Working memory (WM) is the temporary storage and processing of incoming information needed to execute complex cognitive actions (Baddeley, 2007), such as processing a foreign language (L2). Since the 1980s, the role of WM in fi rst language (L1) processing has been investigated extensively, but only more recently has its role in L2 learning been examined keeping in mind that adult second language acquisition (SLA) requires additional computation and activation, which in turn imposes extra load on the already limited WM resources. In particular, two memory components have been considered to play a prominent role in SLA: phonological short-term memory and WM's central executive, which is responsible for information management during processing. Our goal is to describe the theoretical underpinnings of WM's central executive, discuss tests to measure WM's central executive for L2 research, and provide a critical review of studies examining its role in L2 processing and learning. This issue is crucial for explaining individual differences in L2 learning, creating language-aptitude tests that predict L2 abilities, and implementing pedagogical practices that enhance verbal WM through trainability and compensate for WM capacity (WMC) shortcomings.
Working Memory ModelsOf the numerous WM models that have emerged since the term was coined in the 1960s, the following are particularly popular within SLA circles: Just and Carpenter's domainspecifi c single-resource model, Baddeley's and Waters and Caplan's domain-specifi c multipleresource models, and the domain-free connectionist models. Domain-specifi c models claim that limitations in WMC constrain L2 learning, whereas domain-free models rely on limitations imposed by a domain-general attentional construct. In turn, single-resource models emphasize a trade-off between processing and storage, while multiple-resource models assume that processing and storage are independent from each other.Just and Carpenter's (1992) capacity theory posits that processing and storage depend on and compete for a shared pool of resources with a limited capacity that varies among individuals. Tasks that deplete a person's WMC can result in less storage and slower processing. This model differs from Baddeley's (2007) model that depicts WM as a multipleresource system formed by a central executive and three slave systems with independent limited capacities: two short-term storage systems (phonological loop, visuospatial sketchpad) dedicated to a content domain (verbal and visuospatial respectively) and an episodic buffer that relays information between long-term memory and the phonological loop and visuospatial sketchpad. The central executive coordinates the fl ow of information between its slave systems, binds information from a number of sources into coherent episodes, determines what information is selected and inhibited, and decides when to shift between tasks or retrieval strategies. This polifacetic nature has led some scholars to propose separate resources within the central executive. Fo...