Usage-based and constructionist approaches to language see grammar as an inventory of constructions, or complex form-meaning pairings, learned from the input, that is, as a ‘construct-i-con’. This view clearly implies the possibility of variation in individual construct-i-cons. Indeed, when, according to the usage-based view individually variable domain-general cognitive processes are iteratively applied to increasingly diverse linguistic input, we can expect substantial differences between individual versions of the language. To what extent do different individuals infer different regularities from the inputs they receive? It seems reasonable to hypothesise that an individual grammar can be qualitatively different from a communal one since inferring grammar at the communal level involves averaging. An average across lexical constructions (or lexical items) is a schematic construction. What is the level of specificity/schematicity at which individual language users operate? (RQ1) Along which linguistic dimensions do individual languages vary? (RQ2) And finally, to what extent do different individuals rely on different structural, lexical and acoustic cues in language processing? (RQ3) In this project, the relationship between language at the individual level and at the level of the society is modelled using complexity theory and operationalised using a combination of individual and communal corpora drawn from the same discursive situation ensuring comparability. The corpora are examined with the methods of register variation, authorship identification from forensic linguistics and statistical measures employed in collostructional analysis and analysis of constructional change. In addition, RQ3 employs experimental data tapping into chunk boundary perception in online speech comprehension which is collected in the project CLUMP. In this data, individual variation is studied using multiple regression and (G)LMM modelling as well as bootstrapped correlation analyses.

2021