A core finding of sociolinguistic research is that phonological changes may become subject to social evaluation as they propagate through a speech community. Much work has analyzed the social evaluation of innovations by categorizing them in terms of their salience, or the degree of stigma/prestige attached to them. However, recent studies have attempted to develop a more sophisticated understanding of the types of social-indexical meaning which can attach to linguistic variation. In particular, it is argued that speaker-listeners' social perceptions of variation tend to be structured around locally-relevant stereotypical figures such as the valley girl' or burnout' (Eckert, 2008; Moore & Podesva, 2009; D'onofrio, 2015). It has been claimed that innovations which come about due to global processes of change (such as chain-shifting principles) may become associated with these local stereotypes, resulting in divergent behavior on the part of some groups of speakers (Haddican et al. 2013). This raises the question of how the social meanings proposed for a given innovation can be identified, and how the role of social indexicality in constraining the propagation of that innovation can be diagnosed. With a view to addressing this problem, this paper presents an account of the fronting and diphthongization of the tense back vowels /u/ and /o/ in York, Northern England. After presenting evidence of change in these vowels, the paper evaluates a recent account of the role of social indexicality in shaping the outcome of this change. It does this using data from a novel experimental paradigm, which allows the testing of quantitative predictions regarding the degree to which different dimensions of social-indexical meaning are associated with different acoustic dimensions of the vowels under study. The findings contradict previous accounts of the role of social indexicality in constraining /u/ and /o/ fronting in York, highlighting the risk of inferring social-indexical meanings primarily from production patterns. For example, despite its rapid and uniform incrementation and apparent lack of class-stratification in production, /u/ fronting emerges as a robust cue to socioeconomic status in perception. While the absence of fronted /o/ monophthongs in the speech of younger speakers has previously been interpreted as evidence of their association with a stigmatized working-class stereotype, there is no support for this in the perception data. Intriguingly, there is evidence of structured variability in listeners' social-perceptual responses --- the social groups who lead the changes in production appear to respond more consistently in the social perception tasks, suggesting that some speaker-listeners may be more sensitive to the social-indexical information conveyed by variation in these vowels than others.