Response Time Models For Temporal Order Discrimination

Abstract

<p>The results of three experiments requiring the judgment of the temporal order of two continuous visual stimuli were reported. The stimuli were spatially separated and equal in intensity. Although the proportion of correct order judgments increased with the inter-stimulus-interval (ISI) in each experiment, there were differences in the response time (RT) results. When accuracy was stressed, mean RT increased as the ISI decreased. However, mean RT was relatively independent of ISI when RT deadlines were employed. As mean RT increased the slope of a linear relationship between the difference in mean RTs for the two order responses and a measure of performance accuracy decreased from positive to negative. Under accuracy conditions the mean RT for a correct response was less than the mean RT for an incorrect response by an amount which increased with an increase in ISI. When RT deadlines were imposed the mean RT for a correct response exceeded the mean RT for an incorrect response by an amount which increased with an increase in ISI.</p> <p>The covariation between RT and response proportion measures was accounted for by a model which proposed that order discrimination involves three stages. In the stimulus information encoding stage, a sample of stimulus information is stored in a sensory storage buffer. In the decision stage the information obtained in the buffer drives a random walk process which generates quantitative predictions for response proportion and the mean RT for each order response. The response output stage contains processes such as motor actions which are assumed to contribute a constant component to the mean RT for each stimulus condition.</p> <p>A model specifying that the order decision results from whichever stimulus is perceived first was shown to generate mean RT predictions which were not supported by the data obtained under accuracy conditions. It was concluded, therefore, that order discrimination involves a decision process which utilises information obtained from the stimulus pattern.</p> <p>The use of a gap in one of the stimuli of a pair did not generate changes in response proportion which were predicted by a random walk decision process which was driven by information sampled directly from the stimulus pattern. However, the results of each experiment were consistent with the prediction of a model specifying that the parameters of the stored stimulus information vary with time since the commencement of the decision process.</p> <p>When accuracy was stressed and the mean RT was long, the decision process was approximated by a terminal zero drift random walk. This model was shown to fit the data from Experiments I and III with a high degree of accuracy. Estimates of stimulus and response strategy parameters provided a means for assessing their separate contributions to order discrimination performance. However, this approximation failed to account for data obtained when RT deadlines were imposed.</p> <p>The covariation between response time and response proportion measures was similar to that obtained in other psychophysical tasks. It was concluded, therefore, that the model proposed in this thesis has wide applicability to a variety of two-choice tasks.</p>