Summary and Conclusions
The VARBRUL analyses are summarized in Figure 9.1. The five changes are listed on the left, and their time courses run horizontally across the page. At each stage, the social factors studied are presented in order of their relative importance. This order directly reflects the range calculated by the VARBRUL program for each factor group: the wider the spread of the factors within a group, the wider its range and the more significant the variable. In each case we also single out the factor that shows the highest positive correlation with the process of change. When a factor does not have a notable impact (a weighting over 0.550) at a particular stage of a change, it is indicated in italics in square brackets.
We are now in a position to say something about the relative robustness of the three variables and their alignment in promoting the five changes. First, the weakest variable proves to be register. It tops the list only once: when you is an incipient change. But our data also support earlier studies such as Schendl (1997), suggesting that the register affiliation of a change in progress can change in the course of time. This happens with the loss of the -n- allo- morphs of possessive determiners. In that change register makes a significant addition to three out of the four stages of the process.
Overall, the changes however prove to be either region- or gender-driven throughout, or mixed. The third-person -s and the determiners my and thy spread from the North to London and, when nearing completion, are both promoted by the Court. They clearly display the snowball effect proposed by Ogura and Wang (1996): the regions that are later to adopt a change do it more forcefully than those that are affected by it earlier. The phase of the change also matters: by mid-range, at the latest, the changes are promoted by the capital region.11
Gender is the other variable that systematically correlates with the changes examined. With the exception of the relative pronoun which, all of them are promoted by women. The object of the gerund is distinctly gender- driven. The VARBRUL analysis confirms our findings in Chapter 6 that, unlike in regional diffusion, no snowball effect can be detected with gender: a change in progress does not alter its gender affiliation when it is nearing completion. We may therefore argue that, among the literate ranks, women are more likely than men to favour a process of supralocalization in the early modern period. The results also suggest, however, that for a change to be first and foremost promoted by gender at any one stage of its progress it must originate in the capital region: as we have seen, region systematically outstrips gender in precisely those two changes that have their origins in the North.
On the basis of these findings we can advance the following generalization on the range of variation displayed by the supralocalizing features in the personal correspondence data:
Figure 9.1. Summary of VARBRUL analyses of the five linguistic changes. (Region: L = London, C = Court, E = East Anglia, N = North; Gender: w = female, m = male; Register: ff = family and friends, nf = non-family) social variation (regional variation > gender variation) > register variation according to addressee
The implicational scale supports Bell's Style Axiom and suggests that the degree of register variation does not exceed the degree of social variation. It also predicts that within social variation the range of regional variation is usually wider than the range of gender variation. These hypotheses will no doubt be refined in the future when more data are analysed and subtler distinctions made within the categories. As for now, we hope the results will inspire more confidence in the historical sociolinguistic enterprise. Although the data may not be ideal, they are ample enough to produce consistent generalizations.