Abstract

Cognition is a type of mental process: the processes by which an organism obtains knowledge, and by which people take in, hold, transform and use information. Since the 1950s, research has treated immediate memory and controlled processing as sharply limited in capacity. Executive control steers that capacity, judgment leans on shortcuts whose errors are predictable, and whether fast and slow thinking are distinct kinds is disputed. Grounded, predictive and computational approaches compete to explain how knowledge is represented, and abilities change unevenly across adulthood. Three interactive demonstrations cover chunking, automaticity with practice and adult cognitive trajectories.

Keywords: cognition, working memory, attention, executive functions, dual-process theories

Cognition names the processes through which people acquire, retain and use knowledge, and its study is the core of cognitive psychology and a central concern of the wider, interdisciplinary cognitive science. The field acquired its first comprehensive textbook in 1967, when Neisser (1967/2014) laid out how stimulus information is handled on its way into memory and thought; the book is cited here from its classic reissue.

Several earlier works had prepared the ground. Broadbent (1958) published a book on perception and communication, and Chomsky (1959) wrote a review of B. F. Skinner's Verbal Behavior. Before either, Miller (1956) had reviewed evidence that the span of absolute judgment and that of immediate memory are both narrow, casting human information processing in terms of capacity limits that the sections below take up.

Key Takeaways

  • Working memory, the small amount of information held in mind and worked on at one moment, is limited, but the limit is counted in chunks, meaningful units built by recoding, so practice at recoding can stretch it; Miller allowed that the recurring number seven might be a coincidence.
  • The multicomponent model of working memory, which divides it into cooperating parts, has been revised and remains influential, yet several rival definitions of working memory are still in use.
  • Executive functions, the control processes that keep goals in view and override habits, were moderately correlated yet separable in a study of college students, and which tasks measure which function is disputed.
  • Heuristics, mental shortcuts such as judging how likely something is by how easily examples come to mind, are usually economical and effective but produce systematic, predictable errors.
  • Dual-process theories, which divide thinking into fast Type 1 and slower, deliberate Type 2 processing, are contested: critics doubt that the two kinds are exclusive and that a switch moves thinking from one to the other.
  • Processing speed, how quickly simple mental operations are carried out, slows with age, and that slowing was proposed to account for part of age differences in fluid abilities; much of the change across abilities in adulthood is shared, and the shared part grows with age.

What Cognition Is

Intellectual or mental process whereby an organism obtains knowledge.

Scope note for the MeSH descriptor Cognition (D003071), U.S. National Library of Medicine.

MeSH lists two entry terms under this descriptor. Cognitive Function is the clinical phrasing for the same set of capacities, used when they are tested or found impaired. Insight is filed here too, in its sense of suddenly grasping how a problem's parts fit together, which is why research on insight problem solving is indexed under cognition.

The term is broad by design. Perception, learning, language and thinking each have their own literatures, and cognition names what they share, the acquisition and use of knowledge, rather than a separate faculty alongside them.

Because the same process can be described in very different terms, researchers often separate levels of analysis. Working through the case of vision, Marr (1982/2010) distinguished the hardware in which a process is physically implemented, the algorithm and representations it uses, and the computational problem it solves, and he treated seeing as the construction of a series of staged representations. The later edition of the book acknowledged that the core problems of vision remained open some thirty years on.

More recently, the demand has grown that explanations actually run. To explain how the brain carries out cognition, Kriegeskorte and Douglas (2018) called for brain-computational models able to do the tasks themselves, judged against both neural and behavioral data, though they presented this line of work as young, with its literature only beginning to form.

Figure 1

Three Levels at Which a Cognitive Process Can Be Explained

Marr's three levels of analysis, illustrated with mental arithmetic Three stacked boxes. The top box, computational level, asks what problem is solved and why, with the example of adding two numbers. The middle box, algorithmic level, asks what representation and procedure are used, with the example of decimal digits added column by column with carrying. The bottom box, implementation level, asks what physical system carries out the procedure, with the example of neurons in a brain or transistors in a calculator. Arrows run between neighbouring boxes. Computational level: what is computed, and why? Example: produce the sum of two numbers. Algorithmic level: which representation and procedure? Example: decimal digits added column by column, carrying tens. Implementation level: what physical system runs it? Example: neurons in a brain, or transistors in a calculator.
Note. An original schematic. The same task can be described at each level, and a description at one level does not fix the others: many procedures compute the same sum, and the same procedure can run on very different hardware. The arithmetic example is this article's own.

Types of Cognition

In MeSH, cognition sits under mental processes and has twelve narrower descriptors, listed in Table 1. The list records how the biomedical literature is indexed, not a theory of how cognition divides. The subtypes overlap and are not mutually exclusive: metacognition is itself a form of awareness, and intuition and cognitive reflection are often studied together.

SubtypeIn brief
AwarenessTaking account of an object or state of affairs, without necessarily attending to or evaluating it.
Cognitive DissonanceThe motivational tension that arises when beliefs, or beliefs and actions, are inconsistent.
Cognitive FlexibilityAdjusting thinking and behavioural strategies when task or environmental demands change.
Cognitive ReflectionDeliberate reconsideration of past actions and decisions and of the thinking behind them.
Cognitive ReserveThe capacity to cope with, or recover from, the effects of neural injury or a psychotic episode.
ComprehensionGrasping the meaning, nature or importance of something; understanding.
ConsciousnessThe sense of awareness of oneself and of one's surroundings.
ImaginationForming new perceptual or ideational patterns out of past experience.
IntuitionKnowing or understanding without the conscious use of reasoning.
MetacognitionAwareness and understanding of one's own thought processes.
Processing SpeedThe time needed to understand and carry out a mental task, with no motor response necessarily involved.
Rumination, CognitiveRepetitive, obsessive thinking about an idea, a situation or a choice.

Table 1. Direct subtypes of cognition in the MeSH classification (tree F02.463.188), each glossed from its MeSH scope note.

Several of these subtypes return later in the article. Processing speed and cognitive reserve are central to the section on adulthood, intuition to the section on judgment and the dual-process debate, and metacognition to the discussion of control.

Limited Capacity and Working Memory

The modern study of cognition began with limits. Reviewing experiments on absolute judgment and on immediate memory, Miller (1956) found that both spans are narrow; he noted that the number seven kept turning up, while allowing that its recurrence might be nothing more than coincidence. His more durable point concerned recoding. By grouping items into larger units, or chunks, and renaming them, a person can carry more information within the same span, and Miller treated linguistic recoding as one form of this. He also remarked that recoding had received little experimental study. A later review put the central capacity limit nearer four chunks than seven, with estimates averaging three to five chunks (Cowan, 2001). That limit shows only under constrained conditions, where rehearsal and recoding into long-term memory are prevented, because both push the estimates up (Cowan, 2001).

Atkinson and Shiffrin (1968) proposed a system for human memory together with the control processes that operate on it. The multicomponent model of working memory originated in a chapter by Baddeley and Hitch (1974). More than two decades later, Baddeley (2000) argued that the original three-component model missed some phenomena and added a fourth component, the episodic buffer: a store of limited capacity that holds information in a multimodal code and is tied to executive control and conscious awareness. The model has been revised rather than replaced and remains influential; in a later review, Baddeley (2012) distinguished the overall multicomponent framework, which he judged relatively stable, from the more specific models built within it, and closed with speculations on further developments. It is not the only definition in circulation. Cowan (2017) separated nine definitions of working memory in current use, observed that combining them could produce still more, and argued that the variety breeds confusion.

Divergent definitions make shared targets valuable. A group of authors from a broad range of theoretical perspectives, drawing also on a survey of experts, assembled the benchmark findings that any model of short-term and working memory ought to explain (Oberauer et al., 2018). The benchmarks were chosen by consensus, described in terms meant to be neutral between theories, and left open to new proposals. Neuroscience adds a further complication to any single-store picture: Squire (2004) traced the history of the view that memory is not one faculty but several systems housed in separate brain structures, including the division between declarative and nondeclarative memory. An alternative splits memory by processing mode rather than by consciousness, holding that consciousness is a poor criterion for telling memory systems apart (Henke, 2010).

The first demonstration makes Miller's recoding concrete. It shows a binary string briefly and asks for immediate recall, then offers the same string grouped in threes with each group's octal name, so that the reader holds a third as many items. The scores are illustrative, not data.

Demo 1: the same string as 18 digits or 6 chunks

Press Show, read the string, then type it back.

Illustrative, not data. Each group of three binary digits names one octal digit (000 = 0 up to 111 = 7), so recoding turns 18 items into 6 without losing information. The gain depends on having practised the translation until it is quick.

Attention and Automatic Processing

Attention concerns which information receives further processing. Among the book-length treatments of the early information-processing period was Broadbent's Perception and Communication (Broadbent, 1958). Against Broadbent's filter theory, Deutsch and Deutsch (1963) argued that every input receives full perceptual analysis and that selection comes only afterwards, although they could say only that the neurophysiological evidence was compatible with this view. Across three studies using a response competition paradigm, Lavie (1995) found that distractors interfered when perceptual load was low but not when it was high, which Lavie proposed as a way to reconcile the early and late selection views. The distractors in that work were, for the most part, clearly separate from the targets (Lavie, 1995).

Schneider and Shiffrin (1977) separated automatic from controlled processing, setting automatic detection against controlled search. Their experiments on detection and search, scored by accuracy and by reaction time, contrasted consistent mapping, in which an item that serves as a target never appears as a distractor, with varied mapping, in which items exchange roles across trials.

Whether attention is one thing at all is now openly questioned. A theory paper argued that attention is neither a unitary construct nor a single system, and that the word names something to be explained rather than an explanation, so it may be best avoided in explanations altogether (Hommel et al., 2019). The authors concentrated on visual selection, left mechanistic details open, and proposed no alternative theory.

The second demonstration simulates practice on a search task under each kind of mapping, showing how automaticity might be expected to develop when targets are always targets but not when items swap roles. The curves are seeded illustrations, not data.

Demo 2: practice makes search automatic only when the mapping stays fixed

practice sessionsearch time (ms)varied mappingconsistent mapping
After 12 sessions with 3 target items: consistent mapping about 469 ms, varied mapping about 667 ms. The consistent-mapping curve has flattened toward the base time, so holding more targets barely slows it; the varied-mapping curve still grows with every added item.

Illustrative simulation with a fixed seed, not data. Under consistent mapping an item is always a target or always a distractor; under varied mapping items swap roles between trials. The model gives each held item a search cost that shrinks with practice only under consistent mapping.

Executive Functions and Cognitive Control

Executive function refers to the control processes that coordinate other cognitive operations, and the main empirical question has been whether it is one capacity or several. In a latent-variable study of college students, three executive functions, updating, inhibition and shifting, proved moderately correlated yet separable, a pattern the authors described as unity and diversity (Miyake et al., 2000). The study related the three to complex tasks such as the Wisconsin Card Sorting Test and the Tower of Hanoi; a dual-task measure, by contrast, was unrelated to any of them, and the sample was confined to college students.

Diamond (2013) reviewed the field and identified three core executive functions: working memory, cognitive flexibility and inhibitory control, the last spanning both response inhibition and interference control. A cross-sectional study of 325 participants, roughly 30 at each age from 4 to 13 years plus young adults, all given the same computerized battery, found task switching still immature at age 13 and memory and inhibition not independent of each other, with young children responding impulsively on difficult trials (Davidson et al., 2006). Diamond also discussed related capacities, fluid intelligence among them. Her review flagged several unsettled points: the assignment of particular tasks to particular components is disputed, complex span tasks draw on more than working memory, and the debate over whether working memory or inhibition is more basic remains unresolved. She further noted that loneliness, stress and poor sleep impair executive functions.

At the level of the brain, Miller and Cohen (2001) proposed that the prefrontal cortex supports cognitive control by actively maintaining representations of goals, which send bias signals that steer processing elsewhere in the brain. They acknowledged that further issues remained open.

Control also depends on tracking how one's own cognition is going. Flavell (1979) framed metacognition and cognitive monitoring as a new area of developmental research, drawing on studies in which younger children, unlike older ones, did little monitoring of their own memory and comprehension.

Judgment, Heuristics and the Dual-Process Debate

Judgment under uncertainty shows the same economy at work. Tversky and Kahneman (1974) argued that people estimating uncertain quantities rely on three heuristics, availability, representativeness and adjustment from an anchor, which are usually economical and effective but lead to systematic and predictable errors. A rival programme, built on bounded rationality, proposed fast and frugal heuristics that rest on one-reason decision making and do not integrate all the available information (Gigerenzer & Goldstein, 1996). In their computer simulations, Take The Best, which decides on a single reason, matched or beat more demanding rational strategies such as multiple regression, a result they offered as an existence proof from simulation and not as data on human judges (Gigerenzer & Goldstein, 1996).

Kahneman (2003) later organised this research around accessibility, the ease with which particular thoughts come to mind, and used it to explain the biases of intuitive judgment, including attribute substitution and framing effects; the same review also took in choice, discussing prospect theory and its bearing on decision making. He stressed that intuition is generally skilled and successful, and that deliberate reasoning corrects it only occasionally, when a relevant rule happens to be accessible.

These contrasts are formalised in dual-process theory. Evans and Stanovich (2013) defended dual-process theories of higher cognition, which distinguish fast Type 1 processing that supplies default responses from Type 2 processing linked to working memory and hypothetical thinking, against five main lines of criticism. They granted that some criticisms held against some versions of the theories, but argued that critics wrongly treated the theories as a single category and that the critiques were overstated. The theories remain contested. Melnikoff and Bargh (2018) rejected as a myth the typology that sorts mental processes into two bundles, one conscious, controllable and inefficient and the other the reverse; theirs was an argument, not a report of new data. De Neys (2023) reviewed evidence against two assumptions common to fast-and-slow models, that intuition and deliberation are exclusive and that some mechanism switches thinking from one to the other; he found no solid ground for the first and judged the leading switch accounts conceptually problematic. He proposed an alternative architecture, which that paper did not test.

Grounded and Predictive Approaches

Grounded cognition is an approach rather than a single theory. Barsalou (1999) had earlier argued that knowledge is carried not by amodal symbols but by perceptual symbols, simulations run in sensory-motor systems. In a later review, Barsalou (2008) gathered evidence that cognition, including the conceptual knowledge examined in research on concept formation, rests on bodily states, situated action and modal simulations instead of amodal symbols. He also addressed common misperceptions of the area and set out theoretical, empirical and methodological issues that remained open. Critics replied that sensory and motor activation during conceptual tasks does not show that concepts are embodied, because the same evidence fits a disembodied view of conceptual representation equally well (Mahon & Caramazza, 2008). They put forward a middle position, grounding by interaction (Mahon & Caramazza, 2008).

Related ideas travel under the label embodied cognition, but Wilson (2002) showed that the label covers six distinct claims, among them that people offload cognitive work onto the environment, that cognition is for action, and that it is situated, and that their support is uneven. Four of the claims appeared at least partly true, the claim that the environment forms part of the cognitive system was deeply problematic, and the claim that even off-line cognition is body-based had received the least attention, although Wilson judged it possibly the best supported and most powerful of the six.

Predictive processing treats the brain as a hierarchical prediction machine. In a target article, Clark (2013) described brain activity as the work of hierarchical generative models that predict incoming signals and minimise prediction error, the mismatch between what is predicted and what arrives, with perception, action and attention folded into the same scheme. The view connects with a broader proposal, the free-energy principle, which Friston (2010) suggested might unify several existing theories of brain function; he offered that unification tentatively. Predictive processing remains contested. Clark himself devoted part of the article to its pitfalls and challenges, evidential, methodological and conceptual (Clark, 2013). A review of the neurophysiological evidence judged the support for predictive processing to be mixed rather than decisive, while noting that no clear counterevidence had yet appeared (Walsh et al., 2020). Among its reasons, population-level measures blur the separate expectation and error units the view posits, and adaptation can explain repetition suppression (Walsh et al., 2020).

Cognition Across Adulthood

Research on cognitive aging has long asked how much of adult decline reflects a single cause. Salthouse (1996) proposed that age-related slowing in processing speed contributes to declines in fluid cognition, through what he called a limited-time mechanism and a simultaneity mechanism. Salthouse presented slowing as accounting for only some of the age differences, and judged the evidence consistent with the proposal rather than conclusive. A later methodological analysis argued that cross-sectional age variance extraction is not a sound test of developmental hypotheses about whether change in a mediator variable goes with change in a target variable, because the age-related variance two measures share across a cross-sectional sample is not equivalent to a correlation between their longitudinal changes; on that basis the authors discouraged the method for such multivariate hypotheses (Lindenberger et al., 2011). A later multilevel meta-analysis of 89 effect sizes from 22 datasets, covering more than 30,000 individuals, found that a general factor shared across abilities accounts for a large share of individual differences in age-related cognitive change, a share that rose from 45% at age 35 to 70% at age 85 (Tucker-Drob et al., 2019). The analysis also considered how this general factor of change relates to general intelligence; its result held in a sensitivity analysis that controlled for dementia, and different classifications of cognitive domains made little difference.

How age effects are measured changes what they look like. Salthouse (2019) compared cross-sectional data from over 5,000 adults with three-occasion longitudinal data from almost 1,600 adults across four domains: vocabulary, speed, reasoning and memory. Cross-sectional and quasi-longitudinal estimates of age trends were similar, whereas longitudinal estimates were distorted by experience with the tests, which inflated apparent retention up to about age 65. The trajectories differed in shape: Salthouse (2019) traced speed falling almost linearly from early adulthood onward, while losses in memory and reasoning gathered pace as age advanced.

People with similar brain changes do not all show similar symptoms. Stern (2002) noted that the extent of brain pathology and its clinical expression are only weakly related, and proposed cognitive reserve as an active capacity to cope with pathology, in contrast with passive brain reserve; he cited epidemiologic and imaging data in support. A consensus whitepaper from a workgroup later gave working definitions of brain maintenance, brain reserve and cognitive reserve, framed for cognitive aging and Alzheimer's disease, two settings in which the loose link between brain changes and cognitive dysfunction matters in practice (Stern et al., 2020). The workgroup stressed that the definitions were still evolving, that proxies such as education, occupation and IQ are not direct measures of reserve, that brain reserve and brain maintenance cannot be told apart in cross-sectional data, and that the neurobiology of cognitive reserve remains a black box.

The third demonstration draws schematic curves from age 20 to 80 for speed, memory, reasoning and vocabulary, and lets the reader switch between a cross-sectional view and a longitudinal one that adds an illustrative practice boost. The curves are schematic and reproduce no study's estimates.

Demo 3: different abilities, different paths through adulthood

20304050607080age (years)relative score
At 50: speed 0.00, memory -0.16, reasoning -0.03, vocabulary 0.23. Speed falls steadily from the start; memory and reasoning fall faster later; vocabulary holds up longest.

Schematic curves drawn for illustration, not data. Scores are relative to an arbitrary average (dashed line). The shapes follow the qualitative pattern described in the text; the practice boost is an invented constant that fades after 65.

Minds, Models and Machines

Comparisons with artificial systems have sharpened questions about what human cognition does. A theory paper argued that machines built to learn and think like people would need compositionality, learning to learn, and causal models of the world supported by intuitive theories of physics and psychology, and it contrasted that kind of model building with the pattern recognition emphasised in deep learning (Lake et al., 2017). The authors stated plainly that they were not dismissing neural networks.

Binz and Schulz (2023) took the reverse route, running GPT-3, a language model with 175 billion parameters, through experimental tasks from cognitive psychology, including vignette problems such as the Wason card selection task and the Linda problem, bandit tasks such as the horizon task, and a causal-reasoning task contrasting observing a variable with intervening on it; a blicket problem, which GPT-3 solved, was among the vignettes. The results were mixed. The model answered six of twelve vignettes correctly and performed well on the bandit tasks, but it showed no directed exploration and failed at causal reasoning, including inference along a causal chain. The authors tested one model only, cautioned that the vignettes may have been in its training data, and found that small changes to a prompt could derail its answers.

Kriegeskorte and Douglas (2018) placed such comparisons within cognitive computational neuroscience, which seeks brain-computational models that perform cognitive tasks and confronts them with the data of cognitive neuroscience, using methods such as representational similarity analysis applied to fMRI. They considered production systems such as ACT-R, Bayesian cognitive models and deep convolutional neural networks, and cautioned that deep networks use few of the brain's neural dynamics, that the role of backpropagation in the brain is controversial, and that decoding reveals the product of processing rather than the process itself.

Worked Example

This section is an illustrative recoding exercise, and its arithmetic is this article's own rather than any study's. Its premise is Miller's observation that recoding into chunks offers a way past the limit on immediate memory (Miller, 1956).

Consider the 18-digit binary string 101 000 100 111 001 110. Held as 18 separate digits, it is too long to repeat back after a single hearing. Read in groups of three, each group can be renamed as one octal digit: 101 is 5, 000 is 0, 100 is 4, 111 is 7, 001 is 1, and 110 is 6. The string becomes the six digits 504716, which fit comfortably within a span of a handful of items. The information is unchanged; what has changed is that it is now held as six chunks instead of eighteen. The saving is available only to someone who has practised the translation until it is quick. Table 2 sets out the translation step by step.

The lesson is that a capacity limit is counted in chunks, not in raw items, so expertise changes what a fixed span can hold.

Table 2

Recoding an 18-Digit Binary String Into Six Octal Chunks

PositionBinary groupOctal digit
11015
20000
31004
41117
50011
61106

Note. This article's own illustrative example, not data. Each group of three binary digits is read as one octal digit (000 = 0 through 111 = 7), so the 18 items 101000100111001110 become the 6 items 504716 and can be turned back into the original string without loss.

Discussion

Some conclusions about cognition are secure. The span of absolute judgment and that of immediate memory are sharply limited, and recoding into chunks offers a way past the limit (Miller, 1956). Executive functions are related but separable, at least in the college samples in which this was first shown (Miyake et al., 2000). Heuristics are usually serviceable yet bias judgment in predictable directions (Tversky & Kahneman, 1974).

Much else remains open. Working memory itself lacks a single agreed definition, and the confusion this causes is a live complaint (Cowan, 2017). Whether attention is a unitary construct, or a word that should be retired from explanations, is disputed (Hommel et al., 2019). Whether fast and slow thinking are separate kinds of process, or whether the two-category typology is a myth, is argued on both sides (Evans & Stanovich, 2013; Melnikoff & Bargh, 2018). How far grounded and prediction-based views can be put to decisive test is unsettled, since their own proponents list open issues (Barsalou, 2008) and pitfalls (Clark, 2013). And how cognitive reserve is implemented in the brain is unknown, with its definitions still evolving and its common proxies only indirect (Stern et al., 2020). A recurring theme across these disputes is that a single term, whether working memory, attention or intuition, can conceal several different constructs, so that progress often begins with deciding what a word is being asked to explain.

Current Directions

Recent work has concentrated on tightening the field's targets and vocabulary. In working-memory research, consensus benchmarks give competing models a shared set of findings to explain (Oberauer et al., 2018), and an inventory of definitions has made explicit how differently the term is used (Cowan, 2017). The status of attention as an explanation has been challenged directly (Hommel et al., 2019).

The dual-process debate has moved from defending or rejecting the two-category scheme (Melnikoff & Bargh, 2018) toward scrutinising its specific assumptions about exclusivity and switching (De Neys, 2023). In aging research, meta-analytic evidence for a growing shared factor of change (Tucker-Drob et al., 2019) sits alongside work separating age trends from practice effects in longitudinal designs (Salthouse, 2019), and consensus definitions now distinguish cognitive reserve from brain reserve and brain maintenance (Stern et al., 2020).

A third line of recent work compares minds with machines. Arguments about what human-like learning requires (Lake et al., 2017), calls for task-performing models tested against brain data (Kriegeskorte & Douglas, 2018), and experiments that put a language model through the tasks of cognitive psychology (Binz & Schulz, 2023) together raise the question of which features of human cognition an artificial system must reproduce, and how to tell whether it has.

Common Misconceptions

Working memory holds exactly seven items.
The figure comes from Miller's review, but Miller himself allowed that the recurring seven might be a coincidence, and his emphasis fell on chunks: recoding lets the same span carry more information (Miller, 1956). A span measured in chunks is not a fixed count of raw items. Later work put the limit at about four chunks, typically three to five, once rehearsal and long-term recoding are stopped from inflating the count (Cowan, 2001).
Fast intuitive thinking and slow deliberate thinking are two separate brain systems.
The two-category typology behind this belief has been argued to be a myth (Melnikoff & Bargh, 2018), in an argument rather than with new data. Whether the two kinds of thinking are distinct at all remains an open dispute.
Heuristics are simply errors in thinking.
The work that made heuristics famous described them as usually economical and effective; the errors they produce are systematic and predictable rather than the whole story (Tversky & Kahneman, 1974).
Cognitive decline begins only in old age.
Decline in several abilities does not wait for old age: speed drops almost linearly from early adulthood, and losses in memory and reasoning speed up as the years pass, although vocabulary knowledge rose until the 60s (Salthouse, 2019). Longitudinal studies can hide this early change, because experience with the tests flatters retention up to about age 65.

Glossary

Automatic processing.
Processing that runs quickly and with little demand on limited capacity, contrasted with controlled processing.
Chunk.
A meaningful unit formed by grouping and renaming smaller items, so that one chunk occupies one place in a limited span.
Cognition.
The mental processes by which knowledge is obtained, held, transformed and used.
Cognitive function.
The clinical term for the same capacities, used chiefly when they are tested or found impaired; a MeSH entry term for cognition.
Cognitive reserve.
A concept for the active capacity that lets some people cope with brain pathology with fewer symptoms; its definitions are still evolving and its usual proxies are indirect.
Controlled processing.
Effortful processing that draws on limited capacity and can be directed flexibly, contrasted with automatic processing.
Dual-process theories.
Theories that divide thinking into fast Type 1 and slower, deliberate Type 2 processing; contested, with critics doubting that the two kinds are exclusive or that a switch governs them.
Episodic buffer.
A limited-capacity store holding information in a multimodal code, added to the multicomponent model of working memory as a fourth component.
Executive functions.
Control processes that coordinate other cognitive operations; related but separable, with the mapping of tasks to functions disputed.
Grounded cognition.
An approach holding that knowledge rests on bodily states, situated action and modal simulation rather than amodal symbols; several of its issues remain open.
Heuristic.
A mental shortcut that usually works economically but can produce systematic, predictable errors.
Insight.
A sudden grasp of how the parts of a problem fit together; a MeSH entry term for cognition.
Levels of analysis.
Separate descriptions of one process: the problem it solves, the procedure and representations it uses, and the physical machinery that implements it.
Multicomponent model of working memory.
The view that short-term retention is carried out by several cooperating components; revised over time and still influential, though rival definitions of working memory remain in use.
Predictive processing.
The contested view of the brain as a hierarchical prediction machine that works to minimise prediction error.
Processing speed.
How quickly simple mental operations are carried out; its slowing with age was proposed to account for part of age differences in fluid abilities.
Working memory.
The limited capacity for holding information in mind and working on it over short intervals; defined in several competing ways.

Key Researchers

Alan Baddeley

(University of York). Cognitive psychologist who proposed, with Graham Hitch, that short-term retention works as a multicomponent working memory, later adding an episodic buffer. ORCID · Wikipedia · Wikidata

Lawrence W. Barsalou

(University of Glasgow). Cognitive scientist who argues that knowledge is grounded in simulations of perception, action and bodily states rather than abstract symbols. ORCID · Faculty page · Wikipedia

Andy Clark

(University of Sussex). Philosopher and cognitive scientist who has critically examined the view of the brain as a hierarchical prediction machine and what it implies for situated, acting agents. Faculty page

Adele Diamond

(University of British Columbia). Developmental cognitive neuroscientist who studies the core executive functions and how they develop and can be improved in children. ORCID · Wikipedia · Wikidata · Faculty page

Daniel Kahneman

(Princeton University; 1934–2024). Psychologist who, with Amos Tversky, showed how intuitive judgment relies on heuristics that produce systematic biases. Wikipedia · Wikidata · Faculty page

Ulric Neisser

(Cornell University; 1928–2012). Psychologist whose 1967 book gave cognitive psychology its first comprehensive survey and its first true textbook. Wikipedia · Wikidata · Faculty page

Yaakov Stern

(Columbia University Irving Medical Center). Cognitive neuroscientist who developed the concept of cognitive reserve to explain why brain pathology and symptoms are only loosely linked. ORCID · Wikipedia · Wikidata · Faculty page

Frequently Asked Questions

What is cognition?

Cognition is the set of mental processes through which knowledge is acquired, held and used, spanning perception, memory, thought and language. Its study gained a first comprehensive textbook in 1967 (Neisser, 1967/2014).

How many items can working memory hold?

There is no fixed count of raw items. Miller noted that the number seven kept recurring but allowed that this might be coincidence, and he showed that recoding items into chunks lets the same span carry more information (Miller, 1956). When rehearsal and recoding are prevented, the estimate drops to about four chunks, usually somewhere between three and five (Cowan, 2001).

What is the episodic buffer?

The episodic buffer is a store of limited capacity, holding information in a multimodal code, that Baddeley added as a fourth component of working memory. He proposed it to capture phenomena the original model did not readily explain (Baddeley, 2000).

Are executive functions one ability or several?

Evidence from college students points to both. Updating, inhibition and shifting were moderately correlated with one another yet separable when measured as latent variables (Miyake et al., 2000).

Are heuristics irrational?

Not as a rule. Heuristics such as availability are usually economical and effective, although they also produce systematic and predictable errors in judgment (Tversky & Kahneman, 1974).

Is the idea of fast and slow thinking still accepted?

It is contested. A recent review found no solid ground for treating intuition and deliberation as exclusive, judged the leading accounts of a switch between them conceptually problematic, and sketched an alternative that it did not test (De Neys, 2023).

Does cognition slow down with age?

Processing slows across adulthood, and Salthouse proposed that this slowing contributes to declines in fluid abilities. Salthouse presented it as explaining only some of the differences between age groups (Salthouse, 1996).

What is cognitive reserve?

Cognitive reserve is a concept for the active capacity that lets some people cope with brain pathology while showing fewer symptoms, which helps explain why pathology and symptoms are only weakly related (Stern, 2002).

References

Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), Psychology of learning and motivation (Vol. 2, pp. 89–195). Academic Press. https://doi.org/10.1016/S0079-7421(08)60422-3

Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), Psychology of learning and motivation (Vol. 8, pp. 47–89). Academic Press. https://doi.org/10.1016/S0079-7421(08)60452-1

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

Barsalou, L. W. (2008). Grounded cognition. Annual Review of Psychology, 59, 617–645. https://doi.org/10.1146/annurev.psych.59.103006.093639

Barsalou, L. W. (1999). Perceptual symbol systems. Behavioral and Brain Sciences, 22(4), 577–660. https://doi.org/10.1017/S0140525X99002149

Binz, M., & Schulz, E. (2023). Using cognitive psychology to understand GPT-3. Proceedings of the National Academy of Sciences, 120(6), Article e2218523120. https://doi.org/10.1073/pnas.2218523120

Broadbent, D. E. (1958). Perception and communication. Pergamon Press.

Chomsky, N. (1959). A review of B. F. Skinner’s Verbal Behavior. Language, 35(1), 26–58. https://doi.org/10.2307/411334

Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 181–204. https://doi.org/10.1017/S0140525X12000477

Cowan, N. (2017). The many faces of working memory and short-term storage. Psychonomic Bulletin & Review, 24(4), 1158–1170. https://doi.org/10.3758/s13423-016-1191-6

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

Davidson, M. C., Amso, D., Anderson, L. C., & Diamond, A. (2006). Development of cognitive control and executive functions from 4 to 13 years: Evidence from manipulations of memory, inhibition, and task switching. Neuropsychologia, 44(11), 2037–2078. https://doi.org/10.1016/j.neuropsychologia.2006.02.006

De Neys, W. (2023). Advancing theorizing about fast-and-slow thinking. Behavioral and Brain Sciences, 46, Article e111. https://doi.org/10.1017/S0140525X2200142X

Deutsch, J. A., & Deutsch, D. (1963). Attention: Some theoretical considerations. Psychological Review, 70(1), 80–90. https://doi.org/10.1037/h0039515

Diamond, A. (2013). Executive functions. Annual Review of Psychology, 64, 135–168. https://doi.org/10.1146/annurev-psych-113011-143750

Evans, J. St. B. T., & Stanovich, K. E. (2013). Dual-process theories of higher cognition: Advancing the debate. Perspectives on Psychological Science, 8(3), 223–241. https://doi.org/10.1177/1745691612460685

Flavell, J. H. (1979). Metacognition and cognitive monitoring: A new area of cognitive–developmental inquiry. American Psychologist, 34(10), 906–911. https://doi.org/10.1037/0003-066X.34.10.906

Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138. https://doi.org/10.1038/nrn2787

Gigerenzer, G., & Goldstein, D. G. (1996). Reasoning the fast and frugal way: Models of bounded rationality. Psychological Review, 103(4), 650–669. https://doi.org/10.1037/0033-295X.103.4.650

Henke, K. (2010). A model for memory systems based on processing modes rather than consciousness. Nature Reviews Neuroscience, 11(7), 523–532. https://doi.org/10.1038/nrn2850

Hommel, B., Chapman, C. S., Cisek, P., Neyedli, H. F., Song, J.-H., & Welsh, T. N. (2019). No one knows what attention is. Attention, Perception, & Psychophysics, 81(7), 2288–2303. https://doi.org/10.3758/s13414-019-01846-w

Kahneman, D. (2003). A perspective on judgment and choice: Mapping bounded rationality. American Psychologist, 58(9), 697–720. https://doi.org/10.1037/0003-066X.58.9.697

Kriegeskorte, N., & Douglas, P. K. (2018). Cognitive computational neuroscience. Nature Neuroscience, 21(9), 1148–1160. https://doi.org/10.1038/s41593-018-0210-5

Lake, B. M., Ullman, T. D., Tenenbaum, J. B., & Gershman, S. J. (2017). Building machines that learn and think like people. Behavioral and Brain Sciences, 40, Article e253. https://doi.org/10.1017/S0140525X16001837

Lavie, N. (1995). Perceptual load as a necessary condition for selective attention. Journal of Experimental Psychology: Human Perception and Performance, 21(3), 451–468. https://doi.org/10.1037/0096-1523.21.3.451

Lindenberger, U., von Oertzen, T., Ghisletta, P., & Hertzog, C. (2011). Cross-sectional age variance extraction: What’s change got to do with it? Psychology and Aging, 26(1), 34–47. https://doi.org/10.1037/a0020525

Mahon, B. Z., & Caramazza, A. (2008). A critical look at the embodied cognition hypothesis and a new proposal for grounding conceptual content. Journal of Physiology-Paris, 102(1–3), 59–70. https://doi.org/10.1016/j.jphysparis.2008.03.004

Marr, D. (2010). Vision: A computational investigation into the human representation and processing of visual information. MIT Press. (Original work published 1982) https://doi.org/10.7551/mitpress/9780262514620.001.0001

Melnikoff, D. E., & Bargh, J. A. (2018). The mythical number two. Trends in Cognitive Sciences, 22(4), 280–293. https://doi.org/10.1016/j.tics.2018.02.001

Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Annual Review of Neuroscience, 24, 167–202. https://doi.org/10.1146/annurev.neuro.24.1.167

Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. https://doi.org/10.1037/h0043158

Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive Psychology, 41(1), 49–100. https://doi.org/10.1006/cogp.1999.0734

Neisser, U. (2014). Cognitive psychology (Classic ed.). Psychology Press. (Original work published 1967) https://doi.org/10.4324/9781315736174

Oberauer, K., Lewandowsky, S., Awh, E., Brown, G. D. A., Conway, A., Cowan, N., Donkin, C., Farrell, S., Hitch, G. J., Hurlstone, M. J., Ma, W. J., Morey, C. C., Nee, D. E., Schweppe, J., Vergauwe, E., & Ward, G. (2018). Benchmarks for models of short-term and working memory. Psychological Bulletin, 144(9), 885–958. https://doi.org/10.1037/bul0000153

Salthouse, T. A. (1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103(3), 403–428. https://doi.org/10.1037/0033-295X.103.3.403

Salthouse, T. A. (2019). Trajectories of normal cognitive aging. Psychology and Aging, 34(1), 17–24. https://doi.org/10.1037/pag0000288

Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information processing: I. Detection, search, and attention. Psychological Review, 84(1), 1–66. https://doi.org/10.1037/0033-295X.84.1.1

Squire, L. R. (2004). Memory systems of the brain: A brief history and current perspective. Neurobiology of Learning and Memory, 82(3), 171–177. https://doi.org/10.1016/j.nlm.2004.06.005

Stern, Y., Arenaza-Urquijo, E. M., Bartrés-Faz, D., Belleville, S., Cantilon, M., Chetelat, G., Ewers, M., Franzmeier, N., Kempermann, G., Kremen, W. S., Okonkwo, O., Scarmeas, N., Soldan, A., Udeh-Momoh, C., Valenzuela, M., Vemuri, P., Vuoksimaa, E., & the Reserve, Resilience and Protective Factors PIA Empirical Definitions and Conceptual Frameworks Workgroup. (2020). Whitepaper: Defining and investigating cognitive reserve, brain reserve, and brain maintenance. Alzheimer’s & Dementia, 16(9), 1305–1311. https://doi.org/10.1016/j.jalz.2018.07.219

Stern, Y. (2002). What is cognitive reserve? Theory and research application of the reserve concept. Journal of the International Neuropsychological Society, 8(3), 448–460. https://doi.org/10.1017/S1355617702813248

Tucker-Drob, E. M., Brandmaier, A. M., & Lindenberger, U. (2019). Coupled cognitive changes in adulthood: A meta-analysis. Psychological Bulletin, 145(3), 273–301. https://doi.org/10.1037/bul0000179

Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131. https://doi.org/10.1126/science.185.4157.1124

Walsh, K. S., McGovern, D. P., Clark, A., & O’Connell, R. G. (2020). Evaluating the neurophysiological evidence for predictive processing as a model of perception. Annals of the New York Academy of Sciences, 1464(1), 242–268. https://doi.org/10.1111/nyas.14321

Wilson, M. (2002). Six views of embodied cognition. Psychonomic Bulletin & Review, 9(4), 625–636. https://doi.org/10.3758/BF03196322