The problem solving test has three meanings and one dispute

0 The short answer

A problem solving test is three different things under one name: a short timed screen that employers buy, a complex problem solving task from research and from the 2012 PISA study, and a puzzle or case question at an interview. A meta-analysis of 47 studies found that complex problem solving tasks correlate .433 on average with intelligence (Stadler and colleagues, 2015), researchers still dispute whether the tasks add anything beyond reasoning, the employer screens are short timed tests whose listed sections point to reasoning and acquired knowledge, and brainteaser questions at interviews lack evidence for validity. ACIS has no interactive problem solving task, and it can show the reasoning part of your own profile on adult norms: Quick costs 15 dollars and takes about 45 minutes, Optimized 30 dollars and about 110 minutes, and Full Scale 50 dollars and about 175 minutes (prices read October 29, 2026). The decisive limit is that an online, unsupervised ACIS score is not an employer test, does not prepare you for one and does not predict how you will do on one.

1 What does "problem solving test" mean?

The phrase names three unrelated products that share a label: employer screening tests sold as "problem solving tests", the research and PISA tasks called complex problem solving, and the puzzles or case questions that some interviewers ask. Each one answers a different question and has a different body of evidence behind it, which is why a single search for "problem solving test" returns a vendor page, a journal article and a list of interview puzzles side by side.

The first use is commercial. Vendors of hiring software sell short timed tests under the name, and the person who sits one is a job applicant. The second use is scientific. Psychologists built computer scenarios in which a person has to explore an unfamiliar system, work out how its parts affect each other and then steer it toward a goal, and they call the skill complex problem solving. A version of those tasks went into the 2012 round of the OECD's PISA study of 15-year-olds. The third use is conversational: an interviewer asks a logic puzzle, an estimation question or a business case and watches how the candidate reasons aloud.

Three neighbouring topics have their own pages and are not repeated here. The page on the logic IQ test covers deduction and the formats of logic tests, the page on the Cognitive Reflection Test covers the three item puzzle that psychologists use to study intuition, and the page on IQ and creativity covers the question of whether intelligence makes people inventive. This page covers what is left once those are set aside, which is the meaning of the label itself and the dispute about the research tasks.

2 What do employer problem solving tests actually measure?

Employer problem solving tests are short, timed tests whose listed sections point to reasoning with unfamiliar material and some acquired numerical knowledge, and the vendor pages opened for this article name the competencies without publishing what the questions contain. Two vendor pages, both read October 29, 2026, show the market. TestGorilla lists its Problem Solving test as 12 questions drawn from a bank of 97, with a time of 9 minutes and intermediate difficulty, and classifies it as a cognitive ability test; the page does not state the topics or how a score is formed. Mettl lists a problem solving skills assessment of 30 questions in 30 minutes, with three sections called analytical ability and problem-solving, numerical ability and critical thinking, and describes the competence as the ability to outline a problem, break it down, develop a solution and assess its effectiveness. These are product facts from the sellers, they are not linked here on purpose, and neither page is evidence about intelligence.

What do such tests measure by the research evidence? The Cattell-Horn-Carroll (CHC) model, laid out on the page about the CHC model, separates fluid reasoning (Gf), which is solving unfamiliar problems and finding rules, from quantitative knowledge and reasoning (Gq), which draws on learned mathematics, and from comprehension and knowledge (Gc), which draws on language and learned facts. A short screen that mixes number, logic and verbal material samples Gf and Gq with some Gc, and the mix would explain why two tests with the same name can disagree about the same person. The sibling pages on the inductive reasoning test and the deductive reasoning test each describe one of the Gf families these screens borrow from, and the page on abstract reasoning tests shows how much of general intelligence one such test captures.

Does a short employer screen predict job performance? The strongest evidence is about general mental ability tests as a class, not about any vendor's product. Schmidt and Hunter's review of 85 years of research (Psychological Bulletin, 1998) summarized the validity of 19 selection procedures and reported that the combinations with the highest multivariate validity for job performance were general mental ability plus a work sample test (mean validity of .63), general mental ability plus an integrity test (.65) and general mental ability plus a structured interview (.63). Sackett and colleagues revisited those estimates (Journal of Applied Psychology, 2022) and concluded that validity had been overestimated through overcorrection for restriction of range: most of the same procedures stayed high in rank, but their mean validity estimates fell by .10 to .20, and structured interviews emerged as the top-ranked procedure. The page on IQ and job performance explains the workplace literature in more detail, and the page on the pre-employment cognitive test describes the best known employer screens.

Those are group results about test types. They say nothing about a particular 9 minute test. The TestGorilla page reports statistics of its own, which this page does not repeat or evaluate, and neither vendor page offers independent validation of its questions. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014), open to the public on testingstandards.net, make the same point in Standard 11.1: before an employment test is implemented, a clear statement of the intended interpretations of scores for specified uses should be made, and the validation effort should show how well that interpretation holds for all relevant subgroups.

3 What is complex problem solving in research?

Complex problem solving is the name researchers give to the skill of exploring an unfamiliar, changing system, learning how its parts interact and steering it toward a goal, and it grew out of a dissatisfaction with what ordinary intelligence tests ask. Dörner and Funke, in a paper written for Frontiers in Psychology (2017), trace the field back to computer simulations. One famous early study, Lohhausen, had students act as mayor of a simulated town over a simulated ten years, in a scenario with more than 2,000 variables. The authors report that conventional IQ scores did not predict performance in it, and that staying calm under pressure and switching between analytic and holistic thinking did.

The same paper lists five attributes that complex problems share, citing Funke's earlier work. Complexity is usually reflected in the number of variables. Connectivity means that the variables depend on each other. Dynamics means that the system changes over time. Intransparency means that some or all variables and their values are hidden from the solver. Polytely means that the solver has several goals that can conflict. A task with all five resembles the problems a manager or planner meets, and it is also hard to score: a single 2,000 variable town yields essentially one performance per person, which Wüstenberg and colleagues criticized as one item testing, according to the account of Süß and Kretzschmar.

Psychologists therefore shrank the systems. Dörner and Funke describe how minimal complex systems, introduced by Greiff and Funke in 2009, were meant to find the lower bound of complexity, and how small scenarios such as MicroDYN, in which a person has about three minutes to explore a system, replaced the 2,000 variable towns so that one session could contain many short items. In MicroDYN style tasks the person first varies inputs to discover how they affect outputs, and then sets the inputs to reach target values. Dörner and Funke note that such tasks lean heavily on a strategy called VOTAT, "vary one thing at a time", which works on simple systems.

Here is an original illustration, written for this page and not taken from any test. Suppose a panel has three dials called A, B and C and two gauges called X and Y, and you may try different dial settings before you must set the gauges to target values. If you move all three dials at once and both gauges change, you learn almost nothing, but if you move one dial at a time you can see that A moves X, that B moves Y and that C does nothing. The skill has two parts: learning the structure by exploring it, and then using the structure to reach the goal. Wüstenberg and colleagues, whose study appears later on this page, found that rule knowledge and rule application could be told apart as two dimensions of performance.

Whether such short tasks are still complex problems is itself contested. Dörner and Funke argue that small, linear and stable systems should not count as complex problem solving, and they quote Funke's warning that "Minimal complex systems run the danger of becoming minimal valid systems." That warning comes from one of the researchers who founded the field, and it frames the dispute in the section on the evidence below.

4 How did PISA 2012 test problem solving?

PISA 2012 tested problem solving by computer, with interactive scenarios in which a 15-year-old had to explore a system to find out what was missing, and its results correlate strongly with mathematics, reading and science without being the same thing. The details here come from the national report for England by Wheater and colleagues at the National Foundation for Educational Research, a government published report (December 2013, revised April 2014), which draws on the international report (OECD, 2014, Volume V). The OECD website refused automated access when this article was prepared on October 29, 2026, so the international report itself was not opened, and every PISA figure below is stated as the national report states it.

The national report gives PISA's definition of problem solving competence as a person's capacity to engage in cognitive processing to understand and resolve problem situations where a method of solution is not immediately obvious, together with the willingness to engage with such situations. It adds that the problem solving questions were intended to be unrelated to specific areas of the curriculum, so the assessment focuses on general reasoning ability, skills in approaching problems and the willingness to do so. Of the 65 countries in PISA 2012, 44 took the computer based problem solving assessment.

The framework has two aspects. The first is the nature of the situation. In a static problem, the information needed is available at the outset. In an interactive problem, only part of it is available, and other crucial elements have to be uncovered by exploring the situation. The second aspect is the cognitive process, in four kinds: exploring and understanding the information provided, representing and formulating (building representations and hypotheses about the relevant factors), planning and executing (setting goals and carrying out steps), and monitoring and reflecting (tracking progress, reacting to feedback and reflecting on the strategy). The international report, according to the national one, classifies the first two as knowledge acquisition tasks and the last two as knowledge utilization tasks.

Joachim Funke, a researcher in this field, summarized the assessment on his website (read October 29, 2026). He writes that about 85,000 students from 44 countries took the computer based assessment, that the 16 units were grouped into four clusters of 20 minutes with each student receiving one or two, and that most interactive units belonged to two classes from the research literature, MicroDYN systems and finite-state automata, with a single resource allocation exception. Those are the same minimal complex systems described in the previous section, which is why the research and PISA uses of the name are linked.

How strongly does PISA problem solving relate to the school subjects? Table 8.6 of the national report gives OECD average correlations of .81 with mathematics, .75 with reading and .78 with science. Among the subjects themselves the correlations are higher: .85 between mathematics and reading, .90 between mathematics and science and .88 between reading and science. The report concludes that pupils who do well in problem solving are likely to do well in the other subjects, that mathematics is the most strongly correlated and reading the least, and that the correlations with problem solving are weaker than those among the three subjects. In our arithmetic, a correlation of .81 squared is about .66, so roughly a third of the variation in problem solving scores is not shared with mathematics, and that leftover third also contains measurement error. PISA tested 15-year-olds and reports results for groups and countries, so it is a source about how the tasks are built and not a test an adult can sit.

5 What do the studies that find an added value report?

Several studies found that complex problem solving tasks carry information that a reasoning test does not, but the added value is described as modest and depends on how intelligence is measured. Three research groups, working mostly with university students, supply the evidence for the claim that complex problem solving is "more than reasoning".

Wüstenberg, Greiff and Funke published the best known version of the claim in Intelligence (2012). According to the abstract, 222 university students completed MicroDYN, a computer based complex problem solving test, and Raven's Advanced Progressive Matrices, a test of reasoning. A model in which rule knowledge and rule application were separate dimensions fit the data best. Reasoning predicted rule application only indirectly, through its influence on rule knowledge, which the authors read as showing that learning during exploration is a prerequisite for controlling a system. Complex problem solving then explained variance in school grade point average even beyond reasoning, and the authors concluded that it measures aspects of academic performance that reasoning does not and should be considered when predicting real life criteria. The abstract does not state how much additional variance it explained, and the full text was not available for this article, so the size of the gain is not reported here.

Kretzschmar, Neubert, Wüstenberg and Greiff tried to correct the weaknesses of earlier designs in a 2016 study in Intelligence. In 227 German university students, they paired a comprehensive intelligence measure (the Berlin Intelligence Structure test) with two complex problem solving tools, MicroDYN and MicroFIN, and used final school grades as an outside criterion. In the abstract they report that complex problem solving represents unique variance that is not accounted for by established intelligence constructs. They also report that the incremental validity for grades appeared only when intelligence was measured by a commonly used narrow operationalization, figural reasoning, and not when a broad operationalization was used. That condition is the key to the whole dispute: the added value shrinks when the intelligence measure is broader.

A third finding concerns the tasks, not the people. Greiff and colleagues compared one classical microworld, the Tailorshop, with three newer minimal complex systems tests, MicroDYN, Genetics Lab and MicroFIN, and defended the comparison in a 2017 rejoinder in the Journal of Intelligence. They report that the newer tests converged with each other more than with the Tailorshop, even after reasoning was controlled for, which they take as empirically separating the two approaches. Two commentators had raised concerns about how the study was conducted and analyzed, and the rejoinder says the authors respectfully disagree with some of them. For a reader, the practical consequence is that "complex problem solving score" is not one thing: a score from a small system of a few variables and a score from a large simulated business do not rank people the same way.

6 What do the studies that find little added value report?

The studies on the other side find that complex problem solving overlaps intelligence heavily, that reasoning is the ability that predicts it best, and that prior knowledge of the system also matters, so the claim of a unique ability is not supported. The largest single piece of evidence is the meta-analysis by Stadler, Becker, Gödker, Leutner and Greiff, published in Intelligence (2015). It examined 47 studies containing 60 independent samples and a total of 13,740 participants and found a substantial correlation with an average effect size of .433. It notes that over almost four decades empirical studies produced results in support of both arguments, that the two constructs had been hypothesized to be everything from completely separate to identical, and that there were no significant correlation differences by how intelligence was operationalized. How complex problem solving was measured did moderate the correlation, and the most recent approach produced the strongest associations. Süß and Kretzschmar, in a 2018 article in Frontiers in Psychology, describe the same meta-analysis as reporting a correlation of .47 with reasoning, .58 for minimal complex systems and .34 for realistic systems. In our reading, the approach with the cleanest measurement is also the one closest to reasoning.

Süß and Kretzschmar then ran two studies of their own with realistic microworlds. In Study 1, 137 students ran a simulated shirt factory (Tailorshop). In Study 2, 152 students ran a simulated forestry system. Reasoning was the only intelligence construct that reliably predicted performance, prior knowledge contributed as well, and in Study 2 working memory added nothing significant beyond reasoning (a change in explained variance of .02). They concluded that cognitive abilities and prior knowledge are substantial predictors of complex problem solving and that the evidence does not support treating it as a unique ability construct. Kretzschmar is an author of both this paper and the 2016 paper in the previous section, so the two are best read as answers to different questions: whether the tasks contain variance that a figural reasoning test misses, and whether that variance is a separate ability.

A third critique is about validity of content, not of correlation. Dörner and Funke argue that small, linear and stable systems should not count as complex problem solving, that the strategy of varying one thing at a time solves them, and that a business manager facing real problems would not solve problems that way. They describe complex problem solving as a bundle of competencies similar to intelligence but going beyond IQ, and they say the environments are not complex intelligence tests, because the process, including errors and detours, matters as well as the outcome.

Where does that leave a reader? Our summary, and not any author's: the research tasks are real, partly distinct from a matrix test, and strongly related to reasoning; whether the remainder deserves the name of an ability, or reflects learning a specific system and the knowledge a person brings to it, is not settled. That is a more modest statement than a vendor page for a problem solving skills assessment would make, and it is the one the evidence supports.

7 What do logic and case questions at interviews measure?

A puzzle or case question at an interview samples how a person handles an unfamiliar question under observation, and the evidence that it predicts job performance is weak for brainteasers and strongest for structured interviews. The interview use of the name is the least studied of the three. The one peer reviewed source opened for it is Highhouse, Nye and Zhang's paper in Applied Psychology (2019). Its abstract describes brainteaser interview questions as an example of aggressive interviewer behaviour that lacks evidence for validity and is unsettling to job applicants. The paper's own data are about the interviewers, not about the validity of the questions: in a representative sample of 736 working adults, who were shown a list of traditional, behavioural and brainteaser questions, narcissism and sadism explained the likelihood of choosing brainteasers, and a second study found that perspective-taking partially mediated the link between those traits and how helpful or abusive such questions were judged to be. The statement about validity is the authors' framing, and no validity study of brainteasers was opened for this article.

What the research does support is structure. Sackett and colleagues, in the 2022 revision described earlier, found structured interviews to be the top-ranked selection procedure after correcting their estimates, and Schmidt and Hunter's 1998 review reported a mean validity of .63 for general mental ability combined with a structured interview. A structured interview asks every candidate the same job related questions and scores them against a rubric, which is the opposite of an improvised puzzle. For case questions, such as a market sizing exercise or a business scenario, no validity study was found in the sources opened. A case question can be a reasonable sample of work that the job really contains, and Standard 11.3 says that when test content is the main source of validity evidence, a close link between the content and the job requirements should be demonstrated; an arbitrary puzzle has no such link.

For the person being interviewed, the useful facts are modest. A single puzzle answered badly is weak evidence about reasoning, because one task is a very small sample of a broad ability. Standard 11.4 says that when test scores and non test information are combined in a decision, the role of each should be explained and each inference supported by validity evidence, which means an interviewer who reads one puzzle as a measure of intelligence is drawing an inference that the sources opened do not support. The sibling page on the IQ test for a job interview covers the separate question of who sees a score. The page on critical thinking and intelligence explains why a puzzle about judgment and a test of capacity answer different questions, and the sibling page on the critical thinking test covers the employer tests that use that name.

8 Can you practice for a problem solving test, and are free versions worth taking?

Practice raises scores on cognitive ability tests by a measurable amount, more when the practice is coached and when the same form is repeated, so a higher second score does not by itself mean a higher ability. Hausknecht, Halpert, Di Paolo and Moriarty Gerrard summarized 50 studies in the Journal of Applied Psychology (2007). From 107 samples and 134,436 participants, they found an adjusted overall effect size of .26 for practice effects on tests of cognitive ability. Effects were larger when practice was accompanied by test coaching and when identical forms were used. The authors note that as many as 25 to 50 percent of applicants in organizational and educational settings are retested, and that more research is needed on how retesting affects the inferences drawn from scores.

Three consequences follow for a candidate who sees "problem solving test with answers" or "problem solving test questions" in a search. First, this page prints no questions and no answers from any test, because the questions belong to their publishers and a question seen in advance changes what it measures. Second, the format is the transferable knowledge: employer screens are short and timed, the sections can be analytical, numerical or about critical thinking (the Mettl page), and the time per question is brief, 9 minutes for 12 questions in one product and 30 minutes for 30 in another. Third, the retest research says that familiarity with the format can raise a score without changing the ability behind it, and that is a reason to treat a second attempt as a different measurement.

Many searchers want a free version, and a free problem solving quiz is worth taking if the goal is to see the format and the pace. It is worth reading as a measurement of ability only if its page states who the norm group was and what the score means, and a page that states neither supports a feel for the format and not a score. The page on the free IQ test describes what ACIS offers at no cost, and the logic page cited earlier explains in more detail what practice does to a logic score. Practice test sellers are not cited or linked here because their materials are written to sell preparation, not to describe the evidence.

9 Which problem solving test should an adult take?

The right choice depends on what you want to know: a clinician's full battery answers the broadest question, an employer screen answers only what that employer asks, and a normed online battery shows the reasoning part of a profile at a lower price but without a supervisor. The table compares the real options for an adult. Times and prices come from pages opened on October 29, 2026, they can change, and a cell says "not stated" where the page opened for this article does not say.

A few reading notes help. The row for the research task is there because it is the meaning of the name that no ordinary buyer can purchase, so it shows where the words come from. The employer rows describe tests that the employer chooses, which is the reason this page speaks to the candidate about understanding a profile and not about gaming a screen. Readers choosing between ACIS and other normed online batteries can compare them on the page about the best online IQ tests, and the page on where to take an IQ test lists the supervised and unsupervised routes together. The page on IQ tests for adults describes the broader choice.

10 Which ACIS subtests sample the reasoning that problem solving tests touch?

ACIS has no interactive problem solving task, no system to explore and no feedback loop to learn from, and what it offers instead is a set of reasoning subtests that sample rule discovery, deduction and quantitative reasoning in problems that are presented to you whole. This is the honest scope of what an ACIS profile can show. In the realistic microworld studies of Süß and Kretzschmar (2018), reasoning was the only intelligence construct that reliably predicted performance on complex problem solving tasks, so the reasoning part of your profile is relevant to the question, but it is not the exploring part and it is not a score on any employer test.

The fluid reasoning index (FRI) is built from five subtests, and each has its own page that describes the task. Matrix Reasoning measures rule discovery in unfamiliar visual patterns and is closest to fluid reasoning (Gf). Figure Weights measures relational balance reasoning, which the page places between Gf and quantitative knowledge and reasoning (Gq). Visual Number Series measures sequential rule discovery, also between Gf and Gq, and the page says the core demand is usually induction. Logic Grid measures controlled deduction under multiple constraints, and Complex Relations measures relational integration, the ability to keep several relations organized within one problem. The quantitative reasoning index (QRI) is built from Arithmetic, which the page describes as mental quantitative problem solving that is memory loaded and presented as audio, and Mathematical Achievement, which measures acquired quantitative knowledge and the application of learned procedures and so depends partly on instruction. The same CHC model explains the broad abilities behind those names, and the page on fluid intelligence tests explains what Gf tasks have in common.

Two practical points follow. The Quick form contains only two of the five fluid reasoning subtests and no quantitative reasoning subtest, so it gives a taste of the reasoning part, and a full fluid reasoning index needs all five subtests. The Custom Test lets a buyer pick subtests at 5 dollars each and says that any domain completed returns its full index score, so the five fluid reasoning subtests would cost 25 dollars and the two quantitative reasoning subtests 10 dollars (our arithmetic on the listed price). The free trial of the Full Scale form needs no card and is a way to try the platform, and the reasoning subtests described here are in the paid forms.

11 How should you read a reasoning score?

A reasoning score is an estimate with a band around it, it means something only against a described reference group, and a single subtest is less precise than the index or the total built from several. The scales are the same ones used across IQ testing: a scaled subtest score from 1 to 19 with a mean of 10 and a standard deviation of 3, and an index or Full Scale IQ with a mean of 100 and a standard deviation of 15. The page on why a standard deviation of 15 is used explains the scale, and the page on how IQ scores are normed explains why a raw count is compared with a group of adults. A percentile says what share of the reference group scored lower, and the page on the difference between an IQ score and a percentile shows why the two cannot be averaged.

Precision matters more for a part than for a whole. Standard 2.3 of the Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) asks that estimates of reliability and precision be reported for each total score, subscore or combination of scores that is to be interpreted, and its comment warns that total scores can be consistent while subscores have unacceptably low reliability. As a general principle and not an ACIS figure, a difference between two index scores is less precise than either one, which is why a gap between a fluid reasoning index and a quantitative reasoning index should be read as a prompt for a closer look and not as a verdict. The page on index score differences covers how large gaps between indices can be, and the page on reliability and validity explains the terms. The subtest pages add a simple reading rule for this pair: Mathematical Achievement depends partly on instruction, so a gap with a fluid reasoning score can reflect schooling and practice as much as reasoning.

The conditions of the session matter too. Standard 6.3 says unmonitored testing environments should meet standardized administration conditions, and otherwise the report on scores should note that those conditions were not guaranteed, and Standard 6.4 asks for reasonable comfort with minimal distractions. An ACIS session is online and unsupervised, so those conditions are yours to provide: a quiet room, a device with a screen you can read comfortably, and working sound, because Arithmetic presents its items as audio. Breaks are allowed, progress is saved between subtests and a buyer has 30 days to finish, so a long form does not need to be taken in one sitting. Keep the session free of interruptions, and read the result with the knowledge that no one verified those conditions.

12 Where ACIS sits, and what a buyer should do

ACIS is one commercial way to see the reasoning part of your own profile on adult norms, it is sold by the publisher of this page, and its limits are as important as its features: it is online, unsupervised, not clinical and in English only, and it has no interactive problem solving task. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014), published jointly by the American Educational Research Association, the American Psychological Association and the National Council on Measurement in Education, bear on how any score from this page should be read. Standard 1.1 asks that the developer set out how scores are intended to be interpreted and used, notes that no test permits interpretations that are valid for all purposes, and says each recommended interpretation for a given use requires validation. Standard 1.3 asks that, when validity for a likely interpretation has not been evaluated, potential users be strongly cautioned against it. Standard 11.1 asks for a clear statement of intended interpretations for employment uses and validation for them. Together they say what this page says: an ACIS score describes a profile and is not validated as an employer test or as a stand-in for one.

The site states the limit itself. The research page says that ACIS "is not intended for clinical diagnosis, hiring decisions, accommodation determinations or admission to high IQ societies" (read October 29, 2026). That is why this page never presents ACIS as preparation for an employer problem solving test, as a way to predict a result on one, or as something to hand to an employer. If you have a test coming, the useful preparation is the format: its length, its pace and the kinds of material, which the vendor's own materials describe. The useful use of ACIS is separate, and it is self knowledge: to see how your fluid reasoning and quantitative reasoning compare with adults, and whether they are a strength or a valley in your profile. The page on the technical manual is where the documentation for the scores is kept, and it is the place to read before leaning on any one number.

Here is what to do. If your question is clinical, or concerns a diagnosis or an accommodation, start with a licensed psychologist. If you want to see the reasoning part of your profile, the Quick form at 15 dollars (about 45 minutes) is the lowest cost route and gives Matrix Reasoning and Figure Weights, the Optimized form at 30 dollars (about 110 minutes) adds Logic Grid, and the Full Scale form at 50 dollars (about 175 minutes) has all five fluid reasoning subtests and both quantitative ones. Prices were read on October 29, 2026 and can change. It is a one time payment with no subscription, and the refund wording 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." The example reports on the site show what the report looks like before you buy.

13 Sources Behind This Page

Every figure above is traceable to one of the following, and each is linked at the point where it is used. The two vendor pages named in the text (TestGorilla and Mettl, read October 29, 2026) are market facts, are not linked on purpose and are not evidence about intelligence. The one calculation in the PISA section (a correlation of .81 squared) and the 25 and 10 dollar Custom Test totals are our own arithmetic on published numbers and are labeled as such. The PISA figures are taken from the national report for England, because the OECD website refused automated access on October 29, 2026. Prices and product details were read on October 29, 2026 and can change. Where a study is described from its abstract, the text says so.

• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing. Washington, DC: AERA, 2014 (Standards 1.1, 1.3, 2.3, 6.3, 6.4, 11.1, 11.3 and 11.4).
• Dörner D and Funke J. Complex Problem Solving: What It Is and What It Is Not. Frontiers in Psychology, 2017, volume 8, article 1153.
• Funke J. Results for PISA 2012 Problem Solving. Author website, April 1, 2014. Read October 29, 2026.
• Greiff S, Stadler M, Sonnleitner P, Wolff C and Martin R. Sometimes More is Too Much: A Rejoinder to the Commentaries on Greiff et al. (2015). Journal of Intelligence, 2017, volume 5, issue 1, article 6.
• Hausknecht J, Halpert J, Di Paolo N and Moriarty Gerrard M. Retesting in selection: A meta-analysis of coaching and practice effects for tests of cognitive ability. Journal of Applied Psychology, 2007, volume 92, issue 2, pages 373 to 385.
• Highhouse S, Nye C and Zhang D. Dark Motives and Elective Use of Brainteaser Interview Questions. Applied Psychology, 2019, volume 68, issue 2, pages 311 to 340.
• Kretzschmar A, Neubert J, Wüstenberg S and Greiff S. Construct validity of complex problem solving: A comprehensive view on different facets of intelligence and school grades. Intelligence, 2016, volume 54, pages 55 to 69 (abstract read).
• Sackett P, Zhang C, Berry C and Lievens F. Revisiting meta-analytic estimates of validity in personnel selection: Addressing systematic overcorrection for restriction of range. Journal of Applied Psychology, 2022, volume 107, issue 11, pages 2040 to 2068.
• Schmidt F and Hunter J. The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 1998, volume 124, issue 2, pages 262 to 274.
• Stadler M, Becker N, Gödker M, Leutner D and Greiff S. Complex problem solving and intelligence: A meta-analysis. Intelligence, 2015, volume 53, pages 92 to 101.
• Süß H and Kretzschmar A. Impact of Cognitive Abilities and Prior Knowledge on Complex Problem Solving Performance: Empirical Results and a Plea for Ecologically Valid Microworlds. Frontiers in Psychology, 2018, volume 9, article 626.
• Wheater R, Ager R, Burge B and Sizmur J, National Foundation for Educational Research. Achievement of 15-Year-Olds in England: PISA 2012 National Report. December 2013, revised April 2014 (chapter 8 and Table 8.6).
• Wüstenberg S, Greiff S and Funke J. Complex problem solving: More than reasoning? Intelligence, 2012, volume 40, issue 1, pages 1 to 14 (abstract read).

14 Frequently Asked Questions