Loneliness and Interoceptive Accuracy in the Elderly Population

Table of Contents

Introduction and Theoretical Background
Loneliness in the Elderly Population
Loneliness and Alexithymia
Interoceptive Accuracy
Interoceptive Accuracy and Alexithymia
Limitations of Prior Research
Aim of the study and hypothesis

Methods
Participants
Measures
Sociodemographic Information
Loneliness
Anxiety
Depression
Alexithymia
Interoceptive Accuracy
fMRI Data Acquisition
Study procedure
Data Analysis
Behavioral and Statistical analysis
fMRI analysis

Results
Behavioral results
Descriptive Statistics
Correlational Analysis
Analysis of Variance
Analysis of Frequency
fMRI results
Main effect of condition
Interaction loneliness x condition

Discussion

References

Appendix
List of Figures
List of Tables
Abstract
German Abstract

Introduction and Theoretical Background

Loneliness in the Elderly Population

Loneliness is described as a distressing emotion accompanied by the subjective perception that one's needs of social companionship are not being met in light of the quality or quantity of one's actual relationships (Hawkley & Cacioppo, 2010; Peplau & Perlman, 1982). Cacioppo (e.g., 2014) denominates loneliness as perceived social isolation (PSI) and delimits the construct from objective social isolation (OSI). Accordingly, there can be people who have many social contacts (are objectively not lonely) but feel subjectively lonely, whereas there can also be those who lack social companionship and, conversely, not feel lonely.

However, the prevalence of loneliness seems to differ across nations and age (Yang & Victor, 2011). In the German population, loneliness is proposed to form a non-linear trajectory across the lifespan with increased levels of loneliness among the young adults and the oldest old (Luhmann & Hawkley, 2016).

Primary reasons for the occurrence of loneliness among older adults are considered deteriorating health as well as the loss of one's partner (Perlman, 1988).

Baumeister and Leary (1995) argued that the so-called need to belong is a fundamental motivation to form and maintain long-lasting interpersonal relationships. Furthermore, the social control theory (e.g., Umberson, 1987) posits that the impact of the social environment encourages health-promoting behavior and discourages health-impairing behavior. For instance, being married was suggested to promote good health behavior like exercise (Umberson, 1987, 1992). Thus, loneliness is associated with a series of poor health behaviors and might lead to negative consequences such as obesity (Lauder, Mummery, Jones, & Caperchione, 2006), poorer physical health (Cornwell & Waite, 2009), Alzheimer's disease (Wilson et al., 2007) and alcoholism (Akerlind & Hornquist, 1992; Bell, 1956). Therefore, it is often viewed as a risk factor for mortality and morbidity (e.g., Cacioppo et al., 2014; Holt- Lunstad, Smith & Layton, 2010).

Based on a social neuroscience model, social ties helped individuals to reproduce and to survive (Cacioppo et al., 2014). Isolation from others might be a dangerous condition, since no mutual protection is available in case of an assault. As a result, the brain activates a so called “short-term self-preservation mode” (Cacioppo et al., 2014, p. 1467) when being without the assistance of others. The mode evokes multiple effects: (i) implicit hypervigilance for social threats accompanied by elevated hostility, anxiety and social withdrawal; (ii) elevated fragmentation of sleep due to potential predation during sleep; (iii) increased hypothalamic- pituitary-adrenal activity, heightened vascular activity, as well as modified immunity and gene expression to cope with possible assaults; (iv) reduced impulse control for the benefit of responses most elevated in the hierarchy of responses (i.e., prepotent responding) resulting in the display of behavior that led to success in past events; (v) increased depressive symptoms signalizing the need for social reconnection and support. Although these effects might serve to ensure short-term survival, they may also carry costs when being exposed to them chronically, resulting in alterations in brain function and structure (Cacioppo et al., 2014).

According to researches like Hawkley and Cacioppo (2010), the aforementioned hypervigilance for social threat generates cognitive biases: Lonely persons perceive social interactions more negative. They also seem to perceive their social environment to be more threatening when compared to non-lonely individuals. Those cognitive biases might lead to negative social expectations, which in turn result in a self-fulfilling prophecy. Since others might display more frequently negative behavior towards lonely people and thereby confirm the expectations imposed to them beforehand, lonely individuals most likely react with social withdrawal, distancing themselves from others. Consequently, a “self-reinforcing loneliness loop” (Hawkley & Cacioppo, 2010, p. 220) is set in motion, creating the belief in lonely people that the reason for their loneliness is due to others and uncontrollable for them (Newall, Chipperfield, Clifton, Perry & Swift, 2009). Therefore, they actively distance themselves from potential social partners (Hawkley & Cacioppo, 2010). Thus, these mechanisms of loneliness seem to be associated with low self-esteem, as well as with feelings of anxiety, stress, pessimism, hostility (Cacioppo et al., 2006), social withdrawal (Cacioppo et al., 2014) and interpersonal distrust (Qualter, Wagner, Quinton & Brown, 2009).

Loneliness and Alexithymia

Alexithymia is considered a three-dimensional personality trait (Preece, Becerra, Allan, Robinson & Dandy, 2017) that is characterized by difficulties identifying one's own feelings, difficulties describing feelings and externally oriented thinking (Bagby, Parker & Taylor, 1994). The dimension externally oriented thinking is defined as “an absence of the capacity to produce fantasies with the result that thought content [is] restricted to a preoccupation with external objects, people, and environmental events” (Nemiah, 1984, p. 127). Nemiah and Sifneos (1970) were the first who implemented the term alexithymia which originates from the Greek and means “lack of words for emotions” (a = lack, lexis = word, thymos = feeling).Thus, individuals with high levels of alexithymia are able to experience physiological arousal associated with emotional states. Nonetheless, they are unable to distinguish between emotions and their physiological concomitants as well as between affective states as such (e.g. between anger and sadness) (Taylor & Bagby, 2004).

Two studies have examined the direct relationship of alexithymia and loneliness (Frye- Cox & Hesse, 2013; Qualter et al., 2009). Both of them found the two constructs to be significantly associated with each other. Qualter and colleagues (2009) utilized the Social and Emotional Scale for Adults (SELSA; DiTommaso & Spinner, 1993) to assess the levels of romantic loneliness (the perceived lack of attachment with a romantic partner), family loneliness (the perceived lack of attachment with family members) and social loneliness (the perceived deficiency of a peer social network) (Weiß, 1973). Results indicated moderate positive correlations between romantic, family and social loneliness and alexithymia (measured with the TAS-20; Bagby, Parker & Taylor, 1994). Furthermore, interpersonal distrust was considered to partly mediate this association. Frye-Cox and Hesse (2013) conducted a study collecting data of married couples. Aim of the study was to investigate the mediating roles of intimate communication and loneliness in the relationship between alexithymia and marital quality. Results showed that increased levels of alexithymia (assessed with the TAS-20; Bagby, Parker & Taylor, 1994) were related to higher levels of loneliness (assessed with the UCLA Loneliness Scale; Russell, 1996), predicting decreased intimate communication, which was associated with lower marital quality. Taken together, both studies suggest a moderate to high correlation between loneliness and alexithymia.

Possible mechanisms explaining the link between loneliness and alexithymia are considered to be deficiencies in the ability to recognize and understand one's own emotions and also those of others' which might result in problems to form intimate and effective interpersonal relationships (Qualter et al., 2009). Thus, the perceived quality of romantic relationships seems to be decreased in individuals with higher levels of alexithymia (Frye-Cox & Hesse, 2013; Holder, Love & Timoney, 2014). According to the neo-Darwinian affection exchange theory (AET; Floyd, 2006) affectionate communication is an adaptive behavior in the sense that it helps us through the formation and maintenance of relationships to reproduce and to survive. Individuals with higher levels of alexithymia displayed decreased abilities to give and experience affection (Hesse & Floyd, 2008) leading to less satisfactory and fewer close relationships (Frye-Cox & Hesse, 2013; Hesse & Floyd, 2011). In addition, prior research also pointed out that highly alexithymic persons felt disconnected from others and reported having poor networks of social support (Hesse & Floyd, 2011; Humphreys et al., 2009). Another theory explaining the link between loneliness and alexithymia is the personality-behavioral approach (e.g. Maragoni & Ickes, 1989), which posits that personality predispositions such as alexithymia might cause higher levels of loneliness due to the incapability of engaging in certain behaviors, for instance intimate communication, which are necessary to maintain interpersonal relationships (Frye-Cox & Hesse, 2013).

Interoceptive Accuracy

Previous studies defined interoception as the perception of the physiological state of the body (Craig, 2002, 2009). Hence, interoception is said to function as a body-to-brain axis (Cameron, 2002; Sherrington, 1948), “connecting the mind and body pathways” (Forrest, Smith, White & Joiner, 2015, p. 754). Whereas former definitions of the term interoception claim that only visceral information (e.g. hunger and heart rate) are supposed to be considered as interoceptive (e.g. Fowler, 2003), definitions that are more recent have been extended by the inclusion of other physiological signals (e.g. tickle and affective touch), which are processed utilizing the same neurological pathways as visceral stimuli (e.g., Murphy, Brewer, Catmur & Bird, 2017a). Hence, according to some recent definitions all bodily signals that are transmitted (i) via small unmyelinated group C nerve fibres or large myelinated group A nerve fibres, the marginal nucleus of the spinal cord (lamina I) and the spinothalamic tract onto the anterior cingulate cortex (ACC) and the insula (Craig, 2002), or (ii) via cranial nerves (glossopharyngeal and vagus) to the nucleus of the solitary tract (NTS) (Critchley & Harrison, 2013; Saper, 2002) are considered to be interoceptive. While some researchers maintain the original and more restrictive meaning to define the term interoception (e.g., Dworkin, 2007), we use the term more broadly to refer to the perception of any physiological condition that is transmitted by the aforementioned neural pathways (e.g., Craig, 2002; Critchley & Harrison, 2013; Murphy et al. 2017a). According to Sherrington (1984), interoception is distinguishable from proprioception, the body position in space, and exteroception, defined as the perception of external stimuli.

Garfinkel and colleagues (2015) suggested to distinguish between three different dimensions of interoceptive ability. Correspondingly, they defined objective interoceptive accuracy as the extent to which an individual is capable of accurately perceiving physiological signals, subjective interoceptive sensibility as the belief about the own interoceptive accuracy and metacognitive interoceptive awareness as the degree of the equivalence between objective and subjective interoceptive abilities. However, due to an observed moderation of interoceptive accuracy (assessed using heartbeat detection tasks) on the relationship between interoceptive accuracy, interoceptive sensibility (assessed using confidence judgements) and interoceptive awareness (assessed using confidence-accuracy correlations), interoceptive accuracy was proposed to be the construct underlying the other interoceptive measures. To assess interoceptive accuracy two different kind of measurements, which both rely on heartbeat perception, are applied most frequently, namely heartbeat tracking tasks, originally developed by Schandry (1981), and heartbeat discrimination tasks (Whitehead & Drescher, 1980; Whitehead, Drescher, Heiman & Blackwell, 1977). The first ones require the participants to count their own heartbeats during a specified time interval. Later on, the counted number of heartbeats is compared with the actual number of heartbeats. The latter ones ask the participants to detect whether external stimuli (a series of lights, tones or tactile stimuli) are simultaneous or delayed in comparison to their own heartbeats. The term “heartbeat detection task” serves as an umbrella term to refer to both kind of assessments (e.g. Garfinkel et al., 2015). For the present study, we applied a heartbeat tracking task in order to measure interoceptive accuracy (for more details see section measures).

Accurate interoceptive abilities may be essential for homeostasis and thus for the regulation of the bodily condition and the maintenance of the physiological health (Murphy et al., 2017a; Murphy, Geary, Millgate, Catmur & Bird, 2017b). In addition, poor interoceptive accuracy has been found to contribute to a number of mental disorders, such as schizophrenia (Ardizzi et al., 2016), depression (e.g. Harshaw, 2015), somatic symptom disorders, obsessive compulsive disorders, eating disorders (Brewer et al., 2015; Brewer, Cook & Bird, 2016a,b; Khalsa & Lapidus, 2016; Stern, 2014) and addiction (Naqvi & Bechara, 2010; Verdejo-Garcia et al., 2012). Atypically heightened interoceptive accuracy, on the other hand, has been observed in individuals with anxiety syndromes (Paulus & Stein, 2006) and panic disorders (Ehlers & Breuer, 1992).

Previous neuroimaging studies have associated elevated interoceptive accuracy (as assessed via heartbeat detection tasks) with enhanced activation in the somatosensory cortex, the inferior frontal gyrus (IFG), the precentral gyrus and the anterior insula when paying attention to bodily sensations (Caseras et al., 2011; Critchley et al., 2004; Kuehn et al., 2016; Pollatos, Traut-Mattausch, Schroeder & Schandry, 2007b; Stern et al., 2017). Enhanced processing in these brain regions has shown to entail a more accurate perception of internal signals.

Interoceptive Accuracy and Alexithymia

In several emotion theories, the perception of physiological signals plays a crucial role for the emotional experience (Damasio, 1994; James, 1884; Schachter & Singer, 1962). As James (1884) hypothesized, the interpretation of our bodily changes subsequently following the perception of an exciting incident is the emotion that we are feeling. In addition, Damasio's somatic marker hypothesis (Damasio, 1994, 1999) states that bodily signals are integrated in certain brain regions, generating associations that elicit emotional states, which in turn influence cognition. Furthermore, Damasio (1999) argues for the relevance of considering the relatedness of bodily and emotional experiencing when referring to the body as the principal stage of emotions in either a direct way or an indirect way (through bodily representations in somatosensory brain regions). Thus far, a number of studies have shown that bodily sensitivity (interoceptive accuracy) is positively associated with more intense processing of emotion­eliciting stimuli (Herbert, Pollatos & Schandry, 2007; Pollatos, Kirsch & Schandry, 2005). Regions most commonly involved in the processing of non-affective interoceptive states have shown to be the AI and the ACC (Craig, 2002, 2003, 2009; Garfinkel & Critchley, 2013). Interestingly, those brain areas are also proposed to be implicated in the processing of one's own affective states (Bush, Luu & Posner, 2000; Etkin, Egner & Kalisch, 2011; Kober et al., 2008; Lindquist, Wager, Kober, Bliss-Moreau & Barrett, 2012; Phan, Wager, Taylor & Liberzon, 2002).

There is a substantial amount of research demonstrating that alexithymia (emotional blindness) may be associated with impaired interoceptive accuracy. Therefore, many researchers proposed to describe alexithymia by general deficits in interoceptive accuracy rather than by specific deficits in the emotional domain (Brewer et al., 2016a; Gaigg, Cornell & Bird, 2016; Herbert, Herbert & Pollatos, 2011; Longarzo et al., 2015; Näring & Van der Staak, 1995; Shah, Hall, Catmur & Bird, 2016). The underlying idea is that people who have problems identifying their physical state also fail to recognize which emotional state they are in. Brewer and colleagues (2016a), for instance, observed that alexithymia is related to poor affective and non-affective interoception. Two other studies (Herbert et al., 2011; Shah et al., 2016) discovered that highly alexithymic individuals were less accurate in detecting their own heartbeat in comparison to a control group of typical participants. On a neural level, alexithymia has been shown to involve atypical function and structure in the ACC and the anterior insular cortex (Goerlich-Dobre, Bruce, Martens, Aleman & Hooker, 2014; Ihme et al., 2013; Moriguchi et al., 2007). These brain areas also appear to be involved in the processing of interoceptive signals (Craig, 2002, 2003, 2009; Garfinkel & Critchley, 2013).

Limitations of Prior Research

In previous research, aging has not only been associated with increased prevalence of loneliness (e.g., Luhmann & Hawkley, 2016), but also with reduced interoceptive accuracy (Khalsa, Rudrauf & Tranel, 2009; Murphy et al., 2017a,b). Nonetheless, only a limited number of studies (Khalsa et al., 2009) has examined cardiac interoceptive accuracy in late adulthood. None of them investigated the cerebral activity during heartbeat detection. Overall, there is a lack of objective tests and longitudinal data, leaving several questions unanswered, including those on whether the observed decline in interoceptive accuracy is just a cohort effect and cannot be traced back to the process of aging or whether the decline is caused by psychological and physiological factors other than aging (e.g. the BMI or the heartrate frequency) (Murphy et al., 2017b). Furthermore, the relationship between loneliness and interoceptive accuracy has not been examined so far.

Aim of the study and hypothesis

The present study aims to investigate the relationship between interoceptive accuracy and loneliness in the elderly population. Because of the negative consequences of atypical (too low or too high) interoceptive accuracy and the negative consequences of loneliness, an investigation of how these two constructs are linked to and might interact with each other is of major relevance. By combining two large areas, the one of loneliness and the one of interoception, which are both of great psychological interest, we intent to close a gap within the current research literature. Results might give us new insights into the perception (interoeptive and exteroceptive) of lonely persons.

Previous studies observed a positive correlation between alexithymia and loneliness (Frye-Cox & Hesse, 2013; Qualter et al., 2009). Due to the association of alexithymia with deficits in interoceptive accuracy (e.g., Brewer et al., 2016a), we hypothesize that the level of interoceptive accuracy is decreased in lonely individuals. On a neural level, increased interoceptive accuracy was observed to involve enhanced activation in the somatosensory cortex, the inferior frontal gyrus (IFG), the precentral gyrus and the anterior insula (Caseras et al., 2011; Critchley et al., 2004; Kuehn et al., 2016; Pollatos et al., 2007b; Stern et al., 2017). Since we expect lonely participants to show reduced interoceptive accuracy, we hypothesize that they will exhibit decreased neural activity within the aforementioned brain areas (the anterior insula, the somatosensory cortex, the IFG and the precentral gyrus) during the heartbeat perception task.

Methods

Participants

Twenty-eight (17 men, 11 women) participants older than 64 years took part in the study. The mean age of men was 72.2 ± 5.6 years and the mean age of women was 70.3 ± 3.8 years. In total, the age ranged from 64 to 84 years. Participants were recruited via advertisements in Facebook groups, clubs for retirees, fairs targeting the elderly population as well as other public places such as medical practices, pharmacies or supermarkets. Potential subjects who took psychopharmalogical medication, had less than nine years of education, had a current diagnosis of psychiatric disorder, studied psychology, took part in a similar study, or did not use a computer at least once a month were excluded. Only interested persons without any health issues concerning their heart or their brain (e.g. heart attack or stroke) were included. Furthermore, possible participants had to fill in a questionnaire, which assures that they are safe for the MRI scanning (MRI Safety Questionnaire). Other exclusion criteria included colour­blindness and a score below 26 in the Minimental-Mental State Examination (MMSE) (Folstein, Folstein, & McHugh, 1975; Kessler, Markowitsch, & Denzler, 2000). The MMSE is a screening measure to detect initial stages of cognitive deficits. Moreover, all participants were ought to be German-speaking, right-handed and heterosexual. Due to technical errors during the scanning session, three participants had to be excluded for the behavioural part of the study.

Measures

Sociodemographic Information

For each participant sociodemographic information on gender, nationality, year of birth, education, handedness, mental illnesses, cardiac problems, last employment, last income, computer usage and sexual orientation were collected. Moreover, possible participants were required to indicate whether they have studied or are currently studying psychology and if they have already participated in a similar research project. The given information were utilized for further analysis or to exclude persons who were not eligible for the study.

Mini-Mental State Examination

The German version of the Mini-Mental State Examination (Kessler et al., 2000) was used as a tool to screen for cognitive impairment among the elderly population. Higher scores correspond to higher cognitive functioning skills (Folstein et al., 1975). The maximum score is 30. Content of the items regard the local and temporal orientation of a person. Participants were required to answer 28 questions concerning five dimensions of cognitive functioning (Orientation, registration, attention and calculation, recall and language). Example items are, “What is the month?” [Welcher Monat ist gerade?] and “In which town are we?“ [In welcher Stadt befinden wir uns?]. The test-retest reliability of MMSE was satisfactory (rtt = .89) (Kessler et al., 2000). The cut-off threshold for the MMSE was set at 26, whereas persons who scored below 26 were excluded for further analysis.

Loneliness

In order to measure the individual level of loneliness for each participant, the third version of the UCLA (University of California, Los Angeles) Loneliness Scale (Russell, 1996) was applied. The UCLA Loneliness Scale (UCLA-LS) is a self-report questionnaire consisting of 20 items, such as, “How often do you feel that you lack companionship?”, or, “How often do you feel alone?”. The participants were requested to answer all of them on a four-point Likert scale (never, rarely, sometimes and always). The total score can range from 20 to 80, whereat higher scores correspond to higher levels of loneliness. The psychometric properties of the UCLA-LS (Version 3) were also evaluated amongst a sample of elderly individuals, over 65 years of age (Russell, 1996). Results indicate that the UCLA-LS is very reliable, showing a test-retest correlation of .73 (for the elderly population). Internal consistency ranged from Cronbach's a = .89 to .94 for four different populations (college students, nurses, teachers and elderly). Construct validity was provided by significant correlations of loneliness with other instruments measuring well-being, health and the adequacy of interpersonal relationships in each population. Convergent validity was supported by significant associations with other instruments measuring loneliness. The cut-off threshold for loneliness was set at > 40 by the authors of the UCLA-LS (Russell, 1996). For this study, the threshold was set at 33.6 (mean value of the sample) to achieve a balanced distribution of non- lonely and lonely participants. After conducting a mean split, participants with a score > 34 were assigned to the lonely group.

Anxiety

Due to relations between anxiety and both, loneliness (e.g. Cacioppo et al., 2014) and interoceptive accuracy (e.g., Domschke, Stevens, Pfleiderer & Gerlach, 2010), the German version of the State-Trait-Anxiety Inventory (STAI; Grimm, 2009) was applied to measure anxiety. The STAI was designed by Spielberger and colleagues (1970) who distinguished between two different dimensions: anxiety as a personality trait and anxiety as a state. Trait anxiety is described as the general tendency of perceiving anxiety, whereas state anxiety is defined as the current level of anxiety a person experiences (Grimm, 2009). Since the participants filled in the questionnaire on a different day other than the scanning session took place, only the trait and not the state subscale of the STAI was assessed. The trait version of the STAI includes 20 items and is answered on a four-point Likert scale ranging from 1 (“almost never”) to 4 (“almost always”). It provides a score ranging from 20 to 80, with higher scores corresponding to higher levels of anxiety. An example of an item is, “I am happy” [Ich bin glücklich]. Good internal consistency was found for the original trait version in community­dwelling older adults with Cronbach's a ranging from .79 to .90 (Himmelfarb & Murrell, 1984; Stanley, Beck & Zebb, 1996).

Depression

In order to measure the level of depression, a German translation of the second version of the Beck-Depression-Inventory (BDI-II; Kühner, Bürger, Keller & Hautzinger, 2007) was handed to the participants. It a self-report questionnaire consisting of 21 items. Each item is answered on a four-point Likert-type scale, ranging from 0 to 3, whereas higher scores correspond to higher levels of depression within the past two weeks. The items measure diagnostically crucial symptoms of depression such as sadness, pessimism, past failure, crying, worthlessness, loss of pleasure, guilty feelings, punishment feelings, suicidal thoughts, concentration difficulty, etc. An example of an item is, “Sadness: 0 (‘I am not sad'); 1 (‘I am often sad'); 2 (‘I am always sad'); 3 (‘I am so sad and unhappy that I cannot stand it')” [Traurigkeit: 0 (‘Ich bin nicht traurig'); 1 (‘Ich bin oft traurig'); 2 (‘Ich bin ständig traurig'); 3 (‘Ich bin so traurig und unglücklich, dass ich es nicht aushalte')]. Kühner et al. (2007) propose values of the internal consistency ranging from .84 to .94. Scores that range from 0 to 13 denote no or minimal depression, those that range from 14 to 19 denote mild depression, scores of 20 to 28 are indicative for moderate and those of 29 to 63 for severe depressive symptomatology.

Alexithymia

Since previous research suggests an association of alexithymith both, loneliness (Qualter et al., 2009), as well as interoceptive accuracy (e.g., Brewer et al., 2016a), the German version of the 20-item Toronto-Alexithymia-Scale (TAS-20; Bach, Bach, de Zwaan, Serim & Böhmer, 1996) was administered for this study. The TAS-20 measures alexithymia on the basis of self-report and includes three subscales: Difficulty Identifying Feelings (DIF), Difficulty Describing Feelings (DDF) and Externally-Oriented Thinking (EOT). An example of an item is, “I am often confused about what emotion I am feeling” [Mir ist oft unklar, welche Gefühle ich gerade habe]. Items were rated on a five-point Likert scale (doesn't apply at all, does rather not apply, partially applies, largely applies and fully applies). Results of Bach and colleagues (1996) showed an adequate internal consistency of the total scale (Cronbach's a = .70 for a non-clinical sample). The TAS-20 uses a cut-off scoring: scores less than 51denote no alexithymia, scores from 52 to 60 denote possible alexithymia and scores equal to or above 61 denote alexithymia (Bagby, Taylor & Parker, 1994).

Interoceptive Accuracy

To measure interoceptive accuracy a variant of the heartbeat tracking task, originally developed by Schandry (1981), was applied. For this study an fMRI design similar to the one implemented by Wiebking (e.g., Wiebking et al., 2010) was utilized. Wiebkings' design included three different experimental conditions - an interoceptive task (heartbeat counting), an exteroceptive task (tone counting) and rest periods. In the present study, the aforementioned design was modified by replacing the rest periods with a time estimation condition. Anyhow, the time estimation condition was added to the task because both abilities, heartbeat counting, as well as time estimation require an activation of the insular cortex (Wittmann, 2013). Hence, subjective time estimation could be impacted by one's ability to perceive the heartbeat, which might serve as an internal reference (Wittmann, 2013). Since current literature suggests that the tone counting condition is much more common as control task (e.g., Schulz, 2016), only the data of the tone counting condition and not those of the time estimation were analysed. For the sake of completeness, the tone counting condition was nevertheless described within this section.

During the interoceptive task (IT), participants were instructed to count their heartbeat without the help of any external manipulation (e.g. measuring the pulse on the lower forearm or at the throat). Each trial had the same structure: resting period, preparation period, condition and one or two rating periods. At the beginning of the trial a black fixation cross on a white screen was shown to the participant (resting period). Afterwards the preparation period began. Here, the display of a grey symbol (in this case the symbol of a heart) served as a signal for the participant to know which task is next. He/ she had three seconds of time (indicated by dots) to prepare for the task. When the grey heart turned black, the actual condition began and the participant started to count his/ her heartbeat until the black symbol disappeared again. Subsequently, the participant had to rate how many heartbeats he/ she counted (rating period). In case he/ she did not perceive his/ her heartbeat at all, the participant was instructed to answer “0”.

The exteroceptive task (ET) had the same structure as the interoceptive task (resting period, preparation period, condition and rating period), with the difference that the symbol to indicate the task type was a musical note. During this task, the participants were instructed to silently count pulsating tones played by the scanner loud speaker. To make the frequency of the tones distinguishable from the scanner noise, it was set to 440 hz. The duration of the tones was set to 200 ms in order to adjust them to the possible duration of a heartbeat. Since each of the three conditions should be comparable in terms of difficulty, the volume of the tones was calibrated individually by each participant so that it was just audible, like the heartbeat (see study procedure). The stimulus frequency was matched to the participant's heart rate with a jitter of 200 ms to avoid habituation effects. After the condition took place, the participant was asked to report the number of counted tones. The exteroceptive task was chosen to control for neural responses to external stimuli compared to internal stimuli and as well for the control of general effects of counting (see Schulz, 2016).

During the time estimation task (TE) the participants were requested to count the elapsed time of an interval in seconds. The structure of the task was the same as in the other two tasks (resting period, preparation period, condition and one or two rating period), with the difference that the symbol to illustrate the type of the task was depicted by a clock. After the condition took place, the participant was asked to rate the number of counted seconds in the same way as in the other two task.

The total experiment consisted of 18 trials so that each of the three tasks was executed six times. As mentioned before, each trial had the same structure (resting period, preparation period, condition and one or two rating periods). The resting period was jittered and had a duration of 5 ± 2 seconds. The preparation period had a fixed duration of 3 seconds. The condition (IT/ ET/ TE) was jittered and had a variable duration of 20 ± 5 seconds. The duration of the rating period was variable as well (depending on the participant's velocity) and took place after each trial. After the middle and the end trials of each of the three tasks there was a confidence rating. On a continuous visual scale ranging from ‘not confident at all' on one end to ‘very confident' on the other end, the participant was asked to indicate the degree of perceived confidence regarding his/ her rating in the step before. In the middle of the experimental task (after the 9th trial) the participants had a short break of 10 seconds.

fMRI Data Acquisition

The fMRI data was acquired using a Siemens Magnetom Skyra 3-Tesla MRI-system (Siemens Medical, Erlangen, Germany) which was equipped with a 32-channel head coil. It was located at the dental clinic of the University of Vienna. For the MR-scanning, a gradient- recalled echo-planar imaging-system with correction of distortion (echo time = 34 ms, repetition time = 704 ms, voxel size = 1.5 x 1.5 x 5 mm, 32 slices, fractional anisotropy (FA) = 50°) was applied. For the behavioural task and the stimulus presentation (the tones during the sound calibration) MATLAB R2017b (MathWorks) was utilized. Stimuli were presented and recorded using Cogent 2000, a toolbox implemented in MATLAB.

Study procedure

The heartbeat tracking task was conducted within the context of a larger fMRI research project. Therefore, participants were required to execute two further tasks in the MR-scanner (an economic game and a social touch task). Since those tasks are irrelevant for the present research question, they will not be mentioned in the subsequent description of the study procedure.

As a first step, potential subjects, who expressed their interest in participating in the study, were asked to fill out three online questionnaires containing sociodemographic information, the MRI Safety Questionnaire, as well as the UCLA-LS. If they were eligible for the study, they needed to orally confirm during a personal phone call that they had no risk factors regarding MRI scanning (e.g. wearing permanent make-up or having unremovable metal parts such as implants in the body). Next, they were invited to a screening appointment, during which they were asked to fill out several instruments including the STAI, the BDI-II and the TAS-20. Additionally, they were requested to perform the MMSE. After arranging an appointment for the scanning session, they received an e-mail with the informed consent, a document designed to teach about the rights and responsibilities of study participants. On arrival, the MRI Safety Questionnaire and the informed consent were handed to the participant in order to check again for MRI-related risks and to give the subject the opportunity to clarify remaining questions. It was only proceeded with the experiment if the participant signed the two beforementioned forms. Subsequently, the participant was invited to enter the scanning area where the test administrator explained the structure of the experiment. To obtain more detailed instructions and to practice each task, the participant underwent a short computer training (outside of the MR-scanner). Since the stimulus frequency of tones during the exteroceptive condition was individually adapted to each participant's heartrate, the pulse was measured prior to the interoception training. Moreover, before the interoception training a short heartbeat perception training was performed, during which the participant should lean back, relax and try to focus on his/ her heartbeat without manually measuring it. In case the participant couldn't perceive his/ her heartbeat, he/ she was instructed to concentrate on his/ her breath and to try it again. If he/ she couldn't feel his/ her heartbeat at all he/ she should answer “0” in the rating of counted heartbeats. The interoception training consisted of six practice trails (two per condition) that were presented to the participant in a pseudorandomized sequence. Before entering the MR-scanner for the execution of the experimental tasks, the test administrator made sure that the participant did not carry any metal objects with him/ her. In the MR-scanner a head coil was placed over the participant's face. Above, a mirror was installed with which the participant was able to see the instructions of the task on a monitor behind him/ her. Furthermore, the response box, a remote control that enabled the participant to perform the task, was fixated upon his/ her right side. Another device, the pulsemeter, served to measure the participant's heartrate.

[...]