| | |
Summary: Copyright 2002 Psychonomic Society, Inc. 570
Perception & Psychophysics
2002, 64 (4), 570-583
Some categorization tasks are easier to perform than oth-
ers. For example, it is easier to classify Siamese versus
Sphinx cats than Siamese versus Himalayans. Many factors
affect the difficulty of such tasks--for example, stimulus
dimensionality, number of relevant dimensions, correlations
between dimensions, and category overlap. Each of these
factors is determined by category structure. In this study,
we systematically manipulated category structure and
characterized its effect on human performance using ac-
curacy and efficiency measures. In particular, we try to
show that a specific aspect of category structure, covariance
complexity, is an excellent predictor of task difficulty. Co-
variance complexity is an information-based characteri-
zation of the covariance matrix that shapes our categories.
This measure is formally defined below.
The relationship between category structure and task
difficulty has been of interest to psychologists for several
|
