Changes of brain activity in the neural substrates for theory of mind during childhood and adolescence

Abstract

Abstract  Theory of mind (ToM) refers to the ability to attribute independent mental states, such as beliefs, preferences and desires, to the self and others. Neuroimaging studies of normal adults have consistently demonstrated the importance of particular brain regions for ToM, the superior temporal sulcus (STS), temporal pole (TP) and the medial prefrontal cortex (MPFC). However, there are little data showing how ToM develops during childhood and adolescence. Such data are important for understanding the development of social functioning and its disorders. The authors used functional magnetic resonance imaging to study age-related changes in brain activity associated with ToM during childhood and early adolescence (9–16 years). Normally developed children and adolescents demonstrated significant activation in the bilateral STS, the TP adjacent to the amygdala (TP/Amy) and the MPFC. Furthermore, the authors found a positive correlation between age and the degree of activation in the dorsal part of the MPFC; in contrast, a negative correlation was found for the ventral part of the MPFC. The authors also found a positive correlation between the Z coordinate of the peak activation in the MPFC and age. The data indicated that activity in the MPFC associated with ToM shifted from the ventral to the dorsal part of the MPFC during late childhood and adolescence. No age-related changes were found in the STS and the TP/Amy regions. The authors consider that the age-related brain activity observed in the present study may be associated with the maturation of the prefrontal cortex and the associated development of cognitive functions.

INTRODUCTION

Human beings develop the ability to attribute independent mental states, such as beliefs, preferences, and desires, to the self and others in order to explain and predict the behavior of others. This is referred to as the theory of mind (ToM).¹ ToM has been considered to be one of the essential independent cognitive domains for social interaction and cognition. Its impairment has been considered to be associated with deficits in social interaction and social functioning, such as found in the autistic spectrum disorders and schizophrenia.^2,3 Neuroimaging studies have identified the neural substrates for ToM in the medial prefrontal cortex (MPFC), superior temporal sulcus (STS) and the temporal pole adjacent to the amygdala (TP/Amy) in normal adult individuals⁴ and in normal children.⁵ Furthermore, some studies demonstrated abnormal neural activitiesin these areas in individuals with autistic disorders showing impairments of ToM.^6–8 Particularly, the MPFC has been hypothesized to play a crucial role for ToM.^9–19 According to Frith's hypothesis,⁴ the MPFC may be the site of the decoupling mechanism that distinguishes the representation of one's own mental state from the mental states of others and the reality of the physical world. Therefore, the MPFC seems to be responsible for the higher-order processing of ToM.

Although most neuroimaging studies have been conducted with adult subjects, developmental psychologists who focus on the origins and early expressions of ToM have studied infants and young children. Behavioral studies revealed that most children are able to pass the first-order false belief test in their fourth year,^20–22 However, there are little data showing how ToM develops beyond the age of about 7 years, when children have acquired understanding of second-order false beliefs.²³ For example, it is not known whether the social sophistication of children older than 7 years of age is related to the acquisition of general knowledge and the maturation of general cognitive components or reflects the further development of ToM. Particularly, there is no data about how ToM develops during adolescence. Adolescence is a time of profound mental change affecting social awareness and adaptation. Several forms of mental illness, such as schizophrenia, begin during adolescence.²⁴ Therefore, it is important to understand the cognitive and neural maturation during this period.

A recent electrophysiological study demonstrated that the event-related potentials associated with the frontal cortex are still maturing into late adolescence, and that their amplitudes have significant correlations with behavioral capacities.²⁵ Furthermore, several neuroimaging studies have focused on the process of brain maturation during childhood and adolescence.^26–29 These studies have consistently suggested that maturation of the prefrontal cortex occurs during childhood, adolescence, and even during postadolescence.²⁸ Considering these developmental changes of the prefrontal lobe and its functions, the cognitive abilities that rely on the frontal lobes such as ToM may also change during this period.

In a previous functional magnetic resonance imaging (fMRI) study, the authors found that normally developed children (mean age, 10 years) had fully developed neural substrates for ToM.⁵ To clarify the developmental changes of the neural substrates for ToM, the authors studied the age-related changes of brain activity during a ToM task in late childhood and early adolescence. The authors hypothesized that neural activities associated with ToM may change during late childhood and early adolescence, particularly in the MPFC.

METHOD

Subjects

A total of 16 right-handed, healthy children and adolescents (12 males and four females; mean (standard deviation [SD]) age, 13.4 (2.2) years; range, 9–16) participated in this study. All subjects were interviewed by medical doctors and confirmed to have no major medical or neurological or psychiatric disorders and no history of psychiatric and mental disorders including mental retardation. They all passed first and second-order false belief tests^23,30–33 and the normal development of the ToM ability of each subject was confirmed. Written informed consent was obtained from all subjects and their guardians in accord with the ethical guidelines of the local ethical committee.

Stimuli for functional magnetic resonance imaging measurements

The authors used animations as the stimuli for fMRI measurements. The movie animation stimuli were essentially the same as the stimuli that were originally developed and used in a behavioral study by Abell et al.³⁴ and subsequently were used in previous positron emission tomography (PET) studies conducted by Castelli et al.^6,18 (Fig. 1 and the materials on the website http://www.icn.ucl.ac.uk/dev_group/research.htm). Eight silent animations were used during the fMRI scanning. All the animations featured two characters, a big red triangle and a small blue triangle, moving on a framed white background. Each sequence lasted 20 s, modified for the fMRI measurement. Two types of animations (four ToM animations and four random movement animations as a control condition) were matched for length. The scripts for the ToM sequences involved the two triangles persuading, bluffing, mocking, and surprising one another. The random movement animations showed the two triangles bouncing off the walls resembling the movement of billiard balls, or merely drifting about. While the type of movement was by definition different between the two conditions, the basic visual characteristics in terms of shape, overall speed, and orientation changes were as similar as possible.

Figure 1. Example of theory of mind (ToM) task and control animation. Eight silent animations were used during functional magnetic resonance imaging. All the animations featured two characters, a big red triangle and a small blue triangle, moving on a framed white background. Each sequence lasted 20 s, modified for the functional magnetic resonance imaging measurement. Two types of animations (four ToM animations and four random movement animations as a control condition) were matched for length. The scripts for the ToM sequences involved the two triangles persuading, bluffing, mocking, and surprising one another. The random movement animations showed the two triangles bouncing off the walls resembling the movement of billiard balls, or merely drifting about. While the type of movement was by definition different between the two conditions, the basic visual characteristics in terms of shape, overall speed, and orientation changes were as similar as possible.

Magnetic resonance imaging measurements

All MRI measurements were performed with a 1.5 T Magnetom Vision plus MRI scanner (Siemens, Erlangen, Germany) using the standard head coil. Head motion was minimized by placing tight but comfortable foam padding around the subject's head. Cerebral activation was measured with fMRI using blood oxygen level-dependent contrast.^35,36 After automatic shimming, a time course series of 125 volumes was obtained using single-shot gradient-refocused echoplanar imaging (TR 4000 ms; TE 60 ms; flip angle 90 degree; interscan interval 4 s; in-plane resolution 3.44 × 3.44 mm; FOV 22 cm; contiguous 4 mm slices to cover the entire brain).

The fMRI protocol was a block design with two kinds of epochs of the ToM animation condition and the control (random movement) condition. Each epoch lasted 20 s (equivalent to five whole-brain fMRI volume acquisitions) and consisted of one script. Each script of task and control conditions was used twice in one session, so there were eight task and eight control blocks. The first five volumes of each fMRI scan were discarded because of non-steady magnetization and the remaining 120 volumes were used for the analysis. Before each fMRI experiment, subjects were told to watch the animations and think about what the triangles were doing and thinking during the scanning. After each scan, the subjects again watched the same stimuli on a computer screen outside the scanner and were asked to recall what they had thought, during the fMRI scan, that the triangles had been doing. To check that the subject had been engaged well in the task, the subjects were asked to tell the experimenter what they had thought, in response to the experimenter's neutral question: ‘What was happening in this animation?’ Their verbal descriptions were coded on four different dimensions: ‘Intentionality’, ‘Appropriateness’, ‘Certainty’, and ‘Length’. Detailed scoring procedures are given in the report by Castelli et al.^6,18

The stimuli were presented using a Windows PC and back projected onto a screen, ∼50 cm from the subject's head, using a 65 536-color liquid crystal display and an overhead projector. The subjects viewed the screen through a mirror attached to the head coil. A 3-D volumetric acquisition of a T1-weighted gradient echo sequence produced a gapless series of thin sagittal sections using a magnetization prepared rapid acquisition with gradient echo sequence (MPRAGE; TE/TR, 4.4/11.4 ms; flip angle, 15 degree; acquisition matrix, 256 × 256; 1NEX field of view, 31.5 cm; slice thickness, 1.23 mm).

Data analysis

Data were analyzed with the Statistical Parametric Mapping software (SPM2, http://www.fil.ion.ucl.ac.uk/spm). Scans were realigned and EPI BOLD images were summed and coregistered to the subject's T1-weighted MRI. Then the T1-weighted MRI were transformed to the standard stereotactic space of the MNI (Montreal Neurological Institute) using a T1-weighted MRI template whose space can convert to the Talairach space.³⁷ The parameter for affine and quadratic transformation to the T1-weighted MRI template that was already fitted for the Talairach space was estimated by least-squares means. This transformation was applied to the coregistered EPI BOLD images. Data were then smoothed in a spatial domain (full width at half-maximum, 8 × 8 × 8 mm) to improve the signal to noise ratio. After specifying the appropriate design matrix, the hemodynamic response function as a reference waveform, the condition and slow hemodynamic fluctuations unrelated to the task were estimated according to a general linear model taking temporal smoothness into account. Global normalization was performed using proportional scaling. To test the hypotheses about regionally specific condition effects, the estimates were compared by means of linear contrasts for each epoch. The resulting set of voxel values for each contrast constituted a statistical parametric map of the t statistic SPM {t}. Anatomic localization was according to both MNI coordinates and Talairach coordinates, obtained from M. Brett's transformations (http://www.mrc-cbu.cam.ac.uk/Imaging/mnispace.html) and presented as Talairach coordinates. To account for interindividual variance, all group analyses were computed using a random-effects model.³⁸ Group analysis across subjects involved a one-sample t-test on the whole brain images generated by pooling over the session the individual contrasts of task versus control activation for each subject. The significant voxels and clusters were given a threshold at P < 0.001 and the extent of cluster >50 voxels. The authors also applied hypothesis-driven region of interest (ROI) analyses to test age-related changes of degree of activation in the neural substrates for ToM (the STS, Temporal pole, and the MPFC). For these ROI analyses, the authors used the Wake Forest University PickAtlas³⁹ with a correlation analysis between the degree of activation associated with ToM and age. In the correlation analysis, the authors treated age as a covariate of interest and treated gender as a nuisance variable. The significant voxels and clusters were given a threshold at P < 0.005 without a correction for multiple comparisons.

RESULTS

Behavioral scoring of theory of mind task and age

There is no correlation between age and scoring of verbal descriptions of each dimension (‘Intentionality’; r = −0.22, P = 0.41, ‘Appropriateness’; r = −0.03, P =0.91, ‘Certainty’; r = 0.05, P = 0.85, and ‘Length’; r =−0.04, P = 0.89).

Brain activity associated with theory of mind in late childhood and adolescence

Table 1 and Fig. 2 show activities associated with ToM for all subjects as revealed by a one-sample t-test. A significant activation was noted in the bilateral (right > left) STS, the TP/Amy region, and the MPFC. In addition, the right inferior frontal gyrus [Brodmann area (BA) 45], middle frontal gyrus (premotor area, BA6/8), left posterior cerebellum, inferior/middle occipital lingual gyrus (BA18), right caudate and putamen, right precuneus (BA7/31), bilateral fusiform gyri (BA37) and left inferior/middle temporal gyrus (BA21) were also significantly activated.

Table 1.  Brain activation in response to theory of mind-related task compared to control condition by one sample t-test of all subjects throughout whole brain

	Region	BA	Talairach coordinate			Z	Cluster size (voxels)
			x	y	z
Height threshold: T = 3.73, P = 0.001; extent threshold: k = 50 voxels. BA, Brodmann area.
Rt	Superior Temporal Sulcus	22	51	−47	1	5.65	5461
Lt	Superior Temporal Sulcus	22	−51	−48	12	5.48	1965
Rt	Temporal Pole/Uncus	28/38	32	5	−24	4.35	186
Rt	Medial Prefrontal cortex	10	14	63	13	4.17	81
Rt	Medial Prefrontal cortex	9	4	50	29	3.67	72
Lt	Inferior/Middle Temporal Gyrus	21	−55	−14	−16	3.65	82
Lt	Cerebellum Posterior Lobe Tuber	–	−24	−81	−28	4.71	426
Rt	Inferior Frontal Gyrus	45	50	31	4	7.03	798
Rt	Middle Frontal Gyrus (premotor)	6/8	50	10	42	4.31	1214
Lt	Middle Frontal Gyrus (premotor)	6/9	−42	7	22	3.97	120
Rt	Medial Prefrontal Gyrus	8	6	35	42	4.33	50
Lt	Inferior/Middle Occipital, Lingual Gyrus	18	−34	−90	−6	4.32	159
Rt	Precuneus	7/31	10	−51	38	3.82	56
Rt	Caudate body	–	14	12	10	4.1	131
Rt	Fusiform Gyrus	37	−42	−51	−16	4.51	1965
Lt	Fusiform Gyrus	37	44	−51	−16	4.96	5461
Rt	Lentiform Nucleus, Putamen	–	16	11	−4	4.06	131

Figure 2. Activation associated with theory of mind in normal children and adolescents. Statistical parametric mapping {t} is rendered onto a normalized template brain image. Amy, amygdala; MPFC, medial prefrontal cortex; PM, premotor area; STS, superior temporal sulcus; TP, temporal pole.

Correlation analyses

In the correlation analysis for age-associated changes of activities in the MPFC, the STS, and the TP/Amy region, significant correlations between age and the degree of activation were found only in the MPFC. A significant positive correlation with age was noted in the dorsal part of the MPFC (Talairach coordinate: −6, 57, 14; Fig. 3a), whereas a significant negative correlation with age was noted in the more ventral part of the MPFC (Talairach coordinate: 10, 43, 0; Fig. 3b). The ventral medial prefrontal cortex was activated in a child group in a fixed first-level analysis (n = 6, ≤12 years of age; Talairach coordinate: 12, 44, −4; T = 3.81, Z = 3.80, cluster extent k = 51: height and extent threshold; P = 0.001 uncorrected and k = 20). No significant correlation was found between age and the degree of activation in the STS and the TP/Amy. It was supposed that the activation in the MPFC may shift from a ventral to a dorsal part of the MPFC with aging or maturation. Further, the authors added an evaluation of the correlation between the Y and Z coordinate of the peak activation in the MPFC and age in each subject. To reduce the false negatives, the locus of the maximum MPFC peak in each subject was searched at a lenient height and extent threshold (P < 0.05 uncorrected and P < 0.05 volume corrected within the ROI defined by medial part of Brodmann area 9, 10, 11, and 32 using the Wake Forest University PickAtlas software) at the fixed level subject-wise analysis. There was a significantly positive correlation between the Z coordinates and age [(Z coordinate) = 6.78 × (age) − 73.8, r = 0.78, P = 0.00037); Fig. 4], although there is not a significant correlation between the Y coordinates and age (r = 0.33 P = 0.29).

Figure 3. Results of correlation analyses. Statistical parametric mapping {t} is rendered onto T1-weighted magnetic resonance images. (a) A significant positive correlation between age and degree of activation in the medial prefrontal cortex (MPFC) was noted in the dorsal part of the MPFC (Talairach coordinate: −6, 57, 14; t = 4.34; r = 0.72). (b) A significant negative correlation between age and degree of activation in the MPFC was noted in the ventral part of the MPFC (Talairach coordinate: 10, 43, 0; t = 5.49; r = 0.82).

Figure 4. The graph shows the correlation between age and Z coordinates in the Talairach space for each peak activation in the medial prefrontal cortex. A significant positive correlation was noted. [r = 0.78; P = 0.00037; regression line: (Z coordinate) = 6.78 × (age) − 73.8].

There is no significant correlation between ToM-related brain activity and scoring of verbal descriptions of each dimension.

DISCUSSION

In this study, the authors focused on the developmental aspects of the neural substrates for ToM. As the authors demonstrated in a previous fMRI study with normal children,⁵ they found activation in the three major loci: MPFC, STS, and TP/Amy regions. These regions have been consistently shown to be important neural substrates for ToM in previous studies with adult subjects.⁴ In addition, the authors also found ToM-related activation in the cerebellum, dorsal premotor area and the fusiform gyri. Some neuroimaging studies with adults also demonstrated that these regions may be associated with ToM and social cognition.^40–42 The current data indicate that children and adolescents have essentially the same neural substrates for ToM as adults.

Notably, the authors found age-associated changes in the activation pattern in the MPFC, which is a crucial neural substrate for ToM. Activities in other important neural components for ToM, such as the STS and the TP/Amy region, did not show any age-associated changes.

Regarding activity in the MPFC, a significant positive correlation between age and the degree of activation was noted in more dorsal part of the MPFC (BA9/10). This location (Talaraich coordinate (mm); x = −6, y = 57, z = (14)) is similar to that of a previous PET study of ToM with adult subjects (BA9; x = 10, y = 54, z = 30).⁶ In contrast, a negative correlation was found in more ventral part of the MPFC (BA32/10; x = 10, y = 43, z = 0). Furthermore, the Z coordinates of the individual activation peak in the MPFC positively correlated with age. These results indicate that the MPFC activation associated with ToM shifted with age from the ventral to the dorsal portion of the MPFC.

Different regions of MPFC play different roles in thinking about intentions. A neuroimaging study revealed holding in mind an intention to act and at the same time thinking about an intentional action led to reduced activity in a dorsal section of the MPFC, which was a different region from a more anterior, inferior dorsal MPFC region that responded to intentional causality.⁴³ A recent meta-analysis demonstrated that the medial prefrontal functions were segregated into cognitive and emotional functions.⁴⁴ The ventral MPFC was recruited when tasks had emotional contexts, whereas the dorsal MPFC was recruited when tasks required primarily cognitive demands. Furthermore, in previous reports, different loci of the MPFC were engaged by different kinds of mentalizing tasks.⁴⁵ The dorsal MPFC loci reported in several investigations were related to mentalizing trials with more social cognitive demands.^15,19,46,47 Otherwise the ventral MPFC loci reported in previous studies were related to mentalizing with inner self-referential processing.^48–54 Mitchell et al.⁴⁵ hypothesized that the ventral MPFC may guide the understanding of others' mental states through examining the individual's own mental state or simulating in the self the possible mental state of others. Otherwise the dorsal MPFC may instead represent more universally applicable social-cognitive processes that can aid mentalizing even when simulation is inappropriate (e.g. for dissimilar others). Taken together, the shifting of MPFC activity from the ventral to the dorsal portion could be interpreted as the maturation of the ToM system, which is changing from using only the developmentally earlier simulating system to using additional cognitive comprehension of others. The authors speculate that the older subjects in the present study try to ‘read the minds’ of the animated triangles in a more cognitive way than the younger subjects. The authors assume that the anatomical and functional maturation of the prefrontal cortex from late childhood to early adolescence may underlie developmental change of the ToM system. Previous neuroimaging studies have demonstrated heterochronic development in the human cerebral cortex and revealed late maturation of the prefrontal and temporal association cortices.^26–29 One study demonstrated that the maturation of the PFC continued even during postadolescence.²⁸ Also, behavioral studies have revealed age-related improvement in performance in frontal functions, such as the executive function, from childhood to late adolescence.⁵⁵ Further, a recent study⁵⁶ found a marked developmental shift from a predominantly negative correlation between intelligence and cortical thickness in early childhood to a positive correlation in late childhood and beyond, and this trajectory of change in the thickness of the cerebral cortex was found most primarily in frontal regions including MPFC. These results support the authors' notion that the changes in the MPFC activity in the present study seem to be developmental changes of the neural substrates for ToM and the underlying maturation of the PFC. However, it is still inconclusive whether the more dorsal region activated in the older children is functionally homologous to the more ventral region activated in the younger children. Further studies will be needed to look at the relationship between anatomical change in the mentalizing network and actual mentalizing behavior.

In contrast, the activity in the STS and the TP/Amy regions did not show age-related changes. The authors consider that the functions of these areas for ToM may have matured before late childhood. A lesion study of early and late damage to the human amygdala demonstrated that subjects with lesions of the amygdala arising in early childhood showed impaired ToM ability, whereas subjects who acquired damage to the amygdala in adulthood did not.⁵⁷ The posterior STS is a heteromodal association (along with prefrontal and inferior parietal cortices) and is involved with integration of memory, audiovisual association, and object-recognition functions.^58–60 Important functions in the posterior STS involve social perception using visual information, for example biological motion,^61–63 gaze direction,^64–67 and the mirror neuron system.⁵ All of these were found even in monkeys while monkeys are inferred not to have entire ToM ability, and have been regarded as more basic precursors for ToM.⁴ Behaviorally, before the age of four when most children can pass the first-order false belief test, children have acquired these precursors for ToM, such as detection of gaze direction⁶⁸ and joint attention,⁶⁹ which are associated with the STS⁴. Therefore, these facts suggest that the functional maturation of the STS as a part of ToM may have been almost completed before the age of nine. Perhaps this may have accounted for the lack of age-related changes in the STS in the present study.

In the present study, the authors could not find age-related change of ToM scorings and brain activities. One possible explanation is that ToM ability might be more associated with subject's personal achievement of mentalizing ability in individual cases than age-associated developmental effects. Another explanation is that subjects were equally engaged in inferring mental states in the task movies, regardless of its accuracy, which possibly resulted in no association between ToM scorings and brain activities. In future, longitudinal study of ToM-related behavioral performance and brain activity will be needed to clarify more detailed developmental factors of ToM.

A potential weakness of the study is the usage of an adult brain template for normalization of the EPI images, which might affect the correlation between ages and peak Z coordinates in the MPFC activations. Development and usage of a child brain template would be needed to confirm the result for the future.

In conclusion, in late childhood and early adolescence, normal individuals have essentially the same pattern of brain activities associated with ToM as observed in previous studies with normal adults. Importantly, the authors found age-related changes of activities in the MPFC that may be associated with the functional development of the prefrontal cortex. Although a detailed behavioral study is needed, the present data suggest that the neural system for ToM may continue to develop beyond the age of 7 years, when most children can understand second-order false beliefs.

ACKNOWLEDGMENTS

This research was supported by a Health Science Research Grant from the Ministry of Health, Labour and Welfare (H17-kokoro-007). The authors wish to thank Professor Uta Frith, Institute of Cognitive Neuroscience and Department of Psychology, University College London, for providing animations as fMRI stimuli.