ABSTRACT

In Chapter 5, the basic characteristics of TSD were described. TSD was seen to be a powerful model for conceptualizing what observers do when they make decisions about the presence or absence of signals. According to the model, a sensory event arising from the presentation of a signal is fundamentally no different than one that arises from noise. There is no threshold to divide sensory events into those that are suprathreshold and those that are subthreshold, and there are no distinguishing characteristics that would enable the observers to identify the event as resulting from noise or, alternatively, from signal added to noise. Instead the observer experiences sensory events that vary in magnitude along a single continuum. The problem with which the observer is confronted is to decide whether a particular sensory event originated from the distribution of noise or from that of signal plus noise. According to the model, the problem is solved when the observer evaluates sensory events relative to the position of a decision criterion. The decision rule is that if the magnitude of the sensory event is greater than the value of the criterion, a signal is reported and if it is below this value it is not. The position of the criterion relative to the location of the noise and signal-plus-noise distributions is determined by practical considerations such as the probability of the signal being presented and the costs and values of the decision outcomes. The position of the criterion, as well as the detectability of the signal, specified as the d' distance between the means of the noise and signal-plus-noise distributions, can be inferred from experimerttal data by determining the observer's hit rate and false alarm rate in the detection situation. One of the greatest strengths of this approach in psychophysics is that it provides the investi-

gator with separate and independent measures of the location of the criterion and signal detectability. In classical psychophysics, this was not possible, but instead thresholds, as detectability measures, were often contaminated by response bias resulting from the unaccounted for location of the criterion. Thus, TSD represents a substantial improvement in psychophysical methodology.