Divergent Thinking

Abstract

Divergent thinking is the generation of many different responses to a single open-ended prompt, the cognitive process most closely identified with the ideational side of creativity. Guilford introduced the term within his Structure of Intellect model in the 1950s, distinguishing it from the convergent thinking that narrows toward one correct answer. Performance is scored along four dimensions: fluency, the number of responses; flexibility, the number of categories they span; originality, their statistical rarity; and elaboration, their added detail. Standard tasks such as naming unusual uses for a brick, or the Torrance Tests of Creative Thinking, operationalise these dimensions and yield scores treated as indicators of creative potential rather than realised achievement. This article defines the construct, distinguishes its dimensions, works a scoring example, and reviews what the scores predict.

Keywords: divergent thinking, fluency, originality, creative potential

A person asked to name every use they can think of for a brick faces a divergent problem: there is no single right answer, only a space of possibilities the mind must open up and explore. That same person, asked instead what a brick and a loaf of bread have in common and required to give the one best word, faces a convergent problem, one that narrows toward a correct solution. Divergent thinking is the cognitive process behind the first kind of task — the fluent, branching generation of multiple, varied ideas from one starting point (Guilford, 1956). It is the component of intelligence that classical ability tests, built almost entirely from convergent items, were long criticised for ignoring.

The construct matters because it gave creativity research something it had lacked: a way to measure the ideational raw material of creative work with paper-and-pencil tasks that could be scored and compared. That promise has always been shadowed by a caution the field now states plainly — a divergent-thinking score is an estimate of creative potential, not a certificate of creative achievement (Runco & Acar, 2012). The sections below define divergent thinking, trace it to Guilford's model of the intellect, separate it from its convergent counterpart, set out the four dimensions on which it is scored, describe the tasks that measure it, and weigh what those measurements actually predict.

## What Divergent Thinking Is

Divergent thinking is the production of numerous, varied ideas in response to a prompt that admits many acceptable answers. The defining feature is not the quality of any single idea but the breadth of the search: the thinker moves outward from the starting point, generating alternatives that differ from one another rather than homing in on one solution (Guilford, 1967). Where a convergent problem has a criterion that decides right from wrong, a divergent problem has only a space of possibilities that can be more or less fully and more or less unusually explored.

It is important to be precise about what the term does and does not name. Divergent thinking is not a synonym for creativity, and it is not itself a personality trait or an achievement; it is a measurable cognitive ability to generate ideational variety, one that creativity theorists treat as a contributing component of the larger creative act (Runco & Acar, 2012). Creativity, on most accounts, requires that ideas be both novel and useful, and usefulness is a judgement divergent-thinking tasks do not by themselves make. A person can score highly on idea generation and never produce anything the world calls creative, and a genuinely creative product depends on convergent evaluation, domain knowledge, motivation and opportunity as much as on the initial spread of ideas. Holding this line keeps the construct doing the specific work it was built for rather than standing in for creativity as a whole.

## Origins in the Structure of Intellect

The term entered psychology through J. P. Guilford, whose 1950 presidential address to the American Psychological Association reproached the field for having almost wholly neglected creativity in favour of the convergent abilities measured by intelligence tests (Guilford, 1950). Guilford set out to give creativity a place inside a formal model of the mind. In his Structure of Intellect model he classified intellectual abilities along three facets — the mental operation performed, the content operated on, and the product that results — and among the operations he distinguished divergent production from convergent production (Guilford, 1956). Divergent production named the operation of generating multiple, logically diverse answers from given information; convergent production named the operation of arriving at the one answer the information determined.

That distinction is Guilford's lasting bequest to creativity research, and he made it central because he believed the standard intelligence test, by scoring only convergent production, systematically missed the abilities that creative work draws on (Guilford, 1967). The Structure of Intellect model as a whole, with its many dozens of hypothesised factors, has not survived as a theory of intelligence; few psychologists today defend its full factorial architecture. But the convergent/divergent contrast was extracted from it, took on a life of its own, and now organises a large empirical literature that has long outlived the model that bore it. The next section makes that contrast concrete.

## Divergent and Convergent Thinking

Convergent and divergent thinking are best understood as two complementary modes of directed thought, not as opposites in a contest. Convergent thinking applies logic and accumulated knowledge to move from many inputs to the single best or correct answer, the operation tapped by most items on an intelligence test and by puzzles with a determinate solution. Divergent thinking runs the other way, expanding from one input to many candidate outputs, the operation tapped by open-ended generation tasks. A well-known convergent measure, the Remote Associates Test, gives three words such as cottage, Swiss and cake and asks for the one word that links them all — here, cheese (Mednick, 1962). A divergent measure gives one object and asks for as many uses as the respondent can produce. The demonstration below sets the difference between searching for the answer and searching for answers side by side.

Mednick's associative account is worth dwelling on because it supplies a mechanism for why people differ. He proposed that creative solutions come from connecting mental elements that are only remotely associated, and that individuals with flat associative hierarchies — for whom a given cue evokes many associates of roughly comparable strength — can reach more distant and unusual combinations than individuals with steep hierarchies dominated by a few strong, conventional associates (Mednick, 1962). On this view divergent thinking reflects the shape of a person's associative memory, and the tendency of fluent responders to give conventional answers first and unusual answers later, once the obvious associates are exhausted, is a robust and much-studied feature of the tasks. Full creative performance, though, needs both modes in sequence: divergent generation to open a wide field of possibilities, then convergent evaluation to select and refine the ones worth keeping.

## The Four Dimensions

Guilford proposed that divergent production could be scored along several distinct dimensions, and four have become standard (Guilford, 1967). Fluency is the sheer number of appropriate responses a person generates — the size of the idea pool. Flexibility is the number of different conceptual categories those responses fall into; a respondent who lists ten uses for a brick that are all forms of building shows high fluency but low flexibility, because the ideas do not range across categories. Originality is the statistical rarity of the responses relative to some comparison group, so that answers few others produce earn more credit than common ones. Elaboration is the amount of detail worked into each response, the degree to which a bare idea is developed into a fuller one.

The four are conceptually separable but empirically entangled, and the entanglement is the source of the field's most persistent measurement headache. Fluency and originality in particular tend to correlate strongly, because a respondent who generates many ideas is, other things equal, more likely to hit on a rare one; the unusual answers typically arrive after the obvious ones are used up. As a result an originality score computed by counting rare responses can be partly a disguised measure of fluency, a confound that has driven much of the work on how divergent-thinking tasks ought to be scored (Silvia et al., 2008). The demonstration below scores a set of responses to the Alternative Uses task on the first three dimensions at once, so the way fluency, flexibility and originality come apart can be seen directly.

## Measuring Divergent Thinking

Two families of task dominate measurement. Guilford's own battery included the Alternative Uses task, which asks for unusual uses of a common object, along with tasks calling for consequences of hypothetical events and for many ways of completing a figure or a sentence. The most widely used instrument, however, is the Torrance Tests of Creative Thinking, a standardised battery of verbal and figural activities that Ellis Paul Torrance developed and normed for use with children and adults, scoring each on fluency, flexibility, originality and elaboration (Torrance, 1972). The Torrance tests turned divergent thinking from a laboratory curiosity into a practical assessment used in schools and gifted-education programmes for decades, and they remain the most heavily researched creativity measure in existence.

Scoring is where the real difficulty lies, and it comes in two broad styles. The classical approach scores originality by statistical infrequency: responses are compared against a normative sample, and a response earns originality credit in proportion to how rarely it appears in that reference pool. The alternative, developed to escape the fluency confound, has trained raters make subjective quality judgements of a person's best responses, an approach that improves the separation of originality from sheer output but depends on rater agreement and careful instructions (Silvia et al., 2008). Both styles are sensitive to the testing conditions, and that they are is an old finding rather than a new one. Wallach and Kogan (1965) argued that ideation should be assessed in a game-like, untimed and non-evaluative setting, and reported that when it was, divergent-thinking measures cohered into a dimension only weakly related to intelligence, whereas timed, test-like administration pulled the same measures back toward general ability. On that early evidence the separability of creative ideation from intelligence — the property that made the construct worth measuring at all — is partly a product of how the ideation is elicited. The point has since been placed on a quantitative footing: a meta-analysis by Said-Metwaly and colleagues found that time limits and the exact wording of the instructions — in particular whether respondents are explicitly told to be creative — systematically change the scores a task yields (Said-Metwaly et al., 2020). The demonstration below shows the statistical-infrequency logic at work.

Table 1 sets the four dimensions side by side with what each rewards and the measurement risk each carries.

## Divergent Thinking and Creative Potential

The central interpretive claim of the field is deliberately modest: divergent-thinking scores are indicators of the potential for creative thought, not measures of creativity itself (Runco & Acar, 2012). The distinction is not a hedge but a substantive position. Creativity is realised in products and performances judged novel and useful by a domain; divergent thinking is one cognitive contributor to producing them, upstream of the evaluation, selection, elaboration and execution that turn a promising idea into a creative accomplishment. Treating a high fluency score as creativity confuses the capacity to generate options with the achievement of having made something valuable from them.

Read that way, the construct is defensible and useful. Divergent-thinking tasks capture individual differences in ideational productivity that ordinary intelligence tests miss, and those differences carry real if limited predictive weight for later creative behaviour, a point the evidence section takes up. The strongest form of the case is not that divergent thinking equals creativity but that no account of creativity can omit the ideational generation these tasks measure, and that among the available measures of that generation the divergent-thinking tradition remains the most developed. Its limits — the fluency confound, the sensitivity to instructions, the gap between potential and achievement — are limits of measurement and interpretation, not reasons to discard the thing measured.

## The Evidence Base

The most consequential empirical question is predictive: do childhood divergent-thinking scores forecast adult creative achievement? Torrance's own longitudinal studies, following schoolchildren across decades, reported that the tests predicted later real-world creative accomplishment (Torrance, 1972). Those claims were later scrutinised by Plucker, who reanalysed Torrance's longitudinal data and found that the divergent-thinking scores did predict adult creative achievement, and did so more strongly than childhood IQ, while cautioning that the effect sizes were moderate and the methodology of the original follow-ups imperfect (Plucker, 1999). The consensus that emerged is a measured one: divergent thinking has genuine predictive validity for creative accomplishment, incremental to intelligence, but it is one modest predictor among several rather than a decisive test of creative destiny.

Two further lines of evidence sharpen the picture. The first concerns the cognitive machinery behind the tasks. Divergent thinking is not a wholly separate faculty walled off from general cognition: fluid intelligence and executive control predict divergent-thinking performance, and the strategies people use — in particular whether they move beyond obvious responses to search for remote ones — account for much of the individual variation (Nusbaum & Silvia, 2011). Neuroimaging has begun to expose the same machinery directly. Benedek and colleagues found that generating genuinely new ideas, as opposed to recalling old ones, recruited regions associated with controlled memory retrieval and executive processing (Benedek et al., 2014), and Beaty and colleagues showed that individual creative-thinking ability could be predicted from the strength of connections among the brain's default, executive and salience networks — evidence that high divergent-thinking performance depends on the cooperation of spontaneous idea generation with top-down control (Beaty et al., 2018).

The second line concerns change over time. Analysing decades of normative Torrance data, Kim reported that American scores on the tests had been declining since the 1990s even as IQ scores rose, a pattern she termed a creativity crisis (Kim, 2011). The finding is contested, and its causes and even its reality remain debated, but it illustrates the double edge of a well-normed instrument: precisely because the Torrance tests were standardised on large samples over many years, they made a population-level trend in measured divergent thinking visible in a way no single study could.

## Worked Example

Consider a participant given the Alternative Uses task for a brick who produces six acceptable responses: build a wall, build a fireplace, use it as a paperweight, use it as a doorstop, grind it into red pigment, and scrape two together as a percussion instrument. Scoring proceeds one dimension at a time. Fluency is simply the count of appropriate, non-redundant responses: six. Flexibility counts the distinct categories those six span — construction (wall, fireplace), anchoring by weight (paperweight, doorstop), art material (pigment) and music (percussion) — giving four categories. Originality is scored here by statistical infrequency against a norm sample, awarding two points to a response given by fewer than 5 percent of respondents, one point to a response given by 5 to 10 percent, and none to a more common response.

Suppose the normative frequencies are 62, 24, 45 and 38 percent for the four common responses and 3 and 2 percent for grinding into pigment and the percussion use. The four common responses earn zero originality points; the pigment and percussion responses, both below the 5 percent threshold, earn two points each, for an originality total of four. The profile — fluency 6, flexibility 4, originality 4 — captures three genuinely different things about the performance: how many ideas came, how widely they ranged, and how unusual they were (Figure 1). Notice that the two originality-earning responses are also the two that arrived last, after the conventional uses were exhausted, which is exactly why raw originality scored this way is entangled with fluency: more responses give more chances at a rare one. The interactive Alternative Uses demonstration above computes this same profile for any subset of responses selected.

Discussion

The history of divergent thinking is a case study in how a measurement idea can outrun and then discipline the theory that produced it. Guilford proposed divergent production as one operation within an elaborate factorial model of intelligence that few now accept; the operation survived the model because it was operationalised into tasks that could be scored, correlated and validated (Guilford, 1956; Guilford, 1967). Those tasks then exposed their own problems — the entanglement of originality with fluency, the sensitivity of every score to time limits and instructions — and the response was not to abandon the construct but to build better scoring, from trained subjective ratings to controlled testing conditions (Silvia et al., 2008; Said-Metwaly et al., 2020).

What holds the field together is the disciplined modesty of its central claim. Divergent thinking is ideational potential, not creative achievement; it predicts later accomplishment incrementally over intelligence but only moderately (Plucker, 1999; Runco & Acar, 2012). Kept in that frame, the construct is one of cognitive psychology's more durable exports, a way of measuring the generative side of thought that has survived sixty years of scrutiny, migrated from paper tasks into the neuroimaging laboratory, and continues to organise research on where new ideas come from.

## Current Directions

Contemporary work is concentrated on two fronts: fixing the measurement and locating the process in the brain. On measurement, Acar and Runco survey a wave of new methods that move past the classic fluency-bound scores — among them approaches that weight the semantic distance between a prompt and a response, and multivariate models that treat a person's whole set of responses rather than a single summary number — and argue for an extended theory in which originality is estimated more directly and less as a by-product of output (Acar & Runco, 2019). The same impulse drives automated scoring: representing responses in a semantic space lets originality be estimated by how far an idea sits from the prompt, reducing the reliance on scarce human raters and on brittle normative frequency tables.

The neuroscience has matured from asking whether creative thinking has a neural signature to asking what that signature is made of. The finding that creative ability can be predicted from the coordination of the default-mode network with executive-control regions reframes divergent thinking as the controlled steering of spontaneous idea generation rather than either alone (Beaty et al., 2018; Benedek et al., 2014). A third strand studies the context sensitivity of the tasks themselves: Dumas and Dunbar showed that simply asking people to imagine themselves as an eccentric poet rather than a rigid librarian raised their divergent-thinking originality, a demonstration that measured divergent thinking depends on the mindset a respondent adopts and not only on a fixed underlying capacity (Dumas & Dunbar, 2016). Across all three fronts the trajectory is the same: toward measures and models that respect how variable, strategic and context-bound the generation of ideas turns out to be.

Common Misconceptions

Divergent thinking is the same thing as creativity.

It is not. Divergent thinking is one cognitive contributor to creativity, an index of the potential to generate ideas, whereas creativity is a realised product judged novel and useful (Runco & Acar, 2012). Generating many ideas is upstream of the evaluation and execution that creative achievement also requires.

More ideas simply means more creative.

Fluency is only one dimension, and its strong correlation with originality is partly a scoring artefact: a person who lists more uses has more chances at a rare one, so an originality score built from counting rare responses can be a disguised fluency score (Silvia et al., 2008).

A divergent-thinking test measures a fixed, unchanging ability.

Scores shift with the conditions of testing. Time limits and instructions change them systematically (Said-Metwaly et al., 2020), and merely adopting an unconventional persona can raise measured originality (Dumas & Dunbar, 2016).

Glossary

Alternative Uses task.

A divergent-thinking task, devised in Guilford's battery, that asks a respondent to list as many unusual uses as possible for a common object such as a brick or a paperclip.

Associative hierarchy.

The ordered set of associations a cue evokes; a flat hierarchy yields many associates of similar strength and supports remote connections, a steep one is dominated by a few conventional associates.

Convergent thinking.

The mode of directed thought that applies logic and knowledge to move from given information to the single best or correct answer; the complement of divergent thinking and the operation most intelligence tests measure.

Creative potential.

The capacity to produce creative work, which divergent-thinking scores are held to estimate, as distinct from realised creative achievement in a domain.

Divergent thinking.

The generation of many varied responses to a single open-ended prompt, scored for fluency, flexibility, originality and elaboration, and treated as an indicator of the ideational potential for creativity.

Elaboration.

A scoring dimension of divergent thinking capturing the amount of detail developed within a response, the degree to which a bare idea is worked into a fuller one.

Flexibility.

A scoring dimension counting the number of distinct conceptual categories a set of responses spans, distinguishing wide-ranging ideation from many ideas of one kind.

Fluency.

A scoring dimension counting the number of appropriate, non-redundant responses a person generates; the size of the idea pool and the dimension most entangled with the others.

Originality.

A scoring dimension crediting the statistical rarity of responses relative to a comparison group, so that answers few others give earn more than common ones.

Remote Associates Test.

A convergent measure of creative thinking, devised by Mednick, that presents three words and asks for the single further word associatively linked to all three.

Statistical infrequency.

A method of scoring originality by comparing responses against a normative sample and crediting each in proportion to how rarely it appears in that reference pool.

Structure of Intellect.

Guilford's three-faceted model classifying intellectual abilities by operation, content and product, within which the contrast between divergent and convergent production was first drawn.

Subjective scoring.

An approach in which trained raters judge the quality or creativity of a respondent's best responses, developed to separate originality from sheer fluency at the cost of depending on rater agreement.

Torrance Tests of Creative Thinking.

The standardised verbal and figural battery developed by E. Paul Torrance, scored on fluency, flexibility, originality and elaboration, and the most widely used and researched measure of divergent thinking.

Key Researchers

Roger E. Beaty (living). Cognitive neuroscientist at Pennsylvania State University, where he directs the Cognitive Neuroscience of Creativity Lab; his work uses functional connectivity to predict individual creative ability from the interaction of the brain's default, executive and salience networks. Homepage - Google Scholar - ORCID

J. P. Guilford (1897-1987). American psychologist at the University of Southern California; introduced the convergent/divergent distinction within his Structure of Intellect model and defined the dimensions on which divergent production is scored, founding the psychometric study of creativity. Wikipedia

Mark A. Runco (living). Creativity researcher and founding editor of the Creativity Research Journal; reframed divergent thinking as an indicator of creative potential and has developed both the theory and the measurement of ideational originality. Homepage - Google Scholar - ORCID

Paul J. Silvia (living). Psychologist at the University of North Carolina at Greensboro; developed subjective scoring methods for divergent-thinking tasks and studied the fluid-intelligence and executive processes underlying performance on them. Faculty Page - Google Scholar - ORCID

E. Paul Torrance (1915-2003). American psychologist at the University of Georgia; created the Torrance Tests of Creative Thinking, the dominant standardised divergent-thinking battery, and conducted the longitudinal studies that first tested its predictive validity. Wikipedia

Frequently Asked Questions

What is divergent thinking?
Divergent thinking is the cognitive process of generating many different responses to a single open-ended prompt, such as listing all the possible uses for a brick. It is scored for fluency, flexibility, originality and elaboration, and is treated as an indicator of the ideational potential for creativity (Guilford, 1967).

How is divergent thinking different from convergent thinking?
Convergent thinking moves from many inputs toward one correct answer, the operation most intelligence-test items measure. Divergent thinking runs the other way, expanding from one input to many candidate answers. Guilford drew the distinction within his Structure of Intellect model, and full creative work uses both in sequence (Guilford, 1956).

What are the four dimensions of divergent thinking?
They are fluency, the number of responses; flexibility, the number of distinct categories those responses span; originality, their statistical rarity relative to others; and elaboration, the amount of detail worked into each response (Guilford, 1967).

How is divergent thinking measured?
The most common measures are Guilford's Alternative Uses task and the Torrance Tests of Creative Thinking, a standardised battery of verbal and figural activities. Responses are scored either by comparing them against normative frequencies or by having trained raters judge their quality (Torrance, 1972).

Does divergent thinking predict real creative achievement?
Modestly but reliably. Reanalyses of Torrance's longitudinal data found that childhood divergent-thinking scores predicted adult creative accomplishment, more strongly than childhood IQ, though the effect sizes were moderate (Plucker, 1999).

Is divergent thinking the same as creativity?
No. Divergent thinking is one cognitive contributor to creativity, an estimate of the potential to generate ideas, whereas creativity is a realised product judged both novel and useful. The field treats divergent-thinking scores as indicators of potential, not as measures of creativity itself (Runco & Acar, 2012).

Why is originality hard to score?
Because it is entangled with fluency: a person who produces more responses has more chances of hitting a rare one, so an originality score built from counting rare responses can be a disguised measure of output. Subjective rating methods were developed to separate the two (Silvia et al., 2008).

What does the brain do during divergent thinking?
Neuroimaging links divergent thinking to the coordination of the default-mode network, which supports spontaneous idea generation, with executive-control regions that steer and evaluate. Individual creative ability can be predicted from the strength of these connections (Beaty et al., 2018).

References