ABSTRACT

Identification of a particular stimulus selected from a set of easily discriminable stimuli, with these stimuli varying along some unidimensional physical dimension, produces consistent patterns of accuracy and response time performance. Some of the most robust features are the so-called bow effects: accuracy is highest for stimuli at the ends of the physical range and lowest for stimuli in the middle of the range; (mean) response times are fastest for stimuli at the ends of the range and slowest for stimuli in the middle of the range. Marley and Cook (1984) have demonstrated that the magnitude of the bow effect for accuracy, which is dependent upon set size and range, can be accounted for by postulating that subjects rely on a fixed rehearsal capacity that must be allocated over the stimulus range. The present paper extends their model to response time predictions by postulating a decision mechanism based upon the accumulation of discrete units of the memory information, X, proposed by the Marley and Cook model. A separate random walk process is postulated for each of the N possible responses in a N stimulus absolute identification task. The general form of the random variable accumulated in each of these processes equals the value of |(Uj – X)(X – Lj )|, where Uj and Lj are the upper and lower boundaries of response category j, j = 1,2,3,…, N. Link (1992) proposed a similar form to account for similarity and equality judgments. The correspondence between experimental and simulation measures of accuracy, response time, information transmitted and d’ for various set sizes are examined.