This study explores the effect of working memory (WM) on speech production in Japanese as a second language (L2). WM is a cognitive system that holds and manipulates information during different kinds of cognitive tasks including language learning (Baddeley 1986, 2000; Baddeley and Hitch 1974). It is argued that WM plays an important role in an individual’s L2 speech production skills such as accuracy, complexity, and fluency (Fortkamp, 1999; Mota, 2003; Weisshemier and Mota, 2009, etc.). Since WM capacity (WMC) differs individually, it has been claimed that individuals with higher WMC tend to perform better than those with lower WMC. This is because learners with higher WMC can use their space more efficiently when speaking L2. Although speaking is an integral part of speech production, only a handful of empirical studies in L2 speech production concerning WM are reported (Fortkamp, 1999; Mota, 2003; Weisshemier and Mota, 2009; etc.). As for L2 Japanese, little research paid attention to the effect of WM and L2 speech production until very recently (i.e., Hattori and Nakayama, 2022; Hattori, 2023; etc.). Furthermore, findings are quite limited (e.g., the effect of WM on grammatical accuracy in L2 Japanese (Hattori and Nakayama, 2022). To better understand the mechanism of L2 speech production beyond accuracy, this dissertation reports the results of an experiment that investigated the effect of WMC on Japanese language learners’ speech production, primarily focusing on their ability to produce complex structures.
Thirty learners of Japanese and 28 native Japanese speakers participated in the experiment. The learners’ levels of Japanese varied from intermediate to advanced, and they were enrolled in either third -, fourth -, or fifth-year Japanese language courses at the time of the experiment. Learners took a background questionnaire, vocabulary task, Japanese Language Proficiency Test (Itomitsu and Nakayama, 2005), Japanese Speaking Span Test (JSST, Hattori and Nakayama, 2022), and Elicited Production Test (EPT). The control group (native speakers of Japanese) performed every task except the vocabulary task. Every procedure was conducted online, and the entire procedure took approximately one hour for each of the two separate sessions.
The data analyses found that 1) JSST, the Proficiency Test, accuracy in the EPT, and the complexity of the elicited sentences in the EPT all significantly differed across learners’ proficiency groups (levels of Japanese classes), 2) The control group’s mean scores of JSST, the Proficiency Test, and EPT (both accuracy and complexity) were significantly different from the learners’ group, 3) the L2JSST (JSST by L2 learners) and the Proficiency Test scores were significantly correlated, 4) the L2JSST was significantly correlated with accuracy and complexity in the EPT but not with the vocabulary task, 5) the Proficiency Test scores were significantly correlated with complexity but not with accuracy and vocabulary, 6) when learners were divided into two groups (low and high WM groups), the low group’s scores on the L2JSST, the Proficiency Test, accuracy, and complexity were all significantly different from the higher WM group’s. 7) The high WM group learners were better at producing more complex sentences, 8) among learners’ produced sentences, lexical choice, particle usage, and inflection were the most common errors, though particle usage was not the biggest issue for the higher WM group, and 9) among the complex sentences, subordinate and coordinate sentences were more challenging for learners. These findings suggest that WMC affects how learners can efficiently control and store their information in L2 Japanese, leading to varied learners’ speech performance, especially in complexity.
I believe that understanding the effect of WM in L2 speech production would add clarity to the cognitive aspects of L2 learning. In the long term, this study aims to find some pedagogical implementation in language classrooms, curriculums, and lesson plans, allowing learners to acquire the target language effectively regardless of their WMC.