Speed-accuracy tradeoff models for auditory detection with deadlines

Two neural models for response latency in auditory signal detection are considered: the Timing model of Luce and Green (1972) and a modified counting model based on that of McGill (1967). The modified counting model is described in some detail. The experimental situation to which the models are applied is one where a deadline in response time is enforced on signal trials only or on noise trials only, the condition of deadlines on both cases having previously been studied by Green and Luce (1973). The results for mean latencies of the various categories of response, together with response probabilities, favour either the counting model or a dual process model. Data for ROC curves indicate either the operation of a dual process model or that the ‘interval of uncertainty’ of the counting model may vary with bias position. Some consideration is also given to the possibility of differential residual response time components and it is concluded that such components may be important in the deadline situation.