A MAPPING MODEL OF BOW EFFECTS IN ABSOLUTE IDENTIFICATION

We propose that the intriguing bow effects in choice probabilities and response times that are found in unidimensional absolute identificati on might be a consequence of the mapping process required when a unidi mensional psychological representation is mapped to a multidimensional response vector. This idea builds on previous work by Lacouture & Mar ley (1991) which modeled absolute identification using a connectionist feed-forward network. A formal solution of the so-called ''encoder pr oblem'' is the basis of the approach, and the inclusion in the model o f ''noisy'' mappings and integrators allows us to model both the respo nses made and the time to make them. The simulated model reproduces se veral of the phenomenon in unidimensional absolute identification incl uding bow, range, and some sequential effects. (C) 1995 Academic Press , Inc.