INTEROCEPTION, AFFECT, AND COGNITION IN OLDER ADULTS

Abstract

Interoception is the detection of signals that arise from within the body. Interoceptive sensitivity has been found to be associated with affect and cognition among younger adults, and examination of these relationships in older adult samples is beginning to emerge. Here, we take an exploratory approach to determine how demographic, affective, and cognitive variables relate to interoceptive sensitivity in neurologically normal older adults, aged 60–91 years old. Ninety-one participants completed a comprehensive neuropsychological battery, self-report questionnaires, and a heartbeat counting task to measure interoceptive sensitivity. Our findings revealed several relationships: 1) interoceptive sensitivity was inversely correlated with measures of positive emotionality: participants with higher interoceptive sensitivity tended to have lower levels of positive affect and trait extraversion; 2) interoceptive sensitivity was found to positively correlate with cognition: participants who performed better on the heartbeat-counting task also tended to perform better on a measure of delayed verbal memory; and 3) when examining the predictors of interoceptive sensitivity in a single hierarchical regression model, higher interoceptive sensitivity was related to: higher time estimation, lower positive affect, lower extraversion, and higher verbal memory. In total, the model accounted for 38% of the variability in interoceptive sensitivity (R² = .38). These results suggest that, among older adults, interoceptive sensitivity is facilitative for aspects of cognition but perhaps disruptive for certain aspects of emotional experience.

Perception of bodily cues, whether conscious or nonconscious, has long been implicated in the formation of emotions and in the guidance of cognition. Theories of emotion posit that interoception, or the perception of visceral states within the body, is essential to consciously experience emotion (Damasio, Everitt, & Bishop, 1994; James, 1884). The empirical study of interoception has increased in recent years, largely due to compelling evidence of its relationships to affect and cognition. While certain facets of interoception have proven elusive due to their subjective nature, interoceptive sensitivity is considered an objective measure based on an individual’s accurate detection of internal stimuli, with the literature widely regarding it as a measurable dimension of interoception (e.g., Garfinkel, Tiley, O'Keeffe, Harrison, Seth, & Critchley, 2016). Additional dimensions of interoception include interoceptive awareness, which refers to an individual’s metacognitive awareness of their own interoceptive sensitivity; and interoceptive sensibility, which reflects an individual’s subjective tendency to be internally focused (Garfinkel & Critchley, 2013).

Traditionally, interoceptive sensitivity has been measured using a heartbeat-tracking task developed by Schandry (Pollatos, Traut-Mattausch, Schroeder, & Schandry, 2007; Schandry, 1981). One way to investigate interoception has been to use interoceptive sensitivity as a trait-like variable and to examine its relationship with other complex functions, such as affect, cognition, and behavior. Because altered interoception has been demonstrated in patients with anxiety, depression, and eating disorders (Domschke et al., 2010; Eggart et al., 2019; Jenkinson et al., 2018), the majority of research to date has been devoted to the examination of interoceptive sensitivity and emotion or affective variables in clinical populations (Khalsa et al., 2018). For example, numerous authors have demonstrated a relationship between interoceptive sensitivity and anxiety, where individuals with higher levels of interoceptive sensitivity have higher levels of anxiety (Critchley, Wiens, Rotshtein, Ohman, & Dolan, 2004; Dunn, Stefanovitch, Evans, Oliver, Hawkins, & Dalgleish, 2010). Notably, Pollatos et al. (2007) demonstrated that the relationship between trait anxiety and emotional arousal was mediated by interoceptive sensitivity, suggesting that interoception plays a role in the pathophysiology of anxiety. Whereas high levels of interoceptive sensitivity mediate increased arousal, low levels are deactivating. Dunn, Dalgleish, Ogilvie, and Lawrence (2007) demonstrated a relationship between interoceptive sensitivity and depression, where individuals with moderate symptoms of depression tend to have lower levels of interoceptive sensitivity. Interestingly, high levels of anhedonia seem to attenuate the relationship between interoceptive sensitivity and anxious arousal (Dunn et al., 2010). A recent meta-analysis on alexithymia and interoception by Trevisan et al. (2019) indicated an inverse relationship between interoception and understanding one’s own emotions, but only in clinical populations. In non-clinical populations, interoceptive accuracy is associated with better emotion regulation (Füstös, Gramann, Herbert, & Pollatos, 2012).

Additional studies have linked interoception with cognition. For example, Werner, Jung, Duschek, and Schandry (2009) identified a relationship between interoceptive sensitivity and performance on the Iowa Gambling Task (IGT). In a dichotomized sample of individuals with either “good heartbeat perception” or “bad heartbeat perception,” those with good heartbeat perception tended to choose significantly fewer cards from disadvantageous decks in the IGT. Correlations also demonstrated a positive relationship between interoceptive sensitivity and mean number of advantageous card selections, and an inverse relationship between interoceptive sensitivity and mean number of disadvantageous card selections. The results suggest a relationship between interoceptive sensitivity and strong decision-making processes. In the memory domain, Umeda, Tochizawa, Shibata, and Terasawa (2016) demonstrated a significant positive relationship between interoceptive sensitivity and performance on a prospective memory task (i.e., memory for future intentions). Notably, Stevenson et al. (2018) found that interoceptive sensitivity was associated with measures of hippocampal dependent learning and memory—the Rey Auditory Learning Test and Logical Memory. Together, these studies demonstrate a relationship between interoception and several facets of cognition, with interoceptive sensitivity tending to correlate positively with cognitive function.

The aforementioned studies examined interoceptive sensitivity in largely younger adult samples of participants, yet interoceptive sensitivity and its association with emotion and cognition, may be altered by aging, given the fact that multiple sensory modalities become less sensitive across the lifespan. In a study of 2939 older adults from 57 to 85 years of age, only 5.9% did not demonstrate any sensory deficits. The most common impairment was a loss of taste (74%), but 32% of participants had poor sense of touch, and up to 67% of older adults demonstrated impaired function of two or more senses (Correia et al., 2016). Visceral pain has also been shown to decrease with age (Lasch, Castell, & Castell, 1997). Several mechanisms have been proposed for loss of interoceptive sensitivity with aging, for example, cortical loss in brain regions responsible for interoception, reduced peripheral sensation related to nerve axon or myelin loss, and reduced attentional capacity (Khalsa et al., 2009). Each of these changes has individually been shown to occur with aging (Andres, Parmentier, & Escera, 2006; Hanewinckel, van Oijen, Ikram, & van Doorn, 2016; Salat et al., 2004; Resnick, Pham, Kraut, Zonderman, & Davatzikos, 2003). In younger patients, interoceptive sensitivity is correlated with activity of the insular and orbitofrontal cortices, and the findings of Ueno et al. (2020) suggest the salience network is necessary to maintain interoceptive sensitivity in older adults.

A review paper examining interoception across the lifespan (Murphy, Brewer, Catmur, & Bird, 2017) indicated a paucity of literature investigating interoception in adults older than 60 years of age. A small number of authors have begun to address this gap in the literature.

The first study to examine interoceptive sensitivity in aging was conducted by Khalsa, Rudrauf, and Tranel (2009), who demonstrated an inverse relationship between heartbeat detection ability and age, suggesting that interoceptive sensitivity declines with age. However, application of this study to older adults is limited as it included only six individuals over age 60, with the oldest participant being 63 years of age. Murphy and colleagues (2018) evaluated interoceptive sensitivity in adults 20–90 years of age. Theirs was the first article to use an objective measure of interoceptive sensitivity in a large cohort of older adults; it included 136 total participants with 61 participants aged 60 years or older. Interoceptive sensitivity was evaluated using the Schandry task. They identified a significant negative correlation between interoceptive sensitivity and age, and using a mediation analysis showed that age directly affected interoceptive sensitivity. Based on these results, they concluded that the effect of age on interoceptive sensitivity was independent of other variables.

More recent research has investigated the role of aging and interoceptive sensitivity among older adults, but has been limited to in scope to emotionality and metacognition. For example, Mikkelsen et al. (2019) found that younger adults who performed well on measures of interoceptive sensitivity showed significantly lower emotional reactivity, while older adults showed no difference, suggesting that age is a factor in the recognition of internal states and reactions to aversive stimuli. A neuroimaging study of older adults demonstrated that higher interoceptive sensitivity and emotional understanding (an aspect of emotional intelligence) was associated with preserved white matter (Dobrushina et al., 2020). MacCormack et al. (2021) found emotional experience to be less associated with interoception for older adults as compared to younger populations. Changes related to metacognition, necessary for interoceptive sensibility (i.e., subjective beliefs of interoceptive accuracy) were found to be negatively associated with self-estimates of cognition in older adults (Kamp et al., 2021). While the aforementioned studies (Dobrushina et al., 2020; Kamp et al., 2021; MacCormack et al., 2021; Mikkelsen et al., 2019) provide more insight in regard to interoception and aging, they fail to address additional variables of interest in the affective and cognitive domains of a general, older population. Given the significant body of research on interoception and aberrant emotion in clinical populations (e.g. Dunn, Dalgleish, Ogilvie, & Lawrence, 2007; Khalsa et al., 2018; Pollatos et al., 2007) affective variables of interest include emotional valence and personality. Within the cognitive domain, variables of interest include language, memory, and attention, among others.

The present study involved an exploratory examination of the correlates of interoceptive sensitivity in older adults. More specifically, we examined how demographic (i.e., age, sex, education, body mass index (BMI)), affective (i.e., “Big Five,” positive affect, negative affect), and cognitive (i.e., intelligence, language, memory, attention and concentration, psychomotor speed, executive function) variables relate to interoceptive sensitivity, in a large sample of healthy older adults. To our knowledge, previous research has not examined all three domains of interoceptive sensitivity, affect, and cognition in a comprehensive analysis with healthy older adults.

Methods

Participants

Ninety-three participants were recruited from an existing registry of 120 older adults in eastern Iowa. The registry was begun in the early 2000s and as such there has been attrition as a result of relocation and mortality. Secondary to this attrition, the registry replenished itself with an aggressive recruitment in 2015. Participants in the current study were independently living, community dwelling, cognitively healthy older adults with no history of neurological or psychiatric disease as determined secondary to extensive clinical interview (after Tranel, Benton, & Olson, 2009). All participants completed the heartbeat counting task, time estimation task, self-report questionnaires, and a comprehensive neuropsychological battery. Two participants were excluded: one who was non-compliant with the task (resulting in performances that were greater than three standard deviations from the mean performance) and one who was missing neuropsychological data. The remaining 91 participants (52% female) were included in all statistical analyses. Participant age ranged from 60 to 91 years old (Mean_age = 75.16, SD = 6.34; Median_age = 75 years). Demographic, affective, and cognitive characteristics of the sample are presented in Table 1, which indicate normal to above normal, age-appropriate cognitive abilities per normative data (Lezak, Howieson, Bigler, & Tranel, 2012). In addition to neuropsychological data being classified as normal (no measure fell greater than 1.0 SD below age-appropriate norms), scores on the Mini-Mental State Exam (MMSE; Folstein et al., 1975) were near ceiling level (M = 29.2, SD = 1.0; range 27–30).

Table 1.

Demographic, affective, and cognitive characteristics of the sample.

Category	Measure Mean (SD)
Demographics	N	91
	Age	75.2 (6.3)
	Sex	52% Female
	Education	16.4 (2.7)
	BMI	28.5 (5.6)
Affective Self-Report	PANAS-X
	Positive Affect	32.6 (6.8)
	Negative Affect	11.6 (2.3)
	BFI
	Extraversion	3.4 (0.8)
	Agreeableness	4.1 (0.6)
	Conscientiousness	4.1 (0.6)
	Neuroticism	2.4 (0.7)
	Openness	3.7 (0.7)
Mental Status/Intellect	Full Scale IQ	116.6 (12.8)
	Verbal IQ	120.9 (11.4)
	Performance IQ	120.9 (13.7)
	MMSE	29.2 (1.0)
Language	BNT	57.2 (2.8)
Memory	AVLT Delay	10.1 (2.9)
Attention/Concentration	Digit Span Total	18.3 (4.4)
	BVRT errors	3.7 (2.1)
Psychomotor Speed	TM A (Time)	30.7 (8.7)
Executive Function	TM B (Time)	72.6 (23.6)

Note: BMI = Body Mass Index; PANAS-X = Positive and Negative Affect Schedule, Expanded Form; BFI = Big Five Inventory; IQ = Intelligence Quotient; MMSE = Mini Mental State Exam; BNT = Boston Naming Test; AVLT = Rey Auditory-Verbal Learning Test; BVRT = Benton Visual Retention Test; TM = Trail Making Test

Procedure

Participants completed two independent visits. The first visit (Visit #1; 2 hours in duration) consisted of a comprehensive neuropsychological evaluation for cognitive characterization and to confirm that participants were cognitively healthy. The second visit (Visit #2; 1 hour in duration) consisted of the two laboratory tasks to measure interoceptive sensitivity, as well as the completion of self-report instruments measuring affect. Informed consent was obtained for all testing procedures.

Visit #1: Neuropsychological Battery

Intelligence.

Intellectual ability was measured using the Wechsler Abbreviated Scale of Intelligence – Second Edition (WASI-II; Wechsler & Hasio-pin, 2011). Verbal intelligence (Verbal IQ) was assessed with the Vocabulary and Similarities subtests. In the Vocabulary subtest, participants are required to define words of increasingly difficulty. In the Similarities subtest, participants are presented with two related words and asked to indicate how the two words are alike or similar. Performance IQ was assessed with the Block Design and Matrix Reasoning subtests. The Block Design subtest consists of several designs that the participant recreates using red and white colored blocks. The Matrix Reasoning subtest has the participant select one design from a five-alternative forced choice array to complete an incomplete pattern matrix. Full scale intelligence (IQ) was calculated from all four of the aforementioned subtests.

Language.

Language abilities were measured using the Boston Naming Test (BNT; Kaplan, Goodglass, & Weintraub, 2001), a 60-item test of confrontation naming abilities in which participants are shown line drawings, one at a time, and given 20 seconds to correctly name each object. The items consist of commonly encountered objects (e.g., bed) and less commonly encountered objects (e.g., abacus).

Memory.

Anterograde verbal memory was measured using the Rey Auditory-Verbal Learning Test (AVLT; Schmidt, 1996), which begins with five learning trials of a 15-word list. During each trial, the list is read aloud to the participant and he/she subsequently recites as many of the list words as possible. After a 30-minute incidental delay, participants are again prompted to recite as many of the list words as possible, and this served as the memory variable of interest.

Attention and Concentration.

Attention and concentration were measured with both the Digit Span subtest from the Wechsler Adult Intelligence Scale – Fourth Edition (WAIS-IV; Wechsler, 2008) and the Benton Visual Retention Test (BVRT; Benton, 1946). The Digit Span test consists of two parts that are summed together into a total score. In the Digits Forward subtest, participants are read a number sequence and must recite the sequence directly as given. The Digits Backward subtest has participants recite the number sequence in reverse order. These subtests assess attention and concentration, respectively. The BVRT assesses visual attention/retention; 10 images containing one or more figures are individually presented for a 10-second time period, after which they are covered and immediately reproduced from memory by the participant. BVRT responses are scored with a number correct and a number of errors made, and the latter was utilized for data analysis.

Psychomotor Speed.

The Trail Making Test Part A (TMT-A; Reitan, 1958) assesses psychomotor speed. In the task, participants are instructed to draw lines connecting consecutively numbered circles on a sheet of paper as fast as they can without making any errors. The examiner points out any errors made, and redirects the participant to the prior correct circle to resume. The score is the total number of seconds required to complete the task, with time taken for self-corrections as the penalty for committing errors.

Executive Functioning.

Executive functioning was assessed using the Trail Making Test Part B (TMT-B; Reitan, 1958), a set-shifting task in which participants must alternate drawing lines between circles containing consecutive numbers and letters (e.g. 1, A, 2, B). Similar to TMT-A, participants are instructed to complete the task as quickly as they can without making any errors. Errors are identified and corrected as in TMT-A. The score is the total time in seconds required to complete the task, again with time taken for self-corrections as the penalty for committing errors.

Visit #2: Interoceptive Sensitivity and Affective Self-Report

Electrodes were placed on the participant’s right mid-clavicle and left flank. The participant then remained seated for approximately 5 minutes before starting either the heartbeat counting task or the time estimation task. Electrocardiogram (ECG) readings were taken from the time electrodes were placed until the completion of both tasks, digitized at 1000Hz sampling rate using the BIOPAC MP150, ECG100C amplifier, and AcqKnowledge data acquisition software.

Heartbeat Counting Task.

Interoceptive sensitivity was assessed using the mental tracking method originally described by Schandry (1981). Participants were not allowed to take their pulse or perform any body manipulations that would facilitate heartbeat counting during the task; they were instructed to remove their wristwatch prior to beginning the task and to place their hands palms down on the table in front of them as they performed the task. A script in E-Prime® was used to deliver a standardized set of instructions as well as the counting task. Participants were instructed to silently count the number of times their heart beat during three randomized intervals of 25 seconds, 35 seconds, and 45 seconds, but were not made aware of the duration of the intervals. The beginning of the counting phase was initiated by the participant striking a key on the computer keyboard, and was signaled by an audible tone and presentation of a green “COUNT” cue. The counting phase was ended by another audible tone and presentation of a red “STOP” cue. Participants recorded their response within the E-prime® experiment environment. The actual number of heartbeats was measured in AcqKnowledge by locating ECG complex boundaries and then counting QRS peaks. Interoceptive sensitivity was calculated as a mean score across the three testing intervals as by Koch and Pollatos (2014) using the formula:

$$1/3\Sigma \left[1-\left(\left|recordedheartbeats–countedheartbeats\right|/\left(recorded\text{ heartbeats}\right)\right)\right]$$

A perfect score of 1 on this task represents an individual counting every heartbeat that is identified by ECG monitoring.

Time Estimation Task.

To determine whether measured interoceptive sensitivity was related to participants’ ability to estimate time, a time estimation task was administered as has been done in prior studies (Dunn et al., 2007; Dunn et al., 2010; Umeda et al., 2016). This was accomplished using an experiment in E-Prime® identical to the heartbeat counting task. Participants were instructed to silently count the number of seconds in three randomized intervals of 23 seconds, 40 seconds, and 56 seconds. Participants were randomized to performing either the time estimation task or the heartbeat counting task first to mitigate any task order effects. A Time Estimation Index (TEI) was calculated using the formula:

$$1/3\Sigma \left[1-\left(\left|\text{actua time}–\text{counted time}\right|/\left(\text{actual time}\right)\right)\right]$$

A perfect score of 1 on this task represents an individual accurately counting every second that passes.

Affective Self-Report.

The Positive and Negative Affect Schedule – Expanded Form (Watson & Clark, 1994) is a 60-item self-report measure of emotional valence, or the amount of positive and negative emotions that an individual experiences. Items on the PANAS-X are scored on a five-point Likert scale. The PANAS-X can be given with different time instructions to reflect emotional experience either in the present moment or over a period of time. For our purposes, we instructed participants to respond with regard to how they feel generally. For our analyses, we utilized the positive and negative affect scale scores.

The Big Five Inventory (BFI; John & Srivastava, 1999) is a 44-item self-report measure of the Big Five personality traits: Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Openness. It is scored on a five-point Likert scale from 1–5, with the total score for each domain presented as a mean score for its respective items. High trait Extraversion suggests high sociability, positive emotionality, and an energetic approach to social interactions, whereas individuals with low trait extraversion are typically less assertive and less outgoing in social situations. Individuals with high trait Agreeableness tend to be more altruistic, trusting, and modest, whereas individuals who score low on Agreeableness are generally more self-interested. Individuals with high Conscientiousness follow rules and social norms, plan ahead, and think before they act, whereas those with low Conscientiousness tend to be more carefree and less likely to consider the consequences of their actions. Neuroticism reflects an individual’s tendency to experience negative emotionality; individuals with high trait neuroticism experience more anger, sadness, and guilt, whereas those with low trait Neuroticism have a more stable emotional experience. High trait Openness suggests an individual is likely to be more creative and accepting of new experiences, whereas those with low Openness tend to prefer routine.

Statistical Analyses

Because this study was exploratory, no adjustment was made for multiple testing and .05 was used as the level of significance. Preliminary analyses were conducted to examine the data for the presence of outliers and the appropriateness of assumptions of linearity, independence of errors, and multicollinearity. Two-tailed partial correlations were then conducted to examine the associations between interoceptive sensitivity and demographic, affective, and cognitive factors, while controlling for performance on the Time Estimation Task (after Dunn et al., 2010). Next, to control for inflation of the experiment-wise Type I error, we used the Benjamini-Hochberg procedure (Hochberg, 1998) to account for multiple comparisons. After examining the partial correlations and the Benjamini-Hochberg findings, hierarchical linear regression analyses were utilized to explicate the amount of variance accounted for by demographic, affective, and cognitive variables in predicting interoceptive sensitivity, as well as the predictive power of the model overall. This analysis allowed for examination of which affective and cognitive variables relate to interoceptive sensitivity, over and above the demographic variables and the time estimation index. As such, the results provide an indication of the predictive power of each variable (β), as well as the predictive power of each step (sr²). The amount of variability accounted for by the entire model is also examined (R²). With interoceptive sensitivity as the outcome of interest, the following steps were entered: 1) Age, Sex, BMI; 2) time estimation index; 2) PANAS positive affect, BFI extraversion; 3) AVLT 30-minute recall, Full Scale IQ. Beta weights are reported as standardized coefficients. All statistical data analyses were performed using SPSS version 24.0 (IBM Corp., 2016).

Results

Performance on the heartbeat counting task.

Interoceptive Sensitivity score was calculated as noted above, with the score representing average performance over three heartbeat counting trials. The mean interoceptive sensitivity score was 0.68 (SD = 0.19, min = 0.16, max = 0.97). We did identify a significant positive correlation between interoceptive sensitivity and performance on the time estimation task (r = .44, p < .0001), and thus, the time estimation index was used as a covariate in all subsequent analyses, as carried out by Dunn et al. (2010).

Relation of interoceptive sensitivity to demographic variables.

The relationship between interoceptive sensitivity and the demographic variables was assessed with a series of partial correlations, controlling for the time estimation index. We found no significant relationships between interoceptive sensitivity and age (r = –.06, p = .55), sex (r = .08, p = .43), years of education (r = .01, p = .89), or body mass index (r = –.02, p = .88).

Relation of interoceptive sensitivity to affective variables.

Partial correlations were again conducted to explore the relationship between interoceptive sensitivity and the affective variables, while controlling for the time estimation index. Significant negative correlations were observed between interoceptive sensitivity and positive affect, as measured by the Positive and Negative Affect Schedule (PANAS) (r = –.28, p = .008), and between interoceptive sensitivity and trait extraversion, as measured by the Big Five Inventory (r = –.22, p = .036). The remaining affective variables were non-significant: PANAS Negative Affect (r = -.06, p = .52); BFI Agreeableness (r = .04, p = .74); BFI Conscientiousness (r = -.08, p = .46); BFI Neuroticism (r = .08, p = .48); and BFI Openness (r = -.11, p = .29).

Relation of interoceptive sensitivity to cognitive variables.

The relationship between interoceptive sensitivity and the cognitive variables was assessed with a series of partial correlations, controlling for the time estimation index. Significant positive correlations were observed between interoceptive sensitivity and 30-minute delayed verbal memory, as measured by the Rey Auditory-Verbal Learning Test (r = .34, p = .001). There was also a trend for a relationship between interoceptive sensitivity and Full Scale IQ, as measured by the WASI-II (r = .19, p = .07). The remaining cognitive variables were non-significant: Verbal IQ (r = .15, p = .15); Performance IQ (r = .07, p = .51); BNT (r = .15, p = .16); Digit Span Total (r = .04, p = .69); BVRT Errors (r = .17, p = .12); Trails A (r = -.01, p = .93); and Trails B (r = .11, p = .30).

Hierarchical regression analyses.

In order to examine the relative contributions of affective and cognitive variables in predicting interoceptive sensitivity, hierarchical regression analyses were used. Given that the data were cross-sectional, it was most prudent to examine which variables related to interoceptive sensitivity when entered into the same model. In addition, hierarchical regression allowed for the examination of whether affective and cognitive variables¹ were associated with interoceptive sensitivity over and above demographic variables, including age, sex, and BMI, and the time estimation index. Demographic variables did not account for significant variance in interoceptive sensitivity (sr² = .01), and none of the demographic variables were associated with interoceptive sensitivity (ps > .05). The time estimation index accounted for substantial variance in interoceptive sensitivity (sr² = .15), with higher scores on this index associated with greater interoceptive sensitivity (t = 4.94, β = .46, p < .001). Of the affective variables (sr² = .16), positive affect and extraversion remained significant predictors in the final model, with lower positive affect associated with higher interoceptive sensitivity (t = −2.33, β = −.22, p = .02) and lower extraversion associated with higher interoceptive sensitivity (t = −2.16, β = −.22, p = .03). Of the cognitive variables (sr² = .06), only AVLT accounted for significant variability in the final model. Higher 30-minute recall on the AVLT was associated with higher interoceptive sensitivity (t = 2.37, β = .26, p = .02). Full Scale IQ was not significantly associated with interoceptive sensitivity (t = 0.84, β = .08, p = .41). In total, the predictors accounted for 38% of the variability in interoceptive sensitivity (R² = .38).

Discussion

The present study examined how interoceptive sensitivity relates to individual differences in demographic, affective, and cognitive variables in a large sample of cognitively healthy, older adults with no history of neurological or psychiatric disease. The study identified multiple relationships. First, interoceptive sensitivity was inversely correlated with measures of positive emotionality: participants with higher interoceptive sensitivity tended to have lower levels of Positive Affect and Extraversion, by self-report. Second, interoceptive sensitivity was found to positively correlate with measures of cognition: participants who performed better on the heartbeat-counting task also tended to perform better on a measure of verbal memory. Third, when examining the predictors of interoceptive sensitivity in a hierarchical regression model, higher interoceptive sensitivity was related to: lower Positive Affect, lower Extraversion, and higher 30-minute verbal recall on the AVLT. These results suggest that, among older adults, interoceptive sensitivity is facilitative for aspects of cognition but perhaps disruptive for aspects of emotional experience. Importantly, these findings indicate that these relationships exist even when controlling for age, sex, BMI, and the time estimation index. However, in keeping with the exploratory nature of our study, no unequivocal conclusions may be drawn regarding causality.

Our findings that indicate a negative relationship between interoceptive sensitivity and measures of positive emotionality align with those put forward by Alkozei and Killgore (2015), who demonstrated an inverse relationship between interoceptive reactivity to social threats and emotional intelligence. Though not measured directly in our study, emotional intelligence has been shown to correlate strongly with extraversion (Day, Therrien, & Carroll, 2005). The implication is that individuals with greater interoceptive sensitivity are more disturbed by perceived threats in complex social situations, which may result in aversion to social situations, and thus decreased levels of emotional intelligence and extraversion. Individuals with higher interoceptive sensitivity may be more socially anxious, as interoceptive sensitivity has been posited to play a role in the pathophysiology of anxiety (Critchley et al., 2004; Dunn et al., 2010; Pollatos et al., 2007).

Further, Herbert, Pollatos, Flor, Enck, and Schandry (2010) suggest that individuals with higher interoceptive sensitivity tend to pay more attention to unpleasant stimuli (in their case, images), compared to individuals with lower interoceptive sensitivity who pay more attention to pleasant stimuli. If this pattern persists beyond the laboratory setting, it could explain our findings of decreased positive affect in participants with high interoceptive sensitivity. These individuals may be more attuned to aversive stimuli in their environment, resulting in a subjectively less positive life experience compared to individuals with lower interoceptive sensitivity.

Within the affective domain, our findings may be mediated by an individual’s awareness of their own emotional experience. Individuals with lower interoceptive sensitivity may be more detached from their negative emotional experience and subsequently rate valence as more positive. Some evidence suggests that interoceptive sensitivity is inversely associated with alexithymia in typical, healthy populations (Herbert, Herbert, & Pollatos, 2011) but more recent meta-analytic data indicates that the relationship only exists in clinical populations (Trivisan et al., 2019).

It is also possible that personality traits such as extraversion may impact one’s rating of emotional valence. For example, more extraverted individuals may be more disinhibited, and in turn, rate their affect as more positive, given that extraversion and positive affect is a robust finding in personality literature (e.g., Lucas & Baird, 2004).

As alluded to previously, relationships between interoceptive sensitivity and cognition have been demonstrated in prior studies. Werner et al. (2009) identified a positive relationship between interoceptive sensitivity and performance on a decision-making task. A correlation between interoceptive sensitivity and prospective memory has also been established (Umeda et al., 2016). Perhaps most notably, Stevenson et al. (2018) demonstrated that higher interoceptive sensitivity was associated with higher performance on measures of list learning and prose memory. Our results suggesting a positive relationship between interoceptive sensitivity and verbal memory are consistent with these prior studies.

In our sample of elders aged 60–91 years, interoceptive sensitivity did not decline with age. However, it is possible that there is a non-linear interaction between interoception and age, such that age effects may be apparent during middle-adulthood (e.g., prior to age 60) but then level off and remain quite stable after age 60. As an alternative way to address this question, we contrasted our data to published studies for which the same heartbeat-counting task was undertaken in younger adults. The mean score for our older adult sample was 0.68 (SD = 0.19), which is comparable to mean scores in multiple samples of healthy young adults: Ainley, Maister, & Tsakiris, 2015, mean = 0.64 (SD = 0.18); Herbert & Pollatos, 2014, mean = 0.72 (SD = 0.13); Herbert et al., 2010, mean = .74 (SD = 0.19); and Umeda et al., 2016, mean = 0.67 (SD = 0.22). Furthermore, in our own sample of older adults, spanning more than three decades, there was essentially no relationship between age and performance on the heartbeat counting task (r = –.06). Importantly, when controlling for age in the hierarchical regression analyses, the associations between interoceptive sensitivity and cognitive and affective predictors remained significant. Our present findings, in conjunction with published literature, does not support the notion that interoceptive sensitivity declines with normal aging.

We also did not identify reliable relationships between interoceptive sensitivity and demographic variables such as sex or body mass index. We found it particularly interesting that there was no relationship between interoceptive sensitivity and BMI, as prior research has demonstrated diminished interoceptive sensitivity in overweight and obese individuals (Robinson et al., 2021). Herbert and Pollatos (2014) found overweight and obese individuals (n = 75) to have an average interoceptive sensitivity score of 0.62 (SD = 0.19) compared to 0.72 (SD = 0.13) for those within a normal weight range (n = 75). As mentioned above, the average score for our participants is within that range at 0.68 (SD = 0.19). Thirty percent of our participants fall within the normal weight range with BMI between 18.5 and 25 kg/m² (mean = 28.5 ± 5.6 kg/m²). Only one individual fell below that range, with the remaining participants having a BMI greater than 25 kg/m². It is possible this may have influenced our results, but it is important to consider that a healthy BMI for elderly individuals is higher than the typically quoted 18.5–25 kg/m² value (Winter, MacInnis, Wattanapenpaiboon, & Nowson, 2014). The mean age of participants in the Herbert and Pollatos study was approximately 25 years old compared to our group in which the mean age is over 75.

While our study did not address interoception and body representations, research has indicated varying relationships, with higher interoceptive sensitivity corresponding with worse performance on measures of body schema in older adults (Raimo et al., 2021). However, Dobrushina and colleagues (2021) found no relationship between interoceptive accuracy and body perception across several age groups. These mixed results indicate the complexity of mechanisms that mediate the internal and external body.

Interoceptive awareness may be more critical to an older adult’s physical health than to a younger adult’s. Given increased medical comorbidity, including the relatively high frequency of cardiac conditions, data as to how interoception changes with age becomes medically relevant. Physicians are often reliant on a patient’s self-report of symptomatology. If an older adult has altered awareness of bodily signals and changes, they may be unable to alert a medical professional to a new or ongoing condition.

Indeed, Garimella et al. (2015) studied interoceptive awareness and self-reported symptomology. In a sample of patients with atrial fibrillation (AF), a condition that greatly increases one’s risk for heart attack and stroke, these authors found that 15% of patients either over- or under-estimated (with the great majority being underestimators; mean age 72.5 years) how frequently they were in AF, with underestimation more common with increasing age (Garimella et al., 2015). Similarly, elderly patients (mean age 72.0 years) with peptic ulcer disease were significantly more likely than younger adults to present without complaints of abdominal pain that is characteristic of the disease (Clinch, Banerjee, & Ostick, 1984). Both of these studies indicate that older adults have an increasing disconnection with body sensation as they age which results in reduced symptomatology. This, in turn, can impede a medical professional's ability to accurately assess and treat certain significant medical disorders in older adults.

At the same time, limited interoceptive sensitivity could be advantageous to an older individual’s emotional experience. Noticing fewer bodily sensations, such as minor pain or an aberrant heart rate, may contribute to lower anxiety about bodily changes associated with aging, and thus more positive overall affect. This purposeful, advantageous approach could partially account for our findings of lower interoceptive sensitivity and more positive affect in older individuals.

Our study had a number of limitations. For example, the older adults included in our sample were high functioning, with more than 16 years of education and a Full Scale IQ which fell in the high average range. It is possible that the high functioning nature of our sample could have biased the relationship between interoceptive sensitivity and our individual difference variables.

Additional cognitive factors may influence the relationship between interoceptive sensitivity and higher verbal recall in our sample. While our findings of higher verbal memory, as measured by the AVLT, were consistent with those of Stevenson et al. (2018), the nature of memory and cardiac interoception remains unclear. It is possible that interoception may impact verbal learning and memory, or perhaps another unknown factor mediates cognitive function in this regard. Further study is warranted to elucidate the complex mechanisms underlying interoception and cognition.

Use of the Schandry task also presents a limitation, in that it requires significant attention and concentration to perform the task acceptably well. Our pursuit of overlapping cognitive variables related to attention and concentration may have been confounded by the requirements of the Schandry task. Psychometric and methodological limitations of the task have been well documented (Brener & Ring, 2016; Ring, Brener, Knapp, & Mailloux, 2015). Further, Murphy et al. (2018) recently demonstrated that performance on heartbeat counting tasks is mediated by prior knowledge and beliefs, which we did not assess in our participants. Finally, multiple instruments have been used to assay interoceptive sensitivity. For example, Khalsa et al. (2009) use a heartbeat-tracking task in which participants are presented with a set of tones and instructed to determine whether the tones were in sync or out of sync with their heartbeat, while others have used a two-alternative forced choice task as developed by Whitehead, Drescher, Heiman, and Blackwell (1977). Brener and Ring (2016) have called for the standardization of all interoceptive sensitivity data collection with a common instrument. Recently, Suksasilp and Garfinkel (2022) proposed a more comprehensive assessment of interoceptive sensitivity across multiple body systems, including cardiac, respiratory, gastric, and urinary axes. In the future, it would be appropriate to repeat our study using an alternative task (or tasks) to assay interoceptive sensitivity, particularly if a single instrument or perhaps a more comprehensive assessment is adopted as the new standard. Additionally, Chen and colleagues (2021) have called for a unified framework of interoception that acknowledges the bidirectional nature of signals from the brain and internal organs (and vice versa).

In conclusion, the present study is the first to evaluate interoceptive sensitivity in a large sample of healthy older adults, in relation to demographic, affective, and cognitive variables. We did not identify a relationship between interoceptive sensitivity and age, as has been suggested previously. Our results support prior findings that indicate a positive relationship between interoceptive sensitivity and cognitive variables such as verbal memory, while documenting an inverse relationship with measures of positive affect. These exploratory findings represent compelling evidence for future inquiry and research examining the role of interoception and aging.

Table 2.

Hierarchical regression analyses: Role of demographic, affective, and cognitive variables in predicting interoceptive sensitivity.

Variables Entered (Step 1, Step 2, Step 3, Step 4)	β (Final)	sr 2	R 2
Interoceptive Sensitivity
Age	−.11
Sex	.03
BMI	−.07	.01
Time Estimation Index	.46***	.15
Positive Affect	−.22*
Extraversion	−.22*	.16
AVLT	.26*
FSIQ	.08	.06	.38

Note: N = 91.

^*p < .05;

^***p < .001.

BMI = Body Mass Index. AVLT = Rey Auditory-Verbal Learning Test, 30-minute recall. FSIQ = Full Scale Intelligence Quotient.