An investigation of mental imagery in bipolar disorder: Exploring “the mind's eye”

Objectives Mental imagery abnormalities occur across psychopathologies and are hypothesized to drive emotional difficulties in bipolar disorder (BD). A comprehensive assessment of mental imagery in BD is lacking. We aimed to test whether (i) mental imagery abnormalities (abnormalities in cognitive stages and subjective domains) occur in BD relative to non‐clinical controls; and (ii) to determine the specificity of any abnormalities in BD relative to depression and anxiety disorders. Methods Participants included 54 subjects in the BD group (depressed/euthymic; n=27 in each subgroup), subjects with unipolar depression (n=26), subjects with anxiety disorders (n=25), and non‐clinical controls (n=27) matched for age, gender, ethnicity, education, and premorbid IQ. Experimental tasks assessed cognitive (non‐emotional) measures of mental imagery (cognitive stages). Questionnaires, experimental tasks, and a phenomenological interview assessed subjective domains including spontaneous imagery use, interpretation bias, and emotional mental imagery. Results (i) Compared to non‐clinical controls, the BD combined group reported a greater impact of intrusive prospective imagery in daily life, more vivid and “real” negative images (prospective imagery task), and higher self‐involvement (picture‐word task). The BD combined group showed no clear abnormalities in cognitive stages of mental imagery. (ii) When depressed individuals with BD were compared to the depressed or anxious clinical control groups, no significant differences remained—across all groups, imagery differences were associated with affective lability and anxiety. Conclusions Compared to non‐clinical controls, BD is characterized by abnormalities in aspects of emotional mental imagery within the context of otherwise normal cognitive aspects. When matched for depression and anxiety, these abnormalities are not specific to BD—rather, imagery may reflect a transdiagnostic marker of emotional psychopathology.


| INTRODUCTION
Mental imagery comprises the experience of seeing in the "mind's eye," now regarded as "a weak form of perception". 1 No wonder that negative mental images generate strong emotions, indeed stronger than does thinking in verbal language. 2 Bipolar disorder (BD) is characterized by periods of heightened emotion (depression and mania), 3 both during acute episodes and inter-episodically. 4, 5 We have suggested that mental imagery may act as an "emotional amplifier"-fueling mood deterioration, mood elevation, and anxiety symptoms typical in BD. 6 Initial data suggested that patients with BD present with heightened emotional mental imagery compared to non-clinical controls, 7 in particular higher trait imagery use and heightened impact of intrusive mental imagery of future events (prospective imagery). Furthermore, those patients with BD with greatest mood instability reported a greater impact of prospective imagery. 7 Compared to unipolar patients with equivalent levels of depressed mood, patients with BD reported more compelling and preoccupying prospective suicidal images. 8 This is of interest given that BD has the highest suicide rate of all psychiatric disorders. 9 Patients with BD also reported more frequent "flashforwards" to future events at times of positive mood than did people with unipolar depression, and rated these "flashforwards" as more vivid, exciting, and pleasurable. 10 However, a more comprehensive assessment of mental imagery function in BD is lacking. Pearson et al. 11  Previous studies have investigated only selected stages of imageryrelated processing, with evidence of deficits in cognitive tasks of imagery generation and manipulation in depressed individuals 13 and imagery generation in anxious individuals. 14 The subjective domains relate to spontaneous imagery use, 15 the presence of imagery-related interpretation biases and emotional mental imagery [16][17][18] and the phenomenological characteristics of mental imagery in different affective states. 16 "Rediscovering" mental imagery in clinical practice can improve assessment. 17 There is also emerging evidence of imagery as a valid target to reduce mood instability in BD. 18 Therefore, a comprehensive evaluation of cognitive stages and subjective domains of mental imagery in BD could further inform our understanding of BD psychopathology and treatment development, by identifying problematic aspects of mental imagery in BD and refining treatment targets.
The current study aimed to investigate: (i) whether individuals with BD have mental imagery abnormalities compared to non-clinical controls and (ii) whether mental imagery abnormalities (when present) are specific to individuals with BD compared to clinical controls with depression and anxiety. To address these questions, we compared (i) patients with BD and non-clinical controls; (ii-a) patients with BD and patients with unipolar depression with equivalent levels of depressive symptoms; and (ii-b) patients with BD and patients with anxiety disorders with equivalent levels of anxiety symptoms. We also explored whether clinical variables such as depressive and anxious symptomatology, bipolar phenotype traits, affective lability and general functioning levels predicted scores on mental imagery measures in the whole sample combined. A range of tests were used encompassing both cognitive stages of mental imagery and assessment of subjective and emotional domains.

| Participants
Participants completed pre-screening questions via email or phone to assess potential eligibility, based on which 175 were invited to attend a screening session. At the beginning of the session, all participants

| Clinical characteristics
Clinical characteristics were assessed using the SCID for DSM-IV Axis I disorders, including main diagnosis and lifetime and current comorbid disorders, as above. Current medication was recorded. Depressive, (hypo)manic, and anxiety symptoms were assessed using the HAM-D, the Young Mania Rating Scale, 21 the Altman Self-Rating Mania scale, 22 the Quick Inventory of Depressive Symptomology (QIDS), 23 and the Beck Anxiety Inventory (BAI). 24 The Mood Disorder Questionnaire (MDQ) 25 was administered to assess hypomanic experience. The Affective Lability Scale (ALS) 26 was used to measure changeable affect and the Functional Assessment Staging Test 27 to assess functional impairment in areas including occupational functioning, cognitive functioning, and interpersonal relationships.

| General cognitive function
The National Adult Reading Test 28 was used as an assessment of premorbid IQ. Verbal fluency (as a measure of general executive function) and verbal working memory function were assessed using the Verbal Fluency Test with the letters F, A, S 29 and Forward and Backward Digit Span Task, 30 respectively.

Spontaneous imagery use
The spontaneous use of mental imagery in everyday life was assessed via the Spontaneous Use of Imagery Scale (SUIS) 15 and two Visual Analogue Scales (VASs). 7 The SUIS is a 12-item self-report scale measuring the use of non-emotional mental imagery in daily life (e.g., If I am looking for new furniture in a store, I always visualize what the furniture would look like in particular places in my home.). Each item is rated on a five-point scale, with total scores ranging from 12 to 60. Higher scores indicate more use of mental imagery in daily life. The SUIS has an internal consistency of α=0.98 and good convergent validity. 15 Two VASs were used to assess the extent to which participants had been thinking in verbal thoughts or in mental images over the past week on a 1 (not at all) to 9 (all the time) scale.

Imagery interpretation bias
The Ambiguous Scenarios Test (AST-D) 31 and the Homograph Interpretation Task (HIT) 32 were used to measure imagery interpretation bias. The AST-D comprises 24 ambiguous scenarios, which participants were asked to imagine happening to them personally (e.g., 1=You go to a wedding where you know very few other guests. After the party, you reflect on how the other guests behaved.), and then rate each image's pleasantness from 1 (extremely unpleasant) to 9 (extremely pleasant) and vividness from 1 (not at all vivid) to 7 (extremely vivid).
The AST-D has good internal consistency (α=0.82). 31 In the HIT, participants are presented with a word and then asked to generate a mental image. The words were eight threatening/non-threatening homographs, for example, "mug" could cue either a benign (e.g., imagining oneself drinking out of a mug) or negative (e.g., imagining being attacked/mugged) mental image. Participants provided a short written description of each image and then rated their pleasantness (1-9 scale) and vividness (1-7 scale). Average vividness and pleasantness scores were computed for benign, negative and ambiguous mental images.

Emotional mental imagery
Emotional mental imagery was assessed using a Picture Word Cue (PW) task, 33 the Impact of Future Events Scale (IFES), 34  The PW task is a computer-based task examining self-reported spontaneous use of imagery in response to emotional information and emotional context. Participants were presented with 20 ambiguous/neutral pictures with negative word captions and instructed to "combine the picture with the word" (e.g., picture of students sitting an exam and caption word "fail"). They then rated from 1 (not at all) to 9 (extremely) how much they found themselves thinking in mental images, or in verbal thoughts, and how emotional they found the picture−word combination. Average tendency to use images and verbal thoughts and average emotionality of the picture−word combinations were computed.
On the IFES, participants were asked to identify three future events they had thought about/imagined over the past 7 days and state whether each was positive or negative. Participants then responded to 24 statements about prospective imagery in relation to the past week, on a scale from 0 (not at all) to 4 (extremely). The IFES has acceptable test−retest reliability (0.73) and a good internal consistency (α=0.87). 34 The PIT comprises 10 positive and 10 negative hypothetical future scenarios. Participants were asked to generate an image of each and rate each image on a five-point Likert scale for vividness, likelihood of the event happening to them in the near future, and how much they feel as though they are experiencing the event whilst imagining it, with higher ratings indicating more vivid and "real" prospective imagery. All subscales of the PIT have demonstrated good internal consistency (0.83<α<0.90). 36 The MII is a semi-structured interview, which assesses content and characteristics of mental images and verbal thoughts experienced when the participant has been most anxious, most low and most high in mood. Participants are first asked to describe their most significant mental image anchored to each affect state and rate characteristics of the image such as valence, general emotionality of the image and intensity of one specific associated emotion per each affect state (i.e., threatening, demotivating and exciting). They are then asked to rate overall characteristics of mental imagery and verbal thoughts for each affect state (anxious, low, and high) such as frequency, realness, and compellingness. All ratings use nine-point Likert scales, with higher ratings indicating more frequent, real (etc.) imagery or thoughts.

| Cognitive (non-emotional) stages of mental imagery
The following tasks were administered to assess the four cognitive stages of mental imagery. 11

Imagery generation
The Image Generation Task (IGT) 37 measures the ability to generate a mental image based on previously encountered perceptual information.
Participants were asked to memorize the shape of four block capital letters presented in a 4 × 5 grid: "U" and "H", classified as simple (three or fewer segments), and "S" and "J," classified as complex (four or more segments). Participants were then presented with a blank grid with a lowercase letter underneath, indicating which letter the participant should imagine. An "X" was presented in one of the grid squares and participants were asked to respond "True" if the "X" would cover the imagined block letter if it were present in the grid or otherwise "False." Accuracy and reaction time were recorded. Socially anxious participants have previously shown image generation deficits on this task. 14

Imagery maintenance
The ability to maintain mental images in mind was assessed using two visual working memory tasks. The Short Term Memory (STM) task The Visual Patterns Test (VPT) 40 measures visual short-term memory and memory for positional sequences. Participants were presented with a sequence of increasingly complex checkerboard patterns, starting with a 2 × 2 matrix (with two cells filled in) and progressing to the largest 5 × 6 matrix (with 15 filled in cells). Each pattern was shown to the participant for 3 s and then hidden, at which point participants were asked to reproduce the pattern by marking squares in an empty grid of the same size. Accuracy scores were calculated using the maximum difficulty level reached for which two patterns were correctly reproduced.

Imagery inspection
The Letter Corner Classification (LCC) task 41 measures image inspection ability, involving interpretation of an object-based spatial characteristic of the image. Participants were first presented with four block capital letters (F, N, Z, and G), marked with an asterisk in the bottom left corner and an arrow travelling clockwise around the letter. Participants were instructed to memorize the shape of each letter and reproduce it on a blank piece of paper, starting at the point marked by the asterisk and following the direction of the arrow. Participants then categorized the corner of the letters. For each letter, first, for "top and bottom points", participants were asked to go around the shape, starting at the point marked by the asterisk, indicating "yes" if the corner was at the extreme top or bottom of the shape or otherwise "no". The letters were then removed and participants instructed to imagine each letter and categorize the corners. The letters were then presented again and the same procedure followed for "outside points", which required a "yes" response for corners on the extreme left and right of the figure. Accuracy and time taken for each letter in both conditions were recorded.

Imagery manipulation
Two tasks measuring the ability to manipulate mental images were administered. A computerized version of the classic Mental Rotation Task (MRT) 42 measured participants' ability to transform mental images.
Participants were shown pairs of three-dimensional line drawings and instructed to decide whether the two drawings were the same or different by using a mental rotation strategy. Following a practice trial, the task included trials with three difficulty levels based on whether the angular disparity between the two shapes was 50, 100, or 150.

| Statistical analysis
First, we tested (i) if participants with BD had imagery abnormalities by comparing the BD group (euthymic and depressed combined) to nonclinical control participants on measures of cognitive and subjective domains of mental imagery. To test for between-group differences on these aspects of imagery abnormalities, 55 statistical tests were performed. A BD group combining euthymic and depressed individuals was used to test replication of previous data. 7 Moreover, as euthymic individuals with BD present with depression levels greater than those of non-clinical controls (albeit subclinical), we chose to first assess the presence of imagery abnormalities regardless of affect state. Next, we sought to determine the specificity of any group differences by comparing the scores of (ii-a) currently depressed participants with BD to those of currently depressed participants with unipolar depression (this also allows controlling for the impact of depressed mood on mental imagery abnormalities), and (ii-b) currently depressed participants with BD with concurrent anxiety symptoms to those of participants with anxiety disorders (this also allows controlling for the impact of anxiety on mental imagery abnormalities; the two groups were also matched on levels of depression). To limit the number of tests, comparisons of BD depressed to clinical control groups were limited to (i) those variables that showed significant group differences in the initial comparison (BD group combined versus non-clinical controls) and (ii) those comparisons that had yielded significant differences in previous studies. 7,8 To test for differences between depressed participants with BD and unipolar depressed participants and to test for differences between depressed participants with BD with concurrent anxiety symptoms and participants with anxiety disorders, 38 statistical tests were performed.
Pairwise differences between variables in the different groups as outlined in our aims were analyzed using unpaired t tests if the residuals obtained using these t tests achieved normality with P-values above .05 using both the Kolmogorov-Smirnov and Shapiro-Wilk tests. Where the group variances were found to differ using Levene's test, Satterthwaite's correction was applied to the degrees of freedom of the t test. Where the residuals of a pairwise comparison on an untransformed response did not achieve normality, log, square root and reciprocal transformations were applied and normality of the residuals reassessed. Where transformations failed to achieve normal residuals, Mann-Whitney U tests were used to analyze group differences. For the CMS task, Fisher's exact test was used to identify group differences in the number of CMS trials that were judged as recognizable, creative, having good correspondence, having a correct pattern and having a present pattern.
To explore the specificity of differences in mental imagery between diagnostic groups further, we computed correlations be- In all analyses, P-values <.05 were considered statistically significant and no corrections for multiple testing were applied. Normality checks of model residuals allowed any undue influence of outliers to be reduced without losing information by removing them.

| Participants
Demographic and clinical characteristics of all groups are presented in Table 1. There were no between-groups differences in age, gender, ethnicity, level of education, or premorbid IQ.

| Do individuals with BD show mental imagery abnormalities compared to non-clinical controls?
Scores on assessments of cognitive (non-emotional) stages and subjective domains of mental imagery of participants with BD and non-clinical controls, and results of between-group comparisons are summarized in Tables 2-4 (all data referring to the BD combined group).

| Subjective domain of mental imagery
All results related to the subjective domain of mental imagery are detailed in Table 3. Participants with BD did not significantly differ  a Current anxiety disorder types were: social anxiety (n=12), obsessive compulsive disorder (n=9), posttraumatic stress disorder (n=11), generalized anxiety disorder (n=20), specific phobia (n=10), panic disorder (n=12), and agoraphobia (n=2). Please note that some participants presented with multiple anxiety disorders.
reported a higher number of benign homographs which was marginally significant compared to those with BD. The two groups did not significantly differ in other interpretation bias ratings from the HIT.
On measures of emotional mental imagery, participants with BD scored higher on the PW self-involvement scale compared to nonclinical control participants. The two groups did not differ in any of the other PW task scales. Participants with BD reported a stronger impact of emotional prospective imagery on the IFES compared to nonclinical controls. They also reported higher ratings of vividness and sense of experiencing for negative future images, and lower ratings of likelihood for positive future scenarios on the PIT. The two groups did not significantly differ on the remaining PIT scales.
All results on the MII are detailed in Table 3. For the time when their mood was most low, participants with BD rated their most significant mental image as more negative and more demotivating compared to non-clinical controls. For the time when their mood was most anxious, participants with BD rated their most significant mental image as more negative, threatening and emotional compared to non-clinical controls. They also rated overall thinking in mental images to be more frequent and more "real" compared to non-clinical controls. For the time when their mood was most high, participants with BD rated their most significant image as more exciting compared to non-clinical controls. They also rated overall mental imagery as more "real" compared to non-clinical controls. Full results are reported in Table 4. Qualitative examples of significant mental images are reported in Table 5.

| Cognitive (non-emotional) stages of mental imagery
All results related to the cognitive (non-emotional) stages of mental imagery are detailed in Table 2. Participants with BD did not significantly differ in their performance on any part of the imagery generation task, indicating no imagery abnormalities in the BD group in terms of imagery generation in a non-emotional cognitive task.
Of the two imagery maintenance tasks, participants with BD had a higher recall rate on the visual STM task compared to those in the non-clinical control group, indicating that participants with BD in this study had a greater likelihood of remembering visual target cues in a T A B L E 2 Mean differences between participants with bipolar disorder and non-clinical control participants in measures relating to the cognitive (non-emotional) stages of mental imagery

| Are mental imagery abnormalities specific to patients with BD?
Next we tested the specificity of findings to BD compared to individuals with unipolar depression and individuals with anxiety disorders (see the section 'Statistical analysis').

| Cognitive (non-emotional) stages of mental imagery
There were no differences in imagery manipulation (based on performance on the MRT) or in visual short-term memory (based on recall rate scores on the STM task) between (ii-a) BD depressed and unipolar

| Relation between mental imagery measures and depression, anxiety, BD phenotype, affective lability and general functioning
Given the lack of specificity of mental imagery abnormalities present in participants with BD, we next explored whether these imagery ab-  Table S2). Only those imagery variables that showed significant group differences in the comparison between participants with BD and non-clinical control participants (reported in Table 2) were included in these analyses.
T A B L E 5 Example of significant images for each affect state (anxiety and low and high mood) for participants with bipolar disorder and non-clinical controls taken from the Mental Imagery Interview and mean emotional ratings of the significant images

Bipolar disorder Non-clinical controls
Low mood A suicide plan-extensive and intelligent. I would go to the college bar and take one of the CO 2 bottles used to pump Guinness and take it back to my room. I would send an email to tell people not to come in and release the CO 2 (pp 157) Seeing the email rejecting you from the job (pp 131) Thinking about mold growing in the kitchen. The corners of the surfaces having mold, greeny gray mold. General disorder-lots of dirty crockery, lots of food. General horror. Smell of mold (pp 178) What my mother looked like when healthy and well. What she looked like after a series of strokes (pp 269) Picture of a human brain with nasty pathology-fear about my own brain.

BD phenotype
Across all groups, higher levels of hypomanic experiences (measured by the MDQ scores) were associated with worse performance on imagery manipulation as measured by a higher error rate on the MRT task Affective lability Across all groups, higher levels of affective lability (measured by scores on the ALS) were associated with better performance on imagery maintenance as measured by a higher recall rate on the STM task

| Summary of main findings
Our study investigated, first, whether BD is associated with abnor- This finding is particularly interesting given that affect lability (including concurrent anxiety) represents a particularly challenging feature across different mental disorders.

| Emotional mental imagery in BD
We replicated previous findings that individuals with BD (euthymic and depressed combined) experience a greater impact of intrusive prospective mental imagery in everyday life, 7 and perceive prospective negative images as more vivid in an experimental task compared to non-clinical controls. 7 Further, we extended these findings in that our sample of participants with BD also reported more real (greater "sense of experiencing") prospective negative mental images and perceived imagined positive events as less likely to occur compared to non-clinical controls. Consistent with a greater sense of experiencing anxiety for negative prospective images, compared to non-clinical controls, participants with BD also felt more self-involved when spontaneously generating mental images by combining negative pictures and words.
Unlike previous studies, we did not find evidence of greater spontaneous use of non-emotional mental imagery in BD compared to nonclinical controls, although mean values were in the same direction as in a previous study. 7 This suggests that differences in spontaneous tendency to visualize are likely to be small and test of replication in larger samples is needed to verify these inconsistencies.
Overall our BD sample reported imagery abnormalities particularly for prospective imagery and during anxious affect. This is consistent with the relationship between anxiety and future thinking. 46 As prospection plays an important role in regulating emotions and behavior, 47 it is possible that these abnormalities in prospective imagery (although not limited to BD; see below) contribute to emotional and behavioral dysregulation typical of BD. 6 It could be fruitful to investigate the effect of prospective imagery on the presence and severity of comorbid anxiety, which is a key clinical feature in BD. 48,49 For example, one participant with BD reported that when most anxious they repeatedly experience vivid negative images of embarrassing themselves at a social event; the images feel so real that they further fuel anticipatory anxiety to the point of making them avoid attending the event. A better understanding of prospective anxiety-inducing imagery in BD may also have implications for therapy, given the challenge of treating anxiety in this disorder. 18,48 Future studies could investigate whether the experience of emotional mental imagery in BD differs depending on the type of anxiety comorbidity, following current cognitive accounts of anxiety disorders where imagery is predominant, such as social anxiety, 16 or irrelevant if not suppressed, such as general anxiety disorder. 50

| Cognitive stages of mental imagery in BD
Finally, and novel to the literature (as called for by Pearson et al. 11 ), the absence of major dysfunctions in the cognitive (non-emotional) stages of mental imagery suggests that there are no deficits in the ability to generate, manipulate, and recall images. Interestingly, our BD sample also showed a greater likelihood of recalling the target cues in one of the visual short-term memory tasks compared to non-clinical controls. Thus, individuals with BD appear to have an overall intact functioning or even an "advantage" in this aspect of imagery processing. Therefore, drawing on mental imagery techniques 18,51 could be a successful strategy in treatment interventions for BD where other cognitive processes may be impaired (as in our sample with reduced verbal fluency/executive function performance). 52

| Mental imagery abnormalities as a transdiagnostic phenomenon
Unlike previous studies comparing BD and unipolar depressed patients, 8,10 no differences emerged between our clinical groups in prospective imagery measures. In fact, across the whole sample combined, prospective imagery abnormalities (on IFES total and PIT negative scenarios scores) were associated with severity of anxious symptomatology and affective lability traits. This suggests that inconsistencies between studies of clinical groups may be accounted for by the relative distribution of affective lability traits and concurrent anxiety in the samples.
Consistent with previous data, 10 depressed participants with BD rated their most significant image at times of high mood as more exciting compared to participants with unipolar depression. This may reflect both an association between mania and positive mental imagery (even at times of depressed mood) and a deficit in positive mental imagery in unipolar depression 51,53,54 . The finding is also consistent with recent neuroimaging evidence showing that participants with BD and unipolar depression present different neural responses to positive stimuli only while depressed. 55 We did not replicate previous evidence of negative images being more compelling in BD compared to unipolar patients. 8 This discrepancy might be accounted for by less severely depressed samples in the present study or may suggest that greater compellingness might be specific to suicidal flashforwards in patients with BD 8 rather than any image during low mood.
With regard to the cognitive (non-emotional) stages of mental imagery, previous studies have reported biases in imagery generation and manipulation in unipolar depressed individuals compared to controls 13,44 ; however, these depression-related abnormalities were only present in measures that index the sensory/response component of imagery tasks 13 rather than specific imagery (e.g. spatial ability) processing biases. Therefore, discrepancies between studies may be explained by differences in sensory-motor retardation symptoms between the samples.
Overall, our study indicates that mental imagery characteristics representing features of greater emotionality and intensity (e.g., greater intrusive imagery impact, vividness of negative images, and sense of realness of images) may represent a marker for general emotional psychopathology, and general functioning. This supports our idea that "bringing back the mind's eye" to psychiatric assessments 17 could help identify clinical severity. Most importantly it can help clinicians to understand and normalize aspects of patients' subjective experiences that may otherwise feel particularly alien and distressing (such as intrusive highly emotional mental images). Asking about mental images offers an alternative access to capturing distress in those patients who may struggle to communicate their subjective experiences via traditional verbal thoughts. The transdiagnostic relevance of mental imagery also highlights potential avenues for new treatment interventions: e.g. if depression scores relate to how likely positive future images feel, reverting positive imagery biases may be a useful target to improve mood. 36,51,56 Our results on the association between BD phenotype and affective lability traits, and greater imagery frequency and emotionality, are in keeping with previous findings that individuals with a BD phenotype are more susceptible to intrusive imagery and to spontaneous use of imagery. 57,58 Interestingly, better performance in imagery maintenance via visual short-term memory was also associated with affective lability. Future studies should investigate the relationship between biases in emotional mental imagery, visual short-term memory function and emotional instability across psychopathology, including in other conditions where this is relevant such as borderline personality disorder.
Overall, mental imagery biases could be conceptualized as a cognitive psychopathological dimension in line with most recent neuroscience dimensional approaches to understanding mental disorders (research domain criteria 59 ). Future research should investigate how currently established cognitive and neural markers of emotional dysregulation and affective lability 60-62 relate to abnormalities in emotional mental imagery described in our sample. Moreover, as affective lability often represents a therapeutic challenge, treatment innovation should explore the potential for using imagery-focused interventions for emotional instability. 18

| LIMITATIONS AND CONCLUSIONS
A limitation of our study is the absence of statistical correction for multiple comparisons. Moreover, we did not include a (hypo)manic BD group that would allow us to establish the presence of mental imagery abnormalities associated with mania state diagnosis. With regard to results from the MII, it should also be noted that these were based on retrospective accounts of times of intense affect and could have been subject to recall/memory biases. Future qualitative studies are needed to analyze in detail potential differences in the image contents exemplified in Table 5. Our data suggest that mental imagery abnormalities are typical of acute clinical states of anxiety and depression, but are also associated with traits of BD phenotype and affective lability. Future studies should include individuals recovered from unipolar depression and anxiety disorders to clarify if emotional mental imagery abnormalities also persist beyond acute depression/ anxiety across psychopathology, as they do in our BD sample (euthymic and depressed). Moreover, as our clinical groups all presented moderate to high levels of both anxiety and depression, future studies could attempt to tease apart the association between mental imagery abnormalities and anxiety/depression, although this may be a challenge given the high co-occurrence of these symptoms in emotional disorders. Nevertheless, the regression analyses across all groups in our sample suggest a greater impact of anxiety on mental imagery characteristics. Finally, longitudinal rather than cross-sectional studies should further investigate stability and change of mental imagery abnormalities in BD over the course of illness. Future studies could also compare individuals with bipolar I and II disorder using sufficiently powered samples.
In conclusion, this first comprehensive investigation of a range of mental imagery measures in BD compared to both non-clinical and clinical controls confirms that imagery abnormalities are present in patients with BD in the emotional aspects of mental imagery, while the cognitive processes underpinning mental imagery experience remain largely intact. Biases in emotional mental imagery appear as a transdiagnostic feature of our clinical groups matched on depression and anxiety levels related to clinical dysfunction. We suggest that imagery abnormalities are a transdiagnostic processes driving affective lability, and that imagery can be targeted via novel psychological treatment techniques. Imagery-focused techniques hold promise across psychiatric disorders, 17 including adding treatment value to BD. 18