In the regressive pattern of onset, children appear to be developing typically for a year or longer, but then lose communication and social skills that they had previously acquired. Retrospective studies using both parent report and home videotape analysis have documented losses in a wide range of social communicative behaviors.^9–12 Regression is most often reported between the first and second birthdays, with mean ages across samples ranging from 16 to 20 months.^9,11,13–15 The prevalence of regression varies across studies and is dependent upon both sampling methods and definitions of regression. Two large population-based studies found regression in 15.6%¹³ and 27% of samples.¹⁶

Recently, the adequacy of a dichotomous classification of onset has been questioned,¹⁷ based on emerging results from studies that provide evidence of other onset patterns. Some parents report that their child displayed neither early signs of autism nor regression, a phenomenon that has been called developmental plateau^17,18 or stagnation.¹⁹ Other studies have found evidence of both early signs of ASD and regression simultaneously present in the retrospective histories of children with ASD.^9,11,12,20

Questions about when and how behavioral signs of autism emerge may better be answered through prospective, rather than retrospective, studies. Prospective investigations are a powerful method of studying onset, because they reduce errors due to parental recall and biases introduced by selective home videotaping by sampling behaviors directly.⁷ In addition, they provide the opportunity to test specific hypotheses through experimental methods. In the past decade, several research groups have prospectively studied infants at elevated risk for ASD because they have one or more siblings with the condition. These infant sibling studies reliably find differences in the second year of life in the same social communicative domains as retrospective studies.^21–25 However, few group differences predictive of ASD outcomes have been found before the first birthday in prospective studies.^21,22,24,25 Bryson et al,²⁶ in a consecutive case series of infant siblings followed prospectively from 6 months of age, describe several children whose symptoms were not present at their 6-month (and, for some, 12-month) visits, but emerged slowly during the second year of life. None of the children who developed autism displayed marked limitations in social reciprocity at 6 months. All nine infants were described as interested in social interactions and responsive to others, demonstrating sustained eye contact and social smiles. Thus, prospective studies suggest that social behavior may be grossly intact in the first 6 to 9 months of life in infants who are later diagnosed with autism.

The present prospective investigation examined the timing and course of onset of behavioral signs of autism in an infant sibling sample. Two prospective methods were used to examine symptom emergence: precise coding of the frequency of specific social communication behaviors exhibited during the session and examiner ratings of the frequency of infant social engagement. We also collected a traditional retrospective measure of symptom emergence using parent report, permitting us to examine how prospective and retrospective methods of data collection corresponded in terms of determining onset.

The comparability of social behavior at 6 months of age in the ASD and TD groups replicates findings from other prospective studies.^25,26,35 The lack of group differentiation at this age does not appear to result from low power, as the group means were very similar and effect sizes were very small. Most convincingly, the effects were in the opposite direction as predicted, with the ASD outcome group demonstrating (nonsignificantly) better social communication behavior at 6 months than the TD outcome group on all variables. After 6 months, the ASD group shows a rapid decline in eye contact, social smiling, and examiner-rated social responsiveness. Group differences were significant by 12 months in gaze to faces and social smiling and by 18 months on all other variables. Similar declining trajectories in the onset of autism symptoms have also been reported by others.³⁵

Some have suggested that the decrease in social interest in this time period is related, at least in part, to an overly focused attention to objects or difficulty with attentional switches from objects to people.³⁶ However, our data did not support this hypothesis; we found no group differences in looking at objects at any age and duration of looks to objects did not increase over time in the ASD group. Thus, the decline in looking at faces and social smiling could not be accounted for by an increase in interest in objects. Similarly, neither group differences in cognitive functioning, slowing of cognitive growth between 12 and 36 months in the ASD group, nor less successful performance on the Mullen VR subscale (which provided the context for coding) could explain the decrease in social communication behavior over time in the ASD group.

Collectively, the present investigation and recent prospective studies suggest that signs of autism emerge over the first year or so of life in many children with ASD, rather than being present from close to birth, as once suggested by Kanner.³⁷ Although there are likely cases in which behavioral signs are indeed evident at or before 6 months of age, this pattern may be less common than originally thought.

Interestingly, cognitive and language skills, as measured by the Mullen, did not demonstrate the same declining trajectory in the ASD group as social communication skills. Mullen raw scores increased over time in both groups, but showed significantly slower growth in the ASD group starting at 12 months. This suggests that developmental loss of skills may occur specifically in the social communicative domain, rather than generally across all developmental domains. One implication of this finding is that the Mullen and other standardized developmental tests will not be good prospective measures for tracking regression, as they appear to be less sensitive to the kinds of behaviors that decline as autism emerges.

The results of the current prospective study suggest that the traditionally defined categories of early onset and regressive autism do not portray accurately how symptoms emerge, nor does the newer-onset category involving a developmental plateau. In the present study, specific social communicative behaviors clearly decreased rather than failing to progress. Losses were particularly dramatic between 6 and 18 months. This suggests that ASD onset marked by loss of social communication behaviors occurs much more often than has been recognized using parent report methods. However, rather than the rapid and marked losses typically reported in previous studies of regression onset, the declines that could be detected through specific probes of social development every 6 months were relatively subtle and gradual, were often preceded by earlier parental concerns, and were often followed by failures to progress in other areas.³⁸ This may have made the losses difficult for families to identify, as demonstrated by the low rate of parental reports of regression on a standardized measure, the ADI-R, administered at 36 months. It might be expected that parents who have already experienced the developmental course of autism onset in an older child would be more sensitive to these signs than the general population, underscoring how difficult it will be for most parents to report about skill loss. For this reason, it is critical to explore other methods of detecting the earliest signs of autism.

In the present study, a second method of measuring social communication development was examiner ratings, which were conducted at the end of every visit, blind to the group assignment of the participant. We found that these global ratings of the frequency of social communication behaviors were highly correlated with coded rates of the same behaviors, validating both approaches to measurement. Although it is not practical to code behavior frequencies during medical office visits, these results suggest that it may be possible to develop measures that can be rated by physicians or nursing staff during well-child visits that capture “snapshots” of an infant’s current social repertoire. If such measures are found to be valid reflections of infant sociability, then they could be added to developmental monitoring exams and serially rated over time to identify infants at higher risk for ASD and in need of further specialized evaluation.

The present findings suggest that existing definitions of onset patterns will need to undergo further development. One possibility is that we need to expand the number of categories used to describe onset. For example, perhaps there are four rather than two categories of onset, including groups characterized by developmental plateaus and by mixed features of early symptoms plus later regression, as we have suggested elsewhere.¹⁷ The prospective data acquired in the present investigation suggest another possibility, however. We hypothesize that symptom emergence may better be considered dimensionally, as a continuum characterized by the amount and timing of regression. In this conceptualization, at one end of the continuum lie children who display loss of social interest so early that the regression is difficult to see and symptoms appear to have always been present. At the other end of the continuum lie children who experience losses of social interest and communication skills so late that the regression appears quite dramatic. Supporting the idea that there may not be two different processes of symptom onset, but rather one dimension, several recent studies have found few differences in later developmental, functional, and adaptive outcomes between children with traditionally defined early onset and regressive autism.^18,39 Resolution of opposing viewpoints regarding onset (e.g., multiple categories v. a dimensional view that emphasizes timing) is urgently needed for etiologic studies, which have been hindered already by the tremendous heterogeneity of the autism phenotype.