Reflexes of the Proto-Berber Glottal Stop in Nefusa and Ghadames 1 By MARIJN VAN PUTTEN (Leiden) One of the more recent and significant advancements of Proto-Berber recon- struction has been the discovery that the Zenaga Berber Ҍ is a reflex of an old Proto-Berber laryngeal *Ҍ (Taine-Cheikh 2004, Kossmann 2001b, Prasse 2011). Making it possible to now distinguish between three separate ‘laryngeal’ conso- nants, *Ҍ, *ȕ and *h. The Ҍ in Zenaga is a reflex of Proto-Berber *Ҍ or *ћ. It only appears in non- word-final coda position. Reflexes of an original word-final syllable |vݦ| howev- er, are kept distinct from original vowels in the same position. Original word- final vowels terminate in a non-etymological off-glide [h] (cf. Kossmann 2001b: 62; Taine-Cheikh 1999: 309-311; Cohen & Taine-Cheikh 2000: 284), while synchronic word-final vowels go back to an original glottal stop. Word-final Ҍ resurfaces when suffixes follow it, e.g. pf. 3sg.m. y-Ωnnäh /y-ԥnna/ 2 pf. 3pl.m. Ωnnä-n /ԥnna-n/ ‘to say’ pf. 3pl.m. y-ΩҌ/y-ԥݦnaݦ/ pf. 3pl.m. ΩҌnaҌ-n /ԥݦnaݦ-n/ ‘to kill’ For many languages evidence for *Ҍ has been lost completely, but Kossmann (2001b) convincingly shows that the Ghadames retains a reflex of the Proto- Berber glottal stop in the development *ăҌ > o. In the verbal system we find initial o in Ghadames that corresponds to verbs with an initial glottal stop in Zenaga. Ghadames can thus distinguish initial long- vowel verbs from initial glottal stop verbs, which are clearly distinct in Zenaga, but are generally not distinguished in other Berber languages, e.g. 1 I thank Maarten Kossmann, Lameen Souag and Ahmad Al-Jallad for commenting on early drafts of this paper. This article cites lexical material from the following lan- guages and their respective sources: Nefusa (Beguinot 1942, Provasi 1973); Ghada- mes (Lanfry 1973); Zenaga (Taine-Cheikh 2008; 2010); Middle Atlas Berber (Taïfi 1991); Sokna (Sarnelli 1924); El-Foqaha (Paradisi 1963); Awjila (Paradisi 1960, Van Putten 2014); Figuig (Kossmann 1997); Kabyle (Dallet 1982); Mzab (Delheure 1984); Ouargla (Delheure 1987); Tashelhiyt (Stroomer fthc.); Beni Snous, Beni Menacer, Metmata (Destaing 1914); Tuareg (Heath 2006). 2 Taine-Cheikh (2008) employs a highly phonetic transcription of Zenaga. I provide a phonemic analysis with these forms where I assume that the contrast of the short vowels [i] and [u] vowels are mostly allophones of /ԥ/ which ultimately is a merger of Proto-Berber *i, *e, *u and *Ω. Taine-Cheikh (1997: 120-121) shows that the vowel contrast between [i] and [u] is phonemic, but the number of minimal pairs is limited, and it is often impossible to determine the original value.