Is the error-reaction time correlation in category verification tasks evidence of fuzzy limits in categories?

This thesis attempts to review evidence supporting a positive error-reaction time correlation in category verification tasks. All reviewed models predict that categorization errors will increase when the time needed to make a membership judgement increases. This is explained either as a result of the structure of categories (e.g., as another manifestation of category fuzziness), or as a product of the category verification process (e.g., attributed in general memory models to the random nature of the retrieval process). Two specific models that attempt to explain the correlation were tested. One that assumes the correlation is the result of incomplete or inconsistent concept retrieval when subjects are under speed emphasis conditions, and other that assumes the correlation is not a psychological phenomenon, but the result of grouping data across subjects (the common data gathering procedure in the field). Results support this latter explanation of the error-reaction time correlation. It is shown that if the effect of intersubject disagreement in category membership judgements over errors is statistically controlled, the correlation significantly decreases for both categories used. The reduction in the calculated correlation is such that for one category (furniture) the magnitude of the effect is not significantly different from zero, and for the other (vehicle) it accounts for a mere 6% of the variance of categorization errors. The implications for models of category membership decisions are discussed, and a two stage model of the process that does not predict the correlation (but that can explain its rise when accumulated data is used) is suggested