The purpose of this study was to longitudinally examine the relationship between thought and language with 24 normal children from the age of 6 to 36 months. First, the order of emergence among thought and language variables was from the visual representation of a hidden object at multiple places measured by fixation time(range:6-8 months), to the pointing behavior to known objects(range:9-12 months), and then to the first meaningful word observed in laboratory settings(range:10-22 months). However, the correlations between the above thought and language variables were not always significant. Second, prior to the emergence of two-word utterances(range:18-28), not only the percentage of symbolic play was increased at a high rate but also symbolic play became more complex in a sequence of 4 to 8 different acts. Yet, all the variables of symbolic play and language were not correlated significantly. Third, thought and language variables at 36 months showed significantly positive correlations in most cases. The subjects' classification abilities and their results on the Suzuki-Binet Intelligence Test were the thought variables. The language variables, on the other hand, were their verbalization abilities on substitution with colored blocks in free play with mothers, and their results on the Japanese version of the Illinois Test of Psycholinguistic Abilities. In summary, up to 2 years of age, there was a consistant trend of thought development preceding that of language. However, the correlations between thought and language variables were not always significant probably due to other factors, such as qualities of mother-child relationship, which also would influence on the development of language. When the subjects reach 36 months old, their language skills are well established. Thus it is Presumed that there are significantly positive correlations between thought and language variables.