Interoception refers to the sensing of signals concerning the internal state of the body. Individual differences in interoceptive sensitivity are proposed to account for differences in affective processing, including the expression of anxiety. The majority of investigations of interoceptive accuracy focus on cardiac signals, typically using heartbeat detection tests and self-report measures. Consequently, little is known about how different organ-specific axes of interoception relate to each other or to symptoms of anxiety. Here, we compare interoception for cardiac and respiratory signals. We demonstrate a dissociation between cardiac and respiratory measures of interoceptive accuracy (i.e. task performance), yet a positive relationship between cardiac and respiratory measures of interoceptive awareness (i.e. metacognitive insight into own interoceptive ability). Neither interoceptive accuracy nor metacognitive awareness for cardiac and respiratory measures was related to touch acuity, an exteroceptive sense. Specific measures of interoception were found to be predictive of anxiety symptoms. Poor respiratory accuracy was associated with heightened anxiety score, while good metacognitive awareness for cardiac interoception was associated with reduced anxiety. These findings highlight that detection accuracies across different sensory modalities are dissociable and future work can better delineate their relationship to affective and cognitive constructs.
This article is part of the themed issue ‘Interoception beyond homeostasis: affect, cognition and mental health’.
Interoception is the sensing of the internal state of the body, rather than the detection of changes in the environment. The study of interoception has traditionally focused on signals originating within the visceral organs, including the heart and stomach, but it has also been extended, by some researchers, to include respiration and other internally generated processes that may be under greater volitional control [1–3]. In order to determine how accurately changes in internal bodily state are perceived, experimental tests quantify individual differences in sensitivity to internal bodily signals. Historically, the focus of research is dominated by tests of cardiac interoception, perhaps because heartbeats are discrete, easily measurable and performance accuracy can thus be easily quantified [4–9]. One example of a heartbeat detection task involves participants judging whether stimuli (such as tones or a light) are presented synchronously with the heartbeat, or after a fixed delay [10,11]. Individuals demonstrate marked variation in their ability to perform accurately on such tasks.
Interest in quantifying individual differences in interoceptive sensitivity is fuelled by theoretical models, backed by empirical observations, proposing bodily signals exert a fundamental influence upon cognitive and emotion processing. Internal states of bodily arousal shape emotional experience  and greater interoceptive accuracy is associated with emotions being felt with greater intensity . In a related way, interoceptive signals also have the capacity to guide cognitive processes, with heightened interoceptive accuracy being associated with superior intuitive decision making [14,15] and enhanced memory processing [16,17].
(a) Interoception across the senses
While the vast majority of interoceptive research attends to the heart, sensitivity to gastric [18–20], respiratory [21,22] and rectal sensations  has also been investigated, albeit less extensively. These organ-specific axes of interoception are typically studied in isolation. Those studies which have tested sensitivity to responses in more than one visceral organ report moderate relationships between sensitivity across axes, notably between cardiac and gastric systems, where accuracy to both cardiac and gastric sensations is reasonably well correlated [18,24]. Unsurprisingly, both heart and stomach have partially overlapping cortical representations within the ‘interoceptive’ insula [25,26]. Surprisingly however, considering close cardiorespiratory coupling, little is known about the relationship between cardiac and respiratory axes.
(b) Differentiating interoceptive accuracy from metacognitive awareness
Internal physiological changes can affect emotions and thoughts preconsciously or with different degrees of sensory awareness . Accordingly, interoceptive performance accuracy is not identical to interoceptive (metacognitive) awareness, because not all interoceptive sensory information, even that required to perform a heartbeat detection task, necessarily enters consciousness . Early studies highlight discrepancies between participants' self-reported interoceptive sensitivity and their actual (experimentally measured) accuracy, with strong or significant correlations seldom observed [10,29]. These findings, along with the acknowledgement that ‘a sensory impulse need not reach conscious awareness in order to affect behavior’  inspired questions such as ‘to what extent can accurate discrimination of visceral sensations be achieved in the absence of conscious awareness of these sensations?’ . The historic terminology of interoception has contributed to confusion: the terms interoceptive awareness and interoceptive sensitivity have previously been treated synonymously and interchangeably, although metacognitive awareness measures of interoceptive ability have seldom been probed [30,31]. Recent empirical work demonstrates a dissociation between performance accuracy for cardiac interoception and subjective belief about one's own performance aptitude (e.g. probed via confidence or questionnaire measures) . Moreover, both these measures of objective performance accuracy and subjective interoceptive ‘sensibility’ are shown also to be dissociable from interoceptive awareness, a metacognitive measure denoting correct insight into one's own interoceptive aptitude (i.e. correctly knowing whether you are good or poor) [28,32]. As this article deals with both interoceptive and exteroceptive senses, interoceptive awareness is also referred to as ‘metacognitive awareness’, thus extending this construct to both respiratory and somatosensory axes. These three dimensions of cardiac interoception (accuracy, sensibility and metacognitive awareness) become more aligned in individuals with high interoceptive accuracy . However, the delineation of these interoceptive dimensions has, to date, been confined to cardiac interoception. The interrelatedness or divergence of objective performance, subjective belief and conscious access is poorly understood when it comes to other interoceptive signals, such as those accompanying respiration.
(c) Clinical relationship
Defining the relationship between objective performance, subjective and metacognitive dimensions of interoception, and documenting their dissociation has utility when considering psychological disorders such as panic or generalized anxiety disorder, and how aspects of bodily representation are altered in long-term clinical conditions. For example, individuals with autism show a strong double dissociation between their capacity to accurately detect cardiac signals and their subjective belief in their ability to sense interoceptive signals. Overall, performance accuracy on heartbeat detection tasks is markedly impaired, while subjective interoceptive sensibility is inflated in high-functioning adults with autism relative to matched control participants . Anxiety was thought to be associated with increased cardiac interoception. However, there is much inconsistency regarding anxiety and interoception; for example, in panic disorder, data on whether interoceptive accuracy is enhanced under resting conditions are mixed [34–36]. Recent work suggests that it is actually the relative dimensions of interoception (i.e. objective performance relative to subjective belief), which may be key to understanding the relationship between interoception and anxiety symptomatology . As interoceptive awareness is also a metacognitive measure denoting the degree to which objective and subjective interoceptive axes align, it can thus be hypothesized that this (cardiac) metacognitive awareness measure should be associated with lower levels of anxiety.
Prominent theories and empirical work have directly posited a role for altered respiratory interoception in driving anxiety [37–39]. Interestingly, asthma, which is associated with a reduced ability to discriminate different levels of airway resistance [40,41], has high comorbidity with anxiety . Together these results also point to a possible relationship between sensitivity to respiratory symptoms and anxiety.
We sought to characterize the consistency of interoception across distinct organ-specific axes. We therefore undertook a set of experimental studies to test two main questions: First, are the dimensions of interoception (accuracy, sensibility and metacognitive awareness) constant across senses (cardiac, respiratory and touch)? For example, if participants display good interoceptive accuracy towards cardiac signals, do they also display heightened accuracy to detect respiratory and somatosensory stimuli? Second, are the different senses (cardiac, respiratory and touch) associated with different degrees of confidence and metacognitive insight into one's own interoceptive ability. Finally, we also sought to investigate the relationship between interoceptive dimensions and anxiety symptomatology.
Forty-two healthy volunteers (34 females, 8 males) were recruited through poster and online advertisements at the University of Sussex and in the local area of Brighton and Hove, UK. The age of participants ranged from 18 to 65 years, with a mean age of 28.19 years (s.d. = 9.8). The experiment was approved by the Brighton and Sussex Medical School (BSMS) Research Governance and Ethics Committee. All participants were non-smokers or only smoked occasionally with no reported history of cardiovascular disease. Body mass index (BMI) ranged from 17.04 to 28.52 (M = 22.17, s.d. = 2.76). Ten participants reported a history of very mild asthma (see electronic supplementary material, figure S1).
(i) Cardiac interoception
Interoceptive accuracy for cardiac signals was assessed using heartbeat discrimination [10,11], a task designed to determine individual differences in heartbeat detection. Participants were presented with a number of trials. In each trial, the series of tones were either coincident with the beat of their heart (synchronous condition) or presented at a delay (asynchronous condition). Each trial consisted of 10 tones presented at 440 Hz, each with a 100 ms duration, triggered by the participant's heartbeat. Under the synchronous condition, tones were generated at the beginning of the rising edge of the pressure wave. Under the asynchronous condition, a delay of 300 ms was inserted. Assuming an average delay of 250 ms between the R-wave and the arrival of the pressure wave at the finger , this set-up delivers tones around 250 ms or 550 ms after the R-wave, which correspond to maximum and minimum synchronicity judgements, respectively . Participants were asked to state whether the tones were synchronous or asynchronous with their own heart. Approximately half the tones were presented on the heartbeat and half were delayed. The order of these synchronous and asynchronous trials was fully randomized for each participant. The task consisted of 25 trials in total; at the end of each trial participants signalled their confidence in the accuracy of their response on a visual analogue scale ranging from ‘Complete guess/No heartbeat awareness' to ‘Complete confidence/Full perception of heartbeat’.
(ii) Respiratory interoception
Respiratory sensitivity was measured using an inspiratory resistance detection task developed on the basis of a respiratory resistance threshold task . To determine sensitivity, participants breathed through an open breathing circuit and judged whether additional resistance to the airflow was present on targeted trials. The circuit consisted of a two-way non-rebreathing valve (Hans-Rudolph 2600), a disposable mouthpiece, a connecting plastic tube (1.5 m long) and MicroGard® II bacterial filters (V—892383, CareFusion, Germany) that served as resistors at the inspiratory end of the circuit and one filter at the mouth end to keep the circuit hygienic. Each filter had a resistance of 34 cm H2O l−1 s−1. This enabled the measurement of very low detection thresholds. Removal of the nylon mesh from some filters provided dummy-resistors for control trials without additional resistance.
During the task, the experimenter attached filters, hidden from the participant's view, to the breathing circuit in a fixed decreasing order (starting with seven and potentially ending with a minimum of one) for 20 trials each. On every trial, participants had to detect the difference between an open circuit with no extra resistive load and a circuit with added resistance on the in-breath. They were instructed to: ‘Take up to 3 breaths on the open circuit to sense the baseline resistance, [after a signal] continue breathing for up to 3 more breaths to detect whether resistance is present.’ Participants indicated their answer verbally for every trial and how confident they were by giving a value between 0—‘Complete guess' and 100—‘Complete confidence’. If participants were able to discriminate with a minimum of 70% accuracy they progressed to the next stage of filters (N−1); otherwise the experiment moved on to the next task.
(iii) Tactile acuity
Participants' sensitivity for touch was measured with an adapted version of the grating orientation task . Tactile spatial acuity threshold is determined by applying a series of grooved dome shaped pieces of plastic (JVP tactile domes, Stoelting Co., Wood Dale, IL, USA) with decreasing square wave grating widths (3.0, 2.0, 1.5, 1.2, 1.0, 0.75, 0.5 and 0.35 mm) to the finger. Participants then indicated the orientation of the gratings (parallel or perpendicular). The participant's index finger of the dominant hand was checked for abnormalities (e.g. scar tissue or wounds) and used when no abnormalities were found. Participants closed their eyes and looked away from their hand, which rested comfortably in a supine position. For each grating, participants stated the orientation of the grating and indicated confidence by giving a value between 0—‘Complete guess' and 100—‘Complete confidence’. Each grating was presented 12 times and participants only received the next grating down if a minimum of 70% were correct.
Anxiety symptomatology was assessed using the State portion of the Spielberger's state trait anxiety inventory (STAI; ). The STAI consists of a 20-item state-scale to measure how anxious participants feel ‘right now and at this moment’ using normal and reversed four-point Likert-type items. Examples for the state-scale are: ‘I am presently worrying about possible misfortunes' or ‘I am relaxed’ with answers ranging from ‘Not at all’ to ‘Very much so’. This questionnaire was administered at the start of the experiment prior to the tests of interoception.
(c) Data analysis
(i) Accuracy and confidence
For the heartbeat discrimination task, cardiac interoceptive accuracy was taken as the percentage of times a correct decision was made (i.e. tones correctly identified as being synchronous or asynchronous with one's own heartbeat). For both the respiratory and touch acuity tasks, accuracy was calculated based on the percentage of times a correct decision was made on the final series of trials. This ensured levels of accuracy were equated across the senses allowing for comparison of analyses pertaining to confidence and metacognitive awareness. Thus, confidence ratings were averaged over trials for the heartbeat discrimination task and averaged over the trials from the final stage of the respiration and touch acuity tasks. In addition, filter level threshold and spatial acuity threshold were also documented as a second level of accuracy pertaining to both respiratory and touch acuity accuracy. These measures were the final level of filters (between one and seven) where participants could reliably determine the presence of filters or width of the domes where orientation could be reliably ascertained (performance at 70% or greater), with lower values indicating higher accuracy.
(ii) Metacognitive awareness
Metacognitive awareness of cardiac signals, as well as metacognitive awareness for both respiration and touch, were operationalized as the extent to which confidence predicts accuracy . This was quantified using a receiver operating characteristic (ROC) curve analysis . Correct decisions for the heartbeat discrimination task (synchronous or asynchronous with the heart), respiratory (presence of filters) and touch acuity (orientation of grating) served as the state variable, with confidence as the test variable. For both the respiration and touch acuity tasks, decisions and confidence were taken from the final stage of the respective tasks.
The area under the ROC curve gives a precise measure of the extent to which confidence reflects accuracy, independent of the participant's overall propensity to report high confidence. This metacognitive measure was computed separately for cardiac, respiratory and touch acuity responses to provide three measures of metacognitive awareness.
(iii) Statistical analyses
Firstly, a series of analyses assessed whether the dimensions of interoception (accuracy, confidence and meta-awareness) were consistent across the senses (cardiac, respiratory and touch). Specifically, a series of separate Pearson's correlations determined the relationships between (i) accuracy across the senses, (ii) confidence across the senses and (iii) metacognitive awareness across the senses. In all Pearson's correlations, age and sex were entered as covariates. The direct effects of age on interoceptive accuracy across cardiac, respiratory and touch accuracy can be seen in the electronic supplementary material, figure S2. Secondly, a repeated measures 3 (sense) × 3 (interoceptive dimension) ANOVA assessed whether the different senses (cardiac, respiratory and touch) were associated with different degrees of confidence and metacognitive awareness into one's own interoceptive ability. This was broken down with one-way repeated measures ANOVAs and further explored using paired sample t-tests. Finally, to assess the relationship between interoceptive dimensions and anxiety symptomatology, Pearson's correlational analyses assessed for significant relationships between interoceptive dimensions within each sense and anxiety symptoms.
(a) Performance accuracy across sensory axes
Cardiac and respiratory accuracy did not significantly correlate (r = −0.14, p = 0.37). Moreover, both respiratory and cardiac accuracies were not significantly positively related to tactile acuity accuracy (r = −0.07, p = 0.68; r = −0.32, p = 0.04, respectively). Furthermore, respiratory and tactile acuity thresholds were not significantly associated (r = 0.13, p = 0.42) and neither of these two thresholds was related to interoceptive cardiac accuracy (r = 0.06, p = 0.70; r = 0.20, p = 0.21). Together these results speak to dissociable performance in accuracy across the axes of heart, respiration and touch.
(b) Confidence in judgements across sensory axes
Propensity to be confident in one's own performance accuracy was related across the senses. Confidence in cardiac interoception was significantly related to respiratory task confidence (r = 0.38, p = 0.016) although it did not reach significance when compared with confidence for touch acuity (r = 0.27, p = 0.093). Respiration confidence was significantly related to touch acuity confidence (r = 0.57, p < 0.001).
(c) Metacognitive awareness across sensory axes
Metacognitive awareness was correlated across cardiac and respiratory axes (r = 0.36, p = 0.02). Thus, if an individual displayed good insight into his/her accuracy when performing the cardiac interoception task this tended also to be the case for his/her respiratory judgements. This suggests relative stability of metacognitive insight across these axes, with insight transferred across cardiac and respiratory axes, even though performance itself did not transfer. Importantly, this was not a generic effect of metacognitive sensory judgement: the metacognitive awareness measures for both respiratory and cardiac performance did not correlate with metacognitive awareness for performance on the exteroceptive touch acuity task (r = 0.154, p = 0.34; r = 0.019, p = 0.91, respectively) (figure 1).
(d) Relative dimensions of interoception across sensory axes
Performance accuracy was broadly matched across the three sensory tasks: there was no difference in terms of discrimination accuracy (F2,82 = 1.13, p = 0.33). Nevertheless, a marked difference was observed for average confidence levels (F2,82 = 12.19, p < 0.001). Individuals tended to be much more confident in their ability to detect respiratory loads relative to their confidence for both cardiac interoception and tactile acuity (t41 = 4.35, p < 0.001; t41 = 5.26, p < 0.001). Metacognitive awareness was also different across the senses (F2, 82 = 4.33, p = 0.016). Specifically, metacognitive awareness was elevated for tactile acuity, and this reached significance when compared with respiratory metacognition (t41 = 3.21, p = 0.003) but not for cardiac metacognitive awareness (t41 = 1.62, p = 0.11). Thus, participants tended to have greater metacognitive insight into their touch acuity performance. However, this awareness was weaker for the two other senses, particularly respiration (figure 2a).
(e) Confidence–accuracy correspondence within senses
There was no significant relationship between performance accuracy and average confidence for either cardiac (r = −0.03, p = 0.84) or respiratory measures (r = 0.06, p = 0.71). However, for the exteroceptive tactile acuity task, performance accuracy was related to confidence (r = 0.46, p = 0.003) (figure 3). This correspondence for touch between average confidence and overall accuracy is consistent with our earlier finding of elevated metacognitive awareness present for touch.
(f) Exploratory analyses: dimensions of interoception and relationship to anxiety symptoms
Anxiety symptoms were related to cardiac interoceptive metacognitive awareness (r = −0.40, p = 0.01) and respiratory threshold (r = 0.32, p = 0.043). Thus, both good insight into cardiac interoceptive performance, and overall respiratory sensitivity (as a higher threshold is indicative of poorer accuracy) were associated with reduced anxiety symptomatology. No other correlations across interoceptive dimensions (accuracy, confidence and metacognitive awareness) for each of the three sensory axes (cardiac, respiratory and touch) were related to anxiety symptomatology (p > 0.14) (figure 4).
(a) Summary of key findings
Accurate detection of bodily sensations across different sensory modalities (cardiac, respiratory and touch) were not found to be related; thus performance accuracy was not transferable. However, metacognitive insight into one's own performance was related, with a positive association between interoceptive metacognitive awareness for both cardiac and respiratory modalities. This suggests metacognitive awareness may be a ‘trait’ measure, which is transferable across these interoceptive senses, with accurately knowing whether you are good or poor at interoception being a relatively constant individual difference. This stability of metacognitive insight did not extend to exteroception, where metacognitive awareness for touch acuity was not related to cardiac and respiratory metacognitive awareness.
Cardiac interoceptive accuracy can be differentiated from both confidence and metacognitive insight into one's own interoceptive performance. Here, we demonstrate that this dissociation also extends to respiration, where accuracy in detecting respiratory resistance was not associated with either metacognitive awareness or confidence in performance. Instead, individuals tended to manifest an inflated confidence in their respiratory detection performance. By contrast, for the exteroceptive control of touch acuity, participants had greater insight into their ability, as underscored by enhanced metacognitive awareness, also manifesting as a positive relationship between average confidence and accuracy.
Anxiety symptoms were associated with metacognitive awareness for cardiac signals, with good metacognitive awareness related to reduced anxiety. The ability to more ably detect changes in respiratory resistance (i.e. accuracy in determining the presence of low resistance filters) was associated with reduced expression of anxiety symptomatology.
(b) Interoceptive ability across sensory axes
Within our sample, good cardiac interoceptive accuracy did not predict enhanced respiratory and somatosensory detection accuracy. This finding contrasts with previous comparisons of interoceptive accuracy for both cardiac and gastric sensations, where strong relationships have been observed [18,24]. One possible explanation is that there is a shared cortical representation of cardiac and upper gastrointestinal/gastric sensation within the insula cortex [1,7,49,50]. This common architecture potentially supports this shared performance of gastric and cardiac interoceptive accuracy found previously. However, this account is weakened by evidence for topological dissociation of specific cardiac and gastric representations within sub-regions of insula cortex . Moreover, the insula is activated by respiratory chemical and mechanical signals and sensations such as ‘air hunger’ [51–53] and by affectively laden somatosensory sensations such as itch and sensual touch [54–56]. However, the detection of respiratory resistance also engages a network of additional brain regions where, for example, the strength of inspiratory resistance correlates with anticipatory activity within the lateral periaqueductal grey matter, bilateral putamen and caudate ; tactile discrimination likewise engages an array of neural regions including the anterior intraparietal sulcus, right postcentral sulcus and gyrus, and bilateral ventral premotor cortex .
Interestingly across individuals, there was a relationship between measures of metacognitive awareness for detecting cardiac and respiratory signals (but not for touch). This indicates that the degree of insight of an individual into his/her interoceptive performance was transferable across cardiac and respiratory domains. Thus, knowing whether you are good or poor in judging your own internal bodily sensations appears to act as a relatively stable trait across these axes, even if your accuracy in judging these internal bodily signals is distinct. As noted, the metacognitive awareness relating to cardiac interoception or respiration was not significantly associated with metacognitive awareness pertaining to tactile acuity. This suggests that insight is not smoothly transferable from interoceptive to exteroceptive domains, at least regarding the senses tested within this study. This latter finding is consistent with emerging empirical research indicating that the level of metacognitive insight for bodily sensations does not necessarily transfer to other categories of mental processing, such as metacognition of cognition. For example, interoceptive awareness was not found to predict metacognition of memory performance . There is likely to be a neural basis for dissociation between different domains of metacognition, as indicated by the observation of distinct networks anchored in the medial and lateral regions of the anterior prefrontal cortex supporting metacognitive ability for memory and perception . By contrast, there did appear to be a generalized process underlying an individual's reported sense of confidence in his/her judgements. Thus, we observed correlations between confidence ratings across interoceptive cardiac, respiratory and exteroceptive touch discrimination tasks, indicating a trait propensity to be confident that was largely transferable. Thus, both cardiac and respiratory interoception may share subjective and metacognitive representation. In addition, individuals tended to be more confident in their respiration judgements. For respiration judgements, we found a tendency for inflated confidence. As respiration is under greater volitional control, this system may be more sensitive to confidence effects allowing for greater top down regulation.
(c) Dissociated dimensions of interoception: clinical implications
The present findings extend a growing body of literature suggesting that accurate detection of bodily sensations is dissociable from both subjective belief and metacognitive awareness [28,32,61]. Interestingly, we show this dissociation for both respiratory and cardiac measures, but not touch acuity. Variable insight into interoceptive accuracy may represent a particular characteristic of interoceptive processing that could even be the basis for the affective and motivational influences of internal bodily physiology (e.g. states of autonomic arousal) upon cognition and behaviour [28,33]. Discrepancies between objective performance measures and subjective measures are observed in specific domains of emotion [62,63], subliminal perception  and implicit learning . These dissociations raise interesting questions about what it is to process a bodily sensation in either the presence or absence of conscious awareness, and what happens when subjective belief about interoceptive ability deviates from actual interoceptive performance. Clinical populations, notably adults with diagnoses of autism spectrum disorder, can show an exaggerated dissociation between interoceptive sensibility (i.e. self-ascribed proficiency in interoceptive ability as assessed via questionnaire) and actual capacity to accurately detect cardiac signals . The degree of mismatch (termed ‘interoceptive trait prediction error’) predicts the expression of anxiety symptoms and interpersonal emotional difficulties in this patient group. These results highlight the potential importance of the interoceptive signals in combination with error arising from the deviation between actual and expected interoceptive proficiency, even within healthy populations. In particular, emotions and affective feeling states are proposed to originate from interoceptive states of physiological arousal and the precision and predictability of these signals in interpretive representation.
In this study, we observed that metacognitive awareness of cardiac signals was associated with reduced anxiety symptomatology. This is consistent with the weighting of cardiac afferent sensations by higher-level representations of their reliability and expectancy. Anxiety emerges where these top-down and bottom-up signals diverge . This mechanism may be specific for cardiac interoception rather than extending to other internal bodily sensations, because volitional control of the heart is minimal, in contrast, for example, to respiration. Correspondingly, for respiration (which retains a strong volitional component to control), anxiety symptoms were related more to the quality of the afferent signal, i.e. respiratory sensitivity (measured as accuracy in detecting respiratory resistance). Here, a more precise representation appeared to mitigate anxiety. This finding accords with previous work denoting a relationship between respiratory detection measures and anxiety in clinical samples. In patients with asthma, perception of physiological changes during a bronchial challenge is associated with anxiety , and in patients with emphysema, anxiety is associated with both impaired exercise performance and more shortness of breath . Thus, both the precision and the quality of the respiratory signal have implications for anxiety symptomatology.
The concept of prediction has informed how we consider the interaction of different dimensions of interoception and how they impact upon emotion and cognition. Predictive coding has deep-rooted relevance to how the brain represents both the external world and internal bodily state (see [69–71] in this volume). Predictive coding theories propose that prediction and error-correction are integral to the way in which the brain operates [72–75]. Predictive coding revises the established concept of perception as a feed-forward ‘bottom-up’ process for feature detection and building of representations. Instead, perception arises from the generation of ‘best-guess’ top-down predictions about the likely cause of incoming sensory signals. Within a hierarchy of sensory processing, descending predictions meet with sensory signals to generate prediction errors, which feed forward to higher representational levels. Perception arises from prediction and prediction errors. The brain attempts to minimize prediction error through precision-weighting afferent information, changing generative models underlying prediction or modifying behaviour/responses (active inference) to change the sensory afferent input. While predictive coding models are widely applied to exteroception, increasingly the same principles are being applied to the organism itself [72,76,77]. For interoception, prediction errors arise when interoceptive signals deviate significantly from expectations about bodily state. Emotions, including feelings of anxiety, may arise from a short-term mismatch between anticipated and actual state of arousal. However, these processes may also operate at a trait level over longer timespans. Poor interoceptive accuracy may reflect the reliability of interoceptive afferent input, while subjective understanding about what, how and when internal signals represent beliefs/predictions about interoceptive signal processing proficiency. This latter interpretation includes both over and under estimations of confidence in judging one's own internal bodily sensations, and insight/metacognitive awareness into one's ability to make such judgements. Our new observations show that a lower prediction error, i.e. increased metacognitive awareness, for cardiac interoception, is associated with reduced anxiety thus resonating with the perspective that feelings of anxiety may arise from a mismatch in bottom-up signals and top-down predictions in relation to cardiac signals [66,78]. By contrast, we found respiratory accuracy was inversely related to anxiety, with a greater ability to detect resistive loads associated with reduced anxiety. In terms of predictive coding, this latter observation is interesting because there is much greater capacity for volitional control of breathing when compared with cardiac regulation. Meditative practices, such as mindfulness, potentially enhance respiratory accuracy , while meditation-based therapies reduce anxiety . In addition, physical exercise has been used to treat anxiety disorders . It is thought that through eliciting autonomic changes (e.g. elevations in heart rate, shortness of breath), aerobic exercise may constitute a type of ‘interoceptive exposure’. Anxiety sensitivity is reduced with interoceptive exposure through exercise  and it is thought that this exposure may result in a form of ‘extinction’, which teaches individuals that these symptoms are not ‘catastrophic’ . Together these studies indicate that both enhanced respiratory interoception combined with interoceptive exposure may be mechanisms through which meditation and exercise can reduce anxiety.
This study is one of only a few to compare perceptual accuracy beyond cardiac interoception and the first to investigate all three dimensions of cardiac and respiratory interoception in one experiment: accuracy, sensibility (i.e. confidence) and metacognitive awareness. These relationships were analysed both between and within organ specific axes, and relative to an exteroceptive measure, thus yielding a number of analyses. In order to document all effects obtained, these were included without further correction. It is hoped that this will maximize the likelihood of future research based on directional hypotheses. Another shortcoming of the study concerns the characteristics of the sample particularly regarding sex distribution, as only a small number of males (N = 8) participated. Furthermore, a healthy community sample was recruited but was not adequately screened for psychiatric and neurological conditions; thus the presence of co-morbid clinical diagnoses cannot be adequately accounted for. Moreover, the characterization of anxiety relied on state rather than trait assessment, and effects related to these state-anxiety levels might differ in clinical anxiety groups. Furthermore, while efforts were made to equate levels of accuracy across senses, it should be noted that the comparison between respiratory and tactile perturbation above baseline state versus assessment of cardiac signals at baseline (resting) state is somewhat uneven and may have affected results obtained. Future work utilizing different paradigms across organ-specific axes at baseline and using perturbation methods may yield differing results.
Together, our results characterize further the extent to which preconscious and conscious psychological dimensions of interoception are dissociable across different sensory axes, and differ from exteroceptive senses. They point to mechanisms through which interoception shapes emotional style. Understanding how sensitivity to interoceptive signals map onto anxiety symptomatology, both in terms of absolute levels and via the interoceptive error signals inherent within metacognitive measures, can help elucidate mechanisms through which body-centred therapies may help alleviate anxiety.
The experiment was approved by the Brighton and Sussex Medical School (BSMS) Research Governance and Ethics Committee. All participants gave informed consent.
We have no competing interests.
This work was supported by the Brighton and Sussex Medical School JRA programme awarded to M.F.M. and via a donation from the Dr Mortimer and Dame Theresa Sackler Foundation. H.D.C. and S.N.G. are supported by an ERC Advanced grant to H.D.C. (CCFIB AG 234150).
We wish to thank both Phoebe Votolato who collected pilot data for a previous protocol related to this study and Thomas Janssens for helpful initial guidance regarding the respiration task.
One contribution of 16 to a theme issue ‘Interoception beyond homeostasis: affect, cognition and mental health’.
Electronic supplementary material is available online at https://dx.doi.org/10.6084/m9.figshare.c.3469755.
- Accepted June 14, 2016.
- © 2016 The Author(s)
Published by the Royal Society. All rights reserved.