Interoceptive awareness is associated with enhanced neural response to unpleasant images

Abstract

Interoception, or perception of internal bodily signals, is thought to be an important contributor to emotional experience and cognition, and individual differences in interoceptive beliefs may affect cognitive processing of emotional information. Addressing this issue, the present study examined the correlation between self-reported interoceptive awareness and EEG markers of motivated attention to emotional versus neutral images. A sample of undergraduates (n = 115) viewed images of neutral, pleasant, and unpleasant scenes while EEG was recorded, and the late positive potential (LPP) was extracted for each image type. Group results replicate prior findings of enhanced LPP signals for pleasant and unpleasant compared to neutral images. Moreover, a greater degree of enhanced LPP for unpleasant images (versus pleasant or neutral images) was positively correlated with self-reported interoceptive awareness. Results support the role of interoceptive beliefs in cortical processing of emotional information.

Introduction

Psychologists have long posited that bodily states play an important role in emotional experiences and perception (James 1894, Schachter and Singer 1962). Signals from the body, such as cardiovascular changes, respiration, muscle tension, skin sweatiness, and other cues, are thought to contribute to the mental representation of emotional states. Bidirectional feedback between bodily reactions and cognitive representations of the external world allows us to react to emotionally salient information and to modify our representations according to information from the body. In recent years, theoretical perspectives on interoception, or the perception of internal body states, have described neuroanatomical pathways and cognitive processes that may contribute to interoception (e.g. Seth 2013; Critchley and Garfinkel 2017; Feldman et al. 2024). Current models highlight the complexity of interoception across multiple levels of processing, from peripheral sensation through higher levels of cortical processing and interpretation of bodily signals (Greenwood and Garfinkel 2025).

Researchers have attempted to quantify aspects of interoception that may differ across people, especially given that individual differences in interoception are relevant to understanding vulnerability to affective disorders such as anxiety disorders (e.g. Domschke et al. 2010, Mayer-Benarous et al. 2025, Verdonk et al. 2025). Some studies have examined performance-based measures, sometimes referred to as “interoceptive accuracy” tasks, in which participants must, e.g. detect their own heartbeat (Ferentzi et al. 2022). These measures can be considered objective in the sense that the participants’ reports can be compared to actual bodily signals to determine accuracy in perceiving those signals. However, these methods have been criticized because, among other drawbacks, they typically focus on only one bodily cue (such as heartbeats). Recent studies have found that the ability to accurately perceive one internal signal (e.g. heartbeat) does not predict accuracy in perceiving another (e.g. respiration), undermining the idea of a unitary interoceptive accuracy ability (Schoeller et al. 2025; for other critiques, see also Ring and Brener 2018, Ferentzi et al. 2022, Körmendi et al. 2023).

Another approach has been to develop questionnaire measures that assess participants’ own evaluations of their interoceptive sensitivity or awareness (e.g. Mehling et al. 2012, 2018, Todd et al. 2022). While such questionnaires have the strength of addressing a more global construct of sensitivity to bodily signals rather than narrowing to a particular bodily cue, they depend on self-report, which is difficult to validate against putatively objective criteria (Desmedt et al. 2022). Some studies have found that various interoceptive questionnaire measures do not correlate strongly with one another (Vig et al. 2022) or with performance-based interoceptive tasks (Calì et al. 2015, Garfinkel et al. 2015). These findings imply that beliefs about one’s own interoceptive tendencies may be distinct from the accurate perception of bodily signals.

Investigating whether questionnaire measures of interoceptive beliefs correlate with neural responses to emotional information could potentially strengthen the case that interoceptive awareness is an important aspect of individual differences relevant to emotional information processing. The present report aims to address this question using a neural signal of emotion processing derived from EEG measures. Numerous studies have found that when people perceive information that is emotionally salient, such as viewing pleasant or unpleasant pictures, the event-related potential measured with scalp EEG exhibits a sustained positivity, known as the late positive potential (LPP), that reflects attentional engagement with that emotional information (Cuthbert et al. 2000, Hajcak and Foti 2020). The LPP is most evident over posterior scalp regions and reflects widespread activation in a range of cortical regions, including visual cortex and insular cortex, as shown by joint EEG-fMRI studies (Liu et al. 2012, Sabatinelli et al. 2013). An open question is whether interoceptive ability or awareness is related to the LPP signal of emotional engagement.

Supporting a potential link between interoception and the LPP response to emotional information, two early studies (Herbert et al. 2007, Pollatos et al. 2007) found that participants who performed better on a heartbeat detection task showed elevated LPPs to emotional images, compared to those who performed worse at the task. However, these studies were conducted on relatively small samples for the examination of individual differences (ns 30 − 40) and they focused on interoceptive accuracy tasks specific to heartbeat estimation (see also Füstös et al. 2013). A more recent study experimentally primed interoceptive attention through a task that involved tapping to either one’s own heartbeat (interoception) or an external audio recording of a heartbeat (control) and found that interoceptive priming led to an elevated LPP response to emotional images (Chen et al. 2024). This recent study was not focused on individual differences but contributes to the evidence that interoceptive attention can modulate the neural response to emotional information as measured by the LPP.

The goal of the present investigation was to extend these prior findings by examining the emotion-related LPP in individuals differing in self-reported interoceptive awareness. We addressed this question using an existing dataset from our laboratory in which a relatively large sample of participants completed an image-viewing task designed to elicit the LPP. Participants also completed a set of self-reported questionnaires that included the Multidimensional Assessment of Interoceptive Awareness (MAIA; Mehling et al. 2012), which has been identified among self-report measures as best representing the core component of interoceptive awareness according to psychometric measures (Todd et al. 2022). While other aspects of the dataset are reported elsewhere (Compton et al., under review), the present report is the first to address the relationship between individual differences in self-reported interoceptive awareness and cortical sensitivity to emotionality as measured by the LPP.

Materials and methods

Participants

One hundred twenty undergraduate students were recruited from the Haverford–Bryn Mawr college communities. The goal of n = 120 participants was selected because sample-size calculators indicate that a sample of n = 123 can detect a correlation of r = .25 with 80% power. Mean age was 19.8 years (range 18–24), and the sample included 92 women, 20 men, and 7 non-binary participants (summing to 119 due to missing self-report data from one participant). Seventy-two identified as White, 30 as Asian, 10 as Black, 18 as Hispanic/Latinx, 1 as Native American, and 4 as Other, with numbers totaling more than the sample size because 17 participants indicated more than one racial/ethnic identification. All procedures were approved by the Haverford College IRB.

Task

Results reported here are from a passive viewing task, in which 90 images were presented in random order for 1000 ms, with an inter-trial interval consisting of a 1500-ms blank screen. Participants were instructed simply to view the images as they normally would. (An emotion regulation block was also included in the experiment, as reported in Compton et al., under review, but its results did not interact with individual differences in the MAIA.) A total of 30 neutral, 30 pleasant, and 30 unpleasant images were included.

Images were selected from three existing emotional image databases: Nencki Affective Picture System (Marchewka et al. 2014), Open Affective Standardized Image Set (Kurdi et al. 2017), and Geneva Affective Picture Database (Dan-Glauser and Scherer 2011). Candidate images were rated on seven-point scales for valence and arousal by a separate sample of 30 online participants via Prolific. We then retained pleasant images with a mean valence rating >5; unpleasant images with a mean valence rating <3; and neutral images with a mean valence rating between 3 and 5. We controlled across sets for the presence of people in images, such that the neutral, pleasant, and unpleasant sets each included eight images that featured people (seven images with visible faces), as some research suggests enhanced LPPs in response to images of people (Ferri et al. 2012). A list of the images used and their sources is available in Supplementary Table 1.

For the final set of 90 images, mean valence on a seven-point scale was 4.06 for neutral images, 6.00 for pleasant images, and 2.17 for unpleasant images [one-way ANOVA, F(2,87) = 616, P < .001, all pairwise comparisons P < .001 by Tukey’s HSD]. Mean intensity (arousal) on a seven-point scale was 2.27 for neutral images, 2.89 for pleasant images, and 4.66 for unpleasant images [one-way ANOVA, F(2,87) = 57.4, P < .001; pairwise comparisons by Tukey’s HSD, unpleasant > neutral and unpleasant > pleasant, Ps < .001; pleasant > neutral, P = .025).

Self-report measures

To assess interoceptive awareness, participants completed the MAIA questionnaire (Mehling et al. 2012). The MAIA includes 32 items, each of which is rated on a scale from “never” (score = 0) to “always” (score = 5). Items are clustered into eight subscales, including Noticing (e.g. “When I am tense, I notice where the tension is located in my body”), Not-Distracting (e.g. “I distract myself from sensations of discomfort”; reverse-scored), Not-Worrying (e.g. “I start to worry that something is wrong if I feel any discomfort”; reverse-scored), Attention Regulation (e.g. “I can maintain awareness of my inner bodily sensations even when there is a lot going on around me”), Emotional Awareness (e.g. “I notice how my body changes when I am angry”), Self-Regulation (e.g. “When I feel overwhelmed I can find a calm place inside”), Body Listening (e.g. “I listen for information from my body about my emotional state”), and Trusting (e.g. “I feel my body is a safe place”). Cronbach’s α for the full scale was 0.90.

In addition to the MAIA, participants also completed several other questionnaires as part of the larger study (Compton et al., under review): Pediatric ACEs and Related Life-Events Screener (PEARLS) scale for teens/adults (Koita et al. 2018); Center for Epidemiological Studies-Depression (CESD) scale (Radloff 1977); Perceived Stress Scale (PSS; Cohen et al. 1994); Behavior Rating Inventory of Executive Functioning (BRIEF; Roth and Gioia 2005); Positive Childhood Experience survey (PCE; Bethell et al. 2019); and Predictability, Opportunity, and Safety Dimensions of Environmental Variability (POSDEV) measure (Lopez et al. 2026). These measures are more fully reported in Compton et al. (under review).

EEG data acquisition and processing

EEG data were collected continuously during the passive viewing task. Data were sampled at 1024 Hz from Ag/AgCl sensors embedded in a Compumedics neoprene cap, amplified by a Grael amplifier, and recorded with Curry software. Sites included F3, Fz, F4, C3, Cz, C4, CP3, CPz, CP4, P3, Pz, P4, and the left and right mastoid sites. Blinks and eye movements were monitored by bipolar vertical and horizontal electrooculogram electrodes. At the time of acquisition, data were referenced to a separate central site on the cap (labeled REF in the Curry system).

Offline, continuous data were filtered with a 0.1–30 Hz bandpass filter and re-referenced to the average of the left and right mastoid sites. Blink artifact reduction was accomplished for most participants using Curry’s time domain filter applied to the VEOG lead, except for nine participants with faulty VEOG recordings, for whom the F3 site was used instead to build a template of the blink artifact and apply Curry’s principal-components-based reduction. For seven participants, a single bad channel was interpolated based on the two nearest sites. Visual inspection of each file was used to identify and reject segments with any remaining gross artifacts.

Continuous data were epoched from −200 to 1000 ms around the onset of the neutral, pleasant, and unpleasant images. Epochs were baseline-corrected with the 200-ms pre-image window used to define the baseline, and epochs were averaged separately for each trial type. The number of epochs included in averages was M = 28.5 for neutral, M = 28.5 for pleasant, and M = 28.7 for unpleasant images. The LPP for each individual and image type was quantified as the mean amplitude between 500 and 1000 ms. The 12 scalp sites were reduced to four regional clusters (frontal, central, central-parietal, and parietal) by averaging across the 3 sites in each region. Thus, each participant had four regional LPP values for each of three image types.

Data exclusions

For three participants, EEG data were unusable due to excessive artifacts or experimenter error, leaving 117 complete EEG datasets for analysis. Furthermore, responses on the MAIA were missing for 2 of the 120 participants. Therefore, for analyses examining the relationship between the MAIA and LPP responses, the final sample size was n = 115.

Results

Self-report measures

Inter-relationships among MAIA subscales

Table 1 presents correlations among the eight subscales of the MAIA. The number of significant positive correlations suggests a high degree of overlap in the constructs assessed by the subscales. Because the key research question did not differentiate among subscales, we conducted a principal components analysis (with varimax rotation) to extract a global MAIA score. The results of that analysis are presented in Table 2. Six of the eight scales load on a common factor, which accounted for 42.2% of the variance, similar to factor analysis results reported by other researchers (Ferentzi et al. 2021, Todd et al. 2022). We extracted individual scores on the first principal component and refer to these scores as the MAIA global score in subsequent analyses.

Table 1.

Intercorrelations among MAIA subscale scores.

Subscale	1.	2.	3.	4.	5.	6.	7.	8.
1. Body listening	—
2. Emotional awareness	0.684**	—
3. Self-regulation	0.658**	0.551**	—
4. Noticing	0.488**	0.571**	0.356**	—
5. Trusting	0.419**	0.403**	0.545**	0.267*	—
6. Attention regulation	0.428**	0.352**	0.428**	0.371**	0.293*	—
7. Not-distracting	0.139	0.041	0.105	0.114	0.257*	0.107	—
8. Not-worrying	−0.201*	−0.150	−0.108	−0.236*	0.090	0.047	0.028	—

^* P < .05;

^** P < .001.

Table 2.

Results of principal components analysis on MAIA subscales.

Subscale	Loading
Body listening	0.851
Emotional awareness	0.812
Self-regulation	0.804
Noticing	0.689
Trusting	0.642
Attention regulation	0.614
Not-distracting	(0.226)
Not-worrying	(−0.188)

Relationship between MAIA and other self-report measures

Table 3 summarizes the correlations between the MAIA global score and other measures administered as part of the larger study. The general pattern suggests that interoceptive awareness is inversely correlated with the scales indexing negative attributes (adversity, depression, stress, and executive dysfunction) and positively correlated with the scales indexing positive childhood experiences.

Table 3.

Correlations between MAIA global score and other self-report measures.

Scale	Pearson’s correlation	Probability
PEARLS (adversity)	r = −0.179	p = .056
CESD (depression)	r = −0.247	p = .008
PSS (perceived stress)	r = −0.180	p = .054
BRIEF (executive dysfunction)	r = −0.263	p = .005
PCE (positive childhood events)	r = 0.382	p < .001
POSDEV (positive childhood events)	r = 0.202	p = .031

LPP results

Overall group effects

We first report the overall ANOVA results from the LPP data to characterize the results in the group as a whole, before turning to the MAIA effects. These group results have also been reported in Compton et al. (under review). Figure 1 displays the scalp maps and an illustrative waveform from the Pz site, reflecting typical LPP distribution and pattern across image types.

Figure 1.

Scalp maps (from 500 to 1000 ms window) and sample waveform (from Pz site). Patterns demonstrate typical LPP responses to emotional image content.

The 3 × 4 repeated-measures ANOVA included the factors Image Type (neutral, pleasant, and unpleasant) and Region (frontal, central, central-parietal, and parietal). Effects followed the expected pattern for the LPP, as both main effects and their interaction were significant [Image Type, F(2,232) = 9.47, P < .001, η²_p = 0.075; Region, F(3,348) = 131.58, P <.001, η²_p = 0.531; Image Type × Region, F(6,696) hx2009;= 69.49, P < .001, η²_p = 0.375]. Means for the interaction are presented in Table 4 and reflect the anterior–posterior gradient of the LPP and the differentiation of neutral versus emotional images (greater LPP for emotional than neutral images) especially in posterior regions. For the parietal region, where LPP is typically maximal, the main effect of Image Type [F(2,232) = 36.90, P < .001, η²_p = 0.241] reflected that LPP was highest for unpleasant, intermediate for pleasant, and lowest for unpleasant images (all pairs significantly differ, Tukey’s post hoc, Ps < .005; means in Table 4). These data statistically support the patterns evident by visual inspection of Fig. 1 and replicate the expected effect of emotionality on the LPP.

Table 4.

Mean LPP amplitudes (in µV) as a function of image type and region.

Region	Image type			Emotionality effect
	Neutral	Pleasant	Unpleasant	P–N	U–N
Frontal	−8.38	−7.18	−10.20	1.20	−1.82
Central	−6.68	−5.37	−6.10	1.31	0.58
Central-parietal	−5.06	−3.70	−3.71	1.36	1.35
Parietal	−2.31	−0.90	−0.06	1.41	2.25

Emotionality effects are subtractions between pleasant (P), unpleasant (U), and neutral (N) image types.

Effects involving MAIA global score

To address the relationship between MAIA and the effects described above for the group, we added the MAIA global score to the ANOVA reported above as a covariate, i.e. conducting an ANCOVA with Image Type and Region as factors and MAIA as covariate. A significant Image Type × MAIA score interaction emerged [F(2,226) = 6.73, P < .001, η²_p = 0.056], but no other effects involving the MAIA score were significant (Ps > .35).

The Image Type × MAIA interaction that emerged from the ANCOVA indicates that individual differences in MAIA scores predicted the effect of Image Type on the LPP but did not interact with Region. To determine the contribution of each of the image types to the significant Image Type × MAIA interaction effect, we conducted a multiple regression predicting MAIA global scores based on the mean LPP for neutral, pleasant, and unpleasant images (averaged across regions) entered as three separate predictors [full model R² = .095, F(3,111) = 3.87, P = .011]. When entered simultaneously, LPPs for unpleasant images predicted higher MAIA scores (t = 2.943, P = .004), whereas LPPs for neutral images predicted lower MAIA scores (t = −2.503, P = .014) and LPPs for pleasant images did not significantly relate to MAIA scores (t = −0.476, P = .635).

Further illustrating this pattern, in separate correlations we found that MAIA global scores were significantly and positively associated with the unpleasant LPP minus the neutral LPP (r = 0.298, P = .001) and with the unpleasant LPP minus the pleasant LPP (r = 0.204, P = .029), but not with the pleasant LPP minus the neutral LPP (r = 0.150, P = .109). Scatterplots, including estimated slopes and confidence intervals, are depicted in Fig. 2. In summary, participants with higher MAIA scores, indexing greater interoceptive awareness, exhibited stronger LPPs to unpleasant images relative to the other two image types.

Figure 2.

Scatterplots depicting relationship between MAIA interoception scores and LPP emotionality effects. Shaded region reflects 95% confidence interval.

The above analyses examined the MAIA scores as a continuous variable, which is most statistically sensitive and appropriate for a scale in which scores vary continuously. For purposes of illustration, we also divided participants by a median split on MAIA global scores. Means for the relevant interaction are depicted in Fig. 3 and indicate that the LPP response to unpleasant images best differentiated High from Low MAIA participants. This pattern is also illustrated in the waveforms depicted in Fig. 4, in which the difference between the LPP for unpleasant versus neutral images is greater for High versus Low MAIA participants.

Figure 3.

Mean LPP amplitude (in µV) across three image types for participants divided by median split on MAIA global scores. Error bars represent SEM.

Figure 4.

Grand-average waveforms at the Pz site for High and Low MAIA groups. Groups are based on median split of MAIA global scores.

Examination of other individual differences variables

Because scores on the other self-report measures were correlated with the MAIA (see Table 3), we examined whether any of these scores were significant covariates in the analysis of the LPP amplitudes to rule out possible confounds. Each scale was entered as a covariate in the Image Type × Region ANOVA, as we had done for the MAIA. Results of the critical Image Type × scale interaction are listed in Table 5, indicating that none of the scales predicted the LPP as the MAIA scores had. The other scores also did not emerge as main effect predictors nor did they interact in any other effects in the ANCOVA. These results rule out the possibility that the MAIA effect on the LPP is explained by one of the other measured variables.

Table 5.

Summary of F-test results for image type × scale interaction in ANCOVA on LPP amplitudes.

Scale	Test statistics for image type × scale	Effect size (η2p)
MAIA (interoception)	F(2, 226) = 6.73, p < .001	0.056
PEARLS (adversity)	F(2, 226) = 0.38, p = .714	0.003
CESD (depression)	F(2 ,228) = 1.53, p = .219	0.013
PSS (perceived stress)	F(2, 228) = 0.97, p = .382	0.008
BRIEF (executive dysfunction)	F(2, 228) = 0.73, p = .481	0.006
PCE (positive childhood events)	F(2, 228) = 1.96, p = .143	0.017
POSDEV (positive childhood events)	F(2, 226) = 1.16, p = .316	0.010

Denominator df differs slightly between scales due to missing data from two participants for three of the scales.

Finally, when gender was entered into the ANOVA as a between-subjects factor, gender did not interact with any other terms (Ps > .20); and when gender and MAIA global score were both entered into the analysis as predictors, the MAIA global score continued to interact with Image Type [F(2,222) = 6.89, P = .001, η²_p = 0.058], indicating that gender cannot account for the MAIA effect.

Examination of MAIA subscales

As described above, our main analysis approach quantified the MAIA with a global score, derived from the PCA, because hypotheses did not differentiate among the facets of the MAIA and because of intercorrelations among the subscales (Table 1). Nevertheless, following this main analysis, we reran the ANCOVA with each subscale as a predictor in place of the global MAIA score. Results, which should be considered exploratory, are presented in Table 6. The Image Type × Subscale interaction was significant for Emotional Awareness, Self-Regulation, Noticing, and Trusting subscales; at the level of a marginal trend for Body Listening; and not significant for Attention Regulation, Not Distracting, and Not Worrying subscales.

Table 6.

Test statistics for MAIA subscales entered as covariate in LPP ANCOVA.

Subscale	Image type × subscale
Body listening	F(2, 226) = 2.87, p = .059
Emotional awareness	F(2, 226) = 7.74, p < .001
Self-regulation	F(2, 226) = 7.66, p < .001
Noticing	F(2, 226) = 3.482, p = .032
Trusting	F(2, 226) = 4.79, p = .009
Attention regulation	F(2, 226) = 0.63, p = .533
Not-distracting	F(2, 226) = 1.13, p = .325
Not-worrying	F(2, 226) = 0.001, p = .992

Discussion

This study investigated the relationship between individual differences in interoceptive awareness, as self-reported on the MAIA, and neural responses to images differing in their emotional content. Participants who reported high levels of interoceptive awareness showed greater differentiation in their neural response to emotional versus neutral images, compared to participants lower in interoceptive awareness. Results were significant when comparing responses to unpleasant versus neutral images, while pleasant-image responses showed minimal association with interoception. These results, statistically reliable in a relatively large sample for an EEG study, support the role of self-perceived bodily awareness in emotional information processing at the cortical level.

These results build upon existing research associating interoception with enhanced emotional perception using EEG methods. For example, both Herbert et al. (2007) and Pollatos et al. (2007) found that late positivities in the EEG response to emotional information distinguished between emotional and neutral information better for participants who also performed better on a heartbeat detection task, compared to those who performed more poorly. The present results extend those findings to a larger sample and to self-reported interoceptive awareness. Together, the findings indicate enhanced cortical processing of emotional content in participants with greater interoceptive capacities, despite the fact that those capacities were measured in different ways.

Researchers have distinguished between interoceptive accuracy (e.g. as measured through heartbeat detection) and interoceptive awareness, as measured through self-report, finding that these two types of measures of individual differences are not consistently correlated (e.g. Garfinkel et al. 2015). The present finding that interoceptive awareness is associated with increased neural reactivity to emotional images, similar to prior findings with interoceptive accuracy, suggests possible commonality between these aspects of interoception. However, it does not necessarily imply that awareness and accuracy are identical constructs. A future study of the LPP emotionality effect that measures individual differences in both interoceptive accuracy and self-reported beliefs about interoceptive awareness could better address whether variance in the LPP effect is independently or jointly predicted by these separable aspects of interoception. This, in turn, could advance understanding of the underlying structure and components of interoceptive ability.

In the present study, relationships between interoception and the LPP were strongest for unpleasant images, as the MAIA score predicted a stronger LPP response to unpleasant versus neutral pictures and unpleasant versus pleasant pictures, but not pleasant pictures versus neutral pictures. This raises the question of whether a response to unpleasant images is more closely tied to cues that might be detected by interoception, such as a heart rate increase or other autonomic response when viewing negative content. The negative images in the present study were rated higher in arousal than the positive images, although both were rated as more arousing than the neutral images. Moreover, the LPP was also higher for the unpleasant than pleasant images in the parietal region in the group as a whole, consistent with prior findings that highly arousing stimuli produce the highest LPP (e.g. Cuthbert et al. 2000; Leite et al. 2012) especially when they are also unpleasant (Feng et al. 2014). Based on the image sets used in the present study, we cannot determine if interoceptive awareness correlated more strongly with the LPP to unpleasant images because of those images’ negative valence, their intensity, or both. In either case, results support a role for interoception in representing emotional information, but additional research with different image sets would be needed to determine whether valence or arousal is the more important dimension of emotionality driving the effect.

The present study was correlational, so the causal pathway associating interoception and the LPP response to unpleasant images cannot be definitively addressed. Because the study did not measure peripheral responses to the picture content (e.g. heart rate, skin conductance, respiration), it is unknown whether the participants with greater self-reported MAIA scores produced stronger or more differentiated peripheral responses than those with lower scores, or whether the peripheral responses were similar but higher MAIA-scoring participants were more sensitive to them. Alternatively, interoceptive awareness may reflect beliefs about the self that are independent of the actual peripheral bodily signals. Further studies may benefit from including autonomic measures as well as the scalp-recorded LPP and interoceptive measures to better understand their inter-relationships. In addition, future studies could examine the relationship between interoceptive awareness and self-reported emotional experience during image viewing to determine if the interoception–LPP relationship in the present study reflects interoception-related differences in conscious feeling of emotional experience.

Due to the correlational nature of the present study, we cannot fully rule out whether other individual difference characteristics could account for the relationship between the LPP and MAIA scores. For example, perhaps participants scoring higher on the MAIA are more attentive to both internal and external information, contributing to better perception of the external image content even without a causal role for interoceptive awareness. Another possibility is that individual differences in anxiety may account for the findings, as anxiety may plausibly correlate with both neural responses to unpleasant information (e.g. Moser et al. 2008, Kujawa et al. 2015) and interoceptive tendencies (e.g. Domschke et al. 2010). Although we did not measure anxiety directly, our other measured scales—including depression, stress, self-reported executive functioning, and childhood experiences—did not correlate with the LPP response in the passive viewing task and therefore could not account for the MAIA–LPP relationship. Yet, it is possible that other unmeasured variables could account for the relationship, an issue that could be examined in future research. Despite these limitations, the present study adds to empirical findings relating individual differences in interoceptive awareness with the emotion-related LPP, thus supporting the hypothesized role of bodily signals in the cognitive processing of emotional information.

Author contributions

Rebecca Compton (Conceptualization [lead], Data curation [lead], Formal analysis [lead], Funding acquisition [lead], Project administration [lead], Resources [lead], Supervision [lead], Visualization [lead], Writing—original draft [lead], Writing—review & editing [lead]), Tiffany Mikulis (Conceptualization [supporting], Investigation [equal], Writing—review & editing [supporting]), Caroline Frost (Conceptualization [supporting], Investigation [equal], Methodology [supporting], Writing—review & editing [supporting]), Danylo Shudrenko (Conceptualization [supporting], Investigation [equal], Writing—review & editing [supporting]), Rosaleen Cochran (Investigation [equal], Methodology [supporting], Writing—review & editing [supporting]), Trinity Ellis (Investigation [equal], Writing—review & editing [supporting]), Aidan O’sullivan (Formal analysis [supporting], Investigation [equal], Writing—review & editing [supporting]), Tapsi Suryawanshi (Formal analysis [supporting], Investigation [equal], Writing—review & editing [supporting]), and Nadine Tang (Formal analysis [supporting], Investigation [equal], Writing—review & editing [supporting])

Supplementary material

Supplementary material is available at SCAN online.

Conflicts of interest

The authors report no conflict of interest.

Funding

This work was supported by the National Science Foundation under Grant RUI 2416630.

Data availability

Raw data and processing scripts are available at https://osf.io/m7cxg/.

Declaration of generative AI and AI-assisted technologies

No generative AI or AI-assisted technologies were used in the manuscript preparation process.