Tracking the Impact of Family Size and Frequency on L1 and L2 Collocation Processing Over Time

Collocations, frequently co-occurring words, are fundamental to language use, yet it remains unclear whether they are processed holistically or sequentially through their constituents over the timecourse of comprehension. Therefore, using eye-tracking, we investigated how frequency and family size (i.e., the number of collocations sharing a constituent; fatal fatal accident/error/inquiry/disease) influence collocation processing in L1 and L2 speakers. We observed that higher-frequency constituents were read faster than low-frequency constituents in earlier eye-tracking measures, while high-frequency collocations received fewer fixations than low-frequency collocations in later eye-tracking measures. Additionally, high family sizes facilitated processing frequent collocations in L1 speakers, whereas the opposite pattern was observed in L2 speakers. These findings demonstrate that L1 and L2 speakers are sensitive to frequency at both the word and phrase level. However, L2 speakers’ greater reliance on constituent-level cues suggests weaker holistic representations. The study supports the usage-based and statistical learning models.