When Response Selection Becomes Gambling: Post-error Slowing and Speeding in Self-paced Colour Discrimination Tasks

People tend to slow down after committing an error in many tasks. However, some studies failed to observe such post-error slowing. Furthermore, recent work found speeding after another type of sub-optimal outcomes: people often speed up after losses in gambling situations. What features determine whether people slow down or speed up after sub-optimal outcomes (error vs. loss)? To answer this question, we focused on the role of task characteristics and control over the outcome, by making a task where we previously observed post-error slowing more like tasks where we previously observed post-loss speeding. First, we made a color-discrimination task completely self-paced (Experiment 1A) and added reward/punishment (Experiment 1B). In both experiments, post-error slowing was observed, without modulation by reward/punishment. We then manipulated task difficulty to investigate the influence of control over the outcome. Consistent with our predictions, control over the outcome modulated post-error adjustments, as participants slowed down after controllable errors, but sped up after uncontrollable errors (Experiment 3). Importantly, this effect was global as post-error speeding was observed when controllable and ???uncontrollable??? errors were intermixed (Experiment 2), suggesting an influence of overall task context. Thus, responses to sub-optimal outcomes might depend on the control over the outcome.

Automated summary

People tend to slow down after committing an error in tasks, but some studies have found that they speed up after losses in gambling situations. This study investigated the factors that determine whether people slow down or speed up after sub-optimal outcomes (error vs. loss). The results showed that control over the outcome played a role in determining the response to sub-optimal outcomes, as participants slowed down after controllable errors but sped up after uncontrollable errors. This effect was observed regardless of whether controllable and uncontrollable errors were intermixed or not.