Examining Cognitive Control Through Working Memory and Intervention: Electrophysiological Mechanisms and Frontal Midline Theta Neurofeedback

Working memory relies on the recruitment of monitoring, updating, and categorizing processes in the active storage whilst filtering out interfering traces. Frontal midline theta (FMT) oscillations (4–7.5 Hz) are thought to impose cognitive control by the means of error monitoring, conflict detection, and working memory maintenance, by allowing communication across frontoparietal networks. Whether these rhythms can be endogenously regulated to improve working-memory performance remains a matter of debate in the neurofeedback (NFB) literature. Here, we used electroencephalography (EEG) and a NFB paradigm in 38 healthy adults to investigate the electrophysiological signals of working memory processes and whether an intervention can increase this cognitive control function. Participants were divided into two groups tasked with either upregulating or downregulating FMT power across a single, 6-block NFB training session. They performed a 2-back task featuring deceptive "lure" stimuli before and after training. Behavioral models confirmed that lures carried a cognitive cost, significantly slowing response times and triggering a 43.6% false alarm rate. Stimulus-locked ERPs revealed that these memory failures are associated with a false categorization process, characterized by an enhanced centroparietal P300/Late Positive Component complex. Moreover, during target detection, a frontal N2 mismatch signal reflecting the violation of baseline non-target expectations was observed. Response-locked analyses revealed a peri-response frontocentral negativity peaking during motor execution, signalling an immediate detection of conflict during behavioral failure. On the other hand, target hits were uniquely characterised by evidence accumulation that peaks right before response. Regarding NFB, although theta power retained a focal midfrontal topography we found no evidence of group-by-block interactions, and Bayesian model selection provided robust evidence against group-specific NFB learning. Overall, these findings demonstrate that working memory failures were governed early attentional lapses for simple distractors, and late categorization errors and response inhibition mechanisms for interfering lures and true targets. Finally, there was no evidence that the NFB training successfully drove the intended changes; instead, the elevation in FMT during the NFB could be driven by cognitive engagement rather than a training effect.