An integrated approach to incidental collocation learning: learner, item, and context effects

Collocations, pairs of words that tend to occur together in discourse, are an important aspect of second language (L2) proficiency (Schmitt, 2000), but they stand out as being particularly troublesome for L2 learners (Laufer & Waldman, 2011). Given the limited classroom time available for explicitly teaching collocations, learning through meaning-focused input, such as reading, plays an important role in developing L2 learners' collocational knowledge. The majority of the studies that have investigated incidental collocation learning through reading-(while-listening) are limited in that they only focused on specific linguistic, contextual or individual factors, or failed to consider the impact of relevant confounding variables in L2 vocabulary learning (e.g., Pellicer-Sánchez, 2017; Sonbul & Schmitt, 2013; Szudarski & Carter, 2016; Toomer & Elgort, 2019; Webb & Chang, 2020). Additionally, they produced inconsistent findings, partly due to differences in treatment but also because they did not examine how various factors can contribute independently to learning, or even interact with each other. This dissertation adopted a multifactorial approach to researching collocation learning so as to glean a more comprehensive picture of the nature of collocations, while addressing some research gaps that emerge from existing studies. Three experiments were designed with these goals in mind.

Experiment 1 investigated the effect of reading mode (reading-only vs reading-while-listening), taking into account a range of factors (prior vocabulary knowledge, congruency, compositionality, corpus frequency) to examine the influence of such variables. Unlike previous studies, which have largely focused on fictional short stories as reading materials, this experiment used a single academic text to assess whether the text-related factors could affect learning. The results showed that reading-while-listening did not confer an advantage over reading-only, which contradicts prior studies that used narrative texts as reading materials. However, prior vocabulary knowledge and congruency - characteristics of individual learners and individual collocations, respectively - improved performance.

Experiment 2 examined the effect of repetition by seeding a series of contextually-supportive academic texts with various encounters with collocations. The results indicated that learning gains improved as the number of repetitions increased. Prior vocabulary knowledge again emerged as a reliable predictor of learning. Repetition interacted with prior vocabulary knowledge, suggesting that the facilitative effect of repetition decreases for learners with higher existing vocabulary knowledge.

Experiment 3 explored the role of contextual informativity on collocation learning. Specifically, it compared the effect of three levels of context informativity (low, mid, high) to evaluate whether the degree of contextual support affects learning. As in Experiment 2, a series of short academic texts were used as stimulus materials. The results showed that contextual informativity increased performance to a certain extent, but not beyond it. In particular, it showed that mid informative contexts led to higher gains than low and high informativity contexts. Prior vocabulary knowledge significantly predicted the learning gains.

Overall, the results of this thesis suggest that a multi-factor approach to collocation research can provide a comprehensive account of the effects that unfold in incidental learning conditions. This thesis extends the line of inquiry into the moderator variables that influence collocation learning. Taken as a whole, the results provide useful insights for the learning and teaching of collocations in a more systematic and efficient manner.