Psycholinguistic and corpus-linguistic evidence for L2 constructions

Abstract

In Construction Grammar, highly frequent syntactic configurations are assumed to be stored as symbolic units in the mental lexicon alongside words. Considering the example of gerund and infinitival complement constructions in English (She tried rocking the baby vs. She tried to rock the baby), this study combines corpus-linguistic and experimental evidence to investigate the question whether these patterns are also stored as constructions by German foreign language learners of English. In a corpus analysis based on 3,343 instances of the two constructions from the British component of the International Corpus of English, a distinctive collexeme analysis was computed to identify the verbs that distinguish best between the two constructions; these verbs were used as experimental stimuli in a sentence completion experiment and a sentence acceptability rating experiment. Two kinds of short-distance priming effects were investigated in the completion data: we checked how often subjects produced an ing-/to-/’other’-construction after having rated an ing- or to-construction (rating-to-production priming), and how often they produced an ing-/to-/’other’-construction when they had produced and ing- or to-construction in the directly preceding completion (production-to-production priming). Furthermore, we considered the proportion of to-completions before a completion in the questionnaire as a measure of a within-subject accumulative priming effect. We found no rating-to-production priming effects in the expected direction, but a weak effect in the opposite direction; short-distance production-to-production priming effects from ing to ing and from ‘other’ and to to to, and, on the whole at least, a suggestive accumulative production-to-production priming effect for both constructions. In the rating task, we found that subjects rate sentences better when the sentential structure is compatible with the main verb’s collexemic distinctiveness.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 153 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.