Yes. Repeating a cognitive ability test often raises observed scores, but the size varies with the test, form, interval, and examinee. A higher second score describes performance on that sitting; by itself, it does not explain why performance changed or establish a lasting rise in general intelligence. Treat a retake as practice or a performance comparison, and use qualified assessment when the result will affect a consequential decision.
Do repeat administrations usually raise cognitive-test scores?
Yes. Repeating a cognitive ability test often raises observed scores, but the size varies with the test, form, interval, and examinee. A higher second score describes performance on that sitting; by itself, it does not explain why performance changed or establish a lasting rise in general intelligence. A practice effect is a score advantage associated with prior exposure to the same or a similar test. The 2018 meta-analysis “Retest effects in cognitive ability tests: A meta-analysis” combined 174 samples from 122 studies, covering 153,185 participants and 786 test outcomes. The authors found significant retest effects and no further pooled score gains after the third administration. This is not a prediction for one person: cognitive operation and content, form equivalence, retest interval, and participant age moderated effect size. Repeating a reasoning test can change performance, but the pooled pattern cannot supply a universal adjustment to an individual result. The review concerns cognitive ability tests, not emotional intelligence or general cognitive-training programs. Prior exposure can affect performance; a score difference alone leaves its cause open. These findings describe a tendency, not a guaranteed gain.
Sources: Retest effects in cognitive ability tests: A meta-analysis
What might have changed between two sittings?
A repeat score can reflect remembered items, familiarity with instructions or pacing, a more efficient strategy, learning on similar tasks, changed conditions, and ordinary measurement variation. A score alone cannot separate these explanations. The useful description is “performance changed after prior exposure,” unless evidence about the specific test supports more. Calling every increase a memory effect is too narrow; calling it proof of broader ability growth goes beyond what a before-and-after score establishes. Item memory is the most literal route. If a person sees the same questions again, remembering an answer or a previous mistake may help on those items. This can raise observed performance without showing that the person can solve unfamiliar problems of the same kind. A different form may remove direct recall, but not what someone learned about instructions, pacing, or which details to attend to. Format familiarity can matter even when visible content changes. Test-specific skill is another possibility. Someone may learn a procedure that applies across new items on the same test or closely similar tasks: checking one feature at a time, comparing alternatives systematically, or avoiding an unproductive first guess. That learning is real and can be useful. Its scope still needs accurate description. Better use of a task’s demands does not, by itself, demonstrate a lasting change in general intelligence. The distinction is between improved performance on a practiced operation and evidence that the change carries to substantially different tasks. Consider an example, purely illustrative: a student taking a visual reasoning assessment initially scans a crowded display without a plan. After reviewing explanations, the student begins checking each row for one visible relation before considering another. On a later set of unfamiliar items in a similar format, that orderly scan could help. No particular test item or answer is implied. If the student performs better, the change might reflect a learned approach, format familiarity, ordinary variation, or some combination. A score cannot assign shares to those causes; this example is not research evidence. A retest contrasts observations taken at different times, after exposure and under different conditions. Unless the design includes an appropriate comparison, the difference bundles these changes together. The person may have slept differently, felt more rushed, understood instructions better, practiced related operations, or encountered normal measurement variation. That does not make a score arbitrary; it means one change does not identify its cause. A controlled study can estimate effects across conditions, but a group result cannot reveal exactly why one individual’s score moved. “Retest effects in cognitive ability tests: A meta-analysis” reports that test operation and content, form equivalence, interval, and age moderated retest effects, while noting that the literature did not directly settle mechanisms. Prior exposure is associated with changed scores, but the pooled pattern cannot tell a reader whether memory, strategy, familiarity, or another factor explains a particular result. It cannot identify mechanisms for every instrument or diagnose an individual. “Retesting in Employment Selection: A Review and Suggestions for Future Research” provides adjacent evidence from personnel testing: scores tend to rise, variability increases, and retesting may change which constructs scores reflect. The review says causes remain insufficiently delineated. Because employment assessments differ from IQ tests, this does not establish a mechanism for every test. It cautions against assuming remembered answers alone explain a retest difference. For an individual, record the test and form, whether answers or explanations were reviewed, similar practice, time passed, and changed conditions. Describe the result as higher or lower performance on the later sitting. If the same items returned, note that exposure; if the form changed, format familiarity may remain. Avoid naming a cause as fact without suitable evidence. This preserves uncertainty without turning a test event into a claim about identity or fixed potential.
Sources: Retesting in Employment Selection: A Review and Suggestions for Future Research; Retest effects in cognitive ability tests: A meta-analysis
Why is one retest gain not a universal correction?
One retest gain cannot serve as a universal correction because a score change depends on what was tested, who took it, how the second administration differed, and how long passed between attempts. Research can show that retest effects occur across many cognitive ability tests; it cannot turn an average pattern into a guaranteed gain, deduction, or adjustment for an individual. The 1979 study “Test-retest gains in WAIS scores after four retest intervals” offers a clear illustration of why that distinction matters. It administered the then-current Wechsler Adult Intelligence Scale (WAIS) twice to 76 male college students. Different participants were retested after one week, one month, two months, or four months. The publisher abstract reports increases in Verbal, Performance, and Full Scale IQ scores, but the sizes were not uniform across score types or intervals. Verbal scores rose by different amounts at the four intervals, and the four-month verbal increase was the sole reported gain that was not statistically significant. Those results answer a narrow question: in this particular sample, using this particular WAIS edition and study arrangement, retest scores tended to be higher, with variation by score and interval. Nor do they show that waiting four months removes a retest effect. Four intervals were examined; the least pronounced verbal change at one interval does not establish a safe waiting period for a different score, population, or test. The study is historical evidence about one context, not guidance for present-day instruments. A much broader result comes from “Retest effects in cognitive ability tests: A meta-analysis.” Its authors combined 174 samples from 122 studies, covering 153,185 participants. Their longitudinal multilevel analysis found significant retest effects and reported no further pooled score gains after the third administration. They also found that test operation and content, form equivalence, retest interval, and participant age influenced effect size. It does not make the pooled result a forecast for one retake: the evidence combines different tests, samples, forms, and timing arrangements, and the authors note that causes and consequences are not fully understood. The two sources therefore work at different scales. The WAIS study makes variation visible within one bounded example: outcomes differ by score and interval, even when the instrument and sample are held relatively specific. The meta-analysis shows that retest effects recur across a much wider body of cognitive testing, while also identifying moderators that prevent a single estimate from fitting every case. Taken together, they justify a general expectation of possible score change, not a personal correction factor. This distinction matters whenever someone encounters a number from a study and is tempted to subtract it from a later result, or add it to an earlier one. A group average describes the pattern reported for the studied group under its conditions. It is not an adjustment validated for another test-taker. Even when a score type shares a label, differences in edition, content, form, administration, examinee group, and interval can change what comparison is justified. Applying a historical mean as if it were a universal allowance would make the score look more precise while removing the conditions that gave the original result meaning. Repeating a cognitive test can raise observed scores. But neither a broad pooled effect nor a result from a small, specific study identifies the cause or size of change for an individual. Before interpreting a repeat score, ask whether the instrument, score, sample, form, and interval resemble the evidence being cited. If they do not, report the observed change plainly and avoid borrowing a study’s gain as a universal retest penalty or bonus. A retest finding is a limit on interpretation, not a mathematical repair. It warns that prior exposure may matter; it does not tell us how much it mattered in a different case. Without a verified match between the study and test, describe both administrations and leave the difference unadjusted. The evidence earns caution, not a numerical rule for every repeat.
Sources: Test-retest gains in WAIS scores after four retest intervals; Retest effects in cognitive ability tests: A meta-analysis
Do alternate forms or adaptive tests solve the problem?
Alternate forms and adaptive delivery can reduce direct repetition, but neither automatically removes familiarity or makes two scores comparable. A different set of questions may reduce memory for exact items; the person may still know the instructions, pacing, response format, or reasoning operations. A same-form retest presents the same items again, so remembering earlier questions may influence performance. An alternate form uses different items intended to assess the same abilities. “Alternate” describes the design, not demonstrated equivalence. Treating scores as interchangeable requires evidence about how forms relate, how consistently they measure the intended construct, and how retesting affects performance. Exposure can teach more than answers. A test taker might become quicker at deciding which features of a visual display matter, or learn the instructions and response options. Such familiarity may transfer to new items. Thus, different items may reduce exact recall while leaving some practice advantage. An alternate form can also differ in difficulty if its items are not comparable. A score difference then combines exposure, item content, administration, and the examinee’s condition; it cannot identify their separate contributions by itself. The experiment “Reducing the effect size of the retest effect: Examining different approaches” tested these design choices. Researchers randomly assigned 960 test takers to five conditions. Participants completed four cognitive ability tests twice, about one month apart; conditions varied retest form and delivery, including fixed-item and computerized adaptive tests. Retest effects differed across conditions. The authors report results consistent with learning test-specific skills during testing and found that adaptive testing at the initial sitting reduced later score gains in their study. That result concerns those participants, tests, conditions, and interval. It does not show that every adaptive test removes practice effects or makes repeat scores comparable. Adaptive delivery was a feature tested in an experiment, not a general guarantee of validity. The finding does not mean all improvement is irrelevant: it supports the narrower conclusion that test-taking can teach skills affecting later performance, and design choices may change the size of that gain. A computerized adaptive test adjusts item selection during administration according to its design. That does not tell a reader whether a retake uses wholly new items or whether administrations are calibrated for comparison. Interpreting a repeat therefore requires information about item pools, scoring, form construction, and retest evidence, not merely the label “adaptive.” Ask whether the second sitting draws from a documented parallel form and whether the score scale is linked across administrations. The U.S. Office of Personnel Management provides a bounded example from federal employment assessment. It says agencies should document the basis for retesting policies, including alternate-form availability and effects on validity or test integrity. It notes that repeated exposure to cognitive test material can create security concerns and says that, where feasible, applicants should not receive the same form twice in a row when exposure threatens usefulness. OPM also lists timing and score replacement as policy issues. This is hiring guidance, not a universal IQ-test waiting period or individual clinical advice. “Retest effects in cognitive ability tests: A meta-analysis” likewise found form equivalence and retest interval moderated effects across studies. It synthesized 174 samples from 122 studies, totaling 153,185 participants. The tests and contexts varied, so its pooled result is not a personal correction. Together, these studies show forms and delivery can matter. For a formal comparison, ask whether the exact test documents alternate-form comparability and retest behavior for the relevant population and interval. Check how exposure is handled and which score is reported. For practice, describe the outcome as performance after exposure. A retest method can shape observed gains, but only evidence matching the instrument and intended interpretation can show what that means. Without that information, a numerical difference remains descriptive, not explanatory.
Sources: Reducing the effect size of the retest effect: Examining different approaches; Can agencies limit the number of times an individual is allowed to take a test or assessment during a given timeframe? What guidance is available if our agency wants to develop its own retesting?; Retest effects in cognitive ability tests: A meta-analysis
What does a score difference establish?
A score difference establishes one narrow fact: the recorded result changed between two administrations. It does not, on its own, establish why it changed, whether the change exceeds the test’s expected measurement variation, or whether a lasting change in general intelligence occurred. A group-average retest gain and one person’s difference are different claims and require different evidence. A raw score is the number of points earned on one administration before comparison with a reference group. Answering more items correctly describes performance under those conditions. It does not locate the person relative to others or convert by simple arithmetic into a standard score or percentile. Such conversions depend on the test’s scoring rules and appropriate reference data for the relevant version and population. Without those, a raw-score increase remains a raw-score increase. Do not infer a percentile from item counts; percentile ranks express standing within a group and depend on relevant standardization evidence for that exact test, version, and intended population comparison group. A norm-referenced score compares performance with a defined standardization group. Raw and norm-referenced changes are not interchangeable; interpretation depends on the exact test, edition, administration and comparison population. Two item counts alone cannot support a claim of gained IQ points, and a conversion should not be borrowed from another test or invented. Reliability is the consistency of scores across occasions, editions or raters, as explained in the Educational Testing Service memorandum “Test Reliability—Basic Concepts.” It describes scores under specified conditions; it does not promise identical results. Occasions can differ, and scores can vary. Measurement variation makes an observed result an imperfect indication of intended performance or attribute. A small difference may combine changed performance and variation; without measure-specific evidence, these cannot be separated. This does not mean a difference is meaningless. More correct answers are a genuine observation. Interpretation should stop at the level supported by evidence. One can say, “My recorded score was higher on the second sitting,” and note differences in form, timing, setting or exposure. That does not mean “my general intelligence increased.” Such a conclusion requires evidence linking the score change to the broader construct rather than familiarity, strategy, conditions or measurement variation. Reliability concerns consistency; validity concerns whether evidence supports a particular score interpretation and use. Evidence for one interpretation, population or decision does not automatically support another. Reliability evidence in another context does not validate a retest as a measure of change. Ask whether evidence supports interpreting these administrations as measuring change in the intended way. Group findings need the same care. A study may find that people, on average, score higher when tested again. It describes those participants, tests, forms and conditions, not a correction factor for an individual. An average cannot identify the cause or size of one person’s change. Because research combines tests and intervals, it cannot supply a personal adjustment without test-specific evidence. The American Psychological Association’s “Standards for Educational and Psychological Testing” page is relevant to testing practice, but the opened page exposed no substantive standards text and cannot support a detailed rule here. Neither it nor the ETS guide gives a threshold for deciding that an IQ-score change is real. The instrument’s own precision evidence is needed before describing a difference more strongly. For a comparison, record test, version, form, interval, scores, changed conditions and intervening practice. State plainly: “The reported result differed across these administrations.” If it is a raw count, call it that. If norm-referenced, name the scale and comparison only as the instrument’s documentation permits. If those details are unavailable, keep the description narrower, not more confident. This reports the comparison without overclaiming.
Sources: Test Reliability—Basic Concepts; Standards for Educational and Psychological Testing
Does a practice effect mean the person learned nothing—or became more generally able?
Neither conclusion follows. A person can learn a useful operation and improve on similar tasks, while the gain still falls short of evidence that unfamiliar tasks or general intelligence improved. Repeated-task performance and broad transfer are different outcomes. A practice effect is a score advantage associated with prior exposure to the same or a similar test. It describes performance, not all the learner can do across settings. Separate three distances between practice and outcome. With the same item, a person may remember its answer or route through it. Near transfer occurs when someone uses a learned operation on a new item that shares important features with the practice task. Far transfer means an effect on a substantially different task or broader ability, where overlap is weaker. These categories depend on how much the source and target tasks have in common; a score increase alone cannot assign them. Consider an example, not a reported study: a learner who once scanned a visual arrangement feature by feature may later compare rows and columns systematically. That strategy could help on a fresh, similar problem, even if no answer was remembered. If performance improves only on problems built around the same comparison, the evidence supports a narrow description: better performance on similar reasoning tasks after practice. It does not show improvement on unrelated verbal or memory tasks. This distinction gives credit for learning while keeping the conclusion at the level measured. Cognitive-training research helps explain the boundary, though it is not retest research. “Near and Far Transfer in Cognitive Training: A Second-Order Meta-Analysis” combined meta-analyses of working-memory training and other programs, including video games, music, chess, and exercise games. Working-memory models found near-transfer effects on memory; far-transfer models examined reasoning, speed, language, and cognitive control. The authors reported far-transfer effects as small or null; after adjustments for placebo effects and publication bias, the overall far-transfer effect and true variance were estimated at zero. It covered 21,968 participants, 1,555 effect sizes, and 332 clusters. This informs transfer, not retest correction or individual outcomes. A repeat score can reflect more than memory: familiarity with instructions, pacing, formats, or strategy can alter a later approach. “Retest Effects in Cognitive Ability Tests: A Meta-Analysis” found higher cognitive-test performance after prior administration, with test operation and content, form equivalence, interval, and age moderating the pattern. The authors noted substantial variability and limited direct evidence about mechanisms. This leaves room for task-specific learning but does not identify the cause of one person's change. Neither “all memory” nor “my general ability grew” follows from a before-and-after difference alone. A fair account avoids two errors. Calling every gain meaningless dismisses learning that could matter on similar problems. Calling every gain proof of a general increase extends beyond what was assessed. A broader claim needs improvement on new, meaningfully different measures, a suitable comparison group, clear administration, and follow-up. The outcome must match the claim: change on another reasoning task is not automatically evidence of change in all cognitive abilities or of a lasting shift in general intelligence. For reasoning practice, value the operation learned and name it precisely. You may say repeated exposure helped you recognize a format or use a strategy more effectively, if that is what the experience supports. You may report higher performance on a later administration if scores were recorded. Without evidence tied to the exact test and comparison, do not claim the increase measures a durable change in general intelligence. This wording separates an observable achievement from a wider interpretation that needs stronger evidence. The conclusion could change with longitudinal or randomized research that compares retaking with an appropriate control, uses fresh tasks at different distances from practiced material, keeps administration conditions clear, and measures outcomes after a meaningful interval. Until then, a practice effect can signal exposure-linked improvement and genuine task learning while leaving far transfer unresolved. Practice can improve practiced operations; generalization requires separate evidence.
Sources: Near and Far Transfer in Cognitive Training: A Second-Order Meta-Analysis; Retest effects in cognitive ability tests: A meta-analysis
How should you compare a repeat result with professional or online results?
Compare like with like only when evidence for the exact instrument supports the comparison. Two unrelated online quizzes do not create a clean before-and-after measure: their questions, scoring rules, difficulty, timing, and intended interpretations may differ. Describe a browser result as performance under that sitting’s conditions. Even when both screens display an IQ label, labels do not establish a shared scale. A useful comparison begins by asking what stayed constant and what changed. On the same instrument and same form, a score difference is a retest difference after direct exposure; it may reflect familiarity with items and other changes. On the same instrument with an alternate form, direct item repetition may be reduced, but the forms must be shown to support comparison. A different instrument changes more than the questions, so a difference cannot be read as growth or decline without evidence connecting its scales and uses. These situations are distinct, and each requires different evidence before a score difference can carry a stronger interpretation. A changed number alone cannot tell which account applies or whether the comparison is useful. | Comparison | What can cautiously be said | What remains unknown | | --- | --- | --- | | Same instrument, same form | Performance changed on a repeated form | How much reflects exposure, conditions, or measurement variation | | Same instrument, alternate form | Performance differed across forms | Whether forms are sufficiently comparable and how exposure carried over | | Different instruments | Results differ under two assessment systems | Whether scores measure the same thing on comparable scales | | Test IQ Free retakes | Answers differed across attempts on this browser test | A normed IQ change, population standing, or lasting ability change | The distinctions matter because reliability is not a synonym for an unchanged score. In Test Reliability—Basic Concepts, ETS defines reliability as the extent to which scores are consistent across occasions, editions, or raters, and discusses measurement error and alternate-form reliability. The guide does not provide an error band or retest correction for every test. The review Retaking Employment Tests: What We Know and What We Still Need to Know reports that retest scores in employment testing tend to be higher and more varied, and says evidence suggests retesting can change the constructs reflected by scores. The review also says that the factors underlying these differences have not been clearly delineated. In personnel selection, it cannot explain an individual change or predict an online quiz result. For Test IQ Free specifically, the assigned product is an adaptive 50-question browser test covering pattern, number, and spatial reasoning. Its displayed estimated IQ range comes from placeholder conversion parameters and has no validated adult reference population. The band is not a confidence interval, a normed score, or an established comparison with a population. A retake can describe changed performance, but not a standardized IQ change. Nor does a different estimate from another website corroborate it: without evidence that both conversions use comparable measures and reference groups, the numbers are not a common ruler. The optional paid report uses the same provisional conversion for its point estimate and percentile. It adds explanation and practice structure, not an independent measurement or validation. Seeing a point estimate beside a percentile may make the result look more exact, but extra display detail does not supply the missing reference evidence. Interpret professional tests using evidence for the specific edition, administration, scoring, and use. Their protected items and norm tables are unnecessary here. State the test and form, record prior exposure and conditions, and describe results only as far as evidence supports. If a decision has serious consequences, a browser score or cross-site comparison is not a substitute for qualified assessment.
Sources: Test Reliability—Basic Concepts; Retesting in Employment Selection: A Review and Suggestions for Future Research
What should you do before taking the test again?
Decide what a repeat is for before you take it. If you want low-stakes reasoning practice, you can retake and treat the result as performance after prior exposure. If you need a score to support a decision about education, clinical care, or eligibility, do not use an informal repeat as a substitute for assessment. Tell a qualified assessor what you have already taken and what practice followed; the instrument and its retest evidence matter to interpretation. For any comparison, keep a record: the test name and version or form if known; when each sitting occurred; whether you saw answers, explanations, or similar practice between them; and whether the conditions changed. This record does not calculate a correction. It makes the circumstances visible, so a higher or lower result is not mistaken for a clean measure of change. The meta-analysis “Retest Effects in Cognitive Ability Tests” found that retest effects varied with form equivalence, interval, age, and test content across its combined studies. For practice, use worked answers to understand an operation, then try fresh questions that require it. For example, after reviewing how a visual pattern changes across rows, describe the rule in words and apply it to new material. Improvement on similar tasks is a useful learning outcome; it does not, by itself, establish a lasting change in general intelligence. Evidence does not support one universal retest interval for every instrument and purpose. The U.S. Office of Personnel Management’s retesting guidance is written for agencies making employment-assessment policies, not for personal IQ testing. It says those policies should document their basis, account for cognitive-test exposure and assessment integrity, and consider waiting periods and alternate forms. Where feasible, it advises against giving applicants the same form twice in a row. Test IQ Free’s free 50-question adaptive reasoning test can be used for open practice. Its displayed estimated IQ range comes from a provisional conversion with placeholder parameters and no validated adult reference population; it is not a confidence interval or established population comparison. Record a repeat as another performance under those conditions. The optional $9 report adds explanation and practice structure using the same unvalidated conversion, not accuracy or validity.
Sources: Can agencies limit the number of times an individual is allowed to take a test or assessment during a given timeframe? What guidance is available if our agency wants to develop its own retesting?; Retest effects in cognitive ability tests: A meta-analysis
Sources
- Retest effects in cognitive ability tests: A meta-analysis
Supports the pooled retest finding and the role of test content, form equivalence, interval, and age as moderators.
- Retesting in Employment Selection: A Review and Suggestions for Future Research
Supports the bounded personnel-testing discussion of higher retest scores, increased variability, and unresolved causes.
- Test-retest gains in WAIS scores after four retest intervals
Supports a bounded historical example of score changes across intervals in 76 male college students.
- Reducing the effect size of the retest effect: Examining different approaches
Supports the comparison of alternate forms and adaptive versus fixed-item delivery in a retest experiment.
- Can agencies limit the number of times an individual is allowed to take a test or assessment during a given timeframe? What guidance is available if our agency wants to develop its own retesting?
Supports the limited employment-policy discussion of exposure, alternate forms, waiting periods, and score policies.
- Test Reliability—Basic Concepts
Supports general definitions of reliability, measurement error, and alternate-form reliability; it supplies no test-specific retest threshold.
- Standards for Educational and Psychological Testing
Identifies a relevant testing-standards resource; the opened page did not expose substantive standards text for a detailed rule.
- Near and Far Transfer in Cognitive Training: A Second-Order Meta-Analysis
Supports the qualified distinction between near-task training gains and broader transfer; training is adjacent to, not identical with, retesting.
- Test IQ Free 50-question reasoning test
Supports the product description and provisional-estimate limits specified in the assigned publication brief.
Try the difference yourself.
Work through 50 adaptive reasoning questions, then see your estimated IQ band. Unlock a detailed personalized report after purchase.
Take the free test