Abstract
Reddi & Carpenter (2000) proposed that the saccadic latency reflects the reciprocal rate of visual processing for reaching a decision threshold. They observed that the rates (=reciprocal latencies) of 1st saccades were distributed normally. This observation does not apply to saccade sequences, because for visual search tasks van Loon et al. (2002) found that the rate distributions for 2nd and later saccades are skewed, much like Gamma distributions. Gamma distributions arise when many independent stochastic processes contribute to the decision, suggesting the skewing results from a reduced number of processes in later saccades. However, skewed distributions can also be explained by a competition process that pits two rate signals against each other (van den Berg, NS 2001). Such extended decision model, with for instance ‘make saccade’ vs ‘keep fixating’ decision signals, would enable the saccade sequence to stop. Interestingly, the model predicts beta distributions, which typically have more tail at high rates than gamma distributions. Furthermore, the beta function's two parameters represent the thresholds of the two competing signals. Here, we investigated the evidence for a beta rate distribution and for systematic variations in its parameters. Subjects were to saccade as quickly as possible towards a target that deviated in line orientation from distractors consisting of lines arranged in a radial pattern. The number of distractors (Exp. I) or the chance of the target appearing at the fixation point (Exp. II) was varied. In both experiments, the rate distributions for second and later saccades were significantly better fit by beta than by gamma functions. Moreover, significant changes in the beta fit parameters were found in Exp. I for the later saccades, with threshold changes as predicted by the competition model. Our results are consistent with a competition between two decision signals underlying the timing of saccades.