Abstract
Saccade target selection is known to be influenced by bottom-up factors, like salient objects as well as by top-down factors like reward. We wanted to investigate whether saliency and reward can also affect the fine tuning of saccadic landing positions. We presented two luminance-defined, overlapping, blurred patches of opposite contrast polarity in front of a homogeneous gray background. One patch had a fixed contrast of 20%, the contrast of the other was varied to manipulate the saliency. In the saliency baseline condition, we instructed subjects to make saccades to the target configuration. In the reward condition, subjects won points for landing on one target and lost points for landing on the other target. To manipulate the reward, we varied the relative amount of bonus and penalty. The subjects were instructed to make as many points as possible, which were converted into a monetary reward at the end of the experiment. Both saliency and reward influenced the saccade landing positions. Subjects' saccades landed closer to the patches that were more salient and rewarded. A model was able to account for these data by linearly weighting and combining saliency and reward. Saliency was modeled as the average of the two target positions, weighted by their relative contrast. For the reward, we predicted the optimal saccade endpoint that maximizes gain, based on the individual saccade variability and the bonus-penalty ratio. Interestingly, the relative weights were modulated by the latency of the saccades. While fast saccades nearly exclusively used salience to determine the landing point, slower saccades gave a higher weight to reward information. Our results show that rewards do not only affect saccadic target selection, but also the exact landing position within the target. However, integration of this top-down factor is time-consuming and can be overridden in saccades with short latencies.
This work was supported by the DFG Forschergruppe FOR 560.