Autism is characterized by impairment in verbal and nonverbal communication, reciprocal social interaction and a restricted repertoire of activities and interests . The symptoms fall on a continuum of severity referred to as autism spectrum disorder (ASD), in which impaired social functioning is the hallmark feature across the spectrum. Adequate social knowledge relies on the ability to interpret another person's behavior, to interact in both complex social groups and close relationships, to empathize and to predict how others will feel, think and act.
Play is critical for the development of social, cognitive and motor skills [2, 3]. Even though poor reciprocal social interaction is the hallmark deficit in autism, surprisingly few observational studies of play exist [4–7] and they primarily utilize only questionnaires or contrived laboratory conditions. The type of environment and context can affect social interaction, with enclosed environments facilitating imitation and gross motor play in autism [8, 9]. Playground observation can assist in screening for ASD  and may serve as an ecologically valid approach to elucidate social and psychobiological profiles in autism, including social stress.
Stress and anxiety are technically different constructs, and both have been reported from the earliest conceptualizations of autism  and alluded to in classification systems [11, 12]. In this study, stress refers to the response to perceived threat to the physiological or psychological integrity of an organism, often leading to the increased release of glucocorticoids (e.g., cortisol) [e.g., 13, 14]. Anxiety, on the other hand, pertains to a feeling of apprehension or worry that may be state- or trait-based (e.g., ). Many children with autism have significant anxiety [16, 17], and notable physiological stress has been reported [18–20]. Furthermore, these constructs can be closely linked and may co-occur.
Limited research has examined how stress may be associated with aspects of social functioning in autism [21–23]. While many children find social interaction stress-reducing, children with autism often appear to find social interaction stress-inducing. Increased stress and anxiety in autism may be the result of dysfunction of the amygdala, a brain structure involved in the detection of threats and mobilizing an appropriate behavioral response [24, 25]. The amygdala is fundamentally involved in social cognition  and is a key regulator of the limbic hypothalamic pituitary adrenocortical (LHPA) axis and a mediator of processive stimuli .
The LHPA axis is highly regulated, and the system is dependent on the ability to maintain, respond and reset itself through a homeostatic process involving three primary interrelated processes: the maintenance of a diurnal rhythm, activation in response to stress or threat and the restoration of basal activity via negative feedback mechanisms. Once activated, a neuroendocrine cascade is initiated which results in the release of glucocorticoids from the adrenal gland. In humans, cortisol is the primary glucocorticoid, which, once released, results in a suite of metabolic changes and engages in a negative feedback loop to return the system to basal levels. Thus, cortisol is a widely used biological marker of both stress activation and restoration of homeostasis. Although the concentrations of cortisol in saliva are lower, it has been established that the correlation between plasma cortisol and salivary cortisol is high (0.71 to 0.96) and that the temporal changes of cortisol in saliva closely mimic those in blood in response to potentially stressful events [28, 29]. Thus, salivary cortisol provides a useful, noninvasive biomarker for use with children.
The rhythmicity and responsivity of the LHPA axis in autism have been investigated, revealing abnormalities in neuroendocrine function, including an exaggerated stress response to various environmental events [18–20]. However, notable variability, exposure history and other factors may influence LHPA regulation and responsivity in autism [22, 30, 31]. Developmental research shows that stress reactivity is dynamic and that responsivity may vary based on context  as well as age and approach behavior in children . Moreover, social variables can induce, enhance or diminish the stress response .
Despite the growing literature, research has been scant in terms of the influence of social factors, especially under more ecologically valid paradigms [22, 35]. Autism is heterogeneous, and the social behavior that defines it is diverse . It may be the case that differences in social behavior may also reflect distinct underlying psychobiological profiles related to LHPA responsivity .
The purpose of the investigation was to evaluate cortisol responsiveness in a naturalistic playground social setting. A well-established, direct observational measure of natural social behavior with peers in autism has been lacking. This led to the establishment of our peer interaction paradigm, which was developed with colleagues on the basis of transactional behavioral measures of social behavior in nonhuman primates [36–38], social initiation in autism  and clinical expertise in observational techniques in autism [39, 40]. This ecologically valid design permits the careful investigation of social interaction in a play-based paradigm. It also allows real time assessment of social variables to be directly compared to stress reactivity.
In contrast to many studies of physiological responsivity which inherently aim to solicit a stress response , the paradigm was designed to emulate a "real life" playground to determine whether such environments would be deemed physiologically stressful. The peer interaction described below includes only three children and involves solicited but not forced interaction. There was no evaluative threat or peer rejection, both of which have been shown to activate the LHPA [41, 42]. We reasoned that if a participant exhibited social stress under such benign conditions, then responses may be even more notable on a typical school playground with many children, enhanced stimulation and challenging social exchanges. On the basis of previous research showing enhanced stress responsivity in children with autism [18, 22], we hypothesized that many children with autism would show an increase in salivary cortisol following the peer interaction compared to their average afternoon home level, baseline (arrival/acclimation) level and the cortisol values of neurotypical children. Our previous work led us to predict that variability in cortisol responsivity would be evident . Thus, we hypothesized that social and biological phenotypes would be evident; specifically, enhanced social interaction would result in an increase in cortisol in the autism group, whereas more social interaction would be associated with reduced cortisol in the neurotypical group. Since it was a new paradigm, specific a priori hypotheses regarding the behavioral coding variables were not considered beyond the total social interaction time.