The non-initial-syllable vowel reductions from Proto-Uralic to Proto-Finnic 1 . Non-initial-syllable vocalism in Proto-Uralic

According to our Jubilar Juha Janhunen (1981), the Proto-Uralic non-initial unstressed syllables had two vocalic archiphonemes, a low *A and a high *I, both of which had front and back allophones depending on the vowel harmony, that is, whether the initial stressed syllables had front or back vowels, respectively (see also Korhonen 1988; Sammallahti 1988).1 Before this view, however, Janhunen (1982)2 reconstructed reduced *ǝ instead of high *I (cf. *ǝ and *ǝ̑ already in Lehtisalo 1936: 16–17, 22–23). In other words, the Proto-Uralic non-initial unstressed syllables would not have had the qualitative opposition between low *A and high *I but a quantitative opposition between full *A and reduced *Ə, which can still be supported by the following facts: ○ Samoyed: non-initial-syllable *ǝ and *ǝ̑ (= Janhunen’s *i and *ï) were either lost or merged as *ø (= *ǝ), whereas *a and *ä were usually preserved as such (Sammallahti 1988: 485). ○ Mari: non-initial-syllable *ǝ and *ǝ̑ were either lost or merged as *ǝ̑ , whereas *a and *ä were usually only reduced to *ǝ̑ (Bereczki 1994: 121–136). ○ Mordvin: non-initial-syllable *ǝ and *ǝ̑ were either lost or preserved as *ǝ and *ǝ̑ , whereas *a and *ä were at most reduced to *ǝ and *ǝ̑ (Bartens 1999: 60–66). ○ Saami: non-initial-syllable *ǝ and *ǝ̑ usually developed to short *e̮ , whereas *a and *ä usually developed to long *ē or *ā (Korhonen 1981: 99–102). ○ Finnic: non-initial-syllable *ǝ and *ǝ̑ of words of three syllables or more were more likely to be lost than *a and *ä (see Section 3 below).