Characterizing language knowledge as a complex adaptive system helps us to have a clearer understanding of its properties and development. Of particular importance is to take into consideration the fact that the components of the system interact with each other, so that the entire system would be constantly updated. This seems to imply that learning a particular feature of a second language may have consequences for other aspects of the knowledge system. The possibility has been discussed in several studies, though not explicitly in terms of complexity. Several of them have suggested the roles of indirect triggers for system development and the transfer of processing strategies. Thus, the notion of ‘evidence’ in the discussion of language learnability needs to be expanded to take them fully into account. This paper develops the arguments, with reference to the sense network of polysemous words as a manifestation of complex system, and suggests future research directions.

The purpose of this paper is three-fold: (1) to clarify Vygotsky’s theory (mediational development, personality development, and affective development), (2) to investigate Vygotsky’s view of foreign language learning by reviewing his two works, and (3) to propose an L2 concepts developmental model which emphasizes the role of ‘uptake.’ State-of-the-art Vygotsky studies postulate that there are three robust pillars established by Vygotsky: mediational development of human higher mental functions, personality development by new formations, and Spinozistic monistic emotion development. Vygotsky’s view of foreign language learning metaphorically states that the development of the foreign language moves from above to below. ‘Above’ means that foreign language learning begins with conscious awareness and volitional and voluntary mastery of language. His keen insight reveals that both volition and voluntariness develop through the gate of scientific concepts. Lastly but not least, the article proposes an L2 concept formation model of second language appropriation. Although the model replicates a computational model of second language acquisition, it emphasizes the development of not L2 knowledge but L2 concepts and marks the role of ‘uptake’ which is associated with ‘intake.’ Vygotsky has a distant view of personality development which originates in the movement from concept development to self-consciousness arousal.

There has been a growing trend toward conducting second-language acquisition (SLA) research in the framework of usage-based models of language, which posit that grammar emerges through language use. Previous studies that claim to investigate usage-based SLA, however, vary in their research questions and their methods. Because of this diversity, we tend to lose sight of what constitutes usage-based SLA research. To clarify this fundamental issue, the first part of this paper discusses four key underpinnings of usage-based models of language: frequency, construal of the situation, shared meanings in social interactions, and conventionality. Next, some representative SLA studies from usage-based perspectives are reviewed in terms of the first three underpinnings to show that each of the studies focuses on one or two of the multiple dimensions of usage-based SLA. Based on this review, the paper confirms that this line of research is still in its exploratory stage. Furthermore, it proposes that research should be conducted from the perspective of multilingualism to investigate how new construals emerge and are conventionalized through the contact of a second language with the existing language resources including those in the first language. It is concluded that transdisciplinary approaches are needed to tackle the multifaceted nature of usage-based SLA.

Over the past half century, the theory of generative grammar has been one of the main approaches used to analyze language acquisition and has contributed to several major advancements in the field. This paper reviews some of the previous research on Wh-questions from the perspective of the “third factor” in language design (Chomsky, 2005), in order to point out some of the theoretical issues and to explore the direction of future research in second language acquisition.

In this study we explore the acquisition of rendaku voicing by learners of Japanese whose first language is Italian, drawing upon acoustic measurements of the duration of alveolar fricatives. It is well-attested that, in Italian, the alveolar fricative /s/ turns into its voiced counterpart /z/ when followed by a sonorant at a word boundary. The present study, assuming the speech model proposed by Levelt (1989), examines recordings of the learners’ utterances to see how the Italian learners’ word-internal intervocalic-s voicing affects their formation and processing of compounds in Japanese from the viewpoint of rendaku. Specifically, we focus on items containing an alveolar fricative in a word-internal position and the acoustic duration of the phonetically realized target segments. The experiment, involving seven Italian learners of Japanese (L2 participants) and six native speakers of Japanese (L1 participants), elicited sets of compounds consisting of two nouns, of which the second element began with an alveolar fricative. The two participant groups showed no significant differences in average duration and the voiced segments were shorter than the voiceless segments that underwent rendaku. The acoustic realizations of the voiced alveolar fricative /z/ in the data from the L2 participants showed a tendency similar to that of the L1 Japanese participants, but the L2 participants displayed greater individual variability than the L1 participants. Given these results, we surmise that the resources available for the learner to use the Lexicon to process information on the path from the Formulator to the Articulator, as well as to activate the muscles of the speech organs, vary and develop in accordance with level of language acquisition.

It has been shown that language users quickly adapt to statistical properties of language input available in a local environment and that such learning by adaptation can be explained by the error-based learning mechanism that is based on prediction error, i.e., the difference between predicted input and received input. Many previous studies on L1 adult speakers and children demonstrated evidence for learning based on prediction errors. It is, however, not yet known whether the same learning mechanism operates for second language learning. The current study tested this with Japanese learners of English. Using the eye-tracking reading technique, we examined an effect of structural ambiguity, which induces prediction error, on learning of the relative clause structure. The results from both comprehension questions and eye-tracking reading time measures showed that the ambiguity cost was significantly reduced following the experience of temporarily ambiguous reduced relative clause sentences compared to following unambiguous counterparts, demonstrating that prediction error plays a crucial role for second language learning.

This study investigates L2 acquisition of the definiteness marking function of the Japanese demonstrative, sono. Japanese sono optionally marks nominal definiteness in direct and indirect (bridging) anaphoric contexts, but it is incompatible with non-anaphoric contexts. Knowledge of sono is compared between two L2 groups: native speakers of Korean and English. Korean has a demonstrative, ku, with properties corresponding to sono. English, however, does not have such a lexical item—its definite article the and demonstrative that functionally overlap with sono and ku only partially. We examine whether this crosslinguistic variation leads to any difference between L1-Korean and L1-English speakers' acquisition of Japanese sono in direct anaphoric, bridging, and non-anaphoric contexts. The results of an acceptability judgement task confirmed that native Japanese speakers are sensitive to the incompatibility of sono with non-anaphoric contexts, and that they prefer explicit definiteness marking by sono despite its optional nature. The L2 groups demonstrated a degree of knowledge of the properties of sono in all contexts, regardless of L1. Crucially, the L2 speakers showed robust evidence for the incompatibility of sono with non-anaphoric contexts, which suggests that they can acquire this core property of sono. However, native-level sensitivity did not arise in the other two contexts. We propose that this contextual difference may be due to the quality of the evidence for the form-meaning mapping. Specifically, the mapping is less transparent in direct and bridging anaphoric contexts, where sono is optional, than in non-anaphoric contexts, where sono is consistently infelicitous.