Processing Collocations in First (L1) and Second Languages (L2): Effects of Collocation- and Word-level Variables on Speakers Varying in Proficiency and Dominance

Abstract

Collocations are words that have a tendency to co-occur within a few words’ spans, e.g., “drink coffee” and “dark chocolate” in English. Growing empirical evidence suggests that both native (L1) speakers and advanced second language (L2) learners process two-word collocations faster than unconnected word pairs, and that speakers and learners are sensitive to the frequency distributions of linguistic units beyond individual words. Here we investigate this processing advantage of collocations in L1 and L2 and unravel the factors underlying collocation processing in native speakers (of English, Exp 1; of Chinese, Exp 2) and non-native speakers (of L2 English, Exp 3). In a series of double Lexical Decision Tasks (LDT), participants were presented with pairs of letter- or character-strings and were asked to judge if both strings in a pair were correctly-spelled words in English (Exp 1 & 3) or in Chinese (Exp 2). The word pairs of all three experiments fall into four conditions: 1. collocations in English but not Chinese (e.g., honest mistake, 诚实错误), 2. collocations in Chinese (if translated verbatim) but not English (e.g., dead road, 死路), 3. collocations in both languages (e.g., deep sleep, 深睡眠), and 4. baseline controls in which the two words were unconnected in both languages (e.g., bright hand, 明亮手). Exp 1 and 2 demonstrated faster and more accurate processing of collocations in native speakers of both languages. Most interestingly, Exp 3 showed facilitated processing of English collocations for Chinese-English bilingual speakers, and the effect size increased as English experience increased. General linear regression models including distribution variables of different grain sizes in both languages revealed that highly advanced L2 speakers process collocations in a similar way to L1 speakers, with their performance being more impacted by collocation-level factors and less impacted by word-level factors, whereas the reverse pattern is found for beginner L2 speakers. This suggests that as language dominance and proficiency grows, learners become increasingly sensitive to the statistical associations relating to larger chunks – from single words to collocations and potentially other multiword units.