Cues, prototypes, and exemplars: towards efficient learning strategies

In this talk I will present some pedagogical recommendations for teaching linguistic structures (argument structure constructions) to second/foreign learners of English following principles of construction grammar and learning theory. Although most linguists originally discarded learning theory, it was brought back into the field by Ellis (2006a, 2006b) who used principles of associative learning in successful experiments with learners. I follow up on this work and more recent work in the same vein (cf. Romain & Divjak 2024 for an example) with a case study of the causative alternation in English (The bottle shattered vs. The bullets shattered the bottles). I will show how a simple learning algorithm can be used to identify the most reliable elements for learning how to use these structures. Running simulations of learning using carefully manually annotated corpus data (approx. 11,000 instances of 29 verbs in the two constructions) showed that one type of cues is the most reliable to make a choice between the two constructions: individual verb senses. I translate these findings into pedagogical materials that help learners better understand the meaning of each construction. Providing learners with semantically coherent sets of examples of each construction should help them identify constructional meaning and generalise this meaning to new instances (Goldberg, Casenhiser & Sethuraman 2004).