On-line processing of multi-word sequences in a first and second language : evidence from eye-tracking and ERP

A view that has been gaining popularity is that humans are sensitive to frequency information at different levels, and that this information affects the processing of linguistic material, subsequently shaping our mental representations. Frequency effects have been reported extensively in word processing literature, but only a small number of studies have investigated frequency effects in units larger than a word. The question that the present thesis strives to answer is: Do units above the word level, both fully compositional and less so, exhibit frequency effects? In Study 1, using an eye-tracking paradigm, I investigate the comprehension of idioms used figuratively (at the end of the day – 'eventually'), literally (at the end of the day - 'in the evening'), as well as novel phrases (at the end of the war) in a first and second language. In Study 2, which also uses eye-tracking, native and non-native processing of frequent binomial expressions, such as bride and groom, is compared to their infrequent reversed forms, such as groom and bride. Finally, three ERP experiments, which form Study 3, further investigate on-line processing of frequent binomial expressions versus novel phrases in a first language. The results of the studies point to the following. Frequent phrases are processed faster than novel ones by native speakers. Non-native speakers, on the other hand, appear to have a "lexicon in transition", that is, their processing starts to approximate that of natives only with respect to very high frequency items. Overall, the processing of frequent multi-word sequences in a second language is more sequential than that in a first language (this is particularly the case with idioms). The processing advantage for binomials observed in the ERP study with native speakers also suggests that different neural correlates underlie the processing of familiar phrases when compared to novel ones. On the whole, the findings reported in the thesis suggest that the units that language users attend to are not limited to single words, but extend to multi-word sequences as well.EThOS - Electronic Theses Online ServiceGBUnited Kingdo