Proceedings of The Physiological Society
University of Oxford (2011) Proc Physiol Soc 23, PC299
Saccadic latency and information foraging
S. Hänzi1, H. Copley1, R. H. Carpenter1
1. Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom.
Figure 1. Left, average latency reduction across subjects when subjects use information from the target: Trottier and Pratt (appearance and gap tasks), Gabor (appearance and step), and the faces task. Right, reciprobit plot of latency distributions of one subject in the experiment with variable proportions of Gabor stimuli, as indicated. The latency is smaller if the probability of Gabors is larger, and intermediate steps produce intermediate latencies in the expected order, with parallel shift.
The primary purpose of saccades is to acquire information from the environment. In that sense, the foveation of targets that provide more information can be regarded as more rewarding than others, and in accordance with current thinking about neural decision mechanisms and reaction time, should result in shorter saccadic latencies. By analysing how the stochastic distribution of these latencies is affected, we can shed light on the nature of the underlying neural processes. A previous study in which responses did or did not generate useful information  demonstrated the existence of such an effect. Here, we began by exactly replicating this study, confirming their findings, and extending it by LATER analysis of the distributions, and by refinement of the stimuli so as to make them quantifiable in terms of how much information they provided, and also by using a more natural task (reporting the gender of faces). We recorded the saccadic latencies of 12 volunteers (see  for the methods: the procedures had received local ethics committee approval) in response to targets presented randomly at fixed locations to right and left. In some runs, subjects had to respond to features of the target only visible when fixated by pressing a button, in others no response was required. In three different experiments, the tasks were: detecting a small break in a line, replicating the previous study; reporting the gender of a face; detecting the orientation of a high-frequency Gabor stimulus (with the probability of the stimulus occurring being varied). In each case, latencies were shorter when subjects expected to obtain useful information from the saccade, with intermediate values for intermediate proportions of Gabors: this reduction was associated with a parallel shift of the distributions, corresponding to a change in the mean rate of rise,μ, of the underlying decision signal (Figure 1). It is clear that information acquisition reduces saccadic latency in a graded manner. The effect on the distributions was however surprising. Given the primacy of expected utility in determining decisions , it would be expected that reward would have the same effect (swivel) on distributions as is well known for probability . However, a change in the rate of rise of a decision signal has been observed in neurons in the caudate nucleus of monkeys rewarded for making some saccades and not others , which would be in accordance with our findings.
Where applicable, experiments conform with Society ethical requirements