Identification of Multidimensional
Objects and Multiple Observation
In an identification experiment, a single stimulus from a known set is presented on each trial, and it is the observer's job to say which it was—to identify it. The purposes of such experiments vary, but usually include obtaining an overall index of performance, as well as a measure of sensitivity for each stimulus pair and bias for each response.
If there are only two stimuli, identification is simply the yes-no task of chapters 1 and 2, and performance can be summarized by one sensitivity and one bias parameter. The nature of the stimuli is unimportant—it does not even matter if they differ along one physical dimension (lights of different luminance) or many (X-rays of normal and diseased tissue). With more than two stimuli, the task is easily described: One stimulus from a set of m is presented on each trial, and the observer must say which it was. From the participant's point of view, there is nothing more to say, but to extend the analysis to more than two stimuli the dimensionality of the representation must be known. If all stimuli differ perceptually on a single dimension, then m − 1 sensitivity distances between adjacent stimuli and m- − 1 criterion locations can be found along it, as we saw in chapter 5. Perceptual distances for all other pairs of stimuli are easily calculated as the sum of the stepwise distances between them. To characterize overall performance, it is natural to add sensitivity distances across the range.____________________