Complementary versus contrastive classification in preschool children

We compare 3-year-old children's superordinate level classification under two experimental conditions. In the Complementary condition, children were instructed to sort a set of pictures three times, each time extracting a different "target" class (e.g., Animals) from the remaining items (e.g., Clothing and Food). In the Contrastive condition, they formed the three superordinate level classes simultaneously within a single trial (Animals vs Clothing vs Food). Because the probability of assigning items correctly by chance differs under these two conditions, we introduce statistical adjustments to take the different rates of chance success into account. Although children in the Contrastive condition had to divide their attention among three target classes, while those in the Complementary condition had to focus on only one target category per trial, there was no mean difference between these two experimental conditions. There was, however, a striking difference in the distributions under the two conditions: scores in the Contrastive condition were bimodally distributed, while those in the Complementary condition were more normally distributed. A second study, using different categories (Furniture, Vehicles, and Clothing), revealed the same effects. These data suggest that contrast in classification benefits some, but not all, preschool children.