IQ test questions: the main types, with examples

Short answer: IQ test questions come in a handful of families: matrix and pattern puzzles, number series and balance problems, verbal questions about words and ideas, spatial puzzles with blocks and shapes, working memory tasks with digits, and processing speed tasks done against a clock. A full clinical test such as the WAIS-5 mixes several families, while some tests use one. Raven's Progressive Matrices, for example, uses only pattern puzzles (APA task force report, 1995). Real test questions are kept secure and are not published.

The main types of IQ test questions

1. Matrix and pattern puzzles (fluid reasoning). Find the rule in a grid and pick the missing piece. Invented example: a row shows a triangle with one dot, a triangle with two dots, then an empty cell. Answer: a triangle with three dots.
2. Number series (inductive reasoning). Find the next number. Invented example: 12, 10, 11, 9, 10, 8, ? Answer: 9, because the rule is subtract 2, then add 1, and repeat.
3. Balance problems (quantitative reasoning). Work out what keeps a scale level. Invented example: one star weighs the same as two circles, and one circle weighs the same as three squares. How many squares balance one star? Answer: 6.
4. Verbal questions (comprehension-knowledge). Define a word, say how two ideas are alike, or complete an analogy. Invented example: glove is to hand as sock is to ? Answer: foot.
5. Spatial puzzles (visual processing). Rebuild a design, turn a shape in your head, or unfold a folded paper. Invented example: a square sheet is folded in half once and one hole is punched through both layers, away from the fold. How many holes are there when you unfold it? Answer: 2, one on each side of the fold, the same distance from it.
6. Working memory tasks. Hold numbers or letters in mind and change their order. Invented example: hear 7, 2, 9, 4 and say them backward. Answer: 4, 9, 2, 7.
7. Processing speed tasks. Do a simple task as fast and as accurately as you can. Invented example: a key pairs each digit with a symbol, and you copy the right symbol under a long row of digits before time runs out. There is no trick; the score is how many you get right in the time.

Every example on this page was written for this guide.

Where these question types come from

Most modern IQ tests are described with the Cattell-Horn-Carroll model, usually shortened to CHC. McGrew (2023) credits "Carroll’s endorsement of the Cattell–Horn Gf-Gc theory" with cementing the birth of what became CHC. The model now names more than a dozen broad abilities. Biesok and colleagues (2024) write that "The recent CHC theory encompasses 17 broad abilities". A full battery samples about five of them: the WAIS-5 has five primary indexes, according to Winter and colleagues (2024), and the Stanford-Binet 5 measures five factors, according to PRO-ED (read November 4, 2026). The question families above map onto those.

The definitions below come from a table Wilson and colleagues (2023) adapted from Schneider and McGrew (2018):

• Fluid reasoning (Gf): "The use of deliberate and controlled mental processes requiring focused attention to solve novel problems".
• Comprehension-knowledge (Gc): "The ability to comprehend and communicate culturally valued knowledge".
• Visual processing (Gv): "The ability to use mental imagery to perceive, discriminate, and manipulate mental images to solve problems".
• Working memory (Gwm): "The ability to maintain and manipulate information in active attention".
• Processing speed (Gs): "The ability to control attention to automatically, quickly, and fluently perform simple repetitive cognitive tasks".

The APA's 1995 task force report, "Intelligence: Knowns and Unknowns", put the practical side plainly. "Some use only a single type of item or question", it says, but the Wechsler tests and the Stanford-Binet "include many different types of items, both verbal and nonverbal." Our guide to what an IQ test measures explains how those parts become one score.

Matrix and pattern questions

A matrix question shows a grid of shapes with one cell empty. You work out the rules that run across the rows and down the columns, then choose the piece that completes the grid. Preuß and Preckel (2025) describe the task: matrix items "require test takers to analyze figural elements in a matrix, to infer their relationships, and to select the correct response to complete the matrix."

Raven's Progressive Matrices is the best-known test built only from this format. Langener and colleagues (2021) describe the Standard Progressive Matrices: "Each item shows a geometric pattern with a missing piece." The test has five sets, A to E, of 12 items each, 60 in all, and each set is harder than the last. First published in 1938, according to Langener and colleagues (2021), it was John C. Raven's test. The Wechsler tests have their own Matrix Reasoning subtest, and our page on Raven's Progressive Matrices compares the two.

Invented example (described in words, since this page has no drawings): a 3 by 3 grid. Row one holds a circle, a square and a triangle, each with one dot. Row two holds a square, a triangle and a circle, each with two dots. Row three holds a triangle, a circle and an empty cell. Answer: a square with three dots. Rule: each row uses every shape once, and the dot count goes up by one per row.

Matrix questions use no words and no arithmetic, which is why they are often called nonverbal. Our guide to abstract reasoning covers the wider family of pattern puzzles that includes them.

Number series and balance questions

Number series questions give you a run of numbers and ask for the next one. They test inductive reasoning. Loe and colleagues (2018) write that inductive reasoning "is typically measured by series completion tasks, which form a common part of many standardised intelligence test batteries".

Invented example: 2, 6, 4, 8, 6, 10, ? Answer: 8. Rule: add 4, then subtract 2, and repeat.

Balance questions show scales with shapes as weights. In the Figure Weights subtest of the Wechsler tests, McCrea and Robinson (2011) write, "an examinee views scales with missing weights and selects the response option that is best suited to keep the scales balanced." The WAIS-IV technical manual, they report, says the task measures quantitative and analogical reasoning. The invented star, circle and square example in the list above is this kind of question.

Some tests also ask arithmetic word problems read aloud, which you solve in your head. Pearson's mapping of the WAIS-IV places the Arithmetic subtest under both quantitative knowledge and short-term memory, so it draws on holding numbers in mind as well as on number skill. The WAIS-5 product page lists a "New quantitative reasoning index" among its features (Pearson, read November 4, 2026).

Verbal questions: vocabulary, similarities and analogies

Verbal questions ask what you know about words and ideas, and how well you can reason with them. In the Wechsler tests they include the four below; the quoted descriptions are from Pizzigallo and colleagues (2023).

• Vocabulary: say what a word means ("knowledge and definition of words").
• Similarities: explain how two things or ideas are alike ("requiring indication of an aspect shared by two concepts").
• Information: answer general knowledge questions.
• Comprehension: answer questions about everyday situations and common ideas.

Invented example of a similarities question: how are a map and a compass alike? A good answer: both help you find your way.

Verbal analogies ("A is to B as C is to ?") appear in other batteries. The Intelligence Structure Test 2000 R, for example, has verbal analogies, sentence completion and similarities as its verbal task groups (Jokste and colleagues, 2024).

These questions mostly draw on comprehension-knowledge, the knowledge you build through school, reading and culture. Pearson's own WAIS-IV mapping puts Similarities under fluid reasoning, though, so it is a verbal item that also takes reasoning. Our page on crystallized intelligence covers the knowledge side.

A test can also leave words out on purpose. Pearson says the Nonverbal Index of the WAIS-5 "includes subtests from four of the five cognitive domains that contribute to the FSIQ (Verbal Comprehension is not included)" (WAIS-5 FAQ, read November 4, 2026).

Spatial questions: blocks, rotation and paper folding

Spatial questions ask you to see a shape in your mind and change it. Three kinds turn up in tests and research.

• Block design. You copy a pattern with red-and-white cubes against a time limit. Pizzigallo and colleagues describe it as "requiring the reconstruction of visual patterns by appropriately combining faces of a series of cubes".
• Visual puzzles. You look at a finished design and choose the pieces that would build it. McCrea and Robinson write that "the examinee views a completed puzzle and selects three options" that together rebuild it.
• Mental rotation. You decide whether two drawings show the same object turned, or its mirror image. The classic study is Shepard and Metzler (1971), who found that the time it takes rises in a straight line with the angle between the two drawings.

Paper folding is a fourth kind. Mitko and Fischer (2020) describe the Paper Folding Test: participants "imagine a square piece of paper being folded and hole-punched through all of the folded layers", then decide where the holes would be once the paper is unfolded. The invented example in the list above works the same way.

Working memory questions

Working memory questions ask you to hold information in mind and do something with it. The best known is digit span. Pizzigallo and colleagues describe it as "requiring immediate recall and repetition of digit series, progressively increasing in length, both in ascending and descending orders".

In practice that means three versions:

• Forward: repeat the digits in the order you heard them.
• Backward: say them in reverse, as in the invented 7, 2, 9, 4 example above.
• Sequencing: put them in order from smallest to largest. Invented example: hear 6, 1, 8, 3. Answer: 1, 3, 6, 8.

The WAIS-5 changed these tasks. Winter and colleagues (2024) write that Digit Span is "now called Digit Sequencing", and Pearson's FAQ says Digits Forward "is not an FSIQ subtest". The WAIS-5 also added Running Digits, which Pearson describes as measuring "registration and maintenance, including refreshing and updating information held in short term storage while processing additional stimuli." Visual memory tasks such as Symbol Span ask you to hold a design and its place on the page in mind.

Processing speed questions

Processing speed questions are simple on purpose. The question is how many you can do correctly in a fixed time, not whether you can solve them at all. Pizzigallo and colleagues (2023) describe the two classic tasks.

• Coding: copy the symbol that goes with each digit, using a key: "requiring rapid association of numbers to symbols within a grid".
• Symbol search: decide, row by row, whether a target symbol appears in a group: "requiring participants to swiftly determine whether a specific target stimulus is present within a group of symbols".

The WAIS-5 gave speed and working memory, taken together, less weight in the full scale IQ than the WAIS-IV did. Winter and colleagues (2024) write that "the Processing Speed Index and Working Memory Index comprised 40% of the WAIS-IV Full Scale, a value that was reduced to 28.6% on the WAIS-5." Pearson also dropped Picture Completion, partly "to decrease the emphasis on speed in the battery" (WAIS-5 FAQ, read November 4, 2026).

What makes a question harder

Many tests order items from easier to harder. In the Intelligence Structure Test 2000 R, Jokste and colleagues (2024) write, "each task group starts with easier items and finishes with the most difficult". Raven's sets, as above, each become harder than the last.

For matrix puzzles, the research names what drives difficulty. Carpenter, Just and Shell (1990) found that items with more rules were solved less often. Preuß and Preckel (2025) summarize it: Carpenter and colleagues "reported significant negative correlations between the number of rules in matrix items and their solution probabilities". Pallentin, Danner and Rummel (2023) put it this way: "Items usually become more difficult when the number and complexity of their underlying operations increase".

Why that happens is less settled. Carpenter's account gave working memory a large role, but Preuß and Preckel note that "newer research shows that the relationship between working memory and performance on figural matrices does not consistently vary with the number of rules". So more rules make a matrix harder, but the reason is still being studied.

Which tests use which questions

• WAIS-5 (Pearson, 2024), ages 16 to 90. It has 20 subtests in five primary indexes, and the full scale IQ "is a composite of seven subtests": Vocabulary, Similarities, Block Design, Matrix Reasoning, Figure Weights, Digit Sequencing and Coding, according to Winter and colleagues (2024). Pearson gives a completion time of "45 minutes for the 7-subtest FSIQ", per its WAIS-5 product page (read November 4, 2026). Our page on the WAIS IQ test covers it in full.
• Stanford-Binet 5. PRO-ED, its publisher, says "Pairs of SB5 subtests measure five cognitive factors": fluid reasoning, knowledge, quantitative reasoning, visual-spatial processing and working memory. "One half of the test has nonverbal content that requires no (or minimal) verbal responses from the examinee". It has 10 subtests at "Approximately 5 minutes for each" (PRO-ED, read November 4, 2026). See our page on the Stanford-Binet test.
• Raven's Progressive Matrices. One format, matrix puzzles, in 60 items for the standard form (Langener and colleagues, 2021). The APA report calls it "a nonverbal, untimed test that requires inductive reasoning about perceptual patterns".
• Online matrix tests. Some online IQ tests use only matrix puzzles. A one-format test samples fluid reasoning; it does not sample the verbal, memory or speed questions a full battery adds.

Clinical batteries are counted in subtests rather than questions, so there is no single answer to "how many questions are on an IQ test". Our guide to how long an IQ test takes gives the times test by test.

Why you cannot see the real questions

Psychologists are bound to protect test content. The APA Ethics Code, Standard 9.11, defines test materials to include "test questions or stimuli" and says psychologists "make reasonable efforts to maintain the integrity and security of test materials" (APA Ethical Principles of Psychologists and Code of Conduct, as amended effective January 1, 2017).

Publishers also limit who can buy the tests. Pearson's qualifications policy says "Tests with a C qualification require a high level of expertise in test interpretation", and the WAIS-5 is a Level C test (Pearson, read November 4, 2026). Some tests build their forms from an item bank: Pearson says Raven's 2 digital forms "are constructed from item bank to limit item overlap between test takers".

So a page that claims to show "real WAIS questions" is showing practice material at best. Mensa's own online challenge says it is "provided for entertainment purposes only, and your score will not qualify you for Mensa", according to its page (read November 4, 2026). Our page on the Mensa IQ test explains what does count.

What IQ tests do not ask

None of the subtests in the WAIS-5 or the Stanford-Binet 5 is a personality measure. Some batteries do include a general knowledge subtest. The Wechsler Information subtest asks general knowledge questions, though in the WAIS-5 it is not one of the seven subtests in the full scale IQ, according to Winter and colleagues (2024).

IQ tests also leave out large parts of what people mean by intelligence. The APA's 1995 report says: "It is widely agreed that standardized tests do not sample all forms of intelligence. Obvious examples include creativity, wisdom, practical sense and social sensitivity; there are surely others."

Scores on creative-thinking tests and IQ scores are related, but only modestly. A 2021 meta-analysis by Gerwig and colleagues, on intelligence and divergent thinking, pooled "849 correlation coefficients from 112 studies" and found "a significant positive correlation of r = .25." By our calculation, that is about 6 percent shared variance. Our page on EQ vs IQ covers another ability IQ tests do not measure.

Can you practice IQ test questions?

You can get used to the formats, and that usually lifts a score a little the second time. A 2007 meta-analysis by Hausknecht and colleagues, of 107 samples, found "an adjusted overall effect size of .26" for retaking cognitive ability tests. On the IQ scale, with a standard deviation of 15, that is about 3.9 points by our calculation. The gains were larger "when identical forms were used."

Pearson builds Raven's 2 digital forms from an item bank partly for "limitation of practice effects", which suggests familiarity with the items drives some of the gain. Every score is a range rather than a single point, and your result can differ between tests and between sittings. Our guide to studying for an IQ test covers what helps and what does not, and when a retake is fair.

Taking a test yourself

If you want to try the matrix type described above, WhatIQ's test has 24 picture puzzles, takes about 15 minutes and gives a score on the standard IQ scale with the range it most likely sits in (whatiq.com, read November 4, 2026). It is an estimate from one kind of question, not a clinical assessment. If you need a score for a school or a clinic, see a licensed psychologist; our guide to where to take an IQ test lists the options.

Frequently asked questions

What kind of questions are on an IQ test?

Most IQ tests use some mix of matrix and pattern puzzles, number series, balance problems, verbal questions about words and ideas, spatial puzzles, working memory tasks and timed speed tasks. A full clinical battery such as the WAIS-5 samples five ability areas (Winter and colleagues, 2024). A test like Raven's uses only pattern puzzles, which the APA's 1995 report calls a single type of item.

How many questions are on an IQ test?

It depends on the test. Clinical batteries are counted in subtests: the WAIS-5 uses seven subtests for its full scale IQ, according to Winter and colleagues (2024), and the Stanford-Binet 5 has 10 (PRO-ED, read November 4, 2026). The standard form of Raven's Progressive Matrices has 60 items, according to Langener and colleagues (2021). Mensa's online challenge has 35 puzzles in 25 minutes (Mensa International, read November 4, 2026).

How can I pass an IQ test?

There is no pass mark on an IQ test. It places your score against the scores of a norm group, and the score comes with a range. Studying the answers to real items is not possible, because those items are kept secure. Retaking cognitive tests lifts scores by about a quarter of a standard deviation on average (Hausknecht and colleagues, 2007), and the gains were larger when the same form was repeated.

Do IQ tests measure creativity?

Not directly. The APA's 1995 report names creativity among the forms of intelligence standardized tests do not sample. Research finds a modest link: Gerwig and colleagues' 2021 meta-analysis put the correlation between intelligence and divergent-thinking test scores at r = .25, so a high IQ score says little about how a person does on those tests.

Are the IQ test questions online real?

Treat questions you find online as practice material at best, because the real items are kept secure. The APA Ethics Code asks psychologists to protect the security of test questions, and publisher rules let only qualified professionals buy tests such as the WAIS-5, the Stanford-Binet 5 and Raven's. The examples on this page are practice material written to show the format, and Mensa says its own online challenge is for entertainment only.

Are IQ test questions timed?

Some are and some are not. Processing speed tasks are timed by design, and Pearson's WISC-V FAQ says Block Design keeps its time bonuses for speed. For the Matrix Reasoning subtest of the WAIS-5, Pearson uses a 30-second guideline rather than a strict limit (WAIS-5 FAQ, read November 4, 2026). The APA's 1995 report calls Raven's untimed, so the answer depends on the test you take.