Was/were variation with plural subjects in recent British English – towards standard uses?

Talk with Paula Rautionaho.

Abstract

The alternation of was and were (as in they were all muddy vs. they was all muddy) is a linguistic phenomenon that has drawn much attention in sociolinguistic and dialectological studies in recent years. Studies on the variation have observed uses in varying degrees of focus, ranging from surveys examining the variation in specific cities (e.g. Tagliamonte 1998) to those covering nationwide patterns of use (e.g. Hay & Schreier 2004). Different types of regional uses have been identified, such as was-levelling, where was is used in grammatical contexts where were would be normally used (e.g. with plural subjects), as well as were-levelling, where the opposite holds. In some varieties, the spread of one form has been seen to occur in specific grammatical contexts, with e.g. were being preferred in contexts involving negation, a phenomenon which in some dialects is particularly prominent in tag questions.

This study explores recent trends in the was/were variation in British English, focussing on the uses of was/were with plural pronoun subjects in the spoken demographic part of the BNC and the Spoken BNC2014 corpus. Extracting all instances of was and a random sample of 500 instances of were with plural pronoun subjects in both corpora, we annotated the data for intra- linguistic (e.g. verb form, negation, pronoun) and sociolinguistic (age, gender, region, and social class) variables. While a striking decline in the normalized frequencies of was with plural pronoun subjects is undisputable (from 123.5pmw in 1994 to 30.9pmw in 2014), we dig deeper into intra- and extra-linguistic parameters to reveal the changing patterns at hand with generalized linear mixed model tree analysis (GLMM tree; Fokkema et al. 2018). GLMM trees are, in essence, a combination of the recursive partitioning of a tree-based method (such as conditional inference trees and random forests) and a GLMM, accounting for random-effects parameters.

The results indicate that the sociolinguistic parameters override intra-linguistic ones; the major divide is found between upper and lower social classes, the north and the south, and the younger age groups as opposed to the older ones, while pronouns are the only intra-linguistic parameter chosen in the final model. These observed contrasts together with the overall decrease in the use of was with plural pronoun subjects give rise to further considerations on the phenomenon from a sociolinguistic point of view.