Wei Yi · [LSLT] Frequency, Probability and Online Processing of Multiword Sequences: An Eye Tracking Study
This study examined the role of phrase frequency and probabilistic information during on-line processing of multi-word sequences in both native and nonnative speakers. Research has revealed the effect of phrase frequency during the processing of multi-word sequences; however, these studies are inadequate for at least three reasons. First, most of them used high frequency target materials solely; hence we do not know whether the claimed effect can apply to multi-word units of lower frequency. Second, due to a lack of manipulation of probability between constituent words when creating control phrases, it is unknown whether probabilistic information contributed to the observed processing advantage for highly frequent multi-word sequences. Third, current findings are inconsistent as to whether the effect of phrase frequency exists in nonnative speakers. In this study, we had a two-by-two factorial design, with phrase frequency and probability of multi-word sequences (measured by mutual information) as the predictors of interest. 80 two-word Chinese adverbial collocations were selected, ranging from 1 to 64 times of occurrence per million words. Native and nonnative participants were asked to read the target collocations in sentential contexts, while their eye movements being recorded. Data collection was done on five measures: first fixation duration, first pass reading time, total reading time, fixation count, and skipping rate. Using linear mixed-effects regression method, a significant interaction between phrase frequency and mutual information was detected in total reading time in native speakers. For nonnative speakers, significant effects of phrase frequency were found in first fixation time, first pass reading time and total reading time. Such findings provided support for usage-based theories, and we attributed the differences in sensitivity to frequency and probability between native and nonnative speakers to the distinct properties of the two metrics and differences in the linguistic experience of the two groups.