Ensitavun vokaalin kadosta mordvassa

This paper studies the conditions for the loss of the first-syllable vowel (V1) in Mordvin and the chronology. I look at which first-syllable vowels have disappeared, in which environments, and from which dialects. I also investigate the phonetic developments of the word-initial consonant clusters and non-first-syllable vowels closely related to V1 loss. I evaluate etymologies and cognates proposed for words that have experienced V1 loss, and propose a number of new etymologies.I have collected from Heikki Paasonen’s Mordwinisches Wörterbuch (MdWb) all the entries common to both Mordvin languages that begin with a probably non-borrowed, non-iconic consonant cluster. The MdWb materials have large dialectological gaps to be filled in by further research with the help of other sources.I have divided the loss of V1 into three developments with different conditions. The oldest, the “SK loss” already advanced in Proto-Mordvin (PM), occurred before the affricate č and before the fricative consonant clusters st, sn, śk, śt́, št, šk, št́, šn, šń, ft́, or χt́ (41 cases, 24 of which are probably PM and 10 possibly PM). The later “R loss” before sonorants took largely place after the split of PM (17 cases, six of which are probably PM). Both occurred through a shift of stress to the earlier second-syllable vowel (V2), which at the time of the loss was a full vowel. In the oldest cases, the loss is represented across the whole of Mordvin, while in the younger ones it alternates with some cases of retention; the identification of retention is complicated by some dialectal processes where new V1 vowels were inserted, which I try to map here. I also study the relationship of the variation kš- ~ š- (and ks- ~ s-) to V1 loss.A separate development is the reduction in the number of syllables through the fusion *-urə̑- > *-r̥- without shifting stress (six cases, two of which are possibly PM). In the dialects of Erzya and Moksha, this PM *-r̥- is represented variably as -ro- ~ -ru- ~ -or- ~ -ur- ~ -ə̑r-.Any vowel could undergo loss before an open V2; SK loss deleted also *ŕ̥ through consonantizing it to š. SK and R loss occurred before V2 *a, SK loss also before V2 (*ä >) *ε (in three cases of R loss, early V2 *ä > *a after *ŕ must be reconstructed). This *ε has in most of Erzya and Moksha developed into a word-final -i and undergone reduction inside the word, but the variation -ə- : -i has been evened out everywhere: in some Erzya dialects in favor of (*ə >) e, elsewhere in favor of i. In the westernmost dialects of both Erzya and Moksha, V2 *ε was first reduced everywhere and then joined the development *-ə > -ä / -e word-finally.In addition to the chronology of phonetic changes, my work clarifies the picture of PM stress. Contrary to the prevailing opinion, any vowel, including the open *a and *ä, could disappear from the first syllable. Hence, it is no longer justified to consider V1 loss as evidence for the primacy of Moksha stress conditions. Instead, the main stress of the PM word seems to have fallen on the last open vowel, even when V1 was not a close vowel.