Behavioral assessment (TNT task)

Participants performed the TNT task (Anderson et al. 2004), which assesses the ability to suppress unwanted memories. The TNT task consists of 3 main phases: the encoding phase, the TNT phase, and the memory test phase. The TNT phase was completed in the scanner, whereas the encoding phase and memory test phases were completed outside the scanner. Further task details are provided elsewhere for space reasons (Yang et al. 2020). Here, we briefly describe the 3 main phases.

During the encoding phase, participants were asked to study associations between 66 weakly related cue-target word pairs such that when presented each cue word, they would be able to reliably recall the associated target. To advance to the main TNT phase, participants were required to recall at least 50% of the target words when presented with their associated cue. Participants had up to 3 test-feedback cycles to demonstrate this competence after an initial study period.

After participants demonstrated successful encoding of the threshold number of cue-target word pairs, they were asked to practice the TNT task using filler pairs to ensure that participants understood the instructions pertaining to the Think and No-Think conditions. The training phase consisted of 2 blocks. After participants completed each training block, the experimenter administered a standard diagnostic questionnaire to make sure they followed the instructions correctly (Liu et al. 2021). Before entering the scanner to complete the critical TNT task, participants received a 5-min break. Then, prior to scanning, the correct word pairs were presented one final time (in a re-randomized order) to refresh the materials in the scanner environment.

The critical TNT phase consisted of 6 runs in a single session, with each run lasting 6.7 min and involving the presentation of 16 Think cues and 16 No-Think cues (each cue was presented twice in each run, according to blocked randomization, with the condition assignments for the pairings counterbalanced across participants). Cues from the remaining third of the studied word pairs did not appear during the TNT phase, as they were reserved to provide a baseline measure of memory on the final test, given that they would neither have been suppressed nor retrieved during the TNT phase. As in the practice phase, cue words from the Think condition appeared in green for 3 s, indicating that participants were to silently recall the associated target and keep it in mind for the entire time that the cue remained on the screen. For No-Think trials, the cue word appeared in red for an equal duration while participants sought to prevent the associated target word coming into mind; on these No-Think trials, participants were told to directly suppress retrieval to block out the unwanted item, without trying to distract themselves by substituting another word, image, or idea for the unwanted target. Across both conditions, participants were trained to keep their eyes and attention fixed on the presented cues throughout the trial duration. It is worth noting that the word pairs cued in the TNT phase included those that were successfully memorized in the encoding phase, as well as those that were not successfully memorized. Therefore, for the purpose of our analysis, the studied word pairs could be further conditionalized based upon whether participants had successfully recalled the target word in the final test-feedback round of the encoding phase, ensuring that only successfully learned pairs contributed to the analyses we discuss.

After the critical TNT phase, participants completed surprise memory tests for all the studied targets outside the scanner. After asking participants to think back to the original encoding phase (in order to reinstate the encoding phase context), they were asked to recall the targets to 18 filler pairs targets, cued one at a time with the original cue word. This practice test allowed participants to adjust to the instructions to try their best to recall the target words matching the given cues, regardless of what happened in the preceding phase. Participants’ memory for the critical pairings was then tested in a block-randomized fashion with respect to TNT condition, ensuring that the average test position of the Baseline, Think, and No-Think items was equated and that output interference was matched. Each critical target was tested in 2 ways during this phase, each assessing the accessibility of the learned targets. The same-probe (SP) test, involves the presentation of the original cues on the screen for 3.4 s (interstimulus interval 0.6 s) each as test prompts to elicit the learned target words. The independent-probe (IP) test is designed to measure inhibition in a way that bypasses the original cue-target association and any associative interference that the original cue may trigger (Anderson and Green 2001). Similar to the SP test, the IP test presents a category or a semantically related cue on the screen for 3.4 s (interstimulus interval 0.6 s) as test prompts to elicit the target word fitting the cues. The order of the SP and IP test was counterbalanced across subjects.