Zawawi, Abdalkarim and Michel, Marije (2017) Quantifying syntactic priming in oral production : a corpus-based investigation into dyadic interaction of L1-L1 and L2-L2 speakers of English. PhD thesis, Lancaster University.
zawawi2017.pdf - Published Version
Available under License Creative Commons Attribution-NoDerivs.
Download (3MB)
Abstract
It is well established that speakers are inclined to repeat themselves or their interlocutors in native language (L1) and second language (L2) conversations. This phenomenon is largely attributed to priming, whereby exposure to a given language form, i.e. prime, facilitates its processing or that of a related form in a subsequent language production, i.e. target. The present study uses an innovative combination of corpus-linguistic methods and manual analyses to identify syntactic priming of verb-particle constructions (e.g., taking out a paper vs. taking a paper out), the dative construction (e.g., giving Emma a paper vs. giving a paper to Emma) and the caused-motion construction (e.g., putting the money in Emma’s pocket), which shares its constituent structure with the prepositional dative and its semantics with both dative alternation variants. These constructions are studied in task-based free dialogues among native English and L1- German L2 speakers of English. Binary logistic regressions from a generalized linear model (GLM) are employed to disentangle the priming effect from other factors that might be predictors for the target. The analysis of all three constructions controls for interaction between primes and prime-target pair intervening distance, lexical boost and speaker identity. The verb-particle construction results show no evidence for priming as an independent predictor of verb-particle variant reproduction in L1-L1 and L2-L2 conversations. In both language conditions, the reuse of verb-particle constructions can largely be explained by the same set of factors, i.e. the syllable length of the direct object and whether it is new to the discourse. The dative alternation analysis reveals evidence for priming in the L1-L1 and L2-L2 conversations even when controlling for various discourse-related predictors. In the former condition, the difference in length between the target’s themes and recipients and the discourse accessibility of the theme, along with prime-target pair distance and structural similarity, are found to be the best predictors for the target. In the L2-L2 language condition, seven factors are found to best predict the target in the L2 conversations in addition to the identity of the prime, e.g., the pronominality and animacy of the recipient and the concreteness of the theme. The caused-motion analysis shows that it is amenable to priming even though more double object targets follow the caused-motion primes than prepositional dative targets, which are structurally similar to the caused-motion primes. The study finds little support for the relevance of the prime-target distance to the strength of priming across constructions and language conditions. The results also show that the magnitude of priming is unaffected by the identity of the prime-target pair speaker across constructions and language conditions. Finally, (partial) lexical overlap (the so-called lexical boost) is found to encourage the reuse of particle placement primes in the L2-L2 condition (e.g., take out the money – put back the money). For the dative alternation analysis, only some lexical factors (e.g., primetarget main verb lemma match and the semantic class of the target’s main verb) boost the magnitude of dative alternation priming in the L2-L2, but not the L1-L1 language condition. In addition, a shared main verb lemma seems to increase the likelihood of reused caused-motion primes. Taken together, these results indicate that the L1-L1 and L2-L2 reuse of primed constructions is conditioned by the shared constituent structure between prime-target pairs, but also by the mapping of syntactic features to semantic and lexical features of the primed sentence.