This paper reports findings from a written syntactic priming experiment with Korean-speaking L2 learners of English investigating whether the status of the person providing input (primes) as a native or a non-native speaker of English affects the extent to which L2 learners align their own written production (picture descriptions) with those encountered in the input. Additionally, as part of a larger investigation exploring the benefits of forcing L2 learners to predict, or guess, a partner's utterances (Grüter et al., 2021, in press), this study employed the guessing game paradigm to examine whether forced prediction would modulate effects of priming and nativeness of the input. In partial support of the Competence hypothesis (Kim & Chamorro, 2021), which states that listeners are more likely to adapt to speakers they perceive to have native competence, results from this study show that when participants interacted with a non-native partner without having to guess their responses, priming effects were reduced; this reduction was observed only in the post-test phase of the experiment. Priming effects were fully maintained in the post-test among participants interacting with a native partner, regardless of whether they had to guess or not, as well as among participants interacting with a non-native partner in the guessing-game condition. These findings suggest that while alignment is reduced when interacting with a non-native interlocutor, an explicit prediction task can mitigate this reduction, likely through engagement of higher-order processes involving social cognition, explicit awareness and/or deeper processing, an observation that is particularly relevant for the potential implementation of the guessing game paradigm in contexts of foreign language pedagogy, where it is common for language learners to receive input from other non-native speakers.

This paper provides an overview of the design of the acceptability judgment task (AJT), an experimental methodology that has long been a standard tool in theoretical linguistics and theory-driven second language (L2) research. It first argues that the AJT is an effective method for investigating speakers' and learners' grammatical knowledge. It then reviews the factors that should be considered and controlled in designing an AJT, distinguishing between intra-linguistic factors affecting the stimulus sentences (e.g., verb type and lexical familiarity) and extra-linguistic factors related to the experimental procedure (e.g., the judgment scale, scale labels, fillers and distractors, presentation order, and instructions). Finally, in connection with the issue of lexical familiarity, the paper discusses the potential of experimental designs employing nonwords and semi-artificial languages.

The distinction between two classes of intransitive verbs, unaccusatives (e.g., fall, appear) and unergatives (e.g., run, cry), has been argued to exist in L2 interlanguage grammar. This claim is supported by evidence that L2 English learners frequently produce and accept unaccusatives in passive constructions, suggesting an underlying association between NP-movement in unaccusative and passive derivations. Specifically, learners appear to misanalyze NP-movement in unaccusative structures as requiring passive morphosyntax (be + en). While this analysis accounts for a range of learner behaviors, it fails to capture two crucial factors: discourse-pragmatic influences and the role of subject animacy (agentivity). To address these limitations, this study proposes the Theta Misanalysis Hypothesis—the classification of the sole argument in intransitive constructions (as either an internal or an external argument) is determined by the distribution of Proto-Role properties. In L2 grammar, discourse pragmatics and subject animacy contribute to Proto-Role assignments, influencing argument interpretation. To test this hypothesis empirically, an acceptability judgment task was conducted. The experimental results revealed no significant effect of context but showed robust effects of lexical category. The findings are consistent with the Theta Misanalysis Hypothesis and offer new insights into the interaction between lexical semantics and syntax during L2 acquisition.

This study examined how cross-linguistic phonological similarity between Chinese and Japanese affects the recall of Japanese kanji words in Chinese-speaking learners of Japanese. In Experiment 1, Chinese-Japanese cognates differing in phonological similarity (high vs. low) were used as stimuli, while semantic overlap across the two languages was also considered (same-meaning cognates, partially overlapping cognates, and false cognates, defined here as Chinese-Japanese same-script words with different meanings). Participants were instructed to memorize words while subvocally articulating them in Japanese, followed by a written free recall task. A control group of native Japanese speakers was also included to examine whether lexical properties of the stimuli affected recall performance. The results showed that, among learners of Japanese, words with high cross-linguistic phonological similarity were recalled better than words with low phonological similarity, regardless of the degree of semantic overlap. In contrast, no significant effects of phonological similarity or semantic overlap were observed in the native Japanese control group. Experiment 2 further examined the effect of semantic overlap within the low phonological similarity condition in Experiment 1 by including additional stimulus items. The results showed that same-meaning cognates were recalled better than partially overlapping cognates, whereas no significant differences were found between same-meaning cognates and false cognates, or between partially overlapping cognates and false cognates. These findings indicate that, under a memory task involving subvocal reading in Japanese, cross-linguistic phonological similarity condition, the lower recall performance for partially overlapping cognates compared with same-meaning cognates suggests that, for partially overlapping cognates, competition may arise between using L1 semantic representations as recall cues and maintaining the words as distinct L2 semantic representations, making the cues available for retention and recall less stable.

This study investigated how L2 proficiency (higher vs lower group) and the types of cues used during inference affect the accuracy of learners' meaning inference of Japanese onomatopoeia, focusing on unknown words and unknown meanings of known words. Data were collected through individual interviews in which each target word was presented in a single sentence. The results showed that inference accuracy was lower for items with known meanings than for those without, particularly among the lower-proficiency group. For unknown meanings of known words, higher proficiency was associated with higher accuracy, and the higher-proficiency group benefited from using the sentence meaning cue, whereas no cue significantly contributed to accuracy in the lower-proficiency group. Although not statistically significant, both groups showed high accuracy when they inferred unknown meanings based on known meanings of the target word, but the lower-proficiency group used it less frequently. In contrast, for unknown words, no proficiency difference was observed, and both groups relied effectively on sentence meaning and world knowledge. These patterns may be explained by two factors: erroneous inferences caused by being influenced by known meanings, and the possibility that using the known-meaning cue and integrating multiple cues impose greater processing demands on lower-proficiency learners. Furthermore, the use of sound symbolism, a cue specific to onomatopoeia, was limited overall and tended to cluster around certain limited sounds.