As part of an ongoing clinical service program for children with developmental delay in an Asian developing country, we analyzed the cognitive attributes of 362 Taiwanese children (average age 48.5+/-12.9 month-old) with uneven/delayed cognitive development as they were assessed repeatedly with average duration of 39.7+/-22.6 months from preschool through early childhood. The objectives were to determine the stability and related factors in cognitive scores of these 362 children belonging to three diagnostic subgroups: 181 children with non-autistic mental retardation (MR), 95 children with autism spectrum disorder (ASD) and 64 children with mixed type developmental language disorder (DLD); and to contribute to the accumulation of data on cognitive outcome in preschool children with developmental delay. Analysis revealed that mean initial cognitive score (IQ1) was 64.9+/-16.9 while mean cognitive measure at follow-up (IQ2) was 72.2+/-19.7. Whole group analysis showed the correlation between IQ1 and IQ2 was moderate (r=0.73, p<0.001). Analysis by a general linear model showed only male gender (beta=4.95, p=0.02, C.I.=0.8-9.1) and IQ1 (beta=0.79, p<0.001, C.I.=0.68-0.90) to be significant predictors of IQ2. There were differences among three groups in IQ1 (p<0.001), IQ2 (p<0.001) and IQ change (p<0.001). Correlation coefficients of IQ1 and IQ2 were 0.6 for ASD group, 0.7 for MR group and 0.4 for DLD group respectively. The greatest proportion of children remained within the same cognitive range for both assessment points, however, it is noted that a substantial minority of children changed IQ ranges drastically from preschool through early childhood. Our results suggest that measurements of cognitive function at preschool age for children with developmental delay were valid in the context of a developing country, and the observed change in cognitive scores during follow-up emphasized the need to interpret the initial results of cognitive tests with caution.