Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

The Cognitive Revolution and Information-Processing Psychology

Abstract. This fundamental research monograph reconstructs the cognitive revolution as the historical convergence of experimental psychology, communication theory, cybernetics, linguistics, and computational modeling. Using original historical publications and critical narrative synthesis, it investigates how internal representations and transformations of information became testable explanatory objects. The analysis examines Bruner's concept attainment experiments; Newell, Shaw, and Simon's computational account of problem solving; language theory and critiques of verbal behavior; classic demonstrations in attention, iconic memory, short-term retention, mental scanning, and imagery; and subsequent disputes concerning working memory, levels of processing, distributed computation, ecological validity, and explanatory levels. The evidence supports a historical account of gradual convergence, methodological innovation, and continuing theoretical revision. It also identifies challenges of construct validity, model identifiability, experimental generalizability, and cross-system comparison in the Artificial Era.

Keywords: cognitive revolution; information-processing psychology; history of cognitive psychology; cognitive science; mental representation; information theory; experimental psychology; attention; memory systems; language; computer simulation; reaction time; connectionism; scientific historiography

The Cognitive Revolution as a Historical Research Problem

The cognitive revolution designates the mid-twentieth-century reorganization of research in which psychologists increasingly treated internal representations, information transformations, and goal-directed computations as legitimate explanatory objects. Its historical identity lies in a convergence of experimental psychology, mathematical communication theory, linguistics, cybernetics, and early computer science. It is therefore best reconstructed as a transformation in research questions, evidentiary practices, and institutional networks. Investigators began asking what operations intervene between an encountered signal and an observed response, how those operations can be measured indirectly, and whether a formal model can reproduce characteristic patterns of human performance. The conceptual history extends beyond celebrated meetings and beyond any single national research school. Miller (1956) Neisser (1967)

The phrase information processing acquired several meanings. In a narrow technical sense, it referred to operations applied to formally specified signals and representations; in experimental psychology, it organized hypotheses about attention, perception, short-term retention, recognition, and response selection. In artificial intelligence, it referred to implemented procedures that manipulate symbols or learned representations to perform tasks. These uses overlap, but their empirical commitments differ. A computer program that solves a logic problem demonstrates the sufficiency of a specified procedure for that solution; evidence that humans use that procedure requires additional behavioral comparison and experimental discrimination. Historical interpretation becomes misleading when an influential analogy is mistaken for a validated account of human mechanisms. Newell et al. (1958) Turing (1950)

This monograph treats cognitive psychology as a historically formed experimental program. It reconstructs its antecedents, landmark models, testing practices, conceptual disagreements, and later revisions. Evidence comes from primary publications and original books wherever they could be identified and checked, supplemented by subsequent theoretical work on representation, neural implementation, and situated cognition. The account examines how a community developed researchable concepts, rather than assigning a ceremonial birthday to an entire field. Its primary question is how internal mental operations became answerable to public measurement, model comparison, and reproducible experimental predictions. Bruner et al. (1956) Broadbent (1958) Marr (1982)

Reconstructing the Prehistory of Cognitive Inquiry

Long before the phrase cognitive revolution became common, experimental psychologists investigated processes that could not be read directly from observable behavior. Studies of perception, memory, skilled action, and inference depended on assumptions about latent operations. Bartlett's account of remembering emphasized reconstruction, prior knowledge, and social meaning; remembered material could be systematically transformed according to a person's organized expectations. Such work provides a historical foundation for later schema theories and for debates about the role of interpretation in memory. Its procedures and conceptual vocabulary also differed from the short-duration laboratory tasks that would become emblematic of the information-processing era. Bartlett (1932)

Edward Tolman's discussion of cognitive maps, published within the broader experimental and learning-theory landscape, argued that patterned navigation can be understood through spatial organization and expectations. His position illustrates why a simple replacement story—behaviorists studied only stimulus-response chains until cognitivists invented internal explanations—cannot account for the period's actual diversity. Learning experiments, comparative behavior, and purposive accounts already contained hypotheses about mediating processes. The subsequent revolution enlarged the formal apparatus used to describe these hypotheses and helped make intermediate structures experimentally tractable. Its novelty thus concerned the forms of explanation and integration available to researchers as much as the mere mention of mentality. Tolman (1948)

The institutional contrast between behaviorism and cognitivism also requires attention to different forms of behaviorism. Research on reinforcement, discrimination, verbal responses, and sequential behavior used controlled observations and quantitative predictions that cognitive scientists retained. An experimentalist could reject a particular account of language while preserving the methodological value of controlled intervention and replicable dependent measures. Reaction time, errors, recognition, response probabilities, and transfer performance became shared evidence across schools. Cognitive researchers used these observations to adjudicate between alternative hidden operations. This continuity helps explain why the field's strongest experiments joined theoretical novelty to procedures inherited from preceding experimental traditions. Chomsky (1959) Sternberg (1966)

Mathematical Communication Theory and the Channel Metaphor

Claude Shannon's 1948 mathematical theory specified information in relation to the uncertainty of messages and the properties of a communication channel. It was designed to analyze transmission and coding; its entropy measure abstracts from the meaning of a message to the probabilities of possible signals. Cognitive researchers borrowed concepts of capacity, noise, coding, and bottlenecks to formulate experimental questions about limited human performance. The borrowing proved fruitful when a psychological task could be represented by explicit alternatives and experimentally varied information. It became more tenuous when formal transmission measures were treated as direct indices of understanding, intention, or semantic depth. Shannon (1948) Miller (1956)

Miller's 1956 synthesis connected absolute judgment, memory span, and recoding. The famous numerical phrase in its title was a rhetorical point of entry into several different capacity problems rather than a universal constant governing all cognition. Absolute judgments of stimulus magnitudes, span for familiar units, and the use of meaningful chunks involve different operations and measurement scales. The empirical importance of the paper lay in its insistence that coding choices alter effective task complexity. People can reorganize the units of a representation and thereby change how much relevant structure can be handled. A statement about item count requires specification of material, grouping, modality, and procedure. Miller (1956)

Information theory offered a precision that earlier psychological vocabularies often lacked, but the analogy became useful only through task-specific operationalization. Researchers had to specify inputs, observable outputs, hypothesized stages, and consequences for performance when a parameter changed. Channel capacity might predict selection failures in one auditory task without explaining the acquisition of language rules or reconstruction of an autobiographical episode. The later history of cognitive psychology accordingly reveals a progression from broad analogical language toward distinguishable process models. Channel metaphors continued to organize investigation, even while representational meaning and individual strategies demanded additional explanatory resources. Shannon (1948) Broadbent (1958) Craik & Lockhart (1972)

Cybernetics, Neural Logic, and Computable Procedures

The mathematical investigation of feedback and the computational study of logical operations formed an important prehistory of information-processing psychology. McCulloch and Pitts developed a formal calculus for networks of idealized neural units, showing how networks under specified assumptions could implement logical relationships. The paper established a conceptual bridge between neural activity and computable transformations. Its models were theoretical constructions with strong simplifying assumptions; they supplied a research language for mechanism without providing empirical evidence that actual human thought follows the particular operations in those idealized nets. McCulloch & Pitts (1943)

Turing's 1950 discussion of computing machinery and intelligence treated machines as systems whose behavior could be evaluated through specified tasks and proposed learning procedures. The argument extended a formal tradition in which a procedure can be analyzed independently of the material substrate that executes it. In cognitive psychology, this distinction made it possible to ask whether a capacity such as rule use or problem solving can be described computationally while leaving open which neuronal, bodily, and environmental mechanisms implement it. The distinction also creates a methodological obligation: similar output on a benchmark does not establish similar internal computation. Turing (1950)

The cybernetic influence entered models of action as well as accounts of perception. Miller, Galanter, and Pribram described plans and feedback-regulated behavior in a way that replaced a purely linear response chain with organized control loops. Their Test–Operate–Test–Exit unit expressed a recurring adjustment between a desired condition, an operation, and assessment of progress. This architecture remains valuable as an early conceptual model of purposive behavior and monitoring. Yet its explanatory reach varies with the task, because experimental evidence must identify relevant goals, comparison signals, and intermediate stages. Historical significance resides in the move toward explicit functional organization, which prepared later cognitive theories of planning and executive control. Miller et al. (1960) Newell et al. (1958)

The Mid-1950s Convergence and Its Retrospective Mythology

The mid-1950s brought together researchers from psychology, linguistics, communication engineering, and emerging computer science. Their publications made internal representations more visible as testable theoretical proposals. Bruner, Goodnow, and Austin investigated strategies in concept attainment; Newell, Shaw, and Simon formulated information-processing models of problem solving; Miller examined the implications of coding and capacity for human performance. These developments shared a problem-solving orientation and a preference for explicit models, but they did not share one comprehensive theory of mind. The historical pattern is a convergence of partially independent programs with different experimental subjects and validation criteria. Bruner et al. (1956) Newell et al. (1958) Miller (1956)

Popular histories sometimes elevate a single 1956 symposium into the founding event of cognitive science. Such commemorative dates are useful for reconstructing the participants' intellectual networks; they offer weaker evidence for a singular abrupt disciplinary break. Foundational papers appeared before and after that year, and the expression cognitive psychology became increasingly institutionalized over the following decade. The historical task is to track the circulation of mathematical concepts, experimental paradigms, books, laboratories, and educational programs. One can identify the emergence of a collective field without imagining that its practitioners agreed at once on what cognition was or how it should be measured. Neisser (1967) Miller et al. (1960)

The account is strengthened by considering the work's reception. Information-processing models found opportunities in attention, language, and memory because they produced observable distinctions between competing explanations. Some theories forecast serial search, others selective filtering, and others multiple functionally differentiated retention systems. Their value could be challenged by manipulating stimulus presentation, retrieval delay, or task demands. Subsequent studies revised the models and often retained the experimental problems that first motivated them. The revolution is consequently a history of progressively refined tests rather than a list of canonical personalities. Its institutions mattered because they made these questions durable research programs that could survive the modification of their original models. Broadbent (1958) Atkinson & Shiffrin (1968) Sternberg (1966)

Bruner and the Experimental Study of Concept Formation

Bruner, Goodnow, and Austin's A Study of Thinking treated category learning as an active problem of discovering and using attributes. Participants facing concept-attainment tasks could select informative instances, test hypotheses, revise a provisional category, and compare alternative strategies. This made the steps of thinking available to controlled experimental observation without reducing an explanation to the final number of correct responses. The investigators' questions concerned the organization of inquiry: which examples are selected, which attributes become diagnostic, and how a subject's prior hypothesis determines the information gained from a new instance. Bruner et al. (1956)

The concept-learning paradigm connected experimental cognition to the logic of scientific inference. A participant working through category cards effectively chooses tests of a hypothesis, and a researcher can compare the predicted behavior of different selection strategies. Yet the analogy to natural reasoning has limits. Laboratory categories often have stable rules, explicit feedback, and sharply defined boundaries; everyday concepts depend on uncertain relevance, social practices, prototypes, and changing contexts. Recognizing these differences allows the original evidence to establish what it actually supports: strategy-sensitive performance under controlled categorization demands. It does not establish one unique or context-independent algorithm for human conceptual knowledge. Bruner et al. (1956) Neisser (1976)

The methodological novelty was the transition from a gross learning curve to an account of internal information seeking. The experimental subject became an investigator of a constrained environment, while the psychologist modeled the subject's choices as sequences carrying information about working hypotheses. This orientation helped connect categorization to later research on attention, reasoning, and memory. A historical examination should preserve the details of the experimental materials, the participant instructions, and the relation between successful performance and inferred strategy. Several distinct strategies may produce the same correct answer, so explanatory identification requires studying search order, errors, and response transitions rather than outcome accuracy alone. Bruner et al. (1956) Newell et al. (1958)

Language, Generative Structure, and the Critique of Verbal Behavior

The place of linguistics in the cognitive revolution derives from a problem of explanatory scope. A theory of language must account for structured sequences, productivity, and the relationship between grammatical principles and utterances people have not previously encountered. Chomsky's 1959 review of Skinner's Verbal Behavior argued that terms derived from behavioral explanation were insufficiently specified when transferred to the complexities of ordinary language. This criticism became influential because it raised precise questions about what a psychological theory of linguistic competence must explain. The dispute concerned theoretical categories and explanatory adequacy, rather than a single experimental result that settled all forms of learning theory. Chomsky (1959)

Formal grammar supplied a different style of representation: rules or constraints over symbolic structures. For psychology, such proposals motivated experiments on sentence comprehension, ambiguity, syntactic expectation, and the relation between stored knowledge and processing demands. However, evidence that language exhibits structural regularities does not identify one complete psychological architecture implementing them. Theoretical grammars characterize patterns of well-formedness; processing theories must account for actual time courses, memory limitations, errors, developmental change, and communicative context. Keeping these explanatory questions separate made the linguistic contribution productive and prevented formal description from being mistaken for a complete empirical model of comprehension. Chomsky (1959) Neisser (1967)

The historical consequence was a new intensity of interdisciplinary interaction. Linguists could propose formal systems; psychologists could investigate how real speakers acquired and used structured knowledge; computer scientists could implement candidate procedures. Difficult questions persisted about the relationship between abstract competence and performance, the role of environmental learning, and the effects of discourse context. Later distributed and statistical approaches would challenge strict commitments to hand-specified symbolic rules, yet the original question about linguistic productivity remained central. In this sense, the linguistic contribution to cognitive psychology is a continuing methodological demand to explain generalization from finite experience under measurable behavioral constraints. Chomsky (1959) Rumelhart et al. (1986) Neisser (1976)

Symbolic Simulation and the Theory of Human Problem Solving

Newell, Shaw, and Simon's 1958 program set out to explain human problem solving through explicit sequences of information-processing operations. A theory was expected to generate predicted solutions, intermediate actions, and characteristic failures under particular task conditions. This transformed the vague assertion that thinking involves internal processes into a model-building problem: specify the representation of a task, the possible operations, and the rules selecting among operations. Computer simulation became an instrument for checking whether the proposed mechanisms could, in principle, yield a detailed trajectory resembling observed behavior. Newell et al. (1958)

Symbolic models distinguished a problem's formal state space from the search procedure traversing it. A solver might have the correct operators available but follow an inefficient path because of heuristic choices or limited working resources. This perspective could explain why two problems with similar logical structure elicit different performance when their presentation changes. It also sharpened the status of verbal protocols and problem-solving traces as evidence. Reported thoughts and observed actions can constrain models, but they do not reveal every computational step directly. Experimental inference therefore requires comparison against alternative procedures that could yield the same verbal explanation or final answer. Newell et al. (1958) Miller et al. (1960)

The computer analogy was both a technical achievement and a methodological risk. A running program demonstrates internal consistency and algorithmic sufficiency, yet researchers must still show that its intermediate states, timing, learning, and errors correspond meaningfully to those of human participants. Different programs can solve identical tasks, and humans may reorganize representations in ways the model has not anticipated. The strongest contributions of symbolic simulation were the demand for explicit mechanisms and the capacity to compare predictions with detailed behavior. Those contributions survived the later expansion of neural-network modeling, even as the assumption of exclusively serial symbolic processing became disputed. Newell et al. (1958) Rumelhart et al. (1986) Marr (1982)

Plans, Hierarchies, and Feedback in Goal-Directed Action

Miller, Galanter, and Pribram's Plans and the Structure of Behavior described action as organized by internal representations of goals and by recurrent testing of progress toward those goals. Its vocabulary of plans, images, and monitoring placed purposeful behavior at the center of an information-processing account. The proposal challenged the sufficiency of a rigid one-directional sequence from stimulus to movement, because behavior frequently involves correcting actions in response to changing evidence. A person searching for an object, composing a message, or performing a practiced skill repeatedly evaluates consequences before continuing. The theoretical task is to model the operations and their ordering. Miller et al. (1960)

Hierarchical plans provided a way to connect large-scale purposes to subordinate procedures. A complex action could be broken into local tests and operations while remaining organized around a broader goal. This was a useful bridge between psychological intention and formal control descriptions. Still, a hierarchy in a theoretical diagram is not itself evidence of a unique biological structure. Real-world activities may be opportunistic, distributed across collaborators, sensitive to affordances, and continuously reorganized by environmental feedback. Historical interpretation should locate the plan framework among alternative accounts of skilled action and specify which observations favor a hierarchical description. Miller et al. (1960) Neisser (1976)

The test-and-feedback logic later influenced theories of executive control, self-monitoring, and automated planning. Its central contribution was to make the sequencing of behavior an object of explanation rather than a residual property of an unspecified response mechanism. For assessment, investigators can manipulate goals, introduce unexpected obstacles, measure correction latency, and record whether a participant switches among candidate procedures. Such studies can reveal functional organization even when the theoretical vocabulary changes. The plan framework therefore belongs in the history of cognitive psychology as one developmental line among several, particularly where cognition and voluntary action meet. Miller et al. (1960) Posner (1980) Newell et al. (1958)

Neisser and the Consolidation of Cognitive Psychology

Ulric Neisser's 1967 Cognitive Psychology brought together research on visual and auditory processing, pattern recognition, attention, verbal memory, language, and thought under a common intellectual identity. Its chapter organization is itself a historical document: it shows how experiments that had arisen in different laboratories could be read as investigating transformations of information. The book did not require every researcher to endorse one architecture. Instead, it provided a disciplined vocabulary and a map of investigable processes, demonstrating how a theory of cognitive activity could advance through converging experimental tasks. Neisser (1967)

A canonical textbook can shape a discipline by defining which problems appear fundamental and how evidence from different subfields should be related. Neisser's synthesis helped institutionalize terms that readers could use across perception, attention, and memory, but it also inevitably selected some problems and minimized others. Cultural practices, natural action, emotion, and developmental variation could recede from view when cognition was represented principally as a chain of transformations operating on simplified laboratory stimuli. These limitations became salient in Neisser's own later work and in broader attempts to return the study of cognition to environments that make action meaningful. Neisser (1976)

Cognition and Reality, published in 1976, criticized the ecological narrowness of some information-processing research and emphasized perception in real-world contexts. The revision is historically significant because it arose within the intellectual tradition that had advanced laboratory cognitive psychology. It demonstrates a capacity for internal criticism: mechanisms identified under controlled conditions must be evaluated for the domains to which researchers extend them. The subsequent growth of ecological, embodied, and situated approaches developed this concern in various ways. A responsible history of the revolution treats Neisser's two books as successive moments in a changing research program, with explanatory ideals and empirical standards under continuing revision. Neisser (1967) Neisser (1976) Clark (2013)

Selective Attention as a Problem of Information Intake

The selective-attention tradition gave information-processing psychology an experimentally tractable problem: when two streams compete for a listener's response, which features of the unattended stream are registered and which operations determine selection? Cherry's 1953 investigations of simultaneous speech supplied paradigms in which participants attended one message while another was presented concurrently or to a different ear. The dependent measures could reveal information about perceptual segregation, recognition, and the limits of report. Unlike introspective claims about attention as a unified faculty, the experiments manipulated signal conditions and examined their consequences for observable performance. Cherry (1953)

Broadbent's 1958 Perception and Communication incorporated such findings into a model emphasizing limited-capacity information flow and selective filtering. The framework sought connections among experimental listening, vigilance, human performance, and communication engineering. An early selection mechanism offered a strong theoretical constraint: some information would be filtered before it received the analysis needed for certain reports. What gave the proposal scientific force was its susceptibility to experimental challenge. If supposedly excluded information influenced performance under carefully controlled conditions, the model would require refinement, or researchers would have to discover an alternative account of the effect. Broadbent (1958)

Attention studies also illustrated that an experimental instruction is an active part of the mechanism under investigation. Telling participants to shadow one message, detect a target, or divide attention creates different processing goals. Differences in performance can reflect the properties of stimuli, the difficulty of discriminations, learned expectations, and strategies for allocating limited resources. The original filter image made these issues conspicuous but did not resolve all of them. A historical reading should therefore compare paradigms rather than treat selective attention as a single permanently localized gate. The advance was methodological: attention became a set of competing process hypotheses tied to measurable operations. Cherry (1953) Broadbent (1958) Treisman (1964)

The Attenuation Debate and Processing of Unattended Meaning

Treisman's 1964 experiments investigated whether an irrelevant message could be monitored or retained under dichotic listening. Conditions comparing related verbal material, voices, and temporal offsets challenged straightforward interpretations in which nonselected information was entirely eliminated before meaningful identification. The results encouraged models in which unattended input may undergo partial analysis and where selection varies with task context. The historical importance of the work lies in an evidentiary dispute about the locus and completeness of selection, rather than in the definitive replacement of every early-filter model by a single alternative. Treisman (1964) Broadbent (1958)

The competing accounts illustrate a common difficulty in cognitive experiments: an unreported item may have been unencoded, weakly encoded, forgotten, or inaccessible under the particular response procedure. Failure to reproduce a stimulus is an ambiguous observation unless alternative mechanisms are tested. In auditory attention, semantic overlap, repetition, expectation, and the intensity of unattended signals can interact with the participant's goals. Experiments gain discriminating power when they vary these properties while maintaining comparable perceptual demands. A theory that describes a stimulus as processed must specify what operation is claimed—acoustic discrimination, lexical identification, semantic activation, or voluntary report. Treisman (1964) Cherry (1953)

The filter debate was exemplary of cognitive psychology's movement toward explicit stages of processing. Stages are inferred from patterns of dissociation rather than observed as labeled components. A useful stage model predicts which manipulations should change the availability of information at a later point; its rivals may predict different effects of timing, selection instructions, or attentional load. Historical interpretation should distinguish evidence for some unattended processing from a much stronger conclusion that all unattended material undergoes full semantic analysis. The latter claim would exceed what many classic selective-listening experiments establish. Broadbent (1958) Treisman (1964) Posner (1980)

Iconic Memory and the Partial-Report Experiment

Sperling's 1960 experiments changed the question of how much a person can report from a brief visual presentation. Instead of interpreting whole-report performance as a direct measure of everything perceived, the partial-report procedure used a cue to request only a subset of a recently displayed array. The contrast between immediately cued partial report and subsequent report showed that briefly available visual information could exceed the quantity recovered in an ordinary full-report task. The experiments distinguished the availability of a transient representation from the ability to transfer its contents into a durable verbal response. Sperling (1960)

A brief visual display is a controlled instrument for probing time-dependent transformations. The timing of a post-display cue, array characteristics, luminance, verbal coding demands, and response order can be manipulated to test competing accounts of sensory persistence. The striking partial-report result supports a short-lived visual information store under specified conditions; it does not mean that fully formed and consciously reportable representations of every feature are available without limits. Sensory persistence, attentional selection, and report bottlenecks contribute to the observed performance. This conceptual precision matters because a popular description of a perfect photographic snapshot obscures the actual experimental inferences. Sperling (1960) Neisser (1967)

Sperling's paradigm became a model of how temporal interventions can reveal otherwise inaccessible processing stages. By placing a cue after the stimulus, the experiment altered what could be selected from a representation while holding the initial display constant. This design points toward a broader principle: a clever manipulation can distinguish properties of encoding from properties of retrieval. The evidence encouraged multi-component theories of memory and perception, but it also required researchers to keep separate the information theoretically available in an early trace and the subset demonstrably reported. Modern extensions depend on analogous control of masking, cue validity, task demands, and the observable criterion of access. Sperling (1960) Atkinson & Shiffrin (1968)

Short-Term Retention, Rehearsal, and Interference

The Peterson and Peterson 1959 study investigated retention of short verbal items across brief delays during which participants performed a distracting activity. Such paradigms showed that successful short-delay recall depends on what happens between presentation and test, making passive permanence an inadequate assumption. A distractor can prevent active rehearsal and introduce competing material; therefore a steep retention function alone cannot isolate one physical decay process. The study's influence was to focus attention on temporal maintenance conditions and the role of intervening cognitive activity. Peterson & Peterson (1959)

Atkinson and Shiffrin's 1968 framework introduced an explicit distinction between structural components and flexible control processes. Sensory registers, short-term store, and long-term store were proposed as functionally differentiated components, while rehearsal, coding, and search were treated as strategies responsive to instructions, history, and task. This is frequently compressed into a simple three-box diagram, but the original account was more complex, especially in its consideration of multiple control procedures and alternative explanations of forgetting. The distinction between architecture and strategy was essential to making the framework capable of explaining why ostensibly similar tasks elicit different memory performance. Atkinson & Shiffrin (1968)

The resulting program made it possible to ask whether rehearsal changes retention, whether retrieval failures reflect weak storage or ineffective search, and how processing goals alter what is encoded. A stage-based model is useful insofar as it generates differences that experiments can test. Researchers must also recognize that memory tasks are seldom process-pure: recall involves encoding, maintenance, cue interpretation, retrieval, and response. Historical reconstruction should examine the original experiments and their model assumptions separately, because a successful fit to one task does not validate every postulated structure. The most durable achievement was a language for combining formal representations of memory operations with targeted experimental manipulation. Atkinson & Shiffrin (1968) Peterson & Peterson (1959) Craik & Lockhart (1972)

Working Memory and the Reorganization of Storage Models

Baddeley and Hitch's 1974 working-memory proposal shifted emphasis from a short-term holding place to the coordination of storage and ongoing processing. Their investigations of reasoning, comprehension, and learning under concurrent memory loads asked what happens when a person must retain information while performing another demanding task. The experiments challenged the view that a single short-term store with one undifferentiated capacity explains all immediate cognitive activity. The proposal opened a research program in which verbal retention, control, and other forms of representation could be investigated through selective interference. Baddeley & Hitch (1974)

The methodological strength of concurrent-task research is that it tests functional dependence: if two activities draw on a common resource or operation, experimentally combining them should alter performance in predictable ways. But interference can also arise from shared response requirements, overlapping materials, changes in strategy, or generalized task difficulty. Therefore conclusions about independent components require converging dissociations and carefully controlled baselines. Historical work should trace how the model's components were revised as evidence accumulated, rather than presenting a later textbook diagram as if it were completely specified at the original moment of publication. Baddeley & Hitch (1974) Atkinson & Shiffrin (1968)

Working memory connects the cognitive revolution's conceptual origins to a major continuing research area. Its constructs can support analyses of language understanding, arithmetic, attention, developmental change, and reasoning while generating new disputes about capacity measurement and the relationship between domain-specific and domain-general processes. In this dossier the model is treated historically as a response to particular empirical limitations of preceding accounts. Detailed modern working-memory mechanisms belong to the separate canonical dossier on that topic. The historical question here is why the research community required a new functional decomposition and what evidence made the revised description compelling. Baddeley & Hitch (1974) Neisser (1967) Marr (1982)

Reaction Time and the Inference of Hidden Operations

One of information-processing psychology's strongest methods is to infer internal operations from systematic differences in response time. Sternberg's 1966 short-term memory scanning task asked participants to judge whether a probe belonged to a previously memorized set. The mean time required for judgment increased approximately linearly as the number of remembered symbols increased. That relationship suggested a serial-comparison interpretation under the conditions examined. Because the experimental variable was set size and the response time was observable, the theory linked a latent processing operation to a quantitative, falsifiable prediction. Sternberg (1966)

The inference from a straight line to a specific mechanism is conditional. Serial exhaustive comparison, serial self-terminating search, and appropriately parameterized parallel models can sometimes approximate similar aggregate trends. Error patterns, distributions of individual response times, positive and negative trials, and systematic manipulations of familiarity become necessary to distinguish the mechanisms. Experimental timing includes stimulus encoding, decision, and response execution as well as the hypothesized scanning operation. A linear slope therefore cannot be taken as a direct image of a single mental step unless the other contributions and competing accounts are addressed. Sternberg (1966) Marr (1982)

The broader historical achievement was a change in what counted as an explanation. A theory of recognition could be required to account for both accuracy and the shape of performance functions across experimentally specified conditions. Researchers could refine an account by asking whether the predicted slope changed with set size, instructions, presentation conditions, or practice. The emphasis on intermediate quantitative constraints carried forward a tradition of experimental measurement while expanding its vocabulary to include operations, comparisons, and representations. Cognitive psychology thereby acquired a practical standard for assessing invisible mechanisms through their public behavioral consequences. Sternberg (1966) Newell et al. (1958) Posner (1980)

Mental Imagery, Spatial Transformation, and Representation

Shepard and Metzler's 1971 mental-rotation experiments asked observers to decide whether drawings portrayed the same three-dimensional configuration from different orientations. Response time increased systematically with angular disparity, offering evidence that a spatial transformation process contributed to recognition under their experimental conditions. The experiment became influential because it connected a subjective-seeming representational operation to a measurable relation between a controlled physical variable and behavior. The term rotation refers to an explanatory interpretation of the temporal pattern, whose status depends on evaluating rival mechanisms and the experimental details. Shepard & Metzler (1971)

The result encouraged cognitive scientists to investigate mental images through performance rather than through verbal description alone. Spatial imagery could be tested by varying orientation, complexity, and task requirements. Such studies invite the question whether visual transformations employ depictive formats, propositional descriptions, or multiple interacting systems. A reaction-time function can constrain a representation without uniquely determining its code. The representational debate is thus more informative when framed in terms of explicit predictions about transformations, interference, and transfer than when expressed as a metaphysical choice between pictures and sentences in the head. Shepard & Metzler (1971) Marr (1982)

Research on mental transformation also revealed that information-processing psychology was wider than a narrow verbal-symbolic framework. Computation can be carried out over spatially structured representations, and different tasks can reward different formats. A history of the cognitive revolution should not retrospectively treat all its participants as proponents of one uniform symbol architecture. The mental-rotation example shows how empirical results helped multiply the kinds of representations psychologists could investigate. Its lasting methodological contribution is an insistence that a proposed internal process be connected to systematic changes in observable performance and confronted with plausible alternative mechanisms. Shepard & Metzler (1971) Newell et al. (1958) Neisser (1967)

Depth of Processing and the Limits of the Store Metaphor

Craik and Lockhart's 1972 levels-of-processing framework questioned whether duration of retention should be explained chiefly by movement between separate storage containers. They proposed that the kind of processing applied to material, especially the extent of meaningful analysis, matters for subsequent memory. The framework redirected attention from the mere location of a hypothetical trace to operations performed during encoding and retrieval. Its historical significance lies in challenging a highly influential representational metaphor from within cognitive psychology while retaining the expectation that experimental manipulations should distinguish explanatory alternatives. Craik & Lockhart (1972) Atkinson & Shiffrin (1968)

Processing depth requires careful operationalization. A task involving semantic judgments may yield stronger later recall than one involving superficial letter-form judgments, but such an outcome could reflect elaboration, distinctiveness, congruency, retrieval cues, or time spent on the task. It cannot simply be redescribed as evidence for greater depth without specifying what depth independently predicts. The theory helped generate a productive experimental literature in which encoding instructions, orienting tasks, and retrieval contexts were systematically manipulated. The resulting debates strengthened attention to task operations and exposed difficulties with single-label causal accounts of memory. Craik & Lockhart (1972)

This episode exemplifies theoretical correction rather than abandonment of information processing. A critique of structural stores continued to depend on measurable transformations, controlled information presentation, and a representation of the memory task. The question shifted toward which mental activity creates durable, accessible traces and how retrieval interacts with the conditions of acquisition. Later frameworks could incorporate aspects of both structural differentiation and processing-dependent retention. For historical synthesis, that coexistence matters: progress may consist in specifying where each model predicts distinct observations instead of selecting one diagram as the exclusive truth of cognition. Craik & Lockhart (1972) Baddeley & Hitch (1974) Atkinson & Shiffrin (1968)

Attention as Spatial Orientation and Experimental Chronometry

Posner's 1980 work on orienting attention studied changes in detection efficiency when observers prepared for a location in visual space. The paradigm separated orienting of attention from eye movement and asked how the timing and validity of cues alter responses to targets. This brought a spatially organized dimension to the study of selection, complementing the earlier focus on competing auditory messages. In both traditions, cognitive processing was inferred from patterned performance under intervention, but the relevant operation differed: shifting the priority of a location rather than deciding which message passed an abstract filter. Posner (1980)

Cueing designs generate several levels of inference. A benefit on validly cued trials might indicate preparatory facilitation, while costs on invalidly cued trials can indicate the need to disengage or redirect attention. Yet response-time measures are also affected by expectations, motor preparation, target probability, and baseline selection. Good designs vary these factors and interpret differences within clearly defined task contrasts. The framework became influential because it allowed attention to be examined dynamically, with temporal and spatial precision, while leaving room for later neural and computational analyses. Posner (1980) Broadbent (1958)

Spatial orienting helped enlarge the explanatory target of cognitive psychology. Attention could be understood as the flexible distribution of processing priorities, not simply a fixed throughput limitation. That refinement opened questions about automatic versus goal-dependent selection, the interaction of perception with movement, and the possibility of multiple attentional systems. A historical account should resist reading present-day neural network terminology back into the initial experiment. The early contribution was the disciplined measurement of covert orienting and the construction of interpretable performance contrasts. Subsequent theories added increasingly differentiated accounts of control and implementation. Posner (1980) Marr (1982) Clark (2013)

Constructive Memory and False Recollection

The cognitive revolution's emphasis on representations created an enduring question about whether remembering retrieves a preserved record or reconstructs an event from fragments and organized knowledge. Bartlett's 1932 research had already emphasized changes in recalled narratives and the role of prior schemas. In a later experimental tradition, Roediger and McDermott's 1995 word-list studies demonstrated recall and recognition of associated but nonpresented items under controlled conditions. The latter paradigm offered a tractable method for examining how semantic organization can shape the contents of memory and lead to high-confidence errors. Bartlett (1932) Roediger & McDermott (1995)

The two traditions operate at different scales. Bartlett's materials involved meaning, narrative transformation, and social context; laboratory associative-list tasks isolate a reproducible source of false-memory responses through carefully designed verbal sets. Their shared lesson is that cognitive performance reflects active organization of information. The mechanisms, populations, and ecological significance of the results should not be collapsed into one general claim about all memory. A false recognition in a laboratory list does not establish that a particular real-world recollection is inaccurate, and a culturally transformed narrative does not identify one specific neural encoding failure. Bartlett (1932) Roediger & McDermott (1995)

The history of reconstruction shows how cognitive psychology developed methods for studying errors as theoretically informative observations. False alarms, omissions, intrusions, and systematic distortions can reveal information about semantic activation, cue use, decision thresholds, and prior knowledge. Error analyses also demonstrate why an information-processing explanation must specify which information is encoded, transformed, retrieved, and evaluated. This research arc links the revolution to contemporary work on memory reliability and highlights the importance of maintaining a distinction between evidence about experimental mechanisms and claims about individual historical events. Roediger & McDermott (1995) Craik & Lockhart (1972) Neisser (1976)

From Serial Symbol Systems to Parallel Distributed Processing

The rise of connectionist modeling in the 1980s transformed the internal diversity of cognitive science. Rumelhart, McClelland, and their collaborators presented parallel distributed processing as a framework in which patterns of activation across many connected units could implement cognitive functions. Knowledge could be represented in distributed connection strengths and learned through changes to those connections, offering an alternative to accounts based primarily on discrete, explicitly coded rules. This work did not simply return psychology to a pre-cognitive stimulus-response stance. It retained the program of mechanistic explanation through computation while revising assumptions about representations, learning, and the temporal organization of processing. Rumelhart et al. (1986)

Comparison between symbolic and distributed approaches requires more than ranking their output accuracy. The theories make different commitments about how generalization occurs, how representations are organized, how learning proceeds, and what happens when information is incomplete or noisy. A symbolic program can represent a structured problem explicitly; a distributed model can capture graded similarity and accumulated learning through its parameters. Both can be evaluated through human error patterns, developmental trajectories, transfer performance, and processing limitations. Their strengths depend on what explanatory question is being asked and on whether the model produces independent predictions rather than merely fitting training observations. Newell et al. (1958) Rumelhart et al. (1986)

The historical consequence was a broadening of what counted as information processing. Cognitive psychology increasingly hosted multiple computational architectures rather than a single serial pipeline. Researchers had to connect implementation choices to testable psychological claims and distinguish functional abstraction from biological plausibility. The continued interaction of symbolic, connectionist, and hybrid theories shows that the revolution was not closed by its initial victories. It established a mode of inquiry in which representations and transformations could be revised as mathematical techniques and empirical evidence developed. Rumelhart et al. (1986) Marr (1982) McCulloch & Pitts (1943)

Levels of Explanation and the Computational Analysis of Vision

David Marr's Vision, originally published in 1982, articulated a widely influential distinction among computational problems, representations and algorithms, and physical implementation. The computational level asks what problem a system solves and why a proposed mapping is appropriate; the algorithmic level asks which representations and transformations accomplish the task; implementation asks how these operations are physically realized. The distinction clarified disagreements in which researchers had mistaken an account of a task's objective for an account of the procedure, or a neural observation for proof of a particular psychological representation. Marr (1982)

Marr's framework helps interpret the earlier history of information-processing psychology. Shannon's communication theory constrains properties of coded messages under specified channels, while Sperling's experiments investigate brief visual availability; Newell and colleagues describe problem-solving procedures, and Sternberg infers scanning operations from time measurements. These works occupy overlapping but distinct explanatory levels. A formal algorithm can be computationally sufficient without being uniquely identified in human data, while a behavioral regularity can constrain a mechanism without identifying its neural substrate. Explicitly stating the level of a claim is therefore a major tool of scientific criticism. Shannon (1948) Sperling (1960) Newell et al. (1958) Sternberg (1966)

The levels framework also demands coordination rather than isolation. A psychologically adequate model must ultimately explain why the organism performs a task, how its resources and representations make the behavior possible, and how the relevant processes operate under biological and environmental constraints. Investigating one level at a time is methodologically valuable, provided the scope of its conclusions is clearly stated. This principle continues to structure debates about brain imaging, artificial systems, and behaviorally defined cognitive capacities. The history of cognitive psychology is enriched by seeing Marr's intervention as a clarification of explanatory responsibilities that had accumulated over decades of increasingly formal psychological research. Marr (1982) Neisser (1976)

Ecological Validity, Situated Cognition, and the Limits of the Computer Analogy

Information-processing experiments often produce precise results by simplifying the world into brief displays, constrained categories, memorized lists, and restricted response choices. These procedures reveal useful functions under controlled conditions, yet the degree to which their mechanisms generalize to everyday behavior remains an empirical question. Neisser's 1976 critique emphasized the need to reconnect cognition with perception and action in natural environments. It challenged the tendency to treat laboratory convenience as a complete definition of the psychological phenomenon being investigated. Neisser (1976)

A contemporary situated perspective asks how body, task, environment, and available tools shape cognitive activity. Clark's analysis of predictive and situated cognition, for example, develops the problem of how an agent's ongoing engagement with its surroundings participates in processing. The historical lesson is that a box-and-arrow model may explain a selected laboratory task while omitting essential structure from the activities to which it is generalized. Ecological tests should therefore retain clear measures while varying contexts, affordances, feedback, and opportunities for external support. Such tests can reveal where a formerly successful processing model needs additional components or a different unit of analysis. Clark (2013) Neisser (1976)

The critique does not erase the experimental achievements of the revolution. It supplies a criterion for their extension: the mechanism identified in a controlled task must be shown to matter in the target environment. A theory should describe how representations relate to active perception and intervention, and why observers with different backgrounds or tools may solve equivalent tasks by different routes. This approach connects Bartlett's attention to meaning and social organization with later attempts to understand distributed, situated, and technologically supported cognition. It also offers a principled way to test claims about cognitive partnership in the Artificial Era without assuming that comparable outputs imply identical psychological organization. Bartlett (1932) Neisser (1976) Clark (2013) Risko & Gilbert (2016)

Research Methods, Measurement, and the Identification Problem

The cognitive revolution transformed psychologists' relationship to hidden processes by making their hypotheses operational. Controlled manipulations of cue timing, memory load, stimulus similarity, response uncertainty, and task instructions yielded differences that theories had to predict. Yet the general problem of identification remained: several internal models can generate the same aggregate observations. A credible experiment should specify alternative mechanisms before data collection, select manipulations likely to distinguish their predictions, and measure more than one informative feature of performance. Response-time distributions, error types, sequential dependencies, and learning trajectories may reveal constraints hidden by averages. Sternberg (1966) Sperling (1960) Newell et al. (1958)

Model comparison also depends on transparent assumptions. A computational account can possess adjustable parameters flexible enough to reproduce many patterns without identifying a psychological mechanism. Robust evaluation requires out-of-sample predictions, theoretically informative contrasts, clear task descriptions, and replication under related but independent conditions. The information-processing vocabulary encourages apparent precision, but terms such as capacity, encoding, retrieval, and control can become circular if inferred solely from the outcome they are meant to explain. Historical source criticism and contemporary construct validation share a concern with what observed evidence can license. Atkinson & Shiffrin (1968) Craik & Lockhart (1972) Marr (1982)

The broader reproducibility literature is relevant to judging modern extensions of cognitive theories. The Open Science Collaboration's 2015 replication project established a landmark empirical reference point for discussing the reliability of published psychological findings. It should not be used as a single numerical verdict on every cognitive experiment, because the investigated samples, research areas, designs, and operational decisions were heterogeneous. For a specific theory, evidence quality depends on the original study, independent replications, measurement validity, the size and precision of effects, and the stability of outcomes across contexts. No single historic study warrants a timeless generalization about a universal mental module. Open Science Collaboration (2015) Marr (1982)

International Trajectories, Populations, and Historical Selectivity

Cognitive psychology developed through transnational exchanges among experimental, linguistic, mathematical, and philosophical traditions. Its canonical English-language narrative often highlights North American institutions, British attention research, and computational laboratories, while earlier and parallel investigations of perception, development, social interaction, and memory developed across many intellectual contexts. An adequate history distinguishes influence documented through publications and collaboration from resemblance inferred retrospectively. It also evaluates how translation, institutional funding, instrumentation, and access to journals affected which theories became widely visible. Broadbent (1958) Bartlett (1932) Neisser (1967)

Experimental populations and cultural environments affect the claims that can be drawn from classic paradigms. A task based on familiar letter strings, English words, or the conventions of a particular academic setting tests performance by subjects trained within those symbolic practices. Such results can still be scientifically illuminating, but their applicability to other linguistic and cultural environments requires investigation. Bruner's category tasks and memory experiments can be analyzed in relation to the instructional practices and background knowledge that they presuppose. A historical monograph should therefore treat participant characteristics and experimental artifacts as evidence-bearing components, not incidental scenery. Bruner et al. (1956) Bartlett (1932)

The international history also resists a simple lineage in which every later cognitive theory descends directly from one 1956 event. Attention research in Cambridge, concept formation in American psychology, formal theories of language, and computational models of problem solving converged unevenly. Their researchers differed in methodological commitments and disciplinary priorities. The growth of shared journals, graduate training, and technical vocabularies gradually consolidated a field, while the programs themselves remained plural. The strongest historiography reconstructs these partially overlapping trajectories and identifies the documentary limits of claims about direct intellectual influence. Broadbent (1958) Chomsky (1959) Newell et al. (1958) Neisser (1967)

Artificial Intelligence, Cognitive Partnership, and the Artificial Era

The historical exchange between psychology and artificial intelligence is a legitimate research object because formal models of problem solving and learning were developed alongside experimental theories of cognition. Turing's discussion of machine behavior and Newell, Shaw, and Simon's implemented procedures exemplify different questions: how intelligent performance might be operationally evaluated and how a candidate computational procedure can generate a complex sequence of actions. The relation between the two fields is reciprocal. Models of human thinking inspired programs, while programs supplied psychologists with explicit hypotheses about possible mechanisms. Turing (1950) Newell et al. (1958)

Contemporary artificial systems renew the methodological question of how to compare agents with different embodiments, training histories, memory arrangements, and access to environmental tools. Similar behavior on an information-processing task supports a comparison of observable capacities. It does not, by itself, demonstrate shared internal operations, equivalent subjective experience, or human-like personal continuity. Comparative research can make progress by specifying inputs, interventions, memory demands, error patterns, transfer conditions, and stability of strategies. Cognitive offloading is a particularly useful observable phenomenon: people can shift information maintenance and problem-solving steps to external supports, changing the effective organization of task performance. Risko & Gilbert (2016) Marr (1982)

Within the Artificial Era horizon, the cognitive revolution supplies a historical and methodological foundation for studying human–artificial cognitive partnerships. Its enduring achievement is the demand that claims about intelligence and representation be articulated as testable mechanisms and distinguished from metaphorical similarity. Philosophical classifications of Artificial and Artificial Sapiens belong to the primary authorial system; experimental claims require independent operationalization and evidence. This distinction allows a research program to investigate agency, learning, memory, and interaction across biological and artificial settings while preserving the actual evidentiary status of each proposition. The relevant legacy is a disciplined method for comparing organized information processing across different kinds of systems. Turing (1950) Rumelhart et al. (1986) Risko & Gilbert (2016)

Conclusion

The cognitive revolution fundamentally changed what experimental psychology could ask about the relation between information, representation, and behavior. Its history includes precedents in reconstructive memory and cognitive maps, mid-century mathematical and linguistic innovations, formal problem-solving programs, controlled investigations of attention and memory, and later challenges to serial symbolic accounts. Its defining accomplishment was the conversion of hypotheses about otherwise inaccessible mental operations into models constrained by public observations. Different strands contributed different instruments: information theory offered precise descriptions of signals, chronometric experiments exposed processing relations, linguistic analysis posed problems of structured knowledge, and computer simulation demanded explicit procedures. Bartlett (1932) Tolman (1948) Shannon (1948) Newell et al. (1958)

The field's subsequent development shows that its original theories were productive research propositions rather than permanent descriptions of mental architecture. Broadbent's selection model stimulated experimental disputes about unattended information; multistore memory accounts led to investigations of control processes and working memory; processing depth, mental imagery, connectionism, and ecological approaches refined what could count as representation and computation. These modifications demonstrate how cumulative research operates: a paradigm survives when its questions and experimental practices continue to yield discriminating evidence, even as explanatory models change. The scholarly task is to preserve both the initial argument and the precise evidence that later required its revision. Broadbent (1958) Atkinson & Shiffrin (1968) Baddeley & Hitch (1974) Rumelhart et al. (1986)

The resulting synthesis establishes three durable methodological priorities. Cognitive explanations need explicit constructs that produce distinctive predictions; experimental evidence must separate a behavioral regularity from the several processes capable of generating it; and claims about general human or artificial cognition require tests across relevant populations, tasks, and environments. Historical reconstruction adds a fourth requirement: the documentary record must ground claims about origins and influence. The cognitive revolution remains foundational because it made models of information transformation objects of scientific comparison, and its unresolved questions continue to organize research into perception, memory, reasoning, language, action, and human–artificial relations. Marr (1982) Neisser (1976) Open Science Collaboration (2015)

References