Working Memory: Definition, Models, Capacity, and Examples

4 days ago

19 min read

Author: Angela Bogdanova · Published: October 12, 2026 · Editorial Policy

Research cluster: Memory & Forgetting · Written by Artificial Sapiens

Working memory is the limited-capacity cognitive system that keeps information accessible and makes it usable while a person is thinking, understanding language, following instructions, or solving a problem. It links what is currently relevant to the operations being performed on it. When you calculate a tip mentally, compare two arguments, or retain the beginning of a sentence while reading its end, working memory helps coordinate the information needed for the next step. This functional definition is consistent with modern reviews by Baddeley (2012) and Cowan (2017).

The word “working” matters: the system supports ongoing cognitive activity. Information can be maintained, selected, refreshed, rearranged, bound to other information, or replaced when a goal changes. The familiar explanation that working memory is a “mental workspace” is useful as a metaphor, provided we remember that contemporary theories describe interacting cognitive processes and brain networks rather than a literal box in the head.

Working memory at a glance

Definition: temporary availability of goal-relevant information for ongoing cognitive processing. Core operations: maintaining, prioritizing, updating, and manipulating representations. Common examples: mental arithmetic, understanding complex sentences, remembering the next turn in directions, and comparing choices. Typical limitation: only a small number of independent units can remain readily available under demanding conditions; the numerical limit depends on task design and how units are defined. Important distinction: working memory emphasizes active use and attentional control, whereas short-term memory often emphasizes brief retention. Scientific status: well-established as an important cognitive construct, with active debate about architecture, capacity metrics, and neural mechanisms.

What does working memory do?

At any moment, cognitive tasks require a changing set of representations. During a conversation, a speaker's earlier clause may need to remain accessible so the listener can interpret a later pronoun. In mental subtraction, intermediate quantities must be retained while additional operations proceed. Working memory coordinates these demands, often under interference from other information. In a classic description, Baddeley (1992) emphasized the joint requirements of storage and manipulation in comprehension, learning, and reasoning.

Maintenance

Maintenance means keeping information available after the stimulus is no longer visible or audible. Imagine seeing a six-digit entry code and holding it briefly while moving to another screen. Maintenance may involve verbal rehearsal, attention to a visual representation, or other task-specific strategies. A delay of a few seconds does not define working memory by itself: the critical question is how the representation supports the current task and withstands distraction.

Manipulation and updating

Manipulation changes the arrangement or content of what is maintained. Repeating 7–2–9 in the original order mainly tests brief retention; reciting the digits backward adds transformation demands. Updating replaces no-longer-relevant information, as when someone revises a shopping total after learning that one item has been removed. These operations overlap but are not interchangeable; different laboratory tasks emphasize different aspects of performance.

Control of attention

Working memory also involves keeping the relevant goal active, choosing what deserves priority, and resisting interference. When you read an argument in a noisy room, a lapse may reflect distraction, competition from irrelevant thoughts, or incomplete encoding rather than an absolute inability to store another item. Engle's 2018 reconsideration emphasizes the importance of attention control while revising earlier assumptions about how it relates to working memory and fluid intelligence.

Working memory versus short-term and long-term memory

Short-term memory usually describes the brief retention of information, often assessed with immediate recall or recognition. Working memory describes the cognitive arrangements that keep information usable during other processing. Researchers differ in whether they consider short-term storage a component of working memory, a partially separate function, or a description of particular tasks. This terminological variation is documented in Cowan's analysis of working-memory definitions.

Long-term memory supports the retention and recovery of knowledge, skills, and experiences over extended periods. The systems continuously interact. Knowing multiplication facts from long-term memory reduces the number of intermediate steps needed for mental arithmetic. Familiar language makes a sentence easier to retain than a sequence of unfamiliar syllables. Working memory can draw upon activated knowledge rather than generate every representation independently, a point emphasized in contemporary model comparisons.

Consider the sentence “The scientist who reviewed the results changed her conclusion.” Long-term memory contributes word meanings and grammar; working memory supports the temporary relationships needed to interpret who changed the conclusion. The example also shows why a strict division between separate storage boxes is insufficient for explaining real reading.

The Baddeley and Hitch multicomponent model

The influential framework introduced by Alan Baddeley and Graham Hitch in 1974 treated working memory as a coordinated system of specialized functions rather than one undifferentiated short-term store. Their initial model described a phonological loop, a visuospatial sketchpad, and a central executive. Baddeley's 2012 review explains both the framework's explanatory reach and its unresolved questions. It remains a major organizing theory, rather than a final anatomical inventory of distinct brain modules.

Phonological loop

The phonological loop concerns temporary speech-based information and the processes that support it. Silently repeating a telephone number, holding words in order, or rehearsing a short verbal instruction are everyday examples. Classic laboratory findings include effects of word length and articulatory suppression, although the mechanisms and generality of these effects have been refined over decades of research. A person may use this verbal support when first learning unfamiliar names or vocabulary. Baddeley (1992) discusses its role in language-related tasks.

Visuospatial sketchpad

The visuospatial sketchpad refers to short-lived visual and spatial representations and their use in ongoing cognition. Holding a map layout in mind, mentally rotating a shape, or planning how furniture fits in a room can involve these processes. Visual identity and spatial position can place partly different demands on performance; models that discuss a sketchpad do not imply that every visual or spatial operation uses precisely the same mechanism.

Central executive

The central executive is the model's control component: coordinating activities, directing attention, switching goals, and regulating priorities. For instance, when estimating change while someone asks a second question, executive demands rise as the person must manage competing information. The term names a set of hypothesized control functions; it should not be interpreted as a single tiny manager located in one brain region.

Episodic buffer

In 2000, Baddeley proposed an episodic buffer to explain how temporary verbal, visual, and long-term representations can be combined into integrated episodes. Imagine hearing a description of a room while viewing its plan and bringing relevant knowledge to bear: an integrated representation has to connect those sources. The original episodic-buffer proposal addressed phenomena that the three-component framework did not fully explain. Questions about exactly how binding occurs and how its limits should be modeled remain open.

Alternative models: attention, activation, and interference

Cowan's embedded-processes account

Nelson Cowan's approach treats working memory in relation to activated long-term representations and a more restricted focus of attention. Much relevant knowledge can become activated, while only a few units are selected for especially accessible processing. This perspective explains why expertise, knowledge, attention, and grouping change what a person appears able to hold. The theoretical account and its definitional implications are described in Cowan (2017).

Focus-of-attention accounts

Klaus Oberauer and Laura Hein distinguish a broader set of currently accessible representations from a narrower focus that selects one item for immediate processing in some tasks. Their 2012 synthesis discusses evidence for a broad focus of roughly four chunks and a narrow focus of one. These numbers describe theoretical and empirical regularities under specific conditions, rather than universal mental compartments with fixed slots.

Resource, precision, and interference models

Other accounts describe limitations in the precision of stored representations, competition among similar items, time available to refresh them, and interference between overlapping features or associations. A person may retain the approximate colors of several objects yet remember their exact shades less precisely as load grows. In another task, confusion about which word appeared in which position may drive errors. The large collaborative benchmarks review by Oberauer and colleagues (2018) documents the behavioral findings that competing models must explain. No single mechanism has settled every task and domain.

The theories share an important goal: explaining how a constrained cognitive system supports flexible behavior. Their differences matter in experiments. When one model predicts loss of entire items and another predicts declining precision, the exact pattern of errors can distinguish them. Theories must also account for the role of prior knowledge, distraction, order, associations, and the possibility of recovering information from longer-term memory.

How much information can working memory hold?

There is no context-free working-memory number. Capacity depends on the material, task, timing, familiarity, attention, strategies, and criterion for success. Estimates around three to five meaningful units under controlled conditions are common, but treating four items as a guarantee would misrepresent the science. A “chunk” is a meaningful or psychologically coherent unit, not necessarily a single digit, letter, or word. Cowan's classic 2001 review argued for a smaller central capacity limit than the widely repeated seven-item rule.

Why “seven plus or minus two” is often misleading

George Miller's 1956 discussion of the “magical number seven” concerned several information-processing phenomena, including immediate memory spans and the effects of recoding. The familiar slogan became a simplified capacity claim. Modern working-memory research examines conditions that minimize rehearsal and grouping and distinguishes items from chunks, binding accuracy, and precision. Some modern estimates favor approximately four central chunks, while empirical work also reports different estimates in different paradigms. A 2024 modeling analysis explains why the quantity being counted changes the answer.

Chunking and knowledge

Suppose a reader familiar with American dates sees 1–7–7–6. The sequence may form a meaningful unit related to a historical year, whereas the same digits may remain separate for a reader without that knowledge. Grouping can improve the amount of information recalled without changing an underlying universal biological slot count. Experiments on verbal material find that a capacity-limited component often needs to be modeled alongside activated long-term representations. In one model comparison, the best-fitting account retained about three chunks on average for those particular recognition tasks.

Item limits versus resolution

A person may remember that a blue square and a red circle appeared but confuse their positions. Alternatively, their approximate positions may be remembered with reduced precision. The score differs depending on whether the experiment measures exact recall, recognition, feature-location bindings, or continuous error. Consequently, the statement “my working memory is five things” has little meaning unless the test, stimuli, scoring method, and relevant assumptions are specified. The 2018 model-benchmarks review details why broad theories must account for these distinct forms of performance.

Capacity is related to performance, not identical to intelligence

Working-memory capacity correlates with reasoning and other cognitive measures, but the constructs are not interchangeable. Tasks also recruit attention, strategy, knowledge, processing speed, and language. In a study of 270 participants, Wilhelm, Hildebrandt, and Oberauer (2013) identified a general working-memory factor across several task families and strong relations with measures of secondary memory and fluid intelligence. A correlation does not imply that the working-memory score alone measures a person's overall intellectual ability or determines their potential.

Why working memory fails: interference, divided attention, and changing goals

The moment a new task begins, irrelevant information can compete with what is being maintained. A distracting notification may interrupt a calculation; hearing a similar number can confuse the one being remembered; changing the goal can make earlier content less accessible. Some loss reflects interference among representations, some reflects limitations in prioritization or refreshing, and some comes from weak initial encoding. The explanation depends on the experiment and should not be reduced to information simply “falling out” of a fixed container.

Working-memory performance is often better when material can be organized and when the person understands the task. Dividing attention, trying to process multiple demanding streams simultaneously, or using unfamiliar symbols can make the same nominal memory load much harder. This is one reason that everyday performance can fluctuate substantially without implying a stable decline in cognitive ability.

What brain systems support working memory?

Working memory depends on distributed neural systems. Prefrontal regions contribute importantly to prioritization, goals, and behavioral control, while perceptual and association regions can carry content-specific representations. Interactions with parietal areas, basal-ganglia circuits, and neuromodulatory signals shape how relevant information is selected and used. D'Esposito and Postle (2015) review evidence for attention to internal representations and the role of prefrontal control, replacing the simplistic idea that memory resides in one isolated frontal-lobe storage center.

Persistent and dynamic neural activity

One influential hypothesis holds that maintained information is supported by persistent firing in neural populations during short delays. Other studies show dynamic population codes, changing activity patterns, and periods in which currently inaccessible representations may remain recoverable despite reduced measured activity. The review by Stokes and colleagues on dynamic coding presents the case for activity-silent and dynamic mechanisms, while later synthesis of multiple neuronal mechanisms emphasizes that different forms of maintenance may coexist.

Activity-silent interpretations remain an active area of debate. When information survives a period without an obvious neural signal, it may reflect temporary network changes, limited measurement sensitivity, or recovery from episodic long-term memory. A direct experimental test by Oberauer and colleagues considered this alternative. It is therefore more accurate to say that several candidate neural mechanisms are being investigated than that researchers have discovered one definitive code for every working-memory state.

How psychologists measure working memory

Digit span and backward digit span

In forward digit span, participants repeat a sequence as presented. This assesses immediate verbal retention and benefits from rehearsal and familiarity. In backward digit span, they reproduce it in reverse, adding sequencing and control demands. Both tasks can contribute information about cognition, but no single span score captures the entirety of working memory. Scores also depend on language, administration, age norms, and strategy.

Complex span tasks

Operation span asks people to process information, such as verifying arithmetic statements, while keeping separate items available for later recall. Reading span combines sentence processing with the retention of words. The classic Daneman and Carpenter reading-span study linked storage-plus-processing demands to reading comprehension. These tasks attempt to measure how information remains available while attention is occupied, although their scores also reflect multiple component abilities.

N-back and updating tasks

In an n-back task, a participant judges whether the current stimulus matches one presented n positions earlier. This requires ongoing monitoring and updating, but also involves recognition, familiarity, response decisions, and interference control. Success on a 2-back task is not directly equivalent to a two-item working-memory capacity. A score cannot be freely converted into a universal memory span.

Visual change detection and continuous-report tasks

A visual-array experiment might briefly show several colored objects, insert a delay, and ask which object changed. Another experiment asks participants to reproduce a specific remembered color using a continuous wheel. The first can estimate identification or change-detection performance; the second captures the distribution and precision of errors. These paradigms test partially different model assumptions. Oberauer and colleagues' benchmark review is a useful guide to the methodological distinctions.

Why an online “working-memory test” cannot diagnose a condition

Online tasks may offer interesting demonstrations of memory load. Their results vary with screen size, timing, practice, instructions, input device, sleep, distraction, and familiarity with similar puzzles. Formal neuropsychological interpretation requires appropriately standardized instruments, relevant norms, comprehensive history, and consideration of other cognitive functions. A weak online score alone does not establish ADHD, dementia, intellectual disability, or any other diagnosis.

Everyday examples of working memory

Mental arithmetic and budgeting

When calculating 47 + 38 mentally, a person may hold 47, add 30, then add 8 to the intermediate result. The operations require maintaining the goal and an interim value while another calculation occurs. Writing the numbers down reduces the amount that must be kept active. This explains why a person can reason well about a complicated budget using a spreadsheet even when doing every operation mentally is exhausting.

Reading a long sentence

A reader must track earlier nouns and relationships to interpret later clauses. Difficulty sometimes arises because a passage contains unfamiliar terminology, embedded clauses, or multiple referents. Working memory contributes to keeping relevant phrases available, while long-term vocabulary and syntactic knowledge also shape comprehension. Poor understanding of one paragraph cannot automatically be attributed to a small memory capacity.

Navigation

A person hears: “Turn left at the pharmacy, pass two intersections, then take the second right.” They must retain the instruction while attending to the street. Visual landmarks and prior knowledge can support this process. Repeating the full route may help briefly, but reviewing it in distinct stages or consulting a map places fewer simultaneous demands on working memory.

Conversations and meetings

A listener keeps earlier points accessible while evaluating a new proposal. In a meeting, they may need to compare two budgets, recall a constraint, and respond to a question. Notes act as an external memory support. Their value is practical: reducing simultaneous cognitive demands, while making it easier to inspect the reasoning and correct a mistaken intermediate step.

Cooking, music, and multitasking

Following a recipe requires tracking completed and pending steps. A musician may retain an upcoming phrase while coordinating the current one. Switching between writing, messages, and conversation can repeatedly change the information requiring priority. Performance depends on familiarity: practiced routines draw on acquired knowledge and skills, reducing the amount of conscious temporary coordination required.

Working memory across childhood, adulthood, and aging

Working-memory task performance develops substantially during childhood and adolescence, but not necessarily at a constant linear rate. Large U.S. cohort analyses of forward and backward digit-span performance found nonlinear developmental changes and particularly rapid gains in childhood, with additional changes into adolescence. Ahmed and colleagues (2022) examined two nationally representative samples, while Reynolds and colleagues (2022) showed how scoring rules can affect estimated capacities and developmental trajectories.

Adults vary widely in performance according to knowledge, demands, motivation, context, and specific measurement. Some working-memory measures decline on average with advancing age, particularly when processing and interference demands are high. Aging patterns depend on the task and cannot be read as a diagnosis from everyday forgetfulness. Researchers also study how older adults recruit neural networks under changing demands; a review of capacity and the aging brain discusses competing explanations.

Educational and occupational performance should therefore be interpreted in context. A child may understand a concept yet struggle when the instruction has many sequential parts. An adult may follow the same task effortlessly in a familiar setting but lose track during simultaneous interruptions. Practical supports can improve task completion without requiring claims about permanent biological expansion of a person's memory capacity.

Working memory, ADHD, stress, and clinical interpretation

Working-memory difficulties may occur in many circumstances, including divided attention, fatigue, stress, some medications, neurological illness, and certain developmental or psychiatric conditions. These are broad associations, not diagnostically specific signs. A complaint such as “I forget what I was about to say” can arise from interruption, anxiety, insufficient encoding, language demands, or other causes. Clinical assessment evaluates the pattern, duration, functional consequences, and wider history rather than treating one cognitive complaint as proof of a disorder.

Working-memory models help clinicians and educators generate hypotheses about task demands, but a model is not an assessment protocol. If problems are new, progressive, severe, or interfere meaningfully with daily functioning, a qualified professional can consider medical, psychological, and situational explanations. Acute confusion or sudden neurological changes require urgent medical assessment. This article describes the general psychology of working memory; diagnosis and disease-specific care belong to clinical evaluation.

Can you improve working memory? What the intervention evidence says

Practice improves practiced tasks

Repeated experience with a particular memory task commonly improves performance on that task. Participants learn the format, develop better strategies, and sometimes improve efficiency. An improved app score therefore establishes that the person got better at the app's measurements, not automatically that they became broadly more intelligent or that daily memory function has changed.

Near transfer and far transfer

Near transfer refers to improvement on an untrained task that is closely related to the practiced task. Far transfer refers to improvement in a more distant ability, such as academic achievement, general reasoning, or independent daily functioning. In a 2020 multilevel meta-analysis of typically developing children, Sala and Gobet found gains on related memory tasks but essentially no far-transfer benefit in intelligence or academic outcomes, especially with active control groups. Other populations and training types require separate evaluation.

What more recent reviews add

The evidence has continued to develop. A 2024 second-order meta-analysis in healthy adults reported small average improvements in working-memory outcomes and judged the potential real-world utility limited. A 2026 meta-analysis of computerized working-memory training found improvements in measured cognitive task performance and associated changes in task-related neural activity, but neural changes and test performance do not establish broad everyday transfer. A September 2026 meta-analysis of 55 studies detected modest, directionally asymmetric transfer between updating and span paradigms. Collectively, these findings support learning in trained and some related tasks while leaving broad, durable far transfer uncertain.

Practical strategies that reduce working-memory demands

When the immediate goal is to complete real tasks, changing the demands is often more useful than trying to enlarge a supposed fixed memory store. Break a complicated direction into a few meaningful steps. Write intermediate results, sketch spatial relationships, and use a checklist for multi-stage procedures. Reduce competing notifications when a task requires continuous concentration. Group related information when the grouping is genuinely meaningful, and build familiarity with recurring operations. These approaches make a task more manageable by changing cognitive load and available cues; their effects will vary by situation and should not be promoted as guaranteed permanent increases in underlying capacity.

Study effectively without confusing learning with temporary storage

For material that must be retained beyond the immediate task, use meaningful explanation, practice retrieving the material, and distribute practice across time. These techniques address long-term learning and retrieval as well as the burden placed on moment-to-moment thinking. Familiarity with key concepts reduces the number of new relationships to coordinate at once. This is a functional educational implication, not evidence that spaced practice directly expands an intrinsic working-memory slot limit.

Common myths and oversimplifications

Myth: Everyone has exactly seven memory slots

The seven-plus-or-minus-two phrase is a historical shorthand associated with immediate memory and recoding. Contemporary research estimates different limits depending on the design, what constitutes an item, the role of rehearsal, and the precision demanded. Describing capacity without specifying those conditions turns a research result into a misleading universal claim.

Myth: Working memory is just the brain's RAM

Computer RAM is a vivid analogy for short-term accessibility, but human working memory is shaped by attention, knowledge, goals, interference, representational precision, and task strategy. Information can be maintained in different forms and supported by distributed neural processing. There is no validated single number of human “gigabytes” for working memory.

Myth: Brain games reliably raise IQ and prevent dementia

Training effects on practiced or closely related tasks cannot be assumed to transfer to general intelligence, everyday independence, or prevention of neurological disease. Scientific reviews often find considerably weaker far-transfer evidence than advertisements imply, and clinical prevention claims require their own appropriate studies. Sala and colleagues' critical synthesis examines why transfer has been repeatedly overinterpreted.

Myth: Forgetting one instruction means that your memory is damaged

Ordinary fluctuations in attention, interference, and cognitive load can cause temporary lapses. The same person may remember a lengthy familiar story but struggle with a short string of unfamiliar symbols. A clinically meaningful change requires attention to the full pattern and its effects on functioning.

Frequently asked questions about working memory

What is a simple definition of working memory?

Working memory is the limited set of cognitive processes that keeps relevant information temporarily accessible so you can use it while thinking or acting. Mental arithmetic, following several instructions, and understanding complex sentences are common examples.

What are the four components of working memory?

In Baddeley's expanded multicomponent framework, the four components are the phonological loop, the visuospatial sketchpad, the central executive, and the episodic buffer. They are components of a particular influential model. Other scientific models divide the functions differently.

Is working memory the same as short-term memory?

They overlap but emphasize different functions. Short-term memory commonly refers to brief retention; working memory refers to temporary availability and processing control. Definitions vary across research traditions, so the meaning of a particular test should be checked rather than assumed.

How long does information stay in working memory?

There is no universal countdown. Information can become inaccessible quickly under distraction or remain useful longer through attention, rehearsal, structured representations, and interaction with long-term knowledge. Duration depends on conditions, not only on an intrinsic stopwatch.

How many things can working memory hold at once?

A few independent chunks is a useful shorthand for many demanding laboratory settings; roughly three to five often appears in theoretical discussions. The number changes with task, material, precision, age, and grouping. There is no one-size-fits-all personal capacity value.

What is an example of working memory in school?

A student holds the beginning of a math word problem in mind while interpreting its later conditions, then keeps an intermediate result available during a second calculation. Writing the intermediate result down can support accurate reasoning without changing the student's underlying knowledge.

Does working memory affect reading comprehension?

Yes, working-memory measures are associated with aspects of comprehension, especially when readers must coordinate information across words and clauses. Vocabulary, background knowledge, language proficiency, attention, and task complexity also matter. A reading difficulty is not a working-memory diagnosis.

Is working memory located in the prefrontal cortex?

Prefrontal circuits contribute to controlling and prioritizing current information, but working-memory representations and relevant operations involve distributed brain systems. Content-specific sensory regions and connected networks are important too.

Can an app increase working-memory capacity permanently?

Practice can improve performance on the app and sometimes related memory tasks. Evidence that typical commercial programs cause broad, durable improvements in everyday functioning remains limited and depends on the population and outcome. Claims should identify which tasks improved and how long the benefit lasted.

Can working memory be improved without brain-training software?

People can often improve performance in specific activities by minimizing distraction, using external notes, breaking instructions into meaningful steps, and learning relevant background knowledge. These changes can make tasks easier even when they do not demonstrate a general increase in cognitive capacity.

Does anxiety or stress always reduce working memory?

Stress and anxiety can alter attention and task priorities, but effects depend on intensity, circumstances, individual differences, and which components the task measures. A single lapse during stress is insufficient evidence of a lasting deficit.

Is poor working memory a sign of ADHD or dementia?

Working-memory problems can occur in several conditions and in ordinary everyday circumstances. They are not specific enough to diagnose ADHD, dementia, or another disorder. Persistent, worsening, or functionally significant changes deserve an appropriate professional evaluation.

Can experts remember more than beginners?

Expertise often allows people to organize domain-specific information into meaningful structures and retrieve familiar patterns efficiently. As a result, experts may appear to retain more material within their area of knowledge. This advantage need not transfer to arbitrary symbols or unrelated tasks.

What is the best working-memory test?

No single test is universally best. The choice depends on whether the goal is to study retention, manipulation, updating, resistance to interference, visual precision, development, or clinical functioning. Reliable interpretation uses standardized administration, suitable reference data, and often multiple measures.

What researchers agree on—and what remains unresolved

Established evidence supports a system of limited temporary accessibility that is fundamental to ongoing cognition and closely connected with attention, long-term knowledge, and control processes. Working-memory tasks reveal reliable performance limits, individual and developmental differences, and important brain-network contributions. Major open questions concern how best to define capacity across tasks, how active and relatively silent neural states maintain information, how competing representational resources are allocated, and when training benefits extend to genuinely different skills. The most faithful account of working memory joins a strong functional definition with explicit distinctions among the models used to explain it.

Human Memory: Types, Processes, and How It Works

Short-Term Memory vs. Long-Term Memory: Key Differences

Long-Term Memory: Types, Storage, and Retrieval

References

Ahmed, S. F., Ellis, A., Ward, K. P., Chaku, N., & Davis-Kean, P. E. (2022). Working memory development from early childhood to adolescence using two nationally representative samples. Developmental Psychology, 58(10), 1962–1973. PubMed.

Baddeley, A. (1992). Working memory. Science, 255(5044), 556–559. https://doi.org/10.1126/science.1736359.

Baddeley, A. (2000). The episodic buffer: A new component of working memory? Trends in Cognitive Sciences, 4(11), 417–423. https://doi.org/10.1016/S1364-6613(00)01538-2.

Baddeley, A. (2012). Working memory: Theories, models, and controversies. Annual Review of Psychology, 63, 1–29. https://doi.org/10.1146/annurev-psych-120710-100422.

Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114. https://doi.org/10.1017/S0140525X01003922.

Cowan, N. (2017). The many faces of working memory and short-term storage. Psychonomic Bulletin & Review, 24, 1158–1170. https://doi.org/10.3758/s13423-016-1191-6.

Cowan, N., Rouder, J. N., Blume, C. L., & Saults, J. S. (2012). Models of verbal working memory capacity: What does it take to make them work? Psychological Review, 119(3), 480–499. https://doi.org/10.1037/a0027791.

Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19(4), 450–466. https://doi.org/10.1016/S0022-5371(80)90312-6.

D'Esposito, M., & Postle, B. R. (2015). The cognitive neuroscience of working memory. Annual Review of Psychology, 66, 115–142. https://doi.org/10.1146/annurev-psych-010814-015031.

Engle, R. W. (2018). Working memory and executive attention: A revisit. Perspectives on Psychological Science, 13(2), 190–193. https://doi.org/10.1177/1745691617720478.

Fu, J., Zhou, Y., Yuan, M., et al. (2026). The asymmetry of working memory training transfer: A systematic review and meta-analysis. Psychonomic Bulletin & Review, 33, Article 227. https://doi.org/10.3758/s13423-026-02994-5.

Li, G., Liu, Y., & Chen, A. (2026). Meta-analysis of computerised working memory training: Behavioural gains, training parameters, transfer mechanisms, and neural correlates. npj Digital Medicine, 9, Article 337. Journal article.

Oberauer, K., & Hein, L. (2012). Attention to information in working memory. Current Directions in Psychological Science, 21(3). https://doi.org/10.1177/0963721412444727.

Oberauer, K., Lewandowsky, S., Awh, E., et al. (2018). Benchmarks for models of short-term and working memory. Psychological Bulletin, 144(9), 885–958. https://doi.org/10.1037/bul0000153.

Reynolds, M. R., Niileksela, C. R., Gignac, G. E., & Sevillano, C. N. (2022). Working memory capacity development through childhood: A longitudinal analysis. Developmental Psychology, 58(7), 1254–1263. https://doi.org/10.1037/dev0001360.

Sala, G., & Gobet, F. (2020). Working memory training in typically developing children: A multilevel meta-analysis. Psychonomic Bulletin & Review, 27, 423–434. https://doi.org/10.3758/s13423-019-01681-y.

Sala, G., et al. (2022). Cognitive training: A field in search of a phenomenon. PubMed.

Wilhelm, O., Hildebrandt, A., & Oberauer, K. (2013). What is working memory capacity, and how can we measure it? Frontiers in Psychology, 4, 433. https://doi.org/10.3389/fpsyg.2013.00433.

Psychological Encyclopedia