Sibilants in Libyco-Berber.

• Author: Kossmann, Maarten
• Position: Report

stridentihus horrida Unguis verba ferunt (Corippus, Iohannis, IV.351-52) '

1. INTRODUCTION

   There exists little doubt that the Libyco-Berber inscriptions--a large and heterogeneous group of texts from northern Africa, dating from antiquity--are related to the Berber cultural and linguistic sphere. There is a lot of corollary evidence for this, ranging from the fact that a descendant of the script is still in use by the Tuaregs, who speak a Berber language, to the persistence of the Libyco-Berber onomasticon in tribal names from the Islamic period (cf. Jongeling 1994: xi-xxi). Unfortunately, the inscriptions themselves provide only very restricted linguistic material, consisting mainly of names and functions (Fevrier 1956; Rebuffat 2006). Among these few elements, some can easily be equated with Berber terms, such as gld 'king' (cf. the general Berber term agzllid 'king', Nait-Zerrad 2002: 773-74), and the single-letter proclitics w 'son of, n 'of, and d 'and', all of which have exact equivalents in modern Berber. The analysis of most other terms is strongly debated, and although Berber equivalents have been proposed (e.g., Marcy 1936; Rossler 2001a [1958]; Rossler 2001b [1962-1964]; Chaker 1995, 2005), (2) there is no consensus about this.

   Few people therefore doubt a relationship of Libyco-Berber with Berber in general, and even the late Lionel Galand, who was once highly skeptical of Berberocentric interpretations of the texts (e.g., Galand 2002a [1996]), sometimes admitted Libyco-Berber facts when analyzing Berber linguistic history (Galand 2002c [1991], more cautiously 2010: 215).

   Considering Libyco-Berber to be part of the Berber sphere does not automatically imply that it is "simply" a Berber language, i.e., that it would have descended from the same proto-language as that reconstructible on the basis of the modern Berber languages. This is a point made by Louali and Philippson (2004: 114). In their view, Libyco-Berber could have been a sister language of the ancestor of the modern Berber languages rather than a member of the group itself. The little lexical information that we have at our disposal is not very helpful for this question. Elements such as the prepositions n, d, or the word w- could already have existed on a higher genetic node, while the cultural item gld could easily be a loan from Libyco-Berber into proto-Berber or the other way around. In this article, I will analyze one specific element of the phonological system, sibilants, and discuss the possible implications for the larger question of the relationship of Libyco-Berber to modern Berber languages.

   (2.) THE ANALYSIS OF THE GRAPHEMIC SYSTEM

   In 1973, Lionel Galand, who maintained a life-long ambivalence as to the interpretation of Libyco-Berber, vacillating between strong skepticism and cautious optimism (Galand 2002d), endeavored to employ the only alternative to lexical comparison in addressing the status of Libyco-Berber: the analysis of the phonemic system as represented in writing (Galand 2002b [1973]). While the question of the relationship of Libyco-Berber to proto-Berber may have been irrelevant to his thinking, as he did not accept the notion "proto-Berber" (Galand 2010: 13-14), his type of argumentation is certainly relevant to this purpose.

   Libyco-Berber inscriptions are written in various scripts, they are attested over a large area, and they date from different periods (for recent overviews, see Pichler 2007; Springer Bunk 2014; Casajus 2015; Mora Aguiar 2017). Therefore, it does not make much sense to provide a phonological analysis of all Libyco-Berber texts together, as there is no reason to assume that they all represent the same stage and dialects of the language. Instead, it is better to follow Galand's example and focus on the only group of texts that have a clear dating and for which the letter values are largely established. These are the official inscriptions from the city of Dougga in present-day Tunisia (Latin: Thugga; Punic and Libyco-Berber: tbgg), a closely related group of texts using the same script, and different from all other Libyco-Berber inscriptions in that they are written from right to left rather than from bottom to top. Two of these inscriptions are bilingual and contain both a Punic and a Libyco-Berber text. The comparison of the two versions has allowed researchers to establish equivalents between Punic and Libyco-Berber for almost all letters of the Dougga alphabet. Up to now, fourteen texts in the Dougga script have been published. Eleven (3) were included in Chabot's Recueil d'inscriptions libyques (RIL 1-11), while three more were brought to light by Mansour Ghaki (Ghaki 1997, 2002, 2009; Rebuffat 2013). (4) Three texts are short fragments that contain only one or two words; the others contain up to fifty-two orthographical words (RIL 2). RIL 2 is dated to 138 BCE, and the other texts are probably from roughly the same period (Fevrier 1964-1965: 85).

   The script used in the Dougga inscriptions is similar to that in vertical inscriptions from the same region, but it contains a number of letters that are absent in the latter (Pichler 2007: 69).

   The letters and their equivalents in Punic are provided in Table One below. (5) In the transcription and analysis I basically follow Galand (similarly Casajus 2015: 73), with some adaptations. (6) Chabot (1940), (7) Rossler (2001a [1958]), Pichler (2007: 47, 69, 73), and Kerr (2010: 54) use different conventions, which are given for the sake of reference.

   In his comparison of the Libyco-Berber and Berber systems, Galand points to a number of structural similarities. (12) While Galand remains somewhat vague about how he interprets the results of his comparison of the Dougga system and Berber phonological systems, other scholars are more outspoken, and consider the two to be highly similar, e.g., Chaker (1984: 275) and Casajus (2015: 45-46). Most prominent among these similarities are:

   1) the presence of only two labial stops (or fricatives), and 2) the presence of no more and no fewer than two emphatic sounds (excluding from the equation, which Galand does not count as an emphatic).

   The presence of only two labial stops (or fricatives) provides a clear parallel to reconstructible Berber. Even though Kossmann (1999: 249) reconstructs a triple opposition /f/--/b/ - /[beta]/, there is only very scanty evidence for (*)b versus (*)[beta], and a slightly more audacious reconstruction could easily do away with the opposition. Of course, this feature is hardly a decisive argument in favor of a Berber link, as two-way systems are well attested in Afroasiatic (among others, in Semitic), and elsewhere in African languages (cf. Maddieson 2013).

   The presence of two emphatic sounds is hardly a convincing parallel between Libyco-Berber and proto-Berber either. In the first place, the Punic script distinguishes only two emphatic sounds, and ; thus, a further emphatic sound could easily hide behind one of the Libyco-Berber letters corresponding to Punic or .

   The interpretation of is especially problematic. (13) It is far from certain that in Punic the sound written by the letter was merely [s] with some kind of special "emphatic" articulation (pharyngealization or some kind of glottalization). It may very well have been affricated as some scholars have argued (cf. Friedrich and Rollig [.sup.3]1999: 26; Krahmalkov 2001: 24). In the latter case, the choice of the Punic equivalent to Libyco-Berber could have been triggered by a similarity in assibilation and not by the emphatic quality of the sound. Thus, there is little to argue in favor of an interpretation of Libyan as emphatic ([s] or [s']) rather than, for example, non-emphatic [c] or [[t.sup.s]]. Moreover, as Jongeling (1994: 132) has pointed out, Latin renderings of the Libyco-Berber name sdn suggest that was an affricate, cf. sicldin, tziddin, and stiddin (see below); similar transcriptions are found in Latin renderings of Neo-Punic (Krahmalkov 2001: 24; Kerr 2013: 20-21). All in all, the evidence is ambiguous in the case of the emphatic consonants, and Galand's interpretation is less compelling than it may look at first sight.

   (3.) THE SIBILANTS

   As noted by Galand, the situation with sibilants looks very different from modern Berber (Galand 2002b [1973]: 34-35). Most Berber languages nowadays distinguish Isl, Isl, Isl and /z/, /z/, Izl, that is, six sibilants. Among these, /s/ is either a loan phoneme from Arabic or an assimilated variant of /z/. Moreover, as shown in Kossmann (1999: 219-35), there is hardly any evidence for an ancient opposition between Isl, Izl vs. Isl, Izl. Thus, the most probable reconstruction for proto-Berber has only three sibilants, (*)s, (*)z, (*)z (Prasse 2015: 6484).

   There are a few Berber varieties that present complications to this reconstruction. In the first place, in Zenaga of Mauritania the regular correspondent to northern Berber /s/ is [f], while the regular correspondent to /z/ is [[theta]]. There are two correspondents to northern Berber /z/: [[theta]] and [3] (non-strident), both corresponding to [z:] when geminated. Before velars, a further realization, [z], is found (Taine-Cheikh 2008: lxxiii-lxxiv). In the present language the presence of [[theta]] and [3] is unconditioned, and they thus represent different phonemes. There are, however, major tendencies as to their distribution, which suggests that the opposition is the result...