Saccades are rapid eye movements through which foveate animals examine visual scenes. However, while in natural behavior saccades implement an active sensing strategy for sampling task-relevant cues, these strategies have been seldom investigated and very little is known about their neural mechanisms. In the vast majority of laboratory paradigms, participants (humans or non-human primates) make saccades in a purely exploitative fashion - to harvest rewards, as the primary goal of the task, rather than gather information about future task states. The use of saccades for information sampling raises two major questions beyond those arising in traditional paradigms. First and foremost, how are saccades motivated by uncertainty ? or expected gains in information ? in addition to reward expectation? Second, how is the selection of an informative target (before the saccade) coordinated with the processing of the sampled information (after the saccade)? We examine these questions in the monkey lateral intraparietal area (LIP), a cortical area implicated in top-down oculomotor control, using neural recordings and reversible inactivation combined with a novel information sampling paradigm. In this paradigm, monkeys are free to deploy gaze to obtain information about a reward that accrues after a delay, but receive no rewards or punishments for a specific gaze pattern. Preliminary evidence shows that saccades depend on the monkeys? prior beliefs about reward probability and uncertainty, consistent with computational models of gaze allocation in natural conditions; in addition, target selective LIP cells (1) are modulated by reward and uncertainty, (2) show a trans-saccadic integration of the effects in their pre- and post-saccadic responses, and (3) respond specifically in relation to information sampling rather than global changes in arousal/motivation. Starting from these observations, we aim to understand the functional contribution of area LIP to information sampling based on reward probability and uncertainty. The results are significant because (1) they represent the first investigation of top-down deployment of gaze as an active sensing policy, (2) they will clarify the role of LIP in reward decisions, which has been the topic of protracted debate and (3) will help integrate research on attention and decision making, two cognitive functions that are naturally coordinated but have been so far studied separately. Thus, the studies are expected to have a strong impact on our understanding of attention and information constrained decisions, which are critical in complex behaviors and psychiatric diseases including attention deficit disorder, drug addiction, anxiety and depression.
The judicious selection of a source of information is central for decision making, and is the primary function of active sensing behaviors such attention and gaze. However, very little is known about the factors shaping information sampling policies or their neuronal implementation. Here we address this question using a novel oculomotor information seeking paradigm, combined with neural recordings and reversible inactivation in monkeys. We examine how parietal neurons involved in oculomotor control select cues about future task states as a function of the reward probability and reward uncertainty associated with those states.
| Gottlieb, Jacqueline (2018) Understanding active sampling strategies: Empirical approaches and implications for attention and decision research. Cortex 102:150-160 |
