In this study, we longitudinally followed into childhood a cohort of preschool children, initially diagnosed as autistic or non-autistic with developmental delay, to evaluate the stability of cognitive assessments performed during the preschool period. The consistency in group means and intra-individual stability of developmental quotients (DQ) and non-verbal intelligence quotients (IQs) were compared for these two groups, which were matched by chronological age, initial non-verbal IQ/DQs, initial test given, and length of follow-up interval. The case group comprised 16 autistic children with average age at initial assessment of 3 years 8 months. The control group comprised 16 non-autistic developmental delayed children with average age at initial assessment of 3 years 11 months. Mean DQ/non-verbal IQ at initial assessment was 73.9+/-23.9 for the case group and 80.3+/-23.2 for the control group. ANOVA yielded no significant effect of time or time x diagnosis interaction (F=0.183, P=0.675). The absolute difference in scores and group means were equivalent for both groups of children, with no difference in patterns of change. Correlations between DQ/non-verbal IQs at initial assessment and follow-up were significant and high for the two groups (autistic group: r=0.87; control: r=0.77). Intellectual functioning can be a valid measure in Taiwanese preschool children with autism, and has an equivalent meaning for children with autism and for non-autistic children with developmental delay. Though the follow-up period is too short for definite prognostic conclusions to be drawn, we think that non-verbal intelligence should be an essential assessment for preschool oriental autistic children so that sound expectation and treatment plan can be made.