The digit symbol substitution test is sensitive, not specific

0 The short answer

The Digit Symbol Substitution Test (DSST) is a timed task from the Wechsler scales, now filed under the name Coding, in which you use a key to match symbols with numbers, and what sets it apart is that it is sensitive to many influences and specific to none of them. Judith Jaeger's 2018 review of its research reports that the score is the number of correct symbols within an allowed time of usually 90 to 120 seconds, and that the test is sensitive to cognitive dysfunction across many conditions but has low specificity for saying which cognitive domain is affected. The Symbol Digit Modalities Test (SDMT) is the sibling task that can be answered aloud, and both are used in research and in clinics by professionals, never as a self test for any condition. ACIS offers Coding as one normed online option beside a psychologist giving the WAIS-5 (it is in the free trial of the Full Scale form, in Optimized at 30 dollars and about 110 minutes, and in Full Scale at 50 dollars and about 175 minutes, prices read October 29, 2026), and the decisive limit is that an online, unsupervised, non clinical Coding score describes speed on one task and cannot tell you why it is high or low.

1 What is the Digit Symbol Substitution Test?

The Digit Symbol Substitution Test is a symbol matching task that began over a century ago as an experiment on learning, entered the Wechsler intelligence scales, and survives in the current WAIS-5 as the subtest called Coding. Jaeger's review in the Journal of Clinical Psychopharmacology (2018) says the test was developed as an experimental tool by researchers seeking to understand human associative learning, that its clinical usefulness was first recognized in the 1940s, and that it is now among the most commonly used tests in clinical neuropsychology. The record of the earliest forms is thin, but the research software library of Millisecond describes a keyboard version "as described by Thorndike (1919)" and says Arthur Otis included the task in his 1918 Otis Group Intelligence Scale (page read October 29, 2026).

The Wechsler scales gave the task its lasting form. According to Jaeger, David Wechsler first described the test's sensitivity to brain damage in 1936, the Wechsler-Bellevue Intelligence Scale of 1939 incorporated it, and during World War II it was reported to distinguish soldiers with brain damage from healthy ones during screening. The original Wechsler Adult Intelligence Scale of 1955 carried it forward with 90 stimulus items. That history explains why the task still carries a clinical reputation, and it is also why this page treats its clinical uses as research and professional uses only.

The names vary, and the variation causes most of the confusion in search results. The task is called Digit Symbol, Digit Symbol Substitution Test or DSST in the research literature, and Joy, Kaplan and Fein titled their study of the WAIS-III in Archives of Clinical Neuropsychology (2004) with the label Digit Symbol Coding. In the current battery it is simply Coding. The Pearson WAIS-5 page (read October 29, 2026) says Coding did not change from the WAIS-IV, that it is administered on paper with a response booklet shared with Symbol Search, and that it is not available digitally on Q-interactive. It describes the demand as encoding and rapidly retrieving newly encoded associations together with graphomotor speed. Anyone who types "wais coding test" is looking for this subtest, and anyone who lands on a computer programming quiz has simply met a different use of the same word. The page on the subtests of the WAIS-5 places Coding among the other tasks of that battery.

What happens on the page is simple to describe, and describing it needs no key. A row of numbers sits above empty boxes. A reference key pairs each number with a simple symbol. The person works across the boxes and, in the paper form, copies the symbol that belongs under each number, as many as possible before time runs out. Jaeger describes it exactly that way: the subject copies the symbol into spaces below a row of numbers. The SDMT reverses the response. Benedict and colleagues in Multiple Sclerosis Journal (2017) describe a key that pairs the digits 1 to 9 with nine symbols, and test takers match symbols to digits, so the answer is a number that is written or spoken and not a drawing. The two tasks share a design, a key, a stream of items and a clock, and they differ in the direction of the response.

This page does not show any key from any test. The keys belong to their publishers, and a key printed on the open web would spoil the task for anyone who later meets it under standard conditions.

2 How is the Digit Symbol Substitution Test scored?

The raw score is the number of correct symbol and number pairings completed in a fixed time, and versions differ in the clock, the response mode and the treatment of errors, so raw counts from different versions are not interchangeable. Jaeger writes that the number of correct symbols within the allowed time, usually 90 to 120 seconds, constitutes the score. The original WAIS had 90 stimulus items, which equals the number of blanks, and in the WAIS-III the test duration was extended from 90 to 120 seconds. The Pearson page read for this article does not state the WAIS-5 time limit, so this page does not either.

The SDMT keeps a 90 second clock. In the description by Benedict and colleagues, test takers have 90 seconds to give as many correct responses as possible, the score is the number of correct responses, ten guided practice items come first, and the whole administration takes no more than 5 minutes. The standard protocol has a written response task followed by an oral one, and most multiple sclerosis studies use only the oral version, which reduces sensory motor demands.

Research software adds a different scoring idea. The Millisecond page lists a task with 5 practice problems and a 2 minute testing period, scored by two metrics: the correct count and the average time in seconds used to make a correct match (page read October 29, 2026).

The ACIS Coding page states its own rules. You have 120 seconds to match as many symbols as possible, you use on-screen buttons to select the number that matches each symbol, and items appear in a continuous stream. Each correct match scores 1 point, and errors are counted but not penalized. Three practice items come first, they are not scored, the clock starts after them, and the timed test uses a new key. No external aids are allowed, which means no calculator, notes, scratch paper or drawing tools. When the subtest ends the page reports a raw score and a scaled score. The rule about errors is worth noting because the neighboring Symbol Search page says the opposite for its task: there is a penalty for wrong answers there, which makes Symbol Search a measure of speed under an accuracy constraint and Coding a measure of rate.

A raw count has no meaning without a reference group. A publisher converts it into a scaled score by comparing it with adults of the same age, and ACIS reports subtest scaled scores from 1 to 19 with a mean of 10 and a standard deviation of 3, alongside percentiles and a 95 percent confidence interval for the Full Scale IQ. The reason the comparison group matters so much for this task in particular is the subject of a later section on age. The page on how IQ scores are normed explains the logic of comparing a person with a reference group.

3 What does the Digit Symbol Substitution Test measure?

Coding measures processing speed (Gs) in the CHC model, but a score also reflects motor speed, visual scanning, attention and a small contribution from learning the key, so it is a mixed task by design. Jaeger lists what good performance requires: intact motor speed, attention and visuoperceptual functions including scanning. She adds that executive functions are involved, since planning and strategizing decide whether a person learns the pairings or looks them up each time, and that working memory is likely needed to hold the task rules in mind. She also notes that Lezak classed the test as a measure of complex attention. None of those is a reasoning demand. Jaeger describes minimal effects of language, culture and education on performance, which fits a task whose symbols and digits carry almost no language.

The most informative work separates the components. In three studies of adults aged 18 to 84, Salthouse in the Journal of Gerontology (1992) found that age-related declines in digit symbol performance were largely independent of the amount of education and of self-reported health, and that they were greatly attenuated after statistical control of a composite measure of perceptual comparison speed. He also found that young and old adults appeared to use similar strategies and spent nearly equal proportions of time writing responses and searching the code table. Jaeger, summarizing the same line of work, reports that copying speed alone explains 72 percent of the variance in the performance difference between young and elderly groups. Her figure is a reviewer's report of earlier studies, so it is best read as a statement that most of what separates age groups on this task is plain speed of executing the pairing.

Memory plays a smaller part. Joy, Kaplan and Fein analyzed the WAIS-III and WMS-III standardization sample in the Archives of Clinical Neuropsychology study and found that speed, measured with a Digit Symbol copying condition, correlated very strongly with Coding, while memory measures correlated more moderately. Even after the variance explained by copying was removed, a statistically significant share of the remainder was explicable by memory functions, and a small portion of the age-associated decline could not be accounted for by copying. In their words the contribution of memory is relatively small but real.

Where does that leave the CHC placement? The public ACIS Coding page relates the task most closely to processing speed (Gs), which it defines as the efficiency of simple, overlearned or low-complexity cognitive operations performed under time pressure, and it adds that this is not the same as deep reasoning. The technical manual assigns Coding the narrow ability that it calls perceptual speed compare, notes that the task is often sensitive to motor fluency and automaticity, and says it is therefore not a pure cognitive reasoning task. A low Coding score beside strong verbal and fluid reasoning, the manual says, may indicate weaker speeded response execution rather than weak reasoning. The page on the CHC model lists the broad and narrow abilities behind those terms, and the page on processing speed as an IQ domain explains why Gs is usually the part of a profile that moves most with age.

One more distinction keeps the construct clean. Coding asks the examinee to compare a symbol with a key and respond, with the same time pressure on every item. Symbol Search, the other subtest in the ACIS processing speed domain, asks the examinee to scan a group for a target and decide whether it is present. The manual treats them as two forms of perceptual speed, compare and search, and says that when Symbol Search and Coding agree the interpretation of the processing speed index is stronger. The Processing Speed Index page describes the index that the two subtests produce together.

4 Why is the DSST sensitive to so many conditions but specific to none?

A task that needs motor speed, scanning, attention, learning and working memory at once will drop when any of those is disturbed, which makes it a good detector of change and a poor identifier of cause. Jaeger's title states the point: the case for sensitivity over specificity. Her conclusion is that the DSST is sensitive to the presence of cognitive dysfunction and to change in cognitive function across a wide range of clinical populations, that it has low specificity to determine exactly which cognitive domain has been affected, and that it offers a practical method to monitor cognitive functions over time in clinical practice. She also writes that without further testing the DSST says little about the nature of a deficit. In plain terms, sensitivity here means that the score moves readily when something affects cognition, and specificity would mean that a movement pointed to one cause. The DSST has the first property and lacks the second.

The evidence Jaeger collects shows how wide the sensitivity is, and every item in it is a group result from a controlled study. A meta-analysis of depression by Snyder, which she cites, pooled 22 studies and 1,904 participants and found a mean effect size of 0.55. An effect size is a standardized gap between group averages, so 0.55 means that the groups differed by a little over half a standard deviation on average, which describes a difference between groups and not a verdict about any person. In trials of the antidepressant vortioxetine, effect sizes on the DSST were 0.51 for 10 milligrams and 0.52 for 20 milligrams in one study and 0.254 for 10 to 20 milligrams in another, and diphenhydramine, an antihistamine, was distinguished from placebo with an effect size of 0.28. In studies of alcohol, scores at a blood alcohol concentration of 0.11 percent were approximately 10 points lower than with placebo. In a sleep study in which 35 participants were randomized to 4, 6 or 8 hours of sleep for 14 days, the 4 hour group differed from the 8 hour group by more than 12 points by the end of the first week.

Read together, those results are what a mixed task produces. A depressed mood, a sedating drug, alcohol and a short night all lower the same score, plausibly by different routes such as slower motor output, narrower attention, poorer visual scanning or slower learning of the key. That is why researchers use the DSST as an outcome in studies where the cause is already known, because participants were randomized or the dose was fixed, and not as a way to find out what is wrong. It is also why the Jaeger review stresses monitoring over time. A professional who knows a person's baseline can see change. A single score from a single sitting has no baseline.

For a reader who is wondering what a particular result means, the same logic applies in the other direction. The public Coding page says a weaker result can reflect motor speed, visual tracking, perfectionism, distraction or unfamiliarity with timed work, and the manual adds motor fluency and automaticity.

5 What does the Symbol Digit Modalities Test add?

The SDMT is the variant that can be answered aloud, which removes writing from the task, and it is the form that multiple sclerosis researchers have studied most, with evidence that comes from groups followed in clinical trials and not from individuals testing themselves. The test was written by Aaron Smith and is published by Western Psychological Services, according to the Buros Center for Testing directory (page read October 29, 2026), which describes it as designed as an early screening of cerebral dysfunction. Benedict and colleagues report that it was first published as a commercial test in 1982 after Smith began piloting the method in the 1960s. They describe a key that pairs the digits 1 to 9 with nine symbols and a 90 second clock, with the written response task followed by an oral one in the standard protocol. Most multiple sclerosis studies use only the oral version, which reduces sensory motor demands, and the examiner records the answers on a scorer form. Because the oral mode needs an examiner, and the materials are the publisher's, the SDMT is not something a person takes alone at home.

Why is it studied so closely? Strober and colleagues, writing for the Multiple Sclerosis Outcome Assessments Consortium in Multiple Sclerosis Journal (2019), used data from 14 multiple sclerosis disease modifying registration trials with a total sample of 12,776 to examine the psychometric qualities of the SDMT and a competing test, the Paced Auditory Serial Addition Test (PASAT). Test-retest reliability, which asks whether the same person scores about the same on two occasions, was an intraclass correlation of 0.85 for the SDMT and 0.86 for the PASAT. The practice effect after three administrations, expressed as an effect size, was 0.10 for the SDMT and 0.27 for the PASAT. The authors concluded that the SDMT should be considered the measure of choice for trials of multiple sclerosis in assessing cognitive processing speed.

Benedict and colleagues give further numbers from their own review. The test manual reported a test-retest coefficient of 0.76 for the oral version in healthy adults, and 34 patients retested after 2 weeks produced a coefficient of 0.97. They conclude that a raw score change of 4 points, or about 10 percent, is clinically meaningful in this research context.

A digital descendant shows where the field is heading. The Konectom cognitive processing speed test is a smartphone based task. In a study in Multiple Sclerosis Journal (2024) with 80 people with multiple sclerosis and 66 matched healthy controls, remote test-retest reliability ranged from 0.67 to 0.87 as intraclass correlations, and the highest correlation with the SDMT was 0.80. The authors designed the outcome measure, which they call substitution time, to account for visuomotor function. It is a different test from the DSST and the SDMT, and it is cited here because it shows the central problem of a digital substitution task: the response must be separated from the motor act of making it. The page on reliability and validity explains terms such as test-retest reliability with examples.

6 How does age change DSST performance?

Digit symbol performance falls steadily across adult life on average, faster than most abilities, which is why a raw count must be read against a same age group and why group trends say nothing definite about one person. Hoyer, Stawski, Wasylyshyn and Verhaeghen ran a meta-analysis in Psychology and Aging (2004) of the effects of age, education and year of measurement on scores from the WAIS and WAIS-R Digit Symbol Substitution Test. Across 141 studies published between 1986 and 2002 the mean standardized difference was -2.07, and age accounted for 86 percent of the variance in a regression model that also included education and year. There was no association between years of education or year of publication and Digit Symbol scores for either younger or older adults. A standardized difference of about two is far larger than the group gaps in the clinical examples of the previous sections, which is our comparison and not the authors', and it explains why the test is hard to read without age norms.

The shape of the curve matters as much as its size. Hartshorne and Germine, in Psychological Science (2015), combined data from 48,537 online participants with an analysis of normative data from standardized IQ and memory tests. In their analysis of the WAIS-III normative sample of 2,450 healthy, cognitively unimpaired Americans aged 16 to 89, digit symbol coding peaked in the late teens, and in later experiments working memory tasks peaked around age 30, significantly later than processing speed. Their abstract says that some abilities peak and begin to decline around high school graduation while others do not peak until the 40s or later. The page on the Stroop test and IQ uses the same age data to show why a speed score at 18 is measured against a group at or near its raw peak.

Why does a symbol matching task decline early and strongly? Salthouse's processing-speed theory in Psychological Review (1996) proposes that increased age in adulthood is associated with a decrease in the speed with which many processing operations can be executed, and that this slowing impairs other cognitive functioning through a limited time mechanism, where relevant operations cannot be executed in time, and a simultaneity mechanism, where the products of early processing are no longer available when later processing is complete. The abstract says several types of evidence are consistent with the theory, among them the shared age-related variance across speed measures and the large attenuation of age effects on other measures after statistical control of speed. A theory is not a fact, but it is the main reason processing speed tasks sit at the front of discussions of cognitive aging, and the page on how IQ changes with age sets out which abilities hold and which fall.

The public ACIS Coding page says the age curve underneath speed starts descending in the twenties while standing among age peers stays close to where it began, so the score is read against an age matched reference group and not a fixed one. The same question sits behind brain age claims, and the sibling page on the brain age test looks at what they can support.

These are averages across many people, and an average is silent about the individual. A slower rate at 60 than at 25 is the expected pattern, and it is not by itself a sign of anything. Anyone who notices changes in everyday thinking, memory or coordination, or whom others have asked about such changes, should speak to a doctor or a licensed psychologist, because a symbol matching task is not designed to answer that question and a normal score would not settle it.

7 Which digit symbol test should you take?

The right option depends on whether you need a clinician's assessment, a research instrument or a normed score you can read against adults, and the options differ most in what they are normed against and in who watches you take them. The table compares the real choices for an adult. Prices and times come from pages opened on October 29, 2026 and can change.

The Pearson WAIS-5 page lists an age range of 16:0 to 90:11 and says the Full Scale IQ now uses seven subtests and takes about 45 minutes. It says nothing about prices or the logistics of a session, so those cells say so rather than guess. A clinician is the right route when the question is clinical, a point the final sections make again, and the page on where to take an IQ test compares in person and online routes.

Free games deserve a plain description and not a verdict. Many are enjoyable, and a score from one is a fact about that game. The question to ask of any page is who the score is compared with, and whether the page says how a count of matches becomes a number. The page on the Human Benchmark IQ test shows what that site's games measure in the same spirit, and the page on the free IQ test explains how to judge free offers. The page on the best online IQ tests compares the paid options with their prices.

ACIS is one concrete option and not the only one, and it sells the product described in the table, so weigh that disclosure as you read. One caution is specific to this task. ACIS Coding asks you to select the number that matches each symbol with on-screen buttons, which is a different response from copying a symbol by hand under a number. It is a Coding style task scored against its own adult norms, and it is not the Wechsler subtest, so a result from one cannot be swapped for a result from the other. What it offers is a normed speed score in the context of a wider profile. The page on the cognitive ability test frames the whole battery, and the page on the IQ test for adults is the entry page for adults who want all of it.

8 How do you read a Coding or DSST score?

A Coding score is meaningful only against an age reference group, and the numbers worth reading are the scaled score or index with its confidence interval and the comparison with Symbol Search, not the count of matches. ACIS reports subtest results as scaled scores from 1 to 19, with a mean of 10 and a standard deviation of 3, and indices and the Full Scale IQ on the familiar scale with a mean of 100 and a standard deviation of 15. Percentiles accompany them. A scaled score of 7 or 13 sits one standard deviation from the mean, which is roughly the 16th or 84th percentile. The page on why a standard deviation of 15 matters shows why the two scales line up, the page on IQ scores versus percentiles explains why they should not be averaged, and the IQ percentile calculator converts one into the other.

Precision is the second layer. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014), Standard 2.3, asks that estimates of reliability and precision be reported for each total score, subscore or combination of scores that is to be interpreted, and its comment warns that subscores may have unacceptably low reliability even when total scores are consistent. As a general illustration only, and not an ACIS figure, suppose a hypothetical subtest score has a standard error of about 1.5 scaled points. A 95 percent band is then about plus or minus 3 points (1.96 times 1.5 is 2.9, our arithmetic), so an observed 10 would be consistent with a true standing from about 7 to 13. A band that wide is the reason a one point difference is rarely a finding.

Differences need more care than single scores. Standard 2.4 asks that when an interpretation emphasizes the difference between two observed scores of an individual, reliability and precision data including standard errors be provided for that difference, and its comment notes that psychodiagnostic inferences are frequently drawn from differences between subtest scores. In the same hypothetical illustration, the standard error of a difference between two independent scores is the square root of 2 times 1.5, about 2.1 scaled points (our arithmetic). The pairing that matters for this task is Coding with Symbol Search. The ACIS Coding page says a weaker result should be compared with Symbol Search to separate visual decision speed from output or transcription demands, and the manual says that when the two agree the processing speed index is on firmer ground. The page on index score differences shows how often sizable gaps occur in ordinary profiles, which is a useful corrective to reading every gap as a strength or a weakness.

Retaking changes what a score means. Strober and colleagues found a practice effect of 0.10 for the SDMT after three administrations in their trial data, and Jaeger notes that learning or practice effects may occur when the same test is given repeatedly. Benedict and colleagues point to incidental learning of the pairings as the reason alternate forms exist. For an unsupervised online task the sensible rule follows from that: treat the first result under good conditions as the observation, and be cautious about reading a higher second result as a change in ability, because familiarity with the format is a competing explanation.

9 How do you take a coding test online?

An unsupervised Coding task measures what the device and the room let it measure, so fix the screen, the pointer and the quiet before you start the clock. The ACIS Coding page tells mobile users to rotate the device to landscape and says the task stays locked horizontally so the whole workspace fits on screen. Responses are made by clicking or tapping on-screen number buttons. The clock is 120 seconds and starts after three unscored practice items, so use the practice items to find a comfortable hand position, and do not start the timed part until you are settled. No external aids are allowed, which means no calculator, notes, scratch paper or drawing tools.

The response method is not a detail. Research designers treat it as part of the measurement: the Millisecond page offers keyboard forms and mouse or touch forms of the same task, and the authors of the Konectom study built an outcome that accounts for visuomotor function, because the act of tapping can differ from the act of comparing. The ACIS pages do not state whether device type changes scores, so the careful assumption is that it might. Our advice, which is inference and not a published finding, is to use the same kind of device and pointer if you take Coding more than once, and to avoid a trackpad you rarely use. The sibling page on taking an IQ test on a phone discusses the device question for the whole battery.

The Standards speak to this setting directly. Standard 6.3 asks that changes or disruptions to standardized administration be documented and reported, and its comment says that testing environments that are not monitored, such as those on the Internet, should meet the standardized administration conditions, and that otherwise the report on scores should note that standardized conditions were not guaranteed. The APA Guidelines for Psychological Assessment and Evaluation (2020), Guideline 8, ask psychologists to be aware of the need for the appropriate normative comparison, situational influences, effort and standardized administration. A clinician controls those. An online test taker has to control them personally.

Practical conditions are modest and they matter. Take the task when you are alert, close other windows and silence notifications so that nothing interrupts a 120 second clock, and use a break between subtests and not inside one: ACIS allows breaks, saves progress between subtests and gives 30 days to complete a purchased form. If something interrupts the timed part, retake the subtest later and do not trust a doubtful number.

10 What can a Coding or DSST score not tell you?

A Coding score describes speed on one structured task, and it cannot tell you why the score is what it is, whether you have any condition, or how well you reason. The central finding of the research is the one that limits the score: the DSST is sensitive to change across a wide range of clinical populations and has low specificity to say which cognitive domain was affected. Because so many ordinary influences lower the number, a low result on its own leaves almost every explanation open. That is our inference from Jaeger's conclusion and from the ACIS Coding page, which lists motor speed, visual tracking, perfectionism, distraction and unfamiliarity with timed work among the things a weaker result can reflect.

The clinical uses in the literature are professional uses. In the studies reviewed above, researchers measured the same people under controlled conditions, often before and after an intervention, or followed patients with a known diagnosis over time with trained examiners. The SDMT evidence from multiple sclerosis trials concerns groups and concerns change within a person who has a baseline. None of that translates into a rule for reading one online score, and this page offers no threshold, cutoff or checklist for any condition. Group statistics are never individual diagnoses.

ACIS does not claim to detect impairment of any kind. If the reason you are reading this page is a worry about attention, memory, coordination or a change that you or others have noticed, the right step is to talk to a doctor or a licensed psychologist, who can look at the whole picture and use instruments meant for that purpose. A normal online score would not rule anything out, and a low one would not rule anything in.

The same restraint applies to a high score. A fast Coding result is one result on one task, and the technical manual says processing speed should not dominate the interpretation of reasoning ability. Speed is a legitimate part of a broad cognitive profile (the page on what IQ measures places it among the other abilities), and it is also the most method specific part: the way a screen, a pointer and a practice routine work can change the number. Standard 1.1 of the Standards for Educational and Psychological Testing makes the general point that no test permits interpretations valid for all purposes, and each recommended interpretation for a given use requires validation. Reading Coding as a measure of speed on this task, and no more, stays inside that boundary.

11 How does Coding fit the rest of your profile?

Coding is read as one half of the processing speed index and as one input to composites that mix speed with other abilities, so its meaning depends on the profile around it. In ACIS the processing speed index (PSI) is built from Symbol Search and Coding. The manual describes PSI as best interpreted as cognitive efficiency under simple visual symbol demands, as a processing speed score and not a direct proxy for general reasoning. The page on the six cognitive domains shows where processing speed sits among the others, and the sibling page on the Processing Speed Index covers the index in depth.

Two composites from the manual's list handle speed in opposite ways. The General Ability Index is described as emphasizing higher order reasoning and knowledge while reducing dependence on working memory and processing speed, so a person with a slower Coding result can see a summary in which speed counts less; the page on the General Ability Index compares it with the Full Scale IQ. The Cognitive Proficiency Index, in the manual's words, combines working memory and processing speed as cognitive proficiency, so it puts speed together with the ability to hold and use information. The working memory side is the subject of the sibling page on the Letter Number Sequencing Test, a task in which the examinee must hold a mixed stream and reorder it.

The research adds a reason to read speed beside memory and not apart from it. Joy and colleagues found that memory contributes a small but real share to Coding scores even after copying speed is accounted for, and Jaeger lists working memory among the abilities the task draws on. A profile in which Coding is low while verbal, reasoning and working memory indexes are strong is therefore a pattern that the manual reads as possible weaker speeded response execution, not weak reasoning, and it is a pattern worth a second look and not a conclusion.

For a reader who wants the profile and not just the speed score, the free trial of the Full Scale form opens five subtests, among them Coding and Symbol Search, and finished trial subtests are kept if the person later unlocks a form that includes them. The technical manual documents the structure of the indexes and composites for readers who want the detail before they buy.

12 Where ACIS sits, and what a buyer should do

ACIS is one commercial option for a normed speed score from a Coding task, it is sold by the publisher of this page, and its limits are as important as its features: it is online, unsupervised, not clinical and available in English only. The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) bear on how any score from this page should be read. Standard 1.1 asks the developer to set out how scores are intended to be interpreted and used, to delimit the population and to describe the construct. Standard 5.8 asks that norms refer to clearly described populations, and the page on the technical manual is where ACIS documents its own. Standard 2.3 asks for reliability and precision estimates for each score that is interpreted, Standard 2.4 asks for them for differences between scores, Standard 6.3 says unmonitored testing should either meet standardized conditions or note that they were not guaranteed, and Standard 6.10 asks that interpretations describe in simple language what the test covers, what scores represent and the precision of the scores.

The APA Guidelines for Psychological Assessment and Evaluation (2020) add three points that apply directly. Guideline 6 asks psychologists to select tools with sufficient validity evidence for their uses, sufficient score reliability and sound psychometric properties. Guideline 7 asks for multiple sources of relevant and reliable information. Guideline 8 asks for attention to the appropriate normative comparison, situational influences, effort and standardized administration. For a buyer those mean three things: read the interval and not the point, read the subtest page for what the task samples, and read the technical manual before leaning on any score. ACIS is not designed for hiring decisions, school accommodations or admission to high IQ societies, and it does not diagnose.

Here is what to do. If your question is clinical, start with a clinician, who can use the Wechsler Coding subtest or the SDMT as part of a wider assessment. If you want to see how your speed on a symbol matching task compares with adults, and whether it is a strength or a valley in your own profile, the free trial of the Full Scale form, which needs no card, is the lowest cost way to take Coding and Symbol Search. If you want only the processing speed domain, the Custom Test at 5 dollars per subtest returns it for 10 dollars across the two subtests, and the Optimized form at 30 dollars adds eleven other subtests. Prices were read on October 29, 2026 and can change, and the refund wording is exact: "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." Use a quiet room, avoid notes, and treat the first score as an observation to confirm and not a finding to act on.

13 Sources Behind This Page

Every figure above is traceable to one of the following, and each is linked at the point where it is used. The two small calculations in the section on reading scores (the 95 percent band and the standard error of a difference) are our own arithmetic on a hypothetical value and are labeled as such. The reported effect sizes, trial results and age trends are group findings from the studies named. Prices and product details were read on October 29, 2026 and can change.

• American Educational Research Association, American Psychological Association and National Council on Measurement in Education. Standards for Educational and Psychological Testing. Washington, DC: AERA, 2014 (Standards 1.1, 2.3, 2.4, 5.8, 6.3 and 6.10).
• American Psychological Association, Task Force on Psychological Assessment and Evaluation Guidelines. APA Guidelines for Psychological Assessment and Evaluation. Approved by the APA Council of Representatives, March 2020 (Guidelines 6, 7 and 8).
• Benedict R, DeLuca J, Phillips G, LaRocca N, Hudson L and Rudick R, for the Multiple Sclerosis Outcome Assessments Consortium. Validity of the Symbol Digit Modalities Test as a cognition performance outcome measure for multiple sclerosis. Multiple Sclerosis Journal, 2017, volume 23, issue 5, pages 721 to 733.
• Buros Center for Testing. Symbol Digit Modalities Test, directory entry. University of Nebraska-Lincoln. Read October 29, 2026.
• Hartshorne J and Germine L. When Does Cognitive Functioning Peak? The Asynchronous Rise and Fall of Different Cognitive Abilities Across the Life Span. Psychological Science, 2015, volume 26, issue 4, pages 433 to 443.
• Hoyer W, Stawski R, Wasylyshyn C and Verhaeghen P. Adult Age and Digit Symbol Substitution Performance: A Meta-Analysis. Psychology and Aging, 2004, volume 19, issue 1, pages 211 to 214.
• Jaeger J. Digit Symbol Substitution Test: The Case for Sensitivity Over Specificity in Neuropsychological Testing. Journal of Clinical Psychopharmacology, 2018, volume 38, issue 5, pages 513 to 519.
• Joy S, Kaplan E and Fein D. Speed and memory in the WAIS-III Digit Symbol Coding subtest across the adult lifespan. Archives of Clinical Neuropsychology, 2004, volume 19, issue 6, pages 759 to 767.
• Millisecond Software. Digit Symbol Substitution Task (Inquisit library page). Read October 29, 2026.
• Pearson Assessments. Wechsler Adult Intelligence Scale, Fifth Edition (WAIS-5) product page. Read October 29, 2026.
• Salthouse T. What Do Adult Age Differences in the Digit Symbol Substitution Test Reflect? Journal of Gerontology, 1992, volume 47, issue 3, pages P121 to P128.
• Salthouse T. The processing-speed theory of adult age differences in cognition. Psychological Review, 1996, volume 103, issue 3, pages 403 to 428.
• Scaramozza M, Chiesa P, Zajac L, Sun Z, Tang M, Juraver A and colleagues. Konectom cognitive processing speed test enables reliable remote, unsupervised cognitive assessment in people with multiple sclerosis: Exploring the use of substitution time as a novel digital outcome measure. Multiple Sclerosis Journal, 2024, volume 30, issue 9, pages 1193 to 1204.
• Strober L, DeLuca J, Benedict R, Jacobs A, Cohen J, Chiaravalloti N, Hudson L, Rudick R and colleagues. Symbol Digit Modalities Test: A valid clinical trial endpoint for measuring cognition in multiple sclerosis. Multiple Sclerosis Journal, 2019, volume 25, issue 13, pages 1781 to 1790.

14 Frequently Asked Questions