Choosing Your Relatives: Relative Clauses in New Zealand English

Abstract

This thesis is a corpus-based study of relativiser choice in New Zealand English (though with wider applications to the role of prescriptions in English, and to corpus analysis in general). Relativiser choice in written and spoken New Zealand English (NZE) was studied through quantitative analysis of relative clauses in the Wellington Corpus of Written New Zealand English (WWC), and in samples collected for the Wellington Corpus of Spoken New Zealand English (WSC) and the NZ section of the International Corpus of English (ICE-NZ). This material was supplemented by extracts from selected sections of the Brown, Lancaster-Oslo/Bergen (LOB) and FLOB corpora, enabling direct comparison with British and American varieties of standard written English. 15,000 tokens of relativiser choice in the NZE dataset were coded for linguistic, textual/stylistic, and demographic factors, and this data was investigated by variable-rule analysis (a version of logistic regression) using GoldVarb for the Macintosh. A detailed description of coding decisions is provided, with particular attention to the traditional functional dimension of 'restrictiveness' (which is shown to be more accurately defined in terms of the subjective 'relevance' of the relative clause to the discourse), and the stylistic dimension of 'formality'. A general index of linguistic formality was obtained through principal component analysis of 29 wordform-based counts. This index is shown to have considerable value for the evaluation and interpretation of corpus text categories, as well as allowing a check on the comparability of data from different demographic groups or from different corpora. On the basis of a preliminary frequency survey, attention is focussed on three environments of variation: the choice between who and that in personal subjects; the choice between which and that in impersonal subjects; and the choice between 'null' and overt relativisers in object relativisation. For these subvariables, quantitative analysis is used as the basis for a full explanatory account of linguistic and stylistic variation. With minor exceptions (most notably regarding antecedent animacy), the linguistic influences on the distribution of relativisers in these three environments are found to reduce to processing constraints operating on language production: all else being equal, environments that hinder processing favour more explicit relativisers, and environments that facilitate processing favour less explicit relativisers. This explanatory account can be extended to cover the entire distribution of relativisers by clause function and syntactic role, replacing the traditional description of the English relativiser system. The textual distribution of these variants is consistent with their linguistic distribution: the more explicit wh-relativisers are favoured in more 'formal' texts (where there is a general tendency towards more explicit marking of semantic and syntactic distinctions), and the less explicit marker that is favoured in the least 'formal' texts (especially in spontaneous speech). However, object Ø is favoured in writing rather than in speech; this is attributed to channel-specific differences in language production and processing requirements. Demographic patterns in the distribution of relativiser choice by speaker age, gender, ethnicity, and education were investigated within the sample of casual conversation. These observations are compared with predictions derived from published prescriptions concerning English relativisation. The observed differences in usage by different demographic groups, and differences between national standards, are for the most part extremely small, with two exceptions where published prescription is consistently enforced, and reinforced by popular support: the avoidance of that with personal subjects, and the peculiarly American avoidance of restrictive which. The significance of these results for the effectiveness of prescription, and for the teaching of English, are discussed, and directions of further study are suggested.