Though many scholars agree that an unexpected revival of the mandative subjunctive (MS) brought this once-obsolescent English mood back into productive use in the 20th century (Weekley 1952: 36-37; Barber 1964: 133; Visser 1966: 825; Leech et al 2009: 52), no standardized methodology exists for its quantitative analysis. British English has been identified as the variety most resistant to the resurgence of the MS, with high reported rates of a periphrastic should-construction (Quirk et al 1985: 1013f). The indicative has been described as another characteristically British option (Algeo 2006: 263). Modally harmonic periphrases with auxiliaries other than should can also be found (Huddleston and Pullum 2002: 997-998). Owing to methodological diversity, reports differ on the situation between the MS and the should-variant in BrE (Övergaard 1995: 52; Crawford 2009: 262; Serpollet 2001: 541), and many studies ignore the indicative altogether. The present analysis seeks to ascertain the distribution of the three variants, as well as its diachronic trend, in contemporary BrE. Secondly, it tries to identify which intra- or extralinguistic variables are statistically significant predictors of variant choice. Little work has hitherto been dedicated to the latter question, with prior analyses largely limited to the lexical item governing the that-clause. Answers are sought through rigorous variationist methodology and statistical multivariate analysis. The corpus consists of two 9.5-million-word subcorpora of identical design, representing British news from around 1990 (supplied by the BNC) and post-2015 (compiled by the author). The analysis first empirically identifies the pool of over 50 most frequent triggers, i.e., subjunctive-licensing expressions, that collectively account for >90% of all MS’s. Then all inflectional forms of each trigger are searched for mandative that-clauses in the corpus. Relevant observations are coded for the observed variant and explanatory variable values. The subsequent analysis is performed using traditional bivariate crosstabulations combined with multivariate Poisson and multinomial regression (Agresti 2015: 236-243; 203-207). The dataset consists of 3800 observations. Bivariate cross-classification of diachrony and the dependent variable shows a rise in indicatives from 30% to 40% and a decline of should from 29% to 16%, with the rates of other periphrases and the MS essentially unchanged. This overall diachronic change is highly statistically significant, and the multivariate analysis corroborates it. The final quaternary model was chosen through likelihood-ratio tests between nested models (Agresti 2015: 207), comparisons of predictive power in leave-one-out cross-validation (Stone 1974), comparisons of McFadden’s ϱ2 (McFadden 1979: 306-307), and comparisons of different models’ concordance indexes in binary predictions (Agresti 2007: 143-144). The most significant predictor in the model is trigger group. The triggers seem to form a number of fairly homogeneously behaving clusters, corresponding approximately to shared semanto-syntactic characteristics. Monosemous verbs (demand, order) favor the subjunctive. Extraposed-subject adjectives (it is important that…) favor the indicative. Predicative adjectives (he is determined that…) favor periphrases, while nouns and polysemous verbs (requirement, suggest) display the most even distribution. Other highly significant predictors include matrix tense, sub-clause polarity, sub-clause voice, and that-omission. Surprisingly, syntactic contexts where the subjunctive-indicative distinction is lost do not favor periphrases.
|Status||Publicerad - 2 jun 2018|
|Evenemang||ICAME 39 (International Computer Archive of Modern and Medieval English) - Tampereen yliopisto, Tampere, Finland|
Varaktighet: 30 maj 2018 → 3 jun 2018
|Konferens||ICAME 39 (International Computer Archive of Modern and Medieval English)|
|Period||30/05/2018 → 03/06/2018|
- 6121 Språkvetenskaper
Citera det här
Ruohonen, J. K. (2018). Diachronic and Synchronic Trends in BrE Mandative Constructions: A Multivariate Analysis. 1-60. Artikel presenterad vid ICAME 39 (International Computer Archive of Modern and Medieval English), Tampere, Finland.