Abstract
The theoretical view that sound change is selective is attractive to specialists in language variation and change, who often encounter diachronic developments which appear to propagate through phonological systems selectively, frequently by phonetically defined classes. Yet two areas remain ill understood: how innovations affect members within a narrowly defined class, and what effects extrasystemic factors can have on selective change. This study begins with the view that incipience and expansion of sound change is selective within classes, and that the relative propensity of class members to undergo change in specific environments is expected to be constant within a given language. The history and present state of surface-level spirantization of post-vocalic /p t k/ in the Italian region of Tuscany—e.g. /la kasa / → above; [laha:sa] ‘the house’, /diko/ → above; [di:ho] ‘I say’—appears to violate these basic principles. Detailed analysis of change in progress suggests, however, that sociolinguistic factors operative during the variation phase of implementation can be responsible for disturbance of system-internal implicational directionality.