Introduction

Social cognition refers to the perception, interpretation, and processing of all information relating to a person’s social environment and relationships [1]. The term social cognition is interchangeable in the literature with other terms (e.g., Theory of Mind, Mentalization). In fact, the historical interest in how humans are aware of one’s own and others’ mental states is patent in the number of approaches that its study has produced. For instance, constructs such as social intelligence [2], metacognition [3], Theory of Mind (ToM) [4], mind-blindness [5], inter and intra-personal intelligences [6], emotional intelligence [7] and others (e.g., mind-reading or insight) seem to be referred to the same Higher-Order Cognition [8]. In this context, the recent Mentalization paradigm contributes to systematize a field with multiple historical approaches.

Mentalization or mentalizing (MZ) emerges from recent advances in several disciplines (developmental psychology, neuropsychoanalysis, social neuroscience) to address a well-studied social-cognitive capacity from a new multidimensional point of view [9,10]. Thus, MZ paradigm organizes the ability to recognize intentional mental states (i.e., not only thoughts, but also feelings, emotions, desires, impulses) [11] according to four basic dimensions [12]. MZ is a multidimensional construct that refers both to self and others’ mental states, both to cognitive and emotional functioning, both to implicit-automatic and explicit-controlled activation of the ability, and it can be based on internal versus external features of self and others [13,14]. Given that MZ constitutes a new broad paradigm that integrates several constructs and systematizes both self-awareness phenomena (e.g., insight, mindfulness, metacognition), and the wide field of social cognition (e.g., empathy, Theory of Mind (ToM), mind-reading), this is the perspective adopted in the present study in order to assess two dimensions of this capacity: MZ referred to one’s own and to others’ mental states.

The awareness of one’s own and others’ mental states not only allows the understanding and prediction of human behavior, but it is related to good social functioning and mental health [15–17]. By contrast, evidence supports that MZ impairment is related to several psychopathological conditions [18]. For instance, a diminished capacity to imagine others’ mental states (i.e., ToM deficit) is associated to autism spectrum disorders [19–21] and psychosis [17,19,20,22,23]. Deficiencies in complex mental state reasoning and in simple mindreading have been found, respectively, in adolescents and children with psychopathic symptoms [24–28], and the tendency to attribute excessive or unrealistic mental states has been described in borderline personality disorder [12,29–32]. In addition, recent studies showed that problems in MZ are also present in depression [33], anxiety [34,35], trauma-related disorders [36], eating disorders [37], attachment-related problems [11,38] and addictions [39,40]. Thus, problems in MZ capacity are present in several psychopathological conditions. Given the possibility that social cognition may play an important role in the presence of psychopathology, there is growing interest to extend the study of this Higher Order Cognition [8] to other psychopathological domains.

Social cognition is especially interesting to be analyzed in disorders with an important social component. In the Social Anxiety Spectrum [41] the role of cognition (i.e., attention, perception and memory biases) [42,43] and meta-cognition [44] have been widely documented, but the specific role of social cognition remains relatively underexplored [35,45]. This is important because, in contrast to other anxiety phenomena, the basic fear in social anxiety (SA) is precisely “social” (i.e., SA is defined by the fear of scrutiny, social interaction, and negative judgment of others [46].

Interestingly, as Banerjee and Henderson [45] already pointed out, the relationship between social anxiety and social cognition is very probably bidirectional, and this means that they may influence each other along development. Thus, in addition to the contribution of social cognition to social anxiety, it might be reasonable to analyze how social anxiety may contribute to the development of social cognition.

Despite this, among the relatively small number of studies analyzing social cognition in SA, most of them are focused on how social cognition contributes to SA [47–54], but very few of them analyze how SA impacts social cognition development. This is important because if SA and social cognition feed each other in some way, maybe clinical levels of SA in adolescence or adulthood depend to some extent upon this interaction.

Most of the studies about social cognition in SA are based on adult samples with clinical levels of SA [47–52]. In contrast, there are relatively few studies focused on how non-clinical forms of SA in early stages of development [i.e., temperamental shyness or behavioral inhibition (BI); [46,55–57] impact the development of social cognition.

In this matter, the preliminary investigation of Banerjee and Henderson [45] highlighted the interplay between social cognition and SA and established that socially anxious children, particularly those with high levels of shy negative affect, experienced specific social-cognitive difficulties in understanding the links between emotions, intentions, and beliefs in social situations. According to this pioneering study, any condition that leads to the avoidance of social interaction may affect the “practice” and development of social cognition. Therefore, an important question is whether SA along development, which is typically associated to social avoidance, may affect social cognition or, in other words, MZ capacity.

After the correlational study of Banerjee and Henderson [45], two lines of findings report variations in MZ as a consequence of early forms of SA. According to the emotional reactivity hypothesis [58] suggesting that emotional arousal impairs ToM development, there is evidence about an inverse relationship between early SA and social cognition. For instance, Suway, Degan, Sussman, and Fox [59] found that children with high levels of BI at 24 months of age showed less ToM understanding at 36 months, in comparison with non-behaviorally inhibited children, supporting the notion of a possible impairing effect of BI on MZ capacities. This study is consistent with previous ones supporting that the higher the SA (i.e., shyness, BI), the lower is level of social cognition [60,61].

By contrast, there is another line of results pointing that early forms of SA may lead children to an observant position regarding social phenomena that may help to improve social cognition development. In this sense, Mink, Henning, and Aschersleben [62] found that shy temperament at 18 months was associated with more sophisticated ToM abilities at age 3. According to previous studies [63,64], in this case, these authors interpreted that a more observant temperament (shyness) provided better opportunities for reasoning about others’ mental states.

Finally, some studies like those of Colonnesi and colleagues [65] and La Bounty and colleagues [66] report findings partially supporting both lines of results. Thus, it is still not well-known whether shyness might benefit or impair social cognition development. Both findings provide evidence for an influence of early SA in MZ abilities, but maybe the differences in measures of shy temperament and ToM, differences on how ‘shy’ or ‘inhibited’ condition is defined, as well as differences in age sample and time of follow-up may explain discrepant results.

Most of the studies analyzing this topic are focused on an early and narrow period of development usually placed [59,62,63,65,66] or beginning at the preschool age [60]. However, although there is evidence about errors of metaperception in socially anxious adolescents [67], there is no study analyzing whether early forms of SA in childhood (such as BI) predict MZ capacity in adolescence.

Therefore, the first aim of the current study was to analyze if BI in childhood may be associated to MZ in a later period of development. Specifically, we aim to analyze whether MZ in adolescence is associated to BI in childhood, retrospectively assessed. In contrast to some previous findings, we assume that the tendency to avoid social interaction over years, due to a high level of temperamental BI, may impede practice and sophistication of MZ abilities and therefore lead to lower levels of MZ in adolescence. Thus, we predict that a group of adolescents with high level of childhood BI, retrospectively assessed, would show lower MZ abilities than a group with no BI.

However, considering a possible explanation to previous discrepant results, we also hypothesized that, although a high level of BI may impede MZ development—because anxiety leads to social avoidance and excessive arousal blocks mental functioning [68] it is possible that moderate levels of BI may be associated with higher mentalizing capacity because BI may foster a tendency to pause and observe rather than immediate action. This combined hypothesis may explain why some studies find that early SA leads to better ToM, and may help to integrate the previous conflicting results.

Thus, the second aim of this study is to examine whether a group of adolescents with a moderate level of BI in childhood show higher levels of MZ in adolescence than a group with high childhood BI, and even than a non-inhibited group.

In addition, given that most preceding studies have given preference to analyzing the relationship between SA and MZ of others’ mental states, the third aim of the current study was to see whether early SA, operationalized here as a measure of BI, not only affects social cognition but also affects the ability to consider one’s own mental states. Neuroscience research reveals that MZ capacities referred to self and others’ share several neurological routes [69], and other evidence suggests that the impairment of emotional knowledge in SA may be more pronounced in the intrapersonal (i.e., own emotions) than in the interpersonal form (i.e., others’ feelings) [70]. Therefore, we predicted that the group of adolescents with high BI in childhood would show MZ problems both referred to self and others’ mental states.

In light of preceding findings [71] suggesting higher MZ capacity in girls compared to boys [72,73], and evidence of higher prevalence of SA in girls than in boys [74], all the analyses were controlled for sex. Given the wide range of age considered (i.e., 12 to 18 years old), the effect of age also taken into account. In addition, given that 1) anxiety is usually considered a backstage of depression [75,76], 2) there is specific evidence for secondary depression associated to SA [77–80], and the well-known association between MZ problems and depression [33,75,76], depressive symptoms were controlled for in all analyses in order to avoid the confusion effect of depression. For the same reason, given that BI usually leads to SA in adolescence [55], SA in adolescence was also controlled for in order to avoid the confusion of childhood BI and adolescent SA effects on MZ capacities, and thus obtaining a more accurate measure of the relationship between childhood BI and adolescent MZ.

Materials and methods

Results

Comparing adolescent MZ capacities between groups with high and low childhood BI

Table 1 shows descriptive statistics of MZ measures for every group of BI. These groups are based on the level of BI in childhood, retrospectively assessed with the second part of the Behavioral Inhibition Scale. On the base of the participants’ responses, the group with high BI includes 37 of the 256 participants of the whole sample (15%), the group of low BI includes 80 (31%) and the moderate group includes the remaining 138 participants (54%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics of dependent variables by group of behavioral inhibition (BI).

https://doi.org/10.1371/journal.pone.0195303.t001

Table 2 shows the effect (B), the 95% confidence interval (95% CI) of B and the p value, of BI (High vs Low, High vs Moderate, Moderate vs Low) on different MZ measures. The first column of Table 2 shows the effects of having high BI vs. low BI on MZ. Adolescents with high BI in childhood show significantly lower MZ capacity, in almost all measures, compared to adolescents with low BI in childhood. Thus, for global MZ measures (i.e., encompassing both MZ referred to own and others’ mental states), the group with high BI in childhood scored between 2.2 and 5.3 points lower than the low BI group for the BRF (i.e., 10% to 20% lower, given the range of BRF: 0–24), and between 1.8 and 5.4 points lower (i.e., 5–20% lower) in the AMI interview. This result is consistent across global and specific MZ measures, both for MZ referred to own and others’ mental states and both for the self-report and interview. Thus, with the only exception of the measure of MZ deficit provided by the MZQ, these results suggest that a high level of BI in childhood is associated to impaired MZ capacities in adolescence when compared to low level of BI.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Multiple lineal regressions comparing levels of mentalizing among groups with different levels of BI.

https://doi.org/10.1371/journal.pone.0195303.t002

Discussion

The first aim of the current study was to analyze whether retrospectively reported BI in childhood, which is considered an early form of SA [41,55], is associated with impaired MZ capacity in adolescence. Retrospective BI was evaluated with the second part of the Behavioral Inhibition Questionnaire (BIS), which consists of presenting three descriptions and ask the informant which one better identifies his or her profile. The resulting groups, based on the responses of the participants to the BIS—i.e., not based on cut-offs selected by the researchers—suit relatively well the criteria of the group of Harvard [95–97], which establishes that 15% of the general population can be considered qualitatively ‘inhibited’. However, this percentage was established by previous studies of general populations [96,98,99], while the current sample shows an overrepresentation of SA. This suggests that the retrospective self-classification of historically ‘inhibited’ might be underestimated. In fact, to the well-known biases attributed to retrospective measures [100], it must be considered that also variations in MZ, especially in self-referred dimensions (i.e., the capacity to be aware of one’s own mental states or meta-cognition) can affect any self-report.

In any case, the over-representation of SA in the current sample was a decision design of the comprehensive project to make easier the study of SA phenomena. So it should also benefit the current study. Thus, the underestimation of historically ‘inhibited’ children in the current sample suggests that some group differences might be attenuated since some ‘inhibited’ children have been wrongly self-classified in a non-high inhibition group.

Having said that, the analysis of current results supports that having been highly inhibited in childhood is indeed associated to lower MZ capacities in adolescence, both compared to those with low BI and even to those with intermediate BI along development. This is important because the fact that also contiguous groups (intermediate vs. high BI), show differences in MZ capacities reinforces the implications derived from the comparison of extreme groups (i.e., high vs. low BI).

Moreover, results do not only support that high BI in childhood is associated to poorer social cognition in adolescence, but also indicate that inhibited children exhibit impairments in both self- and other- dimensions of MZ. This is reasonable since social interaction seems necessary both to develop social cognition and also meta-cognition, since others’ point of view (i.e., mirroring; [101]) may help to increase self-knowledge in terms of recognition and management of own mental states in the social world [102].

To our knowledge, a relatively small number of studies analyzed how SA may affect social cognition, and their results go in two opposite directions. While some studies provide evidence for Hare’s [58] emotional reactivity hypothesis, which supports that less reactive temperaments lead to more sophisticated social-cognitive skills [59–61], other studies clearly support the opposite, that is: that shy / inhibited temperaments lead to observant attitudes that foster social-cognitive development [62–64]. These conflicting results lead Lane and colleague’s [64] to point the necessity of carefully review Hare’s hypothesis in the case of humans, and this was what motivated current research to consider 3 different levels of childhood BI here.

According to previous studies [59–61], the current results can be interpreted in light of what classic research on BI supports [95–99, 103]: Inhibited children are more socially reticent [104] and less likely to approach or engage others in social interaction [105]. As a consequence, they have fewer opportunities to learn the abilities necessary for competent social cognition [106]. In turn, impoverished social cognition may contribute to further declines in social understanding and therefore more social avoidance [107], thus suggesting that the relationship between SA and social cognition may be bi-directional. This apparent reciprocal influence, already suggested by previous research [45], may be an important point to explain the stability of both SA and social cognition deficit, and allows to combine the current findings with those supporting that problems in MZ may play a role in the development of SA (e.g., [47–52]).

By contrast, the current results do not support those of previous studies reporting better MZ capacities in preschoolers with higher shyness [62–64]. It is worth noting that it was precisely the existence of conflicting results what suggested us that maybe they were different levels of SA that may influence the development of SA differently. Therefore, it was hypothesized that, although a high level of BI could be so arousing to lead to social avoidance and consequently to impair social cognition, maybe an intermediate level of SA may lead children to stop and observe—instead of avoid or flee—thus improving social cognition instead of impairing it. This may go along with previous studies supporting Lane and colleagues’ statement [64], and it is possible to apply Wellman and colleague’s postulate to moderate shyness, considering that moderate levels would allow for a more quietly observant stance toward social interaction, which would in turn yield to better insight (i.e., MZ) of interpersonal processes. However, against predictions, current results are clear in refuting this hypothesis, and indeed support that levels of BI are proportional to the level of impairment in MZ ability.

Interestingly, differences between the intermediate and the low BI groups only appear regarding self-reports and not for interview measures. In this context, it is interesting to note that an interview, which is an emotional-arousing situation for people with SA, did not detect differences in MZ abilities between groups with non-high SA (i.e., low versus moderate BI), while self-reports, reflecting informants’ appraisals of their capacity to MZ, indeed supported lower MZ abilities in people with history of intermediate BI. Although participants with low and moderate childhood BI show no differences in MZ functioning over an interview, it seems as if a history of intermediate BI may influence self-perception of MZ capacity. This is important because one of the expected limitations of this study is precisely the use of self-reports to assess MZ—because an impaired MZ capacity, especially regarding one’s own mental states, may affect all psychological assessments. Thus, it could be expected to find differences with the interview—considered more reliable—and not with self-reports, but not the opposite. However, current results indicate, in light of the interview measurements of MZ, that to have been inhibited is associated to poorer MZ capacity compared to moderate and low childhood BI, but there are no differences between having been moderate- and non-inhibited. This sheds light on an additional issue: the potential differences between MZ measures heighten the debate about the still unsolved difficulty to apprehend this Higher Order Cognition [108,109]. We tried to attenuate this problem using different measures of MZ in the current study—i.e., a multimethod perspective—, but this also obligates researchers to carefully consider the conclusions derived from the current findings until appropriate replication with improved measurement can be carried out.

A final prediction of this study concerned the multidimensionality of the impairment of MZ associated to SA. In light of previous research [35,44] and classical theories about SA [46,55–57], it was predicted that developmental BI may be associated not only with differences in social cognition (i.e., MZ referred to others’ MS), but also to metacognition (i.e., MZ referred to own MS). Results supported this hypothesis with both self-report and interview measures, both when the high BI group is compared to low BI and also to moderate BI group. This finding seems to be robust given the dose-response pattern observed for the influence of SA on MZ abilities.

However, there are two exceptions to this conclusion. The first one is the measure of MZ deficit provided by the MZQ. This is interesting because the MZQ is a specific measure of deficit and this is the only measure in which any difference was found. The MZQ was used because it was the only evidence-based psychometric measure of MZ available at the time when the study was conducted. However, precisely for being an instrument specifically designed to detect deficit (i.e., clinical impairment) in MZ in adults from clinical population, it is possible that its items (e.g., “Sometimes feelings are dangerous for me”, “Most of the time is better not to feel anything”) are not the most suitable to capture differences in a sample of adolescents from general population.

The second exception implies that adolescents with high and moderate BI along childhood do not differ in the meta-cognition specific measure (i.e., TMMS-24), although they do in the other two measures referred to the self-MZ dimension. The TMMS-24 is a general measure of meta-cognition with items referred to attention, recognition and comprehension of own mental states. In contrast to the more precise self-MZ measures (i.e., BRF, interview), in this “G-factor” of meta-cognition, results indicate that the moderate group is clearly nearer to the high BI group than to the non-inhibited group. This reinforces the idea that SA along development (either high or moderate) not only affects social cognition but also metacognition, and that on this regard there is no difference across levels of SA, because differences appear on the basis of presence (either high or moderate) compared to absence of SA (i.e., the low BI group).

In conclusion, this study indicates that retrospective assessments of childhood BI are associated to poorer MZ in adolescence, and that not only the social cognition dimension is impaired (i.e., MZ referred to others’ mental states), but also meta-cognition. Contrary to our predictions, intermediate levels of BI are not associated to an increased MZ capacity in adolescence. By contrast, they are associated to and intermediate impoverishment of MZ. These findings underscore the relevance of taking into account at early developmental stages the potential implications of BI in terms of a potential detrimental effect for MZ abilities beyond its well-known impact on increasing risk for SA.

Supporting information

Acknowledgments

The authors would like to thank all families who contributed to the study, as well as the schools Escola Pia Sarrià-Calassanç, Institut Lluís Vives, Institut Besos, Institut Corbera and Institut Angeleta Ferrrer i Serrat from Barcelona and Institut Francesc Ribalta and Escola Arrels from the center of Catalonia for their collaboration. Special thanks go to Judit López-Berga, Ángel Lorite, Ruth Ocón, and Cristian Lago for their contribution to data collection, and Teresa Kearn for proofreading the manuscript.