The critical role of mental imagery in human emotion: insights from fear-based imagery and aphantasia

Abstract

One proposed function of imagery is to make thoughts more emotionally evocative through sensory simulation, which can be helpful both in planning for future events and in remembering the past, but also a hindrance when thoughts become overwhelming and maladaptive, such as in anxiety disorders. Here, we report a novel test of this theory using a special population with no visual imagery: aphantasia. After using multi-method verification of aphantasia, we show that this condition, but not the general population, is associated with a flat-line physiological response (skin conductance levels) to reading and imagining frightening stories. Importantly, we show in a second experiment that this difference in physiological responses to fear-inducing stimuli is not found when perceptually viewing fearful images. These data demonstrate that the aphantasic individuals' lack of a physiological response when imaging scenarios is likely to be driven by their inability to visualize and is not due to a general emotional or physiological dampening. This work provides evidence that a lack of visual imagery results in a dampened emotional response when reading fearful scenarios, providing evidence for the emotional amplification theory of visual imagery.

1. Introduction

Models of human cognition propose that mental imagery exists for aiding thought predictions by linking them to emotions [1,2], allowing us to simulate and ‘try out’ future scenarios, as if we are experiencing them and their resulting emotional outcomes, aiding us in decision making (see [3] for a review). Models of mental illness commonly identify mental imagery as a key driver of negative emotion [4,5] through the sensory simulation of non-current events which is hypothesized to drive affective reactions to thoughts about non-current events. In other words, imagery is a proposed ‘emotional amplifier’ of mental content [6].

However, theories proposing that imagery exists to link thoughts with an emotional response are largely based on subjective reports of emotions from individuals attempting to think with and without sensory simulations [7]. It is likely that for most of us, sensory simulations in thought are largely involuntary [8], and hence cannot be switched off on cue.

Here, we tested the theory that imagery can function to enhance emotions in thought, by using a group of individuals who do not experience visual imagery, a naturally occurring ‘knock-out’ model of imageless cognition. The subjective absence of visual mental imagery ability in such individuals has been documented since the nineteenth century [9]; however, the condition was only recently given a name, congenital aphantasia [10], and has to date been most commonly defined by subjective reports and floor scores on the self-report Vividness of Visual Imagery Questionnaire (VVIQ) [11].

After using a multi-method verification of the absence of imagery (using scores from the VVIQ and binocular rivalry imagery paradigm; see Participants section below for more details) in self-identified aphantasic volunteer participants, we compared a sample of aphantasic individuals with a general population (imagery-intact) control sample on a task designed to elicit frightening imagery. Participants read a series of custom, speed-controlled, first-person fictitious fearful scenarios (figure 1a; Methods), while their skin conductance level (SCL) was continuously recorded. Skin conductance indexes changes in autonomic nervous system arousal and generally increases in response to frightening stimuli [12], including imagined stimuli [13]. The fearful scenarios were written in visually descriptive terms and intended to engender spontaneous visual mental imagery production in non-aphantasic participants. We predicted that aphantasic participants, who cannot generate visual mental imagery, would exhibit significantly less increase in SCL than controls. This experiment is hereafter referred to as the imagery experiment.

Figure 1.

Skin conductance data for imagery and perception experiments. (a) Imagery experiment. Fifty seconds of baseline SCL was recorded prior to each scenario trial while participants viewed an on-screen instruction. Next, each scenario trial was presented to participants as a succession of 50 on-screen phrases, each displayed for 2 s. (b) (i) Aggregated progressions of baseline-corrected SCL across the duration of scenarios (sampled as average across 5s time bins). (ii) Mean and SEM across time bins. (c) Perception experiment. Baseline SCL was recorded while participants viewed neutral photos, before being presented with a succession of frightening photos. Photos appeared on screen for 5 s each and immediately followed one another. (d) (i) Aggregated progressions of baseline-corrected SCL across the duration of the frightening photos sequence (sampled as average across 5 s time bins). (ii) Mean and SEM across time bins. (Online version in colour.)

2. Material and methods: imagery experiment

(a). Participants

Sample size estimation was based on the theoretical centrality of mental imagery in amplifying emotional responses to thought and the appearance of large effect sizes in the limited existing research comparing aphantasics and the general population on imagery measures [10,14]. For a two-independent-group t-test with a large effect size (Cohen's d = 1) and with α = 0.05 and power = 0.8, the projected sample size necessary was 17 participants per group (GPower 3.1). Conservatively, we set out to recruit at least 20 participants for each group.

Twenty-nine self-identified aphantasics approached the laboratory, wherein we first audited their lack of imagery using the VVIQ with a score cut-off ≤32 [10], leading to the exclusion of four volunteers on the basis of high scores. However, were we to solely rely on self-report measures such as the VVIQ, self-identified aphantasics may simply lack good metacognition of their imagery or have a bias to report it as low or non-existent. We therefore further verified the absence of imagery in this population using a psychophysical task that objectively measures imagery's sensory strength via imagery's ability to systematically prime perception in subsequent binocular rivalry (score cut-off ≤0.60, see description of paradigm in electronic supplementary material, Methods) [14,15]. Using these criteria, a further three self-identified aphantasic participants were excluded, leaving a total sample of 22 verified aphantasics (10 female) qualified to participate in our study. The aphantasics were recruited through a Facebook page, had emailed the laboratory regarding their aphantasia or were referred to the laboratory by Professor Adam Zeman, a UK-based neurologist and aphantasia researcher. Aphantasic participants were financially reimbursed (20 Australian dollars per hour) for their time.

We also recruited a general population (imagery-intact) control group of 24 participants (11 female), who self-identified as having intact imagery and had VVIQ scores in excess of the cut-off. The control participants were psychology students recruited via the UNSW research participation for the course credit system. Although our aphantasic group were on average older than controls (aphantasic mean age 33 years, SD 10.4; control mean age 23.0, SD 9.5), the possibility that age was a significant factor in our results was subsequently ruled out (see electronic supplementary material, figure S3). All experiments were approved by the UNSW psychology ethics committee.

(b). Apparatus

Participants were seated in a blackened, sound-attenuated room with lights off, in front of a BENQ ZL2420-B LCD computer monitor with a viewing distance of approximately 60 cm. All stimuli were displayed against a black background using Matlab v.R2014b with the Psychophysics Toolbox (v.3.0.12) extension (Mathworks, Natick, 2015) on an HP Z240 Tower Workstation computer. A second HP Z240 Tower Workstation, interfaced with the display computer, continuously recorded skin conductance level using ADInstruments hardware (Powerlab Model 16/30, MLT116F finger electrodes, and ML116 GSR Amplifier) and associated Labchart Pro v.8.1.4 software (ADInstruments, Sydney, 2016) at a 1000 Hz sampling rate. The finger electrodes were attached to the distal phalanges of the participant's non-dominant hand. The in-room air conditioning setting was maintained at 20 degrees Celsius. Participants were instructed to keep still throughout skin conductance data collection and were continuously observed by the experimenter to monitor compliance. Where significant movement occurred, the relevant story trial was excluded from analysis for that participant.

(c). Stimuli

All on-screen text stimuli were displayed in white 14-point Courier New font. A text instruction reading ‘relax, clear your mind, and keep still’ appeared during baseline skin conductance recording prior to each fictitious scenario in the Imagery experiment. Fictitious scenarios were presented as a succession of 3–7 words phrases, immediately following one another, each displayed for 2 s, see electronic supplementary material for all scenarios.

Participants’ proper reading and comprehension of each scenario was verified by asking them to briefly summarize the events described in each scenario at the conclusion of the experiment. Participants had no prior warning that they would be asked to do this. If a participant could not freely recall the events of each scenario, it was assumed that they had not adequately attended to that trial and their data from that trial was excluded from analysis. Two aphantasic individuals had one trial excluded, one had two excluded and one had three excluded (total of 7 trials across 4 participants). In the undergraduate population, 13 participants had one trial removed while one participant had two trials removed (total of 15 trials across 14 participants).

3. Results: imagery experiment

For each trial, skin conductance level was baseline-corrected by subtracting the mean SCL recorded during that trial's preceding 50 s baseline period. Average baseline-corrected SCL was then extracted from each trial in 20 consecutive 5 s time bins. Trial data were then averaged for each participant and across participants within each group. Figure 1b shows both the aphantasic (blue) and control participant SCL data (red), during the imagery provoking scenarios. The rightmost bar plot shows the mean SCL collapsed over the full 100 s of the stories' duration. A two-tailed t-test on the difference between this collapsed mean revealed a significantly higher SCL in controls relative to aphantasics (Welch's t-test: t_43.51 = 2.126, p = 0.039, d = 0.628, 95% CI = [0.031, 1.217]), suggesting that aphantasic individuals show a reduced fear response to the frightening scenarios when compared to controls. Underscoring this effect, further one sample t-tests revealed that only the control group experienced a significant increase in SCL compared to zero (t₂₃ = 3.811, p = 0.0009, d = 0.778, 95% CI = [0.312, 1.230]), while the aphantasic group's increase did not significantly differ from zero (t₂₁ = 0.6664, p = 0.5124, d = 0.142, 95% CI = [−0.280, 0.560]).

4. Discussion: imagery experiment

These results suggest that the arousal response to reading the fictitious scenarios may be largely contingent on having intact imagery to simulate the scenario content. This is consistent with existing evidence for imagery's theorized role as an emotional amplifier [6]. However, it is also possible that the difference in SCL change reflects group differences unrelated to imagery. One possibility is that our control group was more prone to general anxiety than our aphantasic group, or that aphantasic individuals were for some reason generally less emotionally responsive. However, a comparison of trait anxiety scores from the State-Trait Anxiety Inventory (STAI) suggested equivalent levels of trait anxiety between the two groups (see electronic supplementary material, figure S2).

Another possibility is that aphantasic individuals have a reduced/dampened fear response in general. To investigate this non-imagery explanation of the data, we ran a follow-up experiment, the perception experiment (figure 1c). In this experiment, SCL was measured in the same manner with the same apparatus; however, participants viewed frightening perceptual stimuli, instead of reading frightening fictitious scenarios. In this case, we predicted that both aphantasic and control participants would experience significant increases in SCL, the magnitudes of which would not differ significantly between the groups.

5. Material and methods: perception experiment

(a). Participants

Sixteen verified aphantasic participants (10 female, 8 from the imagery experiment) and 15 imagery-intact control participants (7 female, 5 from the imagery experiment) participated in the perception experiment. Recruitment procedures were identical to the imagery experiment.

(b). Stimuli

The frightening perceptual stimuli were a series of 18 International Affective Picture System (IAPS) [16] photographs, presented consecutively in the centre of the otherwise black screen in their original JPEG format and size, for 5 s per picture. As shown in figure 1c, the first five photographs served as a neutral baseline (rated least frightening in IAPS norms (e.g. umbrella, dinner plate)), while the remaining 13 were rated very frightening (e.g. snake's mouth, assault, cadaver). IAPS image catalogue numbers used, and the order of presentation for all participants was as below:

Neutral photographs: 7006.JPG, 7010.JPG, 7150.JPG, 7045.JPG, 7150.JPG; Frightening photographs: 3001.JPG, 3061.JPG, 6555.JPG, 3016.JPG, 1932.JPG, 1120.JPG, 1304.JPG, 2120.JPG, 3530.JPG, 5972.JPG, 6370.JPG, 9620.JPG, 9908.JPG.

6. Results: perception experiment

As in the imagery experiment, baseline-corrected skin conductance level during the presentation of the 13 frightening photographs was separated into 5 s time bins for each participant. The resulting data were averaged across participants within each group to again produce two plots (figure 1d). As predicted, both groups showed a monotonic increase in baseline-corrected SCL when viewing frightening perceptual pictures. Again, the rightmost bar plot shows the mean SCL collapsed over all time bins. One-sample t-tests revealed that both groups experienced significant increases in SCL compared to zero, while viewing the frightening pictures (aphantasic t₁₅ = 2.290, p = 0.037, d = 0.571, 95% CI [0.034, 1.095]; control t14 = 2.461, p = 0.0275, d = 0.635, 95% CI [0.069, 1.183]), and a between-subjects t-test revealed no significant difference between the SCL increases observed in the two groups (Welch's t-test: t_28.97 = −0.007, p = 0.994, d = −0.003, 95% CI = [−0.702, 0.707]). A Bayesian analysis was also run to assess the evidence in favour of the null. There was moderate/weak evidence in favour of the null hypothesis (BF₀₁ = 2.938), in line with the t-test, which suggests that there is little to no difference in SCL responses for the two groups when perceptually viewing emotive stimuli.

7. General discussion

Our data demonstrate that aphantasic individuals show significantly less of a physiological fear reaction (SCL) when reading scary stories, as compared to control participants with the ability to visualize. Importantly in a control perception experiment, aphantasic individuals did not have a general reduction in fear response relative to controls, in circumstances where frightening stimuli is perceptual in nature. This suggests that the observed difference in SCL in the imagery experiment is likely to be due to the group difference in visual imagery ability. As aphantasia allows the study of emotional thought without any visual imagery these findings add very strong evidence to imagery's theoretical role as an ‘emotional amplifier’ [3,6,17,18]. These data provide the first assessment of emotional thought in the complete absence of visual imagery and thus highlight how important visual imagery is in amplifying emotional physiological responses, probably due to its similarity to perception. This adds further impetus to the growing clinical appreciation that the imagistic content of cognition is a central determinant of mental wellbeing, and that image-based cognition is more emotionally consequential than its propositional (e.g. verbal) counterparts, despite these often being the explicit focus of ‘talk therapy’.

These data also provide further support for aphantasia being a condition that is characterized by a true lack of visual imagery, rather than a condition of poor metacognition. The results also raise questions about the emotional consequences of being, or not being, aphantasic. It is important to note the results do not suggest that aphantasia is associated with general muted emotionality. The equivalent responses between groups in the perception experiment and to the trait anxiety measure suggest otherwise. Rather, if aphantasics' emotional lives do differ systematically from their imager peers, this difference could well be limited to the reading of fictitious scenarios, which may have implications for education. Future experiments should examine whether aphantasics’ lower emotional response generalizes beyond thinking about fictitious scenarios to more personalized thoughts about distressing episodic memories or future-based worries.

If muted emotionality is observed across all imagery content, then aphantasia might engender resilience to the range of conditions known to involve imagery [1], including depression [19], social anxiety [4], obsessive compulsive disorder [20], PTSD [7,21], addiction [22], schizophrenia [23] and even the hallucinatory symptoms of Parkinson's disease [24]. Conversely, there is evidence that positive imagery may subserve optimism [25] and protect against depression [26], raising the possibility that aphantasia could increase the risk for dysphoric mood states (see [27] for a review of mental imagery and psychiatric disorders). Recent questionnaire studies seem to provide early support for these possibilities, with data indicating that aphantasic individuals report less intrusive imagery in relation to past traumatic episodes; however, they also report less vivid autobiographical memories and a decreased ability to imagine future scenarios [28,29].

It would also follow that insofar as psychological disorders do arise in the aphantasic population, forms of evidence-based therapy that rely on visual mental imagery, such as imagery rescripting and imaginal exposure, may well be inefficacious for aphantasic patients. Future research should examine these possibilities on an epidemiological basis by inquiring as to the prevalence of imagery-linked disorders among the aphantasic population and the efficacy of imagery-based treatments within this population.

An important limitation to note in this research is that only one emotion was assessed here: fear. It may be the case that different types of emotion might be less amplified by visual imagery. For example, aphantasic individuals may show physiological responses when imagining positive scenarios or personally relevant scenarios [30]. Future studies should examine whether the same reduced physiological response is present when imagining other emotional scenarios and more personally relevant content, such as imagery of past upsetting experiences. Another limitation to our study is that our control group was not age matched to our aphantasic sample; however, we think this is unlikely driving the observed results as age does not predict the SCL response to imagined scenarios (see electronic supplementary material, figure S3).

One final limitation is that this study is not capable of a full within-subjects (within-aphantasic) comparison. This is important because it has been shown that skin conductance is highly variable, and some otherwise healthy individuals do not always show a skin conductance response to arousing images. As such it may be the case that, by chance, we ended up with a larger proportion of ‘non-responders’ in our aphantasic group in the imagery experiment compared to the perceptual control experiment, and this is driving our observed effect. However, we think this is unlikely as there is significant variation in the SCL responses within the control group in the imagery experiment, with some individuals not showing a SCL effect (see electronic supplementary material, figure S4), and there is strong evidence that individual SCLs for all scenarios are positively correlated, suggesting these responses are consistent within individuals (see electronic supplementary material, table S1). Further, eight aphantasic individuals participated in both experiments (these individuals do allow a within-subjects design), and these data show the same pattern of response as the main findings. That is, these individuals demonstrated a galvanic skin response to IAP's images, but not to imagined fearful scenarios (see electronic supplementary material, figure S6). Future studies should endeavour to replicate this work using a fully within-subject design with age-matched controls, as well as extending the imagery condition to more personally relevant scenarios.

In conclusion, this study provides the first evidence that physiological responses to fearful imagery are reduced in aphantasia, providing support for imagery being an emotional amplifier of thought, as well as providing a potential new objective measure of aphantasia.

Data accessibility

The raw data reported here are available on request from the corresponding author (R.K.). These data are not publicly available as consent was not obtained from our special population (aphantasia) to share the data in a repository.

Authors' contributions

All authors developed the study concept. M.W. and R.K. built the study design and recruited participants. M.W. collected the SCL data, and performed SCL data analysis. R.K. collected the BR data and performed the BR data analysis as well as the within-subjects and correlational analysis. M.W. drafted the manuscript, and all authors provided critical revisions. All authors approved the final manuscript for submission.

Competing interests

We declare we have no competing interests

Funding

This work was supported by Australian National Health and Medical Research Council grants APP1024800, APP1046198 and APP1085404, J. Pearson's Career Development Fellowship APP1049596 and an Australian Research Council discovery project grant DP140101560.