In conclusion, the published studies concerning children with autism younger than 3 have provided us with an ample database for making important decisions. First, it appears that most children with autism, while they may not be diagnosed until age 3 or older, nevertheless have symptoms much earlier. Parents may report concerns almost from birth and almost universally by 18 months. Retrospective studies of infant videotapes of children later diagnosed with autism have demonstrated significant differences in infants with autism by the age of 10-12 months, differences that distinguish them both from normally developing children and from children who have other kinds of developmental delays. The differentiating behaviors concern social responses to other people: responses to their names being called and gaze directed at other people. It is not clear whether children with autism who have late onsets after a period of normal developmental demonstrate these early precursors; studies from Dawson and Osterling (Werner et al., in review; Osterling & Dawson, under review; Osterling & Dawson, 1994) have indicated that they do not, thus providing some empirical validation of the phenomenon of late onset. In general, these early symptoms of autism involve reduced frequency of expected behaviors-negative symptoms-rather than behavioral excesses or the presence of abnormal behaviors-positive symptoms. This is one of the big differences in autism in infants and toddlers, the predominance of negative over positive symptoms. Developmental differences specific to autism increase as children approach their second birthday and development as a whole is becoming more differentiated. Specific patterns of atypical verbal and nonverbal communication are present that distinguish autism from other developmental disorders. Lack of social and affective reciprocity becomes far more marked. Lack of appropriate language and communicative development, lack of symbolic play, lack of empathic responses to others, and reduced and inaccurate imitation skills are main features of autism in toddlers. The specificity of this early autism profile is supported by the accuracy of clinical diagnosis of autism in 2-year-olds by experienced clinicians. The presence of an autism spectrum disorder, as well as autism per se, can be diagnosed reliably by clinicians by the time children are 20 to 24 months of age, and the vast majority of children diagnosed at 2 years continue to demonstrate the symptoms of autism spectrum disorders when seen 12 to 18 months later. While both the diagnoses of autistic disorder, and the broader diagnosis of an autism spectrum disorder, demonstrate very high stability from age 2 to age 3 and beyond, diagnoses of PDDNOS and atypical autism do not demonstrate either inter-rater agreement or stability over time when made at the age of 2. However, most of the published diagnostic tools are not gauged for this age group, and the published cutoff scores of the most popular tools-the CARS, ADI-R-overdiagnose autism in this age group. Similarly, several of the DSM-IV criteria are not appropriate for diagnosing autism at these young ages. Thus, clinicians need to adjust cutoff scores and use other sources of data in addition to scores on instruments in order to make accurate assessments. The use of standardized parent interview formats and standardized interactive observation formats is strongly encouraged for clinicians, since these tools make sure that all relevent areas of symptomology are addressed in the assessment. An accurate diagnosis of autism requires a thorough parent interview, an interactive assessment of the child, and clinical experience and judgment. The most common diagnostic errors in this age group are to overdiagnose autism in children with very young mental ages and to underdiagnose autism in very verbal children. Given that children with autism are generally identifiable by age 2, pediatricians, other health professionals, and assessment teams need sophisticaiton in early diagnosis. The field needs better screening approaches, so that symptoms apparent at 12 to 18 months are recognized more easily in well-baby checkups. The fields needs strong primary and secondary screening tools, and that work is underway. Finally, the main purpose of early diagnosis is to lead to early treatment. Getting a child through the diagnostic process is not enough; diagnosis does not necessarily lead directly to programming. In addition to the diagnosis of autism, toddlers need careful and thorough individual assessment of their developmental strengths and needs, so that intervention can be individualized, targeting needs and strengths (including family needs, Dunlap & Fox, 1999) upon which to build programs (Marcus & Stone, 1993). The purpose of diagnosis is treatment, and the purpose of early diagnosis is early treatment. While we have the capacity to diagnose autism in children at or before their second birthday, most autism intervention models have been developed for preschool-aged children (though McGee, Morrier, and Daly (1999) have recently described a specific intervention model for 2 year olds with autism). The needs of infants and toddlers with autism may be quite different, and the intervention field needs to develop models of appropriate treatment of infants and toddlers with autism.