The Ritvo Autism Asperger Diagnostic Scale – Revised (RAADS-R) is an 80-item self-report questionnaire designed to identify autism in adults (18 years or older). Developed by Ritvo et al. (2011) as a revision of the original RAADS, it was created to address the growing need for validated assessment tools specifically designed for adult populations, where autism presentations may differ from those typically seen in childhood.
The RAADS-R captures the ways autism manifests in adults who may have developed coping strategies, learned social scripts, or experienced years of adaptation to neurotypical expectations. Originally developed as a clinician-administered tool, it is advised that if using the RAADS-R as a self-report tool that it is then completed with clinician supervision to ensure accurate understanding of items and to provide an opportunity for clarification and clinical observation.
Example RAADS-R Item
The RAADS-R evaluates four symptom domains:
The RAADS-R assesses both current and historical autism-related behaviours through a unique response format that captures developmental trajectories. Each item can be rated as “never true,” “true only when I was younger than 16,” “true only now,” or “true now and when I was young.” This temporal dimension allows clinicians to identify both lifelong patterns characteristic of autism and traits that may have emerged or become more apparent in adulthood.
With increasing recognition that autism is often undiagnosed until adulthood – particularly in individuals who may have developed compensatory strategies – the RAADS-R serves as a valuable screening tool in mental health settings. Adult referrals for autism assessment have increased significantly in recent years (Lai & Baron-Cohen, 2015; Russell et al., 2022), with many adults presenting initially with anxiety, depression, or other mental health concerns before their underlying autism is recognised.
The RAADS-R is best used in conjunction with clinical expertise and other assessment procedures to establish a diagnosis. As a self-report measure, it relies on the individual’s insight and ability to accurately report their experiences. Individuals with limited self-awareness or those who have extensively masked their autistic traits may score below clinical thresholds despite having diagnosable autism.
In clinical practice, the RAADS-R serves multiple valuable functions. As a screening tool, it can help identify adults who warrant a comprehensive autism assessment, particularly those who may not present with obvious autism characteristics. The detailed item-level responses provide rich clinical information that can guide diagnostic interviews and help clinicians understand how autism traits manifest in the individual’s daily life. It can also facilitate discussion about experiences clients may not spontaneously report, such as sensory sensitivities or social camouflaging strategies.
The RAADS-R was originally developed by Ritvo et al. (2011) as a revision of the original RAADS (Ritvo et al., 2008), with the addition of a fourth symptom area (Circumscribed Interests), additional items, and wording clarifications. The original validation study (Ritvo et al., 2011) with 201 adults with autism and 578 non-autistic adults established a cutoff score of 65, reporting sensitivity of 97% and specificity of 100%. Test–retest reliability was high (0.987) and it had high concurrent validity (96%) with the SRS-A.
Given the evolving understanding of autism and concerns raised about the RAADS-R’s performance in clinical populations (Brugha et al., 2020; Folatti et al., 2024; Jones et al., 2021; Sizoo et al., 2015; Sturm et al., 2024), NovoPsych conducted a comprehensive psychometric re-evaluation using a large clinical sample of 63,209 individuals who completed the RAADS-R through the NovoPsych platform. This clinical sample consists of individuals who have completed the RAADS-R as part of routine clinical assessments and assessments that would have been completed due to suspected autism.
Exploratory and confirmatory factor analyses were conducted to examine the underlying structure of the RAADS-R. Despite the theoretical four-factor structure (Social Relatedness, Circumscribed Interests, Language, and Sensory-Motor), NovoPsych’s analyses indicated that a unidimensional model provided the most parsimonious fit to the data (CFI = 0.942, TLI = 0.941, RMSEA = 0.075). This finding aligns with recent research by Sturm et al. (2024), who similarly found the RAADS-R to function as a unidimensional instrument. While the four subscales do not represent statistically independent dimensions, they retain clinical utility as descriptive domains that can help characterise an individual’s autism-related traits. Therefore, it is recommended that subscale scores be used for clinical description and treatment planning purposes, but not interpreted as measuring distinct constructs.
The RAADS-R demonstrated excellent internal consistency in the NovoPsych clinical sample, with a Cronbach’s alpha of 0.97 for the total scale. This high reliability indicates strong inter-item correlations and supports the unidimensional nature of the instrument. Analysis of response patterns revealed that the time-limited response options (“true only when I was younger than 16” and “true only now”) were used infrequently, with less than 30% of respondents endorsing these options. This finding supports previous research (Sturm et al., 2024) suggesting that dichotomous scoring for the RAADS-R (using only “never true” and “true now and when I was young”) may be more parsimonious without loss of clinical information. However, it was decided to keep the existing response options in the present analyses for consistency and familiarity with the original RAADS-R.
The original RAADS-R total score cutoff of 65 has been criticised for producing high false positive rates in clinical populations, with studies reporting specificities as low as 3% (Jones et al., 2021) and over half of mental health clinic patients scoring above this threshold (Folatti et al., 2024). Sizoo et al. (2015) were among the first to identify this issue, finding poor predictive validity for the RAADS-R in adults referred for autism assessment and proposing a higher cutoff of 98. They observed that neurotypical controls in their psychiatric sample scored on average 27 points above the original cutoff, highlighting its inadequacy for clinical populations. Subsequently, Brugha et al. (2020) conducted a comprehensive evaluation in adult mental health services and found that a cutoff of 120 performed substantially better than the original threshold, with improved specificity in distinguishing autism from other psychiatric conditions.
Building on this body of evidence, NovoPsych conducted ROC analyses using a large clinical sample of 37,620 individuals who completed both the RAADS-R and Autism Quotient (AQ) via the NovoPsych platform. Using the established AQ thresholds of 26 or greater (consistent with autism) and 36 or greater (pronounced autism traits) as criterion variables, ROC analyses identified optimal RAADS-R cutoffs of:
These thresholds were established through convergent validity with another screening measure (AQ) rather than criterion validity against clinical diagnoses. While this approach cannot establish true diagnostic sensitivity and specificity, it addresses the practical problem that a large percentage of individuals in clinical services exceed the original threshold, rendering it clinically unusable. These psychometrically-derived thresholds align with independent research using diagnosed samples (Sizoo et al., 2015; Brugha et al., 2020) and provide improved clinical utility for screening purposes.
These thresholds demonstrated 81% sensitivity and 81% specificity and align with percentile distributions showing that a score of 106 falls at approximately the 23rd percentile of the original autism distribution (Ritvo et al., 2011) and a score of 140 at approximately the 57th percentile.
These revised thresholds represent a four-tier interpretation system for the total score:
NovoPsych’s findings are consistent with emerging literature suggesting that the original RAADS-R cutoff may be too sensitive for clinical populations. Jones et al. (2021) found no predictive validity using the original cutoff in a mental health service sample, while Folatti et al. (2024) reported that 57.5% of young adults in a mental health clinic scored above 65, decreasing to 21.2% using their suggested greater than 119 threshold. NovoPsych found that 81.5% of the clinical sample scored above 65, whereas 50.9% scored above Folatti et al.’s 119 threshold (Hegarty et al., 2025). NovoPsych’s updated thresholds improve specificity while maintaining acceptable sensitivity, making the RAADS-R more suitable for use in clinical populations where differential diagnosis is critical.
Given the unidimensional nature of the RAADS-R, subscale thresholds were derived using a percentile matching approach using the NovoPsych data for the total score (i.e., ‘consistent’ at 40th percentile and ‘pronounced’ at 63rd percentile) rather than independent ROC analyses. This method ensures that subscale thresholds are proportionally aligned with the total score thresholds.
The percentile equivalent resulted in the following thresholds for “consistent with autism”:
Similarly, the NovoPsych percentile equivalent to the total score of 140 was applied to derive thresholds for “pronounced autism traits”:
The autism norms, which are used to calculate Autistic percentiles, are based on the initial RAADS-R validation study by Ritvo, et al. (2011). A sample of 201 individuals with a confirmed DSM-IV-TR diagnosis of Autism or Asperger’s had a mean RAADS-R score of 133.83 (SD = 37.74). This combined Autism and Aspergers group had an average age of 31, an IQ of 119, and were 28% female. The Autism group alone had a mean RAADS-R score of 138.46 (SD = 41.4), however, the combined group was used in calculating percentiles, given that it is most representative of DSM-5-TR diagnostic criteria.
The non-autistic percentile is based on a Swedish clinical sample from Andersen et al. (2011), consisting of 197 individuals without autism spectrum disorder. This comparison group included 61 doctors and medical students, 69 university students from three Swedish campuses, 60 participants who served as comparison cases in various research studies, and 7 psychiatric patients who were assessed for ASD but did not meet diagnostic criteria. Among the total sample, 12 individuals had other DSM-IV psychiatric diagnoses, including depression, social anxiety disorder, generalized anxiety disorder, ADHD, bipolar disorder, obsessive-compulsive disorder, schizotypal personality disorder, delusional disorder, and personality disorder NOS. The comparison group had a mean age of 34 years (SD = 13, range 19-75), with 80 males and 116 females. The mean RAADS-R total score for this comparison sample was 33.8 (SD = 27.6). This Swedish normative sample provides a more clinically representative comparison group than the original Ritvo et al. (2011) normative sample, as it includes individuals with other psychiatric conditions who might realistically present for autism assessment in clinical settings, thus offering more ecologically valid percentile comparisons for clinical practice.
Analysis of RAADS-R total scores across gender groups in NovoPsych’s large clinical sample (N = 63,209) revealed significant differences in autism trait presentation. Individuals identifying as nonbinary reported the highest levels of autism traits (M = 144.79, SD = 40.94), substantially exceeding both the “consistent with autism” (106) and “pronounced autism traits” (140) thresholds. This group showed medium to large effect size differences compared to males (d = -0.686) and females (d = -0.543).
Traditional binary gender comparisons revealed that females (M = 117.73, SD = 50.37) scored slightly higher than males (M = 111.28, SD = 49.69), though this difference was small (d = -0.129). Those who did not declare gender (N/A group) also showed elevated scores (M = 124.94, SD = 50.04) compared to males and females. These findings align with emerging literature suggesting higher rates of autism traits in gender-diverse populations (Strang et al., 2014; Kung, 2020) and support the need for clinicians to be particularly attentive to potential autism when working with nonbinary individuals. The elevated scores in the nonbinary group (with 50% scoring above 150) highlight the importance of comprehensive autism assessment in gender-diverse populations, where the intersection of neurodivergence and gender identity may create unique clinical presentations and support needs.
These updates represent not a departure from the RAADS-R, but rather participation in an evidence-based international consensus on appropriate clinical interpretation. The original cutoff of 65 was established by Ritvo et al. in 2011, but multiple independent research teams have since documented serious limitations when this threshold is applied in real-world clinical settings. Sizoo et al. (2015) found that neurotypical controls with psychiatric conditions scored an average of 27 points above this cutoff and recommended increasing it to 98. Brugha et al. (2020) reported specificities as low as 3% in UK mental health services, with over half of all service users exceeding the threshold, leading them to recommend cutoffs between 120-126. Jones et al. (2021) found the original cutoff had essentially no predictive validity (AUC = 0.45) in specialist autism services, concluding it was “not effective” for clinical use. Picot et al. (2021) documented false positive rates exceeding 50% in psychiatric populations.
In response to this mounting evidence, NovoPsych conducted an advanced statistical analysis of 63,209 clinical cases, which independently confirmed these researchers’ findings. Our analysis revealed that 81.5% of the clinical sample scored above 65—aligning closely with other studies’ observations. The updated thresholds NovoPsych derived (106 for “consistent with autism” and 140 for “pronounced traits”) achieve 81% sensitivity and 81% specificity, and notably converge with the thresholds proposed by other teams. NovoPsych retained a “some autistic traits” category (65-105) to acknowledge sub-threshold presentations, including potential masking, where clinical judgement should guide decisions about further assessment.
Multiple independent research teams have reported issues with the original RAADS-R cutoff of 65:
The original validation study tested 201 autistic adults against 578 comparison subjects, including both neurotypical controls and people with other psychiatric diagnoses, reporting 100% specificity at a cutoff of 65. However, when independent researchers tested this cutoff in real-world mental health settings, they found drastically different results, with most people with anxiety, depression, or other conditions scoring above 65. This discrepancy likely reflects differences in sample selection and clinical complexity between research and practice settings. The new thresholds are based on NovoPsych’s advanced statistical analysis, providing more realistic cutoffs for clinical populations where the tool is routinely used. These updated thresholds (106 and 140) better distinguish autism from other mental health conditions in everyday clinical practice. The original cutoff of 65 is now described as “Some autistic traits” to recognise that some individuals could be engaging in masking behaviours but that they are not at the level which is clearly consistent with autism and that clinical judgement should guide decisions about further assessment..
Research confirmed that despite having four subscales, the RAADS-R actually measures a single underlying autism construct rather than four independent domains. Given this unidimensional structure, subscale thresholds were derived using percentile matching from the same large clinical sample that informed the total score thresholds. This ensures all thresholds are proportionally aligned and based on real-world clinical populations where differential diagnosis actually occurs. The subscales remain useful for describing different aspects of autism presentation, even though they’re not statistically independent constructs.
Yes, subscale scores remain clinically valuable despite the unidimensional structure. While the subscales don’t measure completely independent constructs, they provide useful descriptive information about how autism traits manifest for an individual. For example, someone might show pronounced sensory-motor differences but fewer social challenges, which is important for treatment planning and understanding their specific support needs. Think of it like describing someone’s physical fitness – while overall fitness is one construct, it’s still helpful to know their relative strengths in cardio versus flexibility versus strength.
If someone scored 65 or above but below 106 on a previous assessment, this doesn’t mean their autism traits have changed – it means we now have more accurate thresholds for clinical populations. The original cutoff was too sensitive for people with any mental health condition. Someone who scores between 65-105 likely experiences some autism traits (i.e., labelled as “Some autistic traits”), but these may be better explained by anxiety, depression, ADHD, or other conditions. Importantly, scores within the “Some autistic traits” category require clinical judgement to guide decisions about further assessment. These new thresholds don’t invalidate previous assessments but provide clearer guidance about when comprehensive autism evaluation is warranted. Clinical judgment and thorough assessment remain essential regardless of which thresholds are used.
A score below 106 doesn’t rule out autism, particularly in individuals who mask or camouflage their autistic traits. In the original validation study, the only 6 autistic individuals (3%) who scored below the threshold were all young adults around age 20, whom family members described as trying to “appear as normal as possible.” Masking – the conscious or unconscious suppression of autistic behaviours to fit social expectations – can significantly affect self-report scores. This is especially common in individuals who have learned to hide their difficulties through years of social conditioning, those diagnosed later in life, women and gender-diverse individuals who may have different presentations, and people who have received extensive behavioural interventions since childhood. Additionally, some individuals may lack insight into their own differences or may not recognise certain behaviours as unusual if they’ve always experienced them. If clinical observations, developmental history, or informant reports suggest autism despite a low RAADS-R score, comprehensive assessment is still warranted. The RAADS-R is a screening tool, not a definitive diagnostic test, and clinical judgment should always take precedence over any single assessment score. If masking is suspected the CAT-Q is a good option for further assessment.
Analysis of clinical cases revealed that individuals identifying as non-binary (n = 201) report the highest levels of autism traits on the RAADS-R (average score 144.79), with 50% scoring above 150. Only 15.9% of this subsample scored below the 106 threshold with 60.6% scoring at or above the 140 threshold. The scores for this non-binary subsample substantially exceeds scores for males (111.28) and females (117.73), and is well above both clinical thresholds. These findings align with emerging research showing higher rates of autism traits in gender-diverse populations. The intersection of neurodivergence and gender identity may create unique clinical presentations – for instance, the experience of not fitting into binary gender categories may parallel the experience of not fitting into neurotypical social expectations. Additionally, autistic individuals may be more likely to question and reject social constructs, including gender norms. It’s important to note that elevated scores don’t automatically mean someone is autistic; comprehensive assessment is essential. However, clinicians should be particularly attentive to potential autism when working with non-binary individuals, as traditional diagnostic approaches developed primarily with cisgender samples may miss or misinterpret their presentations. The high scores suggest that many gender-diverse individuals seeking mental health support may benefit from autism assessment as part of understanding their full support needs.
Ritvo, R. A., Ritvo, E. R., Guthrie, D., Ritvo, M. J., Hufnagel, D. H., McMahon, W., Tonge, B., Mataix-Cols, D., Jassi, A., Attwood, T., & Eloff, J. (2011). The Ritvo Autism Asperger Diagnostic Scale-Revised (RAADS-R): a scale to assist the diagnosis of Autism Spectrum Disorder in adults: an international validation study. Journal of Autism and Developmental Disorders, 41(8), 1076–1089. https://doi.org/10.1007/s10803-010-1133-5
Andersen, L.M., Näswall, K., Manouilenko, I., Nylander, L., Edgar, J., Ritvo, R.A., Ritvo, E., & Bejerot, S. (2011). The Swedish version of the Ritvo Autism and Asperger Diagnostic Scale: Revised (RAADS-R). A validation study of a rating scale for adults. Journal of Autism and Developmental Disorders, 41(12), 1635-1645. https://doi.org/10.1007/s10803-011-1191-3
Brugha, T., Tyrer, F., Leaver, A., Lewis, S., Seaton, S., Morgan, Z., Tromans, S., & van Rensburg, K. (2020). Testing adults by questionnaire for social and communication disorders, including autism spectrum disorders, in an adult mental health service population. International Journal of Methods in Psychiatric Research, 29(1), Article e1814. https://doi.org/10.1002/mpr.1814
Folatti, I., Santangelo, G., Sanguineti, C., Inci, S., Faggioli, R., Bertani, A., Nisticò, V., & Demartini, B. (2024). The prevalence of autistic traits in a sample of young adults referred to a generalized mental health outpatient clinic. Diagnostics, 14(21), Article 2418. https://doi.org/10.3390/diagnostics14212418
Hegarty, D., Buchanan, B., Kaufman, H., Smyth, C., Baker, S., Bartholomew, E., Phillips, J., Wotherspoon, J., & Marshall, J. (2025). A Review of the Clinical Utility and Psychometric Properties of the Ritvo Autism Asperger Diagnostic Scale – Revised (RAADS-R): Updated Clinical Thresholds. https://doi.org/10.17605/OSF.IO/MEAUS
Jones, S. L., Johnson, M., Alty, B., & Adamou, M. (2021). The effectiveness of RAADS-R as a screening tool for adult ASD populations. Autism Research and Treatment, 2021, Article 9974791. https://doi.org/10.1155/2021/9974791
Kung, K. T. F. (2020). Autistic traits, systemising, empathising, and theory of mind in transgender and non-binary adults. Molecular Autism, 11, Article 73. https://doi.org/10.1186/s13229-020-00378-7
Lai, MC. & Baron-Cohen, S. (2015). Identifying the Lost Generation of Adults with Autism Spectrum Conditions. The Lancet Psychiatry, 2(11), 1013-1027. https://doi.org/10.1016/S2215-0366(15)00277-1
Picot, MC., Michelon, C., Bertet, H. et al. The French Version of the Revised Ritvo Autism and Asperger Diagnostic Scale: A Psychometric Validation and Diagnostic Accuracy Study. Journal of Autism and Developmental Disorders 51, 30–44 (2021). https://doi.org/10.1007/s10803-020-04518-z
Ritvo, R.A., Ritvo, E.R., Guthrie, D., Yuwiler, A., Ritvo, M.J., & Weisbender, L. (2008). A Scale to Assist the Diagnosis of Autism and Asperger’s Disorder in Adults (RAADS): A Pilot Study. Journal of Autism and Developmental Disorders, 38, 213–223. https://doi.org/10.1007/s10803-007-0380-6
Russell, G., Stapley, S., Newlove-Delgado, T., Salmon, A., White, R., Warren, F., Pearson, A. and Ford, T. (2022), Time trends in autism diagnosis over 20 years: a UK population-based cohort study. Journal of Child Psychology and Psychiatry, 63: 674-682. https://doi.org/10.1111/jcpp.13505
Sizoo, B. B., Horwitz, E., Teunisse, J., Kan, C., Vissers, C., Forceville, E., Van Voorst, A., & Geurts, H. (2015). Predictive validity of self-report questionnaires in the assessment of autism spectrum disorders in adults. Autism, 19(7), 842–849. https://doi.org/10.1177/1362361315589869
Strang, J. F., Kenworthy, L., Dominska, A., Sokoloff, J., Kenealy, L. E., Berl, M., Walsh, K., Menvielle, E., Slesaransky-Poe, G., Kim, K. E., Luong-Tran, C., Meagher, H., & Wallace, G. L. (2014). Increased gender variance in autism spectrum disorders and attention deficit hyperactivity disorder. Archives of Sexual Behavior, 43(8), 1525–1533. https://doi.org/10.1007/s10508-014-0285-3
Sturm, A., Huang, S., Bal, V., & Schwartzman, B. (2024). Psychometric exploration of the RAADS-R with autistic adults: Implications for research and clinical practice. Autism, 28(9), 2334–2345. https://doi.org/10.1177/13623613241228329
Also known as: AQ, Autism Spectrum Quotient, AQ-50, Autism screening test
Also known as: EQ-40, EQ, Empathy quotient, Empathy assessment scale, Social empathy measure
