What is the GARS-3? The Gilliam Autism Rating Scale explained

The GARS-3, formally the Gilliam Autism Rating Scale, Third Edition, is a "norm-referenced" behavior rating scale, meaning its scores compare an individual's ratings to those of a large standardization sample. Written by James E. Gilliam and published by PRO-ED in 2014, it is completed in about 5 to 10 minutes by a parent, teacher, or clinician who knows the person well, and it covers ages 3 through 22. The GARS-3 is a screening and evidence-gathering instrument. It is neither a diagnostic test nor an IQ test, and only a qualified clinician can diagnose autism spectrum disorder. This article explains how the scale works, what its six subscales cover, how the Autism Index and probability levels are interpreted, and what the research literature says about its strengths and limitations.

The GARS-3 is a questionnaire, and that is worth stating plainly. Nobody sits the child down to solve puzzles or answer questions. Instead, an adult who has had recent, sustained contact with the individual rates 58 statements describing behaviors associated with autism spectrum disorder. According to the publisher, PRO-ED, the items are based on the diagnostic criteria for autism spectrum disorder adopted by the American Psychiatric Association in 2013 and published in the DSM-5.

Each item is rated on a 4-point scale, from "not at all like the individual" (0) to "very much like the individual" (3). Higher ratings indicate that a behavior is more characteristic of the person being rated. Because the rating takes only about 5 to 10 minutes, the GARS-3 is one of the quickest autism-related instruments in wide use, which is a large part of its appeal to school teams and busy clinics.

PRO-ED reports that the scale was normed on 1,859 individuals with autism spectrum disorder, with data collected in 2010 and 2011 and demographic characteristics keyed to U.S. Census figures. That norming approach matters for interpretation: scores describe how closely someone's rated behavior resembles that of the normative autism sample, and this resemblance is treated as evidence about the probability of autism, never as a diagnosis by itself.

The 58 items are grouped into six subscales. Peer-reviewed descriptions of the instrument, including a psychometric study by Samadi and colleagues published in the journal Children in 2022, list the item counts as follows.

• Restrictive/Repetitive Behaviors (13 items): stereotyped behaviors, unusual preoccupations, and resistance to change in routines.

• Social Interaction (14 items): responding to others, initiating contact, and engaging in back-and-forth social exchange.

• Social Communication (9 items): nonverbal and pragmatic communication, such as understanding gestures, facial expressions, and social cues.

• Emotional Responses (8 items): unusual or intense emotional reactions to everyday situations.

• Cognitive Style (7 items): distinctive patterns of interest, attention, and thinking; rated only for individuals who use speech.

• Maladaptive Speech (7 items): atypical speech patterns such as echoed or idiosyncratic language; also rated only for individuals who use speech.

Because the last two subscales require speech, the GARS-3 produces two composite scores. Autism Index 4 is computed from the first four subscales and can be used for anyone, including people who do not speak. Autism Index 6 adds the two speech-related subscales for verbal individuals.

The GARS-3 converts subscale ratings into "standard scores," which place performance on a common numerical scale, and "percentile ranks," which show the percentage of the norm sample scoring at or below a given level. The composite Autism Index is the headline number, and the manual pairs it with a probability category and a suggested level of support.

Researchers led by Xuejun Kong, writing in Frontiers in Psychiatry in 2020, summarize the four levels as follows.

• Autism Index 54 or below (level 0): autism is classified as "unlikely" based on the ratings.

• Autism Index 55 to 70 (level 1): autism is "probable," with minimal support indicated.

• Autism Index 71 to 100 (level 2): autism is "very likely," with substantial support indicated.

• Autism Index 101 or above (level 3): autism is "very likely," with very substantial support indicated.

Two cautions belong next to any table like this. First, these categories describe the probability that rated behavior resembles that of the autism norm sample; they do not confirm or rule out a diagnosis. Second, all rating scales carry "measurement error," the unavoidable imprecision in any score, so a single Autism Index should be read as an estimate rather than a fixed fact about a person.

According to the Centers for Disease Control and Prevention, there is no medical test, such as a blood test, that diagnoses autism spectrum disorder. A formal evaluation is conducted by trained specialists, such as a developmental pediatrician or child psychologist, who combine direct observation, structured testing, interviews with parents, questionnaires, and a careful developmental history.

In that process, the GARS-3 plays a specific and limited role: it captures how an informant perceives the individual's everyday behavior, in a structured format that can be compared against norms. Evaluators often gather ratings from multiple informants, for example a parent and a teacher, because behavior can look quite different at home and at school. Other informant measures serve a similar purpose, such as the Social Responsiveness Scale, which we describe in our article on what the SRS-2 is. A comprehensive evaluation also typically includes an adaptive behavior measure such as the one covered in our guide to the Vineland-3, along with cognitive and language testing where appropriate.

The value of converging evidence is hard to overstate here. When observation ratings, clinician-administered measures, and developmental history all point in the same direction, a diagnostic team can be far more confident than any single questionnaire allows.

The GARS-3 has genuine strengths. It is fast, inexpensive relative to clinician-administered instruments, aligned with current DSM-5 criteria, and usable across a wide age span. PRO-ED reports strong "reliability," the consistency of scores, with internal consistency coefficients above .85 for subscales and above .93 for the Autism Indexes, and test-retest values above .80 and .90 respectively. The publisher also reports high accuracy figures from its own validity studies, with "sensitivity" (the proportion of true cases the test flags) and "specificity" (the proportion of non-cases it correctly clears) both reported at .97.

The independent literature urges more caution. In a 2002 study in the Journal of Autism and Developmental Disorders, Mikle South and colleagues examined the original GARS in 119 children with carefully established autism diagnoses and found that it consistently underestimated the likelihood of autism: the sample's mean Autism Quotient was 90.1, well below the reference mean of 100, and sensitivity was only .48. That study concerned the first edition, and later editions were substantially revised, but it remains a landmark caution about relying on informant ratings alone.

Reviews of the current edition are more favorable while still noting limits. In a 2017 test review in the Journal of Psychoeducational Assessment, Benjamin Karren concluded that the GARS-3 made significant strides over earlier editions and shows sound psychometric properties overall, yet he recommended caution when assessing individuals ages 20 to 22 and individuals from minority groups, reflecting concerns about how well the norms represent those populations. Similarly, the 2022 study by Samadi and colleagues found that the GARS-3 separated autistic children from typically developing children very well but showed lower specificity when distinguishing autism from intellectual disability or communication disorders, and the authors recommended further assessment to confirm findings in those cases.

The practical takeaway is consistent across this literature: the GARS-3 is a useful structured source of observational evidence, and its results should always be interpreted by a qualified professional alongside other measures.

Confusion between behavior rating scales and cognitive tests is common, so the distinction deserves its own section. The GARS-3 asks how much a person's observed behavior resembles behavior associated with autism. An intelligence test asks how well a person can reason, remember, and solve problems under standard conditions. These are different questions, and autistic people show the full range of cognitive ability, a topic we explore in our article on IQ and autism.

For readers interested in measuring cognitive ability specifically, the Reasoning and Intelligence Online Test, developed by RIOT IQ with psychometrician Dr. Russell T. Warne, is an online IQ test for adults 18 and older. It includes 15 subtests across six cognitive indices (verbal reasoning, fluid reasoning, spatial ability, working memory, processing speed, and reaction time), takes about 52 minutes, and reports scores on the familiar mean-100, standard-deviation-15 scale. To be clear about scope: the RIOT IQ test measures cognitive ability only. It does not screen for or detect autism, and no online test replaces an individually administered diagnostic evaluation. You can learn more or take the test at riotiq.com.

No. It is a norm-referenced screening and evidence-gathering instrument. A diagnosis of autism spectrum disorder is made by a qualified clinician who integrates observation, history, and multiple measures.

A parent, teacher, or clinician who has had recent, sustained contact with the individual rates the items based on observed behavior. The person being evaluated does not answer any questions.

According to PRO-ED, completing the ratings takes about 5 to 10 minutes. Scoring and interpretation by the examiner take additional time.

An Autism Index of 71 or above falls in the "very likely" probability range, with 101 or above indicating a need for very substantial support. An index of 54 or below is classified as "unlikely." These are probability statements based on informant ratings, never a diagnosis.

No. It measures behaviors associated with autism as observed by an informant. It provides no information about intelligence, and cognitive ability must be assessed with separate tests.

The publisher reports sensitivity and specificity of .97, and independent research finds it separates autistic children from typically developing peers well. Studies also report weaker discrimination from intellectual disability and communication disorders, and reviewers advise caution with young adults ages 20 to 22 and with individuals from minority groups.

1. Gilliam, J. E. (2014). GARS-3: Gilliam Autism Rating Scale, Third Edition [Complete kit product page]. PRO-ED. proedinc.com

2. Karren, B. C. (2017). A test review: Gilliam, J. E. (2014), Gilliam Autism Rating Scale, Third Edition (GARS-3). Journal of Psychoeducational Assessment, 35(3), 342-346. eric.ed.gov

3. Samadi, S. A., Noori, H., Abdullah, A., Ahmed, L., Abdalla, B., Biçak, C. A., & McConkey, R. (2022). The psychometric properties of the Gilliam Autism Rating Scale (GARS-3) with Kurdish samples of children with developmental disabilities. Children, 9(3), 434. pmc.ncbi.nlm.nih.gov

4. Kong, X., Zhu, J., Tian, R., Liu, S., Sherman, H. T., Zhang, X., Lin, X., Han, Y., Xiang, Z., Koh, M., Hobbie, C., Wang, B., Liu, K., Liu, J., Yin, Y., & Wan, G. (2020). Early screening and risk factors of autism spectrum disorder in a large cohort of Chinese patients with Prader-Willi syndrome. Frontiers in Psychiatry, 11, 594934. pmc.ncbi.nlm.nih.gov

5. South, M., Williams, B. J., McMahon, W. M., Owley, T., Filipek, P. A., Shernoff, E., Corsello, C., Lainhart, J. E., Landa, R., & Ozonoff, S. (2002). Utility of the Gilliam Autism Rating Scale in research and clinical populations. Journal of Autism and Developmental Disorders, 32(6), 593-599. pure.johnshopkins.edu

6. Centers for Disease Control and Prevention. (n.d.). Screening and diagnosis of autism spectrum disorder. cdc.gov

Photo by Unseen Studio, via Wikimedia Commons, CC0