Test Methodology & Science
IQ test accuracy is one of the most searched but least clearly explained topics in psychometrics. Whether you just received an online score or are evaluating whether to pursue a formal assessment, understanding what accuracy actually means — and what limits it — is essential for reading your result correctly. This guide covers reliability and validity, standard error of measurement, the six major accuracy factors, how clinical and online tests compare, and how to interpret your result with appropriate confidence.
In psychometrics, accuracy is not a single property — it is the combination of two distinct but related concepts: reliability and validity. A test can be reliable without being valid, and both need to be answered before a score can be interpreted with confidence.
Reliability
Does the test produce consistent scores across repeated administrations under equivalent conditions? Reliability minimises random error. A test can be highly reliable (consistently giving the same score) while measuring the wrong thing entirely.
Validity
Does the test actually measure what it claims to measure? Well-designed IQ tests are validated against independent measures including academic performance, occupational outcomes, and other cognitive assessments. See the IQ by country guide for how cross-national data applies these caveats.
Test-retest reliability is measured by administering the same test to the same group at two points in time, then correlating the scores. The WAIS-IV — published by Pearson Assessments — reports a Full Scale IQ reliability coefficient of 0.98, with composite scores ranging from 0.87 to 0.97 across sub-indices. The WAIS-5, normed on data collected after 2020, reports coefficients of 0.90–0.97 across index scores. Values above 0.90 are considered excellent for clinical instruments. Online tests rarely publish equivalent data, making direct comparison difficult.
Internal consistency measures whether all items within the same sub-scale are measuring the same underlying construct. It is reported as Cronbach’s alpha, where values above 0.80 are acceptable for clinical instruments and above 0.90 for high-stakes decisions. A test without published alpha coefficients cannot make a credible reliability claim at all.
One frequently overlooked accuracy issue is that IQ test norms become outdated. The Flynn Effect — the documented generational rise in raw cognitive test performance — averages approximately 2.93 IQ points per decade across major Wechsler and Stanford-Binet instruments, according to a meta-analysis of 285 studies. A test standardised 15 years ago will report systematically inflated scores compared to a freshly normed instrument.
IQ Score Distribution (Mean 100, SD 15)
Every IQ score is a point estimate within a range of uncertainty, not a precise fixed number. Use the slider below to see the 68% and 95% confidence intervals around any score, based on the WAIS-IV Full Scale SEM of ±2.16 points — the most precisely documented clinical benchmark available. Understanding this range is the single most important step in reading any IQ result correctly.
68% Confidence Interval
103 – 107
2 in 3 repeated test sessions would fall within this range under equivalent conditions.
95% Confidence Interval
101 – 109
Percentile span: 53th – 73th. This is the range your true score most likely falls within.
Score range on the IQ scale
Confidence intervals calculated using the WAIS-IV Full Scale IQ SEM of 2.16 points (published reliability: 0.98). IQMog’s SEM is unverified and likely higher — use these figures as a reference baseline, not as IQMog-specific values.
The SEM formula is: SEM = SD × √(1 − r), where SD is 15 (the IQ scale standard deviation) and r is the reliability coefficient. With a WAIS-IV FSIQ reliability of 0.98, this gives SEM = 15 × √(0.02) ≈ 2.12–2.16 points. For a reported score of 115, the 95% confidence interval spans approximately 111–119 — straddling the High Average / Superior band boundary. The American Psychological Association recommends reporting confidence intervals alongside IQ scores for exactly this reason.
On most instruments, SEM is slightly larger at the extremes of the distribution because the item pool provides less differentiating information in the tails. For high IQ scores, this means a reported result of 135 carries more uncertainty than a result of 105, even on the same instrument.
A score of 130 on a clinical instrument might represent a true score anywhere from roughly 126 to 134 at 95% confidence — straddling the line between Superior and Very Superior. Treating a single reported number as a definitive boundary is one of the most common misinterpretations of psychometric results. See the IQ score chart and percentile reference for how each point maps to population standing.
Score variance in any IQ test — clinical or online — comes from a finite set of identifiable sources. Expand each factor below to understand what it is, how large its effect typically is, and what it means in practice for interpreting your result.
The norming sample is the reference population used to calibrate a test’s scoring scale. Accuracy depends directly on how large, representative, and recent this sample is.
Where and how you take a test significantly affects your measured score. Noise, interruptions, device quality, and time of day all introduce variance.
Prior exposure to IQ test item formats inflates scores on retesting — a well-documented measurement artefact that affects both clinical and online assessments.
Elevated anxiety at test time is a documented suppressor of measured cognitive performance, particularly on timed reasoning tasks.
All tests have a maximum and minimum measurable score range. Near these boundaries, precision degrades because the item pool lacks sufficient difficulty gradient.
For clinical tests, the examiner’s training and procedural adherence is a material source of score variance. Self-administered online tests eliminate examiner variance but replace it with uncontrolled environmental variance.
Effect magnitudes are approximate ranges derived from published psychometric literature. Individual instruments and populations vary.
The accuracy gap between a clinically administered IQ test and a self-administered online assessment is significant and predictable. It is not that online tests are useless — it is that they serve a fundamentally different purpose. The table below maps every major accuracy dimension side by side.
| Factor | Clinical Test (e.g. WAIS-IV) | Online Test (e.g. IQMog) |
|---|---|---|
| Norming sample size | 2,200+ stratified participants | Rarely disclosed publicly |
| Administration control | Proctored by licensed psychologist | Self-administered, unproctored |
| Full-scale reliability | 0.98 (WAIS-IV FSIQ) | Unverified for most instruments |
| Typical SEM | ±2.16 pts (WAIS-IV) | Higher; unquantified for most |
| Score ceiling reliability | Validated through IQ ≈145+ | Degrades above ≈130 |
| Accepted in clinical contexts | Yes | No |
| Practice effect controls | 12-month retesting standard | Minimal controls in place |
| Norm recency | Restandardised on defined schedules | Update schedule rarely disclosed |
WAIS-IV figures sourced from the Pearson Assessments technical and interpretive manual. Online SEM estimates are conservative approximations based on published psychometric literature.
The norming sample determines what a score of 100 means, how standard deviations are calibrated, and whether percentile estimates are trustworthy. Clinical instruments invest heavily in stratified sampling that matches national demographic distributions across age, education, ethnicity, and geography. A recent 2025 meta-analysis found that remote and in-person WAIS administrations differed by less than one tenth of a standard deviation — below thresholds that matter clinically — when test conditions were well-controlled. For a full percentile mapping, see the IQ score chart and percentile reference.
Clinical assessments are conducted under controlled, proctored conditions with standardised equipment, scripted instructions, and trained examiners. This eliminates the environmental variance that is unavoidable in self-administered online testing. Test conditions matter because IQ tests measure performance under specific circumstances — and performance is partly a function of context.
Score precision degrades at the high end of any test’s range because the norming sample contains progressively fewer people at extreme scores, providing less statistical basis for precise differentiation. Clinical instruments extend their item difficulty gradient to provide reasonable precision up to approximately IQ 145–150. Most online tests are not designed or normed to reliably differentiate above IQ 130.
IQMog is a fixed-form, 20-question Raven-style pattern test scored on the standard IQ scale (mean 100, SD 15) using Classical Test Theory. It produces an IQ estimate, percentile band, and cognitive profile — useful for personal benchmarking and understanding your approximate position in the distribution. It is not a proctored instrument, does not carry the norming sample depth of a clinical test, and carries higher measurement uncertainty, particularly above 130. It cannot be used in medical, legal, or educational selection contexts. What it can do well is give you a structured, consistent, culture-fair baseline for fluid reasoning performance.
The Flynn Effect is the well-documented phenomenon of rising raw IQ test scores across generations — first identified by James Flynn in the 1980s. A meta-analysis of 285 studies since 1951 found a mean gain of 2.31 standard score points per decade across all intelligence tests, and 2.93 points per decade specifically for major Wechsler and Stanford-Binet instruments. This has a direct, practical consequence for how you should interpret any IQ score.
Norms aged 5 years
≈1.5 pts
Minor impact on most clinical decisions
Norms aged 10 years
≈3 pts
Meaningful at classification boundaries. Renorming is typically due.
Norms aged 15+ years
≈4–5 pts
Results may misclassify individuals near band boundaries.
Recent decades have shown score declines in several developed countries including Norway, Denmark, the United Kingdom, the Netherlands, Finland, France, and Estonia — with decline rates ranging from 0.38 to 4.30 IQ points per decade. This reversal appears to reflect environmental and educational factors rather than genetic changes, and is an active area of research. Its practical implication: the Flynn Effect should not be assumed to still be running in the same direction in all populations.
A well-interpreted online result is more useful than a poorly interpreted clinical one. The key is to extract what the result can credibly tell you while being clear about what it cannot.
IQMog measures fluid reasoning performance on Raven-style matrix pattern items. Research shows that Raven’s Progressive Matrices correlate with broader intelligence measures at .54–.86, sharing roughly 50% of their variance with g — making them a strong but not exhaustive signal. Fluid reasoning is among the most g-loaded cognitive abilities (most closely correlated with general intelligence), which is why matrix-based formats are the dominant design in online cognitive assessment. Your result reflects how you performed on this specific task, under your specific session conditions. It does not capture crystallised knowledge, verbal reasoning, processing speed, or working memory capacity. For context on what different scores mean in population terms, see the average IQ score breakdown or the full IQ score ranges guide.
A single result is a data point. Two results from controlled sessions that agree within a narrow range are a baseline. If you take the test twice — rested, distraction-free, full-screen, not rushed — and the scores fall within 5–8 points of each other, that consistency is meaningful. It suggests the result is capturing something stable rather than session-specific noise.
If two controlled-condition results differ by more than 10–15 points, an environmental factor — fatigue, interruption, anxiety — likely suppressed one session. The higher of the two controlled results is typically a better estimate of your stable baseline.
Environment
Quiet room, no interruptions, desktop or laptop screen preferred
Timing
Morning or whenever you feel most alert, not immediately after meals or strenuous activity
State
Well rested, not anxious, not medicated in ways that affect concentration
Mindset
Treat as exploration, not evaluation. Performance is better under low-stakes framing
Score changes between sessions are almost always a mix of true learning, practice effects, and random error — not a pure signal of cognitive change. Wait at least six months between serious retests for a more stable comparison. Within that window, treat any difference of less than ±8 points as noise rather than signal.
An online assessment is appropriate for personal benchmarking, understanding your position in the distribution, tracking improvement, or satisfying curiosity. There are specific contexts where only a formally administered clinical assessment will serve.
High-IQ society eligibility
Mensa International and other recognised high-IQ organisations require scores from approved, proctored instruments. Online results do not qualify regardless of score. See the Mensa IQ score guide for threshold and qualifying test details.
Academic or occupational selection
If an employer, university, or programme requires cognitive assessment evidence, only a formally administered test from an approved instrument will be accepted.
Medical, educational, or legal contexts
Gifted programme qualification, learning disability assessment, neuropsychological evaluation, and related determinations require clinical testing under controlled conditions.
Consistent very high online scores
If you consistently score above 130 across multiple platforms and controlled sessions, a clinical assessment provides the kind of evidence that online scores cannot.
Significant score inconsistency
If your scores vary by more than 15 points across controlled sessions, a clinical assessment with standardised environmental control will give you a more reliable baseline than further online testing.
To pursue a formal assessment, contact a licensed psychologist in your area. The American Psychological Association’s overview of intelligence testing explains what clinical assessment entails and how to find a qualified assessor.
0.98
WAIS-IV FSIQ Reliability
Values above 0.90 are considered excellent for clinical instruments. Published reliability coefficient for the Full Scale IQ composite.
±2.16 pts
WAIS-IV Full Scale SEM
Standard error of measurement on the most widely used adult intelligence battery. 95% CI spans ±4.24 points around the reported score.
~3 pts/decade
Flynn Effect Rate
Approximate rate of raw score inflation from outdated norms. Tests more than 10–15 years old may overestimate IQ by this amount or more.
Online IQ tests provide a useful directional measure of pattern reasoning but are less accurate than clinically administered tests due to uncontrolled conditions, undisclosed norming methodology, and reduced score ceiling reliability above 130. A consistent result across two controlled sessions is more meaningful than any single run.
The standard error of measurement (SEM) quantifies how much a reported score might differ from a person’s true underlying score due to random error. On the WAIS-IV Full Scale IQ, the SEM is approximately 2.16 points (reliability: 0.98). A reported score of 100 has a 95% confidence interval of roughly 96–104 — meaning the true score falls in that range with 95% probability.
Well-designed IQ tests have documented construct validity — they consistently correlate with cognitive abilities associated with general fluid intelligence. However, Raven’s Progressive Matrices account for roughly 50% of the variance in g (general intelligence), meaning important cognitive dimensions fall outside their scope. Validity also depends heavily on how carefully the test was normed and constructed.
Clinical tests like the WAIS-IV are administered by trained psychologists under controlled conditions, normed on thousands of stratified participants (reliability 0.98, SEM ≈2.16 points), and have published reliability data. Online tests are self-administered with no proctor, typically do not disclose equivalent norming data, and carry higher measurement uncertainty — especially at score extremes. Clinical results can be used in medical, educational, and legal contexts; online results cannot.
Yes. Short-term score gains from practice are well-documented. Familiarity with item formats, improved strategy, and reduced anxiety all contribute to increases that may not reflect underlying cognitive change. Practice effects typically produce 5–15 point gains on a second attempt, declining on subsequent attempts. Clinical standards recommend waiting at least 12 months before re-administering the same instrument for a valid comparison.
Ready to measure yours?
20 Raven-style matrix questions. Instant IQ estimate, percentile band, and cognitive profile on completion. No email required to start. Built on a fixed, culture-fair dataset with consistent scoring methodology — so your result means the same thing each time you take it.
Results are an IQ estimate for personal benchmarking. Not a clinical assessment. Cannot be used for Mensa admission or formal selection purposes.