[previous section] [next section] [table of contents] [back home]


4. THE STUDY OF SOUND CHANGE IN HEBREW


hmynp bblb dvi lk
hymvh ydvhy wpn
hmydq crzm ytaplv
hypvx Nvyxl Nyi

kçl çjd bale@jvçv pnimç@ kol od bale@vav pnima@
nE@fES jehu@di hçjmiç@ nE@fES jehu@di homia@
ulfaa@sej m"@zrçx kçdimç@ ulfaa@te m"@zrax kadima@
a@jin l´c"@çjn cçjfiç@ a@jin lec"@on cofia@

- Left: first stanza of Hatikvah, Hebrew poem by N.H. Imber, as sung by Menke Katz (b. Svintsyan, Lithuania 1906) in Spring Glen, New York, 8 October 1990, as remembered from New York in the 1920s (Katz 1993: 83). Right: first stanza of Hatikvah, Israeli national anthem.




4.1. Ashkenazic, Sephardic, and Israeli
4.2. Grapho-phonology
4.3. Non-normativity and psychological reality


4.1. ASHKENAZIC, SEPHARDIC, AND ISRAELI

Normal linguistic transmission is primarily an acoustic affair. The first interaction any child has with language is in the form of acoustic signs, and under normal circumstances these signs will be part of the language that the child acquires first. This, of course, was not the case for Hebrew, until the eldest son of Eliezer Ben-Yehuda, the 'first Hebrew child,' became the first to know Hebrew in such a way in modern times. Thus the importance of the point made by Weinreich, Labov & Herzog (1968) regarding the structured heterogeneity of a unilingual linguistic competence can hardly be overstated. Linguistic material originating in a Hebrew language of the past has been repeatedly transmitted from speaker to speaker, incorporated into the expression of their linguistic competence. As such, its limited corpus has been subject to some kind of phonological change through time. Weinberg (1966), in fact, presents an exhaustive description of the ways in which Modern Hebrew realizations relate specifically to the phonological rules of the classical language. His observations point to the fact that Biblical phonology per se can hardly be said to exist in Israeli Hebrew.

Yet 'phonology,' in its various conceptions, was a principle concern both of the early revival movement and of those overseeing the subsequent development of the language. The intention to eliminate the salient features of Ashkenazic Hebrew was not merely an attempt to dissociate the new Hebrew culture from Yiddish and Eastern European Jewish life. Aside from the romantic attraction to its 'Oriental' character, the Sephardi accent favored by Ben-Yehuda and most other revivalists had 'scientific' justification. They claimed that its features more accurately preserved those of the Biblical language (Fellman 1973: 84-85). However, even early in the speech revival, some realized that the nature of this concern, expressed in terms of the reading pronunciations of Classical Hebrew texts, was misdirected:

With respect to our own pronunciation, in relation to 'our Ashkenazic society,' it appears from what the supporters of the Sephardic inflection say that this principle is not a principle, and that the influence of imitation [of Germans and Poles] alone governs us. They pay no attention to the fact that in this very manner we could just as well not apply this very sentence to their own pronunciation...it is by no means the original pronunciation which the ancient Hebrews possessed. (Saul Tchernikovski 1912, cited in Saulson 1979)

Sound change in Hebrew is not merely a matter of change restricted to the Whole Hebrew pronunciation of various communities. Within the generative framework, Halle (1962) described phonological change as the loss and addition of rules to the grammar of a language which describes a speaker's linguistic knowledge. In addition, as the sub-field of generative grammar which seeks to describe the transformations of underlying representations to their phonetic realizations, generative phonology usually treats forms in terms of sub-segmental phonological features. Phonemes are 'cover symbols' for sets of features, and sound change operates at different phonological levels, affecting features rather than phonemes. This view allows for an integration of what have historically been competing theories as to the basic mechanism of sound change, i.e. 'phonemes change,' the neogrammarian principle that sound change is a gradual transformation in a continuous phonetic space, vs. 'words change,' the position of lexical diffusionists who argue for sound change as an abrupt substitution motivated by analogy (Labov 1994: 542).

Traditionally, however, the study of Hebrew sound change has been described not in terms of changes in the rules of a grammar or the features of phonemic segments, but in terms of the different pronunciations assigned by different traditions to the orthographic symbols of written Hebrew. Consequently, it does not describe the true nature of Jewish language speakers' phonological integration of Hebraic material into their linguistic competence, either in Merged or Whole forms. Furthermore, as Faber (1987: 20) states with respect to Ashkenazic Whole Hebrew, 'whether one treats the liturgical pronunciation as something that evolved in tandem with the religious and cultural traditions...or independently of them depends in large measure on one's (possibly romanticized) preconceptions about the centrality of religious observance in the Jewish past.' In other words, there is a tendency to understate the importance of linguistic competence in favor of overstating the influence of religious tradition on a speaker's linguistic knowledge. The distinct functional roles of orthography and phonology in the knowledge of Hebrew were thus merged into one. This blurring of functional distinctions between orthographic and phonological representations is especially interesting when we consider the role it played in the development of the native sound system.

In 1913, the Language Committee published the orthoepic grapheme-phoneme correspondences which they declared tydrpsh hrbhh, 'the Sephardic pronunciation' (see Appendix 1). These correspondences were to serve as the ideal pronunciation of Modern Hebrew, one which rejected the abuses of Ashkenazic Hebrew and endorsed the 'Oriental' pronunciation, which preserved, at the very least, more features of the original Hebrew phonological system. In doing so, the Committee intended to reinstate the large variety of phonetic distinctions indicated by the Tiberian tradition of vocalization in Hebrew texts, especially those lost or altered in Ashkenazic pronunciation. In addition to prescribing stress assignment according to traditional rules, they called for the pharyngeal articulation of c and i, which in Ashkenazic Whole Hebrew had merged with the [x] and [/] traditionally associated with k and a respectively. The realization of the latter in fact alternated freely with Ø in Ashkenazic Hebrew, so that the glottal stop was usually realized by i. The Committee also called for the 'emphatic' articulation of e, so as to differentiate it from dageshed t< [t]. Ashkenazic Hebrew speakers did differentiate t [s] from its dageshed form, but the Committee opted for the Judeo-Spanish [T] realization. They prescribed a uvular articulation for q, which Ashkenazic Hebrew speakers did not differentiate from velar k<. They sought to reinstate the [w] articulation of v, which in Ashkenazic Hebrew had merged with undageshed b as [v]. In the case of the articulation r, it was less a matter of reinstating a lost distinction than the outright elimination of a characteristic feature of Eastern European Yiddish speech, the uvular trill, in favor of the alveolar trill to demonstrate Hebrew's unity with the regional Sprachbund (Blanc 1968).

Yet by 1940, Rabin had observed that, except in the most formal registers, the glottal stop (the reflex of historical a and i) was disappearing from all styles of Hebrew speech. He noted that despite the condemnation of this pronunciation by the authorities of orthoepy, 'elle s'est répandue de plus en plus, et les jeunes des groupes sephardi et yéménite l'adoptent de plus en plus' (1940: 77). This trend has indeed persisted, so that the articulation of [/] alternates freely with Ø in most environments and in most native Israeli Hebrew dialects (Bolozky 1978). Davis (1984) found that the pharyngeal segments, very often present in the phonemic inventories of non-Ashkenazic Jews, and prescribed as characteristically Semitic sounds part of ancient Hebrew pronunciation, are in fact stigmatized in Israeli Hebrew speech. Furthermore, based on studies of linguistic change in 'apparent time,' Davis believes that a sound change in progress is eliminating the pharyngeals from all varieties of Israeli Hebrew. In addition, Yaeger-Dror (1993) claims that ycrzm [mizra@xi] 'Eastern,' speakers, i.e. those of North-African and Middle-Eastern background, whose phonemic inventories and Whole Hebrew pronunciation include the prescriptive alveolar /r/, assimilate in their speech to the koine described by Blanc (1968), which has developed with a distinct preference for the non-normative uvular trill or fricative (see section 1.2, footnote 3). And Katriel (1986) contends that children's use of non-normative penultimate stress, which typically gives Hebrew words what she calls a 'Yiddishized' texture, has a distinct function as a pragmatic particle, which has permeated general speech patterns. The re-assignment of ultimate stress to the very same lexical items which in Yiddish Merged Hebrew received penultimate stress was as much a part of revival practice. Yet Katriel (1986: 280) states firmly that 'the ideological connotations originally associated with the Hebrew stress pattern are no longer relevant to the understanding of its current uses.' The same can be said of the other phonological features of Israeli Hebrew. Although they are the concern of normativists to promote, they have not developed as distinctive Israeli Hebrew features.

Other prescriptive efforts, such as that of articulating a well-documented rule of Biblical Hebrew phonology, known in traditional grammar as 'schwa mobile,' i.e. the insertion of [´] in initial consonant clusters, are also almost completely ignored in casual speech, while conspicuously present in careful speech and extremely formal registers (Bolozky 1978). Many of the prescribed articulations, such as [q] and [T], did not persist in any variety of Israeli Hebrew once a native speech community developed. The presence of the pharyngeal phones [÷] and [_], as well as an alveolar /r/ have been maintained to a degree, and for Blanc (1964) they are a defining feature not only of the speech of radio announcers and of more formal registers, but of the Israeli Hebrew dialect he calls 'Arabicized.' However, Blanc states that he knows of 'no case of genuine acquisition of Arabicized Israeli by a speaker of non-Oriental antecedents' (1964: 134). The 'General' Israeli Hebrew phonemic system has not maintained any segmental distinctions that were not part of Ashkenazic Hebrew speakers' phonological inventory, nor has it developed phonotactic patterns foreign to Ashkenazic speech. Hebrew linguists (e.g. Blanc 1965, 1968; Rosén 1958, 1977; Tene 1969, 1996) have stated repeatedly that there is no doubt that the phonological development of the 'General Israeli Hebrew' of the educated native Israeli speaker was conditioned overwhelmingly by Ashkenazic standards of speech. This conditioning is reflected in both the development of the native sound system, and the subsequent changes it has undergone.

It has been widely assumed in the study of sound change in Hebrew that at its ninth-century encoding, the Tiberian system of vocalization was devised to preserve graphically every one of the phonemes historically present in Hebrew, which by that time was no longer in vernacular use. Yet in the same way that the 'rules' of a grammar are the linguists' model of native competence, so the phonemic system is an abstracted formulation of the way native speakers distinguish the sounds of their language. Ornan (1964: 111), for example, takes the very fact of Hebrew's non-native status to mean that it is more likely that the codifiers of the Tiberian system intended it to represent 'everything their ears heard; it was, then, basically a phonetic system, not a phonemic one...a marking system denoting the real situation of the language, not the theoretical one.' The Hebrew Language Committee assigned a distinct phonemic value to each grapheme so that Hebrew would take on its ancient phonetic form. They viewed the Tiberian system as constituting a phonemic system of significant oppositions, and they believed that the Sephardic pronunciation of Hebrew most accurately preserved this system. The correspondences proposed by the Committee did reflect more closely the phonemic inventory of Arabic and 'Oriental' Hebrew speakers, especially in the consonant system. Yet the 'Sephardic' vowel system adopted by the Committee actually maintained fewer distinctions than the Ashkenazic vowel pronunciations, which differentiated, for example, between the diacritics [kama@ts] and [pata@x] where most Sephardic speakers realized both as [a]. In fact, the Committee's prescriptions did not match any extant Whole Hebrew pronunciation (Tene 1996). Inevitably, a new standard of speech, and a new standard for the analysis of this speech, had to emerge.


4.2. GRAPHO-PHONOLOGY

Many linguists have attempted to describe the sound system of the language of native Israeli speakers, objectively recognizing its divergence from the traditional pronunciations which influenced its development. An article entitled 'The Phonology of Sabra Hebrew' by Patai (1953), for example, offers such an analysis. Yet Patai's description is not only far from the generative understanding of a 'phonology,' it is also typical of even the most non-prescriptive attempts to characterize the sound of the native Hebrew speech community. Patai begins his article by showing that with sixty-four phonetic elements potentially represented in Hebrew script, the contemporary spoken language, in its various traditions of pronunciation, necessarily shows fewer phonemic distinctions than it did at its ninth century encoding in the Tiberian system. In a similar vein, Morag describes the vowel systems of three pronunciation systems of Hebrew, traditional Sephardic, General Israeli, and Oriental Israeli, in specific reference to the respective realizations of the diacritic vowel marks he terms 'historical q__Ames1, h1olam and s1eri' (1959: 251). And Kutscher specifically bemoans the phonemic split of b, k, and p, while accepting that the merger of the phonemes a and i 'has not in any way altered the structure of the language' (1956: 40).

It should be noted that in modern phonological descriptions of a non-pedagogical nature, phonological representations are rarely treated in terms of orthographic forms. The neogrammarian theories of language, to which the phonemic principle is heir, were often so concentrated on idiolects and individual linguistic behavior that something as conventionalized as spelling could hardly have been synchronically revealing of a 'language custom' (Weinreich, Labov & Herzog 1968). However, Chomsky & Halle (1968) interpret the phonetic opaqueness and apparent inconsistencies of orthography, such as that of English, as desirable according to the principles of generative phonology, which views speech output as the product of systematic phonological rules operating on 'underlying' representations of lexical items. They claim that in English spelling, phonetic variability is not indicated where it is predictable by a general rule of phonology, so that it maintains a close correspondence between semantic units and orthographic forms. Though in no way a phonological reality, spelling is the custom around which idiolects vary. English orthography is to Chomsky and Halle 'a near optimal system for the lexical representation of English words,' because it represents what native English speakers know about their language (1968: 49). The 'conservative' nature of English spelling maintains what is psychologically real and salient about English words, namely, their etymological and semantic relations.

Hebrew, too, represents what is psychologically salient about its lexical items through a highly conservative orthography. I have shown elsewhere how in reading unvocalized Hebrew texts, which is the form of most modern printed Hebrew, lexical information is extracted primarily via the unpronounceable tri-consonantal wrvw [So@rES], or 'root.' These graphemes preserve distinctions orthographically that existed phonologically in Classical Hebrew, but which are not represented fully in any Hebrew pronunciation tradition, giving the spelling a phonemically (though not morpho-phonemically) 'historical' or 'archaic' character (Rosén 1977: 67). Hebrew is thus considered to have a 'deep' orthography, where semantic information has significantly more psychological salience than phonological or phonetic information (Strolovitch 1996: 5). Moreover, orthographic forms that have existed in the Hebrew corpus since Biblical times have shown extraordinarily little variation in spelling in nearly every phase of Hebrew writing, including their forms in calque translations of texts into other Jewish languages. For example, Hebraisms in Yiddish were readily identifiable in writing as such by their 'un-Yiddish' spelling which, in keeping with their unvocalized Hebrew forms, did not make use of orthographic vowels (combinations of the four matres lectionis a, v, y, and i, some of the diacritic dashes and points) in the Hebraic root. However, a phonological analysis such as Morag's cited above puts two entirely different levels of linguistic knowledge on the same plane of inquiry. It forces a comparison between the different phonetic outputs of a non-native language to two native systems, using terminology which Blanc (1964, 1968) more accurately applies to characterize the dominant synchronic phonemic inventories of Israeli Hebrew.

This tradition of analysis, whereby graphemes function as phonemes, can be reconciled by virtue of the diglossic nature of pre-revival Hebrew. Standardized orthographies fail to reflect phonetic reality not only because of their inefficiency in representing constantly-changing spoken language, but because they are more efficient when they have a cross-dialectical application (Wexler 1971: 336). Having fractionalized into different Whole Hebrew pronunciations, and having been integrated differentially into the Merged Hebrew components of Jewish languages, the written norm of pre-revival Hebrew was not intended to have a super-dialectical function. In this kind of diglossic situation, the orthography is an archetype for H-language performance. Yet the Hebrew Language Committee specifically intended to supersede the super-dialectical function of the written norm in their incipient non-diglossic Hebrew. This is evident from the pronunciation prescriptions which, when adopted by revivalists, were designed not to reflect Chomsky & Halle's orthographic ideals, but to circumvent native phonology. For pre-vernacular Hebrew, whose performance was intimately tied to its visual representation, the absence of native speakers meant that knowledge of Hebrew was very much a textual matter. This conception carried over into the linguistic analysis of post-revival Hebrew, in terms of the locus of sound change in the language revival.

Berman (1978: 9) cites evidence from advanced courses in phonology at Tel-Aviv University, where students with considerable background in contemporary issues in phonological theory seem to have difficulty in distinguishing between sound and letter when discussing the structure of Modern Hebrew. She believes that because of the extremely low rate of illiteracy among native Hebrew speakers, orthographically 'naïve' informants are rare. Bentur (1978) maintains that in Israeli Hebrew, exposure to orthography can in fact lead to modification of the speaker's grammar (in the generativist sense). She refers to the application of a historical /a/-insertion rule whose structural description includes pharyngeal segments. The phonemic status of these segments in Israeli Hebrew is at the very least abstract and underlying, since they do not surface in most informal speech. Based on the results of word formation tests, she determined that such phonological rules can be psychologically real without being extended to all new formations which meet their structural description, because of access to orthographic information. Therefore, 'disregarding the relevance of orthographic data in phonological analyses results...in a misrepresentation of the speaker's knowledge' (1978: 21). In fact, she offers a synchronic description of the rule which includes orthographic constraints on the conditioning environment, namely, that the /a/ is inserted before [/] and [x] only when these phones represent realizations of i and c. Because the 'phonemes' formerly known as i and c are synchronically realized as [/] and [x], which are historically the reflexes of a and k exclusively, without the orthographic condition specifying that only underlying /÷/ and /_/ incur the rule, i.e. the phonemes historically represented by i and c but no longer realized, the rule would not express the 'valid generalization' which is the object of generative phonology.


4.3. NON-NORMATIVITY AND PSYCHOLOGICAL REALITY

Bentur's analysis, though by no means prescriptively oriented, is nonetheless inspired by the classical rules of Hebrew grammar, which are assumed to have persisted in some form throughout Hebrew's diglossic existence and to have adapted to the exigencies of modern usage. To determine the degree to which this is so requires linguistic study beyond the scope of this paper. Nevertheless, the influence of H-language normativism in even the most objective linguistic analysis of Modern Hebrew is typical of other situations of weakened diglossia. Rabin (1986) cites the first grammars of the European Romance vernaculars as an example of this practice, stating that these descriptions 'continued to be written in slavish imitation of Latin grammar' (1986: 548).

The so-called spirantization rule provides an excellent illustration. Faber (1986) states that the rule was productive in Biblical Hebrew, applying across word boundaries and within phrases. Six obstruent consonants, /b g d k p t/, alternated allophonically in post-vocalic environments with the corresponding fricatives, i.e. [v Ÿ D x f T], except when geminated. An acronym formed from the corresponding graphemes, tpkJdgb, shows not only the other name by which the rule is known, the 'Begad-Kefat' rule, but it also shows its limited application in current Hebrew phonology to /b k p/. Furthermore, variable application has resulted in several phenomena, such as the phonemic split of several formerly allophonic pairs, and hence the co-existence of spirantized and non-spirantized forms, sometimes as semantic doublets, e.g. rbcth, where normative [hitxave@r] = 'join, unite' and non-spirantized [hitxabe@r] = 'become friends.' Fischler (1981) discusses how the rule is construed by the Academy and normativists, yet he also offers a plethora of examples of its non-operation, especially in regard to borrowings which conform entirely to native morphological patterning yet show almost no variance in their non-observance of spirantization. Bar-Adon (1977) gives credit to children for 'revolutionizing' the morphophonemics of /b g d k p t/. He cites forms with initial spirantization, e.g. [fixe@d] 'he feared,' post-consonantal spirantization, e.g. [likfo@c] 'to jump,' and post-vocalic de-spirantization, e.g. [Saba@r] 'he broke' as having gained currency in General Israeli Hebrew due to the persistence of these forms as produced by native children. He thus credits the first native speaking children with 'exceptional creativity,' and systematizes non-normative forms as part and parcel of new linguistic intuitions in a native Hebrew competence.

Kiparsky (1971) also sees a role for children in determining the current state of spirantization, although of a less pro-active nature. He believes that the spirantization rule has become 'opaque' in Modern Hebrew, because the surface output of the rule, i.e. the fricatives, occur in environments other than that predicted by the rule. The merger of /_/ ('historical c') with [x] (the spirantized form of /k/), which occurred prior to the speech revival, means that the output of spirantization has another source in the grammar. In addition, the non-articulation of historical schwas means that spirantization seems to have applied in an environment where it should not have, i.e. after consonants, as in /Saf´x+u/ > [Safxu] 'they spilled' (Bolozky 1978: 34). While the spirantization of /p b/ remains 'transparent,' Kiparsky cites evidence that children have more difficulty learning the opaque part of Modern Hebrew spirantization, thereby producing forms such as [lekabe@s] and [xvisa@] where the rule calls for [lexabe@s] and [kvisa@] 'to wash / laundry' (1971: 78). In response, however, Ben-Horin & Bolozky (1972) object to the postulation of any sort of general spirantization rule in Modern Hebrew. They note that the sounds to which the rule applies do not form any kind of natural class of segments, and claim that the spirantization rule may in fact have no psychological reality to Modern Hebrew speakers:

...it took time before psychologically real (i.e. productive) rules were crystallized. We are not sure whether there is a productive spirantization rule in Modern Hebrew. And even if there is such a rule, it is obviously not a direct descendent of the more general spirantization rule, since the latter never existed in Modern Hebrew. (1972: E34)

This tentative conclusion betrays the importance of the difference between the psychologically reality of the rules of diglossic Hebrew and of the internalized system of language developed by native Hebrew speakers. Until 'rules' of this latter sort crystallized in the renativized grammar of native speakers, the Hebrew language did not have an autonomous phonological component which could operate in the grammar of a once-native language. In developing their linguistic competence, children do make 'mistakes' in the sense of over-generalizing, or 'optimizing,' the patterns of the language they are acquiring. Some linguistic change may result from children's optimization of the grammar having persisted into their linguistic adulthood (King 1969). Still, to what extent these over-generalizations may replace 'adult' forms, and to what extent they actually reflect a changed linguistic competence, are difficult to establish. Thus it is equally difficult to determine to what degree the underlying representations and phonological rules of Israeli Hebrew are the result of natural evolution from those of the classical language during its restricted existence. What can be said is that the first native speakers were, by definition, not in contact with the speech of a native Hebrew competence -- their mother tongue was not 'normally transmitted' to them. At the very least, this must have resulted in differences greater than normal between the grammar constructed by the first Hebrew children and the grammar of those whose speech constituted their linguistic experience (Kiparsky 1968).

Having begun an article on the diachronic transition from Classical to Israeli Hebrew with the appropriate hesitation for treating such changes as 'legitimate examples of linguistic change,' Rosén (1964: 832) concludes that 'the processes [of systemic change]...are of a purely internal nature (i.e. features of "diachrony" and not of "contact").' However, this conclusion is unsatisfactory. Changes in the Hebrew system are most definitely features of contact, between the native competence of non-Hebrew speakers and their performance in Hebrew through the ages. Moreover, these changes were accentuated when foreign substrata necessarily became the base upon which native Hebrew speech developed. However, as I have stated, the discontinuity in native competence meant that the rules which generate every grammatical Hebrew sentence have not been transmitted normally through the ages from speakers of Classical Hebrew to speakers of Israeli Hebrew. One cannot simply add current linguistic data to that of past centuries of Hebrew speech as if they were drawn from the same speech community, since the earlier community no longer exists (Labov 1994: 20). Thus the question which the next section addresses is, what were the mechanisms and influences which allowed an abnormally transmitted language to regain its potential for normal transmission, and thus normal linguistic change? In other words, what in the development of Israeli Hebrew can account for the consistent discrepancy not only between normativism and actual usage, but also between the contradictory views on the diachrony of the Hebrew language?


[previous section] [top] [next section] [table of contents] [back home]



© 1997 Devon L. Strolovitch