On the Nature of Proto-Semitic *l

Abstract

BASED ON THE REGULARITY of the correspondences supporting it, it would appear that reconstruction of Proto-Semitic *1 is uncontroversial and, by implication, uninteresting. It is the purpose of this note to demonstrate that, clear-cut though reconstruction of a Proto-Semitic *1 may be, it is not uninteresting. On the contrary, consideration of the phonetic nature of *l and of its reflexes has the potential to illuminate several problems of Semitic phonology, and, hence, is of great interest. In ordinary phonetic and phonological usage, the symbol /1/ denotes a voiced lateral approximant.' Two basic types of voiced lateral approximant are known, one with primary constriction in the alveodental region, and one in which this constriction is supplemented or supplanted by a velar constriction. Both types are found in many dialects of English, in complementary distribution: clear, alveolar [1]2 occurs in pre-vocalic position (e.g., leap), and dark, velarized [1] post-vocalically (e.g., pool) (Catford 1977, 19293). (In some dialects of American English, one variety or the other predominates: some New York urban dialects have velarized [1] in all positions, while some southern American dialects have only alveolar [1], [1] having changed to /w/ or been lost.) When the symbol 1 is used in transliterations, transcriptions, and reconstructions of forms in ancient languages, it is easy to interpret it as referring to a non-velarized, clear, approximant [1] rather than simply to a lateral with unspecified place and manner of articulation.3 In particular, given the apparent marked status of [1], it is easy to assume from consonantal inventories like that given in Moscati (1969) that Proto-Semitic *I was demonstrably a clear [1], despite the fact that this position has not to my knowledge been explicitly argued for. In fact, dispassionate consideration of the facts suggests, rather, that Proto-Semitic *1 was actually a velarized [1]. The primary evidence that Proto-Semitic *1 was actually dark *[I] comes from evidence regarding the nature of /1/ in Akkadian, Hebrew and Aramaic. In all three cases, phonological phenomena are best explained by the assumption that /1/ was [1]. The first case is the most straightforward. Steiner (1987) discusses a hitherto un-noted sound change in Hebrew and Aramaic: *aw becomes / ui/ rather than the regular /a/ when preceded by /1/. This change accounts for the difference between the parallel reduplicated forms 1lz^1v 'sprout' (< *lawlab < *lablab) and koxaiv 'star' (<*kawkab < *kabkab).4 Steiner suggests that this change would be explicable if the Hebrew/Aramaic /1/ were [1], at least in word-initial position. Rabin (1970) and Blau (1977) discuss an additional anomalous and sporadic development in Hebrew. In approximately 30 roots discussed by Rabin,5 *d appears as /d/ rather than the expected /z/. Rabin notes that in ' See Catford (1977, 132-33, 212) for discussion of the articulatory nature of approximant, fricative, and affricate laterals. Essentially, an approximant is produced with a more open articulatory channel than that involved in fricative articulation. As a result of this articulatory difference, both voiced and voiceless fricatives will be produced with sufficient airflow to produce acoustic turbulence, while only in the case of a voiceless approximant will turbulence be generated. To over-simplify only slightly, a voiced approximant will sound more vowel-like than will its voiceless equivalent. Besides /1/, other commonly recognized English approximants are /r w y/. 2 The various diacritic modifications of 1 used herein are to be interpreted as follows: [1] denotes a voiced alveolar lateral approximant and [1] the voiceless equivalent thereof. [I] denotes a voiceless alveolar lateral fricative. [1] denotes a voiced velarized alveolar lateral approximant and [1] its voiceless equivalent. For the distinction between approximants and fricatives, see the previous note. 3 The epistemological basis for this confusion appears to be the systematic ambiguity between 1 as the symbol for any member of the set of laterals, regardless of place and manner of articulation, as well as for the linguistically unmarked member of that set, i.e., a voiced alveolar lateral approximant. 4 Other forms resulting from this change are, according to Steiner (1987) liuah 'tablet', l0z 'almond tree', and 12 'if'. 5 Blau (1977, 110) rejects many of these candidates on semantic grounds. Secure cases include Hebrew ndr 'vow', 33