Set-shifting is one cognitive process that holds promise as an intermediate phenotype in ASDs. Intermediate phenotypes include cognitive processes that serve as a link between the observed behavioral symptoms of neurodevelopmental and psychiatric disorders and their brain and/or genetic bases (Gottesman and Gould, 2003).This connection requires establishing an association between the cognitive process and core symptoms as well as brain and genetic indices. Several independent investigations report a positive relationship between set-shifting deficits and restricted interest/repetitive behavior symptoms in autism (Lopez et al., 2005; South et al., 2007). Furthermore, in a recent functional neuroimaging investigation of set-shifting in adults with ASDs, the activation patterns of anterior cingulate cortex and left intraparietal sulcus correlated negatively with the restricted interest/repetitive behaviors index from the Autism Diagnostic Interview–Revised (ADI–R: Shafritz et al., 2008).This finding connects brain function during a set-shifting task with core symptoms of ASDs. Taken together, these findings suggest set-shifting as a potentially valuable intermediary between ASD symptoms and neurobiological functioning.

To investigate this possibility further, a reliable measure of set-shifting with as few extraneous (cognitive and motor) demands as possible and with an established link to brain function is needed. Investigations of set-shifting in ASD have typically relied on two measures of flexibility: the Wisconsin Card Sorting Test and the ID/ED Test. The former is criticized for its lack of specificity to flexibility (Hill, 2004), social administration (Ozonoff, 1995), and language demands (Liss et al., 2001). In contrast, the ID/ED Test measures flexibility in a systematic fashion that allows for controlled increases in shifting demands (Cambridge Cognition, 1996). It consists of nine stages, which encompass three types of tasks: simple object discrimination; shifting within a single dimension (e.g. shape: intra-dimensional (ID) shift); and shifting attention from one dimension to another (e.g. ignore shape and attend to previously ignored line: extra-dimensional (ED) shift). Embedded in ID and ED shifting is a reversal shift that requires participants to maintain the same rule but select an alternate exemplar (see Figure 1). Success on ID/ED is measured both by the ability to pass a stage (number of stages completed) and by the number of errors made while passing a stage (errors to criterion). Experiments with monkeys have linked lesions to lateral and orbital regions of the prefrontal cortex with impaired performance on ED and ED reversal shifts, respectively (Dias et al., 1996). Atypical development of orbital (Girgis et al., 2007) and lateral (Carper and Courchesne, 2005) prefrontal regions in ASDs highlights the importance of assessing flexibility in ED shift and ED reversal shift conditions.

While ID/ED is a promising task for investigating flexibility in children with ASDs, studies to date have yielded mixed results. Some investigations report deficits (Hughes et al., 1994; Ozonoff et al., 2000; 2004) while others do not (Edgin and Pennington, 2005; Goldberg et al., 2005; Happé et al., 2006; Landa and Goldberg, 2005). Conflicting results may reflect variability in the participants’ age and the stages of the task which are analyzed. Studies focusing on younger and narrower age ranges of high-functioning children have not reported performance on ED reversal shifting and have not found ED shift deficits in ASD groups. Two of the three studies finding deficits on the ID/ED involve ASD groups with mean ages in the teens, and the third involves cognitively impaired children. In addition, both of the studies that analyzed ED reversal shifting found deficits in the ASD group. In summary, previous investigations document deficits in ED shifts and reversal shifts in high-functioning adolescents and adults with ASD, while investigations of primary school age children have failed to find deficits in ED shifts and have not reported on ED reversal shifts (see Table 1 for a summary of ID/ED studies in ASD).

The current study attempts to address this gap in primary school age children (6–13 years) by probing both ED shifting and ED reversal shifting in high-functioning children with ASDs and typically developing (TD) controls matched on age, gender ratio, and IQ. We focus on higher-functioning individuals because ASD-specific difficulties are best observed in individuals whose deficient performance is not confounded with general cognitive impairments (arguments made by Hill and Bird, 2006). In light of previous work with smaller samples in our age range we predict group differences in errors, but not in the number of stages completed (Goldberg et al., 2005; Landa and Goldberg, 2005). Consistent with this hypothesis, Luciana and Nelson (2002) report a lack of sensitivity of the stage success criterion beyond 9 years of age. Given conflicting evidence regarding increased errors during the ED shifting stage, we constrain our hypothesis to an expectation of increased errors in the ASD group on the ED reversal stage, as reported by Ozonoff and colleagues (2004). In addition, in line with recent evidence that shifting relates to restricted interests/repetitive behaviors (Lopez et al., 2005; South et al., 2007), we predict a positive relationship between the number of errors committed during these two stages and restricted interest/repetitive behavior symptoms in the ASD group.

The current investigation not only documents ED reversal shifting deficits but also, for the first time, links these deficits with restricted interest/repetitive behavior symptoms among high-functioning children with ASD. Our finding of an ED reversal shifting deficit for high-functioning children in the full and subset analyses supports the two previous studies which have reported on ED reversal shifting in ASD. The present study extends findings from a study of low-functioning children with ASDs (Hughes et al., 1994) and a study including a combined group of children and adults with high-functioning ASDs (Ozonoff et al., 2004). Our findings are consistent with, and expand on, previous investigations of high-functioning primary school age children with ASD, which reported intact ED shifting but did not report on ED reversal shift performance.

ED reversal involves reversing a valence assignment, switching the reward value from positive to negative for a stimulus, and the orbitofrontal cortex plays a key role in stimulus–reward assignments (Loveland et al., 2008). Associating a stimulus and reward within a social interaction (i.e. reward is meted out by a person) is impaired in young children with ASDs (Dawson et al., 2002). Loveland and colleagues reduced the level of social demands and still reported reversal deficits in primary school aged children (Loveland et al., 2008). The present findings extend this line of investigation by demonstrating ED reversal deficits in a computerized task that has no social demands, suggesting that the orbitofrontal-limbic network may also underlie non-social reversal learning deficits.

The correlation between ED reversal errors and restricted interest/repetitive behavior symptoms highlights the utility of reversal set-shifting as a potential intermediate phenotype for ASDs. As discussed above, intermediate cognitive phenotypes serve as a link between behavioral symptoms and brain and genetic assays. In line with past studies, the present findings confirm the relationship between set-shifting difficulties and restricted interest/repetitive behavior symptoms in children with ASDs (Kenworthy et al., 2008; Lopez et al., 2005; South et al., 2007). Further investigation of ED reversal shifts might include other groups of children with developmental disorders, however, in order to determine the specificity of these findings to ASDs alone. In addition, it would be important to investigate whether this relationship occurs across all ages and functioning levels of individuals with ASDs. Previous studies investigating ED reversal shifts did not investigate the relationship between ED reversal shifting performance and ASD symptoms (Hughes et al., 1994; Ozonoff et al., 2004).

One puzzling aspect of the current findings is the relatively unimpaired ED shifting but deficient ED reversal shifting in our high-functioning ASD group. Past studies of individuals with ASDs of similar ages and functioning levels also report successful ED shifting (Edgin and Pennington, 2005; Goldberg et al., 2005; Happé et al., 2006; Landa and Goldberg, 2005), but the one study probing ED reversal shifting found deficits in both ED shifting and ED reversal shifting (Ozonoff et al., 2004). Two potential interacting confounds may explain this apparent discrepancy of findings: (1) the inclusion of a substitution score of 25 errors for stages not attempted, and (2) the age groups tested in previous studies. The substitution score inflates variance and this may obscure group differences of modest effect size in the ED shift stage. Studies reporting no ED shifting deficits for ASD groups tested mostly primary school aged samples. Primary school aged samples may require a higher number of substitution scores for both ASD and TD groups during the ED shift stage than an older, particularly TD, sample. This discrepancy in the utilization of substitution scores may lead to reduced variability in the older sample which increases the power to detect differences of moderate effect size.

While the present study provides novel insights into the set-shifting abilities of high-functioning children with ASDs, several unanswered questions remain for future investigations. Examination of the relationship between these deficits and restricted interest/repetitive behavior symptoms was coarse due to utilization of a summary score (i.e. the restricted interest/repetitive behavior symptoms scale from the ADI/ADI–R), and would be better investigated using a continuous measure, such as the Social Responsiveness Scale (Constantino and Gruber, 2005). Future investigations of ED reversal shift deficits may also elect to segregate this behavioral symptom construct and probe whether ED reversal shifting deficits relate to higher-order cognitive rigidity, reflected in resistance to change, insistence on sameness, and rituals, or to lower-order repetitive and sensory behaviors, reflected in stereotypies and self-stimulation. Additionally, the high average IQ of our ASD group may limit generalizing the current results to lower-functioning children with ASDs.

Strengths of the study include the large, well-characterized ASD sample within a narrow age range, which provided adequate power to detect group differences and significant correlations. Children with ASDs were well matched group-wise (with a two-to-one ratio) for age, IQ, and gender ratio; this matching was sustained on the subset analysis. The current investigation also presents ID/ED data on the largest exclusively pediatric ASD sample published to date (n = 42), which allowed matching the ASD and TD groups on performance. While the correlation analysis was conducted on a subset of the total sample (n = 27), raising issues of power or sample stability, this sample size was equal to or larger than full samples reported earlier (Edgin and Pennington, 2005; Goldberg et al., 2005; Landa and Goldberg, 2005). Furthermore, the significant correlation observed between the number of ED reversal shift errors and restricted interest/repetitive behavior symptoms was robust and immune to the influence of outliers because we used non-parametric (Spearman’s rho) correlations. Thus, we avoided limitations of previous investigations by recruiting a large group of high-functioning children with ASDs over a narrow age range when dynamic developmental gains in set-shifting skills are observed (Luciana and Nelson, 2002), and by examining group differences in performance on both the ED shift and ED reversal shift tasks.