This paper is concerned with the heterogeneity and home correlates of social isolation in school. Three data sets, each involving 40–42 4‐year‐olds, were studied at home and preschool. Three observational measures of social isolation were used—time alone, time unoccupied and frequency of interaction. These social isolation variables showed few linear relations to measures of temperamental characteristics, other child characteristics or measures of mother‐child interaction. However, categorizing the children into four groups on their status of alone and unoccupied, or into eight groups on all three social isolation variables, yielded categories showing meaningful patterns of interactions in school and meaningful relations to home variables. The dangers of considering any categorization system as absolute are emphasized.