Alcohol and IQ: which way the causal arrow runs

0 The Short Answer

Sustained heavy drinking is associated with a measurable decline in tested ability, but a large share of the difference you see between heavy drinkers and everyone else was already present before the drinking became a disorder. That is not a way of excusing alcohol. It is the finding of the only design that can separate the two, and it changes what the rest of the literature means.

Gronkjaer and colleagues, publishing in Alcoholism: Clinical and Experimental Research in 2019, followed 2,499 Danish men who took the military conscription board's intelligence test in young adulthood and then took the same test again in late midlife. The 8 percent who later acquired an alcohol related hospital diagnosis had scored 95.0 at conscription against 100.5 for everyone else, a gap of 5.5 points that existed before any diagnosis. They then declined further and faster, losing 8.5 points against 4.8 for the rest. Working from the means the authors report, the midlife gap is about 9.2 points, of which roughly 60 percent was already there at conscription.

The honest complication is that neither half of that sentence cancels the other. The prior difference is large, and it means cross sectional comparisons of drinkers and non drinkers overstate what drinking did. The additional decline is also real, survived adjustment for education, smoking, baseline score and comorbidity, and points at genuine harm. Meanwhile the popular mechanism, that alcohol kills brain cells, is false as stated, while the damage it does cause runs through pathways most people have never heard of.

1 The Study That Reverses the Arrow

Almost every widely cited finding about alcohol and intelligence is cross sectional, which means it compares people who drink with people who do not at a single moment and cannot tell you which came first. Gronkjaer and colleagues solved that with a design Denmark makes possible and almost nowhere else does: universal male conscription with a standardized cognitive test on file, plus national health registries that record hospital diagnoses decades later.

The analytic sample was 2,499 men drawn from the Lifestyle and Cognition Follow up study of 2015. Each had taken the conscription board's intelligence test as a young man and sat the identical test again in late midlife. Alcohol related hospital diagnoses came from the registries rather than from self report, and a second exposure, self reported treatment for alcohol problems, was recorded at follow up and affected about 10 percent of the sample. The results were the same for both definitions.

Two numbers matter and they point in opposite directions. Men who would later carry an alcohol related diagnosis averaged 95.0 at conscription against 100.5 for the rest, significant at p below .001. They then lost 8.5 points by late midlife against 4.8 for the comparison group, again significant at p below .001, and the extra decline survived adjustment for year of birth, retest interval, baseline score, education, smoking, alcohol consumption, and psychiatric and somatic comorbidity. The 4.8 points the comparison group lost are the ordinary ones, and they are why a decline figure means nothing quoted alone: in the Lothian cohorts, rank order held at .66 between age 11 and age 79 while raw performance climbed and then fell, so the fall that happens with no exposure at all is a measured quantity rather than a blank. The authors' own conclusion is the one to keep: the low mean seen in people with alcohol related disorders is probably the product of both lower premorbid ability and greater subsequent decline.

Read the limits before you read the meaning. The sample is Danish men only, so nothing here is established for women. Hospital diagnoses capture the severe end of the distribution rather than ordinary heavy drinking. The conscription test is a group administered instrument, not a Wechsler adult scale, so its points are not interchangeable with Wechsler points even though both are scaled to a mean of 100. And a 5.5 point premorbid gap is a group average, not a property of any individual, a distinction that matters throughout score interpretation.

2 Alcohol Does Not Kill Brain Cells, and That Is Not Reassurance

The single most repeated claim about drinking and the brain has been directly tested and it failed. Jensen and Pakkenberg, publishing in The Lancet in 1993, used stereological counting on brains obtained at autopsy from 11 chronic alcoholic men and 11 control men matched on age and height. Total neocortical neuron number was 23.4 billion in the alcoholics and 23.2 billion in the controls. There was no difference. Neocortical volume did not differ either.

Two later studies from the same laboratory tradition pushed in the same direction. Tang, Pakkenberg and Nyengaard reported in Brain Research in 2004 that the total volume, total length and mean diameter of myelinated fibers in subcortical white matter did not differ between alcoholic and control subjects. Fabricius and colleagues reported in Alcohol and Alcoholism in 2007, comparing 11 alcoholics with 10 controls, that neuronal cell volumes and total glial cell numbers in the neocortex were also unaffected.

Now the part that does not get quoted. The same 1993 paper found an 11 percent reduction in the white matter volume to weight ratio, significant at p equal to .013, and a 30 percent reduction in the archicortex, the older cortical tissue that includes the hippocampal formation, significant at p equal to .028. Ventricular volume was 26 percent larger in the alcoholics, though that difference was not statistically significant. The authors' interpretation was not that the brain is fine. It was that the damage sits in the connections rather than in the cell bodies, which is a different claim with different consequences.

Note the sample sizes before you take any of this as settled. Eleven brains per group is very small, these were men, and the counts covered the neocortex and archicortex rather than every structure in the head. What the myth correction buys you is precision about mechanism, not a clean bill of health. If you want a companion case of a durable and equally wrong brain claim, the ten percent of the brain story has the same structure, and the site keeps a wider list of myths about testing that survive correction.

3 Being Drunk, Being Hungover and Being a Heavy Drinker

Three different questions get answered with the same word, and mixing them is the main reason the public literature reads as contradictory. Acute intoxication is a state that lasts hours. The next morning is a second state that lasts most of a day. Chronic exposure over decades is the question people actually mean when they ask whether drinking lowers intelligence, and it is the only one of the three where a lasting change in ability is even on the table.

The acute answer is not in dispute and never has been. Alcohol at intoxicating doses degrades attention, working memory, motor control and judgement while it is present, which is the entire basis of drink driving law. Wetherill and Fromme, reviewing 26 clinical studies from 2010 to 2015 in Alcoholism: Clinical and Experimental Research in 2016, note that memory loss for part or all of a drinking episode is reported by roughly half of drinkers, and that individual differences rather than dose alone predict who experiences it.

The next day is more interesting because most people assume they have recovered. Gunn and colleagues, in a systematic review and meta analysis published in Addiction in 2018 covering 19 studies and 1,163 participants, restricted the hangover condition to a blood alcohol concentration below 0.02 percent, meaning the alcohol was effectively gone. Impairments persisted: Hedges' g of 0.64 for short term memory, 0.59 for long term memory, 0.47 for sustained attention and 0.66 for psychomotor speed. An effect of 0.6 standard deviations, if it landed on a composite scaled with a standard deviation of 15, would move a score by roughly nine points. That is arithmetic on their effect size, not their finding, and the tasks were memory and attention rather than a full battery. In CHC terms the load falls on Gs, processing speed, and Gwm, working memory, which is part of why several subtests are timed at all. Stored knowledge was not among the outcomes the review examined, so nothing in it speaks to Gc, comprehension knowledge, and the distinction between those domains is laid out under processing speed as a measured domain and the split between fluid and crystallized ability.

If your question is the practical one, whether to sit an assessment the morning after a heavy night, that is covered separately in the site's guidance on what to do in the day before a test, and the same logic applies to short sleep, which is dealt with in the page on sleep loss and test day performance. Neither of those is a question about your intelligence. They are questions about whether a particular administration will represent you.

4 What the Large Imaging Cohorts Found

The strongest evidence that even moderate drinking is associated with something is structural, and it comes from samples large enough that the confidence intervals are narrow. Daviet and colleagues, publishing in Nature Communications in 2022, analyzed multimodal imaging from 36,678 generally healthy middle aged and older adults in UK Biobank, controlling for a long list of potential confounds. Alcohol intake was negatively associated with global brain volume, with regional gray matter volumes and with white matter microstructure.

Their translation into age equivalents is what made the study travel. Moving from zero to one daily UK unit corresponded to about 0.5 years of additional aging in gray matter volume; from one to two units, about 2.0 years; from two to three, about 3.5 years; from three to four, about 4.9 years. The associations were already detectable in people averaging one to two units a day, which is why the study is usually summarized as finding no safe threshold for brain volume. Note that the relationship is steeper as intake rises, so the last increment costs more than the first.

Topiwala and colleagues, publishing in the BMJ in 2017, came at it with a longer runway. Their Whitehall II imaging substudy tracked weekly alcohol intake across 30 years in 550 community dwelling adults, then scanned them. Compared with abstainers, people drinking over 30 units a week had 5.8 times the odds of hippocampal atrophy, with a confidence interval from 1.8 to 18.6. Even 14 to 21 units a week carried 3.4 times the odds of right sided hippocampal atrophy, interval 1.4 to 8.1. Light drinking, one to under seven units a week, showed no protective effect over abstaining.

The 2022 iron paper deserves its qualifier stated plainly. Its Mendelian randomization analyses, which use genetic variants as instruments to approximate a natural experiment, gave only weak support for a causal reading, and the associations that appeared did not survive correction for multiple testing. Its own authors flag that the imaging markers can reflect myelin as well as iron and that intake was self reported.

5 Brain Volume Is Not the Same Thing as Measured Ability

A study can find a robust association between drinking and brain structure and still be almost silent about intelligence, because the link between brain volume and test performance is much weaker than most readers assume. Pietschnig and colleagues, in a meta analysis published in Neuroscience and Biobehavioral Reviews in 2015, pooled 88 studies covering 148 samples and more than 8,000 individuals. The association between in vivo brain volume and IQ was r equal to .24, which means brain volume accounts for about 6 percent of the variance in scores. They also found evidence that the literature had overstated it, that reported effects declined over time, and they warn explicitly against treating brain size as a proxy for ability.

The cleanest demonstration of the gap sits inside the Whitehall II cohort itself. Sabia and colleagues, publishing in Neurology in 2014, followed 5,054 men and 2,099 women with a mean age of 56 across repeated cognitive testing. In men, there was no difference in decline between abstainers, quitters, and light or moderate drinkers below 20 grams a day. Only consumption at or above 36 grams a day was associated with faster decline, and the differences were modest: 0.10 standard deviations on the global score across ten years, with memory the worst affected at 0.16.

Put the two Whitehall papers next to each other and the lesson is methodological rather than nutritional. The same broad cohort produced a strong structural signal at moderate intake and no measurable cognitive difference at moderate intake. Structural change is easier to detect than functional change, imaging endpoints are more sensitive than test scores, and a difference in tissue does not automatically become a difference in what a person can do. If you want the general version of that argument, the site's page on measurement quality and what validates a score covers why an endpoint has to be validated against behavior before it means anything about behavior.

None of this makes the imaging results irrelevant. It makes them evidence about brains rather than evidence about IQ. Anyone converting a UK Biobank volume coefficient into a claim about lost points is doing something the data does not license, and the same caution applies to reading a single index score as though it were the whole of the general factor these tests load on. Substituting a convenient endpoint for the contested one is the standard failure in exposure research, and its clearest current example sits outside this topic entirely: the claim that generative artificial intelligence lowers intelligence rests on three studies measuring EEG connectivity, survey responses and perceived mental effort, which is a whole literature standing in for an outcome nobody administered.

6 Dose, Pattern, and What a Unit Actually Is

Most confusion about thresholds is a units problem, because the British studies that dominate this literature count in UK units and American readers count in standard drinks, and the two are not the same size. The NHS defines one UK unit as 10 milliliters or 8 grams of pure alcohol, with a guideline of no more than 14 units a week. The National Institute on Alcohol Abuse and Alcoholism defines a US standard drink as 0.6 fluid ounces or 14 grams of pure alcohol, with binge drinking set at the pattern that brings blood alcohol to 0.08 percent, typically five drinks for men or four for women in about two hours, and heavy drinking at 15 or more drinks a week for men and 8 or more for women.

Once the units line up, a second Gronkjaer paper answers the question people rarely think to ask. Writing in Alcohol and Alcoholism in 2019, the same group analyzed 2,498 men from the same Danish follow up and separated average weekly consumption from pattern. Men drinking more than 28 units a week declined more than men drinking 1 to 14, with a coefficient of 3.6 points for the 29 to 35 unit band. But each additional year of weekly extreme binge drinking, meaning 10 or more units on a single occasion, predicted a 0.12 point decline, and binge frequency explained more variance in score change than average weekly volume did. When both were entered together, the binge term held and the average weekly term lost significance.

That is the practically useful result in this whole literature, and it is the one the headlines skip. How the same weekly total is distributed appears to matter more than the total itself. Whether that generalises beyond Danish men is not established, and the exposure was self reported decades after the fact, which is exactly the kind of recall that degrades.

7 Why Moderate Drinkers Look Good in Cross Sectional Data

The best documented reason moderate drinkers outperform abstainers on cognitive tests is that they were different people before they were drinkers. Corley and colleagues, publishing in Neuropsychology in 2011, had something almost nobody else has: intelligence test scores from age 11 for 922 members of the Lothian Birth Cohort 1936, retested at about 70. Moderate to substantial drinking, more than two units a day, was associated with better cognitive performance than low level drinking or abstaining. Then they adjusted for childhood IQ and adult social class, and most of the associations were removed or substantially attenuated. What survived was small: a residual link to memory in both sexes and to verbal ability in women.

A second and stranger confound runs the other way. Batty and colleagues, publishing in the American Journal of Public Health in 2008, followed 8,170 members of the 1970 British Cohort Study and found that higher mental ability at age 10 predicted more problem drinking at age 30, not less. Per standard deviation of childhood ability, the odds ratio was 1.38 for women, confidence interval 1.16 to 1.64, and 1.17 for men, interval 1.04 to 1.28. Higher childhood ability also predicted a higher average intake and more frequent drinking. Whatever else is going on, the people who drink more are not a random draw from the low ability end of the distribution. That finding sits oddly beside the broader pattern in childhood ability and how long people live, where higher early scores predict better outcomes almost everywhere else.

The third confound is the abstainer category itself. A group labelled non drinkers in a middle aged sample contains lifelong abstainers, people who stopped for religious or cultural reasons, and people who stopped because they became unwell, including people who stopped because their drinking had already damaged them. Any comparison against that mixture inherits its composition. Sabia and colleagues saw the fingerprint of it in 2014: among women, ten year abstainers declined faster on the global score than women drinking 0.1 to 9.9 grams a day, which is not plausibly a pharmacological benefit of small amounts of wine.

This pattern is not unique to alcohol. The site's page on breastfeeding and later scores works through the same problem, where the exposure is heavily confounded with maternal ability and social position, and the page on exercise and cognitive outcomes has a similar shape. Recognizing the structure once makes every subsequent lifestyle claim easier to read.

8 How Alcohol Actually Damages the Brain

The mechanisms that carry real alcohol related brain damage are nutritional, vascular and structural rather than a simple toxic killing of neurons, and the most severe of them is preventable with a vitamin. Chronic heavy drinking interferes with the absorption and utilization of thiamine, and heavy drinkers frequently eat badly on top of that. The result can be Wernicke's encephalopathy, an acute neuropsychiatric emergency, which when untreated or undertreated can leave Korsakoff's syndrome, a profound and largely permanent amnesia.

Sechi and Serra, reviewing the condition in The Lancet Neurology in 2007, make two points that matter for anyone reading this as a risk question. Autopsy based studies indicate the disorder remains greatly underdiagnosed in both adults and children, meaning the clinical incidence understates the true one. And the prevention and treatment of Korsakoff's syndrome in alcohol misuse depends on adequate parenteral thiamine, which is a dosing question with a right answer rather than a matter of general nutrition. This is the single clearest case in which alcohol produces catastrophic cognitive loss, and the causal chain runs through vitamin deficiency rather than through ethanol dissolving cells.

At population scale the association with dementia is large. Schwarzinger and colleagues, publishing in The Lancet Public Health in 2018, analyzed 31,624,156 adult hospital discharges in metropolitan France between 2008 and 2013, of which 1,109,343 carried a dementia diagnosis. Alcohol use disorders were the strongest modifiable risk factor for dementia onset, with adjusted hazard ratios of 3.34 for women, interval 3.28 to 3.41, and 3.36 for men, interval 3.31 to 3.41. Among the 57,353 cases of early onset dementia, 38.9 percent were alcohol related by definition and a further 17.6 percent carried an additional alcohol use disorder diagnosis.

Read the discharge study for what it is. Hospital records capture people ill enough to be admitted, so this is evidence about severe alcohol use disorder rather than about ordinary drinking, and diagnostic coding for dementia is imperfect. What it establishes is that the top of the exposure range is genuinely dangerous, which is compatible with everything above and rules out any reading of the myth correction as an all clear. Alcohol use disorder also travels with other psychiatric conditions, and the site's treatment of how ability relates to mental health explains why comorbidity makes single cause attributions unreliable. Ability differences of this magnitude are not the sort of thing an online administration is designed to detect.

9 Prenatal Exposure Is a Different and Stronger Literature

Alcohol exposure before birth produces effects on cognitive development close to an order of magnitude larger than anything documented for adult drinking, and the two questions should never be answered with the same sentence. The developing brain is a different organ under a different constraint, the exposure is involuntary, and the outcome is a developmental trajectory rather than a decline from an established baseline.

Streissguth and colleagues, publishing in JAMA in 1991, described 61 adolescents and adults with fetal alcohol syndrome or fetal alcohol effects. The average IQ was 68, with a wide range, and average academic functioning sat at second to fourth grade level with arithmetic the most characteristic deficit. That average is more than two standard deviations below the population mean, which is a categorically different magnitude from the 3.7 point extra decline seen in Danish conscripts who developed alcohol related disorders as adults.

At lower exposures the picture is smaller and messier. Flak and colleagues, in a meta analysis published in Alcoholism: Clinical and Experimental Research in 2014, screened 1,593 articles and included 34 cohort studies. Any binge exposure during pregnancy was associated with poorer child cognition, Cohen's d of negative 0.13 with an interval from negative 0.21 to negative 0.05, across 8 studies of 10,000 children. Moderate exposure was associated with poorer behavior, d of negative 0.15, interval negative 0.28 to negative 0.03, across 3 high quality studies of 11,900 children. Their other result is the instructive one: mild to moderate exposure showed a small positive association with cognition, d of 0.04, which did not survive the exclusion of one large study and which nobody interprets as a benefit. It is what residual social confounding looks like when it is left in the model.

Two further framings worth keeping straight. A group mean of 68 describes a group, and individual outcomes within fetal alcohol spectrum conditions vary enormously, which is the same caution that applies whenever a diagnosis is paired with a score, as on the page covering attention conditions and test performance. And the size of a prenatal effect says nothing about heritability, a distinction the page on what heritability actually means takes apart in detail.

10 What Happens When People Stop

Some of what heavy drinking takes comes back, and it comes back faster than most people expect, but the evidence base is small and the recovery is partial. Bartsch and colleagues, publishing in Brain in 2007, scanned 15 recently detoxified, uncomplicated alcoholic patients at admission and again after six to seven weeks of sobriety, alongside 10 age and gender matched healthy controls who showed no change. Automatically detected global brain volume gain averaged nearly 2 percent, concentrated around the superior vermis and the perimesencephalic, periventricular and frontal brain edges.

They also measured brain chemistry and behavior. Cerebellar choline and frontomesial N acetylaspartate rose significantly over the same weeks, and the gains in N acetylaspartate were associated with improved performance on the d2 test of attention. Because water content, creatine concentrations, serum electrolytes and red cell indices stayed stable, the authors argue the change is not simply rehydration. Their conclusion is that the adult brain, and its white matter in particular, retains genuine capacity for regrowth.

Hold that against the sample size. Fifteen patients and ten controls, six to seven weeks, one attention test. This is a demonstration that early recovery is measurable, not an estimate of how much anyone gets back. The 1993 autopsy work points the same way for a different reason: if the cell bodies are preserved and the loss is in connections, then a route back exists in principle that would not exist if the neurons themselves were gone.

The exception is important and it is the severe one. Korsakoff's amnesia, once established, does not typically reverse, which is why the thiamine question in the previous section is a prevention question rather than a treatment question. It is also a memory disorder specifically rather than a general lowering, a distinction the page on how memory relates to tested ability works through. So the honest summary has two halves that people tend to hear as one: early structural and metabolic recovery after stopping is real and documented, and the worst alcohol related cognitive outcomes are permanent. Neither half licenses a prediction about a specific person.

Whether measured ability itself improves with abstinence is a separate empirical question from whether tissue volume does, and it runs into everything the site covers under whether tested ability can be trained: practice effects on a retest, regression toward the mean when the first score was unusually low, and the difficulty of attributing any change to one cause. The Bartsch attention finding is a correlation in 15 people, not a demonstration that a composite score moves.

11 What One Score Can and Cannot Tell You

People arrive at this question wanting to test themselves and find out, and a single administration cannot answer it, for reasons that are about measurement rather than about alcohol. A score describes where you stand now relative to a reference frame. It carries no information about where you stood before, and without a before there is no change to attribute to anything.

The numbers make the problem concrete. The ACIS Full Scale IQ composite has a reliability of .9886 by omega and a g loading of .958, and the standard error of measurement for the Full Scale composite is about 1.60 IQ points, established on a technical analysis set of 2,750 complete records within an adult reference frame of 3,243 English speaking records aged 16 to 90. The higher order model fits well, with CFI .9761, TLI .9726, RMSEA .0406, SRMR .0217 and a chi square of 916.703 on 166 degrees of freedom. Those are strong figures for an unsupervised instrument. They still do not solve the inference problem, because the error of a difference between two administrations compounds the error in each of them, and a difference score is therefore less precise than either score alone.

Three further obstacles sit on top of that. Retaking a battery produces practice effects that inflate the second score independently of any real change. Regression toward the mean pulls an unusually low first score upward on retest whether or not anything improved. And there is no way to isolate one habit as the cause of a difference when everything else in a life also changed across the interval. The site's page on how norms are constructed explains why the reference frame itself moves too, and the page on what makes an unsupervised score representative covers the conditions under which any single administration is worth reading at all.

ACIS is an unsupervised online instrument. It is not a clinical or diagnostic tool, it cannot be used for diagnosis, hiring, accommodations or high IQ society admission, and its normative sample is self selected rather than census based, so the reference frame is a modelled adult frame rather than a representative one. Full methods sit in the technical manual, and the way the composite relates to its parts is set out under what the Full Scale composite summarizes. What the assessment can do is describe a current profile across the CHC domain structure, reporting VCI, FRI, QRI, VSI, WMI and PSI alongside the composite, which is a different and more modest claim than detecting a cause.

12 Reading Your Own Situation Without Overreading It

The evidence supports different answers depending on which version of the question you are actually asking, and most disappointment with this literature comes from expecting an answer to a question nobody has studied. The table below maps the common cases onto what the research above will and will not carry. It is descriptive, not advisory, and none of it substitutes for a clinician.

One framing helps more than any of the rows. If you want to know whether drinking has cost you something, the comparison that would answer it is you against your own earlier score, and almost nobody has one. The comparison available instead is you against a reference frame, which answers a different question: where you currently stand, on the mapping from score to percentile and against the population average.

13 Limits, Open Questions and the Standards This Page Is Written Against

The design that would settle this question cannot ethically be run, which is why the literature looks the way it does. Randomly assigning people to decades of heavy drinking is not permissible, so causal inference has to come from natural experiments, from genetic instruments, or from prospective cohorts that measured ability before the exposure. Only the third of those has produced a clear result here, and it produced two results at once. The cannabis literature pushed the second route much further, comparing twins discordant for use across four genetically informed samples so that shared genes and shared upbringing drop out in a single step, and the four discordant twin samples assembled for cannabis are precisely the design the alcohol evidence has never had.

The genetic route is the most promising and currently the least conclusive. Topiwala and colleagues found, in PLOS Medicine in 2022, that Mendelian randomization gave only weak support for a causal effect of alcohol on brain iron, with associations not surviving multiple comparison correction. Zhou and colleagues, in a genome wide meta analysis of problematic alcohol use across 435,563 individuals published in Nature Neuroscience in 2020, reported that Mendelian randomization suggested causal effects on liability to problematic alcohol use from substance use, psychiatric status, risk taking behavior and cognitive performance. That last term is the reverse arrow again, arriving from a completely different method.

Several things remain genuinely unknown and should be labelled as such. Whether the pattern finding generalises beyond Danish men has not been tested. Whether the structural changes seen at one to two daily units ever produce a measurable cognitive difference has not been demonstrated. Whether recovery after abstinence extends to composite ability rather than tissue volume has not been measured. And no study has compared drinkers and non drinkers on a full modern adult battery with pre exposure baselines, which is the study everyone is implicitly citing and nobody has run.

This page is written against the Standards for Educational and Psychological Testing, published jointly in 2014 by the American Educational Research Association, the American Psychological Association and the National Council on Measurement in Education, whose open access materials set out the validity and fairness requirements a score claim has to meet. Those standards require that any interpretation be supported by evidence for the specific inference being drawn. Reading an ACIS score as a measure of alcohol related harm is exactly such an inference, and the evidence for it does not exist, which is why this page does not offer it. APA testing standards also require that a test not be used for purposes it was not validated for, which for an unsupervised instrument excludes clinical and diagnostic use entirely. For the wider context of how instruments are chosen and compared, see supervised and online testing side by side and, for researchers who need administered links rather than self testing, the ACIS Professional workspace.

14 Frequently Asked Questions