Should You Retake a Cognitive Test on Another Device?

Should you switch devices before paying for another assessment?

Retake on another device only when the first device caused a specific, observable problem that could have affected your answers: for example, part of a diagram was clipped, a control failed, or the timer stopped. Correct that problem and treat a retake as a limited delivery check. A different score does not show that one device measured you more accurately. Familiarity, practice, timing, form, interruptions, and ordinary score variation can all change a second attempt.

That distinction matters before spending money. A new device may make a browser page usable; it cannot supply missing validation, a comparison group, or a professional interpretation. If a result must support an educational, clinical, or other consequential decision, ask what question the assessment is meant to answer and whether a qualified provider can select and interpret an appropriate standardized procedure. If your purpose is curiosity, another browser attempt is still just practice and self-exploration.

The practical rule is three-part: troubleshoot a documented device fault once; do not use a preferred second score as proof; pay for an assessment only when a real decision calls for evidence the first test cannot provide. Test IQ Free’s 50-question reasoning test gives a provisional estimated IQ range from an unvalidated conversion with no validated adult reference population. Its displayed band is not a confidence interval or established comparison with adults. Repeating it on a second screen cannot change that. The optional paid report adds explanation to the same provisional conversion, not accuracy or validity.

A useful test is to state the counterfactual: if the original device had worked as intended, what observable part of the attempt would have been different? ‘The image was cut off’ has a clear answer: the full image would have been visible. ‘I felt the score was too low’ does not identify a delivery condition. This small distinction protects the decision from score-shopping, where the device is changed only after seeing the number and any later movement is then credited to the new screen.

Sources: Standards for Educational and Psychological Testing; Equivalence of Computerized and Paper-and-Pencil Cognitive Ability Tests: A Meta-Analysis

What result are you trying to make comparable?

Before comparing two attempts, identify what each result actually represents. A raw score is the count or pattern of responses scored correct under that test’s rules. A norm-referenced score interprets performance by comparing it with a defined reference group tested under specified conditions. An estimated IQ label is a further interpretation that depends on a conversion supported for the test, population, and intended use. These are different layers, even when a page presents them together as one number.

Suppose the first screen reports 32 correct answers and a second attempt reports 35. That is a difference in observed performance on those administrations. It does not itself reveal why the count changed, whether the questions were equally difficult, whether a new form was used, or how either count maps to a population. If the page turns each count into an IQ estimate, the conversion requires evidence beyond the count. The user cannot infer that evidence from a polished interface or from the fact that the same label appears twice.

Measurement error is the uncertainty that remains in an observed score because a test samples only some tasks and responses under particular conditions. It does not mean the device is broken or that the test is useless; it means the score should not be treated as a perfectly exact reading. The Standards for Educational and Psychological Testing describe interpretation as dependent on evidence for intended score uses, including score precision, test content, response processes, and relationships with other variables. A retake on another device provides none of those missing pieces by itself.

Ask four questions before treating scores as comparable: Was it the same instrument and version? Were the items or forms the same or demonstrably equivalent? Was administration timed and delivered in the same way? Is there evidence for interpreting those scores against the same reference group? If the answer is unknown, record that uncertainty instead of converting the numbers into a story about ability. A device change is one difference among several, not a controlled experiment unless the other relevant conditions are held steady and the exact format comparison has supporting evidence.

For Test IQ Free specifically, the free result’s estimated range is provisional because its conversion has not been validated against an adult reference population. The band is neither a confidence interval nor an established statement about where a reader stands among adults. Its domain feedback can describe performance on pattern, number, and spatial reasoning items in that session. It cannot turn raw performance into a normed score by being repeated, nor can a paid explanation make the underlying comparison more defensible.

The score label also does not settle the question. A number called an IQ estimate can be generated from many scoring systems, and the label alone does not disclose whether a test has a representative adult sample, what its uncertainty is, or which uses were studied. Ask for the source of the comparison and whether the reported score comes from a standard score conversion for that instrument. If those details are absent, the output may still be a convenient summary of task performance, but its population meaning remains unsupported.

Sources: Standards for Educational and Psychological Testing

Which device problems are worth correcting before interpreting a result?

A device problem is worth correcting when you can name what happened and connect it to the task. Examples include a visual item extending beyond the display, a tap target that would not respond, text too small to read comfortably, a browser interruption that lost an answer, or a timer visibly freezing. If the page or item could not be seen or operated as intended, the first administration may not represent an ordinary attempt. Write down the concrete fault while it is still fresh; do not retrofit a vague feeling that the device was ‘off’ after seeing an unwanted score.

The Standards for Educational and Psychological Testing discuss legibility, familiarity with an interface, quiet conditions, and support for web-based administration as factors developers should consider. That is guidance for designing and evaluating tests, not a rule that every low score warrants a retest. It supports a practical distinction: remove an obstacle that prevented access to the intended task, but do not assume that any difference in comfort changes the interpretation. A larger display may help you see a visual stimulus; it does not establish that a different score is truer.

Consider an example: a reasoning item appeared partly beneath a fixed toolbar on a phone, and the response buttons were crowded against the screen edge. A second attempt on a laptop after confirming the full item and controls are visible could answer a narrow question: did the display defect disappear? It cannot isolate the defect’s contribution if the second attempt also contains different items or follows exposure to the first. The honest record is ‘the first layout clipped content; the later layout did not,’ not ‘the laptop corrected my IQ.’

Other conditions deserve separate labels. Low sleep, distraction, stress, rushing, or unfamiliarity may affect how a person approaches a task, but those are not device faults. The recent multi-session study by Robison and colleagues discusses possible influences such as fatigue, motivation, task-specific strategy, and measurement error, while examining computerized measures in young adults. It did not test whether a phone is worse than a laptop. Naming a condition helps describe the attempt; it does not diagnose the cause of a score or justify selecting whichever result feels most plausible.

Use one correction for one observed failure. If the page crashed, fix the browser or connection; if content was obscured, choose a screen where it is fully visible; if an interruption broke the timed session, do not pretend it was uninterrupted. Then stop treating repeated device shopping as a route to certainty. Multiple attempts can add exposure and make the later result harder, rather than easier, to interpret. If you cannot identify a specific fault, switching devices is unlikely to answer a useful measurement question.

A good note is factual and brief: ‘At question 18 the right side of the image was outside the screen; I could not scroll without losing the timer.’ Avoid estimates such as ‘the defect cost me five answers’ unless the scoring record can establish that. After switching screens, check whether the content and controls are available before beginning. If the same fault remains, stop and contact the test provider or abandon the attempt; producing another score under the same defective setup adds no useful comparison.

Sources: Standards for Educational and Psychological Testing; Cognitive Test Scores Vary with Choice of Personal Digital Device

Does changing from paper to screen usually preserve performance?

Research comparing paper-and-pencil and computer administration shows why the answer must be specific to the task. Mead and Drasgow’s 1993 meta-analysis examined 159 correlations from timed cognitive ability tests given to adult and young-adult populations: 123 correlations came from timed-power tests and 36 from speeded tests. After correction for measurement error, the overall cross-mode correlation was .91. The estimated correlation was .97 for timed-power tests and .72 for speeded tests, indicating that speededness moderated correspondence in the studies reviewed.

A correlation describes how strongly two sets of scores move together across people; it does not mean each person gets the same score in both modes. Even high correspondence at group level leaves room for individual discrepancies. Nor does a correlation of .91 mean 91 percent of individuals receive identical results. The distinction is crucial when someone compares one personal pair of scores: a group relationship cannot identify the cause of that one person’s difference.

The review also concerns paper versus computerized administration in the tests included in older studies. It does not establish that every contemporary touch interface is equivalent to a laptop, or that a specific browser quiz has the same timing, stimuli, scoring, and administration on phones and desktops. Newer devices introduce other features: display dimensions, touch input, scrolling, refresh timing, and response latency. A test may be mostly untimed visual reasoning, or it may award speed and depend on fast response registration. Those task demands change what device features could matter.

A 2021 paper, ‘Cognitive test scores vary with choice of personal digital device,’ studied whether device differences could affect observed performance on cognitive tests sensitive to reaction time. Its reasoning is that device latency and input method can influence measured response times; the paper reports a moderate-to-large input-type effect for simple reaction time and a relatively small effect for digit-symbol matching. The researchers discuss standardizing or restricting devices when reaction-time precision is important. This is evidence that device features can matter for some outcomes, not evidence that switching screens generally raises or lowers an IQ score.

The useful takeaway is not ‘screens are interchangeable’ or ‘phones distort tests.’ It is that equivalence belongs to a defined comparison: a named test, specific modes, the population and task studied, and the outcome measured. Paper/computer findings can inform a question about digital delivery, but they cannot settle phone/laptop equivalence without direct evidence. Ask whether the test owner has studied the exact format used and whether timing, scoring, or norms account for any observed format difference.

The 2021 device study is unusually relevant because it examined common personal devices rather than assuming every participant used a laboratory computer. It analyzed visitors to TestMyBrain.org in two samples of 59,587 and 3,818, and reported slower reaction-time performance among mobile-device users, particularly Android smartphone users, after controlling for several demographic measures and performance on untimed vocabulary. The researchers also found interface-related differences for fast-reaction and fine-motor tasks. The study remains observational: people chose their own devices, and device groups differed in characteristics. It is about measured online tasks, not a randomized same-person retake of an IQ test.

Sources: Equivalence of Q-interactive and Paper Administration of Cognitive Tasks: WAIS-5; Retest Effects in Cognitive Ability Tests: A Meta-Analysis

What would count as evidence that two devices are comparable?

Stronger evidence comes from a designed comparison, not one person’s two attempts. Test developers can administer comparable versions under controlled conditions, compare scores and response behavior, inspect whether the same construct is measured, and determine whether scoring or norms need adjustment. The comparison must match the claim: evidence that a specific tablet system resembles paper administration does not establish that any device displaying the same questions will work identically.

Pearson’s WAIS-5 Q-interactive technical report offers a bounded example. It describes a digital sample of 498 and matched paper comparison groups drawn from the publisher’s normative sample. The report compares named cognitive tasks delivered through Q-interactive with paper administration, using its stated equivalence criteria. It reports format effects within the criterion for the studied subtests except Spatial Addition before norm adjustment; the digital Spatial Addition norms were adjusted using equating techniques. The publisher concludes the studied Q-interactive and paper formats were comparable under that procedure.

This report is useful because it shows the work equivalence claims require: a defined instrument, a particular digital platform, a comparison sample, explicit analyses, and attention to a task for which adjustment was needed. It is publisher-produced, and that matters when weighing independence. It is not an independent evaluation of every conclusion, and it cannot validate an unrelated online quiz, a self-administered retake, a phone touchscreen, or a different combination of tasks. The instrument’s content and norms are proprietary; a reader needs none of those details to understand the design principle.

When considering a professional assessment, ask the provider which administration mode they use and whether the instrument’s technical materials support that mode for the purpose at hand. If a provider administers a standardized instrument on a tablet, ask whether it is an approved platform and whether any mode-specific scoring or norm adjustments apply. A general answer that ‘digital tests are validated’ is too broad: evidence should identify the actual instrument and delivery method. When the mode is outside the evidence, the provider should explain what that means for interpretation.

The Standards make a related point: validity evidence is about supporting a particular interpretation and use of scores. It is not a permanent seal attached to a test name. Evidence for an instrument in one population or purpose may not answer whether an altered interface, untimed use, translated instructions, or a different decision is supported. So the reader’s question is not only ‘Is this test valid?’ but ‘What score meaning is supported for this test, delivered this way, for this person and decision?’

A personal retake cannot reproduce this comparison. Even if the same person uses the same test on two screens, the attempt order is fixed: the second follows prior exposure. Unless the test has fresh equivalent forms and a validated cross-device design, the difference is entangled with learning, form difficulty, timing, and state. A device comparison made for curiosity can still be informative about usability, such as whether text is legible. It is not a substitute for technical evidence about score comparability.

Ask for the evidence in concrete terms rather than accepting a broad claim. Which group took the digital form? Was there a comparison with paper or another device? Were forms counterbalanced so that order and familiarity did not always favor one mode? Were examinees matched or randomly assigned? Which outcomes were compared: total score, subtest score, response time, or item functioning? Did the study find a difference that required adjustment? These questions help a reader distinguish an actual equivalence study from a general statement that a platform is widely used.

Sources: Retest Effects in Cognitive Ability Tests: A Meta-Analysis; Standards for Educational and Psychological Testing

Why can a second score move even when ability has not changed?

A retest is a new observation after the person has already encountered the task. The instructions are less novel, the response format is familiar, and strategies may have been discovered. Items may repeat, or a new form may differ in difficulty. The interval can be short or long; the person may be more rested, more distracted, more cautious, or more practiced. A device change occurs alongside these possibilities, so score movement does not identify which one mattered.

The 2018 meta-analysis ‘Retest effects in cognitive ability tests’ combined 174 samples from 122 studies, 786 test outcomes, and 153,185 participants. It examined cognitive ability tests and up to four administrations, finding significant retest effects overall and no further score gains after the third administration in the literature it analyzed. Its moderator analyses found retest effects varied with cognitive operation and content, form equivalence, interval, and participant age. This broad synthesis supports caution about repeated scores, but it does not predict what will happen to one reader on one website.

An important implication is that a second score may rise without a broad change in general intelligence. A person can become more efficient at recognizing what a task asks, remembering the response interface, or applying a strategy to familiar item types. Those gains may matter for that kind of task, and they can be real improvements in performance. They do not automatically transfer to unrelated reasoning tasks or establish a change in a broad underlying ability. The label ‘practice effect’ describes score change associated with repeating an assessment; it does not explain every individual change by itself.

The same meta-analysis found the retest interval mattered, with longer intervals associated with smaller effects in its analyses, though not every comparison followed that pattern. It also distinguishes equivalent and nonequivalent forms. These details make simple advice such as ‘wait a month’ or ‘use a different device’ unjustified unless it comes from the particular test’s technical guidance. Time alone does not erase learning, and a different screen does not erase familiarity with the item format.

The newer study by Robison and colleagues followed 255 young adults across four weekly sessions on 24 computerized measures spanning seven constructs. The report describes generally adequate to high reliability and a stable measurement structure across occasions, while also discussing task-specific strategy, response-mapping automaticity, fatigue, motivation, and measurement error as potential sources of variation. Because the study used a convenience sample, computerized laboratory tasks, and a short interval, it cannot tell us how an adaptive web quiz behaves or whether changing devices would change its score.

The disciplined interpretation of a changed result is therefore modest: performance differed between two administrations. To attribute the difference to the screen, you would need stronger evidence that other factors were controlled and that the test behaves comparably across those modes. To attribute it to general cognitive change would require still more: a suitable measure, reliable change methods, and a relevant assessment context. A higher second number is not a verdict on which device was right.

The possible causes do not carry equal relevance in every test. If a timed task depends on milliseconds, input latency may be meaningful; if a test is untimed and responses are untapped or slowly selected, latency could be less central. If an adaptive test selects later questions based on earlier answers, even one different response may change the path and the subsequent items, so two attempts no longer present the same sequence. The article’s test is adaptive across an item pool, which makes a personal two-device comparison particularly poor at isolating a hardware effect: later questions may differ because earlier performance differed.

Sources: A Comprehensive Psychometrics of Cognitive Ability Measures: Reliability, Practice Effects, and the Stability of Latent Factor Structures across Retesting; Examining 3-Month Test-Retest Reliability and Reliable Change Using the Cambridge Neuropsychological Test Automated Battery

How can stable ability and changing performance both be true?

A test can measure a fairly stable pattern across sessions while the exact score from one sitting changes. Stability and sameness are different claims. Stability usually describes consistency in a measured attribute or the structure of relationships among tasks over time. A single score is an observed result shaped by the selected items, conditions, responses, scoring rules, and measurement uncertainty. The first idea does not promise that the second will be identical every time.

Robison and colleagues’ study is useful for seeing that distinction. Across four weekly sessions with 255 young adults completing 24 computerized measures, the researchers found the factor structure was largely supported across occasions and the factor scores from the first and second attempts were robustly correlated. The paper also reports task-level reliability analyses and discusses systematic variance attributable to latent traits. Those are group-level psychometric findings from the study’s sample and tasks. They do not say that each participant’s scores were identical, or that a person’s online test result can be treated as stable without its own evidence.

Imagine two administrations of a timed visual task. The person may understand the underlying kind of reasoning equally well on both days, yet one attempt includes a moment of distraction or a missed tap. The observed performance changes, while the broad capacity the task samples may remain stable. This is an illustration, not a report about a participant or a claim about how much scores should vary. It helps explain why neither every difference signals real cognitive change nor every similarity proves perfect measurement.

The distinction also prevents a common error in interpreting a retake. If the second attempt matches the first, that repetition does not by itself validate the score: the same unsupported conversion could produce similar estimates twice. If the scores differ, the discrepancy does not prove that one interface was defective. Reproducibility is useful evidence only when the test, interval, sample, administration, and intended interpretation have been studied sufficiently to say what repeatability means.

The study’s boundaries are material. It assessed a set of laboratory-style computerized measures, not a clinical IQ battery, not a browser-based 50-question quiz, and not a phone-versus-laptop comparison. The sample was young adults and the retest schedule was weekly. It cannot tell an adult whether to buy a particular assessment, and it does not provide a personal score adjustment. Its contribution here is conceptual: construct-level consistency and attempt-level change can coexist without contradiction.

For a reader, the practical response is to describe the evidence at the right level. Say ‘my answers differed on a retest’ if that is what happened. Do not jump to ‘my ability changed,’ ‘the device caused it,’ or ‘the higher score is my true score.’ If a consequential question remains, use an assessment selected for that question and ask the qualified provider how repeatability and uncertainty are handled for the exact instrument.

Reliability is often misunderstood as sameness. A reliability estimate describes consistency under a specified design and population; it does not promise identical observed scores or cover a different administration mode automatically. Likewise, a factor structure concerns how measures relate to one another, not whether a score band has been normed for a particular group. These distinctions matter because statements such as ‘the test is reliable’ are incomplete without the population, interval, form, and type of score. Ask what consistency evidence applies to the use you have in mind.

Sources: Examining 3-Month Test-Retest Reliability and Reliable Change Using the Cambridge Neuropsychological Test Automated Battery

What does retesting research say about a changed score?

Retesting research supports the possibility of systematic score change, but the direction and meaning depend on the test and setting. In cognitive ability studies, repeated exposure can be associated with higher scores, with differences across domains, forms, time intervals, ages, and number of administrations. That is a population-level pattern across studies, not a promise that every second attempt rises or an estimate of a particular person’s change.

The 2018 meta-analysis by Scharfen, Peters, and Holling provides a broad cognitive ability evidence base: 122 studies contributed 174 samples and 786 outcomes, with 153,185 participants overall. Its definition concerns score increases after repeated cognitive-test administration. The analysis found significant retest effects and reported no further gains after a third administration in the material analyzed. Moderator findings implicated operation and content, form equivalence, retest interval, and participant age. The authors’ pooled result cannot be imported as a correction factor for a 50-question browser test, especially one without a validated adult comparison group.

The cognitive ability meta-analysis is more directly relevant than research about employment selection, but it is still a synthesis of varied study contexts rather than a direct device experiment. Its result describes retest effects across tests and participants. The authors found no further score gains after the third administration in the analyzed literature and identified moderators; that does not imply that an individual must improve on a second try or that three attempts are useful. The exact task and purpose remain decisive.

The sources contribute a bounded account: the meta-analysis focuses on cognitive ability testing across research settings and moderators; it helps explain why repeat exposure complicates a retake. Neither it nor the other accessible evidence randomly assigns the same people to phone versus laptop while holding test exposure and all other conditions constant. The gap matters because the question is specifically about switching devices. General retest evidence helps explain confounding, but it cannot supply a causal device effect.

A second attempt may show that the person answered differently; it may also reveal improved familiarity with the format or a resolved access problem. Without an appropriate comparison design, these accounts remain competing explanations. A responsible report would not select the preferred account based only on a higher score. In professional practice, the assessment context and instrument guidance matter; the examiner may consider prior exposure when interpreting results, rather than treating the latest number as automatically decisive.

This evidence also resists two overstatements. It would be wrong to say repeated tests are meaningless: repeated measurements can be useful when the instrument and change interpretation are appropriate. It would also be wrong to say practice inevitably inflates every result by a fixed amount. The studies show patterns and moderators, not a universal personal forecast. For a low-stakes browser quiz, use repeated attempts to engage with fresh reasoning practice, not to search for the number that feels most legitimate.

The variation across settings is itself a reason to avoid personal arithmetic from a research average. Meta-analyses pool designs that may vary in test content, participants, intervals, forms, and administration procedures. A pooled effect summarizes those studies under statistical assumptions; it is not an expected point increase for an individual. The fact that moderators matter tells the reader what to inspect, not which adjustment to apply. Without the relevant technical norms for the exact test, adding or subtracting a generic practice allowance would create false precision.

Sources: Retest Effects in Cognitive Ability Tests: A Meta-Analysis; A Comprehensive Psychometrics of Cognitive Ability Measures: Reliability, Practice Effects, and the Stability of Latent Factor Structures across Retesting

Can a score difference show that cognition changed?

A score difference alone cannot establish that cognition changed. To make a claim about change, a professional needs to know the measure’s precision and reliability for the relevant population, the interval between attempts, how much practice is expected, and whether the two forms and administrations are comparable. In clinical neuropsychology, reliable-change methods were developed to distinguish meaningful change from expected score variation and practice effects. Their calculations are tied to specific tests and populations; a reader should not borrow them for an online quiz.

The methods paper ‘Reliable Change Indices and Regression-Based Change Approaches’ in the Archives of Clinical Neuropsychology explains that practice effects can make retest performance look improved when the difference is not necessarily a meaningful change in the cognitive ability of interest. The paper discusses approaches intended to account for measurement error and expected retest effects. Its relevance is methodological: a raw difference is not enough. It does not provide a formula for interpreting Test IQ Free, nor a universal threshold to apply to all cognitive tests.

For a score difference to count as evidence of change, the comparison must first be meaningful. The same construct should be assessed in a way that supports comparison; administration and scoring need to be consistent or adjusted by validated methods; and expected uncertainty must be considered. If one administration was on a small screen with clipped content and the second was not, the data do not form a clean measure of change because access differed. If the test has no relevant standardization or precision evidence, the calculation cannot create it after the fact.

The distinction between statistical change and practical importance also matters. A method may estimate whether a difference is larger than expected from measurement error and practice. That does not by itself say whether the difference affects daily functioning, identifies a diagnosis, or should change a person’s plans. Those conclusions require a relevant assessment question and other evidence. A score is one observation, not a complete account of a person’s reasoning or circumstances.

The Standards for Educational and Psychological Testing frame score interpretation around evidence appropriate to intended uses. The same number may be adequate for informal practice feedback but inappropriate for a high-stakes decision if the evidence does not support that use. If an employer, school, clinician, or other decision-maker needs a defensible conclusion, ask them what evidence they require and use a qualified assessment route. Do not use a sequence of browser retakes to stand in for a formal change analysis.

In the specific case of the site’s free 50-question test, its output is a provisional estimated IQ range from an unvalidated conversion with no validated adult reference population. A repeat on another device might confirm that the site loads and responds on that device. It cannot establish a reliable-change threshold, estimate a confidence interval, show a population rank, or indicate a change in general intelligence. The report’s paid point estimate and percentile use the same provisional conversion, so the extra explanation changes neither the evidence nor the score’s status.

The CANTAB study ‘Examining 3-month test-retest reliability and reliable change’ illustrates the specificity of change methods. In a sample of 75 healthy adults, nine CANTAB tests across visual learning and memory, executive function, and visual attention were repeated after three months. Six of 14 outcome measures showed statistically significant practice effects, ranging from negligible to medium; only three named outcomes had adequate test-retest reliability, and measurement error around differences was large. The study developed change methods for those CANTAB outcomes. It cannot provide a cutoff for a different test, but it shows why change interpretation is built from test-specific data rather than intuition.

Sources: Examining 3-Month Test-Retest Reliability and Reliable Change Using the Cambridge Neuropsychological Test Automated Battery; Standards for Educational and Psychological Testing

When is a new assessment worth paying for?

Paying for a new assessment makes sense when a defined decision needs information that the existing result cannot provide, and the proposed assessment has evidence suited to that question. Examples may include a qualified professional evaluating a specific concern, or an institution requiring a recognized assessment for a stated purpose. The point is not to purchase a more impressive number. It is to obtain a method, administration, interpretation, and supporting evidence appropriate to the decision.

Start by writing the decision in plain language: What do you need to decide, who will use the result, and what kind of evidence will they accept? If the aim is simply to learn how you approach reasoning puzzles, a free activity may be enough. If the aim is to make a consequential clinical or educational judgment, an unsupervised browser quiz is not a substitute for assessment by a qualified professional. The professional can ask about the referral question, choose suitable measures, and interpret results alongside relevant history and observations. This article cannot decide whether any individual needs that assessment.

Before paying, ask the provider what instrument and version will be used, what ages and purposes its evidence covers, how it is administered, how scores are compared, what uncertainty is reported, and who interprets the findings. Ask whether the digital platform is part of the instrument’s supported procedure. Pearson’s WAIS-5 Q-interactive report illustrates that a publisher can study a particular digital format against paper and make a task-specific adjustment. The report does not mean every digital assessment is equivalent, nor does it establish that every provider’s procedure meets the same standard.

Also separate the value of explanation from the validity of measurement. A paid report may be worthwhile when it clarifies a result, shows worked reasoning methods, or provides a structured practice plan that you want. But explanation cannot add a representative reference sample, validate a conversion, or transform a provisional browser estimate into a clinical score. Test IQ Free’s optional $9 report supplies more interpretation and practice material using the same provisional conversion as the free result. Payment adds explanation, not accuracy or validation.

Do not buy an assessment simply because two browser scores disagree. Disagreement may reflect different item sets, test design, scoring, timing, reference samples, or practice, rather than a defective first device. A professional can help when the question is important enough to require better-supported measurement; the decision should depend on the intended use, not on which result is more flattering. If a vendor claims unusual precision or equivalence, ask for the technical evidence and check whether it covers the exact test, mode, age group, and purpose.

Finally, consider practical costs: money, time, accessibility, and whether the result will actually be used. A professionally administered assessment may involve more than a single score and may include an interview or broader evaluation, depending on the referral question and provider. Do not assume those steps or outcomes without asking. The right purchase is the least burdensome method that can responsibly answer the real question; when no consequential decision depends on the result, no paid assessment may be necessary.

A fair comparison between a browser quiz and a professional assessment begins with purpose, not prestige. A browser quiz may give quick reasoning practice and feedback at little cost; a professionally administered assessment may answer a defined question with a qualified examiner and evidence tied to its instrument. They differ in administration, intended use, and supported interpretation. The professional route is not automatically necessary for everyone, and a higher price does not guarantee fit. Confirm that the proposed assessment answers the question you actually have and that the result will be accepted by whoever must use it.

Sources: Standards for Educational and Psychological Testing; Retest Effects in Cognitive Ability Tests: A Meta-Analysis

A practical decision path after the first attempt

Use a short record to decide whether a retake is useful. Note the test name and version if shown, device type, browser, whether timing was active, the date, and the specific problem you observed. Record conditions that clearly interrupted the task, such as a dropped connection or missing content. Do not write a causal conclusion like ‘the phone lowered my score’ unless evidence supports it. The record is there to keep observations separate from explanations.

Then ask: Did a concrete delivery fault happen? If no, a device switch is unlikely to resolve an identifiable problem. If yes, can it be corrected without changing the task beyond that fault? Move to a display where the whole item is visible, use controls that respond, and make sure the timer or connection works. For a casual test, you may repeat once to check usability. Interpret the second result as performance under a new attempt, with prior exposure acknowledged.

Next ask what answer you need. If you want to practice, use worked explanations and fresh examples to understand operations such as tracking a changing visual relation or comparing quantities. A small improvement on similar items may reflect familiarity with that operation. It is not proof that general intelligence has increased. If you need a population comparison, look for a score supported by appropriate norms and documentation for the instrument and population. If you need an educational or clinical decision, ask a qualified provider or the decision-making organization which assessment is appropriate.

Finally, compare the cost of the next step with what it can establish. A device retake can show that a known interface problem is gone. A second self-administered score generally cannot identify a device effect or settle a high-stakes question. A professional assessment may offer a better-supported interpretation, but only when its procedure and evidence fit the use. A paid explanatory report may help with learning, yet it does not validate the underlying conversion. Keep each option matched to its actual job.

If you repeat, avoid turning the exercise into a search for the best outcome. Decide in advance what usability observation you will make, such as whether every visual element fits or every response control works. Do not decide success by whether the final score rises. If the platform uses adaptive items or draws from a larger pool, the second run may not be the same test experience even if it carries the same title. That can be useful for fresh practice, but it makes direct score comparison less transparent.

Sources: Standards for Educational and Psychological Testing; Retest Effects in Cognitive Ability Tests: A Meta-Analysis

What should you do next?

Verdict: change devices to correct an observable delivery failure, not to find a more satisfying score. The strongest exception is a test with documented device-specific problems or an official retake instruction; then follow the instrument’s procedure and disclose the earlier attempt when relevant. Even in that case, the second result is not proof that the first device caused the difference unless the test’s evidence supports that inference.

If the first attempt was usable and your aim is curiosity, keep the result modest and move on to free reasoning practice or worked answers. Test IQ Free’s provisional estimated range has no validated adult reference population, so it is not a normed score, confidence interval, established population comparison, diagnosis, or measure of worth. Its optional report adds interpretation and a plan, not validity. If a consequential decision depends on cognitive assessment, ask what it requires and consult a qualified professional about an instrument and administration suited to that decision.

The next steps are observable: write down any specific device fault; correct that fault before one optional usability retake; label the result as a new performance rather than a verified score; then decide whether your actual question is low-stakes practice or requires supported assessment. Pay only if the proposed assessment can answer that defined question.

One caveat can change the recommendation: when the assessment provider documents a technical failure and supplies a retest procedure, use that procedure rather than improvising. Keep the first attempt information if requested, and tell the provider whether items were already seen. Their procedure may include a new form, a wait interval, or a particular device setup; these details belong to that instrument’s technical guidance and should not be guessed. If no such guidance exists, the prudent claim remains narrower: the second screen made the task accessible, while the score comparison is unresolved.

Sources: Standards for Educational and Psychological Testing; Retest Effects in Cognitive Ability Tests: A Meta-Analysis

Questions readers ask