Manuel F. Pulido, Ph.D.
Assistant Professor, Penn State
Department of Spanish, Italian, and Portuguese
"From Collocations to Constructions: Exploring Generalization During L1 and L2 Processing"
Friday, October 27 9:00–10:30 a.m. EDT
Foster Auditorium, 102 Paterno Library
Speakers constantly innovate in their use of language. However, linguistic innovation is not generated in a vacuum,
but is built through incremental deviations from conventional usage. Previous research indicates that analogy is a productive mechanism for language innovations, with frequent and prototypical exemplars serving as the basis for generalization. Such similarity-based generalization is not only an important part of L1 speakers' language use, but it may also be an important mechanism for L2 users who are still increasing their repertoire in their non-native language. But little is still known about how analogical innovations are processed by L1 and L2 speakers. In this talk, Manuel F. Pulido will present data from three experiments that examined how L1 and L2 speakers process novel multiword units that vary in their degree of relatedness to previously known verb-noun collocations. In Experiment 1, he will present data from acceptability judgements and ERPs in L1 speakers, associated with the processing of conventional collocations and novel multiword units. Two additional experiments examined L2
speakers' processing of novel multiword units that do not have a direct equivalent in their L1. Experiment 2 follows the same methodology as the first experiment, to examine the processing of novel multiword units that are related to familiar L2 collocations. Importantly, these L2 collocations are lexically incongruent with their L1 equivalent (i.e., they have no literal translation), so that generalization must also rely on knowledge of L2 lexical units. Finally, Experiment 3 examines L2 speakers' levels of reliance on frequent prototypes to learn and generalize semantic categories for which no equivalent category exists in their L1. Altogether, the data suggest that successful L2 generalization is possible even shortly after learning, and is partly relies on similarity to prototypes.