The deductive reasoning test: what must follow and what only feels true

0 The short answer

A deductive reasoning test asks which conclusion must be true if the stated information is true, and the version you sit is the one your employer's publisher chose. SHL's fact sheet for its Verify Deductive Reasoning test allows 20 minutes for a maximum of 18 multiple choice questions, and in the CHC model the ability is general sequential reasoning, a narrow part of fluid reasoning (Gf) that differs from induction. ACIS is one way to see your own deductive reasoning on adult norms, through its Logic Grid subtest in the Optimized form (30 dollars, about 110 minutes) or the Full Scale form (50 dollars, about 175 minutes), prices read October 29, 2026, but it is not that employer test and neither prepares for it nor predicts it. The decisive limit is that ACIS is online, unsupervised and not an employer or clinical instrument, and the decisive trap in deduction is that a believable conclusion feels correct whether or not it follows.

1 What Is a Deductive Reasoning Test?

A deductive reasoning test hands you statements and asks which conclusion must be true if those statements are true, which is a different job from asking which conclusion is likely, sensible or familiar. The SHL fact sheet describes the ability it measures as drawing logical conclusions from the information provided, and the same document adds that candidates evaluate arguments, analyze scenarios and complete scenarios when information is incomplete. People search for the same thing under other names, including "logical reasoning test", and publishers use the labels loosely, which is why this page reads the descriptions that publishers print rather than the names on a practice site.

The word that matters is "must". An argument is valid when its conclusion has to be true whenever its premises are true, and a valid argument can still have a false conclusion if a premise is false. Here is an original example, written for this page and not taken from any test or practice site (ours): "All bakers are early risers. Some early risers are cheerful. Therefore some bakers are cheerful." The conclusion sounds plausible, and the argument is invalid, because the two premises are consistent with a world in which every cheerful early riser is something other than a baker. A deductive test rewards the person who asks whether the premises could all hold while the conclusion fails, not the person who asks whether the conclusion sounds right.

Employers use such tests to compare applicants on a skill they say a role needs. The SHL sheet says the test is used for job candidates applying to entry and mid level positions that require deductive reasoning ability, and that this form of reasoning is commonly required to support work and decision making in many jobs. That is the publisher's description, and the sheet we read gives no validity coefficient or norm group, a gap the section on job performance returns to. The candidate who wants evidence, as opposed to a sales description, has to look further.

This page is written for the adult who will sit, or has sat, such a test. It does not advise employers and does not offer ACIS as a way to prepare for a particular employer's test. ACIS sells an adult normed battery, so this page has a commercial interest, and it states its limits near the top and again near the end. What it can honestly offer is a way to see one's own deductive reasoning as one ability inside a profile. Neighboring topics have their own pages: logic as an IQ ability covers deduction inside intelligence testing, the Cognitive Reflection Test covers a three item measure of checking a first answer, and the sibling pages on the critical thinking test and the problem solving test cover other reasoning tests employers send.

2 What Formats Do Test Publishers Describe?

Publishers describe deductive reasoning tests as short, timed and computer delivered, with the SHL sheets also stating they are designed for unsupervised use, and with the number of questions and the question style varying by publisher and by version. The most specific public documents are fact sheets and candidate pages, so the table below uses only what we could read in them on October 29, 2026. Where a document is silent the cell says so, because a guessed time would be worse than a blank. We describe formats and never reproduce a question.

Two details deserve attention. Twenty minutes for at most 18 questions leaves a little over a minute per question, and 18 minutes for at most 12 leaves about a minute and a half (our arithmetic on the posted limits). And the SHL sheets say the tests use computer adaptive technology, so two candidates can see different questions, and a pass mark for "the test" means little without knowing the version.

The SHL fact sheet for Verify Deductive Reasoning prints one example question as a set of stated facts followed by five options, with the instruction to pick the one that must be true. The format is what matters for this page, and the lesson is plain: the answer is fixed by the facts, and the other options are the ones that merely sound possible. The interactive version, described in the Verify Interactive Deductive fact sheet, presents the same ability in an activity based design and says it can run on any device type, including mobile phones. Aon's page for candidates lists two tests under a heading for logic tests, with an untimed introduction and example exercises before each timed test. Neither looks like a verbal syllogism, which is a reminder that "deductive reasoning test" names a family and not one instrument.

A reading of the SHL wording is worth making, and it is ours. The sheet says the test measures the ability to draw logical conclusions from information provided, to identify strengths and weaknesses of arguments and to complete scenarios using incomplete information. The first clause is deduction in the strict sense. The second is closer to argument evaluation, which the sibling page on the Watson-Glaser test treats as a separate test, and the third, completing a scenario from incomplete information, is closer to inference under uncertainty than to a conclusion that must follow. Publishers are free to define the test as they wish, and the practical point for a candidate is to read the invitation from the employer, because the version and its time limit decide what the experience is like. Both SHL documents are dated by their copyright notices, 2018 and 2022, so what an employer sends today may differ.

3 How Does Deduction Differ From Induction in the CHC Model?

In the Cattell-Horn-Carroll model deduction is general sequential reasoning, the ability to reason logically from known premises, and induction is a different narrow ability, the discovery of a rule from observed cases; both sit inside fluid reasoning (Gf). The Institute for Applied Psychometrics codebook of CHC definitions, in slides by Kevin McGrew dated June 14, 2014, defines General Sequential Reasoning (RG) as the ability to reason logically using known premises and principles and lists "deductive reasoning or rule application" as its alternative name. The same slides define Induction (I) as the ability to observe a phenomenon and discover the underlying principles or rules that determine its behavior, and Quantitative Reasoning (RQ) as the ability to reason, either with induction or deduction, with numbers, mathematical relations and operators. Fluid reasoning itself is defined as the deliberate but flexible control of attention to solve novel "on the spot" problems, according to the codebook slides. The page on the CHC model sets out how these narrow abilities sit under the broad ones.

The difference is easiest to see in what the test taker has been given. In a deduction task the rules or premises are stated, and the work is to carry them through correctly. In an induction task the rule is hidden, and the work is to find it from examples before applying it. A sequence of shapes that continues according to a pattern you must discover is induction. A set of statements from which you decide what must be true is deduction. The sibling page on the inductive reasoning test covers the second family and the page on the abstract reasoning test covers the figural genre that employers send in the same batteries.

Real tasks blur the line. A person often has to induce what a term in the premises means before deducing, and quantitative reasoning is defined in the codebook as covering either route, so a numerical task cannot be filed neatly. The page on the fluid intelligence test explains what Gf tasks measure as a group. A reader who sees a score called "logical reasoning" should ask whether the task states its rules or asks the person to find them, because that decides which narrow ability the number speaks to.

Popular culture ties the word deduction to a fictional detective, and the page on what Sherlock Holmes's IQ is claimed to be shows that the number attached to him is not in the texts it searched. A deductive reasoning test measures none of that glamour. It measures whether a person can hold stated premises in mind, apply a rule without importing outside beliefs and stop when the conclusion is forced.

4 What Do Deduction and General Ability Have in Common?

Deduction overlaps with general ability, but the evidence supports a modest, specific statement: reasoning that needs working memory and sustained control tends to track intelligence, and a single deduction score is a clue, not a measure of IQ. Three lines of work say this from different directions. Evans (2003), in a review of dual process accounts of reasoning in Trends in Cognitive Sciences, describes the slower, analytic system as one that permits abstract reasoning and hypothetical thinking, is constrained by working memory capacity and is correlated with measures of general intelligence. The review is about a theory of how people reason, and it does not report a coefficient for a deduction test, so it supports the direction of the relation and not its size.

Kaufman, DeYoung, Reis and Gray (2011) came closer to the question. In Intelligence they report that 112 participants solved 70 contextualized reasoning problems in a computerized Wason selection task, a conditional rule task. At the level of individual differences, performance on all the reasoning tests was significantly correlated and loaded on a single deductive reasoning accuracy factor, and this factor was significantly correlated with general intelligence. They found no relation between general intelligence and the speed of arriving at the correct answer. Two cautions follow. The study sampled one kind of deduction, and a correlation across people does not say what any one person's score means. Read for what it supports, it says that people who deduce accurately across several problem types also tend to score higher on general ability, while a faster answer on its own did not mark higher ability.

Stanovich and West (2000), in Behavioral and Brain Sciences, reported experiments on most of the classic tasks of the heuristics and biases literature. Their abstract concludes that performance errors are a minor factor in the gap between normative and observed responding, and that computational limitations underlie non normative responding on several tasks, particularly those involving some type of cognitive decontextualization. In plain terms, many wrong answers on tasks that strip away familiar context reflect limits in what a person can hold and manipulate, not carelessness. Johnson-Laird (2010), in the Proceedings of the National Academy of Sciences, offers one mechanism for why. On his account people construct mental models of each distinct possibility consistent with the premises and derive a conclusion from them, the theory predicts systematic errors, and the ability to use counterexamples to refute invalid inferences is a foundation of rationality. More possibilities mean more models to keep active, which is why deduction leans on working memory.

The ACIS technical manual makes the same point for its own subtest: it says a low Logic Grid score with weak working memory may reflect difficulty maintaining rules, while a low score with adequate working memory may reflect difficulty using constraints strategically. None of this makes a deductive reasoning test an IQ test. The pages on what IQ measures and on critical thinking versus intelligence draw the line: intelligence tests estimate capacity under structured conditions, while judgment about when to deliberate and whether to trust a conclusion depends on more than capacity.

5 What Is Belief Bias, and Why Do Believable Conclusions Win?

Belief bias is the tendency to accept a conclusion because it is believable and to reject it because it is not, whatever the premises entail, and the classic experiments found it to be large. Evans, Barston and Pollard (1983) reported three experiments in Memory and Cognition on how people weigh logic against belief when judging syllogisms. Their abstract states that substantial belief biases were observed despite controls for possible conversions of the premises, that equally substantial effects of logic were observed despite controls for two possible response biases, and that belief bias was more marked on invalid than on valid syllogisms. The paper gives the percentage of correct judgments in each study, and the contrast between agreement and conflict is the story.

The design crosses two things: whether the conclusion follows, and whether it is believable. When the two agree (a valid conclusion that is believable, or an invalid one that is not), a reasoner can be right by following either logic or belief. When they conflict (a valid conclusion that is unbelievable, or an invalid one that is believable), only logic gives the right answer. In the first experiment, with 24 undergraduates, people were correct 92 percent of the time when logic accorded with belief and 27 percent of the time when it conflicted. In the second, with 64 undergraduates, the figures were 87 and 48 percent. In the third, with 32 psychology students, they were 97 and 43 percent. On the conflict items the only way to be wrong is to follow belief, so the share of belief following answers on those items was 73, 52 and 57 percent in the three experiments (our arithmetic: 100 minus each correct percentage).

Two things in the paper keep this from being a simple tale of irrational people. The authors recorded verbal protocols and interpreted them as showing, in some cases, rationalizations for prejudiced decisions and, in other cases, a genuine process of reasoning from premises to conclusion; where people reasoned from the premises, belief bias was minimal but still present. They also found logic effects as strong as belief effects, summarizing the picture as a conflict between logic and belief at several levels. Our original example above is the worst combination, an invalid conclusion that sounds fine.

Later work asked what the bias is. Trippas, Handley and Verde (2013), in the Journal of Experimental Psychology: Learning, Memory, and Cognition, noted that prevalent theories claim belief affects the quality of syllogistic reasoning, while recent work had suggested it may be a simple response bias. In their receiver operating characteristic analysis, believability affected accuracy for complex but not for simple syllogisms. In a second experiment the effect on accuracy disappeared when judgments were made under time pressure and with participants low in cognitive capacity. The authors concluded that beliefs influence more than response bias when conditions are conducive to the use of certain reasoning strategies, and that individual differences need to be considered. In plain terms, belief does not only bias what people say, it can change how well they reason, and how much it does depends on the task and the person.

For a candidate the practical meaning is narrow and useful. A test that states facts about everyday things invites a reader to bring in what they know, while a test built on arbitrary labels gives belief nothing to hold. The SHL and Aon documents we read do not say whether any item sets belief against logic, so a candidate should not assume either way. These are group findings from experiments with students, and they describe how reasoning behaves on average, not the reasoning of any one reader. The sibling page on how critical thinking is tested looks at the related question of who notices when a conclusion deserves a second look.

6 What Can a Candidate Do About Belief Bias and the Clock?

The research supports a small set of modest habits, and it also says plainly that familiarity with a test raises scores a little on average without telling you how much it will raise yours. Start with what is known about belief. Macchi and colleagues (2019), in Europe's Journal of Psychology, wrote that earlier attempts to reduce the influence of belief bias on deductive thinking had often been unsuccessful and, when they succeeded, had failed to replicate. Their own study changed the pragmatics of the task through a subtle manipulation of the instruction of the syllogism problem and obtained a strong improvement in accuracy in both untrained respondents and respondents trained in logic. It is one study, and we do not claim the improvement transfers to a timed employer test. It does suggest that reading what a task is asking for, and treating it as a request to judge only whether the conclusion follows, is not trivial.

Johnson-Laird (2010) points to a second habit. If reasoning means building models of the possibilities consistent with the premises, then the useful question is whether any possibility makes every premise true and the conclusion false. Such a possibility is a counterexample, and his abstract describes the ability to use counterexamples to refute invalid inferences as a foundation for rationality. These suggestions are ours, drawn from that account, and no study cited here tested them on an employer's test: restate the conclusion with neutral letters, ask whether the premises could all hold while the conclusion fails, and answer only from the stated facts. On a timed test, one more is practical. At roughly a minute per question (our arithmetic from the SHL limits above), a long search for a counterexample on every item is not available, so the habit is to look for one quickly and move on.

The evidence on repetition comes from selection research. Hausknecht, Halpert, Di Paolo and Moriarty Gerrard (2007), in the Journal of Applied Psychology, pointed out that as many as 25 to 50 percent of applicants in organizational and educational settings are retested, according to previous studies, and summarized 50 studies of practice effects for cognitive ability tests. Results from 107 samples and 134,436 participants revealed an adjusted overall effect size of .26. Effects were larger when practice was accompanied by test coaching and when identical forms were used, and the authors wrote that additional research is needed to understand the impact of retesting on the validity inferences drawn from test scores. Aon's candidate page says candidates may be asked to repeat a module in a later retest, so the retest the study describes is not hypothetical.

Three conclusions follow. Practice on a similar format tends to help on average, more so with coaching and identical forms. The meta-analysis covers cognitive ability tests as a group, so it gives no figure for a specific deductive test and cannot say how much practice would help a particular person. And material sold as practice is not the employer's test: sellers write their own items, and this page does not link to them or reproduce any item. The page on how to prepare for an IQ test covers what helps and what hurts on a normed ability test.

7 What Does a Deductive Test Score Say About Job Performance?

A deductive reasoning score is evidence about one narrow ability, and whether it predicts performance in a given job depends on validity evidence that the fact sheets we read do not print. The SHL sheet says its test provides an indication of how an individual will perform when asked to develop solutions when presented with information and draw sound conclusions from data. That is a statement of intent. The sheet gives no validity coefficient, no norm group and no study, and the Aon candidate page, which is aimed at test takers, does not either. The absence from a fact sheet is not evidence that the validation work does not exist, since publishers may hold technical documents elsewhere, but a candidate cannot read it on those pages.

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) say what such evidence would look like. In its chapter on workplace testing, Standard 11.12 says that when personnel selection relies on a relationship between a predictor construct domain and a criterion construct domain, two links need to be established: evidence that the test scores are reliable and that the content adequately samples the predictor construct, and evidence for the relationship between that predictor domain and major factors of the criterion domain. Its comment gives, as an example, a cognitive ability test used for a job that is complex and requires sophisticated analysis of many factors, supported by research findings and by expert judgments about the importance of cognitive ability. Standard 11.4 adds that inferences drawn from test scores should be limited to those for which validity evidence is available, and its comment warns against treating a high score on a test of a single narrow domain as showing overall job suitability.

Standard 2.14 bears on cut scores. It says that where cut scores are specified for selection or classification, the standard errors of measurement should be reported in the vicinity of each cut score. The fact sheets we read do not show that. It matters because a score a point or two above a cut and one a point or two below it can be statistically close, which is a reason not to read a rejection as a verdict on ability. The standards do not say employers should not use such tests; they say each use needs its own evidence.

The research on general mental ability in work settings is much larger than the research on any one narrow test, and the page on IQ and job performance reviews it with its limits. The page on the pre-employment cognitive test covers the short, timed general ability tests employers also use. For a candidate, the sensible stance is neither dismissal nor awe: a deductive test samples a real ability, and a single result is one measurement under one set of conditions, not a ruling on whether someone can do a job.

8 How Do the Real Options Compare?

Only two of the routes to a deductive reasoning score give you an adult normed result you can read yourself, a psychologist's battery and a normed online battery, and neither is the employer's test. The first two rows are the tests an employer orders. A candidate does not choose them, and the documents we read give no norm group or price for a candidate. The remaining rows are what a curious adult can obtain. The table holds only what we read on October 29, 2026, and where a source is silent the cell says so.

The Pearson WAIS-5 product page, read on October 29, 2026, gives the age range, the administration times and the qualification level shown in the table, and names five cognitive domains, Verbal Comprehension, Visual Spatial, Fluid Reasoning, Working Memory and Processing Speed. We found no separate deduction score on it, so a psychologist's report would speak to fluid reasoning as a broad domain. The page on where to take an IQ test compares the in person and online routes in detail, and the page on the IQ test for adults explains what a self administered adult test can and cannot replace.

The free quiz row is where most searches for a deductive reasoning test free of charge end. A quiz can be well made and still cannot report where you stand, because standing needs a comparison group, and a score from a practice site cannot be tied to the comparison group an employer uses, which the documents we read do not name. The page on the cognitive ability test sets out what an adult can take, and if the question is a hiring decision, none of these rows replaces what the employer's own process produces.

9 How Should a Deductive Reasoning Score Be Read?

A deductive reasoning score should be read as one measurement of one narrow ability, made under one set of conditions, set against a stated comparison group and carrying an error band, not as a verdict on how well you think. The first difficulty with an employer's result is that the documents we read do not say how it is reported. The SHL and Aon pages describe the tests and not the score, so a candidate cannot tell from them whether a result is a percentile against a comparison group the employer chose, a band or a pass mark. If the invitation or a feedback report does not say, asking what group the score is compared with is a reasonable question.

On a normed battery the scale is stated. ACIS reports indices and the Full Scale IQ on the standard scale, with a mean of 100 and a standard deviation of 15, and scaled subtest scores from 1 to 19, with a mean of 10 and a standard deviation of 3, plus percentiles and a 95 percent confidence interval. A scaled score of 13 is one standard deviation above the mean (our arithmetic: 13 minus 10, divided by 3), the same distance as an index of 115. The pages on how IQ scores are normed and on why the standard deviation of 15 matters explain the scale, and the page on score versus percentile explains why the two must not be averaged or swapped.

Precision is the second difficulty. A subtest is a short sample of behavior, and a short sample is a less precise measure than a battery that combines several. The Logic Grid page says so directly: one subtest should never be read as the whole construct, and the pattern across tasks is usually more informative than any isolated score. The Standards for Educational and Psychological Testing require that the standard error of measurement be provided in the units of each reported score (Standard 2.13), which is why a score is best read with its interval. The page on reliability and validity explains the terms with examples, and the ACIS technical manual is the place for the documentation of the battery, which this page does not quote.

The third difficulty is profile shape. Index scores in one person often differ, and the page on index score differences explains how to judge whether a gap is meaningful. The Logic Grid page lists three reasons a weaker result may arise: working memory load, impulsive elimination or difficulty coordinating conditions, rather than a general reasoning deficit. Group findings such as the belief bias experiments describe students and volunteers on average, and they do not diagnose an individual.

10 What Can Logic Grid Show You About Your Own Deduction?

Logic Grid is the ACIS subtest that samples deduction directly, and it is a way to see your own deductive reasoning on adult norms, not a copy of any employer's test. The Logic Grid page describes the subtest as one that tests "deductive reasoning by solving logic puzzles". Its instructions state that each puzzle shows a grid with some shapes already placed, that each row, each column and each outlined region must contain each shape exactly once, and that the person uses logic to decide which shape belongs in the highlighted cell. Two practice puzzles come first and are not scored, the clock starts after them, the time limit is 20 minutes in total, the test stops after 3 consecutive errors and no external aids are allowed. The Logic Grid subtest page has the full public description.

The technical manual files Logic Grid under general sequential reasoning within Gf, the CHC ability described earlier. It describes the task as one where the answer follows by elimination from the constraints and not by recognizing a pattern, which separates it from Matrix Reasoning, and where the rules are logical and relational and not numerical, which separates it from Arithmetic. The manual adds that a person may be strong at pattern induction and weaker at stepwise deductive tracking, or the reverse, and that Logic Grid should be read with Complex Relations, Matrix Reasoning and the working memory index. Complex Relations is the ACIS subtest the manual files under induction.

The grid carries no real world content. That is our reading of the format described on the page, and it has a consequence worth stating: there is nothing for belief to attach to, so Logic Grid cannot show how you handle a believable but invalid conclusion, which is the trap the belief bias section describes. An employer's verbal test may sample that and Logic Grid does not, which is one more reason not to treat one as a stand-in for the other. Nothing here says that Logic Grid resembles, predicts or prepares for any test described above.

Logic Grid feeds the Fluid Reasoning Index, which combines five subtests, Matrix Reasoning, Figure Weights, Visual Number Series, Logic Grid and Complex Relations. The manual says agreement across them supports broad fluid reasoning and disagreement suggests that format, quantitative content, general sequential reasoning or relation abstraction may be influencing the profile. Figure Weights, which is in every form, is the quantitative member of that group: its page describes relational balance reasoning, between Gf and quantitative knowledge, and the manual assigns it to quantitative reasoning and not to general sequential reasoning. The page on the Fluid Reasoning Index explains why factor analyses of the Wechsler version of that index are disputed, and the page on how deduction is scored among the Gf abilities goes further than this one.

What the buyer gets depends on the form. Logic Grid is in the Optimized form (13 subtests in 5 domains, 30 dollars, about 110 minutes) and the Full Scale form (all 20 subtests in six domains, 50 dollars, about 175 minutes), and in the Custom Test at 5 dollars per subtest, prices read October 29, 2026 and subject to change. ACIS states that any domain you complete in a Custom Test returns its full index score and that every subtest returns its scaled score, so Logic Grid alone returns a scaled score and not an index. Taking all five Fluid Reasoning Index subtests as a Custom Test would cost 25 dollars (our arithmetic: five at 5 dollars). The free trial needs no card and opens other subtests, so a reader who wants Logic Grid pays for a form. Every form is a one time payment with no subscription, with breaks allowed, progress saved between subtests and 30 days to finish.

11 What Does an Unsupervised Online Session Need?

Both the employer's test and an ACIS session ask the test taker to supply the testing conditions, so a quiet room, no outside help and an honest effort are part of what the score means. SHL describes its deductive tests as designed for unsupervised use, and Aon's page says assessments can be completed at any time, from anywhere, for example from home. Standardization is the reason a test room exists, and the open question for any unsupervised score is whether the conditions matched those the norms assume.

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) address this in several places. A comment under Standard 3.4 says unproctored test administrations, where standardization cannot be ensured, could provide an advantage to some test takers over others, and that in such situations results should be interpreted with caution. A comment under Standard 10.9 says the interpreter of the test scores should be informed if the test was unproctored. The APA Guidelines for Psychological Assessment and Evaluation (2020), in Guideline 8, say that psychologists strive to follow standardized administration to support valid interpretations of norms and that adhering to standardized testing conditions will minimize confounds that could lead to misinterpretations of test results. Those documents are written for professionals, and they apply to a reader as well: the conditions are part of the data.

For ACIS the conditions are concrete. The Logic Grid page says no calculators, notes, scratch paper or drawing tools are allowed, and the 20 minute limit applies to the subtest. Three ACIS subtests present items as audio and need working sound: Digit Span and Alphanumeric Sequencing, which are in the Optimized and Full Scale forms, and Arithmetic, which is in the Full Scale form. The technical manual says verbal and knowledge tasks are affected by English proficiency and educational exposure, and the whole battery is English only. A long form can be spread across days within the 30 days, because breaks are allowed and progress is saved between subtests, and a rested start is a better one.

For the employer's test the same ideas hold with fewer choices. A quiet place, a device that works and the time block the invitation names are the candidate's to supply. Helpers change what is measured: a calculator, a search or a friend reading along turns the result into a description of the helped performance.

12 Where ACIS Sits, and What a Buyer Should Do

ACIS is a normed online battery for adults who want to see their own cognitive profile, including how deduction compares with induction and quantitative reasoning, and it is not the employer's test, a preparation course for it or a hiring tool. The strengths are concrete. It reports a Full Scale IQ and six primary indices from 20 subtests in six CHC domains on its Full Scale form, with percentiles and a 95 percent confidence interval on every form, adult norms for ages 16 to 90, breaks, saved progress and 30 days to finish. The technical manual documents how the battery is built and normed, and this page states nothing about that beyond what the report shows.

The limits are as concrete. ACIS is online and unsupervised, it is not a clinical or diagnostic instrument and it is English only. Its research page states that it "is not intended for clinical diagnosis, hiring decisions, accommodation determinations or admission to high IQ societies". No result should be sent to an employer as a substitute for the test the employer set, and ACIS does not claim to predict, convert to or stand in for any employer test. The refund promise is narrower than people assume and should be read as written: "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 Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) are the yardstick this page tries to meet, not an endorsement of ACIS. Standard 1.0 asks that each intended score interpretation for a specified use be clearly articulated and supported by appropriate validity evidence. Standard 1.3 asks that, when a likely interpretation has not been evaluated, potential users be strongly cautioned, and Standard 1.4 makes the user who interprets a score in an unvalidated way responsible for justifying it. The APA Guidelines for Psychological Assessment and Evaluation (2020), in Guideline 4, ask the psychologist to recognize when the use of an instrument deviates from its expected or intended purpose and to weigh the significance of the deviation. Read together, they say the same thing: use an ACIS score for what it was designed to show, which is your own profile on adult norms, and use the employer's process, a recruiter or a licensed psychologist for the decisions that need them.

A reader can proceed in this order.

1. Read the invitation. Note the version, the time limit and the number of questions the employer states, and use the publisher's own candidate materials for that test.
2. Decide the question. If it is about the employer's test, ACIS is the wrong tool. If it is about knowing your own deduction, continue.
3. Pick the route. Logic Grid alone is 5 dollars in the Custom Test. For the whole fluid reasoning picture, the Full Scale form (50 dollars) contains all five Fluid Reasoning Index subtests. The free trial, with no card, shows the format of the platform first.
4. Set up the room. Quiet, no aids, working sound for the audio subtests and a plan for breaks.
5. Read the result as one source. Look at the interval and at the pattern across subtests, and do not use the number for a decision it was not built to support.

13 Sources Behind This Page

Every figure above traces to one of the following sources, each linked at the point where it is first used, and every date is the day we opened the page, October 29, 2026. Prices of ACIS forms were read that day and can change. Any figure we derived ourselves is labeled in the text as our arithmetic. Where only an abstract was read, the entry says so.

• SHL. Verify Deductive Reasoning Assessment Fact Sheet. Copyright 2022, document CST22_138. Read October 29, 2026.
• SHL. Verify Interactive Deductive Reasoning Assessment Fact Sheet. Copyright 2018. Read October 29, 2026.
• Aon. Prepare for your Online Assessment. Undated page, footer copyright 2026, read October 29, 2026.
• McGrew K. CHC theory codebook 1: cognitive definitions. Institute for Applied Psychometrics slides dated June 14, 2014. Read October 29, 2026.
• Evans J, Barston J, Pollard P. On the conflict between logic and belief in syllogistic reasoning. Memory and Cognition, 1983, volume 11, issue 3, pages 295 to 306.
• Trippas D, Handley S, Verde M. The SDT model of belief bias: Complexity, time, and cognitive ability mediate the effects of believability. Journal of Experimental Psychology: Learning, Memory, and Cognition, 2013, volume 39, issue 5, pages 1393 to 1402. Abstract read.
• Macchi L, Poli F, Caravona L, Vezzoli M, Franchella M, Bagassi M. How to get rid of the belief bias: Boosting analytical thinking via pragmatics. Europe's Journal of Psychology, 2019, volume 15, issue 3, pages 595 to 613. Abstract read.
• Johnson-Laird P. Mental models and human reasoning. Proceedings of the National Academy of Sciences, 2010, volume 107, issue 43, pages 18243 to 18250. Abstract read.
• Stanovich K, West R. Individual differences in reasoning: Implications for the rationality debate? Behavioral and Brain Sciences, 2000, volume 23, issue 5, pages 645 to 665. Abstract read.
• Evans J. In two minds: dual-process accounts of reasoning. Trends in Cognitive Sciences, 2003, volume 7, issue 10, pages 454 to 459. Abstract read.
• Kaufman S, DeYoung C, Reis D, Gray J. General intelligence predicts reasoning ability even for evolutionarily familiar content. Intelligence, 2011, volume 39, issue 5, pages 311 to 322. Abstract read.
• Hausknecht J, Halpert J, Di Paolo N, 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. Abstract read.
• Pearson Assessments. Wechsler Adult Intelligence Scale, Fifth Edition (WAIS-5) product page. Age range, administration times, qualification level and list prices, read October 29, 2026.
• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing, 2014. Standards 1.0, 1.3, 1.4, 2.13, 2.14, 3.4, 10.9, 11.4 and 11.12, read in the open access edition.
• American Psychological Association. Guidelines for Psychological Assessment and Evaluation, 2020, Guidelines 4 and 8, read October 29, 2026 in an archived copy of the apa.org document.

14 Frequently Asked Questions