The phonetics of anaphonesis between production and perception

In Florentine Italian high vowels /i/ and /u/ occur before a velar nasal instead of the expected /e/ and /o/, e.g. vinco ‘I win’ < VĬNCO, punto ‘point’ < PŬNCTUM (cf. lembo ‘hem’ < LĬMBUM, rompo ‘I break’ < RŬMPO). This phenomenon, referred to as ‘anaphonesis’, represents a tricky phonetic process of the history of Italian, which has been claimed may have some relationships with the universals of vowel nasality and nasal consonants. It has been observed that there is an evident front-back asymmetry in the words triggering anaphonesis: front vowels are more prone to anaphonesis than back vowels. Moreover, there appears to be an effect of the postnasal, since back vowels show anaphonesis only in the case of voiced postnasal (fungo ‘mushroom’ < FŬNGUS vs. tronco ‘trunk’ < TRŬNCUS). In order to explore the parallels between documented synchronic and diachronic variation, the complexity and diversity of nasal vowel evolution in Romance are considered to have probative value and are reconsidered in the light of the results of two production and perception experiments. These experiments have been designed in order to verify: i) whether the two asymmetries historically associated with anaphonesis (front-back vowels, voiced-unvoiced postnasal) are present in dialects for which anaphonesis is historically attested, absent in those for which it is not; ii) if present, whether the presence of these asymmetries is perceptually salient. Production data show that in the triggering contexts for anaphonesis, Florentine vernacular speech synchronically mirrors the asymmetry between front and back vowels which has been observed all along the history of Italian. Crucially, the simultaneously presence of onk and ung clusters in Italian lexicon is mirrored both in synchrony and to a certain extent also in perception. In Florentine onk clusters are admitted in Florentine vernacular speech, while enk clusters are considered almost impossible. In the perceptual domain, onk/unk clusters showed the highest rate of misperception: they are precisely the clusters which appear to be somehow singular with respect to the others (ink, ing, ung).