29.14.02 · psychology / history-of-psychology

The cognitive revolution and the computational theory of mind

shipped3 tiersLean: none

Anchor (Master): Fodor, The Language of Thought (Harvard UP, 1975); Turing, 'Computing Machinery and Intelligence' (Mind, 1950); primary sources McCulloch & Pitts 1943, Miller 1956, Chomsky 1959, Neisser 1967, Rumelhart & McClelland 1986

Intuition Beginner

For nearly half a century, American psychology refused to look inside the head. The behaviorists, led by John Watson and B. F. Skinner, insisted that a real science could measure only what an organism does — its responses to stimuli — and that talk of "mind," "idea," or "image" was unscientific. By the 1950s this had become a cage. Psychology's most interesting questions — how we speak, remember, plan, recognize a face — could not even be asked inside the vocabulary behaviorism allowed.

Then a new idea arrived from an unexpected direction: the computer. If a machine could take in information, hold it, transform it, and act on it, then perhaps a person could be described the same way — not as a bundle of stimulus-response links, but as a system that processes information. The mind was back. This change, gathered between 1956 and 1967, is the cognitive revolution. Ulric Neisser's 1967 book Cognitive Psychology gave it a name and a textbook.

The revolution's boldest claim is the computational theory of mind: thinking is computation, and the brain runs something like a program over something like a code. This unit follows that claim from its roots — Turing's computer metaphor, Miller's measurements of memory, Chomsky's attack on behaviorist accounts of language — through Fodor's "language of thought," and on to its challenger, the connectionist tradition of artificial neural networks.

Visual Beginner

Stage	Function	Everyday analogy	Capacity limit
Sensory memory	briefly holds raw input	an echo or afterimage	decays in under a second
Attention	selects what gets through	a spotlight	roughly one stream at a time
Working memory	holds and operates on current items	a mental workbench	about seven chunks
Long-term memory	stores knowledge	a vast library	effectively unbounded
Response	produces action	the output	shaped by all prior stages

The table is a caricature, but it is the caricature the cognitive revolution made canonical. Behaviorism kept only the two end columns — the stimulus and the response — and treated the middle as a sealed box. The revolution opened the box and gave each stage inside a name, a measurable capacity, and a model.

Worked example Beginner

In 1956 George Miller measured how much a person can hold in mind at once, and the result became a founding fact of the new psychology ^{[miller1956 Miller 1956]}.

Imagine you are read a string of random digits at one per second: 7 3 9 1 4 8 2. When the list ends you repeat it back. Most adults manage about seven digits, sometimes six, sometimes nine. Miller called this the "magical number seven, plus or minus two." A behaviorist would record this as a response rate and stop there. The new psychology asked what the limit revealed about the machinery inside.

Now try letters: C I A F B I I B M T N N. That is twelve letters, well past the limit — yet most people reproduce them easily. The trick is that the letters re-group into four familiar acronyms: CIA, FBI, IBM, TNN. Miller called this regrouping chunking. The limit is not on raw items but on chunks — and a chunk is whatever your prior knowledge has already bundled together.

What this tells us: the seven-item limit presupposes internal representations doing the grouping, because the same physical input yields different capacities depending on what the listener already knows. Behaviorism's stimulus-response vocabulary had no place for "what the listener already knows." The information-processing framework did, and that single move reopened the mind as a scientific object.

Check your understanding Beginner

Formal definition Intermediate+

The cognitive revolution supplied a vocabulary that the rest of this chapter inherits. The definitions below fix the central terms.

An information-processing model ^{[neisser1967 Neisser 1967]} is a functional decomposition of a cognitive act into a sequence of stages — encoding, storage, transformation, retrieval, response — each characterized by an operation on an internal representation and by measurable parameters (capacity, decay rate, latency). The model is a flow: stimulus enters at one end, is transformed by a cascade of intermediate processes, and emerges as behavior at the other. A flow diagram is not a metaphor tacked onto prose; each box commits its author to a distinct process with its own properties.

The computational theory of mind (CTM) is the stronger thesis that mental processes are computations. On the standard statement, a system has a mind, or a mental state, insofar as it implements the right sort of computation over the right sort of representation; the physical substrate matters only in that it must be capable of implementing the computation. The brain's biological details are relevant to how the computation is realized, not to whether there is a computation at all. This is the sense in which the theory makes the computer more than a metaphor: thinking and computing are claimed to be the same kind of thing.

A physical symbol system, in the formulation of Newell and Simon ^{[newell1958 Newell, Shaw, & Simon 1958]}, is a physical device that manipulates structured patterns (symbols) according to rules. Their hypothesis states that a physical symbol system has the necessary and sufficient means for general intelligent action — that intelligence is, at bottom, rule-governed symbol manipulation. This is the symbolic pole of the computational theory.

The language of thought hypothesis, due to Fodor ^{[fodor1975 Fodor 1975]}, holds that thought takes place in a Mentalese: an internal, compositional system of representations in which the meaning of a complex thought is a function of the meanings of its parts and their mode of combination. Thinking is then computation over Mentalese sentences, and the productivity and systematicity of thought — you can think a thought you have never thought before, and anyone who can think "John loves Mary" can think "Mary loves John" — are explained by the compositional syntax of the medium.

Connectionism, or parallel distributed processing ^{[rumelhart1986 Rumelhart & McClelland 1986]}, is the counter-tradition. A connectionist model is a network of simple units organized in layers, weighted links between units, and a rule for adjusting the weights; information is stored not in discrete symbols but in the pattern of weights across the whole network. Knowledge is distributed, retrieval is the reactivation of a pattern, and learning is the gradual reshaping of weights by experience. The neural inspiration is overt; the explanatory primitive is the weighted connection, not the symbol.

The contrast that organizes the rest of the unit is between the symbolic tradition (CTM, physical symbol systems, the language of thought) and the subsymbolic tradition (connectionism). Both descend from the revolution, and both take the mind to be an information-processing system; they disagree about the basic currency of the processing.

Key model: the computational theory of mind and the information-processing framework Intermediate+

The revolution's model is best seen as two claims layered together: an architectural claim about what the mind is and a methodological claim about how to study it. Treating them separately exposes where the framework earns its strength and where it inherits behaviorism's unfinished business.

The architectural claim is the computational theory of mind. Turing's 1950 paper proposed that the question "can a machine think?" could be recast as a behavioral test — the imitation game — and that the test admitted no principled line between human and machine reasoning ^{[turing1950 Turing 1950]}. McCulloch and Pitts had already shown, in 1943, that networks of simple threshold units could implement any Boolean function and so, in principle, any finite computation ^{[mcculloch1943 McCulloch & Pitts 1943]}. By the late 1950s the Newell-Simon programs — the Logic Theory Machine and the General Problem Solver — were proving theorems and playing chess by searching a space of symbol structures under rules ^{[newell1958 Newell, Shaw, & Simon 1958]}. These threads converge on the claim: cognition is computation; the brain is the hardware; the mind is the program.

The methodological claim is the information-processing framework ^{[neisser1967 Neisser 1967]}. Broadbent's 1958 flow diagram of selective attention — a sensory buffer feeding a selective filter feeding a limited-capacity channel feeding long-term storage — set the template ^{[broadbent1958 Broadbent 1958]}: draw boxes for processes, arrows for the flow of information, and attach to each box a measurable property. Miller's magical number seven then quantified one box ^{[miller1956 Miller 1956]}; the Sperling partial-report paradigm quantified another; reaction-time methods decomposed a third. The framework turned "the mind" from a philosophical abstraction into a collection of named stages, each admitting an experiment.

The framework's first triumph was that it redescribed behaviorism's own data at a deeper level. A Skinner-box reinforcement schedule, read as a stimulus-response curve, is also the output of an animal that encodes the stimulus, retrieves a value, and selects an action; the information-processing account predicts, and in many cases explains, features the flat curve leaves opaque. The critique that Chomsky leveled at Skinner's Verbal Behavior made the gain vivid: children produce sentences they have never been rewarded for and make systematic errors like "I goed," which no schedule of imitation and reinforcement can generate ^{[chomsky1959 Chomsky 1959]}. Within the behaviorist vocabulary these facts are anomalous; within the computational vocabulary they are exactly what a rule-governed system should produce. This is the sense in which the revolution was a shift and not merely an accumulation: the same datum meant a different thing before and after.

The framework's acknowledged limit is the implementation question it leaves open. To say that cognition is computation is to say what the mind does; it leaves nearly untouched how wet neural tissue manages to do it. The connectionist tradition, taken up in the Master section, arose precisely to supply the missing layer — and in doing so it reopened the question of whether the symbol is the right primitive at all.

Bridge. The information-processing framework builds toward every cognition unit that follows: attention, memory, language, and decision are each modeled as stages of this same pipeline, and the computational theory of mind appears again in the cognition chapter 29.05.01 and the learning-and-memory chapter 29.04.01. This is exactly what separates a cognitive account from a behaviorist one — the science begins at the box behaviorism refused to open. The foundational reason the framework took hold so fast is that it handed psychologists a shared architecture: stimulus, encoding, storage, computation, response, each measurable and each modelable. Putting these together, the pipeline generalises to non-human animals and to machines, and the bridge is that a flow diagram is itself a theory — every box commits its author to a representational claim.

Exercises Intermediate+

Exercise 1 (easy, multiple choice).

Which event is most often cited as the conventional date for the cognitive revolution's opening, when Miller presented the magical number seven and Newell and Simon presented the Logic Theory Machine at the same symposium?

A. The 1913 publication of Watson's manifesto

B. The 1956 MIT symposium on information theory

C. The 1938 publication of Skinner's Behavior of Organisms

D. The 1990 publication of Danziger's Constructing the Subject

Hint

The date falls in the mid-1950s, the decade in which the pieces of the new synthesis began to converge.

Answer

Option B. The 1956 MIT symposium.

The September 1956 Symposium on Information Theory at MIT brought together Miller (the magical number seven), Newell and Simon (the Logic Theory Machine), Chomsky (early work on syntax), and Bruner and Goodnow — a convergence retrospectively read as the founding moment of cognitive science.

Exercise 3 (medium, multiple choice).

Chomsky's 1959 critique of Skinner's Verbal Behavior turned on the observation that:

A. children cannot learn language without explicit reinforcement

B. language is acquired by simple imitation of adult speech

C. children produce novel, rule-governed utterances that reinforcement cannot explain

D. nonhuman animals acquire grammar as readily as humans

Hint

Consider what a child says that no adult has said to the child.

Answer

Option C.

Chomsky argued that children produce sentences they have never heard and make systematic errors such as "I goed" — the mark of an internal rule system overgeneralizing, not of imitated reinforced responses. This "poverty of the stimulus" argument showed that the input alone cannot determine the grammar the child acquires, and it became a centerpiece of the case against behaviorist accounts of language.

Exercise 4 (medium, short answer).

Distinguish the symbolic and subsymbolic (connectionist) traditions. State one empirical regularity each is held to explain well.

Hint

One tradition excels at rules and productivity; the other at graceful failure and statistical learning.

Answer

The symbolic tradition (CTM, physical symbol systems, the language of thought) takes the rule-governed symbol as its primitive and excels at the productivity and systematicity of reasoning — the fact that whoever can think "John loves Mary" can think "Mary loves John," and that novel inferences compose. The connectionist tradition takes the weighted link as its primitive and excels at pattern completion, graceful degradation after damage, and statistical generalization from examples — a network that has learned "robins fly" will tend to infer that sparrows fly without storing any rule.

Exercise 5 (medium, short answer).

Fodor argues that thought has a compositional syntax (the language of thought). State the two phenomena he takes this to explain, and say in one sentence why a pure associationist network struggles with them.

Hint

The phenomena are productivity (you can think infinitely many thoughts) and systematicity (the capacity to think related thoughts comes in clusters).

Answer

Fodor cites productivity — the open-ended capacity to entertain novel thoughts — and systematicity — the fact that capacities come in families, so that an ability to think a given content brings with it the ability to think its systematic neighbors. A compositional Mentalese explains both: complex thoughts are built from reusable parts by structure-sensitive rules, so new combinations are available for free and the parts can be recombined in lockstep. A pure associationist network, Fodor argues, has no combinatorial syntax to underwrite this lockstep, so the systematic clustering of capacities is left unexplained.

Exercise 6 (hard, short answer).

The computational theory of mind is often glossed as "the brain is hardware, the mind is software." State the strongest philosophical objection to this gloss, and then say why the gloss can nonetheless be retained as a regulative ideal for cognitive psychology.

Hint

Consider whether the same program can be run on substrates that share none of the brain's biological structure, and what that implies about explanatory levels.

Answer

The strongest objection is the worry about implementation: a mere equivalence-class of input-output behavior does not fix a unique computation, because any finite behavior can be described under indefinitely many computational interpretations, so "the mind is software" threatens to underdetermine the very states it posits. If multiple programs are compatible with the same behavior, the software description seems to float free of the biology that would adjudicate between them, and the theory risks vacuity.

The gloss survives as a regulative ideal because cognitive psychology does not traffic in unconstrained input-output descriptions. It posits specific intermediate representations (working-memory buffers, object files, syntactic trees) whose existence is tested by converging operations — reaction-time dissociations, interference paradigms, neuropsychological deficits, and neural signatures. These constraints pin down particular computations well enough to make them explanatory, even if the philosophical underdetermination remains. The metaphor is a research program, not a finished ontology.

Exercise 7 (hard, short answer).

Explain why the connectionist revival of the 1980s ^{[rumelhart1986 Rumelhart & McClelland 1986]} is sometimes described as a return to associationism rather than a break with it. Then give one reason it nonetheless belongs to the cognitive revolution and not to behaviorism.

Hint

Ask what basic explanatory primitive each tradition uses, and then ask whether the new model still posits internal representations.

Answer

Connectionism revives associationism in its explanatory primitive: where behaviorism explained behavior by learned associations between stimuli and responses, connectionism explains cognition by learned associations encoded as weights between units. In both cases the basic currency is the associative link, strengthened by experience, and neither needs a symbolic rule as a basic posit. That is the continuity.

It nonetheless belongs to the cognitive revolution because the weights are internal parameters of a model that posits hidden representations and intermediate transformations. The network is a hypothesis about the machinery between input and output — about what is computed and how it is stored — and its predictions concern internal states (patterns of activation, distributed representations, error gradients) rather than stimulus-response rates alone. Behaviorism refused the inner box; connectionism fills it, just with weights instead of symbols. The disagreement with the symbolic tradition is therefore a family quarrel inside the revolution.

Symbol-processing versus connectionism Master

The revolution's internal fault line is the dispute over the mind's basic currency. The symbolic tradition, descending from Turing and McCulloch-Pitts through Newell and Simon to Fodor, takes the rule-governed symbol as its primitive. The subsymbolic tradition, descending from the same McCulloch-Pitts neurons but developed by Rumelhart, McClelland, and the Parallel Distributed Processing group, takes the weighted connection as its primitive. Both are computational; they disagree about what computation looks like inside the head.

The symbolic tradition's strongest argument is systematicity. Fodor pressed the case that thought exhibits a combinatorial structure that any adequate psychology must explain ^{[fodor1975 Fodor 1975]}: whoever can think "John loves Mary" can think "Mary loves John," and this covariance is not an accident to be learned case by case but a structural feature of what it is to possess the constituent concepts. A compositional Mentalese accounts for the covariance directly — the constituents are reusable slots in a structured representation — whereas a network trained to produce the first thought must, on the face of it, learn the second as a separate fact. Productivity, the open-ended capacity to entertain novel thoughts, runs on the same engine: a finite stock of constituents and rules yields an unbounded stock of thoughts, and no finite training set can approximate the unbounded by enumeration.

The connectionist reply is that the appearance of separate learning is an artifact of how early networks were trained and evaluated. A network with distributed representations can share structure across training instances: the weights that encode "love" and the weights that encode its argument roles can compose in activation space, so that learning one ordering partially privileges the other. The famous Rumelhart-McClelland model of the English past tense ^{[rumelhart1986 Rumelhart & McClelland 1986]} was exhibit A: a network with no explicit rule learned to produce regular past-tense forms and then overgeneralized them to irregulars ("goed," "runned") — reproducing, without a symbolic rule, the very developmental pattern that had been taken as evidence of an internal rule system. The lesson drawn was that rule-like behavior need not issue from explicit rules; the appearance of a rule can be the surface of a distributed statistical regularity.

Each tradition explains a regularity the other finds stubborn. The symbolic side accounts for the productivity, compositionality, and inferential coherence of explicit reasoning — the fact that a single counterexample defeats a universal claim, that logical form tracks validity, that novel sentences are understood on first hearing. The connectionist side accounts for the gradedness, context-sensitivity, and damage-resilience of cognition — the fact that recognition degrades gracefully rather than collapsing, that prototypes rather than definitions seem to govern many categories, that statistical generalization across examples is the rule rather than the exception in learning. A theory that captures logic's crispness struggles with perception's softness; a theory that captures perception's softness struggles with logic's crispness.

The contemporary resolution is level-of-description pluralism. A connectionist network can implement a symbol system — the symbol is a stable pattern of activation, the rule is a reliable transformation of such patterns — and a symbol system can be approximated by a sufficiently expressive network. On this reading the dispute was never about which primitive is real but about which level of description carries the explanatory load for a given phenomenon. Language parsing and logical inference sit closer to the symbolic level; perception, motor control, and many forms of learning sit closer to the subsymbolic level. The unification, where it has been achieved, treats the brain as connectionist hardware whose operations admit a symbolic description at a higher level of abstraction — a modern echo of the hardware-software distinction Turing's metaphor first imported.

Synthesis. The symbol-processing–connectionism debate appears again in every computational treatment of cognition — from language parsing 29.05.01 to learning theory 29.04.01 to contemporary deep learning — and it builds toward the modern reconciliation in which brains are modeled as connectionist hardware running operations that admit a symbolic description at a higher level. This is exactly the level-of-description question the cognitive revolution inherited but could not settle: whether the mind's basic currency is the rule-governed symbol or the weighted connection. The foundational reason the dispute persists is that the two traditions explain different regularities — systematic, productive reasoning on one side, graceful degradation and statistical learning on the other. The central insight of the contemporary view is that these need not compete: a network can implement a symbol system, and a symbol system can be approximated by a network. Putting these together, the revolution's legacy generalises across the cognitive sciences, and the bridge is the recognition that "how the mind works" admits more than one true level of description.

Connections Master

This unit is the depth companion to the history-of-psychology overview 29.14.01, which places the cognitive revolution as one episode in the longer arc from Wundt to cognitive neuroscience. Where the overview identifies the revolution, this unit opens the black box: the information-processing pipeline, the computational theory of mind, and the symbolic-versus-connectionist dispute that the overview could only name.

The information-processing framework flows directly into the cognition and intelligence chapter 29.05.01, which inherits Miller's magical number seven, the multi-store model of memory, and the flow-diagram method as its working vocabulary. Every stage named in this unit's key model — encoding, storage, retrieval, capacity — reappears there as a measured object rather than a historical commitment.

The learning-and-memory chapter 29.04.01 inherits the revolution's family quarrel: behaviorist reinforcement schedules on one side, cognitive representations and the connectionist account of learning on the other. The Chomsky critique and the Rumelhart-McClelland past-tense model both resurface there as the central evidence the chapter must reconcile.

The neuroscience chapter 29.02.01 carries the implementation question this unit leaves open. The claim that cognition is computation tells what the mind does; cognitive neuroscience asks how neural tissue does it, and the modern answer — populations of neurons encoding distributed representations — is the biological face of the connectionist tradition formalized here.

Historical & philosophical context Master

The revolution has a precise genealogy, and its dating is a genuine historiographical question rather than a convenience. The symbolic-computational root was planted in 1943, when McCulloch and Pitts proved that networks of simple threshold units could compute any Boolean function and so, in principle, realize any finite automaton ^{[mcculloch1943 McCulloch & Pitts 1943]}. The information-theoretic root matured in the late 1940s with Shannon's theory of communication and Wiener's cybernetics, which made "information" a measurable currency available to psychology. Turing's 1950 paper then imported the computer as a model of mind ^{[turing1950 Turing 1950]}, replacing the old metaphors of mind-as telephone-exchange or mind-as hydraulic-press with a metaphor of a kind that could, in principle, be implemented.

The conventional birthday is the Symposium on Information Theory held at MIT in September 1956, at which the threads visibly converged. George Miller presented the magical number seven ^{[miller1956 Miller 1956]}; Allen Newell and Herbert Simon presented the Logic Theory Machine, a program that had derived theorems from Whitehead and Russell's Principia; Noam Chomsky presented a formal treatment of syntactic structure; and Jerome Bruner and Jacqueline Goodnow presented their work on thinking and categorization ^{[bruner1960 Bruner, Goodnow, & Austin 1956]}. Historians of cognitive science, notably Howard Gardner, have taken this simultaneity as the field's founding moment: the disciplines that would compose cognitive science — psychology, linguistics, computer science, philosophy, neuroscience — found, in the same year, a shared vocabulary.

The consolidation followed across a decade. Broadbent's Perception and Communication (1958) gave the information-processing framework its canonical flow diagram for attention ^{[broadbent1958 Broadbent 1958]}. Chomsky's 1959 review dismantled the behaviorist account of language and supplied, in the poverty-of-the-stimulus argument, the anomaly the old paradigm could not digest ^{[chomsky1959 Chomsky 1959]}. In 1960 Miller and Bruner founded the Center for Cognitive Studies at Harvard, institutionalizing the movement. Neisser's Cognitive Psychology (1967) then named the field and assembled its findings into a textbook ^{[neisser1967 Neisser 1967]}, and the discipline that had spent forty years studying only inputs and outputs turned back to the machinery between them.

The revolution's philosophical apotheosis arrived with Fodor's The Language of Thought (1975), which argued that the computational theory required an internal, compositional medium of representation — Mentalese — and that the productivity and systematicity of thought could be explained by no other means ^{[fodor1975 Fodor 1975]}. The counter-tradition crystallized a decade later in the two-volume Parallel Distributed Processing (1986), in which Rumelhart, McClelland, and their collaborators showed that networks of neuron-like units could learn structured tasks without explicit rules ^{[rumelhart1986 Rumelhart & McClelland 1986]}. The latter re-opened the question the symbolic tradition had thought settled — whether the symbol is the right primitive — and the debate it ignited still organizes the field.

Bibliography Master

@article{mcculloch1943,
  author  = {McCulloch, Warren S. and Pitts, Walter},
  title   = {A Logical Calculus of the Ideas Immanent in Nervous Activity},
  journal = {Bulletin of Mathematical Biophysics},
  volume  = {5},
  year    = {1943},
  pages   = {115--133}
}

@article{turing1950,
  author  = {Turing, Alan M.},
  title   = {Computing Machinery and Intelligence},
  journal = {Mind},
  volume  = {59},
  number  = {236},
  year    = {1950},
  pages   = {433--460}
}

@article{miller1956,
  author  = {Miller, George A.},
  title   = {The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information},
  journal = {Psychological Review},
  volume  = {63},
  number  = {2},
  year    = {1956},
  pages   = {81--97}
}

@book{bruner1960,
  author    = {Bruner, Jerome S. and Goodnow, Jacqueline J. and Austin, George A.},
  title     = {A Study of Thinking},
  year      = {1956},
  publisher = {John Wiley \& Sons},
  address   = {New York}
}

@book{broadbent1958,
  author    = {Broadbent, Donald E.},
  title     = {Perception and Communication},
  year      = {1958},
  publisher = {Pergamon Press},
  address   = {London}
}

@article{newell1958,
  author  = {Newell, Allen and Shaw, John C. and Simon, Herbert A.},
  title   = {Chess-Playing Programs and the Problem of Complexity},
  journal = {IBM Journal of Research and Development},
  volume  = {2},
  number  = {4},
  year    = {1958},
  pages   = {320--335}
}

@article{chomsky1959,
  author  = {Chomsky, Noam},
  title   = {A Review of {B. F. Skinner}'s \textit{Verbal Behavior}},
  journal = {Language},
  volume  = {35},
  number  = {1},
  year    = {1959},
  pages   = {26--58}
}

@book{neisser1967,
  author    = {Neisser, Ulric},
  title     = {Cognitive Psychology},
  year      = {1967},
  publisher = {Appleton-Century-Crofts},
  address   = {New York}
}

@book{fodor1975,
  author    = {Fodor, Jerry A.},
  title     = {The Language of Thought},
  year      = {1975},
  publisher = {Harvard University Press},
  address   = {Cambridge, MA}
}

@book{rumelhart1986,
  author    = {Rumelhart, David E. and McClelland, James L. and {the PDP Research Group}},
  title     = {Parallel Distributed Processing: Explorations in the Microstructure of Cognition},
  year      = {1986},
  publisher = {MIT Press},
  address   = {Cambridge, MA},
  note      = {2 volumes}
}

@book{eysenck2020,
  author    = {Eysenck, Michael W. and Keane, Mark T.},
  title     = {Cognitive Psychology: A Student's Handbook},
  edition   = {8},
  year      = {2020},
  publisher = {Psychology Press},
  address   = {London}
}

Prerequisites

29.14.01

Tier anchors

beginner: Eysenck & Keane, Cognitive Psychology: A Student's Handbook (8e, Psychology Press, 2020), Ch. 1 — narrative of the cognitive revolution and the information-processing framework
intermediate: Stillings, Weisler, Chase, Feinstein, Garfield, & Rissland, Cognitive Science: An Introduction (2e, MIT Press), Ch. 1-3; primary sources Miller 1956 and Neisser 1967
master: Fodor, The Language of Thought (Harvard UP, 1975); Turing, 'Computing Machinery and Intelligence' (Mind, 1950); primary sources McCulloch & Pitts 1943, Miller 1956, Chomsky 1959, Neisser 1967, Rumelhart & McClelland 1986

References

Eysenck & Keane, Cognitive Psychology: A Student's Handbook (8e, Psychology Press, 2020) · Ch. 1 (the cognitive revolution, the information-processing framework, the decline of behaviorism). Used as the narrative spine.
miller1956 · the magical number seven; chunking as the unit of short-term memory capacity. · source being verified
chomsky1959 · the poverty-of-the-stimulus argument; rule-governed creativity of language. · source being verified
neisser1967 · opening chapter that names and frames the field; the information-processing framework. · source being verified
fodor1975 · the language of thought hypothesis; the computational theory of mind. · source being verified
turing1950 · the imitation game; the computer metaphor for mind. · source being verified
mcculloch1943 · formal threshold neurons realizing Boolean functions. · source being verified
broadbent1958 · the flow-diagram model of selective attention; the information filter. · source being verified
newell1958 · the physical symbol system hypothesis; thinking as heuristic search. · source being verified
rumelhart1986 · connectionist models; distributed representation; the challenge to symbolic cognition. · source being verified
bruner1960 · the Harvard symposium era; categorization and strategies of inquiry. · source being verified

Estimated time

beginner: 20m
intermediate: 50m
master: 80m