This paper explores the link between a synchronic change in progress in Sicilian, where the diphthong /w-open O/ is monophthongizing to /open O/, and a related diachronic change in the corresponding Spanish diphthongs (/w-open O/>/wo/>/we/). It is proposed that both these changes are motivated by the resolution of a perceptually non-optimal sequence of segments. Experimental evidence is presented showing a difference in perceptual salience between diphthongs consisting of /j/ plus a mid front vowel (i.e. Sicilian /j-open/ and Spanish /je/) and diphthongs consisting of /w/ plus a mid back vowel. It is argued that this difference makes the former more stable over time and the latter more susceptible to change, either through dissimilation (Spanish) or deletion (Sicilian).