Thinking is the mind at work on purpose. This page explains how people build internal models of the world, search those models for solutions, leap to conclusions with fast rules of thumb, and sometimes catch their own errors. Three interactive tasks let you test the classic effects on yourself as you read.
Cognitive psychology defines thinking as the goal-directed manipulation of mental representations. On this view, the thinker encodes a situation as an internal structure of concepts, images, and propositions, applies operations such as search, inference, and mental simulation to that structure, and reads off a judgment, a decision, or a plan. The definition is deliberately broad: it covers solving a geometry problem, deciding between two job offers, diagnosing a strange noise in an engine, and imagining how a couch will look against the far wall. What unites these activities is that information is transformed, step by step, in the service of a goal, an idea formalized in the information-processing framework that treats problem solving as a search through a space of possible states (Newell & Simon, 1972). The sections below follow that framework outward: first the representations thinking operates on, then the operations themselves, and finally the systematic shortcuts and biases that reveal the machinery at work.
What Thinking Is
The modern study of thinking begins with a rejection of two older pictures. Against behaviorism, cognitive psychology insists that something lawful happens between stimulus and response and that it can be measured, chiefly through response times, error patterns, and verbal protocols. Against the introspectionist tradition, it insists that thinkers have limited access to their own machinery, so the machinery must be inferred from behavior rather than simply reported. The replacement picture is computational: thinking is the processing of information over internal representations (Newell & Simon, 1972).
That picture has three load-bearing parts. First, representations: the mind does not operate on the world directly but on coded stand-ins for it, and the format of the code matters for what is easy or hard to compute. Second, operations: processes such as memory search, rule application, analogy, and mental simulation transform one representation into another. Third, constraints: attention and working memory are sharply limited, so real thinkers cannot consider everything and must allocate processing selectively (Miller, 1956). A complete account of any episode of thought names the representation, the operations, and the constraints that shaped which operations actually ran. Herbert Simon called the resulting compromise bounded rationality: behavior that is reasonable given the organism's computational limits and the structure of its environment, rather than optimal in the economist's sense (Simon, 1955). Figure 1 sketches the loop.
The Building Blocks: Concepts and Mental Representations
Every act of thinking presupposes a vocabulary of mental units. The most studied unit is the concept, the mental representation of a category such as bird, tool, or fair price. Classical theories treated concepts as definitions, lists of necessary and sufficient features, but most everyday categories resisted definition. Eleanor Rosch showed instead that categories are organized around prototypes: membership is graded, and items closer to the central tendency of the category are verified faster, named earlier by children, and generated first in listing tasks. A robin is simply a better bird, psychologically, than a penguin is, and this typicality gradient predicts performance across a remarkable range of tasks (Rosch, 1975). Because concepts feed inference, prototype structure shapes thought downstream: properties true of typical members are projected onto the whole category more confidently than properties of atypical members.
Concepts do not float free; they combine into propositions, language-like units that encode who did what to whom, and they connect into larger relational structures that support inference. Alongside this descriptive code, the mind also maintains an analog, picture-like code, mental imagery, examined in the next section. The dual codes matter because format determines cost: reading a spatial relation off an image can be nearly free, while deriving the same relation from propositions may take a chain of inferences, and the reverse holds for abstract relations with no spatial signature. Whatever the format, capacity is the binding constraint. Working memory holds only a handful of chunks at once, which is why mental arithmetic with large numbers fails, why intermediate results must be offloaded to paper, and why expertise so often consists of recoding many small units into fewer, richer chunks (Miller, 1956).
Thinking in Images: Mental Imagery
Some thinking looks less like sentence processing and more like seeing. Roger Shepard and Jacqueline Metzler asked participants to judge whether two perspective drawings of block figures showed the same object in different orientations or mirror-image objects. Response time rose as a strikingly linear function of the angular difference between the views, at a rate of roughly 60 degrees per second, exactly as if participants were rotating an internal object through intermediate orientations (Shepard & Metzler, 1971). Stephen Kosslyn and colleagues found the spatial analog of the same result in mental scanning: after memorizing a map, people took longer to shift attention between imagined landmarks that were farther apart, indicating that the image preserved the metric structure of the original (Kosslyn et al., 1978).
These chronometric signatures fueled the imagery debate. Zenon Pylyshyn argued that the data do not force a picture-like format: participants know how the world works, and that tacit knowledge could generate distance and angle effects even if the underlying representation were propositional all the way down (Pylyshyn, 1973). Defenders of depictive representation answered with converging evidence, including findings that imagery recruits visual cortex, while critics maintained that the format question is harder to settle than the behavioral curves suggest. The dispute remains one of the field's clearest examples of how difficult it is to infer representational format from behavior alone, and the most defensible summary is that imagery is functionally spatial, whatever its ultimate code. You can generate the rotation signature yourself in the task below; the block figures are constructed fresh on each run, so the angle effect you see comes from your own processing, not from memorized answers.
Try It: Mental Rotation
Two block figures appear side by side. Decide whether the right figure is the same object as the left, only rotated, or a mirror image. Work quickly but accurately; the task times each response across ten trials.
Reference
Comparison: same object, rotated
Figure 2. Mental rotation demonstration. Block figures are generated procedurally in your browser and are original to this page; responses are timed locally, and no data is stored or transmitted.
Problem Solving
A problem exists when a goal is blocked and no routine response removes the block. Allen Newell and Herbert Simon formalized this situation as search through a problem space: a set of states, an initial state, one or more goal states, and operators that transform states into other states. Because the space of even a modest puzzle explodes combinatorially, solvers cannot search exhaustively. They prune with heuristics, the most general being means-ends analysis: compare the current state with the goal, identify the largest difference, and apply an operator that reduces it, setting subgoals when the operator's preconditions are unmet (Newell & Simon, 1972). Verbal protocols of solvers working through logic and cryptarithmetic problems matched the step-by-step traces of computer programs built on these principles, the strongest early evidence that thinking could be modeled as symbol manipulation.
Search succeeds or fails with the representation of the problem, and two classic obstacles are representational. Functional fixedness is the inability to see an object outside its customary role: in Karl Duncker's experiments, participants struggled to use a small box as a candle platform when the box was presented full of tacks, because its representation was locked to the container function (Duncker, 1945). Mental set is the analogous rut for procedures, where a practiced method persists even when a shortcut exists. Restructuring the representation, sometimes experienced as sudden insight, dissolves such problems at a stroke.
The other royal road past a search impasse is analogy: importing the relational structure of a solved problem into the current one. Mary Gick and Keith Holyoak gave participants a difficult medical problem, destroying a tumor with rays strong enough to damage surrounding tissue, after they had read a structurally parallel military story in which an army split into small groups that converged on a fortress from different directions. Few participants spontaneously connected the two; given a hint to use the story, most solved the problem with a convergence solution (Gick & Holyoak, 1980). The finding is double-edged: analogical transfer is powerful, but retrieval of the right analog is the bottleneck, since surface dissimilarity hides structural identity. Expertise helps precisely because experts index problems by deep structure rather than surface features.
Reasoning: Deductive and Inductive
Reasoning is thinking constrained by evidence and logic. In deductive reasoning the conclusion follows necessarily from the premises; in inductive reasoning the premises make the conclusion probable but never certain, as when a clinician generalizes from a run of cases. Most everyday inference is inductive, and even ostensibly deductive tasks recruit knowledge and pragmatics rather than formal rules alone.
The signature finding is that people are poor falsifiers. In the four-card selection task, reasoners decide which cards to turn over to test a conditional rule of the form if p, then q. The informative cards are p, the named case, and not-q, the only case that can refute the rule, yet most participants choose p alone or p together with q, checking the cases that match the rule rather than the case that could disprove it (Wason, 1968). Reviews put solution rates for abstract versions near 10%, while versions with familiar, rule-enforcing content are solved far more often, a content effect that any complete theory of reasoning must explain (Evans, 2008). Mental model theory offers one account: reasoners construct mental models of the possibilities the premises allow and read conclusions off those models, so errors arise when working memory limits force some possibilities to be represented and others neglected (Johnson-Laird, 1983). Try both versions below; the contrast between them is the content effect in miniature.
Try It: The Selection Task
Rule: If a card has stripes on its pattern side, then it has an even number on its other side.
Each card shows one side; the other side is hidden. Select every card you must flip, and only those, to find out whether the rule holds for these four cards.
Figure 3. The four-card selection task with an abstract rule and a matched concrete rule. All rules, card faces, and scenarios are original to this page.
Judgment Under Uncertainty: Heuristics and Biases
Frequencies, risks, and probabilities rarely arrive labeled, so people estimate them. Amos Tversky and Daniel Kahneman argued that these estimates lean on a small set of heuristics, operations that substitute an easier judgment for the target judgment (Tversky & Kahneman, 1974). The representativeness heuristic judges probability by similarity: an outcome seems likely to the degree that it resembles its parent population or the process that produced it. It works when similarity tracks probability and fails when it does not, yielding base-rate neglect, insensitivity to sample size, and misperception of randomness. The availability heuristic judges frequency by ease of retrieval, so vivid, recent, or heavily reported events feel more common than they are. Anchoring and adjustment starts from an initial value, even an arbitrary one, and adjusts insufficiently, which is why a first asking price shapes a negotiation long after everyone has dismissed it as a number someone invented.
These shortcuts can override elementary logic. Consider an original version of a classic problem. Dana trained at a conservatory, spends weekends producing electronic music, and keeps meticulous spreadsheets for fun. Asked whether it is more probable that Dana is an accountant or an accountant who produces electronic music, many people rank the second option higher, although a conjunction can never be more probable than either of its components. The added detail raises similarity while lowering probability, and similarity wins (Tversky & Kahneman, 1983).
Decision Making: Framing and Prospect Theory
Classical decision theory assumes description invariance: preferences should depend on outcomes and probabilities, not on wording. Framing experiments show otherwise. When options are described in terms of what will be saved, most people prefer a sure thing; when the identical options are described in terms of what will be lost, most prefer the gamble (Tversky & Kahneman, 1981). Prospect theory explains the reversal with three assumptions: outcomes are coded as gains and losses relative to a reference point rather than as final states; the value function is concave for gains and convex for losses, making people risk averse over gains and risk seeking over losses; and losses loom larger than equivalent gains, an asymmetry called loss aversion (Kahneman & Tversky, 1979). The theory also weights probabilities nonlinearly, overweighting small chances, one reason lottery tickets and insurance policies coexist in the same household.
Herbert Simon supplied the broader frame decades earlier. Real decision makers operate under bounded rationality, with limited information, computation, and time, so rather than optimize they often satisfice, searching until an option clears an aspiration level and stopping there (Simon, 1955; Simon, 1956). On this view heuristics are not defects bolted onto an otherwise rational mind but the operating strategy of an organism adapted to environments whose structure usually rewards them. The demonstration below presents one decision under two descriptions; choose naturally in each, then compare your pattern with the published one.
Try It: One Decision, Two Descriptions
An aggressive fungal blight has reached a heritage orchard of 800 rare apple saplings. The conservancy can fund exactly one response program. Choose the program you would fund.
Figure 4. A within-subject framing demonstration with identical outcomes described as gains or as losses. The scenario and quantities are original to this page.
Two Systems of Thinking: Dual-Process Theory
A recurring idea organizes much of the above: thinking involves processes of two broad types. Type 1 processes are autonomous; they run whenever their triggering conditions are met, demand little working memory, and deliver intuitions, defaults, and practiced skills. Type 2 processes support hypothetical thinking; they are working-memory dependent, largely serial, and capacity limited, and they allow the mind to decouple representations from belief in order to simulate possibilities (Evans, 2008; Evans & Stanovich, 2013). Daniel Kahneman popularized the labels System 1 and System 2 for roughly this division and recast the heuristics-and-biases program as the study of what happens when System 1 outputs go unchecked (Kahneman, 2011).
The framework has sharpened under criticism. Responding to charges that dual-process theories are vague, that their attribute lists confuse correlates with definitions, and that single-process accounts fit the data, Jonathan Evans of the University of Plymouth and Keith Stanovich narrowed the defining features to two: autonomy for Type 1 processing, and working-memory-dependent hypothetical thought for Type 2. Speed, conscious awareness, and evolutionary age are typical correlates, not criteria, and the same authors distinguish the capacity for Type 2 thought from the disposition to deploy it, two traits that come apart in measurement (Evans & Stanovich, 2013). Debate continues over whether the types are discrete or ends of a continuum, but the working memory criterion gives the framework empirical teeth: concurrent load selectively impairs the performances attributed to Type 2 (Evans, 2008).
Language and Thought
Does the language you speak shape what you can think? The strong claim, that language determines thought, finds little support: people reason about possibilities their language does not encode, and mental model theory treats inference as operating over representations of situations rather than over sentences (Johnson-Laird, 1983). Weaker effects are real. Russian obligatorily distinguishes lighter blues (goluboy) from darker blues (siniy), and Russian speakers discriminate colors across that boundary faster than within it; the advantage disappears under verbal interference, indicating that language is recruited online during perception rather than permanently rewiring it (Winawer et al., 2007). A fair summary is that language acts as a tool that directs attention and supplies categories, biasing habitual thought without imprisoning it.
Comparing the Two Systems
Table 1 summarizes the contrast as current theory draws it, separating the single defining difference from the typical correlates.
| Feature | Type 1 (intuitive) | Type 2 (deliberative) |
|---|---|---|
| Working memory (defining) | Minimal demand; runs autonomously when cued | Loads working memory; capacity limits performance |
| Speed and effort | Fast, effortless, parallel | Slow, effortful, largely serial |
| Initiation and control | Mandatory once triggered; hard to suppress | Deliberately engaged; can be sustained, revised, or abandoned |
| Conscious access | Only the output reaches awareness | Intermediate steps are often reportable |
| Typical work | Pattern recognition, practiced skills, default judgments | Hypothetical simulation, mental arithmetic, checking and overriding defaults |
| Characteristic failures | Systematic bias when cues mislead | Lapses under load, fatigue, and time pressure |
Note. Only the working memory row is treated as defining; the remaining rows are typical correlates that admit exceptions (Evans & Stanovich, 2013).
Worked Example
Maya needs a car for a new commute. The episode begins with categorization: reliable commuter car activates a prototype nearer a compact sedan than a vintage roadster, and candidates are judged by resemblance to it (Rosch, 1975). Scanning a listing, she imagines reversing the hatchback into her narrow driveway, mentally rotating the scene to check clearance, an operation whose duration tracks the transformation required (Shepard & Metzler, 1971). The search itself is movement through a problem space: the initial state is no car, the goal state is a sound car within budget, and the operators include searching listings, booking inspections, and negotiating, with means-ends analysis applied whenever a gap opens between where she is and where she needs to be (Newell & Simon, 1972).
Judgment enters with the evidence. A friend's breakdown story makes that model's failures feel common, the seller's tidy garage makes well maintained feel nearly certain despite the model's base rate of transmission trouble, and the listed price of 9,800 anchors every counteroffer that follows (Tversky & Kahneman, 1974). When the seller notes the price already reflects a 700 discount, the deal is framed as a gain; when Maya reframes the pending brake work as paying 700 more, the same arithmetic registers as a loss and looms larger (Tversky & Kahneman, 1981; Kahneman & Tversky, 1979). She does not optimize over the full market; she satisfices, setting an aspiration of sound mechanics, under 9,000, this week, and taking the first option that clears it (Simon, 1956). Throughout, fast impressions arrive first and a slower check sometimes revises them: the gleam of the paint is Type 1, the call to a mechanic is Type 2, and the quality of the outcome depends less on banishing intuition than on knowing when to audit it (Evans & Stanovich, 2013).
Why It Matters
The findings on this page are load bearing in any field where judgment has consequences. Diagnostic errors track availability and representativeness, negotiations and sentences track anchors, and treatment uptake tracks frame. Institutions respond most effectively not by exhorting effort but by redesigning the judgment environment, with checklists, structured protocols, and defaults that insert a Type 2 check at known failure points (Kahneman, 2011). Debiasing the procedure is more reliable than debiasing the person, since Type 1 outputs arrive unbidden and subjective confidence is a poor signal of their accuracy (Evans, 2008).
For the individual reader, the practical skill is metacognitive: learn the signatures of situations where intuition misfires, including unfamiliar statistics, salient anecdotes, arbitrary numbers on the table, and options worded as losses, and slow down exactly there. The demonstrations above are small drills in that recognition.
Key Researchers
Herbert A. Simon (1916–2001). With Allen Newell, founded the information-processing study of problem solving; introduced bounded rationality and satisficing; Nobel laureate in economic sciences, 1978.
Allen Newell (1927–1992). With Simon, built the first symbolic models of human problem solving and formalized search through a problem space as its core mechanism.
Amos Tversky (1937–1996). With Kahneman, mapped the heuristics of judgment under uncertainty and co-authored prospect theory and the framing studies.
Daniel Kahneman (1934–2024). Shared the 2002 Nobel in economic sciences for the work with Tversky on judgment and choice; later synthesized the field for a general audience in Thinking, Fast and Slow.
Peter C. Wason (1924–2003). Invented the four-card selection task and seeded the modern psychology of reasoning, including the study of confirmation strategies.
Karl Duncker (1903–1940). Gestalt psychologist whose monograph on problem solving introduced functional fixedness and the analysis of solution as re-representation.
Eleanor Rosch. Psychologist at the University of California, Berkeley, who established prototype theory and the graded structure of natural categories. Berkeley faculty page.
Keith Holyoak. Distinguished Professor of Psychology at UCLA; with Mary Gick, established analogical transfer as a central mechanism of problem solving. Reasoning Lab · Google Scholar.
Philip Johnson-Laird. Stuart Professor of Psychology, Emeritus, at Princeton University; architect of the mental model theory of reasoning. Google Scholar.
Keith Stanovich. Professor Emeritus of Applied Psychology and Human Development at the University of Toronto; codified dual-process theory with Jonathan Evans and pioneered the measurement of rational thinking dispositions. Personal site · Google Scholar.
Key Terms
| Term | Definition |
|---|---|
| Mental representation | An internal structure that stands in for objects, properties, or situations and supports computation over them. |
| Concept | A mental grouping of entities treated as equivalent for purposes of inference and naming. |
| Prototype | A summary representation of a category centered on its most typical features; membership is graded by resemblance to it. |
| Mental imagery | Quasi-perceptual representation in the absence of the stimulus, preserving properties of perception such as spatial extent. |
| Mental rotation | Continuous transformation of an image through intermediate orientations, with time proportional to angular distance. |
| Problem space | The set of states reachable from an initial state by applying operators; solving is search through this space toward a goal state. |
| Means-ends analysis | A search heuristic that repeatedly selects the operator that most reduces the difference between the current state and the goal. |
| Functional fixedness | Difficulty seeing an object as usable outside its customary function. |
| Analogical transfer | Solving a new problem by mapping the relational structure of a previously solved problem onto it. |
| Deductive reasoning | Inference in which true premises guarantee the conclusion. |
| Inductive reasoning | Inference in which premises make a conclusion probable without guaranteeing it. |
| Mental model | A representation of a possibility whose structure mirrors the situation it describes; reasoning proceeds by constructing and inspecting such models. |
| Heuristic | A shortcut operation that substitutes an easier judgment for a harder target judgment, trading accuracy for speed. |
| Representativeness heuristic | Judging probability by similarity to a prototype or to a generating process. |
| Availability heuristic | Judging frequency or likelihood by the ease with which instances come to mind. |
| Anchoring and adjustment | Estimating by adjusting, usually insufficiently, from an initial value. |
| Framing effect | A change in preference produced by redescribing the same outcomes, typically as gains versus losses. |
| Loss aversion | The greater weight of losses than of equivalent gains in evaluation and choice. |
| Satisficing | Choosing the first option that meets an aspiration level rather than searching for the optimum; the signature strategy of bounded rationality. |
| Type 1 processing | Autonomous processing that runs when cued and makes minimal demands on working memory. |
| Type 2 processing | Working-memory-dependent processing that supports hypothetical thinking and can override defaults. |
Frequently Asked Questions
What is thinking in cognitive psychology?
Thinking is the internal manipulation of mental representations in the service of a goal. It covers forming concepts, simulating possibilities in imagery, solving problems by searching a problem space, reasoning from premises, and choosing among options (Newell & Simon, 1972).
What is the difference between System 1 and System 2 thinking?
System 1, or Type 1, processing is autonomous: it runs when its triggering cues appear and places little demand on working memory. System 2, or Type 2, processing supports hypothetical thought and depends on working memory, so it is slower, effortful, and capacity limited. Current theory treats speed and awareness as typical correlates of the two types rather than defining features (Evans & Stanovich, 2013).
What are heuristics, and why do they produce biases?
Heuristics are mental shortcuts that substitute an easier judgment for a harder one, such as judging probability by similarity or by ease of recall. They are fast and usually serviceable, but because they ignore relevant information such as base rates and sample size, they produce systematic, predictable errors rather than random ones (Tversky & Kahneman, 1974).
What is the framing effect?
The framing effect is a reversal of preference when the same outcomes are described in different terms. People tend to avoid risk when options are framed as gains and to seek risk when the same options are framed as losses, which violates the assumption that choices depend only on outcomes (Tversky & Kahneman, 1981).
What is functional fixedness?
Functional fixedness is the tendency to see an object only in its customary role, which blocks solutions that require an unusual use. Duncker demonstrated it with problems in which participants failed to repurpose everyday objects because the conventional function dominated their representation of the object (Duncker, 1945).
Why do most people fail the four-card selection task?
The rule names two items, and people tend to pick the cards that match those items instead of the cards that could prove the rule false. Checking the named case feels like testing, but only the potentially falsifying combination is fully informative, so most unaided reasoners verify rather than falsify (Wason, 1968).
Is mental imagery picture-like or language-like?
Behavioral evidence shows that imagery preserves spatial and metric structure: rotation time grows with angular difference and scanning time grows with distance (Shepard & Metzler, 1971; Kosslyn et al., 1978). Critics reply that these results could reflect tacit knowledge applied to abstract, language-like representations, so the format question is harder to settle than the curves suggest (Pylyshyn, 1973).
Does relying on heuristics mean people are irrational?
Not by the standard the mind actually faces. Simon argued that real agents have limited time, information, and computation, so strategies that reach good enough outcomes are adaptive rather than defective (Simon, 1956). Dual-process theorists add that people differ in the disposition to check intuitive outputs, and that disposition is measurable and trainable (Evans & Stanovich, 2013).
References
| 1 | Duncker, K. (1945). On problem-solving (L. S. Lees, Trans.). Psychological Monographs, 58(5), i–113. https://doi.org/10.1037/h0093599 |
| 2 | Evans, J. St B. T. (2008). Dual-processing accounts of reasoning, judgment, and social cognition. Annual Review of Psychology, 59, 255–278. https://doi.org/10.1146/annurev.psych.59.103006.093629 |
| 3 | Evans, J. St B. T., & Stanovich, K. E. (2013). Dual-process theories of higher cognition: Advancing the debate. Perspectives on Psychological Science, 8(3), 223–241. https://doi.org/10.1177/1745691612460685 |
| 4 | Gick, M. L., & Holyoak, K. J. (1980). Analogical problem solving. Cognitive Psychology, 12(3), 306–355. https://doi.org/10.1016/0010-0285(80)90013-4 |
| 5 | Johnson-Laird, P. N. (1983). Mental models: Towards a cognitive science of language, inference, and consciousness. Harvard University Press. |
| 6 | Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux. |
| 7 | Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291. https://doi.org/10.2307/1914185 |
| 8 | Kosslyn, S. M., Ball, T. M., & Reiser, B. J. (1978). Visual images preserve metric spatial information: Evidence from studies of image scanning. Journal of Experimental Psychology: Human Perception and Performance, 4(1), 47–60. https://doi.org/10.1037/0096-1523.4.1.47 |
| 9 | Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. https://doi.org/10.1037/h0043158 |
| 10 | Newell, A., & Simon, H. A. (1972). Human problem solving. Prentice-Hall. |
| 11 | Pylyshyn, Z. W. (1973). What the mind's eye tells the mind's brain: A critique of mental imagery. Psychological Bulletin, 80(1), 1–24. https://doi.org/10.1037/h0034650 |
| 12 | Rosch, E. (1975). Cognitive representations of semantic categories. Journal of Experimental Psychology: General, 104(3), 192–233. https://doi.org/10.1037/0096-3445.104.3.192 |
| 13 | Shepard, R. N., & Metzler, J. (1971). Mental rotation of three-dimensional objects. Science, 171(3972), 701–703. https://doi.org/10.1126/science.171.3972.701 |
| 14 | Simon, H. A. (1955). A behavioral model of rational choice. The Quarterly Journal of Economics, 69(1), 99–118. https://doi.org/10.2307/1884852 |
| 15 | Simon, H. A. (1956). Rational choice and the structure of the environment. Psychological Review, 63(2), 129–138. https://doi.org/10.1037/h0042769 |
| 16 | Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124–1131. https://doi.org/10.1126/science.185.4157.1124 |
| 17 | Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453–458. https://doi.org/10.1126/science.7455683 |
| 18 | Tversky, A., & Kahneman, D. (1983). Extensional versus intuitive reasoning: The conjunction fallacy in probability judgment. Psychological Review, 90(4), 293–315. https://doi.org/10.1037/0033-295X.90.4.293 |
| 19 | Wason, P. C. (1968). Reasoning about a rule. Quarterly Journal of Experimental Psychology, 20(3), 273–281. https://doi.org/10.1080/14640746808400161 |
| 20 | Winawer, J., Witthoft, N., Frank, M. C., Wu, L., Wade, A. R., & Boroditsky, L. (2007). Russian blues reveal effects of language on color discrimination. Proceedings of the National Academy of Sciences, 104(19), 7780–7785. https://doi.org/10.1073/pnas.0701644104 |