An apple for a table, a car for a cow: a simple count on category fluency is sufficient?

Abstract

OBJECTIVE: This study attempted to demonstrate the use of conventional simple ‘‘count’’ of category in semantic fluency test may not sufficiently reflect the actual performance of patients with schizophrenia and normal controls. METHODS: A cross-sectional design was adopted with a total of 187 patients with schizophrenia and 98 normal controls. Participants were asked to generate as many exemplars from within a semantic category as they could in 3 min. Four types of common categories encountered in daily life were selected for this study; they were animal, means of transport, food, and furniture. The Rasch model was undertaken for data analysis. This model is a measurement model which provides probabilistic additive conjoint measurement. Data which fit the model, given certain testable assumptions, are unidimensional, have order, additivity and specific objectivity. A logit transformation is obtained, providing interval level measurement for the trait under consideration. Results: The findings indicated that the psychological distance, in terms of logits unit, was not at an equal interval within each category across the 3-min interval in both patients and healthy controls. The logits between different categories were also found to be inconsistent in both groups. CONCLUSION: These preliminary findings suggest that we should not simply equate different category scores when we study semantic categorization. Such a simple ‘‘count’’ of categories may not truly reflect the actual performance of patients and normal controls. Future study should further examine a more representative and accurate measurement of semantic categorization.