While register variation has been widely studied in terms of frequency distributions of linguistic features and their co-occurrence patterns, for example by means of Multidimensional Analysis (e.g., Biber 1988), the role of the situational context in probabilistic, linguistic choice-making is less clear. Within variationist linguistics, register has often been treated as a “nuisance factor” (Szmrecsanyi 2019: 77), which should be accounted for in the analysis. As a result, evidence is still scarce as to how it influences language-internal constraints that co-determine the choice of one (grammatical) variant other another. Although previous studies suggest that mode of communication (Theijssen et al. 2013) and genre (Grafmiller 2014) influence the effects of language-internal constraints substantially, systematic studies into the effect of register differences on alternation phenomena are lacking.

In this paper, we present results from a case study on the English future marker alternation (This will happen vs This is going to happen), involving interactions between register and nine language-internal constraints such as grammatical person or animacy of the subject, sentence type, clause type, verb type, and proximity of future time. The registers included in this study are defined at the intersection of communication mode and formality: spoken informal (Spoken BNC2014) vs spoken formal (House of Commons debates) vs written informal (British English GloWbE, blogs) vs written formal (The Independent newspaper articles). The central question is how the effect size and direction of language-internal constraints on variation interacts with register as a language-external parameter. We submitted a balanced dataset of 2,600 observations (650 tokens per register, half of which with will and the other half with be going to) to a mixed-effects logistic regression analysis. Results show five interactions with register, indicating that the choice of the future marker is considerably influenced by the situational context. For example, the size of the effect for grammatical person is not only reversed but also much larger in spoken formal, written informal and written formal registers compared to in spoken informal conversations (see Figure 1).

To investigate the language users’ sensitivity to these register-specific probabilistic effects, the corpus model was complemented with a rating task experiment. Participants (n = 114, mean age = 53 years) indicated on a continuous scale by means of a slider bar how natural the variants sound to them given the context (cf. Bresnan & Ford 2010). Analysis shows that participants are sensitive to register-specific effects. The results will be discussed with regard to the importance of methodological diversity in investigating usage-based theories of grammar (Klavan & Divjak 2016) as well as against the backdrop of customary theorizing in variationist sociolinguistics that language-internal are stable in the face of stylistic variation (Labov 2010: 265; but see Guy 2015).