Substantial recent progress in understanding the typical development of EF provides further reason to investigate individual differences in EF as a possible source of the heterogeneity in outcomes in autism. First, it is clear that EF is intimately tied to another aspect of neurocognitive development known to develop atypically in autism: theory of mind (ToM: mental state awareness). Numerous studies report robust correlations between individual differences in tasks tapping ToM and EF, independent of the effects of age and IQ, in typical preschoolers [24–27] and toddlers [28, 29]. Several theorists have further proposed a direct, functional link between EF and ToM. They assert that the abilities to monitor one's actions and to act with volition are critical for reflecting on the mental states of self and other, that is, that ToM emerges from EF [30, 31]. Research findings have repeatedly shown that early EF skills are predictive of later ToM but early ToM skills are not predictive of later EF [27–29], providing overwhelming support for the view that EF is critical to developmental changes in ToM.

Second, early EF is also predictive of typical preschoolers' later school readiness [32, 33] and academic success in reading and mathematics [34–36]. The transition to school relies on mastery of basic EF skills, including remembering and following instructions, completing tasks independently and smoothly transitioning between tasks, and inhibiting inappropriate behaviors. EF, therefore, is held to play an important role in the acquisition of knowledge; the better children are at focusing and refocusing their attention, holding information in mind and manipulating it and resisting distraction, the better placed children should be to acquire knowledge and skills in the classroom.

Third, it is well known that the prefrontal cortex, which mediates EF, shows a protracted development trajectory: it begins to develop very early in life, has a boost during the preschool period, and continues to develop well into adolescence [37]. The extended postnatal developmental of prefrontal cortical networks means that they should be particularly sensitive to exogenous influences. Several studies now report direct evidence of the malleability of EF: “exercising” young children's EF skills results in substantial improvements in their ability to regulate their behavior in both at-risk [38] and typical [39] populations.

Taken together, these findings provide compelling grounds for suggesting that one source of the heterogeneity in autistic individual's functional outcomes are individual differences in emerging EF. Some researchers have begun to speak to this possibility, with several studies reporting links between EF and a range of concurrent behaviors, including social competence (e.g., [18, 40]) and everyday adaptive behavior (e.g., [41, 42]). Yet full consideration of this developmental issue has not yet been possible due to a paucity of studies analyzing individual differences within a longitudinal design. The handful of studies using such an approach have done so, though, with dramatic success. Griffith et al. [20] found links between early executive skills (performance on a spatial reversal task tapping cognitive flexibility) and sociocommunicative behavior (bids for joint attention) one year later in young children with autism, but not in nonautistic children with developmental delay. Also, individual differences in set-shifting ability predicted the social competence scores of cognitively able adults with autism 3 years later in one study [43] and, in another, predicted real-life adaptive skills between 11 and 27 years later [5].

A recent study has further demonstrated the influence of autistic children's early EF skills on their sociocognitive awareness. Pellicano [18] investigated the EF and ToM skills of 37 cognitively able children with autism and 31 typical children (M age = 5 yrs 6mths). As expected, children with autism showed difficulties in both domains compared to typical children. There were also significant EF-ToM correlations, suggestive of a functional link between domains. Analysis of patterns of “atypicalities,” however, revealed striking dissociations in one direction only: poor ToM coupled with intact EF [18]. In line with Russell [31], these findings suggested that EF skills might be one important ingredient for the development of ToM but that ToM does not play this role for the development of EF.

Three-year follow-up data on the same samples of children further supported this claim [54]. For children with autism, individual differences in early EF skills (Time 1) were longitudinally predictive of developmental change in ToM skills (Time 2), independent of age, language, nonverbal intelligence, and early ToM skills. Yet there were no predictive relations in the opposite direction—a finding that is entirely consistent with longitudinal studies of the EF-ToM relation in typical children [27–29]. These results provide initial evidence that autistic children's cognitive skills emerge within a dynamic developing system where EF skills play a critical role in shaping the developmental trajectories of ToM.

Studies using similar methods with school-age children in part support this integrated framework [63, 64], although recent studies with 2- to 6-year-old children have instead reported evidence in line with a unitary model of EF [65, 66]. These latter results question the apparent continuity in the structure of EF during development but are consistent with a dynamic, neuroconstructivist approach in which cognitive functions begin relatively undifferentiated and become progressively modularised or specialised over time [67] (see Figure 1). This framework suggests that individual differences in development itself might be key to explaining the wide variation in findings both within [68] and across [12, 16] studies on EF in autism.

No study has yet explicitly investigated the nature of EF in autism. Early fractionation of EF makes it plausible for distinct EF components, such as cognitive flexibility, to be specifically affected in autism. Yet if, as the evidence suggests, EF is a single, unitary construct during early childhood (at least in typical children), then it becomes more difficult to see how a distinct profile of EF difficulties might emerge in autism. Given the prolonged development of EF and the degree of neural plasticity during childhood [69, 70], it is likely that emerging prefrontal cortical networks affect, and are affected by, the development of other key cognitive functions. In this case, then, disruption to distinct EF components in autism might be driven by other factors.

One goal therefore should be to understand precisely which mechanism(s) drives the development of EF in typical children and in children with autism. Some theorists propose that progress in EF occurs via the development of the prefrontal cortex [60] and the strengthening of prefrontal representations [58] in an experience-dependent manner [71]. An influential yet contrasting view suggests that development in children's attentional control—the ability to focus on a task and ignore irrelevant information—is the source of common variance in EF [61, 72]. Posner et al. have demonstrated significant advances during the preschool period in the central “attention network,” which includes alerting, orienting, and executive attention processes [73, 74]. Developmental changes in attention are considered to provide children with greater executive control over action. On this view, then, rate of growth in EF should be predicted by developments in attentional capacities (Figure 2(a)) and such developments might even mediate the fractionation of EF.

Impairments in core attentional processes have been reported in autism, including problems with disengagement or so-called “sticky” attention [75–77]. Fundamental problems in critical aspects of attention could therefore place limits on the rate of EF development, which could in turn hinder the emergence of autistic children's social and learning outcomes. It is of course possible that the causal relationship might exist in the opposite direction, such that early developments in EF might influence the emergence of attentional networks. The relationship between aspects of attention and components of executive control has, however, been hitherto unaddressed in autism.

Yet Posner and Rothbart's [72] model (see also [61]) neglects the potentially mediating role of another key function: language. Children's verbal skills can affect the expression of EF, for example, by limiting their ability to store phonological information in working memory [78]. Yet language might play a more fundamental role, affecting the very development of EF. For several theorists [31, 79–81], language provides an internal plan for behavior. Vygotsky [81] emphasized the importance of self-directed speech during early childhood, which becomes increasingly internalized during the preschool years, allowing children (verbally) to “think through” problems and to guide future-oriented behavior. Similarly, Zelazo et al. [59] stress that language is the medium through which higher-order (if-if-then) rules are formulated and is key to recursive thought. Developmental gains in young children's language skills (specifically their ability to formulate hierarchical rules) are therefore directly implicated in the rate of EF development (Figure 2(b)).

Difficulties with communication are a core characteristic of autism [82] and has been previously implicated as a potential limiting factor on the development of EF (e.g., [83]). Furthermore, children with autism are less likely to use verbal rehearsal strategies on executive tasks [84–86] suggesting that they, unlike typical children, may not be using internal language in the service of executive control. Pellicano's [54] longitudinal work showed that autistic children's initial receptive-vocabulary skills were not predictive of EF performance three years later, suggesting further that verbal skills may not influence the emergence of EF in autism as they do in typical development [26, 28]. Individual differences in the growth trajectories of autistic children's verbal skills therefore might partially mediate (or fail to mediate at all) the rate of progress of EF in autism. Further still, it is possible that attention and language could mediate the development of EF. In this case, both functions could make independent contributions to the rate of growth of EF (Figure 2(c)) and both could be limiting factors in autism. Importantly, evidence for any one of these patterns (Figures 2(a)–2(c)) would suggest that the potential influence of EF on children's functional outcomes is indirect rather than direct.

All of these models suggest that the development of EF itself might be shaped by certain endogenous factors, which in turn could mediate children's developmental outcomes. Yet the developmental trajectory of EF, and its resulting neurocognitive architecture, will be an emergent property of interactions within the children and between the children and their environment. Alternative explanations therefore place the development of EF squarely in the social realm. For example, Luria emphasized that “we must go beyond the limits of the individual organism and examine how volitional processes are formed for the child in his/her concrete contacts with adults” [87, page 89] (see also [88]). Hughes [27] (see also [80]) extended this view to suggest that the effect of EF upon the development of ToM should be indirectly influenced by the child's social environment. Since negotiating social interactions requires children to regulate their own behaviors (e.g., turn-taking, following rules in games), peer relations are likely to have positive effects on children's developing executive skills, which in turn will foster their developing ToM.