Eye movements and the maximization of value
Only the central region of the retina, the fovea, can provide us with high-acuity details of our visual environment. In the periphery however, resolution fades away with increasing eccentricity. As a consequence, humans and other animals with a foveated visual system move their eyes to redirect t...
|Online Access:||PDF Full Text|
No Tags, Be the first to tag this record!
|Summary:||Only the central region of the retina, the fovea, can provide us with high-acuity details of our visual environment. In the periphery however, resolution fades away with increasing eccentricity. As a consequence, humans and other animals with a foveated visual system move their eyes to redirect their gaze towards objects of interest. And with each saccadic eye movement, we choose a different region of the visual field for high-acuity processing. In the recent decades, the eye movement system has thus evolved as a role model to study decision making (Glimcher, 2003), which is also because the oculomotor system is sensitive to valuation processes. Moreover, our eye movements are tightly linked to visual perception, because where we look determines what we see and every eye movement poses a major challenge to the visual system as it shifts the whole visual image on the retina. In three studies, this dissertation project examined whether the eye movement system can adjust saccade latencies to maximize informational and motivational value and whether the visual system can maximize all the information available despite making eye movements. The first study investigated whether the eye movement system is sensitive to the information that can be gained by executing an eye movement. Participants saccaded to a peripherally appearing target and perform a perceptual task. By exchanging the target while the saccade was in flight, we could independently manipulate the pre-saccadic peripheral and the post-saccadic foveal visibility and thus create conditions where participants either lost or gained information by making an eye movement. In the loss condition, the probability of correctly identifying the target increased with saccade latency because participant could benefit longer from high resolution peripheral vision. The opposite pattern was observed in the gain condition. However, eye movement latencies did not differ no matter whether participants could gain or lose information and thus could not maximize the all the information available. Instead, latencies decreased with the probability that visual information at the saccade target was task-relevant, suggesting that saccade eye movements are influenced by the motivation to foveate task-relevant information, but not by the information that can be gained by saccade execution. In Study II, we tested whether the visual system is able to integrate pre-saccadic peripheral and post-saccadic foveal information and whether it weighs the incoming visual information according to its reliability, that is, according to how well something can be seen. This optimal integration would minimize the perceptual uncertainty and thus maximize all the information available to the visual system. For every individual, we separately measured discrimination performance in the fovea and the periphery. Using maximum-likelihood integration (Ernst & Bülthoff, 2004), we predicted the optimal weight given to peripheral information as well as the optimal uncertainty associated with the trans-saccadic percept. Both, in terms of weighting and uncertainty, trans-saccadic performance was not distinguishable from optimality. We thus could show that the visual system is able to integrate information across saccades and that it is close to optimal in doing so. This highlights that the visual system is able to maximize all the visual information available despite making eye movements. Study III investigated whether the influence of expected motivational value on saccades (Milstein & Dorris, 2007, 2011) can only be found in contexts where participants additionally have to choose between multiple rewarded targets. We recorded saccade latencies to rewarded targets differing in reward and manipulated the proportion of interleaved choices within one block. In choice-trials, two targets were displayed and participants could choose between the two to obtain the corresponding reward. Without choices present, we found no evidence for single target saccades to be affected by reward. When choices were interleaved, latencies to less rewarded targets were delayed and the magnitude of this delay increased with the proportion of choices. This delay was elicited by the expectation of an upcoming choice-trial as well as inter-trial priming: After a choice, saccadic reactions to the non-chosen target were delayed. We thus could show that there is no direct relationship between expected motivational value on the one hand and saccade latencies on the other hand. Rather, this relationship only persists in contexts where humans can maximize their reward outcome by preferring one target over the other. In sum, the present dissertation shows that there is no direct relationship between saccade latencies on the one hand and motivational value (Study III) or informational value (Study I) on the other hand. Instead, saccade latencies are sensitive to the probability that information acquired at the saccade target becomes task-relevant (Study I) and the preference of one target over the other (Study III). For perception we could show that the visual system can optimally integrate information about saccades and thus that vision does not correspond to disconnected snapshots, but rather to an integrated stream of continuous information (Study II).|
|Physical Description:||135 pages.|