To look or not to look? Reward, selection history, and oculomotor guidance

Abstract

The current eye-tracking study examined the influence of reward on oculomotor performance, and the extent to which learned stimulus-reward associations interacted with voluntary oculomotor control with a modified paradigm based on the classical antisaccade task. Participants were shown two equally salient stimuli simultaneously: a gray and a colored circle, and they were instructed to make a fast saccade to one of them. During the first phase of the experiment, participants made a fast saccade toward the colored stimulus, and their performance determined a (cash) bonus. During the second, participants made a saccade toward the gray stimulus, with no rewards available. On each trial, one of three colors was presented, each associated with high, low or no reward during the first phase. Results from the first phase showed improved accuracy and shorter saccade latencies on high-reward trials, while those from the second replicated well-known effects typical of the antisaccade task, namely, decreased accuracy and increased latency during phase II, even despite the absence of abrupt asymmetric onsets. Crucially, performance differences between phases revealed longer latencies and less accurate saccades during the second phase for high-reward trials, compared with the low- and no-reward trials. Further analyses indicated that oculomotor capture by reward signals is mainly found for saccades with short latencies, while this automatic capture can be overridden through voluntary control with longer ones. These results highlight the natural flexibility and adaptability of the attentional system, and the role of reward in modulating this plasticity. NEW & NOTEWORTHY Typically, in the antisaccade task, participants need to suppress an automatic orienting reflex toward a suddenly appearing peripheral stimulus. Here, we introduce an alternative antisaccade task without such abrupt onsets. We replicate well-known antisaccade effects (more errors and longer latencies), demonstrating the role of reward in developing selective oculomotor biases. Results highlight how reward and selection history facilitate developing automatic biases from goal-driven behavior, and they suggest that this process responds to individual differences in impulsivity.

Keywords: antisaccade; attentional bias; eye-tracking; goal-driven; reward-driven.

Fig. 1.

Task display sequence and trial types. In this image, colors are represented as lined circles for the sake of clarity, but solid colors were used on the actual experiment. Note that reward-color associations were counter-balanced across participants, colored circles were equally likely to appear at either side of the saccade display, and trial types were intermixed within blocks, all equally likely to occur on any given trial.

Fig. 2.

Phase × Reward interactions. A and C: latency of saccades made to the target (i.e., correct saccades only) and % of error saccades (respectively) by phase and reward level. B and D: computed differences in latency and % of error saccades (phase II and phase I). Error bars represent within-subjects confidence intervals (Morey 2008). Lines emphasizing significant effects represent t-tests results on the computed differences. ***P ≤ 0.001; *P ≤ 0.05.

Fig. 3.

Phase × reward × block interactions. A and B: presents the latency of saccades made to the target (i.e., correct saccades only) and % of error saccades (respectively) by phase, reward level, and block. Shaded areas represent within-subject confidence intervals (Morey 2008).

Fig. 4.

Vincentized proportion of saccades to distractor for each phase (solid and dashed lines for phase I and II, respectively) and reward level. Shaded regions correspond to within-subject confidence intervals (Morey 2008). Each data point represents a decile in the corresponding latency cumulative distribution. ***P ≤ 0.001; *P ≤ 0.05; +P ≤ 0.06.

Fig. 5.

Spearman ρ correlations between second-order motor impulsivity and the proportion of saccades to the distractor by phase and reward (low reward data not shown, as it did not significantly correlate with motor impulsivity).