Cognitive Psychology

Abstract

Cognitive psychology is the scientific study of the mental processes that intervene between stimulus and response — attention, perception, memory, language, and thinking. It emerged in the 1950s and 1960s as the cognitive revolution displaced behaviorism, reclaiming the internal events the earlier program had ruled out of bounds and modeling the mind as a limited-capacity processor of information. Its defining framework treats cognition as a sequence of representational stages through which information flows, and its defining method, mental chronometry, uses response time to decompose those stages. This article traces the field from its founding through the information-processing framework, surveys its core domains, and follows its computational turn, in which Bayesian models and connectionist learning systems recast cognition as inference and adaptation under real-world limits. Interactive demonstrations illustrate mental chronometry, mental rotation, and the capacity limits of short-term memory.

Keywords: cognitive psychology, information processing, mental chronometry, attention, memory

Cognitive psychology is the branch of psychology that studies how the mind acquires, represents, transforms, stores, and uses information. Where the behaviorism that preceded it confined the science to observable stimulus-response relations, cognitive psychology takes the intervening mental processes themselves as its subject, treating them as real, structured, and open to experiment (Neisser, 1967). Its rise was less a single discovery than a convergence: work in information theory, linguistics, and the new science of computation supplied a common vocabulary in which the mind could be described as a processor of information, and a generation of experiments showed that internal operations leave measurable traces in behavior (Miller, 1956; Simon, 1955). The field it created remains the core of how psychology explains the thinking human being.

What Cognitive Psychology Is

Cognitive psychology occupies the middle ground between the stimulus that reaches the senses and the behavior that eventually follows, and it insists that this middle ground has a structure worth describing in its own right. The guiding metaphor is that the mind is an information-processing system: sensory input is encoded into internal representations, those representations are transformed by a series of operations, and the results are stored, retrieved, and used to guide action (Neisser, 1967). A representation is simply an internal stand-in for something in the world, and a process is an operation that creates, changes, or acts on one. The field's explanatory task is to specify the representations the mind uses and the processes that manipulate them, and to test those specifications against behavior.

Figure 1

The Mind as an Information-Processing System

This commitment marks cognitive psychology off from the two positions it grew up between. Against behaviorism, it holds that the internal events discarded as unobservable are precisely what a science of mind must explain, and that they can be inferred rigorously from patterns in behavior even when they cannot be seen directly (Neisser, 1967). Against introspectionism, the older program it also rejected, it holds that those events are not to be read off from conscious report, since much of cognition is rapid, automatic, and unavailable to awareness. The route to the mind runs instead through controlled experiment, in which the structure of the unseen process is reconstructed from the timing, accuracy, and errors of observable performance.

The Cognitive Revolution

For the first half of the twentieth century, behaviorism held that a science of psychology could admit only what could be observed, and that talk of internal mental states was unscientific. The position was productive but confining, and by the 1950s it was straining against phenomena it could not comfortably address, above all the structure of language. Three developments then converged to break it. Information theory offered a precise way to quantify how much a signal carries and how much a channel can transmit; the new theory of computation showed that symbol manipulation could be described exactly, so that thinking might be a species of computation; and linguistics demonstrated that the productivity of language demands internal rules rather than chains of conditioned responses. Together they made internal representation respectable again.

The revolution's landmarks came in quick succession. George Miller argued that the span of immediate memory is limited to about seven items, but that the limit is defined over chunks — meaningful units the observer constructs — so that recoding expands what a fixed capacity can hold (Miller, 1956). Noam Chomsky's review of B. F. Skinner's Verbal Behavior dealt the behaviorist account of language its decisive blow, arguing that the unbounded productivity of speech — the endless supply of novel sentences a child comes to produce and understand — cannot arise from conditioned response chains and instead requires a system of internalized generative rules (Chomsky, 1959). Herbert Simon argued that human rationality is bounded — that a real mind reasons within the limits of finite time, knowledge, and computation rather than reaching the unconstrained optimum of ideal theory (Simon, 1955). The synthesis arrived in 1967, when Ulric Neisser gathered the scattered experimental work under a single banner and gave the field its name and its charter, defining cognitive psychology as the study of how sensory input is transformed, reduced, elaborated, stored, recovered, and used (Neisser, 1967). Within a decade the study of the mind had been re-established as an experimental science. Table 1 sets the displaced program against the one that replaced it.

Table 1
Behaviorism and Cognitive Psychology Compared
Note. The cognitive revolution did not deny the behaviorists' data but enlarged the science to include the internal processes they had set aside (Miller, 1956; Neisser, 1967).

The Information-Processing Framework

The framework that organized the new field treats cognition as the flow of information through a series of stages, each performing a distinct operation and each subject to its own limits. Its canonical statement is the modal model of Richard Atkinson and Richard Shiffrin, which fixed the architecture that Figure 1 depicts: sensory information is briefly registered, selectively attended, and passed into a limited-capacity short-term store, from which a set of control processes such as rehearsal governs whether it is transferred to a durable long-term store and later retrieved (Atkinson & Shiffrin, 1968). The model's lasting contribution was to separate the fixed structural stores from the strategic control processes that operate within them, a distinction later refinements preserved even as they replaced the passive short-term buffer with an active working memory. The power of the approach lies not in the boxes but in its testability: if cognition really is organized into stages, then experimental manipulations that target different stages should have separable, measurable effects on performance.

Saul Sternberg turned this idea into a precise tool. In his memory-scanning task, participants hold a short set of items in memory and judge whether a probe belongs to it; the striking result is that response time rises linearly with the size of the memorized set, and by the same slope whether the probe is present or absent (Sternberg, 1966). The linearity implies that the comparison is made item by item, and the equal slopes imply that the scan is exhaustive rather than self-terminating, because a search that stopped on a match would be faster on present trials. From the shape of a reaction-time function alone, Sternberg thus inferred the hidden organization of a process no one can observe — the founding demonstration of mental chronometry, the use of time to expose the architecture of thought. The demonstration below varies the set size and the per-item comparison time to trace the linear scanning function the model predicts.

The additive-factors logic Sternberg introduced generalizes far beyond his task. If two experimental factors affect different stages, their effects on total response time should add independently; if they affect a common stage, they should interact. The pattern of additivity and interaction across manipulations therefore becomes evidence about how many stages there are and what each one does. This inferential strategy — reconstructing an unobservable processing architecture from the fingerprints it leaves in timing and accuracy — is the methodological signature of cognitive psychology, and it recurs across every domain the field studies.

Core Domains of Cognition

The information-processing framework is realized differently in each of the mind's faculties, and cognitive psychology's substance lies in the detailed models it has built for them. In attention, Anne Treisman's feature-integration theory proposed that the visual system registers simple features such as color and orientation in parallel across the field, but that focused attention is required to bind those features into an object at a given location; the theory explains why a target defined by a single feature pops out effortlessly while one defined by a conjunction of features must be searched for serially (Treisman & Gelade, 1980). Michael Posner's work recast attention as a set of separable networks — for alerting, for orienting to locations, and for executive control — each with its own operations and neural substrate, giving the informal notion of paying attention a decomposable structure (Posner & Petersen, 1990).

In perception and mental imagery, Roger Shepard and Jacqueline Metzler produced one of the field's most elegant results. Asked whether two pictured three-dimensional shapes are the same object seen from different angles, people take longer to decide the more the shapes differ in orientation, and the time rises linearly with the angular difference — as though they rotate a mental image at a constant rate to bring the shapes into alignment (Shepard & Metzler, 1971). The finding gave imagery a claim to being an analog process with genuine spatial structure, not merely a figure of speech. The demonstration below reproduces the linear relationship between angular disparity and rotation time.

In memory, cognitive psychology drew distinctions that reorganized the field. Endel Tulving separated episodic memory, the record of experienced events located in time and place, from semantic memory, the store of general knowledge stripped of its learning context, arguing that they are functionally distinct systems (Tulving, 1985). The structure of semantic memory itself was probed by Allan Collins and Ross Quillian, who showed that the time to verify a fact such as a canary can fly depends on the conceptual distance the query must traverse in a network of concepts, evidence that knowledge is stored in an organized, hierarchical structure rather than as an unstructured list (Collins & Quillian, 1969). And the limited-capacity store at the center of the framework has itself been reconceived. Alan Baddeley and Graham Hitch recast it as working memory — not a passive buffer but a multi-component system that actively holds and manipulates information, with separate subsystems for verbal and visuospatial content coordinated by a central executive (Baddeley & Hitch, 1974); more recent work suggests that even this store is less a fixed location than a distributed activity, its contents held across the same sensory and association areas that process the material in the first place (Christophel et al., 2017). Miller's insight that capacity is defined over chunks, not raw items, remains the practical key to that limit; the demonstration below shows how grouping the same material into fewer, larger units brings an over-long sequence back within span (Miller, 1956).

In thinking and reasoning, the field mapped the systematic ways human judgment departs from formal norms. Amos Tversky and Daniel Kahneman showed that people judge probability and frequency by a small set of heuristics — representativeness, availability, anchoring — that are efficient but produce predictable biases, so that the errors themselves become a window onto the underlying processes (Tversky & Kahneman, 1974). Across these domains the pattern is the same: a well-chosen experiment converts an invisible mental operation into a measurable regularity, and the model that predicts the regularity is the field's product.

The Computational Turn

The computer metaphor that launched cognitive psychology has since become a working method, as the field increasingly states its theories as explicit computational models and tests them against detailed behavior. Two traditions dominate. The Bayesian or probabilistic approach treats cognition as inference under uncertainty: the mind combines prior expectations with incoming evidence to infer the hidden structure of the world, and many perceptual, memory, and reasoning phenomena fall out as approximately optimal solutions to the inference problem the environment poses. The connectionist approach models cognition as the propagation of activation through networks of simple units whose connection strengths are gradually tuned by experience, capturing how graded, statistical knowledge is learned and generalized.

The modern synthesis of these strands is a central current in the field. Brenden Lake and colleagues argue that human learning still outstrips even powerful artificial systems in exactly the respects cognitive psychology has long emphasized: people build structured, causal, compositional models of the world and learn new concepts from very few examples, and a genuine theory of intelligence must explain those abilities rather than only pattern classification (Lake et al., 2017). From the other direction, Matthew Botvinick and colleagues show how the slow, statistical learning of deep networks can be reconciled with the fast, one-shot learning that cognition also displays, by equipping learning systems with memory and with the capacity to learn how to learn (Botvinick et al., 2019). Running through both is a renewed appreciation of the point Simon first pressed: that intelligence is shaped by the limits within which it operates, so that many apparent quirks of human thought are the rational adaptations of a system making the best use of bounded time and computation (Griffiths, 2020). The computer metaphor has thus come full circle, from a loose analogy that made the mind respectable to study into a precise instrument for saying what the mind computes.

Worked Example

Mental chronometry turns the architecture of a hidden process into arithmetic. In Sternberg's memory-scanning task, the total response time is modeled as a sum of stage durations: a fixed overhead for encoding the probe and executing the response, plus a comparison time that is paid once for every item in the memorized set. Writing the overhead as an intercept and the per-item comparison as a slope, the model is simply RT = intercept + slope times set size.

Suppose the encoding-and-response overhead is 400 milliseconds and each comparison takes 38 milliseconds. For a memorized set of four items, the model predicts a response time of 400 + 38 times 4 = 400 + 152 = 552 milliseconds. For a set of two items it predicts 400 + 38 times 2 = 476 milliseconds, and for six items 400 + 38 times 6 = 628 milliseconds. The crucial quantity is not any single time but the slope: the difference between the six-item and two-item predictions is 628 minus 476 = 152 milliseconds, which is exactly 38 milliseconds for each of the four additional items. Because that per-item cost stays constant as the set grows, and because it is the same whether the probe is in the set or not, the data imply a serial, exhaustive scan — a conclusion about unseen mental structure read directly off the slope of a line. The mental-chronometry demonstration above computes this function for any overhead, comparison time, and set size, and plots the linear scanning function the model predicts.

Discussion

Cognitive psychology succeeded because it found a disciplined way to study the unobservable. Its wager — that the mental processes behaviorism had banned could be inferred rigorously from the timing, accuracy, and errors of behavior — proved extraordinarily productive, yielding precise, quantitative models of attention, perception, memory, and thought where the previous program had offered only stimulus-response associations (Neisser, 1967; Sternberg, 1966). The information-processing framework gave these models a common form, and mental chronometry gave them a common currency, so that findings from different laboratories and different faculties could be compared and combined (Miller, 1956).

The framework's limits have shaped its later development. The strict stage models of the 1960s, with information passed forward through discrete boxes, gave way to accounts in which processing is graded, interactive, and distributed, and in which the sharp line between memory and processing blurs (Christophel et al., 2017). The heuristics-and-biases tradition complicated the picture of the mind as a general-purpose computer, showing that human judgment runs on efficient shortcuts whose systematic errors reveal the machinery beneath (Tversky & Kahneman, 1974). And the field's alliance with neuroscience has grounded its abstract stages in measurable neural systems, most clearly in the decomposition of attention into distinct networks (Posner & Petersen, 1990). What has not changed is the core commitment: that the mind is an information-processing system whose representations and processes are the proper objects of an experimental science.

Current Directions

The field's most active frontier is computational, and it revives an old idea in a new key. The resource-rational program treats the mind as making the best use of limited computation, deriving the strategies a bounded but rational agent should adopt given the cost of thinking, so that the heuristics and biases catalogued half a century ago reappear as optimal adaptations to real constraints rather than as defects (Griffiths, 2020). The program restates Simon's bounded rationality in the precise language of expected value of computation, and it connects the levels cognitive psychology had studied separately, letting a claim about a memory limit or an attentional bottleneck be derived rather than merely described.

A second direction runs between cognitive psychology and artificial intelligence, in both directions at once. Findings about human learning — its reliance on structured, causal, compositional models and its capacity to generalize from a handful of examples — are used to diagnose what machine learning still lacks, while advances in machine learning supply new, testable models of how human cognition might be implemented (Lake et al., 2017; Botvinick et al., 2019). The traffic has made the boundary between the two fields increasingly porous, and it has returned cognitive psychology to its founding ambition: not merely to catalogue the mind's operations but to say, precisely enough to build, what those operations compute.

Glossary

Additive-factors method.

Sternberg's inferential strategy in which two manipulations that affect separate processing stages add independently in response time, while two that affect a shared stage interact, revealing the number and identity of the stages.

Attention.

The set of processes that select some information for privileged processing while filtering the rest; treated in cognitive psychology as a limited resource with separable alerting, orienting, and executive components.

Behaviorism.

The earlier program that restricted psychology to observable stimulus-response relations and excluded internal mental states; the position the cognitive revolution displaced.

Bounded rationality.

Simon's principle that real minds reason within the limits of finite time, knowledge, and computation rather than achieving unconstrained optimality.

Chunk.

A meaningful unit into which several items are recoded, so that the fixed capacity of immediate memory, defined over chunks rather than raw items, can hold more information.

Cognitive revolution.

The mid-twentieth-century shift, driven by information theory, linguistics, and computation, that displaced behaviorism and re-established internal mental processes as the subject of an experimental psychology.

Connectionism.

The approach that models cognition as activation spreading through networks of simple units whose connection strengths are gradually tuned by experience, capturing graded, statistical learning.

Episodic memory.

Tulving's term for memory of specific experienced events located in time and place, distinguished from general knowledge.

Feature-integration theory.

Treisman's account in which simple visual features are registered in parallel across the field but focused attention is required to bind them into an object.

Information-processing framework.

The organizing model of cognitive psychology that treats cognition as a sequence of limited-capacity stages through which information is encoded, transformed, stored, and retrieved.

Mental chronometry.

The use of response time to measure the duration and organization of mental processes, the signature method by which cognitive psychology infers unobservable processing structure.

Mental rotation.

The imagined turning of a visual object, evidenced by response times that rise linearly with the angular disparity between two shapes, taken as evidence that imagery is an analog spatial process.

Modal model.

Atkinson and Shiffrin's multi-store account of memory, which separates fixed structural stores — sensory register, short-term store, long-term store — from the strategic control processes, such as rehearsal, that govern the flow of information between them.

Representation.

An internal stand-in for something in the world; specifying the representations the mind uses and the processes that manipulate them is cognitive psychology's central explanatory task.

Resource-rational analysis.

A framework that derives the strategy a rational but computationally limited agent should adopt, recovering observed heuristics and biases as optimal uses of bounded resources.

Semantic memory.

The store of general knowledge and facts, stripped of the context in which it was learned; probed through the time to verify statements across a network of concepts.

Working memory.

The limited-capacity system that holds and manipulates information in the service of ongoing cognition; now viewed as a distributed activity across sensory and association areas rather than a single buffer.

Key Researchers

Matthew M. Botvinick (b. 1968). Senior Director of Research at Google DeepMind and Honorary Professor at the Gatsby Computational Neuroscience Unit, University College London; he works at the boundary of cognitive psychology and machine learning, showing how the fast and slow forms of learning can be reconciled in a single system. Google Scholar - ORCID

Thomas L. Griffiths (b. 1979). Professor of Psychology and Computer Science at Princeton University; he develops Bayesian and resource-rational models of cognition, deriving human strategies and biases from the assumption that the mind makes optimal use of limited computation. Wikipedia - ORCID

Daniel Kahneman (1934-2024). Nobel laureate and Eugene Higgins Professor of Psychology, Emeritus, at Princeton University; with Tversky he mapped the heuristics and biases of judgment under uncertainty, showing that systematic error reveals the machinery of thought. Wikipedia - Google Scholar

George A. Miller (1920-2012). A founder of cognitive science and psycholinguistics at Princeton and Harvard; his analysis of the magical number seven established that the span of immediate memory is defined over chunks, and he later created the WordNet lexical database. Wikipedia

Ulric Neisser (1928-2012). Sometimes called the father of cognitive psychology; his 1967 textbook named and organized the field, and his later work pressed it toward ecological validity and real-world cognition. Wikipedia - Google Scholar

Michael I. Posner (b. 1936). Professor Emeritus of Psychology at the University of Oregon; a pioneer of mental chronometry, he decomposed attention into separable alerting, orienting, and executive networks with distinct neural substrates. Wikipedia - Google Scholar

Herbert A. Simon (1916-2001). Nobel laureate and Turing Award recipient at Carnegie Mellon University; he framed bounded rationality and, with Newell, the information-processing account of problem solving that gave cognitive psychology its computational vocabulary. Wikipedia - Google Scholar

Joshua B. Tenenbaum (b. 1972). Professor of Computational Cognitive Science at the Massachusetts Institute of Technology; he builds probabilistic models of how people learn structured, compositional concepts from very few examples. Wikipedia - ORCID

Anne Treisman (1935-2018). Professor of Psychology at Princeton University and recipient of the National Medal of Science; her feature-integration theory became the dominant account of how attention binds visual features into objects. Wikipedia

Endel Tulving (1927-2023). University Professor Emeritus at the University of Toronto; he drew the episodic-semantic distinction that reorganized the psychology of memory. Wikipedia - Google Scholar

Amos Tversky (1937-1996). Professor of Psychology at Stanford University; with Kahneman he identified the heuristics that govern judgment under uncertainty and the biases they systematically produce. Wikipedia

Frequently Asked Questions

What is cognitive psychology?
Cognitive psychology is the scientific study of mental processes such as attention, perception, memory, language, and thinking. It treats the mind as an information-processing system and reconstructs its unobservable operations from the timing, accuracy, and errors of behavior (Neisser, 1967).

How is cognitive psychology different from behaviorism?
Behaviorism restricted psychology to observable stimulus-response relations and set internal mental states outside science. Cognitive psychology takes those internal processes as its subject, holding that they can be inferred rigorously from behavior even when they cannot be observed directly (Neisser, 1967).

What was the cognitive revolution?
The cognitive revolution was the shift in the 1950s and 1960s in which information theory, linguistics, and the theory of computation gave psychology a way to describe internal representation. It displaced behaviorism and re-established the mind as an experimental subject (Miller, 1956).

What is the information-processing framework?
It is the model that treats cognition as information flowing through a series of stages, each with a limited capacity, that encode, transform, store, and retrieve it. Its value is that it is testable, because manipulations aimed at different stages produce separable effects (Sternberg, 1966).

What is mental chronometry?
Mental chronometry is the use of response time to measure the duration and organization of mental processes. Sternberg used it to show that memory scanning is serial and exhaustive, inferring hidden structure from the slope of a reaction-time function (Sternberg, 1966).

What are the core areas of cognitive psychology?
The main domains are attention, perception and mental imagery, memory, language, and thinking and reasoning. Each has detailed models, such as feature-integration theory in attention and the episodic-semantic distinction in memory (Treisman & Gelade, 1980; Tulving, 1985).

How does cognitive psychology relate to artificial intelligence?
The two fields exchange ideas: findings about human learning diagnose what machine systems still lack, while advances in machine learning supply testable models of how cognition might be implemented. Human learning still exceeds machines in building structured, compositional models from few examples (Lake et al., 2017).

What is the field working on now?
A leading current direction is resource-rational analysis, which derives the strategies a rational but computationally limited mind should adopt, recovering long-catalogued heuristics and biases as optimal adaptations to bounded resources rather than as defects (Griffiths, 2020).

References