The inter-relationship between word learning, native phonology and production practice through cross-situational statistics

Second language (L2) acquisition requires learners to associate the correct meaning with novel words and acquire the non-native phonology. According to the Revised Speech Learning Model (SLM-r), learners may not produce L2 words in a target-like manner due to interference from their native phonological system (Flege & Bohn, 2021). To develop this L2 knowledge, the skills specificity hypothesis suggests that comprehension and production proficiency is guided by practice of the skill itself (Li & DeKeyser, 2017). This original study tests the inter-relationship between L2 word learning, native phonology and practice modality from a single implicit learning task, which trained participants on novel word-referent mappings from cross-situational statistics. We determined whether production practice improved word production, and whether this was beneficial to acquiring the words’ meanings. Eighty native speakers of British English (aged 18–30 years) with no prior exposure to Portuguese learned to associate eight disyllabic pseudowords with eight novel objects via a cross situational word learning task (CSWL, e.g., Ge et al., in press). Four pseudowords contained sounds that exist in both the participants’ native language (English) and in the target language (Portuguese); the other four contained a nasal vowel that only exists in Portuguese. Participants were randomly assigned to either the listen-only (n=40) or listen-and-repeat (n=40) condition. In every trial of the CSWL task, all participants observed two objects, listened to a pseudoword and selected the object to which it referred, without feedback. Whilst participants in the listen and-repeat condition repeated the word aloud after selection, those in the listen-only condition remained silent. After this task, participants completed a comprehension test (word recognition) and a production test (picture naming).
For the CSWL task, both groups rapidly acquired the novel word-object associations. This was evidenced by their above-chance performance in selecting the correct object, as predicted. However, the listen-and-repeat group’s production accuracy declined as participants completed the CSWL task. Contrary to our predictions based on the SLM-r, participants produced non native sounds more accurately than native sounds in the production test. We also predicted that the listen-only group would outperform the listen-and-repeat group in the comprehension test and vice versa in the production test, following the skills specificity hypothesis. However, this
only occurred in the production test where there was a significant effect of production practice on accuracy. Our findings challenge preconceived expectations of the learnability of non-native sounds across different modalities, whilst opening the door to further production-based CSWL research.