Understanding the working memory model

Short-term memory, as described in the multi-store model, was conceived as a passive temporary store — a waiting room that holds information briefly before it either decays or is rehearsed into long-term memory. Working memory, as Baddeley and Hitch (1974) proposed, is an active system that not only stores information but manipulates it during ongoing cognitive tasks. The key differences: working memory is multicomponent (central executive + slave systems), whereas STM was unitary; working memory supports active processing (reading, reasoning, arithmetic), not just passive storage; and working memory components are more specialised — the phonological loop handles verbal material, the visuospatial sketchpad handles visual-spatial information, and the episodic buffer integrates across systems. Working memory capacity — particularly central executive function — is a stronger predictor of cognitive ability than STM capacity alone.

There is no single capacity number for working memory. Capacity depends on the component, the task, the material, and how information is grouped into meaningful units. Miller (1956) observed that many immediate-memory tasks show a capacity of roughly 7 ± 2 meaningful units and introduced the concept of chunking. Cowan (2001) reanalysed the evidence and argued that when chunking and long-term memory support are controlled, the number of independently maintained items is closer to four. The phonological loop is further constrained by rehearsal rate — it holds roughly what can be articulated before decay. The episodic buffer integrates multimodal information, with Baddeley describing its capacity as approximately four chunks. Working memory performance can be substantially improved by prior knowledge and chunking, which is why an expert chess player can hold far more board positions in mind than a novice. Individual differences in working memory span are among the strongest cognitive predictors of reading comprehension, fluid intelligence, and academic outcomes.

Working memory capacity — particularly the central executive — shows strong correlations with fluid intelligence (the ability to reason with novel problems) and with academic outcomes in reading and mathematics. Daneman and Carpenter (1980) showed that reading span (a dual-task measure requiring processing and storage simultaneously) predicted reading comprehension better than simple word span. Individual differences in working-memory performance are associated with reading comprehension, mathematical reasoning, fluid intelligence, and academic performance. These relationships are correlational, however, and working memory is only one factor contributing to educational outcomes. The relationship is thought to reflect the central executive's role in focusing attention, suppressing distractors, and binding information from different sources.

Yes. Baddeley's multicomponent model is the most widely taught account, but it is not the only one. Cowan's embedded-processes model, for example, emphasises activated long-term memory and a limited focus of attention (roughly four items) rather than several dedicated temporary storage systems. In Cowan's framework, working memory is not a separate module but the currently activated portion of long-term memory, with a spotlight of attention that can hold only a few items simultaneously. Other frameworks — such as those emphasising interference-based forgetting rather than trace decay, or resource-competition models — have also shaped the field. The existence of competing models is itself an important lesson: working memory is not one settled, agreed-upon mechanism, and researchers continue to debate the architecture, capacity limits, and mechanisms involved.

The evidence is mixed and the research has been contentious. Commercial working memory training programmes (such as Cogmed) do improve performance on the specific trained tasks. Transfer to untrained working memory tasks is sometimes observed (near transfer) but is less consistent. Evidence for broader transfer to untrained cognitive abilities, academic performance, or everyday functioning (far transfer) is much less consistent: a systematic review by Melby-Lervåg and Hulme (2013) found near-zero far-transfer effects that fade over time. This distinction between near transfer and far transfer is important when evaluating claims about cognitive training. The current consensus is that intensive working memory training produces practice-specific improvements, but does not increase the underlying capacity of the working memory system in a way that generalises broadly. Strategy-based interventions — teaching people how to use working memory more efficiently through chunking, rehearsal, and organisation — may be more practically useful than capacity-focused training.