When subjects learn to categorize new stimuli adequately,
they have to segment these stimuli into relevant features
for categorization. In the experiments reported here,
children had to discover a rule for categorization.
Preliminary experiments have shown that depending on the
nature of the irrelevant features, children could find the
relevant features from age four or could not find them before
the age of eleven or twelve. A central question is whether
children aged four or six who have discovered the rule in a
simplified version of the relevant features would generalize
to a “complex” version (i.e., in which there is more
background noise) of the relevant features, i.e., a version
that they would be unable to learn before twelve without
pre-training. Conditions promoting the generalization
from the simple version to the complex version were also
investigated. Two conditions were compared: relearning
with or without feedback. Results showed that children aged
4 and 6 could generalize the “simple” version of the target
concept to a more complex version of the same concept,
either with and without feedback in the generalization
phase.