The present experiment was the first fMRI study to investigate whether and how depressed individuals differ from their nondepressed counterparts in suppressing unwanted memories. Although we did not observe the expected behavioral impairment in memory suppression for MDD participants, we did find indications in the neuroimaging data concerning how they achieved memorial control. First, depressed individuals were more likely than controls to recruit the right MFG when required to suppress a memory. Second, when we considered the valence of the to-be-suppressed information, the MDD and CTL groups differed in how they modulated activity in the targets of memorial control, specifically the amygdala and hippocampus. These findings point to potential differences between depressed and nondepressed individuals that expand on prior work documenting deficits in inhibitory control in depressed individuals engaged in WM tasks (see Gotlib & Joormann, 2010).

Researchers have consistently reported that lateral prefrontal cortex (PFC) is engaged during memory suppression (Anderson, 2004; Benoit & Anderson, 2012; Benoit et al., 2015; Butler & James, 2010; Depue et al., 2007; Gagnepain et al., 2014; Levy & Anderson, 2012). This activation often spans both the ventral and dorsal aspects: the inferior frontal gyrus and MFG, respectively. Investigators have generally interpreted these lateral PFC activations as reflecting control processes that are engaged to prevent the unwanted memory from entering awareness (see, e.g., Anderson & Hanslmayr, 2014; Anderson & Levy, 2009). Although we found evidence broadly consistent with this general pattern in both the depressed and nondepressed groups, we observed engagement of a specific region within the right MFG that was specific to depressed individuals. Interestingly, this region appears to be anterior and superior to the right-MFG activations that are often seen in univariate contrasts in nondepressed participants. One possibility is that this is a distinct region that is recruited by depressed participants to suppress unwanted memories. Another possibility, which is not mutually exclusive of the first, is that the difference could reflect differential use of suppression strategies by the two groups. Participants in the TNT task report using a wide range of strategies (e.g., thinking of diversionary thoughts or attempting to let their mind go blank; see Levy & Anderson, 2008), and when participants are instructed to use different strategies, we observe differences in the recruitment of specific prefrontal regions (Benoit & Anderson, 2012). Benoit and Anderson found that participants selectively recruited a region near this area when they engaged in a direct suppression strategy in which they attempted to stop retrieval without generating alternative thoughts. Because in this study we did not instruct participants to use any specific strategy or ask them to report which strategies they used, we do not know whether the two groups used similar strategies to complete the task. One surprising aspect of this finding is that it reflects increased activity in MDD participants, whereas we predicted that PFC should be hypoactive for these individuals. In retrospect, however, we note that there have been widespread observations of heightened PFC recruitment in depression (e.g., Bär et al., 2007; Diener et al., 2012; Grimm et al., 2008; Harvey et al., 2005; Wagner et al., 2006; Walter, Wolf, Spitzer, & Vasic, 2007). Interestingly, Harvey et al. 2005 suggested that this hyperactivity is most likely to be observed when behavioral performance is matched to healthy controls, as was the case here, suggesting that this is a form of compensatory recruitment (e.g., perhaps due to strategy differences, as suggested above). Further research will be needed to replicate this unexpected pattern and to gain greater leverage on what this region contributes during the task.

In addition to increased PFC activity during suppression, it is also typical to observe decreased activity in the medial temporal lobes (Anderson, 2004; Benoit & Anderson, 2012; Benoit et al., 2015; Depue et al., 2007; Gagnepain et al., 2014; Levy & Anderson, 2012). This finding is often interpreted as indicating that the medial temporal lobes are a target of frontal cognitive control, and that their activation is down-regulated, which results in an override of their role in recovering information from long-term memory (e.g., Anderson & Hanslmayr, 2014; Anderson & Levy, 2009). In addition, there is evidence that when participants attempt to suppress negatively valenced information, they also down-regulate activity in the amygdala (Butler & James, 2010; Depue et al., 2007). In the present study, we observed a cluster in both hemispheres that included the amygdala as well as the hippocampus and that activated differentially in the depressed and nondepressed groups to the different valences. Across all combinations of valence and group, think activity in this region was numerically greater than no-think activity, consistent with prior evidence that these regions are modulated during suppression. The three-way interaction appears to reflect the differential effectiveness of this modulation between the groups for the two different valence conditions: Specifically, nondepressed participants were relatively more successful at modulating this region when suppressing negative materials, and depressed participants were relatively more successful when suppressing neutral stimuli. We expected that depressed individuals would have the most difficulty suppressing memories of negative experiences, so this pattern is broadly consistent with our predictions.2

It is interesting to note that we did not observe any significant behavioral differences between the depressed and nondepressed participants. This may not be surprising, given that other studies have also failed to observe significant SIF impairments in depressed and dysphoric populations (Hertel & Calcaterra, 2005; Joormann et al., 2009). Interestingly, the studies that have reported no significant differences in SIF as a function of depression status have tended to have smaller sample sizes than the studies that have reported significant differences, a concern that is also relevant to the present study. Interestingly, we did observe numerically greater SIF in CTLs than in MDDs on the IP test (7.3 % vs. 5.0 %, respectively, collapsed across valences), which has been argued to be more sensitive than the SP test for assessing inhibitory control ability (Anderson & Levy, 2011). When considering all of the behavioral evidence, it appears that the inhibitory impairments in this paradigm may be relatively modest. Similarly, despite our emphasis on the regions that differed between controls and depressed individuals, there was considerable similarity in the general patterns of brain activity observed in these two groups of participants, which is again consistent with the position that differences on this task due to suppression are subtle.3 That does not mean, however, that any differences would be unimportant for increasing our understanding of depression. Indeed, even small impairments in the moment-to-moment regulation of retrieval could compound dramatically, given the many retrieval opportunities that are present in everyday life. For example, we know that retrieval itself acts as a powerful learning event (see, e.g., Karpicke, 2012); thus, each instance of failed retrieval suppression could lead to that unwanted thought being encoded better in long-term memory and, in MDDs, contributing to negative schemas that have been posited to perpetuate the negative cognitions that are fundamental to this disorder (Beck, 1976). It is also important to note that, consistent with prior studies using this paradigm (Hertel & Gerstle 2003; Joormann et al., 2005), we did not observe any evidence of diminished forgetting of negative material by the depressed individuals. Although individual null results can be inconclusive, the repeated failure to confirm this prediction seems to be at odds with other evidence suggesting that depressed individuals have particular difficulty suppressing negative stimuli (for a review, see Gotlib & Joormann, 2010). Interestingly, all of the extant TNT studies have used verbal stimuli, which may not produce as strong an affective modulation as contextually richer stimuli (e.g., photographs). Perhaps future studies could use valenced picture stimuli (see Depue, Banich, & Curran, 2006; Depue et al., 2007) with these populations to further explore this issue.

In closing, our findings suggest that depressed and nondepressed individuals differ in their brain activity when they are asked to suppress memory retrieval. A fruitful next step would be to further explore the regions implicated in this study, to elucidate their mechanistic contributions to this task. This area of inquiry could also benefit from a behavioral study with a larger sample, in order to better characterize the magnitude and consistency of depression-associated behavioral impairments in this task. It would also be instructive to examine the brain activity as depressed individuals are provided with an explicit strategy to perform the suppression task. Previous research has suggested that “aided” strategies, in which an alternative diversionary thought is used, can help depressed individuals overcome deficits in SIF (Hertel & Calcaterra, 2005; Joormann et al., 2009), and can also change the pattern of brain activity in nondepressed individuals (Benoit & Anderson, 2012; Bergström, de Fockert, & Richardson-Klavehn, 2009). Finally, this line of research could help bridge two features that are often discussed as distinct aspects of depression: impaired inhibitory control and rumination. As we discussed earlier, most of the evidence for impaired inhibition in depression has focused on how items enter or are removed from WM. This has obvious implications for how individuals might encode new experiences, but it tells us relatively little about how they deal with existing long-term memories of their own experiences, which are precisely the kinds of events about which depressed individuals are likely to ruminate. Indeed, in an unselected sample of undergraduate students, levels of rumination were found to be related to impaired suppression-induced forgetting, even when controlling for deliberate reprocessing, which is related to both rumination and suppression (Fawcett et al., 2015; Joormann & Tran, 2009). Furthermore, Hertel and Gerstle (2003) found that dysphoric individuals who were high in rumination exhibited reduced SIF as compared to lower-ruminating dysphoric individuals. These findings implicate rumination as a key component of SIF abnormalities in MDD. Gaining a better understanding of how depressed individuals struggle to control access to memory will not only help elucidate the core features of depression, but may be relevant for other disorders that also involve difficulty controlling unwanted thoughts (e.g., Brewin, Gregory, Lipton, & Burgess, 2010; Catarino, Kupper, Werner-Seidler, Dalgleish, & Anderson, 2015; Depue, Burgess, Bidwell, Willcutt, & Banich, 2010; Koob & Volkow 2009; Marzi, Regina, & Righi, 2014; Verfaellie & Vasterling, 2009).