How transparent is this IQ test about its scoring?
This page explains the site's scoring approach using psychometric concepts such as Item Response Theory, calibration and measurement error. Every step from your answers to your score is documented.
Online IQ tests that explain their scoring model, item difficulty and measurement precision are the most useful ones. This test explains all three.
What scoring methods does this IQ test use?
Five commitments shape how a score is produced and reported.
Question difficulty shapes how each answer affects your estimate.
Ability is estimated from the response pattern, improving accuracy across difficulty levels.
Standard Error of Measurement and confidence intervals are reported with the score.
Inconsistent patterns and rapid responding are detected and flagged.
Every formula and check behind the score is documented openly.
The science behind your IQ score
Intelligence testing is not just counting correct answers. It is a measurement problem: estimate ability from a finite set of responses, while being clear about uncertainty and the limits of interpretation. Three frameworks inform the approach.
Cattell–Horn–Carroll (CHC) theory
A commonly cited framework for describing broad and narrow cognitive abilities. It is why the test reports several domains instead of treating intelligence as one undifferentiated score.
Fluid reasoning (Gf), crystallised knowledge (Gc), working memory capacity (Gwm), processing speed (Gs), visual-spatial thinking (Gv).
Over 70 specific cognitive skills sit within the broad domains.
Spearman’s g-factor theory
A classic theory proposing a general factor behind performance across many cognitive tasks. It is why one overall IQ score is meaningful alongside the domain breakdown.
A shared ability underlying intellectual tasks, accounting for roughly 40–50% of performance variance across cognitive domains.
Domain-specific skills and knowledge: verbal, mathematical, spatial and memory.
Modern psychometric theory (IRT & CAT)
Item-response ideas, estimation and adaptive difficulty explain why a score is more than a raw total. They let every answer count according to how informative it is.
Models how item difficulty, discrimination and guessing influence score estimation from a response pattern.
CAT-inspired: question selection can be guided by your responses and estimated difficulty.
The four cognitive domains this IQ test measures
Logical Reasoning
Your ability to identify patterns, solve novel problems and think abstractly without relying heavily on prior knowledge.
- Pattern recognition and completion
- Deductive and inductive reasoning
- Abstract problem solving
- Logical consistency analysis
Spatial Intelligence
Your ability to visualise, manipulate and reason about objects in space: often linked to technical, design and visual problem solving.
- Mental rotation of 3D objects
- Spatial visualisation
- Pattern transformation
- Geometric reasoning
Verbal Comprehension
Language understanding, vocabulary depth and verbal reasoning: knowledge and skills accumulated through education and experience.
- Vocabulary and word meaning
- Verbal analogies and relationships
- Reading comprehension
- Linguistic pattern recognition
Working Memory
Your capacity to hold and manipulate information in mind at once, which matters for reasoning, learning and multi-step problems.
- Information retention capacity
- Mental manipulation of data
- Attention control
- Cognitive processing efficiency
How to read what the score means
Three ideas that make an IRT score more meaningful than a raw total.
Question calibration concepts
Question difficulty and response patterns can affect score interpretation more than a plain raw total.
3PL-MAP scoring model
An item-response style scoring concept: ability is estimated from a response pattern instead of the raw correct count alone.
Precise percentile rankings
Percentiles are computed on the familiar 100-mean, 15-SD scale using the standard normal distribution.
How your IQ score is calculated
Your score is not a simple raw total. Question difficulty and your response pattern are combined, through adaptive scoring, into one result in four steps.
Response pattern analysis
Which questions you got right, their difficulty and how you answered, not just how many.
IRT ability estimation
The 3PL-MAP model turns the response pattern into an ability estimate θ.
Age-adjusted normalisation
Normalisation and interpretation context, rather than treating every raw total as directly comparable.
Scale transformation
θ is expressed on the familiar 100-mean, 15-SD IQ scale, so the result is easy to read and compare.
Every question has its own curve
The curve shows the chance that someone at ability θ answers one question correctly. Move the three item parameters to see why a hard, sharp question tells us more than an easy one.
Every question calibrated on thousands of real test-takers
Item parameters aren't guesses. Each question's difficulty, discrimination and guessing rate is estimated from over 5,000 completed tests, and re-estimated as new responses arrive.
Where your score sits on the bell curve
Percentiles are derived from the standard normal distribution (mean 100, SD 15) using the cumulative distribution function. That gives an exact ranking for every possible score.
How we maintain test integrity
Four layers of checks run on every result, so a rushed or careless sitting is flagged instead of reported as your ability.
Person-fit analysis
Detects inconsistent response patterns that may mean random guessing, carelessness or invalid testing conditions.
- →Guttman scalogram analysis for response consistency
- →Lz statistic for aberrant response detection
- →Response-time outliers: answers under 2 seconds flagged
Validity indicators
Quality flags monitor test-taking behaviour and warn when a result may not reflect true ability.
- →Rapid responding detection with validity penalties
- →Poor likelihood fit (minimum 8 calibrated items required)
- →FSIQ–GAI discrepancy over 8 points triggers a flag
Precision measurement
Confidence intervals and measurement uncertainty come from Fisher information in the IRT model.
- →SEM = 1/√I(θ) from Fisher information
- →95% confidence interval: θ ± 1.96 × SEM
- →Test information function I(θ) to optimise precision
Continuous calibration
Item parameters are stored in a database and updated regularly from new response data.
- →Database-backed item calibration (PostgreSQL)
- →Dynamic parameter estimation
- →Regular psychometric audits and updates
More informative items, a tighter range
Each answered item adds Fisher information I(θ). Precision is SEM = 1/√I(θ), and the 95% range is θ ± 1.96 × SEM. Drag to see how the range narrows as items add up.
Illustration assuming 0.35 information per item near your ability level. The minimum of 8 calibrated items is the site's validity threshold.
Getting the most accurate result
Continuously refined rankings
Rankings use the μ = 100, σ = 15 distribution plus our calibration sample of 5,000+ completed tests, so they sharpen as more people take the test.
Take it rested and focused
Finish in one sitting without distractions. Fatigue and interruptions can pull any result a few points off your true ability.
Available in nine languages
Take the test in English, Turkish, Arabic, Chinese, German, Indonesian, Russian, Spanish or Hindi.
Track your trend
Take the test again after training and compare results on the same consistent scale.
What this test is good for
Alignment with testing standards
The methodology is built on four core principles of psychometric measurement. The same principles guide professional cognitive assessment.
The math behind the score
For researchers and curious readers: the IRT model, parameter estimation and scoring formulas.
3-parameter logistic (3PL) model
P(X=1 | θ,a,b,c) = c + (1 − c) × 1 / (1 + e^(−a(θ − b)))θ is latent ability, a is item discrimination, b is item difficulty and c is the pseudo-guessing parameter.
Maximum a posteriori (MAP) estimation
θ̂ = argmax L(responses | θ) × N(θ; 0, 1)Newton–Raphson iterations with a Bayesian prior (μ = 0, σ = 1), maximising the posterior probability given the response pattern.
Standard error of measurement
SEM(θ) = 1 / √I(θ)95% CI = θ ± 1.96 × SEMPrecision comes from the test information function I(θ) and is used to build 95% confidence intervals.
Person-fit analysis
A multi-part validity check: Guttman scalogram analysis (response consistency), mean log-likelihood statistic (model fit) and response-time outlier detection (rapid responding).
Methodology version 1.0 (January 2025) · refined continuously from psychometric research and response data.