Purchase this article with an account.
Christian Wolf, Anna Heuer, Anna Schubö, Alexander Schütz; The necessity to choose causes effects of reward. Journal of Vision 2017;17(10):909. https://doi.org/10.1167/17.10.909.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Humans can maximize reward by choosing the option with the largest expected value (probability × magnitude of reward). Expected value is taken into account for the preparation of saccadic eye movements, as latencies to single-targets are negatively correlated with expected value (Milstein & Dorris, 2007). Here we show that this relationship only holds when responses to single targets are embedded in a context where participants sometimes have to choose between different options. Participants were rewarded differently for saccades to targets appearing to the left or right from fixation. In any trial, either one less or highly rewarded target (single-trial) or two targets (choice-trial) were displayed. In choice-trials, participants had to decide for one of the two targets and received the corresponding reward. We show that in single-trials, saccade latencies to the less rewarded target were elevated only when choice-trials were present. This effect increased with the amount of choice trials within a block. When changing the reward congruency between choice- and single-trials, single-trial latencies only depended on which target was preferred in choice-trials but not on the single-trial reward, showing that choice-trials causally determined the latency delay. Two further experiments emphasize the functional role of this delay: First, its magnitude scaled with the difficulty to make a reward-maximizing response in choice-trials. Second, it was reduced when single-trials were cued in advance. Moreover, fitting the LATER model (Carpenter & Williams, 1995) implied that prolonged latencies to the less rewarded target are caused by a reduced baseline activity in the decision signal. Taken together, our results suggest that there is no direct connection between expected value and saccade preparation. Instead, expected value affects latencies only when there is a functional reason to do so, for instance when participants have to choose between options differing in expected value.
Meeting abstract presented at VSS 2017
This PDF is available to Subscribers Only