Making sense of what you sense: Disentangling ...

International Journal of Psychophysiology 109 (2016) 71–80

Contents lists available at ScienceDirect

International Journal of Psychophysiology journal homepage: www.elsevier.com/locate/ijpsycho

Making sense of what you sense: Disentangling interoceptive awareness, sensibility and accuracy Thomas Forkmann a,⁎, Anne Scherer a, Judith Meessen a, Matthias Michal b, Hartmut Schächinger c, Claus Vögele d, André Schulz d a

Institute of Medical Psychology and Medical Sociology, University Hospital of RWTH Aachen University, Germany Department of Psychosomatic Medicine and Psychotherapy, University Medical Center Mainz, Germany Institute of Psychobiology, University of Trier, Germany d Institute for Health and Behaviour, University of Luxembourg, Luxembourg b c

a r t i c l e

i n f o

Article history: Received 4 March 2016 Received in revised form 27 September 2016 Accepted 28 September 2016 Available online 01 October 2016 Keywords: Interoception Heart rate Interoceptive accuracy Interoceptive sensibility Interoceptive awareness

a b s t r a c t Garfinkel and Critchley (2013) recently proposed a three level model of interoception. Only few studies, however, have empirically tested this theoretical model thus far. The present study aimed at investigating (1) the central assumptions of this model, i.e. that Accuracy, Sensibility and Awareness are distinguishable facets of interoception and that Interoceptive Accuracy is the basic level of interoception, and (2) whether cardiovascular activation (as indexed by heart rate) is differentially related to the three facets of interoception. Analyses were conducted on a total sample of N = 159 healthy participants (118 female [74.2%]; mean age = 23.9 years, SD = 3.3, range = 19–45) who performed either the heartbeat tracking task, the heartbeat discrimination task or both. The results suggest that Accuracy, Sensibility and Awareness are empirically distinct facets of interoception, showing no correlation when based on heartbeat tracking, but moderate correlations when based on heartbeat discrimination. The assumption that Interoceptive Accuracy is the basic level of interoception could only be partially confirmed. Instead, we conclude that the level of objective physiological states should be considered as the most basic level of interoceptive signal processing. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Interoception, i.e. the sense of the physiological condition of the body (Craig, 2003), is related to various mental and behavioral processes, such as emotion processing, emotion regulation, and intuitive decision-making (Bechara and Naquvi, 2004; Pollatos and Schandry, 2008; Fustos et al., 2013; Sütterlin et al., 2013; Terasawa et al., 2013). Previous research on interoception, however, often suffered from conceptual vagueness. Only recently, Garfinkel et al. (2015) suggested clear and concise definitions for terms such as “interoceptive awareness” and “interoceptive accuracy”, which previously were often used interchangeably (Dunn et al., 2010; O'Brien et al., 1998; Terhaar et al., 2012), whereas others treated these terms separately, which may impede comparisons between outcomes of different studies (Ceunen et al., 2013). Moreover, Garfinkel and Critchley (2013) proposed a multilevel conceptualization of interoception that distinguishes between three facets of the construct: 1. Interoceptive Sensibility, 2. Interoceptive Accuracy and 3. Interoceptive Awareness.

⁎ Corresponding author at: Institute of Medical Psychology and Medical Sociology, University Hospital of RWTH Aachen, Pauwelsstraße 19, 52074 Aachen, Germany. E-mail address: [email protected] (T. Forkmann).

http://dx.doi.org/10.1016/j.ijpsycho.2016.09.019 0167-8760/© 2016 Elsevier B.V. All rights reserved.

Interoceptive Sensibility refers to a dispositional tendency to be internally focused. This term captures self-reported beliefs about body sensations, which are typically assessed via self-report measures such as questionnaires (Mehling et al., 2012; Porges, 1993). Garfinkel et al. (2015) also use the mean of individual confidence ratings in interoceptive accuracy tasks as indicator of interoceptive sensibility. The second level of Garfinkel and Critchley's (2013) model is named Interoceptive Accuracy, which refers to “objective” behavioral tests of interoception. There are two main approaches for the assessment of Interoceptive Accuracy: a) the tracking method, originally proposed by Schandry (Dunn et al., 2007; Herbert et al., 2012; Schandry, 1981), which assesses a person's accuracy in detecting their heartbeats by counting them in a given time interval, and b) the signal discrimination method, which presents a series of external stimuli (tones, lights or tactile stimuli) and asks the participant to judge whether the stimuli are simultaneous or delayed relative to one's own heartbeat (Whitehead and Drescher, 1980; Whitehead et al., 1977). The tracking task has been criticized for its results being influenced by expectancies or guesses about heart rate or other factors such as attention or motivation (e.g., Windmann et al., 1999; Ring et al., 2015). Even though both established heartbeat perception tasks involve different processes (e.g. attention focusing on visceral sensations to perform heartbeat tracking, but attention focusing on visceral and external sensations to perform heartbeat discrimination;

72

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

Schulz et al., 2013), many studies have found moderate correlations between indices derived from both tasks (Knoll and Hodapp, 1992; Schaefer et al., 2012; Hart et al., 2013). In addition to Interoceptive Sensibility and Interoceptive Accuracy, Garfinkel and Critchley (2013) consider Interoceptive Awareness as a third level. This aspect of interoception can be assessed via metacognitive judgments on Interoceptive Accuracy and refers to the extent of an individual's confidence ratings predicting their own actual interoceptive performance (Garfinkel and Critchley, 2013). Garfinkel and Critchley (2013) point out that most of the existing literature refers to the first or second facet of interoception (Sensibility or Accuracy), and particularly Interoceptive Accuracy has been demonstrated to be related to several mental disorders such as eating disorders (Pollatos et al., 2008), panic disorder (Ehlers and Breuer, 1996), depersonalization/ derealization disorder (Schulz et al., 2015) and depression (Terhaar et al., 2012). Recently, Yoris et al. (2015) have reported that metacognitive beliefs and worries about interoception and not interoceptive performance in a heartbeat detection task discriminated between patients with panic disorder and healthy controls, pointing at the importance of metacognitive interoceptive processing for the understanding of the role of interoception in mental disorders. There are only two studies so far which have investigated the question as to whether the three facets of the model are distinguishable and how they relate to each other empirically. The authors of the model themselves investigated the relationship between the three facets of interoception in a sample of 80 healthy participants (Garfinkel et al., 2015). In a stepwise forward linear regression analysis, Interoceptive Sensibility and Awareness predicted Interoceptive Accuracy, and these associations were lowered when Interoceptive Awareness was entered as the dependent variable instead. The authors interpreted their results as underpinning the potential independence of the three facets of interoception as proposed in their model and as in line with their assumption of the primacy of Interoceptive Accuracy: only after a basic accuracy threshold is overcome, a correspondence between the different facets of interoception would emerge (Garfinkel et al., 2015). A second study (N = 25) by Meessen et al. (2016) found the three facets of interoception to be uncorrelated. The generalizability of the results of this study, however, may be limited because the sample investigated was small and no data on heartbeat discrimination paradigms was assessed. Taken together, both studies suggest a partial independence of the three facets of interoception as proposed by the model introduced by Garfinkel and Critchley (2013), and larger associations between these facets in people high in Interoceptive Accuracy. Several unresolved issues remain, however. Firstly, with only two published studies the empirical evidence is sparse. Additional studies with reasonably sized datasets on the relation between different facets of the model are, therefore, needed. Particularly data on the association between Interoceptive Accuracy assessed using a heartbeat discrimination paradigm and Interoceptive Awareness are lacking. Secondly, neither Garfinkel et al. (2015) nor Meessen et al. (2016) controlled whether the three facets of interoception were related to indices of visceral-afferent neural signal transmission, which represent a necessary prerequisite for its perception (Vaitl, 1996). Cardiovascular activation by exercise (Pollatos et al., 2007; Schandry et al., 1993) or laboratory stressors (Fairclough and Goodwin, 2007; Schulz et al., 2013) has been shown to affect Interoceptive Accuracy, which may be due to altered transmission of visceral-afferent neural signals during cardiovascular activation (e.g., as indexed by increase of heart rate, blood pressure, etc.). Furthermore, interoception is affected by indices of baseline cardiovascular activation, in that lower resting heart rate is associated with higher Interoceptive Accuracy (Knapp-Kline and Kline, 2005). If empirical results support the assumption that visceral-afferent signal transmission is related to higher levels of interoception, the objective physiological state or process (such as resting heart rate) may constitute the most basic level underlying the processing of internal signals, which

is required for interoception, and could, therefore, be integrated as the fourth element in the model introduced by Garfinkel and Critchley (2013). The present study investigated the relationship between Interoceptive Accuracy, assessed with both a heartbeat tracking and a heartbeat discrimination task, Interoceptive Sensibility and Interoceptive Awareness. As primary hypothesis we expected (I) to replicate previous findings (Garfinkel et al., 2015) in support of the three-facet model of interoception (Garfinkel and Critchley, 2013). Support for this model would be found in three distinguishable facets of interoception with Interoceptive Accuracy as basic construct. In line with Garfinkel et al. (2015), distinctness is indicated by zero to mild correlations. In hypothesis (Ia) we, therefore, expected zero to mild correlations between the three facets and higher correlations between “Accuracy” and “Sensibility” than between “Accuracy” and “Awareness”. As we expected “Accuracy” as the most basic construct, we hypothesized (Ib) associations between interoceptive facets (Accuracy, Sensibility and Awareness) to be higher in individuals high in Interoceptive Accuracy compared to individuals low in Interoceptive Accuracy. Thus, Interoceptive Accuracy is expected to act as moderator of these associations. According to hypothesis (Ic) we expected Interoceptive Sensibility and Awareness to predict Interoceptive Accuracy in a linear regression analysis and that these relations should be weakened when a reversed regression model was calculated. In hypothesis (II), we expected heart rate (as an index of cardiovascular activation) to be differentially related to the three facets of interoception. If significant relations between these variables would emerge, implications for a potential integration of heart rate into the model proposed by Garfinkel and Critchley would be discussed. As secondary hypotheses, we expected (III.) accuracy and sensibility to be related across tasks (heartbeat tracking and heartbeat discrimination) and (IV.) that the use of a specific formula to calculate Interoceptive Accuracy would not affect outcome. For this purpose, we compared the formula proposed by Hart et al. (2013), that was used by Garfinkel et al. (2015) with the formula proposed by Schandry (1981). 2. Material and methods 2.1. Participants Participants were 159 healthy individuals (118 female [74.2%]; mean age = 23.9 years, SD = 3.3, range = 19–45) who took part in the five separate experiments contributing to this study. See Table 1 for details on the demographic characteristics of the subsamples. Data from experiments three and four have been published previously, addressing research questions other than the relationship between Interoceptive Accuracy and Awareness (Schulz et al., 2013; Michal et al., 2014). The present study was conducted in accordance with the declaration of Helsinki and was approved by the local ethics committees of the Universities of Trier and Mainz (Germany), where the studies were conducted. All participants gave written informed consent. 2.2. Interoceptive accuracy tasks 2.2.1. Heartbeat discrimination task An interoceptive-exteroceptive discrimination task was initially developed by Brener and Jones (1974), but further elaborated by others (Brener and Kluvitse, 1988; Katkin et al., 1981). Participants are asked to judge whether exteroceptive stimuli appear either “synchronously” or “delayed” to their own heartbeats. Exteroceptive stimuli were elicited with a latency of 230 ms (S+ trials, synchronous) or 530 ms (S− trials, delayed) after the participants' actual R-wave. The task was implemented in an auditory version in which a sinus tone (440 Hz) of 80 ms duration was used. During each trial, six consecutive stimuli with the same latency were presented. Participants completed test trials

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

73

Table 1 Descriptive statistics of study variables. Heartbeat discrimination task (HDT)

Significance HDT Heartbeat tracking task (HTT)

Exp 1

Exp 2

Exp3

Overall

Sample (N) Number of trials Seconds per trial

33 – –

33 – –

38 – –

104 – –

% Female (n)

X2 = 14.939 (p = 0.001) 22.88 (3.3) 23.48 (2.6) 23.00 (2.6) 23.12 (2.8) F = 0.418 (p = 0.659) 0.53 (0.86) 0.83 (0.90) 0.40 (1.1) 0.58 (1.0) F = 1.843 (p = 0.164)

Mean age (SD)

100 (33)

Accuracy d′/HPS (M/SD) (S = Schandry, H = Hart) Sensibility (M/SD) 5.2 (1.2) Awareness (S = Schandry, H = Hart) Heart rate

63.6 (21)

67.5 (27)

76.4 (81)

4.5 (1.1)

4.2 (1.5)

4.6 (1.4)

71.82 (12.93)

72.49 (10.14)

73.00 (11.38)

– – –

F = 5.574 (p = 0.005) 0.54 (0.13) 0.63 (0.11) 0.54 (0.14) 0.56 (0.13) F = 5.016 (p = 0.008)

74.26 (11.37)

F = 0.337 (p = 0.687)

and 40 experimental trials each (20 synchronous and 20 delayed). Discrimination performance was calculated using the d′ parameter derived from signal detection theory (Wickens, 2001). Correctly identified synchronous trials (S +) were defined as “hits”, delayed trials (S −) that were incorrectly judged as S + trials were defined as “false alarms”. Thus, the parameter of interoceptive accuracy was calculated using the following formula: 0

Exp 4

Exp5

Overall

33 3 30, 45, 60

27 6 25, 35, 45, 55, 65, 75 48.1 (13)

26 4 25, 35, 45, 55

86 – –

92.3 (24)

67.4 (58)

According to Schandry (1981): HPS ¼

  1 ðjrecorded heartbeats−perceived heartbeatsjÞ n ∑k¼1 1− n recorded heartbeats

The HPS typically takes on values from 0 to 1, where 1 depicts perfect accuracy. Negative values are possible if number of perceived heartbeats exceeds 200% of recorded heartbeats. According to Hart et al. (2013): HPS ¼

2.2.2. Heartbeat tracking task For the Heartbeat tracking task by Schandry (1981), participants were asked to monitor their heart beating and count the beats silently. They were not allowed to take their pulse and watches had to be removed beforehand. The task instruction was presented on a computer screen. An acoustic cue signaled the beginning and end of each trial. Again, after each trial, participants were asked to indicate the number of perceived heartbeats and subsequently were asked to give a judgment of confidence concerning how sure they were that their estimation was correct (JOC-Tracking) on a scale ranging from zero (not sure at all) to eight (absolutely sure). They received no feedback about their performance and participants were not told the lengths of the tracking phases. Depending on the experiment, three to six test trials of this task varying in length were performed. Participants also started each experiment with a test trial that was not used for analysis. The order of presentation of trials was randomized with trial duration ranging between 25 and 75 s (see Table 1 for details on the number of trials and duration per experiment). Heartbeat perception scores (HPS) were calculated according to two slightly different formulas (with n = number of trials in the respective experiment).

– – –

X2 = 12.116 (p = 0.002) 23.48 (2.6) 27.00 (4.1) 24.04 (1.9) 24.8 (3.3) F = 11.26 (p b 0.001) S: 0.77 (0.20); S: 0.73 (0.19); S: 0.75 (0.23); S: 0.75 (0.20); S: F = 0.225 H: 0.70 (0.28) H: 0.65 (0.27) H: 0.68 (0.32) H: 0.68 (0.29) (p = 0.799) H: F.211 (p = 0.810) 4.2 (1.6) 3.9 (1.8) 4.0 (1.6) 4.1 (1.6) F = 0.376 (p = 0.688) S: 0.25 (0.74) S: 0.29 (0.46) S: 0.31 (0.48) S: 0.27 (0.58) S: F = 0.088 H: 0.24 (0.75) H: 0.29 (0.46) H: 0.30 (0.48) H: 0.28 (0.58) (p = 916) H: F = 0.084 (p = 0.920) 71.82 74.59 83 76 F = 6.942 (12.93) (7.67) (15.09) (13.14) (p = 0.002) 63.6 (21)

d ¼ zhits –zfalse alarms

After each trial, participants were asked to give a judgment of confidence concerning how sure they were that their decision was correct (JOC-Discrimination) on a scale ranging from zero (not sure at all) to eight (absolutely sure). In congruence with the method to determine the accuracy score, the JOC score was only used for trials judged – correctly (“hits”) or incorrectly (“false alarms”) – to be synchronous. Participants received no feedback on their performance.

Significance HTT

Exp 2

  1 ðjrecorded heartbeats−perceived heartbeatsjÞ n ∑k¼1 1− n ½ðjrecorded heartbeats þ perceived heartbeatsjÞ=2

Here, HPS ranges between − 1 and + 1. Moderate differences between number of recorded and perceived heartbeats result in values between 1 and 0, whereas negative values are possible with extreme differences (e.g., perceived ≤33% or ≥300% of recorded heartbeats). 3. Apparatus The ECG signal was monitored using CG Tyco Healthcare H34SG Ag/AgCl electrodes (diameter: 45 mm), recorded with a Biopac MP150 amplifier system, high pass filtered (0.5 Hz) and stored on a disc (sampling rate: 1 kHz) for analysis. R-waves were identified online by a DASYLAB (National Instruments Inc.) based algorithm, since online R-wave detection was required for the heartbeat detection paradigm. Accuracy of R-wave detection in sinus rhythm was higher than 99.8% with a latency below 3 ms (internal lab report). Heart rate as indicator of objective physiological state, was extracted from the ECG signal via automatic R-wave detection algorithm (WinCPRS, Absolute Aliens Oy, Turku, Finland) and subsequent manual inspection. 3.1. Procedure Upon arrival at the laboratory room, participants gave written informed consent to participate in the study, and were fitted with the electrodes. Electrodes for ECG-measurement were placed according to a standard lead II configuration. Headphones (Sennheiser Electronic GmbH & Co. KG, Wedemark, Germany) were placed, and participants were informed about the experimental procedures on a computer display. Participants then completed either the heartbeat discrimination task (Experiments 1 and 3), the heartbeat tracking task (Experiments 4 and 5) or both in permutated order (Experiment 2). Depending on

74

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

the experiment, they afterwards completed other tasks that are not reported here. 3.2. Data analyses Analyses were conducted using SPSS 22 for Windows (2013, IBM, Somers, NY). Since the present study is partly a replication of the study by Garfinkel et al. (2015), data analysis was largely based on that study. Setting α at 0.05 and 1 − β at 0.80, power analyses indicated a minimum sample size of N = 104 for the heartbeat discrimination task, to detect effects of r ≥ 0.270, and a minimum sample size of N = 86 for the heartbeat tracking task to detect effects of r ≥ 0.297. Where applicable, all reported results reflect two-tailed p-values. To determine a priori group differences between the subsamples from the different experiments with respect to demographic variables, interoceptive accuracy and confidence judgments, we employed a one factorial analysis of variance (ANOVA) and a chi-square test with “source experiments” (1–5) as between-subjects variable. Interoceptive Accuracy was determined by calculating mean performance in the heartbeat tracking task and the heartbeat discrimination task separately. Interoceptive Sensibility was expressed as the mean of confidence judgments (Garfinkel et al., 2015). To obtain a score of Interoceptive Awareness as metacognitive measure as defined by Garfinkel et al. (2015), we calculated the congruency between accuracy and judgments of confidence. For the heartbeat discrimination task, the area under the curve (AUC) in a receiver operating characteristics curve analysis was calculated for each participant using the accuracy of the answers as binary outcome and the confidence judgments as dimensional predictor. For the heartbeat tracking task, we calculated Pearson's correlation coefficients between judgments of confidence and performance over all trials. One-sample t-tests were conducted to determine whether Accuracy, Sensibility and Awareness were significantly different from zero (or different from 0.50 for the AUC scores, respectively). To investigate whether Interoceptive Accuracy was more strongly correlated with Sensibility than with Awareness (hypothesis Ia) and whether Accuracy and Sensibility correlated across tasks (hypothesis III), correlations were calculated between all three levels according to Garfinkel and Critchley's (2013) model of interoception, and each participant's mean heart rate. Garfinkel et al. (2015) stipulate a threshold of Interoceptive Accuracy to be passed, i.e. a basic level of Interoceptive Accuracy is a prerequisite for “higher” levels of interoception (Sensibility and Awareness) to be implemented. They further assume that correlations between the three dimensions of their model would be larger in people high in Interoceptive Accuracy relative to those low in Interoceptive Accuracy. In contrast to Garfinkel and Critchley (2015) in the current study we tested hypothesis (Ib) with a hierarchical multiple regression model, which allowed us to keep Accuracy as continuous variable. In a first step, only the predictor (X) was introduced into the model (Model I: X = Accuracy; criterion [Y] = Sensibility; Model II: X = Accuracy; Y = Awareness; Model III: X = Sensibility; Y = Awareness). As we expected a threshold effect in that the association between X and Y was higher in participants showing high Accuracy than in those showing low Accuracy, a quadratic term of Accuracy (i.e. relationship between Accuracy and Y at different levels of Accuracy) was introduced in the second step of the regression models (Models I and II: Y = X + X2). For Model III, in the second step Accuracy as a moderator (M) and interaction effects between X and M were introduced (Y = X + M + X × M + M2 + X × M2). All data (Accuracy, Sensibility, Awareness) were z-transformed before they were entered in these analyses. Models I to III were calculated for heartbeat tracking task data and repeated for heartbeat discrimination task data. Finally, two linear regression analyses with a stepwise forward inclusion procedure were performed to examine whether Interoceptive Accuracy was the basic construct underlying Sensibility and Awareness

(hypothesis Ic). In the first analysis, Interoceptive Accuracy was the criterion and Interoceptive Awareness and Sensibility were used as predictors. In the second analysis, Interoceptive Awareness served as criterion, and Interoceptive Accuracy and Sensibility as predictors. According to Garfinkel et al. (2015) the empirical relationships should be larger in the first than in the second analysis and both Interoceptive Sensibility and Interoceptive Awareness should be significant predictors of variance in Interoceptive Accuracy. To examine mean heart rate as the most basic level of interoception (hypothesis II), these regression analyses were repeated with mean heart rate as fourth element. Firstly, mean heart rate was the criterion and Interoceptive Accuracy, Sensibility and Awareness were the predictors. Secondly, Interoceptive Awareness was the criterion and Sensibility, Accuracy and mean heart rate were the predictors. 4. Results Descriptive statistics (means, standard deviations) of all variables used in the analyses are summarized in Table 1. There were significant differences between the subsamples (i.e., the samples from the five different source experiments) with regard to age, sex distribution and resting heart rate (heartbeat discrimination task: sex distribution: χ2 = 14.939, df = 2, p = 0.001; heartbeat tracking task: sex distribution: χ2 = 12.116, df = 2, p = 0.002; age: F = 11.26, df = 2, p b 0.001; heart rate: F = 6.942, p = 0.002). There were also differences in the key variables “Interoceptive Sensibility” and “Interoceptive Awareness” across groups for the heartbeat discrimination task (sensibility: F = 5.574, df = 2, p = 0.005; awareness: F = 5.016, df = 2, p = 0.008) and in heart rate for the heartbeat tracking task: F = 6.942, df = 2, p = 0.002. 4.1. Correlation analyses between interoceptive accuracy, sensibility, awareness and objective physiological state (heart rate) Means and standard deviations (SD) for Interoceptive Accuracy, Sensibility and Awareness are summarized in Table 2. Mean performances in all three facets of interoception were significantly different from random performance (heartbeat tracking task: Accuracy: Schandry: M = 0.75, SD = 0.20; t = 34.335, df = 85, p b 0.001; Hart: M = 0.68 SD = 0.29; t = 21.837, df = 85, p b 0.001; Sensibility: M = 4.05, SD = 1.62; t = 23.231, df = 85, p b 0.001; Awareness: Schandry: M = 0.28, SD = 0.59; t = 4.258, df = 79, p b 0.001; Hart: M = 0.28, SD = 0.58; t = 4.202, df = 79, p b 0.001. Heartbeat Discrimination Task: Accuracy: M = 0.58, SD = 0.97; t = 6.133, df = 105, p b 0.001; Sensibility: M = 4.59; SD = 1.33; t = 35.520, df = 105, p ≤ 0.001 Awareness: M = 0.57, SD = 0.13; t = 5.209, df = 104, p b 0.001). To investigate hypothesis Ia, correlation analyses were conducted. Table 2 gives an overview of all correlations between the three different levels of interoception regarding both tasks, heartbeat tracking and heartbeat discrimination, and mean heart rate. For Interoceptive Accuracy, both scores (according to Schandry (1981) and according to Hart et al. (2013)) are included. For the heartbeat discrimination task, there was a significant correlation between Accuracy and Interoceptive Sensibility (r = 0.386, n = 106, p b 0.001). For the heartbeat tracking task, no such correlation could be found (Schandry: r = 0.176, n = 86, p = 0.105; Hart: r = 0.186, n = 86, p = 0.086). Among the small subsample (sample 2) that completed both tasks, a significant correlation between both mean confidence judgments (Sensibility) could be found (r = 0.348, n = 33, p = 0.047). Accuracy indices from either task did not correlate significantly in this subsample (Schandry and Hart both: r = 0.072, n = 33, p = 0.691; see Fig. 1). For Interoceptive Awareness on the heartbeat discrimination task, there was a significant correlation with Interoceptive Accuracy on the same task (r = 0.577, n = 105, p b 0.001), but the correlation

Table 2 Correlations between the three facets of interoception and heart rate.

Accuracy heartbeat tracking (according to Schandry) Accuracy heartbeat tracking (according to Hart) Awareness heartbeat tracking (according to Schandry) Awareness heartbeat tracking (according to Hart) Sensibility heartbeat tracking Accuracy heartbeat discrimination Awareness heartbeat Discrimination Sensibility heartbeat discrimination Mean heart rate

Mean and standard deviation

r p n r p n r p n r p n r p n r p n r p n r p n r p n M SD

Accuracy heartbeat tracking (according to Hart)

Awareness heartbeat tracking (according to Schandry)

Awareness heartbeat tracking (according to Hart)

Sensibility heartbeat tracking

Accuracy heartbeat discrimination

Awareness heartbeat discrimination

Sensibility heartbeat discrimination

Mean heart rate

1

0.987** 0.000 86 0.157 0.166 79 0.158 0.163 79 0.176 0.105 86 0.072 0.691 33 −0.191 0.286 33 0.195 0.277 33 −0.168 0.123 86 0.75 0.20

1

0.132 0.246 79 0.132 0.245 79 0.186 0.086 86 0.072 0.691 33 −0.191 0.286 33 0.212 0.236 33 −0.175 0.108 86 0.68 0.29

1

1.000 0.000 79 −0.023 0.842 79 −0.063 0.743 30 −0.160 0.399 30 −0.237 0.208 30 0.236⁎ 0.036 79 0.28 0.59

1

−0.023 0.843 79 −0.063 0.743 30 −0.160 0.399 30 −0.233 0.215 33 0.238⁎ 0.034 79 0.28 0.58

1

0.098 0.588 33 0.283 0.111 33 0.348⁎ 0.047 33 −0.346⁎⁎ 0.001 86 4.05 1.62

1

0.577⁎⁎ 0.000 105 0.386⁎⁎ 0.000 106 0.09 0.350 106 0.58 0.97

1

0.145 0.139 105 −0.02 0.837 105 0.57 0.13

1

0.061 0.532 106 4.59 1.33

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

Accuracy heartbeat tracking (according to Schandry)

1

75 12.10

75

76

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

Fig. 1. Correlations between the two measures of Interoceptive Accuracy (A) and the two measures of Interoceptive Sensibility (B) across tasks for subsample 2.

with Interoceptive Sensibility was not significant (r = 0.145, n = 105, p = 0.139). Interoceptive Awareness of heartbeat tracking showed neither a significant correlation with Interoceptive Accuracy (Schandry: r = 0.157, n = 79, p = 0.166; Hart: r = 0.132, n = 79, p = 0.245), nor with Interoceptive Sensibility (Schandry: r = − 0.023, n = 105, p = 0.842; Hart: r = −0.023, n = 105, p = 0.843). Comparing the correlation between Interoceptive Accuracy and Interoceptive Sensibility with the correlation between Interoceptive Accuracy and Interoceptive Awareness in the heartbeat discrimination

task, we found the latter correlation to be larger, and for this effect to be marginally significant (Fisher's z = 1.8, p = 0.072). For the heartbeat tracking task, no significant difference could be found for the correlations between Interoceptive Accuracy and Interoceptive Sensibility and the correlation between Interoceptive Accuracy and Interoceptive Awareness (Schandry: Fischer's z = 0.12, p = 0.91; Hart: Fisher's z = 0.35, p = 0.73). In the subsample that performed both Interoceptive Accuracy tasks (sample 2), heartbeat discrimination Awareness and heartbeat tracking

Table 3 Results of hierarchical linear regression analyses. Model

steps

Heartbeat discrimination task I X = Accuracy Y = Sensibility

step I step II

II

X = Accuracy Y = Awareness

step I step II

III

Y = Awareness X = Sensibility M = Accuracy

Heartbeat tracking task I X = Accuracy Y = Sensibility

step I step II

step I step II

II

X = Accuracy Y = Awareness

step I step II

III

Y = Awareness X = Sensibility M = Accuracy

step I step II

stand. β constant X constant X X2 constant X constant X X2 constant X constant X M X×M M2 X × M2

constant X constant X X2 constant X constant X X2 constant X constant X M X×M M2 X × M2

0.39 0.33 0.11 0.58 0.54 0.13 0.15 −0.09 0.52 0.06 0.08 0.08

0.18 0.19 0.02 0.16 0.24 0.18 −0.02 −0.25 0.25 −0.01 0.28 0.28

t

p

adj. R2

F

p

0.00 4.27 −0.54 3.22 1.04 0.30 7.17 −0.69 6.37 1.52 0.04 1.49 −0.62 −0.85 5.29 0.57 0.79 0.67

1.00 0.00 0.59 0.00 0.30 0.77 0.00 0.49 0.00 0.13 0.97 0.14 0.54 0.40 0.00 0.57 0.43 0.50

0.14

18.22

0.00

0.14

9.66

0.00

0.33

51.35

0.00

0.33

27.17

0.00

0.01

2.22

0.14

0.32

10.95

0.00

0.00 1.64 −0.11 1.50 0.16 −0.02 1.40 −1.04 1.92 1.44 0.02 −0.20 −1.43 −1.40 1.88 −0.11 1.98 1.40

1.00 0.11 0.91 0.14 0.87 0.99 0.17 0.30 0.06 0.15 0.99 0.84 0.16 0.17 0.06 0.92 0.05 0.17

0.02

2.68

0.11

0.01

1.34

0.27

0.01

1.95

0.17

0.03

2.02

0.14

−0.01

0.04

0.84

0.02

1.37

0.25

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

Awareness did not correlate significantly (Schandry and Hart both: r = −0.160, n = 30, p = 0.399). In the heartbeat tracking task, mean heart rate was positively associated with Interoceptive Awareness (Schandry: r = 0.236, n = 79, p = 0.036, Hart: r = 0.238, n = 79, p = 0.034), but negatively with Interoceptive Sensibility (r = − 0.346, n = 86, p = 0.001). No significant bivariate correlation was found with Interoceptive Accuracy. No substantial differences were found between measures calculated based on the formula proposed by Schandry (1981) compared to the formula proposed by Hart et al. (2013). In the heartbeat discrimination task, no significant bivariate correlations between heart rate and the three facets of Interoception could be found. 4.2. Threshold effect of interoceptive accuracy – hierarchical multiple regression analyses 4.2.1. Threshold effect of interoceptive accuracy in heartbeat discrimination task To examine whether Interoceptive Accuracy was the basic level of interoceptive processing (hypothesis Ib), hierarchical multiple regressions were calculated with Accuracy as quadratic term to indicate a possible threshold effect (i.e. higher association between predictor and criterion in individuals with high than with low Accuracy). In Model I (Y = Sensibility; X = Accuracy), both in the first step including Accuracy as predictor only (adj. R2 = 0.14; F[1104] = 18.22; p b 0.001) and in the second step including the threshold effect in addition (adj. R2 = 0.14; F[2103] = 9.66; p b 0.001), the criterion was significantly predicted by its predictors. In both steps, however, only the predictor Accuracy (linear) was significantly associated with the criterion (step 1: stand. β = 0.39; T = 4.27; p b 0.001; step 2: stand. β = 0.33; T = 3.22; p = 0.002). The same pattern was observed in Model II (Y = Awareness; X = Accuracy). Again, the overall models were significant (step 1: adj. R2 = 0.33; F[1103] = 51.35; p b 0.001; step 2: adj. R2 = 0.33; F[2102] = 27.17; p b 0.001), but only the predictor Accuracy (linear) was significantly associated with the criterion (step 1: stand. β = 0.58; T = 7.17; p b 0.001; step 2: stand. β = 0.54; T = 6.37; p b 0.001). Regarding Model III (Y = Awareness; X = Sensibility; M = Accuracy), in step 1 the overall model was not significant if only including the predictor X (adj. R2 = 0.01; F[1103] = 2.22; p = 0.14). In step 2 the overall model including the moderator and all interaction effects was significant (step 2: adj. R2 = 0.32; F[5,99] = 10.95; p b 0.001), but again, only Accuracy (linear) was significantly associated with the criterion (stand. β = 0.52; T = 5.29; p b 0.001) (for a summary see Table 3). Taken together, in all models only the predictor Accuracy showed statistical significance, whereas all other terms did not significantly contribute to the prediction of the criterion. 4.2.2. Threshold effect of interoceptive accuracy in heartbeat tracking task When the same regression analyses were based on heartbeat tracking, none of the Models I–III was significant (all ps N 0.10; see Table 3 for details). 4.3. Is interoceptive accuracy the basic level? – Regression analyses 4.3.1. Heartbeat tracking task To further analyze whether Interoceptive Accuracy was the basic construct of interoceptive processing (hypothesis Ic) a series of regression analyses was conducted. In the first regression analysis, Interoceptive Accuracy was the criterion variable and Interoceptive Awareness and Sensibility were entered as predictor variables. Neither Interoceptive Awareness (β = 0.161, p = 0.154) nor Interoceptive Sensibility (β = 0.157, p = 0.165) were significant predictors in this model (adj. R2 = 0.024, F(df = 2/76) = 1.971, p = 0.146). In the second regression analysis, Interoceptive Awareness was the criterion variable and Interoceptive Sensibility and Accuracy served as predictors. Again, neither

77

Interoceptive Sensibility (β = −0.048, p = 0.677) nor Interoceptive Accuracy (β = 0.165, p = 0.154) significantly predicted Interoceptive Awareness (adj. R2 = 0.001, F(df=2/76) = 1.055, p = 0.353). In a second pair of regression analyses, mean heart rate was added to investigate whether cardiovascular activation (as indexed by heart rate) was differentially related to the three facets of the model (hypothesis II). Firstly, mean heart rate was entered as criterion variable and Interoceptive Accuracy, Sensibility and Awareness were entered as predictors. Interoceptive Sensibility (β = − 0.319, p = 0.003) and Interoceptive Awareness (β = 0.257, p = 0.016) but not Interoceptive Accuracy were significant predictors (adj. R2 = 0.175, F(df = 3/75) = 6.502, p = 0.001). Secondly, Interoceptive Awareness was entered as criterion variable. In this model, only mean heart rate (β = 0.292, p = 0.016) was a significant predictor for Interoceptive Awareness (adj. R2 = 0.064, F(df=3/75) = 2.776, p = 0.047). 4.3.2. Heartbeat discrimination task In the first regression analysis with Interoceptive Accuracy as criterion variable, both Interoceptive Sensibility (β = 0.250, p = 0.002) and Interoceptive Awareness (β = 0.541, p b 0.001) were significant predictors of Interoceptive Accuracy (adj. R2 = 0.382, F(df =2/102) = 33.117, p b 0.001). In the second regression analysis with Interoceptive Awareness as criterion variable, in line with the expectations, a slightly different picture emerged: Interoceptive Sensibility was still no significant predictor (β = − 0.049, p = 0.565), but Interoceptive Accuracy significantly predicted Interoceptive Awareness (β = 0.593, p b 0.001). This model explained slightly less variance in the criterion variable (adj. R2 = 0.322, F(df=2/102) = 25.667, p b 0.001). Again, in a second pair of regression analyses (examining hypothesis II), mean heart rate was added. Firstly, mean heart rate was entered as criterion variable and Interoceptive Accuracy, Sensibility and Awareness were entered as predictors. None of these variables were significant predictors of heart rate (adj. R2 = −0.019, F(df=3/101) = 0.362, p = 0.781). Secondly, Interoceptive Awareness was entered as criterion variable. In this model, only Interoceptive Accuracy (β = 0.597, p b 0.001) was a significant predictor for Interoceptive Awareness (adj. R2 = 0.319, F(df=3/101) = 17.219, p b 0.001). 5. Discussion The aim of the current study was (1) to investigate the relationship between the interoceptive facets ‘accuracy’, ‘sensibility’ and ‘awareness’ as introduced by Garfinkel and co-workers (2015) and (2) to investigate their relationship to cardiovascular activation, as indexed by mean heart rate. We also examined whether differences in Interoceptive Accuracy depend on the algorithms used for the calculation of this measure (Hart et al., 2013; Schandry, 1981). 5.1. Study aim I: replication of previous findings in support of the threefacet model of interoception Garfinkel et al. (2015) postulated three distinguishable facets of interoception, which show no or if any, only mild inter-correlations, with Interoceptive Accuracy being the most basic component. Hypothesis I aimed at replicating previous findings in support of this theory. In particular, in hypothesis (Ia) we expected zero to mild correlations between the three facets and higher correlations between Accuracy and Sensibility than between Accuracy and Awareness. In partial support of this hypothesis, we observed moderate correlations between Accuracy and Sensibility (r = 0.386), as well as Accuracy and Awareness (r = 0.577) when evaluating indices of the heartbeat discrimination task. We found, however, no correlations between any facets as assessed by the heartbeat tracking task. Accuracy in heartbeat discrimination has previously been reported to also show positive correlations with Sensibility as assessed by self-reports in the Body Perception Questionnaire (BPQ) (Fairclough and Goodwin, 2007; Porges, 1993; Schulz et

78

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

al., 2013). Based on the present results we would argue that the three interoceptive facets posited by Garfinkel et al. (2015) are distinct, yet share more common variance than has previously been reported, depending on the interoception task employed. The differences in both interoceptive tasks with regard to the relationship between the interoceptive facets may be explained by the psychological processes that are involved to fulfil the respective task: in the heartbeat tracking task, attention needs to be focused on visceral sensations, and its indicators may be affected by beliefs about one's actual heart rate (Ring et al., 2015). In the heartbeat discrimination task attention has to be focused on both visceral and external sensations at the same time, and both need to be integrated effectively (Suzuki et al., 2013). One may argue, therefore, that the higher correlation between behavior (Accuracy), confidence judgments (Sensibility) and their convergence (Awareness) is due to the multisensory integration component in the discrimination task. In particular, confidence judgments in heartbeat discrimination may be related to Awareness of multisensory (interoceptive-exteroceptive) integration, whereas confidence judgments in heartbeat tracking may solely be based on Awareness of interoceptive signals, but could be biased by beliefs about one's own heart rate and the subjective difficulty of judging the length of time intervals (Ring and Brener, 1996; Yates et al., 1985). In contrast to the current study, Garfinkel et al. (2015) found low correlations between Sensibility and Accuracy only for heartbeat tracking, whereas no associations were observed within the indicators based on heartbeat discrimination. Although both Garfinkel et al. (2015) and the current study report positive correlations (if any), potentially diverging findings may still be due to the differences in methodology between both studies (e.g., number of trials; Likert scale vs. VAS; paperpencil vs. electronic confidence assessment, etc.). In contrast to our hypothesis, the correlation between Accuracy and Sensibility was not higher than between Accuracy and Awareness. This finding does not support the model's assumption that Accuracy and Sensibility represent subsequent hierarchical levels, whereas Awareness is located above the others (Garfinkel and Critchley, 2013; Garfinkel et al., 2015). This suggests that the Awareness of multisensory integration plays a seminal role for the inter-relationships of the three interoceptive facets. Regarding the heartbeat tracking task, it seems that task specific aspects such as the ability to estimate time intervals might be important so that no correlations occurred. The central assumption of the model of interoception as proposed by Garfinkel et al. (2015) posits Interoceptive Accuracy as the basic construct. This assumption could be reflected in the relations between Interoceptive Accuracy, Sensibility and Awareness, which should be larger for people high compared to people low in Interoceptive Accuracy (“threshold effect”). This relationship was tested via a set of hierarchical regression models including a quadratic term of Interoceptive Accuracy reflecting this threshold effect. In disagreement with hypothesis (Ib), none of the regression models showed effects that suggest that there is a “threshold” in Interoceptive Accuracy in the current data that may moderate the association between the three facets Accuracy, Sensibility and Awareness. This threshold effect of Interoceptive Accuracy was previously reported by Garfinkel et al. (2015). Discrepancies between both studies could be explained by differences in methodology (Garfinkel: median split analysis; this study: regression model) or composition of the study sample, as in contrast to the Garfinkel study, in the current study there were several participants with an almost perfect Accuracy. Previous findings on higher Interoceptive Accuracy in panic disorder (Ehlers and Breuer, 1992, 1996; Ehlers et al., 1995) were later attributed to a subgroup of individuals with particularly high Interoceptive Accuracy (Willem Van der Does et al., 2000). It remains for future research to clarify if this subgroup of individuals with high Interoceptive Accuracy show a different pattern of association between interoceptive facets (which may account for vulnerability for or resilience against mental disorders with physical symptoms) than individuals with Interoceptive Accuracy within the medium range.

Furthermore, in all hierarchical regression models with heartbeat discrimination as index of Interoceptive Accuracy, only Accuracy emerged as significant predictor of Sensibility and Awareness. This finding, however, still supports the thesis that Accuracy could be interpreted as most basic construct among the three interoceptive facets, although a threshold effect was not observed in the current study. The assumption (Ic) that Interoceptive Accuracy is the basis of all interoceptive processing and that a minimal extent of Accuracy is a necessary requirement for “higher order” interoceptive processes like Interoceptive Sensibility and Awareness, was further tested using another linear regression analysis. This analysis was similar to that reported by Garfinkel et al. (2015) although we did not collapse data across the two interoceptive tasks (heartbeat tracking and heartbeat discrimination) because most participants did not complete both tasks, but reported separate analyses instead. In contrast to the results presented by Garfinkel et al. (2015), in the present study neither Interoceptive Awareness nor Interoceptive Sensibility significantly predicted Interoceptive Accuracy in the heartbeat tracking task, but both significantly and independently predicted Interoceptive Accuracy in the heartbeat discrimination task. When the regression model was reversed and Interoceptive Awareness was entered as criterion variable a similar picture emerged. Again, neither Interoceptive Sensibility nor Accuracy were associated with Interoceptive Awareness in the heartbeat tracking task but both significantly and independently predicted Interoceptive Accuracy in the heartbeat discrimination task. 5.2. Study aim II: Investigation whether an index of cardiovascular activation should be added as fourth level of the model In hypothesis (II) we proposed that heart rate as index of cardiovascular activation should be differentially related to the three facets of interoception. In line with this assumption, an additional set of regression analyses based on the heartbeat tracking task suggested that heart rate was independently and significantly predicted by Interoceptive Sensibility and Awareness but not Accuracy. When the regression model was reversed and Interoceptive Awareness was entered as dependent variable, mean heart rate was the only significant predictor. This pattern of results, however, was limited to the heartbeat tracking task. Based on the heartbeat discrimination task, no significant associations between heart rate and the levels of interoception could be found. These results suggest that mean heart rate has to be taken into account as important variable. Mean heart rate reflects an objective physiological state associated with cardiovascular activation that is required to consciously process and perceive afferent signals from the cardiovascular systems in all interoceptive tasks conducted in the present study. The higher participants' mean heart rate, the higher their Interoceptive Awareness and the lower their Interoceptive Sensibility. One possible explanation for this finding could be the inverse relationship of heart rate and stroke volume (Eichler and Katkin, 1994; Ring et al., 1994), the latter of which represents a major determinant of Interoceptive Accuracy (Schandry et al., 1993). Individuals with higher heart rate and lower stroke volume may perceive less afferent neural signals per single heartbeat, which results in a decrease of confidence ratings. In fact, these lower confidence ratings better predict the actual accuracy in the tests than those observed in individuals with lower heart rate. Although further research is needed to shed light on the detailed implications, it appears possible to speculate about the integration of an additional, fourth level to Garfinkel and Critchley's model (2013), i.e. a level of objective physiological states, such as the activity of the cardiovascular system, as previously suggested by Vaitl (1996). Since ‘objective’ physiological states may also be the result of central control mechanisms (e.g., sympathetic tone), it may be further speculated that interoception is not a unidirectional process from the body to the brain, but is embedded into a bidirectional regulatory circuit between the brain and the body (Schulz and Vögele, 2015). Future studies should integrate subjective and behavioral indicators of interoception with

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

psychophysiological indicators of visceral-afferent signal transmission, which do not require the conscious perception of visceral sensations, such as heartbeat-evoked potentials (Schandry and Montoya, 1996) or visceral modulation of startle (Schulz et al., 2009). The question remains as to why the relationship of ‘objective’ physiological state with interoceptive facets can only be observed for the heartbeat tracking task. One possible explanation could be that indicators derived from heartbeat tracking (i.e. Accuracy, Sensibility, Awareness) are more closely associated with actual visceral processes (despite that they may be biased by beliefs about one's heart rate), whereas indicators derived from heartbeat discrimination tasks also reflect processes of multi-sensory integration to a larger degree (Schulz et al., 2013; Schulz and Vögele, 2015). In partial support of hypothesis (III), the two measures of Interoceptive Sensibility, one based on the heartbeat tracking task and one based on the heartbeat discrimination task, correlated significantly with each other (r = 0.348), but markedly lower than previously reported (r = 0.711) (Garfinkel et al., 2015). This moderate correlation may be partially determined by a ‘global’ confidence related to interoception, independent of the assessment method, but it may also be due to task-specific aspects of confidence, such as related to the multisensory integration that is implied by the discrimination task. Nevertheless, there was no significant correlation between the two measures of Interoceptive Accuracy. Previous studies yielded mixed findings, with some showing positive correlations between heartbeat tracking and discrimination tasks (Hart et al., 2013; Knoll and Hodapp, 1992; Schaefer et al., 2012), whereas others did not find significant correlations (Michal et al., 2014; Phillips et al., 1999; Schulz et al., 2013). One may conclude that both tasks only share a low amount of variance, which requires a larger study sample to show significant positive correlations. There was no evidence for a large impact of the calculation method on Interoceptive Accuracy. The only exception was that in the sample of “good” performers in the heartbeat tracking task, the correlation between Interoceptive Accuracy and Interoceptive Sensibility was no longer significant when accuracy was calculated based on Schandry's formula (Schandry, 1981) instead of the formula proposed by Hart et al. (2013). Thus, it can be concluded that the results reported by Garfinkel et al. (2015), Meessen et al. (2016) and our own data are comparable although different formulae were applied.

5.3. Strengths and limitations According to Cohen (1988), the present study is sufficiently powered to detect medium effects. Thus, if any of the hypothesized but unsupported relationships exist but are relatively small in magnitude (smaller than 0.27–0.30), the present study was not adequately powered to detect them. Consequently, future studies should try to examine the present hypotheses in even larger samples. One limitation may be the fact that the present data were generated from sub-studies, which were not identical, but still comparable. In particular, the heartbeat perception paradigms were identical in terms of wording, instruction, items, and R-wave detection method and thus results did not differ between subsamples in the key variables “Sensibility”, “Accuracy” or “Awareness”. We here used heart rate as indicator of cardiovascular activation and assumed that increased peripheral activation is strongly translated into altered central nervous system (CNS) representation of these afferent signals, as previously observed in healthy individuals and patients with depersonalization disorder (Shao et al., 2011; Schulz et al., 2015). It needs to be acknowledged that some differential factors, such as intact visceralafferent signal transmission (Pauli et al., 1991; Leopold and Schandry, 2001; Schulz et al., 2009), may mediate the relationship between cardiovascular activation and CNS representation of visceral-afferent signals.

79

6. Conclusions The results of the present study give partial support for the assumption that interoception is a multi-facetted construct. In line with results presented by Garfinkel et al. (2015), the present results show that Interoceptive Accuracy, Sensibility and Awareness are mostly uncorrelated based on heartbeat tracking, but show moderate correlations based on heartbeat discrimination. Nevertheless, one central assumption of the model proposed by Garfinkel et al. (2015), i.e. Interoceptive Accuracy as the basic level of interoception, could only partially be confirmed. Instead, parts of our results suggest that a fourth level of processing of internal bodily signals should be considered to be added to the model: a level of objective physiological states, such as the activity of the cardiovascular system. Future research is needed to verify this proposition. The integration of subjective and behavioral indicators of interoception with psychophysiological indicators of visceral-afferent signal transmission, which do not require the conscious perception of visceral sensations, is recommended. Conflict of interest statement The authors declare that they have no conflicts of interest. Acknowledgments Experiment 3 was funded by the German Research Foundation (Deutsche Forschungsgemeinschaft: DFG), grant GRK 1389/1 to the International Research Training Group (IRTG). The funding body had no further role in study design, in the collection, analysis and interpretation of data, in the writing of the report, and in the decision to submit the paper for publication. References Bechara, A., Naquvi, N., 2004. Listening to your heart: interoceptive awareness as a gateway to feeling. Nat. Neurosci. 7, 102–103. Brener, J., Jones, J.M., 1974. Interoceptive discrimination in intact humans: detection of cardiac activity. Physiol. Behav. 13, 763–767. Brener, J., Kluvitse, C., 1988. Heartbeat detection: judgments of the simultaneity of external stimuli and heartbeats. Psychophysiology 25, 554–561. Ceunen, E., Van Diest, I., Vlaeyen, J.W.S., 2013. Accuracy and awareness of perception: related, yet distinct (commentary on Herbert et al., 2012). Biol. Psychol. 92, 426–427. Cohen, J., 1988. Statistical Power for the Behavioural Science. Erlbaum, Hillsdale, NJ. Craig, A.D., 2003. Interoception: the sense of the physiological condition of the body. Curr. Opin. Neurobiol. 13, 500–505. Dunn, B.D., Dalgleish, T., Ogilvie, A.D., Lawrence, A.D., 2007. Heartbeat perception in depression. Behav. Res. Ther. 45, 1921–1930. Dunn, B.D., Stefanovitch, I., Evans, D., Oliver, C., Hawkins, A., Dalgleish, T., 2010. Can you feel the beat? Interoceptive awareness is an interactive function of anxiety- and depression-specific symptom dimensions. Behav. Res. Ther. 48, 1133–1138. Ehlers, A., Breuer, P., 1992. Increased cardiac awareness in panic disorder. J. Abnorm. Psychol. 101, 371–382. Ehlers, A., Breuer, P., 1996. How good are patients with panic disorder at perceiving their heartbeats? Biol. Psychol. 42, 165–182. Ehlers, A., Breuer, P., Dohn, D., Fiegenbaum, W., 1995. Heartbeat perception and panic disorder: possible explanations for discrepant findings. Behav. Res. Ther. 33, 69–76. Eichler, S., Katkin, E.S., 1994. The relationship between cardiovascular reactivity and heartbeat detection. Psychophysiology 31, 229–234. Fairclough, S.H., Goodwin, L., 2007. The effect of psychological stress and relaxation on interoceptive accuracy: implications for symptom perception. J. Psychosom. Res. 62, 289–295. Fustos, J., Gramann, K., Herbert, B.M., Pollatos, O., 2013. On the embodiment of emotion regulation: interoceptive awareness facilitates reappraisal. Soc. Cogn. Affect. Neurosci. 8, 911–917. Garfinkel, S.N., Critchley, H.D., 2013. Interoception, emotion and brain: new insights link internal physiology to social behaviour. Commentary on: “anterior cortex mediates bodily sensibility and social anxiety” by Terasawa et al. (2012). Soc. Cogn. Affect. Neurosci. 8, 231–234. Garfinkel, S.N., Seth, A.K., Barrett, A.B., Suzuki, K., Critchley, H.D., 2015. Knowing your own heart: distinguishing interoceptive accuracy from interoceptive awareness. Biol. Psychol. 104, 65–74. Hart, N., McGowan, J., Minati, L., Critchley, H.D., 2013. Emotional regulation and bodily sensation: interoceptive awareness is intact in borderline personality disorder. J. Personal. Disord. 27, 506–518.

80

T. Forkmann et al. / International Journal of Psychophysiology 109 (2016) 71–80

Herbert, B.M., Muth, E.R., Pollatos, O., Herbert, C., 2012. Interoception across modalities: on the relationship between cardiac awareness and the sensitivity for gastric functions. PLoS One 7, e36646. Katkin, E.S., Blascovich, J., Goldband, S., 1981. Empirical assessment of visceral self-perception: individual and sex differences in the acquisition of heartbeat discrimination. J. Pers. Soc. Psychol. 40, 1095–1101. Knapp-Kline, K., Kline, J.P., 2005. Heart rate, heart rate variability, and heartbeat detection with the method of constant stimuli: slow and steady wins the race. Biol. Psychol. 69, 387–396. Knoll, J.F., Hodapp, V., 1992. A comparison between two methods for assessing heartbeat perception. Psychophysiology 29, 218–222. Leopold, C., Schandry, R., 2001. The heartbeat-evoked brain potential in patients suffering from diabetic neuropathy and in healthy control persons. Clin. Neurophysiol. 112, 674–682. Meessen, J., Mainz, V., Gauggel, S., Volz-Sidiropoulou, E., Sütterlin, S., Forkmann, T., 2016. The relationship between interoception and metacognition: a pilot study. J. Psychophysiol. 30, 76–86. Mehling, W.E., Price, C., Daubenmier, J.J., Acree, M., Bartmess, E., Stewart, A., 2012. The multidimensional assessment of interoceptive awareness (MAIA). PLoS One 7, e48230. Michal, M., Reuchlein, B., Adler, J., Reiner, I., Beutel, M.E., Vögele, C., Schachinger, H., Schulz, A., 2014. Striking discrepancy of anomalous body experiences with normal interoceptive accuracy in depersonalization-derealization disorder. PLoS One 9, e89823. O'Brien, H.O., Reid, G.J., Jones, K.R., 1998. Differences in heartbeat awareness among males with higher and lower levels of systolic blood pressure. Int. J. Psychophysiol. 29, 53–63. Pauli, P., Hartl, L., Marquardt, C., Stalmann, H., Strian, F., 1991. Heartbeat and arrhythmia perception in diabetic autonomic neuropathy. Psychol. Med. 21, 413–421. Phillips, G.C., Jones, G.E., Rieger, E.J., Snell, J.B., 1999. Effects of the presentation of false heart-rate feedback on the performance of two common heartbeat-detection tasks. Psychophysiology 36, 504–510. Pollatos, O., Schandry, R., 2008. Emotional processing and emotional memory are modulated by interoceptive awareness. Cognit. Emot. 22, 272–287. Pollatos, O., Herbert, B.M., Kaufmann, C., Auer, D.P., Schandry, R., 2007. Interoceptive awareness, anxiety and cardiovascular reactivity to isometric exercise. Int. J. Psychophysiol. 65, 167–173. Pollatos, O., Kurz, A.L., Albrecht, J., Schreder, T., Kleemann, A.M., Schopf, V., Kopietz, R., Wiesmann, M., Schandry, R., 2008. Reduced perception of bodily signals in anorexia nervosa. Eat. Behav. 9, 381–388. Porges, S., 1993. Body Perception Questionnaire. Laboritory of behavioral assessment: University of Maryland, Maryland. Ring, C., Brener, J., 1996. Influence of beliefs about heart rate and actual heart rate on heartbeat counting. Psychophysiology 33, 541–546. Ring, C., Liu, X., Brener, J., 1994. Cardiac stimulus intensity and heartbeat detection: effects of tilt-induced changes in stroke volume. Psychophysiology 31, 553–564. Ring, C., Brener, J., Knapp, K., Mailloux, J., 2015. Effects of heartbeat feedback on beliefs about heart rate and heartbeat counting: a cautionary tale about interoceptive awareness. Biol. Psychol. 104, 193–198.

Schaefer, M., Egloff, B., Witthoft, M., 2012. Is interoceptive awareness really altered in somatoform disorders? Testing competing theories with two paradigms of heartbeat perception. J. Abnorm. Psychol. 121, 719–724. Schandry, R., 1981. Heart beat perception and emotional experience. Psychophysiology 18, 483–488. Schandry, R., Montoya, P., 1996. Event-related brain potentials and the processing of cardiac activity. Biol. Psychol. 42, 75–85. Schandry, R., Bestler, M., Montoya, P., 1993. On the relation between cardiodynamics and heartbeat perception. Psychophysiology 30, 467–474. Schulz, A., Vögele, C., 2015. Interoception and stress. Front. Psychol. 6, 993. Schulz, A., Lass-Hennemann, J., Nees, F., Blumenthal, T.D., Berger, W., Schachinger, H., 2009. Cardiac modulation of startle eye blink. Psychophysiology 46, 234–240. Schulz, A., Lass-Hennemann, J., Sütterlin, S., Schächinger, H., Vögele, C., 2013. Cold pressor stress induces opposite effects on cardioceptive accuracy dependent on assessment paradigm. Biol. Psychol. 93, 167–174. Schulz, A., Koster, S., Beutel, M.E., Schachinger, H., Vogele, C., Rost, S., Rauh, M., Michal, M., 2015. Altered patterns of heartbeat-evoked potentials in depersonalization/derealization disorder: neurophysiological evidence for impaired cortical representation of bodily signals. Psychosom. Med. 77, 506–516. Shao, S., Shen, K., Wilder-Smith, E.P., Li, X., 2011. Effect of pain perception on the heartbeat evoked potential. Clin. Neurophysiol. 122, 1838–1845. Sütterlin, S., Schulz, S.M., Stumpf, T., Pauli, P., Vögele, C., 2013. Enhanced cardiac perception is associated with increased susceptibility to framing effects. Cogn. Sci. 37, 922–935. Suzuki, K., Garfinkel, S.N., Critchley, H.D., Seth, A.K., 2013. Multisensory integration across exteroceptive and interoceptive domains modulates self-experience in the rubberhand illusion. Neuropsychologia 51, 2909–2917. Terasawa, Y., Fukushima, H., Umeda, S., 2013. How does interoceptive awareness interact with the subjective experience of emotion? An fMRI study. Hum. Brain Mapp. 34, 598–612. Terhaar, J., Viola, F.C., Bar, K.J., Debener, S., 2012. Heartbeat evoked potentials mirror altered body perception in depressed patients. Clin. Neurophysiol. 123, 1950–1957. Vaitl, D., 1996. Interoception. Biol. Psychol. 42, 1–27. Whitehead, W.E., Drescher, V.M., 1980. Perception of gastric contractions and self-control of gastric motility. Psychophysiology 17, 552–558. Whitehead, W.E., Drescher, V.M., Heiman, P., Blackwell, B., 1977. Relation of heart rate control to heartbeat perception. Biofeedback Self-Regul. 2, 371–392. Wickens, T.D., 2001. Elementary Signal Detection Theory. Oxford University Press, Oxford. Windmann, S., Schonecke, O.W., Frohlig, G., Maldener, G., 1999. Dissociating beliefs about heart rates and actual heart rates in patients with cardiac pacemakers. Psychophysiology 36, 339–342. Willem Van der Does, A.J., Antony, M.M., Ehlers, A., Barsky, A.J., 2000. Heartbeat perception in panic disorder: a reanalysis. Behav. Res. Ther. 38, 47–62. Yates, A.J., Jones, K.E., Marie, G.V., Hogben, J.H., 1985. Detection of the heartbeat and events in the cardiac cycle. Psychophysiology 22, 561–567. http://dx.doi.org/10. 1111/j.1469-8986.1985.tb01651.x. Yoris, A., Esteves, S., Couto, B., Melloni, M., Kichic, R., Cetkovich, M., Favaloro, R., Moser, J., Manes, F., Ibanez, A., Sedeno, L., 2015. The roles of interoceptive sensitivity and metacognitive interoception in panic. Behav. Brain Funct. 11, 14.