Thinking, Intelligence and Creativity

Overview

Thought can be studied as internal representation and process. Associationist views emphasise links between ideas built by experience; Gestalt views emphasise restructuring and insight, where a problem is suddenly seen in a new organisation; information-processing views model thought as operations on symbols, with working memory limits shaping what can be held and manipulated; connectionist views model thought as patterns across networks of simple units. Language is both a tool of thought and a constraint on it: it allows abstract reasoning, planning and communication, but categories embedded in language can also channel attention and inference. The relation is therefore bidirectional, not a simple claim that language determines thought or that thought is independent of language. Inner speech, imagery and propositional reasoning are different formats in which thought can occur, and skilled thinkers switch formats to suit the problem.

Theoretical perspectives on thought

Behaviourist perspectives treated thought as covert behaviour or as verbal mediation, testable only through its observable correlates, and they struggled to explain generative, novel reasoning. Gestalt perspectives treated thought as reorganisation: a problem solver restructures the problem representation until the solution becomes apparent, as in insight problems where the barrier is a fixed representation rather than lack of knowledge. Cognitive perspectives model thought as computation over representations, distinguishing declarative knowledge, knowing that, from procedural knowledge, knowing how, and explaining reasoning errors through limits of working memory, use of heuristics and failure to represent the problem correctly. Developmental perspectives show thought changing qualitatively with age, from concrete operations tied to present objects to formal operations that manipulate hypotheses systematically, though later research shows this progression is more gradual, domain-specific and culturally variable than a strict stage account suggests.

Sociocultural perspectives add that thought is mediated by tools, language and social interaction: children learn to think by participating in joint activity with more skilled partners, internalising dialogue as inner speech and strategies as mental tools. Dual-process perspectives distinguish fast, intuitive, associative thinking from slow, deliberate, rule-based thinking, explaining why a quick judgement can be both efficient and systematically biased. No single perspective explains all thought. Insight problems favour Gestalt accounts, syllogistic errors favour cognitive and dual-process accounts, and the development of planning favours sociocultural and developmental accounts. The examination skill is to match perspective to phenomenon and to state what each perspective predicts that the others do not.

Concept formation, reasoning, language and thought

Concepts are mental categories that organise experience by grouping instances that share features and separating them from non-instances. Classical views defined concepts by necessary and sufficient features, but most natural concepts have graded structure: some instances are prototypical, such as a robin for bird, while others are marginal, and membership is judged by similarity to a prototype or by resemblance to stored exemplars. Concept formation proceeds through hypothesis testing, where learners test rules against feedback, through abstraction of common features across examples, and through instruction that names the concept and its boundaries. Overgeneralisation and undergeneralisation are the twin errors of category learning, and concept attainment improves when examples and non-examples are contrasted and when learners explain why an instance belongs or does not.

Reasoning moves from given information to new conclusions. Deductive reasoning draws conclusions that follow necessarily from premises, as in syllogisms, and errors arise from belief bias, where a believable conclusion is accepted regardless of validity, and from atmosphere effects and illicit conversions. Inductive reasoning draws probable conclusions from examples, and its strength depends on sample size, diversity and representativeness. Analogical reasoning maps structure from a familiar source to a new target and is powerful when structural relations, not surface features, are mapped. Language and thought interact through inner speech in planning, through labels that aid categorisation, and through discourse that makes reasoning public and correctable. Claims that language wholly determines thought overstate the case, since prelinguistic infants and non-human animals show categorisation and problem solving, but language undeniably extends thought by supporting abstraction, hypothetical reasoning and cultural transmission.

Problem solving and decision-making

Problem solving moves from an initial state through a problem space of possible moves toward a goal state, guided by a representation of the problem. Stages commonly distinguished are recognising the problem, defining and representing it, generating strategies, organising information, allocating resources, monitoring progress and evaluating the solution. Strategies include algorithms, which guarantee a solution if followed, and heuristics, which are efficient shortcuts that usually help but can mislead. Means-ends analysis reduces the difference between current and goal states step by step; working backwards starts from the goal; analogy transfers a known solution structure. Obstacles include functional fixedness, seeing an object only in its usual function; mental set, persisting with a previously successful strategy; irrelevant information that overloads working memory; and incorrect representation that makes the solution invisible until restructuring occurs. Expertise differs from novice performance less in raw capacity than in organised knowledge, pattern recognition and representation quality.

Decision-making chooses among options under uncertainty, weighing probabilities, values and costs. Normative models such as expected utility prescribe choosing the option with the highest weighted value, but descriptive research shows systematic departures. Heuristics such as availability, judging likelihood by ease of recall, representativeness, judging by similarity to a prototype, and anchoring, over-relying on an initial value, produce predictable biases. Framing effects show that the same outcome described as a gain or a loss changes choice, violating strict utility consistency. Prospect theory captures loss aversion, where losses weigh more than equivalent gains, and diminishing sensitivity to larger amounts. Group decision-making adds social dynamics: groupthink, polarisation and diffusion of responsibility can degrade decisions that individuals would make better alone. Good decision support therefore restructures representation, makes base rates explicit, separates generation of options from evaluation, and builds in dissent and review.

Metacognition

Metacognition is cognition about cognition: knowledge about one's own mental processes and regulation of those processes. Metacognitive knowledge includes knowledge about persons, including oneself as a learner; about tasks, including their demands and difficulty; and about strategies, including when and why a strategy works. Metacognitive regulation includes planning before a task, monitoring during it, and evaluating after it. Monitoring accuracy matters more than confidence: learners who overestimate their understanding stop studying too early, while learners who monitor through self-testing calibrate better. Feelings of knowing, judgements of learning and ease-of-processing fluency are cues to monitoring, but fluency can mislead, as when rereading feels easy without improving retrieval.

Metacognition explains individual differences in study, problem solving and reading comprehension that ability alone does not. Skilled readers monitor comprehension, reread when meaning breaks down, and summarise to check understanding; unskilled readers may continue without noticing confusion. In problem solving, metacognitive control means noticing an impasse, switching strategy, and checking a solution rather than persisting blindly. Metacognition can be taught: explicit strategy instruction, self-questioning, prediction before reading, and reflective evaluation after tasks improve regulation, especially when combined with subject knowledge rather than taught as abstract slogans. Its limits are real: monitoring is effortful, it competes with task processing under load, and young children and novices may lack the knowledge base needed to monitor accurately. The practical lesson is to design tasks that make monitoring cheap and informative, through worked examples, checklists and low-stakes self-testing.

Intelligence: major theories

Spearman proposed a general factor, g, underlying performance across diverse tests, alongside specific factors for each test, based on positive correlations among abilities. Thurstone countered with primary mental abilities, a set of relatively independent factors such as verbal comprehension, numerical ability, spatial visualisation and memory. Cattell distinguished fluid intelligence, the capacity to reason with novel problems, from crystallised intelligence, accumulated knowledge and skill, with fluid ability more vulnerable to ageing and brain injury. Guilford proposed a structure-of-intellect model with operations, contents and products, generating many narrow abilities, a model criticised for over-fragmentation and weak evidence for its full structure. Hierarchical models reconcile these views: a general factor at the top, broad group factors below, and narrow abilities at the base, a structure supported by factor analysis of large test batteries.

Gardner proposed multiple intelligences, including linguistic, logical-mathematical, spatial, musical, bodily-kinaesthetic, interpersonal, intrapersonal and naturalistic, arguing that intelligence is plural and culturally valued in different domains. Its educational influence is large, but its psychometric evidence is contested, because the proposed intelligences correlate and lack independent measurement. Sternberg proposed a triarchic theory of analytical, creative and practical intelligence, emphasising adaptation to real-world contexts, and later a theory of successful intelligence balancing those components. Emotional intelligence, popularised as the ability to perceive, understand and regulate emotions in self and others, overlaps with personality and social skill, and its measurement and distinctness from general intelligence remain debated. Intelligence testing must therefore be judged by purpose: tests predict school and job performance moderately well, but they are sensitive to schooling, language, culture and opportunity, and they measure developed abilities, not a pure innate capacity. Heritability estimates for intelligence are population statistics that rise with age and vary with environment, and they do not imply immutability or justify ranking groups.

Creativity

Creativity is the production of ideas or products that are both novel and useful or appropriate within a domain. Novelty alone is eccentricity; usefulness alone is routine competence. Creative thinking is often described through divergent thinking, generating many varied ideas, contrasted with convergent thinking, narrowing to one correct answer. Guilford distinguished fluency, the number of ideas; flexibility, the variety of categories; originality, the rarity of ideas; and elaboration, the detail of development. The creative process is commonly described in stages: preparation through knowledge and problem finding; incubation, where the problem rests while unconscious processing continues; illumination, the moment of insight; and verification, where the idea is tested and refined. These stages are not strictly sequential, and incubation effects are strongest when prior work has thoroughly engaged the problem.

Creativity depends on person, process, product and press, the environment. Person factors include knowledge, because creativity without domain knowledge is usually shallow; cognitive style, including tolerance for ambiguity and willingness to take intellectual risks; motivation, where intrinsic interest sustains persistence better than extrinsic reward alone; and personality traits such as openness to experience. Process factors include analogy, combination of distant ideas, restructuring and deliberate practice. Press includes cultural values, educational climates that tolerate dissent and failure, and resources that allow exploration. Assessment uses divergent thinking tests, ratings of products by domain experts, biographical and personality measures, and real-world creative achievement, each with limits: divergent tests predict creative potential weakly, expert ratings are costly and domain-specific, and achievement reflects opportunity as well as ability. Teaching for creativity therefore combines knowledge building, divergent practice, safe risk-taking, feedback on usefulness as well as novelty, and time for incubation and revision.

Intelligence theories compared

Creative process and assessment