem to have been four classes, the fourth class surviving intact only in Gothic. Prokosch ( 1 9 3 8 : 1 9 3 ) lists the four classes as follows: I. II. III.

Stem in ja/U/i/ Gothic nas j a n , sokjan Stem in o, Gothic salbon Stem in ai , e, Gothic haban (pret. habaida) , OHG haben

IV.

Stem in no, Gothic fullnan

(pret. fullnoda)

The following table illustrates the weak classes in the older Germanic dialects:

class I

class II

class III

class IV

present

Go ON OE OS OHG

nas jan telja nerian f remmian fremmen

salbon kalla luf ian salbo(ia)n salbon

haban haf a habban hebbian haben

f ullnan -

past sing habaida

ON

nasida talo"a

salboda kallaäa

OE

nerede

lufode

OS

f remida

salboda

hafai haefde hab5a

OHG

f rumita

salbota

habeta

Go

fullnoda —

8

2.

The Old English verb system

2.1

Major characteristics

2 . 1 . 1 Strong and weak verbs Old English maintains the distinction between strong and weak verbs. There are seven classes of strong verbs, which form their past tense with vowel variation (ablaut) in the stem, and three classes of weak verbs, which form their past tense with a stop segment usually considered dental (or alveolar) and which follows the actual stem. The stem of the weak verbs is commonly unchanged throughout the verb paradigm but several notable exceptions exist ( e . g . , brengan, bröhte,

(ge)broht; pyncan, fauhte, (ge)buht,

wyrcan, worhte,(ge)worht, e t c . ) . The strong verb classes were no longer productive during the OE period and fluctuations in class membership are making themselves felt because of the extremely high productivity of the weak verbs/ which, in fact, has been retained into Modern English, where newly formed verbs will always form their past tenses on the weak verb pattern. In OE, strong verbs occasionally have weak past tenses, slaepan

'to sleep 1 , a class VII verb, has (Camp-

bell: 3 2 0 ) the form

slepte in the Vespasian Psalter, the

Mercian Rushworth Gospels and in Bede. In the Lindisfarne Gospels and the Durham Ritual slepde is found. The Lindisfarne Gospels also contain a past tense plural -slepedon. Perhaps more surprising are several strong verbs which have weak present tenses that follow class I weak verbs. Among those verbs are biddan 'to a s k 1 , sittan 'to s i t ' , licgan 'to l i e ' , faicgan 'to p a r t a k e " , fricgan 'to a s k ' , hebban 'to r a i s e " , swerian "to swear 1 , and scefaban 'to i n j u r e * . The present tense of biddan, for example,

is

just like the paradigmatic model for class I weak verbs given below, fremman; 1sg bidde, 2sg bitst, 3sg b i t ( t ) , pi. biddap. Remarkably, faicgan, besides its strong past peah, paegon

and its

weak present, also has a weak past

tense and past participle pigde, pigede, faiged.

2 . 1 . 2 The -j-

formative

Of the four weak verb classes that existed in Germanic only three survived in OE (see section 1, a b o v e ) . The fourth class as it existed in Gothic contained verbs having "inchoative meaning and derived either from verbs (mostly strong) or adjectives"

(Prokosch: 1 5 6 ) . It

is

often impossible to trace verbs of this class, but it

is

f a i r l y safe to assume that they passed, at least in OE, into other weak classes, waecnian class II,

'to awake 1 , OE weak

is one of the few verbs that seems to preserve

a r e f l e x of the IE stem-formative s u f f i x -na/-rio- of the fourth class ( c f . Gothic g a w a k n a n ) . The three surviving weak classes are mostly derived verbs. And like the fourth class they are traditionally distinguished by their derivative s u f f i x e s in the present tense: the f i r s t weak class is the IE jo-

(Gmc. jji-) class.

According to Brunner ( 3 1 0 ) , the original present tense s u f f i x IE *-e-jo developed to G m c . * - i j a which became -jaafter a short stem vowel and -iaThe second class is the IE adisputed present stem s u f f i x

a f t e r a long stem vowel.

(Gmc. o-) class. The much (see below) seems to have

been Germanic - o ( j ) a , - o ( j ) i (see Wright 1 9 1 0 : 1 5 6 ) . is obvious, judging from the OE paradigm, that it

It

con-

tained a high front, non-consonantal element at least in some forms of the paradigm. Finally, the third class is also referred to as the IE e- (Gmc. -aej-, -a_i-) class. Only four OE verbs are conjugated according to this class. Although much of its development is far from clear, there can be no doubt that the stem s u f f i x included a -j- element. It is rather striking that all three OE weak verb classes appear to have had a -j_- or -_i- element in their stem s u f f i x e s and the question arises as to how this ment can be reconstructed

ele-

from evidence in OE and other

languages. Verbs such as settan "to set 1 or scieppan "to create 1 (OE weak class I) provide no surface m a n i f e s t a t i o n s that

10

actually contain a -j_- s u f f i x .

It is relatively easy,

however, to trace their development in Gothic and Old Saxon, which possess cognates satjan, skapjan and settian, skeppian, respectively. Since

the

same historical pro-

cesses of gemination and umlaut are evident in Old Saxon and OE, the conditioning factor ~2.-r missing in OE, can be recovered. In nerian 'to save 1 , another OE class I weak verb, metric evidence (Brunner: 3 3 2 ) shows that orthographic

i.

must have represented a glide, at least in early OE, which made nerian, derian 'to h u r t 1 , onhyrian 'to emulate' dissyllabic rather than trisyllabic words. Evidence for a phonetic glide [j] is even more convincing in the Mercian Glosses of the Vespasian Psalter (ca. 8 5 0 ) where nerian is consistently spelled nergan, with g_ = t j l - This occurs not only in the infinitive, but in all the other forms that contain orthographic i. in West Saxon, as well: 1 sg. pres. ind. nergu, pi. pres. ind. nergap, sg. pres. s u b j . nerge, pres. part,

(ge)nergende, etc.

Similarly,

other words that are conjugated like nerian also show up with g instead of ,ί: dergan, biscergu, hergende, swergap. The long stem verbs of class I ordinarily give no clue as to the former presence of a stem s u f f i x -j-. There is, however, one exception, the verb ciegan, which has a -j_troughout its paradigm ( c f . Brunner: 3 2 4 , 3 2 6 ) . Class II weak verbs have orthographic i_ in some forms like the infinitive,

1 sg. pres. ind., pi. pres. i n d . ,

the subjunctive, and others. Contrary to class I nerian whose spelling, except for the stem, is identical, class II verbs like l u f i a n or hopian are generally (Brunner 332: "auch nach Ausweis der Metrik und des Me.") considered to be trisyllabic. Other f o r m s , such as 2, 3 sg. pres. ind. or the imperative, contain no hint of a former -j_-, or -i.-. Gothic, Old Icelandic, and Old High German are entirely without surface evidence of -j_- or -_i- in the entire paradigm. This and the fact of the great ortho-

11

graphic similarity between dissyllabic nerian and trisyllabic l u f i a n seem to indicate a problem area of some importance which will be investigated below. Whatever the status of orthographic i_ in OE weak class II verbs,

the

presence of a high front glide or vowel following the stem of some of the forms of the paradigm cannot be disputed. OE class III

weak verbs, which behave like class I

verbs in some respects and like class II in others, also evidence the former presence of a stem-formative -j- in many forms. WS libban 'to live 1 follows in its

geminated

forms class I settan and the same arguments apply. Mercian lifgan 'to live" even retains a glide in several forms of its paradigm. The presence of this high f r o n t non-consonantal element must have played an extremely important role in the prehistoric development of the OE weak verbs. That it was also decisive in documented OE times remains to be shown. 2.2

Weak verb classes In the preceding paragraphs the various OE weak verb

classes have been alluded to many times. In the following sub-divisions of this chapter the paradigms of all

three

classes and their subclasses will be presented in greater detail. The designations class I, class II, class I I I originally referred to the relative frequency of occurence of these verbs, but by the time of standard A l f r e d i a n Old English considerable

changes in class membership had

taken place and class II had become the most numerous and productive class. In this paper the traditional designations will be retained. 2.2.1

Class I

The verb of class I do not constitute one homogeneous block; and although there are some consistent t r a i t s , we have to recognize at least three major subclasses. The first subclass is represented by verbs such as fremman

12

"to d o 1 , trymman 'to strengthen 1 , cynssan 'to k n o c k 1 , jpennan 'to s t r e t c h 1 , and many others. The verbs are characterized as originally short-stem verbs in which the Gmc. stem-final consonant has undergone gemination under certain conditions. A typical WS paradigm would be as follows: pres. indie. 1 fremme

pres. s u b j . 1 fremme (optative)

sg 2 f r e m ( e ) s t

sg 2 fremme

3 frem(e)b pi past indie.

3 fremme

fremmap 1 fremede

sg 2 fremedest

pi

past s u b j . 1 fremede (optative) sg 2 fremede

3 fremede pi

3 fremede

freraedon

imperatives 2sg freme

pi infinitives

2pl fremmap participles present

fremmende

past

fremed

fremmen

fremeden fremman to fremmenne

The second subclass of class I is made up of short stem verbs in r , which did not participate in WGmc.gemination, nerian "to save' is typical: pres. indie. 1 nerie

pres. s u b j . 1 nerie (optative)

sg 2 nerest

sg 2 nerie

3 nereb pi neriap

3 nerie pi nerien

past indie.

1 nerede

sg 2 neredest

past s u b j . (optative)

sg 2 nerede

3 nerede pi

neredon

imperatives 2sg nere 2pl neriap participles present neriende past nered

1 nerede

3 nerede pi infinitives

nereden nerian to nerienne

13

The third subclass of class I is exemplified by the long-stem verb deman 'to j u d g e " . pres. indie. 1 deme

pres. s u b j . 1 deme (optative)

sg 2 demst

sg 2 deme

3 demjp pi past indie.

3 deme

demap

pi

1 demde

past s u b j . (optative)

demen 1 demde

sg 2 demdest

sg 2 demde

3 demde pi demdon

3 demde pi demden

imperatives 2sg dem

infinitives

2pl dema£>

deman to demenne

participles present demende past

demed

Besides these three major subclasses of class I, there are several other verbs or groups of verbs that are normally included in the first weak class but which are irregular. The largest group contains verbs which formed their past tense and the past participle with the dental preterite s u f f i x immediately adjoining the stem without the so-called medial vowel, which is i. in the oldest texts (cf.

Brunner: 3 2 ) , as in neridae and e_ later: nerede. In

some instances the absence of the medial vowel seems to be an IE phenomenon, in others the vowel appears to have been syncopated in Gmc. times. This is of no relevance here, but the missing connecting vowel is responsible for the unumlauted past tenses and past participles characteristic of this group. Brunner (318-321) lists 24 of these weak class I verbs. Some examples will s u f f i c e here: bycgan, bohte, (ge)boht 'to b u y 1 ; cweccan, cweahte, (ge)cweaht 'to s h a k e ' ; pencan, frohte, (ge)poht ' t o t h i n k ' ; secan, sohte, (ge)soht 'to seek'; sellan, s ( e ) a i d e , ( g e ) s ( e ) a l d 'to g i v e ' ; tellan, t ( e ) a l d e , ( g e ) t ( e ) a i d 'to t e l l ' . T h e forms including the vowels in parentheses are WS; excluding the vowels they are Anglian variants

14

which exemplify more clearly the absence of umlaut in the past tense and past participle. It should also be noted that -ccj-, as in bycgan, is the geminated form of -g_- so that the f i r s t three persons present indicative are quite regularly 1 bycge, 2 bygst, 3 bygp in WS. (Likewise, -bb- as in swebban 'to k i l l ' represents gemination of f_.) gierwan ' t o prepare' and several other verbs that retain a high back glide in the infinitive are another subgroup of class I. Their conjugations correspond exactly to fremman in that w appears in those forms in which fremman has geminated consonants: 1 sg. pres. ind. fremme, gierwe, 2 sg. pres. ind. fremest, gierest. Consistent paradigms are relatively rare, however, since these verbs frequently restructure their conjugations by either extending w to all

forms or by eliminating w in all

forms

( c f . Campbell: 3 2 7 ) . It is relatively easy to demonstrate the correspondences that exist between the subclasses of class I: 1. all

stem vowels in the present tenses are umlauted.

The existence of nouns corresponding to deman, sendan, secan : dpm ' j u d g m e n t ' , sand " m i s s i o n 1 , socn 'search' without umlaut and the existence of verb cognates, especially in Gothic (domjan, sandjan, s o k j a n ) , provide clear evidence. 2. the past system is uniform throughout as far as stem and inflectional

endings are concerned with

the

exception of the long stems like deman which syncopate the medial connecting vowel in prehistoric times: demde; but note the. past participle demed, probably due to syllable structure restrictions. The differences and irregularities in class I, on the other hand, are much more telling, particularly about the synchronic state of a f f a i r s . The very fact that class I has so many subclasses and subgroups shows that the classic WS paradigms (Brunner: 4: "streng westsächsisch") are part of a verbal system in transition. Germanic class membership in class I is based on a common -j- or -i-

15

s u f f i x followed by a connecting vowel in the present tense,but in attested OE, even the earliest Northumbrian, this is no longer obvious. A comparison of the conjugations of fremman, nerian, deman points up two problem areas: 1. the synchronic relation of the fremman type to the nerian type or deman type as sub-members of one common class is not immediately evident, especially since the nerian type is a small group. 2. A consistent segmentation of class I verbs into stem and inflectional ending is not possible because of a variety of stem alternants. The two points are really to a considerable degree interdependent because it is precisely a consistently realized stem and inflectional ending segmentation that would, aside from inflectional vowels, help establish a psychologically real class membership in OE. fremman and gierwan, whose handbook paradigms coincide, with the only exception that fremman has geminated consonants in those forms where gierwan has w, are a case in point. If in the present tense -e-, -est/ -ep, -ab are regarded as the inflectional endings, then we have a stem alternation frem- - fremm-. I f , in addition to that, 2 sg. imperative frerne is considered the base form (Erickson 1 9 7 0 : 6, Wagner 1 9 6 9 : 2 1 7 , Keyser 1 9 7 5 : 388, 3 9 9 ) , there are even three stem alternants. 2 , 3 sg. pres. indie, could also be segmented freme-st, freme-b, but this would still leave us with freme- and fremm-. The situation is similar for nerian where, according to the first suggestion above, the stem alternants are ner-, neri-, nere, or, according to the second, neri, nere. deman is more regular. The inflectional vowel in 2, 3 sg. pres. indie, and the vowel of the past tense marker were syncopated in prehistoric times (Brunner 132:" in der Zeit vor unserer Ü b e r l i e f e r u n g " ) , dem- emerges as the stem. But then the past participle demed causes problems. The usually very close connection between the past tense and the past participle is here disturbed: demde, demed. Compared to the other two major subclasses of class I

16

the deman type stands out as a paradigm that would seem to uphold a freme-st, nere-st, freme-b, segmentation, which is,

freme-de, nere-de

at least from a diachronic point

of view, not very likely. A divergence also exists as far as 2 sg. imp. is concerned: freme, nere, dem. The tentative and informal segmentations presented above are not an attempt at a real analysis,but they probably show fairly clearly the variation that exists within and between subclasses of one verb class. It not too d i f f i c u l t

is

to see that the f i r s t weak class was

in an undesirable and untenable condition. If we add to this the complexities of the type bycgan, bohte, with its principal parts nearing suppletion,

(ge)boht,

it becomes

evident that a speaker of OE must have found it

virtually

impossible to recognize a common class membership. 2 . 2 . 2 Class I I The weak verbs of class II present a much more u n i f o r m picture. There are no subclasses with significant membership as in class I, although a handful of contracted verbs follow the class II pattern. The number of verbs in this class is very large, the largest of all weak verbs in OE; and throughout Germanic class II is extremely productive as far as word formation is concerned: "Die eigentliche Produktivität der Verbalbildung liegt im Bereich des schwachen Verbums, in historischer Zeit vor allem bei der 2. Klasse ( I n f i n i t i v auf -on) und ihren Untergruppen." (Krähe/Meid 1969: HI/231 ) . Krähe/Meid also point out that, for instance, in OHG and OS the more productive second class, consisting mainly of denominative and deverbative verbs, attracted verbs of the first weak class. Verbs that belong to class I in Gothic have transferred to class II in OHG and/or OS: Goth, stainjan 'to stone 1 , OHG steinon; Goth, timrjan "to construct 1 , OHG zimbaron, OS timbron; Goth, meljan 'to write, to d r a w 1 , OHG malon. This is a remarkable transfer that we will return to later. Examples of the very many verbs conjugated accord-

17

ing

to class II are l u f i a n 'to love 1 , bodian 'to

a n n o u n c e 1 , endian 'to e n d ' , locian

'to look 1 , hopian

'to h o p e ' . pres. indie. 1 lufige

pres. s u b j . 1 lufige (optative)

sg 2 lufast

sg 2 lufige 3 lufige lufigen

3 lufap pi lufiap past indie.

1 lufode

past subj. (optative)

sg 2 lufodest

sg 2 lufode 3 lufode pi lufoden

3 lufode pi lufodon imperatives 2sg lufa infinitives 2 pilufiab pariciples present lufiende past

1 lufode

lufian to lufienne

lufod

Smeagan 'to think 1 is one of a total of about seven contracted verbs. These verbs are of minor significance, but their paradigm is given below for the sake of completeness : pres. s u b j . 1 smeage (optative) sg 2 smeage

pres. indie. 1 smeage sg 2 smeast 3 smeab

3 smeage

pi smeagap past indie. 1 smeade sg 2 smeadest 3 smeade pi smeadon imperatives 2sg smea

pi

past subj. 1 smeade (optative) sg 2 smeade 3 smeade pi smeaden smeagan infinitives

2pl smeagap participles present past

smeagen

to smeagenne

smeagende smead

Unlike class I, class II verbs do not distinguish between long and short stems. Their stems, at least super-

18

ficially, do not seem to o f f e r particular problems for segmentation because there are no geminated forms in class II, and verbs with umlauted and unumlauted vowel in present and past tense, respectively, do not exist. In f a c t , verbs in class II characteristically have an unumlauted stem vowel throughout the paradigm as a comparison between the verb lufian and the noun luf u "love 1 shows. Yet, OE class II has several complex forms that have troubled linguists for some time. The internal development of the forms is problematic. A look at the paradigm reveals the striking differences between 1 sg pres. ind. lufige on the one hand, and 2, 3 sg. pres. ind. lufast, lufab on the other. All class I verbs have an early OE i. and a late OE e as inflectional vowel for 2, 3 sg pres. ind., and often this vowel is syncopated, not only in long-stem verbs. In class II this vowel is always a and, as Brunner puts it: " [ d ] a s a der 2. und 3. Sg. Ind. ist im Ws. ganz [my emphasis—DS] fest" (333) . This difference has led to much speculation about the prehistoric development of class II. The two most important positions will be presented in the following paragraphs because of their fundamental importance for an understanding of class II. The discussion centers around the question wether the inflections of the present represent a mixture of thematic and athematic forms or whether Gmc., and thus OE, class II verbs can be explained on the basis of Proto-Indo-European ( P I E ) thematic present stems alone. If a verb in PIE is conjugated athematically, the inflectional endings immediately follow the root or verbal stem, as for instance 3 sg pres. ind. *es-ti "he is'. In the case of the thematic conjugation, a theme or connecting vowel is inserted *bher-e-ti 'he bears'. Krahe/Meid ( 1 9 6 9 : I I / 1 2 1 - 2 ) , who argue for a mixture of conjugations in weak class II, suggest the following conjugations of the PIE present forms:

19

Athematic 1 -a-mi

Thematic 1

-a-jo

sg 2 -a-si

sg 2 -a-je-si

3 -a-ti

3 -a-je-ti

1 -a-mes

1 -a-jo-mes

pi 2 -a-te 3 -a-nti

pi 2 -a-je-te 3 -a-jo-nti

In the thematic paradigm the theme vowels

and e_ are

preceded by a j-element. The reason why a mixture of thematic and athematic inflections may have come about is that it is "anzunehmen, daß bei den thematisch gebildeten Verba in den Formen mit idg.

e_ > germ. i. als

Bindevokal das

j früher schwand als in den Formen mit Bindevokal o > a., daher 2. Sg. germ. *salbojis > *salbois

(ent-

sprechend die 3. Sg. und 2. P L ) , und daß in diesen Formen dann weiter -oi- zu-o-kontrahiert wurde, also 2 . Sg. -ois > -os usw. Da aber auch die 2. Sg. der athematischen Verben (idg. -a-s_i) zu -os geführt hatte, fielen beide Typen in dieser Form (und entsprechend in der 3. Sg. und 2. P l . ) zusammen. So war der Weg für eine Vermischung der thematischen und athematischen Flexion freigemacht."

(Krähe/Meid 1969:

11/122) Based on the thematic vowel, a convenient distinction can be made between e-forms and o-forms in PIE and, correspondingly, between informs and a-forms in Germanic. In OE weak class II paradigms these forms are clearly reflected, no matter what their origin was. The o-forms comprise 1. sg. pres. ind. inf.

l u f i g e , pi. pres. ind.

lufiap,

l u f i a n , pres. part, lufiende, 2. pi. imp. lufiap,

sg. pres. s u b j . l u f i g e , pi. pres. subj. lufigen. The £-forms are 2. sg. pres. ind. l u f a s t , 3. sg. pres. ind. l u f a p , 2. sg. imp. l u f a . Krahe/Meid, then, are of the opinion that OE preserved thematic forms at least in

20

1. sg. pres. ind. and pi. pres. i n d . : "Das Angelsachs. hat in der 1. Sg. und im Pl. deutlich thematische Formen bewahrt" ( 1 9 6 9 : 11/123). Cowgill ( 1 9 5 9 ) , on the other hand, claims that the inflection

of the present forms of the Germanic second

weak conjugation can be explained by positing in PIE an exclusively thematic -aje-/ajo-. Loss of

intervocalic *-j-

in pre-Gmc. and subsequent contraction then brought about a seemingly athematic and u n i f o r m paradigm preserved

in Gothic ( s g . pres. ind.

pi. pres. ind.

1 labo, 2 labos, 3 lapop;

1 lapom, 2 lapop, 3 lapond; i n f .

lapon

' t o i n v i t e 1 ) . But in W G m c . , and therefore OE, this conjugation was replaced with an-o -/-oja-conjugation. In the PIE e-forms PIE *-aje-went via *-oji-to Gmc. *-o-. W G m c . , and more precisely Ingveonic (Old English, Old F r i s i a n , Old Saxon) *-oja- does not derive directly from PIE *-ajo- as had been generally assumed but, according to Cowgill, "is an innovation, one more in an area already characterized by such notable innovations in verb morphology as loss of distinct 1st and 2nd person plural forms and almost complete absorption of the third weak conjugation into the second". ( 1 9 5 9 : 7 ) . What had happened was that the conjugation to which the OE weak class II verbs were to belong had been reformed on the analogy of the frequent and numerous

long-stem verbs of the weak class

I.

The stem s u f f i x e s of the pre-historic f i r s t weak class in WGmc. are reconstructed as *-i- and *-ija- for the PIE e_-forms and o-forms, respectively. Analogical remodeling of the o_-forms then took place according to the proportion: domi : domija = l u f ο

: X. X is

l u f o j a . A similar

analogy, without details, is also adduced for the third weak class, in which changes had been even more dramatic. In OE only four verbs remain in this class, the others having most likely passed into class Brunner ( 3 2 9 )

II.

and Campbell ( 3 3 3 , 1 4 8 ) assume that -o-

in the e-forms remained unchanged from Germanic times. The second weak class is therefore also referred to as

21

the o-class. -oja- of the o-forms developed to -ejawith the help of umlaut, became shortened to -eja- and was assimilated to -ija-. -ija-then contracted 1

locian "to l o o k ,

to -ia-

:

etc.

The past tense and past participle are quite regular. They all

contain a stem s u f f i x which in West Saxon is

-o- : lufode. This connecting vowel again distinguishes class II from class I where the vowel is generally -efrom older - L - , or syncopated: fremede, demde. One important aspect of the weak class II verbs is the phonetic realization of the -_ig_-/-_i- element in the o-forms. This element, quite consistently spelled i in the infinitive and the pi. pres. ind. and -ig-

in 1 sg.

pres. ind. and the subjunctive in EWS, was always labic, i.e.

syl-

[ i ] , so that locian, l u f i a n , bodian, endian

and a large number of similar class II verbs were

all

trisyllabic words. Keeping this in mind, we now look at the class I nerian type and f i n d that the orthographic similarities, especially in later WS where the -ig-

is

f r e q u e n t l y s i m p l i f i e d to -i.-, between, for instance, nerian and l u f i a n , are indeed remarkable. Except for the i n f l e c t i o n a l vowel in 2, 3 sg. p r e s . ind. and the connecting vowel in the past tense and the past participle the paradigms are identical. Both i n f l e c t i o n a l and connecting vowels are,

however, in unstressed position and

are therefore very unlikely to retain their f u l l vocalic value in spoken OE. It

is a well-known f a c t that ortho-

graphy is more conservative than the spoken language,and it

can be s a f e l y assumed that unstressed vowels in medial

syllables or elsewhere in the word have a tendency to move toward the mid central position in the vowel t r i a n g l e , i.e.

toward s c h w a [ 0 ] . If we take this development into

account, the nerian and l u f i a n paradigms become v i r t u ally identical. If an informal segmentation stem versus inflectional ending is attempted, the problems are again the same as in nerian where the forms exhibiting orthographic -_i- or -ig-

are odd. Aside from this ques-

22

tion, luf- as. the stem (or ner- for nerian) seems to be an obvious choice if we are only interested in the regularity of the stem proper. The considerable uniformity of the class II conjugation, the conspicuous similarity of the nerian type of class I and the absence of divergent subclasses are important factors in diachronic and synchronic analyses which will be the topics of chapters 2 and 3. 2 . 2 . 3 Class I I I The four verbs which make up the third weak class in OE are: habban 'to have 1 , libban 'to l i v e ' , secgan 'to s a y 1 , hycgan 'to t h i n k ' . All four WS paradigms of this highly irregular and non-productive class will be given. pres. indie,

1 hasbbe

pres. subj. (optative)

1 hasbbe

sg

2 haefst

sg

2 hasbbe

3 haefp

pi past indie,

pi

habbab 1 haefde

sg

3 haebbe past subj. (optative)

2 haefdest

sg

3 haefde pi imperatives 2sg 2pl infinitives

haefdon hafa habbab

1 haefde 2 haafde 3 hasfde

Pi participles

present

habban past inflected infinitive not attested

hasbben

haefden haebbende haafd

23

pres. indie. sg

1 libbe 2 leofast

pres. subj. (optative) sg

libbab 1 lifde

past indie, sg

2 lifdest

pi past subj . (optative) sg

3 lifde

pi imperatives2sg 2pl

infinitives

lifdon leof a libbab

1 secge

sg

2 saegst

sg

pi participles present

pres. subj. (optative) sg

imperatives2sg 2pl infinitives

lifden libbende

secgab 1 saegde 2 saegdest

saegdon saege secgab

1 secge 2 secge 3 secge

pi past subj . (optative) sg

secgen 1 saegde 2 saegde 3 saegde

3 saegde

pi

2 lifde 3 lifde

3 saegb

past indie,

libben 1 lifde

libban inflected infinitive not attested

pres. indie,

pi

2 libbe 3 libbe

3 leof ab

pi

1 libbe

pi participles present past

secgan inflected infinitive not attested

saegden secgende saegd

24

pres. indie.

1 hycge

pres. s u b j . (optative)

1 hycge

sg

2 hygst

sg

2 hycge

3 hygp

pi past indie,

pi

hycgap 1 hogde

sg

3 hycge past s u b j . (optative)

2 hogdest

sg

3 hogde

pi imperatives 2sg 2pl infinitives

1 hogde 2 hogde 3 hogde

pi

hogdon hycge

hycgen

hogden

participles

hycgap

present

hycgende

past

hogod

hycgan inflective

infinitive

not attested Similar to class I, but to a much higher degree, class III

verbs in WS are at a stage in their development

that seems to require imminent reduction of allomorphy. A look at ME shows that verbs of this class had transferred to one of two categories of weak verbs, namely to that subgroup of class I weak verbs that formed

its

past tense without a connecting vowel between stem and past tense marker -d-. It is quite striking, of course, that this highly irregular OE class comprises words that must have been extremely common: have, live, say,

think.

This was probably the only reason that made it possible for these verbs to survive into late OE times. It is frequently assumed that, like class II weak verbs, the Gmc. class III

is a mixture of thematic and athematic

forms ( c f . Krahe/Meid 1 9 6 9 : I I / 123-4) which are most clearly represented in the Gothic paradigm (see section I ) : 1 sg. pres. ind. pres. ind.

haba, 2 habais, 3 habaip; 1 pi.

habam, 2 habaib, 3 haban; i n f .

haban ' t o h a v e ' .

In the non-Gothic dialects of Germanic, and thus in OE, an intrusion of the f i r s t weak class is often seen as being responsible for those forms which contain gemination and/or umlaut (Bennett 1 9 6 2 : 1 4 0 ) . Forms with gemination

25

in class III correspond exactly to those which have a geminated consonant in class I, a group that was earlier referred to as the o-forms; i-umlaut is not quite so regular: in the geminated forms only hycgan and secgan are umlauted, whereas habban according to Brunner ( 3 3 9 ) is not affected in those forms, ae in 1 sg. pres. ind. haebbe is explained in Brunner ( 7 3 ) as typical of words "in denen in verschiedenen Flexionsformen ... ae und a nebeneinanderstehen, so daß nachträglich a an Stelle von ae eindringen konnte und durch i-Umlaut verändert wurde." If habban had followed class I consistently, the expected stem vowel in the geminated forms would habe been e_, as in the secgan paradigm. The a in i n f .

habban and pi. pres.

ind. habbab "muß erst nach [my emphasis] der Zeit des i-Umlauts übertragen worden sein." (Brunner: 3 4 0 ) . The most consistent feature of all

four class III

verbs in OE is the syncopated medial vowel in the past tense and also in the past participle, except hogod. In class I only the originally long stem vowels (deman) syncopate the medial vowel in the past tense but never in the past participle. Dialectal variation in this class is more comprehensively treated in 3 . 2 . 3 . It is, however, necessary here to refer to the doublets libban, lifgan 'to l i v e ' , of which only libban is standard WS while lifgan is found in the Mercian Vespasian Psalter and Rushworth Gospels . The presence of a non-syllabic stem s u f f i x -j_- in these dialectal forms is usually considered a conclusive form feature of class III,

where they appear in the so-called

o-forms. The e_-form 3 sg. pres. ind.

in the Vespasian

Psalter is leofap. The third characteristic of class III is, as Campbell points out ( 3 3 9 ) : "the appearance side by side of forms with and without (a) i-umlaut of the root vowel;

(b) con-

sonant gemination," although "3 (a) and (b) by themselves may point only to the existence of Class I and Class II conjugation side by side, e . g . , f y l g a n , f o l g i a n ; t e l l a n ,

26

talian."

(339, fn. 4).

By the same standards that were applied for class I, it

is possible to state with certainty that secgan,

hycgan, libban, habban must have once had a stem formative -_j_- which caused consonant gemination in the stem and also umlauted the stem vowel, although this may have been later obscured by other processes, as for instance in habban. The Mercian lifgan is more d i f f i c u l t to explain but one version that is accepted by Campbell ( 3 4 1 ) is that the stem formative was -aej-

with "the preceding

vowel [ i . e . , ae] having been syncopated too late for consonant gemination to take place, but s u f f i c i e n t l y early for -j- to cause u m l a u t . " A third stem s u f f i x -deis found in 2, 3 sg. pres. ind. g u l a r . The inflectional vowel is

and the imperative

sin-

syncopated in WS hce:fst,

saegst, and hygst but in habban and secgan the 2 . 3 sg. pres. ind.

have no umlauted stem vowels (haefst,

saegst),

a fact that d i f f e r e n t i a t e s them from the f i r s t weak class. The problem here is of course the number of s u f f i x e s posited for class III. fix -_j_-, class III

Where class I needs one stem suf-

needs three. Two d i f f e r e n t

suffixes

are required in l i f g a n of the Vespasian Psalter: 1 sg. pres. ind.

l i f g u is seen as containing the s u f f i x -aej-

but third person singular pres. ind. paradigm has -ae-.

leofab of the same

If WS is considered as well, the

third s u f f i x -j_- is required in WS libban in contrast to Mercian -333- in l i f g a n . Bennett ( 1 9 6 2 : 1 3 4 ) who proposes, at least for Gothic, one common Proto-Gmc. s u f f i x -aja-, sums up the situation by stating that " [ a ] m o n g the types of weak verbs occurring in Germanic, none has provoked more d i f f e r e n c e of opinion than that of Class III,

and no other has so consistently eluded a believable

analysis."

27

3.

Variations in the OE system

3.1

Early Old English versus Late Old English The handbooks ordinarily distinguish four dialects

in OE: Northumbrian ( N o r t h . ) , Mercian ( M e r c . ) , Kentish ( K e n t . ) , and West Saxon ( W S ) . Mercian and Northumbrian together are frequently referred to as Anglian. The earliest OE documents are of Northumbrian origin and probably date from around the year 700 A . D . The latest are West Saxon from the eleventh century. West Saxon occupies a central position in the study of OE because of

the wealth of documents preserved in this dialect

and because of the political and cultural importance of Wessex in the years 900 - 1000 A . D . All standard handbooks (Brunner, Campbell, Wright, Bülbring) use WS as a grammatical norm and the paradigms and forms given usually represent Early West Saxon (EWS) from around the time of A l f r e d the Great, roughly the last third of the ninth century and the beginning of the tenth. The more important works of this era include the Parker MS of the Old English Chronicle, A l f r e d ' s translation of the Cura Pastoralis, and the Lauderdale MS of A l f r e d ' s translation of Orosius. Late West Saxon (LWS) can be found in numerous works of the Abbot Aelfric

(around

1000 A . D . ) . A term frequently used in German handbooks is "streng westsächsisch", classical West Saxon, which comprises texts that go back to A l f r e d ' s and A e l f r i c 1 s spheres of influence and which are characterized by shared orthographic and scribal features. These texts fall roughly into the time between 900 and 1000 A . D . In the following paragraphs several differences between Early Old English ( E O E ) , which need not always be West Saxon, and Late Old English ( L O E ) , which is practically always West Saxon, will be pointed out to the extent that they are important for the development of the OE weak verb system.

28

3 . 1 . 1 EOE versus LOE variations in class I One of the most noticeable traits of the earliest texts, particularly if compared to LWS or Middle English (ME) texts, is the existence of orthographic f u l l vowels A' fE' —' — ' — ' — ' X i n un ^ccented (unstressed) syllables. By ME times, in some instances much earlier, almostall these vowels have become orthographic e_ in unstressed position, although some seem to be holding out better than others. Campbell ( 1 5 6 ) notes that in contrast to ae,

i., u unaccented a. is f a i r l y stable in OE till the

late period. The front vowels ^, e, ae had already coalesced to e_ when a was still consistently realized, for example in the inflectional vowels of the 2, 3 sg. pres. ind. of the second weak class. If we look at the earliest attested forms that constitute the paradigm of weak class I fremman, it will be noted that 1 sg. pres. ind. ended in u: frenunu, or its

later equivalent -o. In the 2 sg. pres. ind. the

oldest texts have the inflectional vowel _i followed by s: fremis. Similarly, 3 sg. pres. ind. is fremib. The pi. pres. ind. appears as in classical WS: fremmap. The singular and plural of the present subjunctive are attested in the oldest texts as fremmas,

fremmaen,

respec-

tively. In the past tense the 2 sg. ind. is f remidaest (t) with the medial vowel of the past tense s u f f i x still represented as -i.- and an inflectional vowel -ae-. The plural ends in -un, which is later replaced by -on and also -an. The past subjunctive endings a r e , like in EWS, -e_ in the singular and -en in the plural. In the imperatives freme is attested for the singular, fremi is called "uraltenglisch" (primitive OE) by Brunner ( 1 2 5 ) but does not seem to be documented although evidence in Old Saxon (OS) clearly points in this direction (see Krahe/Meid 1 9 6 9 : 11/132). The plural imperative is like the pi. pres. ind. fremmap and remains f i r m for a con-

29

siderable time. Likewise, the uninflected i n f i n i t i v e is fremman in the oldest texts and the inflected infinitive is f remenne, although EWS has frequent f remmanne, and occasionally fremmonne. Fremmaendi or fremmendi is the EOE form of the present participle, fremid is common and frequent for the past participle in the earliest documents. Compared to the forms of the standard paradigm ( 2 . 2 . 1 ) , the earliest forms display a diversity of vowels that is relatively quickly reduced and weakened to orthographic e and thus phonetic [9] by the f i r s t half of the tenth century. In EWS only unstressed a, as for instance in the i n f i n i t i v e , and unstressed o in the pi. past ind. survive. In LWS and early ME even these orthographic contrasts have been eliminated and e is common throughout. These vowels must have become [9] in speech long before this was reflected in orthography. Modern English shows analogous processes. The pronunciations of [ ' e v r i ] 'every',

['bxzms]

'business 1 and similar words attest

to the continuing productivity of vowel reduction, which proceeds from f u l l vowel to schwa to zero. In the nerian type of class I, which is generally like fremman in the forms discussed there, an interesting change, "parasiting", can be observed in the forms containing [j] in later WS texts: EWS nerian appears frequently as nerigan and nerigean. This change can also be observed in many other words (see Campbell: 1 5 2 ) . The conclusion is unavoidable that these spellings represent [ i j ] or simply [ i ] , which is also documented (Campbell: 152: byriweard < byrigweard < byrgweard 'townguard' and other examples). In other words, the spelling n e r i g ( e ) a n is indicative of a trisyllabic pronunciation that must have been current much earlier. When the medial -i- became syllabic, the pronunciation of nerian and other verbs of this subclass must have become identical with class II l u f i a n , hopian, etc.

This striking similarity

between a subclass of class I and class II points up a

30

tendency among the weak f i r s t class which considerably changed the OE verbal system. In WS verbs with an originally short stem vowel, the types fremman and nerian transfer to weak class II so early that geminated forms are not even attested in classical WS, except for fremman itself as well as trymman and cynssan, all

three of

which maintain their original class membership even in LWS. But even here forms like fremian and trymian exist side by side with fremman and trymman, and indeed the WS Cura Pastoralis has only gremian, lemian, temiari, behelian, etc., all of which are verbs that should actually have geminated stem consonants in the o-forms. The subgroup bycgan, bohte, geboht, which adds the past s u f f i x without medial vowel, does not participate in the transfer to class II but keeps its class I

conjugation.

The whole question of the transfer from class I to class II will be discussed in detail in Chapter 4. 3 . 1 . 2 EOE versus LOE variations in class II What was said about class I as far as the earliest attested unstressed vowels are concerned is also true for class II,

with the exception of those unstressed

syllables which contain a and o. Thus for instance, 2, 3 sg. pres. ind. l u f a s t , l u f a b , imp. sg. l u f a , and the medial vowel o in the past tense and past participle: lufode, lufod. The a in particular is very f i r m throughout the WS period, as Brunner ( 3 3 3 ) , with uncharacteristic ist

emphasis, testifies: "Das a der 2. und 3. Sg. Ind. im WS ganz f e s t . " The past tense and past participle have, in the earli-

est texts, u as medial vowel, so that in the past singular we have lufudae

( 1 , 3 person) and in the plural

l u f u d u n . Forms like these are found in the Mercian Epinal Glossary. In EWS the medial vowel is very generally o, but all dialects have examples of -ed-, even as early as the ninth century in EWS and in Mercian in the Vespasian Psalter, and, of course, much more frequently in

31

LWS. Campbell ( 1 5 8 - 9 ) writes that "there is a very strong tendency for the f i r s t of two successive unaccented back vowels to be reduced to a sound written e and hence forms are frequent like fugelas, roderas, heoretas, gedwime.ru from f u g o l , rodor, heorot, gedwimur." This is of course the beginning of the tendency r e f e r r e d to several times above and which is productive even today in words such as every and business. The inflected i n f i n i t i v e and the present p a r t i c i p l e show an early development which proves that long-stem and short-stem verbs were not always conjugated indistinguishably in class II. The original endings for long-stem verbs for the inflected i n f i n i t i v e and the present participle must have been -enni and -endi, respectively, whereas the short-stems had -genni and -gendi. The Corpus Glossary has two forms, tacnendi ( i n f . tacnian "to indicate, m a r k 1 ) versus dobgendi ( f r o m *dofian 'to be d o t i n g ' ) / that show this early state. 3 . 1 . 3 EOE versus LOE variations in class

III

Class III follows, in general, those patterns outlined in 3 . 1 . 1 . The transfer to class II, portant feature of class III

which is such an im-

development, must have oc-

curred in pre-OE times because in OE only four verbs are attested in the various documents. In class I the transfer to class II occurred in historical times,

i.e.,

during the OE period. This must be considered a significant d i f f e r e n c e between the two verb classes. The prehistoric transfer will not be discussed in this study. The major distinctions between various forms ( e . g . , libban vs. l i f g a n ) most often have a dialectal rather than a temporal basis and will be treated in 3 . 2 . 3 . Yet there are some forms that seem to suggest a temporal tendency toward reanalysis in the third class. In LWS the present system of secgan has levelled e_ throughout: 1 secge, 2 s e g ( e ) s t , 3 s e g ( e ) b , pi. secgap. In EWS hycgan has 3 sg. pres. ind. hogap although hygp, as in the paradigm, seems to be as f r e q u e n t . In LWS past hogode and

32

past part, hogod are normal. 3.2

Dialectal variation in the OE period In 3.1 the major dialect divisions were discussed;

they shall not be repeated here. Since EWS is the grammatical norm for all modern treatments of OE, it will be used as the form against which forms from other dialects will be compared. Some of these forms identify a tendency that appears to involve more and more forms as the end of the OE period approaches. Only the most noteworthy variants will be covered in the following paragraphs. 3 . 2 . 1 Class I In class I only a few points are of significance. In the oldest Anglian ( i . e . , Mercian and Northumbrian) texts, the types fremman and nerian are

distinctly conjugated

subclasses of class I. This state is clearly reflected in the EWS standard paradigms,although it

is important

to note that in the Southern dialects, WS and Kentish, this clearly identifiable division became gradually obscured mainly because of the transfer of many class I verbs to class II and the increasingly noticeable reduction of unstressed vowels to [ 3 ] . In the Vespasian Psalter, the quality of the orthographic i. in the nerian subclass is conspicuously revealed to be [ j ] , which is the standard paradigm norm, because nerian is orthographically nergan, containing the letter g ( = [ j ] )

in all

forms that have -_i- in the

standard paradigm. This is usually considered the best proof that nerian was dissyllabic and that orthographic i_ in WS for instance is a high front glide. (Interestingly enough, in the same MS class II verbs, which also have orthographic -JL- in the standard paradigm, as in l u f i a n , are consistently rendered with -i.- pointing toward a vocalic pronunciation, in contradistinction to the orthographic c j . ) Later the glide quality of -_i- in nerian becomes more and more ambiguous because of variant spel-

33

lings nerigan and nerigean (see above 3 . 1 . 1 ) . These spellings in LWS leave little doubt that the original [j]

had become syllabic.

3 . 2 . 2 Class II EWS and LWS practically always have -ig- before e? and -j.- before a.. This corresponds to the standard paradigm. The Vespasian Psalter has -_i- before all Rushworth Gospels

( M e r c . ) has -ig-

vowels but the

rather consistently

preceding vowels, just like N o r t h , texts. The Rushworth Gospels

are also remarkable because they seem to have

reduced -a- to -e- in 2, 3, sg. pres. ind. and imp. sg. before this can be detected in texts of other dialectal origins: l u f e s t , l u f e p , l u f e . The medial vowel in the past tense divides WS on one hand and Kentish and Anglian on the other. WS has prevailing -o- there: lufode and very rarely -a-,

whereas -a_- is the rule in Kentish

and Anglian. Reduced forms (-ed-) are relatively but can be found in all

rare

dialects. This also applies to

the past participle. One of the main d i f f e r e n c e s

bet-

ween Anglian and the other dialects are the forms of the inflected infinitive

and the present participle. While

WS has -ienne for the inflected i n f i n i t i v e , and -iende for the present participle for all class II verbs, the Anglian dialects distinguished long-stem and short-stem verbs by writing -enne and -ende a f t e r long stems and -genne and -gende a f t e r short stems, -genne and -gende were later replaced with -ienne and -iende and extended to all verbs. 3 . 2 . 3 Class I I I So many dialectal variants exist in class III that it

is not practical or necessary to discuss these forms

in detail here. Also, many of the attested dialectal variants do not constitute complete paradigms so that a great many forms have to be guessed, even within one dialect or one given document (see a p p e n d i x ) .

34

The most important verb is perhaps libban (VJS) . Unlike the other three verbs it has corresponding forms in all three non-VvS dialects that contain - f- in all

forms.

In other words, there is no alternation between -f_- and -bb- as in WS. Vespasian Psalter has the following documented forms:

1 sg. pres. ind. l i f g u , 3 sg. pres. ind.

leof afc, pi. pres. ind. 1 if gap/ sg. pres. sub j . l i f g e , pres. part, l i f g e n d e , past part. l i f d . In LWS there is one occurrence of the verb l i f g a n / l i b b a n that has gone over completely to class II and whose infinitive leofian,

is

lifian.

habban, secgan, and hycgan are without non-WS doublets that preserve a d i f f e r e n t stem. They all

have geminated

forms but their distribution is not consistent as in the f i r s t weak class. For example, the Lindisfarne Gospels ( N o r t h . ) have 1 sg. pres. ind. hafo but pi. pres. ind. habbap. 4.

Summary

The reader will undoubtedly have noticed the rather forbidding nature of the data—forbidding in two ways: first of all,

the sheer mass of variant forms which can-

not always be unambiguously assigned to one particular dialect at one particular time; and, on the other hand, the relative paucity of data from older periods and from the northern dialects, which would allow one to follow the history of the development of the standard forms with a much greater degree of certainty. Superimposed on this purely factual problem is the problem of the traditional analyses, which group OE weak verbs into three classes. Even a cursory look at the data we have presented in this chapter shows how inappropriate this classification is for describing the actual OE situation. In f a c t , those linguists who, like 2 Quirk and Wrenn (1957 ) , write for a student audience rather than for the specialist, simply describe two classes of weak verbs, and list the four members of the

35

traditional class III under the category of irregular verbs. As we shall see in the deeper analysis in the next two chapters, this is really a far more satisfactory descriptionof the OE situation, since in a major lexical category, a class which has only four members which do not systematically share any semantic features, can hardly deserve the name class. Moreover, this 'class 1 has no features unique to i t s e l f , but is merely a mixture of certain features of class I and class

II.

One might also question the assignment of the nerian subclass of class I to the same class as fremman and deman It might, on synchronic grounds, be better to consider class I as j u s t consisting of fremman and deman, and class II as consisting of sub-types l u f i a n and nerian. We will have much more to say about this later. Another important point to realize is that the orthography does not actually represent the spoken language. The time lag between the phonetic features at a given stage and their representation in the written language was alluded to earlier. But it is also important to consider that certain stages of the development may never be recorded at all, simply because the inventory of orthographic symbols was fixed prior to the change and could not be represented uniquely. We believe that this is an important factor in trying to establish the chronology of vowel reduction in the back vowels and will discuss it

in the analysis of class II verbs in chapter two.

Here it should simply be noted that while the orthography is always the starting point in the investigation of the sound system and its development, one must be prepared to draw reasonable conclusions about the meaning of spelling conventions on the basis of universally accepted phonetic developments. The conclusion to be drawn from all data as we have it,

this is that the

and, more to the point, as we do not

have it, lends itself to a wide variety of interpretations no matter what linguistic model one is operating in. The

36

various attempts that have been made to describe the classical OE paradigms and their development from pre-OE times have all

had to face the mass of variation in the

data. Some analyses have tried to cope with the data by opting for very abstract phonological representations often going back as far as Proto-Indo-European, at least for the stems. The validitiy of such analyses will concern us in chapters two and three.

37 CHAPTER TWO: THE DIACHRONIC DEVELOPMENT OF THE OLD ENGLISH WEAK VERBS

1.

Introduction In the preceding chapter the system of the Germanic

weak verbs and in particular that of the OE weak verb as it

is commonly given in the standard handbooks were

discussed. The present chapter attempts to relate the Germanic and the EOE stage on an exclusively historical (diachronic) basis. Phonological alternations and variations within attested Old English and the comparative evidence of other Germanic languages (mostly Old High German [OHG] and Old Saxon [ O S ] ) f r o m about the

same

period will allow us to establish with some degree of certainty a set of major rules which r e f l e c t the historical development up to the handbook paradigms. In the section following this introduction each of these phonological rules will be described and discussed individually. There will be frequent references to the data at hand or the treatment of related topics in recent articles and books. The diachronic order of rules must, of course, remain to some degree tentative, but it is possible to establish a reasonably accurate succession of rules that can be verified against the background of OE and other Germanic data. In the third section, the weak verbs will be shown to undergo the diachronic rules given in section 2. Full derivations will be given for five representative forms of each paradigm: the i n f i n i t i v e , 1 sg. pres. i n d . , 2 sg. pres. i n d . , 2 sg. imperative, and 2 sg. past indicative. These derivations, it should be emphasized again, are actually the somewhat generalized development of these verbs between a Proto-Germanic and an e a r l y Old English

38

stage. In most generative phonological analyses those derivations could also be shown to be synchronically adequate, if fairly abstract underlying forms and the appropriate rules are believed to represent psychologically real states for the native speaker of OE. This view of a synchronic grammar is rejected in the present paper (see section 4 ) . As we will also see, the rules of section 2 are s u f f i c i e n t to account for the development of the large majority of class I weak verbs. Those verbs of class I which cannot be f u l l y accounted for by these rules will be shown to involve one of the crucial forces in the development of OE weak verbs: the j formative

al-

ready alluded to in chapter one. In class II this s u f f i x also plays an important role, in the sense that its absence or presence is responsible for the divergent appearance of class II as compared to class I.

Finally,

the irregular verbs of class III provide some evidence for the processes partially obscured in class I. The final section of chapter two will be concerned with some basic theoretical issues. The standard approach to generative phonology, as represented by Chomsky/Halle, The sound pattern p_f English ( 1 9 6 8 ) will be discussed, as well as its

applicability to problems of historical

phonology. Underlying representations (= proto forms) and diachronic rules from sections 2 and 3 of this chapter will be investigated as to their validity and plausibility in synchronic analyses in the framework of generative phonology, or, put in another way: what is the relationship between historical development and underlying forms in a synchronic analysis? Most importantly, it will be claimed that recoverability of historical processes is not identical with their productivity at a given stage and that such processes may not be j u s t i f i a b l y considered as part of the native speaker's competence. This is directly related to the abstractness controversy (Kiparsky 1968 b) in generative phonology. The distinction between recoverability and productivity will lead us, in chapter

39

three, to postulate underlying phonological representations which are maximally concrete without sacrificing those generalizations which can reasonably be assumed to characterize processes that had psychological reality for the speaker of OE. 2.

Diachronie rules All references in this section will be to forms given

and discussed in chapter one, especially the earliest attested data, which are crucial for the determination of the diachronic rules and the protoforms to which they apply. All rules will be discussed with reference to the weak verbs of OE, especially the class I weak verbs which require a f u l l e r set of rules than class II,

because they

are the ones that exhibit the greatest variety of stems and inflectional endings. The special problems associated with the historical development of class II will be brought up in section 3.2 below. Many or all of the diachronic rules discussed here w i l l , of course, also be of relevance in other areas of OE ( e . g . , its noun morphology) , but their analysis will in this study be limited to the weak verbs. The precise absolute, and even relative, chronology of many of these rules has been a controversial matter which is unlikely to be solved unequivocally. For while some rules must be regarded as belonging to a specific sequence of changes, the scarcity and incompleteness of the data makes such determination impossible for many other rules. For example, the evidence available makes it clear that gemination was conditioned by the j. s u f f i x of the present stem of the verb. The deletion of this s u f f i x must therefore have been a later change. On the other hand, it

is virtually impossible to determine from

the OE evidence alone the relative order of palatalization and i-umlaut. Both are attested from the earliest documents and both share similar conditioning factors; for i-umlaut it is a high front non-consonant, for palatalization

40

the conditioning factor must at least be front and nonconsonantal. The order in which the rules will be discussed reflects the diachronic sequence to the best of our knowledge. 2.1

Glide Vowel Coalescence The term "glide-vowel coalescence" is probably a mis-

nomer in that the rule itself will be formulated as a deletion rule. Yet it

can be assumed that in a situation

where a high front vowel and a high front glide come to stand next to each other, the vowel as the hierarchically stronger segment ( c f . Lass 1971: 17) will not just eliminate the weaker glide but the glide will assimilate to the vowel to such a degree that the f i n a l e f f e c t is one of deletion.

(See also Mahlow's statement

[ 1 8 7 9 ] , quoted

in Cowgill [ 1 9 5 9 : 6 ] : jedes j_ vor . . . . i_ verloren gehen muss, ausgenommen vielleicht im A n l a u t . " ) At any rate, in 2, 3 sg. pres. ind. j_ and i, were contiguous when the stem formative j and the i_ of the earliest inflectional vowel were as in * frame + j + is, which is the form from which OE fremest developed. Kiparsky and Ο ' N e i l in their paper on "The phonology of Old English inflections" absorption

( 1 9 7 6 ) state a rule of

i-

( 5 4 2 ) where in the same environment—except

that they assume the stem-forming s u f f i x to be i_--the second element is deleted. Aside from the fact that the deletion of one or the other of the elements is not motivated in this rule since both are i, Modern English evidence, as for instance in two [ t u ] , seems to indicate that in a sequence glide plus vowel the f i r s t of the two is deleted, which in this case happens to be a high back glide followed by a high back vowel. High vowels and glides, whether front or back, often react in a parallel manner. It is also interesting that, although Kiparsky/ O ' N e i l ' s formulated rule deletes the second element, they carelessly delete the first in their derivation ( 4 6 ) ( 1 9 7 6 : 5 4 4 ) . In a footnote ( 5 4 8 ) , they claim that since

41

it is immaterial which element is deleted, they "arbitrarily" delete the second, "for at this stage of the language there is

no way to choose between the two alter-

natives." The point here is that they refer to various stages of OE (earlier j_+i_ versus later i.+^J that are never made explicit throughout or even touched upon before or after this reference.

In f a c t , in their section

4 ( 5 3 7 ) , they simply state: "In the first class of weak verbs, we assume that the stem-forming s u f f i x is / i / . " That they d i f f e r e n t i a t e it from a stem formative ^ in the Germanic j_a-stem nouns ( e . g . , cynn) has nothing to do with an earlier or later stage of OE. Their "marginal distinction between underlying /y/

[= j_ in the present

study] and / i / " ( 5 3 8 ) consists in the occurrence of gemination in cynn for which they postulate underlying /kun + y/ and the absence of gemination in 2 sg. imperative of fremman: freme for which they postulate / f r a e m + i + 0/.

The

same result can be achieved, more convincingly because of a historically more plausible stem formative j for the verbs, if the rule of glide vowel coalescence is followed by vocalization and then by gemination (see 2.2 below and section 3 ) . The rule of glide vowel coalescence, ancient though it i s r must be a rule which is operative even today on a purely phonetic level because of the obvious articulatory configurations involved. It is also a rule which works either way: i.e., it makes no d i f f e r e n c e whether the glide is positioned before or a f t e r the vowel as long as they share the same highness and frontness f e a t u r e s . The glide will be assimilated to the vowel and the vowel will emerge as the dominant element. This is borne out by the LWS forms herigan, herigean ( f o r standard WS h e r i a n ) which are trisyllabic. OE has many examples where EOE [j] combined with a front vowel belonging to the same syllable: early frignan

'to a s k 1 , brigdels ' b r i d l e ' are later

written f r i n a n , bridels ( c f . Campbell 1 1 4 , Brunner 1 2 4 ) . Glide-vowel coalescence is the earliest recoverable rule

42

for weak verbs (long before WGmc. gemination, which does not apply if glide-vowel coalescence applied in 2, 3 sg. pres. i n d . ) . It will be formulated without an environment bar to indicate that the deletion may take place following or preceding the vowel: + α

syll cons high back

+ + α

syll cons high back

(As is obvious, the alpha notation will also allow the coalescence of a high back vowel and a high back glide. This is not important for the argument at h a n d . ) Vocalization

2.2

A vocalization rule which changes a high front (and presumably also a high back) glide into a vowel in

cer-

tain environments must certainly be a rather old rule, which has most probably remained productive throughout the centuries. Take Gothic as an example: the f i r s t class weak verb nasjan "to s a v e 1 , which corresponds to OE nerian, has the stem-forming s u f f i x j_ preserved in

all

forms of the present indicative and subjunctive. In 2 sg. imperative nasei, however, where the original j came to stand in word-final position, it became vocalized to [i] (Orthographic ei_ = [ i : ] ) . It is possibly even more obvious in OS where fremmian (OE fremman) has 2 sg. imp. fremi. In the OE weak verb nerian, which preserves the stem formative j in the o-forms, 2 sg. imperative is nere where the final orthographic e_ can most certainly be assumed to have been original j_ which vocalized to _i and was later reduced to £ (= [ β ] ) because it occurred in an unstressed syllable. There is no reason to doubt a parallel development in Gothic, OS and OE. Likewise,

the

formative j served as a connecting vowel between stem and dental s u f f i x in the past tense and past participle of most of class I weak verbs. Although this position was

43

not word-final,

it was morpheme-final and the following

sound was a consonant. This again vocalized the high front glide: for

instance, 1 sg. past ind.

*nar+ j+d+ae -*-

* nar+i+d+ae ->· OE nerede (omitting irrelevant rules) . In the oldest texts the medial vowel _i, later also reduced to e, is frequently attested. Gothic also o f f e r s support: 1 sg. past ind. nasida

Ί saved 1 . The rule will be for-

mulated as follows: - cons"! + high] high -> _- back]

[+ syll] / /

Γ
1

2

2

Condition: 2 * r Their underlying form is

fraem+i + e

to which a rule of

glide formation applies which converts the stem-forming s u f f i x i_ into j

( = [ j ] > before a following vowel. This

then allows the application of gemination. At least from an historical point of view both rules are problematic because they replace the geminating factor with the geminated consonant. This saves both of them a rule which later deletes / j / . Even though K e y s e r ' s and Kiparsky/ Ο " N e i l ' s rules are synchronic rules which, according to standard generative phonology, may be d i f f e r e n t

from

a diachronic rule (although synchronic rules often reflect the diachronic formulation in this theoretical framework), they must be considered linguistically and phonologically unsatisfactory. Synchronic rules too should be phonetically and phonologically plausible and should not just be successful in that they "work" within a given theory. A

45

highly formalistic simplicity metric as that in standard generative phonology (exemplified by Chomsky/Halle, The sound pattern of English) which depends on f e a t u r e counting within the lexicon and the phonological rules may not take into consideration rather basic phonetic processes (see also section 4 of this c h a p t e r ) . Thus it

is unlikely from a phonetic point of view that

a conditioning factor is deleted at the same time that

it

brings about a change, as in K e y s e r ' s and Kiparsky/ 0 ' N e i l ' s gemination rule. I-umlaut and gemination in the OS example above are a case in point: the conditioning factor remains for some time, even a f t e r umlauting the vowel and geminating the consonant. Their gemination rules also force Keyser and Kiparsky/ O'Heil to order i-umlaut, an historically later rule, before gemination b e c a u s e / j / , which also causes i-umlaut, is eliminated in those forms in which gemination applies. Yet they still need a rule deleting / j /

(or / i / ) in the

originally long-stem verbs ( e . g . , d e m a n ) , which do not have geminate consonants. Their formulation raises two questions which will be explored in more detail in section 4 of this chapter and section 1 of chapter three, but which should at least be mentioned here:

( 1 ) how abstract

(or concrete) should a phonological rule, especially a synchronic one, be formulated in order to still represent phonetically plausible processes; and (2) can the inclusion of a rule such as WGmc. gemination be j u s t i f i e d in a synchronic grammar of OE? .Inclusion would give this rule considerable status in the OE native speaker's competence and is equivalent to the claim that WGmc. gemination is operative in OE times. The complexity of the rule and the fact that in OE proper no trace of the gemination causing factor is left in those verbs that undergo gemination (although, as we have seen, evidence exists in OS) make this a highly dubious and unsubstantiated claim. Another problem that was only hinted at at the beginning of 2 . 3 is whether gemination is a geminating,

46 i . e . , doubling rule which inserts an identical consonant under the conditions described or whether it

is more

accurately a lengthening rule. Epenthesis or insertion rules are relatively common only with vowels. They insert vowels to break up certain consonant clusters (see Schane 1973:

54: Latin patris ' f a t h e r ' gen. sg. versus pater

' f a t h e r 1 nom. sg., where the e in pater is presumably i n s e r t e d ) . Consonant insertions, on the other hand, are quite rare in the Germanic and Italic languages. A formulation of gemination as an insertion rule would also create the additional problem of having to specify the identity of the two segments, since it r e f e r only to

Sy I"·"tw\Jilο I

is not enough to

. It seems likely therefore that

gemination is technically a lengthening of the consonant in question, which is indicated by the orthographic device of doubling the written character. Lengthening of consonants is of course a normal process that

is

basically more plausible than consonant insertion. The rule will be formulated as follows: + + j-

, ·*·*· [ + long] L-I- consj ^ Condition:

, [+ consj

#

syll cons stress long __

"- syll" + high .- back_

r

For convenience, the doubling of the consonant to indicate lenghtening will however be retained in the formal presentations of section 3 and elsewhere. It does not a f f e c t the validity of the argument. 2.4

Aufhellung This change, simple though it

appears, has had some

influence on the development of OE and is one of the prerequisites to Breaking (see 2.5 below), and therefore has to precede it. Aufhellung basically means Gmc. *a > ae

before a non-back vowel or glide. Campbell ( 5 2 )

r e f e r s to it

as "a very early change ... in O E " ; Cassidy/

Ringler ( 1 9 7 1 : 2 2 ) call it prehistoric. Aufhellung

is

47

sometimes simply called ' f r o n t i n g 1 and Lass/Anderson ( 1 9 7 5 : 59) translate it

literally as 'Anglo-Frisian

1

brightening , but the German term shall be retained here because of its common acceptance and vivid image. a. > ae seems to have occurred quite regularly, except in the following environments:

( 1 ) preceding a nasal,

therefore land, lond ' l a n d ' ; man ( n ) , mon ( n ) 1

lomb ' l a m b , e t c . ;

(2)

before a

+

' m a n ' ; lamb,

environment,

e.g. faran ' g o ' ; clawu ' c l a w ' ; nacod ' n a k e d ' ,

etc.

A problem arises, however, when we consider words such as f remman, men ( n ) , sendan , drencan, etc. which have obviously undergone i-umlaut but whose Gmc forms must have contained a stem vowel a ( c f . G o t h i c ) . If we accept that nasals prevented the operation of A u f h e l l u n g , and if we also accept, as I think we m u s t , that i-umlaut did not front and raise a vowel in one step (compare the data in all find it

other cases of i-umlaut, 2.7 b e l o w ) , we will

d i f f i c u l t to explain the existence of e_ in f rem-

man, men ( n ) , sendan, etc.

in classical OE. On the other

hand, there is clear evidence in the oldest manuscripts of such forms as f raemman, masnn , saendan (Brunner: 74) . This can either be j u s t i f i e d as a regular application of the rule of i-umlaut (because all

words contain the um-

lauting factor i. or j_ in the following syllable) , or as an irregular one of A u f h e l l u n g . The former reason seems phonetically far more plausible, also in view of the fact that i-umlaut is a pervasive change that occurred in virtually all environments that could possibly be a f f e c t e d , whereas Aufhellung is a rule with important exceptions. This, however, is only the f i r s t step. Should i-umlaut be responsible for a > ae,

we need a later additional

raising rule ae -»· e, which is historically clearly documented (Brunner: 7 4 ) , but which cannot be part of the classic i-umlaut rule because it occurs much later, namely in historical times when i-umlaut had been completed (see

2 . 7 ) . And even if we accept an irregular application

48

of A u f h e l l u n g , we still need a rule as->· e, which cannot be the regular umlaut rule for the reasons just cited. Lass/Anderson ( 1 9 7 5 : 67 f n . ) ical facts can be explained

suggest that "the histor-

just as easily on the assump-

tion that nasal retraction was at some point reordered to follow [ A u f h e l l u n g ] and u m l a u t . " This means, in

effect,

that only those ae which did not undergo ae > e would be retracted in a nasal environment. Diachronically, this is not a satisfactory solution if the data given above are taken into account. Thus we have as our only choice the assumption that in fremman, sendan, etc. a > ae through i-umlaut, and ae > e through a later process which cannot be classified as i-umlaut because the umlauting environment was no longer existent. Campbell . ( 7 4 ) attempts to j u s t i f y a > ae by stating that the consonant or consonant group following the stem vowel and palatalized through original i-umlaut caused this further change. This seems to be pure speculation, because such a palatalization is not manifested anywhere. A u f h e l l u n g , then, is part of the diachronic rule inventory. It did not apply before nasals and l + i, a c k en ~ vironments. This includes cases like fremman, sendan, etc. where, following a > ae through i-umlaut, we have to accept a later raising, even though its scure.

cause appears to be ob-

(As a minor rule, it will not be considered in the

derivation of fremman. For the sake of a clear exposition, i-umlaut there a-*· e.) Nerian, secgan, sellan and similar OE words did of course undergo Aufhellung and later i-umlaut quite regularly, as we will se below. Aufhellung will here be formulated in its form: + + +

syll cons low stress

[- back] ,

„0 ,_

most general

backj

49

2.5

Breaking For a limited number of OE weak class I verbs the

process of breaking must be discussed.

It concerns the

past tenses of cwellan, dwellan, sellan, stellan, tellan, a group of verbs which shows no evidence of a medial connecting vowel in the past tense. The early loss of this connecting vowel (which must have existed originally because its non-vocalized form caused gemination in the present tense system) yielded the OE past tense forms cwealde, dwealde, sealde, stealde, tealde which in prehistoric OE had undergone breaking. Breaking a f f e c t s the front vowels ae, e_, _i to which "a vocalic glide" (Campbell: 54) is added when these vowels precede geminate 1^ or !_ plus consonant, geminate r or r plus consonant (excluding - r j - as in the nerian type: j_ is not a consonant), and h ( = [ x ] ) or h plus consonant. Only as, however developed this glide in all

these

environments; the other vowels broke only in certain environments. There was also considerable variation in the various dialects, which need not concern us here. While the chronological order gemination-breaking is not disputed, a problem develops in standard generative phonology when a synchronic analysis is attempted. A case amply discussed in the literature ( e . g . , Postal 1968: 2 6 3 - 4 , King 1973: 561-2) deals with the existence of an OE verb fiellan

'to f e l l 1 and the verbs mentioned above:

cwellan, dwellan, sellan, etc.

The claim is that breaking

did not occur in those verbs where _! was doubled by WGmc. gemination

(cwellan, e t c . ) , but that it did occur if,

as

in f iellan, the 1^ had been doubled before WGmc. gemination. This then constitutes something like original 11. The underlying pre-gemination forms would therefore have been (according to King and Postal): / f a e l l j a n / versus /cwasljan/ /saeljan/. In order to account for this

ir-

regularity, King claims, breaking would have to be ordered before gemination in a synchronic grammar, contrary

50

to the historical development. If ordered in the

his-

torically correct order, breaking would result in both f i e l l a n and seallan since the breaking rule cannot

dis-

tinguish between recent and original geminated forms. It is remarkable in this respect that fiellan is not listed in Bosworth-Toller ( 1 8 9 8 ) and Clark Hall ( 1 9 7 0 ) and that the Oxford English Dictionary lists fiellan only as one of several variants in the etymology of MnE f e l l . Other examples of breaking before 11 are rare: eall ' a l l 1 , feallan

'to f a l l 1 . On the other hand, there

is at least one instance where a verb which underwent WGmc.gemination also underwent breaking: geminate h ( = [ x x ] ) in hliehhan 'to laugh" does cause breaking (see Campbell 5 4 ) . Whereas breaking occurs in a great many cases, the number of exceptions is minimal and probably does not exceed much the instances quoted in this section. Lass/Anderson ( 1 9 7 5 : 2 7 1 ) , like Postal and King before them, take up this dilemma and state that " [ t ] h e easier and less interesting [explanation] is that the [11] clusters in sellan, tellan and the like are marked with some indiosyncratic f e a t u r e like [ - b r e a k i n g ] . . . . The second possibility, which we prefer on grounds of principle, and which contradicts no known synchronic f a c t s , is that at the point in the grammar where breaking applies, the present-tense forms of those verbs do not s a t i s f y the proper analysis of the rule; but at this point the [Id] clusters are present in the preterite forms, so that the rule applies. This means whatever the historical order might have been, gemination is synchronically a later rule than breaking." Although the facts presented above will not unambiguously r e f u t e Lass/Anderson's statement, they seem at least to establish a certain balance in the two views. Even if gemination is a synchronic rule in the inventory of a native speaker of OE, which is highly questionable because of the complex and unusual nature of the rule and because of the inconclusive--to the speaker of OE--environ-

51

ments attested in OE, it might indeed be the case that a small and limited number of words do not undergo breaking and are marked idiosyncratically. While it may be impossible in this context to prove that the reordering solution is clearly wrong, it should also be noted that Lass/Anderson partially base their contention on the weak claim that "it known synchronic facts"

contradicts no

( 2 7 1 ) . The principle that they

allude to in the same paragraph (see above) is the principle that an alternation, as for instance sellan-sealde, can and should be accounted for phonologically in the rule inventory of the synchronic grammar. In the epilogue of their book, in a section "Some afterthoughts on theory" ( 1 9 7 5 : 2 2 5 - 2 4 7 ) , Lass/Anderson call their adherence to standard generative phonology into question: "In looking back over these studies, a year a f t e r having completed the basic work, we are struck by the extent to which the structure of our arguments has been determined by implicit presuppositions. And even more, by the lack of genuine empirical support for many of them. The most disturbing thing, ultimately, is the immense power of 'standard 1 generative phonological theory, and the somewhat ad hoc and even circular nature of portions of the theory that enshrine this powerful machinery as canonical" ( 2 2 5 - 2 2 6 ) . They continue: "But is the mere fact that a phonological solution works any guarantee that it is correct? In the light of our present knowledge this would seem to be at least arguable. In f a c t ,

since even the

internalized grammar of a speaker of a living language (let alone one likeOE) is not available for direct inspection, we cannot tell wether its

internal organiza-

tion is anything at all like what the theory (whose framework is largely aprioristic) makes it out to be" ( 2 2 6 ) . Statements like these allow the conclusion that a reordering of rules, even in a framework of standard generative phonology, does not represent a completely convincing solution.

52

In the investigation of the weak verbs breaking will not be a major rule. The diachronic rule will therefore be given here only in approximate form. + syll - back + stress

0

For a discussion of breaking, in its

historical context,

s e e Lass/Anderson ( 1 9 7 5 : 7 4 - 1 1 2 ) . 2.6

Palatalization What is often referred to as palatalization represents

in reality two distinct diachronic rules which form a relatively close unit: palatalization proper and assibilation. Palatalization is an early prehistoric process, assibilation occurs later. Palatalization, generally speaking, is an assimilatory process in which the consonants [k] and

[g]

(orthographically mostly c_ and g_)

developed an a f f i n i t y to palatal sounds following or preceding them. Runic evidence suggests a distinction between a palatal and a velar pronunciation because four d i f f e r e n t runes are used f a i r l y consistently to indicate [k-'] and [g ] versus [k] and [ j ] .

(The runes denoting a

palatal pronunciation, however, seem also to have been used for velar consonants [see Campbell 1 7 3 ] ) . Aside form the runes, the orthographic realization (which naturally occurs later than the actual pronunciation) of the process of palatalization is frequently e, as in geat 'door 1 ( c f . gatu " d o o r s " ) and ceaf " c h a f f " . This is also done in medial position: laece ' p h y s i c i a n ' , laeceas "physicians' nom. p i . , secean 'to seek 1 , licgean 'to lie' Infrequently orthographic k is used in some manuscripts to indicate the velar stop, for instance in the Mercian Rushworth Gospels. There k stands word-initially before ae,

e:, i. to indicate [k] . Assibilation, the step which converted [k- 3 ] and [g- 3 ]

to [t/] and [ d j ] , respectively, must have occured later, possibly a f t e r the f i r s t surviving documents, and has

53

never been satisfactorily explained. It is quite possible that there is no one single rule of assibilation but that several complicated stages ( i . e . rules) are involved a f t e r palatalization took place. Likewise,

it

is impossible to tell when assibilation started to take effect.

Brunner ( 1 7 1 ) states: "Zu Beginn der me. Zeit

war aber die Stufe / t / / , /dz,/ j e d e n f a l l s bereits erreicht. In der zweiten Hälfte des 12. Jahrhunderts

beginnen die

Schreibungen ch für / t / / nach f r a n z . Schreibgewohnheit und g_, dg für / d z / , die allerdings weiterhin zweideutig ist,

weil g_ auch den stimmhaften Verschlusslaut

... be-

deutet. " In two recent studies on OE verb morphology (Keyser 1 9 7 5 , K i p a r s k y / O ' N e i l 1 9 7 6 ) , the problem of palatalization and assibilation is not elaborated upon. The palatalization rules are simply stated. Keyser's rule ( 3 8 8 ) is this: + obst] [+ highj

- back] [+ cor J

/ /

f

+ sylll |_- backj

It becomes obvious from the context that it is also supposed to account for assibilation. K i p a r s k y / O " N e i l ' s rule is identical, except that their structural change is [-back, -cor]. Presumably, their rule is designed to yield [ k ^ ] , n o t [ t / ] . If we look only at palatalization, the earlier phenomenon, we can observe that the initial velar stop k was palatalized before primary (= original) f r o n t vowels, as in cirice ' c h u r c h 1 . K f a i l e d , however, to palatalize before front vowels which had come about as a result of i-umlaut and remained velar: cyning ' k i n g 1 , cemban 'to c o m b ' . A well-known example here is the case of cene 'keen, brave 1 and cen ' t o r c h 1 . The palatal pronunciation of cen is documented by the use of the rune which signifies palatalization. The velar pronunciation of cene is established by its

subsequent development. A common

solution to this problem in standard generative phonology ( c f . Lass/Anderson 1 1 6 ) is to postulate a synchronic

54

palatalization rule, a synchronic i-umlaut rule and the underlying forms /koni/ for cene and /ken/ for cen.

palatalization

/koni/

/ken/

-

k-'en

i-umlaut

koani

unrounding

keni

reduction

ken3

phonetic output

[kenS]

[k^en]

King ( 1 9 7 3 : 5 6 3 ) disputes that the order palatalization i-umlaut is the historically correct one by pointing out that there seems to be no palatalization in OHG and OS and because "the highly variable generality of the palatalization of /k/

in OE dialects suggests a change of

recent origin there." K i n g ' s arguments are not very persuasive, especially in view of his own examples: cemban and cyning. Had palatalization been the later process, it would have most certainly applied to the front vowels produced by i-umlaut. From the point of view of the native speaker of OE, it seems very likely indeed that he was not able to differentiate between, for instance, original [e] and [e] caused by i-umlaut, particularly if the umlauting factors had begun to become obscured. If cemban, cyning, cene retained [ k ] , this could very well mean that palatalization had ceased to apply as a phonological rule when these surface forms were commonly used. Palatalization in the strictest sense, i . e . , fronting of velar stops, and no more, before

as a

[-low, -back]

non-consonants may certainly have preceded i-umlaut or existed simultaneously with it. The fact that this is not manifested in other Germanic dialects is not conclusive evidence that palatalization did not occur in them. It may simply mean that the palatal allophones of the velar stops never developed further and never achieved independent status

( c f . the alternation between velar and

palatal articulations of the velar stops in Modern Eng-

55

lish keel or cool, or in Modern German Kiel, k hle,

or

K u h l e ) . Later, in OE, assibilation sets in and continues the development begun by palatalization. The orthographic changes with regard to the original velar stops, which become more and more general through the course of the OE period, can thus be seen as the written evidence for the f u r t h e r development of the palatalized stops in the OE dialects. In conclusion,

it can be said that a diachronic order

palatalization, i-umlaut is not contradicted by the known facts and that there is some positive support for it.

As for the development of the weak verbs to the EOE

stage, it need not concern us here because of its relatively late occurence. The diachronic palatalization rule, which applies to k , g_, looks as follows: + ObS 1 , , + back Γ cor J

2.7

->

r , , , [-back]

// /

//

ι-

Γ- consΊ I I - Lback

i-umlaut i-umlaut is a rule whose e f f e c t has been felt through-

out OE, ME and MnE and all

old and modern Germanic

dia-

lects, except Gothic. It affected long and short a, o, u, which it

fronted and also short as which it

raised, i-um-

laut was caused by the raising and fronting influence of a high front glide or vowel in the following syllable or even the following sound, i-umlaut can therefore be called an assimilatory process and should be seen as such. The results of i-umlaut are as follows: a_ > ae fore m and n ) ; o > oe

(and to e be-

(which is unrounded to e in most

d i a l e c t s ) ; u > y_ (which is unrounded to i_ in L W S ) ; short _ae > e^. While it is quite simple to list related OE forms which show alternations between umlauted and unumlauted vowels, it is much more d i f f i c u l t to point out attested OE forms which preserve the umlaut causing segment [i] or [ j ] . Lass/Anderson ( 1 9 7 5 : 1 1 7 ) provide t h e f o l l o w i n g list of alternations:

56

u γ cup ' k n o w n ' : cypan 'make known' u y burg " c i t y 1 : byrig ' c i t y 1 , dat.

sg.

ο $ (e)dom ' j u d g m e n t ' : deman " j u d g e " " h a s t e 1 : efstan 'hasten'

ο ?5 (e) ofost

a ae hal ' w h o l e ' :

haslan ' h e a l '

1

a e mann " m a n : menn " m e n " a as faran

' g o " : faerst "go 1 2 sg. pres. ind.

The only form exhibiting a high front vowel in the syllable following the umlauted vowel is byrig but as Campbell ( 1 5 2 ) points out, this is a parasite vowel which is quently inserted before [j]

fre-

in LWS. Other Gmc dialects,

especially OS as in settian, skeppian, preserved the umlauting element for

some time. OHG has very early ruckie

' ( h u m a n ) b a c k " , phonetically

[ ' r u k j e ] . OE in even the

earliest documents either deletes the umlauting segment or reduces it to schwa. OE vowel alternations like the ones above are therefore all that is l e f t to show that i-umlaut was operative as a rule in pre-OE. It is therefore not clear why these vowel alternations should provide "morphophonemic evidence suggesting that umlaut was productive in OE" ( L a s s / A n d e r s o n : 1 1 6 ) . Later they refer to these alternations as being of "obvious and productive types"

( 1 1 8 ) . The evidence here rather suggests that

these forms are merely inherited inoperative relics of the past. Brunner ( 6 8 ) states: "Die Ergebnisse des i-Umlaut erscheinen in den Schreibungen der

ltesten ae.

Texte bereits voll durchgef h r t . " i-umlaut had been completed in OE by the time of the earliest written records. Also, there is no evidence that a process like i-umlaut regained its productivity

in any Gmc. dialect after

it

ceased to be operative in pre-OE. Taking this lack of empirical support into account,

it

is surprising to see that practically all linguists working on a generative analysis of OE phonology include a synchronic rule of i-umlaut in their rule inventory ( c f . Keyser 1 9 7 5 , Lass/Anderson 1 9 7 5 , K i p a r s k y / O " N e i l

57

1 9 7 6 ) . On the other hand, the inclusion of such a rule is consistent, given the nature of standard generative phonology as expounded by Chomsky/Halle 1968. Thus it was natural to derive the alternation cup - cypan from underlying / k u p / , with i-umlaut applying in the case of cypan (see Lass/Anderson: 2 2 7 ) . Erickson ( 1 9 7 0 : 5) even says that it

is "apparent" that processes like WGmc

gemination and i-umlaut are "very much operative in Old English." Again, it may be interesting to quote from the epilogue of Lass/Anderson ( 1 9 7 5 : 227 - 2 2 8 ) : " [ O ] n the basis of the obvious derivation of [ y ( y ) ] in such pairs as those given above from [ u ( u ) ] [ i . e . , cub~cyban —DS] , we then claimed that there is no need, at the lexical level, for front round vowels—even in those forms that are not members of an alternating set. What held for trymman also holds for cyning, what held for cypan holds also for hydan. This is not, given the structure of current phonological theory, at all out of line; it is simply an application of the ' f r e e ride' principle. Exactly the same kind of thinking holds for the claim in SPE for a synchronic vowel shift in N E . " A diachronic rule of i-umlaut would account for the major changes given above (a > ae, ο > ae, u > y , ae > e) :


e, shorter all the others.

the

58

There is also another aspect to i-umlaut that is somewhat problematic.

Campbell ( 7 1 ) writes that it

demonstrable "whether e^ > _i. " Yet there are

is not

forms like 2 ,

3 sg. pres. ind. b i r ( e ) s t , b i r ( e ) p of the strong class IV verb beran 'to bear 1 in WS, or h i l p s t , h i l p p for helpan. The process at work here must be a type of i-umlaut because the change in vowel quality occurs in those forms that in the earliest attested forms have an i as their inflectional vowel. 1 sg. pres. ind.

is bere, helpe,

respectively, are without umlaut, because the original ending is u for 1 sg. pres. ind. Elsewhere, Campbell ( 7 6 ) then states "an i-umlaut of e_ to i_ must be assumed for O E . " Therefore we will formulate an extra rule, which 2 we will call "i-umlaut ": + syll - low - back

[+ high]

+ -

It might also be possible to include i-umlaut

cons high back stress in the

regular i-umlaut rule (as suggested by Peter A. Schreiber) + /-

but it

syll back\ long \ high

- back - low χ. - α high/


xtaj ) . This is a point that should be kept in mind when a synchronic analysis is attempted Sellan is a d i f f e r e n t

(see below).

case. While bycgan had a con-

ventional present tense and an unconventional past tense (compared to the majority of class I weak v e r b s ) , sellan has a straightforward

past tense but an odd present tense.

As already discussed in section 2.5 of this chapter, the present tense forms of sellan and four other verbs like it

do not undergo breaking although 11 is a Breaking

environment. If Breaking is disregarded for a moment, the present tense forms correspond almost exactly to those of the fremman subclass (except A u f h e l l u n g ) ;

in

other words, all other rules apply regularly. But Breaking which should apply, does not. This appears to be the only case where the established diachronic sequence of rules is not capable of yielding the forms to be expected from the proto-forms. The presence of Breaking in the past occurs because the medial vowel must have been syncopated very early. Thus the breaking cluster Id, as in sealde, had to be formed before Breaking set

in.

73 I sal 1 . gvc 2. voc 3. gem 4. Auf h 5. brk 6 . pal 7 . uml 8. j-del 9 . apo 1 0. sync 1 1 . red output phonetic orthographic

+ j + an | sail + j + an saelL + j + an does not apply sell + j + an sell + an — sell + an [ 'sellan] sellan

| sal 1 . gvc 2 . voc 3. gem 4. Auf h 5. brk 6. pal 7 . uml 8. j-del 9 . apo 10. sync 1 1 . red output phonetic orthographic

+ d + aes saal + d + aes seal + d + ass seal + d + aes [ 'seaidees] sealdaes

If Breaking is disregarded, the development of sellan, sellu, selis, seli, etc. is perfectly regular. V7hy Breaking did not apply, we do not know. Since there are environments which were caused by WGmc. gemination and in which Breaking later applied ( h l i e h h a n ) , there is a chance that Breaking did once apply in sellan but was later obscured. It seems not unreasonable to assume that a minor irregularity is to be blamed for the absence of Breaking in the present tense of sellan, especially when one compares it to the vast numbers of class I weak verbs (many of which appear to have the same phonetic structure as s e l l a n ) . In all

languages there are bound to be irreg-

ularities which can practically always be attributed to historical factors. Therefore the irregular percentage of verbs in class I is quite small indeed. 3.2

The weak verbs of classes II and III

3 . 2 . 1 Class II The exposition in section 2 made it

clear that al-

though the rules described there found their widest application in class I verbs, these rules were by no means limited to class I. However, unlike class I, the verbs in class II constitute a mixed paradigm in that the second and third person singular and the imperative do not contain the stem formative j.

The mixed nature

of this class is described by Campbell (332 - 333) as

74

follows:

"This conjugation is a mixture of one in which

endings were added to Gmc. -o- < I-E -a- ( 2 n d and 3rd sg. pres. ind. and i m p e r . ) and of one in which they were preceded by -i-

[i.e.

-j-] as in class I, but this -i-

followed -o-." Other handbooks agree

(Brunner; 3 2 9 , Krahe/

Meid 1 9 6 9 : 11/121, but see also Cowgill 1 9 5 9 ) . Thus the proto-forms of class II for the five forms considered in class I can be described as follows: i n f i n i tive: |luf + o j + a n | ; 1 sg. pres. ind. 2 sg. pres. ind.

|luf + oj + u | ;

|luf + ο + s | ; 2 sg. imperative |luf +

ο + 0 ; 2 sg. past. ind.

| luf

+ ud + aes .

that contain the stem formative

The forms

| j | follow class I, i . e . ,

i-umlaut applies. The change in the stem formative then follows roughly these steps: oj ·> ej ->· i j -»· i. The stage ij is frequently attested as orthographic -ig- and serves as a particularly clear indication of the vocalic quality of this feature of class II verbs (see the end of this section for the complete derivations). One particularly interesting aspect of this class is the formative -p- which originally appeared in alle the forms of the present. Where the -o- was followed by -j_it was, as we have just seen, umlauted. If not followed by - j_- ι the classical OE paradigm shows orthographic a. It is worth noting in this context that Gothic, OS, and OHG all have attested orthographic o as an inflectional vowel in the form of 2 sg. pres. ind. We would like to claim that the orthographic change from o to a is to be seen as part of the process of vowel reduction as it applied to back vowels and that this process followed along lines similar to the process of vowel reduction already shown for front vowels. We have already seen the tendency to reduce the vowel of unstressed endings in our examination of the reduction rule in 2 . 1 1 . See Campbell ( 1 5 3 ) : "ae, e and i^ fell together in a sound written e_ in unaccented syllables." Whereas one gets the impression from Campbell that there were no distinct stages in the reduction of the front vowels, it is quite

75

clear that there were at least two distinct stages in the reduction of back vowels with unaccented u f i r s t becoming q ( c f . Brunner:270: "Als ursprüngliche Endung der 1. Sing. Ind. Präs, der starken und schwachen Verba ist

für alle Mundarten -u ... bzw. jüngeres -q anzuset-

z e n " ) , and then o/u and a falling together: "As a result of the changes described above, OE had virtually two unaccented back vowels, o/u and a" (Campbell: 1 5 6 ) . He then goes on to describe the weakening of the distinction between these back unaccented vowels and ends up by stating: "In the eleventh century unaccented e ( < ae, e_, _i) and the unaccented back vowel in which a, q, u had largely coalesced became confused" (Campbell: 1 5 7 ) . Apart from the fact that the front vowels reduced earlier than the back vowels, two interesting points emerge, particularly from this last paragraph of C a m p b e l l ' s : (1)

whereas Campbell explicitly identifies the

front

unaccented vowels as orthographic e, he is vague about the identity of the back unaccented vowel. (2) There seems at some point to have been two unaccented reduced vowels, one back and one front.

(Compare here, for

stance, Modern German bitte 'please' bitter 'bitter

1

in-

t ' b i t a ] versus

[ ' b i t A ] , where orthographic -er reduces

to a back vowel because of the uvular quality of the liquid.) These facts point to a certain parallel development in the front and back vowels and it will be claimed that the occurrence of orthographic a in the present tense of class II weak verbs is a strong indication of this vowel reduction process. The two-stagedness of the reduction is shown clearly by comparison of the prehistoric and classical forms of the past tense and the present tense of class II verbs. Thus 1, 3 sg. past ind. which is originally lufudae appears as lufode, whereas 2 sg. pres. ind.

lufos appears

as lufas. There are no special symbols to indicate reduced vowels in OE any more than in MnE,but since orthographic e among the front vowels was closer to the

76

raid central schwa sound than orthographic i. or as, a likely symbol to represent the merger of all

it was

three

sounds. With the back vowels, however, the situation was more problematic because of the distinction between rounded and unrounded vowels. Here ο would be a natural symbol to represent the reduced rounded vowels u_ and ο while a would be the only symbol available to represent a reduced unrounded vowel. Therefore it

is likely that

in a pre-final stage of the reduced vowel merger ο represented the rounded unstressed back vowel [o] already lowered to mid position from [u] whereas a represented the unrounded back vowel [Λ] , also in mid position but more centralized than [ o ] . The next step would have been for

[o] to lose its

rounding and to become more

centralized but to still retain its

backness.

[3] and [Λ]

would then have been the two unstressed vowels, one front and the other back. The apparent confusion which Campbell alludes to ( 1 5 7 ) follows naturally from the phonetic closeness of [9]

and [ Λ ] . it is, however, probable that this

confusion set in earlier than the eleventh century. The eleventh century j u s t produced the documents in which this confusion shows up orthographically. The f i n a l stage was the f a l l i n g together of

[9] and

[Λ] in

[ 9 ] . The entire

development of the unstressed vowels follows a persuasive articulatory pattern in which sounds from the extremes of the vocalic range tend to move toward a more centralized, neutral position which requires no such special features as roundness and backness. If we now look at the medial vowel in the past tense as in WS lufode (from earlier attested lufudae) and at the inflectional vowel in 2 , 3 sg. pres. ind. l u f a s t , lufab, 2 sg. imperative l u f a , we will find that the medial vowel had been lowered to the mid [ ο ] , whereas the inflectional vowel a is simply the symbol used for the sound [ Λ ] . In other words, in the latter case the change from o (as in the proto-forms) to a does not represent a lowering but an unrounding, one step f u r t h e r

in the

77

direction of a mid central articulation. The change to

[Λ] which must have also taken place at some time

does not concern us here

it

has nothing to do with the

present developments. It is interesting to note that quite a few WS texts also display the medial vowel a in the past tense ( f o r examples see Brunner: 3 3 5 ) and the reduced vowel e_. These are facts which support the reasoning that the

orthographic development u ·> o -> a

-*- e in reality represents the phonetic development [u] -*· [ο] -> [Λ] -> [3] , as claimed above. In many other cases, especially toward the middle and the end of the OE period when vowel reduction gained momentum, it was common for all

vowels to go directly to [3] . It can also

be considered quite normal that in trisyllabic words, as in lufode, the medial vowel [o] went directly to

[9]

since it occurred in a virtually stressless syllable, if compared to syllables 1 and 3. The fact that syllable 3 received more stress than 2 but less than 1 is also borne out by the fact that in pi. past ind.

the last syllable

undergoes an attested change: early -un, later -on. In view of these facts the question arises whether front vowel reduction was really an earlier process than back vowel reduction. It seems likely that both started at the same time, long before documented OE , but that back vowel reduction was really a more complicated development with more stages that took longer to complete. Earliest attested OE forms like aslacudae, suicudas

(in

the Epinal Glossary) with unreduced front and back vowels (as, u) which later, in classical OE, became aslacode, etc., i.e., orthographic e and o, and later again aslacede, confirm this. The various, diachronic reduction rules as well as a diagram of the development of unstressed vowel are given here informally:

78 V

(D

[-back] ae, e, i

[β]

Unstressed syllable

V

Γ+high "J [_+back J u

Unstressed syllable

V

(3)

(4)

-high -low +back ο V Γ+low

Unstressed syllable

Unstressed syllable

w. From this point on the difference in underlying representations, v i z . / h w . . . / versus / w . . . / which used to distinguish the two types disappeared, because a child hearing his elders use only [w] would no longer be able to differentiate between the two originally separate sets. It would then construct a new, and simpler, grammar which contained only the underlying form /w/ for both types of words. The underlying forms / h w . . . / and / w . . . / had been restructured to / w . . . / , and the rule would no longer be required and would consequently be eliminated. In well attested modern languages the tracing of such a restructuring (as in the case discussed above) is relatively easy. In more ancient languages, with an often highly limited number of documents and'thus data, establishing an instance of restructuring is often a rather speculative matter. Since the rule triggering the eventual process of restructuring (as the deletion rule above) is later lost, and since the original alternation disappears, the reconstruction of a pre-restructuring stage of the language is either

100

impossible or can only be brought about with the help of relic forms or, more probably, with the help

of data

from a related language. The last case, the reconstruction of early language stages with data from other Gmc. dialects, can frequently be found in OE handbooks and grammars. But because generative phonologists generally work on the assumption that no restructuring of underlying lexical representations can occur as long as there exists a s u r f a c e alternation which can be accounted for by a phonological

rule, the intervals between restructurings

are extremely long. Such phenomena as gemination or umlaut or G r i m m ' s Law or V e r n e r ' s Law are well

i-

explored,

but practically never within the context of just one language, in this case, Old English. This is to say that many or all

generative analyses of OE phonology depend

on crucial information that was not available to the speaker of OE, that of other Gmc. languages. It is thus relatively easy to account for alternations that are as remote as G r i m m ' s Law and V e r n e r ' s Law and K i p a r s k y / 0 ' N e i l do indeed include these two processes in their synchronic rule inventory ( 1 9 7 6 : 5 5 1 ) . That restructuring of underlying forms in generative phonological analyses of OE is practically unheard of under these

cir-

cumstances comes as no surprise. At some point in the f u t u r e , with more and more comprehensively attested modern languages, this development will lead to interminably long synchronic grammars, because all or most of the historical changes are part of the grammar. At the same time these grammars are unlikely to bear any resemblance to the average speaker's native competence. Let us take gemination and i-umlaut as examples of rules which are commonly included in synchronic generative analyses of OE. In those verbs which display gemination there are no distributional hints which signal to the OE speaker when to expect, within the OE evidence, the occurrence of a geminated or a non-geminated verb form. Geminated and single consonants, for instance, both occur

101

before vowels. Not j u s t d i f f e r e n t fremede)

vowels ( c f . fremmap,

but also before the same vowels ( c f . fremme,

f r e m e ) . Neither does the original gemination causing factor come into play here. There are no s u r f a c e forms with j_. Ironically, those f o r m s , which in the oldest texts do display a high front element a f t e r the stem consonant

(2, 3 sg. pres. ind.

fremist, f r e m i f a ) ,

are

the ones that never underwent gemination. Though an alternation mm ~ m obviously exists in

fremman, and of

other sounds in other verbs, the principle of this alternation is not reconstructable from OE data alone, and we therefore have to conclude that to the speaker of OE, gemination was not a productive rule. Put d i f f e r e n t ly,

gemination in the OE period is a relic of the past

and not operative as a process. Similarly, i-umlaut cannot be called a productive rule of OE because the alternations lack a principled foundation that is visible to the unsophisticated er.

speak-

If we take the same example as above, fremman (which

has umlauted vowels in all

f o r m s ) , i-umlaut may perhaps

be transparent in the old forms f r e m i s t ,

fremifa but

cer-

tainly not in equally old fremmu, or later fremme. Verbs like

ha^lan "to h e a l " , deman "to j u d g e " , 1

'to p u r s u e , cyfaan 'to make known 1

1

ehtan

or nouns like cynn

1

' k i n d ' , ende ' e n d , see

' s e a , rest ' r e s t ' and many

other words effectively

conceal the earlier working of

i-umlaut. As in the case of gemination, alternations do exist but in many cases they do not even occur within the same paradigm but only if,

for instance, noun and

verb are compared (see 2 . 7 of chapter two, dom versus deman, e t c . ) . But these alternations are not superficially transparent and are therefore

devoid of meaning

to the speaker. Possibly the most convincing evidence that i-umlaut is a non-productive rule in OE is the existence of so many class II weak verbs that meet the required conditions for i-umlaut and yet do not undergo i - u m l a u t . Examples

102

of a long list of words are bodian, hoiian, lopian, macian, labian, fund ian, locian, etc. The reason why these verbs do not undergo umlaut is not that underlyingly a synchronically u n j u s t i f i e d back vowel intervenes between stem-final consonant and umlauting factor

(as claimed

by Wagner 1 9 6 9 , Keyser 1 9 7 5 , and K i p a r s k y / 0 ' N e i l 1 9 7 6 ) , but the fact that i-umlaut was no longer operative. Surely, supremely transparent umlauting environments like the ones in the verbs above would have brought about umlauted forms had i-umlaut been a viable process at the time. For appropriate arguments concerning the inoperativeness of palatalization in classical OE see section 2.6 of chapter two. Arguments like the ones above and those presented in the discussion of the diachronic rules (section 2 of chapter two) seem to me to be a clear indication that certain phonological rules are no longer productive and cannot be included in a synchronic rule inventory of OE. The opacity of the rules discussed is of course one factor why they lose their productivity or are lost from the grammar, as Kiparsky ( 1 9 7 1 : 630) puts it: "Opacity will ... be one of the factors that bring about the loss of rules from a grammar by historical change." This, in turn, leads to a lexical reanalysis of the underlying forms which have absorbed the e f f e c t s of the now defunct rules. In what manner specifically the weak verbs of OE were relexicalized and what the synchronic derivations of the various verbs are will be discussed in section 6 of this chapter. 3.

The j_ formative In chapter one and especially in chapter two it

became

apparent that the j formative must have played a very important part in the prehistoric development of the OE weak verbs, although even the earliest attested stage of the weak verbs in OE does not consistently display a surface manifestation Of this formative in the various

103

verb paradigms. The fact that Gothic and Old Saxon, on the other hand, are consistent in the presence of j_ in those verbs that historically contain it is a f i r s t indication that even as early as the beginning of the OE period the significance of this formative in OE had begun to decrease. In standard Middle English none of the verbs concerned retain an element that can be independently and unequivocally

identified as a reflex of the original

j_. This development has to be accounted for somehow and ought to be reflected in a synchronic grammar of the weak verbs of OE. In the course of the OE period the J_ formative gressively lost its identity, either because it appeared completely

pro-

dis-

in certain verb paradigms ( e . g . ,

fremman, deman, and also habban, libban of class I I I )

or

because it participated in such processes as glide vowel coalescence or vocalization in such verb paradigms that otherwise retain some evidence of j ( c f . i n f .

nerian

versus 2 sg. pres. ind. nerest versus 2 sg. imp. n e r e ) . It is not s u r p r i s i n g , therefore, that the j formative which is so apparent in such verbs as Gothic s a t j a n , OS settian by classical OE times did not have what has been called transparent surface structure in section 1. On the basis of deman and fremman, where there are no forms in the OE paradigm which contain ^, it

is clear

that this j had ceased to be at least a surface characteristic of the OE weak verbs in general. It is equally clear that, whatever the original function this j_ formative might have had in proto-Gmc· in those places where it is attested in OE it was, in the OE period, an obligatory formative with no specifiable function whatever: what might be called a relic of a former stage (or former stages) of the language. In such a situation the learner of classical OE, confronted with a segment to which no meaning could be assigned, would interpret this segment either as part of the stem or as part of the personal ending, these being the only two possibilities. Since

104

this phonetic segment does not occur in all

members of

the paradigm, assignment of the surface form of j to the stem would require the postulation of two underlying forms, one for each stem alternant. This would create an undesirable situation because, certainly, a description that would allow for only one stem type instead of two or even more alternants ( c f . the discussion of this problem in section 2 . 2 . 1 of chapter o n e ) would be able to capture a significant generalization which would make the description less costly. Therefore it that this

is more likely

high front segment would be interpreted as

part of the personal ending which is that part of the verb which is expected to vary in the individual forms of the paradigm anyway. The consideration of how the remaining surface manifestations of the original j_ formative should be interpreted

in a synchronic ana-

lysis of OE will again play a role in the discussion of the t r a n s f e r of weak class I verbs to weak class I I . For this see chapter f o u r . 4.

Degemination

It is clear from what has been said in this chapter so far that in the synchronic analysis of classical OE verbs like deman and fremman the stem cannot be entered in the lexicon with an underlying j[ formative. Since there no alternation o ~ e in the

is

case of deman (< * domjan)

or a ~ e in the case of fremman

(< * f r a m j a n ) , one would

also rule out the back vowels as possible underlying vowels in these verb forms. Back vowels would have to be considered an impossibility in this case anyway because the absence of the j-formative

(see above)

in all

forms

of deman and fremman would preclude a consistent derivation of these surface f o r m s . There are, so be sure, nouns like dom which are etymologically related to deman, but as was said b e f o r e , the additional complexity which would result from trying to give both the verb and the noun the same underlying representation would be far

greater

105

than a mechanism which would relate these surface forms in a more direct way. deman, the representative of the long stem verbs of class I, belongs to the subclass which is by far the most regular of all

class I weak verbs and

a segmentation stem versus ending is comparatively uncontroversial. I would therefore postulate an underlying lexical

re-

presentation of /dem +/ for the stem of the long-stem verb of weak class I,

deman. In the case of fremman,

however, one is faced with the problem that the stem appears in two alternants

( a l l o m o r p h s ) , fremm- and

frem-. Before this point is discussed in more detail, the basic forms

(present and past tense) of the fremman

paradigm are here repeated for convenient reference: pres.indie, sg.

pi.

1 fremme

past indie, sg.

1 fremede

2 frem(e)st

2 fremedest

3 frem(e)b

3 fremede

fremmab

pi.

fremedon

Note that in the past tense the stem is uniformly frem-, while in the present tense there is an alternation ~ f_£em~· If we were to assume frem-

fremm-

( i . e . , / f r e m +/) as

the underlying form, we would have to motivate gemination of the stem-final consonant in 1 sg. pres. i n d . , pi. pres. i n d . , pi. imperative, pres. p a r t . , and the i n f i n i t i v e .

It

is evident that, whatever the rule might be, it would have nothing in common with the historical rule of gemination (as discussed in section 2.3 of chapter two) which produced these forms from the non-geminated protoGmc.forms, because the j formative which motivated the original rule cannot be postulated in the underlying forms of the stem in the kind of concrete analysis which I am proposing here. On the other hand, if we postulate underlying /fremm +/ as the stem, the lack of gemination in most of the forms of the paradigm can be explained by a rule which is a more natural rule than an unmotivated gemination rule and also one which reflects one aspect

106

of an empirically demonstrable historical process of simplification. This process is evidenced by the general tendency to shorten words by reduction, and in many cases syncopation, of unstressed vowels. The further extension of this process eventually meant the loss of a large number of inflectional endings in English. Moreover, the original gemination through the j_ formative was never an OE process but rather a West Germanic one, which was later undone independently in all

of the

his-

torically attested languages. The timing, however, was not the same in all

the Germanic dialects. As we have

seen, OS proved to be more conservative than OE and preserved the j_ formative at a time when OE had already lost it. This rule, degemination, reflects a well attested process by which geminates were shortened mainly in two environments, before consonants and word-finally. Brunner writes about degemination before consonants: werden vereinfacht, z . B . fyllan-fylde f ü l l e n ,

"Geminaten ws .

af ierran-af ierde e n t f e r n e n , wemman-wemde beflecken" ( 3 1 5 ) . Likewise, degemination is common at the end of words where hyl ' h i l l ' , dyn ' d i n ' , eal ' a l l ' , bed ' b e d 1 , cyn ' k i n 1 , hlyn "noise" are frequent variants of the standard forms hyll , dynn, eall , bedd, cynn, hlynn . In view of the formulation of the rule of degemination it

is interesting

to note here that whereas the word-final geminate is frequently deleted, it is always retained in the declined forms whose inflectional endings start whith a vowel, e . g . , n. sg . gen. cynnes , n. sg. dat. cynne , n. pi. gen. cynna , n. pl. dat.

cynnum.

In accordance with the rule of gemination in section 2 . 3 of chapter two, which is a lengthening rule, degemination will be formulated as a shortening rule:

[

+ cons] - syllj

For obvious reasons, this rule will only play a role in

107 the synchronic derivation of those verbs and their forms which underwent the historical rule of gemination, v i z . class III

and a large number of class I. The actual

derivations and a further discussion of the processes will be presented in section 6 of this chapter. 5.

Schwa-Insertion The rule of degemination which was proposed in the

preceding section is actually generally accepted. Notice that in the forms quoted above (fylde, afierde, wemde), in the past tense of the other long-stem verbs of class I, in all class III verbs, and many others the past tense marker -d- directly follows the stem. That is to say, d is the complete past tense marker and clearly identifies these verb forms as past tense forms. This fits in well with the diachronic development in which a dental element marks the past tense and the orthographic e which precedes it

in classical OE (i_ in EOE)

is simply the reflex of the old stem formative j_. In other words, preceding orthographic e or d. (or o, for that matter, in class II) is not an integral part of the past tense marker and is not required in order to unequivocally indentify a weak past tense. In Modern English this early development has been brought to a consistent conclusion because, of course, the dental (or alveolar) stop is the only audible weak past tense marker despite archaic spellings which retain £: grabbed [graebd] , kissed [ k i s t ] , sawed [so.d] , bribed [braibd] , walked [ w o k t ] , etc. Those verbs in Modern English with word-final t: or d actually insert [9] in the past tense in order to make the word pronounceable: rated I ' r e i t 3 d ] , braided I ' b r e i d 9 d ] , etc. That this unstressed vowel is inserted becomes also clear if we look at such forms as learned [ ' I s r n & d ] , beloved [be'lAVgd] , rugged [ f r A g 9 d ] , etc. Here schwa is epenthesized in order to assure the adjectival quality of these verb forms and to avoid confusion

( c f . learned [ l a r n d ] , past tense of to l e a r n ) .

108

Today, schwa-insertion before the past tense marker d is the easily identifiable exception rather than the rule. In OE times, a millennium before today's English andbefore an enormous number of changes in the linguistic system of English, schwa-insertion in this position was common, if not entirely predictable ( c f . Campbell: 323, Brunner:313 - 3 1 7 ) . In the weak past tense empirically supported rules of degemination and schwa-insertion in a synchronic OE analysis can be easily j u s t i f i e d .

A look at the present

tense of the weak verbs, however, suggests another area in which schwa-insertion must have applied in classical OE. I am referring to the endings of 2 , 3 sg. pres. ind. In the earliest attested forms the personal endings clearly contain a vowel before the consonantal segment(s) (see also the diachronic derivations in section 3.1 of Chapter t w o ) , but this original ( i . e . , in many cases prehistoric) condition becomes increasingly obscured before and during the OE period. Thus forms such as framest and fremst, fremep and f remfa, selep and selb, trymefr and trymb, etc.

come to exist side by side. Even

the remarkably stable deman shows occasional forms like demest, demep ( c f . Brunner:326) beside the normal demst, demb. It is obvious that the forms of the classical OE paradigm, i.e.,

the variants fremst, fremfci, selb, trymb,

demst, demb, etc.

are the result of a process of syn-

cope, details of which were discussed in section 2 . 1 0 of chapter two. By the time of the classical OE texts, the only segments that can be consistently isolated as markers for the second and third person singular are -st and -b_, respectively. This conclusion is nicely supported by verbs such as faencan, byncan, secan, sengan. All the infinitives of these verbs, and indeed all

the other

forms, except 2, 3 sg. pres. i n d . , are assumed to have had palatalized or even assibilated stem-final consonants. 2 , 3 sg. pres. i n d . , however, possessed velar stem-final consonants (see Campbell: 1 7 7 ) . The absence

109

of the inflectional vowel i prevented the palatalization of the stem-final

consonants. It is also possible that

the once palatalized stem endings reverted to velar stops once the inflectional vowel had disappeared. The point is that the inflectional vowel is clearly not present

in,

for instance, bencb, byncp, secfa, sengp; otherwise palatalization would have been retained. For these reasons we postulate as the underlying forms of 2 , 3 sg. pres. ind. /st/

and / ρ / , respectively. Like-

wise, for the reasons given at the beginning of the

sec-

tion, the underlying form of the past tense will be / d / . Schwa-insertion, which resembles syncope in its unsystematic nature and indeed is closely connected with it,

is here formulated as an optinal rule, in its most

general form: 0

-

I·]

/

/C

+

°

The problematic nature of this rule is taken up again in the next section. 6.

Synchronic derivations of the OE weak verbs

6.1

Class I The following discussion of the synchronic analyses

of classes I, II,and III will depend to a considerable degree on a knowledge of the phenomena treated in chapter two. The diachronic analysis had to precede the synchronic analysis because a relatively concrete analysis with a minimum of rules intervening between underlying forms and surface structure can only make sense if one is already acquainted with all the rules that must be taken into consideration, even if not all of them actually play a part in the synchronic derivations. Constant reference to a later chapter with the diachronic analysis would have made it

d i f f u c u l t for the reader to follow

the arguments in the present chapter. We will first look at the three subclasses of class I

110

which are represented by fremman, nerian, deman. It will be recalled that in the diachronic analysis the stemformative j_ was of crucial importance because it was directly involved in almost all the changes that occur up to the OE period. In the synchronic analysis presented here, however, jj_ will only play a marginal role. The obvious reason for this development is the complete surface opacity of this formative in verbs of the fremman and deman subclasses, and its inconclusive presence (from a synchronic viewpoint) in nerian. This and the additional arguments in the previous sections of this chapter will lead us to postulate / f r e m m + / , / n e r + / , and /dem+/ as the synchronic underlying representations for the stems of the verbs fremman, nerian, and deman, respectively. Again, for reasons discussed above, the underlying form for the past tense will be /d/. The inflectional ending for 2 sg. pres. ind. will have /st/ as its

underlying form, and 3 sg. pres. ind., / p / . Notice,

however, that unlike the diachronic form for 2 sg. past i n d . , the synchronic inflectional ending will be identical with 2 sg. pres. ind., namely / s t / . As in chapter two, five forms and their derivations will be given for each paradigm because they cover all

alternations and endings:

infinitive, 1 sg. pres. ind., 2 sg. pres. ind., 2 sg. imperative, and 2 sg. past. ind. (For a justification of the postulation of /# ner + jan #/ and / # ner + je #/ see section 2 . 2 . 1 of chapter one, sections 3 and 6 . 2 . 1 of chapter three, and section 5 of chapter f o u r . ) UF

/#fremm+an#/

/#fremm+e#/

/#fremm+st#/

1. degem 2. schwa-ins 3. red

fremm+Λη

fremm+β

frem+st frem+^st

output phonetic orthographic

fremm+Λη ['fremmAn] fremman

fremm+9 ['fremm9] fremme

frem+ist ['fremast] fremest

111

/#fremm#/

/#fremm+d+st#/

1. degem 2. schwa-ins 3. red

f rem frema

frem+d+st frem+dd+dst

output phonetic orthographic

frem9 ['freme] freme

frem+dd+ast t'fremeddst] fremedest

UF

UF

1. degem 2. schwa-ins 3. red output phonetic

orthographic

/#ner+st#/

/#ner+jan#/

ner+Sst ner+jAn

ner+je

ner+jAn ['nerjAn] nerian

ner+jd I'ner nerie

nera

ner+dd+8st

ner+8st ['ner»st] nerest

UF

1. degem 2. schwa-ins 3. red

ner+3d+3st ['neriddst] neredest

output phonetic orthographic UF

/#dem+an#/

1. degem 2. schwa-ins 3. red

dem+An

dem+d

output phonetic orthographic

dem+An [ "de:mAn] deman

dem+6 ['de:me] deme /#dem+d+st#/

UF

1. degem 2. schwa-ins 3. red output phonetic orthographic

dem+st [^dermst] demst

dem+d+Bst dem [de:m] dem

dem+d+est demdest

The synchronic derivations of the three major subclasses of the class I weak verbs are straightforward. The restructured underlying representations are now highly

112

transparent because they contain all those features that can reasonably be expected not to be handled by rules in a synchronic grammar. The appropriate arguments were presented in this chapter and in chapter two. The number of rules that are actually required is minimal, three, and there are cases where the underlying form is also the surface form. Writing about phonological change and its dependence on performance-related targets, Bynon ( 1 9 7 7 : 138) makes the following statements which appear quite relevant in the present context: "The basic principle governing systematic restructuring must clearly be sought in the transparency of surface structure with regard to the underlying functional relationships, the ideal being a simple direct correlation between the two--which tallies well with the traditional interpretation of analogical change. As Kiparsky has pointed out, this principle appears to be closely connected not only with the requirements of language acquisition but also with those of speech perception and production, all of which demand that underlying relationships be unambiguously expressed in surface structure." A few forms and derivations may require some comment. The occasional variants fremst-fremest, fremb-fremeb in the short-stem verbs of class I and the rare demest-demst, demeb-demp are due to the relatively unpredictable application of schwa-insertion (see above, section 5 ) . This is nowhere clearer than in the 2 sg. imperative of the subclasses fremman and nerian. Here schwa is epenthesized in word-final position. Recall that the final -e_ of f r erne cannot be considered a marker of the imperative since it would then prevent degemination from applying. Thus the final -e is indeed to be seen as a result of a rule of schwa-insertion. It is interesting in this connection to note that in a 1971 article Lass presents arguments for treating boundaries as obstruents. He mentions the fact that boundary environments rather than sonorant environments account for the voiceless fricatives in forms like

113 healf

' h a l f 1 , heals ' n e c k ' , a5 O a t h * . If these environ-

ments were sonorant environments, we would expect voicing word-finally. He writes: "This suggests that in some sense the boundary H is higher in some scale or hierarchy of ' s t r e n g t h ' than sonorants and that the presence of # in an environment will always override the e f f e c t of a sonorant in a given case" ( 1 6 ) . The only segment higher than a sonorant in a strength hierarchy is of course an obstruent. In that sense the environment in the derivations of 2 sg. imperative freme, nere after degemination can be seen as an environment that corresponds to the schwa-insertion environment and schwa-insertion therefore applies. However, if this were f u l l y applicable, we might expect forms like *fremeste, etc., which do not in fact occur. The insertion of schwa in 2 sg.

imperative must

therefore be seen as conditioned by a morphological rather than a phonological environment for the rule of schwa-insertion. It is interesting to note in passing that the morphological environment for the schwa-insertion rule is a direct result of the necessity to posit a single underlying form for the stem of the fremman type verb—more specifically an underlying stem with geminated final consonant. As I have already pointed out in chapter two, section 3.1, the original imperative form of fremman was fremi. The final i_ in this case was not an original imperative ending but rather a vocalization of the j formative. This vocalization prevented gemination and in a synchronic analysis where the j_· formative no longer exists as an independent morpheme it would be natural for the i.', which had now been reduced to schwa and written e, to be considered as an independent marker of the imperative. It is assumed, however, that the general principle of assigning only one underlying representation in cases like these is valid and that the arguments that were presented in favor of degemination in preference to gemination in a synchronic grammar of OE are sound. That

114

there was a conflict between the underlying forms with geminated final consonants and the morphological environment for schwa-insertion is not to be doubted. We will see in chapter four how this conflict was resolved. Finally, it

should be noted that the deman paradigm is

by far the simplest of the three. Only in the past tense f o r m , demdest, does the rule of schwa-insertion apply because the consonant cluster revealed in the underlying form ( / d e m + d + s t / ) would be too large for the syllable structure restrictions of OE. As a general rule we can assume that the more complex the environment is

(i.e.,

the longer the consonant sequence) the less optional

is

schwa-insertion. In all other cases the underlying form is also the surface form, if we disregard for the time being the relatively minor e f f e c t of the reduction rule. This fits in well with the observation that deman is the only type of weak verb which, in OE times, is completely devoid of any surface forms of the stem formative j and which is therefore the most regular of the weak verbs of class I. This regularity was maintained beyond the OE period, where in Middle English the OE long-stem verbs of class I provide the model for one of the categories according to which the weak verbs are divided by then: the absence or presence of a medial vowel before the dental element in the past tense. 6 . 1 . 1 byegan, sellan bycgan and sellan in their synchronic derivations are somewhat d i f f e r e n t form the other verbs of class I, but in general their analysis is the same. The main difference is the unumlauted vowel in the past tense stem and the absence of a medial vowel before the dental s u f f i x in the past tense of classical OE. As in the diachronic analysis

(section 3 . 1 . 1 of chapter t w o ) , the stems pos-

tulated for the present tense and the past tense of bycgan will d i f f e r , i.e., bohte is considered a very old suppletive form of bycgan for reasons detailed in the

115

diachronic analysis. It was also pointed out in chapter two that orthographic eg represents the geminated form of g_ and that in the earliest forms, because of the unclear range of assibilation and palatalization, eg is most reasonably assumed to have been a palatalized long voiced stop [gg ] rather than a f u l l y assibilated a f f r i c a t e . That this is indeed the case is evidenced by ME by en, MnE buy. When, at the end of the OE period, verb forms ceased to have geminated stem-final consonants, the infinitive of the verb to buy met the general OE requirement that original voiced velar stops became voiced palatal glides in the environment of front vowels ( c f . synchronic derivation of bycgan, 2 sg. pres. ind. and 2 sg. imperative) . These glides then coalesced with the preceding front vowel to form a long vowel which then participated in the Great Vowel Shift. Had orthographic eg represented, in this case, a single sound ( i . e . , voiced palatal a f f r i c a t e ) , we would expect to have Modern English [ b i d z l , just as we have bridge [ b r i d a l / ridge [ r i d z ] , etc. For this reason we will assume in the synchronic analysis that the underlying representation for the present stem of bycgan is /bygg^+/ with a geminated stem-final consonant, in accordance with the other verbs of class I. In the past the stem will be /boh+t+/. The derivation for the five basic forms of the paradigm will be as follows: /tfbygg^+antf/

/#byggj+e#/

1 . degem 2. schwa-ins 3. red

τ bygg-'+An

τ bygg-'+J

output phonetic orthographic

bygg-^+ΛΠ ['bygg^An] bycgan

bygg-'+^ ['bygg^3] bycge

UF

byg-' + st -

bygst

116

UF

/tfbygg^/

/#boh+t+st#/

1 . degem 2. schwa-ins 3. red

byg·? byg-^3

boh+t + dst

output phonetic orthographic

byg-'e ['byJ9] byge

boh+t+ast ['boxtdst] bohtest

The synchronic analysis of sellan is not much d i f f e r ent from that of bycgan. The present tense stem, which in the diachronic analysis was remarkable because of the absence of Breaking, poses no problems in synchrony because it

corresponds exactly to that of all the geminated

verbs of class I ( f r e m m a n ) . The past tense stem, as in bycgan, is suppletive and is / s e a l + d + / : UF

/#sell+an#/

/#sell+e#/

/#sell+st#/

1. degem 2. schwa-ins 3. red

sell+Λη

sell+9

sel+st sel+9st

output phonetic orthographic

sell+Λη ['sellAn] sellan

sell+6 ['seile] seile

sel+9st ['seiest] seiest

UF

/#sell#/

/seal+d+st#/

1. degem 2. schwa-ins 3. red

sei seid

seal+d+9st

output phonetic orthographic

sei ['se!9] sele

seal+d+9st [*se9ld9st] sealdest

6.2

The weak verbs of classes II and III

6 . 2 . 1 Class II The synchronic derivations of weak class I verbs show a remarkable uniformity of stems and inflectional endings in OE times, despite an historically large variety of subclasses and subgroups. Class I I , although without this variety of subclasses, is different its if

and more complex in

OE surface forms. The OE stems of the class II verbs, compared to the Gmc. proto-forms presented in Chapter

117

two, are basically unchanged because breaking and i-umlaut did not affect the stem vowels. And neither did the stem-final

consonants undergo gemination or palata-

lization. Thus the underlying representation for the stem of the quintessential class II weak verb, l u f i a n , will be / l u f + / . The Gmc. proto-form was | l u f + | . The situation as far as the verbal endings are concerned is rather d i f f e r e n t in OE. In proto-Gmc. the presence of a stem formative was an obvious feature of all

weak

verbs. In class I this stem formative, amply attested, was j_ throughout. In class II two different

formatives

had to be postulated for the present tense, oj and o, and one for the past tense, u, followed by the past tense marker d. As we showed in section 3 of this chapter and the synchronic derivations of class I verbs in section 6 . 1 , the original stem formative j_ had completely lost an identifiable function as a stem formative in OE times because of its occurrence.

irregular and inconclusive surface

OE nerian which, as the only subclass of

class I, retains a reflex of the stem formative j in some forms reanalyzes this j in such a manner that the orthographic representation .i, as, for instance, in 1 sg. pres. ind. nerie, pi. pres. ind. neriap, is in OE times the first segment of the personal ending rather than the last of the stem. That this must indeed be the case is evidended by the complete regularity of the stem which results: it is /ner/ throughout. This tendency to regularize the stem, sometimes at the cost of the regularity of the endings, is documented by such class I verbs as gierwan which should have no w in those forms in which fremman has a non-geminated stem consonant. Frequently, however, these verbs extend w to all

forms so that, for instance, 2, 2 sg. pres. ind.

appear as gierwest, gierwep instead of the expected gierest, gierep. But they may also follow a pattern which eliminates w in all

forms, even those that should

118

historically contain it.

This holds also true for class

II verbs in which in certain MSS i_ is either extended to all

forms or deleted in all

3 3 4 ) . The result is in all

forms

(see Campbell: 3 2 7 ,

cases a more regular stem.

This clearly supports our analysis of nerie as / n e r + j e / and the postulation of /ner/ as the common OE stem for nerian.

(See also section 3 of this chapter).

Going back to class II,

we are faced with a situation

which, for example, in 1 sg. pres. ind. l u f i ( g ) e , pi. pres. ind. l u f l a p , i n f . l u f i a n , is very similar to that in class I nerian. It is easy to see why, as in nerian, the regular synchronic stem / l u f / can be established. In accordance with the nerian subclass, the personal endings of the forms just listed above are / i e / ,

/iab/,

and / i a n / , now without a stem formative intervening between stem and ending. But in the past tense, 2 sg. imperative and 2, 3 sg. pres. ind. the situation is not quite so clear cut: lufodest, l u f a , l u f a s t , lufap pose problems because the occurrence of ο and a in the endings that have e in class I can probably not be accounted for in the same manner. The existence of back vowels in the endings suggests that, at least in EOE, these back vowels must have been phonetically manifested in some form. This raises the question how these vowels should be treated in a synchronic analysis of classical OE. If we begin by looking at the past tense, as in 2 sg. past ind. lufodest, we will find that there are four possibilities of how to analyze the ο preceding the past tense marker: 1. 2. 3. 4.

ο is part of the stem. ο is an independent morpheme. ο is epenthetic. ο is part of the past tense morpheme of class II weak verbs.

Case 1 has to be ruled out because ο appears nowhere in the present stem and, considering the general tendency of verb stems to become regularized in OE, because this

1 19

would not represent a consistent solution. Case 2 appears somewhat more plausible, o was indeed an independent morpheme in prehistoric times (see chapter two), but this is impossible to j u s t i f y synchronically . Its synchronic meaning is not clear since the past tense is sufficiently marked by -d- . In case 3 we have to assume that schwa was epenthesized and then rounded in the environment of a back rounded vowel . That this is not entirely impossible is shown by such forms as fugol 'bird 1 masc. sg. nom. and f ugeles 'bird* masc. sg. gen. which Campbell ( 1 5 1 ) mentions in connection with parasite vowels: "This vowel, if developed, was i^ (later e) after a front vowel, u (later o) after a back vowel." But this too can be disproven since many cases exist ( e . g . , waestum ' f r u i t ' masc. pi. dat.) where this process did not occur, thus showing an inconsistency that is not at all discernible in class II verbs. And, of course, all verbs in class II have -od, even if their stem vowels are front vowels. This leaves case 4. Since we have exhausted all

other alternatives, we are forced to assume

that in OE ο is an integral part of the past tense formative of weak class II verbs. We will therefore postulate /+od+/ as the past tense s u f f i x of class II. For 2 sg. imperative luf a (cf . nere , f reme of class I nerian, f remman) , it

could also be suggested that a. is

part of the stem or that it

is epenthetic. But these

proposals have to be rejected for the same reasons as those given for the past tense. The most likely analysis of a is that it

is indeed an independent morpheme which,

if word-final, is the sg. imperative marker of class

II:

Finally, a in the personal endings of 2 , 3 sg. pres. ind. will have to be considered part of these personal endings. The reasoning here is the same as in the case of the past tense. It is evident from these considerations that the class II weak verbs were able to maintain a great deal of individuality well into OE times. And it

is also

120

obvious that they were stabler, from a diachronic point of view, than class I: a minimum of changes occur from proto-Gmc. to OE times. The synchronic derivations of lufian are as follows: /#luf+ian#/ 1 . schwa-ins 2. red output phonetic orthographic

luf +iAn [ ' luviAn] lufian

/#luf+ie#/

/#luf+ast#/

luf+i8

luf+Ast

luf+ia ['luvid] luf ige luf ie

luf+ASt

[ ' luvAst] lufast

/#luf+od+st#/ 1. schwa-ins 2. red

luf+A

luf+od+9st luf+Ad+3st

output phonetic orthographic

luf+A f'luvA] lufa

luf+Ad+ast t ' l u v A d 9 s t ] or [ ' l u v e d O s t ] lufodest

Since gemination never played a role in the development of class II verbs, its counterpart degemination is not a part of these synchronic derivations either. Schwa-Insertion does apply to a limited degree at this stage but will apply more frequently as the unstressed vowels lose their back vowel identity (on this point see also chapter f o u r ) . The various reduction rules treated in chapter two come into play again because the process of reducing every unstressed back vowel to schwa is of course a gradual one that must have gone on throughout the OE period, as indeed it

is operative today. To determine

the exact stage of this reduction process in classical OE, around the year 9 0 0 , is naturally very d i f f i c u l t or not at all possible. For reasons stated in chapters one and two, we can, however, assume that the reduction

pro-

cess had advanced further than is evidenced by the conventional OE orthography. The stage at which all

un-

stressed vowels are reduced to schwa is forseeable

from

the synchronic derivations given here. That this is likely to have occurred quite some time before the be-

121

ginning of the ME period is also evident from the many OE dialectal spellings that actually contain e in those environments that historically have back vowels. See Brunner (335) for details on this point. 6 . 2 . 2 Class III The difficulties posed by the weak verbs of class III have already been discussed in chapter two. While it was relatively unproblematic to propose common Gmc. protoforms for these verbs, it is virtually impossible to postulate one underlying representation for the stem of all verb forms of a given paradigm of class III. Forms like 1 sg. pres. ind. libbe and 2 sg. pres. ind. leofast (libban) or i n f . secgan and 2 sg. past ind. saegdest within one WS paradigm point up the difficulties involved. Even if the alternations could be taken care of by certain underlying forms and the appropriate phonological rules, the question remains, especially within the framework of the views put forward in this paper,whether this would represent a viable analysis of the data from a synchronic standpoint. Chances are that each paradigm of class III, habban, libban, secgan, hycgan, contained a number of suppletive forms in OE that were, however, reasonably well associated with each other because of the extremely high frequency of these verbs. But even so these verbs were soon absorbed by class II, as many class II verbs before them had been in prehistoric times. LWS already has orthographic evidence of a complete class II conjugation of libban, which appears there as leofian, lifian. All this points in the direction of a "class" whose conjugational patterns had lost their productivity and whose great complexity in classical OE represents an intermediate stage in the efforts of this small group of verbs to realign themselves with a productive class.

122 CHAPTER FOUR: THE TRANSFER OF OE WEAK VERBS OF CLASS I TO CLASS

1.

II

Introduction

In chapter two we traced the development of the weak verbs from roughly proto-Gmc. to classical OE times. In chapter three a synchronic analysis was proposed which is actually simpler, i.e., less abstract, than other analyses which are current within the standard generative phonological framework. In this chapter we will attempt to show that one of the strengths of our synchronic analysis is that it provides a principled—and in some way natural—explanation of the reshaping of the weak verb system during the OE period. This reshaping is evidenced by such forms as fremian (attested already in EOE times) and nerigan (attested in LOE; cf. l u f i g a n ) . Such forms show that already in the classical OE period, which is taken as the basis of all of the generative descriptions, the traditional four paradigms divided into large classes (viz., the three subclasses of class I plus class II) had given way to an essentially twoparadigm system: the deman type on the one hand, and the lufian, nerian, fremian type on the other. The fact that fremman, and not fremian, is taken as the standard paradigm by the handbooks would seem to indicate that its incorporation (and that of many similar verbs) into the lufian, nerian type was only fully accomplished during the later OE period. In fact, if we look ahead for only one century to early ME times, we find a complete leveling in the present tense, the only difference in the paradigms of the weak verbs being whether they have a medial vowel in the past tense. Compare for this the following table (after Mosse:76, 81) which gives

123

the southern forms of the weak verb, i.e., the direct descendants of the WS forms of classical OE. lufian will be taken as the verb presenting the uniform present tense inflections of all weak verbs in ME. 1

pres. ind.

sg 2 3 pi imperatives 2 sg 2 pi present participle

lufe luf(e)st lufep lufep luf (e) lufep

pres. subj. (optative) sg

1

lufe

2 3

lufe lufe lufe(n) lufe(n)

pi infinitive

lufinde lufing(e)

In the past tense the two types, with and without a medial vowel, are exemplified by lufian and deman. 1

past ind.

sg 2 3 pi past participle

lufedest lufed(e) lufed(e)(n)

past subj. (optative) sg

1

lufed(e)

2 3

lufed(e) lufed(e) lufed(e)(n)

1

demde

2 3

demde demde demde ( n )

Pi

(i)lufed

1

past ind.

sg 2 3 past participle

lufed(e)

pi

demde demde s t demde demde (n)

past subj • (optative) sg

pi

(i)demd

The abstract representations which Kiparsky/O'Neil and others postulate for the OE verbs—which necessitate derivations recapitulating the historical development from PGmc.—do in fact provide an adequate description of the alternations attested in the standard paradigms. What they fail to do is to recognize that these paradigms already represent a system far removed from the original Germanic system. It is virtually impossible to explain, on the basis of these analyses, how and why this reshaping took place, and the whole pattern of development for the OE period is obscured. Any synchronic analysis

124 of the ME forms, if compared to the abstract analysis of the OE paradigms, would imply a massive restructuring within a very brief space of time—a most unlikely event. In this chapter we will begin by demonstrating the actual stages of development in the structure of the OE weak verb system, stages which fully justify the concrete synchronic analysis which we have posited on the basis of a restructured verb system. We will then show that there are arguments of various types which, taken all together, provide a plausible explanation of why the reshaping of the weak verb system took place and why it took the form it did. This last point will lead us into a discussion of the theoretical position of analogy in historical generative phonology. 2. 2.1

Stages of transfer Loss of the j_ formative Throughout this paper I have emphasized the catalytic

role played by the IE j. formative in the phonological development of the OE weak verbs. In previous chapters we have concentrated on the phonological rules which were triggered by the presence of the j_ formative but we have not systematically discussed the stages by which the j_ formative became first a relic of the IE past and then, by ME times, disappeared completely from the grammar of weak verbs. We believe that the j formative was lost in the following stages. First, it was lost in the deman paradigm. Despite related non-derived nouns like dorn, there is not a single trace of an alternation in the verbal paradigm itself. In other words, there is no distributional evidence whatsoever in OE which might hint at the former existence of j in the long-stems verbs of weak class I. Next, or possibly at the same time as in deman, the formative j was lost in the fremman paradigm. Again, as in deman, the total absence of j_ on the surface is an important indication of its loss but, unlike deman, the

125

alternation between geminated and non-geminated stems still shows one of the historical e f f e c t s of the j_ formative in this class of verbs. That this a l t e r n a t i o n , however, is not s u f f i c i e n t grounds for postulating the continued existence of the j formative in the underlying form of the stem in classical OE time was demonstrated in section 6 of chapter three, j had no semantic or morphological

function

in OE. Of the two main rules

which the j_ triggered historically, there are no synchronically relevant alternations in the OE weak verb system based on umlaut; and alternations based on gemination are restricted to the present tenses of fremman, verbs of the bycgan type, and class III hycgan. The appearance of j_ in l u f i a n and nerian can easily be reinterpreted as part of the ending rather than the stem, although this means that allomorphy in the verb endings is increased. It was shown, however, in sections 3 and 6 . 2 . 1 of chapter three that there was a tendency in OE to regularize the stems and not the e n d i n g s . The greater allomorphy which resulted in the i n f l e c t i o n a l endings was undoubtedly a short-lived phenomenon, as evidenced by the Middle English verb paradigms in section 1 of this chapter. 2.2

A two-paradigm system As we have seen, the deman paradigm, without

surface

j in OE, remains completely stable. It is indeed the one weak verb paradigm which is quite uncontroversial and whose derivation requires little explanation. The problemswhich have arisen have all

been within the ana-

lysis of the other three paradigms. Our thesis in this section is that the basic e f f e c t

of the development of

these paradigms during the OE period was that the

fremman

and nerian subclasses came to be conjugated like l u f i a n . This, in f a c t , constituted a transfer of class I verbs to class II. This is by no means speculative. The increasing tend-

126

ency in LOE times for the frenunan and nerian subclasses to conform to the l u f i a n pattern is well attested. fremman becomes fremian, of which the following forms are attested: 1, 2, 3 sg. pres. ind. fremie,

fremast/

fremap, pi. pres. ind. fremiap, 1, 3 sg. past ind. fremode/

etc.

Similarly, nerian has 1, 2 , 3 sg. pres.

ind., nerie, nerast, nerab, etc.

The result of this

transfer is a subcategorization of the LOE verb system according to the conjugational patters of two types of verbs, deman and lufian that all

(see section 1 ) . This is to say

the other conjugations had completely

dis-

appeared at the end of the OE period. 3.

Reasons for the t r a n s f e r Having shown through late attested forms that a trans-

fer did take place and having shown its prehistory through the progressive loss of the j formative, I would now like to speculate on the mechanics of the transfer process. It will probably never be possible to unequivocally ascertain just how and why this transfer took place but the evidence which will now be presented contains a variety of factors that will allow some reasonable guesses. An explanatory approach to the problem of transfer can be divided into two questions:

1. why did the three paradigms

fremman, nerian, l u f i a n become one paradigm in the first place?; and 2. why did the lufian paradigm become the model to which the other verbs

transferred?

The most natural answer to the first question would logically be based on a high degree of formal similarity so that the conflation of paradigms would be seen as nothing more than the smoothing away of certain minor anomalies in an otherwise unitary paradigm. The fact that this was the case with lufian and nerian has been alluded to several times in this paper. This is also supported by Campbell's statement ( 3 2 5 ) that in the nerian subclass "all

forms of the present system are

ambiguous except 2nd and 3rd sg. pres. indie, and imper.

127

sg." The source of the ambiguity quoted by Campbell is something which was mentioned as early as chapter one, namely the question of how many syllables the verbs in question actually had. Very early in the OE period nerian [ ' n e r j a n ] and l u f i a n [ ' l u v i a n ] had clearly been distinguished

as dissyllabic and trisyllabic, respec-

tively. In EOE a tendency to pronounce the high f r o n t glide in nerian (orthographic jj like a high f r o n t vowel (orthographically also JL) became more and more evident until, in later OE,

[i] seems to have been the only

pronunciation. This is clearly shown in such forms as nerigan and nerigean in which the orthographic cluster -ig- and -ige- is the best and most reliable indication of the pronunciation of this subclass of class I verbs as trisyllabic words. Likewise, this must have been the case in forms that contained surface i,

e . g . , 1 sg.

pres. ind. nerie, pi. pres. ind. neriap, pi. imperative neriab, etc.

In several cases forms are attested that

are completely identical,

for instance, i n f . nerigan,

lufigan or 1 sg. pres. ind. nerie, l u f i e , and others. This phonetic similarity was made possible by the relic nature of the stem formative j_. The interpretation of j_ as part of the ending in nerian led to allomorphic variation in the endings of the i n f i n i t i v e 1 sg. pres. ind. other forms

[ J 9 , 13], pi. pres. ind.

[jan, ian], [ j a p , iap] and

(see paradigm in chapter o n e ) . The so-called

parasite vowel in the nerian paradigm ( c f . n e r i g a n , nerigean) may be seen as an attempt to resolve the allomorphy in forms which were already nearly identical. It was resolved in favor of the trisyllabic model represented by l u f i a n , and not surprisingly so, because, as Campbell ( 1 5 2 ) notes, there was a tendency especially in WS, and increasingly frequent in later texts, "to develop Q i and u a f t e r a short syllable to ii, herigan praise, byrig d.s.

u u , e . g . herigas armies,

city, bylig bag, mirigb j o y ,

fyligan follow, byrigan bury, gearuwe n . p . ready, beaduwe d.s.

battle, seonuwa n . p . sinews, swaluwe swallow." This

128

is of course a crucial point because it

shows clearly

that there was indeed a trend to vocalize glides in certain environments and that this trend favored the adoption of lufian as a model for class I verbs because class II verbs had a vocalic high front segment throughout the entire OE period. In those forms that did not contain a surface .i, like 2 , 3 sg. pres. ind. and 2 sg. imperative, the clear cut surface level distinctions between the two paradigms in the derivations were obscured by the vowel reduction rule so that in LOE the native speaker of English could not readily assign a given surface form to one class or the other. [Λ] are phonetically very close.

[9] and

These circumstances are s u f f i c i e n t to answer, at least for nerian and l u f i a n , the f i r s t question and to some degree the second. But why do we increasingly find forms like nerast instead of forms like lufest in West Saxon? (The reader may recall that Brunner calls the presence of the inflectional a in 2 , 3 sg. pres. ind. "ganz fest" [ 3 3 3 ] . ) Aside from the vocalization tendency cited above, the answer to this question lies, I believe, in a nonphonological aspect of the OE weak verbs, namely that of sheer numbers. The nerian subclass was always a very small group in OE, whereas lufian was the result of a highly productive process of denominative verb formation ( c f . the discussion of Krahe/Meid's views in section 2 . 2 . 2 of chapter one) which is productive even today, e . g . , ground ^ plane, contact a person, etc. Thus the number of lufian type verbs which the OE child was confronted with outnumbered by far the number of nerian type forms. Class II was the largest weak verb class in OE and it was a productive class. Under these circumstances it was not surprising that the gravitational field of this class, together with the additional

factors

of phonetic similarity and regular stem, grew larger and larger. In the case of fremman, answers like phonetic simi-

129

larity and sheer numbers do not apply. The phonetic similarity between nerian and l u f i a n derive from a common CVC stem type plus an ending beginning, in the crucial cases, with a high front segment. In fremman, the stem alternates between CVC and CVCC structures and none of

the

endings begin with a high front segment. With regard to numbers, although the fremman type verbs were not as numerous as the l u f i a n type verbs, they still constituted a fair proportion of OE weak verbs. Part of the answer to the fremman problem was the result of the

increasing

simplification of geminated consonant clusters in later OE--a simplification which aided the establishment of a single CVC stem type for the fremman verbs. These verbs would then be, in every particular but one, identical to the nerian-lufian model. This particular is the presence of a high front element in the endings which was supplied to the appropriate forms since we do get forms like fremian for the infinitive. Moreover, 2, 3 sg. pres. ind. become fremast, fremap, respectively, indisputable proof that transfer to class II had taken place. Unlike the case of nerian, where the forms with a high front segment already existed, there is only one way to account for the presence of a high front element—insertion on the analogy of the nerian-lufian type. That forms like fremian are indeed the result of analogy and not a return to some historically motivated underlying form ( a s , I am sure, many defenders of standard generative phonology would claim) is shown by the fact that the stem of fremian has the umlauted vowel e, i.e., we never find the form *framian. In other words, there is no reversal of historical developments, but an entirely new development. Before drawing some final conclusions it

is important

to consider the status of analogy as a viable concept within the framework of this paper and generative phonology in general.

130

4.

Analogy The term 'analogy 1 goes back to the nineteenth-century

neo-grammarians. In its

strictest sense it was intended

to provide an explanation for all

those changes which

could not be shown to have a phonetic basis. The concept is defined as follows by Hermann Paul ( 1 9 2 0 : 1 1 0 ) : Die Wörter und Wortgruppen, die wir in der Rede verwenden, erzeugen sich nur zum Teil durch blosse gedächtnismässige Reproduktion des früher Aufgenommenen. Ungefähr eben so viel Anteil daran hat eine kombinatorische Tätigkeit, welche auf der Existenz der Proportionengruppen basiert ist. Die Kombination besteht dabei gewissermassen in der Auflösung einer Proportionengleichung, indem nach dem Muster von schon geläufig gewordenen analogen Proportionen zu einem gleichfalls geläufigen Worte ein zweites Proportionsglied frei geschaffen wird. Diesen Vorgang nennen wir Analogiebildung. Although P a u l ' s description of analogy seems intuitively clear, the assignment of any specific change to this category (as opposed to "sound change") depends crucially on what one understands as belonging to the phonological component of the grammar. The current favor in which the term 'analogy

1

dis-

is held by adherents

to standard generative phonology is not a result of any basic disbelief in the phenomena it is intended to account for.

The term rather than the concept has come

to s u f f e r as a result of its

continued use by the

structuralist and traditional schools of phonology. In the massive polemic leveled at the structuralists by generative phonologists in the late 1 9 6 0 ' s the traditional concept of linguistic change as describable in terms of the two principles of sound change and analogy received some heavy, sometimes erratic, criticism. Postal ( 1 9 6 8 : 234 fn.

3) writes:

'Analogy' is really an unfortunate term. There is reason to believe that rather than being some sharply defined process, analogy actually is a residual category into which is put every kind of linguistic change which does not meet some set of a

131 priori notions about the nature of change. In particular, I think that the term 'analogy 1 has been used very misleadingly to refer to cases of perfectly regular phonological change in which part of the conditioning environment involves Surface Constituent Structure, i.e. changes which happen only in nouns, or only in verb stems, etc. I suspect that any analytic survey of cases which have been referred to as 'analogy 1 would yield many instances of regular phonological change with nonphonetic environments. It is interesting to note that Postal presents nor arguments against analogy, but merely o f f e r s some suspicions that there is no true concept behind the term, that it

is merely a catchall for changes which are

neither phonetic nor autonomous phonemic. What Postal is really attacking here is not analogy as such but the structuralist model of phonology.Through criticism such as Postal's the term 'analogy 1 has become almost a taboo word among generative phonologists. Analogy as a principle, however, has not been abandoned. But it emerges with different labels. When r e f e r r i n g to the fremman subclass of weak class I verbs and its transfer to class II, it is simply not possible to propose a watertight phonological solution. K i p a r s k y / O " N e i l ( 5 4 6 ) state that " [ t ] h e great variety of dialectal variants in the third class as well as the tendency for

all

weak verbs to move into the second class might indicate morphologization of the system." Similarly, they suspect that some of their rules are "partly morphologically conditioned," without, however, elaborating on this point. This morphological conditioning is of course some form of analogy. Kiparsky ( 1 9 7 2 : 208 - 2 0 9 ) suggests a functional factor in phonological and morphological analysis, which he calls 'paradigm coherence" or 'paradigm u n i f o r m i t y ' , a f t e r realizing that in one particular example (the generalization of r throughout the inflectional

paradigm of a class

132

of Latin s-stems, e . g . , honos -> honor) a phonological solution is not adequate. The grammar cannot be simplified if only formal concepts like bleeding and feeding orders or opacity and transparency are used. Instead, he states that "allomorphy within a paradigm tends to be minimized." This is a new principle in generative phonology and, as Kiparsky states, it

amounts to what would have been

called analogical leveling or analogical extension in traditional philology. Kiparsky recognizes quite well that "this principle is

[sometimes] implemented at the

cost of complicating the system of rules"

( 2 0 8 - 9 ) and

that "paradigm coherence is an independent factor, and not wholly reducible to formal properties of rules and relations between rules" ( 2 0 9 ) . As has been shown, the concept of analogy has been gradually reintroduced into generative phonology but it is, even today, far from being generally accepted. Kiparsky himself does not refer to it

in K i p a r s k y / 0 ' N e i l

( 1 9 7 6 ) . We will go back to another aspect of analogy in the next section but it

is clear now that what has

been said in sections 2 and 3 of this chapter and elsewhere, viz. the various points referring to phonetic similarity between the paradigms, the change in the morphological structure where the reflex of the stem formative j becomes assigned to the inflectional

ending

in nerian and l u f i a n , or the regularization of the stems of the weak verbs, all this points in the direction of analogy as a necessary concept in linguistic change. 5.

Conclusion One of the problems connected with analogy is the fact

that the concept has so far escaped rigorous definition or formulation. This is particularly d i f f i c u l t because the proportional equations proposed by Herrmann Paul (see section 4) do not allow a prediction of the directionality of analogical change, which would be highly desirable. Although this formal definition,

especially

133

in individual cases, will remain d i f f i c u l t for some time to come, it

is possible to make a few statements

on the operation of analogy as a general trend. If. we look at the development of the OE weak verbs from the earliest times to the ME period and particularly the transfer of class I to class II, it becomes obvious that the most striking trait of this development is the elimination of verb classes or verb forms that are in some manner irregular. Thus, the regularization of the stem of the fremman type verbs where the stem alternation between geminated and non-geminated forms was eliminated and, in very early OE, the regularization of the inflectional endings of nerian and lufian that contain i_ are actually simplifications of the entire weak verb system that in the long r u n , namely in ME times, leaves only two verb types. This is the deman group on one hand which is historically straightforward and which underwent practically no changes throughout its OE and ME history (except for some unstressed vowel r e d u c t i o n ) , and the lufian group on the other which also undergoes a minimum of changes (no gemination, no umlaut, no palatalization, no breaking) but retains its distinctiveness because of a continuously productive process of denominative word formation and its already originally large number. Occasionally, analogy produces paradigms in which, for instance, the ±_ of the luf ian paradigm is extended to all verb forms of the present tense, or others in which the i. is deleted from all

forms. In both cases, however,

the deletion or extension represents a regularization or simplification of the paradigm, even though the f i r s t

is

objectively more complex than the second. But the further development in ME shows that there too the

objectively

simpler paradigm, that without ^, wins out. Standard ME preserves no forms of former

( O E ) class II verbs with d..

Perhaps more speculative but not implausible is the explanation that may be given in this context for the failure of the present tense stem of the sellan subgroup of class I to show the effects of breaking. Again ana-

134

logy with its

tendency to simplify stems and endings of

the verbs can be cited. Had the present tense stem undergone breaking, it would have resulted in a further complication of the already complex geminated class I stem. The past tense stem, however, with its

simpler

stem structure was better able to accommodate an additional change. There is,

of course, the possibility that

the present tense of sellan did undergo breaking and that it was leveled, already in prehistoric times, to conform at least with the rest of geminated class I verbs. The results would have been the same in both cases. The general tendency of analogy then, at least in the case of the OE weak verbs analyzed in this paper, is to regularize and simplify the grammar of a language and to indeed minimize allomorphy, as Kiparksy put it. In this sense, and in admittedly rather general terms, it is possible to predict the directionality of analogy. In sections 3 and 4 of this chapter several types of evidence for the reasons and manner for the transfer of class I weak verbs to class II were presented. It was my

intention

to

show how this transfer supports the

claim that the system underlying the classical OE weak verb paradigms had undergone considerable restructuring from its Proto-Germanic stage. In chapter two a diachronic analysis of the data of the OE weak verbs provided the broad background for a simpler and more concrete synchronic analysis of the same data in chapter three. It was found that although most alternations were preserved into historical OE times, many rules, and thus processes, can no longer be claimed to be viable in a synchronic grammar of classical OE. There can be no doubt that parts of the highly abstract underlying forms and phonological rules used in many previous linguistic studies of OE are no longer descriptively adequate and that such concepts as surface transparency and analogy will have to play a far more important role in f u t u r e analyses.

135

APPENDIX

A survey of some of the non West-Saxon verb forms of weak class III ( a f t e r Brunner and Campbell)

136 rH

fO 3

4J •rH

0)

fO XJ

Q CO H

0) D.

ω ω

ο 4-1 to

l|

co

ο

Λ

ω c 3m M

o 4-1 10

^ x:

4-1 fO

M-l

C tO XJ XJ (0

*O ^ XJ (H

rtJ "^

XJ

cn

-H 4-1 ft)

χ;

CU XJ XJ ft)

x;

0)

cn

^

^M ^O

* ft) **"·''

(0 (0 XJ XJ XJ XJ (0 (0 X! XJ

^

XJ (JJ

c

Ο Ό 4-1 ft) XJ

cn 0)

-H 4-1 (0

W

δ x:».. u indis

tj j

x: x:

XJ

co

10 4-1

1

c

QJ X} XJ XJ (0 ft) x; X!

S

-*D

1) y , (0

x: x: ^ft)

0) 4-1 ft)

χ:

Λ

«Ό tO X5 Λ (0

x;

Q) XJ XJ ft) XJ

(0 x;

-»Ό 0) 4-1

ft)ft) xJ

CO

ctO

XJ XJ tO

o XJ

-H •

Φ c

to (0

4-» -~

(0 XJ XJ (0 X!

β·· •Η

υ J-|

φ

S

Λ Λ

m β

-Ρ

*Ό

β

Λ

~

id ίθ *4-ι Μ-Ι

rO Λ

Τ3 m

τ3 m

Λ

σ> •d 4-1

χ: χ:

χ:

χ;

χ:

W ^Ο

π3 nj

Ό

OJ

ίΰ

(D

τ— (Ν no

• &1 W

• T3

β

(0 Λ Λ (0 Χ!

•H to• Φ Μ ft

ί^

(0

0)

Λ (0

r4

CO

ft

4-> Μ

•Η

(0 ft

-Ρ •Η β

ft

e •Η

•

4J (0 ft

•Η

05

+J

-Η

nft

(0

4-1

φ

to

ft

138

ΙΌ -*οrd

rH

(β

·*Ο ΜΗ

3 -Ρ

(0 ·Η Ι Η

•rl

Λ

εid

ω·*ο

(Ο (d ΜΗ ΜΗ 0 0

χ; V-|

3 Q

ΜΗ (0 0 •Η υ>·Η

-—σ>·

» ΜΗ >—' • rH

•Η Μ-Ι •Η

rH rH

rH

Μ

Η

·ιΗ rH

•Η

•Η Η ΜΗ •Η ι-Η

ν.

^Ό Φ Ι

Ο

υ φ

-»Ό

c

"*Ό

id

en

φ

(Λ -Η φ ΜΗ Ι -Η

(0

(0

^

>0

*οΦΟ* m ΜΗ ΜΗ .,-)

α ω

τ) c

(U-

Φ £ Ό Φ C •Η Φ CP

•Η ι-Η

•Η ·Η

ω Φ

Μ-Ι CO •ιΗ

Φ

ΜΗ

Φ

Cn

Ο) * -*0 ΜΗ w •Η

CD φ

4-> W

ft

ft

-H

ft

·

-H ΜΗ

·Η

-Η

ε

•

!H (d ft

n

h

•Η

φ Τ3

β DI β φ ·—· φ tP Ή CP

4 - 1 4 - 1

4-| ΜΗ J3

•Η rH

-Η rH

-Η -Η ·Η rH rH rH

CN CO Φ

α 10 0 Ο Χ!

.ρ SH Ο [«s Χ!

φ p, p,

CO

rH

Ό td MH tp

φ Ό 4H -H

4H -H

rH rH

rH

rH

O 4H Φ ·Η

rj

PS

α; tn

Μ Φ 1—Ι

ΐΰ co

^^&

(Χ

MH

s

r^1

f

m φ

•Η LQ (0

•t)OJ

α ω

MH

oφ

0)

>

β

(0 -Η

3 Cn Μ ΜΗ

ϋ

φ s

φ

" *ΌΌ (0 (0 Φ ΜΗ DI Cn Ο ΜΗ ΜΗ

•Η

rH

·Η ·Η

·Η

rH rH

rH

r- OJ ΓΠ

• t7>

» O ^ r H rH W eX

•

·

&» co

•r~>

Π3 Λ Λ

•Η Η

·

rHcoexcocx ex

·

Λ

•Η

W

W

0)

-p

JJ

cn

^

Μ

(β

(0

cx

rH

rH

rH

-Η

•Η

« . . . DlrH tT»rH

Ό

Φ

ΜΗ

.,Η

«— fNJ ΓΊ

« e x β

β 03 οι ΜΗ

Ό c Φ en 4Η

3

Φ

ex

·

-η

Τ3

Λ

•Η

03

co

ex

CN CN

φ

·

>

-Ρ

-Ρ

.p

r

(0

ex

-Η

ΜΗ

03

6

β

>Η

-Ρ CO (Ο

·Η

3

ex

co

-Η .

-Η

>Η

(Χ

φ

-π ex

ex

ex

140

H rd

•P

•H

φ

φ >O •+0 υ

Φ T3 cn

(U (U (U ig (U ( (U ΓΟ CQ CQ CO CO

rH cn ai di Cn ·η Λ

c

-H

W

Cn

cn• Φ

rd

-)

a>

Ό Cn ( cn

ίο

cn

CO

ird

t— cn

•P

a

o
-ι

C

rd

CO (0 ft

CN CN

• φ +J ^ ^ -H rd CQ

-P

•H β •H ft β -H

m (5 •H

ft •

•

4-1 X_|

rd ft

cn 4J φ >_ι cn rd ft ft

141

Λ

cn

-Η -Ρ

(D

Cn Φ cn

C

φ

* cn (L) φ i d u cn Φ ο cn Ο ushwor

-Ρ

ω

—.

g1 g W in

Λ o3 Cn Φ

2

&

"""-P 0 Φ cn-Ό o3 Cn Φ Φ cn Cn i

™ w

«

fll

Π) »Ό Η) ίο ΠΙ ίο Π) ίο

Φ Cn U

Φ 13 i

cn

cn

Π) ίο

cn cn cn cn

Λ π3 "Cn cn

ί) Oj

Cn

Π)3 0Π)3 Π

f

cn

cn cn

ω

-Ρ 03

03

φ

•Η

cn

C G

03

α

cn Φ ρ> .. C 03

-Η

υ

Μ Φ S

Q)

Cn

Oi Φ

O

cn

Ι

- cn ·*ο

C

O

j3 ο3·Ό 03 Φ Φ Φ Οι 0* Φ ο ι Ο ΐ φ ι ' ΰ Uo3Cn i

r-ππ rH

cn

W

T)

•n Λ

•H

cn

C

dl

. ·

cn

>

·

.

.

lrH

Φ

Φ £>

-Ρ (-ι

-P

-P

di

(0

C

M

-H

di

Π3

cn

T~>

C

0

(N (N

S

·Π

cn -p cn Π3

β

di

-H

J-l

M

cn 0)

di

di

di

03

.

> rH

iHcndicndi di cn

·

•H -P co

cn Φ

φ cn

^rsm

c n d i

G 03 D U Φ

φ 01 Ό U Cn

φφ cn cn

>rH Cn

C

π3 ΦΟ^ CnU

φ ω φ φ φ φ φ cni 03 cn cn cn cn cn

Φ Ό

-Η

Λ

· di

03

.

-H dl C -H cn -P Ή Φ cn dl

142

H

(d

•H

ε(β

ω

"Ό f D^ 0

Χ! M 3

α

x;

G

0 *"0 1 " O

ο

^O f ΟΊ O

x;

0

χ: •Η

x;

Λ

CO

Q)

tn

o

Cn 0 M

ω C C

m-*O •^ (d cn Cn id

Cn

αtn

rd

U Φ

O—

c In ι

*Ό υ ι Φ

>1

Cnxl

C

XJ Ο Π3 >-ι m Di Ο 0 «W Χ!

Ό Cn Ο Χ!

ο Μ·*Ό

•H

Ό C -H

CN

(d

fd

χ;

0)

ro

c φ

DI Di 0 υ >ι Χ! Xi

tn Φ

α XJ -P 1-1 0

u

φ

J^

c

PH

(d Di

cn 3

φ c φ

T) ·· r)

^1 Xi

*0 rd

-P ^ O

O X!

X!

«— CN ro Cn

tn

C rd tjl ϊ>1

X!

U

^

ι—ι

ft

,— CN m • * tn ft · · Cn i-H

en

13 C -H

•n Λ 3 W

t

·

tn d) s-i ft

in d) ft

^

CO

DI i-H

Htnfttoft ft

CO

•

Ό C

· ·η Λ 3

CN CN 0) -P • > -Ρ ·Η (d Μ -P ft

-P ω

jj w

. ft

. -H ω m φ }_|

·Η

-Η

•H

ro

ft

)

ft

(0

g

-H g

ft

C

-Ρ CO (0 ft

ft

143

Λ Φ Cn

Λ tfl

ω

•Η -Ρ C

ω

(Λ Η

ω α en Ο Ο

0) -Ρ W

Λ -Ρ Μ Ο

ω

4J -Ρ (Ο

Λ en

-ρ ο

C

(Ο

tn

PL,

-Η (Λ

Λ

τ- CN 00

PL,

U) DJ (Α

•n Λ

Λ 3 υ)

en 0) n

W

θ) H

w

tn ft

•H -P •H

C •H

(

(A

-P iH

-P (0

Οι

Q)

•H

04 α

E

144

BIBLIOGRAPHY

Anderson, James M. 1973. Structural aspects of language change. London: Longmann. Anderson, John ., and Charles Jones. 1 9 7 7 . Phonological structure and the history of English. Amsterdam: North Holland. Anderson, Stephen R. 1 9 7 4 . The organization of phonology. New York: Academic Press. Barrack, Charles Michael. 1975. A diachronic phonology from Proto-Germanic to Old English stressing West Saxon conditions. The Hague: Mouton. Bennett, William H. 1 9 6 2 . The parent s u f f i x in Germanic weak verbs of class III. Language 38, 135-141. Bosworth, J . , and T. N. Toller. 1898. An Anglo-Saxon dictionary. Oxford: Oxford University Press. Braune, Wilhelm. 1 9 7 5 . Althochdeutsche Grammatik. 13. Auflage, bearbeitet von Hans Eggers. Tübingen: Max Niemeyer. Brunner, K a r l . 1 9 6 5 . Altenglische Grammatik. Max Niemeyer.

Tübingen:

Bülbring, Karl D. 1902. Altenglisches Elementarbuch. 1. Teil: Lautlehre. Heidelberg Bynon, Theodora. 1 9 7 7 . Historical linguistics. Cambridge: Cambridge University Press. Campbell, A. 1959. Old English Grammar. University Press.

Oxford: Oxford

Cassidy, Frederic F . , and Richard N. Ringler, eds. 1 9 7 1 . B r i g h t ' s Old English grammar and reader. Third edition. New York, etc.: Holt, Rinehart and Winston Chomsky, Noam. 1965. Aspects of the theory of Cambridge, Mass.: The M . I . T . Press. .

syntax.

, and Morris Halle. 1968. The sound pattern of English. New York: Harper & Row.

145

Christie, W. M . , ed. 1 9 7 6 . Current progress in historical linguistics. Amsterdam: North Holland. Clark Hall, John R. 1 9 7 0 . A concise Anglo-Saxon dictionary. Cambridge: Cambridge University Press. Cowgill, Warren. o-presents.

1959. The inflection of the Germanic Language 35, 1-15.

Dingwall, William Orr, ed. 1 9 7 1 . A survey of linguistic science. College Park, M d . : University of Maryland Press. Dresher, B. Elan. 1978. Review of Lass/Anderson ( 1 9 7 5 ) . Language 54, 4 3 2 - 4 4 6 . Erickson, Jon L. 1970. A reexamination of Old English verbal base forms. Papers in Descriptive, Historical and Applied English Linguistics. Madison: Department of English, The University of Wisconsin. Flasdieck, Hermann M. 1935. Untersuchungen über die germanischen Verben III. Klasse (unter besonderer Berücksichtigung des Altenglischen). Anglia 59, 1-192. Harms, Robert T. 1968. Introduction to phonological theory. Englewood C l i f f s : Prentice-Hall. Hass, Wilbur A. 1969. different types of the Fifth Regional Chicago Linguistic

The psychological reality of phonological change. Papers from Meeting, 343-347. Chicago: Society.

Hogg, Richard M. 1976. The status of rule reordering. Journal of Linguistics 12, 103-124. Holthausen, F. 1 9 7 4 . Altenglisches etymologisches Wörterbuch. Heidelberg: Carl Winter. Hooper, Joan B. 1976. An introduction to natural generative phonology. New York: Academic Press. Howren, R. 1967. The generation of Old English weak verbs. Language 43, 674-685. Hymann, Larry M. 1975. Phonology: theory and analysis. New York: Holt, Rinehart & Winston. Keyser, Samuel Jay. 1 9 7 5 . Metathesis and Old English phonology. Linguistic Inquiry 6, 3 7 7 - 4 1 1 . King, Robert D. 1 9 6 9 . Historical linguistics and generative grammar. Englewood C l i f f s : Prentice-Hall.

146

1973.

Rule insertion.

Language 49, 551-578.

Kiparsky, Paul. 1968a. Linguistic universals and linguistic change. In Universals in linguistic theory, Emmon Bach and Robert R. Harms, eds. New York: Holt, Rinehart & Winston. 1968b. How abstract is phonology? Reproduced by the Indiana University Linguistics Club. . 1971. Historical linguistics. ( 1 9 7 1 ) , 577-649. . 1 9 7 2 . Explanation in phonology. ( 1 9 7 2 ) , 189-227.

In Dingwall In Peters

, and Wayne Ο ' N e i l . 1976. The phonology of Old English inflections. Linguistic Inquiry 7, 527-557. Krahe, Hans and Wolfgang Meid. 1 9 6 9 . Germanische Sprachwissenschaft. 3 vols. Berlin: Walter de Gruyter & Co. Laferriere, Martha. 1972. Analogical change: Problems form Germanic dialects. Unpublished doctoral dissertation, Brown University. Lass, Roger. 1970. Palatals and umlaut in Old English. Acta Linguistica Hafniensis 13, 75-98. . 1971. Boundaries as obstruents: Old English voicing assimilation and universal strength hierarchies. Journal of Linguistics 7, 15-30. 1976. English phonology and phonological theory: Synchronic and diachronic studies. Cambridge: Cambridge University Press. , and John M. Anderson. 1975. Old English phonology. Cambridge: Cambridge University Press. Luick, Karl. 1 9 6 4 . Historische Grammatik der Englischen Sprache. Reprint. Oxford: Basil Blackwell. Mosse, Fernand. 1968. A handbook of Middle English. Translated by James A. Walker. Baltimore: The Johns Hopkins University Press. Nessly, Larry. 1 9 7 3 . The weakening chain in natural phonology. Papers from the Ninth Regional Meeting, 4 6 2 - 4 7 4 . Chicago: Chicago Linguistic Society. Paul, Hermann. 192O. Prinzipien der Sprachgeschichte. Halle: Max Niemever.

147 Peinovich, Michael Peter. 1972. A re-examination of morphological change in Old and Middle English.. Unpublished doctoral dissertation, University of Wisconsin-Madison. Penzl, Herbert. 1 9 4 7 . The phonemic split of Germanic k in Old English. Language 23, 34-42. Reprinted in Scott/Erickson ( 1 9 6 8 ) 173-183. Peters, Stanley, ed. 1 9 7 2 . Goals of linguistic theory. Englewood C l i f f s : Prentice-Hall. Pilch, Herbert. 1 9 7 0 . Max Hueber.

Altenglische Grammatik.

München:

Postal, Paul. 1968. Aspects of phonological theory. New York: Harper and Row. Prokosch, Eduard. 1938. A comparative Germanic grammar. Baltimore: Linguistic Society of America. 2 Quirk, Randolph, and C. L. Wrenn. 1957 . An Old English grammar. London: Methuen. Rhodes, Richard A. 1972. Natural phonology and MS conditions. Papers from the Eighth Regional Meeting, 544557. Chicaco: Chicago Linguistic Society. 1973. Some implications of natural phonology, Papers from the Ninth Regional Meeting, 530-541. Chicago: Chicago Linguistic Society. Schane, Sanford A. 1973. Generative phonology. wood C l i f f s : Prentice-Hall.

Engle-

Scott, Charles T . , and Jon L. Erickson. 1968. Readings for the history of the English language. Boston: Allyn and Bacon. Stampe, David. 1973. A dissertation on natural phonology, Unpublished doctoral dissertation, University of Chicago. Stockwell, Robert P. and Ronald K . S. Macaulay. Linguistic change and generative theory. Bloomington: Indiana University Press.

1972.

Vennemann, Theo. 1 9 7 4 . Phonological concreteness in natural generative grammar. In Toward tomorrow's linguistics, R. Shuy and C. J. Bailey, eds. Washington: Georgetown University Press. Voyles, Joseph B. 1 9 7 4 . West Germanic inflection, derivation and compounding. The Hague: Mouton.

148

Wagner, Karl Heinz. 1969. Generative grammatical studies in the Old English language. Heidelberg: Groos. Wright, Joseph. 1910. Grammar of the Gothic language. Oxford: Oxford University Press. , and Elizabeth M. Wright. grammar. 3rd edition. Oxford: Press.

1925. Old English Oxford University