Gothic, Germanic, and Northwest Germanic 3515035737, 9783515035736

Table of contents :
CHAPTER I: Introductory Remarks 1
CHAPTER II: Gothic 4
CHAPTER III: Germanic and Northwest Germanic 104
References 154

Citation preview

ZEITSCHRIFT FÜR DIALEKTOLOGIE UND LINGUISTIK BEIHEFTE

HERAUSGEGEBEN VON JOACHIM GÖSCHEL • WERNER H. VEITH

HEFT 39

FRANZ STEINER VERLAG GMBH WIESBADEN 1981

GOTHIC, GERMANIC, AND NORTHWEST GERMANIC BY JOSEPH B. VOyLES

FRANZ STEINER VERLAG GMBH • WIESBADEN 1981

Eine Veröffentlichung in Verbindung mit dem Forschungsinstitut für deutsche Sprache „Deutscher Sprachatlas“ der Philipps-Universität Marburg/Lahn

TP-Kurztitelaufnähme der Deutschen Bibliotek 'oyles, Joseph B.: Gothic, Germanic, and Northwest Germanic / by Joseph B. Voyles. - Wiesbaden : Steiner, 1981. (Zeitschrift für Dialektologie und Linguistik : Beih.; H. 39) ISBN 3-515-03573-7 ÏE: Zeitschrift für Dialektologie und Linguistik / Beihefte Jle Rechte Vorbehalten >hne ausdrückliche Genehmigung des Verlages ist es auch nicht gestattet, das Werk oder einzelne Teile axaus nachzudrucken oder auf photomechanischem Wege (Photokopie, Mikrokopie usw.) zu vervieliltigen. © 1981 by Franz Steiner Verlag GmbH, Wiesbaden, rinted in Germany

DEDICATED TO THE MEMORY OF MY FRIEND EDUARD WESTERICH (JUNE 26, 1900 - APHIL 6, 1974)

C o n t e n t s CHAPTER I * Introductory Remarks .... l Frequently Used Abbreviations

3

CHAPTER H i

4

1. 2

•

Gothic .................... ........... .

The Form of the Description

4

The Phonemes of Gothic ......

6

3.

Derivation and Compounding

13

3.1#

Derivational Prefixes

.........

13

.............

14

3.2

• Derivational Suffixes

3-3.

Compounding ..............

16

4.

Inflectional Morphology ..........

18

4.1.

The Sb-rules ....................

19

4.2.

The Pn-rules ..............

24

4.3.

The Vb-rules .....................

26

4.4.

The De-rules ....................

29

5.

Phonological Rules ............

30

5*1*

The Minor Rules ••••......

31

1.1.

31

Vocalic ..................................

1.1.1. Ablaut alternations (within inflectional paradigms) ......................

31

1 .1 .2 . Deletion from /anot/ (Pn3 morpheme) ••».. 36 1.1.3* Sievers* Law ........ 1.1.4.

36

/i/ (Vbl? morpheme)-deletion .............39

1.1.5« /i/ (Sbl4 morpheme)-deletion ..... .

40

1 .1 .6 . /ii/ (Vbl9 morpheme)-shortening

40

1.1.7. / 0 «/ (Sb24 and Sb26 morphemes)-to-/a/ rule ......

41

Contents

VlU

1.1.8. / 01 / (Sb26 morpheme)-deletion ............ ^1

1 .1 .9 . /u/ (Sb2? morpheme)-to-/i/ rule .......... *42 1.1.10. Word-stress rule

........................

*42

1.2.

Consonantal

............. ............. *45

1.2.1.

Consonantal change in A and V stems ....

1.2.2.

Reduplication ............................. *47

1.2 .3 .

Thurneysen#s Law ......................... 50

1.2.*4.

Verner's Law (within inflectional parad igms ) • • • ....................... . • • • • 5*4

1.2.5*

/n/-insertion in St V .................... 55

1.2.6.

/s/-insertion in V ............

1.2.7.

V suffixes Vbl2 /d/ and derivative suffix 52 /t/ realized as / s / ........... 55

1 .2 .8 . /d/ (Vbl2 and Vbl3 morphemes) realized 1.2 .9 .

*4*5

55

as / t f þ/ ................................

56

/z/ (Sb31 morpheme)-deletion ............

57

1 .2 .1 0 . ##-deletion ............

56

5*2. The Major Rules ....................................56

2 .1 .

Vocalic .......................

2.1.1.

Like-vowel contraction .................... 58

2.1.2.

Vowel d e l e t i o n .... ••••...... ........... 59

2 .1.3 .

Breakings / i f u/-/e, 0/ alternation ..... 6 l

2 .1 .*4.

/i, u, j, w/-alternation ........... ••••• 6^

2 .1.5 .

Monophthongization of /ai, au/ to / E i , 01 / .....

2.1.6.

Vowel lowering ........................... 67

2.1.7«

Glide insertion ••••;•••................... 68

56

66

2 .1 .8 . /u/-t 0-/ 0/ rule ...............

68

2 .1.9 . /it/-/ei/ alternations ..................

69

Con t e n t s

IX

2.2.

Consonantal ....................... ....

69

2.2.1.

Consonantal devoicing .........

69

2.2.2.

Distribution of voiced stop and continuant consonants

71

Geminate-consonant simplification .....

72

2.2.3« 2.2.

^.

Consonant-cluster simplification ......73

2.2.5«

/h/-assimilation ..... ....

73

2 .2 .6 .

/h/-deletion ... ............

73

2.2*7«

Nasal assimilation

7k

2.2.8.

Nasal deletion and vowel lengthening

2.2.9«

/n/-deletion ... .........

..

7k 75

2 .2 . 1 0 . /z/-to-/r/ rule .....

75

Appendix It Paradigms .............. .......

76

1• Nouns (Sb-rules)

76

2 . Adjectives (Sb-rules) ...........

81

3 « Pronouns (Pn- and Sb-rules) ................ 8^ 4.

Verbs (Vb-rules)

................ .

Appendix 2 * Rule Ordering ..... .

88

9k

Footnotes to Chapter 2 ................. .

102

CHAPTER IIIi Germanic and Northwest Germanic ......

10k

1 •

Introduction............................. .... 104

2.

The Problems of Umlaut and/ei2/ .............. 106

2.1. Preliminaries .................

106

2 .2 . The Problem of Umlaut ...................

107

2.3. The Remaining G m c . Vowels

11^

................

2.k. The Problem of /ei^/ and /eig/ ................lié

Contents

X

3.

The Reduplicating Verbs in Northwest Germanic »..... ........... ..................... 120

3.1.

The Problem ............... •.......... ....... » 120

3 .2 .

A Solution ...••.... ...... .................... 123

3'2*1. The Gothic S i t u a t i o n ................. .

123

3.2.2. The Germanic and Northwest Germanic Situation .......... ••••••••••.................124 3.2.3. An Account of the Developments ...... .

129

3.3«

143

Concluding

Remarks ....................

Appendix 1: The Reduplicating Verbs in Germanic ....................

145

Appendix 2: Attested Changes Analogous to those Posited in 3*2.3............. 146 Footnotes to Chapter 3 ..........................148 References ..........

154

Chapter I Introductory Remarks In the next chapter we give a detailed analysis of the greater part of the derivational morphology, the inflectional morphology, and the phonology of Wulfilian Gothic (as reproduced in Streitberg, 1950). In spite of the number of good traditional handbooks on the subject and the frequency of the more recent considerations of isolated phenomena of Go. phonology, a fairly exhaustive and detailed account of this corpus still seems necessary. This is so because much of the work on Gothic, dealing as it does with one or only a few rules considered in iso­ lation, is certainly incomplete and quite often erroneous. This is of course to be expected since morphological and phonological rules are related in various respects and interact in so many ways that the formulation given to any one of them usually affects the form of numerous others. Accordingly, we have found that almost every one of our Go. rules in chapter 2 - when considered in the light of the other rules of the language - has taken a form not congruent with any of those previously proposed in the literature. One example out of many is our version of a rule formulated early in the history of Go. philology and commonly referred to as Thurneysen#s Law. (This is phono­ logical rule 1.2.3 in section 5 of chapter 2). We argue in chapter 2 that this rule has been misstated since its initial discovery and formulation by Thurneysen (1898)« Another such case is Sievers* Law (phonological rule 1.1.3 in section 5 of chapter 2). Section 3 in chapter 2 on derivation and compounding and section 4 on inflectional morphology are perhaps best considered for reference instead of for reading through from beginning to end. (A more synoptic view of the mor­ phological rules may be obtained by consulting the para­ digms given in appendix 1 of chapter 2.) Whether or not our views on the nature and function of morphological rules are correct remains to be seen. We formulate them explicitly so that the reader can at least know what we have considered to be morphological as opposed to phono­ logical phenomena, and can evaluate our phonological rules in section 5 of chapter 2 accordingly. The principal em­ phasis of the chapter is on these rules of Go. phonology. Chapter 3 is an attempt to formulate a coherent

2

Introductory

Remarks

account - based in large part on the material (in partic­ ular section 5) of chapter 2 - of three of the salient problems of historical G m c . phonology. These are the question of umlaut in Germanic, that of the so-called /e:-j/ and /e:2/i and finally that of the reduplicating verbs in Gothic and in the NWGmc. languages. In accordance with M arc h a n d é dictum (lg?0il22), ••Das Gotische bleibt nach wie vor der Grundstein für die Erforschung der germanischen Sprachen.", we believe that some of the conclusions reached in chapter 2 can be brought to bear in the resolution of the questions posed in chapter 3. Although frequent reference is made in chapter 3 to the content of chapter 2, chapter 3 can nonetheless be read independently without one first having gone through chapter 2. Similarly, chapter 2 can be considered in and for .itself without reference to chapter 3. And of course there will be many who will not choose to read either chapter 2 or chapter 3. Finally, in the appendix to the present chapter we give a list of the abbreviations frequently used throughout the book.

Frequently

Used A b b r e v i a t l o n s

Frequently Used Abbreviations A Ac Act Art Be-rules Def Dt Du ev

Gmc. Gn Go * IE Imp Ind Int MHG Mn MS N Nm NWGmc• OE OF OHG ON OS Pers PI Pn Pn-rules

Pp Prep Prs Prt Pst Psv

adjective accusative active article morphological rules applying to the verb *bef definite dative dual eventually, i.e. through the application of rules unspecified in a derivation Germanic genitive Gothic Indo-European imperative indefinite pronoun, also indicative interrogative pronoun Middle High German masculine morpheme structure noun nominative Northwest Germanic Old English Old Frisian Old High German Old Norse Old Saxon person plural pronoun morphological rules applying to pronouns preterite-present preposition present participle past passive

Ref Sb-rules

sg SPC St Sub V Vb-rules WG Wk 1» 2, 3 /.../ //...//

reflexive pronoun morphological rules applying to adjectives and nouns singular surface phonetic constraint strong subjunctive verb morphological rules applying to verbs West Germanic weak first, second, third person phonemic transcription phonetic transcription

3

Chapter II Gothic 1. The Form of the Description This chapter contains a portion of the grammar of Gothic - the major part of its morphology and phonology. Before considering the form in which we have cast these rules, we shall first consider briefly how such rules might fit into the grammar as a whole. .Put in very general terms, we consider a grammar to be a set of rules or principles on the basis of which the grammatical utterances or sentences of a language are pro­ duced. Such rules are of three basic typesi 1. Rules of remote or deep structure which generate the basic sentence structures. 2. Transformational rules which map or modify the basic structures into derived structures. 3* Interpre­ tive rules which take the output of 1 and 2 and assign to this output a phonological configuration. Here we shall be concerned with rules of type 3# the interpretive rules of Go. morphology and phonology. To give a concrete example, let us consider the deri­ vation of the following Go. sentence (Luke 2i7)# jah gabar sunu seinana þana frumabaur jah biwand ina jah galagida Ina irT"uzetin7 **And she bore her son the f irs t -born and wrapped him and laid him in a manger." Let us assume that the remote structure of this sentence might be something like this I A. £

S1

AND T MARY. BORE MARY'S. SON T MARY'S SON WAS THE

S2

S3

FIRST-BORN J

J

S3 S2 H

S5

AND Q MARY^ WRAPPED MARY’S^ SON 1

Sk

AND

%

MARY. LAID MARY’S. SON IN A MANGER H

1

S5 S1

After the application to structure A of the lexical inser­ tion rules and transformations like anaphoric pronominalization, relative-clause reduction, conjunction reduction, case marking, agreement, and the' like, the resultant structure would be something like thisi

5 5

The F o r m of the De s c ri p t i o n

B.

^

"(a) AND

1 ~(a) BORE

(b) Conj

~|

(b) V, Pst, St ^th class, 3rd Pers, Sg, Ind

_(c) /jah/ J I (c) /ga-bir/ 1

2

"(a) SON

“ T(a) HER

(b) N, Sg, u-class, Ac, Sg, Mn

_(c) /sun/

(b) Possessive, Pn, 3rd Pers, Ac, Sg, Mn

J |_(c) /sitn/

3

^

'(a) THE

t r i a ) FIRST-BORN

(b) Def Art, Ac, Sg, Mn

IRa)

(b) N, Compound, Ac, Sg, Mn

AND

(b) Conj L(c) / jah/ _

(c) /frum/, /bur/

|_(c) /þ/

J 1-

5

6

?

~(a) WRAPPED

(a) HIM

(b) V, Pst, St, 3rd Class 3rd Pers, Sg, Ind

(b) Pn, 3rd Pers, Ac, Sg, Mn

_(c) /bi-wind/

_ _(c) /i/

8 ~(a)

AND

(c) /jah/

9

l [ ( a ) LAID

(b) Conj

J

(b) Prep

" |"(a) HIM

(b) V, Pst, Wk, j-Class, 3rd Pers, Sg, Ind 1(c)

/lag/

10 "(a) IN

~j

(b) Pn, 3rd Pers, Ac, Sg, Mn

J 11

l_(c) /1/ 12

J

J

f(a) MANGER (b) N, Sg, n-Class, Ac, Sg, Mn

S1

_(c) /in/ J |_(c) /uzest/ 13

14

The structure under B consists of a sequence of four­ teen morphemes represented in fourteen matrices specified

Gothic

6

as to their (a) semantic, (b) morphosyntactic, and (c) phonological features. The morphological and phonological rules of Gothic given in sections 4 and 5 of this chapter apply to structures like B. The morphological rules applying in this particular instance are thesei Rule Sb27 inserts /u/ after matrix 3 (/sun/) since it is a u-class noun; Sb8 inserts /ana/ after matrix 4 (/si in/) since it is marked as a Pn, 3rd Pers, Ac, Sg, Mn; the same rule also inserts /ana/ after matrix 5 (the definite article /þ/) and after matrices 9 and 12 (the Ac Sg Mn of the 3rd Pers Pn /i/) • To the weak verb /lag/ (matrix 11), three Vb-rules of morphology applyi Vbl7 which inserts /i/ after /lag/ because it is marked as a j-class weak verb, Vbl2 which inserts /d/ after /lag/ + /i/ since /lag/ is marked as a weak verb in the past, and rule Vb3 which inserts /a/ after /lag/ -h /i/ + /d/ since /lag/ is marked as a weak verb, in the past, third person singular indicative. Finally, two rules of noun morphology apply to /uzeit/ since it is marked as an n-class Mn Sg Ac noun* Sbl2 which inserts /i/ after /uzeit/ and Sb20 which inserts /n/ after /uzest/ + /i/. After the application of these morpholog­ ical rules (including the rule for compounds given under 3 «3 below), the resultant phonological structure — dis­ regarding the presence of the word boundaries ## — is this * C. /jah ga-bir sun-u s i m - a n a þ-ana frum-a-bur jah bi-wind

1

2

3

4

5

6

7

8

i-ana jah lag-i-d-a i-ana in uze«t-i-n/ 9 10 11 12 13 14 The segments in C are the fully specified systematic phonemes of Gothic, an enumeration of which is given under A in section 2 of this chapter. The phonological rules of Gothic apply to structures like C to produce the final phonetic output. The phonological rules applying to structure C are theses Rule 1.1.1, ablaut, applies to matrices 2 and 8 to produce /ga-bar/ and /bi-wand / 1 rule 1.1.10, the rule of word-stress assignment, applies to all the matrices; rule 2.1.2, vowel deletion, applies to matri­ ces 9 and 12; rule 2.1.3, breaking, applies to matrix 6; and rule 2.2.2 for the distribution of voiced stop vs. continuant consonants applies to matrices 2, 6, 8, and 11. This results in the final phonetic outputs yJ,' ' ' / ✓ / \ / / / D. //jah ^abar sunu sisnana þana frumabor jah biwand ina ✓ / s / / \ jah lapida ina in uzeitin// where ' denotes primary and s secondary stress.

The P h o n e m e s

of

Gothic

2. The Phonemes of Gothic A. consonant sonor ant vocalic coronal anterior high low back round nasal laterai continuant tense voiced strident long stress

f þ h h w p t k k w b d g s z l m n r + +++ ++ ++ + + +++ + 4 +4 ----------------------------f + ^ + ----------------------------------— f---- + ---— h + + + — i- -f + + ---- + ------+ + - + -I- + + + + -----+ ------+ --------------------- h - H ----------------------------+ + ----+ -+-----+ ----------------+ ------- h ----------------------- --------É c ­ ---------------------------- f-----+ + + + ------- -------- h + + + + + + + -h-t + + + + ------ --------------------------+ + + + + + + + + + ----- ----------------- f- + --------------------------------------------------------------------------j wa

consonant sonorant vocalic coronal anterior high low back round nasal lateral continuant tense voiced strident long stress

e i ouar

e: i: o: u: E: 0 :

--------------- — - + ++ + + +-C + + + + ---+ 4 1 - + + + + + + + ------------------ - - --------------- --- - - + + -- + -+ - - + - + ---- h -------- b - - - — + + — — + ■+ + — — + -+ — h ---- *t + - - - -+ + --------------- — - --------------- -- - - + + + + + + + + + + + + ------------- + ■+ + ■+ -+ + + + + + + + + + + -f + --------------- — - — — — — — — — + + -+ -+ -t --- Integral values assigned segments by rule 1.1.10.

- + + + + - - - — + - + — — - + + - + + — -+ + to vocalic

The phonetic segments are those under A with the addition of the following: B. l. //tScffg// like /b, d, g/ under A except that they are +continuant by phonological rule 2.2.2 below. 2 , //}// like /n/ under A except that it is +high and +back by phonological rule 2.2.7 below. 3 . Possibly, //x// like /g/ under A except that it is -t-high, -hcontinuant, -ftense, and -voiced through

Gothic

8

phonological rule 2.2.1. 4. Possibly» //?// like /o/ under A except that it is 4tense by phonological rule 2.1.8. The phonotactics, in this case the so-called ••surface phonetic constraints** (hereafter SPC) which are the admissable phonological sequences after the application of the phonological rules, are in C below. The admissable morpheme-initial consonantal sequences are given under C.l.-j the admissable word-final consonantal sequences under C.2.— .1 C.l.l. In ##ClV, i.e. in any morpheme (or word) beginning with only one -vocalic segment (C1 ), C1 may be any -vocalic segment except /z/.2 1.2. In ##C2V, morphemes beginning with two —vocalic segments, the following conditions prevail. C2 = f 4- (1# r); þ + (1, r, w); h + (1, r, n)*

1

2

1

2

1

2

p 4 (1. r)i t 4 (r, w) i k 4 (1, r, n) ; b 4 CLr,n)i

1

2

1

2

1

2

1

2

d 4 (r, w)j g 4 (1, r); s 4 (p# t, k, 1, m, n, w)> 1 2 1 2 1 2 or w 4 (If r ) . 1 2 That is, a sequence may begin with /f/ followed by either /1/ or /r/ as in flodus •flood* and fragan •ask*, etc. There is one possible instance or a sequence /kwr/ in qrammiba #dampness#, which may however be an orthographic error. 1.3. In ##c3v, c3 = s + (p, t, k) +■ r. 1 2 3 2.1. In VC1##» i.e. in any word ending in only one -vocalic segment, C1 may be any —vocalic segment except /j/. 2.2. In VC2##, C2 = f -f (t, 1, n)i þ + (þ, 1, r, m, n,w)i

1

2

1

2

h + (t, s, 1, r, m, n)j t + (t, s, 1, r, w)j k + 1 2 1 2 1 (k, 1, r, n) 5 b + (1, r, n); d -I- (d, r, w) t g 4-

2

1

2

1

2

1

(d, 1, r, m, n, w); s + (t, k, kw , s, g, m, n);

2

1

2

z + (d, g, n, w)» 1 + (f, þ, h, p, t, k, b, d, g,

1

2

1

2

s, z, 1, m, w)» r + (f, p, h, hw , p, t, k, b, d, g, 1 2 s, z, r)i m + (f, p, b, s, z, 1, r, m, n)j n + , 1 2 1 (þ, k, kw , d, g, s, n, w ) .

The P h o n e m e s

of

Gothic

9

There is no word- or morpheme-final sequence *VpCl##. 2 .3 . In VC3##, C3 = f + t -f r, 3 -f t, i.e. /f/ followed

1

2

3

2

3

by /tr/ or /st/j h + t + (r, w), s + (t, n, w ) ;

1 g+g+w; 1

2

3

2

3

s-t-t-t-rj 1 + 1

2

3

1

2

s + k> r 4- (f* h, s) + t,

2

2

3

(þ»d) + r, h -t- m, 3

2

3

(k, z) + n, þ + r*

3

I 2 3 2 -I-(t,1), b f ri n + (t, 1 2 3 2 3 1 (1, w), s+ (t, 1), d + w.

m + a

3 2 3 d, g) + r, g +

2

3 2

3 2 3 2 3 , , There are no sequences in *VpC2##, *V(p, t, kJC^##, or *V(b, d, g)C2##. 2 A . In VC**##,O** = 1 + (f, s) + t t r ,

1 r-hs+t+w;

1

2

3

4

2

3

k

h

2

f s + n;

3

k

m + f + s + l .

1

2

3

^

In addition to the vocalic segments given under A above, we posit the underlying diphthongs /ai, au, iu/. If followed by a —vocalic segment or if unstressed, the first two are realized as //Es, 0:// by phonological rule 2 .I . 5 in section 5 below. Otherwise they are realized as //aj, aw// by 2.1 A. Hence the diphthongs //aj» aw// never occur unstressed. The diphthong /iu/ if stressed and followed by a vocalic segment is realized as //iw// by phonological rule 2.1.4. If stressed and followed by a —vocalic segment, it is //iu//. If unstressed and followed by a vocalic segment, it is //iw// by 2.1.^s and if un­ stressed and followed by a —vocalic segment, it is //ju// by the same rule. Hence the diphthongs occurring under stress are //aj, aw, iw, iu//, the only true diphthong in the sense of having both segments vocalic being //iu//. And in unstressed position no true diphthongs occur, only the diphthongoids //ju// and //iw//, (Examples of forms may be found under the rules mentioned,) Other distributional facts about the vowels are these« 1. According to rule 2.1,3> the short vowels /e, o/ cannot occur unstressed and after the principal stress in a w o r d . They can occur under main stress, and /e/ - as opposed to /o/ - can occur pretonically. (The reason for this is that Go. pretonic /e/ is a reflex of IE /e/, while IE /o/, which would presumably have also been retained in Gothic preton­ ically just as /e/ was, had already changed to /a/ in Germanic.) 2. Long /a«/ is produced only by rule 2.2,8. Thus there is some question as to whether it should be con­ sidered among the phonemes of the language. In any event, long /a:/ occurs only under main stress in native Go. words.

10

Gothic

The distribution of the Go. vocalic segments given in A above as well as the diphthongoids //aj, aw, iw, ju// and the diphthong //iu// may be summarized as followsi D.i. In C q VC q --, i.e. before the main stress in a word, the vocalic nucleus V may be any single vocalic segment except /o/ or /at/, but no diphthong-like sequences except possibly /iw/ or /ju/. 2. In V, i.e. under the main stress in a word, V may be any single vocalic segment and any diphthong-like sequence• 3 . In -'-C0VCo##, i.e. after the main stress in a word, V may be the short vowels /a, i, u/ and any long vowel except /ai/. The only diphthong-like sequences occur­ ring here are /iw/ and /ju/. Although we have included the reasons for our partic­ ular phonemicization mainly in our discussions of the indi­ vidual phonological rules, we shall here touch on a few general questions brought up in such sources as Antonsen (1972), Beade (1971, 1972), Beck (1973), Marchand (1970, 1973), Moulton (19^8, 195^, 1961, 1972), Vennemann (1971), and Wurzel (1975). There is comparatively little disa­ greement in these sources on the consonants. First, con­ cerning our analysis, we have posited /hw/ and /kw/ as unit phonemes instead of /h/ -f /w/ and /k/ -I- /w/ since they function as unit phonemes in the environment of pho­ nological rules like 1 .2. 2 (reduplication) and 1 .2 . 3 (Thurneysenfs Law). In this way /hw/ and /kw/ differ from /g/ + /w/, which do not function in any of our rules as a unit segment. Second, we have considered the stops /b, d, g/ as the underlying segments and then formulated rule 2 .2. 2 to realize them as continuants in specific environ­ ments instead of considering the underlying segments con­ tinuants /fe, brangda (rules 1.2.1, 1.2.8, and 2.2.8) //brashta//. (Ad) greipan, Pst PI gripum, Pst Prt gripans. (Ae) bindan 'tie* vs. Pst PI bundum, Pst Prt bundans; bliggwan, Pst PI bluggwum, Pst Prt bluggwans where this rule applies because the stem vowel / 1 / is followed by at least two nonvocalic segments g likewise intrusgjan *implant*, Pst PI intrusgum, Pst Prt intrusgans; niman *take *, Pst Prt numans because the basic stem vowel /u/ is followed by /m/. (A f ) Prs PI witum •know*, Prs Sg 1 w a i t ; the St V digan 'knead*, Pst Sg pre­ sumably daig. (B) letan /leitan/ ylet*, Pst Sg lailot /lelo:t/T (C) NmSg fon 'fire*, GnSg funins. (Da) NmSg sunus or sunaus. (Db) NmSg broþar 'brother', DtSg brobr. The ablaut rule applies to forms entered in the lexi­ con with their basic vowels specified. As mentioned above, the basic vowel of the St V is that of the Prsg and the basic vowel of the Pp V is that of the Prs PI, Prs Sub, and all Pst forms. Thus the basic vowel for any V is the one which occurs in the most forms within the paradigm. This fact, however, is not the only reason for considering this particular vowel as basic. Another argument for considering the vowel of the pre­ sent the basic one in St V is that in the history of Gothic, whenever St V have changed their classes as the result of some sort of phonological change, the resultant newly formed present-tense form has invariably been considered basic. The verb with its new Prs-tense vowel has then realigned itself and formed its other tenses according to the ablaut rule as we have formulated it above• For in­ stance, V #s like le ihxan 'lend*, peihan 'thrive*, preihan •push', and weihan 'fight' were all originally in Germanic

Th e

Minor

Rules

(Vocalic)

35

/linhwan/, /þinhan/, and the like with short /i/ followed by a nasal and another consonant in the stem* Accordingly, their paradigm was — with the application of ablaut Ac and Ae above and of Verner*s Law — this: Prs /linhwan/, Pst Sg /lanhw/, Pst PI /lungwum/, Pst Prt /lungwanaz/. With the addition of the Gmc. version of rule 2.2,8 (nasal deletion and vowel lengthening), this paradigm became /li:hwan, la:hw , lungwum, lungwanaz/. In Gothic the base form was invariably considered that of the Prs, /li:hw/, which then automatically underwent the ablaut alternations Ac and Ad: / l i :hw/ (Ac) /laihw/ (e v ) -^> / / l E : h V / laihi , and /li:hwum/ (Ad) — /lih^um/ (ev) — £> //lehwum// laihi urn, Pst Prt laihi a n s . Similarly, /þrinhan/ became Go. þreihan, braih, þraihum, praihans. It is significant that a Pst Sg form like / l a i h v was never considered the basic form for such V #s in Gothic. In that case, /la:hw/ and the like would automatically have become reduplicating St V: *lahian, *lailahi, »lailahium, »lahia n s . Similarly, in all the other Gmc. languages whenever a St V becomes Wk, the base form is invariably the Prs-tense form. Thus in English help, holp, holpen became help, helped and not *holp, »holped, The V*s diwan *die* and sniwan •hasten* present a prob­ lem. The attested paradigm of the latter is sniwan, snau, sniwum or snewum, sniwans. If the base form is considered /sniw/, alternations Ab and Ac apply resulting in sniwan, snau for /''snOt// from /snaw/ by rule 2.1,5* snewum for / s n e iwum/ by Ab, and sniwans. The additional Pst PI sniwum may mean that alternation Ab was optional for the V*s /diw/ and /sniw/. The V standan, stop, stopum, standans ‘stand* under­ goes alternation Aa. Sccordingly, the base form is /staþ/ which undergoes alternation Aa for the nonparticipial Pst forms and rules 1,2.4 (Verner*s Law) and 1,2,5 (/n/-inser­ tion) for the Prs and Pst Prt, As for the ablaut alternations under D, it is not ob­ vious which forms should be considered basic; it is thus quite possible that they should read 0 : — £> u and a -~2> 0' These are to be considered ablaut alternations because they are not phonologically conditioned and occur only in the given morphosyntactic environments. Hence while both sunus and sunaus *son* can occur, an alternation like nemup and ♦nemaup *you all took* does not.

We have described only the intraparadigmatic ablaut alternations and not those occurring in derivation such as draibjan *to make go* from dreiban *drive* wegs *movement* from wigan *to shake*, or raups *red* and ga-riups •honor­ able*. Most such alternations would be descnbable through the inclusion of additional environments to the ablaut alternations already given. The only one not given in the ablaut rule above is o: —> 0 in fidur-falps ‘four-fold*

36

Gothic

from fidwor •four*. Finally» we emphasize that we have been able to state all the ablaut alternations applying to St V stems in terms of the MS of the stem and the morphosyntactic features of tense (Prs, Pst) and P r t . We have not had recourse to additional morphosyntactic classification such as lst-class St V, 2nd-class St V, and the like. And those St V which do not meet the MS constraints of the ablaut rule under A automatically undergo the rule of reduplication (1.2.2). 1.1.2.

Deletion from /anoi/ (Pn3 morpheme)

a n o: -£> 0 n oi, and possibly optionally 0 0 o: 12

3

1 2 3

1 2 3

/ *~kc Sg Mn of /E:n/ *no one*! ----E.g., Ac Sg Mn / E sn-ano:-hun/ ainnohun, possibly (if not an orthographic error) ainohun. Otherwise this morpheme is /ano i/ as in Ac Sg Mn hi arj-anoh *every*. 1.1.3« Sievers* Law 1. i ^ 1

i i 12

/ a . FA or N, ja-class, Nm ! TSbl4 morpheme! fSb31 aor-! LSg Mn J [jpheme /z/J b.

St or Wk V, j-class TVbl? morpheme (i.e.f taking the Vbl? morpheme, but not /hir/ *come here*,

Vb21 mor-! pheme /n/

_Imp rwk V, j-class, L_nominalized

!

J ----

c. “ (Nm PI of i-class N) “ (“Sbl?11 or Pnltf] rsb30 mor-I or (Nm PI Mn of /þri/ morpheme j pheme /z/ •three* or of 3 Pn L ’they* ) _ J |_ That is, the morphemes Sbl^, Vbl7, Sbl7, and PnlO (all realized as /i/) are geminated in the specified morphosyn­ tactic environments by inserting another /i/ in the immedi­ ately following position.

j

The

2,

i i

Minor

Rules

(Vocalic)

37

i:

1 2 / a, "(A or N, optional optional or Wk V,

ja-class, St or (i) Wk Mn or (ii) Nt) or (St j-class).

Phonological condition_ingi long syllable

Sb]> or Vbl?“1" morpheme j

“]

"

J

1

2

b. "(St or Wk V , j-class," r~Vbl? ^ " "* /TVb21 "’S Imp) or (Wk V, jmor; l mor\ class, nominalized pheme pheme with the Vb21 mor\ /n/ jpheme /n/) |_---------

J

JL

1

_ \L

2

c. [~(Nm PI of i-class N) 1 "Sbl7 "IT or (Nm PI Mn or /þri/ or •three1 or of 3 Pn PnlO ^ t h e y ’) J morpheme

1

Jj

" TSb30 morpheme

/%/

2

That is, the sequence /ii/ is realized as /is/ in the spec­ ified morphosyntactic environments. The second /i/ in the sequence in (2a) may be the Sbl5 morpheme, the Vbl8 mor­ pheme, or the /Í/ inserted by part (la) of Sievers* Law. (Parts 1 and 2 of this rule are ordered.) In the first optionality in (2a) the rule applies very seldom! in the second, about half the time. The second /i/ in (2b) and (2c) is inserted by parts (lb) and (lc). A "long syllable" is defined as follows: "4-vocalic"] _+long

J

Q-vocalicJ L

Q-vocalicQ^

Q v o c a l i c J L-vocalicJ iLtvocalicJ _ I.e., a syllable with a long vowel and ending in at least one nonvocalic segment, or a syllable ending in at least two nonvocalic segments, or a syllable with at least two vocalic nuclei and ending with at least one nonvocalic seg­ ment •

Gothic

38

E .g.* (la) and (2a)i NmSgMn /herd-i-z/ •shepherd* ( = stem 4 Sbl4 4 Sb3l) hy Sievers* Law 1.1.3 (la) - > herdiiz (1 .1 . 3 (2 a)) -4 > herdi:z (ev) -£> //herdiis// hairdeis as opposed to NmSgMn /har-i-z/ farmy* (1.1.3 (la)) -4> hariiz, to which 1 .1 . 3 (2 a) does not apply because /har/ is not a long syllable, but by 2.1.4 harjiz (ev) -4> //harjis//. Also the NmSgMn of the A /wilþ-i-z/ •wild* ( = stem 4 Sbl4 4 Sb31) by 1.1.3 (la) -£> wilþiiz and by 1.1.3 (2a) wilþisz (ev) -4 > //wilþiis// wilþeis. A single example of optionality i in 2a is attested, the GnSg /unse:l-i-i-n-z/ •evil* ( = stem 4 Sbl4 4 Sbl2 4 Sb20 4 Sb29) by 1.1.3 (2a) unse:li:nz (ev) — J> //unse :li :ns// unseleins (Ephesians 6 «l6 ) as opposed to the more frequent unseljins from the same source by rule 2.1.4. An example of optionality ii in 2a is the GnSg of the Nt N /andbaht-i-i-z/ *office* (2a) — > andbahtisz (ev) — > //andbahti:s// andbahteis (attested once) as opposed to andbahtjis (attested three times) to which 2a does not apply and 2.1.4 does. The same optionality in the GnSg of other ja-class N is in gawairþeis •peace* attested four times vs. gawairpjis attested six times, and waldufneis *power* attested once vs. waldufnjis attested twice. Part 2a can apply only if the N or V stem ends in at least one nonvocalic segment. Hence the 3 Sg Prs /sto:-i-i-d/ *judge* where neither any part of 1 nor 2 a can apply, but by rule 2.1.4 — sto:jid (ev) — ?> //stosjiþ// vs. the Imp /sto;-i/ ( = stem 4 Vbl7 ) by lb sto:-ii (2 b) — sto:-is (2 .1 .6 )— > //stO; i i// staue i . E.g., (lb) and (2b); 3 Sg Prs /so:k-i-i-d/ *seek* (2b) — $> so:ki:d (ev) -£> //soskiiþ// sokeiþ as opposed to /nas-i-i-d/ *save* where 2 b does not apply, but by rule 2.1.4 — nasjid (ev ) - > //nasjiþ//. But the Imp /soik-i/, /nas-i/ (lb) — > so:kii, nasii (2 b) — > //so:ki:// sokei, //nasi:// n asei. Also the nominalized /so:k-i-n-z/ ( = stem 4 Vbl7 4 Vb 21 4 Sb31* the i-n listed as derivative suffix 1 6 ) by lb -4 > so:kiinz (2 b) — $> so:ki:nz (ev) — //soikisns// sokeins *search*, similarly naseins *salvation* from /nas-i-n-z/ with the same suffixes and the same deri­ vation. The Imp /hir-i/ ( = stem + Vbl7) does not undergo lb or 2b, hiri *come hereî* Nor do other types of nominalizations of Wk j-class V*s such as compounds like /þiuþ-i-kw iþ-t-z/ *blessing* ( = stem 4 Vbl7 4 stem 4 derivative suffix 52 4 Sb31) — í> þiuþ-i-qiss, not #piupei-qiss. E.g.,

(lc) and (2c): NmPl i-class N /balg-i-z/ *bag* balgitz (ev) //balgiís// balgeis; NmPIMn /þri-i-z/ *three* (lc) — > þri-ii-z (2 c) þriisz (2 .1 .1 ) — £> þri:z (ev) — £> //þri!s// þreis? NmPIMn of 3 Pn /i-i-z/ ( = PnlO 4 Sbl7 4 Sb30) by lc — > i-ii-z (2c) — > i-i:-z (2.1.1) i:z (ev) //i:s// eis *they*.

( = stem 4 Sbl7 4 Sb30) by lc — > balg-ii-z (2 c) — >

The ja-class A Nm Sg Mn (also probably the same form

The

Minor

Rules

(Vocalic)

39

in the Gn Sg Mn) freis *free* presents an unusual case, if the stem is considered to consist of only two consonants, /fr/, then the derivation is as followsi Nm Sg Mn /fr-i-z/ ( = stem + Sbl4 + Sb31) by la fr-ii-z by 2a (where the £honological_condition for a long syllable is fulfilledi [_-vocalicjL l_-vocalic3if namelY ending in, and in this case consisting entirely of, two nonvocalic segments) fr-i :-z (ev) //friis// freis. A form like the Dt Sg Mn would be derived as /fr-i-amma/ ( * stem -j- Sbl4 + Sb7) by 2.1.7 (glide insertion) — > //frijamma//. The //j// in such forms as frijei ‘freedom* ( = stem -f Sbl4 + derivative suffix 18) is also inserted by rule 2,1.7. Presumably, the NtSgNm or Ac would be fri and optionally frijata, The IE antecedents of Sievers* Law have been described by Sievers (1878s esp. p. 129) and in a later addendum by Edgerton (1934* esp. p. 235)» According to Sievers, IE prevocalic /j, w/ were realized as //i, u// after long syllables and as //j, w// after short ones. Edgerton pro­ posed the additional proviso that IE prevocalic /j, w/ were realized as //ij, uw// after long syllables. Our formu­ lation of Sievers* Law under 1 (whereby /i/ — /ii/) is reminiscent of Edgerton*s hypothesis whereby /j/ — /ij/, while the alternation under 2(/ii/-~> /is/) may well be the reflex of Sievers* original /j/-to-/i/ rule after long syllables (i.e., /ii/ to /is/ after long syllables, other­ wise /ii/ to /ji/ by rule 2.1.4 below). Whatever the IE situation may have been, it is clear that in Gothic the rule applies only to /i/, not to /u/, and that it has be­ come in large part morphologically conditioned. It was thus a dying rule, having been to a large extent super­ seded by Go. phonological innovations like rule 2.1.4 (/i, u, j, w/-alternation) below. Our version of Sievers* Law is more complex than other recent formulations such as those in Vennemann (1971), Fullerton (1977)» and Schmierer (1977)» The complexity of our formulation is ascribable to two factsi First, we have included all morphosyntactic and lexical-exception infor­ mation in the rule instead of considering individual lexi­ cal items marked with features like "-hSievers * Law” or ••-Sievers* Law". Second, our rule applies to a larger domain than do the other formulations and is to our knowl­ edge the only descriptively adequate version. 1.1.4. /i/ (Vbl7 morpheme)-deletion i (Vbl7) -£> 0

/ r ( a ) All Pst forms of the follow-"] ----ing Wk V*s: /bug/ *buy#, /brusk/ 'use1, /þank/ fthink*, /þunk/ •seem*, /work/ *work*; or (b) Pst but not Prt of all V ’s formed _with the V stem -f /at/ _

Gothic

40

E.g., Pst 1 Sg Ind /bug-i-d-a/ ( s stem + Vbl7 + Vbl2 + Vb3) by 1.1.4 -$> bugda (1.2.1) -£> buhda (1.2.8) — buhta (2.1.3) //bohta//# also Pst Prt //bohts// bauhts from /bug-i-d-z/; Pst 1 Sg Ind /þank-i-d-a/ ( = stem +■ Vbl7 + Vbl2 +■ Vb3) by 1.1.4 — > þankda (1.2.1) -> þanhda (1.2.8) þanhta (2.2.8) //þa:hta//, also Pst Prt //þa:hts//. Also Pst 1 Sg Ind /kO:p-at-i-d-a/ ( = stem -I- derivative suffix 7 + Vbl7 +■ Vbl2 + Vb3) by 1.1.4 kO»p-at-d-a (1.2.2) kOspasda (1.2.8) -£> //kOspasta//, written kaupasta# but Pst Prt where 1.1.4 does not apply: kaupatiþs. This rule would also apply to /bring/ #bringf (Pst /brang/ by 1.1.1 ablaut) if it is considered for the purposes of this particular rule a Wk V. 1.1.5« /i/ (Sbl4 morpheme)-deletion i (Sbl4) —

0

[”A# i-class.

1

Phonological conditioning:

_long syllable

fQ-vocalic] [##

J

where ••long syllable" is as defined in rule 1.1.3 (part 2a). E.g.# NmSgMn /hrE:n-i-z/ ( = stem + Sbl4 + Sb31) by 1.1.5 — > hrE:nz (ev) //hrEins//, hr a ins *pure#; also NmSgNt /hrE:n-i/ ( * stem + Sbl4) by 1.1*5 —> //hrE:n//. On the other hand# the GnSgMn /hrEin-i-z/ ( = stem +- Sbl5 4- Sb29) — $> //hrErnis// hrainis where the Sbl4 is not in­ volved. Likewise, the DtSgMn / h r E :n-i-amma/ — > //hrEinjamma//, where the Sbl4 morpheme is retained since it is immediately followed by a vowel. Likewise NmSgMn /sut-i-z/ ( = stem -f- Sbl4 + Sb3l) //sutis// #peaceful*# where the Sbl4 morpheme is retained since /sut/ is a short syllable. 1.1.6. /i:/ (Vbl9 morpheme)-shortening i: -£>

i

/ \y stemj “Vbl9 morpheme“ (Qnrocalic]

.----

JU#

E.g.# Pst Sub 3 Sg /ne:m-i:/ 'took* — //ne :mi// nemi, Pst Sub 1 Sg /ne :m-i :-0:/ (1.1.6) — > ne:miO: (2.1.4) — > //nesmjO:// nem j a m but Pst Sub 2 Sg /ne:m-i:-z/ (ev) — > //ne:mi:s// nemeis. This rule is morphologically condi­ tioned in that other /ii/*s remain long word-finally# e.g. /manag-i:/ ( = stem •+ Sbl8) //manaqi:// managei# not ♦managi . a

The

Minor

Rules

(Vocalic)

41

1.1.7. /o:/ (Sb24 and Sb26 morphemes)-to-/a/ rule o: —

a

/ a. []Nm or Ac PI Nt]] ~Sb24 or Sb26 "Fn. (os-class 1 A, N, or 3Pn)

stress

JL

~~ >

J

##

b. ”Nm or Ac PI Nt~! ~~Sb24 morpheme ~ ; —stress /tw/ *two*

J _-----

_

J

E.g^, (a) NmPINt /word-oi/ #words* ( = stem +■ Sb24 > - > //worda// waurda v s . the NmPINt of the Art /þ-o:/ ( = Pn24 + Sb24) //þo://, not *//þa// since, the Sb24 mor­ pheme is stressed. Also Nm or Ac Sg Fn ,/gib-o:/ •gift* -> gíba (ev) //a Íða/V, but not GnSgFn /gíb-os-z/ ( = stem -h Sb26 + Sb29) //gíBo^Js//# nor the Nm or Ac Sg Fn of the Int /hw -os/ — /yhwcf:// since the Sb26 morpheme is stressed. The rule does not apply to^the AcSgFn ni... haeilohun *not for a single hour* /hw i :l-o:-hun/ T~- stem +■ Sb26 4 /hun/) because the Sb26 /os/ is not word-final. The rule is morphologically conditioned in that there are numerous word-final unstressed /o:/#s to which it does not apply# e.g. GnPIFn /gib-os/ ( = stem + Sb25)# ImpSg /sálb-o:/ *anoint* ( = stem + Vb25). 1.1.8. /ot/ (Sb26 morpheme)-deletion oî

ft /

Nm Sg Fn jo«-class N.~ rsbl4 mor-“j TSb26 j pheme /i/ :j morPhonological conditi j ! pheme __ioningi long syllable^ j j _ _ _ _

##

where ••long syllable* is basically as defined in rule 1.1.3 (part 2a)» except that the syllable need not necessarily end in a nonvocalic segment«

"fvocalicH (or diphthong) _+-long

~

J

C-vocalic^ 2

L-vocalicV2

Qhvocalic]] Q-vocalic^ L+vocalic]] E.g., NmSg /band-i-o:/ *fetterf — //bandi//» but AcSg /band-i-o«/ where 1.1.8 does not apply# but by 1.1.7 and 2.1.4 //bandja//. NmSg /mau-i-o:/ #maiden* (1.1.8) -^> maui (2.1.4) //mawi// written mawi vs. AcSg /mau-i-o:/ where 1.1.8 does not apply, but by 1.1.7 mau-i-a (2.1.4) mauja (2.1.5) — > //mO«ja// written mauja. Rule 1.1.8 does not apply to short syllables such as NmSg halja ’hell*»

Gothic

42

not *hali. 1.1.9. /u/ (Sb27 morpheme)-to-/i/ rule u (Sb27)

i

/ [u-class kj ----- [VvocalicQ

E.g., NmSgMn or Fn /hard-u-z/ •hard1 ( = stem -f- Sb27 + Sb31) ev //hardus//, but Dt Sg Mn /hard-u-amma/ ( = stem -fSb27 + Sb7) by 1.1.9 ->> hard-i-amma (2.1A) - > //hardjamma//. Likewise» the Nm or Ac Sg Nt can be either hardu or hardjata. Rule 1.1.9 probably represents an extension in the environment of morphological rule Sbl4 which inserts /i/ immediately after the stems of j-class A. This rule tended to supplant Sb2? in the A paradigm so that by the time the later Gmc. languages are attested, u-class A had for the most part disappeared, becoming either regular (as in para­ digm 2.1 in appendix 1 to this chapter) or j-class A (para­ digm 2.1.1) . 1.1.10. Word-stress rule ["-♦-vocalic”!

r-hVocalicH

_-t-stress J

/ ## T - v o c alic^ ["Not: Verbal derivational

prefixes, en­ clitic parti­ cles, or redu­ plicative pre­ fixes from rule 1.2.2 below.

That is, the first vowel occurring in a word takes the main stress except if this vowel occurs in a verbal derivational prefix, in an enclitic particle, or in the reduplicative syllable in the Pst of reduplicative St V #s. The verbal derivational prefixes are given in section 3*1 of this chapter. These are all prefixes which can occur with a V. Hence /bi/, which may be added only to V*s, is such a pre­ fix as is /ga/, which can be added to V*s as well as to A*s and N's. On the other hand, the prefix /un/, which may be added only to A*s — including Prt — is not a verbal pre­ fix in the sense intended here. The verbal prefixes are thus /anda/, /bi/, /diz/ /fer/, /fra/, /fri/, /ga/, /tuz/, /twis/, and /unþa/. Enclitic particles are all constitu­ ents which can occur between a verbal prefix or the first element of a verbal compound (of which more directly) and a verbal stem, e.g. at-uh-gaf #handed to* or bi-u-gitai •meet *.

The

Minor

Rules

(Vocalic)

43

One example of the application of the stress rule is in the Pst forms of reduplicating V's such as haihait 'ordered* /hehErt/. In deriving such forms one can assume that rule 1.2.2 for reduplication applies after 1.1.10* /##hE:t##/ (1.1.10) ->> /##h£:t##/ (1.2.2) - > /##hehEst##/. Other examples of the stress rule are ##tuz##wer.jan## 'doubt' - > ##tuz##werjan## and ##unfra##b 1iuhan## 'flee from1 — > ##unþa##Þ 1 1 uhan## where neither vowel in unþa is stressed since they both occur in a verbal prefix. Likewise the N ##ga## baurg ja## 'fellow citizen' is stressed ##ga##bamgja## (/^abórgja//). The ga is unstressed because, even though it occurs in a N stem here, it is nonetheless a verbal pre­ fix in the sense defined above, However, the prefix un in unbarnahs 'childless* is stressed since it is not a verbal prefix. The structure of this A is something like [^## un £ [_## barnj ahs J • That is, it is an A (A2 ), which itself consists of the N /barn/ and the derivative suffix /ah/ (suffix 2 in section 3.2 of this chapter). The stress rule applies cyclically to such constructions in that it first appliesyto_the innermost constituent (here the N), J## un T L## barnj ahs Ax A2 N N ##^|

^ .

The rule then applies to the next outer constitu­

ent, which is the A2 , and in doing so restresses the mor­ pheme /barn/. Finally, the rule applies to the outermost constituent A , . This results in the stress pattern — / — ^ s un barn]J a^s • For such cases we adopt the convention that the vowel stressed last in the cycle receives the primary stress and any other stressed vowels inside the construction are reduced by one, Thus the final

1

2

stress configuration of this word is un-barnahs with pri­ mary stress on un, secondary on barn, and minimal stress on ahs ( = derivative suffix 2 + morpheme Sb3l) . For N compounds like busundifaþs 'leader of 1000 m e n ’ we assume a structure like the following* ^

þu: sündig

^## faþs ##^j

.

I,e,, it is a N (N^)

consisting of two other N's (N^ and N2 ),

The stress rule

again applies cyclically in that it first applies to the innermost N's resulting in the stress configuration /## þ u :sundi##fáþs ##/. The rule then applies to the outer­ most N, namely N^, restressing the /u;/ in /þu:sundi/.

Gothic

44

1 2 This is then by our convention /##þu:sundi##faþs##/. Some compounds have as constituents prepositions or adverbs like /ana/. Thus a word like unanasiuns •invisible' has a structure like this* £## un £## ana

^ .

siuns

I.e., it consists of

1 2 2 1 a N transformed into an A (A2 )» which itself is a constitu­ ent of a negative A (A1 ).

The stress rule applying cyclic­

ally stresses first /siuns/, then /ana/. According to our 1 2 convention, this is /ana-siuns/. Then the stress rule ap­ plies to /un/, again reducing by convention the other two stresses already in the construction. The final result is 1 2 3 thus /un-ana-siuns/. We add in connection with this exam­ ple that diphthongs in Gothic apparently can occur under primary, secondary, and at the least tertiary stress. Under less than 3 stress, diphthongs seem never to occur* Under such stress, phonological rule 2.1.5 realizes some diphthongs as monophthongs; and under similar conditions, rule 2.1.4 realizes certain other diphthongs as sequences of a vowel and a nonvocalic sonorant segment. Verb compounds are similar to the immediately preced­ ing example in that they consist of an independently oc­ curring preposition or adverb and a verbal stem. For ex­ ample, atgiban #hand to* has the structure L## [atl T## gib-an ##~l 1 . V x P F V2 V2 Vx

I.e., it is a V (V1 ) which

consists of a preposition (P) followed by a V (V2 ).

*n ^ s

cyclic application, the stress rule applies to the P and V 2 constituents first, then to V^. This results in the 1 2 final configuration /##at##giban##/. In formulating the stress rule we have posited the presence of word boundaries (##) between the prefixed con­ stituent and the verbal stem. Corroborative evidence for the presence of a ##-boundary in this position is the fact that enclitic particles u, uh, ban, or þau can be inserted in V constructions only after ##, E.g., they can be in­ serted word-finally as in was-uh •was* /##was##uh/ and after derivative or compound prefixes as in bi-u-gitai •meet* /##bi##u-git-E:##/, ga-þau-laubidedeifr *believe * /##ga##þ0:-10:b-i-d-e sd-i;-d##/, at-uh-gaf *hand to* /##at##uh-gab##/. But forms like »bi-git-u-ai */##bi##gitu-E*##/ or »ga-laubi-þau-dedeiþ */##ga##10sb^T-pO:-d-e:di : d ^ / are impossible. This means that enclitic-particle

Th e

Minor R ules

45

insertion in V fs is possible only at a ## morpheme juncture and not at other junctures such as between a stem and an inflectional ending. We should add in this connection that enclitic-parti­ cle insertion does not apply between the reduplicative pre­ fix added by rule 1.2.2 and a V stemi Hence only saiso •sowed* can occur, not *sai-þau-so or the like. This probably means that the reduplication rule 1.2.2 adds the re­ duplicative syllable with no intervening ##-boundary be­ tween it and the V stem. Enclitic particles also do not occur between a prefix and an A or a N stem. This may mean that the syntactic rule of enclitic insertion applies only word-finally to A or N. (It may also mean that there is no ##-boundary between a verbal prefix like /bi/ if followed by an A or a N stem.) In any event, we assume that such prefixes, whether occurring with A, N, or V stems, were un­ stressed . The evidence for most aspects of the stress rule is comparative* That is, the other attested Gmc. languages have stress rules like this one. However, there can be no comparative evidence for our assumption about the absence of stress from reduplicative prefixes and from enclitic particles since the other Gmc. languages do not have a re­ duplication rule like 1.2.2 below nor do they evince encli­ tic particles. We give in our discussion of rule 1.2.2 our reasons for assuming that reduplicative syllables were un­ stressed. We may assume that enclitic particles like -uhin at-uh-gaf cited above were unstressed since if they had been stressed, the breaking rule 2.1.3 would certainly have applied to them to produce forms like *at-auh-^af instead of at-uh-g a f . Other evidence internal to Gothic for posit­ ing the stress rule as formulated here is that it provides plausible environments for several other rules, among them the /o:/-to-/a/ rule 1.1.7, Thurneysen#s Law 1.2.3# the rule of vowel deletion 2.1.2, the rule of /i, u, j, w/-alternation 2.1.1*, monophthongization 2.1.5# and the unstressed /u/-to-/o/ rule 2.1.8. 1.2. Consonantal 1.2.1. Consonantal change in A and V stems “-hconsonant

~| -^> T-t^ontinuant

-sonorant

» C-backH or [>baclGb J a a b

H

+tense -voiced _ __ L+stridentl or ! -high a a

I

L+low J *

L respectively

b

b

Gothic

46

/ nA or V stem, but not”1 "“Morpheme : fAny nominalizing de-“l i PrS> 2 Sg Ind of Pp i_boundaryJ rivative suffix or ! V #s /mag/ *can* or any verbal suffix /o:g/ 'fear* íwhatever, deriva­ tional or inflec-

----

J

tional. sonor ant

I.e., Tp# b, f 1 t, d, þ

_

"! - > ^f-1 / in the environi ment specified, j !3

_k, g ( = //g. x//), h J

j_h_

E.g., /ga-skap/ ’create* -I- /t/ (derivative suffix 52 in section 3-2) - > gaskaft ’creation*, /fra-gib/ ’give* -h /t/ (derivative suffix $i) frag if t *conferral*, 2 Sg Pst /bi-gat-t/ *you found* -^> bigast, Prs 2 Sg of Pp V /wE:t-t/ *you know* //wE sst// wa 1st, /gild/ *pay* +• /str/ (derivative suffix ^9) — >> gilsstr (rule 2.2*3) gilstr ’tax*, 2 Sg Pst /warþ-t/ *you became1 -~> warst, /work/ *work* +- /stw/ (derivative suffix 5 0 ) — > worhstw (2.2.6) — £> //worstw// waurstw *work*, /suk/ *sick* (ab­ laut ing form of the A /siuk/) + /t/ (derivative suffix 52) suht (2.1.3) -£> //soht// sauht *sickness*# /anda-bug/ •buy* + /t/ (derivative suffix 5 2 ) — > anda-buht (2.1.3) — £> //andaboht// andabauht *ransom*, /mag/ *can* +■ /t/ (derivative suffix 5 2 ) — > maht *power*, /fram-gang/ fto go forth* + /t/ (derivative suffix 5 2 ) fram-ganht Í2.2.8 and ev) — > //fram^a:ht// *progress*, Pst of Pp V /mag-d-a/ *could* //mahta//. This rule does not apply to the Prs 2 Sg of the Pp V*s /mag-t/ *can* magt //maxt// by rule 2.2.1 or /o:g-s/ •fear* — £> ogs //o:xs//. (The negative environment for the rule may be the Prs 2 Sg Ind of all Pp V*s ending in /g/.) The rule also does not apply if the derivative suffix begins with a -fsonorant segment, e.g. /so:k / *dispute* -+ /n/ (derivative suffix bO) —>> //so:kn// *disputât ion*, not ♦//so:hn//. The rule must apply over a morpheme boundary; hence the dual ending Vb28 is /ts/, not #/ss/. And the N •money* is skatt, not #skast. Finally, the rule does not apply over a N stem and a nominal inflectional ending (Sb morpheme). Hence the NmSg is guþs *God*, not *guss. Subject to the restrictions given above, rule 1.2.1 seems to have applied generally, van Coetsem (1972:188) cites 2 Sg Pst halpt *helped* as a Go. form. We assume such a form would have to be halft and have found no form ♦halpt attested. van Coetsem might have been led astray by Krause*s remark (1953*^5) on the **Typus go.-an. halpt.** The form halpt is of course Old Norse (where, it might be added, the phonetic value of the £ was doubtless //f// or

The

Minor Rules

// ----- ]/ *stand •_ c. Optional* P A U

forms of

“

L /E« Ch ----- J/ 1have•_ d. Optional! pPst, but not Prt forms of _ / C9 ----- C le:P/ fsleep#_ where the optionality under c applies in 12 out of 19 instances or at about 63% and that under d in 2 out of 5 instances or k0% of the time. E.g., (a) Prs 1 Sg Ind þarf #I need* vs. 1 PI Ind þaurbum, Prs 1 Sg Sub þaurbjau, Prs Prt paurbands, Pst 1 Sg Ind /þorf-t-a/ ( = stem -h Vbl2 + Vb3) by 1.2.4 þorb-t-a (1.2.1) — £> //borfta// þaurfta; (b) Prs Inf /staþ-an/ •stand1 ( 1.2.M sta?an (1.2.5) s t a n A n (2.2.2) //standan// vs. Pst 1 Sg Ind /staþ/ (1.1.1) //stoip//, also the Pst Prt /stap-anz/ (1.2.^) — > stací-anz (1.2.5) — > stancJ-anz (ev) — ^ standans ; (c) Prs 1 Sg Ind /Eih/ •have' (1.2A) — either //E:h// aih or E:g- (2.2.1)

The

Minor

Rules

(Consonantal)

55

//Eix// aig, also 1 PI either a ihum or aigum, Prs Prt^ either aihands or aigandsi (d) anasaislepun //anasesle:pun// •they sleptr (Thessalonians vs* gasaizlep //jasezle:p// (John 11 ill). We have included here only the alternations occurring within inflectional paradigms. There are other alternations occasioned historically by Verner's Law which occur in derivational paradigms: e.g., the A fulgins 'hidden* vs. the V fulhan 'hide*» alþs 'old* vs. aids 'epoch'» etc. 1.2.5. /n/-insertion in St V

0

n

/ a. QPrs of /freh ----- / 'ask'J b.

QPrs or Pst Prt of /sta ---- þ/ 1 stand'J

That is, /n/ is inserted as specified. E.g., Prs Inf fraihnan /freh-n-an/ vs. Pst Sg frah; Prs Inf standan, Pst Sg stoþT Pst PI stoþum, Pst Prt standans. This rule could also be formulated as an /n/-deleticn rule in the appro­ priate environments. 1.2.6. /s/-insertion in V

0

s

/ “V stem,

*“ ----- “Derivative suffix 52

i

!

[_4-coronal, +*iasal_

|_in section 3*2, /t/

I

J

E.g., /brunn/ 'burn* +• /t/ (derivative suffix 52) by 1.2.6 — > brunnst (ev) — > //brunst// 'burning', also anst •favor' from /ann/ 'grant' + the same suffix. 1.2.7. V suffixes Vbl2 /d/ and derivative suffix 52 /t/ realized as /s/ “-hconsonantl -^> “ fcontinuantH -sonorant _-hcoronal

-voiced _

[_-hstrident

_

/ a. rPst of /wit/ “ Flnflectional suffix Vbl2“| |_'know'

_ _ /d -----

(continued on next page)

/

J

Gothic

56

b. TV stem,

”1 ["Derivational suffix 52*"]

4-consonant

__ /t ----

/

J

—sonorant

4-coronal _-strident _ That is, (a) the inflectional suffix Vbl2 is realized as /s/ when occurring with the V /wit/; and (b) the derivative suffix 52 is realized as /s/ when affixed to any V stem ending in /t, d, þ / . E.g., (a) Pst 1 Sg Ind /wit-d-a/ ( = stem 4- Vbl2 + Vb3) by 1 .2 . 7 — witsa (1.2.1) — > //wi 9 sa//. Other V*s do not change Vbl2 /d/ to /s/, such as Pst 1 Sg Ind /kO:pat-d-a/ fhit# (1.2.8) — > kOrpatta (1.2.1) — > //kO ipasta// kaupasta 1 nor the Pp V, Pst 1 Sg Ind /moit-d-ð/ •have room* (1 .2 .8 ) — > moitta (1 .2 .1 ) — > //moista//. (b) /ab##stad-t-z/ •apostasy* ( = prefix +■ stem +■ deriva­ tional suffix 52 4- Sb31) by 1 .2 . 7 ab##stadsz (1 .2 .1 ) — $> ab##stassz (2 .2 .1 ) — j> af##stasss (2 .2 .3 ) — > //afstass//; /ga##kw iþ-t-z/ •agreement* ( = prefix 4- stem 4- same suffixes) by 1.2.7 ga##kw iþsz (1 .2 .1 ) — > ga##kw issz (2 .2 .1 ) ga##kw isss (2 .2 . 3 and ev) —> //^akw iss// gaqiss. We note that rule 1.2.7 does not apply to V stems end­ ing in /s/. Hence the construction /ur##ris-t-z/ •upris­ i n g * ^ = prefix +■ stem + derivative suffix 52 + Sb31) is realized as urrists, not *urriss.

1 .2 .8 . /d/ (Vbl2 and Vbl3 morphemes) realized as /t, þ/ /d/, the Vbl 2 and” — _Vbl3 morphemes

_

a. “-continuant”] / "-vocalic” ----_—voiced

J

—voiced

b. Ttcontinuant”] / [”Pst of -voiced _-strident

”1 -----

/kunn/

J

|_#can#

(a) Pst of Pp V /mojt-d-a/ *have room* (1.2.1) ->> moisda (1.2.8) - > //mo:sta//, Pst of Wk V /þunk-d-a/ •seemed* (1.2.8) þunkta (2.2.8 ^>//þu :hta//, Prs 3 Sg l1)0 of ’be* /i-s-d/ ( = Bel 4- Be3 + Vbl3) by 1 .2 . 8 -£> //ist//J (b) Pst of /kunn-d-a/ (1.2.8) — > kunnþa (2.2.3) -$> //kunþa//. We have probably conflated here two separate

The

Minor

Rules

(Consonantal)

57

phonological processes. Rule 1.2. 8 .a is probably the mani­ festation of a phonologically conditioned rule by which nonback obstruent consonants assimilate for voicing to a preceding segment:

1 .2 .8 .a* ~-f-consonant”l

[j^voiced]] / ^-vocalic H -----

-sonorant back

—sonorant _

jxvoiced

We limit 1.2. 8 .a* to nonback obstruents in view of the form lntrusgjan *graft onto* instead of *intruskjan. Rule 1.27ö.b may well be part of Verner's Law (1.2.4). 1.2 .9« /z/ (Sb31 morpheme)-deletion z ->

ft / [A or N, Nm Sg Mn

““ r’Sb31

“1

I

a. HfconsonantH = /s, z/

morpheme

-hcoronal

[_-----

_

|_+strident b. ”~+-vocalicH T-f-sonorantH - /Vr/ _-long

J

-Coronal —nasal

_

[^-lateral _

__

That is, the Sb31 morpheme /z/ is deleted from N or A stems ending with the segments specified. E.g., (a) /drus-z/ •fall* — $> drus, not *druss which would be phonologically possible (cf. afstass 'apostasy*)j (b) NmSg /wer/ wair 'mart, not *wairs (which form can occur however, meaning 'worse *)5 and the NmSgMn of the A anþar 'other*, unsar *our#, not ♦anþars, etc. Also the NmSgMn stiur 'steer* for /stiur/, which also fits environment (b) since the second segment of the diphthong is a short vowel. However, if the vowel of the V stem is long, /z/ remains as in hors 'adulterer* from /hotr-z/ and swers 'honored' from /swe:r-z/. This rule thus furnishes some additional evidence that the digraph iu represented a diphthong, not a long vowelt In the latter case, one would expect *stiurs. A further morphosyntactic condition on this rule may be that it does not apply to i-class A's, e.g. the NmSgMn gafaurs 'moral', not »gafaur, and unfaurs 'talkative', not *unfaur. In these cases, however, it is not clear if represents short /o/ or long /o:/.

Gothic

58

1 .2 .1 0 . ##-deletion ##

ff / a . ----- [""Enclitic particles /i;/, /uh/, ” jand perhaps others• b. [""Unstressed steins such” _____ as the Pn /þata/

Prs of /ist/

•that*, /kara/ 'care', jand perhaps others.

_ V, 3 Sg, Ind _*is*

^

_

That is, the word boundary ## is deleted in the environ­ ments specified. Although optional» this rule (particu­ larly part (a)) applies often, perhaps 99# of the time. Its effects are manifested in the application or non-appli­ cation of other rules, in particular rule 2 .2 . 1 (the rule of consonantal devoicing) and rule 2 .1 . 2 (vowel deletion). E.g., hi a zuh •every* from /hwa-z##uh/, which would have been realized as *hiasuh by rule 2 .2 . 1 if 1 .2 . 1 0 had not applied; the relative ^Pn /þata##ii/ ( = Pn + enclitic particle) by 1 .2 . 1 0 — þataii (2 .1 .2 ) — //þatii// a t e i . There are also a few attested instances where .2 . 1 0 has apparently not applied, thus triggering rule 2 .2 . 1 of consonantal devoicing; bid jand ansuþþan #asking# (Matthew 7*7) from /bid-i-a-nd-a-n-z##uh-þan/ instead of the more usual bidjandanzubfran. Similarly, in some cases the non-application of 1 .2 . 1 0 prevents the application of rule 2.1.2 of vowel deletion; huouh *whof (II Corinthians 6 ;1 5 ) from /hw o;##uh/ instead of *hu oh, which would have arisen from the same source if 1 .2 . 1 0 and 2 .1 . 2 had applied.

Ï

5.2. The Major Rules

2 .1 . Vocalic 2.1.1. Like-vowel contraction -t-vocalic H ” +vocalic "1 — > pfvocalic features

J31 ong

J

^features ong

_

^features [_+long

That is* two like vowels, one long and the other short, are realized as the corresponding long vowel. E.g., NmPIMn /þri-i;-z/ #three# ( = stem + Sbl7 + Sb30) by 2.1.1 þri;z (ev) -£> //þri;s// þreis; NmPIMn /i-i;-z/ •they* ( = PnlO -+• Sbl7 + Sb30) by 2.1.1 — > i;z (ev) — //i;s// eis; the relative Pn /s-i-i;/ *she who* ( = Pn20 + PnlO + particle /i */) by 2 .1 . 1 —p //si;// sei, but *she# is si ( = Pn20 + PnlO).

Th e

Major

Rules

(Vocalic)

59

Rule 2 .1 . 1 does not apply over the boundary between a stem and a derivative suffix, nor over a word-boundary (##). E.g., friei or frijei ffreedom# from /fr-i-i:/ ( = stem 4- Sbl4 f derivative suffix 18). The form fri jei arises by the rule of glide insertion (2 .1 .7 )* 2.1.2.

Vowel deletion

QfvocalicJ — >

fi

/ a. [""Interrogative Pn

“j [Vvocalic^]

_ /hw [a ----- 3 / _ b. r+vocalic

~! HSnclitic particle ”1 _ / Lu ---- U (h)/__

T-hlong, or -stress,-

J

Lor both c. '"’Pn, 3 rd Person,

“I “flow

_PnlO morpheme /i/_

d. “flow

H Qhconsonant]

-long

“j "Vvocalic

-long

pblong, or -stress,“j

-stress

j^or both

J

e. “fvocalic“] X “fvocalic”j “-stress“ _+stress

J

4-long

L--

-stress where fstress is 1 , 2 , or 3 stress, -stress less than 3 stress; and where no word boundary (##) occurs be­ tween any of the segmental matrices nor within the variable X. E.g., (a) GnSgMn /h*a-i-z/ ( = Pn? f Sbl5 f Sb29) by 2 .1 .2a — £> hw iz (ev) — $> //hw is//, DtSgMn /hwa-amma/ ( = Pn7 f Sb7) — >> //hwamma//. Part (b), GnPIMn /þ-i-z-e :##uh/ ( = Pn2^ f PnlO f Sb28 f Sbll) by 1.2.10 -> þ-i-z-e:uh (2.1.2b) — £> //þize:h// since /e :/ is flong; NmSgMn of the demonstrative /sá##uh/ (1 .2 .1 0 ) sáuh (2 .1 .2 b) //sáh// since /a/ is stressed; and the NmSgFn of the demonstrative /só:##uh/ (1 .2 .1 0 ) — > so:uh (2 .1 .2 b) — >

60

Gothic

//so :h// since /os/ is -hstress and 41ong. Part (c), DtSgMn of Pn /i-amma/ ( = PnlO 4* Sb?) by 2.1.2 //imma//, NmSgNt /i-ata/ //ita//, but not NmPINt /i-o»/ ( = PnlO 4- Sb2*0 by 1.1.7 ia, where the /a/ is not deleted by 2.1.2 since no consonant fçllows it, (ev) //ija//. Part (d), DtSgMn of Art /þamma##uh/ (1.2.10) -^> þammauh to which 2.1.2b does not apply since the second /a/ is neither stressed nor long, but 2.1.2d does since the /u/ is —stress, hence //þámmuh//; the NmSgNt of the relative Pn /þata-ii/ — > //þátit// patei since the particle /i:/ is -hlong and -stress, but 2^1.2d does not apply to the NmSgMn of the relative Pn /sa-i:/, saei, since the /a/ is stressed. Also the Pst 1 Sg Ind /íddja##uh/ 'I went1 (1.2.10) — > iddja-uh (2.1.2d) -£> //iddjuh// (attested in John 18:30 )% /bata##ist/ *that isf (1.2.10) þataist (2.1.2d) //þatist//* similarly karist *there is care* (John 10:13) from /kara##ist/ by the same rules. The Pst /fra##e*t/ •ate* apparently undergoes 2.1.2d to be realized as //frett//, vs. the Prs 3 Sg Ind fraitiþ (II Corinthians 11:20) for //fraitib//, to which the rule does not apply since the first /i/ is neither -hlong nor -stress. The fact that the rule applies here may indicate that 2.1.2d can apply over ## if the unstressed /a/ is in a V prefix and the following vowel in a V stem. (The rule cannot apply in cases like ga-u-laubjats *you two believe* (Matthew 9*28) to produce *guTaubjats,~ apparently because the /u/ follow­ ing the deletable /a/ in /ga/ is not in the V stem.) Finally, part (d) applies quite possibly in the derivation of /hánd-u-ag-z/ •smart1 ( = stem -h Sb2? -h derivative suffix 1 + Sb31) by a slightly modified version of 2.1.2d handugz (ev) handugs. The modified version of part (d) would have to specify that the deletable /a/ could be either preceded or followed by the segment Qhvocalic, L+long or -stress, or both]] J . There are numerous examples of part (e), which Beade (1973*316) also posits. E.g., numerous forms of the osclass Wk V paradigm such as the Prs Sub 1 Sg /sálb-oi-0:/ •anoint* ( = stem + Vb25 + VblO) - > //sálboi//, 2 Sg /salb-oi-Ei-z/ ( = stem + Vb25 + Vb7 4* Vb31) by 2.1.2e salb-o:-z (ev) //salboss//, Prs Psv Ind 1 Sg /sálb-o:a-d-a/ ( = stem 4- Vb25 + Vb5 + Vbll + Vb5 again) by 2.1.2e salb-o:-d-a (ev) - > //sálbo:ía//» Prs Act Ind 1 Du / salb-o :-o :3 / ( = stem + Vb25 + Vb26) -^> //salboss//. We note in connection with the immediately preceding that 2.1.2e does not apply if the preceding vowel is stressed. Hence forms occur like Pst 1 Sg Ind stauida //s t O :i a// *1 judged* from /sto:-i-d-a/ ( = stem + Vbl7 + Vbl2 4- Vb3), not *stauda; bauaida //bOiEiaa// *1 built* from /bC:-Ei-d-a/ ( = stem 4- Vb6 + Vbl2^+ Vb3), not »baudai and the Prs Prt saiand *sowing* for //sE:and// from /sé Ír^d/ ( • stem *4 Vbl + Vb2^), not *saind. In view of all this, we assume that the reduplicative forms like saisoum sowed*, waiwoun *blew*, and lailoun *despised* discussed

The

Major

Rules

(Vocalic)

61

under rule 1 .2 . 2 above must have been stressed saisoum, waiwoun, and lailoun. If they were stressed on the reduplicative syllable, then by 2 .1 .2 e they would assuredly be *salsom, *waiwon, and »lailon. Finally, we emphasize that 2.1.2e does not apply over word boundaries (##). Hence the 3 Sg Prs Sub form qimaiu •let him come* is attested (Matthew 27*^9) from /kw im^ E:##u/ to which the ##-deletion rule 1.2.10 has not applied. Similarly, the form armaio •mercy* is attested, possibly from something like /arm##Ei##o «/ ( = stem + Vb 6 + Sb26) where the /E:/ and /o:/ function as derivative suffixes separated from the stem by word boundaries (##), over which the rule does not apply. 2.1.3» Breaking: /i, u/-/e, o/ alternation -hvocalic” -^> 1. Qhhigh] / ##X “Vstress~"Yr-stressn Z## —low _-long

_-----

J

-----

A

B

where the variables X, Y, and Z contain no word boundaries ##: and Y and Z may contain instances of nonstressed vowels to which the rule can also apply.

2 . £-high^] /

F (-back) ~ a a +stress

_---

J

f

h

I i \

hw (t)r a a

where no morpheme boundary of any kind occurs between the stressed vowel and the following /h, hw , r, tr/. The two parts of this rule are ordered. Part 1 says that any —low and — long vowel (/e, o, i, u/) must be realized as /i, u/ if it occurs as the main stressed vowel within a word (matrix A) or if it occurs as an unstressed vowel any­ where after the main stressed vowel within the word (matrix B ) • We adopt the convention that the rule applies repeated­ ly whenever its environmental conditions are fulfilled: Hence part 1 can apply repeatedly not only to a stressed vowel (matrix A), but also to any number of posttonic vowels (matrix B ) . Part 2 says that the vowels /e, o, i, u/ must be realized as /e, o/ if stressed and immediately followed — with no intervening morpheme boundary — by /h, hw , r/. Further, stressed /i/ is realized as /e/ before /tr/.

62

Gothic

E,g.» (l) the morpheme fidwor •four* (from /fíduosr/ by 2 .1 .4 )» when occurring in compounds» appears in an ablauted form without the / 0 :/: fidur-dogs *four-day* /fídur##dosg-z/, which would appear as *fidaurdogs if the /u/ before /r/ were stressed. Likewise» the /u/ in tiuhan •drag* /tíuhan/ is retained before a following /h/ since the /u/ is unstressed. Part 1 also applies in the Gothicization of loan words such as Go. spaikulatur 'spy' /spekulaitur/ (with retention of the latin stress) from Latin speculator. In the Gothicization of this form, the unstressed / 0/ is realized by 2.1.3(1) as Go. /u/» but the pretonic /e/ is retained since pretonic vowels are not affected by either part 1 or part 2 of this rule. A similar Gothicization is diabulus *devil* from Gk. ó Go. /diábulus/ by part 1 of this rule. (The Go. form dlabaulus /diabolus/ is also found» which probably indicates that the form was not yet completely assimilated.) A Gothicization effected by both parts 1 and 2 of this rule is Go. aurkjus /órkjus/ *jug* from Latin urceus /urkeus/. The derivation of this form is /urkeus/ (2.1.3(1)) — > urkius (2.1 .3 (2)) — > orkius (2.1.4) — > //orkjus//. Another such form is paurpuroþs 'purple* from /porpur-oid-z/ ( = stem -f Vb25 +- Vbl2 +^Sb31) » which derives in turn from Latin purpura (Gk. rropfúp porpur-o:-d-z (ev) — > //porpuroíþs//. Other instances of part 2 of the rule are to be found in St V paradigms such as the Pst PI 1 Ind /þith-u-m/ *we throve* which by the ablaut rule 1.1.lAd is realized as þihum (just as the corresponding form of /griipu-m/ *we seized* is realized as //gripum//) and by 2.1.3(2) as //þehum//» þaihum (as opposed to gripum in which the /i/ is retained). Similarly, the Pst PI 1 Ind /wirp-um/ *we threw* is by the ablaut rule 1.1.lAc wurpum (just as the corresponding form of /bind-um/ 'we tied* is //bundum//) and by 2.1.3(2) //worpum//, waurpum. A possible in­ stance of part 2 of this rule whereby /i/ before /tr/ is realized as /e/ is in the A baltr 'bitter*, probably /betr/. In the other G m c . languages the corresponding A is /bitr/. Now the orthographic configuration baitr may also represent /bEitr/, in which case the Go. form is from Gmc. /baitr/ (by rule 2 .1 . 5 below) and is an ablauting form of /bitr/. However, an additional argument for considering rule 2 .I.3 (2 ) to apply to /i/ before /tr/ is the fact that the Go. version of the name 'Peter*, Paitrus /petrus/ from Gk. TTerpos, always appears with the vowel /e/, never as *Pitrus as would be derived by 2 .1.3(1)* On the other hand, the Go. version of a name like 'Pontius*, DtSg Pauntiau /pontiOi/ from Gk. TTo v T/os , is occasionally realized by 2 .1 .3 (1 ) as Puntiau with /u/ instead of / 0/. The fact that stressed /e/ is always realized as /e/ in a frequently occurring name like Paitrus (and never as /i/ as would occur by 2 .1 .3 (1 ) in *Pitrus) indicates that 2 .1 .3 (2 ) has applied in the environment before /tr/.

The

Major

Rules

(Vocalic)

$3

Vennemann (1972*871-2) has noted that part 2 of this rule is, viewed phonetically, in fact assimilation. The features we have used capture this in part in that /h/ and /hw/ are the only 4-low nonvocalic segments. It would seem that this feature acted upon stressed /i, u/ to realize them as /e, 0/. As for the /r/, this segment has frequent­ ly been seen to have a lowering effect on preceding vowels in a number of languages, e.g. ModHG fettig *fat* //fetix// vs. fertig *ready* //fErtix//. As for the sequence /tr/, the /t/ is a stop made in approximately the same place as the /r/, which in Gothic was probably tongue-trilled. Hence the /t/ in this position could easily have become the first flap in the series of flaps which constitute trilled /r/. That is, /tr/ may have been phonetically something like //Rr// (where //R// * a voiceless /r/) and is thus not entirely outlandish as one of the environments for break­ ing. This environment was perhaps a comparatively new ad­ dition to the rule: It only affects the front vowel /i/, not /u/ as in snutrs *wise*, not »snautrs. Part 2 does not apply over any kind of morpheme bound­ ary. Hence it does not apply to the forms duhi e *why* /du-hw e:/, nuh #and now?* /nu-h/ (nu *now* occurs independ­ ently) vs. nauh •still, yet* /noh/”with no morpheme bound­ ary between the vowel and /h/, nih *not* /ni-h/, and the defective V 2 Sg, Du, PI Imp hiri, hir.jats, and hirjiþ •come herei* These forms are derived from the deictic particle /hi/ (as in hi-ta *this here* and hi-dre *over here*) 4- the adverbial locative particle /t J ~(which also occurs in the paradigm pa-dei *to there*, þa-þro *from there*, ba-r *there, at that place*) + Vbl7 +• the respec­ tive endings. Thus there is a morpheme boundary between the /i/ in /hi/ and the /r/; and part 2 consequently does not apply here. Neither part 1 nor part 2 of 2.1.3 applies to the reduplicative syllable of reduplicating St V*s. As already indicated in our discussion of the reduplication rule 1.2.2, this is because the vowel /e/ in such forms is in pretonic position, i.e. before the main stress within the word. And the rule of breaking applies only to stressed vowels (matrix A in part 1, as well as part 2) or to post­ tonic vowels (matrix B in part 1). Our account of breaking has as possible exceptions only these forms* jain *that*, aipþau *or*, waila *well*. The first of these, if it is from G m c . /jen/, might be expected to occur as Go. *jin by part 1 of our rule. Interesting in this regard is that Jellinek (1926*50) cites this same form as a possible exception to the rule of word stress (our rule 1.1.10 above): Ja%^es scheint nicht unmöglich, daB unter Um­ ständen auch im einfachen Wort der Hauptton vom

Gothic

64

Anfang wegrucken konnte. So wurde sich die Syn­ kope des -a- in jainþro *von d o r V , jaifldre •dorthin* aus *jáinaþr6, »jainadre erklären. Der Gegensatz der Bedeutungen wurde die Betonung der differenzierenden Elemente bewirkt habenj vgl. dialektisch herein» hinein. That is» if the adverbs with /jen/ were indeed stressed /jenþró:/ and /jendre:/, then part 1 could not apply to /e/ because it occurs pretonically. And the morpheme /jen/ could then have been lexicalized as such (perhaps with a feature like "No breaking rule") and would occur everywhere as /jen/. The remaining problematical cases are aiþþau and wailau The former may stand for /EíþþOt/, in which case it is irrelevant for the breaking rule. Or the form may have been completely unstressed /eþþo/, in which case neither part 1 nor part 2 of the breaking rule would apply. As for walla, it may represent /wÉ:la/; or as an exclamation in the sense of #wohlan!*, it may have been stressed /welá/. In either case it would not have been affected by the breaking rule. (A number of other possible explanations for these forms are given in Braune-Ebbinghaus (1961:23)•)

2.1.^. /i, u, j, w/-alternation -h8onorant

—>

4-high

la. Q+vocalicJ (applying before the wordstress rule 1.1.10) _

/ Q-vocalic]] ") [“-back -long

J

(

Q-vocalic]]

J

## L— J l ## —

lb. Q-vocalicJ (also applying before the

word-stress rule 1.1.10) / rVback™] [Vvocalic]]

2. Q-vocalic]] (applying after the wordstress rule 1.1.10) /a.

~Not: short~ [“A or N 1 [Morpheme“] X stressed /i/ '

stem

_bourdary _

a-cla8s 4-back

(continued on next page)

The

Major Rules

(Vocalic)

65

where X is an inflectional ending or ##, i.e. word-final position or the first constituent of a compound, b. P-stress”! Q-vocalicJ

Our formulation of these alternations is not very satis­ fying. There seem to be at least two rules involved here. By 2.1.4.1a, /j/ is realized as /i/ if both preceded and followed by either a word boundary or a nonvocalic segment. By 2.1.4.1b, /u/ is realized as /w/ if followed by any vocalic segment whatever. By 2.1.4.2a, /u/ is realized as /w/ if it is the final segment in the stem (and not pre­ ceded by stressed /i/ and hence not part of the diphthong /iu/) of an a-class (historically a wa-class) A or N. And finally by 2.1.4.2b, both /i/ and /u/ are realized as /j/ and /w/ if unstressed and occurring before a vocalic segment. E.g., (la) 2 Pn Ac PI /j-zw-iz/ *you* ( * Pnl4 4 Pn 30 4 Pnl3) izwiz (ev) — £> //izwis/y vs. NmPl /j-u-z/ ( = Pnl4 -+■ Pn26 4- Sb30) (ev) -4 //jus//; (lb) 1 Pn Nm PI /u-ii-z/ ( a Pn25 4 Pn9 4 Sb30) wiiz (ev) — //wiis// weis vs. AcPl /u-ns/ ( » Pn25 4 Pnl9) //uns//. For (2a), Nm Sg a-class N /aiu-z/ ( = stem 4 Sb31) by stress rule 1.1.10 — 4 aiuz by 2.1.4.2a — 4 aiwz by 2.1.5 — 4 Eiwz (ev) — 4 //Eiws// aiws. Another derivation from the same stem is /aiu-k-duiþ-z/ ( = stem 4 derivative suffix 34 -h derivative suffix 13 - Sb31) *eternity* to which 2.1.4.2a does not apply since it is followed immediately by deriva­ tive suffix 34 and not by an inflectional ending or by ##, but 2.1.4.2b does apply producing ajukduiþz (ev) //ajukduiþs//. Like aiws, all other a-class A #s or N's^ fitting the MS conditions of 2.1.4.2a end in /w/il5 lasiws /lasiw-z/ 'weak*, snaiws •snow*, waurstw #workf. However, those a-class A #s or N fs whose stems end in /u/ but whose MS conditions do not fit 2.1.4.2a undergo 2.1.4.2b instead! NmSg /þiu-z/ ( = stem 4 Sb31) •servant* — 4 //þius//, but the GnSg /þiu-i-z/ ( = stem 4 Sbl5 4 Sb29) by 2.1.4.2b — 4 þiwiz (ev) —4 //þiwis//. Other instances of 2.1.4.2b are 1 Sg Pst Sub /neim-it-0i/ *take* ( = stem 4 Vbl9 4 VblO) by 1.1.6 — 4 neimiOi (2.1.4.2b) -4> //neimjO»// nemjau, NmPl /sun-i-u-z/ 'sons* ( * stem -+■ Sbl3 4 Sb2? 4 Sb30) by 2.1.4.2b sunjuz (ev) -4> //sunjus//, GnPl /sun-i-u-ei/ ( = stem 4 Sbl3 4 Sb27 4 Sbll) by 2.1.4,2b //suniwei//. In the derivation of the latter form, one may assume that the rule applies from right to left, first changing /u/ to /w/ but not changing /i/ since it is now followed by the nonvocalic /w/. (One might also assume that the entire rule applies — in this case parts la and 2 b — until it can no longer applyi Then the derivation would be /sun-i-u -ei/ (2b) — > sunjweI (la) — //suniwei//.) Still other instances of 2.1.4.2b are the NmSg /þiu-i-oi/ 'female

Gothic

66

servant 1 (1.1.5) — :> Þiui (2.1.4.2b) — > //þiwi// vs. GnSg /þiu-i-o«-z/ ( = stem -h Sbl4 +• Sb 26 + Sb29) by 2.1.4.2b — > þiujoiz (ev) — > //þiujo:s//; and NmSg /mau-i-o:/ •maiden* (1.1.5) — > maul by 2.1.4.2b — //mawi// mawi vs. GnSg /mau-i-os-z/ (2.1.4.2b) — > maujosz (2.1.5) -^> mO:jo;z (ev) —> //mOsjots// maujos. In a few instances the distributions of /i/-/j/ and /u/-/w/ are unpredictable and so we have considered them separate phonemes. E.g., iupa #above* vs. juk ’yoke*! as well as wrikan •persecute* and wlits 'face*, while forms like urikan and ulits (cf. ulbandus *camel') would cer­ tainly seem to be phonologically possible. In such cases we posit as underlying forms /iupa/, /juk/, /wrik/, /wlit/, and /ulband/. 2.1.5. Monophthongization of /ai, au/ to /E:, 0:/ ~+vocalic"l r+sonorant"' — *> "tvocalic”

+low _-long

J

+high

—high

-long

-low

o60), our rule 2.1.6 applies to all long stressed vowels, i.e. not just to /e:, o:/, but to /ii, ui, a:/ as well. Since /a:/ occurs in native Go. words only through rule 2.2.8 below, it does not occur before a vocalic segment and so cannot undergo 2.1.6. ^ Rule 2.1.6 also does not seem to apply to /ia, u*/ in any inflec­ tional paradigms, although there is etymological evidence in the occurrence of forms like trauan *trust* for /trO:-an/ from earlier /trus-an/ and bauan *build* for /bO:-an/ from earlier /bu:-an/ that the rule applied to /u:/. There seem to be no attested Go. forms with stressed / i :/ to which 2.1.6 could have applied.

Gothic

68

2.1.7. Glide insertion Optional: 0 — £

j

/ r+vocalicH ----- [“Word-internal _+stress J

H Qhvocalicj

[jmorpheme boundary^

1

2

where either vowel 1 or vowel 2 or both must be +-high and -back, i.e. /i/ or /i:/. That is, the glide / 3/ is inserted in the environment specified. The optionality is twofold: If vowel 1 is / i (:)/, the rule applies at about 90#; if vowel 2 is /i(*)/, then it applies at about 60# of the time. E.g., GnPl /þrí-e:/ #three* — > //þríje://, 1 PI Ind Prs /sí-u-m/ *we are* — > //sijum// if the rule applies and //sfum// if it does not. (Both forms occur in I Cainthians 1£:19*) And the 3 §g Ind Prs /set-i-d/ 'sow* (2 .1 .6 ) — sEiid (2.1.7) — > sEijid (ev) — > //sÉijiþ// saijib; /fr-í-i:/ ( = A stem +■ SblA -h derivative suffix 1 ÖJ •free­ dom* //friji:// frijei. The rule also applies to nonGo. proper names (with retention of the original stress): Beþanijin from , Abi jins (perhaps Abij ins) from JA/3i "—high +high -fback -long —stress _

" / ----- Ç Q-vocalicJ

j_? t-tense^

##

The

Major

Rules

(Vocalic)

That isf unstressed /u/ — ^ //9// (a tense* short / 0/) in the environment specified. E.g.* fraistobnjo *temptation* (Luke *+il3) instead of the more usual fraistubnjo* sunjos •sons* (Luke 16:18) instead of the more usual sunjus* faiho •cattle* (Mark 10*23) instead of faihu. This alternation may possibly not be the result of a phonological rule* but rather an orthographic variation. (See on this BrauneEbbinghaus 196ltl9.) If it is a legitimate phonological rule* then it applies very rarely. 2 .I.9 .

/i:/-/es/ alternations

Optional* ~-hvocalic~j —

Q-ahighj

ctfiigh -low —back +tense i 1

_+long

J

That is* long /e:/ may be realized optionally as /i:/ and /i:/ as /e:/. E.g.* qeins *woman* (Luke 2:$) instead of the more usual qens* Gnftl dalei •valleys* (Luke 5 55) in­ stead of dale* wehsa *villages* (Mark 8:26) instead of w e ihsa* ize *he who* (Mark 9*1) from /iz-i:/ instead of

Izei. According to Braune-Ebbinghaus (196l:l6* § 8 * Anm. 2), the alternation of /e:/ to /is/ occurs **ziemlich häufig**; that of /is/ to /e:/ **nicht ganz selten** (Ibid.* p. 21* §17, Anm. 1). Like rule 2.1.8* rule 2.1.9 may be reflect­ ing merely an orthographic variation. However, if these alternations are merely graphemic, it is curious that similar alternations do not occur among other Go. graphemes that are formally similar such as a-d- 1 * small e-small f* m-u, au-a* ai-a* and the like. There are still other alternations of /i/* /i*/, and /es/ as well as of /u/ and / 0/ which our rules 2.1.8 and 2.1.9 do not capture. They may reflect infrequently applying optional rules. Examples of them may be found in Streitberg (1920:^8-50) and in Marchand (1956, esp. pp. 1^4-9). 2.2. Consonantal 2.2.1. Consonantal devoicing This rule takes the following form:

Gothic

70

“fconsonant“! — > —sonorant

^-voiced!

/ ----- f [>voicedJ L ##

-hcontinuant _? -back That is# //ht z// and possibly //$// are realized as Þt St x//, respectively, in the environment specified. E.g.# GnSg hlaibis •bread* vs. NmSg hlaifs, GnSgMn frod is ‘wis e • vs. NmSgMn froþs, weizuh sw e in T = ~ P n 2 5 + Pn9 + Sb30 +- /uh/) vs. weis ( - the same without /uh/).

//ft

The fact that //x// from /g/ is always written £ is striking. As noted in the literature# the difference be­ tween the alternations /b/-/f/, /d/-/þ/, and /z/-/s/ on the one hand and that of /g/-//x// on the other is that the first three pairs of segments are independent phonemes in that the latter member of each pair can also occur intervocallys E.g., afar ‘after* vs. aba •man*, haifriwisk •heathenish* vs. haidus *type## nasjan *save* vs. hazjan •praise*. But there are no native Go. words with an Intervocalic //x// such that a contrast like *//daxa// vs. DtSg //daagyy *day# could exist. Since the contrast of //#// withJ//x// is subphonemic — the traditional argument runs — and since the scribe would naturally record only the phonemic contrasts in his language# it is to be expected that the scribe would not record the //&-//-/M / alternation. Hence one has the NmSg dags *day* for /ydaxs// and the GnSg dagis for //da^is//. However# subphonemic orthography is found on occasion in Gothic, e.g. the //rj// which arises by rule 2.2.7 below is usually transcribed with g as in briggan #bring* for //briogan//# but it is sporadically transcribed with n as bringiþ (Luke 15*22). The graphemes gg for //f]g// might well be expected since Gk. orthographic practice upon which Wulfila in great part based his system designates this sequence in this way, e.g. oiyjr^Aos *angel* for ^äijgelos//. Now in view of the facts that subphonemic orthography does occur sporadically# that x existed in the Gk. alphabet# and that this sign is frequently used in the transcription of non-Go. names like xristus 'Christ* as well as for the number *600 *# the question arises, why out of the numerous occurrences of g for /7x//, the sign x does not occur even once to represent //x//. A possible explanation for this state of affairs is that /g/ was simply not devoiced by this rule (hence the specification —back in our formulation). In that case# the rule appliod only to /b, d# z/. This would be not alto­ gether surprising since /g/ has often been seen to elude phonological processes which affect front consonants such

Th e

Major

Rules

(Consonantal)

71

as /b, d / . For example, as can be clearly seen in the attested West Gmc. languages, /b, d/ became stops wordinitially, while /g/ remained a continuant» (A reflex of this is to be seen in rule 2.2.2 below.) Another such in­ stance is the devoicing rule 1 .2 .8a # above which apparently applied to the front obstruents, but not to /g/. Yet an­ other is the rule of consonantal change in A and V stems (1.2.1) which applies to the 2 Sg Ind of Pp V*s like /wEst-t/ 'know* — //wE:st// waist, but not to those whose stems end in /g/, ogs •fear*, magt 'may*, not »ohs or ♦m a h t . If rule 2.2.1 does not apply to /g/, it may well be optional for /b, d, z/i Orthographic variants occur to which the rule may not have applied, e.g. hlaibs 'bread* (Luke 4:3), gods 'good* (Luke 6:35)» and riqiz ^darkness* (Matthew 6:23) • (A complete listing of such forms is in Streitberg, 1905-6.) There are of course two possible in­ terpretations of such forms: They may be instances of morphophonemic writing; or they may be phonetically accur­ ate, in which case rule 2.2.1 is indeed optional, Con­ sidering for the moment the latter possibility, it is in­ teresting that in Streitberg*s enumeration of the graphemic exceptions to rule 2 .2 .1 , z (instead of s) occurs ten times in the environment of rule 2 .2 ,1 , b 23 times, and d 191 times. If one assumes that the number of total occurrences of /z/, /b/, and /d/ in this environment is approximately the same for each of them, then the ratio of exceptions to total occurrences of the segment would be 10/X for /z/, 23/X for /b/, and 191/X for /d/. Thus rule 2 .2 .1 , if op­ tional, applied most often to /z/, less often to /b/, and least often to /d/. If this interpretation of the graph­ emic evidence is correct, then the rule of obstruent de­ voicing so prevalent among Gmc, languages probably began by applying optionally to front continuant consonants, later applied to back ones, and finally in a number of Gmc, languages applied to both continuant and stop consonants. 2.2.2. Distribution of voiced stop and continuant conso­ nants prconsonant- — > -sonorant

_-hvoiced

L+continuant^)

/ a. L+vocalicJ ----b. ? ##

J

-hback^ I

_---- J

That is, (a) /b, d, g/ ~ > //fc, cf, %// after a vocalic segment; and (b) /g/ — ^ //$// word-initially » E.g., Inf /bindan/ //bindan//, but the GNSg /hlE:bis/ 'bread* //hlE:feis//, and the DtSg /##ga##de:-d-E:##/ 'deed* ( = prefix + stem + derivative suffix 11 +-Sb6) by 2.2.2 — >> //#^.a##de :-d“- E *##//. The geminates in triggws ’true* and

Gothic

72

daddjan *suckle* are //tri^gws// and //daddjan//. It is also possible that these sequences were //gg// and //dd//. The formulation of part (a) of this rule is based on the assumption that the rule of consonantal devoicing 2.2.1 affects only continuant consonants. Hence band *tied* (not »taanþ) is presumably //band//, gazds •goad* (not *gasbs) //cjazds//, and gahugds •thought* (not »gahugfrs) //^ahugds//. There is no Go. orthographic evidence for part (b): It is based solely on the evidence from the other Gmc. languages in whose earliest attested stages /g/ in this environment seems to have been //q //. (See also on this Moulton 19^8:77.) Finally, it is possible to describe the alternations of this rule as the realization of underlying continuant segments /b, j 9 %/ as stops //b, d, g// in the appropriate environments, (a; Q—vocalicJ ----- and (b) ## back, ----- H • However, such a rule would need the extra feature -striaent in its formulation to exclude /z/ from its domain. 2.2.3* Geminate-consonant simplification "’-^consonant""!

pfconsonant^

^features

|_j*features

/ a. ~+sonorantH Obligatory: T+consonant

_--

J

~~

-sonor ant _-f-continuant _ Optional:

pHconsonant, except for the~ ^immediately preceding.

_

b. [^consonant!! ----That is, (a) geminate sonorant consonants are simplified obligatorily before a continuant obstruent and optionally before other types of consonants; and (b), all geminate consonants, sonorants as well as obstruents, are simplified when following a consonant. E.g., (a) NmSg /mann-a/ •man* //manna//, DtSg /mann/ //mann//, but Nm or Ac PI /mann-z/ (2.2.3) — $> manz (ev) — //mans//j the V /full-n 4- endings/ •fill* ( » stem + derivative suffix 4-1) is realized both as fulnand fulln- ; the 2 Sg Prs Ind of the Pp V /kann-t/ •can* Ts realized both as kant and kannt; (b) NmSg /##ab##staþ-tz##/ •apostasy* ( = prefix +■ stem + derivative suffix 52

The

Major

Rules

(Consonantal)

73

+ Sb31) by 1.2.1 ##ab##stas-t-z## (1.2,8) ##ab##s tas s-z## (2.2.1) — > ##af##stass-s## (2.2.3) //afstass//, aldom 'old age*, probably from /ald-do:m/ ( = stem 4- derivative suffix 12), NmSg /hals-z/ *neck* ( = stem + Sb31) by 2.2.1 halss (2.2.3) //hals//. Braune-Ebbinghaus (1961i57-8, §80) remark that gemi­ nates are **...in der Regel vereinfachti kant, kunba, rant p Sg PstJ » urruns...dagegen gewöhnlich fullnan, nur einige Male fulnan.** Thus the optionality in part (a) may apply more often to the nasal resonants /nn/ and /mm/ than to /11/ or /rr/. 2.2.^. Consonant-cluster simplification Optional: Qhc on sonant]] —>

0

/ Qfvocalic]] Qf-consonant]] ----- Qhconsonant]] E.g., fimtiguns *fifty* from /fimf-tig-u-n-z/, timrjan •build* from /timbr-i-a-n/. 2.2.5. /h/-assimilation Optionali r+consonant~ ^

[jxfeatures]

_+low / ----- ## T-consonant” jxfeatures E.g., /##ja h##þan##/ *and then* -$> ja£ ^an or jah þan, /##jah##lib-a-n##/ *and to live* jal liban or jah liban, /##nuh##kann-t##/ *neither do you know1" - > nuk kannt or nuh kannt, but not huhrus 'hunger* as »hurrus since there is no word boundary ## between the /h/ and the /r/. This rule applies less than half the time. It applies most frequently if the /h/ precedes a particle or a pronoun (probably unstressed) beginning with /þ/, e.g. /##jah##þ e :##/ jaþ b e . 2.2.6. /h/-deletion /h/ — >

0

/ a. Obligatory: -----

vocalic

b. Optional: Qt-vocalic, -stress]] ----- ## c. Optional: ----- p-vocalic]]

Gothic

74

E.g., (a) /work-stw/ *work# ( = stem + derivative suffix 50) by 1.2.1 — > worhstw (2.2.6) //worstw// waurstw, (b) hi ileiku #what kind of* from /hw ili:k-uh/j (c) läibos *loaves1^ (Mark 8:8) from / h l E :b-o:-z/, hiuma •crowd* (Luke 6:17) instead of the more usual hiuhma. 2.2.7. Nasal assimilation L+nasal]

Rxcoronal "1 / ----- ~t-consonant /3ant er ior _^back

-s onorant

J

o^coronal /3anterior ^back Except for: -bcoronal _+strident, i.e. /s, z/_

where no morpheme boundary of any kind occurs between the two segments in the environmental statement. That is, any nasal consonant is assimilated to an immediately following (and with no intervening morpheme boundary) nonsonorant consonant, except for /s, z / . E.g., gaggan *to go* //^aQgan//, fimf *five*, f inþan *find*. But the rule does not apply over a morpheme boundary: andanumts •acceptance* from /##anda##num-t-z/ ( = prefix + stem + derivative suffix 52 + Sb3l), not »andanunts. It also does not apply if the following segment is /s/ or /z/: ams •shoulder* not »ans, also gramsta *piece of wood* and þramstei •locust1 Cassuming for these cases that no morph­ eme boundary intervenes between the /m/ and the /s/)• Finally, nasals do not assimilate to following sonorant consonants: namnjan •name* not *nannjan, and simle •once* not »sinle. In accordance with Gk. orthographic practice, //q// is usually written g; briggan *bring* for //briqgan//, igqis •you two* for //irjkw is/7. But, as mentioned under rule 2.2.1 above, forms written with n for //q// also occur: bringan (Luke 15*22), inqis (Luke 19:31), banke ip *thank* (Luke Ik :31) . 2.2.8. Nasal deletion and vowel lengthening

The foï'm taken by this rule is the following:

The

Major

Rules

L+vocalicJ n+-nasal]] /h/ — >

(Consonantal)

[“V-vocalic”1 0 /h/

J

_-hlong 1

2

75

3

1

2

3

That is, the sequence vowel + nasal consonant + /h/ must be realized as the corresponding long vowel -+- /h/. E.g., 1 Sg Ind Pst /þunk-i-d-a/ •seemed* ( = stem +- Vbl7 + Vbl2 + Vb3) by 1.1.4 — > bunkda (1.2.1) — > þunhda (1,2.8) þunhta (2,2.8) — £> //þuihta//, /##un##at##gang-t-z##/ 'unapproach­ able* ( » derivative prefix 10 + preposition ■+■ V stem *go* 4- derivative suffix 54 + Sb31) by 1.2.1 — > ##un##at##ganh-t-z## (2.2.8) — > ##un##at##gaih-t-z## (ev) //una t^aiht s / / . The vocalic segment 1 may have been -hnasal by this rule as well as -f-long. This seems to have been the case in at least one other Gmc. language (see on this Haugen, 1950). 2.2.9 /n/-deletion n —>

0

/ p5b20 morpheme"! ["DtPl, Sbl9 morpheme”

_____

__ / V

E.g., DtPl /gum-a-n-m/ 'man* ( = stem t* Sbl + Sb20 -+■ Sbl9) gumam (ev) — > //gumam//. This rule may possibly be completely phonologically conditioned in that /nm/ may always be realized as /m/ word-finally, 2.2.10 /z/-to-/r/ rule ”-hconsonant”’ — ?> r+sonorant~ / "“Obligatoryi Prefix -hcoronal _-i-strident _

—tense

Optional i Prepo-

+voiced

sition

strident_

[_-----

##/r/

_

That is, /z/ (or /s/ if rule 2.2.1 is considered to have applied) is realized as /r/ obligatorily if in a prefix followed by /r/ and optionally if in a preposition followed by /r/. E.g., /##uz##run-z##/ ’exit* ( = prefix + stem ■+ Sb31) by 2.2.10 —$> urrunz (ev) — ^ //urruns//. In one instance this rule applies with the preposition /uz/, ur riqiza 'out of darkness’ (II Corinthians 4t6) instead of *us riqiza.

Gothic

76

Appendix 1 Parad igms The paradigms under 1.- below include the N endings as generated by the Sb-rules in section 4.1 above in this chapter. The paradigms under 2.- are the A*s and numerals (also generated by the Sb-rules). The paradigms under 3*are the pronouns as produced by the Sb-rules and the Pnrules in section 4.2 above. Finally, the paradigms under 4.- include the V endings as produced by the Vb-rules (section 4.3) and theBe-rules (section4.4). The forms are given in terms of the systematic phonemes (figure A in section 2 of this chapter). 1. Nouns (Sb-rules) 1.1. Mn, a-class 31 NmSg dag z (ev) — £ 15 24 GnSg dag i z 2 DtSg dag a

23 30 dag oi z 11 GnPl dag e* 3 19 DtPl dag a m 21 30 AcPl dag a n z

dags fd a y f.NmPl

AcSg dag 1.2. Mn, ja-class !4 31 l6 NmSg herd i z (1.1.3) —> 15 29 GnSg herd i i z

herdiiz (ev) — > //herdiis// hairdeis #shepherd*.

2 DtSg herd i

a (2.1.4) — >

//herdja//.

AcSg herd i NmPl herd i

23 30 oi z

11 GnPl herd i DtPl herd i

et 3 19 am

AcPl herd i

a n

21 30 z

Short-syllable stems of this class have the same morphology as /herd/, but are realized differently by phonological rules 1.1.3 and 2.1.4i NmSg har-i-z — //har jis// •army•.

Appendix

1

77

1.3» Nt, a-class

2k NmSg word

waurd ’word*.

GnSg word i 2 DtSg word a

z

NmPl word o* (1.1.7) -£> GnPl word el //worda//* 3 19 DtPl word a m

2k AcSg word

AcPl word os (like NmPl).

1.4. Nt, ja-class Like paradigm 1.3 above except that rule Sbl4 applies, e.g. 14 NmSg kun i ’tribe*. 1.5. Fn, os-class

26 NmSg gib 0 : (1.1.7 and ev) -£> //gifea// ’gift*. GnSg gib 0 : z

6

DtSg gib Es —>

gibai.

26 AcSg gib o: (1.1.7 and ev) — >

30 NmPl gib os z 25 GnPl gib os

//aifea//. J 26 19 DtPl gib os m 30 AcPl gib os z

1.6. Fn, jos-class Like paradigm 1.5 above except that rule Sbl4 applies, 14 26 e.g. NmSg band i os (1.1.8) — > //bandi// ’fetter*, 14 26 26 GnSg band i os z (1.1.8) -^> //bandjo:s //. 1.7. Mn, i-claas 31 NmSg balg z — 15 29 GnSg balg i z 2 DtSg balg a

balgs ’bag’.

17 30 NmPl balg is z — > 11 balge is. GnPl balg e s 13 19 DtPl balg i m

21 30 AcSg balg

AcPl balg i

1.8. Fn, i-class 31 NmSg anst z ’favor*. NmPl anst 6 29 GnSg anst Es z —&■ anstais. GnPl anst

n

17 30 is z — 11 ei

z

ansteis.

Gothic

78

6 DtSg anst Es

13 19 DtPl anst i m

AcSg anst

AcPl anst i

21 30 n

z

1.9. Fn, i- and os-class N*s with the syntactic structure V stem (j-class Wk) - /in/ ( = derivative suffix 16 = Vbl7 + Vb21) + N endings form their Sg according to paradigm 1.8 and their PI according to 1.5, e.g. NmSg naiteins •blas­ phemy*, NmPl naitelnos. (The N /hEsm/ 'village* also inflects according to this paradigm. Similarly de­ rived verbal N*s from the other Wk V classes inflect according to paradigm 1.8s NmSg lapons *invitation*, NmPl laþoneis.) 1.10. Mn and Fn, u-class 27 10 31 NmSg sun (u or Os) z 29 GnSg sun (u or Os) z DtSg sun (u

or Os)

sunus or sunaus *son*.

sunu or sunau.

AcSg sun (u or Ot) 13 27 30 NmPl sun i u z (2.1.4)

//sunjus//.

11 GnPl sun i

u

DtPl sun

u

e: (2.1.4) //suniwei//. 19 m —$> //sunum//.

AcPl sun

u

n

21 30 z

1.11. Nt, u-class 27 10 NmSg feh (u or 0s) — > 29 GnSg feh (u or 0s) z DtSg feh

(u or 0s)

AcSg feh

(u or 0:)

faihu or faihau ‘money*.

The PI forms of this class are not attested. 1.12. Mn, n-class

1

1 20 30

NmSg gum a *man*.

12 GnSg gum i

NmPl gum a n

z

20 29 n

z

GnPl gum a n

11 es

Appendix

1

79

12 DtSg gum i n

20 QtPl gum a

n

1 20 AcSg gum(i or a) n

AcPl gum a

n

1 20 19 m (2.2.2, 2.2,9) 30 -^> //tfumam//. z J

The N /ab/ •man* inflects according to this class except that it does not have Sbl in the PI. The attested forms are GnPl abne and DtPl abnam ( = stem +■ Sb20 +* Sb3 f Sbl9). A subclass of 1.12 is the Mn j-n-class which is like 1.12 except that rule Sbl4 applies« NmSg wilja •will1 ( = stein +■ Sbl4 +■ Sbl), GnSg wil jins ( - stem -t- Sbl4 -+ Sbl2 + Sb20 •+■ Sb29) . 1.13. Nt, n-class

22 NmSg hert o« fheart# hairto.

12 20 29 GnSg hert i

n

DtSg hert i

n

z

22 AcSg hert o« 20 24 NmPl hert o« n oi (1.1.7) — >

1 GnPl hert a

//hertoina//,

11 n

e« 19 DtPl hert a n m (2.2.9) — ?> //hertam//, 22 24 AcPl hert o« n o« (1.1.7) — * //herto ma//. The N #s /nam/ •name* and /wat/ 'water* are like /ab/ immediately above in that they do not have the Sbl or Sb22 rules in the PI, Hence the PI of /nam/ is namna, namne, namnam. The N / f o m / ’fire* does not have the Sb22 rule in the Sg. 1.14. Fn, n-class

22

22 20 30

NmSg tung oi •tongue* NmPl turig o« n 20 29 GnSg tung o« n z GnPl tung oi n DtSg tung oi n

DtPl tung oi n

AcSg tung o« n

AcPl tung o« n

z 25 oi 19 m (2.2.7» 2.2.9 )*> 30//tu go in//. z

1.15« Fn, i«-n-class This paradigm is like 1.14 except instead of Sb22 / 0 1 / , rule Sbl8 /ii/ applies« E ,g., NmSg

18 18 20 29 manag i« 'crowd*, GnSg manag i« n z — >

manage ins.

Gothic

30

I.l6. Mn, Fnf Nt, r-class NmSg bro:þr (1,1.1) — > //bro:þar// ’brother*. 29 GnSg bro:þr z DtSg bro:þr AcSg broiþr (1.1.1) — > //bro:þar//. 13 2? 30 NmPl broiþr i u z (2.1.^, 2.2.1) —

//bro ;þr jus//.

11 GnPl bro:þr e* 27 19 DtPl bro:þr u m

21 30 AcPl broJþr u

n

z

1.17• Mn, nd-class 31 z •saviour*. 15 29 GnSg nasjand i z NmSg nasjand

DtSg nasjand AcSg nasjand

30 z 11 GnPl nasjand e: 3 19 DtPl nasjand a m 30 AcPl nasjand z

NmPl nasjand

1.18. Fn, consonant class 31 NmSg borg z — > baurgs *city*. 29 GnSg borg z DtSg borg AcSg borg

30 NmPl borg z 11 GnPl borg e : 13 19 DtPl borg i m 30 AcPl borg z

The N /naht/ #night* is in this class except that the DtPl is nahtam ( = stem + Sb3 + Sbl9). The N*s /weht/ ’thing* and /dulþ/ *festival* can follow either this paradigm or 1.8. 1.19» Mn, mixed class 1 1. NmSg mann a *man*. 29 GnSg mann z DtSg mann

1 20 AcSg mann a n 30

NmPl mann (a n or 0) z —$> //mannans// or //mans//.

Appendix

1

81

11 GnPl mann e s 3 19 DtPI mann a m 1 20 30 AcPl mann (a n or 0) z (same as NmPl). 31 30 2. NmSg risk z'ruler*. NmPl risk z 15 29 11 GnSg risk i z GnPl risk es 3 19 DtSg risk DtPI risk a m 30 AcSg risk AcPl risk z The N /mesnosþ/ 'month* is in this class except that the DtPl is menoþum ( = stem + Sb2? 4- Sbl9). 2. Adjectives (Sb-rules) 2.1.

Strong declension 31 NmSgMn blind z 'blind'. 15 29 GnSgMn blind i z 7 DtSgMn blind amma

5 NmPIMn blind E;

blindai. 28 11 GnPIMn blind Es z es 19 DtPIMn blind Es m

8

3 21 30

AcSgMn blind ana AcPIMn blind a n z 26 NmSgFn blind os (1.1.7) //blinda//. 6 28 26 29 GnSgPn blind Es z os z (ev) — > //blindEszoss//. DtSgFn blind Es 26 AcSgPn blind o: (same as NmSgFn). 30 NmPIFn blind os z 5 28 25 GnPIFn blind Es z os 19 DtPIFn blind Es m 26 30 AcPIFn blind os z 9 NmSgNt blind 0 or ata NmPINt blind 15 29 GnSgNt blind i z GnPINt blind 7 DtSgNt blind amma DtPINt blind

AcSgNt

9 blind 0 or ata

2k o: (1.1.7) — 5 28 ll//blinda// Es z es 19 Es m

2k AcPINt blind

os (same as NnPlNt)

Gothic

82

2.1.1. Strong declension, j-class 14 31 NmSgMn mid i z (1.1.3» 2.1.4, ev) — ^ fr i

z

(1.1.3» ev) — >

//micíjis//

//fri:s// freiè^

wilþ i z (1.1.3» ev) -y> //wilþi:s// iree 14 15 29 fwild\ GnSgMn mid ii z (2.1.4, ev) — > //micTjis//. fr

z

(1.1.3» ev) — >

//fri:s//.

(1.1.3» ev) — >

//wilþits//.

mid

ii z 7 i amma

(2.1.4, ev) — $> //mid"jamma//.

fr

i amma

(2.1.7) — >

//frijamma//.

wilþ

i amma

(2.1.4) — >

//wilþjamma//.

wilþ DtSgMn

ii

That is, this paradigm is the same as 2.1 except that rule Sbl4 applies. 2.1.2. Strong declension, i-class 14 31 NmSgMn sut i z (ev) — > //sutis// fsweetf. hrE:n i GnSgMn

sut i hrE:n i

DtSgMn

sut i hrEsn i

z (1.1.5» ev) -^> //hrE:ns// hrains 29 'pure*. z — //sutis//. z — > //hrE:nis//. 7 amma (2.1.4) //sutjamma//. amma (2.1.4) — >

//hrE;njamma//.

This paradigm is the same as 2.1.1 except that Sbl4 does not apply in the GnSgMn or Nt. 2.1.3. Strong declension, u-class. 27 31 NmSgMn hard u z (ev) — //hardus// *hard*. 15 29 GnSgMn hard u i z (1.1.9) — > hardiis, realized either as //hardi:s// if 1.1.3 applies or as //hardjis// if 1.1.3 does not apply and 2.1.4 does. 27 7 DtSgMn hard u amma (1.1.9» 2.1.4) — //hardjamma//. 9 NmSgNt hard u (0 or ata) — $> //hardu// or by 1.1.9 //hardjata//.

Appendix

1

83

2.2. Weak declension, n-class This class has the same endings as the N paradigms 1.12, 1.13» and 1.14 above, e.g. 1 NmSgMn blind a —£> //blinda//. 14 mid i a (2.1.4, ev) //midja//. sut i

a (2.1.4) — >//sutja//.

hrE:n i a (2.1.4) — > //hrEsnja//. 27 hard u a (1.1.9» 2.1.4) ~ > //hardja//. 2.3# Present participles

1

31

NmSgMn giband (a or z ) ’giving*.

18 NmSgFn giband i:

Otherwise this paradigm is like 2.2 above except that Sbl8 /i:/ instead of Sb22 /o:/ applies to the Fn forms. 2.4.

Numerals

2.4.1. /tw/ ’two* 4 11 GnPIMn tw addj e:

24 NmPINt tw o : (1.1.7) — >

25 GnPIFn tw addj o:

AcPINt tw o: (same as NmPINt).

//twa//.

11 GnPINt tw addj e: Otherwise this paradigm has the same endings as 2.1; /tw/ occurs only in the PI. 2.4.2. /þri/ ’three* 17 30 NmPIMn þri i: z (2.1.1, ev) — >

//þri:s// breis.

11 GnPIMn þri e: (2.1.7) — ?> //þrije://. 19 DtPIMn þri m

21 30 AcPIMn þri n 26 NmPIFn þri os 11 GnPIFn þri e: 19 DtPIFn þri m 26 AcPIFn þri o; 24 NmPINt þri o:

z z (2.1.7» ev) — >

//þrijo:s//. b 25 (presumably, possibly þri os).

30 z (same as NmPIFn). (1.1.7, 2.1.7)

//þrija//.

Gothic

84

11 GnPlNt þri e ; 19 DtPlNt þri m 24 AcPlNt þri os (same as NmPINt). 3.

Pronouns (Pn- and Sb-rules)

3.1.

First Person Pnll NmSg ik Pnl6 Pnl2 Pnl GnSg m i:n a — meina. Pn8 Pn29 DtSg m i z Pnl5 AcSg m i k Pn25 Pn22 NmDu u i t (2.1.4) //wit//. Pnl7 Pn4 Pnl GnDu u nk ar a Pnl 3 DtDu u nk iz AcDu u NmPl u GnPl u DtPl u AcPl u

nk iz Pn9 Sb30 i: z (2.1.4, ev) — //wi:s// weis. Pnl9 Pn4 Pnl ns ar a Pnl 3 ns (0 or ïz ) by 2.2.1 / / u n s / / or //unsis//. ns ( 0 or iz )

3*2. Second Person Pn23 Pn27 NmSg þ u: Pnl2 GnSg þ i:n Pn26 DtSg þ u

Pnl a Pn29 z Pnl 5 AcSg þ u k Pnl4 Pn22 NmDu j u t Pnl8 Pn4 Pnl GnDu j nkw ar a Pnl 3 DtDu i nkw iz (2.1.4, 2.2.1, 2.2.7) — i> //inkwi s// igqis. AcDu j nkw iz Pn6 Sb30 NmPl j u z

Appendix

GnPl j DtPl j AcPl j

1

Pn30 Pn4 Pnl zw ar a Pnl 3 zw iz (2.1.4, 2.2.1) zw

85

//izwis//.

iz

3*3» Reflexive Pn21 Pnl2 Pnl Gn s i:n a

Pn21 Pn8 Pn29 Dt s i z

Pn21 Pn8 Pnl5 Ac s i k

3 A . Third Person NmSgMn GnSgMn DtSgMn AcSgMn NmPIMn GnPIMn DtPIMn AcPIMn

PnlO Sb31 i z (2.2.1) — //is//. Sb29 i z Sb7 i amma (2.1.2) //imma//. Sb8 i ana (2.1.2) — > //ina//» Sbl? Sb30 i i: z (2.1.1» 2.2.1) Sb28 Sbll i z e: Sbl9 i m Sb21 Sb30 i n z

//its// e is.

Pn20 NmSgPn s i GnSgPn

i

DtSgFn

i

AcSgFn

i

NmPIFn

i

GnPIFn

i

DtPIFn

i

AcPIFn

i

NmSgNt

i

GnSgNt

i

DtSgNt

i

AcSgNt

i

NmPINt

i

Sb28 Sb26 Sb29 z oi z Sb6 z E: Sb26 o: (1.1.7» 2.1.7) //ija//. Sb30 oi z (2.1.7, 2.2.1) /ijo:s//. Sb28 Sb25 z o: Sbl9 m Sb26 Sb30 o: z (same as NmPIFn). Sb9 ata (2.1.2) //ita//. Sb29 z Sb7 amma (2.1.2) — //imma//. Sb9 ata (same as NmSgNt). Sb24 o: (1.1.7, 2.1.7) — > //ija//.

Gothic

86

GnPINt DtPINt AcPINt

PnlO Sb28 Sbll i z es Sbl9 i m Sb2i+ i os (same as NmPINt).

3.5. Definite Article Pn20 Sbl NmSgMn s a Pn2l* Sbl5 GnSgMn þ i Sb? DtSgMn þ amma Sb8 AcSgMn þ ana Pn20 Sb26 NmSgFn s os Pn2^ PnlO GnSgPn þ i DtSgPn þ AcSgPn þ NmPIPn þ GnPIPn þ DtPIPn þ AcPIFn þ NmSgNt þ GnSgNt þ DtSgNt þ AcSgNt þ

i Sb26 0: o: PnlO i Sb5 Es Sb26 os Sb9 ata Sbl5 i Sb7 amma Sb9 ata

Pn2^ Sb5 þ Es PnlO GnPIMn þ i Sb5 DtPIMn þ Es Sb3 AcPIMn þ a NmPIMn

Sb29 z

Sb28 * Sbl9 m Sb21 n

Sbll e; Sb30 z

Sb28 Sb26 Sb29 z os z Sb6 z Es Sb30 z Sb28 Sb25 z os Sbl9 m Sb30 z Pn2*+ Sb2k þ o: PnlO GnPINt þ i Sb5 DtPINt þ Es

NmPINt Sb29 z

Sb28 Sbll z e; Sbl9 m

SbZk AcPINt

3.6. Demonstrat ive Pn20 Sbl NmSgMn s a uh (2.1.2) Pn2*+ Sbl5 Sb29 GnSgMn þ i z uh Sb? DtSgMn þ amma uh (2.1.2)

þ

os

//sah//. //þizuh//. //þammuh//.

That is, this is the same .paradigm as 3*5 with the enclitic particle /uh/. 3.7. Relative Pn20 NmSgMn s

Sbl a i:— >

//sai:// saei.

Appendix

PnlO NmSgFn s i: (2.1.1) —

1

87

//si:// sei.

That is, this is the enclitic particle /i:/ added either to the forms of paradigm 3 »5 or on occasion to those of paradigm 3*^. 3.8.

Interrogative Pn? Sb33 NmSgMn hwa z Sbl5 GnSgMn hwa i Sb7 DtSgMn hwa amma Sb8 AcSgMn hwa ana Sb26 NmSgFn hwa o: PnlO GnSgFn hwa i

Sb29 z (2.1.2, 2.2.1) - > (2.1.2)

//hw is//.

//hwamma//.

(2.1.2) — ^ //hw o://. Sb28 Sb26 Sb29 z o: z (2.1.2, 2.2.1) — > Sb6 //hw izo:s//. DtSgFn hwa i z E: Sb26 AcSgFn hwa o: (same as NmSgFn). NmSgNt hwa Sbl5 Sb29 i z Sb7 Pn6 DtSgNt hwa (amma or e : ) by 2,1.2 — GnSgNt hwa

//hwamma// or //hw e ://.

AcSgNt hwa 3.9« Indefinites Sb31 3.9.1. NmSgMn E:n z hun (2.2.1) — >

//Esnshun// ainshun

(occurring with ni). The endings of this paradigm are like those of 2,1 except for the following: Pn28 DtSgMn or Nt E:n u m m e : hun Pn3 AcSgMn E:n ano: hun (1.1.2) //E inno :hun// or //Esnoihun//, Pn7 Sb33 3.9.2. NmSgMn hwa z uh 'every* The endings of this paradigm are like those of 2.1 except for the following: Pn7 Pn2 DtSgMn or Nt hwa a m m e : uh (2.1.2) //hwamme:h / / . Pn3 AcSgMn hwa ano: uh (2.1.2) //hwano:h//.

Gothic

88

3.9.3.

Sbl4 Sb31 NmSgMn hwar i z uh (1.1.3* 2.1.4) — £• Sbl5 Sb29 //hwarjizuh// 'every*. GnSgMn hwar i i z (2.1.4) //h^ferjizuh//. The endings of this paradigm except for the following: Sbl4 Pn2 DtSgMn or Nt hwar i amme: Pn5 Nm or AcSgNt hwar i ato: Pn3 AcSgMn hwar i anos

4.

are like those of 2.1 uh (2.1.2* 2.1.4) —$> //hwar jamme ;h / / . uh (2.1.2, 2.1.4) — £> //hwarjato sh / / . uh (2.1.2, 2.1.4) — $> //hwarjano:h //.

Verbs (Vb-rules)

4.1.

Strong verbs

2 PrsActlndSgl nim a 18 31 2 nim i z 13 3 nim i d (2.2.1, 2.2.2) — $> //nimiþ//. 26 Dul nim o:s 1 28 2 nim a ts

20 Pll nim a m 18 14 2 nim i d

1 22 13 3 nim a n d — >> //nimand//. 10 7 28 SubSgl nim 0: - > nimau. Du2 nim Es ts 7 31 20 4 2 nim E: z - > nimais. Pll nim Es m a 14 3 nim Es 2 nim Es d 30 22 4 Dul nim Es wa 3 nim Es n a 1 20 ImpSg2 nim Pll nim a m 1 13 9 18 14 3 nim a d 0s 2 nim i d 28 1 22 13 9 Du2 nim a ts 3 nim a n d 0: -5> nimandau.

21 Inf nim a n 24 Prt nim a nd + Sb-rules for A's.

Appendix

5 11 5 PsvIndSgl nim a d a 31 2 nim a z a

1

89

7 11 SubSgl nim E« d 3 2 nim Et z

11

9 0* 1 Oi

11

3 nim a d a 23 Pli nim a nd a

3 nim E« d 23 Pll nim Ei nd

Oi 0«

2 nim

a nd a

2 nim Ei nd

Oi

3 nim

a nd a

3 nim Ei nd

Oi

29 PstActlndSgl nam

— > nimaidau -------- *

Du2 n e im u

27

28 ts

20

2 nam t

Pll n e im u

3 nam

m 14 d 22 n

2 nein u 29 Dul n e im u 3 ne tm u 19 10 SubSgl ne im it Oi (1*1.6, 2.1.4) 31 2 n e im it z 3 neim it (1.1.6) 30 Dul n e im it wa

//neimjOi// nemjau.

//neimi//.

28 2 netm it ts 20 4 Pll ne tm i t m a 14 2 netm it d 22 4 3 netm it n a PsvPrt

num ■+■ Sb morphemes for A*s.

The PstActInd 2 Sg or V #s whose stems end in a vowel 31 27 has the Vb31 morpheme, e.g. sesot z t (2. 2 . D //sesotst// saisost *sowed#. The Vbl7 morpheme /i/ occurs in the ft's of certain St V*s (listed under rule 17 1 21 Vbl7)# e.g. the Inf bid i a n (ev) //bi jan// •ask*. 4.2.

Weak verbs

4.2.1.

j-class 17 The Prs forms are nas i 'save' *+■ the same endings as those of the St V paradigm 4.1. The 2 Sg Imp

Gothic

90

17 is nas i (1.1.3) -^> //nasii// nasei.

17 12 3 PstActlndSgl nas i

d

a 15 31 et z 3 a

2 nas i

d

3 nas i

d

Dul nas i

d

nas i

d

eidH + the corresponding Psttense endings of the e«d St V paradigm 4,1.

Pli nas i

d

eid_

16

Etc., down to 17 12 nas i d -t the Sb morphemes for A #s.

PsvPrt 4.2.2. oi-class

25 2 PrsActlndSgl salb oi a (2.1.2) //salboi// #anoint#. 18 31 2 salb ot i z (2.1.2, 2.2.1) //salbois//. 13 3 salb oi i d (2.1.2, ev) — //salboiþ//. That is, this class has the same endings as class 4.2.1 except that instead of Vbl7 /i/> rule Vb25 / 01 / applies. Phonological rule 2.1.2 also applies, as in the followingi

25 26 PrsActlndDul salb oi oss (2.1.2) — > //salbois//. 7 SubSgl salb oi Ei (2.1.2) //salboi//. 31 2 salb oi Ei z (2.1.2, ev) — > //salbois//. 3 salb oiEi (2.1.2) //salboi//. 5 H 5 PsvIndSgl salb oi a d a (2.1.2, ev) //salbo ida//. 7 11 9 SubSgl salb oi Ei d Oi (2.1.2, ev) — £> //salbo icfoi// 4.2.3. Ei-class 2 PrsActlndSgl hab a

6

1 28 Du2 hab a ts

31

2 hab Ei z 13 3 hab E i d

26 Dul hab ois

20 Pll hab a m 6 14 2 hab E t d

1 22 13 3 hab a n

d

Appendix

10

3 hab E: 30 Dul hab E * wa 6 ImpSg2 hab Ei 1 13 9 3 hab a d 0: 28 Du2 hab a ts

4.2.4.

Du2 hab E: ts 20 4 Pli hab E: m a 14 2 hab E: d 22 4 3 hab E s n a 1 20 Pll hab a m 6 Ik 2 hab Es d 1 22 13 9 3 hab a n d 0:

Inf

hab a

Prt

24 hab a nd + the Sb morphemes for A*s. The same endings as the St V paradigm 4.1.

6 Pst

91

7 28

SubSgl hab Oi 7 31 2 hab Es z

Psv

1

12

hab Es d

+ the same endings as the Wk V paradigm 4.2.1.

nos-class The Prs has the same endings as the St V paradigm 4.1 above (but without a Psv); and the Pst has the same endings as Wk V os-class 4.2.2. E.g.,

PrsActlndSgl full n a *become full* ( = stem + derivative suffix 41 + V b 2 ). 25 12 3 PstActlndSgl full n o: d a 4.3. Preterite-present verbs PrsActlndSgl wE:t ~ > wait •know*. 27 2 wEst t (1.2.1) //wEsst//.

29 28 Du2 wit u ts 20 Pll wit u m

14 3 wEst

2 wit u

29 Dul wit

22

u

i9

d

3 wit u

n

10

SubSgl wit i: 0: (1.1.6, 2.1.4) -^> //wit.jQ:// witjau, 31 2 wit i: z (2.2.1) -^> //witi:s// witeis. 3 wit i: (1.1.6) — >> //witi// witi. 30 Dul wit i: wa 28 2 wit i : ts

Gothic

92

19 20 4 Pll wit i: m a 14 2 wit i : d 22 4 3 wit i: n a ImpSg2 wit (possibly)

1 21 Inf

wit a n 24 wit a nd + the Sb morphemes for A*s.

Prt

12 3 PstActlndSgl wit d

a (1.2.1# 1.2.7) — $> //wissa//.

16 Pll wit d

29 20

e:d u

m

(1.2.1, 1.2.7, ev) //wisse «4um//. That is, this class has the same endings in the Pst as those of the Wk V paradigm 4.2.1 above, but without the rule Vbl7 /i/. Also, no PstPrt is attested: It was presumably wit d ( « Vbl2) + the Sb-rules for A #s (1.2.1, 1.2.7) — > //wiss//. One Prslmp 2 Sg form is attested with the Vb31 ending, o:g z (2.2.1) — > //osxs// ogs *fear*. 4.4.

The verb •be* (Vb- and Be-rules) ^ Bel Be2 PrsIndSgl i m

Be4 29 28 Du2 si u ts

31

20

2 i

z (2.2.1) //is//. Be3 13 3 i s d (1.2.7) — > //ist//. Be4 29 Dul si u (2.1.7) — > //siju//•

Pll si 2 si 3 si

u

m 14 u d 22 13 n d

10 SubSgl si 2 si

0: (2.1.7) //sijO:// sijau. 7 31 E: z (2.1.7» 2.2.1) -^> //sijBis// sijais.

3 si

E:

Dul si 2 si Pll si 2 si 3 si

30 E: wa 28 E: ts 20 E: m 14 E: d 22 E: n

4 a 4 a

In the Pst and the other forms of the Prs this V is /wis/ which follows the St V paradigm 4.1 above.

Appendix

1

93

4.5. The verb •will* 19 10 PrsIndSgl wil i; 0t (1.1.6, 2.1.4) — ^> //wiljOx// wiljau. 31 , ~ 2 wil i: z (2.2.1) //wiliis// wileis. 3 wil i: (1.1.6) — ^ 30 Dul wil i : wa 28 2 wil i: ts

20

//wili// wili.

4

Pll wil i: m a 14 2 wil is d 22 4 3 wil i: n a In all other forms /wil/ is a Wk j-class V and follows paradigm 4.2.1 above. 4.6. The verb *go* For the Prs and PstPrt forms, the stem is /gang/ which follows the St V paradigm 4.1 above. For all the Pst forms — except the Prt — the usual stem is /iddj/ which follows the j-class Wk V paradigm 4.2.1 above, except that rules Vbl2 /d/ and Vbl7 /i/ do not apply. Examples of such forms are : 3 PstlndSgl iddj a

16 Pll iddj e:d u

29 20 m

As a Pst form the stem /gang/ is found once (Luke 19*12). It follows the j-class Wk V paradigm 4.2.1. E.g., 17 12 3 PstlndSgl gang i d a

Gothic

94

Appendix 2 Rule Ordering In the following overview of the precedence relations obtaining among the phonological rules, the symbol — means "must precede" and 0 "no rule". Each ordering is numbered? and derivations illustrating each numbered order­ ing are given below. 1.1.1 Ablaut

— $> (1) 2.1.3 Breaking. (2 ) 2 .1 . 5 Monophthongization of /ai, au/ to /Et, Oi/• (3) 2.2.8 Nasal deletion and vowel lengthening.

1.1.2 Deletion from /anoi/ (the Pn3 morpheme) 1.1.3 Sievers* Law

—> —»

0• (4) 2.1.1 Like-vowel contraction.

1.1.4 /1/ (Vbl? morpheme)deletion

— $> (5) 1.2.1 Consonantal change in A and V stems.

1.1.5 /i/ (Sbl4 morpheme)deletion

—>

1 .1 . 6 /ii/ (Vbl9 morpheme)- — > shortening

0. (6 ) 2.1.4 /i, u, j, w/alternat ion.

1.1.7 / o %/ (Sb24 and Sb26 —> morphemes)-to-/a/ rule

0.

1.1.8 / 01 / (Sb26 morpheme)- — > deletion

(7) 2.1.4 /i, u, j, w/alternation.

1.1.9 /u/ (Sb27 morpheme)to-/i/ rule

(8 ) 2.1.4 /i, u, j, w/alternation.

1.1.10 Word-stress rule

—> —>

(9) 1.1.1 Ablaut. (10) 1.1.7 / oi/ (Sb24 and Sb26 morphemes)-to/a/ rule. (11) 1.2.2 Reduplication.

Appendix

95

2

(1 2 ) 1.2.3 Thurneysen's Law. (13) 2.1.2 Vowel deletion. (14) 2.1.3 Breaking. (15) 2.1.4 (part 2 ) /i, u, j, w/-alternation. (1 6 ) 2 .1 , 5 Monophthongization of /ai, au/ to /Ei, Oi/. (17) 2.1.7 Glide insertion, (18) 2 .1 . 8 /u/-to-/o/ rule. 1.2.1 Consonantal change in ->( 1 9 ) 1 .2 . 8 /d/ (Vbl2 and A and V stems Vbl3 morphemes) real­ ized as /t, þ/, (20) 2.1.3 Breaking. (21) 2.2.3 Geminate-consonant simplification. (2 2 ) 2 .2. 6 /h/-deletion. (23) 2.2.8 Nasal deletion and vowel lengthening,

0.

1.2.2 Reduplication

—>

1.2.3 Thurneysen's Law

— > (24) 2.2.1 Consonantal devoicing•

1.2.4 Verner's Law

— > (25) 1.2.1 Consonantal change in A and V stems (26) 2.2.1 Consonantal devoicing.

1.2.5 /n/-insertion in St V

— > (27) 2.2.2 Distribution of voiced stop and cont inuant cons onants,

1.2.6 /s/-insertion in V

— > (28) 2.2.3 Geminate-con­ sonant simplification.

1.2.7 V suffixes Vbl2 /d/ and derivative suffix 52 /t/ realized as /s/

—> (29) 2.2.1 Consonantal devoicing.

Gothic

96

(30) 2.2,3 Geminate-con­ sonant simplification.

1 .2 . 8 /d/ (Vbl2 and Vbl3 mor-- > ( 3 D 2 .2 . 3 Geminate-conphemes) realized as /t, Þ/

sonant simplification.

0*

1.2.9 /*/ (Sb31 morpheme)deletion

1 .2 . 1 0 ##-deletion

—>

(32) 1 .1 . 1 0 Word-stress rul e . (33) 2 .1 . 2 Vowel deletion. (3*0 2.2.1 Consonantal devoicing.

2.1.1 Like-vowel contraction — £> (35) 2.1.7 Glide insertion.

0*

2.1.2 Vowel deletion

—>

2.1.3 Breaking

—>

2.1.*f /i, u, j, w/alternation

— J> (37) 1.1.10 Word-stress rule (i.e., part 1 of 2 .1 .** must precede).

(36) 2.1.** /i, u, j, w/alternation.

(38) 2.1.2 Vowel deletion. (39) 2.1.5 Monophthongization of /ai, au/ to /E:, 0 :/. (**0) 2.1.6 Vowel lowering. 2.1.5 Monophthongization of /ai, au/ to /Bs, 0 t/

0.

2.1.6 Vowel lowering

(*+l) 2.1.7 Glide insertion.

2.1.7 Glide insertion

—>

0.

2 .1 . 8 /u/-to-/o/ rule

—>

0.

2.1.9 /i:/-/e«/ alternations — > 2.2.1 Consonantal devoicing

0.

— > (**2 ) 2.2.3 Geminate-consonant simplification.

2.2.2 Distribution of voiced — stop and continuant consonants

(**3) 2.2.1 Consonantal devoicing.

Appendix

2.2*3 Geminate-consonant simplification

2

97

—

2.2.4 Consonant-cluster simplification

2 .2 . 5 /h/-assimilation

—>

2 .2. 6 /h/-deletion

0.

0.

2.2.7 Nasal assimilation

—

0.

2.2.8 Nasal deletion and vowel lengthening

—>

0.

2 .2 . 9 /n/-deletion

—>

0.

2 .2 . 1 0 /z/-to-/r/ rule

—>

0.

Illustrative examples of the above orderings are;

1 . lPIPstInd /þi:hum/ 'throve* (1 .1 .1 )

þihum (2 .1 .3 )

— í> //þehum// baihum. 2. 3SgPstInd of St V /sniw/ 'hasten* (1.1.1) — >> snaw (2.1.5) — $> //snOs// snau. 3. 3SgPBtInd /bring-d-a/ 'brought* (1.1.1) — (2 .2 .8 ) —$> brathda (ev) -■> //brashta//.

brangda

4. NmPlMn /þri-i-z/ 'three* ( = stem -f Sbl? + Sb30) by 1 .1 . 3 — ï> þri-i:-z (2 .1 .1 ) — $> þri:z (ev) -$> //þriss// þreis. 5. 3SgPstInd /bug-i-d-a/ 'bought' (1.1.4) — > bug-d-a (1 .2 .1 ) — $► buhda (ev) — p- //bohta// bauhta.

6 . lSgPstSub (after the application of 1.1.1 ablaut) /nesmij- 0 t/ 'take* ( = stem +- Vbl9 +■ VblO) by 1 .1 . 6 — netmiO: (2.1.4) — ,> //nesrnjOs// nem.jau. 7. NmSgPn /mau-i-oí/ 'maiden* ( = stem + Sbl4 +- Sb26) by 1.1.8 — > mau-i (2.1.4) — > mawl v s . the AcSg /mau-i-o:/ to which 1 .1 . 8 does not apply, but by 2.1.4 and ev — //mOsja// mauja.

8 . DtSg Mn or Nt /hard-u-amma/ 'hard* (1.1.9) —

hardiamma (2.1.4) — //hardjamma//. If the order were 2.1.41 .1 .9 , the form #//hardiamma// would occur.

9. Rule 1.1.1 Ablaut is formulated in terms of stressed and unstressed vowelsi 2SgPstSub /gri:p-i:-z/ 'seize* ( = stem + Vbl9 -I- Vb31) by 1 .1 . 1 0 grisp-i:-z (1 .1 .1 ,

applying only to the stressed /i»/) — £> grip-ii-z (ev) //gripiis// gripeis . 10. Rule 1.1.7 /o«/ (Sb24 and Sb26 morphemes)-to-/a/ is formulated in terms of stressed and unstressed vowels 1 NmPINt /word-oi/ #words* ( = stem +. Sb2^) by 1.1.10 —± word 0 1 (1.1.7) //worda// waurda vs. the NmPINt of the Art /þ-ó«/ ( = Pn2*+ +■ Sb2k) by“1.1.7 //þoi// to which 1 .1 . 7 does not apply because the /01/ is stressed. 11. 3SgPstInd /hEit/ *ordered* (1.1.10) -£> hÉit (1.2.2) 7/hehÉit// haihait. If the order were 1.2.2-1.1.10, the form *//hêbE it/ would result. 12. Rule I. 2 . 3 Thurneysenfs Law is formulated in terms of stressed and unstressed vowels. Hence the rule does not apply to the stem of a N like /hóih-ii/ hauhei •height* to produce #/hÓig-i«/ »haugei since the stem vowel /Oi/ has been stressed by 1 .1 .1 0 . 13. Rule 2.1.2 Vowel deletion is formulated in terms of stressed and unstressed vowels 1NmSgMn of the demon­ strative /sa-uh/ (1 .1 .1 0 ) — > sa-uh (2 .1 .2 b) —> //sah// vs. the NmSgNt /þata-uh/ (1.1.10) —> þáta-uh (2 .1 .2 d) //þátuh//, not *//þktaih/ / . 1**. Rule 2 .I .3 Breaking is formulated in terms of stressed and unstressed vowels. The compound A /fidur##doig-s/ •four-da^* is stressed cyclically by 1 .1 . 1 0 as fidur##doig-s and appears finally as fidurdogs by 2 .1 . 3 — not »fidaurdogs for V/fidordoixs// with //o// before //r//. riere the //u// remains before //r// be­ cause it is unstressed. On the other hand, /u/ before /r/ in a form like lPIPstInd /wurþum/ •became* is by 1 .1 . 1 0 — ^ wurþum, and since the /u/ is stressed is realized by 2 .1 . 3 as //wórþum// waurþum. 15* Rule 2.1.1* (part 2) /i, u, j, w/-alternation is formu­ lated in terms of stressed and unstressed vowels 1 the Inf /tiuhan/ fdrag* (1.1.10) — ^ //tiuhan//, not ♦//tjuhan// by 2 .1 .1*(2 ) because the /i/ is stressed. On the other hand, the unstressed /iu/ sequence in the NmPl /sun-i-u-z/ •sons* is realized by 2.1.1*(2) as //ju// — sunjus. 16. Rule 2 .I .5 Monophthongization of /ai, au/ to /Ei, 0i/ is formulated in terms of stressed and unstressed vowels. Thus a diphthong like //aj// (from /ai/) can occur only under stress in a form like bajoþ *both*, but never unstressed. I?. Rule 2.1.7 Glide insertion is formulated in terms of stressed and unstressed vowels* 3SgPrsInd /sei-i-d/ •sow* (1.1.10 and ev) — > sÉtid (2.1.7) — > sÉtjid (ev)

Appendix

2

99

— > //sE:jiþ// saij iþ « But 2,1.7 does not apply to un­ stressed vowels I E.g., the DtSgFn of the relative Pn is /þizEsi:// þizaiei, not #/þizEjji:/ »þizaijei . 18. Rule 2.1.8 /u/-to-/o/ rule is formulated in terms of stressed and unstressed vowels. It can apply to the un­ stressed /u/ in sunjus *sons‘ to produce sunjos, but not to the stressed /u/ to produce *sonjus. 19. 3SgPstInd /bug-i-d-a/ fbought# (ev) -^> bugda (1.2.1) — > buhda (1.2.8) — > buhta (ev) -^> //bohta// bauhta. If the order were 1.2.8-1.2.1, the form V/bohda// ♦bauhda would occur. 20. 3SgPstInd /bug-i-d-a/ ‘bought* (ev) — > bugda (1.2.1) buhda (2.1.3) — > bohda (ev) — > //bohta// bauhta. If the order were 2.1.3-1.2.1, the form *//buhta// would occur. 21. /gild-str/ ‘tax* ( = stem + derivative suffix 49) by 1.2.1 — > gilsstr (2.2.3 and ev) — > //^ilstr//. 22. /work-stw/ ‘work* ( = stem -+- derivative suffix 50 ) by 1.2.1 — > worhstw (2,2.6) — > //worstw// waurstw. 23. 3SgPstInd /þank-i-d-a/ ‘thought1 (ev) — ^ —> þanhta (2.2.8) — > //þaihta//.

þankta (1.2.1)

24. NmSg /rikw is/ ‘darkness* (1.2,3) -£► rikw iz (2.2.1) //rikw is// ric^is vs. the DtSg /rikw isa/ (1.2,3) — //rikw iza// riqTza to which 2,2,1 does not apply. 25. lSgPstlnd /þarf-d-a/ ‘needed* (1.2.4) þarbda (1.2.1) — > þarfda (ev) //þarfta//. If the order were 1.2.1- 1.2.4, then the nonoccurring *//þarbda// would result. 26. lSgPrsInd /Esh/ ‘have* (1.2.4) -> (2.2.1) //Eix// a i g .

E:g (ev)

E:g.

27. Prslnf /staþan/ ‘stand* (1.2.4) stactan (1.2.5) —> stancfan (2.2.2) — ?> //standan//. If the order were 2.2.2- 1.2.5» the nonoccurring *//stan ab##stastz (1.2.7) — > ab##stassz (2.2.1 and 2.2,3# applying in that order) — > //afstass//.

100

Gothic

31. lSgPstlnd /kunn-d-a/ *knew, could* (1.2.8) kunnþa (2.2.3) -£► //kunþa//. If the order were 2.2.3-1.2.8, then two forms would occur, //kunþa// and optionally the nonoccurring *//kunnþa//. 32. /## idd ja##uh##/ •and I went* (1.2.10) ##iddja uh## (1.1.10) - > ##iddja uh## (ev) //itfdjuh//. If the order were 1 .1 .1 0 - 1 .2 .1 0 , the derivation would be /## id d ja##uh##/ (1 .1 .1 0 ) ##iddja##uh## (1 .2 .1 0 ) ##idd ja uh## (ev) — > *//i //hwazuh//, to which 2 .2 . 1 now cannot apply to produce #//hwasuh//. 35« NmPIMn /þri-ii-z/ ’three* ( = stem + Sbl? ■+ Sb30) by 2.1.1 —> þriiz (ev) — > //þriis// þreis. If the carder were 2 .1 .7 -2 .1 .1 , then the derivation would be /þriii-z/ (2 .1 .7 ) þrijiiz (ev) *//bri$Lta// ♦þrijeis to which 2 .1 . 1 could not apply. 36. Latin urceus /urkeus/ •jug* (2.1.3) — > Go. //órkjus// aurkjus.

orkius (2.1.4)

37. Second-person PnAcPl /j-zw-iz/ *you* ( = Pnl4 + Pn30 -tPnl 3 ) by 2 .1.4(1) — > izwiz (1 .1 .1 0 ) — > izwiz (ev) //izwis//. If the order were 1.1.10-2.1.4(1), then the form would be stressed as *//izYfis//. 38. GnSg /band-i-oi-z/ *fetter* ( = stem + Sbl4 +■ Sb26 + Sb29) by 2.1.4 — > bandjoiz to which 2 .1 . 2 cannot apply# but eventually — > //bandjois//. If the order were 2 .1 .2 -2 .1.4, then 2 .1 . 2 would apply to /band-i01 -z/ to produce eventually *//band 01 s//. 39. GnSg /mau-i-ot-z/ ’maiden* ( = stem -t- Sbl4 + Sb26 + Sb29) by 2.1.4(2) — > mawjoiz (2.1.5) — > mOijoiz (ev) — > //mOijois// maujos. If the order were 2.1.52 .1.4(2), then the nonoccurring *//mawjois// »maw jos would result. 40. Prslnf /stoiian/ ’judge* (2.1.4) — > //stoijan// to which 2 .1 . 6 cannot apply because the /j/ is nonvocalic. But lSgPstlnd /stoiida/ to which 2.1.4 cannot apply is by 2.1.6 stOiida (ev) — * //stOii a//. If the order were 2.1.6-2.1.4, then the Prslnf *//st0ijan// ♦staujan would result. 41. 3SgPrsInd /sei-i-d/ 'sow* (2.1.6) — > sEiid (2.1.7) — » sEijid (ev) — > //sEijiþ// salj i þ . If the order were 2 .1 .7 - 2 .1 .6 , then the form #/ / s e 1 jiþ// *sejib would

Appendix

2

101

result to which 2.1*6 could not apply since /j/ is nonvocalic.

k2. NmSg /hals-z/ #neckf ( = stem + Sb31) by 2.2*1 — > halss (2.2*3) -?> //hals//. ^3. AcSg /hlB«b/ 'bread* (2.2.2) hlE:fc (2.2.1, applying only to continuants) —p //hlE:f// hlaif.

Gothic

102

Footnotes to Chapter 2 ^Moulton (1972 1 I67 -7 O) cites examples of these SPC's.

p

We consider here only native Go. and completely assim­ ilated loans such as aurkjus •jug*. There are of course non-Go. words beginning with /z/ attested in our corpus# e.g. proper names like Zakarias♦ ^The only recent exceptions are to our knowledge Beck (1973) and Wurzel (1975)* ^The one apparent exception to this# uhtiugs •oppor­ tune*# of which Feist (1939*515) says **Bildung dunkel.** is either an orthographic error for uhteigs (which also occurs) or is stressed uhtiugs. ^According to Haugen (1950)# such vowels are to be found in Old Norse. ^To our knowledge, one of the most complete listings of Go. derivative suffixes is in Buckalew (196*0. n

The Vb6 morpheme /Ei/ also probably occurs in the word arm-ai-o 'alms-giving*. Q

The /n/ is possibly suffix kO below. Another possi­ bility is suffix 48. The /s/ is realized as /z/ by rule 1.2.3# Thurneysen's Law. o As part of our formalism we assume that after the application of the morphological rules# any remaining optional M*s are automatically deleted; hence ##lag-i-da(M)(M)(M)(M)## =£> ##lag-i-d-a##. ^°The notation ## ----- M^## of rule Bel means that the word does not contain a lexical or stem morpheme# but rather is generated in its entirety by the morphological rules. In the case of Bel# these forms are ##/i/ (Bel) + /m/ (Be2)## = /im/ *1 am*, ##/i/ (Bel) * /z/ (Vb31)## = /is/ 'you are'# and ##/i/ (Bel) ^ /s/ (Be3) +■ /d/ (Vbl3)## = /ist/ 'is*. The notation ##Mj ----- ## of Be2 is similar except that the inserted morpheme occurs word-finally• ^ T h e Sbl7 morpheme may be considered to be either /i/ or /ii/. If the latter, then it does not undergo this rule,

12 13

Instead of Q - v ocalic^ as in 1.1.3.

One of several possible formalizations of this rule is in Vennemann (1971:123).

Footnotes

to

Ch e p t e r 2

10 3

lb

A laudable exception is Schmierer (1977)*

"^This seems to have been first pointed out by Beade (1971:36 and 1972:456). The numbers in parentheses refer to phonological rules in section 5. 17

The Vb-rules are here designated by number only. There is no Psv paradigm for the V*s fbe# or •will*,

Chapter III Germanic and Northwest Germanic 1. Introduction In a systematic and exhaustive reconstruction of the grammar of Germanic based on the material in the preceding chapter, it would be necessary to consider such aspects as the underlying segments of Gothic, the morphemic rules, the phonological rules, and to compare these with the corre­ sponding segments of the grammars from the other early Gmc. languages. Then, based on these comparisons and on various assumptions on the nature of language change, one could form hypotheses as to what the corresponding parts of the grammar of Germanic might have been. To consider for a moment only the phonological rules, each one as it appears in Gothic incorporates three possibilities for Germanici (1) The rule may be a total innovation of Gothic and not have been present in any form in Germanic. (2) It may have been inherited from Germanic without change. (3) It may have been inherited from Germanic with some sort of modi­ fication. An example of possibility 1 is very probably the rule of vowel lowering (2.1.6) since the other Gmc. languages do not evince such a rulei Cf. Go. bauan /bOian/ •dwell* vs. ON búa, OE bu tan, OHG bu tan, OS bu »an, OF buiwa. Instances of 2 may well be the rule of word stress (1.1.10), the rule of consonantal change in A and V stems (1.2.1), and redu­ plication (1.2.2). And examples of 3 are probably the Go. version of Verner#s Law (1.2.^) and consonantal devoicing (2.2.1). The former seems to be in the process of grad­ ually contracting its domain and dying out in Gothici The corresponding rule of Verner's Law in the other early Gmc. languages applies to many more forms. On the other hand, the rule of consonantal devoicing may well have been ex­ panding its domain in Gothic. It applies to both stops and continuants in the other early Gmc. languages and quite possibly began by applying to only some continuant con­ sonants in Germanic, say to /fc/ and /Í/. In this chapter we shall attempt a reconstruction of certain parts of the grammar of Germanic. Although our re­ construction cannot be exhaustive in that only some of the rules of Gothic can be considered, it will hopefully be systematic in that Gmc. versions of these rules will be posited and their subsequent changes into the NWGmc. lan­ guages will be plotted. In surveying the literature we

I n troduction

10 5

have selected for consideration what seem to be three of the major problems in the reconstruction of Germanic, These are the problem of umlaut, that of the so-called /ei,/ and /ei?/, and finally the problem of the reduplicatingbt V ^ . Before proceeding to our reconstructing, we should perhaps make a few remarks of a methodological nature. First, we can assume that the first major split in the Gmc. speech community was between Northwest Germanic on the one hand and East Germanic (as represented by Gothic) on the other, Bennett (196^13^1) describes the situation in these terms « A. The Goths, together with such kindred peoples as the Vandals, Burgundians, and Rugians, emigrated from the Proto-Germanic homeland in the course of the last two or three centuries B.C. and established colonies along the southeast shorelands of the Baltic, the Goths occupying the district about the lower Vistula, This move,,, separated the Goths from the North and West Germanic peoples by about four hundred miles.., In brief, the North and West Germanic peoples were still relatively contiguous up to the middle of the 5th century, but by this time the Goths were far away, living in settlements extending at least from the Crimea to the Atlantic, Accordingly, when in the following we posit changes from Germanic into the attested Gmc, languages, we shall see some which are peculiar only to Gothic and others which all the older NWGmc, languages — namely Old Norse, Old English, Old Saxon, Old Frisian, and Old High German — have in common A We shall also assume that incipient phonological change has the following characteristics which have been observed in empirical investigations of such change in pro­ gress, Hooper (1976 ilO^) notes, "At first a rule is always variable; the rule may or may not apply in any given situ­ ation," And Labov, in one of the most detailed accounts of language change in progress, states the following (1972« 223, 231, and 3 2 0 )« B. It is important to note that in the course of language evolution, change does go to completion, and variable rules have become invariant. There is a good empirical basis from the study of lin­ guistic change to see most rules as tending to apply maximally — to be generalized to all environments and to go to completion in a given environment,,,and to be reordered to apply to the maximum number of cases, ,, As the linguistic feature develops within the original group of speakers, it becomes generalized in several

Germanic

106

and N o r t h w e s t

Germanic

senses. Over the course of time (three or four decades) a wider range of conditioned subclasses may be involved, and more extreme (less favored) environments. Further­ more, the structural symmetry of the system leads to generalization to other vowels or consonants, or members of the same natural class... In sections 2 and 3 of this chapter we shall have occasion to see these facets of phonological change manifested in early Germanic. 2. The Problems of Umlaut and /e»2/ 2 .1. Preliminaries The IE vowel system from which that of Germanic ulti­ mately devolved is generally considered to have been thisi C. i

u e

iI

o

uI eI

a

oI at

and diphthongs with the structure ”fvocalicH r+vocalic~

_-high

J

-t-high _-long

That is, /ai, ei, oi, au, eu, ou/ and /a;i, eii, oii, aiu, eiu, oiu/. (For our purposes we can ignore the possibility of an IE schwa /0/, which appears in Germanic and all other Western IE languages as /a/.) Among the Gm c . changes in this system, the following are generally assumed! D. l. The introduction of Gmc. stress (as reflected in the Go. phonological rule 1.1.10 in chapter 2. Hereafter when referring to stress, we shall mean Gmc. stress.) 2. Vnh ~>

Vih (as reflected in the Go. phonological rule

2 .2 .8 ).

3c IE syllabic /1, m, n, r/— .

c

.

Gmc, /ul, urn, un, ur/.

«

4. IE /o/ Gmc. /a/ everywhere (i.e., in both stressed and unstressed positions). 5. IE /ai/

Gmc. / 01 / everywhere.

6. The unstressed long diphthongs, except possibly those beginning with /e i/ (/eti/ and /eiu/), are shortened

The P r o b l e m

of

Umlaut

and

/ e : 2/

10 7

and fall together with the short diphthongs, i.e. /aii/ — /ai/, etc. (We shall consider the stressed long diphthongs directly.) The last four of these changes are not reflected in any of the phonological rules of Gothic. They began as optional rules which when completed and obligatory were manifested only in the inventory of underlying segments of the lan­ guage. Regarding change 6, the IE diphthongs /eii/ and /eiu/ are generally considered to appear in Gothic in the ending of the DtSg of Fn i-class N (e.g., anstai /anstEi/ •favor*) and that of the DtSg of Mn u-class N (sunau /sunOi/ *son#). The corresponding endings in the early NWGmc. languages are /-i/ and /-iu/ as in ansti and suniu, The usual explanation given for this is that these endings were ablaut variants inherited into Germanic as i-class DtSg /eii/-/ei/ and u-class DtSg /eiu/-/eu/, The NWGmc. endings are from the /e/-gradei /ei/ — > /ii/ —> /i/ and /eu/ — *> /iu/. And the Go. endings derive from /e «/-grade /eii/ and /eiu/. If this account is correct, then the IE diphthongs /eii/ and /etu/ — in unstressed position — did not terminate in Germanic like /ei/ and /eu/. If the latter had been the case, then the Go. reflexes would have been #anstei */&natit/ (from /anstei/) and »sunju (from /suneu/). 2.2. The Problem of Umlaut The only obvious trace of umlaut in Gothic is the phonological rule of breaking ( 2 . 1 . 3 in chapter 2), w h i c h for convenience of reference we repeat herei E. 2.1.3. Breaking 1. First, /e, o/ are realized as /i, u/ if stressed or if unstressed and posttonic — i.e., after the main stressed vowel within the w o r d . 2. Second, stressed /i, u/ are realized as /e, o/ if immediately followed — with no intervening morpheme boundary — by /h, hw , r / . Further, stressed /i/ is realized as /e/ if followed by / t r / . In reconstructing the G m c . v e r s i o n o f 2 . 1 . 3 » o ne m u s t first consider the earliest attested NWGmc. v e r s i o n of the rule. This seems to have been the following«2

Germanic

10 8

and N o r t h w e s t

Germanic

F. NWGmc. Breaking 1. /e, o/

/i, u/

/ Qt-stress^ X [“-stress-

a. C Q "Vocalic

_-- J

( = /i(i)/)

-hhigh back b. C^##, where

_ cannot be

/h, h", r / . 2. /e, o/

/i, u/

/ ["VstressH

CQ " v o c a l i c ”

_____

-hhigh _— back Q+nasalH Qtconsonant[]

Although rule F could apply to /o/ as well as to /e/, it does not affect /o/ since all /o/*s had by this time become / a / #s by change h under D. Part 1 of F applies to / e / fs occurring after the main stress in a wordi part 2 to stressed /e/*s. E.g., (Fla) DtSg /hánini/ #roosterf (cf. OHG hanin) from earlier /haneni/ vs# the GnSg /hánenas/ (cf, OHG hanen) and the GnSg /dágesa/ *dayf (cf. OHG tages), /ubiri/ and /úber/ *over* (cf. OHG ubir vs. ubar) from /úber-i/ and /úber/, 3SgPrsInd /fár-edL/ *gor ->/faridi/i (Fib) 2SgPstInd /betr-ez/ *carry# — £► /beiriz/ (ev NWGmc. /basri/, cf. OHG barri). But Fib cannot apply^to forms like /after/ *afterr, /ánþer/ •other*, and /hwáþer/ •whether* because C, • /r/ (cf. OHG after, ander, OE hwae^er) . Although we have found no instances of /h, hw/ in this environment, we can assume that, as in Gothic, they would have functioned like /r/ to inhibit the application of Fib. An example of part 2 of F is the 3SgPrsInd /beridi/ •carry1 (from /ber-edi/ by Fla) — > /bir-idi/ (cf. OHG birit) . As this derivation shows, parts 1 and 2 of F, like parts 1 and 2 of the Go. version 2.1.3 under E, are ordered. Another example of F2 is the Inf /bendan/ *tie* /bindan/• Parts 1 and 2 of F can apply to unstressed and stressed /ei/, e.g. /steigan/ *climb * - » /stiigan/ = /sti igan/. Later versions of rule F as they are severally mani­ fested in the various NWGmc. languages exhibit expanded domains. Part lb lost the proviso that C, could not be /h, hw » r/, so forms like /aftir/ 'after* begin to occur instead of /after/. And in some NWGmc. languages like Old Saxon, part 2 of F lost the proviso —back and the rule began

The P r o b l e m

of

Umlaut

109

applying if the following vowel was /u(i)/» Hence OS biru •I carry* from West Gmc. /beru/ vs. earlyJDE beru. Also in Old Saxon the environment ----- [j-nasal^ |_-hcon sonant]] in F2 had begun to lose optionally the proviso L-i-consonantI] > so forms like the Inf niman *take* occur alongside the earlier form neman. Still another kind of expansion which rule F evinces in the various NWGmc. dialects is best illustrated if F is formalized as follows« F * . NWGmc. Breaking (parts 1 and 2) ["-t-vocalic”! — £> L+high]] ; —high -low

/

in "the environments specified under F above, inparticular before« C

"Vvocalic O

_-long

J

;

thigh (The feature —bacx was probably being deleted from the rule by this _time.)

J

The expansion at issue here may be formalized by substitut­ ing high for -high in F * . That is, rule F* not only raises /e/ to /i/, but now begins — doubtless at first optionally — to lower /i, u/ to /e, o/ when followed by C and a —high vowel. This expansion is carried out only sporadically in the attested NWGmc. languages, as opposed to the change of /e/ to /i/ which rule Fo effected first and which was carried through to completion. Forms illustra­ tive of this phenomenon are OHG quec 'alive* vs. OE cwic, OHG skef and skif #ship*f the Pst Prt of lst-class St V*s with /i/t not /e/, as in /ridan/ *ridden*, not */redan/, OHG foil #full# vs. OE full, CHG wolf #w o l f vs. OE wulf, and OHG furhten #fear* as well as forhten. These alterna­ tions indicate that the rule F* had begun to extend its environment optionally and to varying degrees in the various NWGmc. languages. Now the Gmc. proto-language may well have had an in­ cipient umlaut rule. If we consider the fate of IE /ei/ in all the Gmc. languages as well as those environments which the Go. breaking rule under E and the NWGmc. rule under F have in common, the Gmc. rule may have looked something like this«

Germanic

110

and N o r th w e s t

Germanic

G . Gmc• Umlaut /e/

/i/

/ a. ----- /i/ b. L+stressJ X ’"-stress” a, CQ /1 ( i)/

_----

J b. C1##, where cannot be /h, hw r/.

That is, (a) both stressed and unstressed /e/ is raised to /i/ if immediately followed by /i/. This realizes the IE diphthong /ei/ as /ii/ = /ii/. And (b), unstressed /e/ in posttonic position is raised to /i/ if (ba) followed by any number of consonants and /i/ or /i 1/, or (bb) if followed by at least one consonant (except for /h, hw , r/) in wordfinal position# According to the principles enunciated under B above, we can assume that G was at first optional. When part (a) became obligatory, this led to a restructuring of the in­ ventory of underlying segments in that /ei/ now appeared everywhere as /ii/. Hence there was no need for a phono­ logical rule /ei/ /it/, and part (a) was accordingly dropped from rule G. Rule G must have occurred rather late in the common Gmc. period.^ For by the time part (b) began its expansion, the Gmc# proto-language had already begun its initial bi­ furcation into East and NW Germanic. And part (b) extended its domain in somewhat different ways in these two branches# We have already seen under F and F # above how the NWGmc. languages tended to form their versions of the rule on extensions of environment (ba). But in East Germanic — as represented by Gothic — environment (bb) of G was extended in the following wayi Gbb. [VstressJ X ""-stress"" C^##

where C, cannot be /h, hw , r/, becomes through deletion of the proviso the following! Gbbf. [_+stress] X ""-stress”

And probably at about the same time the original environ­ ment with its restrictive conditioning /h, hw , r/ was transferred to stressed vowels, resulting in

The P r o b l e r n o f

Gbb**.

-hstress

_---- J

Umlaut

111

CQ , where CQ is any segment except /h, r/.

The further Go. expansion of Gbb,# is difficult to capture in our system of features. However, if we reformu­ late Gbb** using a feature like w-t-loweringM to describe the effect of /h, hw , r/ on preceding short high vowels, then Gbbfl can be stated as follows« Gbb*'. ""-f-vocalicl — >> Qhhigh^] -high

/ ""tstress

Any segment" except :

L-----hconsonant

-low [flowering

__

_-long That is, /e/ — /i/ except before /h, hw , r/, Now if in Gbb** the specification t-high is replaced by cxhigh and -hlowering by -^lowering, and if the proviso MAny segment except** is stricken, then the rule is generalized. It now not only raises /e/ to /i/, but it also lowers /i, u/ to /e, 0/ before /h, hw , r/. The exact shape such a formal­ ization should take is not at issue here: The point is that a change in a rule from raising /e/ — and / 0/ if it had been present — to /i, u/ except before /h, hw , r/ to one which also lowers /i, u / to /e, 0/ before /h, hw , r/ is a plausible type of rule simplification and generalization. Perhaps the main reason why the expansion of rule G was fairly regular in Gothic (in the breaking rule 2.1.3) and sporadic in NW Germanic (in the rules F and F ' ) is that the Go. version turned out to be a MS rule as opposed to a regular phonological rule in NW Germanic. That is, the conditioning environment in the Go. rule turned out to be the immediately following sequence of consonants. Thus the rule did not apply over morpheme boundaries and there were no paradigms in Gothic which exhibited alternations result­ ing from the rule such as baira *1 carry* vs. *biris fyou carry0. But in all the NWGmc. languages the condittoning environment / i ( «)/ and the affected vocalic segment could be separated by any number of consonants (Cp). This meant that there were numerous NWGmc. paradigms which evinced alternations resulting from the rule, e.g. the Prs 1 Sg Ind /ber- 01 / •carry* vs. Prs 2 Sg Ind /bir-is/, the NmSg /fih-u/ ’cattle* vs. GnSg /feh-es/, among many. Under these circumstances it is not surprising that the umlaut rule F in NW Germanic should have been morphosyntactically — and on occasion lexically — conditioned and as such sporadic in its application. As the G m c . short vowel system at the time of the addition of rule G (and the becoming obligatory of environ­ ment a), we posit /i, e, a, u/ with /i/ and /e/ in

112

Germanic

and N o r th w e s t

Germanic

complementary distribution in the few posttonic environments specified in rule G (environment b), but contrasting else­ where. This has not been the view in almost all recent considerations of Germanic. Most of these such as Antonsen (1965:26) posit a much richer vowel system with numerous allophonic variants« /1/ = //!.“*. i* e//

/u/ = //u, y, o//

/e/ = //e, e, ~ / /

/a/ = //a, ae, d, 3 / /

The following rationale for positing all these vowels is given by Antonsen (1965*20, 24-5): With the refined methods available to us by the structural approach to historical linguistics..., it is entirely possible for us to posit approximate phonetic values for the various phonemes in different environments in Proto-Germanic. There can be no doubt that the umlaut allophones arose at a time when the full endings were still present, and there is no reason to assume that the assimilations involved in a-umlaut occurred before those involved in i- or u-umlaut, nor is there any reason to assume that the i-umlaut of /e/ preceded the i^-umlaut of / a / • The presence of reflexes of mutation allophones in all \yicj dialects and the absence of any "Ruckzugsgebiete"5 demand that we consider the various umlauts to be the result of a single phonetic tendency which was active in the Proto-Germanic period. We have cause to be wary of assertions that umlaut is a common Germanic phenomenon, but not Proto-Germanic, or that mutation arose in the individual dialects.® To rephrase this account in our terms, Proto-Germanic had subphonemic umlaut rules like the following: H.l. / a ( 0 , o ( :), u( :)/ — > //e(;), o ( 0 # u(*)// in the appropriate environment, say ----- C0 /i(:)/ (i-umlaut). 2. / a ( :), e (:), i(*)/ //o(;), o(:), u(:)// in the appropriate environment, say ----- CQ /u(:)/. The vowels produced as allophonic output of these rules then occur in the later attested NWGmc. languages as inde­ pendent phonemes. This occurs as the result of the loss or reduction of the triggering vowels /i(:)/ and /u(:)/ — which, it should be emphasized, occurred separately in the individual languages. However, this account of umlaut en­ counters some difficulties. One of them is that some of the earliest attested Gmc. languages such as Gothic, Old Saxon, and Old High German either evince absolutely no trace of

The

Problem

of

Umlaut

113

the type of umlaut under H above (Gothic) or umlaut at only an incipient stage, namely that of short /a/ to /e/ before C /!(:)/ (Old High German and Old Saxon). In this connec­ tion it might be added that Modern Dutch has no umlaut ex­ cept the remnant of the NWGmc. breaking rule F above in forms like Sg schip •ship* vs. PI schepen and /a/-to-/e/ umlaut in Sg stad *city* vs. PI steden. Umlaut of /o(:)/ and%%/u(:)/ does not occur« E.g., groter ’larger*, not »groter, and küssen *kiss*, not »küssen J The usual explanation offered to account for the ab­ sence of umlaut in the older languages is given by Moulton (1961:506) as follows: ...unsere Erfahrung mit Orthographien im allgemeinen erlaubt uns, eine ganz feste Regel aufzustellem In einer normalen Orthographie (d.h. abgesehen von ge­ lehrten phonetischen Transcriptionen usw.) werden die Allophonen ein und desselben Phonems nie und nimmer [jBmphasis supplied 1 JBV[] schriftlich unterschieden. Der Grund dafür ist leicht zu finden: der normale Sprecher ist sich der Allophone seiner Muttersprache einfach nicht bewußt und was er nicht bewußt hört, schreibt er nicht. Marchand, in taking basically the same line, remarks _ (I97 O 1 II 2 ), **...das Alphabet des Wulfila [Verzeichnetj unterphonemische Varianten nicht.** Statements such as these must strike anyone who has concerned himself at all with the texts of these languages as utterly absurd. Rather, the textual evidence indicates the opposite to be the case: Namely, that most of the subphonemic distinctions — and virtually all of the frequently occurring ones — are sporadically indicated in some fash­ ion or another in the texts. Scribes will often vacillate between a taxonomically phonemic and a redundantly phonetic orthography. One example of this in Gothic is the use of both gg and ng for //q g// as in bringan and briggan *tring* . (See rule 2.2.7 in chapter 2.) If tne scribe had used only graphemic ng for //r)g// and n for //n//, he would have been writing phonemically. However, the fact that he also (and more frequently) uses the digraph gg for //og// indicates that he was aware that there was a nasal preceding /g/ (and /k/) different from that in a word like bindan *tie* (never written *bigdan)• He was therefore aware of the for him subphonemic difference of //r)// vs. //n// and designated it orthographically. Examples of subphonemic writing are even found in Runic inscriptions where forms like sba *prophecy*, gasdiz 'guest*, and aisgz 'challenger* occur instead of spa, gastiz, and aiskz .0 Here graphemic b, d, and g have been used instead of the more usual £, t, and k to repre­ sent unaspirated — and subphonemic — 7/p~» t“ , k~//• And there are numerous other instances of subphonemic

114

Germanic

. orthography attested m

and N o r t h w e s t

Germanic

the early Gmc. languages.

Q

Now if indeed Gothic or any of the other early NWGmc• languages had had umlauted vowels like / 0(1)/ or /u(i)/, then there almost certainly would have been some sort of sporadic indication of this frequent occurrence in the texts. In Gothic, for example, if the 3 Sg Sub Pst drusi •fall* really were *//drusi// instead of //drusi//, It surely would have been written on occasion as »drwsi or the like. The fact that there is not a single graphemic hint of umlauted vowels in several early G m c . languages combined with the%>fact that %Modern Dutch evinces no reflexes of um­ lauted /ö(«)/ or /u(i)/ provides a strong indication that umlaut rules like those under H simply were not present in Germanic. The structuralist motivation for assuming umlaut rules for Germanic is elucidated by Antonsen (19^5>25) as followsi "...we should seek to identify the genes which are respon­ sible for this Germanic hereditary disease p . e . , umlaut^ , and this we can do. The mutated genes are the umlaut allophones which must have been present in Proto-Germanic." While we can agree that the "genes'* of umlaut probably existed in Proto-Germanic, they were not present in the form of actually occurring allophones, but rather as a ten­ dency, and quite possibly as rule G above. We have already seen how this rule was expansionist in that it tended to extend its domain in various directions and at various rates in the various Gmc. languages. Further, extension of this rule — or at least of this assimilatory tendency — to in­ clude /a/ in its domain, then /o(i)/, /u(i)/f and /ai/, and partial assimilation of these vowels to the triggering / 1 ( 0 / (which was retained into all%%the NWGmc. languages) led to the eventual appearance of /o(»)/, /u(i)/, and /aei/ in some, but not all (e.g. Dutch) of these languages. It is also possible that so-called u-umlaut arose in other of these languages by extension of the triggering environment from /i(i)/ to /u(i)/. In sum, while a Gmc. rule like G of vowel assimilation probably existed which could have been elaborated upon and extended in the daughter languages, there is absolutely no evidence that the mutated vowels themselves had appeared in Germanic. It is interesting that in attempting to offer a unified explanation of this phenomenon, the extreme empiricism and concomitant super­ ficiality inherent in the structuralist approach make it necessary to posit the actual occurrence of such vowels when the data available indicate that none were in fact present.1° 2.3- The Remaining Gmc. Vowels As mentioned in change 6 under D, the long unstressed diphthongs — with the possible exception of /eii/ and /eiu/ — fell together with the short ones at a fairly early date. The same also seems to have occurred with the

The

Remaining

Germanic

Vowels

115

stressed long diphthongs. This, along with the other changes mentioned under D, resulted in the Gmc. diphthongs /ai, au, eu/ and for a time /eii/ and /eíu/.11 Now the relfexes of the unstressed diphthongs /ai/ and /au/ appear in all the attested Gmc. languages as monoph­ thongs, generally as /Ei/ and /Oi/ in Gothic12 as specified by phonological rule 2.1.5 in chapter 2. The same diph­ thongs in unstressed position are realized as /ei/ and / o «/ in NW Germanic. E.g,, the Gc, 2 Sg Prs Ind habais /habEis/ 'have* vs. OHG habe is and Go. ahtau /ahtOi/ 1eight * vs. OHG ahto for /axtoi/ (or perhaps /axto/ from earlier /axtoi/). These facts have led some, e.g. Reis (197^), to posit a change of unstressed /ai, au/ to /Ei, Ci/ in Germanic and from thence to /ei, 0 1 / in NW Germanic. If this was in fact the case, then — as emphasized by Reis — this results in an almost unheard-of situation for Germanic« Namely, an inventory of unstressed vowels including / E i, Oi/ and as such greater than that of the stressed vowels. This would seem possible only if the Gmc. rule changing unstressed /ai, au/ to /Ei, Oi/ is considered to have persisted in that form for just a short time. That is, the rule must have been in late Germanic, probably just prior to the initial split into East and NW Germanic. Then in Gothic the Gmc. rule whereby unstressed /ai, au/ — ^> /E «, Ci/ was general­ ized to apply to stressed /ai, au/ as well (to result ir. rule 2.1,5 in chapter 2). Hence Gothic did not violate the constraint against having a larger inventory of stressed than of unstressed vowels. And in NW Germanic a new rule was soon added whereby unstressed /Ei, 0«/ from Germanic became /ei, o«/ — i.e., vowels which also can occur under stress. Hence NW Germanic did not violate the constraint either. The other possibility here is that the monophthongization of unstressed /ai, au/ was an independent de­ velopment in early NW and East Germanic such that in Gothic rule 2.1.5 was added, while NW Germanic added a rule where­ by unstressed /ai, au/ — > /e«, o«/. In either case, the monophthongization of /ai, au/ must have occurred either fairly late in common Germanic or fairly early in the NW and East Gmc. periods,13 Therefore Germanic — for most of its existence — had the diphthongs /ai, au, eu/ and /e«i, e «u/ in both stressed and unstressed positions. (The rule of Gmc. umlaut under G cannot apply to /eu/ to give /iu/,) So in view of the early Gmc. changes cited under D above and considering the immediately preceding discussion, we may assume that the^ Gmc. vowel system throughout much of the common Gmc. period was something like this« I. Short vowels« /i, e, a, u/. Long vowels « /i«, e«, ai, o«, u «/ (with / a «/ possibly occurring only under stress).

Germanic

n e

and N o r t h w e s t

Germanic

Diphthongs: /ai, au, eu/ and very lively /e:i, e:u/. The sole source of /a:/ is from the sequence /anh/ by change 2 under D. A number of sources such as Antonsen (1965:26) posit numerous allophonic umlaut alternations of long vowels: /i :/ = //i:,iu://, /e :/ = //e :, e ://, / u í / = //u:, y ://, / 0 :/ = //o;, /://. And for the reasons cited in 2.2 above, we do not believe such allophones to have existed in Germanic. The vowel system under I is reminiscent in some re­ spects of the systems posited for Germanic in earlier, pre­ structuralist sources. This should not, we believe, be considered evidence against I. The structuralist approach, far from providing the "refined methods" lauded by Antonsen, has in fact perpetuated a fundamentally erroneous account of Germanic and of many of its developments. Structuralist methodology therefore cannot be hailed as having provided any particular insights into the nature and development of Germanic, but must rather be recognized for what it in fact was, an exercise in obfuscation which has beclouded a num­ ber of important issues in the field for well over a quar­ ter of a century. 2.4. The Problem of /e:^/ and / e ^ / A common variation on the long-vowel inventory in I above is found in most of the handbooks and in more recent sources such as Moulton (1961) and Reis (1974). It is this: I*. i :

u: e:2

o: e5l

The system under I* differs from ours under I by first ig­ noring the few occurrences of /a:/ and second by positing two /e:/fs, a higher / e u / and a lower /e:,/ (hereafter written /æ:/). The reason for this is that most Go. words like /je:r/ fyearf which evince / e :/ have NWGmc. corre­ spondents with /a;/, e.g. ON /a:r/ (earlier /ja:r/) and OHG /ja:r/. But four Go. words — two inherited from Germanic (/he:r/ #heref and /fe:ra/ *area*) and two borrowings (/kre:ks/ fGreek* and /me:s/ *table* — have NWGmc. cog­ nates with /e;/, e.g. OHG / h e :r/, /fe:ra/, /kre:x/, and /me:s/. Soon after the attestation of these NWGmc. lan­ guages, the following change occurs: J. Stressed /e:/ /ie/ (later /ia, io/) in certain environments (to be discussed directly). Hence later OHG forms are hier, fiera, Kriach, and m i e s . The traditional account for this has been the

The P r o b l e m

of

/ e : ^/

and

/ e : 2/

117

following « K.l. IE /e i/

Gmc. /æi/ (i.e., /e«-,/).

X 2. IE /eii/ — ^ Gmc. /et/ (i.e., /ei^/) , 3. Gmc* /æi/ — >

Go. /et/.

Gmc. /æi/ — > NWGmc. /ai/. A simpler and more plausible account of the data would be to assume that IE /et/ remained in Germanic as did the long diphthong /e«i/. As we have already noted in our dis­ cussion of change 6 under D, there is evidence that un­ stressed /eti/ and /eiu/ remained for a time in Germanic. Although stressed /eiu/ probably fell together with /eu/ fairly early, there is good evidence that stressed /eii/ remained throughout the Gmc. period. There are numerous forms such as Go. /heir/ which alternate with other forms like Go. /hir(i)/ fcome here’ and which thereby evince a plausible type of IE ablaut which indicates a source having /eti/i IE /keiir/ — Go. /heir/ and IE /kir/ — Go. /hir(i)/. Further, a number of NWGmc. forms occur which have no attested Go. correspond­ ents, but which also evince ablaut alternations with other forms indicating their origin in IE /eii/, e.g. OHG stiega •stairway* (eariier /steiga/) from IE /steiigh/, the same root also occurring in different ablaut grades in the 1stclass St V /stiigan, staig, etc,/ from IE /steigh , stoigh , etc./. Likewise MHG schief #crooked* (earlier /skeif/ from IE /skeiip/) vs. OE sea if (from /skaif/ from IE /skoip/); and OS ti tr fglory# (from Gmc. /tiir/ from IE /deir/) vs, OHG zeiri 'glorious* (from Gmc. /teiir/ from IE /deiir/). Additional examples of this kind are cited in van Coetsem

(1956i28). In view of this evidence, a simpler and more straight­ forward account than K would be this 1 K*.l. IE /et/ remains in Germanic (i.e., no rule). 2. IE /eii/ remains in Germanic (i.e., no rule). 3. Gmc, /ei/ remains in Gothic (i.e., no rule). Gmc. /ei/ — ^

NWGmc. /a 1/ (perhaps via /æ«/).

5# Gmc. /e «i/ Go. and NWGmc. /et/ (through two separate changes). That is, the IE long diphthongs only gradually fell together with the short ones in Germanic, the /eii/ — and perhaps /e«u/, which need not concern us further here —

n e

Germanic

and N o r t h w e s t

Germanic

being retained throughout the G m c . period. Retention of only some of the IE long diphthongs is by the way not un­ precedented* In Greek, for example, all the IE long diphthings were lost except those ending in the nonback vowel /i/. (See on this Buck, 1962*90-1.) Similarly in Germanic, all IE long diphthongs became short except those beginning with a nonback vowel — namely /eii/ and /etu/ — the single long diphthong to have both segments in nonback vowels, /eii/, remaining longest. Regarding change K'5 (/eii/ — /e */) into Gothic, it is interesting that the only two native Gmc. forms in Gothic, /heir/ and /feira/, are automatically derived by our rules in chapter 2. Thus Gmc. /feiira/ by the Go. breaking rule 2 . 1 . 3 ^ /feiera/, which then by the Go. rule of like-vowel contraction (2.1.3) — > /feira/. Like­ wise Gmc. /heiir/ by breaking 2.1.3 —> /he*er/15 and by like-vowel contraction 2.1.1 — > /heir/. Regarding change K*5 into NW Germanic, it was preceded by K*^, a more precise formulation of which being this* K*U. Stressed /e */ —$> /ai/ (perhaps by way of /æi/), unless followed by /i/. Thus Gmc. /jeir/ ’year* is NWGmc. /ja»r/, while a form like Gmc. /heiir/ 'here* remains. It is of course quite natural for /ei/ to be retained before a high vowel. Rule K*^ does not affect unstressed /ei/ — many of these being from Gmc. /ai/ — in forms like the 2 Sg Prs Sub /neme is/ *take*, not */n em a . is/. After the application of rule K*4 and of NWGmc. K f5 by which /e*i/ became / e */, there came to be three main classes of words in NW Germanic evincing /ei/ in stressed position. One of these consisted of borrowings like early OHG /me is/ 'table* (later mies), / k r e ix/ 'Greek* (later Kriach) from earlier /kreik/, and /breif/ *letter* (later brief). All these words are loans from late Latin forms with /ei/ — /meisa/, /greikus/, and /breivis/. It was of course quite natural to employ the NWGmc. /ei/ to repre­ sent the vowel of these foreign words. The second class of words with /ei/ under stress in NW Germanic — and one which is usually ignored in the liter­ ature — consists of the so-called athematic verbs for *go* and 'stand*. These verbs evince throughout the Prs Ind paradigm optional alternations between /ai/ and /ei/. Thus in Old High German the Inf /gain/ *go* alternates with / g e m / and /stain/ 'stand* with / stem/; likewise, the 2 Sg Prs Ind /gais/ alternates with /ge*s/, etc. These verbs have been traditionally termed •’athematic” for morphologic reasons, namely the fact that the Prs tenses are formed without the intervention of an intermediary morpheme (the

The

Problem

of

/e : ^/

and

/ e :2/

119

••thematic vowel") between the stem and the inflectional ending. For example, the Gmc. 2 Sg Ind /ges-s/ *go* ( = stem - the reflex of the Go. Vb31 morpheme) differs morph­ ologically from the corresponding form of other St V*s like Gmc. /nem-i-s/ ftake# ( - stem + the reflex of Go. Vbl8 the reflex of Go. Vb31)* In addition, the 1 Sg Ind Prs of such verbs is formed with /m/ (the reflex of the Go. Be2 morpheme), /gei-m/ *1 go*. There seem to have been three such verbs in Germanic, /doi/ fd o #, / g e i/ •go1, and / s t e i/ (possibly /stai/) •stand*. They have had a precarious existence throughout the history of the Gmc. languages in that they have tended either to be replaced by other verbs or to be assimilated into the morphology of regular St V*s. In Gothic, /doi/ was replaced by taujan and the Prs-tense forms of /get/ and /stei/, probably because of their MS (unusual for St V*s), were replaced by the suppletive stems /gang/ and /stand/. (In Crimean Gothic an athematic form geen *go* is attested.) Exactly the same thing happened in Old Norse except that /doi/ was replaced by g/ra. In West Germanic, Old English replaced /stei/ with /stand/ while retaining /doi/ and /gei/. And Old Saxon and Old High German retained all three. But even those lan­ guages retaining the stems /dot/, /get/, and /ste«/ have throughout their attested histories tended to assimilate these forms into the morphology of the other St V*s. One instance of this is the fact that the Gmc. reflex of the Go. rule Be2 which puts /m/ onto the stem in the 1 Sg Prs Ind of athematic V #s as opposed to other St V*s sometimes does not apply. Thus in Old English 1 Sg Prs Ind dot and ga » 'go* occur, and in Old High German both to »m and toi Td o *, gaim and gai, and staim and stat. Another manifest­ ation of this tendency is the fact that morphological rules inserting thematic vowels between stem and ending which apply to St V*s tend to apply optionally to athematic as well. One of these is the reflex of the Go. rule Vbl8 which inserts /i/ between the stem and ending of the 2 and 3 Prs Sg Ind of St V*s. Thus in Otfridian Old High German the 3 Sg form duit ’does* from /doi-i-t/ (with the Vbl8 morpheme) occurs alongside duot from /doi-t/ (without it), in Old Saxon ste tid #standsf with the thematic /i/ along­ side. steid without it, and in Old Frisian ge «ith ’goes* with /1/ alongside geith without it. Now there is no reason to believe that this optional application of St V morphology to the athematic V*s did not occur even during the earliest Gmc. times. If so, there must have occurred already in Germanic alternations such as the 2 Sg Prs Ind /ge«-s/ and /gei-i-s/, the 3 Sg Prs Ind /ge«-d/ and /gei-i-d/, etc. And if we apply to these forms our NWGmc. rules K*4 (/ei/ — > /a«/ unless followed by /i/) and K*5 (/eii/ — > /e«/), we arrive at alternating forms

120

G ermanic

and N o r t h w e s t

Germanic

like /gais/-/geis/ and /gaid/-/geid/ such as occur through­ out the Prs paradigm of these verbs in all the West Gmc. languages in which these verbs have been retained.16 Finally on this subject, the later rule J given above whereby stressed /ei/ from /eii/ appears as /ie/ and the like does not apply to the /ei/*s in athematic V*s like /geis/ (not */gies/ ) . This means that J was morphologically conditioned from the time of its earliest application. This is not unprecedented. Labov (1972i322) cites an anal­ ogous example of an incipient phonological rule in some American English dialects inhibited by morphological fac­ tors. This is a rule of consonant-cluster simplification whereby the /t/ in a word like last /læst/ is deleted much more often than in a word like passed /pæs-t/ because the /t/ in the latter form is an independent morpheme. Since our rule J did not apply to Prs Sub endings like the 3 Sg Prs Sub /háb-ei/ *have* because the /ei/ was unstressed, the rule also tended not to apply to the Prs Sub forms of athematic V #s as well, even though in this case the /et/ was stressed as in the OHG 3 Sg Prs Sub /géi/ #g o # (possi­ bly from an underlying /gei-ei/ or /gai-ei/ by an OHG rule similar to that of Go. like-vowel contraction 2.1.1 in chapter 2). And since rule J did not apply to the Sub paradigm of athematic V*s, it tended to by-pass all the other forms of the paradigm as well. Finally, the third class of words with /ei/ in NW Germanic and the one which has received the most attention in the literature consists of the nonparticipial Pst forms of formerly reduplicating V #s. It is this class to which we shall direct our attention for the remainder of this chapter • 3. The Reduplicating Verbs in NW Germanic 3.1. The Problem A satisfactory account of the original status and sub­ sequent development of the so-called seventh class of reduplicating St V fs in Germanic has been termed by Prokosch (19391176) "one of the most difficult problems of Germanic grammar." The situation is in outline thisi In Gothic, the past — exclusive of the participle — of such verbs is formed through reduplication (Go. rule 1.2.2) or reduplica­ tion with ablaut (rule,1.1.1, part B), e.g. haitan •com­ mand* vs. haihait /hehEit/ and letan •let* vs. laîlot /leloit/. In the other Gmc. languages — i.e. NW Germanic — the corresponding nonparticipial past is formed through some sort of mutation of the root vowel, e.g. OE haitan vs. OHG heizzan vs. he iz (later 'hiez ), OS he ttan /hEitan/ v s . he it /he it/, OF heita v s . he it (later hi it), ON heita vs. het /he it/, all meaning •command1 or •call*. Since the time of Jakob Grimm until the appearance of van Coetsem*s 1956 monograph on the subject, numerous

The P r o b l e m

121

explanations had been proposed to account for these phe­ nomena.1? They all tended to take one of two tacks, as summarized by Lehmann (1955*58)* Two theories have been proposed* 1. that the seventh class preterites are from reduplicated forms like those of Gothic; 2. that they and the seventh class presents developed from forms with ablaut grades dif­ ferent from those of the first five classes... The strength of the first theory lies in the presence in Gothic of reduplicated preterites for seventh class verbs; its weakness lies in the impossibility of sug­ gesting a development in accordance with sound laws from such forms to the preterite forms that occur in NWGmc. Accordingly, Lehmann opts for a theory originally pro­ pounded by Karl Brugmann whereby the NWGmc. past forms have no genetic relationship to the reduplicated forms of Gothic. Instead, the NWGmc. forms devolve from ablauting verbs in Germanic which existed side-by-side with the reduplicating ones. These G m c . ablauting verbs Lehmann derives, with some assistance from his laryngeals, from IE ablauting forms. This solution has generally been felt to be unsatis­ factory, first because it merely relocates the problem into Proto-Germanic and second because the type of ablaut which must be posited is not found in any of the other IE lan­ guages.10 In 1956 van Coetsem proposed an account according to which seventh-class St V*s in Germanic originally did in fact form their nonparticipial pasts through reduplication as in Gothic. But then, under the influence of the major­ ity of St V fs like class 2 /beugan/ *bendf vs. Pst /baug/ and class 3 /werþan/ #become* vs. Pst /warþ/, an /e/-/a/ ablaut alternation was analogously introduced into the seventh class. But according to van Coetsem (1956*55)* this ablaut was "in umgekehrter Richtung". That is, where­ as in the other St V classes the /e/-to-/a/ alternation marked present vs. past, in the seventh class it marked past vs. present. Thus from the Prs /haitan/ •command’ the Pst /heit/ was formed; from /haldan/ ’hold* the Pst /held/ was formed; and the Pst /euk/ was formed from the Prs /aukan/ *increase*. Van Coetsem’s theory has found wide acceptance, par­ ticularly in view of the general dissatisfaction with the other theories that had previously been proposed. However, this acceptance has not been universal. As Bech (1969« esp. **9-5*0 and Antonsen (1 9 6 5 * 52, footnote) have pointed out, positing the diphthong /ei/ as the antecedent of / e */ in the NWGmc. forms like OE he it, OHG he *z, etc., runs afoul of some fairly well substantiated Gmc. phonological changes. One of these is incorporated in our rule G (part a) above

122

Germanic

and N o r t h w e s t

Germanic

whereby IE (and early G m c .) /ei/ appears later as /ii/ as in IE /steigh/ to Gmc# /stiig/. According to van Coetsemfs account, /ei/ followed by /i (i), j/ in the next syllable became / i 1/ and elsewhere developed as /ei/ to /ee/ * /ei/. Hence the Pst form /heit/ became /he it/, etc# However, all the Pst Sub forms of /heit/ had endings with /ii/, which makes it remarkable that a form like •commanded* does not occur occasionally in one of the earlier attested NWGmc. languages. And if one considers the first-class St Y*s like /stiig/, the only place in the entire Prs paradigm where / 1 / occurs in the following syllable is the 2 nd and 3rd Sg Prs Ind, /-is/ and /-id/. This leaves thirteen other forms of the present system where /i(i), j/ do not occur and where one would accordingly expect */ste*g/ — which verb is not attested in any Gmc# language, even as a remnant# Nor is any other St V of class 1 attested in any Gmc. language with the root vowel /ei/. More recently van Coetsem has defended his account of reverse ablaut in more or less transformational terms (19?2 i207) i ...there was already in the Qîmc.]] e-a period a sort of inversion a [_i#e., /a/ from IE /o/"~] ^ e in the reduplicating class (in comparison with e ~ a of the e-group Qi.e., the other St V classes, particularly the first fivej )#.. This situation may have operated as a trigger to the application of polarity inversion, which as a general phenomenon can be formulated in the simple ruleI QxfeatureJ — > [_-*featurel (ChomskyHalle 1968«355-6 Q The Sound Pattern of English, Harper pan *callf vs. hreiop. A few verbs with the MS of 'C0 /ann/ also change the /a/ to /eio/, e.g. bannan *ban* vs. belonn. (3) In addition to these two main classes, the following 0E remnants of reduplication are attested* beoft and beft2? from be satan *beat#, dreord from drgsdan

The

Germanic

and

Northwest

Germanic S i t u a t i o n

127

•fear*, heht (perhaps heiht) from haitan •command*, leolc from la ican ’jump*, leort from laeitan ‘let’, reord from rætdan 'advise*^ (Til these ec^s may be either /eio/ or /eo/. If the latter, then /eo/ arises by the rule of OE breaking from /e/ followed by /1/ or /r/ and .) Also speoft from spa »tan •spit 1 and blefla^Q rrom bio»wan *blow* . There are very few OE aberrations from the above classification» Campbell cites a swe »op instead of »swe »p as the Pst of swa »pan •sweep1^ Änd Sievers (1951 0^0* 8396 , Anm. 10) notes, ••Bei den Verben auf w ^such as cna»wan •know'H findet sich altws. und mere, im Prat. bisweilen e_»_ statt e »0 .** ii. In Old High German, these verbs tend to fall into the same two classes as the first two given for Old Eng­ lish. (1) Verbs with the root vowels /a/, /a»/ (from /e»/ in the forms under GF,*), or /ei/ (from /ai/) change the root vowel to / e »/ (later /ie/ by rule J above), e.g. haltan 'hold* vs. he »It (also healt, hialt, and hielt — probably all versions of rule J), r a »tan *advisef vs. re»t, heizzan •command* vs. hetz. The /e »/ of the Pst is on occasion written ejl as in firleizss 'left* and biheilt 'held*. The spelling beheialt also occurs, probably for /behealt/._ Verbs of this class with the MS of C /a/ J_-hnasal CJ + Q-nasal C] may, apparently froS the earliest OHG times, optionally change the root vowel to short /e/ as opposed to /e«/, hence gangan *go* vs. geng, ha Ihan *hang* vs. h eng. (2) Verbs with the root vowel /ou/ (from /au/T, / 0 »/ (written 0^, also from /au/), or /uo/ (from / 0 */) change the root vowel to /eo/ (later becoming Vio/ and /ie/). E.g., hloufan •run* (from /hlaupan/) vs. hleof,31 skro»tan •cut* vs. skreot, hruofan •call* vs. hreof. (3) As in Old English, there are in addition to these two classes the following remnants of reduplication. As we shall trace in detail directly, these OHG forms contain re­ flexes of an /-ez-/ infix which derives from Gm c . forms like /sezo»/ 'sowed* in GF,* above. E.g., the Pst PI pleruzzun and Pst Sub pleruzzi, both from the Inf pluozzan *sacrifice * (NWGmc. / b í ö * t a n / ) The Pst forms have the infix /-er-/ (from earlier /-ez-/) inserted after the initial consonant cluster, £l +er -h uzz •+- endings. Also, the Pst PI biruun /biru»un/ and the Pst 2 Sg Sub biruuuis /biru»wi»s/ from bu »an •dwell* have this same /-er-/ infix after the initial consonant. In this particular OHG dialect (the South Rhenish Franconian of the Otfrid text), stressed /e/ before any high vowel is regularly raised (see rule F* above) as in gibu *1 give* vs. the Inf geban. So /-er-/ is realized in these environments as /-ir-/» /b-er-ui-un/ — ^ /biruiun/ biruun, and /b-er-u»-i »s/ — > /biru»i»s/ (with /w/ inserted as a glide between

12 8

Germanic

and N o r t h w e s t

Germanic

/ui/ and /ii/) — /biruiwiis/. The Pst Sg screrot^^ from scroitan ‘cut* also has the /-er-/ infix. There are also some later Bavarian forms with the same infix i steraz, stiriz, sterc (i.e., sterz), and stirz — all Pst Sg forms of stoizan *pushT7 These forms reflect the progressive weakening and eventual disappearance of unstressed /Oi/ in /st-er-Oiz/. Finally, the rule of /-er-/-insertion seems to have extended its earlier domain to apply to a few lst-class St V*s with the MS of C, V, namely the Pst PI scrirun, the Pst Sub scriri, and xhe Pst Prt giscriran, all from scri »an fshoutr, as well as the Pst Prt pespiren from pespiian *spit*. iii.

In Old Saxon, two classes of these verbs appear which correspond to classes 1 and 2 of Old English and Old High German. (1) Verbs with the root vowel /a/, /ai/, (from /ei/), or /Et/ (from /ai/ and written e O change the root vowel to /ei/ (also written e_»., later becoming /ie/). E.g., haldan *hold* vs. hield, la»tan flet* vs. le »t or liet, he »tan /hEitan/ •command* vs. he »t /he it/ or hiet^ Verbs of this class with the MS of CQ -h /a/ -h /1# m, n, or r/ + may optionally (apparent­ ly from the earliest attested OS times) change the root vowel to short /e/ instead of long /e»/, e.g. gangan fg o # vs. geng, haldan •hold* vs. held, faihan •catch* (from /fanhan/) v s . feng, spannan kspan* vs. spenn. (2) Verbs with the root vowel /0»/ (from NWGmc. /au/ and written o») or / 0 1 / (later /uo/) change the root vowel to /eo7~(later /ie/), e.g. hlo»pan /hlO»pan/ •run* vs. hleop, hro »pan /hroipan/ *callf vs. hreop. One verb with the syllabic nucleus /au/ and the Pst still in /eu/ is attested, hauwan *hew* vs. heu /heuw/. Here NWGmc. /au/ and /eu/ were retained instead of be­ coming /0»/ and /eo/ because of the following /w/.

iv. The Old Norse classes la and lb below correspond to the 0E-0HG-0S class 1. ON class 2 corresponds to 0E-0HG-0S class 2; and ON class 3 corresponds to 0E-0HG class 3 in that it develops from a remnant of reduplication, namely the /-ez-/ infix mentioned above, (la) Verbs with the root vowel /a»/ (from NWGmc. /e«/) or /ei/^ (from /ai/) change the root vowel to /e »/ (written e). e.g. blasa fblow# vs. blés, heita •command* vs. h é t .3^ One verb of this class has the Prs vowel /o»/, blóta •sacrifice* vs. blét, which may be an instance of / o »//e»/ ablaut, (lb) Verbs with the MS of C + /a/ + /1, m, n, or r/ + C, change the root vowel t o / e / , e.g. fa •catch* (from /fanhan/) vs. Pst Sg fekk (from /feng/T” vs. Pst PI fengom, falla *fall* vs. fell, hanga *hang* vs. Pst Sg hekk vs. Pst PI hengom. Verbs of this class with the MS of C ■+- /a/ -I- /ng/ may also form the Pst PI with /i/ as in fingom and hingom. (2) Verbs with the root vowel /au/ or V u »/ or with the MS of C ■+■ /pggv/ (from NWGmc. +- /auw/ by Holtzmann*s Law)°change the

The

Germanic

and N o r t h w e s t

Germanic

Situation

129

root vowel or /pggv/ to /jo«/ (from NWGmc• /eu/ and written jo). E.g., hlaupa •run* vs. hijóp, bua flivef vs. Pst Sg bjó vs. Pst PI bjoggom (as a regular ON de­ velopment from /beuw-um/)f hQ^gva *hit# vs. h jó and hjoggom (from /heuw-um/) . (3; Verbs with the MS of dr i- root vowel# where the consonant cluster C., = /s/ or /r/ or a cluster with at least one sonorant1/!* m, n, r / in it insert after the cluster C-, the infix /-er-/ (from /-ez-/). Such verbs delete the Prs-tense vowel and take the Wk V inflection in the Pst« gnua #rub* vs. gnera, gróa •blossom* vs. grera, roa ’row* vs. rera, sa ^ o w ^ vs. sera. Class 3 takes preced­ ence over classes la# lb, and 2 in that a verb like gnua, even though it has the root vowel /ui/, does not inflect according to class 2 as *gnjó. The infix /-er-/ has also been extended to one oth-class St V which meets the MS conditions of class 3» C, ■+■ root vowel I Prs sla •hit* vs. Pst slera. v. In Old Frisian, the latest attested of these larguages, there are two classes« (1) Verbs with the MS of CQ + root vowel ■+■ nasal consonant +change the root0 vowel to /e/. Verbs in which C-, is a nonnasal con­ sonant may optionally change the root vowel to /i/. E.g., h w a « fhang# vs. heng or hing, gain or gunga •go* vs. geng or ging, banna 'ban* vs. benn. (2) All other verbs change the root vowel to /ei/ (later /i«/), e.g. heita *be called* vs. he it or h i 11 , le «ta •let* vs. le «t or li »t , ha »Ida 'hold* vs. heild or hi «Id, hla «pa #run' vs. hie «p or hli »p. The task which now confronts us is that of giving a plausible account of the changes which occurred in the early NWGmc. grammar GF, which produced the output GF,f, changes which resulted in grammars which produced the forms cited under M.i to M.v above. This has indeed proved to be the traditional stumbling block, as noted by Lehmann in his re­ mark above on "...the impossibility of suggesting a devel­ opment in accordance with sound laws from such forms Qas those given under GF-. to the preterite forms that occur in NWGmc. Las given under M] ." 3.2.3. An Account of the Developments In considering the changes in the forms under GF,* which gave rise to those under M, we shall refer to trie following model of linguistic change« N.

EARLIER .GRAMMAR ---- 2-- — £

•

EARLIER*OUTPUT

3—

""^1

LATER .GRAMMAR

• * LATER^OUTPUT

That is, change between two stages of a language may be viewed in terms of at least three relations obtaining

130

Germanic

and N o r t h w e s t

Germanic

between an earlier grammar and its output as opposed to a later grammar and its output. Relation 1 in N is that be­ tween an earlier output and a later one. Traditional Hsound laws" of the kind alluded to by Lehmann such as Grimm*s or Verner#s Laws have usually been formulated in these terms. Relation 2, which has in recent years re­ ceived some attention, is that between the configuration of an earlier grammar vis à vis that of a later one. In de­ scribing change in terms of this kind of relation, one tends to speak of rule deletion, rule expansion, rule reordering, rule simplification, and the like. Finally, relation 3 has as yet received comparatively scant attention in the liter­ ature. It has to do with how the output of a grammar at an earlier stage is reanalyzed by speakers at a succeeding stage when they are in the process of constructing the rules of their own grammars. In the following we shall be con­ cerned exclusively with relations of type 2 and 3 ^ 5 The first in the series of changes in grammar GF^ in­ volved a relation of type 2. This was a change in the stress rule 3 of G F , . After this change the stress was placed on the reduplicative syllable instead of the root vowel of reduplicating verbs. Meid_(1971s9?) gives a re­ motely similar account: "Der^Grund !_for the NWGmc. changes at issue herej liegt in dem Übergang vom beweglichen zum festen, an die Anfangssilbe des Wortes gebundenen Akzent.” (Original italicized.) However, Meid also posits the Go. stress as having been located on the reduplicating syllable, which was not the case. This change in the NWGmc. stress rule was to be ex­ pected. Already in Gothic the reduplicating prefix as op­ posed to other kinds of V prefixes like /bi-/ or /ga-/ must have been felt to be more closely bound to and in fact an integral part of the V root. This is indicated by the fact that the insertion of particles like /u(h)/ and /þO:/ occurs freely between prefixes like /bi-/ or /ga-/ and the V root, while such particle insertion between the reduplicative prefix and the V root never occurs. E.g., ga-u-laubjats and ga-þau-laubidedeib •believe1, but never »tai-u-tok •touched1 or ♦saí-þau-slep •slept1. Since the reduplicative syllable was considered part of the V root, a shift of stress onto that syllable was really only a matter of time. Formally, this change may be accounted for in at least two ways. It may be considered either a reordering in an original sequence of stress rule 3 (minus the proviso about reduplicative syllables) —>> reduplicative rule 1 to the new sequence, reduplicative rule 1 —> stress rule 3» Thus the earlier derivation would be /hald/ fheld# (3) — /háld/ (1) /hehaId/; and the new derivation /hald/ (1) — > /hehald/ (3) — /héhald/. Or, the change may also be de­ scribed as a simplification, i.e. an expansion in the domain of the stress rule 3 in that the proviso "Not; reduplicative

An A c c o u n t

of

the

D evelopm ents

131

prefixes from rule 1 above.** was deleted. After this change, the resulting new grammar GF« was as follows« GF2 . The same as GF. except that the proviso **or reduplica­ tive prefixes rrom rule 1 above** is deleted from rule 3• And the output of GF2 would be this« GF 2 '. The same as GF^* above except that the Pst forms are stressed on the reduplicative syllable, e.g, 1. /éaik/, 2. /éalþ/, 5* /bébut/, 15. /héhait/, 16. /héhlaup/, 22. /séso«/, /sézo«/, /sése«/, or /séze«/, etc. Many of the subsequent changes in GF2 must now be viewed in the light of relation 3 in figure N above — that is, in terms of how the forms under GFp* were reanalyzed and reinterpreted by succeeding generations of speakers. The principal and immediate effect of this shift in stress was the reinterpretation of the former reduplicating prefix as the onset of the verbal root and the remaining segments as some sort of internal modification of that root. This reanalysis has been posited by others, e.g. by Meid (1971«99), "Betonung der Anfangssilbe » Signalisierung des Wortanfanges...'*, and by Bech %(l969«19), "Indem die reduplizierende Bildung des Prat. als Infixation aufgefaBt wird, erscheint sie ja als eine Konjugation durch flex­ ivische Mittel im Inneren des Verbalstamms von ähnlicher Art wie die ablautende Flexion der nicht reduplizierenden starken Verbalklassen.** Under this reanalysis of the forms under speak­ ers would now have to deduce the following grammatical rules for the production of these forms« GF^. To form the nonparticipial Pst of 7th-class St V*s ^ (i.e., those St V*s meeting the MS conditions set out in footnote 21)« 1. Root vowel —

/e/ +• root vowel

/ for verbs with the MS of ## vowel +4# — i.e. verbs whose roots begin with the root vowel and end in a single consonant or consonant cluster. This rule applies in the derivation of forms 1 to 5 under G F p * ( * the corresponding forms under GF^', but with the stress on the reduplicating syllable). 2. The following rules under 2 apply to verbs with the MS of ## C1 -h vowel -h CQ ## — i.e., to v erbs whose roots begin with at least one consonant. They apply in the

132

Germanic

and N o r t h w e s t

Germanic

derivation of forms 5 to 28 under GF2 # ( = GF^#). 2a. Root vowel

/e/.

2b. Copy into the verbal root immediately after the root vowel (/e/ by rule 2a) the morpheme-initial consonants) of the r o o t • 2c. Insert the original Prs-tense root vowel into the root after the Pst-tense root vowel (/e/ by rule 2a) and immediately after the morpheme-initial consonants (inserted by 2b). 3. Delete sonorant consonants occurring in morphemeinitial consonant clusters after obstruents. Rule 3 applies in the derivation of forms 7* 8, 13* 17* 2 3 , Zk9 and 25 under GF2 1 ( » G F ^ ) . 4. If the consonants inserted by 2b are /s/ or /s/ ■+ /1* m, n, r / # change the /s/ to /z/. Rule ^ applies, per­ haps optionally, in the derivation of forms 22, 2kp and 25 under GF2 f ( - G F ^ ) . 5* The stress rule is as under GF2 (and without the pro­ viso excluding reduplicative prefixes). 6. The ablaut rulei The original Prs-tense vowel /ei/ (inserted in Pst-tense forms by rule 2c) is changed in some verbs to / 01 /. Rule 6 applies in the derivation of forms 20 and 22 under GF2 # ( = GF^*). Rules 1, 2a, 2b, and 2c in GF~ correspond to rules 1 and 2 in GF^ and GF2 . Rules 3 and # in GF~ * rule 2 in GF, and GF2 • ^Rule 5 In GFo is basically rule 3 in GF1 and GF2 . And rule 6 in GFo * rule ^ in GF, and GF2 • (The reflex of Verner#s Law — which need not concern us any further here — applies in the derivation of a few forms like /fathan/ •catch* vs. Pst /féfang/.) Some examples of derivations by the rules of GF^ are thesei (i) /aik/ •assert* (1) — /eaik/ (5) — > /eaik/, (ii) /groi/ *grow# (2a) — >> /gre/ (2b) — > /gregr/ (2c) /gregro•/ (3) /gegroi/ (5) /gegroi/, (iii) /hait/ •command* (2a) — > /het/ (2b) — /heht/ (2c) — > /hehait/ (5) — > /héhait/, (iv) /sleip/ •sleep* (2a) —> /slep/ (2b) — 3> /sleslp/ (2c) — £> /slesleip/ (3) — £> /sesleip/ (4) — > /sezleip/ (5) — > /sézleip/, (v) /sei/ •sow* (2a) — > /se/ (2b) —*> /ses/ (2c) — > /sesei/ (4) /seze«/ (5) — î> /sezei/ (6) — > /sézoi/. We emphasize that this output pro­ duced by GF^ is exactly the same as that produced by GFp given under-^GFg* above. But the rules of GF^ are of coarse very different from those of GF2 because of the reanalysis of the original reduplicative vowel as a root vowel. That is, speakers now thought of these forms in different terms

An A c c o u n t

of

the

D evelopm ents

133

than they had previously. Bech (196 9 ) has quite correctly noted an additional instance of reanalysis which can be incorporated into the framework of grammar GF^. This was the development of the OE-OHG-ON infix /-er-/ \from earlier /-ez-/ by a familiar rule of NWGmc. phonological change whereby /z/ became /r/ ) . We have noted all the NWGmc. occurrences of /-er-/ under M above. The occurrence of such forms indicates that at an early stage the phonological sequence /ez/ wa reinterpreted as a Pst-forming infix. This means in our terms that rules 2 and 4 in GF^ were reanalyzed as follows« 2*. The rules under 2 continued to apply to verbs with the MS of ## -t- vowel H- C ##, but added the proviso that could not be /s/ or / s/ 4- /1# m f n, r/. The rules under 2 thus ceased applying to such verbs. 4*. Rule 4 now applied to verbs with the same MS as under 2*, but C-, had to be /s/ or /s/ 4 /1, m, n , r/. Rule 4 now inserted /ez/ after the initial consonant (? con­ sonant cluster) of such verbs. Again we emphasize that the output of GFo with the rules 2* and 4 # is exactly the same as that of^GF^ with the earlier versions 2 and 4. The difference is that speakers with rules 2* and 4* in their grammars now thought of the forms derived in terms other than they had previously. For example, the derivation of the Pst /sézoi/ 'sowed* from /set/ and /sezleip/ •slept* from /sleip/ now differed from that given in the preceding paragraph« /set/ (4*) — £> /s-ez-e«/ (5) — > /sézet/ (6) — 5> /sézo«/, and /sletp/ (4*) -> /s-ez-letp/ (5) /sézletp/. The environment of rule 4* was soon generalized by a number of speakers to insert /ez/ not just after one ini­ tial consonant (C±), but after the initial consonant clus­ ter (C,). This type of insertion is indicated by the oc­ currence of forms like ON gr-er-a 'blossomed*, not ♦g-er-ra, and OHG pl-er-u z z u n ^ sacr if iced *, not *p-erluzzun, as cited under M above. And for some speakers rule consonant clusters beginning with /s/ and followed by /1, m, n, r/, but to clusters beginning with any consonant followed by /1, m, n, r/. The grammar GF^ with rules 2* and 4* was also ripe for some other changesin that certain of its rules proved to be tenacious while others were ephemeral and soon dropped from the grammar. Two of the tenacious rules were 1 (root vowel — > /e/ -t- root vowel) and 2a (root vowel — $> / e / ) . These rules were retained probably because of their similarity in form and function to the NWGmc. ablaut rules (cf. Go. rule 1.1.1 in chapter 2).Two other tenacious

134

Germanic

and N o r t h w e s t

Germanic

rules in GF^ were 2b which inserted the root-initial con­ sonants) into the root and rule 4* of /ez/-insertion. These rules had analogues in the NWGmc. grammar and thus constituted grammatical processes familiar to NWGmc. speakers. For example, an already existing rule of con­ sonant insertion into a root was that of /n/-insertion for the Prs and Pst Prt of /stad/ fstand* as in the Inf /standan/ vs. the Pst /sto»d/ (cf. the Go. rule 1.2.5 in chapter 2). This could of course also be considered a rule of /n/-deletion. But the point is that there existed in NW Germanic morphemic alternates with and without spe­ cific morpheme-internal phonological material, so rules like 2b and ^* which inserted such material would not have seemed unduly bizarre. On the other hand, rules like 2c and 3 in GF^ had no analogues in the grammar of NW Germanic (nor in Gothic). Rule 2c required for the Pst the re-insertion into an un­ stressed position within the root of the original stressed vowel of the Prs, a vowel which had already been "ablauted" by rules 1 or 2 a . And rule 3 deleted sonorant consonants morpheme-initially after obstruents. There are no rules like these in Gothic and were doubtless none in NW Ger­ manic. Hence these rules were exotic in the extreme and gradually dropped from G F ^ . By "gradually" we mean that rules 2c and 3 became at first optional so that for a time two kinds of Pst forma­ tions like /héhait/ *commanded* with 2c in its derivational history as well as /héht/ without^it occurred. ON forms like sera *sowedf from NWGmc. /s-éz-oi/ indicate that be­ fore the final loss of rule 2c, verbs with the MS of ## -h vowel ## (i.e., verbs whose roots terminated in the root vowel) were susceptible of reanalysis. Since most of these verbs had /ei/ or / 01 / as the root v o w e l , 3° these vowels tended while rule 2c was still in the grammar to be interpreted as Pst-tense Wk e n d i n g s . 37 Hence /sézoi/, •sowed* was analyzed as /séz/ + Pst Wk ending and /blebloi/ (earlier /bébloi/) *blossomedf as /blébl/ + Pst Wk ending. Then unstressed and word-final /ei, o*/ became /a/ through normal NWGmc. phonological change. This accounts for the development of ON class 3 under M.iv above with Pst form 3 like sera conjugated as Wk V*s and also for the occurrence OB blefla *bloomed* noted in footnote 30 above. The form /blébloi/ was realized as //blébloi// by the NWGmc. correspondent of Go. rule 2.2.2 whereby voiced obstruent consonants were realized as continuants after vowels and as stops elsewhere. Hence /bléblo»/ - / / b l e M o i / / becomes OS //blébla// or perhaps //blévla//, written blefla. After these changes, grammar GFn had become the following« D

An A c c o u n t

of

the

D evelopm ents

135

GF^. To form the nonparticipial Pst of ?th-class St V*s: 1. Root vowel —

/e/ -I- root vowel

/ for verbs with the MS of ## vowel +

##.

(The same rule as in GF^.) 2. The rules under 2 apply to verbs with the MS of ## C, -I- vowel + C ##, where C, is not /s/ or /s/ + /1, m, n, r/. 2a. Root vowel — >

/e/.

(The same rule as in GF^.)

2b. Copy into the verbal root immediately after the root vowel (/e/ by rule 2a) the morpheme-initial consonants of the root. (The same rule as in G F ^ .) 2c. 0. (The rule inserting the Prs-tense vowel into the Pst form is being lost or is completely lost.) 3. 0* (The rule deleting sonorant consonants morphemeinitially is being lost or is completely lost.) 4. Insert /ez/ into the root after the initial conson­ ant (s) / for verbs with the MS of ## /s/ ( 4- op­ tionally /1, m, n, r/) + vowel + CQ ##. 5. The stress rule.

(The same rule as in GF-j.)

6. 0. (The ablaut rule applying to vowels inserted by rule 2c, which is lost, is also lost.) In considering the changes in GF^, we must turn to some recent work by Shibatani (1973) and by Cooley (1978), who have investigated the role played by so-called "surface phonetic constraints" (hereafter SPC) in phonological change. The following remarks of Shibatani (pp. 88, 96, and 10^, respectively) will concern us here: ...SPC#s state possible and impossible combinations of phonetic features at the phonetic level... My claim is that there is a single structure, and that it is the SPC's of a language upon which a native speaker bases his judgment when confronted with a nonsense form... The procedure for determining the well-formedness of a given nonsense word is quite simple. The word is checked against all the SPC's of a language, and unless it conflicts with any SPC, the word is judged to be possible. Many phonological rules have the function of reconcil­ ing a phonetically unpermitted sequence of segments,

136

Germanic

and N o r t h w e s t

Germanic

created by affixation etc., with what is compatible with SPC*s. The observation that SPC’s — or something like them _ can trigger phonological change is not new. Luick, for example, in treating the changes in Eng. consonant clusters caused by the reanalysis of nominal compounds, notes the following (1964 1965 -0 )« Besonders starke Konsonantenhäufungen in einer ur­ sprünglichen Kompositionsfuge treten dann auf, wenn , der erste Teil des Kompositums selbst ein Kompositum war, also in Verbindungen von der Form (a + b) +■ c, und wenn das Gefîihl £Ür die Komposition schwand. Dann wurde der ursprüngliche Nebenton auf b zu voll­ kommener Tonlosigkeit reduziert, was...zur Folge hatte, daß die Vokale dieses Gliedes ausfielen. Weiter wurde die nun entstandene Konsonantengruppe reduziert,„bis eine geläufige Folge oder nur ein Konsonant ubrlgblieb. (Emphasis supplied.) There are numerous instances of such change attested in Germanic. Luick (op. cit.) mentions the development of OE healfpenigwurcT fhalf-penny worth1 as being to *halfpnwurth to »halpnwurth to halporth to ha*p*orth //héTpJþ/7. A number of English place names evince a similar development such as Alscot from Alvescot, Ardley from Eardwulfesle »ah, and Buscot from Burgwardescot. Another such change which occurred independently in a number of Gmc. languages is the development of the word w a s p . The word was originally /wef-s-a/, where /-s-/ was a derivative morpheme (probably the same as the Go. derivative suffix 45 /s/ in section 3 «2 of chapter 2), the /-a/ the inflectional ending, and the root /wef/ from /we / *weave*. When the / - S - / was no longer perceived as a derivative suffix, the morpheme boundary between /f/ and /s/ was lost and the form was reanalyzed as /wefs-a/• But there existed in these lan­ guages no morphemes with the MS of ## C + vowel ■+■ /f/ + /s/ ## or of ## C -b vowel +■ /s/ +• /f/ $#.3° The closest approximation to w# C +■ vowel +■ /f/ +■ /s/ ## which con­ stituted an acceptable MS in these languages was to per­ mute the /f/ and the /s/ and to make /f/ a stop. This re­ sulted in Eng. wasp and independently ModHG Wespe with the MS ## CQ + vowel -h / s/ +- /p/ ## as in words like lisp. What seems to occur in change of this kind is essen­ tially a threefold process. First, a morpheme boundary of some sort, say that between the elements of a compound or between a stem and a derivative morpheme, is lost so that the resultant form is considered monomorphemic• Second, the new morpheme is subject to the SPCfs of the language. Third, this entails either the deletion of segments or their change into an acceptable cluster — as Luick puts 1** "eine geläufige Folge". When change occurs under these

An A c c o u n t

of

the

D evelopm ents

137

circumstances, there invariably seems to be a sort of sim­ plicity criterion involved in that the new sequence is formed from the older one by changing the fewest features possible to fit it into the SPC*s of the language* Hence in the example of wasp given above, the permuted sequence /sf/ was changed to /sp/ and not, say, to /st/, although morphemes with the structure ## C + vowel ■+• /s/ + /t/ ## (fast, etc.) did of course occur in these languages. Fi­ nally, we note parenthetically that the effect of the SPC*s on the phonological level is analogous to that of so-called ••popular etymology" on the lexical level. In change of this kind, unfamiliar lexical items are replaced by the phonologically closest extant lexical items already present in the language, e.g. in some Eng. dialects sparrow-grass from asparagus, Standard Eng. gooseberry from grose-berry, and MoáHCr Seehund fsealf, i .e• ’sea-dog* from earlier Seelhund •seal-dog *• Such SPCfs effected a number of changes in G F^. One of these resulted from the gradual loss of rule 2c. This loss meant that some unusual consonant clusters would be produced, e.g. the Pst of /speit/ fspit# (2a) — ^> /spet/ (2b) —$> /spespt/, to which 2c would not apply to derive Vspespeit/. But forms like /spespt/ are not attested i All the attested NWGmc. remnants of reduplication given under M above are congruent with the SPC*s of their respec­ tive languages. Hence OE heht •commanded* has the struc­ ture C + vowel «+■ /h/ + /t/, which structure occurs often in other forms of the language like lioht *light*. Like­ wise, leolc •jumped* is structurally analogous to many other forms such as folc *people*. In all cases where the application of rule 2b of consonantal insertion would have produced an unacceptable sequence, the rule has either not applied or the unacceptable sequence has been modified as economically as possible to be in accordance with the SPC*s of the language. An example of the latter kind of change is the OE Pst speoft *spit* from the Inf spaitan. The Pst form /spespt/, which would have been derived by 2b, has had the second /s/ deleted; and in the resultant /spept/ the /p/ has been made into a continuant. This occurred because there were no OE morphemes with the MS of ## C +- vowel + /s/ -t- /p/ -h /t/ ## and indeed no native G m c . ones in Old English with the MS ## C + vowel +• /p/ +- /t/ ##.3° And the phonologically nearest possible structure was ## C +vowel + /f/ +- /t/ ## as in OE loft *air* . Hence the un­ acceptable clusters resulting from 2b were obviated either by being changed to agree with the NWGmc. SPC*s39 0r by being deleted in their entirety. This latter change meant in effect that rule 2b gradually ceased applying. At this juncture then there were three rules in the grammar for forming the Pst of these verbs and GF^ had be­ come the following*

138

GF^.

Germanic

and N o r t h w e s t

Germanic

To form the nonparticipial Pst of 7th-class St V fsi

1. Root vowel ->

/e/ 4 root vowel

/ for verbs with the MS of ## vowel + 2. Root vowel

##.

/e/

/ for verbs with the MS of ## C, + vowel 4- C ## where C, is not /s/ or /s/ 4* /1» m, n, r / . The rootinitial cluster C1 may be optionally inserted into the verb after the root vowel /e/, but the resulting cluster is subject to the SPC*s of the language. 3. Insert /er/ (from earlier /ez/) after the root-initial consonants (or consonant) / for verbs with the MS of ## /s/ ( +■ optionally /1» m, n, r/) +- vowel 4- CQ ##. Rule 1 in GF^ is the same as rule 1 in GF^i rule 2 in GF^ corresponds to rules 2, 2 a , and 2b in GFf/î and rule 3 in GF^ is essentially the same as rule 4 in GF^. (The stress rule 5 in GF^ still applied in GF^5 but this rule need no longer concern us here.) ^ The NWGmc. forms given under M above indicate that the three rules in GF* were in effect competing with one an­ other. For example, the OHG, OE, and ON forms with the reflex of /er/ indicate that rule 3 tended in time to apply to more verbs than it does in GF^. The ON version of the rule inserts /er/ not only if the initial consonant cluster begins with /s/ (followed optionally by /1, m, n, r/), but if the cluster begins with any other consonant as well. The ON rule has also extended its domain to apply to a St V meeting the MS conditions of rule 3* but not originally in class 7, namely to sla #hit* (Pst slera), a 6th-class verb. Through the same process of extending the original MS conditions, rule 1 in GF^ expanded its domain to apply to verbs with the MS of ## C + vowel + C ## — that is, to verbs whose roots begin with consonants, not only those beginning with vowels. This expansion was doubtless facil- , itated by the fact that the addition of prefixes to verbs was common in Germanic. If the prefix ended in a vowel and the verb root began with a vowel or a sonorant consonant, the vowel of the prefix was often elided, e.g. Go. bnauan •rub* from /bi-nuian/, Go. fret #a t e # from /fra-eit/ by the Go. rule of vowel deletion (2.1.2 in chapter 2), likewise ON gnua #r u b f from /ga-nuian/. Hence a reduplicating verb like /galþan/ •become old# could easily be formed from the prefix /ga-/ + the verb /alþan/. The Pst would have orig­ inally been /ga-ealþ/, then by elision of the préfixai vowel, /gealþ/. Finally, the /g/ in /gealþ/ was no longer

An A c c o u n t

of

the

D evelopm ents

1 39

interpreted as a prefix, but rather as part of the verb root. Hence the way was free for rule 1 in GF^ to apply to verb roots beginning with consonants. ^ In the initial stages of this expansion, rule 1 very probably applied obligatorily to verbs with the MS ## vowel + CQ ## and optionally to those with the MS ## C -t- vowel + C ##. In any event, rule 1 in G F q could now produce the following output« D GF^l1• Output of the expanded version of rule 1 in GF^i 1. /aikan/ •assert* —

/eaik/.

2. /alþan/ 'become old* — >

Pst /ealþ/.

3. /aukan/ ’increase' — £> Pst /eauk/. /bloitan/ 'honor*

Pst /bleoit/.

5. /buian/ *live' — >

Pst /beui/.

6. /haitan/ 'command* —>

Pst /heait/.

7. /leitan/ 'let*^0

Pst /leeit/.

8. /skraudan/ *cut'

Pst /skreaud/.

9. /staldan/ 'gain*

— > Pst /steald/.

These Pst forms under GF^l* are all susceptible to change since they are not in acdord with the SPC's of NW Germanic. For one thing, the vocalic nuclei of these forms do not occur among the vowels or diphthongs of Germanic (and early NW Germanic) as given under I above. For another, the Pst forms 1, 3» 6, and 8 under GF^l* have either /ai/ or /au/ in unstressed position. As indicated in section 2.3 above (see also footnote 13)» there must have been by early NWGmc times a monophthongization of unstressed /ai, au/ to / ei, oi/, Thus Pst forms like /héait/ and /skréaud/ cited above were soon realized as /héeit/ and /skréoid/. Whether this was the result of the NWGmc. SPC's or of some sort of •persistent** phonological rule of the language is immaterial The point is that such change is plausible and attested. ^ We may assume then that the affected Pst-tense forms under GFcl' were soon realized as (1) /éeik/ (2) /e'oik/, (6) /heeit/, and (8) /skréo»d/.^2 This makes the inventory of vowel nuclei in the Psttense forms under GF^l* to be /ea, eei, eoi, eu»/. Now there is some disagreement in the literature as to just what diphthongs NW Germanic might have had. Because of his views on umlaut noted above, Antonsen (1965*33) posits the most of any source, /ai, ai, ao, au, eo, eu, iu/. Some of

140

Ge r m a n i c

and N o r t h w e s t

Germanic

the handbooks mentioned in footnote 27 above posit /ai, au, eo, eu, iu/. And other of the handbooks (as well as the present writer under I above) consider the diphthongs in late Germanic and early NW Germanic to have been /ai, au, eu/ . We may assume that the new diphthongs /ea, ei, eoi, eu*/ in the Pst forms under GF*1* were soon realined accord­ ing to the SPC*s of NW Germanic, i.e. according to the in­ ventory of long vowels, short vowels, and diphthongs already in the language (as given under I above). We may further assume that these new diphthongs were transformed to their phonologically closest counterparts by changing as few fea­ tures as possible, (This type of change is well attested! In appendix 2 to this chapter we give some instances of vocalic changes of precisely this type from various NWGmc. languages.) For our purposes here, it is immaterial whether early NW Germanic had the large number of diphthongs posited by Antonsen or, as we deem likely, only /ai, au, e u / . ^3 If the latter, then the new vocalic nuclei /eoi, eui/ in Pst forms like /bleoit/ ^honored* and /beui/ •lived* would both have been reanalyzed most economically as /eu — /bleut/ and /beu/. Then this /eu/ would have been variously realized as /eo, eu, iu/, depending on the rules of the various NWGmc. languages into which it was inherited. (If, on the other hand, the diphthong inventory was as broad as that envisioned by Antonsen, then /eoi/ would have been reanalyzed as /eo/ and /eui/ as /eu/. Then these /eo/*s and /eu/*s would have been variously realized as /eo, eu, iu/, again depending on the rules of the various NWGmc. languages into which they were inherited.) This account still leaves the new diphthongs /ea/ as in /ealþ/ •became old* and /eet/ as in /lee it/ *let* unas­ similated into the early NWGmc. vowel system under I. It seems likely that by this time the early NWGmc. rules given above, K*4 (/ei/ — > (? /æi/) — > / a i / unless followed by /i/) and K*5 (/eii/ — > /ei/), had applied to the vowels under I. Under these circumstances /eei/ was interpreted most economically as /ei/. Similarly, /ea/ was also reana­ lyzed as /ei/. This is not implausible if the long vowel /ei/ is considered similar to a sequence of two short ones, /ee/. Then the difference between /ea/ and /ee/ ( « /et/) is two features, /a/ differing from /e/ by low and back. (A possible competitor might have been /at/ - /aa/, which differs from /ea/ by the same two features. The fact that the /e/ in /éa/ was stressed may well have influenced the choice of /ei/ as its representative. And the fact that / e */ was the obvious choice to represent /eei/ may also have contributed to the choice of */e i/ to represent /ea/.) After the new vowel nuclei /ea, eei, eoi, eui/ had been reanalyzed in terms of the SPC*s of NW Germanic as / e i, et, eu, eu/, respectively, and after the expansion of rule 1 in

An A c c o u n t

of

the

Dev e l o p m e n t s

141

GF^ to apply to verbs whose roots began with consonants, the new grammar looked like thisi GFg. To form the nonparticipial Pst of 7th-class St V #si 1. ~Root vowel of /a, ai, ai or æs (from /ei/)/ ”| — > _Root vowel of /au, ot9 ut/ ~ /ei/ 1

LAu/ J

_

/ for verbs with the MS of ## C -h vowel + C, ##. °

2. Root vowel — >

1 /e/

/ for verbs with the MS of ##

+ vowel + CQ ##.

(For rules 1 and 2, the following conditions prevail« (i) The root-initial consonant cluster cannot be /s/ or /s/ -+■ /it m, n, r/. (ii) The root-initial consonant cluster may be optionally inserted into the verb after the root vowel of the Pst /e, et, eu/, but the result­ ing cluster is subject to the SPC's of the language.) 3. Insert /er/ after the root-initial consonants / for verbs with the MS of ## /s/ ( + optionally /1, m, n 9 r/) + vowel + CQ ##. The Pst forms produced by GF^ were generally the same as those of GF-* except the output of rule 1 given under GF^l' above was now this« GFgl*. Output of rule 1 in G F g « 1. Pst /eik/ 'asserted* instead of /eaik/. 2. Pst /éilþ/ 'became old* instead of /ealþ/. 3. Pst /euk/ 'increased' instead of /eauk/. 4. Pst /bleut/ 'honored' instead of /bleo«t/. 5. Pst /béu/ 'lived* instead of /beui/. 6. Pst /he«t/ 'commanded* instead of /heait/. 7. Pst /le«t/ 'let* instead of /lee«t/. 8. Pst /skreud/ 'cut' instead of /skreaud/. 9. Pst /ste«ld/ 'gained* instead of /steald/. It will be noted that rules 1 and 2 in GF^ could at

142

Qërmanic

and N o r t h w e s t

Germanie

this point apply to the same verbs and were in effect in competition with each other. Now the other nonreduplicat­ ing St V*s all tended to form their Pst tenses on the basis of the MS of the particular verb. This can be clearly seen in the Go. ablaut rule (1.1.1, part A) in chapter 2, where the verbs undergo various vowel alternations contingent upon the MS of the verbal root — whether or not the root vowel is followed by one and only one consonant, by a sonorant consonant, etc. (It is thus not necessary to state the ablaut rule in morphological terms like lst-class St V, 2nd-class St V, etc.) The Go. reduplication rule (1.2.2) is similar in that its application to particular St V*s can also be stated purely in terms of their MS. We can assume that this was also the situation in Germanic and in early NW Germanic. It was thus quite natural for speakers to sort out the domains of rules 1 and 2 in GFg in terms of the MS of the verbs involved. Now a MS consisting of a long root vowel or a diphthong followed by more than a single consonant was unusual for Gothic, Germanic, and NW Germanic. In fact, it was impossible at this early stage for any St V 5 and there are consequently none attested with the MS of ##...long vowel + C2 ##. E.g., the Pst of /bindan/ •tie* was /band/, /bundum/, not */beind/, */botndum/, or the like. Since the output of rule 1 in GF^ was a long vowel or a diphthong, it tended for this r e a s o n t o apply to verbs whose roots ended in not more than a single consonant, i.e. to verbs with the MS of ## Cç + vowel •+ Cg ##. (Probably for the same reason and to avoid a structure like ## C -r long vowel or diph­ thong -I- Cp ##» condition (ii) in GFg for the insertion of the root-initial consonant(s) into xhe root tended to be dropped from rule 1.) Conversely, rule 2 in GFg which changed the root vowel to short /a/ tended to apply to verbs with the MS complementary to that of rule 1, namely to ## CQ 4- vowel + C2 ## or verbs whose roots ended in at least two consonants. After these readjustments, the grammar GF^, while not completely stabilized, reflects in the main xhe situa­ tion in the attested NWGmc. languages given under M above. As for Mi, the OE classes 1 and 2 reflect rules 1 and 2 in GFx. Only a few OE verbs whose roots end in two consonants still form the Pst by rule 1, e.g. fe ing “caught* as well as feng by rule 2. Some OE verbs with two root-final con­ sonants form their Pst in a diphthong which arises from /e/ as the result of OE breaking, e.g. f e 1 oil •fell* from earlier /fell/, fe told “folded* from /feld/. Otherwise the OE Pst-tense vowels reflect more .or less directly those produced by GFg, e.g. belatan *beat* from /bautan/ vs. Pst fre »ot from /beut/• OE class 3 has the remnants of the reduplicative condition (ii) of rule 2 and those of /er/insertion by rule 3 in GF^. When reduplication occurs, the Pst vowel was probably short /e/ by rule 2, e.g. heht or

An A c c o u n t

of

the

D evelopm ents

14 3

he it •commanded1. Finally, the Pst of verbs with the root vowel /ai/ followed by /w/ may reflect ablaut, e.g. Prs cnaiwan, 'know' from /knei-an/ vs. Pst cneiow from

7HÍU7T44. *5

--------

Mii. The OHG classes 1 and 2 also follow rules 1 and 2 in G F / . As in Old English, a few verbs whose roots end in two Consonants can still form the Pst by rule 1, e.g. eing •went* as well as geng by rule 2. Otherwise, the st-tense vowels of these verbs are those prescribed by GF^, e.g. Prs hloufan *run* from /hlaupan/ vs. Pst hleof from /hleup/. Finally, OHG class 3 contains the remnants of /er/-forms derived by an expanded version of rule 3 in GFg, a rule which is in the process of being dropped from the OHG grammar.

f

Miii. The OS classes 1 and 2 are basically like the corresponding classes in Old High German in that they also follow rules 1 and 2 in G F x . As in Old High German and Old English, there are a few OS verbs whose roots end in two consonants but which can still form the Pst by rule 1, e.g. hield #held* from /heild/ as well as the alternative Pst form held by rule 2. The reduplicative provisos i and ii of rules 1 and 2 as well as rule 3 in its entirety have been lost from the OS grammar. Miv. The ON class la corresponds to rule 1 in GF^; class lb to rule 2; class 2 corresponds also to rule l in GF/j and class 3 respresents an extended version of rule 3 in°GF/. The lone ON attestation of a verb with the root vowelD/ot/ in the Prs forming its Pst in /ei/ (not the usual /joi/ from NWGmc. /eu/; is biota •sacrifice* vs. Pst blet, not *bljot. This may go back to a NWGmc. form with ablaut* Prs /bioit/ (ablaut) — > /bleit/ (later /blait/), then by rule 1 in GF^ — ^ /bleit/. On the other hand, blet may represent an ON innovation in that the expected ON Fit” #/bljo*t/ from NWGmc. /bleut/ would have had the initial consonant sequence ##blj-, a sequence which seems to have been extremely rare in ON words. Mv. Finally, the OF class 1 is a reflex of rule 2 in GF^. Class 2 is a reflex of rule 1 in GFg as it applied to the Prs root vowels /a, ai, at/. The part of rule 1 apply­ ing to /au, oi, u*/ to derive /eu/ as well as rule 3 in its entirety has been lost in Old Frisian. (The NWGmc. MS constraint inhibiting the sequence long vowel + C% was also lost in Old Frisian through regular phonological change, e.g. ha «Ida *hold* from earlier /haldan/ where /a/ before /16/ was lengthened.) The OF and ON Pst forms with /i/ such as ging *went* probably result from later developments of / g e n g / a s derived by rule 2 in G F ^ . 3.3* Concluding Remarks To sum up the preceding, we have argued that Gothic

144

Germanic

and N o r th w e s t

Germanic

and Germanic had reduplicating 7th-class St V's which developed, into the NWGmc. "ablauting" St V #s of grammar GFg above. In Gothic (and Germanic) the stress rule placed the accent on the root of the Pst forms of these verbs. The unstressed reduplicative prefix was thus felt by speakers to be just that, a prefix. But with the shift of the stress onto this prefix in NW Germanic, speakers began to reana­ lyze the former préfixai vowel as some sort of modification of the root. The subsequent changes in NW Germanic then followed quite naturally and as it were almost automatically. After this shift in stress, the subsequent sequence of de­ velopments as resprsented in GF2 to GFg does not necessar­ ily reflect the precise historical chronology of the changes depicted. We have described the stages GFp to GFg in order to show how the reduplicated forms of Proxo-Germanic could through reasonable and attested types of change eventually become the ablauting forms of NW Germanic. The grammars GF2 to GFg are intended to illustrate the types of change which must have occurred — rule expansion, rule loss, reanalysis, and the effect of the SPCfs of NW Germanic•

Appendix

1

145

Append ix 1 The Reduplicating Verbs in Germanic The following is a list of the reduplicating verbs of Germanic not given under GF,f above* Our enumeration of these verbs is based on Feist (19071*44 8 - 57 )» who recon­ structed the Gmc. forms from 7th-class verbs attested in various Gmc* languages. Our enumeration differs slightly from Feist#s» We do not include here /dreidan/ •dread1 since the OE verb develops from a reanalysis of a verb formed from the prefix /and/ + the verb /reidan/ (21 in GF, *) to /an/ + /dreidan/. We also exclude Feist*s forms /bfluian/ and /gnu«an/, both meaning •rub*, since they de­ velop from the prefixes /bi/ and /ga/ + the verb /nutan/ (28 below). Some of our reconstructions differ from Feist’s, e.g. /nutan/ which Feist posits as nauan, perhaps assuming that the Go. rule of vowel lowering (2.1.6 in chapter 2) was Germanic. All reduplicating verbs with /ei/ appear in later NW Germanic with /a«/. Finally, a verb /speitan/ 'spit* not included by Feist should probably be included in our listing. (1) /aikan/ 'rage*, (2) /ausan/ 'pour*, (3) /auþan/ •be fated*, (4) /bannan/ 'ban*, (5) /beigan/ ’fight*, (6) /betsan/ •bark*, (7) /blandan/ ’m i x ’, (8) /bletan/ ’blow’, (9) /bleisan/ ’blow*, (10) /breidan/ ’roast’, (11) /brotkan/ 'build*, (12) /fetan/ 'criticize*, (13) /floian/ ’flow’, (14) / f l o Ikan/ ’complain’, (15) /fraisan/ ’tempt’, (1 6 ) /greitan/ 'weep*, (17) /hauwan/ ’hew’, (18) /hloian/ ’low (such as cattle)', (19) /hnaupan/ 'pluck’, (20) /hweisan/ (possibly also /hw oisan/) 'cough', (21) /hwe«tan/ ’drive’, (22) /hwotpan/ ’brag', (23) / k n e tan/ 'know’, (24) /kreian/ 'crow', (25) /loian/ 'despise’, (26) /maitan/ ’cut’, (27) /me «an/ 'mow', (28) /nuian/ 'rub', (29) /prangary' ’push', (3 0 ) /roian/ *row', (31) /saltan/ ’salt*, (3 2 ) /skaiþan/ ’separate', (33) /skaldan/ ’criticize', (34) /skeiþan/ •yield', (35) /so«an/ 'kill*, (3 6 ) /spaldan/ ’split’, (37) /spannan/ 'span', (38) /spoian/ 'help', (39) /swaipan/ 'swoop', (40) /swoigan/ 'howl*, (4l) /taisan/ 'pluck*, (42) /teikan/ ’touch’, (43) /þlaihan/ ’love’, (44) /þretan/ •turn*, (45) /walkan/ ’roll’, (46) /wallan/ ’flow’, (47) /waltan/ 'roll*, (48) /we«an/ 'drift', (49) /woipjan/ 'weep', (50 ) /wroitan/ 'dig up'.

146

Germanic

and N o r t h w e s t

Germanic

Appendix 2 Attested Changes Analogous to those Posited in 3*2.3 The first three changes given below were triggered by the loss of a morpheme boundary. This was followed by the reanalysis of the two resultantly adjacent vocalic segments as a single monomorphemic vocalic nucleus according to the SPC*s of the particular language involved. These three changes thus constitute attested instances of the type of change posited above whereby GF ModHG Feuer. 4. In early Old High German there was a rule which inserted the same suffix /ir/ mentioned in the immediately preceding example between the root and the inflectional endings of certain neuter nouns in all cases except the Nm or Ac Sg, e.g. NmSg lamb •lamb*, GnSg lemb-ir-es. In later OHG times, this suffix was reinterpreted as a PI marker. Thus the rule by which /ir/ was inserted into Sg forms was deleted from the grammar. At this time the GnSg of lamb became lamb-es instead of the earlier lemb-ir-es. It should be emphasized that this does not constitute phonological change in the usual sense in that elsewhere the phonological sequence /ir/ was retained, e.g. in the comparative of adjectives such as eltiro *older*, not »alt o . Thus the dropping of the rule of /ir/-insertion for the Sg forms of these neuter nouns occasioned the loss of certain phonological segments. It thus constitutes a change of the same type as that whereby rule 2 c was dropped from grammar GF 3 , as a result of which forms like /heht/ •commanded* were produced instead of the earlier /heh-ai-t/»

148

Germanic

and N o r t h w e s t

Germanic

Footnotes to Chapter 3 ^One change, Holtzmann's Law, was formerly thought to be a common Go.-ON innovation. According to this change, Gmc. /ww/ appears in both Gothic and Old Norse as /ggw/, while Gmc. /jj/ appears as Go. /ddj/ and ON /ggw/. These have later been shown to be independent changes as noted by Marchand (1973*87), MThe fact that a Primitive Norse in­ scription (Niuwila on the Naesbjaerg bracteat of the 5th cent.) and a Finnish loan (kuva *picture*, c f . Goth. skuggwa) do not show this development j_of A*w/ to /ggw/ indicates. • .that it Qioltzmann*s Law^] cannot have arisen during a period of Goto-Norse unity, but that it is a separate development in Gothic and in Norse...w 2 Our version of this rule and the illustrative examples are based on the material in the standard handbooks mentioned in the bibliography and on that in Lloyd (1961) and Reis (197*0. ''This fact has been remarked on frequently in the literature, e.g. Bremer (1912*150) "Idg. i. is$ nur in Ausnahmefällen, zumeist einzelsprgchlich, zu e geworden.•.", Moulton (1961*8) "Noreen Q1970J fuhrt nicht weniger als 25 solche Doppelgänger [i.e., forms with both /eA o/ as well as /i, u/ such as wulafr and wolafr #wolf*^ fur das Anord. an.", Benediktsson (1967*18*0 ** [TÍtíe change of /i/ to /e/]] occurs sporadically, though quite distinctly... The cor­ responding change |_of /u/ to / 0/ is much more Qthough by no means completely^ regular." The philological evidence tends to support this con­ tention* Lehmann (1961*71)# HThe oldest proper names pre­ serve -e- in stressed syllables, even before vowels which subsequently became -i- and caused change of e to i^, e.g. Tacitus* Venedi as compared with OHG Winlda, Erminones as compared_with OHG Irmin-sul, Gepldos as compared with OE GifeJas Learlier w Similar examples are Tacitus* Segimerus for later Sigmar and Segimundus for Sigmund. The forms Teiva-variaz and Aläteiviae cited"by Antonsen (1965»52, ftn. 49) probably arise from the time in Germanic before IE /ei/ became /i*/. All these attestations in­ dicate that rule G of Gmc. umlaut must have been fairly late — unlike the vowel changes cited under D above for which there are (with the possible exception of the /a*/to-/o*/ change) no attestations of Gmc. forms as they occurred before the rules applied. ^But see our observations on Dutch given below.

F ootnotes

to

C hapter

3

149

^Similar views are to be found in Twaddell (1948), van Coetsem (1970 and 1972), and elsewhere. ^Rounded front vowels //ox// and //ui// do of course occur in Dutch, but they are from different historical sources, not from vowels followed by CQ ■+■ / i (i)/• Q

See on this Antonsen (1975*13» 30, 46, and 88). Q

Some of these are given in Voyles (1976a«21-22 and 1976b*27^-5). ^°It might be added in this connection that independent, but nonetheless similar changes have frequently been seen to occur in the Gmc. languages. Among these are the change of unstressed vowels into schwa, the change known as Holtzmann*s Law (see ftn. 1 above), the unrounding of front rounded vowels in certain German dialects and in English, the change of /it» ui/ to /ai, au/ in both German and Eng­ lish (e.g., Haus, house from /huis/ and Eis» ice from /iis/), the similar development of the indefinite article /9/-/on/ in Yiddish and a-an in English, and finally the loss of nasal consonants before continuant obstruents in the Northern WG languages as well as in certain South German dialects (e.g., Eng. five, Swabian foif vs. Standard German fünf). n 0f course, sources like Antonsen (1965*32-3) posit numerous umlauted allophonic variants so that /ai/ = //æi, ae, æe,ðw// and /au/ - //au, ao, ay//. For essentially the same reasons as those given in section 2.2 of this chapter, we do not consider these allophones to have existed in Germanic. 12

In some environments such as unstressed vowel +----- ## the diphthong /ai/ was realized as Go. /a/, e.g. the 3 Sg Prs Psv /béradai/ - ^ G o . /bérada/ *is carried1 . ^ T h e r e is philological evidence to support this in that there exist NWGmc. runic inscriptions where graphemic ai is used instead of the more usual e to represent un­ stressed word-final /ei/ (see on this Antonsen 1975*5)» This sporadic use of the etymological spelling ai for /ei/ indicates that the monophthongization of /ai, au7 must have occurred during or after the establishment of a writ­ ing system — i.e. in late common Gmc. or early NWGmc. times • ^ T h e breaking rule must have been slightly different in its earlier version to apply here in that it could apply to unstressed /i/ if it occurred in a stressed diphthong. (Such diphthongs do not occur in Wulfilian Gothic because of the monophthongization rule 2.1.5.) But breaking does not apply to /u/ under similar circumstances, e.g. /tiuhan/

Germanic

ISO

and N o r th w e s t

Germanic

•drag*, not */tíohan/. However, the breaking rule in any event applies to /i/ in some instances where it does not apply to /u/, for instance before /tr/. ^ A s s u m i n g that there was no morphemic boundary be­ tween the /heii/ and the /r/, unlike the case of the Go. verb /hi-r-i/ ‘come here’ .1 In the latter example, /hi/ occurs with various other morphemes, e.g. hi-ta •this* and hi-dre •hither*. But Go. /he«/ (Gmc. /he«I7) does not occur with any other endings. ■^The NWGmc. / g a «/ devolves from IE / g h e 1/, while /stai/ possibly comes from IE /sta/. The NWGmc. vowel alternation occurring in /gais/-/geis/ and /ga «d/-/ge «d/ was transferred to original /sta/ resulting in /stais//steis/, etc. It was also transferred to the entire Prs Ind paradigm resulting in such forms as 1 Sg Ind /ga«m//geim/, etc.

17

Extensive summaries of previous research are in Feist (19 0 7 ), Karstien (1921), Flasdieck (1936), Prokosch (1939)» and van Coetsem (1956« esp. 4-7-50) • ^ O n this see Hirt (1931-4-* vol. 2, p. 14-4).

19

Fullerton (1977«107) also accepts in the main van Coetsemfs account and posits an analogous MSeventh-class Pattern Readjustment Rule".

20

Including the most recent one by Bech (1969)» who posits an unmotivated analogical generalization of the in­ fix /-ez-/ into all the NWGmc. forms. Among many other difficulties, this explanation encounters problems in accounting for the attested Pst-tense forms in /eu/ (for which see below).

21

Because of rules like 2.1.5 (monophthongization of /ai, au/) and 2.1.6 (vowel lowering), the Go. MS conditions were somewhat different from those of Germanic. The Gmc. MS conditions on St V fs undergoing the reduplication rule were that the Prs stem should be of the form C /a/C2 or that the Prs-tense vowel should be a diphthong°or /a«, e«, otf u«/ — i.e. any long vowel except /i*/. (See the forms cited under GF^f below and in appendix 1 to this chapter.)

22

Bech (1969 ) assumes that Vernerfs Law applied con­ sistently to the morpheme-initial consonant in all V*s beginning w i t h / f , þ, h, s/. Under our account it does not matter if it did or not. See on this footnote 39 below.

23

A complete listing of the attested reduplicating V*s is given in appendix 1 to this chapter. The forms cited with /e«/ as in / l e «tan/ have /a«/ in later NW Germanic by

F ootn otes

to

C hapter

3

15 1

change K*4 above. 24 This would constitute an exception to the MS con­ dition posited for the reduplication rule 1 under GF, in that the stem vowel /a/ is not followed by two —vocalic segments. Hence this St V was possibly nonreduplicating. If so, it must have been of the 6th classi /aran/, /o:r/, etc .

26

The root-initial consonants m the reduplicated forms of V fs beginning with /b, d, g/ were realized by the Gmc. correspondent of Go. rule 2.2.2 for the distribution of voiced stop and continuant consonants as //fc, j, q //. Hence /bebú:/ w a s //behui//. 26

The latter form is produced by Verner*s Law.

2^These are Campbell (1964) and Sievers (1951) for Old English, Braune (1963) for Old High German, Gallée (1910) for Old Saxon, Steller (1928) for Old Frisian, and Noreen (1970) for Old Norse.

28

This is possibly /eo/. According to Sievers (1951» 338# §396, Anm. 1), "Die Quantität des e:o im Prät. ist nicht direkt zu ermitteln." 2^The latter form cited by Flasdieck (1936 i264), 30 J Because of its apparently bizarre morphology (which we account for below), this particular form has been con­ sidered a scribal error, e.g. by Karstien (1921il50): "Sievers...sieht blefla fur einen Schreibfehler an, was urn so wahrscheinlicher wird, als es das darunter stehende lat. insu-fla-vit übersetzt; der Schreiber is in gedanken versunken gewesen und ist durch das lateinische beeinflußt gewesen. Ja, er hat an der stelle so geträumt, daB er das fla nicht einmal zuende geschrieben hat." Such an error Is highly improbable. It seems much more likely to us that the form blefla did in fact exist in Old English as the Pst of bio:wan and that the scribe was in fact right, Sievers and Karstien wrong. ^ T h e Pst hliuf occurs in Upper Old High German be­ cause /eo/ before labials appears there as / i u / . 32Braune (1963:289, §35^. Anm. 3) erroneously derives these forms from bluoan *bloomf . The z in the OHG forms, however it may have been pronounced, stands for the reflex of Gmc. /t/. The vocalism pleruzz instead of »plerouzz in these forms is the result either of an OHG phonological rule whereby the diphthong /ou/ cannot occur unstressed and is thus reduced to /u/, or the result of the V having taken on the vocalism of 2nd-class St V's like the Pst Sg boug- 'bent* vs. the Pst PI or Pst Sub bug- .

Germanic

152

and N o r t h w e s t

Germanic

^ C i t e d erroneously as scerot in Braune (ibid.).

J One verb originally in this class, sveipa •sweep1, is usually Wk. When St, it follows in the nonparticipial Pst the lst-class St V paradigm« sveipa, sveip, svipom, sveipenn. ^ T h i s is not to say that our description will not in­ corporate certain of the explanations found in the earlier literature on the subject; and we shall of course try to give credit where it is due. But in view of the mass of this literature, we can only echo Karstien's sentiment (1921«ix), ,fWenn mir bei meiner arbeit trotz redlichstem bemühen das eine oder die andere ferner liegende behandlung einer teilfrage entgangen ist, so bitte ich um nachsicht.** Speaking of giving credit where it is due, diagram N has already appeared in Andersen (1973»76?). The type of change involved is called Mabductive". 36 "' See the forms cited under GF-. * as well as those in appendix 1. Apparently only three such verbs had a root vowel other than /e «, o«/. These were /bui/ fdwell#, /nui/ #rub#, and /snui/ •turn*.

37

^'These endings were, according to Johannesson (1923* 71), /-o«/ for the 1st person and /-e«/ for the 3rd person Pst Sg Ind, e.g. /talid- 01 / *1 told* and /talid-ei/ *he told'. 3 fl

J Such sequences did not exist in Gothic either.

See

C in section 2 of chapter 2.

39

For this reason Bech's assumption mentioned in foot­ note 22 above that the Pst of /haitan/ would have been by Verner's Law /hegait/ instead of /hehait/ is immaterial. With the loss of rule 2c, a form like /hegt/ might have been produced from /hegait/. But then by the NWGmc. SPC#s, this would have been automatically realized as /heht/ (phonetically probably //hext//), the same form as would have been derived from /hehait/. i^O

This form was by this time possibly /læitan/ or / l a «tan/ by the NWGmc. rule K'^ given earlier. This rule could not apply to the /e«/ in the Pst /lée«t/ because the /e«/ here was unstressed.

in

Campbell (196 ^ «1^8-9) remarks on the development of certain OE compounds, MIn the second element of such forma­ tions, ai^> aej_ |_more likely /e «/* ] as in unaccented sylla­ bles..., if the reduction in the semantic force of the second element was early...e.g. e »ored 'troop* (< eoh + ♦raid ) ... A development of au to oj_ and hence before m with shortening to u, appears in fultum 'help* Qfrom /full/ -+- /taum/ ^ ..."

F ootnotes

to

C hapter

3

15 3

Similar suggestions are found in Hoffory (1 8 8 5 1 5 9 7 ) and Holthausen (1 8 8 5 *620 ). ^Supportive of our view is Campbell#s observation (196^*53> 0136) that "In 0E the earliest texts have eu for Prim. Gmc. evi, and show no sign of the West G m c . differen­ tiation of the sound into eu and eo which is traceable in s ome language s ."

kk

A few of these verbs may have continued to undergo the ablaut rule in the Pst, e.g. /knai/ •know* (earlier /knei/) by ablaut — /knot/ (then by rule 1 in GF/) — > /kneu/.

I4,5

''The form cne *ow could possibly be the result of the 0E breaking of an original /et/ before /w/. Hence the NWGmc. Pst could have been /kne«w/ from /knaiw/ by rule 1 in GFg and without ablaut.

154

R efe ren ces

References Andersen, Henning. 1973* MAbductive and Deductive Change." Language 49:765-794. Antonsen, Elmer H. 1961. "Germanic Umlaut Anew." Language 37*215-239. Antonsen, Elmer H. 1965« "On Defining Stages in Pre­ historic Germanic." Language 41 :19-37. Antonsen, Elmer H. 1972. "The Proto-Germanic Syllabics (Vowels)." Toward a Grammar of Proto-Germanic. (Frans van Coetsem and Herbert L. Kufner, eds.), pp. 117-141. Max Niemeyer Verlag. Tubingen. Antonsen, Elmer H. 1975» A Concise Grammar of the Older Runic Inscriptions. Max Niemeyer Verlag. Tubingen. Beade, Pedro. 1971. Gothic Phonology« A Generative Approach. Cornell University Ph.D. Dissertation. University Microfilms International. Ann Arbor, Michigan. Beade, Pedro. 1972. "Sievers* Law in Gothic and Other Related Matters." Lingua 30:449-459# Beade, Pedro. 1973# "A New Look at Gothic Verb Mor­ phology." Leuvense Bijdragen 62:313-337# Beade, Pedro. 1973# "Remarks on Gothic Consonantism." Leuvense Bijdragen 62:125-137. Bech, Gunnar. 1969# Das germanische reduplizierte Präteritum. Munksgaard. K^benhavn. Beck, Richard. 1973# "Length and Monophthongization in Gothic." Indogermanische Forschungen 78:113-140. Beeler, M.S. 1966. "Proto-Germanic Qi^ and Qe^ 1 One Phoneme or Two?" Language 42:473-475. Benediktsson, Hreinn. 1967# "The Proto-Germanic Vowel System." To Honor Roman Jakobson. Volume 1, pp. 174-197. Bennett, William H. 1949# "The Monophthongization of Gothic ai, au." Language 25*15-21. Bennett, William H. 1964. "The Earliest Germanic Um­ lauts and the Gothic Migrations." Language 28:339-343# Bennett, William H. 1970. "The Stress Patterns of Gothic." Publications of the Modern Language Association 85:463-472.

R e fe re n ces

155

Braune, Wilhelm. 1961. Gotische Grammatik (1 6 . Auflage neu bearbeitet von Ernst A. Ebbinghaus). Max Niemeyer Verlag. Tubingen. Braune, Wilhelm. 1963* Althochdeutsche Grammatik (1 1 . Auflage bearbeitet von Walther Mitzka). Max Niemeyer Verlag. Tubingen. Bremer, Otto. 1912. MDie germanische Brechung.*• Indo­ germanische Forschungen 26:148-173. Buck, Carl Darling. 1962. Comparative Grammar of Greek and Latin. The University of Chicago Press. Chicago, Illinois. Buckalew, Ronald Eugene. 1964. A Generative Grammar of Gothic Morphology. University of Illinois Ph.D. Disserta­ tion. University Microfilms International. Ann Arbor, Michigan. Campbell, Alistair. 1964. Old English Grammar, Clarendon Press. Oxford. Carr, Charles T. 1939* Nominal Compounds in Germanic. St. Andrews University Publications No. XLI. Oxford Univer­ sity Press. London. Chomsky, Noam and Halle, Morris, 1 9 6 8 . The Sound Pattern of English. Harper and Row, New York. Coetsem, Frans van. 1956. "pas System der Starken Verba und die Periodisierung im alteren Germanischen.** Mededelingen der Koninklijke Nederlandse Akademie van Wetenschappen. Amsterdam. Coetsem, Frans van. 1968. **A Syntagmatic Structure in Development* fUmlaut* and *Consonantal Influence* in Germanic.** Lingua 21:494-525. Coetsem, Frans van. 1970. MZur Entwicklung der german­ ischen Grundsprache.** Kurzer Grundri/3 der germanischen Philologie bis 1500. Band li Sprachgeschichte. (Erich Schmitt, ed.), pp. 1-94. Walter de Gruyter & Co. Berlin. Coetsem, Frans van, 1972. MProto-Germanic Morphophone­ mics.** Toward a Grammar of Proto-Germanic (Frans van Coetsem and Herbert J. Kufner, eds.), pp. 175-211. Max Niemeyer Verlag. Tubingen. Coetsem, Frans van and Kufner, Herbert L. (eds.). 1972. Toward a Grammar of Proto-Germanic. Max Niemeyer Verlag. Tubingen. Cooley, Marianne. 1978. "Phonological Constraints and Sound Changes.** Glossa 12:125-137.

156

R eferences

D #Alquen, Richard J.E. 1974. Gothic ai and aui A Possible Solution. Mouton. The Hague. Dieter, Ferdinand (ed.). 1900. Laut- und Formenlehre der Altgermanischen Dialekte. O.R. Reisland. Leipzig. Edgerton, Franklin. 1934. "Sievers* Law and IE WeakGrade Vocalism." Language 10:235-266. Feist, Sigmund. 1907. "Die sogenannten reduplicierenden verba im germanischen." Beitrage zur Geschichte der deutschen Sprache und Literatur 32*40-7-517. Feist, Sigmund. 1939* Vergleichendes Wörterbuch der gotischen Sprache (3* Auflage). E.J. Brill. Leiden. Flasdieck, Hermann M. 1936. "Die Reduplizierenden Verben des Germanischen." Anglia 40:241-366. Fullerton, G. Lee. 1977. Historical Germanic Verb Morphology. Walter de Gruyter. Berlin. Gallee, Johan Hendrik. 1910. Altsächsische Grammatik (2. Auflage). Max Niemeyer. Halle. Gamillscheg, Ernst. 1934-6. Romania Germanica. 3 volumes. Walter de Gruyter & Co. Berlin und Leipzig. Hàugen, Einar. 1950. First Grammatical Treatise. Language Monograph 25. Linguistic Society of America. Baltimore. Haugen, Einar. 1976. The Scandinavian Languages. Harvard University Press. Cambridge, Massachusetts. girt, Hermann. 1898. "Grammatisches und Etymologisches." Beitrage zur Geschichte der deutschen Sprache und Literatur 23*288-358. Hirt, Hermann. 1931-2-4. Handbuch des Urgermanischen. 3 volumes. Carl Winters Universitätsbuchhandlung. Heidelberg. Hoffory, Julius. 1885. "Die%%reduplizierten praeterita im altnordischen." Zeitschrift fur vergleichende Sprach­ forschung 27*593-602. Hofier, Otto. 1970. "Die Germanischen Reduplizierenden Verba im Lichte der Entfaltungstheorie." Folia Linguistics 4:110-120.

R eferen ces

157

Holthausen, Ferdinand. I 8 85 . "Die reduplicierenden verba im germanischen." Zeitschrift fur vergleichende Sprachforschung 27*618-623. Hooper, Joan B. 1976. An Introduction to Natural Generative Phonology. Academic Press. New York. Janko, Josef. 1906. "Über germanisch ei 2 und die sog. reduplizierenden Praeterita." Indogermanische Forschungen 20:229-316. Jasanoff, Jay. 1973. "The Germanic Third Weak Class." Language 49*850-871. Jellinek, Max Hermann. 1926. Geschichte der Gotischen Sprache. Walter de Gruyter & Co. Berlin und Leipzig. Johannesson, Alexander. 1923. Grammatik derw%Urnordischen Runeninschriften. Carl Winter*s Universitatsbuchhandlung. Heidelberg. Jones, Oscar F. 1958. "Gothic iu." Language 34:353-358. Jones, Oscar F. 1965* "The Case for a Long u-Phoneme in Wulfilian Gothic." Orbis 14:393-405. Karstien, Carl. 1921. Die reduplizierten Perfektg des Nord- und Westgermanischen. Volume 1: Giessener Beiträge zur Deutschen Philologie, v, Munchow*sche UniversitätsDruckerei. Giessen. Kieckers, Ernst. 1928. Handbuch der Vergleichenden Gotischen Grammatik. Max Hueber Verlag. München. Kienle, Richard von. I960. Historische Laut-%%und Formenlehre des Deutschen. Max Niemeyer Verlag. Tubingen. Kluge, Friedrich. 1886. Nominale Stammbildungslehre der altgermanischen Dialekte. Max Niemeyer. Halle. Kluge, Friedrich. 1921. Die Elemente des Gotischen. Walter de Gruyter k Co. Berlin und Leipzig. Krause, Wolfgang. 1953* Handbuch des Gotischen. C.H. Beck* sehe Verlagsbuchhandlung. München. Labov, William. 1972. Sociolinguistic Patterns. University of Pennsylvania Press. Philadelphia. Lane, George S. 1963. "Bimoric and Trimoric Vowels and Diphthongs« Laws of Germanic Finals Again." Journal of English and Germanic Philology 62:155-170.

158

Re f e r e n c e s

Lass, Roger and Anderson, John M. 1975» Old English Phonology. Cambridge University Press. Cambridge, England. Lehmann, W.P. 1954• "Old English and Old Norse Second­ ary Preterites in -r-." Language 30:202-210. Lehmann, W.P. 1955* Proto-Indo-European Phonology. The University of Texas Press. Austin. Lehmann, W.P. 1961. "A Definition of Proto-Germanic." Language 3 7 • 67-75» Lloyd, Albert L. 1961. MIndo-European Unstressed Short e in Germanic." Modern Language Notes 76:847-852. Lloyd, Albert L. 1966. "Is there an a-umlaut of i^ in Germanic?" Language 4 2 :738 -7 4 6 . LÎïdtke, Helmut. 1957» "$er Ursprung des germanischen ei2 und die Reduplikationspraterita." Phonetica 1:157-183» Luick, Karl. 1964. Historische Grammatik der Englischen Sprache. Volume 1, parts 1 and 2 . Harvard University Press. Cambridge, Massachusetts. Marchand, James W. 1956. "Dialect Characteristics in our Gothic MSS." Orbis 5*1^1-151Marchand, James W. 1957» "Germanic Short *i and *ei Two Phonemes or One?" Language 33 046-355» Marchand, James W. 1970. "Gotisch." Kurzer Grundriß der germanischen Philologie bis 1500. Band 1: Sprach­ geschichte, pp. 94-123. Walter de Gruyter & Co. Berlin. Marchand, James W. 1973» The Sounds and Phonemes of Wulfila#s Gothic. Mouton. The Hague. Meid, Wolfgang. 1971. Das Germanische Praeteritum. Volume 3 of the Innsbrucker Beitrage zur Sprachwissenschaft. Innsbruck. Moore, Samuel. 1967» Historical Outlines of English Sounds and Inflections. (Revised by Albert H. Marckwardt.) George Wahr Publishing Co. Ann Arbor, Michigan. %% Motsch, Wolfgang. 1967. "Zum Ablaut der Verben in der Fruhperiode Germanischer Sprachen." Studia Grammatica VIt Phonologische Studien, pp. 119-«l44. Akademie-Verlag. Berlin. Moulton, William G. 1948. "The Phonemes of Gothic." Language 24:76-87.

Re f e r e n c e s

159

Moulton, William G. 195^. "The Stops and Spirants of Early Germanic.” Language 30:1-42. Moulton, William G. 1961. ”Zur Geschichte des deutschen Vokalsystems.” Beiträge zur Geschichte der deutschen Sprache^und Literatur 83:1-36. Our version is from Vorschläge fiîr eine Strukturale Grammatik des Deutschen (Hugo Steger, ed .), pp. 480-518, 1970. Wissenschaftliche Buchgesellschaft. Darmstadt. Moulton, William G. 1972. ”The Proto-Germanic Nonsyllabics (consonants).” Toward a Grammar of Proto-Ger­ manic (Frans van Coetsem and Herbert L. Kufner, eds.), pp. 142-173« Max Niemeyer Verlag. Tubingen. Noreen, Adolf. 1970. Altnordische Grammatik. The Uni­ versity of Alabama Press. University, Alabama. Penzl, Herbert. 1950. "Orthography and Phonemes in Wulfila*s Gothic.” The Journal of English and Germanic Philology 44:217-231. %% Pokorny, Julius. 1959* Indogermanisches Etymologisches Wörterbuch, I. Band. Francke Verlag. Bern und München. Prokosch, Eduard. 1939* A Comparative Germanic Grammar. Linguistic Society of America. Philadelphia. Reis, Marga. 197^. "Phonologie des Spätgermanischen Vokalismus unter besonderer Berücksichtigung der Neben­ vokale.” Probleme der Historischen Phonologie, pp. 2 3 -6 8 . Franz Steiner Verlag. Wiesbaden. Schmierer, Richard J. 1977. Theoretical Implications of Gothic and Old English Phonology. University of Massa­ chusetts Ph.D. Dissertation. Amherst. Shibatani, Masayoshi. 1973. ”The Role of Surface Phonetic Constraints in Generative Phonology.” Language 49:87-107. Sievers, Eduard. 1878. ”gur accent- und lautlehre der germanischen Sprachen.” Beiträge zur Geschichte der deutschen Sprache und Literatur 5 ;63-l64. Sievers, Eduard. 1951. Altenglische Grammatik (neu­ bearbeitet von Karl Brunner). Max Niemeyer Verlag. Halle (Saale). Sonderegger, Stefan. 1959. "Die Umlautfrage in den germanischen Sprachen.” Kratylos 4:1-13. Steller. Walther. 1928. Abriss der Altfriesischen Grammatik. Max Niemeyer Verlag. Halle (Saale).

160

R eferences

Streitberg, Wilhelm« 1905-6. "Gotica." Indogermanische Forschungen 1 8 t383-407. Streitberg, Wilhelm. 1920. Gotisches Elementarbuch. Carl Winter#s Universitatsbuchhandlung. Heidelberg. Streitberg, Wilhelm (ed.). 1950 (reprint of 1908 edition). Die Gotische Bibel. Carl Winter. Heidelberg. Szemerenyi, Oswald J.L. 1972. "A New Leaf of the Gothic Bible." Language 48:1-11. Thurneysen, R. 1898• ••Spirantenwechsel im Gotischen." Indogermanische Forschungen 8:208-214. Twaddell, W.F. 1948. "The Prehistoric Germanic Short Syllables." Language 24:139-152. Vennemann, Theo. 1971« "The Phonology of Gothic Vowels." Language 47:90-133. Vennemann, Theo. 1972. "Phonetic Detail in Assimi­ lation: Problems in Germanic Phonology." Language 48:863893. Voyles, Joseph B. 1967* "Simplicity, Ordered Rules, and the First Sound Shift." Language 43*636-661. Voyles, Joseph B. 1968. "Gothic and Germanic." Language 44:720-746. Voyles, Joseph B. 1976(a). The^Phonology of Old High German. Beiheft 18 of Zeitschrift fur Dialektologie und Linguistik. Franz Steiner Verlag. Wiesbaden. Voyles, Joseph B. 1976(b). "Old High German Umlaut." Zeitschrift fur vergleichende Sprachforschung 90:271-290. Voyles, Joseph B. 1980. "Reduplicating Verbs in NorthWest Germanic." Lingua 52*89-123. Wienold, Gotz. 1967. "On Umlaut in Gothic." Orbis l6: 185-196. Wright, Joseph. 1954. Grammar of the Gothic Language. Clarendon Press. Oxford. Wurzel, W.U. 1975. "Der Gotische Vokalismus." Acta Linguistica Academiae Scientiarum Hungaricae 25:263-338.