Time: Monday 11\/30, 5pm, 32-D461
\nSpeaker: Sverre Johnsen (Harvard)
\nTitle: Contrast maintenance effects in Norwegian retroflexion<\/P><\/p>\n
\n
\nNorwegian contrasts morpheme initial \/s\/ and \/ʂ\/ before a vowel (\/sV-\/
\n- \/ʂV\/), but not before a consonant (\/sC-\/ - *\/ʂC-\/). Initial \/s\/
\nundergoes retroflexion to \/ʂ\/ when preceded by a morpheme ending in
\n\/-r\/. Before a vowel, the contrast between \/s\/ and \/ʂ\/ is then lost
\n(\/sV-\/, \/ʂV-\/ > \/ʂV-\/), but no contrast is lost before a consonant
\n(\/sC-\/ > \/ʂC-\/). I present data from two experiments showing that
\nNorwegian speakers apply this retroflexion significantly more often
\nbefore a consonant than before a vowel. The effect is that a contrast
\nis better maintained. Retroflexed \/ʂV\/-tokens have, however, more
\nlexical neighbors and a higher phonotactic probability than
\nretroflexed \/ʂC\/-tokens. Experiments in lexical access have shown that
\nitems with a high lexical neighborhood and phonotactic probability are
\nmore prone to be misidentified. I show that under an exemplar model of
\nspeech perception\/production, the asymmetry in retroflexion between
\n\/sV-\/ and \/sC-\/ is a direct consequence of such misidentifications.<\/P>\n<\/p><\/blockquote>\nUpcoming schedule: <\/p>\n