This chapter abandons the very aggregational, holistic perspective that has dominated the empirical discussion in the three previous chapters with the exception of Section 5.2.3, in which we correlated MDS dimensions with feature frequencies. Instead, we now adopt a more feature-centered and thus nicely complementary view, asking: How can the patterns we have seen so far be traced back to the distributional behavior of each of the morphosyntactic features on which the present study's analysis is based? Section 7.1 sets the scene by scrutinizing six measuring points (Banffshire, Denbighshire, Dumfriesshire, Middlesex, Leicestershire, and Warwickshire) that have struck us as exceptional to varying degrees in the previous chapters, and we will be interested in whether their status can be pinned down on text frequencies of individual morphosyntactic features. Section 7.2 adopts a more encompassing and systematic approach that marshals Principal Component Analysis to uncover feature bundles and to explore the layered nature of geographically conditioned variability in Great Britain. Section 7.3 is a chapter summary.

Revisiting the outliers

This section revisits six FRED measuring points (Banffshire, Denbighshire, Dumfriesshire, Middlesex, Leicestershire, and Warwickshire) which, throughout the previous chapters, did not fit in the picture in various ways and to varying degrees. Recall also that in Chapter 5's series of regression and correlation analyses, excluding these six outliers from consideration substantially boosted the explanatory power of language-external predictors such as linguistic gravity, which amounts to further evidence that there is something odd about these measuring points.