Divergent Association Task: Intelligence Test or Creativity Task?

Is the Divergent Association Task an intelligence test or a creativity task?

The Divergent Association Task (DAT) is a brief verbal divergent-thinking task with creativity-related evidence, not an IQ test. Its result should not be converted into an IQ score or treated as a measure of general intelligence. The distinction is about what the task and its evidence support: a score from this exercise does not become an intelligence estimate simply because divergent thinking and intelligence show some association in broader research. That wider relationship is relevant, but it does not establish what an individual DAT result means. For a reader deciding how to describe a result, the defensible label is performance on a verbal divergent-thinking task; the evidence does not warrant an IQ interpretation. The word ‘creativity’ describes the research context for the task, not a license to treat one score as a complete account of someone’s creative work or capacity. Keep the result attached to the task that produced it.

What does the DAT score actually record?

In the original task described in “Naming unrelated words predicts creativity,” a participant generates ten words intended to be as different from one another as possible. The score is based on semantic distance: how far apart the meanings of the chosen words are according to the scoring approach. The response-to-score chain therefore begins with a person selecting words, not with answers to a broad set of questions sampling every kind of ability. The score summarizes a particular feature of that set of verbal responses under the task’s instructions and scoring model.

The creators’ “Divergent Association Task: Frequently Asked Questions” explains that the scoring approach compares word meanings and that the score can be affected by deliberate response strategies. In practical terms, the task depends on which words a participant knows and retrieves, which ten they choose, and how they understand the instruction to make them unrelated. Someone may select familiar words that are easy to retrieve; another person may search longer for unusual combinations. Those different response paths can shape the submitted set even when both people attempt the same operation. The task language also matters: word meanings and their distances are language-dependent, so a score is tied to the language and scoring resources used rather than being a language-free reading of thought.

Scoring constraints further define what the number captures. A semantic-distance model can consistently apply its rules to submitted words, but the resulting number is still a summary of those words under that model. Consistent scoring answers how responses were processed; it does not, by itself, establish that the score represents overall creativity, general verbal ability, or intelligence. Those are broader interpretations requiring evidence matched to each claim. The careful reading is narrower: the score records how semantically separated a participant’s self-generated words were under a specified task, language, instruction, and scoring procedure.

This also explains why the score should travel with a description of how it was obtained. A task translated into another language, administered with different instructions, or scored with a different model may no longer represent the same response-to-number process. Even within one version, knowing the scoring rule can influence how a participant selects words. These are interpretive conditions, not proof that any particular response is poor or that the score is unusable. They tell the reader what information is needed to understand the number: the word set, task language, instructions, and scoring procedure that produced it. A simple automated calculation may be easier to apply uniformly than human judgments of many responses, but uniform application alone cannot settle the meaning of the resulting score.

Sources: Naming unrelated words predicts creativity; Standards for Educational and Psychological Testing

What does the DAT validation evidence support?

The primary paper, “Naming unrelated words predicts creativity,” tested whether the semantic-distance score from the Divergent Association Task corresponded with other measures used to study divergent thinking and creativity. In Study 1, the authors report a moderate-to-strong correlation with the Alternative Uses Task and the Bridge-the-Associative-Gap Task. These comparisons matter because they place DAT performance alongside established tasks that ask for forms of divergent or associative thought. Correspondence among measures designed to engage related operations is convergent evidence: the DAT score behaves in ways that are consistent with a narrow interpretation as a measure of divergent thinking. It is evidence about how this score relates to these comparison tasks in the reported study, rather than proof that every activity called creative is captured by the same number. The paper’s title should be read in that evidentiary sense. Its reported result is about associations with creativity-related measures, not an experiment showing that a high DAT score causes creative work or that the score ranks a person’s overall creative accomplishment.

The same paper also reports positive correlations between DAT performance and a range of problems known to predict creativity. This extends the comparison beyond the two named tasks: DAT scores were associated with performance on other problems selected for their relationship to creativity. That pattern gives a reader a reason to treat the DAT as more than an arbitrary word game, because its output tracked several relevant comparison measures. But the outcome remains performance on those problems. The abstract does not establish that the DAT predicts an individual’s later artistic, scientific, or practical achievements, nor does it show that a particular score forecasts success in a real-world domain. Predicting creativity-related task performance and predicting consequential accomplishment are different claims, with different outcomes to measure. The former is what the reported comparisons can support; the latter would require evidence directly relating scores to defined achievements over an appropriate period and in a stated population. The available account does not supply that basis, so it would be inaccurate to turn the word “predicts” in the paper title into a broad promise about what a person will create. The word “predicts” can describe a statistical relationship with an outcome in a study; it does not necessarily mean that a tool has been shown to forecast an individual future achievement. Here, the outcomes described are other creativity-related tasks and problems. That boundary preserves the paper’s meaningful finding without silently changing its outcome from test performance to life accomplishment.

The authors characterize the DAT as a brief, reliable, and objective measure of divergent thinking. That conclusion is relevant: brevity describes the task’s compactness, objectivity reflects its scoring approach, and the authors’ reliability characterization speaks to consistency in the evidence they report. It helps explain why they propose the DAT as a practical measure for the task domain they studied. Yet those properties do not make the measure interchangeable with every creativity assessment or establish that its score represents intelligence. Nor does a conclusion based on the reported studies automatically transfer to a different language, population, administration, or decision. Each proposed use needs evidence matching its own conditions and interpretation. A result can support one limited research purpose while remaining unsuitable for a different one, even when the same name and scoring procedure are involved. In other words, reliability and objective scoring can strengthen the case for using a brief divergent-thinking measure while leaving open whether a given score supports a broader conclusion. The useful reading is conditional: these comparisons support an interpretation of DAT performance as related to divergent-thinking and creativity-task performance in the studied context; they do not establish a general measure of creative achievement, a measure of IQ, or an individual forecast beyond those outcomes.

Sources: Naming unrelated words predicts creativity

Does the link between divergent thinking and intelligence make the DAT an IQ test?

There is a real association to account for. “The Relationship between Intelligence and Divergent Thinking—A Meta-Analytic Update” combined evidence from 112 studies, bringing together 849 correlation coefficients and an overall sample of 34,610. Its three-level meta-analysis found a positive average relationship between intelligence and divergent thinking across the varied measures represented in those studies. In plain terms, a correlation describes how two measured characteristics vary together across observations: when one differs, the other tends, on average, to differ in a patterned way. A positive average relationship means the measure families overlap to some degree in the literature. It is fair counterevidence to any claim that divergent thinking and intelligence are wholly unrelated. It does not mean they are identical, that one causes the other, or that the relationship can be used as a conversion formula. The study’s scope is broad and heterogeneous: it synthesizes many measures of intelligence and divergent thinking, rather than testing the DAT as a specific instrument against an IQ test. The three-level approach accounts for the fact that a research synthesis may contain multiple correlation estimates within its evidence base, while its large combined count does not make all measures or samples interchangeable. The accessible abstract reports an overall positive relationship, but does not provide a DAT-only estimate that could answer the reader’s score question.

That distinction controls what the meta-analysis can tell someone looking at a DAT result. A relationship across broad measure families is a finding about how those measures tend to co-vary across studies and samples. It does not, by itself, provide a score mapping for the DAT, establish a reference group for an individual DAT score, or show that the score estimates a person’s IQ. To make any of those individual claims, one would need evidence about this named task, its scoring, the relevant population, and the interpretation being proposed. The meta-analysis, as described in the accessible record, does not report such a DAT-specific validation or a method for converting DAT performance into IQ. Its positive average association therefore matters as background: it says the constructs are not wholly disconnected. But a background association cannot supply missing task-specific comparisons, precision, or individual interpretation. A person’s score on a verbal word-generation exercise remains a score on that exercise unless evidence supports a more ambitious reading.

The inference that a broad average association cannot validate a particular person’s DAT score is the reasoning applied here; it is not a result the meta-analysis directly tested. The authors synthesized relationships among intelligence and divergent-thinking measures. They did not, on the evidence available in the abstract, ask whether a DAT score accurately estimates IQ for an individual, whether a particular score band corresponds to an IQ range, or whether users can make defensible decisions from that estimate. This difference between construct-level association and score-level interpretation is easy to miss because both claims use the language of a “link.” The first concerns overlap in patterns across studies; the second requires evidence that a particular score supports a particular conclusion for an individual. Accepting the meta-analysis’s positive average result does not require collapsing these claims. It supports saying that divergent thinking and intelligence are associated to some extent across the studied measures, while leaving the DAT’s individual IQ meaning unestablished. That is a narrower conclusion than either “the abilities are unrelated” or “the DAT is an IQ test,” and it preserves what the evidence actually shows.

Sources: The Relationship between Intelligence and Divergent Thinking—A Meta-Analytic Update

What evidence would justify an intelligence interpretation?

Validity is the evidence supporting a particular interpretation and use of scores. The Standards for Educational and Psychological Testing make the proposed claim the starting point: before calling a score evidence of intelligence, specify what “intelligence” is meant to describe and what decision the score is meant to inform. A general label such as “cognitive” or “creative” does not settle that question. The label names a topic or task family; validity concerns whether the evidence supports the conclusion drawn from this score under stated conditions.

For the Divergent Association Task, the first check is the construct: does the proposed meaning concern verbal divergent-thinking performance, or is someone making the broader claim that the score estimates general intelligence? Those interpretations require different evidence. The second check is the people: were the relevant participants similar to the population for whom the conclusion is now being made? A result observed in one group does not automatically describe adults, younger students, or another population. Third, compare the administration conditions. Language, instructions, scoring, and response procedure must be sufficiently aligned for the earlier evidence to bear on the current score. A translated prompt or a new interface can change the response being measured, even if the task keeps its familiar name.

The fourth check is the comparison: does evidence connect this task’s scores to a suitable measure of the intended construct, in the population and conditions that matter? A correlation can show that two measured outcomes vary together in a study. It cannot alone show that one score is an accurate individual estimate of the other, supply a conversion, or support a decision. The comparison must be close enough to the proposed use, and the evidence must address the interpretation rather than merely attach the task to a broad category. For an intelligence interpretation of the DAT, the key question is therefore whether relevant task-specific evidence supports that score meaning for the intended people and setting. A task name, an abstract resemblance, or one association cannot stand in for that chain. The standards also direct attention to whether the evidence applies in the actual setting, rather than assuming that a finding transfers unchanged whenever a familiar construct label appears.

Finally, consider the consequence. A tentative description of performance in a research exercise demands less than a score used to make an educational, clinical, or employment decision. As the Standards for Educational and Psychological Testing emphasize, evidence and theory must support the proposed use; the burden rises when a mistaken interpretation could materially affect someone. On the evidence considered here, the responsible description remains task-specific rather than an IQ estimate. That is a scope judgment, not a claim that the DAT has no value: evidence could support a narrow interpretation without supporting every possible use. If a future study supports one limited intelligence-related interpretation, that would not automatically establish a conversion for individuals or justify a high-stakes decision. The reader should ask whether each link—meaning, people, conditions, comparison, and consequence—matches the particular claim being made.

Sources: Standards for Educational and Psychological Testing

What changes across language versions and response formats?

A translation does not by itself show that two versions produce scores with the same meaning. The development of the Japanese version, described in “Development and validation of Japanese version of divergent association task,” illustrates why an adaptation needs evidence tied to its own language and procedures. The team built DAT-J using word vectors drawn from a Japanese corpus, rather than assuming that English word relationships could simply be carried over. This matters because the task’s score depends on semantic distances among the words a participant supplies; the language resource is part of how those responses become a number. An adapted score therefore needs its own account of how the language-specific material and scoring relate to the intended interpretation.

The accessible abstract describes a sequence of three surveys rather than a single translation check. An initial survey involved 95 art majors. A modified-version survey involved 358 students and included a two-week retest. The researchers also ran an experiment comparing response formats. Across this sequence, the study reports development work, retesting, and evidence the authors characterize as basic validity and reliability evidence for DAT-J. That process supports a more specific conclusion than “the Japanese text was translated”: this adaptation was examined with Japanese-language materials and the reported participants and procedures. The samples and steps are relevant to judging what was studied, but they do not turn the results into a universal statement about every Japanese speaker or setting. The abstract identifies art majors in the first survey and students in the modified-version survey; those descriptions tell readers something about the participants, while leaving other populations unrepresented in the reported summary. Retesting within the Japanese version also addresses repeat performance under its described conditions, not whether a Japanese score and an English score can be placed on one common scale. That cross-version question needs evidence that directly compares the language versions for the intended interpretation. This distinction follows from the design question: evidence for developing and examining one adaptation is not automatically evidence for equating it with its source version.

The response-format result deserves its own precise reading. In “Development and validation of Japanese version of divergent association task,” the authors report no score effect across the response formats they compared. That is useful evidence that the tested formats did not produce a detected difference under that experiment. It answers a practical question about those particular procedures; it does not show that every screen, interface, or way of submitting words is interchangeable. Nor does the result compare Japanese scores directly with English scores. A format comparison within an adaptation and a cross-language equivalence study answer different questions, even though both concern whether scores can be compared.

Before comparing two reported results, record four things: the version, the language, the scoring model or language corpus, and the response procedure. Then ask whether the available comparison actually covered those same features and the relevant participants. If a reader compares an English DAT result with a DAT-J result, the Japanese study supplies context about the adaptation and its tested formats, but its abstract does not establish that equal scores mean equal levels of divergent thinking across languages. It also does not establish that either score measures IQ. Those conclusions would require direct evidence designed for those interpretations and comparison conditions. The practical rule is to preserve the version and conditions alongside the score, and to avoid treating a shared task name as proof of score equivalence.

Sources: Development and validation of Japanese version of divergent association task

What changes when someone retakes the DAT?

A repeat DAT score has to be read in light of what the participant learned between attempts. In “Divergent Association Task: Frequently Asked Questions,” the task creators report a two-week retest observation in which participants did not receive score feedback between administrations. That is a specific observation under a stated condition: the FAQ does not present it as a universal reliability estimate, a guarantee that every person will repeat the same result, or evidence that the score measures intelligence. It does, however, counter the idea that every repeat attempt is automatically meaningless. When the task is repeated without feedback, the creators’ reported observation offers a limited reference point for considering whether performance can remain consistent over that interval. The claim should travel with its conditions: same named task, short interval, and no score feedback described by its creators.

A different interpretive situation arises after a participant sees how a score was produced. The same FAQ warns that knowing the scoring approach can make it easier to choose responses intended to raise or lower the result. Since the score is based on semantic distances among submitted words, a person who understands that feature may construct a response set strategically instead of responding as they did before learning the rule. That possibility concerns the conditions under which the answers were generated; it does not establish that the participant’s general intelligence or creativity changed. Nor does the warning show that every informed retake is manipulated. It means that a later number can reflect both the task performance and the participant’s knowledge of how the scoring works, so apparent change cannot automatically be assigned to a change in the underlying ability of interest.

For practical interpretation, record what happened between administrations: the interval, whether feedback was shown, and whether the participant learned the scoring logic or a response strategy. Treat a no-feedback retest as a repeat under the creators’ described condition, and a feedback-informed retake as a different response context. Do not select the highest score as the most truthful one, or treat the newest score as independent confirmation merely because it came later. Those choices discard information about how the responses were produced. A practice effect is a score change arising from prior exposure or practice; the term describes a possible source of difference, not proof that ability has improved or declined. Here the useful distinction is concrete: repeat exposure without scoring feedback and a retake after learning the scoring mechanism do not carry identical interpretive weight. If the conditions are unknown, the score history is incomplete, and strong claims from the difference are unwarranted.

Sources: Divergent Association Task: Frequently Asked Questions

What can a DAT result not tell you about one person?

A DAT result can describe one person’s verbal responses on one administration, interpreted through the task’s scoring procedure. It cannot, by itself, turn those responses into a claim about the person’s IQ, diagnosis, fixed creative identity, or future achievement. That boundary follows from keeping the observation and the conclusion at the same scale: the observed unit is a set of words produced under particular instructions and conditions; the broader trait or future outcome is not directly observed in that moment. This is an interpretation rule derived from the scope of the task, not a finding from a new study. It also avoids making the opposite mistake of treating the result as empty. The score may be informative as one sample of verbal divergent-thinking behavior, while remaining too narrow to settle who someone is or what they can become.

A compact way to keep those levels separate is to ask what the administration actually records before deciding what the result means: | The result can describe | The result alone cannot establish | | --- | --- | | The words a person submitted under the task instructions | The person’s general intelligence or IQ | | How those words scored under the specified procedure | A diagnosis or clinical status | | Performance in this task context at this time | A permanent creative identity or a future achievement | These contrasts are practical limits on interpretation, not claims that the task has no useful signal. They prevent a narrow performance observation from silently acquiring meanings that require separate evidence. A score can prompt a question about the person’s approach to this particular exercise; it does not answer every question that might have motivated the person to take it.

Consider a clearly hypothetical case: someone receives a result they regard as unusually strong and asks whether that proves they are highly intelligent. The defensible reply stays with the observable event: they produced this response set, and it received this score under the task’s rules. The result may encourage reflection on how they generated varied words, but the step from that performance to an IQ conclusion is not supplied by the score itself. The same boundary applies to a disappointing result: one administration does not establish low intelligence, lack of creativity, or a ceiling on future work. These examples are illustrations of the reasoning rule, not reported participants or test outcomes. When a real decision about education, clinical care, or employment is at stake, the task result should not stand in for an appropriate assessment and the other evidence that decision requires. Keep the conclusion proportional to what was observed: describe the administration, avoid converting it into an identity or forecast, and seek evidence suited to any consequential question.

What should a reader do with a DAT result?

If you took the DAT out of curiosity, keep the result as a description of your verbal performance on that task and administration. It can be an interesting prompt to reflect on how you generated unrelated words; it does not need to become a judgment about your intelligence or a reason to make a consequential choice. A browser exercise can have value as an engaging activity or a way to explore one kind of response, provided you keep that low-stakes purpose separate from an IQ interpretation.

If your real question is whether you or someone else has a particular level of intelligence, or the answer could affect education, clinical care, or another important decision, bring that question to a qualified professional. Ask what assessment is appropriate for the person and decision, what evidence supports interpreting its scores for that population, and what limits apply. A DAT result alone is not a substitute for that assessment. The useful next step depends on the decision you need to make, not on whether one informal result feels high or low.

For low-stakes reasoning practice, you can take Test IQ Free’s 50-question test at [/test](/test). Its displayed estimated IQ range is provisional, comes from an unvalidated conversion with no validated adult reference population, and is neither a normed comparison nor a confidence interval. The optional $9 report adds explanation, worked examples, and a practice plan using that same provisional conversion; payment adds explanation, not validity. Take the free test if practice is your goal, and treat its range accordingly.

Sources: Test IQ Free 50-question reasoning test

Questions readers ask