Generative phonology and analogical change: the case of the Hebrew suffix 'you(r).'

• Author: Malone, Joseph L.

1.1 Tiberian Hebrew evidences a phonosyntactic complex whereby the lexical accent of a word in an intonationally prominent syntactic position may correlate with certain phonological characteristics of that word not present under other (unmarked) structural circumstances. The traditional name for this phonosyntactic prominence is pausal stress, while the unmarked counterpart is called contextual stress; equivalently, a form may be said to occur in pause or in context. In what follows, a pausally stressed vowel will be symbolized by and a contextually stressed vowel by.(1)

The foremost correlates of pause are either or both tonic lengthening and penultimate stress,(2) though, as will be seen, neither mark by itself, nor their combination, necessarily signals a form as pausal. Thus tonic lengthening alone marks the difference between pausal 'he observed' (Ez 18:19) and contextual (loc. cit.);(3) penultimate stress distinguishes pausal 'I' (Gn 24:24) from contextual (Gn 24:27); both marks distinguish 'you' (Gn 3:11) from (Gn 3:14) or, with vowel reduction, 'take ye!' (Gn 43:14) from (Gn 43:12); while neither mark distinguishes pausal 'my heart' (Ps 9:2) from contextual (Ps 13:6).

Taking this latter pair of forms as point of departure, the focal problem for this paper may now be broached, referring to the representative paradigm (1), with the question: why, of all the genitive pronouns suffixed to the noun, is it only the 2nd person masculine singular suffix (lb) that evidences contextual-pausal alternation, |is similar to~?(4)

TABULAR DATA OMITTED

The pure and simple datum of asymmetry between the 2nd masculine singular and the other suffixes might not much exercise the Semitist, but the fact has long been recognized that the disposition of the 2nd masculine singular ill accords with the predictions given by the comparative method. In particular, the fact that the context variant evidences stress postpositive to the pronominal consonant |x~ (|is less than~ *|k~) is at odds with the situation in all other Semitic languages, and even some Hebrew dialects (cf. Steiner 1979); see the sample in (2).(5)

(2) a. (Classical Arabic) b. (Geez) c. (Akkadian) d. (Biblical Aramaic) e. (Mishnaic Hebrew)

Given the joint testimony of the comparative spectrum sampled in (2) and the actually occurring Tiberian Hebrew shapes, the expected development of the 2nd ma. sg. suffix would have been to one shape for both pause and context--a shape moreover certainly like the actual pausal form with respect to its stress and syllable structure, and plausibly likewise with respect to the vowel color of the bridge (cf. note 5).

So it is only the contextual that is judged to be clearly deviant, and in explanation of that deviance various prima facie plausible analogical accounts have traditionally been offered. A composite of such accounts, to be called Analogical Stress Move (ASM), is given in (3):(6)

(3) Analogical Stress Move (ASM)

1. The stress on a form ending in the 2nd ma. sg. genitive or accusative suffix shifts to the ultima under the analogical pressure of both

2. the corresponding 2nd ma. pl. form, and

3. the majority of other singular suffixes:.

   And yet, despite the apparent cogency of this line of argument, ASM as baldly

   formulated in (3) is weakened by the concomitance of at least two ad hoc factors:

   1.2. If the analogical catalysts of (3b-c) were purely and thoroughly consequential, we should expect the operant part of the Analogical Stress Move to be statable more or less simply as (3a), repeated here:

   (3

1. The stress on a form ending in the 2nd ma. sg. genitive or accusative suffix shifts to the ultima.

   But it is immediately apparent that (3a) fails to go through in the case of pausal forms, as we have seen; thus in (1b) does not shift to (7) despite the fact that the pausal forms of both catalysts themselves evince ultimate stress: (cf. (3b)) and (cf. 3c). Thus the simplicity of the analogical change (3) must be compromised by an exceptional rider:

   (4) Rider I. ASM (3) does not affect pausal forms.

   Moreover, contextual forms with stems ending in long vowels resulting from earlier coalescence of diphthongs show mixed reaction to the analogical shift, some shifting (e.g., 5a), others failing to (e.g., 5b), and yet others vacillating (e.g., 5c):

   (5) a. 'chattel' (Gn 46:32) |is less than~ *miqnay; (Gn 30:29) |is similar to~ (Gn 4:20)

2. 'father' (Gn 44:19), construct state (Gn 4:20) |is less than~; (Gn 27:6) = (Gn 12:1)

3. 'field' (Gn 25:27) |is less than~ *saday; (Dt 11: 15) |is similar to~ (Dt 24:19) =

   Thus a second rider:

   (6) Rider II ASM (3) affects some but not all nouns with a long final stem-vowel.(8)

   The thrust of this section of the paper may be summarized. If the intuitively likely analogical hypothesis of (3) is to be adopted, it will apparently be necessary to coopt two riders to that hypothesis, (4) and (6). The trouble with this sort of result is simply that each such rider adds an ad hoc element to the basic hypothesis, so that what started out to be a simple and elegant idea runs the risk of ending up as a passel of rationalizations. These kinds of results, frequent in the annals of historical linguistics, finally helped to cast the notion of analogical change itself into undue disrepute.

   In the next section, ASM (3) will be looked at in a rather different way, which might contribute toward the rehabilitation of analogical change to its rightful prominence in historical-linguistic theory and practice.

   2.1. Before introducing a novel way of dealing with ASM in 2.2, we will consider a fragment of a synchronic generative phonological analysis of some Tiberian Hebrew verbs, in (7).

   With but two qualifications to be noted below, the synchronic derivations in (7) are arguably isomorphic to the diachronic sequence of linguistic changes which gave rise to them. That is, the phonological forms (given in / / on the top line) are in certain respects similar to pre-Tiberian phonetic forms, and the applicational order of rules (abbreviated in the left margin) arguably reflects the chronological succession of historical changes which over the centuries transformed the pre-Tiberian forms to the Tiberian phonetic shapes given (in | ~) on the bottom line.

   TABULAR DATA OMITTED

   While scholars have always recognized that brute phonetic data cannot be taken, per se, as identical to the phonological system of which those data are manifestations,(10) expression of the phonologic-phonetic relation in terms of layered derivations like (7) has become commonplace among linguists only within the past thirty years or so. But despite the popularity of such derivations among linguists, this sort of phonology remains to a significant extent arcane to philologists. A particular problem stems from the "distance" between phonological and phonetic representations: the more layers of rules separating the former from the latter, the more difficult it is to get a feel for how the derivation in question is supposed to relate to the language of which it purports, in a sense, to be a part.

   A good deal of such quandry may be dispelled as soon as it is recognized why synchronic derivations tend to be isomorphic with the sequence of diachronic events which gave rise to them. When a sound change is conditioned by its phonetic environment, it will characteristically alter the pronunciation of a given morpheme in some contexts but not others. For example, if in a given language a sound change arises whereby unaccented vowels are affected in some way, and if in the verb conjugation of that language a stem is accented preceding certain subject suffixes while various other subject suffixes themselves receive the accent, a given stem may be affected by the sound change in the former case but not in the latter.(11)

   Now this diachronic nexus immediately becomes synchronic as well, because though the sound change has altered the pronunciation of a morpheme in some but not all contexts, it will not normally have affected the identity of the morpheme, which remains constant across contexts. Hence the pattern whereby the same morpheme is pronounced one way in some contexts but another way in others is necessarily a synchronic fact of the language, quite independently of how that pattern may have arisen in diachrony.

   But what if subsequently another sound change arises? Should it be a change of the run-of-the-mill type, it will affect the very class of forms which themselves had been affected by the earlier sound change; and so again for subsequent changes, making for a layered appearance both in models of diachronic-change sequences and in synchronic derivations isomorphic to them.

   However, synchrony is not diachrony and, accordingly, synchronic analysis of phonological patterns need not always be isomorphic with historical events. This brings us to the two qualifications on the derivations of (7) alluded...