The IQ test for friends: how to compare scores that mean something

0 The short answer

Friends can compare IQ fairly when each person takes the same normed test alone, in a separate session, and compares index profiles with their intervals instead of one number. ACIS is one concrete option: its Quick, Optimized and Full Scale forms cost 15, 30 and 50 dollars and take about 45, 110 and 175 minutes (prices read October 29, 2026), and each person needs a separate purchase, a free trial on their own email, or a gift code from a friend. The decisive limit is the size of ordinary gaps: even friends who resemble each other in ability differ by about 15 points on average at an assumed correlation of 0.2 (our arithmetic), so a gap of 10 points between two friends is expected and says nothing by itself about who is smarter. The test is online, unsupervised and not clinical.

1 What Is an IQ Test for Friends, and What Is It Not?

There is no separate IQ test for friends: the phrase covers a party quiz, a head to head game and a real normed test taken by each person alone, and only the last supports a comparison. The same is true of the searches for an IQ test for two people, for siblings or for a family. A quiz built for sharing often returns a number without publishing the group it was compared against, so two friends who take different quizzes, or even the same quiz on different days, have nothing in common to compare. A head to head game, where two people race through the same puzzles, measures who was faster that evening. A normed test is different in kind: it places each person on a scale defined by a reference group of adults, so two results taken apart still sit on one scale.

That scale is the whole point of doing this properly. The page on what IQ measures explains that an IQ battery samples several abilities and combines them, and that the number is a position relative to other people, not a quantity of cleverness. When two friends compare positions on the same scale, the comparison has meaning. When they compare numbers from different instruments, it does not, in the same way that two people comparing heights in different, unlabeled units would learn nothing.

Three practical facts frame everything that follows.

• Each person must take the test separately. A test taken together, with one person reading over the other's shoulder, measures neither person cleanly. The sections on fairness and on form choice below set out how to run separate sessions.
• Age matters for who can join. ACIS adult norms cover ages 16 to 90, so a family test can include a 16 year old but not a 12 year old. The gift page asks the buyer to confirm that the person receiving the code is 16 or older.
• A free look is possible, a free comparison is not. The ACIS free trial needs no card, and the page on the free IQ test explains what it includes. It opens only part of the battery, so it helps a friend decide whether to buy, and it is not a basis for ranking five people.

A group is therefore choosing one instrument and agreeing a protocol, not shopping for a game. Friends who would rather not create an account at all can read the sibling page on taking an IQ test without signing up.

2 Which Tests Can Friends Use, and What Does Each Cost?

Five routes are open to a group of friends, and only two of them, a psychologist with an adult Wechsler battery and a normed online battery, return an IQ score with a stated reference group. The table lists each route with the facts we could read on October 29, 2026. The university clinic fee is one example of how a service is priced and not a survey, and we did not find a published private practice fee that we could verify, so we state none.

Read the table by what each route is for. A psychologist evaluation answers a clinical or educational question with professional judgment and is priced as a whole evaluation. Pearson's WAIS-5 product page lists the age range as 16:0 to 90:11, which means 16 years 0 months through 90 years 11 months, and the administration times above. The University of South Florida Psychological Services Center says its fees depend on income and on the type or complexity of the assessment, and lists a gifted assessment with the WISC or WAIS at 100 to 400 dollars, with half of the quoted fee due as a deposit. For five friends that is five appointments, and the route is meant for a clinical or educational question, not for settling a bet.

American Mensa is a different kind of answer again. Its testing page says a qualifying score means testing at or above the 98th percentile of the general population, that it cannot provide a detailed report with scores, percentile ranks or an IQ score, and that it cannot test anyone under 14. The regular prices on the page, 60 dollars for a Local Group session and 99 dollars for private electronic testing, were read on October 29, 2026, and the page showed a promotion that ends October 31, which we do not quote. A friend group that wants numbers to compare cannot get them from a pass or fail society exam. The page on where to take an IQ test lists the in person and online routes in more detail, and the ranking on the best online IQ tests compares the online options.

Job screening practice tests train for employer formats and are not IQ tests. The page on IQ tests for adults is the entry point for an adult who wants the standard route.

3 How Should Friends Take the Test So the Comparison Is Fair?

A fair comparison needs the same instrument, comparable conditions, independent work and no advance coaching, and every one of those is something the friends control and a score report cannot. The measurement literature states the principle in general terms. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) say in Standard 6.4 that the testing environment should furnish reasonable comfort with minimal distractions to avoid construct irrelevant variance, and the comment adds that noise, disruption, poor lighting and malfunctioning computers are among the conditions to avoid, and that in general the testing conditions should be equivalent to those that prevailed when norms and other interpretative data were obtained. The APA Guidelines for Psychological Assessment and Evaluation (2020), in Guideline 8, ask 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. An online test has no examiner to apply these, so the people taking it apply them to themselves.

In practice that comes to six habits.

1. Use the same form. Quick, Optimized and Full Scale cover different numbers of domains, so a Quick result and a Full Scale result share only part of a profile. The next section sets out which domains overlap.
2. Take separate sessions on separate devices. The free trial is one per person, and the site refuses a second trial from the same email address or the same browser, so each person starts on their own device with their own email. The same separation protects the scores: nobody should be reading over a shoulder, whispering an answer or calling out how hard a task was.
3. Do not discuss tasks until everyone has finished. Describing a format afterward is harmless. Describing what to expect before a friend starts is a form of coaching, and the retest research cited later shows how much familiarity alone can add. The Standards also treat the protection of test materials as a responsibility, in Standard 6.7, and the ACIS pages describe formats and never reproduce items for the same reason. Friends should do the same and not post or paste task content.
4. Match the conditions you can control. A quiet room, a charged device and a rested mind are the plain version of Standard 6.4. Three subtests present their items as audio and need working sound: Digit Span, Alphanumeric Sequencing and Arithmetic. Headphones remove the argument about a noisy room. Breaks are allowed between subtests and progress is saved, so a tired person should stop and resume instead of pushing through.
5. Mind the language. The test is in English only, and the ACIS technical manual says verbal and knowledge tasks are affected by English proficiency and educational exposure. A friend who learned English as a second language may score lower on a verbal index for reasons that have more to do with English than with verbal ability in their strongest language. The manual that documents the battery sets out the details.
6. Agree beforehand what the result is for. If the purpose is curiosity, nobody should use the result to settle who is cleverer.

None of this makes an unsupervised online session equal to a supervised one. It makes two unsupervised sessions comparable with each other, which is the realistic goal for friends.

4 Which Form Should a Group Choose?

Everyone in the group should take the same form, because the forms cover different domains and sample different numbers of subtests, so mixing forms leaves only a partial profile that can be compared like for like. ACIS has three forms, each a one time payment with no subscription, with breaks allowed, progress saved between subtests and 30 days to complete. Prices were read on October 29, 2026 and can change.

Every form reports a Full Scale IQ with its 95 percent confidence interval, and the Full Scale form measures all 20 subtests. The domains each form covers are not the same, and that is what decides whether two friends on different forms can be compared at all.

• Quick covers three domains: verbal comprehension (Similarities and Vocabulary), fluid reasoning (Matrix Reasoning and Figure Weights) and working memory (Digit Span and Alphanumeric Sequencing).
• Optimized adds visual spatial (Spatial Comprehension and Layer Rotation) and processing speed (Coding and Symbol Search), and adds Antonyms, Synonyms and Logic Grid inside the verbal and fluid domains, for five domains in all.
• Full Scale adds the quantitative reasoning domain (Mathematical Achievement and Arithmetic) and the other subtests that appear only in this form. Seven subtests are exclusive to Full Scale: Information, Arithmetic, Mathematical Achievement, Visual Number Series, Complex Relations, Visual Puzzles and Visual Sequence.

If one friend takes Quick and another takes Full Scale, the only indices they share are the verbal, fluid reasoning and working memory indices, and even those are built from more subtests in the larger form. The comparison can still be made on those three, with the caution that Quick builds each of them from fewer subtests, and each friend's Full Scale IQ rests on the subtests that friend's form contains. The simplest rule for a group is to pick the form the least patient member will finish, and for everyone to take that one.

A fourth route suits friends who want to compare only one ability. The Custom Test costs 5 dollars per subtest, and the buyer picks any subtests. According to the product description, any domain you complete returns its full index score, and every subtest returns its scaled score, while the Full Scale IQ needs the Full Scale battery. At 10 subtests the Full Scale form costs the same. Two friends who argue about who is quicker at visual tasks could each buy the two subtests of the processing speed domain and compare that index alone, and the page on the processing speed test describes an index that takes about 10 minutes. Friends who play games together may want exactly that comparison, and the sibling page on the IQ test for gamers is written for that audience.

Four friends on Quick spend 60 dollars together and four on Full Scale spend 200 dollars. A gift code moves the cost to the giver, and the free trial shows a task before anyone pays.

5 Can You Give a Friend the Test as a Gift?

Yes: a gift is a code for one ACIS assessment that the friend redeems on their own schedule, and the results stay private to the friend, so the giver learns the score only if the friend chooses to share it. The gift page, read on October 29, 2026, describes each gift as a code for one assessment, Quick at 15 dollars, Optimized at 30 dollars or Full Scale at 50 dollars, on the same platform and with the same report as any purchase. The code arrives by email, and the page says it never expires. The friend opens the Take the Test section, chooses the assessment written on the gift, and enters the code at checkout, where the price drops to free. The 30 days to complete the assessment start at the moment the code is redeemed, not when it is bought, which matters for a gift given at a birthday and opened months later.

Three details of the gift flow bear on friends in particular.

• Age. The buyer confirms that the person receiving the gift is 16 or older. The page explains that ACIS norms start at 16 and that the recipient confirms their age when they redeem the code. A gift for a younger sibling is therefore not possible.
• Privacy. The page says the results stay private to the person who takes the test. That is a design choice worth respecting: a gift of a test is not a request for the result, and a friend who keeps the number to themselves has done nothing unusual.
• A message and a card. The giver can add the names of the sender and the recipient and a short message, which the form limits to 200 characters. The page also offers the code as a card to print, and as a code and a redeem link to copy.

The gift page also carries its own guarantee: if the code does not work, or a technical issue keeps the friend from their results, the page says the purchase is refunded in full. That is a different statement from the refund terms of a direct purchase, which this page quotes in its last section.

A gift raises a social question that no page can answer for you, which is whether the friend wants to be tested. A person can be curious about their own profile and still not want an IQ number as a present, and a gift with an implied comparison can feel like a verdict. The kind approach is to ask first, to say you hope it gives them a profile of strengths and not a rank, and to offer the free trial as a low stakes first step. Friends choosing a first test can also read the sibling page on IQ tests for young adults.

6 Compared With Whom? Why Friends Share One Scale

Friends are not scored against each other: each person is compared with a reference group of adults that the test defines, and because the same instrument defines it for everyone, two reports are two positions on one scale, which is what makes a comparison possible. ACIS reports indices and the Full Scale IQ on the standard scale with a mean of 100 and a standard deviation of 15, scaled subtest scores from 1 to 19 with a mean of 10 and a standard deviation of 3, percentiles, and a 95 percent confidence interval. The page on how IQ scores are normed explains what a norm is and why scores from different tests are not interchangeable, and the page on the standard deviation of 15 explains why that unit was chosen.

The consequence for a group is that points and percentiles do not line up evenly. A 15 point gap sounds the same everywhere on the scale, and the percentile difference it produces depends on where on the scale it falls. Our arithmetic on a normal curve with a mean of 100 and a standard deviation of 15 shows the point. A score of 100 sits at the 50th percentile and a score of 115 at about the 84th, so a 15 point gap in the middle of the scale separates two people by about 34 percentile points. A score of 130 sits at about the 98th percentile and 145 at about the 99.9th, so the same 15 point gap at the high end separates two people by about 2 percentile points. Two friends who sit at 130 and 145 are separated by 15 points of IQ and almost nothing in rank. The page on IQ score versus percentile explains why percentiles should not be averaged, which matters when a group wants a single figure for itself, and the IQ percentile calculator converts any score to its percentile. The IQ rarity calculator states the same position as one person in some number.

For siblings and family members, the reference group raises one more point. A parent and a child are not compared with each other, and each report places its own person on the scale with its own interval. A higher number for a parent than for a teenager is therefore not a statement about who is more able in some absolute sense; it is a statement about two positions, each with its own error. How the reference group is defined for people of different ages is a question for the documentation, and a family comparing across generations should read the technical manual before treating a gap as one between like and like. This page does not quote figures from it.

7 Do Friends Resemble Each Other in Measured Ability?

In the studies of adolescent best friends that we could open, friends resembled each other in measured intelligence, the resemblance was partial and not a match, and the studies cannot say whether friends chose each other for it, grew alike, or simply attended the same school. The first piece of evidence is a large British school study. Burgess, Sanderson and Umaña-Aponte (2011) analyzed a network of 6,961 friendship links among adolescents in the West of England and reported that adolescents tend to befriend peers who resemble them, and that friends chosen at random are much less alike than actual friends. The abstract on the Bristol page names academic achievement, personality and aspirations as features that matter in forming friendships, and it does not mention IQ. The IQ result from the same data is described by a later paper: Meldrum and colleagues (2019) write that Burgess and colleagues reported children's IQs to be significantly correlated with their current friends' IQs at ages 8 and 16 in a sample of almost 7,000 individuals. We have the later paper's description of that result and not the original table.

The second piece is an American study of best friend pairs. Boutwell, Meldrum and Petkovsek (2017) asked whether friendships form partly on shared cognitive ability. Its abstract, as listed in the RePEc index, reports that preadolescent best friends are robustly correlated on general intelligence and that the finding held after controlling for potentially confounding variables. The later paper by Meldrum and colleagues summarizes the study as an analysis of 810 fourth graders and their best friends, with a coefficient of 0.22 after accounting for family and demographic characteristics, in both race and sex subgroups. We cite the 810 and the 0.22 as that paper's summary of the study, because the original paper was not open to us. The coefficient is reported after adjustment, so it is not the same quantity as a plain correlation, and we use it only as an indication of order of magnitude.

The third piece asks the harder question. Do friends make each other smarter, or only start out alike? Meldrum and colleagues analyzed 892 adolescents and their best friends from the Study of Early Child Care and Youth Development, a United States cohort. The friend's intelligence was measured in sixth grade with the Peabody Picture Vocabulary Test, a vocabulary test, and the adolescent's was measured at age 15 with three Woodcock-Johnson tests: picture vocabulary, passage comprehension and applied problems. The authors note that variation in these scores largely captures crystallized intelligence. They report that a large bivariate association between friend and adolescent intelligence was reduced to non-significance once prior levels of adolescent intelligence and other background variables were controlled, and they conclude that their study does not find evidence supporting a socialization effect of peers on intellectual ability. The paper's title calls it exploratory, and the result is one study.

What these studies support is narrow. Best friends in adolescence resemble each other in measured intelligence, at least on vocabulary and knowledge weighted measures. The paper we read finds no evidence that the resemblance comes from friends raising each other's scores. And the limits are real: all of these are samples of adolescents, the best friends in the American cohort were same sex friends of similar age whom the child had known for at least six weeks and spent time with weekly, and the measures leaned on knowledge. We did not find, in the sources we opened, a study of adult friends, and we state no figure for them. A group of adult friends who take the same test may well differ more or less than the adolescent pairs did.

Family resemblance has its own, larger literature, and two neighbor pages own the detail. The page on what heritability actually means explains how twin and family studies are read, and the page on birth order and IQ covers the question of whether older siblings score higher. For a sibling comparison we cite one national study. Kendler and colleagues (2015) worked with Swedish male conscripts whose IQ came from the military conscription examination at about ages 18 to 20, rescaled to a mean of 100 and a standard deviation of 15, and report an intraclass correlation among biological full siblings of +0.30. That is a figure for one design, in men, and it combines genetic and rearing influences, so it is a data point and not a universal sibling constant. The page on the couples IQ test covers the partner case.

8 How Big a Gap Should Friends Expect?

Even when friends resemble each other in ability, the typical gap between two of them is large: at a correlation of 0.2 the average gap is about 15 points, and about 43 percent of pairs differ by 15 points or more. The arithmetic below is ours. It starts from three assumptions: each person's score is normally distributed with a mean of 100 and a standard deviation of 15, the two scores are related by a bivariate normal distribution with a stated correlation, and scores are measured without error. The standard deviation of the gap between two people is then 15 times the square root of 2 times (1 minus the correlation), and the average size of the gap, ignoring direction, is about 0.80 times that standard deviation.

The rows are what-if values and not measurements of any group. The 0.2 row is the order of magnitude of the best friend coefficient of 0.22 reported above, which was adjusted and is not a plain correlation. The 0.3 row matches the sibling intraclass correlation of +0.30 in the Swedish conscript study. The 0.5 row is a deliberately generous case. Real measurement error widens every gap, so the true shares would be somewhat larger than the table says. The point of the table is the pattern. Even at 0.5, which is a strong resemblance for a psychological trait, a typical pair of friends or siblings differs by about 12 points, and about one pair in three differs by 15 points or more.

Groups of friends make the pattern stronger, because with more people there are more chances for two of them to be far apart. For a group of unrelated people with a standard deviation of 15, the expected distance between the highest and lowest score is about 17 points for 2 people, 25 for 3, 31 for 4 and 35 for 5 (our arithmetic). If the group resembles each other at 0.2, with equal resemblance among all members, the expected top to bottom distance shrinks by a factor of about 0.89.

The expected distances come from the standard calculation for the range of a normal sample, and the shares from a simulation of 200,000 groups for each size (our arithmetic). In a group of five it is close to certain that the highest and the lowest score will differ by a figure that looks dramatic, and that figure is mostly the shape of the statistics and not a fact about the people. A ranking among friends will always produce a last place, and the last place says little about the person.

9 How Big Must a Gap Be Before It Is Real?

A gap between two friends' scores is real only if it exceeds what measurement error alone would produce, and because each score carries its own error, the gap needs to be about 2.8 times the standard error of measurement of one score before the 95 percent threshold is met. The Standards for Educational and Psychological Testing address the principle in Standard 2.4: when a test score interpretation emphasizes differences between two observed scores of an individual or two averages of a group, reliability and precision data, including standard errors, should be provided for such differences. The comment adds that the reliability of change scores can be much lower than the reliabilities of the separate scores involved. The standard speaks of two scores of one person, and applying the same logic to two different people tested on the same instrument is our extension, though a conventional one.

The arithmetic is short. If each person's score has a standard error of measurement of s points and the two errors are independent, the standard error of the difference is s times the square root of 2, about 1.41 times s. The 95 percent band for the difference is 1.96 times that, about 2.77 times s. We use illustrative standard errors and not ACIS figures, which this page does not state.

Two lessons come from the last two columns. First, a gap well above the threshold is unlikely to be error, and a gap below it could easily be error alone: with an illustrative standard error of 3 points, a 10 point gap clears the 8.3 point threshold, and a 6 point gap does not. Second, intervals can mislead in one direction. Two reports may show 95 percent intervals that still overlap even though the gap is real, because the threshold for a real difference is lower than the point at which the intervals stop touching. Overlap is a quick check that errs toward calling two people equal, and the square root rule is the sharper one. The page on reliability and validity explains where the standard error comes from.

Two cautions complete the picture. The arithmetic describes the reliability of the gap, not its meaning: a real gap of 12 points says that two scores differ by more than error, and it does not say why, or whether it matters in anyone's life, or which friend had the better day. And with a group, comparisons multiply. Six indices compared across every pair of friends give 18 comparisons for 3 friends, 36 for 4 and 60 for 5 (our arithmetic). If no true difference existed and each comparison used a 5 percent threshold, about 1 comparison in 20 would still look significant by chance, which is roughly 1.8 flagged gaps among 4 friends. The comparisons are not independent, so that figure is an approximation. The page on index score differences works through the same problem inside a single profile, and the Standards add, in Standard 1.14, that when interpretation of subscores, score differences or profiles is suggested, the rationale and relevant evidence in support of such interpretation should be provided. For friends, the practical rule is to decide in advance which one or two comparisons matter and to treat the rest as texture.

10 Why Compare Index Profiles Instead of One Number?

Two friends can share a Full Scale IQ and differ on several indices, and two different Full Scale IQs can hide a shared pattern, so the profile says more about how two people work than the headline number does. In the Cattell-Horn-Carroll framework that ACIS follows, a Full Scale IQ summarizes performance across several broad abilities. The page on the CHC model defines them: comprehension-knowledge (Gc), fluid reasoning (Gf), quantitative knowledge (Gq), visual processing (Gv), short-term working memory (Gwm) and processing speed (Gs). ACIS reports one index per domain a form covers: the Verbal Comprehension Index (VCI), Fluid Reasoning Index (FRI), Quantitative Reasoning Index (QRI), Visual Spatial Index (VSI), Working Memory Index (WMI) and Processing Speed Index (PSI). The page on the six cognitive domains maps the subtests to each.

Four subtests show how different the domains are in practice, using only what their pages state. The Vocabulary subtest, in the verbal domain, asks how precisely a person understands words, and its page notes that word knowledge grows through years of reading, schooling and conversation. The Matrix Reasoning subtest, in the fluid reasoning domain, has a person study a pattern of shapes and infer the missing element from the underlying rule, which its page describes as flexible rule discovery rather than learned verbal knowledge. The Digit Span subtest, in the working memory domain, asks how many items can be held in immediate awareness and, in harder versions, reversed or reordered. The Coding subtest, in the processing speed domain, has a person convert items with a simple symbol key as quickly and accurately as possible. Two friends can differ widely on one of these and not on another, and that is the ordinary case.

For friends the profile matters for a specific reason that the research above supplies. The studies of best friends measured intelligence mostly with vocabulary and knowledge weighted tests, so they say little about whether friends resemble each other on reasoning, memory or speed, and we found no study in the sources we opened that checked. A pair of friends who both love words may land close on the verbal index and far apart on the speed index, and a single Full Scale number averages that story away. Reading the profile is also fairer to the person with the lower verbal index who learned English late, since the technical manual says that verbal and knowledge tasks are affected by English proficiency and educational exposure.

A profile invites a second kind of over-reading, which is to explain every gap. The Standards caution in the comment to Standard 1.14 that evidence for the validity of interpretations of two or more separate scores would not necessarily justify a statistical or substantive interpretation of the difference between them. In plain terms, a friend scoring higher on fluid reasoning than on verbal comprehension has a pattern, and the pattern does not mean a type of person. The sensible use of two profiles side by side is descriptive: it shows which abilities the test sampled and where each person's scores fell.

11 What Should Friends Not Do With the Numbers?

Friends should not rank each other, bet on a guess, retake the test to close a gap, or publish a result, because each of those turns a measurement into a contest it was not built to settle. The first is ranking. The arithmetic above shows that a group of five has a last place by construction, and that the distance between first and last is mostly the shape of the statistics. A ranking invites exactly the simplistic interpretation the Standards warn about in Standard 9.6, where the comment notes that untrained audiences may adopt simplistic interpretations of test results or attribute high or low scores to a single causal factor. A friend group that lists names from highest to lowest has done that.

The second is guessing and betting. People do estimate their own ability, and the estimates are informative but loose. A meta-analysis by Freund and Kasten (2012), covering 154 effect sizes from 41 published studies, estimated the overall relationship between self-estimated and psychometrically measured cognitive ability at r = .33, and found that it was higher when relative scales with clearly specified comparison groups were used. A correlation of .33 is far from useless and far from accurate, so a friend who is confident they would score higher than another is guessing more than they know. The study concerned self-estimates, and we make no claim about estimates of a friend.

The third is retaking. A friend who sees a lower number than expected may want a second go, and scores rise with repetition for reasons unrelated to ability. Scharfen, Peters and Holling (2018) analyzed 174 samples from 122 studies, with 153,185 participants, and report significant retest effects, a standardized gain of 0.33 for the first repetition, and no further score gains after the third administration. Those were healthy adults retaking cognitive ability tests, and their gains reflect practice. A friend who retakes to catch up is measuring familiarity, which is also why friends should not coach each other beforehand.

The fourth is publishing. The gift page says results stay private to the person who takes the test, and the same courtesy applies to a friend's number once they have chosen to share it with you. A score posted to a group chat or a social feed cannot be taken back, and the people most likely to be hurt by a public number are the ones who score below the group average, which statistically includes about half of any group. A low or flat result carries no verdict on a person. The page on scores in the top 5 percent covers the other end of the scale for a friend who lands there, and the same restraint applies.

Finally, do not turn a comparison into a diagnosis. An online, unsupervised test cannot diagnose or rule out anything, and nobody should draw clinical conclusions from a score, a profile or a gap. A real concern about attention, learning or memory belongs with a qualified professional.

12 Where ACIS Sits, and What a Buyer Should Do

ACIS is a normed online battery that gives each friend a Full Scale IQ and index profile on one scale, and it is an online, unsupervised, English only test that is not clinical and not for hiring, school accommodations or admission to high IQ societies, so it suits a friendly comparison and nothing heavier. The strengths are concrete. Every form reports a Full Scale IQ with a 95 percent confidence interval, and the Quick, Optimized and Full Scale forms cost 15, 30 and 50 dollars one time, with no subscription, breaks allowed, progress saved between subtests and 30 days to complete (prices read October 29, 2026). The same form can be given as a gift, and the results stay private to the person who takes it. The guarantee reads, literally, "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." That is a conditional guarantee, and not an unconditional money back one.

The limits are as concrete. The session is online and unsupervised, so conditions vary between people in ways a report cannot see. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) say, in the introduction to the chapter on test administration, that technology and the Internet have made it possible to administer tests in settings where conditions may not be strictly controlled or monitored, and that those who allow a lack of standardization are responsible for providing evidence that it did not affect test taker performance or the comparability of scores. Friends cannot supply that evidence, which is why this page asks them to match the conditions they control and to read a gap against its error. The APA Guidelines for Psychological Assessment and Evaluation (2020) make the related point in Guideline 8 that interpretation should reflect the appropriate normative comparison, situational influences and effort. The test is in English only. It is not a clinical or diagnostic instrument. The documentation is in the technical manual, and this page quotes none of its statistics.

A buyer who has read this far can act in five steps.

1. Decide what the result is for. Curiosity and a shared profile are good reasons, and a ranking is not.
2. Choose one form for everyone. Pick the longest form the least patient person will finish.
3. Run separate sessions. Each person on their own device and email, in a quiet room, with working sound, and no talk about tasks until all are done.
4. Read profiles and gaps together. Compare the indices you share, use an interval or the 2.8 times rule for any gap, and be cautious with a gap under about 10 points, which error and chance often produce on their own.
5. Keep the numbers private unless each person chooses to share.

The next step is small. Start the free trial with no card, which opens a limited set of subtests and costs nothing, then decide on a form. Anyone who wants to give the test to a friend can use a gift code, and anyone who wants to see a report before paying can open an example report.

13 Sources Behind This Page

Every third party fact and figure on this page comes from the documents below, opened or verified on October 29, 2026, and any figure we derived ourselves is labeled in the text as our arithmetic. Prices and policies on commercial and clinic pages change, so check the page before you pay. Figures about ACIS come from the published ACIS pages as read on October 29, 2026.

1. American Educational Research Association, American Psychological Association, and National Council on Measurement in Education (2014). Standards for Educational and Psychological Testing. Washington, DC: AERA. Open access edition: https://www.testingstandards.net/open-access-files.html
2. American Psychological Association (2020). APA Guidelines for Psychological Assessment and Evaluation. Approved by the APA Council of Representatives, March 2020. https://www.apa.org/about/policy/guidelines-psychological-assessment-evaluation.pdf
3. Boutwell, B., Meldrum, R., and Petkovsek, M. (2017). General intelligence in friendship selection: A study of preadolescent best friend dyads. Intelligence, 64, 30 to 35. https://doi.org/10.1016/j.intell.2017.07.002
4. Meldrum, R., Young, J., Kavish, N., and Boutwell, B. (2019). Could peers influence intelligence during adolescence? An exploratory study. Intelligence, 72, 28 to 34. https://doi.org/10.1016/j.intell.2018.11.009
5. Burgess, S., Sanderson, E., and Umaña-Aponte, M. (2011). School ties: An analysis of homophily in an adolescent friendship network. Centre for Market and Public Organisation, University of Bristol, Working Paper 11/267. https://bristol.ac.uk/cmpo/publications/papers/2011/abstract267.html
6. Kendler, K., Turkheimer, E., Ohlsson, H., Sundquist, J., and Sundquist, K. (2015). Family environment and the malleability of cognitive ability: A Swedish national home-reared and adopted-away cosibling control study. Proceedings of the National Academy of Sciences, 112(15), 4612 to 4617. https://doi.org/10.1073/pnas.1417106112
7. Freund, P., and Kasten, N. (2012). How smart do you think you are? A meta-analysis on the validity of self-estimates of cognitive ability. Psychological Bulletin, 138(2), 296 to 321. https://doi.org/10.1037/a0026556
8. Scharfen, J., Peters, J., and Holling, H. (2018). Retest effects in cognitive ability tests: A meta-analysis. Intelligence, 67, 44 to 66. https://doi.org/10.1016/j.intell.2018.01.003
9. Pearson. Wechsler Adult Intelligence Scale, Fifth Edition (WAIS-5), product page, read October 29, 2026. https://www.pearsonassessments.com/en-us/Store/Professional-Assessments/Cognition-%26-Neuro/Wechsler-Adult-Intelligence-Scale-%7C-Fifth-Edition/p/P100071002
10. University of South Florida, Psychological Services Center. Clinic policies, read October 29, 2026. https://www.usf.edu/arts-sciences/departments/psychology/psychological-services-center/clinic-policies/
11. American Mensa. Testing, read October 29, 2026. https://www.us.mensa.org/join/testing/

14 Frequently Asked Questions