Interoception can be described as the ability to sense one’s own physiological condition of the body [1
]. While some researchers referred to interoception as a single construct [2
], others examined different facets of interoception [3
]. These inconsistent terminologies and their use have been criticized by Garfinkel et al. [4
], who demanded consistent and clear definitions. By examining the structure of interoception, Garfinkel, Seth, Barrett, Suzuki and Critchley [4
] showed that interoception is a multi-facetted phenomenon including (at least) interoceptive accuracy, awareness and sensibility.
Interoceptive sensibility can be described as the self-evaluation of someone’s subjective interoception assessed by using self-reports [5
]. Interoceptive accuracy describes someone’s actual performance in an objective interoceptive task, e.g., a heartbeat perception task, whereas interoceptive awareness comprises the metacognitive awareness of one’s own interoceptive accuracy. Garfinkel, Seth, Barrett, Suzuki and Critchley [4
] showed that interoceptive awareness and sensibility could only partly predict interoceptive accuracy. All three dimensions were distinct and separable. Since there was a relationship between the facets, but only within the group of individuals with the greatest interoceptive accuracy, interoceptive accuracy has been highlighted as the core construct of interoception. The authors suggested that the relationship between the interoceptive facets is stronger for individuals with high interoceptive accuracy compared to those with low interoceptive accuracy. In a pilot study of 24 healthy students, Meessen et al. [6
], in fact, demonstrated that all three facets of interoception are uncorrelated. Forkmann et al. [7
] confirmed the three-dimensionality of interoception by reporting no correlations between the facets when accuracy was measured with the heartbeat perception task [8
] and moderate correlations when accuracy was measured with the heartbeat discrimination task.
Research suggests that the facets of interoception seem to be associated with psychological distress, e.g., depression and anxiety [9
]. For anxiety, mixed results have been reported [12
]. While Garfinkel et al. [13
] and Dunn et al. [14
] found that interoceptive accuracy independently contributed to anxiety symptoms, other studies state that patients with panic disorder show similar or even better interoceptive accuracy [3
] but worse interoceptive sensibility [3
] compared to a control group. Interoceptive awareness was positively related to trait anxiety [17
]. For depression, deficits in interoceptive accuracy generally seem to be associated with depressive symptoms [11
]. A review by Eggart et al. [19
] suggested a u-shaped relationship between depression and interoceptive accuracy, with largest interoceptive deficits coinciding with moderate depression severity.
Only recently, it has been proposed that interoceptive deficits might also be related to suicidal ideation and behavior [20
]. Interoceptive sensibility measured with the self-report measure Multidimensional Assessment of Interoceptive Awareness (MAIA) [5
] and assessed with the subscale “interoceptive deficits” from the Eating Disorder Inventory (EDI-3) [21
] was negatively associated with current suicidal ideation and past suicide attempts [22
]. MAIA scores could differentiate between suicide attempters, suicidal ideators and a control group [25
]. The EDI even differentiated recent and distant suicide attempters [20
]. There are only two studies available so far that investigated interoceptive accuracy (i.e., the objective performance in detecting body sensations) in relation to suicidal ideation or behavior. Results showed no differences in interoceptive accuracy (but in interoceptive sensibility) for participants with compared to those without suicidal ideation [27
] and lower heartbeat perception accuracy in suicide attempters compared to non-attempters [28
]. However, there has been no study investigating associations between interoceptive awareness and suicidality [29
Regarding the assessment of interoceptive sensibility, there are two critical points: on the one hand, many studies assessed interoceptive sensibility with a subscale of the EDI [20
], which was originally developed for participants with eating disorders and assesses interoception in relation to food intake and the gastrointestinal system [21
]. Thus, it appears at least questionable whether conclusions based on EDI-data can be generalized to patients with no eating but other mental disorders. On the other hand, in self-report questionnaires, interoceptive sensibility was usually assessed retrospectively and the timeframe respondents are asked to refer to was not specified. For example, the MAIA refers to “the general daily life” [5
], while the EDI refers to how often each statement applies with no time frame at all [21
]. It has been argued that questionnaire-based retrospective assessments are compromised by memory bias and a lack of ecological validity [30
It is unclear whether interoception is best understood as being state-like or trait-like. There is evidence from few studies suggesting within-person change in interoception across time. A study by Wittkamp et al. [31
] using latent state-trait analysis of interoceptive accuracy assessments on three consecutive measurement occasions showed that 40% of variance in one single interoceptive accuracy measurement could be explained by trait, whereas 27% was traced to effects of situation and person-situation interactions–suggesting some variability in interoceptive accuracy over time. Some further evidence comes from studies that investigated whether facets of interoception could be trained. Studies showed that interoceptive awareness trained by daily practices of “Body Scans” and “Breath Meditation” [32
] and accuracy trained by daily “Body Scans” over eight weeks [33
] and by contingent cardiac feedback [34
] could be improved, suggesting that interoception can generally be affected by situational or behavioral manipulations (i.e., training), which implies a certain temporal variability. Only interoceptive sensibility appeared not to be affected by “Body Scan” interventions [33
]. It is unclear, however, whether the mere repeated execution of the various interoception measurements already results in a practice effect as the studies mentioned above explicitly trained the facets through interventions. It is also unclear how the facets of interoception behave over time and whether they are measurable over short time intervals of minutes or hours.
A viable alternative assessment method that promises the possibility to overcome memory bias, lack of ecological validity and allows for the assessment of within-person variation across short time frames, is Ecological Momentary Assessment (EMA). EMA refers to the repeated sampling of subjects’ current behaviors and experiences in real time and in their natural environments [35
], for example, via smartphones [36
]. There is empirical evidence on the within-person variation and temporal trajectories of clinical variables such as suicidal ideation, depression and anxiety [38
]. Although as noted above, interoception has been shown to be related to all these clinical variables, to date, we know virtually nothing about the temporal course of the facets of interoception across short intervals of minutes or hours.
While interoceptive sensibility is usually assessed via self-report, which can rather easily be adopted to the EMA-setting (such as already implemented for, e.g., suicidal ideation or negative affect [42
]), the assessment of interoceptive accuracy and awareness is more challenging, since it requires the collection of both self-reported information and the number of heartbeats in given time-frames.
Therefore, the aim of this study was to measure interoceptive accuracy, awareness and sensibility using EMA. Since the facets of interoception have never been investigated in such a study design, this study should be treated as a pilot study. The main goal of this study was to test the general feasibility of such a study design and to find out whether the facets of interoception are subject to intraindividual fluctuations. Because of the novelty of the design and potential test burden associated with repeated EMA, we abstained from including patients with mental disorders but decided to aim for a non-clinical sample to prove the study concept. Based on prior findings on interoception, we hypothesized that (a) all three facets of interoception fluctuate over time. Additionally, we hypothesized that (b) there will be no practice effect for the facets of interoception through mere interoceptive task repetition, since positive practice effects have only been shown for targeted interventions and not solely for mere task repetition.
Descriptive statistics of all study variables can be found in Table 1
-tests for dependent samples revealed no significant difference between mean interoceptive accuracy measured in the laboratory at T0 (pre-HPT: M
= 0.59; SD
= 0.09) and mean interoceptive accuracy at day 1 measured via EMA (day 1 HPT-EMA: M
= 0.59; SD
= 0.27; t
(28) = −0.064, p
= 0.95). In addition, there was no difference between mean interoceptive accuracy measured in the laboratory at T2 (post-HPT: M
= 0.60, SD
= 0.11) and mean interoceptive accuracy measured at day 7 via EMA (day 7 HPT-EMA: M
= 0.60, SD
= 0.23; t
(27) = 0.138, p
= 0.89). This suggests that results of EMA-based assessments of interoceptive accuracy (HPT-EMA) do not differ from assessments of interoceptive accuracy in the laboratory, supporting the validity of assessing interoceptive accuracy based on EMA outside the laboratory.
3.1. Short-Term Variability of Interoception during EMA
For interoceptive awareness, 37% of the variance was accounted for by within-person variability (over time, see Table 1
). For the EMA-HPT score (as a measure of interoceptive accuracy), ICCs indicated that 58% of variance was due to within-person variability (over time). Between 38% and 62% of variance in the MAIA subscales as a measure of interoceptive sensibility during EMA was due to within-person variability (over time), depending on the subscale. For “Emotional Awareness” and “Self-Regulation” the proportion of variance due to within-person variability was 38%, while for “Attention Regulation”, 62% of variance was due to within-person variability. Additionally, MSSDs, as can be seen in Table 1
, demonstrated a wide range, indicating large differences between individuals in intra-individual variability across time. For the MAIA subscales during EMA, attention regulation seemed to have the highest point-to-point variability in comparison to the other scales. Figure 2
, Figure 3
and Figure 4
show the individual trajectories of interoception during EMA for each of the facets separately over 35 assessments for the participants of the interoception group, sorted by subject ID.
3.2. Practice Effects of Interoception from Baseline to Post Assessment
The mixed factorial ANOVA did not show a statistically significant interaction between time (baseline assessment vs. post assessment) and group (interoception group vs. control group) for interoceptive accuracy (pre- vs. post-HPT). There was also no significant main effect for time or group. For interoceptive awareness, there was neither a statistically significant interaction between time and group nor a significant main effect for time or group (see Table 2
For interoceptive sensibility results of each MAIA scale is reported separately. For four of the subscales, there were significant results. There was a significant main effect of time (F(1, 59) = 5.17, p < 0.05) for the scale “Noticing”. Independently of group, participants improved from the baseline to the post assessment (MBaselineInteroception = 3.31, SDBaselineInteroception = 0.79, MBaselineControl = 3.26, SDBaselineControl = 0.81, MPostInteroception = 3.65, SDPostInteroception = 0.74, MPostControl = 3.39, SDPostControl = 0.68).
For “Attention Regulation”, there was a significant interaction effect of time*group (F(1, 59) = 3.96, p < 0.05). Simple ANOVAs showed that there were only significant differences between the interoception and the control group at the post assessment (F(1, 59) = 6.65, p < 0.05) but not at the baseline assessment (F(1,59) = 1.18, p = 0.28). At the post assessment, participants in the interoception group showed significantly higher scores in “Attention Regulation” than participants in the control group (MPostInteroception = 3.21, SDPostInteroception = 0.83, MPostControl = 2.68, SDPostControl = 0.78).
In regard to “Emotional Awareness”, there was only a significant main effect of time (F(1, 59) = 5.48, p < 0.05). Independently of group, participants improved from the baseline to the post assessment in their self-reported “Emotional Awareness” (MBaselineInteroception = 3.44, SDBaselineInteroception = 0.98, MBaselineControl = 3.42, SDBaselineControl = 0.91, MPostInteroception = 3.76, SDPostInteroception = 0.76, MPostControl = 3.58, SDPostControl = 0.79).
In regard to “Body Listening”, there was a significant main effect of time (F(1,59) = 4.04, p < 0.05). Independently of group, participants improved from the baseline to the post assessment in their self-reported ability to listen to their body (MBaselineInteroception = 2.76, SDBaselineInteroception = 0.99, MBaselineControl = 2.56, SDBaselineControl = 0.99, MPostInteroception = 3.18, SDPostInteroception = 0.93, MPostControl = 2.62, SDPostControl = 1.06).
The aim of the current study was to assess interoception in an EMA-setting to examine its temporal course. We hypothesized that the facets of interoception fluctuate (a) and that there is no practice effect for the facets of interoception based on repetition of interoceptive tasks (b).
4.1. Fluctuations of Interoception during EMA
Given the current findings, the first hypothesis that interoceptive accuracy, awareness and sensibility fluctuate could be confirmed. Around 50% of the variance in all three facets of interoception was due to within-person variability. It could also be confirmed that this variability differed considerably between individuals, which complements findings of Wittkamp, Bertsch, Vögele and Schulz [31
The present results that interceptive accuracy, awareness and sensibility fluctuate across time are in line with results of EMA-based studies measuring clinical variables. Similar to the facets of interoception, it has been shown that depression, anxiety, and suicidal ideation fluctuate across time [38
]. All these variables have been shown to be related to interoception [13
]. Thus, results of this study call for future studies investigating the longitudinal association between clinical variables such as depression or suicidal ideation and interoception across time. The high temporal resolution of EMA facilitates the identification of differential relations between clinical variables and improves the understanding of an individual patient’s dynamic symptom change. Since compliance was excellent in the present study and in prior studies investigating clinical samples [42
], future EMA-studies assessing interoception in clinical samples are warranted. Results of this study support the general feasibility of assessing interoception in EMA-studies.
4.2. Practice Effects of Interoception from Baseline to Post Assessment
Regarding the second hypothesis of different facets of interoception not being improved by repetitive interoceptive tasks, our findings were mixed. Most importantly, for interoceptive accuracy as well as awareness, there were no significant practice effects. At first glance, this result appears to be contrary to findings of Bornemann, Herbert, Mehling and Singer [32
] and Fischer, Messner and Pollatos [33
], who showed that interoceptive accuracy and awareness could be improved by training. However, their trainings contained interventions such as the daily practice of “Body Scan” and contingent cardiac feedback during the HPT, which were both specifically targeted at improving interoception, whereas in the present study, no specific training was applied but only the effect of repeated practice was examined. In the interoception group, only the interoception tasks were presented repeatedly compared to the control group. Participants in this study did not receive feedback and, therefore, could not notice whether they improved or not. Conclusively, interoceptive tasks, especially the HPT as a performance task, can be repeated and assessed multiple times without noticeable practice effects. Fluctuations and interindividual differences appear to be interpretable and should not be traced to practice effects only.
For future studies with a similar setup, it would be interesting to examine how such highly repeated contingent cardiac feedback in short time intervals over several days would influence the interoceptive accuracy and awareness. This is especially interesting in regard to clinical practice. Interoceptive EMA interventions would allow patients to autonomously train repetitively by themselves wherever they are, which could also potentially improve their self-efficacy. In the light of recent studies, the training of interoceptive skills could have a positive impact on e.g., depression [59
] and chronic worry [60
] and could, therefore, be of high benefit in clinical practice.
For interoceptive sensibility, the MAIA scale “Attention Regulation” showed significant practice effects. Participants in the interoception group showed significantly higher scores at the post assessment than participants in the control group. Since “Attention Regulation” seems to be improvable through practice, one could speculate whether a targeted training of “Attention Regelation” might have beneficial effects on mental and/or physical health. However, it is important to keep in mind that in the current study participants could have only improved in their self-reported “Attention Regulation” because they were forced to give more attention to themselves than usually through the repetitive EMA.
Since there were also significant main effects of time for the scales “Noticing”, “Emotional Awareness” and “Body Listening”, participants also improved in those scales measuring interoceptive sensibility independently from group. Thus, improvements in interoceptive sensibility should not only be attributed to repeated assessments of interoception in EMA. Instead, only one repetition of these measurements (T0–T2) seems sufficient to increase interoceptive sensibility, at least for certain scales. However, results could be different in a clinical sample and need replication before further conclusions should be derived.
4.3. Strengths and Limitations
The results of the present study should be appreciated in the light of some strengths and limitations. The major strength of the study was being the first to assess the facets of interoception in EMA and, therefore, providing important findings about how the facets of interoception behave over time. Since there were no substantial practice effects for accuracy and awareness, it seems to be valid to measure the facets of interoception repeatedly. Furthermore, t-tests did not reveal differences in means between the assessments of interoceptive accuracy in EMA and at the baseline and post assessment supporting the validity of the assessment of interoceptive accuracy outside a laboratory setting. Additionally, for the baseline and post assessment in the laboratory, heartbeats were recorded via electrocardiogram (ECG) using a BIOPAC MP150.
As already stated, the current study is the first to assess interoception in an EMA design. Therefore, the primary aim of the study was to show the feasibility of assessing interoception within EMA and to investigate its temporal course. Since EMA is associated with some test burden in daily life, we deliberately abstained from including participants with mental disorders. We were able to show, in this study, the principal feasibility of the assessment strategy and the temporal fluctuation of interoception. In our view, this pattern of results warrants further investigation, ideally in clinical populations.
The first limitation of this study is that resting heart rate of participants was only assessed before the pre- and post-HPT at the baseline and post assessment. During the EMA phase, we did not control for arousal, which is negatively correlated with interoceptive accuracy [61
]. However, participants were familiar with the HPT task and practiced this task in our laboratory. When giving instructions to participants for the EMA phase, we explicitly reminded participants not to perform excessive activities before or during the EMA phase. We also asked participants during EMA what they had been doing before, where they were and if anything noteworthy had happened since the last assessment. There was no statistical control for context effects in the analyses of this study, since answers to these questions were heterogeneous and did not allow quantitative analyses. We did calculate the average heart beats per minute of our sample with M
= 84 and SD
= 10 (min. = 65, max. = 112). For future studies, one possibility could be to use bolus infusions of isoproterenol, a non-selective beta adrenergic agonist, which elicits rapid increases in heart rate and has been shown to overcome a major limitation of HPTs [62
]. Another possibility would be to instruct participants additionally at the beginning of each assessment to sit down and rest for 5 min before starting the assessment. However, compliance for such an extended approach might be low, since it would take longer for the participant.
Second, the HPT [8
] seems to be influenced by non-interoceptive processes [63
] and interoceptive accuracy scores derived from it are potentially problematic [64
]. Even though different psychological processes such as emotion regulation capacities [65
] and decision-making [66
] have been related to interoceptive accuracy scores derived from the HPT highlighting its role in psychological research, Zamariola et al. [64
] propose four criticisms, with which Ainley et al. [67
] insistently disagreed arguing that three of the four criticisms are not valid (for more detailed information see: Ainley et al. [67
Zamariola, Maurage, Luminet and Corneille [64
] state that interoceptive accuracy measured with the HPT depends on the error of participants’ undercounting of their perceived heartbeats due to their beliefs about their heart rate. Ainley, Tsakiris, Pollatos, Schulz and Herbert [67
] counter that participants’ beliefs do not explain why participants would rate their heart rates lower than they are and, thereby, their beliefs would not particularly explain their possible undercounting, which is contradictory to the first criticism.
The number of recorded heartbeats and the number of perceived heartbeats does not correlate (in Zamariola, Maurage, Luminet and Corneille’s data [64
]). However, Ainley, Tsakiris, Pollatos, Schulz and Herbert [67
] found that Zamariola, Maurage, Luminet and Corneille [64
] made this assumption due to arithmetic misunderstanding, which disproves this second criticism.
Zamariola, Maurage, Luminet and Corneille [64
] state that a measure for interoceptive accuracy should not depend on heart condition. However, this should actually be treated in favor of the HPT’s construct validity, since it is clear that the perception of interoceptive signals is depending on one’s specific physiology [67
Last but not least, there seems to be a tendency to poorer performance on the longer trials of the HPT. This statement is rejected by Ainley, Tsakiris, Pollatos, Schulz and Herbert [67
] arguing that mean recorded heart rates significantly differed between the three lengths of the trials in the data of Zamariola, Maurage, Luminet and Corneille [64
], which is in contrast to their assumption that the heart rate is constant across intervals and the poorer performance for longer trials is traced to participants’ undercounting.
In conclusion, the HPT is significantly connected to the activity of interoceptive neural networks, which has been shown in multiple studies [68
], the HPT has been shown to provide information about the associations between an individual’s interoceptive accuracy and psychological distress such as depression [11
], anxiety [13
] and even suicidality [28
] and most of its critical points have been disproved. Thus, we feel secure about the use of this measure for assessing interoceptive accuracy.
Third, participants had to complete the post assessment to the latest 14 days after the EMA phase. This was a wide time frame and should be kept shorter in future studies. It could be possible that practice effects already vanish within short periods of time.
Fourth, the results could also be influences by the high percentage (88.5%) of female participants, since men seem to be better in perceiving interoceptive processes [70
]. However, Pennebaker and Roberts [72
] suggest that those sex differences vanish in a non-laboratory setting. For future studies, it would be interesting to examine sex differences in interoception as well as hormonal changes and their influence on interoception during EMA, which is not a typical laboratory setting.