Abstract
Keywords
Sensory processing problems are common among people with autism spectrum disorders (ASD), with estimates ranging from 30% to 94% (Baranek, David, Poe, Stone, & Watson, 2006; Crane, Goddard, & Pring, 2009; Dunn, Myles, & Orr, 2002; Leekam, Nieto, Libby, Wing, & Gould, 2007; Rogers, Hepburn, & Wehner, 2003; Rogers & Ozonoff, 2005). Most descriptions of these sensory problems implicate dysregulation of behavioral responses to sensory experiences, such as overreacting to ordinary stimuli or failing to notice stimuli that would be salient to peers.
Such atypical behavioral responses to sensations are thought to reflect underlying disruptions in physiological sensory modulation. Sensory modulation is the complex process by which neural messages about the intensity, frequency, duration, complexity, and novelty of sensory stimuli are adjusted by the central nervous system to enable adaptive behavior (Miller & Lane, 2000). When sensory modulation difficulties occur, the person may react adversely or fail to respond to sensations in daily life. Consequently, children with sensory modulation difficulties are hindered from full participation in daily occupations that provide them with a means to learn skills, develop relationships, and meet biological needs that support health and well-being (Jasmin et al., 2009; Lane, Young, Baker, & Angley, 2010; Parham & Mailloux, 2005).
Distinct patterns of sensory modulation difficulties, including over- and underresponsiveness, are commonly identified in occupational therapy behavioral assessments of children with emotional and behavioral problems (Parham & Mailloux, 2005). Overresponsiveness refers to a state in which the person is “disturbed by ordinary sensory input and reacts defensively to it, often with strong negative emotion, avoidance, and activation of the sympathetic nervous system,” whereas in underresponsiveness, the person “tends to ignore or be relatively unaffected by sensory stimuli to which most people respond” (Parham & Mailloux, 2005, pp. 410–411).
Behavioral manifestations of sensory modulation difficulties are thought to be mediated by the autonomic nervous system (Miller, Reisman, McIntosh, & Simon, 2001). Measures of skin conductance activity are well-accepted methods for studying autonomic responses, specifically sympathetic nervous system (SNS) responses (Dawson, Schell, & Filion, 2007). In this article, the term skin conductance (SC) refers to a measure of tonic SNS activity at rest, whereas the term skin conductance response (SCR) refers to a measure of phasic SNS activity in response to a specific stimulus.
The study described in this article examined SC of children with ASD at rest and their SCRs following the administration of controlled auditory stimuli. We focus on auditory responses in this article because the auditory system is the sensory system that is most frequently affected by modulation difficulties among children with a variety of developmental disorders, including ASD (Baranek, Foster, & Berkson, 1997; Gomes, Rotta, Pedroso, Sleifer, & Danesi, 2004; Greenspan & Wieder, 1997; Volkmar, Cohen, & Paul, 1986).
Several research groups have studied SNS activity of children with ASD using measures of SC at rest and SCRs following auditory stimulation, with inconsistent findings. Some researchers reported SCR data indicating that children with ASD demonstrated overresponsiveness to auditory stimuli (Barry & James, 1988; James & Barry, 1984; Palkovitz & Wiesenfeld, 1980), whereas SCR data from others indicated that children with ASD were underresponsive or nonresponsive to auditory stimuli (Stevens & Gruzelier, 1984; van Engeland, 1984). Researchers using the Sensory Challenge Protocol (SCP; Miller et al., 2001) found that some children with ASD demonstrated overresponsiveness, indicated by unusually high SCR magnitudes following presentation of a variety of sensory stimuli, including auditory, whereas others demonstrated underrespon-siveness, reflected in low SCR magnitudes (Schoen, Miller, Brett-Green, & Hepburn, 2008). When these researchers compared children with ASD with children with sensory modulation dysfunction but no other diagnosis, the ASD group showed significantly lower SC at rest and relatively low SCRs following sensory stimuli, suggesting a prevalence of physiological underresponsiveness (Schoen, Miller, Brett-Green, & Nielsen, 2009).
Purpose and Research Questions
The purpose of this study was to contribute to further understanding of the relationship between autonomic and behavioral sensory responses of children with ASD. Although occupational therapists generally assume that behavioral responsiveness reflects the degree of underlying sympathetic activation, research support for this assumption is very limited. We do not yet know whether sympathetic activation is different for children with ASD compared with typical children, whether subgroups exist of sympathetic over- and underresponders among children with ASD, or whether severity of behavioral responses systematically varies with degree of sympathetic responses to sensory experiences.
This study was designed to examine the association between autonomic and behavioral responses of children with ASD to auditory stimulation. The specific aims were to determine whether children with ASD differ from typically developing (TD) children in sympathetic activity at rest and following auditory stimuli and whether their sympathetic responses to controlled sound in a laboratory are associated with behavioral responses to sounds in everyday environments. The research questions were as follows:
Do children with ASD differ from TD children in their SC at rest and their SCRs to auditory stimuli?
For children with ASD, is there a relationship between their SCRs to sound in a lab and their usual behavioral reactions to sounds as reported by parents?
Method
Research Design
This study used a case-control research design comparing children with ASD with TD children. This study is part of a larger program of research designed to elucidate the physiological basis of the sensory behaviors seen in children with ASD. Clarification of how autonomic responses of children with ASD are similar to or different from TD children is an important first step in this research program. Therefore, we chose TD children as the comparison group because our primary interest was in discerning whether the SCRs to auditory stimuli of children with ASD differ from those of TD children. The University of Southern California Institutional Review Board (IRB) approved this study, and we obtained signed consent forms and assent forms from all participants before beginning data collection.
Participants
Participants were ages 5–12 yr. Normative data (N = 1,051) on the Sensory Processing Measure (SPM; Parham, Ecker, Miller-Kuhaneck, Henry, & Glennon, 2007) indicate that behavioral responses to sensory experiences do not change meaningfully across this age range, because effect sizes for ages 5–8 versus 9–12 were small. Effect size of age for auditory processing (the Hearing scale) was particularly small (Cohen’s d = 0.13). Therefore, our selection of this age range minimizes the potentially confounding effects of maturation on behavioral responses to auditory stimuli that might be evident in a younger age group.
Children with ASD were able to communicate verbally and were previously diagnosed with autistic disorder by a psychiatrist or psychologist using the Diagnostic and Statistical Manual of Mental Disorders (4th edition; American Psychiatric Association, 2000). They were recruited from private occupational therapy clinics, autism associations, and state-funded centers for families of children with developmental disabilities using IRB-approved fliers. To be included in the ASD group, a child was required to score 15 or higher on the Social Communication Questionnaire (SCQ) Lifetime Form (Rutter, Bailey, & Lord, 2003), which we used to verify that social communication functioning was consistent with a diagnosis of ASD. Children with ASD were excluded if they had a condition known to affect neurological development, such as seizure disorder or Fragile X syndrome, or were taking medications that might affect autonomic activity, such as dextroamphetamine and amphetamine (Adderall), beta blockers, oxybutynin (Ditropan), selective serotonin reuptake inhibitors, or methylphenidate (Ritalin).
None of the TD participants had a history of sensory processing problems or disability. To verify typical sensory functioning, children in this group were required to have a T score no higher than 59 for both the Hearing scale and the total score of the SPM Home Form (Parham & Ecker, 2007). The TD children were a convenience sample recruited by word of mouth from the families and family friends of participants with ASD and through IRB-approved fliers at community sites.
Procedures
Data were collected for each child in one laboratory visit. After obtaining informed consent from the parent and assent from the child, the experimenter introduced the child to the laboratory by framing it in a child-friendly way as a pretend spaceship where the child would experience different sensations. While the parent completed the SPM, the child was seated in the lab, where electrodes were placed on the thenar and hypothenar eminences of the right hand to collect SC data. Laboratory procedures were video recorded to aid in detection of artifacts that might occur from child movement or incidental sensory stimuli outside the lab procedures.
The experimenter administered the SCP in the laboratory, a procedure that took approximately 45–90 min, including preparation. It included three phases: baseline, sensory challenge, and recovery. Baseline data were recorded for 3 min without any stimulation. Next, six different sensory stimuli were presented, including two auditory stimuli (standard tone at 84 dB and siren at 78 dB). Each stimulus lasted 3 s and was presented in 8 contiguous trials. After each stimulus was presented 8 times, a pseudorandom pause of 12–17 s occurred before the next stimulus was initiated. A final 3-min recovery period followed the presentation of all stimuli.
Instrumentation
Sensory Processing Measure.
We used the SPM Home Form to measure behavioral responses to naturally occurring sensations. This standardized norm-referenced parent questionnaire was designed to identify sensory processing problems in children aged 5–12. Parents answer questions about frequency of specific child behaviors using a 4-point Likert-type scale. In this study, we analyzed SPM T scores for auditory processing (Hearing scale) and composite sensory functioning (Total scale). Higher T scores indicate greater frequency of atypical behavioral responses to sensory experiences. As reported in the SPM manual (Parham et al., 2007), internal consistency is high for the Hearing scale score (α = .84) and excellent for the Total scale score (α = .95). Test–retest reliability is excellent for both scales (r = .95 for the Hearing scale and .98 for the Total scale). Validity of items and scale scores is supported by factor analysis, item–scale correlations, and Rasch analysis. Scale scores demonstrate strong effect sizes (Cohen’s d = 1.05 for the Hearing scale and 1.15 for the Total scale) in differentiating between TD children and children with known clinical disorders (Parham et al., 2007).
For the purposes of this study, we created two SPM subscales to measure behavioral over- and underresponsiveness to auditory stimuli. Each subscale was scored by summing item ratings for each participant. To construct the overresponsiveness scale, we compiled the six Hearing scale items that the SPM manual categorizes as measuring overresponsiveness:
Does your child seem bothered by ordinary household sounds, such as the vacuum cleaner, hair dryer, or toilet flushing?
Respond negatively to loud noises by running away, crying, or holding hands over ears?
Seem disturbed by or intensely interested in sounds not usually noticed by other people?
Seem frightened of sounds that do not usually cause distress in other kids?
Seem easily distracted by background noises such as a lawn mower outside, an air conditioner, a refrigerator, or fluorescent lights?
Show distress at shrill or brassy sounds, such as whistles, party noisemakers, flutes, and trumpets?
Because the SPM manual classifies only one item in the Hearing scale as measuring underresponsiveness (“Does your child appear not to hear certain sounds?”), we added two new questions to this existing question to create the underresponsiveness subscale: “Does your child seem to underreact to loud noises? Ignore you when you call his or her name?”
Skin Conductance Measures.
SC measures detect changes in skin conductance that occur due to the activity of eccrine sweat glands, which are innervated by the SNS (Dawson et al., 2007). When SNS activity increases or decreases, the eccrine sweat glands increase or decrease their production of sweat. The corresponding change in amount of water on the skin surface directly affects the capacity of the skin to conduct electricity—that is, SC. In this study, SC was measured at rest during the baseline and recovery periods of the protocol, and SCRs were measured after presentation of each sensory stimulus during the SCP. A change in SC level following a stimulus was counted as an SCR when the amplitude of SC change was ≥ 0.05 μS and latency of the change was between 0.8 and 4.0 s after the stimulus was delivered. If the change was less than 0.05 μS, it was too small to represent a valid measure of a response (Boucsein, 1992; Dawson et al., 2007). A response occurring sooner than 0.8 s or later than 4 s after a stimulus was treated as unrelated to the stimulus because it appeared too soon or too late to reflect a valid response (Boucsein, 1992; Dawson et al., 2007).
As shown in Table 1, we examined five measures of SCR: amplitude, magnitude, habituation, onset latency, and rise time. Amplitude is the amount of change in SC from baseline to peak of response to a stimulus. In this study, amplitude was operationally defined as the average of the square root of each response from the eight trials of a particular stimulus, not including zero responses. Magnitude was operationalized in the same way as amplitude except that it included zero responses in the average for each set of eight trials. Habituation is the decrement in SCRs toward baseline level after repetition of the same stimulus. It was operationalized as the number of trials required before the first of two consecutive nonresponses to a stimulus. Onset latency is the time between onset of the stimulus and initiation of an SCR. Rise time is the time between the initiation and peak of an SCR.
SC at Rest and SC Responses to Tone and Siren
Note. ASD = autism spectrum disorder; SC = skin conductance; SCR = SC response; SD = standard deviation.
p < .05 after Rom adjustment.
Data Analysis
We first converted SC data to graphs and compared them with video recordings to detect and delete artifacts. The data were then transformed to compensate for skew according to recommendations by Dawson and colleagues (2007). For each participant, SC at baseline and recovery was logarithmically transformed as log (SC), and SCR amplitude and magnitude were calculated respectively as 
To examine whether children with ASD differed from control participants in autonomic functioning, we conducted t tests to detect group differences in the five measures of interest. We further analyzed the effects of age, gender, and ethnicity on SC and SCR using general linear models with least square means. To examine whether autonomic responses were related to behavioral responses for each study group, we computed Pearson’s correlation coefficients for SC and SPM variables, with age, gender, and ethnicity partialed out. In all of our analyses, the criterion for statistical significance was α < .05.
Results
We recruited 39 children with ASD and 31 TD children, but only 25 children in each group completed the study. Reasons for dropout included failure to complete the protocol, use of medications that might have affected SC, excessive movement artifacts, and ineligibility because of scores of ≤15 on the SCQ Lifetime Form. Parent reports indicated that all children with ASD were considered to be high functioning and attended regular classes at school. Table 2 describes participant characteristics. Groups did not significantly differ in age, but the proportions by gender and ethnicity were significantly different.
Participants’ Age, Gender, Ethnicity, and SPM Scores
Note. ASD = autism spectrum disorder; SPM = Sensory Processing Measure; TD = typically developing.
χ2 test applied.
Group Differences in Autonomic Responses
Table 1 presents results for our first research question: Do children with ASD differ from TD children in SC at rest and SCRs to auditory stimuli? Both of the tonic SC measures differed significantly between study groups. The ASD group demonstrated significantly higher SC at baseline and recovery compared with the TD group. For SCRs to the tone, the ASD group demonstrated significantly higher amplitude than the TD group. However, SCRs to the siren did not significantly differentiate between groups.
Because we performed multiple tests of significance, we applied the Rom adjustment, a sequentially rejective test procedure (Rom, 1990), to control for Type 1 errors. After Rom corrections, differences in SC at baseline and recovery and in amplitude of SCR to tone remained significant.
Three children with ASD did not respond to any of the tone trials, and 2 TD children did not respond to any of the siren trials. These participants were excluded from amplitude analyses because amplitude is operationalized to include only nonzero responses, and they were also excluded from the onset latency and rise time analyses, which are appropriate only when there is a response to a stimulus.
Confounding Effects of Gender, Age, and Ethnicity on Autonomic Activity
Gender, age, and ethnicity distributions differed significantly between study groups, raising concern that these variables may have affected SC measures in such a way as to create spurious group differences. We analyzed the effects of age, gender, and ethnicity using t tests and general linear models with least square means. These variables were dichotomized, allowing us to examine their interactions with the study groups. Results are in Table 3.
Effects of Gender, Age, and Ethnicity on SC Measures
Note. ASD = autism spectrum disorder; SC = skin conductance at rest; SD = standard deviation; TD = typically developing.
Effects of Gender.
Girls tended to have higher SC values than boys, but t tests indicated no significant gender differences within the TD group. The number of girls in the ASD group was too small to test for gender differences. Adjustments for gender and for the Gender × Study Group interaction yielded significant study group differences for amplitude and magnitude of SCRs to tone but not to siren and no significant main effects of gender or Gender × Study Group interactions. If gender had confounded these findings on SCR group differences, the study group with more girls would have had higher SCRs, but this was not the case. The ASD group had higher reactivity but significantly fewer girls, suggesting that gender differences did not account for the ASD and TD differences.
Effects of Age.
No significant Age × Study Group interaction was found, but younger children tended to have higher SC and were more reactive than older children for all measures except tone amplitude. Additionally, for the same age groups, ASD children showed higher SC and SCRs than TD children. A significant age effect on siren magnitude was found after adjusting for study group and Age × Study Group interaction, suggesting that age may be a confounder for this variable alone.
Effects of Ethnicity.
Ethnicity was significantly unbalanced, with more White than Asian children in the ASD group and the inverse in the TD group. We found no significant main effect or interaction effect on SC for ethnicity. However, as seen in Table 3, the ethnicity effect on magnitude of the SCR to the siren was significant, and the Ethnicity × Study Group interaction approached significance. The Asian TD children stood out as having a low response to the siren because of 3 nonresponders, whereas for the White children, the ASD and TD groups did not differ significantly. For both study groups, the White children had higher SC than the Asian children and were more reactive on three of the SCR measures. Thus, for the siren magnitude SCR, the marginally significant effect of study group seen in Table 1 may have resulted from the ethnicity imbalance between the groups.
Relationship Between Autonomic and Behavioral Responses
Our second research question was, Is there a relationship between SCRs to sound and behavioral reactions to sounds in daily life? We examined this question for each study group by computing intercorrelations among SC and SPM variables.
SPM Scores of Children With ASD.
As noted earlier, children in the TD group all had SPM scores below 60 (within normal limits). As Table 2 shows, 88% of the children with ASD scored at least 1 standard deviation beyond the normative mean on the Hearing scale, and 96% fell above 1 standard deviation on the Total scale score.
Intercorrelations Among SC and SPM Variables.
Table 4 depicts intercorrelations among SPM and autonomic variables for each group with age, ethnicity, and gender partialed out. For the ASD group, auditory overresponsive behavior was significantly correlated with amplitude and magnitude of SCR to tone, whereas auditory underresponsive behavior was significantly correlated only with magnitude of SCR to tone. The SPM Hearing T score and the Total scale score were significantly correlated with amplitude of SCR to tone. For the TD group, we found no significant correlations.
Intercorrelations Among SPM Scores and Autonomic Measures
Note. Amp = amplitude; ASD = autism spectrum disorder; Mag = magnitude; SC = skin conductance at rest; SPM = Sensory Processing Measure; TD = typically developing.
p < .05 after controlling for age, gender, and ethnicity.
Discussion
This study produced evidence that autonomic activity of children with ASD is related to their behavioral responsiveness. We found evidence of stronger sympathetic activation at rest and in response to auditory stimulation for children with ASD compared with TD control participants. We also found significant associations between sympathetic reactivity to an auditory stimulus and parent reports of problematic behavioral responses to sounds in natural environments. Specifically, we found that parent reports of problem behaviors on the SPM Hearing and Total scale scores and on both the overresponsiveness and underresponsiveness subscales were positively and significantly correlated with SCRs to tone for the children with ASD. These results indicate that the frequency of unusual sensory responses in everyday life is related to the degree of sympathetic activation displayed in the lab.
Our finding of overresponsiveness in the ASD group is consistent with older studies in which children with ASD displayed significantly higher SC levels and were more reactive to auditory stimuli than TD control participants (Barry & James, 1988; James & Barry, 1984; Palkovitz & Wiesenfeld, 1980; Stevens & Gruzelier, 1984; van Engeland, 1984). In two of these studies, a subgroup of children who were underresponsive was identified (Stevens & Gruzelier, 1984; van Engeland, 1984). Similarly, we found that some children with ASD were underresponsive to auditory stimuli: 3 participants did not produce SCRs after any presentations of the tone. All TD children in our study responded to this stimulus.
Our findings contrast with those of Miller and her colleagues (Miller et al., 2001; Schoen et al., 2009), who also used the SCP. These researchers averaged SCR across sensory modalities (auditory, visual, tactile, olfactory, and vestibular) and found a lower magnitude for children with ASD than for TD children. The current study examined only responsiveness to auditory stimuli, which may account for the different results. Miller and colleagues (2001) found evidence of a subgroup of overresponsive children with ASD, so it is possible that they recruited participants with ASD from sources with different distributions of over- and underresponsive children compared with our study.
The contrasting responses of typical children to the tone and siren were unexpected and may relate to the sensory contexts of everyday life. Although all TD children responded to the tone, 4 (16%) did not respond to the first presentation of the siren, and 2 of those children failed to respond to the siren across all trials. The occasional absence of an SCR to a stimulus appears to be a normal occurrence for 10%–25% of typical adults (Dawson et al., 2007). The children in this study lived in a densely populated urban area where sirens are usually heard several times daily, so the 2 TD children who did not respond to any presentations of the siren may have habituated to this sound.
An unexpected finding was that, for children with ASD, the correlation of the underresponsiveness subscale with SCR magnitude to tone was significant (partial r = .54) and positive. In other words, behaviors reflecting underresponsiveness were associated with higher sympathetic responses to the tone in the lab. This was a surprise because we expected that underresponsive behavior would be associated with lower sympathetic reactivity. Instead, our results indicate that children with ASD who have strong sympathetic reactivity may demonstrate underresponsive as well as overresponsive behaviors, perhaps reflecting general dysregulation in sensory modulation processes.
Limitations and Future Research
One limitation of our study is that 80% of the children with ASD were receiving occupational therapy, raising the possibility of clinical bias. Perhaps in the geographical area in which the study was conducted, children with ASD who are overresponsive to auditory stimuli tend to be selectively referred to occupational therapists. It may also be that parents who volunteer to participate in this kind of study are more likely to have children with disturbing overresponsive behaviors. Studies with careful controls for recruitment bias are needed to clarify whether over- or underresponsiveness is more common among children with ASD.
Group imbalances in gender, age, and ethnicity also posed limitations to our study. However, differences in SC and SCRs between the children with and without ASD remained after the influences of gender, age, and ethnicity were statistically extracted. Similarly, Venables and Mitchell (1996) found no significant effects of age, gender, or their interaction on SC of 640 Mauritian participants aged 5–25 yr. Still, our results should be interpreted with caution.
In our study, younger children tended to have higher SC at rest than older children, and girls tended to display higher resting SC and SCRs to auditory stimuli than boys. Gender is unlikely to have biased our results because the preponderance of girls in the TD group should have made it more difficult to detect the significant group differences we found. However, age differences may have biased results toward finding elevated SC among the children with ASD.
Ethnicity imbalance may have biased results, as well. We found that White children had higher SC and SCRs than Asian children, although these differences were not statistically significant. Boucsein (1992) suggested that differences in SC across different ethnic or racial groups may be attributable to the density of active sweat glands, which decreases as skin darkness increases. That is to say, ethnicity or race may appear to affect SC because of group differences in skin color. For example, researchers have reported that African-American people have lower SC levels than White people (Boucsein, 1992; Fisher & Kotses, 1973; Johnson & Corah, 1963) and that SC levels of Native Americans fall midway between those of White and African-American people (Korol & Kane, 1978). These reports suggest that darkness of skin color, rather than race or ethnicity, influences SC. Given the potential effects of age, gender, and darkness of skin, future research using SC measures should strive for a balance of these variables across diagnostic groups.
Implications for Occupational Therapy Practice
Occupational therapists who use a sensory integration or sensory processing approach for children with ASD may find the following implications of this study useful:
This study showed that children with ASD who demonstrate strong behavioral overreactivity to sounds, as measured by an elevated SPM Hearing T score, are likely to have high sympathetic reactivity to sound. Consequently, therapists should consider interventions for these children that address modification of daily routines and environments, as well as self-regulation strategies, to reduce the impact of disturbing auditory stimulation in daily life.
Children with ASD who demonstrate general sensory processing difficulties, as measured by the SPM Total T score, may have elevated sympathetic activation at rest and strong sympathetic reactivity, so for these children, therapists might consider interventions to facilitate development of sensory modulation and self-regulation across multiple sensory systems, to create predictable daily environments and routines, and to plan for anticipated changes in routine or environment.
Children who demonstrate both behavioral under- and overresponsiveness to sounds in everyday life may present strong sympathetic reactivity to sound. Until this phenomenon is better understood, we recommend that practitioners carefully assess the contexts, situations, and types of stimuli involved when sensory questionnaire items indicate that a child both underreacts and overreacts to auditory stimuli to tailor intervention to the needs of the individual child. Individual occupational therapy based on Ayres Sensory Integration® may be particularly appropriate in such cases, with special attention to development of self-regulation of alertness and affect for optimal engagement in occupations.
When educating parents, teachers, and policymakers about the sensory characteristics of children with ASD, this study could be cited as providing evidence that behavior problems related to auditory and other sensory stimuli are associated with physiological events that are not under the child’s voluntary control.
Footnotes
Acknowledgments
We are most grateful to our study participants, without whom this research would not have been possible. We also thank the following clinics, institutions, and associations for their help with recruitment: Center for Developing Kids, Dr. Susan Spitzer Clinic, Gallagher Pediatric Therapy, Play to Learn Center at the Glendale Adventist Hospital, Pediatric Therapy Network, Therapy West, Family Resources Center at the East Los Angeles Regional Center, Foundation for Disabled Youths, Chinese Parents Association for the Disabled, Parent’s Place of West Covina, Talk About Curing Autism, and Interactive Autism Network Research Database at the Kennedy Krieger Institute and Johns Hopkins Medicine–Baltimore sponsored by the Autism Speaks Foundation. This study was funded by the California Foundation for Occupational Therapy.