ABSTRACT
In this chapter, we quantitatively assess sociolinguistic coherence across multiple variables in two datasets, phonetic variation in Toronto English and grammatical variation in Bequia English. We compare the results of analysis using three statistical procedures that are commonly used for reducing multiple variables to a smaller number of underlying factors: Principal Components Analysis (PCA), Constrained Correspondence Analysis (CCA), and factor analysis (FA). For both datasets, FA accounts for a small proportion of the variance and provides results that are difficult to interpret in social or linguistic terms. While both PCA and CCA account for a higher proportion of the variance, CCA provides more easily interpretable results. CCA allows us to map the linguistic variables into a multidimensional space in which (groups of) speakers can be viewed as positioning themselves, either through orientation toward or away from ongoing changes and stereotypical features, or through features that define different linguistic systems.
