The rise of gemination in Celtic.

This study investigates systematically the emergence and establishment of geminate consonants as a phonological class in the Celtic branch of Indo-European. The approach of this study is comparative historical linguistics, drawing on diachronic structuralism combined with aspects of language contact studies and functional approaches to language usage. This study traces the development of geminates from Proto-Indo-European (fourth millennium B.C.), which did not allow geminate consonants, to the Common Celtic period (first millennium B.C.), when almost every consonant could occur as a singleton or as a geminate, and on to the earliest attested stages of the Insular Celtic languages (first millennium A.D.). Although they were prominent in the phonology of Proto- and Ancient Celtic (Gaulish, Celtiberian), ultimately geminates were gotten rid of as a phonological class in the individual Insular Celtic languages. This is probably due to the fact that the contrast between lenited and unlenited sounds took on a central role in Insular Celtic phonology, making gemination a phonetically redundant category. Most instances of geminate consonants in Celtic can be explained by regular sound change operating on inherited clusters of consonants. Each sound change will be discussed in a separate section in a rough chronological order. Effectively, gemination is largely a strategy to reduce the number of allowed consonant combinations. To a limited degree, gemination also had a morphological function, especially in the formation of personal names and in the creation of adjectival neologisms. However, there is a residue of words, especially nouns, in the Insular Celtic languages that defy any attempt at etymologising. They are prime suspects of having been borrowed from prehistoric, substratal languages.