Basic Neurosciences · Paper A
Working memory neuroanatomy
Baddeley's working memory model splits short-term memory into separate specialised components rather than treating it as one single store. Remembering a phone number while also picturing a route on a map is possible because these tasks use separate subsystems, e.g. the phonological loop and the visuospatial sketchpad. Interference is greatest when two tasks compete for the same subsystem rather than two different ones.
The central executive directs attention and coordinates the storage subsystems; the phonological loop holds verbal and acoustic information; the visuospatial sketchpad holds visual and spatial information. The episodic buffer integrates information from these subsystems, plus long-term memory, into a single coherent episode. Repeating digits while rehearsing a second spoken list loads the phonological loop twice and interferes strongly, e.g. unlike pairing digit rehearsal with a simple visual-tracking task, which loads two separate subsystems. In practice, a task that overloads one subsystem doesn't necessarily impair the other, since they operate largely independently.
Suppose a junior doctor is clerking a patient while also trying to remember a bleep number just given by a nurse. If a colleague then asks a verbal question, the doctor is likely to lose the bleep number: but if the colleague instead simply gestures toward a whiteboard, the doctor can glance at it with far less disruption. That asymmetry is not a coincidence; it is the direct, predictable output of how working memory is built. Working memory is the brain's short-term workspace: the small amount of information you hold in mind and actively manipulate over a few seconds, for example remembering a phone number just long enough to dial it, or holding a patient's last few sentences in mind while formulating a follow-up question. Understanding its actual architecture, not just the label "short-term memory," is what lets you predict which two tasks will interfere with each other and which two will coexist easily: a favourite exam theme and a genuinely useful clinical skill.
Working memory is different from long-term memory (the store of information kept from minutes to a lifetime) and different from the old idea of short-term memory as one single passive buffer. The most widely taught account in psychiatry exams is Alan Baddeley and Graham Hitch's multi-component model, first proposed in 1974 and still the reference model examiners use today: short-term memory is not one box but several specialised subsystems that can work in parallel or compete with each other.
Before Baddeley and Hitch, psychologists treated short-term memory as one unitary store, and it is still tempting to think of "working memory" as a single that is either good or bad in a given patient. It is not. Damage to one subsystem, for example after a stroke or a injury, can selectively impair one kind of short-term function while sparing the others completely: which is exactly why the model has four separate components rather than one.
The central executive is the model's attentional controller (the part that decides where attention goes and coordinates the other subsystems); it does not store information itself, it directs attention, switches between tasks, and pulls information in from long-term memory when needed. It is the component most consistently linked to function, particularly the dorsolateral , and it is the component most often impaired in (attention-deficit hyperactivity disorder), frontal dementia, and traumatic brain injury.
The phonological loop is the subsystem that holds verbal and acoustic information (sounds and spoken or subvocalised words) for a few seconds. It has two parts: a phonological store, which holds sound-based information passively and decays within roughly two seconds unless refreshed, and an articulatory rehearsal process, essentially silent subvocal repetition that refreshes the store, for example repeating a phone number under your breath until you dial it. The visuospatial sketchpad holds visual and spatial information (images, shapes, and locations in space) rather than sounds or words: imagine mentally tracing a route from the ward to the pharmacy, or picturing where a rash is distributed across a patient's torso. Because it is a genuinely separate store from the phonological loop, a person can rehearse a spoken word list and track a moving visual object at roughly the same time with only modest interference, whereas two verbal tasks together interfere heavily. The episodic buffer was added later, in the year 2000, once experiments showed that verbal and visual information sometimes needs to be bound together into a single, coherent, time-stamped memory (an episode) rather than staying as two separate streams; it links the phonological loop and visuospatial sketchpad to each other and to long-term memory, and is thought to depend heavily on the given its broader role in binding information into episodes.
Working memory is easy to muddle with several neighbouring concepts, and telling them apart is exactly what exam stems test. It is not simple attention or concentration, which is the to sustain focus on a single stream of information rather than to actively hold and manipulate multiple pieces at once. It is not long-term memory, which is consolidated, stored for far longer, and depends far more heavily on the and medial structures for encoding than on the frontal and parietal circuits working memory relies on moment to moment. And within the model itself, the phonological loop is often mistaken for the whole of working memory when it is only one of four components: a scenario describing two simultaneous VERBAL tasks that clash badly is testing recognition of shared-subsystem overload in the phonological loop specifically, while two tasks that coexist easily (one verbal, one visual) are testing recognition that they use two separate subsystems.
There is no single blood test or scan for working-memory function; it is identified through standardised cognitive tasks that isolate each subsystem. Digit span forward (repeating a lengthening string of numbers in the order given) mainly taxes the phonological loop, while digit span backward (repeating the same string in reverse order) additionally loads the central executive, since the numbers must be manipulated, not just repeated: this forward-versus-backward distinction is a genuinely high-yield exam point. Digit span is part of a standard bedside cognitive examination and features in the Addenbrooke's Cognitive Examination and similar screening tools used in dementia and delirium assessment. Visuospatial working memory can be probed with tasks such as the Corsi block-tapping test (touching a sequence of blocks in a given order), which is the visuospatial analogue of digit span.
There is no drug that directly "treats" working memory as an isolated target, but recognising which subsystem is impaired shapes real management decisions across psychiatry. The source body to keep in mind is the APA Dictionary of Psychology: it describes working memory through the Baddeley-style phonological loop and visuospatial sketchpad architecture, which is why this page treats it as a cognitive model rather than a single symptom. In , working-memory impairment is one of the most consistently replicated cognitive deficits and is linked to prefrontal dysfunction; it is part of why cognitive remediation therapy (structured exercises aimed at strengthening cognitive skills) is sometimes offered alongside antipsychotic treatment rather than relying on medication alone. In , central-executive weakness contributes to the classic pattern of losing track of multi-step instructions, and practical management includes breaking instructions into single steps and reducing simultaneous verbal demands, alongside standard pharmacological treatment of the underlying ADHD. In delirium and dementia, digit span testing helps localise which components are affected and tracks change over time, which in turn guides how information is communicated to that patient going forward: favouring the visuospatial sketchpad over the phonological loop, for example, if verbal short-term memory is the more impaired system.
Remember that the central executive itself has no storage at all (every stored item lives in the loop, the sketchpad, or the buffer, while the executive purely allocates attention and control) and that digit span BACKWARD, not forward, is the one that specifically stresses the central executive rather than simple rehearsal.
Working memory is four separable subsystems, not one box, so two tasks interfere badly only when they share a subsystem: and testing which subsystem is impaired (via digit span forward/backward or a visuospatial equivalent) is what actually localises the deficit in practice.