Successful Goal-Directed Memory Suppression is Associated With Increased Inter-Hemispheric Coordination Between Right and Left Frontoparietal Control Networks

Abstract

The neural basis of suppressing conscious access to one’s own memories has recently received considerable attention, with several studies suggesting this process engages frontal-parietal cognitive control regions. However, researchers to date have not examined the way right and left hemisphere cognitive control networks coordinate with one another to accomplish this. We had 48 participants (25 female) complete a Think/No Think (T/NT) task for memories of emotionally unpleasant visual scenes while undergoing functional magnetic resonance imaging. We used generalized psychophysiologic interaction analyses to examine functional connectivity between right and left hemisphere frontal-parietal regions during memory suppression. Participants who were better at memory suppression, as assessed by greater numbers of forgotten memories in the NT than T conditions, also showed greater functional connectivity between multiple right and left hemisphere control regions. This suggests that individual differences in memory suppression ability may be partially explained by differences in task-specific inter-hemispheric coordination.

Keywords

Retrieval suppression dorsolateral prefrontal cortex inferior parietal cortex precuneus emotion

Introduction

Recent work within cognitive neuroscience (Anderson & Green, 2001; Anderson & Hanslmayr, 2014) has demonstrated that humans can reduce the accessibility of memories they wish to keep out of awareness. Specifically, multiple studies using “Think/No Think” (T/NT) tasks have shown that the right dorsolateral prefrontal cortex (rDLPFC) and ventrolateral prefrontal cortex (rVLPFC) are engaged when subjects attempt to suppress memory retrieval across a range of stimulus domains (Anderson et al., 2004; Benoit & Anderson, 2012; Butler & James, 2010; Depue, Curran, & Banich, 2007; Gagnepain, Henson, & Anderson, 2014; Levy & Anderson, 2012; Paz-Alonso, Bunge, Anderson, & Ghetti, 2013). Hippocampal (and sometimes surrounding medial temporal lobe) activation simultaneously decreased in the aforementioned T/NT studies, suggesting that retrieval suppression engages top-down inhibitory control. Effective connectivity analyses have also found an influence of rDLPFC on the hippocampus, and reductions in hippocampal activation were shown to predict later forgetting (Benoit & Anderson, 2012; Gagnepain et al., 2014). Finally, rDLPFC-mediated inhibitory control also appears to interact with content-specific posterior neocortical regions. For example, when the to-be-suppressed memory is visual in nature, cortical regions of the visual system appear to be inhibited (Gagnepain et al., 2014). DLPFC has also been observed to have significant effective connectivity with these visual cortical regions during suppression, and this suppression effect even appears to reduce later visual priming effects (associated with perceptual identification). Thus, this type of goal-directed retrieval suppression appears to involve rDLPFC-mediated inhibition of both hippocampal and neocortical regions associated with memory retrieval (Gagnepain et al., 2014; Kim & Yi, 2013).

In contrast, left VLPFC activations have been associated with intentional retrieval, and these also predict greater hippocampal activation (Benoit & Anderson, 2012; Bergström, de Fockert, & Richardson-Klavehn, 2009; Kuhl, Dudukovic, Kahn, & Wagner, 2007). Retrieval also initiates a well-characterized event-related potential response in parietal cortex, and this event-related potential is reduced during suppression (Bergström et al., 2009). Successful coordination between left and right hemisphere frontoparietal control networks (associated with goal-directed retrieval and retrieval suppression, respectively) may therefore represent an important variable in the cognitive control of memory. This need for coordination is also supported by a recent electroencephalography study examining brain oscillation frequencies in a T/NT task, which found preparatory increases (i.e., in response to a pre-reminder instructional cue) in both theta frequency power and alpha frequency phase synchronization within frontoparietal cortex in the NT condition (Waldhauser, Bäuml, & Hanslmayr, 2015). They further observed that successful suppression was associated with greater preparatory increases in theta frequency power across both left and right prefrontal cortex (in contrast, during the T/NT reminder phase, a decrease in theta power within the medial temporal lobes was observed to correlate with successful suppression). While these findings suggest that successful suppression may involve better inter-hemispheric coordination, no functional magnetic resonance imaging (fMRI) studies have directly assessed the functional interactions between left and right hemisphere control networks during retrieval and retrieval suppression. One functional connectivity study did find that successful suppression was associated with greater average network coupling between the hippocampus and left PFC, cingulate, and posterior parietal regions (Paz-Alonso et al., 2013); however, the focus of that study was not on inter-hemispheric network coordination.

In the present study, we therefore sought to use fMRI to examine the interactions between left and right hemisphere frontoparietal control regions within a previously validated T/NT task (Depue et al., 2007). Based on the studies described above, we first hypothesized that a right frontoparietal network would increase in activation during retrieval suppression in comparison to retrieval, and that a left frontoparietal network would be engaged by retrieval compared to retrieval suppression. Second, we hypothesized that increased functional connectivity between left and right hemisphere control networks would predict increased forgetting of suppressed items relative to retrieved items. This second hypothesis was based on the idea that, since retrieval is more left lateralized (Benoit & Anderson, 2012; Bergström et al., 2009; Kuhl et al., 2007), if an item gets retrieved during an NT trial, then better inter-hemispheric coordination would allow the right hemisphere network to exert control over the left hemisphere network and suppress the intruding item that is being wrongfully retrieved.

Methods

Participants

Forty-eight healthy, right-handed adults (25 female) participated in the present study. These participants ranged in age from 18 to 45 years (M = 29.8, SD = 7.62). Participants did not have any history of psychiatric, neurological, or substance use disorders.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The research protocol of the present study was also reviewed and approved by the U.S. Army Human Research Protections Office, as well as by the Institutional Review Board of McLean Hospital. All participants provided written informed consent.

Procedure

T/NT paradigm

Both prior to and during fMRI, participants completed a variant of a previously validated T/NT paradigm (Depue, Banich, & Curran, 2006; Depue et al., 2007). This T/NT paradigm has three phases: (1) the training phase; (2) the task phase; and (3) the recall phase.

The training phase was done prior to entering the scanner. In this phase, participants learned associations between 40 face-scene image pairs. Scene images were drawn from the negatively valenced subgroup (mean valence rating ≤3.0) of the commonly used International Affective Picture System (Lang, Bradley, & Cuthbert, 2008). Face images were drawn from the freely available Aberdeen 2D face set (http://pics.stir.ac.uk/2D_face_sets.htm). This training phase (on a laptop computer) involved the repeated simultaneous presentation of each face-scene pair, followed by a test in which participants were shown one face at a time and then asked to choose which of two scenes it was previously paired with. Feedback with regard to accuracy was given after each choice, and participants cycled through this training and testing procedure a minimum of two times. Training continued until either 97.5% accuracy was achieved or until four cycles were completed. The mean number of cycles completed was 2.73 (SD = 0.82). The mean accuracy achieved was 97.9% (SD = 3.8%).

Immediately after completion of the training phase, participants began the task phase. In this phase, participants were told they would perform a task in which they were shown a subset of the previous faces without their associated scenes. The instruction was given that if the face was surrounded by a green border, they were to try their best to “think of the picture that was previously associated with that face.” If the face was instead surrounded by a red border, they were instructed to “not let the picture that was previously associated with that face enter your consciousness,” and to do so without taking their eyes off of the face. Prior research with similar T/NT tasks has utilized 12 repetitions of the stimuli in the scanner (e.g., Depue et al., 2007). In order to minimize participant burden in the scanner and optimize the scan toward the suppression period, they were asked to perform the first 75% of the task (i.e., three repetitions) outside the scanner (on a laptop), lasting 8 minutes each. Each of these 8-minute runs provided three exposures to the same subset of previously learned faces, for a total of nine repetitions of each face outside of the scanner. Approximately 1 hour later, participants were then asked to perform the T/NT task once again during fMRI scanning (i.e., 3 repetitions of each face in the scanner). During each of the four 8-minute task runs (three outside the scanner, followed by one inside the scanner), each face was shown for 3.5 seconds (followed by a 500-millisecond black screen) in a pseudo-random order, with a crosshair trial (3.5 seconds duration) interspersed pseudo-randomly (on average once every five trials). Each presentation of a particular face was always shown with either the T or the NT instruction (i.e., faces were consistently assigned to only one type of trial). A two-sample t-test confirmed that the normative valence ratings of the International Affective Picture System scene images associated with these faces were not significantly different between the T and NT conditions (mean normative valence ratings were 2.31 (SD = 0.47) and 2.41 (SD = 0.48), respectively; t = .59, p = 0.56). Task stimuli both inside and outside the scanner, as well as during training-testing cycles, were all presented using ePrime2 presentation software (http://www.pstnet.com/eprime.cfm).

After completing the fMRI task, participants began the recall phase. In this third phase, participants were immediately given a free recall test (while still laying in the scanner) in which they were again shown each face (including 8 “control” faces that were trained but never used in any T/NT task runs) and asked to verbally describe the image it was associated with. Participants were encouraged to try their best to remember and describe the details of each image. Two independent blind raters then compared these (transcribed) verbal descriptions to the pictures and judged whether each image had been remembered or forgotten. Similar to what has been done in previous T/NT studies (e.g., Depue et al., 2007), if participants clearly remembered an image with two or more words correctly describing its main elements (e.g., “electric chair,” “guy holding knife to woman’s neck”), the image was scored as “remembered”; whereas if participants had no recollection of an image, or provided an incorrect description of it, then it was scored as “forgotten.” If discrepancies arose between the two raters, a blind five-member panel of researchers examined the items and settled the decision based on a majority vote.

Based on these results, a “suppression index” (SI) was calculated for each participant (i.e., as we have done previously; Smith, Alkozei, Lane, & Killgore, 2016) by dividing the number of scenes forgotten in the NT condition (NT_f; i.e., intentional forgetting) by the number of images forgotten in the T condition (T_f; i.e., unintentional forgetting):

SI = {NT}_{f} / T_{f}

Thus, SI represents the ratio of forgetting in these two conditions for each participant, such that higher SI scores indicate a greater ability to intentionally forget the emotional scenes presented during the training phase (i.e., relative to each participant’s own unintentional forgetting level). That is, these scores capture the idea that the number of emotional scenes forgotten in the NT condition over and above those unintentionally forgotten in the T condition can be attributed to the effects of intentional retrieval suppression; therefore, the greater this ratio was in an individual, the more effective intentional suppression was in that individual at producing later forgetting.

Neuroimaging methods

Neuroimaging was performed using a 3T Siemens Tim Trio scanner (Erlangen, Germany) with a 12-channel head coil. T1-weighted structural 3D MPRAGE images were acquired (TR/TE/flip angle = 2.1 s/2.25 ms/12°) covering 128 sagittal slices (256 × 256) with a slice thickness of 1.33 mm (voxel size = 1.33 × 1× 1). Functional T2*-weighted scans were acquired over 34 transverse slices (3.5 mm thickness); 240 volumes were collected with an interleaved sequence (TR/TE/flip angle = 2.0 s/30 ms/90°). The voxel size of the T2* sequence was 3.5 × 3.5 × 3.5 mm. The field of view was 22.4 cm, with a 64 × 64 acquisition matrix.

Image processing

Preprocessing steps on all MRI scans, as well as subsequent statistical analyses, were performed using SPM8 (Wellcome Department of Cognitive Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/spm ). Raw functional images were realigned, unwarped, and coregistered to each subject’s MPRAGE image in accordance with standard algorithms. Images were then normalized to Montreal Neurological Institute coordinate space, spatially smoothed (6 mm full-width at half maximum), and resliced to 2 × 2 × 2 mm voxels. The standard canonical hemodynamic response function in SPM was used, low-frequency confounds were minimized with a 128-second high-pass filter, and serial autocorrelation was corrected using the AR(1) function. As is commonly done in similar studies of neural activation and functional connectivity (e.g., Ellard, Barlow, Whitfield-Gabrieli, Gabrieli, & Deckersbach, 2017; Keller et al., 2015; Smith et al., 2016), the Artifact Detection Tool ( http://www.nitrc.org/projects/artifact_detect/ ) was also used to regress out scans as nuisance covariates in the first-level analysis exceeding 3 SD in mean global intensity and scan-to-scan motion that exceeded 1.0 mm.

Statistical analysis

For each participant, a general linear model was specified to contrast activation between the NT condition and the T condition. Each trial was modeled as a 3.5-second interval. Motion regressors (generated by Artifact Detection Tool—see image processing above) were also added to each of these first-level designs. These contrast images were then entered into a second-level SPM analysis (one-sample t-test) to assess the main effect of each of our two contrasts of interest (i.e., NT > T and T > NT). These analyses were thresholded at p < .05 (false discovery rate (FDR) corrected) and k (extent) ≥ 5 contiguous voxels. Contrast estimates across subjects were also extracted from the peak voxel of specific significant clusters (see results section) found within these analyses (based on a priori hypotheses), and correlated with SI scores to test whether activation differences in these regions were related to goal-directed suppression ability.

To examine our further hypothesis about functional connectivity and goal-directed suppression ability (described above), the first-level results for each participant were then imported into the publicly available CONN functional connectivity toolbox (version 15.C; https://www.nitrc.org/projects/conn) in order to perform a generalized form of seed-to-voxel psychophysiologic interaction (gPPI) analyses. This method was chosen because of its improved sensitivity and specificity in detecting connectivity effects (McLaren, Ries, Xu, & Johnson, 2012). After further preprocessing the data according to a standard pipeline of algorithms commonly used to remove artifactual correlations in functional connectivity analyses (Whitfield-Gabrieli & Nieto-Castanon, 2012), the CONN toolbox uses the following equation to estimate PPI effects:

Y_{k} = H (x_{a})

Y_{i} = [H (x_{a} * g_{p})] * β_{i} + [Y_{k} H (g_{p}) G] * β_{G} + e_{i}

In these equations, H refers to the HRF in Toeplitz matrix form; Y_k refers to the BOLD signal observed in the seed region; x_a refers to the estimated neural activity from the BOLD signal in the seed region (Gitelman, Penny, Ashburner, & Friston, 2003); Y_i refers to the BOLD signal observed at each voxel in the brain; β_i is a matrix of the beta estimates of the psychophysiological interaction terms; β_G is a matrix of the beta estimates of the seed region BOLD signal (Y_k), covariates of no interest (G), and convolution of psychological vectors H(g_p); e_i is a vector of the residuals of model; and g_p is a matrix of N columns, where N refers to the number of conditions in the experiment (formed by separating the time when the conditions are present into separate columns).

For our analysis, we chose specific significant clusters (see results section) found within the second-level contrasts of the NT and T conditions described above. These were extracted from SPM (by saving the individual cluster images and importing them into the CONN toolbox as regions-of-interest) and used as seed regions within gPPI, in order to assess changes in regional inter-hemispheric connectivity between T and NT task conditions. We also ran correlation analyses to test whether connectivity differences across subjects were related to SI scores, and therefore related to differences in goal-directed suppression ability. For these connectivity analyses, we used a peak threshold of p < .001 (uncorrected) and a cluster-level threshold of p < .05 (FDR-corrected).

Results

Behavioral measures

A paired t-test first confirmed that the number of scenes forgotten during the NT (M = 50.4%, SD = 19.9%) condition was significantly greater than during the T condition (M = 45.7%, SD = 22.6%) across participants (t(47) = 2.13, p = .04, 2-tailed; Cohen’s d = .31). The mean SI across subjects was 1.38 (SD = 0.98), also indicating that, on average, participants forgot more scenes within the NT condition than within the T condition. However, there was considerable variability across participants, with some showing greater forgetting in the T than the NT condition. One participant was identified as an outlier, with an SI score more than three standard deviations from the mean, and was removed from all further analyses examining neuroimaging correlates of SI.

As is standardly done in T/NT paradigms, we also compared forgetting for both the T and NT conditions with forgetting of “control” images (that were trained, but never subsequently retrieved or suppressed). Contrary to previous findings using this paradigm (Depue et al., 2007), paired t-tests revealed that forgetting was significantly greater for control images (M = 65.8%, SD = 22.9%) than for either the T (t(47) = 15.35, p < .001, two-tailed; Cohen’s d = 1.14) or NT (t(47) = 14.95, p < .001, two-tailed; Cohen’s d = .87) conditions.

fMRI activation contrasts

NT > T

As predicted, this contrast revealed significant activations within multiple right hemisphere regions, including large clusters spanning the rDLPFC, rVLPFC, insula, inferior parietal lobule (rIPL), posterior cingulate (PCC), and precuneus. Smaller clusters within the left IPL, DLPFC, precuneus, rostral anterior cingulate, as well as bilateral subgenual anterior cingulate and right lateral temporal lobe were also observed (Table 1(a), Figure 1(a)).

Table 1.

fMRI activation results.

Brain region	Peak voxel coordinate	Cluster size (k_E)	T-score
(a) No Think > Think (whole-brain, FDR-corrected peak threshold, p ≤ 0.05)
Right inferior parietal lobule (Supramarginal gyrus)	64, −34, 38	1687	6.52
Right middle frontal gyrus (DLPFC)	30, 34, 28	929	6.02
Right insula/inferior frontal gyrus (VLPFC)	48, 14, 6	820	5.71
Right superior frontal gyrus	10, 14, 60	576	5.30
Left inferior parietal lobule	−60, −38, 42	145	4.80
Right precuneus	6, −58, 54	340	4.64
Right mid-/posterior cingulate cortex	14, −28, 38	134	4.20
Right middle/superior temporal gyrus	56, −16, −4	38	4.11
Left middle frontal gyrus (DLPFC)	−30, 28, 32	29	3.77
Bilateral subgenual anterior cingulate cortex	0, 30, −4	25	3.63
Left rostral anterior cingulate cortex	−4, 32, 8	6	3.61
Right middle frontal gyrus	42, 40, −2	7	3.49
Right middle temporal gyrus	52, −32, −6	10	3.44
Right insula	36, −14, 0	11	3.42
Right superior temporal gyrus	38, −36, 16	9	3.36
Right superior temporal gyrus	44, 2, −14	11	3.32
Left precuneus	−12, −60, 56	6	3.25
(b) Think > No Think (whole-brain, FDR-corrected peak threshold, p ≤ 0.05)
Left inferior parietal lobule	−30, −60, 42	413	5.64
Left inferior frontal gyrus (VLPFC)	−42, 22, 24	581	5.54
Left occipital cortex	−6, −98, 2	168	5.13
Left precuneus	−6, −68, 34	81	4.55
Left posterior cingulate cortex	−4, −32, 34	89	4.55
Right occipital cortex	16, −96, 4	151	4.42
Right occipital cortex (Lingual gyrus)	10, −80, −16	19	4.33
Right parahippocampal gyrus	8, −40, 0	14	4.12
Left retrosplenial cingulate cortex	−6, −50, 14	31	4.00
Left cerebellum	−4, −42, −2	5	3.97
Left middle frontal gyrus (DLPFC)	−36, 4, 56	12	3.93
Right cerebellum	14, −54, −18	9	3.90
Right occipital cortex (Lingual gyrus)	18, −86, −10	6	3.76
Left hippocampus^a	−22, −30, −4	4	3.72

Lowered cluster size threshold to check for activations within hippocampus (based on a priori hypothesis).

Figure 1.

(A) Right frontal-parietal regions activated by the “No Think” greater than “Think” (NT > T) contrast (shown in red). Left frontal-parietal regions activated by the “Think” greater than “No Think” (T > NT) contrast (shown in blue). (B, C) Illustration of greater positive functional connectivity differences in the “No Think” than “Think” (NT > T) condition (based on gPPI analyses; arrows directed from seed to target regions) found in participants with better voluntary suppression ability (as measured with SI scores). (B) Illustrates this connectivity pattern found between right DLPFC and both the left retrosplenial cingulate cortex (RCC) and bilateral occipital cortex. (C) Illustrates this connectivity pattern found between the right and left precuneus (PreC), as well as between the Right Inferior Parietal Lobule (IPL) and a left hemisphere region encompassing both the Superior (SPL) and Inferior Parietal Lobule.

T > NT

As predicted, this contrast revealed significant activations within multiple left hemisphere regions, including large clusters spanning the left DLPFC, VLPFC, IPL, posterior/retrosplenial cingulate, and precuneus (Table 1(b), Figure 1(a)). Occipital (visual) cortex activation (bilaterally) and right parahippocampal activation was also observed in this contrast. Due to the limited size of the hippocampus, we also chose to lower our cluster size threshold; under this reduced threshold, we also observed a four-voxel cluster within the left hippocampus (which still survived FDR correction).

SI correlations

Based on a priori hypotheses, we extracted contrast estimates for the peak voxel of the rDLPFC (Table 1(a)) and left hippocampal (Table 1(b)) clusters from the above contrasts of the T and NT conditions. As predicted, rDLPFC activity (during NT > T) (r(45) = .31, p = .017, one-tailed) and left hippocampal activity (during T > NT) (r(45) = .28, p = .028, one-tailed) were both significantly positively correlated with SI. While not one of our a priori hypotheses, based on previous research suggesting that the left precuneus plays an important role in visual memory retrieval (Cavanna & Trimble, 2006; Lundstrom et al., 2003) we also analyzed correlations between peak voxel contrast estimates in this region (listed in Table 1(b)) and SI scores across participants. Left precuneus activations from the T > NT contrast were also significantly correlated with SI scores (r(45) = .33, p = .02, two-tailed).

gPPI analyses

Our initial gPPI analyses revealed several significant functional connectivity differences between the T and NT conditions across participants. However, as our focus here is on predictors of successful suppression, these results are detailed within supplementary materials.

Our further analyses examining the relation between connectivity and SI scores revealed multiple functional connectivity patterns that predicted higher SI scores (Table 2). As seed regions for these analyses, we chose to examine the rDLPFC, rIPL, and right Precuneus clusters from the NT > T contrast, based on previous research implicating each of these structures in episodic memory and retrieval suppression (Anderson & Hanslmayr, 2014; Cavanna & Trimble, 2006; Depue, 2012; Lundstrom et al., 2003). Each of these clusters was exported from our second-level SPM results of the NT > T contrast and used as a seed region within the CONN toolbox (see methods). We found that higher SI scores were predicted by greater positive inter-hemispheric connectivity in the NT than T conditions between the following regions: rDLPFC (seed) and Bilateral Occipital Cortex/Left Retrosplenial Cingulate (Table 2(a), Figure 1(b)); rIPL (seed) and Left Superior/Inferior Parietal Lobule (Table 2(b), Figure 1(c)); Right Precuneus (seed) and Left Precuneus/Superior Parietal Lobule (Table 2(c), Figure 1(c)).

Table 2.

gPPI suppression index results (NT > T).^a

Brain region	Peak voxel coordinate	Cluster size (k_E)	T-score
(a) Seed region = Right middle frontal gyrus/DLPFC (peak voxel = 30, 34, 28); positive contrast
Bilateral occipital cortex	14, −78, −4	289	4.76
Left occipital cortex/retrosplenial cingulate gyrus	−10, −68, 14	86	4.31
(b) Seed region = Right inferior parietal lobule (peak voxel = 64, −34, 38); positive contrast
Left superior/inferior parietal lobule	−24, −68, 46	88	4.47
(c) Seed region = Right precuneus (peak voxel = 6, −58, 54); positive contrast
Left superior parietal lobule/precuneus	−18, −66, 58	77	5.00

Peak threshold = P ≤ 0.001, uncorrected; Cluster threshold = P ≤ 0.05, FDR-corrected.

Discussion

In this paper, we used a previously validated T/NT paradigm (Depue et al., 2007) within a relatively large-sized sample (i.e., compared to most previous neuroimaging studies using the T/NT paradigm) to test the hypothesis that better suppression ability corresponded to greater coordination between left and right hemisphere frontoparietal control networks (associated with retrieval and retrieval suppression, respectively) (Benoit, Hulbert, Huddleston, & Anderson, 2015). First, as predicted, we observed strong lateralized effects when contrasting the T and NT conditions (Figure 1(a)). While NT > T revealed mainly right frontal and parietal regions (rDLPFC/rVLPFC, rIPL, PCC, insula, and precuneus), T > NT revealed a left hemisphere network (left VLPFC, IPL, posterior/retrosplenial cingulate, and precuneus). To our knowledge, no other studies of retrieval suppression have reported such starkly lateralized findings, perhaps due to smaller sample sizes (i.e., with less statistical power) compared to the present study. Less activation in hippocampal, parahippocampal, and visual cortical regions was also observed in the NT (compared to the T) condition, consistent with previous literature suggesting that memory suppression involves right hemisphere-mediated top-down inhibition of medial temporal lobe memory systems, and associated perceptual representations (Anderson & Hanslmayr, 2014; Depue, 2012). Also consistent with this interpretation, the contrast estimates we extracted from the peak voxels of both rDLPFC (in the NT > T contrast) and left hippocampal activation clusters (in the T > NT contrast) were significantly correlated with suppression ability (higher SI scores) across participants. This suggests that greater rDLPFC activations and greater deactivations of the hippocampus were both related to successful goal-directed forgetting (relative to unintentional forgetting of T items).

Second, our gPPI analyses also largely confirmed our a priori hypotheses (Figure 1(b) and (c)). We found that greater suppression ability across participants (higher SI scores) corresponded to greater increases in positive functional connectivity during the NT condition (relative to the T condition) between multiple homotopic cognitive control regions. This pattern was observed between the right and left precuneus as well as between right and left IPL. As left precuneus activation (in T > NT) was also found to be significantly correlated with SI scores, this suggests that successful retrieval suppression may involve the successful gating of left precuneus-mediated retrieval mechanisms by the right precuneus during the NT condition. This is also consistent with previous evidence suggesting that the precuneus plays an important role in episodic memory retrieval and visual imagery (Cavanna & Trimble, 2006; Lundstrom et al., 2003). Our findings regarding IPL are also consistent with previous studies of memory suppression that have implicated this region (Bergström et al., 2009; Paz-Alonso et al., 2013; Waldhauser et al., 2015). In tandem with prefrontal regions, lateral parietal regions have been implicated in the top-down attentional amplification associated with global neuronal workspace models of consciousness (Dehaene, 2014). Within these models, left parietal activations within the T > NT contrast may plausibly represent attentional amplification of representations within visual cortex and hippocampus associated with allowing a memory to enter awareness. As we find that higher SI scores are associated with increased connectivity between right and left parietal regions, this suggests that successful retrieval suppression may require inter-hemispheric coordination between these regions to prevent attentional amplification of NT item representations.

Our gPPI analyses also revealed that greater suppression ability was associated with greater increases in functional connectivity in the NT condition (relative to the T condition) between rDLPFC and both left and right visual cortex (as well as the retrosplenial cortex). As visual cortex activity was also suppressed in the NT condition (relative to the T condition), this suggests that successful retrieval suppression was associated with a right DLPFC-mediated top-down inhibition of visual imagery. As many studies have found retrosplenial activation to be associated with successful recollection (Rugg & Vilberg, 2013), it also appears plausible that successful rDLPFC-mediated coordination of retrosplenial function could play an important role in successfully suppressing retrieval. These results may also support one plausible mechanism underlying the recently observed effect of retrieval suppression on subsequent visual priming (Gagnepain et al., 2014). That is, consistent with the results of their study, our results suggest that reduced visual priming may involve the long-term effects of direct suppression of visual cortex (and perhaps retrosplenial cortex) by the rDLPFC during NT trials.

Two further considerations regarding these gPPI results are important. First, it may appear counterintuitive that increases in positive functional connectivity were found between regions with opposite activation profiles with respect to the NT and T conditions (e.g., the right and left precuneus). However, gPPI analyses identify voxels that are more strongly related to a seed region in one condition than another over and above the variance explained by the main effect of task (i.e., NT > T). Thus, our results show that when task-related changes in activity are accounted for, there remains a positive relation between the remaining signal variance in these regions. This suggests that signals may continue to pass between the regions we observed in our gPPI analyses, even though their overall activation profiles differ between the NT and T conditions. A second issue is that, although our gPPI results are consistent with the directional interpretations we have suggested above, it is important to emphasize that gPPI analyses only allow the firm conclusion that there is a relationship between the activations observed in two regions. Thus, we can only claim with confidence that the relation between the regions we found is significantly associated with suppression ability. Future work will need to test the specific directional interpretations that we suggest are most consistent with previous literature.

Limitations and conclusion

One limitation of the present study is the fact that we did not confirm the exact strategy participants used to suppress retrieval. It is possible, for example, that some participants may have used a “thought substitution” strategy as opposed to a “direct suppression” strategy. However, our instructions to participants during the “No Think” condition were more consistent with direct suppression, and, based on previous findings (Benoit & Anderson, 2012), our right hemisphere-lateralized findings strongly suggest that direct suppression, and not thought substitution, was employed. Future studies should expand on the present findings by incorporating thought substitution as a third condition. It is also important to highlight that, while participants remembered more scenes from the T than NT condition, the T/NT task is unable to provide any behavioral data to confirm that participants completed the task in line with the instructions they were given. Therefore, we cannot rule out the possibility that some participants were more vigilant than others in following task instructions.

Another limitation of the present study is that, unlike previous T/NT studies, we found significantly greater forgetting of the “control” face-scene pairs (i.e., those that were trained, but never subsequently retrieved or suppressed) than for those in either the T or NT conditions. This is important because it represents a failure to replicate one standard behavioral effect previously reported in the T/NT literature. Further, as we did successfully replicate many of the neuroimaging results reported in this literature (e.g., increased rDLPFC and decreased hippocampal activation during the NT condition), this suggests that these behavioral and neuroimaging effects may be dissociable. When considering that all of our other results came out as expected based on previous studies, and that we made use of a previously validated T/NT task, we suggest that this dissociation is most likely attributable to the one specific difference between the present study and previous uses of this task. That is, unlike other T/NT studies, we instructed participants to perform the first T/NT trials outside the scanner prior to performing the final trial inside the scanner. This resulted in increased time between the three T/NT task repetitions outside the scanner and the final T/NT task repetition inside the scanner (approximately 1 hour), and therefore also led to a greater delay between encoding and testing of the control face-scene pairs (i.e., before and after T/NT task performance, respectively). This increase in time between encoding and testing of the control face-scene pairs likely contributed to greater forgetting in this condition specifically, which would also explain why our other behavioral and neuroimaging results came out as expected. We therefore suggest that, despite this increase in forgetting, the T/NT task was most plausibly performed appropriately, and therefore that our interpretations of our imaging results remain defensible. However, our findings highlight the need for future studies to further examine the effects of variations in the timing of T/NT task runs, training, and testing on levels of forgetting between task conditions. Future work should also address the possibility that the individual differences in inter-hemispheric coordination we observed interact with differences in the timing of T/NT runs.

It should also be highlighted that previous T/NT studies suggest that the number of NT intrusions is reduced across cycles (Anderson & Hanslmayr, 2014; Benoit et al., 2015; Depue et al., 2007). As such, the neuroimaging results we have captured from the final T/NT task repetition may reflect a relatively low level of intrusions from NT items (and perhaps relatively less retrieval effort for T items). Thus, the patterns of inter-hemispheric coordination we have identified might also reflect retrieval control strategies learned over the course of repeated task performance.

Finally, it is important to highlight that this study focused on negatively valenced memories, and did not examine memories of other valences. Previous studies comparing negative to neutral memories in this context have found mixed results, such that levels of forgetting for negative memories have been reported to be greater than (Depue et al., 2006; Lambert, Good, & Kirk, 2010), less than (Chen et al., 2012; Nørby, Lange, & Larsen, 2010), or equivalent to (Murray, Muscatell, & Kensinger, 2011; van Schie, Geraerts, & Anderson, 2013) levels of forgetting for neutral memories. Some neuroimaging work has also suggested that regions of the amygdala, insula, and cingulate gyrus may be more activated during suppression of emotional memories (Butler & James, 2010). It is therefore unclear whether the same results we observed would be found in a T/NT task using non-emotional memories, and future work should examine this further.

In conclusion, we found evidence of strong lateralized effects of goal-directed retrieval and retrieval suppression. We also found evidence that greater forgetting during goal-directed retrieval suppression (compared to during goal-directed retrieval) was associated with greater increases in functional connectivity during the NT condition between right and left precuneus and between right and left lateral parietal regions (that were activated during NT and T conditions, respectively). Finally, we demonstrate that rDLPFC may directly suppress the representation of visual imagery within visual cortex during retrieval suppression, and that greater increases in functional connectivity during retrieval suppression is correlated with later forgetting. In conjunction, these results suggest that individual differences in the ability to successfully suppress negative emotional memories may be partially explained by differences in the dynamic coordination ability of left and right hemisphere frontoparietal cognitive control networks. As current evidence suggests that deficits in this top-down suppression ability may play a role in psychiatric disorders involving unpleasant and intrusive memories (Catarino, Küpper, Werner-Seidler, Dalgleish, & Anderson, 2015), future research should address the possibility that failures in inter-hemispheric coordination might help explain such deficits.

Footnotes

Acknowledgments

For their help in task design and data collection, the authors would also like to acknowledge Zachary J. Schwab and Melissa R. Weiner.

Article Notes

Supplementary Material

Supplementary material is available for this article online.

References

Anderson

Green

(2001) Suppressing unwanted memories by executive control. Nature 410(6826): 366–369. Retrieved from http://doi.org/10.1038/35066572.

Anderson

Hanslmayr

(2014) Neural mechanisms of motivated forgetting. Trends in Cognitive Sciences 18(6): 279–282. Retrieved from http://doi.org/10.1016/j.tics.2014.03.002.

Anderson

Ochsner

Kuhl

Cooper

Robertson

Gabrieli

(2004) Neural systems underlying the suppression of unwanted memories. Science 303(5655): 232–235. Retrieved from http://doi.org/10.1126/science.1089504.

Benoit

Anderson

(2012) Opposing mechanisms support the voluntary forgetting of unwanted memories. Neuron 76(2): 450–460. Retrieved from http://doi.org/10.1016/j.neuron.2012.07.025.

Benoit

Hulbert

Huddleston

Anderson

(2015) Adaptive top-down suppression of hippocampal activity and the purging of intrusive memories from consciousness. Journal of Cognitive Neuroscience 27(1): 96–111. Retrieved from http://doi.org/10.1162/jocn_a_00696.

Bergström

de Fockert

Richardson-Klavehn

(2009) ERP and behavioural evidence for direct suppression of unwanted memories. NeuroImage 48(4): 726–737. Retrieved from http://doi.org/10.1016/j.neuroimage.2009.06.051.

Butler

James

(2010) The neural correlates of attempting to suppress negative versus neutral memories. Cognitive, Affective, & Behavioral Neuroscience 10(2): 182–194. Retrieved from http://doi.org/10.3758/CABN.10.2.182.

Catarino

Küpper

Werner-Seidler

Dalgleish

Anderson

(2015) Failing to forget: Inhibitory-control deficits compromise memory suppression in posttraumatic stress disorder. Psychological Science 26(5): 604–16. Retrieved from http://doi.org/10.1177/0956797615569889.

Cavanna

Trimble

(2006) The precuneus: A review of its functional anatomy and behavioural correlates. Brain 129(Pt 3): 564–583. Retrieved from http://doi.org/10.1093/brain/awl004.

10.

Chen

Liu

Huang

Cheng

Luo

Y.-J.

(2012) Suppression of aversive memories associates with changes in early and late stages of neurocognitive processing. Neuropsychologia 50(12): 2839–2848. Retrieved from http://doi.org/10.1016/j.neuropsychologia.2012.08.004.

11.

Dehaene

(2014) Consciousness and the Brain, New York, NY: Viking Press.

12.

Depue

(2012) A neuroanatomical model of prefrontal inhibitory modulation of memory retrieval. Neuroscience and Biobehavioral Reviews 36(5): 1382–1399. Retrieved from http://doi.org/10.1016/j.neubiorev.2012.02.012.

13.

Depue

Banich

Curran

(2006) Suppression of emotional and nonemotional content in memory: Effects of repetition on cognitive control. Psychological Science 17(5): 441–447. Retrieved from http://doi.org/10.1111/j.1467-9280.2006.01725.x.

14.

Depue

Curran

Banich

(2007) Prefrontal regions orchestrate suppression of emotional memories via a two-phase process. Science 317(5835): 215–219. Retrieved from http://doi.org/10.1126/science.1139560.

15.

Ellard

Barlow

Whitfield-Gabrieli

Gabrieli

Deckersbach

(2017) Neural correlates of emotion acceptance vs worry or suppression in generalized anxiety disorder. Social Cognitive and Affective Neuroscience 3(1): 2–17. Retrieved from http://doi.org/10.1093/scan/nsx025.

16.

Gagnepain

Henson

Anderson

(2014) Suppressing unwanted memories reduces their unconscious influence via targeted cortical inhibition. Proceedings of the National Academy of Sciences of the United States of America 111(13): E1310–E1319. Retrieved from http://doi.org/10.1073/pnas.1311468111.

17.

Gitelman

Penny

Ashburner

Friston

(2003) Modeling regional and psychophysiologic interactions in fMRI: The importance of hemodynamic deconvolution. NeuroImage 19(1): 200–207. Retrieved from http://doi.org/10.1016/S1053-8119(03)00058-2.

18.

Keller

Hedden

Thompson

Anteraper

S. A.

Gabrieli

Whitfield-Gabrieli

(2015) Resting-state anticorrelations between medial and lateral prefrontal cortex: Association with working memory, aging, and individual differences. Cortex 64: 271–280. Retrieved from http://doi.org/10.1016/j.cortex.2014.12.001.

19.

Kim

D.-J.

(2013) Out of mind, out of sight: Perceptual consequences of memory suppression. Psychological Science 24(4): 569–574. Retrieved from http://doi.org/10.1177/0956797612457577.

20.

Kuhl

Dudukovic

Kahn

Wagner

(2007) Decreased demands on cognitive control reveal the neural processing benefits of forgetting. Nature Neuroscience 10(7): 908–914. Retrieved from http://doi.org/10.1038/nn1918.

21.

Lambert

Good

Kirk

(2010) Testing the repression hypothesis: Effects of emotional valence on memory suppression in the think—No think task. Consciousness and Cognition 19(1): 281–293. Retrieved from http://doi.org/10.1016/j.concog.2009.09.004.

22.

Lang

Bradley

Cuthbert

(2008) International affective picture system (IAPS): Affective ratings of pictures and instruction manual (Technical Report A-8), Gainesville, FL: University of Florida.

23.

Levy

Anderson

(2012) Purging of memories from conscious awareness tracked in the human brain. The Journal of Neuroscience 32(47): 16785–1694. Retrieved from http://doi.org/10.1523/JNEUROSCI.2640-12.2012.

24.

Lundstrom

Peterssona

Anderssona

Johanssone

Franssona

Ingvar

(2003) Isolating the retrieval of imagined pictures during episodic memory: Activation of the left precuneus and left prefrontal cortex. NeuroImage 20(4): 1934–1943. Retrieved from http://doi.org/10.1016/j.neuroimage.2003.07.017.

25.

McLaren

Ries

Johnson

(2012) A generalized form of context-dependent psychophysiological interactions (gPPI): A comparison to standard approaches. NeuroImage 61(4): 1277–1286. Retrieved from http://doi.org/10.1016/j.neuroimage.2012.03.068.

26.

Murray

Muscatell

Kensinger

(2011) Effects of emotion and age on performance during a think/no-think memory task. Psychology and Aging 26(4): 940–955. Retrieved from http://doi.org/10.1037/a0023214.

27.

Nørby

Lange

Larsen

(2010) Forgetting to forget: On the duration of voluntary suppression of neutral and emotional memories. Acta Psychologica 133(1): 73–80. Retrieved from http://doi.org/10.1016/j.actpsy.2009.10.002.

28.

Paz-Alonso

Bunge

Anderson

Ghetti

(2013) Strength of coupling within a mnemonic control network differentiates those who can and cannot suppress memory retrieval. The Journal of Neuroscience 33(11): 5017–5026. Retrieved from http://doi.org/10.1523/JNEUROSCI.3459-12.2013.

29.

Rugg

Vilberg

(2013) Brain networks underlying episodic memory retrieval. Current Opinion in Neurobiology 23(2): 255–260. Retrieved from http://doi.org/10.1016/j.conb.2012.11.005.

30.

Smith

Alkozei

Lane

Killgore

(2016) Unwanted reminders: The effects of emotional memory suppression on subsequent neuro-cognitive processing. Consciousness and Cognition 44: 103–113. Retrieved from http://doi.org/10.1016/j.concog.2016.07.008.

31.

van Schie

Geraerts

Anderson

(2013) Emotional and non-emotional memories are suppressible under direct suppression instructions. Cognition & Emotion 27(6): 1122–1131. Retrieved from http://doi.org/10.1080/02699931.2013.765387.

32.

Waldhauser

G. T.

Bäuml

K.-H. T.

Hanslmayr

(2015) Brain oscillations mediate successful suppression of unwanted memories. Cerebral Cortex 25(11): 4180–4190. Retrieved from http://doi.org/10.1093/cercor/bhu138.

33.

Whitfield-Gabrieli

Nieto-Castanon

(2012) Conn: A functional connectivity toolbox for correlated and anticorrelated brain networks. Brain Connectivity 2(3): 125–141. Retrieved from http://doi.org/10.1089/brain.2012.0073.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.11 MB