Classification aided analysis of oscillatory signatures in controlled retrieval

Control processes are critical for both facilitating and suppressing memory retrieval, but these processes are not well understood. The current work, inspired by a similar fMRI design (Detre et al., in press), used a modified Think/No-Think(TNT) paradigm to investigate the neural signatures of volition over enhancing and suppressing memory retrieval. Previous studies have shown memory enhancement when well-learned stimulus pairs are restudied in cued recall (“Recall or think of studied pair item”), and degradation when restudied with cued suppression (“Avoid thinking of studied pair item”). We used category-based (faces vs. scenes) multivariate classification of electroencephalography signals to determine if individual target items were successfully retrieved or suppressed. A logistic regression based on classifier output determined that retrieval activation during the cued recall/suppression period was a predictor for subsequent memory. Labeling trials with this internal measure, as opposed to their nominal Think vs. No-Think condition, revealed the classic TNT pattern of enhanced memory for successful cued-retrieval and degraded memory for cued-suppression. This classification process enabled a more selective investigation into the time-frequency signatures of control over retrieval. Comparing controlled retrieval vs. controlled suppression, results showed more prominent Theta oscillations (3 to 8 Hz) in controlled retrieval. Beta oscillations (12 to 30 Hz) were involved in high levels of both controlled retrieval and suppression, suggesting it may have a more general control-related role. These results suggest unique roles for these frequency bands in retrieval processes.