1 The Origins of Personal Agreement Clitics in Caucasian Albanian and Udi Wolfgang Schulze (Munich) 1. Introduction Udi, belonging to the Southeast Caucasian (Lezgian) language family (Eastern Samur branch), represents one of the best studied minority languages of this family (see Schulze (in press) for a more comprehensive survey on the history of Udi linguistics). From a typological point of view, Udi has found much interest because of its system of so-called floating agreement markers that is said to be unique among the autochthonous languages of the Eastern Caucasus. In the present paper, dedicated to the jubilee with whom I had the honor to discuss over times issues of Caucasian Albanian and Udi grammar, I want to present some new thoughts on the origins of Udi and Caucasian Albanian patterns of personal agreement. The issue has become a hotspot not only in the linguistics of East Caucasian, but also in general linguistics due to the study by Alice Harris (Harris 2002) that has served as a starting point for several theory-driven proposals to interpret these patterns (e.g. Crysmann 2000, Luís & Spencer 2006). Most of these studies are based on the analyses and hypotheses put forward by Harris (2002) and do not offer new data or new arguments concerning the history and motivation of agreement constructions in Udi. Moreover, Harris' analysis and hypotheses could not yet consider data stemming the Mount Sinai palimpsests that include texts written in Caucasian Albanian (~ 600 AD). Jost Gippert and the author of the present article who had edited these palimpsests in collaboration with Zaza Aleksidze and Jean-Pierre Mahé (Gippert et al. 2009) could show the appropriateness of older claims according to which Udi is (directly) related to Caucasian Albanian (CA). In fact, the CA data shed new light on the question of how the agreement patterns of Udi may have emerged. In my paper, I have to confine myself to telling only 'half of the story'. I will concentrate on morphological morphosemantic issues, addressing the syntactic and pragmatic dimension occasionally only. Moreover, space does not permit to discuss the problem of clitic placement except for mentioning some more general observations related to statistics (see Schulze (forthcoming a) for a detailed presentation). The paper is organized as follows: In section 2, I present the 'basics' of Caucasian Albanian and Udi addressing positional and formal issues relevant for the topic of this paper. Section 3 turns to a specific problem of the Udi paradigm, namely the so-called third person singular Q-clitic -a. I will argue that -a is not the result of grammaticalizing a borrowed particle ya 'or', but represents an older pragmatic marker that had once encoded a verificational focus. The history of the general paradigms is discussed in section 4. Here, I combine formal and functional arguments in order to suggest that the agreement clitics do not stem from constructions that involve a focal cleft (Harris 2002), but from strategies of 'local' focus marking.