Examining differences in cognitive and affective theory of mind between persons with high and low extent of somatic symptoms: an experimental study

We used a quasi-experimental design with the between-subjects factor extent of physical symptoms, divided into two groups (see details below): participants who reported a high extent of somatic symptoms (HSR) and participants who reported a low extent of somatic symptoms (LSR). The experimenters were blind to the group membership of the participants throughout data collection and Faux Pas Recognition Test (FPRT) ratings (see below).

A written screening, which included the SL [31], demographic items (age, sex, highest education), the informed consent and some health-related questions (chronic diseases, use of medication, number of medical consultations and hospital stays in the last 12 months) was sent by post to potential participants. Participants who met the inclusion criteria (see Recruitment and Sample Characteristics) were invited to the laboratory part of the study which took place in the Georg-Elias-Müller Institute for Psychology in Göttingen and was conducted by two female experimenters, who were blind to the participants’ group membership. The study lasted approximately 90 min and was divided into two parts. In the first part, the participants’ aToM and cToM were assessed using the paradigms described (ERaToM, FPRT). In order to minimize sequence effects, the order of the two paradigms was balanced between participants and a break of 5 min between the two tasks was implemented. In the second part, the participants completed a set of questionnaires including the TAS-26 [39], the ECQ [36], the HADS [37] and the PANAS [38]. In the end, the participants were informed about the goals of the study and had the opportunity to ask questions.

The differences between the two groups regarding the extent of somatic symptoms, emotional awareness, positive and negative affect, alexithymia, scores in the ERaToM and the FPRT were examined using independent analyses of covariance (ANCOVAs) with age and sex as covariates. The covariates were included because the sample size was rather small and variations in age and sex may influence the results. The groups showed an age difference (M _HSR  = 49.76 ± 10.66; M _LSR  = 42.71 ± 12.01; t (47) = 2.19, p = .034; Table 1). In addition, prior research suggested female superiority in theory of mind and the number of women was not equally distributed between the groups (even though the proportions did not differ significantly; HSR: 20 females, 9 males; LSR: 17 females, 4 males; χ ²  = 0.91, p = 0.34) [40]. For effect size, Eta squared is reported [41].

As the groups showed significant differences in anxiety/depression and positive and negative affect (see below), we conducted further exploratory analyses: ANCOVAs were employed to test whether the difference between the HSR and LSR in self-reported emotional awareness, affective ToM and alexithymia were maintained after controlling for the differences in anxiety/depression and negative and positive affect (the latter can be found in Additional file 1: Table S1). In order to analyze the relationships of the variables among each other, we calculated Pearson correlations. The significance level was set at p = 0.05.

The correlations between the variables were high (see Table 2), for example self-reported emotional awareness showed a strong negative correlation with anxiety/depression (r = −.53, p < .01), with positive and negative affect (r = .57, p < .01; r = −.57, p < .01) and with the sum score of the TAS-26 to measure alexithymia (r = −.76, p < .01).

We calculated two ANCOVAs to test whether the HSR and the LSR groups differed with regard to the two subscales of the ECQ while using age and sex as covariates. HSR participants reported significantly lower emotional awareness of their own emotions than LSR participants (F (1, 46) = 7.33, p = 0.009; Table 6). The covariates were not related to the participants’ self-reported emotional awareness (F _age (1, 46) = 2.71, p = 0.11; F _sex (1, 46) = 0.12, p = 0.745).

The groups did not differ with regard to the recognition of other’s emotions (F (1, 46) = 0.19, p = 0.662) and the covariates were not related to the participants’ recognition of other’s emotions (F _age (1, 46) = 0.72, p = 0.400; F _sex (1, 46) = 2.30, p = 0.136; Table 3, Table 6).

In order to calculate whether the groups differed in depression, anxiety, affect or alexithymia ANCOVAs were calculated using age and sex as covariates. Compared to the LSR participants, HSR participants reported significantly higher depression and anxiety ratings (F (1, 46) = 18.44, p < 0.001), more negative and less positive affect (F _{positive affect} (1, 46) = 12.42, p = 0.001; F _{negative affect} (1, 46) = 14.14, p < 0.001) and higher alexithymia scores (F (1, 46) = 8.02, p = 0.007). Neither age nor sex were related to these variables (Table 3, Table 7).

As the multidimensional construct of ToM can be differentiated in two subcomponents (cToM and aToM), we used two corresponding tests – the FPRT and the ERaToM – which were specifically designed to assess these components, respectively [32, 34]. As a precondition for the aToM task, we also tested whether the groups differed in their ability to recognize the emotions from facial expressions and found no differences. There were no group differences between HSR participants compared to LSR participants regarding their cToM, but participants of the LSR group showed higher aToM than participants of the HSR group. This divergent result concerning aToM and cToM is in line with Stonnington and colleagues [17] who failed to find impaired cToM functioning in patients with somatoform pain disorders (as compared to healthy controls), but whose results also indicated reduced aToM functioning in their patient group. Subic-Wrana and colleagues [15] found reduced ToM functioning in patients with somatoform disorders when compared to healthy controls and Zunhammer and colleagues [16] reported the same for patients suffering from chronic somatoform pain. However, in the latter two studies, affective and cognitive ToM were not differentiated hindering a direct comparison of the results. With regard to aToM the current study is in line with their results and indicates the need for a differentiated approach that incorporates both facets of ToM. The current study extended previous results as it demonstrated that aToM differences can be observed not only by contrasting patient groups with healthy controls but such differences in functioning can be found already in an analogue sample. This is of particular importance as it may indicate that aToM deficits precede and, perhaps, promote the development of a somatic symptom disorder.

The fact that the high symptom reporters differed from the low symptom reporters in affective but not in cognitive ToM may also be rooted in different requirements of the tasks: whereas the ERaToM is a speeded computer based test, the Faux pas test allows more time and context information; both of which may help to solve the task. Possibly, existing slight deficits of the high symptom reporters could be compensated in the cognitive Faux pas test but not in the speeded ERaToM.

So far, studies that found impaired ToM in patients with somatoform disorders [15‐17] assessed ToM with the AT. The AT is a projective measure that uses visual animations featuring two moving triangles in a framed background. The triangles’ movements can be interpreted as ‘interactions’ that ‘make sense’ if the observer considers that the triangles think, feel or intend (e.g. one triangle wants to surprise the other). The participants’ task is to describe what happens in the animation. The answers are then coded firstly with regard to the degree to which the participants interpreted the purpose of the actions of one triangle as influencing the mental state of the other triangle and how appropriately the script had been captured. We conducted the present study to broaden the empirical basis by including paradigms that assess ToM with ecologically more valid stimuli. Firstly, this allows excluding the participants’ verbal fluency as a confounding factor. In addition, and more crucially, as ToM refers to the ability to attribute mental states to persons (oneself and other people; [23]) it seemed desirable to assess this construct by testing whether people assign feelings to real persons who indeed showed emotional facial expressions rather than to geometrical figures, which in reality do not have any feelings at all. The fact that differences in aToM were found with a different experimental paradigm contributes to showing that the differences found may indeed be robust.

To sum up, the current results indicate a deficit in aToM in participants reporting a high amount of somatic symptoms and clearly highlight the need for future research that uses different and varied paradigms to assess the two components of ToM.

At first glance, HSR participants reported less emotional awareness regarding their own emotions than LSR participants, corresponding well with previous results [15, 16, 19]. However, self-reported emotional awareness was strongly correlated with symptoms of anxiety and depression. In line with previous research [8‐10, 12], HSR participants reported higher anxiety and depression scores in the HADS [37] than LSR participants. For this reason, we explored the relationship between self-reported emotional awareness, somatic symptoms and depression/anxiety more closely and found that HSR participants did not differ in their awareness of their own emotions when we controlled for their elevated depression/anxiety scores. The same held true when we used alexithymia as a dependent variable and depression/anxiety as a control variable. Thus the observation that differences in self-reported emotional awareness are abolished when controlling for depression and anxiety, does not seem to be specific to the questionnaire used, but appears to represent a more robust finding. These results correspond with those by Stonnington [17] and Lane [20], but are in contrast with Subic-Wrana [15], who found significantly lower emotional awareness in patients with somatoform disorders when compared to healthy controls, but no relationship between depression and emotional awareness. However, in their paper, they did not report the respective statistical measures.

One possible explanation for the diverging results may lie in the investigated populations. In [15] and [16] the emotional awareness of hospitalized patients was compared with healthy controls, whereas in the current study the sample comprised healthy patients who differed in their extent of somatic symptoms. It is possible that the differences in emotional awareness may be too small to be detected in a healthy population. However, Stonnington [17] who investigated outpatients whose level of suffering might lie considerably above healthy participants but slightly below that of inpatients also failed to find impairments in the emotional awareness. Differences between hospitalized patients and healthy controls are not limited to somatic symptoms; hence it is also possible that other confounding factors may explain the group differences found in these studies.

A strength of this study lies in the robust methodical procedure: ToM was assessed by two different experimental paradigms that were specifically developed to assess the two subcomponents of ToM, were multimodal (computer tasks, text vignettes), required different sensory approaches (auditory, visual) and were conducted by experimenters who were blind to the participants’ extent of somatic symptoms. The sample was recruited from the general population of Göttingen and the surrounding area. In spite of the fact that we compared an analogue sample formed by using extreme groups with regard to their extent of somatic symptoms, the participants were all healthy, which decreases possible confounding factors. However, the fact that two key variables (sample and measures) differed from previous studies, limits the direct comparability of the results. The negative results concerning cToM and emotional awareness have to be interpreted in light of the limited sample size, which – although comparable to prior studies [15, 16] – would still benefit from replication in a larger sample.