Controlling our thoughts is central to mental well-being, and its failure is at the crux of a number of mental disorders. Paradoxically, behavioral evidence shows that thought suppression often fails. Despite the broad importance of understanding the mechanisms of thought control, little is known about the fate of neural representations of suppressed thoughts. Using fMRI, we investigated the brain areas involved in controlling visual thoughts and tracked suppressed thought representations using multivoxel pattern analysis. Participants were asked to either visualize a vegetable/fruit or suppress any visual thoughts about those objects. Surprisingly, the content (object identity) of successfully suppressed thoughts was still decodable in visual areas with algorithms trained on imagery. This suggests that visual representations of suppressed thoughts are still present despite reports that they are not. Thought generation was associated with the left hemisphere, and thought suppression was associated with right hemisphere engagement. Furthermore, general linear model analyses showed that subjective success in thought suppression was correlated with engagement of executive areas, whereas thought-suppression failure was associated with engagement of visual and memory-related areas. These results suggest that the content of suppressed thoughts exists hidden from awareness, seemingly without an individual's knowledge, providing a compelling reason why thought suppression is so ineffective. These data inform models of unconscious thought production and could be used to develop new treatment approaches to disorders involving maladaptive thoughts.
Selecting and controlling the contents of one's thoughts is regarded as paramount for achieving goals, learning, controlling emotions, and for psychological well-being (Østefjells et al., 2017; Shapiro & Astin, 1998; Carver & Scheier, 1982). However, unwanted thoughts can haunt us, leading to outcomes that range from mild discomfort, in the case of thinking about that expensive repair the house needs, to debilitating mental disorders, such as revisualizing violent scenes from the battlefield, in the case of posttraumatic stress disorder (Stander, Thomsen, & Highfill-McRoy, 2014).
Significant effort has been made in characterizing the behavioral mechanisms of thought control. Seminal work from Wegner et al. has shown that, ironically, one's attempts to suppress thoughts lead to a rebound in the occurrence of the same thought after the suppression effort (Wegner, 1994), coined the “ironic process theory.”
Brain imaging has shed light on the areas responsible for thought control. Using a task in which participants had to suppress an object or think freely about anything including the object (Mitchell et al., 2007), sustained suppression was linked to activations in dorsolateral pFC, whereas transient activation in bilateral regions of ACC was observed during the emergence of suppressed thoughts. ACC has also been shown to be more active during sustained thought suppression compared with free thought (Wyland, Kelley, Macrae, Gordon, & Heatherton, 2003). In more recent work, suppression and maintenance of visual thoughts was linked to networks encompassing right-lateralized frontolateral areas for suppression (Aso et al., 2016).
In a recent behavioral study, we investigated the sensory traces of suppressed visual thoughts using psychophysics (Kwok, Leys, Koenig-Robert, & Pearson, 2019). We discovered that the visual representations of suppressed thoughts still led to perceptual priming, although the participants reported successful thought suppression. This intriguing result suggests that suppressed visual thoughts might still exist in visual brain areas despite the judgments of successful suppression. In other words, nonconscious visual representations of the thought content might exist in the visual cortex, without the individual ever knowing. Whereas previous work has studied the areas responsible for thought suppression, the goal of this study was to investigate the possible existence of nonconscious thought representations during successful thought suppression.
To investigate the fate of suppressed visual representations, we used a paradigm comparing imagery to suppression. In each trial, participants were instructed to either imagine or suppress the visual thought of a fruit or a vegetable (red apple or green broccoli). This design was adapted from a previous behavioral study in which we used conveniently green- or red-colored vegetables and fruits as primes for red/green binocular rivaling gratings (Kwok et al., 2019). Participants reported visual thought intrusions, which we called suppression breaks, when the to-be-suppressed objects appeared in their minds. Importantly, we instructed participants not to replace the visual thoughts of the suppressed item with other visual content, as we did in our previous psychophysics study (Kwok et al., 2019; see Methods section for details). We thus have continued with the term “thought suppression” as we did in our previous study, which means avoid thinking about the item, while not replacing that visual thought with another one or distracting oneself.
As a first goal, we sought to identify the areas engaged by suppression and imagery as well as the areas engaged by successful suppression and suppression breaks. To identify these areas, we used mass-univariate general linear model (GLM) contrasts between conditions as well as multivoxel pattern analysis (MVPA) to discriminate between tasks in visual ROIs from V1 to V4. Our second aim was to investigate whether successfully suppressed visual and imagined representations could still be decoded despite participants reporting subjective absence of these visual thoughts. To test this, we used MVPA in a cross-decoding generalization design to decode the content of the suppressed thought (apple or broccoli). The generalization design allowed us to reveal whether and where there was representational overlap between perceptual images and successfully suppressed objects as well as between imagined objects and successfully suppressed ones.
Our results show that suppression recruits a right-lateralized network, including pFC and ACC as well as the superior parietal and temporal cortices. Failure at suppressing visual thoughts (suppression breaks) was associated with enhanced recruitment of memory and higher-level visual areas, whereas successful suppression was correlated with middle prefrontal and insula activations. These results are consistent with previous studies in thought suppression using different paradigms, thus highlighting the reproducibility and generalization of these effects. Finally, our decoding analysis revealed that, in a similar vein to the behavioral work, although participants judged the suppression to be successful, visual imagery-like representations were present in the lateral occipital cortex (LOC). These results provide new neural evidence of the pervasiveness of suppressed thoughts and unveil a network of brain areas to be targeted to treat intrusive thought disorders.
Experimental procedures were approved by the University of New South Wales human research ethics committee (HREC No. HC12030). All methods in this study were performed in accordance with the guidelines and regulations from the Australian National Statement on Ethical Conduct in Human Research (www.nhmrc.gov.au/guidelines-publications/e72). All participants gave informed written consent to participate in the experiment. We tested 15 participants (four women) aged 29.6 ± 1.4 (mean ± SEM) years. For analyses discriminating successful from failed suppression trials, a subset of eight participants (two women, aged 32 ± 2.1 years) was considered using a selection criterion of having at least 25% of failed or successful suppression trials. In other words, participants with failed/successful trial ratios more skewed than 1:3 were discarded.
We selected the sample size based on previous studies (Koenig-Robert & Pearson, 2019; Aso et al., 2016; Soon, Brass, Heinze, & Haynes, 2008) to meet standard criteria of statistical power. We conducted a post hoc power analyses to ascertain the power achieved using G*Power (Faul, Erdfelder, Buchner, & Lang, 2009). For GLM analyses, the power achieved was at least 0.91 at the voxel level (n = 15). For the decoding analysis discriminating the content of successful suppression, we achieved a power of at least 0.79 to detect differences in the paired t test for relevant conditions (n = 8).
Functional and Structural MRI Parameters
Scans were performed at the Neuroscience Research Australia facility, Sydney, Australia, in a Philips 3-T Achieva TX MRI scanner using a 32-channel head coil. Structural images were acquired using turbo field echo sequence consisting of 256 T1-weighted sagittal slices covering the whole brain (flip angle = 8°, matrix size = 256 × 256, voxel size = 1 mm isotropic). Functional T2*-weighted images were acquired using EPI sequence with 31 slices (flip angle = 90°, matrix size = 240 × 240, voxel size = 3 mm isotropic, repetition time = 2000 msec, echo time = 40 msec).
We adapted the behavioral task from a previous study from our group (Kwok et al., 2019) to satisfy fMRI requirements. We instructed participants to either imagine or suppress (avoid imagining) the visual thought of either a red apple or a green broccoli (see Figure 1A). Each trial started with a written object cue reading “green broccoli” or “red apple” for 2 sec. After this, a task cue was shown, reading either “imagine” or “suppress” for 2 sec. Participants were instructed to either visualize (imagery period) as vividly as they could the cued object in the imagine condition or avoid thinking about the cued object (suppression period) in the suppress condition for 12 sec. Importantly, we extensively instructed participants not to use object substitution (imagining another item to avoid imagining the cued object) as we did in our previous behavioral study (Kwok et al., 2019), where we tested both suppression and substitution strategies. We previously observed that object substitution led to very different outcomes compared to object suppression (i.e., substitution led to better thought control). A fixation point was shown on the screen, and participants were required to fixate. In the suppression condition, participants were instructed to press a button as soon as they detected that the visual thought of the to-be-suppressed object appeared in their minds. We labeled such events as “suppression breaks,” and the trial was labeled as a failed suppression trial. The suppression break button could be pressed multiple times within the 12 sec (Supplementary Figure S41), thus representing multiple suppression breaks. After the imagery/suppression period, a prompt asking to rate vividness (from 1 to 4, with 4 = strongest vividness) was presented after imagery or failed suppressed trials. Participants responded by pressing one of the four buttons on two response boxes. No vividness question was shown after successful suppression trials, which were automatically labeled as having vividness = 0. In failed suppressed trials, whenever multiple suppression breaks were reported, we instructed participants to rate the highest vividness suppression break event. After reporting the vividness of the visual thought (if required), an intertrial interval of 10 sec was observed during which the word “rest” appeared on the screen. In each run of 5 min, three trials of each type (imagine/suppress apple/broccoli) were tested, yielding 12 trials. Trials were pseudorandomized within a run.
We presented flickering natural images of a broccoli and an apple against a black background at 4.167 Hz at three different perceptual intensities (40%, 60%, and 80% transparency) to maximize subsequent classifier generalization ability (Bannert & Bartels, 2013). Natural images of a broccoli and an apple against a black background were retrieved on Google image search for images labeled for reuse with modification and were presented inside a rectangle (the same that was used in the imagery/suppression task; Figure 1) including a fixation point at the center. Within a run of 3 min, we presented the flickering images in a block manner, interleaved with fixation periods of 15 sec each (apple: 15 sec, rest: 15 sec, broccoli: 15 sec, rest: 15 sec, etc.). Importantly, an attention task was performed consisting of detecting a change in fixation point brightness (+70% for 200 msec). Fixation changes were allocated randomly during a run, from one to four instances. Participants were instructed to press any of the four buttons as soon as they detected the changes. Participants showed high performance in the detection task (d′ = 2.89 ± 0.15 SEM).
Functional Mapping of Retinotopic Visual Areas
To functionally determine the boundaries of visual areas from V1 to V4 independently for each participant, we used the phase-encoding method (Warnking et al., 2002; Sereno et al., 1995). Double wedges containing dynamic colored patterns cycled through 10 rotations in 10 min (retinotopic stimulation frequency = 0.033 Hz). To ensure deployment of attention to the stimulus during the mapping, participants performed a detection task: pressing a button upon seeing a gray dot anywhere on the wedges.
We performed the three experiments in a single scanning session lasting about 1.5 hr. Stimuli were delivered using an 18-in. MRI-compatible LCD screen (Philips ERD-2, 60-Hz refresh rate) located at the end of the bore. Participants held one dual-button response box in each hand (Lumina, Cedrus) that was used to record all responses. All stimuli were delivered, and responses were gathered employing the Psychtoolbox 3 (Brainard, 1997; Pelli, 1997) for MATLAB (The MathWorks Inc.) using in-house scripts. Participants' heads were restrained using foam pads and adhesive tape. Each session followed the same structure: first, the structural scanning followed by the retinotopic mapping (10 min). Then, the perception task was alternated with the imagery/suppression task until completing three runs of the perception task (3 min per run). Then, the imagery/suppression task was repeated until completing eight runs in total (5 min per run). Pauses were assigned in between the runs. The first four volumes of each functional runs were discarded to account for the equilibrium magnetization time, and each functional run started with 10 sec of fixation.
We interleaved perception and imagery/suppression blocks to remove order effects and to increase participants' engagement. For the same reason, we opted against organizing suppression and imagery trials in different blocks. Pseudorandomizing imagery and imagery trials within a block ensured that participants would not ignore the cue in suppression trials, because the task is only revealed after the cue. In addition, by randomizing the order of suppression and imagery trials, we avoided putting suppression trials together, which revealed to be exhausting based on postexperiment interviews from our previous study.
Visual ROI Functional Definition
fMRI retinotopic mapping data were analyzed using the fast Fourier transform (FFT) in MATLAB. The FFT was applied voxel-wise across time points. The complex output of the FFT contained both the amplitude and phase information of sinusoidal components of the BOLD signal. Phase information at the frequency of stimulation (0.033 Hz) was then extracted, using its amplitude as threshold (≥ 2 signal-to-noise ratio), and overlaid them on each participant's cortical surface reconstruction obtained using FreeSurfer (Fischl et al., 2004; Fischl, Sereno, & Dale, 1999). We manually delineated boundaries between retinotopic areas on the flattened surface around the occipital pole by identifying voxels showing phase reversals in the polar angle map, representing the horizontal and vertical visual meridians. In all participants, we clearly defined four distinct visual areas labeled V1, V2, V3 (specifically its ventral part known as VP), and V4. All four retinotopic labels were then defined as the intersection with the perceptual blocks (broccoli/apple > fixation, p < .001, false discovery rate [FDR] corrected), thus restricting the ROI to the foveal and parafoveal (∼5.5° of visual angle) representations of each visual area.
Suppressed Object Information Containing ROI Definition
We used a decoding approach to define ROIs bearing information about the content (apple vs. broccoli) of suppressed trials. We used these ROIs to test whether similar representational content was shared between imagery and successful suppression (Figure 4) and between perception and successful suppression (Supplementary Figure S5). To define these ROIs, regressors for apple and broccoli were extracted from every run from the suppression trials (independent of success) using 12-sec boxcars (locked to the beginning of the suppression period). We used a leave-one-run-out cross-validation scheme (see MVPA section for details) and a searchlight approach (3 voxels of radius). ROIs containing information about the contents of suppression were defined as those reaching a classification accuracy of Z > 2 (one-sample t test against chance: 50%) at the voxel level. We then corrected for multiple comparisons using cluster-extent based thresholding employing Gaussian random field theory at p < .05. Only two ROIs satisfied these statistical criteria: one in the inferior frontal gyrus (IFG) and one in the LOC (Figure 4). Importantly, this ROI definition is orthogonal to the target cross-decoding analysis where we tested the mutual informational content between imagery and successful suppression, as the training sets of both analyses are independent from each other: suppression trials for the ROI definition and imagery trials for the cross-decoding analysis.
Thus, LI = −1 represents fully left-lateralized effects; whereas LI = +1, fully right-lateralized effects.
fMRI Signal Preprocessing
All data were analyzed using SPM12 (Wellcome Trust Centre for Neuroimaging). We realigned functional images to the first functional volume and high-pass filtered (128 sec) to remove low-frequency drifts in the signal.
Imagery versus Suppression GLM Analysis
Data were spatially normalized into the Montreal Neurological Institute template and spatially smoothed using an FWHM 8-mm 3-D Gaussian kernel. We generated regressors for each condition (imagery and suppression, independent of the imagined/suppressed object) for each run independently. We used boxcar functions of 12 sec, time-locked to the beginning of the imagery or suppression periods, to model each trial with the canonical hemodynamic response function as basis function. Vividness of the imagery and suppression trials was modeled using parametric modulators as it has been shown that brain activations are dependent on the vividness of imagery (Dijkstra, Bosch, & van Gerven, 2017). GLMs were used to test differences between imagery and suppression conditions. Participants' (n = 15) estimates (betas) of the mass-univariate GLM were fed into a second-level two-sample t-test analysis.
Successful versus Failed Suppression GLM Analysis
The analysis was performed as described in the previous paragraph except for the following differences. We generated regressors for successful and failed suppression (independent of the suppressed object) for each run independently. We used boxcar functions of 12 sec (locked to the beginning of the suppression period) and 1 sec (time-locked to the suppression break events) to model successful suppression and suppression breaks, respectively, to capture their respective sustained and transient natures (Mitchell et al., 2007). Only participants having at least 25% of successful or failed suppression trials (n = 8) were considered to have enough data to estimate the parameters.
GLM Analysis for MVPA
Data were analyzed in their native space, without spatial normalization and smoothing to avoid disrupting information contained in the spatial patterns of activation (Hebart, Görgen, & Haynes, 2015). For the task decoding (Figures 2C and 3B), we estimated GLM for imagery versus suppression and successful versus failed suppression as described above. For the content decoding (Figure 4), regressors for apple and broccoli were estimated using boxcar functions (15 sec for the perception trials and 12 sec for the imagery and successful suppression conditions). This analysis was performed on the subset of participants having at least 25% of successful or failed suppression trials (n = 8).
We used a well-established decoding approach to extract information related to each grating contained in the pattern of activation across voxels of a given participant using the The Decoding Toolbox (Hebart et al., 2015). For the task decoding (Figures 2C and 3B), we used a leave-one-run-out cross-validation scheme. We trained a linear supporting vector machine on all runs except one and then tested on the remaining one. We repeated this procedure until all runs were used as test and then averaged the results across validations (eightfold). Using this approach, we tested whether information about the task nature could be decoded from functionally defined visual areas (from V1 to V4; see Visual ROI Functional Definition section for details). For the content decoding analysis (Figure 4), we employed cross-classification to generalize information between imagery and the successful suppression trials. We thus trained on the ensemble of the imagery runs and tested on the ensemble of the successful suppression trials. The same was done for the perception-successful suppression analysis (Supplementary Figure S5) where the ensemble of perception runs was used for training. No cross-validation was used here as the data sets were independent; thus, there was no risk of overfitting. We employed an ROI to test common representational content in functionally defined areas as containing suppressed object information (see Suppressed Object Information Containing ROI Definition section for details). Decoding accuracies were averaged across runs and tested against chance level (50%) using a one-sample t test across participants.
Statistical Analysis on Brain Images
All second-level (across participants) brain statistical images (derived from SPM or decoding) were subjected to a threshold at the voxel level p < .001, as recommended in previous studies (Woo, Krishnan, & Wager, 2014). We then corrected for multiple comparisons using cluster-extent based thresholding employing Gaussian random field theory (Worsley et al., 1996; Friston, Worsley, Frackowiak, Mazziotta, & Evans, 1994) at p < .05, as implemented in FMRIB Software Library (Smith et al., 2004). Importantly, these thresholds have been shown to be valid within the nominal false-positive ratios (Eklund, Nichols, & Knutsson, 2016).
Participants were instructed to either imagine or suppress visual thoughts of an apple or broccoli (Figure 1A; see Methods section for details). Importantly, participants subjectively rated the strength of the visual thought for both imagined and suppressed trials with suppression breaks (Supplementary Figure S1). Successful suppression trials outnumbered failed trials by about 1:3, with some trials showing multiple suppression break instances (up to five; Supplementary Figure S2). Vividness was correlated with the number of suppression breaks (adjusted R2 = .83, p = 1.45 × 10−6; Supplementary Figure S3) when considering all suppression trials. However, this effect vanished when only failed suppression trials were considered (adjusted R2 = −.03, p = .45; Supplementary Figure S3), suggesting that the correlation was driven by the successful suppression trials where vividness is 0. Suppression vividness ratings (0.85 ± 0.17 [mean ± SEM]) were largely variable among participants, with some participants having very little suppression breaks (e.g., Participants 4, 6, 9, and 14; Figure 1B), whereas others reported a large number of suppression breaks (e.g., Participants 1, 3, and 12). This was in contrast with the imagery vividness ratings (2.78 ± 0.09 [mean ± SEM]) that were more homogeneous across participants (Figure 1C). This is consistent with the individual differences in thought control revealed in our previous study (Kwok et al., 2019). For analyses exploring suppression success and/or failure (Figures 3 and 4), only participants having at least 25% of suppression breaks or successful suppression (i.e., no more of 1:3 ratio in either sense) were considered, which corresponded to eight participants (marked with a ★).
Imagery and Suppression Engage Differently Lateralized Networks
A GLM analysis (see Methods section for details) revealed a left-lateralized network associated with imagery production and a right-lateralized network associated with suppression (Figure 2A). This is consistent with previous neuroimaging and lesion studies (Aso et al., 2016; Garavan, Ross, & Stein, 1999; D'Esposito et al., 1997; Farah, 1984).
Brain areas activated by both imagery (on the left hemisphere) and suppression (on the right hemisphere) were the superior frontal, prefrontal, and temporal cortices (Figure 2A). This is consistent with results highlighting the role of the right pFC in inhibitory control (Garavan, Ross, Murphy, Roche, & Stein, 2002; for reviews, see Levy & Anderson, 2002; Duncan & Owen, 2000; however, see Aron, Robbins, & Poldrack, 2014, for a more nuanced role of pFC). On the other hand, the left prefrontal and frontal cortices have been associated with the production of imagery content in different sensory modalities (Sabatinelli, Lang, Bradley, & Flaisch, 2006; Lundstrom et al., 2003; Yoo, Freeman, McCarthy, & Jolesz, 2003).
We found brain areas specifically associated with imagery in the inferior temporal cortex bilaterally (Figure 2A). These areas are known to code for high-level object representations (Rust & DiCarlo, 2010; Grill-Spector, 2003); thus, visualization of real-life objects (i.e., apple and broccoli) is expected to engage these areas. On the other hand, suppression specifically engaged right superior temporal areas engaged by a variety of tasks, such as face processing, semantic processing, and audiovisual integration (for a review, see Hein & Knight, 2008), and also superior parietal areas recruited in tasks including verbal/semantic processing, memory retrieval, and conflict resolution (for a review, see Seghier, 2013). In addition, suppression was associated with activations in the right ACC, an area involved in inhibitory control, presumably by detecting and signaling error (Kolling et al., 2016; Botvinick, Cohen, & Carter, 2004; Matsumoto & Tanaka, 2004; Bush, Luu, & Posner, 2000), such as suppression failure in our paradigm.
A formal analysis of lateralization (Figure 2B; see Methods section for details) of imagery and suppression across hemispheres revealed that imagery activations were significantly left lateralized (LI = −0.3539, p = .0031, one-sample, two-tailed t test against 0, t(14) = −3.5701, 95% CI [0.1423, 0.5706], uncorrected). On the other hand, suppression-driven activations were significantly right lateralized (LI = 0.3565, p = .0204, one-sample, two-tailed t test against 0, t(14) = 2.6146, 95% CI [−0.6441, −0.0636], uncorrected).
Differences between imagery and failed thought suppression were found in memory, executive, and higher-level visual areas (Supplementary Figure S4). Imagery-related activations were only found in the right hemisphere (inferior temporal and parahippocampal cortices). This is consistent with specialization for low-level representations in the left inferior temporal and categorical representations in the right one (Meng, Cherian, Singal, & Sinha, 2012; see Discussion section for details). On the other hand, suppression failure showed noticeably less lateralization compared to the imagery versus suppression contrast. Suppression fail activations were found in both hemispheres (inferior parietal and frontal areas) except for the right ACC.
Imagery and Suppression Can Be Discriminated Using Spatial Patterns of Activations in Visual Areas
We then investigated whether retinotopically organized visual areas contain information that can be used to discriminate between imagery and suppression. Our previous psychophysics study comparing imagery and thought suppression suggested that thought suppression leaves a visual trace that is likely to follow some retinotopic organization (Kwok et al., 2019). We thus tested whether differences in imagery and suppression appear early in the visual hierarchy. Indeed, although no significant differences in activations in lower-level visual areas (V1–V4) were detected using GLM univariate analysis (Figure 2A), the informational content in these areas revealed by MVPA allowed us to discriminate imagery and suppression trials with great accuracy (decoding accuracy > 80%, p = 10−6, one-sample t test against 50%, FDR corrected; Figure 2C). MVPA takes into account the spatial pattern of activation associated with imagery and suppression to discriminate between these two tasks rather than relying on overall differences in BOLD response. Therefore, this analysis is more sensitive and better suited to reveal subtle differences that are beyond the sensitivity of univariate analyses (Kriegeskorte, Goebel, & Bandettini, 2006). This result shows that the spatial patterns of activations in visual areas differ consistently between imagery and suppression, although their average activation level does not differ significantly.
Failed and Successful Suppression Are Associated with Posterior versus Anterior Activations
We employed a GLM analysis to compare activations in failed versus successful suppression trials. This analysis revealed that successful suppression was associated with bilateral activations in the medial pFC, ACC, and the insula (Figure 3A). The engagement of these areas is consistent with their involvement in control and monitoring. The medial pFC has been implicated in inhibitory control (Aron et al., 2014), whereas the insula is implicated in conflict monitoring (Gehring & Knight, 2000; Carter et al., 1998; see Botvinick et al., 2004, for a review).
On the other hand, failed suppression was associated with activations in visual areas: the middle occipital cortex and the medial temporal cortex. Interestingly, we also found bilateral activations in memory-related areas such as the parahippocampal cortex and the hippocampus. The hyperactivity in visual and memory areas during suppression breaks is consistent with the access and generation of visual sensation (Dijkstra et al., 2017; Ishai, Ungerleider, & Haxby, 2000) as suppression breaks can be conceptualized as a case of involuntary imagery (Pearson, 2019; Pearson & Westbrook, 2015).
To test whether information about the success at suppressing thoughts is present early in the visual hierarchy, as our previous study suggested (Kwok et al., 2019), we used decoding in functionally defined visual ROIs from V1 to V4. Using an MVPA, we thus found that visual areas, from V1 to V4, contain information about the success or failure in thought suppression (Figure 3B). Classifiers decoded moderately well the spatial activation patterns associated with failed and successful suppression (p < .005, one-sample t test against 50%, FDR corrected, q = 0.05), with the highest accuracy found in Area V3 (decoding accuracy: 80.5%). This result shows that reliable information about the success and failure of suppression can be found in lower visual areas.
Decoding the Contents of Successful Thought Suppression Using Imagery Information
To investigate the fate of suppressed thoughts, we used a cross-decoding generalization approach to track successfully suppressed representations. To do so, we trained classifiers to discriminate between the content (green broccoli or red apple) in imagery trials from the main experiment. We then tested supporting vector machine classifiers on successful suppression trials. (In this manner, significant above-chance decoding accuracy would show that imagery representations are generated although participants rate their thought suppression as successful.) We applied this analysis on two independently predefined ROIs, where the contents of suppression (independent of success) were more readily decodable: the left IFG and the right LOC (see Methods section for details).
We found significant decoding of successfully suppressed thought content only in the right LOC. Imagery successful suppression cross-decoding accuracy reached 65.6% in the LOC (one-tailed t test, t(7) = 2.53, 95% CI [50.2, inf], p = .0443, FDR-corrected q = .05). This result suggests that imagery-like representations of the suppressed stimuli in visual areas are still present despite the subjective success at suppressing them.
We also performed a cross-decoding analysis using visual perception information from visual perceptual blocks (see Methods section for details) to decode the contents of successful suppression in the same ROI, as previously shown. These results, however, did not survive multicomparison correction (Supplementary Figure S5).
We further sought to investigate whether perception and imagery patterns would also generalize the representational content of successful suppression in visual ROI (from V1 to V4). We, however, did not find significant above-chance decoding in functionally defined visual areas (Supplementary Figure S6).
Suppressed Thoughts Are Still There
Using MVPA, we detected imagery-like neural representations in the LOC despite subjective reports that the thoughts were suppressed from awareness. A number of previous results have shown that suppressed thoughts influence behavior despite subjective success at censoring them. Classic studies (Wegner, 1994; Wegner, Schneider, Carter, & White, 1987) have shown that suppressing thoughts leads to a rebound of thought intrusions after withdrawing suppression efforts, pointing to the pervasiveness of suppressed thoughts. In a recent experiment, we objectively measured the strength of suppressed visual thoughts as a bias on subsequent binocular rivalry (Kwok et al., 2019). Our results revealed that the perceptual trace of successfully suppressed thoughts was as effective as voluntary imagery in biasing subsequent binocular rivalry. The current results add to these behavioral findings by showing that neural representations of suppressed visual thoughts are still present in the visual cortex. This finding sheds light on how successfully suppressed thoughts can interact with perception and affect behavior, despite the feeling that these thoughts have been successfully suppressed out of mind.
It is important to note that the presence of decodable information in the brain does not necessarily mean that that information is used by the brain (Ritchie, Kaplan, & Klein, 2019). We, however, believe that this piece of evidence can help in identifying the neural underpinnings of thought suppression and its failure as a method to control thoughts. In addition, the limited number of participants included in the successful suppression decoding experiment (n = 8) should be noted, and future work should expand on this.
The Nature of the Suppressed Thought Representations
Suppressed thought representations were found to be imagery-like representations in the right LOC, which is known to house visual object representations (Pourtois, Schwartz, Spiridon, Martuzzi, & Vuilleumier, 2009; Grill-Spector, Kourtzi, & Kanwisher, 2001). Interestingly, imagery representations have been shown to generalize nicely to perceptual representations in the LOC (Cichy, Heinzle, & Haynes, 2012), thus suggesting that the information we found in the LOC might also be perceptual in nature. Why did suppressed representations not generalize to visual perception (Supplementary Figure S5)? The most parsimonious explanation is that perceptual visual representations and (nonconscious) successfully suppressed ones differ both anatomically and functionally. Another explanation is that, although there could have been representational overlap between perception and suppressed thoughts, classifiers were unsuccessful in cross-decoding because of the difference in signal strength and thus pattern reliability across both modalities (Sterzer, Haynes, & Rees, 2008).
Interestingly, classifiers using information contained in retinotopically organized visual areas from V1 to V4 were able to discriminate with great accuracy between imagery versus suppression and between successful and failed suppression. These results indicate that lower visual areas house reliable information about these cognitive states, although we did not find differences in the global level of activation as shown in the GLM analyses, thus supporting previous psychophysics results suggesting that retinotopically organized visual areas play a role in visual thought suppression (Kwok et al., 2019).
Awareness of the Suppressed Thoughts
Although participants reported not having visual sensation associated with the suppressed objects during the successfully suppressed trials, we cannot, however, rule out the notion that participants may have failed to report visual intrusions of suppressed thoughts. Interestingly, our previous behavioral study strongly suggests that participants report intrusions faithfully, as shown by a series of control experiments in a similar suppression/imagery paradigm (Kwok et al., 2019).
Importantly, subjective criteria about the existence or absence of thought intrusions are of clinical relevance to ascertain symptoms associated with psychopathology as well as treatment success (Holmes & Mathews, 2010; Ehlers & Clark, 2000; Reynolds & Brewin, 1998). We thus believe that subjective reports about thought suppression are a legitimate way to measure them and represent a suitable way to relate the present results with clinical research.
A Network of Areas as Potential Targets for Therapy of Intrusive Thoughts
The working assumption in our paradigm is that voluntary visual imagery and visual thought suppression are related processes (Kwok et al., 2019; Pearson, 2019; Pearson & Westbrook, 2015). Whereas voluntary visual imagery involves the maintenance of a visual object in one's awareness, visual thought suppression strives to prevent the selected object from entering awareness. Importantly, failure of thought suppression shares a common phenomenology with voluntary imagery, as conscious visual content is present in both cases.
By comparing imagery to suppression, we were able to identify areas involved in voluntary imagery generation versus visual thought suppression. This analysis revealed two lateralized networks—a left-lateralized network associated with imagery and a right-lateralized network associated with suppression—consistent with previous results (Aso et al., 2016; Garavan et al., 1999; D'Esposito et al., 1997; Farah, 1984). These networks included prefrontal, frontal, and parietal areas. A body of literature has linked activations in the right pFC with several forms of inhibitory control, such as error detection, correction, and inhibition implementation (Aron, Fletcher, Bullmore, Sahakian, & Robbins, 2003; Garavan et al., 1999, 2002; see Aron et al., 2014, for a review). In contrast, activations in the left prefrontal and frontal cortices have been associated with the production of imagined content across different modalities (Sabatinelli et al., 2006; Lundstrom et al., 2003; Yoo et al., 2003). Suppression also selectively recruited the right ACC, which has been implicated in conflict monitoring (Gehring & Knight, 2000; Carter et al., 1998; see Botvinick et al., 2004, for a review), in which engagement would be important to monitor and detect thought intrusions during thought suppression.
Imagery selectively recruited the inferior temporal gyrus, an area associated with high-level visual representations (DiCarlo & Cox, 2007; Grill-Spector, 2003; Grill-Spector et al., 2001), which is likely to be important to visualize the stimuli.
The failed versus successful suppression GLM contrast highlighted brain areas responsible for adequate control of thoughts versus those associated with thought intrusions. Again, our analyses found pFC and ACC as areas implicated in thought control or successful thought suppression. This result suggests that engagement of this network is important for keeping suppressed thoughts suppressed. On the other hand, thought intrusions were associated with hyperactivity of memory areas: the parahippocampal cortex and the hippocampus as well as visual areas such as the LOC and the inferior temporal cortex. This result suggests that, during thought suppression, a control/inhibition network in frontal areas would oppose the activation of a sensory/memory network located posteriorly. Imbalances in the activation of these networks might have a tangible effect on thought control and intrusions, thus identifying them as potential therapeutical targets to treat thought intrusion disorders. Interestingly, classifiers trained on spatial activation patterns in visual areas (V1–V4) were successful at discriminating between successful and failed thought suppression. This result shows that visual areas contain reliable information differentiating these conditions. In summary, these results suggest that successful thought suppression prevents or holds at bay the memory contents and visual areas, as shown in other thought control studies, likely by recruiting inhibitory mechanisms triggered in executive areas (Schmitz, Correia, Ferreira, Prescot, & Anderson, 2017; Aso et al., 2016; Depue, Curran, & Banich, 2007).
Interestingly, the comparison of failed suppression versus imagery showed activations in right temporal areas (Supplementary Figure S4) instead of bilateral activations as in suppression versus imagery. Although both imagery and failed suppression share the production of visual qualia, the former is volitional, whereas the latter is automatic or intrusive. It is possible that the left engagement is associated with visual internally generated qualia, no matter its origin (volitional or automatic), whereas the right engagement is preferentially triggered by deeper categorical representations. This hypothesis is somewhat supported by evidence on face processing indicating that the left inferior temporal cortex is involved in low-level processing whereas the right one is engaged by deeper categorical processing (Meng et al., 2012). Left activations would thus cancel out in the failed suppression versus imagery contrast as both would recruit low-level representations.
Despite their differences, studies using memory-suppressing paradigms (think/no-think paradigms) in which pairs of words are learned and cued for recollection or suppression (Anderson & Green, 2001) have identified areas consistent with our results such as suppression linked with recruitment of the dorsolateral pFC (Benoit & Anderson, 2012; Depue et al., 2007; Anderson et al., 2004) and ACC (Anderson et al., 2004), while reducing activity in memory areas (Depue et al., 2007; Anderson et al., 2004) and visual areas when tasks involved visual content (Depue et al., 2007). Interestingly, in our paradigm, the semantic content of the item was not suppressed as it was shown as a cue (e.g., “broccoli”), whereas the visual thought of it was. This hints at a difference in the mechanisms of inhibitory control between these paradigms, as different representational content gets suppressed. Despite the modality of the suppressed item in both paradigms (semantic vs. visual for think/no-think and our paradigm), the recruited and suppressed areas seem to overlap, mainly in frontal areas. This is consistent with a previous study investigating the general mechanisms of inhibitory regulation in different processes, which found a common network responsible for inhibition in executive areas (Depue, Orr, Smolker, Naaz, & Banich, 2016).
It is important to note that the activations associated with thought suppression could reflect an alternative strategy: self-distraction or object substitution. We did, however, instruct participants to refrain from using substitution strategies (see Methods section), as we did in our psychophysics study, where we tested both suppression and substitution strategies (Kwok et al., 2019). Although we cannot absolutely ascertain that participants did not use any substitution, our previous study showed that participants instructed to use either substitution or suppression strategies led to drastically different outcomes, thus indicating that participants are able to follow these instructions accurately.
It is up to future studies to compare direct visual thought suppression and substitution strategies. In a previous psychophysics study of our group, we found that visual thought substitution strategies led to increased thought control (Kwok et al., 2019), whereas a previous neuroimaging study has shown that direct suppression and substitution recruit dissociable neural mechanisms in a memory forgetting paradigm (Benoit & Anderson, 2012).
Finally, our decoding analysis revealed that the content of the successfully suppressed thoughts is stored in the right LOC. Targeting this area with noninvasive brain stimulation (Sparing & Mottaghy, 2008) could help to minimize the unwanted effects of suppressed thoughts even when subjectively suppressed thoughts are successfully censored.
We believe that these results will shed light on the mechanisms of involuntary visual thoughts and the pervasiveness of their effects, despite subjectively successful suppression, by informing therapeutical strategies to prevent unwanted visual thought intrusions.
This work was supported by the Australian National Health and Medical research council (NHMRC) project grants APP1046198 and APP 1085404 as well as Australian Research Council grant DP140101560. Joel Pearson is supported by a NHMRC career development fellowship (APP1049596).
Reprint requests should be sent to Roger Koenig-Robert, School of Psychology, The University of New South Wales, High Street, Kensington, Sydney, NSW 2052, Australia, or via e-mail: firstname.lastname@example.org.
Supplementary materials for this paper can be retrieved from https://doi.org/10.6084/m9.figshare.12858842.v1.