Successive failure triggered by motor exploration in a reinforcement-based reaching task

Motor exploration, a key process in reinforcement-based motor learning, is triggered by suboptimal performance and leads to increased movement variability. Consequently, failure in one trial may increase the likelihood of failure in subsequent trials. If this pattern emerges, successive failures (SFs) are not merely events to be avoided but part of the trial-and-error process of identifying optimal movements, which can facilitate motor learning. This study investigated whether and how SFs occur above chance levels during a motor learning task with binary success/failure feedback and explored the relationship between SFs and motor learning outcomes. Thirty-three healthy young adults participated in the study, performing a reaching task to pass through a target without visual feedback of their hand position. Binary feedback was provided after each trial, indicating whether the hand trajectory overlapped with the target. The results showed that the probability of two SFs was significantly higher than the square of the overall failure probability, indicating that failure streaks occurred above chance levels. A computational model, in which motor variability comprised constant motor noise and exploratory variability regulated by recent reward history (where variability increased following failure), successfully explained SF emergence. However, the learning index, defined as the difference in failure probability between the first and latter halves of the experiment, was unrelated to SF emergence. Additionally, failure streaks disappeared when participants were asked to reach one of seven randomly selected targets in each trial, suggesting that introducing environmental variability can help alleviate SFs.