Category learning of schematic faces in children: Relationships between attribute knowledge and modes of processing

In a pluralistic view of cognitive functioning, it has been proposed that, when having to learn categories, individuals can use two forms of processing: either an analytic mode in which the necessary and sufficient properties are extracted, or a holistic mode in which objects are processed as indivisible webs of attributes and relations. To further investigate this issue, children were observed during a perceptual categorization task based on the Kemler Nelson (1984) paradigm. Besides the usual error pattern, response times and attribute knowledge were used to diagnose what process was elicited to learn family resemblance categories. After a study phase of two sets of schematic faces which varied on six attributes (shape of eyes, nose, etc.), 9-year-old children were administered a category assignment task during which response speed and accuracy were recorded. Subsequently, knowledge of individual attributes was evaluated using a specifically designed task, the attribute identification task. Convergent data from the two tasks show that most of the children were multiple attribute learners, but used either one attribute (about 60% of the subjects) or more attributes (the remaining 40%) to make category decisions. Reasons for the absence of genuine Gestalt-like processes in our study are discussed.