Old English Breaking and its Germanic Analogues 9783111356501, 9783484302532

Citation preview

Linguistische Arbeiten

253

Herausgegeben von Hans Altmann, Peter Blumenthal, Herbert E. Brekle, Hans Jürgen Heringer, Heinz Vater und Richard Wiese

Robert B. Howell

Old English Breaking and its Germanic Analogues

Max Niemeyer Verlag Tübingen 1991

CIP-Titelaufnahme der Deutschen Bibliothek Howell, Robert B.: Old English breaking and its Germanic analogues / Robert B. Howell. Tübingen : Niemeyer, 1991 (Linguistische Arbeiten ; 253) NE:GT ISBN 3-484-30253-4

ISSN 0344-6727

© Max Niemeyer Verlag GmbH & Co. KG, Tübingen 1991 Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Printed in Germany. Druck: Weihert-Druck GmbH, Darmstadt

Prefece

In the course of my work on the general topic of consonantally conditioned vocalic mutations in Germanic I have benefitted greatly from the criticisms, remarks and logistical support of a number of friends and colleagues. Without the generous help of these individuals, the results of this study would have been much poorer and far longer in coming. My first work on breaking was inspired and encouraged by my doctoral advisor, Frans van Coetsem. In the course of my early research on this topic I was fortunate to benefit from his unique blend of encouragement and critical commentary. Without Professor Van Coetsem's assistance I can state with some certainty that the present study would never have materialized. I know I speak for many of his former students when I say that I will always take pride in being able to say "I studied with Van Coetsem". I should also make particular mention of the specific and highly important contributions to my current interpretation of Old English breaking made by Jay Jasanoff. Professor Jasanoff repeatedly forced me to reevaluate and reinforce my argumentation in the early stages of research. As a result of his insistence on consistent interpretations of internal and comparative evidence I have made a concerted effort to support my arguments with as much hard evidence as possible. Professor Jasanoff specifically warned me of the pitfalls inherent in interpretations of breaking assuming some sort of velar or uvular articulation common to all breaking conditioners. He suggested instead that lowering and backing of vowels can be caused by a variety of potential reflexes of PGmc. */r/ and */x/ which would not, at first blush, be characterized as "back". Therefore I have Professor Jasanoff to thank for several important arguments which make this study deviate from traditional interpretations of breaking. Whatever weaknesses persist in my argumentation are, of course, strictly my responsibility. I should also mention several individuals who have contributed to the development of my thoughts on this topic. Linda Waugh, Wayne Harbert and Thomas Young read and commented extensively on the proto-version of this study, the second chapter of my doctoral dissertation (Howell 1983). Numerous substantive reinterpretations in the present study can be traced directly to their commentary. In the past three years I have also been assisted by several graduate students. Brad Holtman and Greg Hanson have both read and criticized sections of this study. Greg Hanson and Lauren Lewis have generously volunteered their services for proofreading large sections of computer-generated page proofs.

11

Research for this study has been supported by substantial grants from the German Academic Exchange Service (DAAD) and the Graduate School at the University of Wisconsin, Madison. I would like to thank the directors and staff of the Forschungsinstitut für deutsche Sprache in Marburg, FRG for graciously opening their archives to me in the summer of 1985. Finally I would like to acknowledge a decade of invaluable editorial and moral support from my wife, Barbara Marshment, who understands everything.

Ill

Contents 0. 0.1 0.2 0.3

Introduction Characterizing breaking The Old English digraphs and the 'Digraph Controversy' The place of Old English breaking in early Germanic

1 1 2 4

Chapter 1. LC Clusters: Breaking, Svarabhakti and Other Related Phenomena 1.0 Introduction 1.1 Proto-Germanic */r/ .1.1 Evidence for apical [r] .1.2 Arguments for early non-apical r .1.3 Uvular r .1.4 Velar co-articulation .1.5 "Molar r " .1.6 Retroflex r .1.7 Evidence against back r theories .1.8 Objections to uvular [R]/[B] .1.9 Objections to velar co-articulation .1.10 Objections to retroflex r 1.2 A possible solution: liquid reduction .2.1 Liquid reduction and syllable structure .2.2 Restrictions on breaking before syllable-final r 1.3 Svarabhakti and its relationship to breaking 1.4 Smoothing and vowel epenthesis 1.5 The question of breaking before / 1.5.1 Restrictions on breaking before 1C 1.5.2 Determining the phonetic realization of Old English /!/. 1.5.3 Accounting for breaking before 1C 1.5.4 Accounting for the restrictions on breaking before 1C 1.6 Conclusion

6 10 10 13 14 16 17 19 20 21 36 39 40 47 55 60 66 69 70 71 74 78 81

Chapter 2. Proto-Germanic */x/ and Old English breaking 2.0 2.1 2.2 2.3 2.4 2.5 2.6

Introduction Arguments for [x] as a breaking conditioner Early reduction and loss of */x/ Reconciling the behavior of */x/ and */k/,*/g/ in Old English Vocalic mutations conditioned by velar [x] Reduction of [x] and its effect on adjacent vowels Conclusion

83 84 86 95 99 101 105

Concluding Remarks

107

Reference

110

0.

Introduction

0.1.

Characterizing breaking.

Although the process known as "breaking1 has received extensive treatment in the literature on Old English phonology, the scholarly discourse on the nature of the breaking process is characterized by a general lack of consensus. Investigators have long recognized that the appearance of the graphemic diphthongs , and for *£ , *£, and *i before reflexes of Proto-Germanic */x/, *M and */!/ represents some sort of consonantally conditioned backing mutation. The phonetic value of the vocalic digraphs and the phonetic quality of the conditioning consonants have, however, been the source of much debate, for it is neither immediately clear what phonetic value the vocalic digraphs depict nor is it obvious why */x/, */r/ and */!/ should act together as conditioners exerting a backing influence on preceding vowels. This study will focus on the development of the consonants conditioning breaking in Old English in the hope that a better understanding of the conditioning environment will lead to a more coherent description of the process as a whole. It will differ from earlier investigations of the breaking process in its broad comparative approach. Rather than concentrating solely on the specific and often apparently contradictory evidence provided by our extant Old English sources, we will pay close attention to the development of */x/, *M, and */!/ in Germanic as a whole and use this comparative evidence in the evaluation of competing analyses of the breaking process. This comparative evidence, used in concert with the available internal evidence, can yield insight into the general developmental patterns of the conditioning factors in Germanic and sheds new light on the well-known facts surrounding breaking that have proven so resistant to plausible interpretation. Traditionally, discussion of the breaking factors in Old English has taken a retained velar [x] derived from original Proto-Germanic */x/ as a point of departure for analysis of the breaking environment This assumption seems to arise from the fact that [x] is one possible reflex of one of the breaking conditioners that is unambiguously back in nature. Various proposed back variants of r and / which appear in Germanic or other languages are then supposed to have

developed prior to breaking in Old English. The [x] and the back r and / variants then form a 'natural' class of back consonants which then exerts the breaking effect The / forms conditioning breaking have been variously described as "dark", "guttural", "velar", the r forms as "dark", "guttural", "velar", "molar", "uvular" and "retroflex". Yet no discussion of the breaking process offers proof that the liquid forms proposed actually exert the mutating effect envisioned. As a result, we are forced to agree with Kühn (1970.37) who concludes his discussion of the realization of the breaking and non-breaking variants of /r/ and l\l as follows: "Although articulatory differences of some sort probably existed, in the present state of our knowledge, we can only conjecture what they might have been. All of the guesses are legitimate attempts to account for the phonetic effects of /!/ and /r/, but they are only guesses." In the following pages we will attempt to improve the accuracy of our "guesses" about the Old English breaking process by comparing the Old English facts to developments similar to breaking in both conditioning and effect found elsewhere in Germanic. This comparative evidence will not be limited to the well-known consonantally conditioned breakings found in early Germanic language such as Old Frisian and Old Norse (cf. Rom 1937), but will also make extensive use of modern evidence of vocalic mutations before conditioning environments similar to those found in Old English. Since this modern evidence is based on relatively reliable phonetic descriptions and because diphthongizations conditioned by variants r, / and reduced forms of

are among the most common phonetic processes found in the modem dialects of the

Germanic languages, a plausible description of the nature of breaking in Old English begins to emerge. 0.2.

The Old English digraphs and the "Digraph controversy".

Any discussion of Old English breaking requires that the phonetic value of the digraphs , , and be established. The traditional interpretation of these digraphs as diphthongs with long and short realizations has been challenged in a series of articles claiming that the digraphs in fact represent monophthongs. The double grapheme is seen as a reflection of coloration of an original monophthongal front vowel effected by an adjacent consonant This monophthongal interpretation was first proposed by Daunt (1940) and was subsequently accepted (Mosso 1945) and modified (Stockwell and Barritt 1951). Stockwell and Barritt's proposal that the digraphs depicted allophones of Old English /at/, /e/ and /i/ was sharply rejected by Kühn and Quirk (1953). Hockett (1959) and Antonsen (1967) defend the monophthongal interpretation assuming that the digraphs represent back unrounded monophthongs. The Hockett/Antonsen

interpretation of the digraphs is rejected point by point in a careful evaluation by Ney (1969), and Giffhorn (1974) provides evidence derived primarily from place names indicating that the interpretation of the digraphs as diphthongs proves most consistent with their subsequent development. In recent work on Old English phonology, a diphthongal interpretation of the digraphs seems to predominate (e.g., Pilch 1970, Lass and Anderson 197S, Nielsen 1981, Suzuki 1982). In this study it will be assumed that the digraphs represent phonemic short and long diphthongs with a front first element and a back, probably rounded second element The reasons for rejecting the monophthongal interpretation are neatly presented in Ney (1969.49). It is, however, perhaps necessary to emphasize the fact that the diphthongal interpretation of the digraphs in breaking environments is strongly supported by the development of diphthongs in precisely the same positions elsewhere in Germanic from the earliest attestations to the present day. Ney states that the monophthongal interpretation "does not seem to account for the parallel development of what appears to be diphthongs arising under similar conditions...in the related Germanic languages." As correct as this statement may be, it rather understates the frequency of diphthongizations "arising under similar conditions." Close analysis of the large number of vocalic mutations in Germanic with conditioning environments similar to those found in the Old English breaking provide extensive evidence that diphthongizations of the sort erC > ea(r)C are not only common in Germanic, but are among the most frequent conditioned vocalic mutations in the modern dialect literature. On the other hand, the development of back unrounded monophthongs in the breaking environments is virtually absent from either the available historical evidence or the modern dialect literature. Rejection of the diphthongal interpretation of the Old English digraphs therefore requires that we believe that the Old English breaking represents a nearly unique development in Germanic totally unrelated to the numerous diphthongizations which seem to occur in precisely the same conditioning environment in both North and West Germanic languages throughout the entire historical period. Of course the fact that the Old English digraphs themselves seem prima facie to represent diphthongs does not argue against the diphthongal interpretation. It is, however, interesting that the Old English graphemic representation of the breaking of *-erC- as provides a fair transcription of the common development of erC in modern Germanic dialects to ea(r)C, eo(r)C, ed{r)C given the limitations of the alphabets) available to the early scribes.

0.3 ΊΤϊβ position of Old English breaking in early Germanic. Diphthongizations conditioned by r or / followed by a consonant or by reflexes of PGmc. */x/ are found not only in Old English, but in Old Norse and Old Frisian as well. Although the vowels affected and the actual environments conditioning breaking in the early Germanic languages show significant differences, a number of investigators have proposed that these early consonantally conditioned vocalic mutations result from shared features of North/West Germanic prior to the Anglo-Saxon invasion of England in the fifth century. Flom (1937), Samuels (1952) and Dyvik (1978) make strong arguments for the claim that Old Norse breaking in long syllables before rC and 1C (e.g., ON beork, hearta; OE beorc, heorte) is conditioned by the following consonant cluster, not by a following back vowel. Flom and Samuels also link Old Frisian breaking before ht (riuht, siuht, etc.) to the Old Norse and Old English diphthongizations. On the basis of the clear similarities exhibited by Old English and Old Norse breakings Rom (1937.129f.) concludes that the breakings are "not only similar in nature but directly related in origin" and that breaking was found in the language of the Jutes, Angles and Frisians before the Anglo-Saxon migration to England. Dyvik (1978.35) gives a more restrained assessment stating simply that "the phonetic beginnings of breaking...was a common development in the North Gmc. area and the northernmost part of the West Gmc area." Samuels (1952.32) arrives at the most conservative and most probable judgment by assuming simply that such diphthongizations can occur in different languages at different periods as long as the proper conditions for the mutating effect are present Philippa (1982.127) essentially accepts Samuels' position, but suggests that the spread of breaking from its original source dialect(s) may have been promoted by contact, "vooral op handelsgebied", among the Germanic peoples inhabiting the North Sea area. Yet despite the interesting similarities snared by the breaking processes in Old English, Old Norse and Old Frisian, there is little evidence favoring a common North/West Germanic origin of the breaking phenomena. Nielsen (1984) convincingly outlines problematic chronological anomalies to such a theory and lays bare clear differences in the conditioning and effect of breaking in the individual languages that would make assumption of a common origin of the processes difficult to envision. Still more damaging, however, to the hypothesis of a single early source of breaking is the fact that breaking processes as similar to Old English breaking as those found in Old Norse and Old Frisian are found throughout the Germanic linguistic area from the earliest period to the present Breaking before h is found in Neuter's Alemannic of the 11th century : Ithti, d hta, ruh > liehte, duohte. ruoh 'leicht', 1.3. sg. preterit ofdunken , 'rauh*

(Braune/Eggers 1987.§154 A.9.) and breaking before r and reduced χ is found in 14th century Bavarian sources (Weinhold 1867.§167). Diphthongs have developed in numerous modem dialects of West and North Germanic languages in similar environments as well. Obviously none of these changes can be linked to some early North/West Germanic kerngebied for breaking. On the contrary the extent and frequency of such consonantally conditioned breaking processes in Germanic serve only to support Samuels' assumption mat breaking can and does readily develop in various dialects at various points chronologically. Therefore while careful comparison of the numerous consonantally conditioned diphthongizations in Germanic which reflect phonetic developments clearly parallel to the Old English breaking process can prove invaluable to the evaluation of the phonetic nature of the Old English developments, the very frequency and geographic dispersal of these breaking phenomena make assumption of some common source both unnnecessary and undesirable. The presence of breaking processes in the early Germanic languges only indicates that the structural tendencies leading to consonanataUy conditioned vocalic mutations were present in Germanic at a very early date. Just exactly what these tendencies might be will be discussed in the chapters that follow.

l.

LC Clusters: Breaking, Svarabhakti and other related phenomena

"Wanneer wij ons verdiepen in de kwestie van de vocaalontwikkeling voor r plus consonant, bekmipt ons het gevoel dat we de hogere rijschool van de linguistiek betreden." -G.G. Kloeke (1950.111) 1.0.

Introduction.

Any survey of conditioned sound changes in early Germanic languages will provide dramatic evidence that clusters of a liquid followed by a consonant play an important role in the diachronic development of Germanic phonological systems (see for example Rauch 1967). Yet despite the repeated mutating effects exerted by liquid clusters (hereafter LC clusters), very little attention has been focused on exactly why liquids in this particular environment exert such frequent and often strikingly different effects as the mutation of preceding vowels typified by Old English breaking and the development of an epenthetic vowel between the two members of the LC cluster found in Old High German and Old English vowel epenthesis:

fig. 1.1 OE breaking:

pre-OE

*/

>

*e >

hiorde 'shepherd*

l_rC

weorban "become* (OHG werdan) beam 'child* (Goth, barn )

I_IC

healf "half (OHG halb)

*ee > *ee>

OHG vowel epenthesis -rC-, -1C- >-rVC-, -IvC-

uuerach Werk* (Alemannic dwuft "Not* (Benedictine rule) uolagen 'folgen' (Monsee glosses) halap *halb' (4th Reichenau gloss) (all OHG examples from Reutercrona 1920.97, 116, 139, 137 respectively)

OE vowel epenthesis -rC-, -1C- > -rVC-, -IVC-

berig 'mountain* (Franks Casket) wylif 'she-wolf " jferih 'through* (Leiden Riddle) aerig- 'arrow* " gewarahta 'worked* (Mortain Casket) (OE examples from Campbell 1959.150f. and Brunner 1965.135Q

When we add to the examples above processes such as r- metathesis in Ingvaeonic dialects (see Campbell 1959.84) and the lengthening, raising and lowering of short vowels before r followed by a consonant in Middle Dutch (see Van Bree 1977.193-194; Van Loey 1976 ν.Π.35) it becomes quite clear that LC clusters prove highly active from early Germanic onward and that they can effect a variety of quite divirgent phonological changes. Furthermore, the very number of developments conditioned by LC clusters from early Germanic to the present suggests that all of these processes arising from original contact of a liquid and a consonant represent, in the broadest sense, related reactions to the same phonotactically difficult situation. Attempts to explain individual phenomena conditioned by LC clusters have traditionally centered on the assignment of specific articulations or phonological features to the conditioning liquid. These features are assumed to be consistent with the backing/diphthongizing effect exerted on the preceding vowel. In the neogrammarian handbooks, the r causing Old English breaking is variously described as "retroflex" (Sweet 1957.3; Mosse* 1945.114; Brunner 1965.146), "burred" (Quirk and Wrenn 1955.15), "guttural" (Luick 1914[1965].144) and "reverted" (= retroflex? Wright 1925.17). Later formulations using distinctive features characterize the breaking factor as [+back] (e.g., Lass & Anderson 1975; Philippa 1982) or as [+grave] (Str0jer 1986). These feature-based analyses represent little more than notational variants of suggestions by the neogrammarians. Because of their tendency to view the breaking factor purely on a segmental level, previous discussions of beaking generally suffer from the same shortcomings. First of all, virtually no independent evidence is introduced demonstrating that the posited r or / breaking factor might have existed phonetically in Old English or even that such liquid variants tend to develop in LC clusters in Germanic. Secondly, investigators seem satisfied that once the backness of the liquid is accepted, the operation of the breaking process is self-explanatory. Whether the liquid is velar, retroflex, uvular, [+back], [+grave] or whatever, the backing diphthongization process is supposed to proceed naturally from some sort of a priori law of assimilation. This approach leaves some rather compelling questions unanswered. One wonders, for example, if the various back r forms proposed actually do seem to cause the kind of mutating effects assumed. Furthermore, if some variety of back r does in fact cause breaking, why don't other back consonants such as /k/ and /g/ participate in the breaking process? Finally, the fact that only postvocalic liquids followed by consonants condition breaking remains completely unaddressed. Of course the LC environment can be stated in a formalized rule (cf. Lass and Anderson 1975.104f.), but this sheds little light on the central question of why liquids in this particular

8

environment are phonetically different from other contextual variants of /]/ and /r/. A strictly segmental approach can offer little in the way of a principled explanation of the processes underlying breaking because it provides no way of treating the obvious fact that the breaking variants of /!/ and /r/ are determined by the liquid's role in the syllable. The discussion that follows will focus on the fact that the liquids involved in breaking 1) occur in the syllable rhyme and 2) invariably are followed by a consonant The following consonant may be either tautosyllabic (e.g. *heelp > healp 'helped'; *eerm > earm 'arm*) or not (e.g. *wer.pan > wear.pan 'to throw*). From this follows that the breaking liquids develop as a result of the demands of Old English internal syllable structure and from the phonotactic relationship between syllable offset and the following onset. The phonetic difference between liquids causing breaking and those not involved in the process can therefore be directly linked to the liquid's role in the syllable. The argument below will make the claim that LC clusters are diachronically marked in Germanic and that the various daughter languages have employed from a very early date two basic strategies to mitigate the phonotactic problems inherent in liquid plus consonant sequences. The first of these strategies could be termed liquid reduction , a process which involves the stripping away of some or all of the complex coarticulations associated with liquids, especially with rhotics. Liquid reduction affects liquids in the syllable coda and yields contextual variants which are much more vowel-like than syllable-initial variants such as the trill or flap common in ft/. The segmental boundary between the preceding vowels and these less abrupt reduced liquids can become very unclear, fostering the development of transitional glides which result in a diphthongization. This transitional diphthongizing effect becomes more dramatic as the aru'culatory characteristics of the two adjacent segments (vowel + reduced liquid) divirge to a greater extent We will posit Old English breaking as an example of this strategy of liquid reduction. Interestingly enough, the primary point of articulation of the syllable-initial liquid variant (e.g., apical, velar, uvular for /r/), a primary concern of previous treatments of breaking, becomes essentially irrelevant— for it is precisely this primary articulation that tends to be stripped away in the process of liquid reduction. The second and in effect complementary strategy for dealing with LC clusters involves the insertion of an epenthetic vowel between the liquid and the following consonant, a process of ^syllabification which removes the liquid from the syllble rhyme and places it in the onset of a newly formed syllable. This restructuring dissolves the LC cluster and hence also the environment for liquid reduction:

flg. 1.2

OE vowel epenthesis. berg > berig

'mountain'

σ / \

σ / 0 1 b

\ r /l\ er g

>

j o l b

\ Γ

1 e

r 1 /\ r i g

0

σ = syllable ο = onset r = rhyme

The Old English and Old High German vowel epentheses represent this second strategy. The various r-metatheses in Germanic would also provide related examples of syllable restructuring resulting in the dissolution of troublesome LC or CL clusters. Given the distribution of the liquid breaking conditioners it seems only logical to undertake a broad comparative survey of the behavior of liquids in LC clusters in Germanic dialects in order to determine exactly what kind of / and r forms in which environments tend to exert breaking effects apparently similar to those found in Old English. This comparative survey will in most instances be based on modem dialect studies which provide relatively accurate phonetic descriptions. Since vocalic mutations induced by LC clusters turn out to be among the most common of phonetic developments found in the extant dialect literature, a plausible evaluation of the Old English breaking process will emerge. Although this evidence of the behavior of LC clusters gleaned from modem dialect studies will provide a major component of the discussion that follows, the extensive use of modem data to interpret archaic phonetic changes should not be understood as an attempt to deemphasize internal evidence. It must be conceded, however, that our extant early Germanic records provide precious little insight into the actual phonetic properties of the consonants conditioning vocalic mutations such as breaking. Nevertheless by collecting modem examples of developments which appear to parallel vowel epenthesis or breaking processes it does prove possible to arrive at a set of phonetic and phontactic conditions which seem to yield the mutating effect This typological approach benefits from the reasonable accuracy of the phonetic information provided by the modem studies. The developmental tendencies derived from the modem data can then be employed when evaluating the relative probability of competing interpretations of the Old English phenomena. The modem evidence should therefore be viewed simply as an important supplement to the necessarily limited internal evidence at our disposal and should help to provide a more convincing description of the breaking process. A fundamental assumption throughout

10

will be that similar phonetic environments in structurally similar languages tend to develop diachronically in parallel ways (for similar aplication of modern data see Howell 1988a, 1988b, 1988c).

1.1.

Proto-Germanic *A/

1.1.1.

Evidence for apical [r].

The inconclusive nature of internal evidence in the older Germanic languages is probably best exemplified by the controversy surrounding the correct phonetic description of Proto-Germanic */r/ and its immediate reflexes in the daughter languages. Standard handbooks have traditionally described PGmc. */r/ as an apical trill [r]. This value is derived primarily from comparative evidence provided by other Indo-European languages where /r/ is generally apical and from rhotacisms within Germanic resulting in the merger of PGmc. */z/ with */r/. Aside from rhotacism, the conditioning of early Germanic vowel epenthesis also speaks in favor of an apical value [r] for early A/ phonemes. Although comparative data indicate that PIE "/r/ represented an apical flap or perhaps trill, one prefers to base judgments about the phonetic value of PGmc. */r/ on more immediate evidence if at all possible. The unconditioned North/West Germanic merger of */zJ and */r/ as */r/ and the Gothic rhotacism of *uz-r to ur-r (*uz-reisan > urreisan 'to arise1) would appear to provide just such evidence. The merger of */z/ with */r/ implies that the two phonemes share at least some crucial phonetic characteristic/s. Clearly the phonetic similarity resulting in the merger need not be a common alveolar point of articulation — no phonetic law precludes the possibility that the original voiced sibilant could have fallen together with a uvular or velar r. Neveretheless, comparative evidence provides examples of apparently similar assibilation and rhotacism specifically involving alveolar consonants and apical [r]: fig. 1.3

a) rhotacism Latin *honozis > honoris gen.sg. 'honor, esteem' ([z] > [r]) Spanish dos reales > [doareales] 'two reales'(Navarro 1966a. 108) Odenwald (Hessian) MHG -VdV-, -VtV- > -VrV-: mird 'mieten1; fleramaus Tledermaus' (Bauer 1957.63) ( = [r]) b) assibilation French chaire > chaise 'chair'; bericles > besides 'spectacles' ([z] > [r])

11

Spanish (Mexico) - Cr- > -Cz- in otra Other'; tropa 'crowd, troop, herd' (Delattre 1966.194-5).1 Any claim that the early Germanic rhotacisms were not similar to the processes illustrated above must be justified. The fact that a svarabhakti vowel often develops between liquids and following labial or velar consonants in Uper German dialects of Old High German and in Old English leads one to believe that the liquids and the consonants separated by the epenthetic vowel must have been nonhomorganic. The insertion of an epenthetic vowel alleviates the phonotactic difficulties encountered in the pronunciation of sequences of a liquid or nasal and a non-homorganic obstruent Since the vowel epenthesis occurs only very rarely when a dental consonant follows the liquid it seems likely that r and / were alveolar (see fig. 1.1 for examples ). This hypothesis is supported by the occasional appearance of similar epenthetic vowels between nasals and clearly non-homorganic obstruents: fig. 1.4 pesamo < besmo 'Besen' (Alemannic, 8th cent., Vocabularius St. Galli) suanichil < suankil 'Schwengel' (Alemannic, 12th cent) (Reutercrona 1920,144,149) The phonetics of the processes of epenthesis and their relationship to breaking phenomena will be discussed in detail below. Within Old English proper the behavior of positional variants of /r/ in processes of metathesis seems to provide further evidence that apical r lived on into the Old English period. Metathesis in Old English (and in the Ingvxonic dialects in general) typically occurs when r precedes a short vowel followed by an alveolar consonant, particularly s and n . Furthermore, a consonant originally preceding a metathesized r is overwhelmingly a nonalveolar obstruent' heern 'wave'; sern 'house'; weerna 'wren'; sernnan 'run, ride (causative)'; beernan 'to (cause to) burn1; beerstlian 'to crackle'; burna 'stream'; cyrps 'cuny'; first 'period'; dxrstan 'dregs'; cerse 'cress'; forsc 'frog';/or« 'frost'; hors "horse"; birdas 'young birds'; dirda 'third'; geers

'grass' (Campbell 1959.184; Brunner 1965.146f.). The examples of

Northumbrian biornan and West Saxon birnan (vs. OHG brinnan) 'to burn' clearly indicate 1

Delattre describes this non-phonemic assibilation of M as resulting in a fricative [j] phonetically similar to [z] "but more intense". Despite the fact that this particular change is non-phonemic it provides a model for the confusion of [z] and original apical [r] which clearly finds parallels in Germanic. The non-etymological rune in the Istaby stone (ca. 600-650 A.D. in for /aftr/ 'after' (Antonsen 1975.84) provides a clear example of a parallel neutralization of the */z/ - */r/ opposition after apical obstruents.

12

that metathesis could both precede and follow breaking chronologically and was therefore quite early in some instances. Regardless of whether one views r metathesis as a gradual change of the soitfrist position (fnst

>fdr(st

>ftrest > first

or as a sudden jump from prevocalic to postvocalic

> first), it is clear that the r is attracted to the following alveolar consonant.

Since the r typically moves away from an orignal preceding non-alveolar consonant, a logical conclusion would be that the r is apical and that it tends to move away from non-homorganic segments and toward alveolar segments (cf. Alexander 1985). Even in those examples of metathesis which show original r in word-initial position, the consonant toward which r is attracted is in every instance alveolar. In addition to the movement of r toward an alveolar consonant, Northumbrian and late West Saxon show a limited leftward metathesis of r in original clusters of -rht-: worhte > wrohte 'worked'; berht > breht "bright'; forht >froht 'afraid, timid'. Sporadic leftward metathesis also occurs in clusters of r followed by a labial: prop beside porp 'farm' (Campbell 1959.184f.). While it is conceivable that the r involved in this type of metathesis could have been something other than apical, it seems far more likely that the leftward movement, like vowel epenthesis and rightward metathesis, represents a reaction to an unstable phonotactic situation resulting from the contact of a flaped or trilled apical r with non-homorganic consonants. Middle Dutch, in which r was very probably apical in most dialects, shows metatheses quite similar to those found in Old English. The apical articulation of r is implied by the occasional insertion of f otd between r and a following s as in arc > cards 'hers'; siaerts (< 'sjaars) Of the year' (Van Loey.l976.f 103). In addition, an epenthetic vowel tends to develop between r and non-alveolar consonants, just as it does in many varieties of modem Dutch. The metatheses in Middle Dutch occur essentially in the same environments as those in Old English cited above: vers (NHG frisch) 'fresh'; kerst- 'Christ-'; barsten (OLF breston) 'to burst'; -born (< -bron) 'spring'; vrucht (NHG Furcht) 'fear'; nootdruft (Goth, fxturfts) 'need'; wrackte (< warchte) 'worked'; verde (< vrede) 'peace1 (Van Loey 1976b.90f.). Lasch (1914[1974].133f.) notes exactly equvalent types of metathesis in Middle Low German, where developments such as -rd- >-r-(waren for warden, 67) point to an apical r. The parallel behavior of OE, MDu and MLG r in such processes infers a phonetic similarity shared by the phoneme ft/. Given the conditions which tend to yield r- metathesis and other independent evidence, it seems highly likely that at least one aspect of that shared similarity is apical articulation.

13

l. l .2.

Arguments for early non-apical r .

Despite the prima facie evidence for positing an apical [r] as the probable phonetic realization of PGmc. */r/ cited above, a number of investigators have argued that PGmc. */r/ or the r phonemes in the early Germanic dialects must have had a velar, uvular or retroflex articulation. The primary motivation for assuming some son of back articulation of r is provided by the widespread lowering, diphthongizing, and backing influences exerted by reflexes of */r/ early in the development of the individual Germanic languages. The hypothesis that PGmc. */r/ represented a velar/uvular trill or flap appears to be buttressed by the fact that the same vocalic mutations conditioned by reflexes of */r/ are often effected by reflexes of PGmc. */x/ (< PIE */k/) as well. Hut (1931. §37,2) states, "r und h müssen einen besonderen Lautcharakter gehabt haben, da sie verschiedentlich auf die vorhergehenden Vokale einwirken." The common effect of */r/ and */x/ in a number of Germanic languages leads to the conclusion that the feature unifying these two consonantal conditioners must have been backness, i.e., velar, uvular or retroflex articulation.

Although the various consonantally conditioned vocalic mutations in

early Germanic are well known, a few representative examples will serve to underscore the frequently identical effect of reflexes of */r/ and */x/. Original / and w ([y]) also often appear as conditioning factors:

fig. 1.5 (in all instances h < */x/) Gothic *i, *e > e () /_ h,h,r: laihum (OHG Itwuni) 'we lent1; saihan (OHG sihan) 'to see'; fxiihum (OHG digum) 'we grew'; raihts (OHG rtht 'right, straight'; carpa (OHG grda) 'earth'; wairpan OHG wgrfan) 'to throw'. *u > () /_ h, h, r : tauhum (OHG zugum) 'we led'; sauhts (OHGjuAf) 'iUness'; bauhta (1.3.pretsg. of bugjan) Tjought'. OHG *ai > e /_ r, A, w : er (Goth, air) 'early'; seo (Goth, saiws) 'sea'; &a (Goto. aihts) 'possession'. ON * ai > a /_ r, h (h is subsequently tost): sar (Goith. sair) 'wound'; far (Goth. faihs) 'shimmering'. OE *ee, *e, *i >ea,eo,io l_h,rC : meaht 'might';/eoA 'cattle';«arm 'arm'; eorpe 'earth'; hiorde 'herdsman'. *e, *i > eo.io I w : aseowan past part of seon 'to sift1; hweowol 'wheel1. *ee> ea I_IC : sealt 'salt1; eald Old'; healf "half; wealh 'foreigner1.

Investigators seeking the phonetic feature common to the consonantal conditioners exemplified above uniformly base their interpretation of the quality of r on the assumption that reflexes of

14

*/x/ in all instances represented velar fricatives, hence the hypothesis that the r was strongly velar or uvular. Obviously w ([y]) also fits nicely into any schema positing the velarity of conditioning factors. The discussion below will show that the development of / to a dark velar [}] is easy to envision as well. 1.1.3.

Uvular r.

Although older handbooks and grammars tend to avoid detailed discussion of the phonetic processes underlying the Old English breaking, the idea that some son of back r was responsible for Old English breaking has a relatively long history, finding expression as early as Koch (1870).

Koch proposes that Old English r must have been very much like the

Northumbrian burr ([R] or [κ]). The breaking effect exerted by this uvular r is explained as follows: "Mit den dunklen vocalen l sst sich dies r bequem sprechen, mit den helleren dagegen (und ee wurde wol mehr nach e als nach a hin gesprochen) weniger leicht. Es schiebt sich deshalb vor r ein erleichterndes α ein"(154). Since Koch, a number of investigators have posited a uvular value for /r/, either in ProtoGermanic (Runge 1973,1974; Antonsen 1975; Ramat 1981) or in specific languages such as Old High German (Kranzmeyer 1956.§50b; Penzl 19612) or Old English (Lass & Anderson 1975). The primary factor motivating the assumption that M in early Germanic languages was uvular is the repeated common effect exerted by reflexes of PGmc. */r/, */x/ and */w/, not to mention */!/. Since the influence exerted by these consonants can be described in the broadest terms as backing and/or lowering, any indication of either backing or lowering of vowels preceding r is taken as evidence that the r conditioning factor must have been uvular. The other conditioning factors, reflexes of */x/ and */w/, are seen as unquestionably velar and this velarity is interpreted as the phonetic factor responsible for the vocalic mutations. Investigators who present arguments for uvular /r/ already in Proto-Germanic generally interpret the Germanic vocalic mutations conditioned by consonants as parallel processes, even when the output of these mutations is, at least superficially, quite different Ramat (1981.28) sees Old Norse and Old English breaking on the one hand and Gothic breaking on the other as essentially the same type of phonetic process despite the fact that the former processes seem to

2

Penzl (1961) does not posit uvular [R] in all cases in Old High German. Instead he takes influences exerted both by /r/ and /h/, /w/ such as monophthongization of *ai to e and the retardation of primary umlaut as indications that at least the specific positional variants of /r/ involved were uvular.

15

involve epenthesis of a back vowel to a front vowel while the Gothic breaking apparently reflects a lowering.

Ramat simply states, "es gibt ... genügend Anhaltspunkte, um

anzunehmen, daß auch ai und au des Got... durch Brechung entstanden sind: die phonetischen Voraussetzungen des Überganges i, u > ai, au

sind im Prinzip genau dieselben wie

diejenigen, die für Brechung im Angelsächs. und im Nord, bestimmend sind.", Ramat goes on to say that the common factor resulting in the breaking effect is the velar character, "uvulares [R], velares [t] und [ ]" is anticipated by "einen velaren... oder zumindest weiter hinten realisierten Gleitvokal " (p.29).3 Extensive arguments supporting early uvular [R] in Germanic found in Runge (1973,1974) and Penzl (1961) are also based almost entirely on the common effect of M and reflexes of */x/ and */w/. These studies make the repeated claim that vocalic mutations conditioned by these consonants demonstrate that the r form involved must have been uvular. Typical of the argumentation is Penzl's (1961.495f.) discussion of the OHG monophthongization of *ai to e: "In the development of Germanic *ai to e,

again appears with /h/ and /w/ as a conditioning

environment. The removal of the second, high palatal component of the diphthong must have been favored by the velar character of the consonant This points to uvular [R]." The phonetic link between the lowering/ backing mutation of vowels and the velarity of the conditioning consonants seems to be assumed tacitly. One final bit of evidence favoring uvular /r/ in early Germanic is introduced by Antonsen (1975.17). In his discussion of vowel epenthesis in the older runic inscriptions Antonsen states that the quality of epenthetic vowels developing between r and originally adjacent obstruents sheds light on the articulation of the variants of AA The development of epenthetic between original -dr- in hed erA, hAid erA (= /haedra/ 'hither1) indicates that this r form is apical whereas the development of epenthetic before word-initial r in gin Arunaz shows that initial r

"was still uvular". The exact phonetic motivation for this conclusion remains

unelaborated. Lass and Anderson (1975.84-90) focus specifically on the phonetic value of OE /r/, arriving with some misgivings (cf. 89-90n.) at the conclusion that the r must have been uvular. The

3 Compare also Ramat 1966,1967. The idea that the Germanic consonantally conditioned vocalic mutations conditioned by reflexes of */r/ and */x/ share common phonetic properties both in terms of the conditioning and the resulting mutation has been around for quite some time. Weinhold (1867.92) sees a relationship between 13th/14th century diphthongization of toie before A andr in Bavarian as parallel to Gothic breaking: "Neben diesem alten echten ie zeigt der Dialekt ein andres, das als konsonantische Brechung von i zu nemen ist und sich dem gotischen ai aus i vor r und A vergleicht"

16

primary basis for assuming this uvular realization of M is neatly summed up in Lass' own critique of Lass & Anderson: "Basically we argued that if (a) breaking before /r V and breaking before /x/ are 'the same process'; and if (b) N is clearly a velar fricative; then there must be some specific phonetic feature common to A V and a velar, within a feature-framework such as the generally reasonable Chomsky and Halle one which we used. Given this framework, /x/ is [+obs, -ant, -cor, +high, +back, +cont, -voice], the requisite specification must lie in this set" (Lass 1977.11). Lass and Anderson then find a 'typical' back r attested in Germanic, namely uvular [R] or [ ]. Breaking is seen as a rule of [u] epenthesis with a subsequent 'Diphthong height harmony* rule to account for the lower second element found in actual texts.

1.1.4.

Velar co-articulation.

Luick (1914-21 [1964]. 144) assumes that the initial development of the vowels affected by breaking must have involved the epenthesis of a [u] to the original front vowels: i, e.ce > iu, eu, seu . This process is seen as the result of the velarity of the conditioning factors: "Die Brechung setzt voraus, daß die bewirkenden Konsonanten ausgesprochen gutturale Färbung hatten, d.h. mit einer gewissen Hebung der Hinterzunge und wohl auch Lippenrundung gesprochen wurden." This guttural articulation is posited for r "nur in gewissen Stellungen" (i.e., in breaking environments). Luick's view is accepted by Van Haeringen (1922a.253f.) who adds that "de r (en / ) met sterke bij-articulatie" must have existed in Proto-Germanic given that fact the PIE sonant liquids f and / yield Germanic ur and ul. Van Haeringen also makes the important point that in preconsonantal position, the r and the vowel preceding it are by nature tautosyllabic. Although this tautosyllabic r is perceived as a factor in the breaking process, the failure of breaking before word-final r remains a problem which Van Haeringen ascribes to the fewer trills of word-final r. This reduction of word-final r is seen as reducing the degree of tongue raising toward the velum, and hence the failure of the breaking effect4

Also assuming a velar r for Old English is Manganella 1958. Rom (1937.126) characterizes the breaking process as follows: "Consonantal breaking of a front vowel into a diphthong with a velar second element is of course caused by the velar articulation of the consonants that cause the vowel change. These consonants may have been velar in quality by their own nature, or else they may themselves have become velarized by their environment-" Exactly what a "velar second component" of a diphthong might be is unclear. Nevertheless, the observation that breaking conditioners might only exhibit the

17

1.1.5.

"Molar r".

Hogg (1971), Philippa (1982) and Lass (1977,1983) argue that the r form responsible for Old English breaking could have been a velarized r or a velar approximant [j]. This velarized/velar approximant r could be easily grouped with the assumed values of the other breaking conditioners, [x] and [1]. Hogg provides very little justification for assuming a contextually determined (i.e., pre-consonantal) velarized r in Old English other than the fact that it allows a more economical formulation of his breaking rule. The only real motivation Hogg can see for a velarized r is provided by the symmetry that would result between /r/ and /!/, with /!/ and HI being velarized in similar environments. Hogg's ambivalence about the r- velarization is clear in a footnote in which he expresses willingness to accept the uvular [κ] value proposed in Lass and Anderson "since [κ] is [+back]" (Hogg 1971.64n.). Apparently Hogg is above all interested in arriving at an r- form that can be characterized with the feature needed for the breaking rule, not in the precise phonetic specifics. Philippa (1982.120f.) accepts the general concept of a velar articulation for r put forward by Van Haeringen and Hogg, but makes the additional argument that parallel nature of the dihpthongization caused by back mutation (or 'velar umlaut1) provides an additional argument for the velar articulation. The fact that back mutation is conditioned by a following u or a ( ae /_ rC : ksenwe 'Kerbe'; äteenwe 'sterben'; leenne 'lernen'; ksenl TCerl'; wsenff 'werfen'; eenbs "Erbse1; meene Old nag'( = MHG merhe 'märe1); wsent "Wirt, wird'; sen 'irre';Afen3 Hirsch'; wsenwj Wirbel'; heent 'Hirt' (Heilig 1898.91f.) ( = uvular r) Pfälzisch colony on the lower Rhine , e > , f / r : h?rt "Hirt1; h$r3 'Hirsch'; (HG island in LG area) khfcex lürche'; kh$ti TÜrsche'; b$r&x Berg'; ^t "Erde1; php* Pferd'. u, o > 9 /_ r : d$rn "Dom1; d$rf 'Dorf; Aprn Horn'; v?rt "Wort; ph?ft "Pforte1; d?r5t Durst'; khyts Tcurz'; v$rm 'Wurm'; st^rm 'Sturm'; vqrtsdl 'Wurzel'; d$tx 'durch' (Böhmer 1909.60f.) ( = uvular r ).

23

Lowering and/or Dipthongization (syllable-initial /r/ = [R]/[B]) e >$&, a, ä l_rC : w$9rd& "werden1; wfart Wert'; f$drä Terse'; hart 'Herz'jAart 'Sterz'; daräe 'dreschen1; päräe 'pressen'; bärSe "bersten1. i > & ~ id /_ rC : kl&rä 'Kirsche'; wisrt *Wirt'; hi&.rt 'Hirt'. (Frings 1913.12f.) Darmstadt (Rhine Franconian) ir > f : b$og Berg'; pp 'ihr'; A^pn 'Hirn'; kh$oä Türche, Kirsche'; mfp 'mir';Sfp£> 'Scherbe'. or > 9p : d$cf "Dorf; dqon TXxri;fpdQCW9 'verdorben'; fyod 'fort'. ur > 90 : thqom Turm'; w