Definition
Working memory is the cognitive system that temporarily holds and manipulates a limited amount of information during active use — the mental workspace for reading comprehension, arithmetic, reasoning, planning, and the effortful self-regulation of behaviour.
The term replaced "short-term memory" in the 1970s when Alan Baddeley and Graham Hitch demonstrated that this system is not a passive store but an active processing workspace. In Focus, Goleman treats working memory as the primary neural substrate of the "cool system" — the deliberate, reflective capacity that can override the automatic "hot system" impulses of the amygdala and reward circuits.
Why it matters
How it works
Baddeley's model
Alan Baddeley's multi-component model (1974, revised 2000) describes working memory as having four components. The central executive is the attentional controller — it coordinates, allocates, and manages the other components and is closely associated with the prefrontal cortex. The phonological loop holds verbal and acoustic information (inner speech, word sequences). The visuospatial sketchpad holds visual and spatial information (mental images, spatial layouts). The episodic buffer integrates information from the other components with long-term memory into coherent episodes.
The central executive is the critical component for self-regulation: it is the mechanism by which deliberate goals and rules are held active in the face of competing automatic responses. When you are on a diet and reach for the biscuit tin and then pause — that pause is the central executive maintaining the diet goal long enough to override the automatic reach.
Capacity and limits
Working memory capacity is routinely estimated at 3 to 5 meaningful chunks (not 7±2 as Miller's popular 1956 figure suggested — that estimate has been substantially revised downward by subsequent research). "Chunks" are the key variable: a chess expert's chunk is a complex board position, not an individual piece; an expert reader's chunk is a phrase, not a letter. Expertise is partly the process of rechunking raw information into higher-order units that fit within the same capacity constraint.
The finite capacity is why divided attention during complex tasks degrades performance: two demanding tasks compete for the same executive resource. Texting while driving, trying to follow a lecture while worrying about a personal problem, or attempting to have a nuanced conversation while emotionally flooded — all reflect the same depletion of working memory by competing demands.
Working memory and emotional regulation
Goleman's argument in Focus turns critically on the relationship between working memory and emotional regulation. The cool system — the capacity to deliberate, plan, and inhibit impulsive responding — is implemented in prefrontal circuits that overlap heavily with working memory circuits. When emotional load is high (fear, anger, anxiety), these circuits are partly recruited for processing the emotional content, leaving less capacity for deliberate cool-system regulation.
This is not a metaphor. Neural imaging studies show that the dorsolateral prefrontal cortex — the primary working-memory region — shows reduced activation during high-stress conditions even for tasks unrelated to the stressor. The amygdala effectively commandeers resources from the prefrontal executive, degrading the very capacity needed to regulate the amygdala's output. The hot-cool feedback loop operates at the level of circuit competition for a shared cognitive resource.
Development and education
Working memory shows a protracted developmental trajectory: capacity increases substantially from age 4 through adolescence and reaches adult levels in the early 20s, paralleling the maturation of the prefrontal cortex. This developmental timeline explains why young children are more impulsive (less cool-system resource available), why adolescents are more risk-prone (adult emotional activation, incomplete prefrontal regulation), and why late adolescence is a sensitive period for establishing self-regulatory habits that will rely on mature working-memory infrastructure.