Sound systems in diachrony: Sibilants and affricates in Udi Jost Gippert Goethe-Universität, Frankfurt am Main 1 Preliminaries It is a well accepted fact that the Khoisan languages of Southern Africa possess the largest consonant systems worldwide, with inventories partly exceeding 100 phonemes. It is as well accepted that the next richest inventories are met with in the Caucasus region, in the languages of the (North-)West Caucasian and (North-) East Caucasian families. Different from the Khoisan languages, the richness is not determined by clicks in these languages; instead, we usually find at least three different articulation types of stops and affricates here (voiced, voiceless-aspirat- ed, voiceless-glottalised), extending from labials down to uvulars, pharyngeals, and laryngeals, and partly comprising palatal or labial coarticulations. In this way, a total of 85 phonemes was determined for Ubykh, a West Caucasian language now extinct (the last known speaker died in 1994), and a not much smaller inven- tory of 82 has been assigned to the Archi language of the East Caucasian stock, which is in general especially characterised by sets of lateral affricates. 1 A peculiar position within the latter language family is held by Udi, an East Caucasian language spoken by about 5,000 speakers in North-West Azerbaijan, South-East Georgia, North-East Armenia, and elsewhere in states of the former Soviet Union. Udi is not only comparatively poor with respect to laterals as well as uvular and pharyngeal consonants, but it is also the only East Caucasian language whose history can be traced back for about 1,500 years, given that it has been ascertained to be the modern offshoot of the medieval language of the Cau- casian “Albanians”. With the detection of a large amount of written text in the underwriting of palimpsests from St. Catherine’s Monastery on Mt. Sinai, the sound system of this latter language (as spoken by the middle of the first millen- nium C.E.) has been established with high certainty, and among the 52 letters of the peculiar alphabet used for writing “Caucasian Albanian” (hereafter: CA), 45 have been determined to represent consonant phonemes. In contrast to this, the inventory of modern Udi is slightly smaller, being confined to something between 32 and 38 consonants depending on the source used. A peculiar problem in this 1 It was in 1995, on the occasion of Anthony Traill visiting Frankfurt that I first discussed the question of the comparability of Khoisan and Caucasian sound systems with Rainer Vossen – the present article is meant to be a small gift in return to him for drawing my attention to the fascinating world of clicks.