Raven's Advanced Progressive Matrices and what free really means

0 The short answer

Raven's Advanced Progressive Matrices (APM) is a nonverbal reasoning test of 48 figure matrix items in two sets, Set I with 12 items and Set II with 36, that its publisher sells to qualified buyers as a paper test, so a free online "APM" or a "23 questions pdf" is not the same thing as a normed APM score. Set I is a short opener and Set II holds the 36 items that Arthur and Day (1994) call a power test of 40 to 60 minutes. Pearson lists a raw score and a percentile rank against a norm group, not an IQ, and the 23 item form belongs to the newer APM-III, which TalentLens sells to employers from an item bank. ACIS Matrix Reasoning is one concrete option for an adult who wants a normed matrix task (it is in every form: Quick at 15 dollars and about 45 minutes, Optimized at 30 dollars and about 110 minutes, Full Scale at 50 dollars and about 175 minutes, prices read October 29, 2026), it is not Raven's APM and uses none of its items, and the decisive limit is that an online, unsupervised, non clinical score describes reasoning on a task and is not a hiring, school or diagnostic result.

1 What is Raven's Advanced Progressive Matrices?

Raven's Advanced Progressive Matrices is the hardest of John Raven's matrix tests: a nonverbal task in which the person finds the missing piece of a figure pattern, built to tell apart people who are already near the top of the ability range. Pearson's own page for the test (read October 29, 2026) names John C. Raven as the author and describes the test as measuring "high-level observation skills, clear thinking ability, and intellectual capacity" and as a non-verbal estimate of abstract reasoning or fluid intelligence. It adds that, given untimed, the test differentiates high levels of ability with harder items in the top 20 percent of the population. That last phrase states the test's niche: the publisher places it at the high end of the range and not as a general screening test.

The record of its origin is thin in public sources, and this page uses only what could be opened. A 1936 paper by Lionel Penrose and John Raven, titled "A new series of perceptual tests: preliminary communication," appeared in the British Journal of Medical Psychology; the encyclopedia article on the matrices cites it for their origin, and the Crossref record confirms its title, venue and year, but this page did not open its text and takes nothing else from it. A Wellcome Collection archive record lists a work titled Progressive Matrices Sets I and II (1947) by J. C. Raven and G. A. Foulds, published privately at Crichton Royal in Dumfries and dated 1948 in the catalogue; the same record says Raven was Director of Psychological Research at Crichton Royal Hospital from 1944 to 1964. Sets I and II are the names researchers use today for the two sets of the Advanced test, so the late 1940s is the first dated appearance of those two sets that this page could confirm.

Later citations name other editions. Carpenter, Just and Shell, in their 1990 analysis in Psychological Review, describe the test they studied as the Raven Progressive Matrices Test (Raven, 1962), and a 2015 study of practice effects cites Raven and colleagues (1998) for the Set II items it used. Those two citations show which editions researchers lean on; they do not describe what changed between them, and this page does not guess.

The Advanced test is also not the only Raven product. The page on what Raven's 2 is and how it differs covers the current Raven's 2 line, and the page on culture-fair IQ tests covers the Culture Fair test. This page stays with the APM.

Two limits shape everything that follows. First, the items belong to the publisher, and this page neither shows nor describes any of them; it describes only the format. Second, a test can keep the same name while its norms, its form and its administration change. Raven's own summary of the norms across cultures and decades, in Cognitive Psychology (2000), says the data on norms are first summarized for different cultural, ethnic and socioeconomic groups and then for variation over time, and offers a possible explanation for the variation. Reading an APM score therefore always means asking which edition, which norm group and which date.

2 How many items does the APM have, and what are Set I and Set II?

The Advanced Progressive Matrices has 48 items split into Set I with 12 items and Set II with 36 items, and the 36 items of Set II are the scored core in the studies cited on this page. Chen and colleagues, in a 2017 study in Scientific Reports, write that the test is conventionally composed of 48 nonverbal picture items that can be divided into one set of 12 (Set I) and another of 36 (Set II). In their administration the 12 items of Set I were used to practice the instructions, and the 36 items of Set II then had to be completed within 30 minutes. Arthur and Day, in their 1994 paper, call the APM a 36-item power test with an administration time of 40 to 60 minutes, which is the same Set II seen from the scoring side.

The format is the same for every item. A matrix of figures is shown with one piece missing, and the person chooses the piece that completes it from eight candidate answers. Chen and colleagues describe picking the missing piece of a picture matrix out of eight candidates, and Hayes, Petrov and Sederberg, in a 2015 study in Intelligence, put it the same way: a Raven problem consists of a matrix and 8 response alternatives. Pearson's UK page (read October 29, 2026) says examinees identify the missing element in a series of figural patterns and that items increase in complexity, requiring stronger encoding and analysis skills. Nothing more about the items is needed to understand the test, and nothing more is given here.

Set I and Set II behave differently in data. Bors and Stokes, in their 1998 norming study of 506 first-year university students, found that Set II scores ranged from 6 to 35, with a mean of 22.17 and a standard deviation of 5.60, and that the first 12 items of Set II added little to the discriminative power of the test. Two readings follow, and both are our reading rather than theirs. The earliest items of Set II are plausibly easy enough for a university sample that they separate few people. And the ceiling is close: the best score in the group of 506 first-year students was 35 of 36, one point below the maximum.

The paper kit is simple. Pearson's US store lists the Exam Kit Sets 1 and 2 with two test booklets, answer documents, an answer key and manual sections, and lists an age range of 12 years 0 months to 16 years 0 months and 17 years and older (read October 29, 2026). A printed or screen copy of an item is an exposed item, and that matters for the free copies discussed later.

3 Is the Advanced Progressive Matrices timed or untimed?

The APM was designed as an untimed power test of 40 to 60 minutes, and the clock is a choice that changes what the score means: untimed it differentiates the top of the ability range, and timed it assesses what the publisher calls intellectual efficiency. Pearson's page (read October 29, 2026) lists the completion time as untimed for individual or group use, 40 to 60 minutes, says the untimed test differentiates high levels of ability in the top 20 percent of the population, and says the timed test assesses intellectual efficiency. Arthur and Day call it a power test; in that tradition the question is how hard a problem a person can solve and not how many easy ones fit into a fixed time.

Researchers have shortened the clock often, because 40 to 60 minutes is long. Hamel and Schmittmann, in their 2006 paper, say the APM can be administered as a 30-minute or 40-minute timed test, or with one of two developed short versions, and they compare a 20-minute timed version with the untimed test in first-year psychology students. Their conclusion is that the 20-minute version proves an adequate predictor of the untimed score. Adequate prediction is a statement about how people rank, not a statement that the two scores are the same number: a person cut off at 20 minutes may earn fewer points than the same person given the full time, and a norm table built from one condition should not be applied to the other.

That is exactly the point of standardized administration. Guideline 8 of the American Psychological Association's Guidelines for Psychological Assessment and Evaluation (2020) asks psychologists to be aware of the need for test selection, scoring and administration to reflect the appropriate normative comparison, situational influences, effort and standardized administration, and the guideline's rationale says that adhering to standardized procedures supports valid interpretation of norms. For an APM result, standardized means: the stated edition, the stated time condition, the stated instructions.

Two practical consequences follow for a reader who meets a time figure in the wild.

• A "20-minute APM" or "30-minute APM" is a different condition from the 40 to 60 minute untimed administration, even if the items are drawn from the same set.
• A result reported with no time condition, no edition and no norm group cannot be compared with the publisher's percentile table.

If your question is how the clock changes reasoning in general, the page on untimed IQ tests covers it; if your question is how fast you work on a simple task, a speed measure such as the one on the processing speed test page answers it, and the two questions are different ones. The APM is not designed to answer the second.

4 What does the Advanced Progressive Matrices measure?

The APM samples fluid reasoning (Gf in the CHC model) by asking for the rule behind an unfamiliar figure pattern, and the research shows that it also leans on working memory and on visual strategy, so it is a strong indicator of reasoning and not a pure one. Pearson describes it as a non-verbal estimate of abstract reasoning or fluid intelligence (page read October 29, 2026). In the CHC taxonomy that behaviour is fluid reasoning, and its narrow ability is induction: finding the relation that a set of figures shares and applying it to a new case. The page on the CHC model names the broad and narrow abilities, and the page on fluid intelligence tests covers what Gf tasks do and do not capture.

The best process evidence comes from Carpenter, Just and Shell. Their Psychological Review paper analyzed the Raven test with verbal protocols, eye-fixation patterns and errors, and expressed the theory as a pair of computer simulation models. They report that the processing common to all subjects is an incremental, reiterative strategy for encoding and inducing the regularities in each problem, and that the processes which distinguish among individuals are primarily the ability to induce abstract relations and the ability to dynamically manage a large set of problem-solving goals in working memory. That second ability is why a matrix score cannot be read as reasoning alone: holding several candidate rules in mind at once is a working memory demand as well as a reasoning one.

Whether the test measures one thing or two is still argued. Chen and colleagues describe the APM as embedding two types of reasoning, visuospatial and verbal-analytic, and report distinct resting-state neural correlates for the two. Bors and Stokes ran exploratory and confirmatory factor analyses in their student sample and found that the two-factor solution of Dillon and colleagues was not confirmed, and that a single factor best represented performance on Set II. Those results come from different samples and methods, and both can hold for their own data. For a reader, the practical message is modest: the APM total score is a defensible single number for reasoning on figure matrices, and the claim that it splits cleanly into a spatial score and a verbal score is not settled. The page on spatial reasoning tests covers the separate construct that hiring tests under that name aim at.

The APM is also described as a measure of general ability. Arthur and Day open their paper by calling it a popular measure of higher order general cognitive ability (g), and say that its use is partly attributable to its apparent low level of culture loading, and Hamel and Schmittmann call it a well-known measure of higher order general mental ability. Describing a test as a measure of g is a statement about where its scores sit among other tests, and it does not say the test is g. The page on what g means in IQ testing explains that distinction, and the page on pattern recognition tests and Raven's covers how much of a Raven-style score is general ability and how much is specific to the task.

The word nonverbal needs the same care. The TalentLens FAQ for the APM-III (read October 29, 2026) says the test needs only minimal verbal instructions and no written or spoken response, so that the effect of language skills and cultural background on performance is reduced to a minimum. Reduced is the publisher's word and the right one: the instructions are still given in a language. The page on nonverbal IQ tests explains what nonverbal does and does not buy. Where employers use tests of this kind under the name inductive reasoning, the page on inductive reasoning tests is the one that covers the hiring side.

5 Who uses the APM, and who is allowed to buy it?

The APM is used where a standard matrix test would be too easy for the group, in research on reasoning, in educational decisions about high ability, and in assessment for senior occupational roles, and its publisher lists it at qualification level B, as kits for qualified buyers. Pearson's US page (read October 29, 2026) lists the benefits as use with culturally diverse populations where English proficiency varies, identifying giftedness and predicting educational success, and its UK page says the test helps identify giftedness and support educational decision-making. The same US page lists every product at qualification level B and prices the Exam Kit Sets 1 and 2 (print) at 374.80 dollars, a pack of 50 answer sheets for Sets 1 and 2 at 82.50 dollars and the answer key at 45.70 dollars; the UK page lists the kit at 345.50 pounds. Those are prices for a buyer who is a professional or an institution. They are not the price of taking the test.

Research is the largest user in the open literature. Bors and Stokes normed the test on first-year university students, Hamel and Schmittmann used first-year psychology students, Chen and colleagues paired it with brain imaging, Hayes and colleagues used it to test whether practice gains reflect reasoning or strategy, and Carpenter and colleagues used it to model problem solving. All of these treat the APM as an instrument for studying reasoning in groups, which is a different job from producing a verdict on one person.

Occupational use is real and is the use that matters most to an adult who meets the test as a job applicant. The TalentLens FAQ says the APM-III can be used to assess a candidate's potential for the highest-level posts, such as recent graduates, potential leaders, managers and department heads, and Hogan's page for the APM-III lists four levels of best use: individual contributors, first-level managers, middle managers and executives. Hogan's page also says the test should not be used as the sole basis for making an employment decision, and says Hogan offers it under license from NCS Pearson Inc. (page read October 29, 2026). If you are an applicant, the test is the employer's instrument and you receive what the employer shares. The page on the SHL test covers the other large family of employer tests, and the page on diagrammatic reasoning tests covers diagram based tests that employers also use and that are a different product from the APM.

Questions about societies come next, because the APM has a place in the folklore of high IQ clubs. This page does not state which tests any society accepts, because that list is set by each society and changes. The page on the Mensa IQ test covers how adults estimate a qualifying score, and the page on high range IQ tests covers what tests aimed above the usual ceiling do and do not show. For an adult who wants a check on giftedness without a society, the page on gifted tests for adults lays out the options, and the page on where to take an IQ test covers the route through a psychologist.

6 How is an APM score normed and reported?

An APM score starts as a count of correct items out of 36 and means nothing until it is compared with a norm group, and the publisher reports a percentile rank against that group, not an IQ. Pearson's page (read October 29, 2026) says the test produces a single raw score as well as a percentile rank, compared with a norm group, and lists interpretation as summary and detailed percentiles; the page does not describe the norm group. The UK page uses the same words and likewise does not describe it. The store pages are product listings and the norm tables are not among their contents, which is a reason to be careful about any number that arrives without the publisher's documentation: a percentile is only as good as the group behind it.

The open literature gives one concrete norm sample. Bors and Stokes titled their study "Norms for First-Year University Students," and in the 506 students they tested, Set II scores ranged from 6 to 35 with a mean of 22.17 and a standard deviation of 5.60. Read plainly, a score of 22 in that group is an average first-year student's score, and the group is a sample of first-year students, not adults in general. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) say what a user should check. Standard 5.8 asks that norms refer to clearly described populations that include the individuals or groups with whom test users will ordinarily wish to compare their examinees, and Standard 5.9 asks that reports of norming studies specify the population sampled, the sampling procedures and participation rates, any weighting, the dates of testing and descriptive statistics. A student norm answers a student question.

Norms also age. Raven's own review in Cognitive Psychology summarizes variation in norms over time, which is the territory of the Flynn effect, and the rationale for Guideline 8 of the APA assessment guidelines says the agedness of the norms can be a particularly important consideration in test selection. The page on the Flynn effect covers why a table from one decade can flatter or penalize a person in another.

The newest APM family reports differently. The TalentLens FAQ for the APM-III (read October 29, 2026) says its percentile ranks rest on theta scores, estimated from which items a person answers correctly and how hard and how discriminating those items are, not on the number of correct answers. Theta scores run from minus 4 to plus 4 in the publisher's description, they are not shown in the reports, and they are converted to percentile ranks against reference groups; the FAQ says that for France only a managers norm was available at first, and that the highest percentile that can be obtained is the 99th. Because candidates see different sets of 23 items, a raw count would not be comparable across them, which is the reason for the method.

None of these publisher descriptions gives an IQ. A percentile can be moved onto the familiar scale with a mean of 100 and a standard deviation of 15 only if the norm group is the population you mean, and the arithmetic belongs to the person doing it. The pages on how IQ scores are normed, on why the standard deviation of 15 matters and on IQ scores versus percentiles explain the scales, and a percentile calculator does the conversion when you already know which scale you are on. Anyone who quotes "an APM IQ" without naming the norm group and the date has performed a conversion that the pages read for this guide do not show the publisher offering.

7 Which short forms of the APM exist?

Two 12-item short forms of Set II were published, by Arthur and Day in 1994 and by Bors and Stokes in 1998, and Bors and Stokes report that the two differ considerably in content yet do not differ in concurrent validity or predictive power. Arthur and Day's abstract states the motive plainly: the APM is a 36-item power test of 40 to 60 minutes, which is a drawback, so they developed a 12-item short form with psychometric properties similar to the long form and a substantially shorter administration time. Bors and Stokes extracted 12 items from the 36 and found their short form to possess acceptable psychometric properties, and they note that it differed considerably in content from the one devised by Arthur and Day while the two did not differ with respect to concurrent validity and predictive power.

Short forms are a trade, and the arithmetic of the trade is easy to see. With 12 items the raw score can take only 13 values, from 0 to 12, against 37 values from 0 to 36 on the full Set II, so a short form separates fewer levels of ability, and at the high end, where the APM is supposed to work, a short form reaches its ceiling sooner. That is our arithmetic and not a finding of either paper; the APA assessment guidelines list ceiling and floor effects among the reasons that normative data or decision rules may not be accurate. A short form is therefore a reasonable research instrument for ranking a group in limited time and a weaker basis for a fine statement about one person at the top.

There is a second warning inside the table. A short form is a different test, so it needs its own norms; a percentile from a 36-item table does not transfer to a 12-item sum. Anyone who scores a short set and then reads a long-form table is combining two instruments, and the result has no documented meaning.

8 What is the 23 item APM-III?

The APM-III is the item-banked, online version of the Advanced Progressive Matrices that Pearson's TalentLens sells for hiring and development, and its 23 items are a session drawn from a bank, so "the 23 questions" are an event and not a document. The TalentLens FAQ for the APM-III (a French language FAQ read October 29, 2026) says the test is the new item-bank version of the Raven matrices and that each administration presents a series of 23 items selected at random from a large pool, a random assignment that preserves the integrity of the items and prevents overexposure of content. The same FAQ says each session has 4 practice items before the 23 scored ones, a limit of 40 minutes and eight response options per item, and that every candidate starts with an item of equivalent difficulty selected from the bank.

The FAQ also explains where the number 23 comes from. The item bank grew out of an international short version of the APM, validated in France and called the APM-II, which is a fixed form of 23 items developed to shorten the long version; the new bank offers forms of 23 items whose psychometric properties are based on that short version, and the FAQ calls the two parallel versions. The publisher reports a correlation of 0.97 between the APM-III and the APM-II, which it describes as extremely high. That figure is the publisher's, from a marketing FAQ, and the technical manual that the FAQ says holds the validity studies was not available for this page. The pages read do not say whether the APM-II is related to either of the academic 12-item short forms, and this page does not assume that it is.

A paper form exists only for the older fixed version: the FAQ says there is no paper version of the APM-III itself, and that the 23-item fixed APM-II is available on request, with a CD-ROM version of that fixed form. A fixed form suits supervised sessions, the FAQ says, while the bank suits online use supervised or unsupervised. Hogan's page for the APM-III (read October 29, 2026) describes it as item-banked, language-free and designed for unsupervised settings.

Unsupervised does not mean unquestioned. The TalentLens FAQ says results from unsupervised sessions should be interpreted with caution unless it is certain that the test was completed without outside help, and that they should be confirmed by a supervised retest given to all candidates at the last stage of selection, or by other sources of information such as a structured interview or an assessment centre exercise that measures the same abilities. It recommends a desktop computer, a laptop or a tablet with a reliable connection and a screen large enough to read, and not a phone. For an applicant, those lines carry a practical message: the employer's own documentation expects to check an unsupervised result.

9 What is a free online "APM", and what is a "23 questions pdf"?

A free online "APM" is, unless it can show a norm group and a licence, a matrix quiz that uses the name, and a "23 questions pdf" cannot be an APM-III session, because the publisher says that session is drawn from an item bank and that its paper 23-item form is supplied on request. This page did not open or test any free APM site or pdf, names none, and links none, for two reasons. The items belong to the publisher, and a page that reproduces them gives away the content that the publisher builds item banks to protect. And the question a reader actually has, what does my result mean, does not depend on which free page is better; it depends on what the result is compared with.

The questions to put to any page that offers an "APM" are the ones the Standards put to every test. Does it name a norm group, with the sample, the sampling and the date of testing (Standard 5.9)? Does it say which edition and which time condition it follows? Does it report a count, a percentile or an IQ, and if an IQ, who made the conversion and from what table? If the answers are missing, the number is a count of items solved on a puzzle quiz, and the useful thing it gives you is a sense of the format, with the cost described in the next sections: an exposed item is less informative the second time.

A pdf adds a separate problem. A fixed list of questions with an answer key is a form that, once it circulates, anyone can rehearse. The publisher's design answer to that, in the FAQ, is the bank: no two candidates receive exactly the same series, and sharing screenshots between candidates is of little interest. A pdf is the opposite design. It also removes the other half of a test, the scoring system, since the APM-III scores items by accuracy, difficulty and discrimination rather than by a count. So a reader who wants to know what a normed matrix result is can use this simple test: if you were not told the norm group and the date, you were not given a normed result.

There are honest uses for a free matrix quiz: seeing the format, seeing how a time limit feels, deciding whether the task is something you want measured properly. The page on free IQ tests explains what a free, normed starting point looks like, and the page on the best online IQ tests compares the options by what they norm against and what they return.

10 Which options let an adult take a normed matrix task?

An adult has four real routes to a normed matrix result, a psychologist, a qualified buyer's paper APM, an employer's APM-III and a normed online battery, and they differ on who controls the session, who receives the score and what it is compared with. The table sets them next to each other, with what each publisher or provider states on pages read October 29, 2026.

Two lessons come out of the table. The first is that the pages read present the APM as a product for professionals and organizations and not for an individual who simply wants a score: the paper test is sold in kits at qualification level B and the APM-III is offered for hiring and development, so for an adult the live choice is between a clinician and a normed online battery. The page on IQ tests for adults lays out that choice by purpose, and the page on cognitive ability tests separates the employer tests from the personal ones. The second lesson is that supervision is a variable and not a virtue: the Standards say unmonitored testing environments, such as those on the Internet, should meet the standardized administration conditions or the report on scores should note that standardized conditions were not guaranteed, which is the honest statement for any online result, including ours.

11 How much do practice and retesting change a matrix score?

Practice raises scores on cognitive tests in general, and one eye-movement study of the APM suggests that a substantial share of the gain on a matrix test comes from a changed strategy and not from higher reasoning, so a first attempt and a rehearsed attempt do not mean the same thing. Hausknecht and colleagues, in their 2007 meta-analysis of 50 studies of practice effects for tests of cognitive ability, covering 107 samples and 134,436 participants, report an adjusted overall effect size of .26, with larger effects when practice was accompanied by coaching and when identical forms were used. Those studies cover cognitive tests in selection and education in general, not the APM alone, so the figure says that retesting matters and cannot be read as the gain on Raven's.

For the APM itself, Hayes, Petrov and Sederberg tracked eye movements as 35 participants solved matrix problems in two sessions about a week apart. The abstract reports that one-third of the variance of score gains could be attributed to test-taking strategy alone, as revealed by characteristic changes in eye-fixation patterns, and that when the strategic contaminant was partialled out the residual gains were no longer significant. In the full text, a group that received paper-and-pencil training on similar problems did not gain significantly more than a no-contact control group, so the mere repetition of the test was enough to raise scores, even with short forms. Two cautions apply: the sample was 35 university students, and the sessions were about a week apart. Even so, the finding is plain enough for a reader to use: a repeated matrix score may reflect a better plan of attack on the format.

That is a reason to treat an online score, especially a retake, as one measurement under stated conditions. It is also a reason for the practical conditions that a matrix task needs. The Standards say the testing environment should furnish reasonable comfort with minimal distractions to avoid construct-irrelevant variance (Standard 6.4); in practice, sit on a full-size screen in a quiet room with no notes or calculators, because the TalentLens FAQ recommends a screen large enough to read and not a phone, and because anything that changes the conditions changes what the score is compared with.

On reading the number itself, the sensible habit is to read the confidence band and not the point. A single task has measurement error, which is why a battery reports intervals and why one subtest is less precise than an index built from several; the page on reliability and validity explains how error and meaning are separated in a test report.

12 Where ACIS sits, and what a buyer should do

ACIS offers one option for an adult who wants a normed matrix task, the Matrix Reasoning subtest inside a battery, with adult norms and a stated confidence interval, and it carries limits that a reader should weigh as plainly as the strengths. Matrix Reasoning is in every form, Quick at 15 dollars (about 45 minutes), Optimized at 30 dollars (about 110 minutes) and Full Scale at 50 dollars (about 175 minutes), prices read October 29, 2026 and subject to change. The Matrix Reasoning page describes visual pattern problems that sample abstraction, logic and nonverbal reasoning. The page relates the task most closely to fluid reasoning (Gf). It feeds the Fluid Reasoning Index, and the technical manual describes a composite, the IRI, that "focuses on Induction through Visual Number Series, Matrix Reasoning, and Complex Relations"; the Fluid Reasoning Index page explains the index and why factor analyses dispute it. The documentation is in the technical manual.

The limits are these. The session is online and unsupervised. ACIS is not a clinical or diagnostic instrument, it is in English only, and, in its own words, it "is not intended for clinical diagnosis, hiring decisions, accommodation determinations or admission to high IQ societies." Matrix Reasoning is not part of the free trial, which is a separate set of tasks and needs no card, so a buyer who wants this task starts at Quick. The refund policy is: "Full refund within 5 days if the report does not deliver the features described at checkout or a technical issue prevents you from accessing your results." Breaks are allowed, progress is saved between subtests and a buyer has 30 days to finish.

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) say how to hold all of this. Standard 5.8 asks users to compare a person with a clearly described population, and Standard 5.9 asks that norming be documented by population, sampling, dates and statistics; the comment to Standard 6.3 says unmonitored testing environments should meet standardized administration conditions or the report should say they were not guaranteed. The APA's Guidelines for Psychological Assessment and Evaluation (2020), Guideline 8, add that normative comparison, situational influences, effort and standardized administration should shape how a score is read. For a reader of this guide, that means: treat a matrix score as a measurement of reasoning on one task under unsupervised conditions, read it with its interval and next to other domains, and do not use it for a decision that a clinician, an employer or a society makes under its own rules.

The next step depends on the reader. If you were asked by an employer to sit the APM-III, do it under the employer's conditions and use the employer's report. If you want a normed matrix task of your own, start at Quick or the Optimized form, and the example reports show what you receive before you pay. If you want a result a clinician or a society will accept, go to the clinician or the society.

13 Sources Behind This Page

Every figure on this page comes from one of the sources below, opened or confirmed on October 29, 2026; the publisher pages are dated, and no source here reproduces a test item.

• Raven, J. (2000). The Raven's Progressive Matrices: Change and Stability over Culture and Time. Cognitive Psychology, 41(1), 1 to 48. Abstract read via PubMed.
• Penrose, L. and Raven, J. (1936). A new series of perceptual tests: preliminary communication. British Journal of Medical Psychology, 16(2), 97 to 104. Record confirmed in Crossref; text not opened.
• Wellcome Collection. Raven, J. C., Progressive Matrices Sets I and II (1947). Archive record DGH1/6/18/157, Records of Crichton Royal Hospital. Read October 29, 2026.
• Arthur, W. and Day, D. (1994). Development of a Short form for the Raven Advanced Progressive Matrices Test. Educational and Psychological Measurement, 54(2), 394 to 403. Abstract read.
• Bors, D. and Stokes, T. (1998). Raven's Advanced Progressive Matrices: Norms for First-Year University Students and the Development of a Short Form. Educational and Psychological Measurement, 58(3), 382 to 398. Abstract read.
• Hamel, R. and Schmittmann, V. (2006). The 20-Minute Version as a Predictor of the Raven Advanced Progressive Matrices Test. Educational and Psychological Measurement, 66(6), 1039 to 1046. Abstract read.
• Carpenter, P., Just, M. and Shell, P. (1990). What one intelligence test measures: A theoretical account of the processing in the Raven Progressive Matrices Test. Psychological Review, 97(3), 404 to 431. Abstract read.
• Chen, Z., De Beuckelaer, A., Wang, X. and Liu, J. (2017). Distinct neural substrates of visuospatial and verbal-analytic reasoning as assessed by Raven's Advanced Progressive Matrices. Scientific Reports, 7(1). Open access full text read.
• Hayes, T., Petrov, A. and Sederberg, P. (2015). Do we really become smarter when our fluid-intelligence test scores improve? Intelligence, 48, 1 to 14. Open access full text read.
• Hausknecht, J., Halpert, J., Di Paolo, N. and Moriarty Gerrard, M. (2007). Retesting in selection: A meta-analysis of coaching and practice effects for tests of cognitive ability. Journal of Applied Psychology, 92(2), 373 to 385. Abstract read.
• Pearson Assessments. Raven's Advanced Progressive Matrices, US store page. Read October 29, 2026.
• Pearson Clinical UK. Raven's Advanced Progressive Matrices, UK store page. Read October 29, 2026.
• Pearson TalentLens France. Advanced Progressive Matrices APM-III, Foire aux Questions (FAQ in French, copyright line 2017). Read October 29, 2026.
• Hogan Assessments. Raven's APM-III. Read October 29, 2026.
• American Educational Research Association, American Psychological Association and National Council on Measurement in Education (2014). Standards for Educational and Psychological Testing. Standards 5.8, 5.9, 6.3 and 6.4 read in the open access edition.
• American Psychological Association (2020). APA Guidelines for Psychological Assessment and Evaluation. Approved by the APA Council of Representatives, March 2020 (Guideline 8).

14 Frequently Asked Questions