ABSTRACT
The comparison of sociolinguistic variables in various speech communities or in socially defined sub-communities typically shows that constraint hierarchies are conserved across the different corpora but that overall rates differ. This is cited as diagnostic of language change, of social stratification, of geographical divergence, or of similar processes. Travis showed that differences in corpus construction led to differences in token distribution across contextual factors, thus exaggerating the differences in overall subject expression in the two Spanish varieties. To see if there is a significant difference between the two corpora, the author repeats the estimation that is, analyzing the two sets of data grouped together in each context. Tabulation is solely for purposes of naming the contexts and comparing the contexts in the two different sub-corpora. It is thus not an additive and is not affected by interactions of factor effects, dependence of factor groups, knockout problems, and the like.
