The goal of temporal difference (TD) reinforcement learning is to maximize outcomes and improve future decision-making. It does so by utilizing a prediction error (PE), which quantifies the difference between the expected and the obtained outcome. In gambling tasks, however, decision-making cannot be improved because of the lack of learnability. On the basis of the idea that TD utilizes two independent bits of information from the PE (valence and surprise), we asked which of these aspects is affected when a task is not learnable. We contrasted behavioral data and ERPs in a learning variant and a gambling variant of a simple two-armed bandit task, in which outcome sequences were matched across tasks. Participants were explicitly informed that feedback could be used to improve performance in the learning task but not in the gambling task, and we predicted a corresponding modulation of the aspects of the PE. We used a model-based analysis of ERP data to extract the neural footprints of the valence and surprise information in the two tasks. Our results revealed that task learnability modulates reinforcement learning via the suppression of surprise processing but leaves the processing of valence unaffected. On the basis of our model and the data, we propose that task learnability can selectively suppress TD learning as well as alter behavioral adaptation based on a flexible cost–benefit arbitration.

Dual-task scenarios require a coordinated regulation of the processing order of component tasks in light of capacity limitations during response selection. A number of behavioral and neuroimaging findings suggest a distinct set of control processes involved in preparing this task order. In this study, we investigated electrophysiological correlates of task-order preparation in a variant of the overlapping dual-task paradigm with cue-determined task order that resulted in trials with blockwise fixed task order as well as trials with repeated and switched task order in blocks with variable task order. During the cue–stimulus interval, we found an earlier centroparietal order-mixing positivity and a later parietal order-switch positivity. A decoding approach based on multivariate pattern analysis showed that the order-mixing positivity is a necessary prerequisite for successful order selection, whereas the order-switch positivity appears to facilitate the implementation of a new task order after its selection. These correlates of order preparation share striking similarities to commonly found potentials involved in the preparation of individual tasks in the (single-)task-switching paradigm, which is strong empirical support for the account that the underlying preparatory processes are to be considered as higher-level control signals that are implemented independently of specific task representations.

The present study tested error detection theories of the error-related negativity (ERN) by investigating the relation between ERN amplitude and error detectability. To this end, ERN amplitudes were compared with a behavioral measure of error detectability across two different error types in a four-choice flanker task. If an erroneous response was associated with the flankers, it was considered a flanker error, otherwise it was considered a nonflanker error. Two experiments revealed that, whereas detectability was better for nonflanker errors than for flanker errors, ERN amplitudes were larger for flanker errors than for nonflanker errors. Moreover, undetected errors led to strongly reduced ERN amplitudes relative to detected errors. These results suggest that, although error detection is necessary for an ERN to occur, the ERN amplitude is not related to error detectability but rather to error significance.