Do Higher Ability Scores Mean Fewer Reasoning Errors?

What is the short answer?

A cognitive-task score and choices in a risky-choice task are different observations. The first records performance on a measure; the second records selections among options with uncertain outcomes. Researchers may infer risk preference from those selections, but that inference depends on task design and scoring. It does not show how often someone makes reasoning errors. “Cognitive abilities affect decision errors but not risk preferences” synthesized 30 studies and 62 effect sizes. It found no credible association between measured ability and risk aversion in the risky-choice studies included. The authors argue task designs can classify inconsistent or random choices as one preference, creating an apparent link. This supports a task-specific error explanation, not proof that higher ability prevents mistakes generally or that preferences never differ. A broader 2019 review, “Cognitive ability and risk aversion: A systematic review and meta-analysis,” reported a weak association in gain choices, but no significant association in mixed gain-and-loss or loss choices. The reviews do not justify a universal answer: measured ability, task performance, and inferred preference must remain distinct. A score alone cannot establish someone’s general error rate or risk appetite, and these associations do not show that changing ability would change either.

Sources: Cognitive abilities affect decision errors but not risk preferences: A meta-analysis | Psychonomic Bulletin & Review | Springer Nature Link; Cognitive ability and risk aversion: A systematic review and meta-analysis | Judgment and Decision Making | Cambridge Core

What does the strongest error-focused synthesis actually test?

The 2022 systematic review and Bayesian meta-analysis asks whether cognitive ability is associated with risk aversion in risky-choice tasks, and whether decision errors can explain that association. It combines 30 studies, represented by 62 effect sizes. Across those studies, the pooled evidence did not support a credible overall association between measured cognitive ability and elicited risk aversion. That does not establish that ability has no relationship to every kind of choice, or that people with different scores make the same number of mistakes in daily life.

A meta-analysis is a statistical synthesis of results from multiple studies. Here, the authors reviewed studies reporting correlations between cognitive measures and behavioral risk-preference tasks, then used a Bayesian model to estimate the overall relationship and examine features that might explain differences across findings. They compare two explanations: a risk-preference account, which proposes a systematic link between ability and underlying preferences, and an error account, which proposes that inconsistent responding can be mistaken for preference when researchers infer it from choices. The abstract reports that apparent correlations can be explained by biased tasks in which errors are classified as a particular preference. The authors favor the error explanation for the pooled pattern in their task set.

This is evidence about associations in study results, not an experiment that increased or reduced participants’ cognitive ability. The review cannot show that raising a person’s score would reduce errors, or determine how accurately anyone reasons outside these tasks. Its conclusion also depends on which tasks were eligible and how choice behavior was classified. A choice that departs from a model’s prediction may reflect inconsistency, but the measure alone does not establish why it happened or prove the person has no relevant preference.

The result is a warning about interpretation: a risk-aversion label may partly reflect task design and response consistency, rather than a clean measure of preference. The review does not settle every domain, measure, or individual case, and a null pooled association is not proof of universal absence. Before applying its headline conclusion, check which cognitive measure and choice outcome were studied, how the task inferred preference, and whether the finding is an association or a causal change.

Sources: Cognitive abilities affect decision errors but not risk preferences: A meta-analysis | Psychonomic Bulletin & Review | Springer Nature Link; Cognitive abilities affect decision errors but not risk preferences: A meta-analysis | PubMed

How can task design make an error look like a preference?

A risky-choice task can turn inconsistent responding into an apparent preference when its choices are unevenly arranged. If more options are scored as risk averse than as risk seeking, random or noisy choices will tend to collect the risk-averse label. Different error rates could then look like different preferences, even if underlying preferences were alike. A recorded choice is observed behavior; preference is inferred from a pattern under a model. In “Cognitive abilities affect decision errors but not risk preferences: A meta-analysis,” the authors describe a task with ten two-option choices. In seven, the riskier option has higher expected value; in the other three, the safer option does. Expected value means the probability-weighted average payoff. Imagine, purely as an illustration of the scoring mechanism, a respondent who selects randomly between the two options on each choice. Across ten choices, that person would average five selections of each type. Because the task offers seven opportunities in which the riskier option has higher expected value, the respondent’s five safer selections can be read as a tendency toward risk aversion, depending on the task’s scoring rule. This example explains how an imbalance can shape a label. It does not estimate how many actual participants respond randomly, or establish that any particular person did so. Three ideas need to stay separate. A decision error is a response treated as mistaken under a specified model or task rule. Inconsistency is a pattern that fails to follow the task’s expected structure, such as switching back and forth when one switch is expected. Preference is a relatively systematic inclination inferred from choices, after assumptions about how the respondent understood and evaluated the options. An irregular pattern may reflect error, but it might also reflect a strategy the model does not capture, difficulty understanding the instructions, inattention, or another feature of the task. The observed choice alone does not identify which explanation applies. The task can also distort in the other direction. The meta-analysis explains that an unbalanced set can be designed so random responding more often earns a risk-seeking classification instead. And a respondent using a simple strategy may consistently choose the option with the best possible outcome or the best worst-case outcome. Depending on the choices offered, such a strategy may look strongly risk seeking or risk averse, even though it is not a direct measure of a stable preference. Thus, random error and task architecture matter when interpreting ability–choice associations. This describes possible effects in risky-choice measures, not a rule that simpler strategies always imply one preference. A useful study should make this interpretation inspectable: explain how its choices are arranged, how it defines and detects inconsistent responses, and whether the scoring method models those responses separately from preference. Without that information, a reported group difference in risk aversion may partly reflect how the task turns answers into categories. The meta-analysis identifies this as a measurement problem in the risky-choice studies it reviewed; it does not show that every apparent preference is an error or that ability scores reveal how a person will reason in everyday decisions.

Sources: Cognitive abilities affect decision errors but not risk preferences: A meta-analysis | Psychonomic Bulletin & Review | Springer Nature Link; Cognitive abilities and risk-taking: Errors, not preferences

Why does the broader review retain a small gain-domain result?

The 2019 systematic review and meta-analysis found a weak association between cognitive ability and risk aversion in choices involving gains, but no observed association in mixed gain-and-loss choices or loss-only choices. The gain finding matters because it qualifies a broad claim that ability is unrelated to risky choice. The result is an average association across studies, not a rule for predicting any one person's decisions.

Here, the domains describe the possible outcomes in the choice task. A gain task presents alternatives with possible positive outcomes; a loss task presents possible negative outcomes; and a mixed task includes both gains and losses. The review included 97 studies in the gain domain, covering 90,723 participants; 41 studies in the mixed domain, covering 50,936; and 12 studies in the loss domain, covering 4,544. It reported a weak negative relationship in gains, with a pooled correlation of ρ = −.07, and no statistically significant relationship in the other two domains. The authors describe the pattern as domain-specific and weaker than some earlier claims suggested. The pooled estimates summarize relationships across studies; they do not isolate a causal effect of ability or show how a participant would respond in a new decision.

Why might the domain matter? A choice score is produced by a particular description of outcomes, probabilities, and alternatives. A person deciding between two possible gains faces a different frame from someone deciding between possible losses or a mixture of the two. Framing can change how the same person evaluates the alternatives, The 2018 review On the Relationship between Cognitive Ability and Risk Preference makes a related distinction across a wider set of settings: it reports that cognitive ability is often associated with avoiding harmful risks, while also being associated with less risk aversion in advantageous situations. These are not opposite readings of one universal trait. They concern choices with different consequences, and the review itself calls for richer ways to elicit and measure risk preferences.

The 2019 review tested potential moderators, but none consistently influenced the relationship across gains, mixed choices, and losses. That cautions against a neat explanation in which one task feature fully accounts for the domain pattern. It does not erase the pooled gain result; nor does it establish why that result appeared. A weak correlation can be statistically detectable across a large literature while still offering little guidance about a particular person's choices.

So when a study says cognitive ability predicts risk preference, check the outcome domain before carrying the result elsewhere. The 2019 evidence supports a narrow conclusion: a small average association appeared in gain-only tasks, while comparable evidence was not observed in the mixed and loss domains. It cannot establish that higher scores reduce general reasoning errors or that a person's risk preference stays unchanged across situations.

Sources: Cognitive ability and risk aversion: A systematic review and meta-analysis | Judgment and Decision Making | Cambridge Core; On the Relationship between Cognitive Ability and Risk Preference

Why do the reviews sound as if they disagree?

They ask overlapping but different questions. The 2019 review, “Cognitive ability and risk aversion: A systematic review and meta-analysis,” maps associations by outcome domain across risky-choice research. The 2022 review, “Cognitive abilities affect decision errors but not risk preferences: A meta-analysis,” narrows its task set and tests whether choice architecture and inconsistent responding can explain apparent links. Their conclusions are not a simple contest between a correct and incorrect review: they select and analyze evidence for different purposes.

The 2019 review used random-effects meta-analyses to summarize correlations separately for choices involving gains, losses, or both. It included 97 gain-domain studies with 90,723 participants, 41 mixed-domain studies with 50,936 participants, and 12 loss-domain studies with 4,544 participants. For gains, it found a weak negative association between cognitive ability and risk aversion (ρ = −.07); it found no significant association in mixed or loss domains. These are pooled associations, not estimates of how a particular person will choose, and the gain result is small. Still, a fair account cannot say that every review found no relationship.

The 2022 review included 30 studies and summarized 62 effect sizes. It asked whether errors and option arrangements could account for apparent links between ability and risk preference. Its main synthesis focused on description-based tasks, where options and probabilities are presented. Experience-based tasks instead provide feedback through which participants learn about outcomes and their likelihoods. Learning becomes part of the decision process, so observed choices can reflect acquired information as well as evaluation of stated gambles. The 2022 authors excluded experience-based tasks to make described choices easier to compare and code; this narrower scope produced fewer eligible studies.

The analyses differ too. The 2019 synthesis reports separate pooled results for gains, mixed outcomes, and losses. The 2022 authors modeled whether choice-set features were related to ability–preference correlations, including whether random responding would tend to be classified in one direction. Their Bayesian meta-analysis found no credible overall association in its primary set and evidence that task imbalance could shape apparent associations. That does not erase the earlier small gain-domain estimate: the studies and analytical questions differ.

The 2022 paper also applied its model to a merged dataset from the 2019 review and found a result consistent with its task-bias account. But this reuses evidence; it is not an independent replication. Nor should its combined estimate replace the 2019 domain-specific results: combining domains answers a different question from estimating gains, mixed outcomes, and losses separately.

Before calling review headlines contradictory, check which tasks were eligible, whether participants learned outcomes or received them in advance, and whether results were pooled across or separated by domain. Those differences help explain how both findings can stand: one review retains a weak association for gains in its broader analysis, while the other tests errors and architecture in a narrower set. Neither establishes that higher ability causes fewer everyday reasoning errors or that a score predicts an individual’s stable appetite for risk.

Sources: Cognitive abilities affect decision errors but not risk preferences: A meta-analysis | Psychonomic Bulletin & Review | Springer Nature Link; Cognitive ability and risk aversion: A systematic review and meta-analysis | Judgment and Decision Making | Cambridge Core

Which ability measure is being compared with the choice?

“Cognitive ability” names a family of capacities, not one interchangeable test score. Here it means performance on tasks that require processing, retrieving, evaluating, or manipulating information. A study’s result applies first to the measure it used. A correlation between one narrow task and choices in a lottery experiment does not automatically describe a general intelligence score or predict how a particular reader handles everyday decisions. Two score terms help keep that boundary clear. Raw performance is the answers a person gives under a particular test’s conditions: its items, instructions, timing, and scoring. A normed score interprets performance against a specified reference group tested under defined conditions. A raw count alone does not show where someone stands in a population. A normed score supports only comparisons its reference group and evidence justify. Neither is a complete account of reasoning or a stable personal trait. The 2022 meta-analysis, “Cognitive abilities affect decision errors but not risk preferences,” grouped measures into the Cognitive Reflection Test (CRT) and standardized ability tests, including Raven’s Progressive Matrices. Its authors note that these measures can correlate while differing in content and in the traits they assess. In their moderator analysis, studies using the CRT did not credibly differ from studies using other included tests in the ability–risk-aversion association. That finding does not establish equivalence. A moderator test asks whether the relationship varied reliably between categories in the available studies; no detected variation does not erase the measures’ distinct meanings. The CRT can be easy to overread because it is used in decision research. “Cognitive Reflection and Decision Making” introduces it as a three-item measure of one type of ability or disposition: reflecting on a question and resisting the first response that comes to mind. Its relationship to decision-theory choices concerns that specific reflective tendency. It does not make the CRT a general intelligence test or a direct measure of every reasoning operation. The 2019 systematic review, “Cognitive ability and risk aversion,” drew on a wider mix: CRT, Raven matrices, numeracy tests, working-memory measures, and cognitive-ability test batteries. Numeracy tasks assess numerical ability; working-memory tasks ask participants to retain information while doing another demanding task; a battery combines measures. A finding pooled across several tests is not automatically a finding about any one of them. Before carrying an association to a reader’s score, check which ability measure a study administered and what it asked participants to do. Do not assume a conversion between scores unless evidence supports it for the specific instruments and population.

Sources: Cognitive abilities affect decision errors but not risk preferences: A meta-analysis | Psychonomic Bulletin & Review | Springer Nature Link; Cognitive ability and risk aversion: A systematic review and meta-analysis | Judgment and Decision Making | Cambridge Core; Cognitive Reflection and Decision Making

What does a direct experiment add beyond pooled correlations?

A experiment can measure choice consistency alongside risky choices, testing whether an apparent preference association tracks decision noise. A 2021 study found no systematic link between cognitive measures and risky decisions in one incentivized sample, while higher ability was associated with fewer noisy choices. This supports an error-based explanation for some task results, but does not show that increasing ability causes fewer mistakes or that the pattern holds in everyday decisions.

In “Cognitive abilities and risk-taking: Errors, not preferences,” 556 first-year Business Economics students in Spain completed two risk-related tasks and three cognitive tasks, with real monetary incentives. The cognitive tasks assessed number additions, the Cognitive Reflection Test, and the Remote Associates Test. The risky tasks elicited choices in gain and mixed gain-loss settings. The design tested whether ability related to preference, irregular responding, or both.

The authors report no systematic role for cognitive measures in risky decision-making. Instead, higher ability, particularly reflection and convergent-thinking performance, was associated with fewer noisy or inconsistent choices. Inconsistent choices were also associated with more risk-averse and less loss-averse classifications. Errors could therefore make lower-ability participants appear more risk averse and less loss averse than their underlying preferences warrant. The abstract also reports more errors when the two payoffs had similar probabilities. These findings concern the tasks studied, not a universal feature of risk judgment.

The researchers used structural equation modeling and factor analysis to estimate how cognitive performance related to noisy or inconsistent choices while estimating risk-taking separately. Someone can prefer one option yet sometimes answer inconsistently, so observed choices need not be a pure readout of preference. But the model does not directly reveal a hidden preference or error rate. They are inferred from task definitions and assumptions, not directly observed mental processes.

The sample consisted of Spanish first-year business-economics students, and outcomes came from two incentivized lottery-style tasks. The findings cannot establish behavior when probabilities are unclear or consequences are learned over time. The authors caution that real-world probabilities or payoffs may be ambiguous. The experiment did not manipulate cognitive ability, so it cannot show that raising a test score would reduce errors.

The experiment and the 2022 meta-analysis, “Cognitive abilities affect decision errors but not risk preferences,” support a useful rule: inspect how a study identifies errors before interpreting choices as preference. The experiment examines consistency in one student sample; the meta-analysis tests a wider set of description-based risky-choice studies. Neither establishes a person’s general reasoning-error rate. A score or single risky choice cannot settle an individual’s risk appetite. Check which measure and task produced the result, and whether inconsistent choices were modeled; keep the inference within that task’s boundaries.

Sources: Cognitive abilities and risk-taking: Errors, not preferences; Cognitive abilities affect decision errors but not risk preferences: A meta-analysis | Psychonomic Bulletin & Review | Springer Nature Link

What does risk preference mean outside one lottery task?

A lottery choice measures what someone does in one described situation; it does not reveal a complete, stable appetite for risk. The 2018 review On the Relationship between Cognitive Ability and Risk Preference surveys evidence from incentivized lottery choices, behavior outside experiments, and self-reports. It reports a recurring contrast: cognitive ability tends to correlate with avoiding harmful risky situations, yet with less risk aversion when outcomes may be advantageous. Those are different outcomes, so calling both “risk tolerance” hides the distinction.

The stakes, framing, information, and outcome domain matter. A gain task concerns possible benefits; a loss task concerns possible harms; a mixed task includes both. Choosing among stated probabilities also differs from deciding after learning through feedback. In a described lottery, probabilities are supplied; in an experience-based setting, a person encounters outcomes over time and may need to learn what is likely. A choice that appears cautious when losses are possible cannot automatically be read as the same preference governing choices among gains.

The review cautions that observed choices and ability-test results are measures used to represent underlying traits. Its abstract says that interpreting their correlation as a relationship between latent preference and latent ability requires substantial assumptions; causal conclusions require more. The bounded conclusion is that risk taking varies with the decision measured, and ability may relate differently to harmful and advantageous situations. One score or lottery choice cannot label someone's enduring style or guide financial, educational, employment, or clinical decisions.

Sources: On the Relationship between Cognitive Ability and Risk Preference; On the Relationship between Cognitive Ability and Risk Preference | PubMed

What can a browser test result responsibly tell you?

A browser result describes how you performed on that test under its particular conditions. It cannot establish your general rate of reasoning errors or tell you how much risk you prefer. Studies discussed above compare ability measures with choices in specified economic tasks; they do not validate this site's conversion or measure your risk preference. The official “IQ Test Methodology” page says the assessment draws 50 questions from a larger pool across pattern reasoning, number patterns, and spatial reasoning, and adapts difficulty to answers. Its conversion uses placeholder parameters and has no validated adult reference sample. The estimated range is therefore not an established population comparison or clinical measurement. A conventional IQ scale does not supply missing validation. The displayed band is not a confidence interval or validated margin of measurement error. Use the result for low-stakes curiosity, noticing which formats felt familiar, and reviewing particular answers. Do not use it to label yourself as careful or careless, predict a risky choice, or make a consequential decision. It is not interchangeable with the different ability measures in the research. The “Free IQ Assessment” page says the range is free and the optional report adds interpretation, worked examples, and a practice plan; it explicitly says the report does not make the estimate more accurate. Review worked answers to identify an operation to practice, not as evidence that general intelligence or real-world judgment has changed. For an important assessment decision, seek a qualified professional.

Sources: IQ Test Methodology; Free IQ Assessment

What should you do next with a score or study?

Treat a score as performance on its stated measure, then inspect the task, domain, and scoring before drawing a conclusion. Ask: What ability measure was used? Did the choice involve gains, losses, or both? How were inconsistent responses handled? Was the score compared with a reference group relevant to the person assessed? These checks matter because the 2022 meta-analysis, “Cognitive abilities affect decision errors but not risk preferences,” concerns specific risky-choice tasks; it does not establish an individual's error rate or stable risk style. For low-stakes practice, use errors to identify an operation to work on, such as tracking a sequence or comparing quantities. This guides practice only; it does not establish transfer. Test IQ Free offers a free 50-question test across pattern, number, and spatial reasoning, with worked answers. Its estimated IQ range is provisional, produced by an unvalidated conversion with no validated adult reference population. It is not a normed comparison, confidence interval, or prediction of risk choices. Review the explanations, and keep the result low stakes. The optional paid report adds interpretation and practice structure, not validity. If an education, work, health, or financial decision depends on assessment, seek qualified advice suited to that decision. The evidence supports a conclusion: task accuracy and preference are separable, and findings can shift with task design and outcome domain.

Sources: Free IQ Assessment; IQ Test Methodology; Cognitive abilities affect decision errors but not risk preferences: A meta-analysis | Psychonomic Bulletin & Review | Springer Nature Link

Questions readers ask