Learning Within-Category Attribute Correlations in a One-Attribute Visual Search Classification Paradigm

In this study, participants categorized stimuli in a oneattribute rule visual search classification paradigm. The stimuli were six-shape displays that included a rule attribute and five diagnostic attributes. In Experiment 1, attribute values were changed at transfer. Slower RTs were obtained when attribute values from conflicting categories were used. In Experiment 2, the rule attribute (and up to two other attributes) were removed at transfer. The results showed that several attributes (color, texture, and size) of varying diagnosticity were used to correctly classify the stimuli. These experiments provide evidence that within-category attribute correlations can be learned in a classification task without intentional or inference learning instructions.