Social Responsiveness Scale

Synonyms

SRS

Abbreviations

AUC:

Area under curve

ADHD:

Attention deficit hyperactivity disorder

ADI-r:

Autism diagnostic interview-revised

ADOS:

Autism diagnostic observation scale

ASD:

Autism spectrum disorder

BPVS:

British picture vocabulary scale

CBCL:

Child behavior checklist

CCC:

Children’s communication checklist

OCD:

Obsessive compulsive disorder

PDD-NOS:

Pervasive developmental disorders-not otherwise specified

ROC:

Receiver operating characteristics

SCDC:

Social and communication disorders checklist

SCQ:

Social communication questionnaire

VABS:

Vineland adaptive behavior scales

Description

The Social Responsiveness Scale (SRS) is a 65-item rating scale that measures the severity of autistic symptomatology as a quantitative trait, among children clinically affected by autism spectrum conditions as well as among children in the general population. It is particularly useful for characterizing milder autistic syndromes that lie at the boundary between the normal population distribution and clinical-level affectation. The SRS can be completed by a parent, a teacher, a spouse (in the case of the adult version of the SRS), or another adult informant. The validity of self-report is still under study. The SRS involves ratings of children in their natural social contexts and reflects what has been consistently observed over weeks or months of time rather than in a single clinical or laboratory observation. In this way, it capitalizes on both direct observation and on the accumulated history of behaviors observed by the informant over time. The SRS generates quantitative scores for the severity of autistic traits and symptoms, and distinguishes Autism Spectrum Disorder (ASD) from other psychiatric conditions (Constantino, Przybeck, Friesen, & Todd, 2000). Norms have been published by gender and rater type (parent versus teacher) in order to standardize ratings, which otherwise differ as a function of these parameters. SRS scores are highly heritable (Constantino & Todd, 2003), stable over time (Constantino et al., 2009), exhibit high inter-rater reliability (Constantino et al., 2007a), are continuously distributed in the general population (Constantino & Todd, 2003), are nonsignificantly correlated with IQ among children representing the normal range of IQ in the general population (Constantino et al., 2007a), and exhibit a unitary factor structure (Constantino et al., 2004), which supports the use of a single index score as a quantitative measure of autistic severity. SRS scores greater than 75 T (98.8th percentile) indicate a level of autistic social impairment that is generally highly clinically significant.

Historical Background

Over the course of some 25 studies involving over 10,000 individuals (including Constantino et al., 2006, 2007a,b, 2009; Constantino & Gruber, 2005; Constantino, Hudziak, & Todd, 2003b; Constantino & Todd, 2000; Lee et al., 2010; Levitt & Campbell, 2009; Pine, Luby, Abbacchi, & Constantino, 2006; Virkud, Todd, Abbacchi, Zhang, & Constantino, 2009), research involving the SRS has yielded information about the distribution of autistic traits and symptoms in children and families affected by autism, as well as children and families in the general population. Complementary research involving an array of other quantitative rating scales has converged with findings from studies involving the SRS in demonstrating that quantitative autistic traits exhibit a continuous distribution in nature (from very mild to very severe, see Fig. 1 panel A below) and are substantially heritable (estimates on the order of 0.60–0.80). The traits and symptoms captured by the SRS are extremely stable over time (5-year test-retest correlations on the order of 0.60–0.70 in clinical samples and 0.70–0.80 in nonclinical populations, as depicted in panel B below), yet capable of fluctuating within individuals over time and in response to intervention. The SRS is currently in use in a broad range of settings, including schools, clinical services, and research programs. It offers the capability of feasible and reliable quantitative characterization of core components of the autistic syndrome across informants and across environmental contexts, and has been translated into over 20 foreign languages at the time of this writing.

Social Responsiveness Scale, Fig. 1

School-age SRS studies: (a) General population distribution (b) 5-year Stability (Constantino et al., 2009)

Psychometric Data

Each version of the SRS can be completed in 15 min and generates both scale scores for specific symptom domains relevant to the characterization and treatment of autistic syndromes, as well as a singular total score for autistic social impairment, empirically validated via factor, cluster, and latent class analysis (Constantino et al., 2004, 2007a). Higher total scores on the SRS indicate greater severity of social impairment; and inter-rater reliability is high (parent-teacher correlation 0.72, n = 1,200). The SRS exhibits nonsignificant correlations with IQ, and substantial agreement with the Autism Diagnostic Interview (ADI-r) and the Autism Diagnostic Observation Scale (ADOS) (Constantino et al., 2003, 2004, 2007a; Lee et al., 2010). Parent-report scores on the SRS distinguish children with ASDs (including Autistic Disorder, Asperger Disorder, and Pervasive Developmental Disorders [PDD-NOS]) from those with other child psychiatric conditions (Constantino et al., 2004, 2000, 2007b; Constantino & Gruber, 2005). Autistic traits and symptoms measured by the SRS are continuously distributed not only in clinical populations but in the general population (Constantino & Todd, 2003). Furthermore, traits measured by the SRS aggregate in the male first-degree relatives of subjects with ASD in multiple-incidence families (Constantino et al., 2006; Virkud et al., 2009). In addition, there is evidence from molecular genetic studies that SRS scores map to specific autism susceptibility loci (Duvall et al., 2007; Levitt & Campbell, 2009; Coon et al., 2010).

Internal Consistency

Internal consistency is a form of reliability that can be estimated from a single administration. Studies using the SRS have reported very high estimates. The earliest study on the SRS (Constantino et al., 2000) assessed internal consistency in regular education children aged 4–7 years (N = 197). An alpha .97 was obtained. In a report on a clinical sample using the German translation, Bolte, Westerwald, Holtmann, Freitag, and Poustka (2011) obtained an alpha .94 in a mixed clinical group (N = 255) and alpha .96 in an ASD group (N = 148).

Retest Reliability/Temporal Stability Construct Validity

Retest reliability and temporal stability are closely related psychometric constructs. While the actual study principle is the same – an individual is asked to fill in an instrument twice so that results can be compared after time delay – the meaning of the result changes substantially when the two procedures are separated over increasingly long time periods. When the interval is short, say a few weeks to a few months, the study is said to address test-retest reliability and findings are attributed narrowly to the reliability of the instrument. When the interval is longer however, say on the order of 6 months to several years, then the study is said to address temporal stability and the results are attributed to the validity of the underlying construct (autism in the case of the SRS) and of the instrument.

Research interest has focused on longer intervals in clinical populations. It is a testimony to both the high validity of the autism construct and to the SRS in measuring that construct that correlations have been very high. Constantino et al. (2000) reported on a clinical group (N = 30) with a retest interval averaging 137 days and found a correlation r = .88. Bolte, Dziobek, and Poustka (2009) assessed a clinical group (N = 49), with a 3–6 month retest interval and found a retest correlation of r = .95 was obtained. Constantino et al. (2009) tested 95 male twin pairs ASD affected and 95 males typically developing (total N = 285) longitudinally over a 5-year period and found that retest correlations maintained around r = .90 in both groups, as shown in Fig. 1 Panel B.

In general, the reliability of an instrument is seen as a strong constraint on its validity. Thus, the evidence of longer interval construct stability above, with correlations on the order of 0.80–0.90, tends to obviate the need to put resources into documenting short interval retest reliability. Bolte, Poustka, and Constantino (2008) did evaluate retest reliability in conjunction with the development of the German edition, but only in the normative group. With a retest interval ranging from 3 weeks to 4 months, they found retest reliability was r = .80 for mothers and r = .72 for fathers.

Inter-rater Reliability/Convergent Validity

Inter-rater report comparisons serve two functions. In a narrow sense they are an aspect of reliability, since they involve applying the same items to the same child. Yet inter-rater comparisons also go beyond that to provide a broader aspect of validity, since they involve passing the items through different experience bases. Thus, mother and father, even while living in the same home and sharing many family experiences, do each have separate personal experiences with a child and, moreover, a unique prior personal history that can influence how each will interpret the meaning of the 65 test items. And this sense of difference in background, experience, and perspective expands even more when a parent report is compared to that of a teacher; for one a teacher and a parent have virtually no overlapping observations of a child’s behavior and for another a teacher, while seeing the child in a more limited setting than the parent, can draw on professional training and direct experience with scores and even hundreds of other children when judging the child’s behavior.

Despite these differences in perspective, studies reporting on different observers have found highly concordant reports. With regard to mother-father inter-rater agreement, Constantino et al. (2003a) collected mother and father reports in a sample of ASD-diagnosed individuals (N = 61). They found mother-father reports correlated r = .91. Bolte et al. (2008), using the German edition, reported on 527 mixed clinical cases (160 with ASDs) and also found mother-father reports correlated r = .91. In the normative sample, Bolte et al. (2009) obtained a mother-father correlation r = .61. Prior comments regarding results in normative or other less-affected groups apply again here.

Parent reports have also been compared to teacher reports in clinical samples. Regarding parent-teacher comparisons, Constantino et al. (2003a) found mother-teacher reports correlating r = .82 and father-teacher reports correlating r = .75 (both N = 61). In a later study Constantino et al. (2007a) compared teacher- and parent-report SRS scores in a larger sample (N = 577) of clinically referred children and found parent-teacher reports correlated r = .72. As noted in the introductory comments, differences in experience bases are much larger between teachers and parents than between parents. These finding on parent-teacher comparisons are also highly indicative of strong reliability and validity.

Comparisons with Other Instruments: Concurrent Validation

Two broad issues can be addressed from reports where the SRS has been used with other instruments. Of these, convergent validity compares the instrument to others that have the same focal purpose. For the SRS, this involves comparisons with other instruments that are designed to evaluate symptoms and behaviors on the Autism Spectrum. In these comparisons, higher correlations are desirable. In contrast, divergent/discriminant validity addresses the instrument’s performance in contexts where a broader variety of symptoms and behavior problems need to be addressed. For the SRS, this could involve comparison to instruments that assess other commonly identified psychological disorders and behavioral problems – for example, behavioral expressions of internalizing problems of anxiety, depression, and Obsessive Compulsive Disorder (OCD) or behavioral expressions of externalizing problems such as Attention Deficit Hyperactivity Disorder (ADHD), conduct disorders, and Bipolar disorder – or those that assess less specifically linked problems such as developmental delay. In these comparisons, moderate- to lower- or even zero-order correlations are desirable, demonstrating that the SRS explains variance that is unique to the social behavior associated with ASDs and not strongly overlapping with that due to behavioral difficulties associated with other disorders and identified by other instruments.

Convergent Validity

Parent and Teacher Report Behavior Assessments. Comparison with other ASD-directed measures have largely involved instruments using a similar approach to assessment: multiple item objective questionnaires with a multipoint Likert-scale or a true-false response format that is completed by informants familiar with the child’s behavior in natural living contexts such as parents and teachers. Good agreement has been reported with the several instruments that have been investigated.

The most widely investigated comparison has been with the Social Communication Questionnaire (SCQ; Rutter et al., 2003), an instrument that uses a true-false response format with 40 items derived from the ADI-r. Findings with the SRS have been relatively consistent. In studies with mixed clinical samples as cited earlier, Charman et al. (2007) report a correlation r = .68 (n = 119); Bolte, Poustka, and Constantino (2008) report r = .58 (n = 107). In addition, also in a mixed clinical sample, (Pine, Guyer, Goldwin, Towbin, & Leibenluft, 2008) found r = .65 (n = 352) and in a small sample of children affected by epilepsy Granader et al. (2010) report r = .61 (n = 21). A slightly lower value was reported by Bolte et al. (2011), r = .50 (n = 480), in a sample that included typically developing as well as clinical subjects.

There have also been several studies correlating SRS scores with those from the Children’s Communication Checklist (CCC; Bishop, 1998), a 70-item instrument with a Likert scale response format. Pine et al. (2008) reports correlations or r = −.49 and −.72 with the two ASD-focused (and positively valanced) subscales. Similarly, Charman et al. (2007) report a SRS to CCC correlation r = −.75.

The SRS has also been compared to the Social and Communication Disorders Checklist (SCDC, [Skuse, Mandy, & Scourfield, 2005]) a brief, 12-item scale using a 3-point Likert scale response format. Using German translations, Bolte et al. (2011) found the instruments moderately correlated r = .49.

The ADI-r and the ADOS. The SRS has also been compared to instruments that are less similar in design, in particular the Autism Diagnostic Interview, revised (ADI-r; Lord, Rutter, & Le Couteur, 1994) and the Autism Diagnostic Observation Schedule (ADOS; Lord et al., 2000).

The design characteristics of these instruments are quite different from that of the SRS and other behavioral parent and teacher reports cited in the previous section. The ADI-r is a long structured psychiatric interview from which a subset of questions are selected and coded; the ADOS is a structured observation of behavior conducted by a trained professional under standard conditions from which pre-identified behavior is identified and coded. Both instruments were designed to facilitate categorical, diagnostic decision making and, consequently, do not produce the same kind of dimensional results produced by the objective questionnaires. It has long been recognized that differences in method tend to produce evidence of validity at lower levels of correlation (e.g., Campbell & Fiske, 1959). Given the differences, results provided clear support for the validity of the SRS.

ADI-r results include “domain scores,” sums of coded item subsets. Constantino et al. (2003a) reported an early investigation of SRS validity, comparing SRS to ADI-r lifetime scores in a sample ASD-diagnosed individuals (N = 61). We note that ADI-r domain scores characterize level of severity around the age of 4 years, at a relative peak in manifestations of autistic symptomatology from the standpoint of developmental history. Despite this fundamental difference between the SRS (which measures current dysfunction) and the ADI-r (which indexes historic symptomatology in early childhood), correlations between parent-report SRS scores and ADI-r domain scores ranged from .65 to .77 for mother reports, from .52 to .70 for teacher reports, and from .60 to .74 for father reports. In a subsequent report on a larger sample, somewhat lower, though highly statistically significant correlations were reported between SRS Parent reports and ADI-r domain scores (range of r = .31 – .36) and SRS Teacher reports (range of r = .26 – .40) (Constantino et al., 2007a). These correlations were on the order of what was observed for correlations between ADI-r and ADOS domain scores. In the study associated with the development of the German edition, Bolte et al. (2008) collected 133 clinical cases and report more specifically on parent reports for the separate domains: Social Interaction Domain score r = .46, Communication Domain Score r = .40, Stereotypic Behavior Domain score r = .38. Charman et al. (2007) used an ADI-r results that combined scores across domains and found a correlation r = .59. As noted in the introductory paragraph, the ADI-r was designed to facilitate categorical analyses and its brief domain scores do not provide a strong basis for correlational analyses, for example, Bolte et al. (2009) used standard adjustments to compensate for attenuated ranges in short scales. The relatively stronger finding in the Charman et al. (2007) study may in part reflect the combination of domains into a single, more dimensionalized score.

The same sets of investigators produced parallel findings on comparisons to the ADOS. Constantino et al. (2007a) found ADOS domain scores correlated with SRS Parent reports (range of r = .37 to .58) and SRS Teacher reports (range of r = .15 – .43). The Bolte et al. (2008) study reports Communications/Social Deficits score r = .35. In a subsequent study Bolte et al. (2011) found correlations in the r = .32 – .35 range with the ADOS Domain scores. Again, using a single combined domain score, Charman et al. (2007) found an SRS to ADOS correlation r = .48. The earlier comments regarding brief domain scores and Charman et al.’s use of a combined domain score also apply here.

Divergent/Discriminant Validity

The task of differentiating diagnostic groups is addressed by studies cited in a later section. This section provides a more general sense of overall validity from comparing correlations with different instruments and their scales. The comparison scales reported here are directed at a wider variety of mental health problems and diagnoses, some of which may have overlapping symptoms with ASD and others thought or known to have no systematic relation to ASDs. Reported correlations will reflect a validation of the SRS by showing moderate- to low- or zero-order correlations.

As noted earlier, differences in methods can have an impact on correlational evidence (e.g., Campbell & Fiske, 1959). Most of the reports in the following discussion involve parent and teacher behavioral reports, similar in design to convergent measures like the SCQ, CCC, and SCDC. The appropriate comparison in this section is with those reported on the parent and teacher report behavioral measures in the previous section.

Studies using the Child Behavior Checklist (CBCL; [Achenbach & Ruffle, 2000]) support the view that the SRS is more sensitive to behaviors that can sometimes be associated with ASD-related problems, less sensitive to behavior seldom seen in ASDs, and – perhaps most critically – sufficiently independent of the CBCL to indicate sensitivity to behavioral problems not assessed by the CBCL. Two studies report correlations between the SRS and CBCL in clinical samples, with Constantino et al. (2000) reporting on 84 clinical cases and Bolte et al. (2008) reporting on 119 clinical cases from the German validation studies. Despite the differences in location and language, the findings were quite parallel. Both studies found moderate correlations for the SRS with CBCL subscales that have some overlap with the kinds of symptoms seen with ASDs: Social Problems, Thought Problems, and Attention Problems (correlations range r = .48 to .64). As expected, results were somewhat mixed but generally lower for correlation with CBCL subscales directed at less related behavior: Withdrawn, Delinquent Behavior, and Aggressive Behavior (correlations range r = .34 – .54). Both studies found zero-order correlations for the SRS with the CBCL Somatic Complaints subscale (r = .11 and .12 ns). In another study involving a general population twin sample, Constantino et al. (2003b) examined overlap in the constructs captured by the scales, suggesting about 16% total shared variance.

Using the Vineland Adaptive Behavior Scales (VABS; Sparrow, Balla, & Cicchetti, 1984) studies have reported on the relation of development to SRS scores. Charman et al. (2007) report a correlation of −44 for the positively valenced composite VABS scores with the SRS and Bolte et al. (2008) found a correlation r = −.36 for the composite correlations ranging from −.34 to −.43 for the subscales. This level of correlation appears reasonable for a developmental disorder like ASD. Regarding narrower cognitive ability, Charman et al. (2007) reported no significant correlation of the SRS to the British Picture Vocabulary Scale (BPVS; Dunn, Dunn, Whetton, & Burley, 1997).

Clinical Uses

The first question for a clinical assessment is whether affected individuals show scores elevated enough for the assessment to have clinical utility. Independent reports providing information on both typically developing children with those diagnosed as autistic have consistently shown a statistically significant and clinically meaningful separation. Children not affected by ASD or any other disorder are typically reported to have SRS Total Scores in the narrow range from 0 to 35. This is true in studies where children have been drawn to be representative of typically developing populations (Bolte et al. 2008, 2011; Coon et al., 2010) and in studies where matched controls have been drawn (Reiersen, Constantino, Volk, & Todd, 2007; Pine et al., 2008). In contrast, results for groups of children with PDD-NOS, ASD, or autistic disorder find SRS Total Score group means in the range 86–116 (Charman et al., 2007; Constantino et al., 2000; Coon et al., 2010; Kalb, Law, Landa, & Law, 2010). Even given the rather large standard deviations reported for autistic groups (ranging 27–33), the findings indicate a separation of two standard deviations or more. These differences are substantial and have a clear practical utility.

Clear and useful separation can also be seen when contrasting autism with other non-autistic diagnoses. Studies that have presented SRS scores for mixed or specific non-ASD diagnosis clinical samples have reported group SRS Total Score means in the range 40–75, that is, consistently higher than those reported for typically developing groups and lower than those reported for autism-affected groups (Bolte et al., 2008, 2011; Charman et al., 2007; Constantino et al., 2000; Pine et al., 2008; Puleo & Kendall, 2011; Reiersen et al., 2007; Towbin, Pradella, Gorrindo, Pine, & Leibenluft, 2005). It must be noted that studies will produce different results depending on the specific symptoms associated with the contrast diagnosis and on the severity of clinically affected subjects who are ascertained or recruited.

Screening and Receiver Operating Characteristics (ROC) Reports

The SRS has had a number of clinical applications: qualifying subjects for research studies, providing support for clinical diagnoses, assessment of treatment effects, etc. Wide international concern about autism, however, has also led to an interest in developing and applying screening tools, and the SRS has been studied for this purpose. These studies evaluate an instrument’s sensitivity (the proportion of actually affected individuals who are correctly identified) and specificity (the proportion of nonaffected individuals who are correctly identified). In general, there is a trade-off so that increasing specificity degrades sensitivity and vice versa.

The studies reported here reflect the complexity of the screening task (e.g., differing contrast diagnoses, population pools, comparison instruments, and cut points that may favor sensitivity in some cases and specificity in others). Even with the variation, however, the findings to be reported indicate that the SRS has a useful power to properly identify affected children and to accurately discriminate children with ASD conditions from typically developing children and also from those with varied non-ASD disorders. For a psychiatric sample of 133 children with ASD diagnoses and 126 mixed non-ASD diagnoses (sample described in Constantino et al. (2004)), a ROC analysis was reported. With the standard recommended research cutoff of 75, sensitivity was reported as .85 and specificity as .75. Charman et al. (2007) reported on a sample of children previously identified with developmental or special education risk factors, and contrasted 49 children with confirmed ASD diagnoses with 70 children with confirmed non-ASD diagnoses. Under these conditions, Area Under Curve (AUC) was .77, sensitivity was .78, and specificity was .67.

A larger and more diverse study by using the German translation involved a sample of 480 children, including 148 with ASD diagnoses, 255 with non-ASD clinical diagnoses, and 77 typically developing children (Bolte et al., 2011). Results were reported based on the recommended research cutoff for the SRS of 75 that was used in the prior two reports. Contrasting ASD cases with typically developing children they report AUC = .98 with sensitivity = .80, specificity = 1.0. Contrasting ASD cases with the non-ASD clinical group, they report AUC = .81 with sensitivity = .80, specificity = .69. Contrasting ASD with a clinical sub-sample of ADHD affected children, they report AUC = .86 with sensitivity = .80, specificity = .78. And finally, contrasting ASD with a clinical sub-sample of children with anxiety diagnoses, they report AUC = .82 with sensitivity = .81, specificity = .74.

The findings across the three studies and groups are fairly consistent with regard to sensitivity, ranging from .78 to .85 and clustering around .80. With regard to specificity, findings are more varied and depend on the nature of the contrast group, ranging from .69 to 1.0, with perhaps .75 as an adequate single estimate in clinical settings where there are mixed diagnoses in the contrast group. Co-equal high rates, as seen for the SRS, are generally desirable, but there can be applications where favoring one or the other many be preferred, for example, an instrument with higher sensitivity for use in early, broad population studies or one with higher specificity for use in aiding differential clinical diagnosis. The SRS is highly versatile in this sense; given the fact that it is fundamentally a quantitative trait measure, users can adjust the cutoff values used for screening to optimize sensitivity or positive predictive value.

The results above are reported on the conventional procedure, where a single administration of an instrument is used to identify children with potential problems. The SRS has two features that make it relatively easy to add to screening power. As a behavioral report, it does not require clinician time to conduct the assessment and as a report with several forms, it is relatively easy to collect more than one score on a child, for example, a parent and a teacher or a mother and a father. Taking advantage of these features, Constantino et al. (2007a) compared teacher- and parent-report SRS scores in a large sample (N = 577) of clinically referred children. A PDD-affected sample (n = 271) was compared to nonaffected siblings (N = 119). Parent and teacher forms were both administered and the screening power reported when both reports on a given child were elevated to the diagnostic cut point. Under these conditions the ROC analyses showed an AUC = .95 with sensitivity 0.75 and specificity 0.96.

Results in Other Clinical and Behavioral Research

While the SRS has the identification and characterization of ASDs as its central purpose, it has been used in studies of other behavior difficulties and clinical diagnoses. In all of the studies reported below, children with non-ASD clinical diagnoses are seen to have clinically relevant weakness in the kinds of social behavior assessed by the SRS. As noted by one of the teams (Pine et al., 2008), this can raise the question of whether ASDs are being under-assessed when there are comorbid conditions. More broadly, however, all of the findings show that the behavioral and social problems associated with ASD can have relevance in other areas. The SRS measures these problems in a way that is sensitive to their presence, differentiated in elevation from that seen in actual ASD diagnoses, and may be highly sensitive to treatment effects (Puleo & Kendall, 2011).

Heritability/Genetic Epidemiology

In an extensive line of research employing the SRS in large genetically informative samples of twins (Constantino & Todd, 2000, 2005; Ho, Todd, & Constantino, 2005), siblings (Constantino et al., 2006; Schwichtenberg, Young, Sigman, Hutman, & Ozonoff, 2010), and families (Virkud et al., 2009), representing both the general population Constantino and Todd (2003) and clinically affected families (Constantino, Zhang, Frazier, Abbacchi, & Law, 2010), including studies involving molecular genetic markers (Duvall et al., 2007; Campbell, Warren, Sutcliffe, Lee, & Levitt, 2010; Coon et al., 2010), it has been shown that the quantitative traits measured by the SRS are highly heritable across the entire range of severity in which they occur in nature (from mild to severe), and that subclinical autistic traits characterized by mild elevations SRS scores constitute candidate endophenotypes, that is, genetically related to the cause(s) of autism itself.

Brain Imaging

Several neuroimaging studies have elucidated brain imaging phenotypes that relate closely to variation in SRS scores, both in the general population and in clinically affected samples (Assaf et al., 2010; Di Martino et al., 2009; Kaiser et al., 2010; Paul, Corsello, Tranel, & Adolphs, 2010).

Longitudinal Course

Constantino et al. (2009) tested N = 285 (95 male twin pairs ASD affected and 95 male typical developing) longitudinally over a 5-year period, and found that test retest correlations maintained around r = 90 in both groups. Individual trajectories varied as a function of severity at baseline. Moderate improvement with age was noted, underscoring the instrument’s potential utility for ascertaining incremental response to intervention. Ongoing research on the longer-term longitudinal course of autistic traits and symptoms among clinically affected children and their siblings is underway in NIH-funded research.

See Also

Attention Deficit/Hyperactivity Disorder

Obsessive-Compulsive Disorder (OCD)

Pervasive Developmental Disorders

Vineland Adaptive Behavior Scales