Attachment Style Predicts Cortical Activity in Temporoparietal Junction (TPJ)

Abstract

Results of the behavioral studies suggest that attachment styles may have an enduring effect upon theory of mind (ToM). However biological underpinnings of this relationship are unclear. Here, we compared securely and insecurely attached first grade university students (N = 56) in terms of cortical activity measured by 52 channel Functional Near Infrared Spectroscopy (fNIRS) during the Reading the Mind from the Eyes Test (RMET). The control condition involved gender identification via the same stimuli. We found that the ToM condition evoked higher activity than the control condition particularly in the right hemisphere. We observed higher activity during the ToM condition relative to the control condition in the secure group (SG), whereas the overall cortical activity evoked by the two conditions was indistinguishable in the insecure group (ISG). Higher activity was observed in channels corresponding to right superior temporal and adjacent parietal cortices in the SG relative to the ISG during the ToM condition. Dismissive attachment scores were negatively correlated with activity in channels that correspond to right superior temporal cortex. These results suggest that attachment styles do have an effect on representation of ToM in terms of cortical activity in late adolescence. Particularly, dismissive attachment is represented by lower activity in the right superior temporal cortex during ToM, which might be related to weaker social need and habitual unwillingness for closeness among this group of adolescents.

Attachment theory posits that individuals build emotional and cognitive representations (internal working models) of the self and others, based on their previous interactions with caregivers and significant others (Bowlby, 1973). Further on, these representations constitute a model that predicts how social information is appraised throughout the lifespan (Collins & Feeney, 2004). Bartholomew and Horowitz (1991) proposed a 4-group model of attachment styles in adulthood that are defined using combinations of a person’s self-image and image of others. These attachment styles have been shown to predict mental and physical health (Widom, Czaja, Kozakowski, & Chauhan, 2018) and social functioning in daily life (Sheinbaum et al., 2015). They do have an influence on the occurrence and outcome of psychiatric disorders (Ikeda, Hayashi, & Kamibeppu, 2014; Tasca & Balfour, 2014; Quijada, Kwapil, Tizón, Sheinbaum, & Barrantes-Vidal, 2015) as well as the course of psychiatric treatment (Byrow & Peters, 2017; Fowler, Groat, & Ulanday, 2013).

Theory of mind (ToM) refers to the ability to attribute mental states – beliefs, desires, and intentions – to others and understand that others can have such states that are different than one’s own. The ToM therefore has an important predictive role in understanding social behaviors among humans and non-human social animals (Premack & Woodruff, 1978). The neural representations of these mental states can serve as models that can be utilized in planning behaviors during social interactions with others. Some authors argued for an intrinsic link between attachment style and the acquisition of ToM ability (Dykas & Cassidy, 2011; Villachan-Lyra et al., 2015). That is, understanding the minds of others may occur, on the basis of understanding the self through creative imitation, a mechanism that is favored by the quality of attachment relationships established by the child with his mother figure (Valsiner & van der Veer, 2000; Villachan-Lyra et al., 2015). Supporting this view, securely attached children have better performance than insecurely attached ones in socio-affective, cognitive, and ToM tasks (Laranjo, Bernier, Meins, & Carlson, 2010; Steele, Steele, Croft, & Fonagy, 2001; Villachan-Lyra et al., 2015).

The relationship between attachment and ToM may also have neurobiological implications. Increasing body of evidence suggests that adversity in early life shapes maturation of rapidly developing brain circuits underlying emotional functions. For example, environmental stress in early life may impact cortical synaptic plasticity via epigenetic mechanisms, such as changes in DNA methylation pattern, histone acetylation, and microRNA expression (Provencal et al., 2012; Ramo-Fernández, Schneider, Wilker, & Kolassa, 2015). Functional neuroimaging studies reveal that irrespective of the task and stimulus formats, ToM is represented by cortical activity in the inferior and medial parts of the prefrontal cortex (mPFC) and temporoparietal junction (TPJ) (Schurz, Radua, Aichhorn, Richlan, & Perner, 2014). Particularly, synaptic plasticity within the mPFC may be sensitive to effects of early life experience (Chocyk, Majcher-Maślanka, Dudys, Przyborowska, & Weôdzony, 2013). Convergent with the development of social cognition, it has been demonstrated that both mPFC and TPJ have a similar protracted maturation pattern and continue to develop structurally across adolescence before relatively stabilizing in the early twenties (Mills, Lalonde, Clasen, Giedd, & Blakemore, 2014). Based on the proposed association between attachment styles and ToM, it is plausible to think that young individuals with different attachment styles also have different representations of ToM in terms of cortical activity in late adolescence. However, a literature survey fails to identify a functional neuroimaging study of ToM depending on the style of attachment.

Functional near infrared spectroscopy (fNIRS) is a non-invasive cortical imaging method which is particularly suitable for socio-emotional tasks. FNIRS is capable of measuring brain activity in a more naturalistic environment than is possible in fMRI experiments (Rolfe, 2000). Since fNIRS is relatively insensitive to motion artifacts, subjects can be examined in a natural sitting position, without any surrounding distraction (Takizawa et al., 2008). For example, fNIRS has been used in real-world situations (Dresler et al., 2009) such as face-to-face communication (Suda et al., 2010) and interpersonal cooperation (Cui, Bryant, & Reiss, 2012).

We aimed to examine the relationship between attachment style and mPFC and TPJ activity during a ToM task in healthy university students. We used the Reading the Mind from the Eyes Test (RMET) (Baron-Cohen, Wheelwright, Hill, Raste, & Plumb, 2001) as a reliable measure of ToM, and we preferred fNIRS to measure cortical activity in a setting close as possible to the real world. We hypothesized that securely and insecurely attached children have different activity within these cortical regions during ToM.

Materials and Method

The Study Samples

Of the total number of 494 first year medical students invited to take part in the study at the Ankara University School of Medicine in May 2016, 71 consented. Students with a current or prior history of psychiatric or neurological disorders, head trauma (resulting in loss of consciousness longer than 30 min), as well as alcohol and/or substance use disorder (except tobacco) were excluded (N = 15). Since the data acquisition was extended to April 2017, nine study participants had advanced to second year medical studies. The mean age of participants was 19 ± 1.6 years; 39 (69.6%) of the participants were female.

Ankara University Ethics Board Commission approved the study procedure (Approval ID: 24.03.2016/8/107). Written informed consent was obtained from each participant. Students were informed that participation was voluntary, refusal had no consequences, but also that they will not be rewarded any extra credit through participation.

Study Measures

Attachment Style

The Relationship Scales Questionnaire (RSQ) (Griffin & Bartholomew, 1994) was used to assess participants’ attachment styles. The RSQ consists of 30 items that are rated on a 7-point scale from 1 (= not at all like me) to 7 (= very like me). Participants rate the extent to which each statement best describes their general orientation in close relationships. The questionnaire allows discriminating between secure and insecure (ambivalent, dismissive, and avoidant) styles of attachment. It has been shown that RSQ is a valid and reliable tool to assess attachment styles in Turkish young adults (Sümer & Güngör, 1999).

The Neuroimaging Task

RMET (Baron-Cohen et al., 2001) was adopted for the fNIRS environment by the Matlab Psychophysiology Toolbox software as the cortical activation paradigm (Figure 1). The neuroimaging task consisted of two conditions, namely the ToM condition and the control condition. During the ToM condition the subjects were expected to guess the correct mental state expressed via the eye photographs (A blocks in Figure 1). The subjects, were allowed to respond to as many photographs as they could/or to pass to the next photograph during the 30-s intervals. There was no predetermined time limit per photograph. During the control condition the participants were presented the same eye photographs as the ToM condition, but this time, they were expected to guess whether the eyes presented were the eyes of a man or a woman during 30 s (B blocks in Figure 1). Again, they were allowed to respond to as many photographs as they could/or to pass to the next photograph during the 30-s intervals of the control task. Both the ToM and the control conditions were presented in four blocks.

Figure 1.

Reading the Mind from the Eyes Task as the cortical activation paradigm (A blocks represent ToM judgments, B blocks represent judgments that require gender identification).

The two task conditions, A and B were consequently presented and were preceded and followed by 45-s rest periods. Since the participants responded verbally to the task, they were asked to repeat Turkish vowels (/a/, /e/, /o/) during those rest periods in order to control the effect of articulation on cortical activity.

Functional Near Infrared Spectroscopy (fNIRS)

We used the 52 channel fNIRS device (ETG-4000; Hitachi Medical Co., Tokyo, Japan) located at Ankara University Brain Research Center fNIRS Laboratory to measure cortical activity during the neuroimaging task. The fNIRS instrument measures relative changes in oxygenated Δxy-Hb) and deoxygenated hemoglobin Δdeoxy-Hb) through optodes (emitters and detectors) of two wavelengths (695 and 830 nm) of infrared light, on the basis of the Beer–Lambert law. As a result, it is possible to noninvasively probe the human cerebral cortex using near infrared light and monitor the cerebral concentration of hemoglobin, which is the dominant near-infrared absorbing species in the brain (Boas, Dale, & Franceschini, 2004). The distance between emitter/detector pairs was set at 3.0 cm, and the channels were defined as the area between these pairs. The 3-cm space between the optodes allows the device to measure Δoxy-Hb and Δdeoxy-Hb at 2–3-cm depth from the scalp that corresponds to the surface of the cerebral cortex (Okada & Delpy, 2003; Toronov et al., 2001). The optodes were fixed to scalp via a thermoplastic 3 × 11 shell. On the left side, Channel 11 was positioned on F7 and on the right side, T4 was positioned in the middle of the line between Channels 8 and 9. The placement of the measurement channels according to the international 10–20 system used in electroencephalography is presented in Figure 2. In order to visualize cortical projections of the channels on the cortex, the 52 optode channels derived from 33 optodes in the 3 × 11 shell were projected on the rendered cortex (Figure 3). To achieve that, the 3D digitizer file was spatially registered with the standard Montreal Neurological Institute (MNI) template cortex using “NFRI_-functions” in the NIRS SPM toolbox (Singh, Okamoto, Dan, Jurcak, & Dan, 2005; Ye, Tak, Jang, Jung, & Jang, 2009).

Figure 2.

Position of the measurement channels on the scalp according to the international 10–20 electroencephalography system.

Figure 3.

Optode channel projections on the standard MNI cortex.

The fNIRS device measures relative changes in oxyhemoglobin (oxy-Hb) concentrations. Therefore, a baseline activity is needed. The activity during repetition of Turkish vowels was used as the baseline activity. The pre-task baseline was determined as the mean over a 9-s rest period just prior to the task period, and the post-task baseline was determined as the mean over the last 7 s of the post-task rest period; linear fitting was applied to the data between these two baselines. The time resolution of fNIRS was set at 0.1 s. The fluctuations of fNIRS signals were known to be related to physiological activities such as the systemic arterial pulse oscillations (0.1 Hz) and respiration (0.2–0.3 Hz) (Hoshi, 2003). Thus, moving average methods were applied to remove short-term motion artifacts and to correct such fluctuations in the analyzed data (moving average window: 5 s). A sharp signal change over 0.4 mM/L in over twenty successive samples was labeled as a body movement artifact by the fNIRS device. A researcher blind to the study groups re-examined these artifacts, in order to detect the individual channels responsible for those artifacts. Data from channels with body movement artifacts were removed from the analyses. Since oxy-Hb change is assumed to reflect cognitive activation more directly than deoxy-Hb change as shown by a stronger correlation with blood oxygenation level dependent signal measured by fMRI (Strangman, Culver, Thompson, & Boas, 2002), we focused on the mean change in oxy-Hb during the task periods relative to the pre and post-task baseline periods.

Statistical Analysis

Socio-demographic and clinical data were compared between the groups with the independent samples t-test and the Mann–Whitney Test where appropriate.

The number of correct responses during the ToM task (ToM score) and the control task (Gender identification score [GI score]) were compared between the two groups with the Mann–Whitney U-test.

ΔOxy-Hb measurements from 52 channels during the ToM and the control conditions were analyzed with 2 (Groups: Secure vs. Insecure) × 2 (Condition: ToM vs. Control) × 52 (Channels) mixed ANOVA design. Group was the between-subject independent variable; Condition and Channels were within-subject independent variables. To prevent Type-1 errors resulting from sphericity violations, Greenhouse–Geisser corrections (F-test); and to prevent Type-1 errors resulting from multiple comparisons in the post hoc tests, Bonferroni corrections (t-test) were applied.

We also performed Spearman correlation tests to examine the relationship between RSQ subscale (secure, dismissing, fearful, and preoccupied) scores with cortical activity during the ToM condition. False Discovery Rate method (Benjamini & Hochberg, 1995) was used to control type-1 errors that may result from multiple comparisons.

Results

The two-factor solution of the RSQ revealed that 18 (32%) students were securely attached (the secure group, SG hereafter) while 38 (68%) were insecurely attached (the insecure group, ISG hereafter). The SG and the ISG were not different in terms of gender: the male to female ratio was 4:14 for the SG; and 13:25 for the ISG groups, X² = 0.83, p = .36.

Comparison of the two groups for the ToM and GI scores is presented in Table 1. The groups were not different in terms of ToM and GI performance.

Table 1.

Comparison of ToM and GI scores between the securely and insecurely attached students

	SG (N = 18)	ISG (N = 38)
ToM score [Mdn (min – max)]	18 (8–26)	18 (10–30)	Z = −0.35, p = .73
GI score [Mdn (min – max)]	77 (49–97)	78 (50–97)	Z = −0.97, p = .33

Note. ToM = Theory of Mind, GI = Gender identification, SG = Secure Group, ISG = Insecure Group, Mann–Whitney U-test revealed that the groups display comparable performance in ToM as well as GI.

Mixed ANOVA revealed that “Group” main effect was not significant while main effect of “Channel” was significant, F(14, 733) = 2.91, p < .001, η²_p = 0.051. “Condition” main effect was also significant, F(1, 54) = 8.57, p = .005, η²_p = 0.14. Post hoc tests indicated that the cortical activity during the ToM condition was significantly higher than the control condition (Mdn = 0.015, SE = 0.005, p = .005). The channels with different activity between the ToM and the GI conditions in both groups are presented in Table 2.

Table 2.

Channels with different activity between the ToM and the GI conditions in both groups.

Channel	Hemisphere, Brodmann (Percent overlap)	MNI			Mean difference	Standard error	p
		X	Y	Z
Ch 7	R, BA 6, (0.72)	−16	65	27	0.037	0.014	.010
Ch 8	R, BA 43, (0.45)	70	−13	20	0.048	0.023	.042
Ch 16	R, BA 8, (1.00)	41	25	54	0.026	0.01	.011
Ch 18	R, BA 6, (0.64)	66	−14	40	0.076	0.017	.000
Ch 21	R, BA 37, (0.90)	58	−58	−30	0.037	0.018	.046
Ch 24	L, BA 8, (0.90)	−11	40	57	−0.036	0.017	.034
Ch 26	R, BA 6, (0.83)	28	20	64	−0.023	0.011	.044
Ch 29	R, BA 40, (0.33)	68	−26	42	0.042	0.016	.010
Ch 30	R, BA 22, (0.66)	69	−46	19	0.082	0.023	.001
Ch 31	R, BA 37, (0.75)	65	−59	−7	0.037	0.018	.046
Ch 32	L, BA 46, (0.84)	−53	33	24	0.073	0.023	.003
Ch 42	R, BA 19, (0.71)	56	−74	−5	0.069	0.018	.000
Ch 44	L, BA 8, (0.73)	−38	21	58	−0.039	0.013	.003
Ch 49	R, BA 3, (0.36)	45	−25	68	0.061	0.027	.029

Note. Ch = Channel, MNI = Montreal Neurological Institute, Bonferroni correction was applied for all comparisons. This table lists the channels that are differentially activated in the two task conditions in the whole study sample.

The “Channel × Group” interaction was not significant, on the other hand the “Group × Condition” interaction was significant, F(1, 54) = 5.86, p = .019, η²_p = 0.098. Post hoc tests revealed that only among the SG group, the cortical activity was higher during the ToM condition than the GI condition (Mdn = 0.027, SE = 0.008, p = .002).

The “Condition × Channel” interaction was also significant, F(15, 823) = 2.71, p < .001, η²_p = 0.051. Post hoc test results are presented in Table 2. All the channels where activity was significantly higher during the ToM condition than the GI condition (Channels 7, 8, 16, 18, 21, 29, 30, 31, 42, 49) were located in the right hemisphere (Figure 2).

Finally, “Group × Condition × Channel” interaction was significant, F(15, 823) = 2.07, p = .009, η²_p = 0.037. When we analyzed this triple interaction effect with respect to group membership, cortical activity in Channels 8, 9, 19, 20, 30, 41, 42 were higher among the SG than the ISG during the ToM condition (Table 3). Nevertheless, from the Condition perspective, the channels with higher activity during the ToM condition than the GI condition were channels 7, 8, 18, 29, 30, 31, 32, 41, 42, 49, 51 in the SG. On the other hand, activity only in channel 44 was higher during the GI condition than the ToM condition in the SG. Among the ISG, activity in channels 1, 11, 17, 21 were higher during the ToM condition than the GI condition, whereas activity in channels 4 and 26 were higher during the GI condition than the ToM condition (Table 4). Cortical activity in the two study groups during the ToM condition is presented in Figure 4.

Table 3.

“Channel × Condition × Group” interaction from the perspective of “Group” comparisons at the each level of “Channel” and “Condition”

Channel	Hemisphere, Brodmann (Percent overlap)	MNI			Condition	Group	Mean difference	Standard error	p
		X	Y	Z
Ch 8	R, BA 43, (0.45)	70	−13	20	ToM	SG > ISG	0.117	0.040	.004
Ch 9	R, BA 21, (0.88)	73	−25	−8	ToM	SG > ISG	0.110	0.054	.045
Ch 19	R, BA 42, (0.41)	71	−26	19	ToM	SG > ISG	0.135	0.047	.006
Ch 20	R, BA 21, (0.71)	71	−45	−6	ToM	SG > ISG	0.118	0.039	.004
Ch 30	R, BA 22, (0.66)	69	−46	19	ToM	SG > ISG	0.134	0.041	.002
Ch 41	R, BA 39, (0.39)	58	−25	56	ToM	SG > ISG	0.121	0.050	.020
Ch 42	R, BA 19, (0.71)	56	−74	−5	ToM	SG > ISG	0.151	0.044	.001
Ch 1	L, BA 10, (1.00)	−30	66	9	Control	SG > ISG	0.059	0.020	.005
Ch 18	R, BA 6, (0.64)	66	−14	40	Control	ISG > SG	−0.124	0.061	.045
Ch 44	L, BA 8, (0.73)	−38	21	58	Control	SG > ISG	0.102	0.040	.013

Note. Ch = Channel, MNI = Montreal Neurological Institute, ToM = Theory of Mind, SG = Securely Attached Group, ISG = Insecurely Attached Group. Bonferroni correction was applied for all comparisons. This table lists the channels that show activity difference between the two study groups.

Table 4.

“Channel × Group × Condition” Interaction from the Perspective of “Condition” Comparisons at the each level of “Channel” and “Group”

		MNI
Channel	Hemisphere, Brodmann (Percent overlap)	X	Y	Z	Condition	Group	Mean difference	Standard error	p
Ch 7	R, BA 6, (0.72)	−16	65	27	SG	ToM > Control	0.055	0.023	.021
Ch 8	R, BA 43, (0.45)	70	−13	20	SG	ToM > Control	0.094	0.038	.017
Ch 18	R, BA 6, (0.64)	66	−14	40	SG	ToM > Control	0.116	0.028	.000
Ch 29	R, BA 40, (0.33)	68	−26	42	SG	ToM > Control	0.079	0.026	.003
Ch 30	R, BA 22, (0.66)	69	−46	19	SG	ToM > Control	0.144	0.038	.000
Ch 31	R, BA 37, (0.75)	65	−59	−7	SG	ToM > Control	0.060	0.030	.048
Ch 32	L, BA 46, (0.84)	−53	33	24	SG	ToM > Control	0.097	0.038	.015
Ch 41	R, BA 39, (0,39)	58	−25	56	SG	ToM > Control	0.128	0.045	.007
Ch 42	R, BA 19, (0.71)	56	−74	−5	SG	ToM > Control	0.122	0.030	.000
Ch 49	R, BA 3, (0.36)	45	−25	68	SG	ToM > Control	0.121	0.045	.009
Ch 51	R, BA 40, (0.91)	60	−56	44	SG	ToM > Control	0.072	0.029	.017
Ch 44	L, BA 8, (0.73)	−38	21	58	SG	Control > ToM	−0.092	0.021	.000
Ch 1	L, BA 10, (1.00)	−30	66	9	ISG	ToM > Control	0.028	0.090	.002
Ch 11	L, BA 10, (0.87)	−42	57	8	ISG	ToM > Control	0.041	0.014	.007
Ch 17	R, BA 6, (0.55)	51	11	51	ISG	ToM > Control	0.030	0.015	.043
Ch 21	R, BA 37, (0.90)	58	−58	−30	ISG	ToM > Control	0.044	0.021	.004
Ch 4	R, BA 8, (0.51)	18	52	45	ISG	Control > ToM	−0.017	0.007	.025
Ch 26	R, BA 6, (0.83)	28	20	64	ISG	Control > ToM	−0.049	0.013	.000

Note. Ch = Channel, MNI = Montreal Neurological Institute, ToM = Theory of Mind, SG = Securely Atached Group, ISG = Insecurely Attached Group. Bonferroni correction was applied for all comparisons. This table lists the channels that show different activity in the two task conditions for each group.

Figure 4.

Cortical activity in the two study groups during the ToM condition. (A) Right-sided view of activity map in the SG, (B) Front view of the activity map in the SG, (C) Right-sided view of activity map in the ISG, (D) Front view of the activity map in the ISG.

We further performed Spearman correlation tests to examine any correlations between the RSQ scores and the cortical activity during the ToM condition in channels where group comparisons revealed significant differences. We found that the activities in Channel 20 (r = 0.333, p = .012) and Channel 30 (r = −0.351, p = .008) were negatively correlated with RSQ-dismissing scores. RSQ-secure, RSQ-fearful, and RSQ-preoccupied scores were not correlated with cortical activity.

Discussion

Our observation that the rate of insecurely attached students (68%) is higher than securely attached (36%) students is consistent with two other studies that involving Turkish university students. Using the same assessment tool, both Sümer and Güngör (1999) and Kiralp and Serin (2017) reported 36% secure attachment rate among Turkish university students. The higher rate of insecure attachment among the medical students may therefore also reflect a universal pattern of increasing insecure attachment styles reported by students, over time, in American colleges (Konrath, Chopik, Hsing, & O’Brien, 2014). A recent meta-analysis attributed this to the emergence, in particular, of the dismissive style of insecure attachment. Since medical schools in Turkey, and in Ankara University in particular, are highly competitive environments, students are admitted to study medicine from across the country following a rigorous national examination. Like similarly aged American college students, they begin their studies directly following high school and may share similarities with contemporary cohorts of college students exposed to group biases, behaviors, and competition, that may be common across different World regions (Choi, Nisbett, & Norenzayan, 1999).

Contrary to what we expected, ToM scores as ascertained by the RMET were comparable among the two groups. Theoretically, attachment styles are likely to have an enduring effect upon ToM, as the ability to process minds of others may depend on the quality of the previously established internal working models. Studies that were carried out in children support this view by demonstrating the importance of secure attachment in the acquisition of ToM (Fonagy, M. Steele, H. Steele, Moran, & Higgitt, 1991; Meins et al., 2002; Symons & Clark, 2000). Nevertheless, the relationship between attachment and ToM in adolescence has been investigated only in few studies. Humfress, O’Connor, Slaughter, Target, and Fonagy (2002) found that in early adolescence, secure attachment style is associated with a better performance in inferring mental states of others, as assessed by the Strange Stories test. In another study, Hünefeldt, Laghi, Ortu, & Belardinelli, 2013 found that attachment anxiety, but not avoidance of mother, was associated with less accurate mindreading in the RMET among 14–19-year-old adolescents (Hünefeldt et al., 2013). Taken together with the results of the present study, these findings suggest that the relationship between attachment and ToM as observed in childhood and early adolescence may diminish in late adolescence. The increasing importance of interpersonal relationships especially among peers (Bosco, Gabbatore, & Tirassa, 2014; Valle, Massaro, Castelli, & Marchetti, 2015) as well as the membership to social networks (Sebastian, 2015) may directly interfere with ToM ability within this period. Nevertheless, we found that attachment style does have an effect on representation of ToM in terms of cortical activity in this group of young adults.

The “Condition” main effect revealed by ANOVA suggests that irrespective of group membership and cortical topography, the task and the control conditions lead to different cortical activity. Supporting the validity of the imaging task, the ToM condition that presumably necessitates more cortical load was associated with higher activity than the GI condition. The “Channel” main effect was significant, thus the experimental paradigm produced different activity across the measurement channels. The “Condition × Channel” interaction was also significant suggesting that independent from group membership, the two conditions lead to different activity across the channels. This is also an expected finding and refers to validity of the design of the experimental paradigm. The post hoc analyses showed that the ToM condition was associated with higher activity than the control condition particularly in the right hemisphere including right inferior frontal, bilateral middle frontal, inferior and superior temporal, and the adjacent parietal cortices. Schurz et al. (2014) performed a meta-analysis of fMRI studies during the RMET and examined activity contrasts for mental state judgments versus physical judgments (gender and age). The authors found that the activity map involves left inferior frontal area, bilateral posterior temporal cortices, middle frontal gyri, bilateral TPJ-posterior, and right TPJ-anterior areas. Although the activated areas are in similar regions, in the current study we found that the activity in the inferior frontal cortex was predominantly in the right hemisphere. Contrary to our findings, Sato et al. (2016) analyzed structural magnetic resonance images using voxel-based morphometry in healthy participants and concluded that the ability to read the mind in the eyes relies more on representations in the left, than the right, fronto-temporo-parietal regions. This discrepancy may be explained by methodological differences among the studies, that is, hemodynamic response versus structural alterations. In line with this view, the effect sizes were greater in the right than the left hemisphere for the parietal and the posterior temporal regions of interest in the meta-analysis done by Schurz and colleagues (2014).

The significant “Group × Condition” interaction suggests that independent from cortical topography, the ToM and the GI conditions lead to different cortical activity in the two study groups. Post hoc analyses showed that this interaction stems from the higher activity during the ToM condition relative to the control condition in the SG, whereas the overall cortical activity evoked by the two conditions was indistinguishable in the ISG. This finding supports the hypothesis that attachment style has an effect on cortical activity during our experimental paradigm. To our knowledge, there have been two other studies that have examined the effect of attachment related factors on brain activity. Similar to the present study, Vrticka and Vuilleumier (2012) showed that attachment style regulates activity during social emotion perception and regulation. The authors found that avoidant attachment is associated with increased prefrontal and anterior cingulate activation to social negative scenes when making spontaneous emotion judgments. In another study, DeWall and colleagues (2012) showed that neural responses to social rejection differ according to attachment anxiety and avoidance. Therefore, attachment seems to play a role in individual differences in brain activity particularly in socio-emotional tasks.

The significant “Group × Condition × Channel” interaction suggests that securely and insecurely attached participants show different cortical activity across the measurement channels in the two task conditions. From group membership perspective, this significance stems from higher activity in channels corresponding to right temporo-parietal junction in the SG relative to the ISG during the ToM condition (Table 3).

From the condition perspective, the activity difference between the ToM and the control conditions was observable in a larger cortical area including right inferior frontal, precentral, right superior temporal and inferior parietal in the SG, and a smaller cortical area corresponding to left inferior frontal and right inferior temporal areas among the ISG. This underscores that the ToM is a higher order social cognitive function. The control condition task we employed is a gender identification task which also involves a simple social judgment. Therefore, the post hoc results from the condition perspective may imply that the secure compared to insecure attachment is associated with a more profound and/or differentiated cortical representation of complex versus simple social judgments.

From group membership perspective, secure relative to insecure attachment was associated with higher cortical activity in the right posterior superior temporal and the adjacent inferior parietal cortices during the ToM task. These areas were previously associated with mentalizing and intention attribution (Blakemore, Sarfati, Bazin, & Decety, 2003). Experience related cortical plasticity via genetic and epigenetic mechanisms may be involved in this activity difference between the two study groups. For example, oxytocin receptor gene promoter hypermethylation was linked to greater activity in the superior temporal gyrus and supramarginal gyrus at the temporo-parietal junction during a social perception task (Jack, Connelly, & Morris, 2012).

We also found that dismissive attachment scores in RSQ were negatively correlated with activity in channels 20 and 30 that correspond to right superior temporal cortex. DeWall and colleagues (2012) found that avoidant attachment is associated with lower activity in dorsal anterior cingulate cortex and insula during social rejection evoked by a cyber-ball paradigm. Similarly, Suslow et al. (2009) showed that avoidant attached participants display lower activity in the somatosensory cortex when they see sad faces. Together with our findings these results may imply that avoidant/dismissive attachment may be represented by lower activity in correspondent brain regions to socio-emotional stimuli. Theoretically these findings may be related to weaker social need and habitual unwillingness for closeness (Vrtička & Vuilleumier, 2012).

An important limitation of the study is that it involves a group of highly capable medical students who represent top percentile of success at university entrance and academic success. The results therefore cannot be generalized to a general or adolescent population. Secondly, The ToM and the control conditions were consequently presented. This kind of presentation may be associated with habituation of brain activity. We therefore made additional paired comparisons of cortical activity between the first and the last epoch of each block in channels that show significant activity change in the ToM condition with the paired samples t-test. None of those comparisons revealed significant results minimizing the possibility of habituation. Finally, ToM is represented by a broader network including some brain regions that were not investigated in this study, such as the precuneus or the orbitofrontal cortex (Schurz et al., 2014), and the activity difference between the study groups may indeed be higher in these areas. Nevertheless, this study is the first to show that the previously suggested relationship between ToM and attachment style may also be evident in terms of cortical activity.