Children with autism were recruited as part of a longitudinal study using magnetic resonance imaging (MRI) to investigate early brain development in autism. Fifty-three children who enrolled between 18 and 35 months of age (i.e., 2 years; Time 1) participated in the present investigation. provides chronological age and Mullen age equivalence for groups at each time point. As part of the longitudinal MRI study, children with autism were invited to participate in a 24 month follow-up assessment (i.e., 4 years of age; Time 2). Twenty-seven of the initial fifty three children participated at time 2. As will be reported below, there were no time 1 differences in age, IQ, or social orienting rates for the children with autism who dropped out after time 1 and those who participated at time 2. Thus, despite the high dropout rate, our longitudinal sample is representative of the initial cohort.
Sample data for autism and typically developing (TD) children
Thirty-five typically-developing (TD) children were recruited separately for the present study (i.e., they did not participate in the longitudinal MRI study). These children served as cross-sectional controls (15 TD children participated at time 1; 20 different TD children participated at time 2); no longitudinal control data was available. TD children were matched to participants with autism on age and gender.
Children with autism were referred to the longitudinal MRI study from nine specialty clinics for pervasive developmental disorders. Children were included in the study if they met Autism Diagnostic Interview-Revised (ADI-R) algorithm criteria for autism, obtained ADOS scores consistent with autism, and met DSM-IV criteria for Autistic Disorder. As part of the MRI study, children were administered the Mullen Scales of Early Learning and a standardized neurodevelopmental examination. They were excluded from the study if they had evidence of a medical condition thought to be associated with autism (including fragile X syndrome and tuberous sclerosis), gross central nervous system injury (e.g., cerebral palsy, significant perinatal or postnatal complications or trauma, drug exposure), seizures, or significant motor or sensory impairments. Because 18–35 months is younger than the usual age of a diagnosis of autism, many of the children recruited for the present study may have been more severely affected than the general population of children with autism.
TD children were recruited from a mailing sent to local families of newborns. Families were contacted if they returned a postcard indicating that they wished to participate in future research projects and had a child of appropriate age. Parents were interviewed informally to determine if they had any indication that their child was developmentally delayed or had a history of neurological injury or disorder. Parents who had concerns about their children, whose children currently were receiving developmental evaluations, or whose children had been identified as developmentally delayed were excluded from the study. Children with a history of neurological injury or disorder (e.g., cerebral palsy, significant obstetric complications or perinatal or postnatal trauma, drug exposure, seizures, or significant motor or sensory impairments) also were excluded.
As part of the protocol for the longitudinal MRI study, children with autism were administered the ADOS and Mullen Scales of Early Learning. Additionally, one of their parents or caregivers participated in the VABS interview. TD children were recruited for the present study and then administered the ADOS and Mullen Scales of Early Learning in a single laboratory or in-home session. Parents of TD children did not complete the VABS.
The Mullen Scales of Early Learning. (Mullen, 1995)
The Mullen Scales of Early Learning assess language, motor, and visual perception abilities for children from birth to age 5 years, 8 months. The Mullen was used in the present investigation to obtain a single, reliable, and valid estimate of IQ for participants at both time 1 and time 2. A limitation of the Mullen is that it has a restricted distribution of standardized scores for lower functioning individuals (i.e., subscales only provide a < 50 Standard Score). Several of the lower functioning participants in the present investigation failed to reach a basal on the Mullen. As a result, mental age equivalents were used for all participants. In order to provide a greater range of IQ scores in the present sample, an average mental age equivalent across four subscales (Visual Perception, Fine Motor, Receptive Language, and Expressive Language) was used as an overall IQ measure.
Vineland Adaptive Behavior Scales (Vineland; Sparrow, Balla, & Cichetti, 1984)
The VABS is a standardized caregiver interview that yields information on several domains of adaptive functioning. These domains include: Communication, Daily Living Skills, Socialization, Motor Skills, and Problem Behaviors. Ratings for VABS items are categorical: DK = don’t know, N = No opportunity, 0 = no never, 1 = sometimes or partially, and 2 = yes, usually.
Social Orienting Continuum and Response Scale (SOC-RS)
The SOC-RS is an observational coding system applied in this study to previously recorded ADOS sessions. In the present study, all time 1 children with autism and 23 of the children with autism studied at time 2 were examined with module 1 of the ADOS (the most basic module). Module 2 of the ADOS is intended for children with phrase speech; four children with autism and 4 TD children were administered Module 2 at time 2. ADOS sessions were included only if children were observable on camera for more than 10 minutes. ADOS sessions were converted to CD and then coded with the SOC-RS using NOLDUS Observer 5.0 Video Pro software (International Headquarters 2003
). The Observer Video analysis program allowed for on-line continuous coding of each event or trial.
The SOC-RS provides ratings for 4 social orienting behaviors (i.e., social referencing, joint attention, orienting to name, social smiling) elicited during ADOS sessions. Each variable of the SOC-RS will be described in general here, and Appendix A
contains detailed coding guidelines.
Social referencing is defined as an event in which the target child fixates his or her attention on the face of another individual. Children must fixate for greater than 2 seconds in order to be coded. Referencing is necessary for all SOC-RS codes, although to avoid redundancy, events are categorized as referencing only if they do not meet criteria for joint attention, orienting to name, or social smiling. Social referencing was coded as rate per minute of observable time and then converted to z-scores.
Joint attention (JA) is coded when children either follow someone else’s attention towards an object (responding) or evoke another person’s attention and direct it towards an object (initiating). Children are scored as JA ‘responders’ only if they follow the examiner’s shifts in eye-gaze during JA trials of the ADOS. Because the SOC-RS focuses on behaviors involving attention to faces, children are scored as ‘non-responders’ if they only follow bids using pointing or if they do not follow the examiner’s JA bids. JA responding was included to increase variability in JA scores; few children with autism initiate JA, although a greater number will respond to a JA bid. JA initiation was coded as rate per minute of observable time. A ‘JA total’ (JAT) score is computed by standardizing JA responding and JA initiating scores separately and then averaging the two z-scores.
Orienting to name is scored only during the ‘response to name’ press of the ADOS. The ADOS response to name press includes 4 trials in which the examiner calls the child’s name, and then 2 trials in which the parent attempts to acquire the child’s attention by calling his/her name. Because the number of trials administered and the method for acquiring the child’s attention are standardized within the ADOS, this press was examined. The press is discontinued as soon as the child orients to the examiner’s or parent’s voice. SOC-RS orienting to name scores reflect the trial on which children first respond to their name being called.
Social smiling events are recorded when a child references another person while smiling in order to share enjoyment. Children must fixate for greater than 2 seconds and the smiling must be judged by the rater to be appropriate to the context in order for social smiling to be coded (i.e., self-stimulatory laughing combined with social referencing was not scored). Social smiling was coded as rate per minute of observable time and then converted to z-scores.
A social orienting composite was calculated by averaging the standardized ratings of children’s social referencing, JAT, name trial, and social smiling. This composite was examined along with the four constituent variables.
To establish reliability of the SOC-RS, raters independently coded 15 videotaped ADOS sessions two times. Reliability was calculated using intra-class correlation coefficients (ICC’s). After establishing an intra- and interrater reliability score of >0.8 across these 15 cases, each rater independently coded cases for final analysis. Reliability was calculated separately for each SOC-RS item. Agreements were scored if both raters identified the same behavior within a 3 s window. Then, raters met and discussed their ratings and coding decisions. If disagreements occurred, raters viewed the session together and discussed their ratings until a consensus was reached. Once raters were reliable, they independently coded the remainder of the videos.
ANCOVA models were used to test group differences in social behavior. Main effects of autism versus TD were examined with age, IQ, and gender entered as covariates. Age and IQ were centered for all analyses to control for multicollinearity. All 2-way interactions with group were included initially to check the assumption of parallelism, but non-significant interactions were dropped. Univariate analysis was used as a first step to examine differences in the social orienting composite between groups. Multivariate analyses were used to compare group rates on the individual SOC-RS variables.
Because longitudinal data were not available for the TD group, within group differences between time 1 and time 2 were examined with separate ANCOVA models for each group. IQ and gender were entered as covariates and time (1 vs. 2) was entered as the predictor variable. The social orienting composite and the individual variables were entered as outcomes.