Mind Reading: Neural Mechanisms of Theory of Mind and ... - CiteSeerX

NeuroImage 14, 170 –181 (2001) doi:10.1006/nimg.2001.0789, available online at http://www.idealibrary.com on

Mind Reading: Neural Mechanisms of Theory of Mind and Self-Perspective K. Vogeley,* ,§ P. Bussfeld,* A. Newen,† S. Herrmann,* F. Happe´,‡ P. Falkai,* W. Maier,* N. J. Shah,§ G. R. Fink,§ ,¶ and K. Zilles,§ ,㛳 *Department of Psychiatry and †Department of Philosophy, University of Bonn, Germany; ‡Institute of Psychiatry, King’s College, London, United Kingdom; §Institute of Medicine, Research Center Juelich, Germany; ¶Department of Neurology, RWTH Aachen, Germany; and 㛳C. & O. Vogt Brain Research Institute, University of Duesseldorf, Germany Received August 16, 2000; published online May 11, 2001

Human self-consciousness as the metarepresentation of ones own mental states and the so-called theory of mind (TOM) capacity, which requires the ability to model the mental states of others, are closely related higher cognitive functions. We address here the issue of whether taking the self-perspective (SELF) or modeling the mind of someone else (TOM) employ the same or differential neural mechanisms. A TOM paradigm was used and extended to include stimulus material that involved TOM and SELF capacities in a two-way factorial design. A behavioral study in 42 healthy volunteers showed that TOM and SELF induced differential states of mind: subjects assigned correctly first or third person pronouns when providing responses to the stimuli. Following the behavioral study, we used functional magnetic resonance imaging (fMRI) in eight healthy, right-handed males to study the common and differential neural mechanisms underlying TOM and SELF. The main factor TOM led to increased neural activity in the anterior cingulate cortex and left temporopolar cortex. The main factor SELF led to increased neural activity in the right temporoparietal junction and in the anterior cingulate cortex. A significant interaction of both factors TOM and SELF was observed in the right prefrontal cortex. These divergent neural activations in response to TOM and SELF suggest that these important differential mental capacities of human self-consciousness are implemented at least in part in distinct brain regions. © 2001 Academic Press

Key Words: theory of mind; self-perspective; simulation theory; theory theory; social communication; fMRI.

INTRODUCTION Self-consciousness includes consciousness of ones own mental states, such as perceptions, attitudes, opinions, intentions to act, and so forth. Representing such mental states into one combined framework that 1053-8119/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.

allows us to maintain the integrity of our own mind is a metarepresentational cognitive capacity, for which the ability to develop and apply a “self-perspective” (SELF) is essential (Vogeley et al., 1999). SELF in this context refers to the subjective experiential multidimensional space centered around ones own person. In this basic sense, SELF is a constituent of a “minimal self“ defined as “consciousness of oneself as an immediate subject of experience, unextended in time” (Gallagher, 2000). The correct assignment and involvement of SELF is reflected by the use of personal pronouns (“I,” “my,” e.g., perception, opinion, and so forth). Closely related to the ability to assign and maintain SELF is the capacity to attribute opinions, perceptions or attitudes to others, often referred to as “mindreading” (Baron-Cohen, 1995). The latter ability is an essential social skill as it plays a crucial role in interindividual communication. The ability to read another person’s mind can be reliably assessed in so-called “theory of mind” (TOM) paradigms (Fletcher et al., 1995), in which mental states or propositional attitudes of an agent with regard to a particular set of information or propositions need to be modeled (e.g., “Person A knows, believes, etc., that p”). SELF in this context refers to the special situation, in which oneself (“I”) is the agent (e.g., “I know, believe, that p”). When Premack and Woodruff (1978) introduced the concept of “theory of mind” (TOM), it referred to the attribution of mental states to both oneself and others. However, to date, it remains a theoretically and empirically open issue, as to what extent taking SELF is involved in modeling someone else’s state of mind. In the course of this debate two different concepts, based on developmental psychology and cognitive science have been proposed and are usually referred to as “simulation theory” and “theory theory.” According to “simulation theory” (ST), TOM capacity is based on taking someone else’s perspective, and projecting ones own attitudes on someone else (Harris, 1992). Thus, the capacity to develop SELF would be

170

171

SELF PERSPECTIVE AND THEORY OF MIND

reduced to a subcomponent of a more general TOM capacity. Both capacities would then be functionally closely related and should employ the same neural mechanisms. By contrast, according to “theory theory” (TT), TOM capacity is based on a distinct body of theoretical knowledge acquired during the individual ontogenic development (Gopnik and Wellman, 1992; Perner and Howes, 1992), which is different from acquiring a concept of a “SELF.” On a purely behavioral level, an independent cerebral implementation of the two capacities could only be inferred on the basis of a double dissociation. Arguments based on information of simultaneous or subsequent development of the two differential cognitive capacities are nonconclusive with regard to their putative differential cerebral implementation which is reflected by the current controversial debate (for more detail see, e.g., Gopnik and Wellman, 1992; Gopnik, 1993; and Carruthers, 1996). Neurophysiological evidence relevant to this debate was recently provided by Gallese et al. (1996) who demonstrated a “mirror neuron system” in macaques, which matches observation and execution of goal-related motor actions in inferior area 6 (corresponding to F5). Interestingly, the neurons in this area respond to both observation of a goal-directed action performed by other animals and execution of the same movement. More recent studies have provided strong evidence for the existence of an equivalent “mirror neuron system” in man (Buccino et al., 2001; Iacoboni et al., 1999). It must be emphasized, however, that studies concerned with the “mirror neuron system” do not address the issue whether or not there is at least in part an independent cerebral implementation of the two distinct mental capacities of SELF and TOM. The question whether taking someone else’s perspective (TOM) as opposed to taking ones own perspective (SELF) rely on overlapping or differential brain mechanisms, is directly amenable to functional imaging of neural activations in normal volunteers. We accordingly designed a functional magnetic resonance imaging (fMRI) study to explore this issue. To do so we used a well-characterized collection of short stories (Fletcher et al., 1995; Happe´ et al., 1996, 1999; Gallagher et al., 2000), which comprised “unlinked sentences” (high-level baseline as a control condition, hereafter: baseline), “physical stories,” “TOM stories,” and two newly developed groups of stories introduced to allow subjects to engage SELF with and without engaging TOM at the same time. This enabled us to employ a fully factorial design with two factors: TOM (factor TOM; present ⫽ ⫹/not present ⫽ ⫺) and SELF (factor SELF; present ⫽ ⫹/not present ⫽ ⫺). A behavioral pilot study prior to fMRI scanning demonstrated that a different group of subjects successfully assigned SELF and TOM associated with the stories. The design of the fMRI experiment then addressed the issue of,

whether taking SELF or modeling the mind of someone else (TOM) employ the same or differential neural mechanisms. Differential activations observed can then be attributed to the distinct cognitive operations performed in SELF and TOM conditions. Thus, our fMRI study constitutes the first explicit exploration of the cerebral implementation of SELF and the first combined TOM study in which the cerebral implementation of representing ones own mental states (SELF) and the mental states of others (TOM) are investigated. METHODS Subjects First, in order to obtain behavioral results on the expanded story collection, a purely behavioral pilot study was performed. Forty-two healthy male and female volunteers (age 21 to 32 years) with no known history of neurological or psychiatric illness were studied. Thereafter, eight right-handed, healthy male volunteers (age 25 to 36 years) with no history of neurological or psychiatric illness participated in the fMRI study. Only male participants were studied in order to avoid variation in brain size and shape between sexes and hence improve image normalization. Informed consent was obtained before participation. Stimulus Material, Tasks, and Study Design The original collection of short studies (Fletcher et al., 1995; Happe´ et al., 1996, 1999; Gallagher et al., 2000), comprising “physical stories” (condition 2: TOM⫺SELF⫺, hereafter: T⫺S⫺), “TOM stories” (condition 3: TOM⫹SELF⫺, hereafter: T⫹S⫺), and “unlinked sentences” (condition 1: high-level baseline, hereafter: baseline), was translated into German (K.V., A.N.) and checked for word count and syntactical and semantic complexity by a linguist. There were no significant differences in word counts between the stories of each particular condition. With regard to syntax, only sentences consisting of one principal sentence and one subordinate clause were used. Concerning semantic complexity, the texts of the stories were judged qualitatively, as there are no quantitative measures for semantic complexity available in German language. The stimuli of the control condition “unlinked sentences” consisted of a series of sentences with no semantic consistency or coherence between them (Fletcher et al., 1995). To ensure attentive reading, subjects were asked details on one of the sentences of the texts. In the “physical story” condition (T⫺S⫺), short consistent texts were shown presenting a short “story” on a certain physical event. No perspectivetaking was involved, neither was TOM capacity as an

172

VOGELEY ET AL.

TABLE 1 Two-Way Factorial Design of the Study Theory of mind TOM⫹

TOM⫺

Self perspective SELF⫺ Condition 3 TOM stories (T⫹S⫺)

Condition 2 Physical stories (T⫺S⫺) SELF⫹ Condition 4 Condition 5 TOM and SELF stories SELF stories (T⫹S⫹) (T⫺S⫹)

Note. This schema demonstrates the two-way factorial experimental design applied, in which both factors TOM and SELF were varied systematically.

assignment of mental states to an actor in the story nor to oneself. In the “TOM story” condition (T⫹S⫺) stories were presented in which agents play a particular role, to which a mental state (e.g., perception, judgment) had to be ascribed. Two newly developed conditions which engaged the capacity of SELF in the presence or absence of TOM were added. To achieve this, these self-perspective conditions incorporated the study participant as one of the agents in the story. In the “self and other ascription stories” (condition 4: TOM⫹SELF⫹, hereafter: T⫹S⫹), subjects were asked to ascribe adequate behavior, attitudes, or perceptions to themselves in the given plot, similar to the task performed in the “TOM stories.” Correct task performance necessitated the correct assignment of both to the other agent and to oneself. In the “self ascription stories” (condition 5: TOM⫺SELF⫹, hereafter: T⫺S⫹), persons were asked to report their behavior, attitudes, or perceptions in inherently ambiguous situations in a specific plot. The correct assignment of another person’s mental state in the TOM conditions (T⫹S⫹, T⫹S⫺) was tested by asking the participants to infer a specific behavior or attitude in the given context of the story, judged as adequate or inadequate according to Fletcher et al. (1995) and Happe´ et al. (1996). Correct assignment of SELF (T⫹S⫹, T⫺S⫹) was monitored by the use of personal pronouns in the documented answer of the particular story. Thus, this experimental design systematically studies the two factors of TOM and SELF independently in four different conditions in a two-way factorial manner (see Table 1), plus baseline. In the behavioral pilot study, stories were presented using a slide projector. Story texts were presented for 25 s. Subsequently questions were presented for 15 s. Examples of the stories and the related questions are given below. Answers had to be given in writing. Subjects were informed about the particular type of story (condition) before presentation of the stimulus material.

In the fMRI study, story texts were presented on a display in black on a light gray background during the fMRI BOLD contrast EPI measurements in the center of a 29 cm screen (horizontal angle of 60°, vertical angle of 30°). Each story was presented for 25 s on the display, with the question being presented subsequently for 15 s. Again, subjects were informed about the particular type of story (condition) before the stimulus material was presented and before scanning was started. Subjects were instructed to read the story carefully and to read and answer the subsequent question silently (covertly). Volumes were acquired continuously every 5 s over the whole period of 40 s, while subjects performed the experimental tasks. After 40 s the presentation and the MR image acquisition were stopped and the subjects were asked to give the answers overtly. Note, this “off-line” assessment of task performance allowed us to check the correct assignment of SELF or TOM but avoided putative movement artifacts due to overt speech during the acquisition of MR images. For each particular story the answer was scored for adequacy thus providing a behavioral control for each experimental condition. Reading each story and the subsequent related question corresponded to 8 MR scans, of which the first two were discarded prior to analysis to allow for T1 saturation effects (see Image Analysis for more detail). In each of the four experimental conditions and the baseline, eight trials were presented. The overall duration of the scanning session was approximately 40 – 45 min. Example of “unlinked sentences” (baseline). The two countries had been at war. A housewife is about to enter the supermarket. Today he is going to buy an expensive new stereo. Mrs. Brown, the postmistress, receives a special parcel. Mrs. Pearson wouldn’t harm a fly. Mary’s birthday is in February. Late one evening the old man was watching television. Question: Is Mary’s birthday in February? Example of “physical stories” (T⫺S⫺). A burglar is about to break into a jeweler’s shop. He skillfully picks the lock on the shop door. Carefully he crawls under the electronic detector beam. If he breaks this beam it will set off the alarm. Quietly he opens the door of the store-room and sees the gems glittering. As he reaches out, however, he steps on something soft. He hears a screech and something small and furry runs out past him, towards the shop door. Immediately the alarm sounds. Question: Why did the alarm go off? Example of “TOM stories” (T⫹S⫺). A burglar who has just robbed a shop is making his getaway. As he is running home, a policeman on his beat sees him drop his glove. He doesn’t know the man is a burglar, he just wants to tell him he dropped his glove. But when the policeman shouts out to the burglar, “Hey, you! Stop!” the burglar turns round, sees the policeman and gives


himself up. He puts his hands up and admits that he did the break-in at the local shop. Question: Why did the burglar do that? Example of “self and other ascription stories” (T⫹S⫹). A burglar who has just robbed a shop is making his getaway. He has robbed your store. But you cannot stop him. He is running away. A policeman who comes along sees the robber as he is running away. The policeman thinks that he is running fast to catch the bus nearby. He does not know that the man is a robber who has just robbed your store. You can talk quickly to the policeman before the robber can enter the bus. Question: What do you say to the policeman? Example of “self ascription stories” (T⫺S⫹). You went to London for a weekend trip and you would like to visit some museums and different parks around London. In the morning, when you leave the hotel, the sky is blue and the sun is shining. So you do not expect it to start raining. However, walking around in a big park later, the sky becomes gray and it starts to rain heavily. You forgot your umbrella. Question: What do you think? Functional Magnetic Resonance Imaging Functional magnetic resonance (fMRI) was carried out using echo planar imaging (EPI) with whole brain coverage. FMRI imaging was performed on a 1.5 T MRI system (SIEMENS Magnetom VISION, Erlangen, FRG), using the standard head coil for radiofrequency (RF) transmission and signal reception. Sequences with the following parameters were employed: repetition time (TR) ⫽ 5000 ms, echo time (TE) ⫽ 66 ms, field-of-view ⫽ 200 ⫻ 200 mm 2, ␣ ⫽ 90°, matrix size ⫽ 64 ⫻ 64, voxel size ⫽ 3.125 ⫻ 3.125 ⫻ 4.4 mm 3. Using a midsagittal scout image, 32 axial slices (0.3 mm interslice gap) were positioned to cover the whole brain. During one trial with the presentation of a story and the subsequent question, eight whole-brain data sets were acquired. The scanning procedure was performed continuously over one trial and was restarted, after the test person answered. In addition, anatomical images of the entire brain were obtained by using a strongly T1-weighted, 3-D-gradient-echo pulse sequence (MP-RAGE, magnetization-prepared, rapid acquisition gradient echo) with the following parameters: TR ⫽ 11.4 ms, TE ⫽ 4.4 ms, 15° flip angle, FOV ⫽ 256 ⫻ 256 mm 2, matrix size ⫽ 200 ⫻ 256, 128 sagittal slices with 1.33 mm thickness. Image Analysis The entire image analysis including realignment, normalization, and statistical analysis was performed using Statistical Parametrical Mapping (SPM99, Wellcome Department of Cognitive Neurology, London, UK) implemented in MATLAB (Mathworks Inc., Sher-

173

born, MA). The first two images of each trial, during which the MR signal reaches a steady state, were discarded. The image time series was realigned using the first image and spatially normalized to the stereotactic space of Talairach and Tournoux (1988) using templates provided by SPM99. Data were subsequently smoothed with an isotropic Gaussian kernel of 7 mm at full-width-half-maximum to compensate for normal variations in brain size and individual gyral patterns. Analysis was carried out using the general linear model and a boxcar waveform convolved with a hemodynamic response function accounting for the delayed cerebral hemodynamic response after stimulus presentation. Subject-specific, low-frequency drifts in signal changes were removed by a high pass filter and global signal changes were treated as a covariate of no interest. The mean activity of each voxel throughout the whole trial was used as a dependent variable. Specific effects for each voxel were tested by applying appropriate linear contrasts to the parameter estimates for each condition, resulting in a t statistic for every particular voxel. The resulting set of voxel values for each contrast constituted a statistical parametric map of the t statistic SPM {t}, which was subsequently transformed to the unit normal distribution SPM {z}. Statistical inferences were based on the theory of random Gaussian fields (Friston et al., 1995). Throughout, we report activations significant at P ⬍ 0.05 corrected for multiple comparisons at an extent threshold of a minimum of 17 pixels (Tables 2 and 3, Figs. 1–3). For the fMRI group data analysis, all images of all subjects were analyzed in one design matrix, generating a fixed-effect model, thus limiting inference of activation to the subjects and scans of this study. The stereotactic coordinates of the voxels of local maximum significant activation were determined within regions of significant relative activity change associated with the different tasks. The anatomic localization of these local maxima was assessed by reference to the standard stereotactic atlas (Talairach and Tournoux, 1988) and superimposed on a mean anatomical image of all subjects participating in the study. The data were first analyzed for the specific effects of each condition against the baseline. Results are given in Tables 2a–2d and the corresponding Figs. 1a–1d. Second, data were analyzed for the main effects of TOM [(T⫹S⫹ plus T⫹S⫺) relative to (T⫺S⫹ plus T⫺S⫺); and vice versa], and SELF [(T⫺S⫹ plus T⫹S⫹) relative to (T⫹S⫺ plus T⫺S⫺); and vice versa]. In addition, the contrast of SELF relative to TOM (T⫺S⫹ relative to T⫹S⫺) was calculated to assess the significance of the specific differences between SELF and TOM. Finally, we assessed whether the neural mechanisms underlying TOM and SELF interacted with each other [(T⫹S⫹ relative to T⫹S⫺) relative to (T⫺S⫹ relative to T⫺S⫺)]; and vice versa]. These comparisons were concerned with the

174

VOGELEY ET AL.

question of whether there are regions specifically activated by the combination of TOM and SELF. RESULTS Behavioral Data of the Pilot Study Participants performed comparably well on all tasks. Individual test persons made up to four errors of the eight presented stories per condition. Stories not answered were taken as wrong answer. Mean values (m) and standard deviations (SD) of correct answers of the participants of the pilot study were: physical stories (T⫺S⫺): m ⫽ 6.5, SD ⫽ 1.1; TOM stories (T⫹S⫺): m ⫽ 6.57, SD ⫽ 0.82; TOM factor in self and other ascription stories (T⫹S⫹): m ⫽ 6.88, SD ⫽ 0.88; SELF factor in self and other ascription stories (T⫹S⫹): m ⫽ 7.5, SD ⫽ 0.7; SELF factor in self ascription stories (T⫺S⫹): m ⫽ 7.38, SD ⫽ 0.72. Behavioral Data of the fMRI Study Again, participants performed comparably well on all tasks. Test persons made no more than two errors out of the eight presented stories per condition. Mean values (m) and standard deviations (SD) of correct answers of the participants of the fMRI study were: physical stories (T⫺S⫺): m ⫽ 7.25, SD ⫽ 0.66; TOM stories (T⫹S⫺): m ⫽ 6.8, SD ⫽ 0.7; TOM factor in self and other ascription stories (T⫹S⫹): m ⫽ 6.75, SD ⫽ 0.43; SELF factor in self and other ascription stores (T⫹S⫹): m ⫽ 7.5, SD ⫽ 0.71; SELF factor in self ascription stories (T⫺S⫹): m ⫽ 7.38, SD ⫽ 0.67. The subjects’ performances were close to ceiling as expected and reported in previous similar studies on normal healthy individuals (e.g., Gallagher et al., 2000). Behavioral results obtained in the fMRI study corresponded well to behavioral data of the previously performed pilot study. No significant differences of the mean values of correct answers in all different conditions could be detected as assessed by t tests for unpaired samples. Simple Effects of Experimental Conditions Relative to Baseline The results of the simple effects of all experimental conditions (T⫺S⫺, T⫹S⫺, T⫹S⫹, T⫺S⫹) against the baseline are summarized in Tables 2a–2d. Figures 1a–1d provide corresponding SPM {z} maps of the areas with increased neural activity corresponding to each experimental condition (relative to the baseline). T⫺S⫺ relative to the baseline did not show any significant increases in neural activity (Table 2a, Fig. 1a). T⫹S⫺ relative to the baseline was associated with significant activation in the right anterior cingulate cortex (Table 2b, Fig. 1b). T⫹S⫹, which required both SELF and TOM relative to the baseline, showed right

TABLE 2 Simple Effects of Experimental Conditions against Baseline Region (a) Effect of T⫺S⫺ relative to baseline No significant activations at this statistical level (b) Effect of T⫹S⫺ relative to baseline Right anterior cingulate cortex (c) Effect of T⫹S⫹ relative to baseline Right anterior cingulate cortex Right premotor cortex Right motor cortex (d) Effect of T⫺S⫹ relative to baseline Right temporoparietal junction Right motor cortex Right anterior cingulate cortex Right premotor cortex

x

y

z

Z

6

56

⫺8

5.62

8 54 54

56 6 ⫺10

⫺6 38 46

6.75 6.07 5.66

58 60 2 52

⫺56 ⫺12 46 6

12 42 ⫺6 40

7.28 6.04 5.80 5.16

Note. This table provides the summary of volume statistics under the calculation of the simple effects of experimental conditions against the baseline. Only those regions are reported that reach the significance level of P ⫽ 0.05 (voxel-level corrected) and the extent threshold of 17 voxels. Data in a– d correspond to Figs. 1a–1d.

anterior cingulate cortex activation in the same region as observed during T⫹S⫺; however, there were additional significant increases in neural activity in the right premotor and motor cortex (Table 2c, Fig. 1c). T⫺S⫹, in which only SELF but no TOM relative to the baseline was required, revealed increased neural activity in the right temporoparietal junction, in addition to the increases in neural activity already observed in the right premotor and motor cortex associated with T⫹S⫹ (relative to baseline) and the right anterior cingulate cortex associated with T⫹S⫹ (relative to baseline) and T⫹S⫺ (relative to baseline) (Table 2d, Fig. 1d). Main Effect of TOM The main effect of the TOM is summarized in Table 3a. Figure 2a provides the corresponding SPM {z} map of the areas with significantly increased neural activity associated with the experimental conditions during which TOM was required relative to those experimental conditions during which no TOM was required [(T⫹S⫹ plus T⫹S⫺) relative to (T⫺S⫹ plus T⫺S⫺); and vice versa]. Increases in neural activity associated with TOM were observed predominantly in the right anterior cingulate cortex and left temporopolar cortex. Figure 3a provides an overlay of the principally activated voxel in the anterior cingulate cortex superimposed on the mean anatomical image of all subjects. The reverse comparison [(T⫺S⫹ plus T⫺S⫺) relative to (T⫹S⫹ plus T⫹S⫺)] revealed increased neural activity in the right temporoparietal junction and superior parietal cortex.


175

FIG. 1. Specific effects of all experimental conditions relative to baseline. (a) T⫺S⫺ relative to baseline. There is no significant differential brain activation. (b) T⫹S⫺ relative to baseline. Significant activation is observed in the right anterior cingulate cortex. (c) T⫹S⫹ relative to baseline. There is significant activation in the right anterior cingulate cortex and the right premotor and motor cortex. (d) T⫺S⫹ relative to baseline. The most significant activation is found in the right temporoparietal junction. Further activations are observed in right motor and premotor cortex, and in the right anterior cingulate cortex. All images: P ⬍ 0.05, voxel-level corrected; extent threshold ⫽ 17 voxels.

Main Effect of SELF The main effect of SELF is summarized in Table 3b. Figure 2b provides the corresponding SPM {z}-map of the areas with increased neural activity associated with the conditions during which SELF was required [(T⫺S⫹ plus T⫹S⫹) relative to (T⫹S⫺ plus T⫺S⫺)]. This comparison revealed increased neural activity in the right temporoparietal junction (Figs. 2b and 3b) and in the anterior cingulate cortex. However, the increase in neural activity in the anterior cingulate cor-

tex was now bilaterally, and more inferior than the one observed as a main effect of TOM. Further significant increases in neural activity associated with SELF were observed in the right premotor and motor cortex and in the precuneus bilaterally. Figure 3b provides an overlay of the principally activated voxel in the temporoparietal junction superimposed on the mean anatomical image of all subjects. The reverse comparison [(T⫹S⫺ plus T⫺S⫺) relative to (T⫺S⫹ plus T⫺S⫹)] revealed increased neural activity in conditions, which did not

176

VOGELEY ET AL.

TABLE 3 Main Effects of the Factors TOM and SELF and Their Interaction Region (a) Main effect of TOM [(T⫹S⫹ plus T⫹S⫺) relative to (T⫺S⫹ plus T⫺S⫺)] Right anterior cingulate cortex a Left temporopolar region Left precentral cortex Right superior frontal cortex Right anterior cingulate cortex Right frontopolar cortex Left lateral prefrontal cortex (b) Main effect of SELF [(T⫺S⫹ plus T⫹S⫹) relative to (T⫹S⫺ plus T⫺S⫺)] Right temporoparietal junction a Right anterior cingulate cortex Right premotor cortex Right motor cortex Left precuneus Right precuneus Left anterior cingulate cortex (c) SELF relative to TOM (T⫺S⫹ relative to T⫹S⫺) Right temporoparietal junction Right precuneus Left precuneus Right motor cortex Left temporoparietal junction Right superior parietal lobule (d) Interaction of TOM and SELF [(T⫹S⫹ relative to T⫹S⫺) relative to (T⫺S⫹ relative to T⫺S⫺)] Right prefrontal cortex a

x

y

z

Z

6* ⫺58 ⫺62 6 4 22 ⫺56

56* 10 4 56 28 46 26

2* ⫺10 26 26 30 46 16

7.73* 6.42 6.18 5.95 5.48 5.27 5.10

58* 6 22 60 ⫺10 8 ⫺12

⫺56* 54 2 ⫺12 ⫺48 ⫺46 50

12* ⫺4 68 42 64 64 ⫺4

6.27* 5.93 5.78 5.70 5.61 5.30 5.15

56 16 ⫺10 38 ⫺46 64

⫺58 ⫺38 ⫺46 ⫺42 ⫺44 ⫺28

14 72 64 64 22 64

7.67 6.47 6.30 6.21 5.92 5.44

⫺4*

5.64*

56*

32*

Note. This table provides the summary of volume statistics under the calculation of the main effects of the factors TOM and SELF, the contrast SELF relative to TOM and the interaction of SELF and TOM according to a two-way factorial design. Only regions are reported that reach the significance level of P ⫽ 0.05 (voxel-level corrected) and the extent threshold of 17 voxels. Data in a– d correspond to Figs. 2a–2d and 3a–3c. Data in a, b, d and Figs. 2a, 2b, and 2d correspond to Figs. 3a, 3b, and 3c. a Figures 3a–3c provide an overlay view of these voxels on the mean structural anatomic image of the subjects of this study.

employ SELF in the left temporoparietal junction only. When contrasting SELF with TOM directly (T⫺S⫹ relative to T⫹S⫺), activation of the right temporoparietal junction and bilateral precuneus was found, thus corroborating the specific difference between SELF and TOM (Fig. 2c, Table 3c). Interaction of TOM and SELF The interaction of TOM and SELF [(T⫹S⫹ relative to T⫹S⫺) relative to (T⫺S⫹ relative to T⫺S⫺)] was calculated to identify those areas activated specifically as a result of the presence of both TOM and SELF, i.e., to demonstrate the neural activations evoked specifi-

cally due to assigning SELF in the presence of modeling another person’s mind or vice versa. This analysis revealed an isolated area with increased neural activity in the right lateral prefrontal cortex (Table 3d). Figure 2d provides the corresponding SPM {z}-map. Figure 3b provides an overlay of the principally activated voxel in the right prefrontal cortex superimposed on the mean anatomical image of all subjects. DISCUSSION Our results demonstrate that the ability to attribute opinions, perceptions, or attitudes to others, often referred to as “mind-reading” (Baron-Cohen, 1995), and the ability to apply SELF rely on both common and differential neural mechanisms. The cerebral implementation of TOM capacity is located predominantly in the anterior cingulate cortex, whereas the capacity for taking SELF leads to additional neural activations in the right temporoparietal junction and the medial aspects of the superior parietal lobe, i.e. , the precuneus bilaterally. The fact, that differential brain loci in different brain lobes are activated associated with the attribution of SELF or with “mind-reading” of others (TOM), suggests that these components are implemented at least in part in different brain modules and thus constitute distinct cognitive processes. This view is supported by the observation of a significant interaction between TOM and SELF in the right prefrontal cortex, a region which has previously been implicated in “supervisory attentional” mechanisms (Shallice and Burgess, 1996) or monitoring situations that involve conflict of senses (Fink et al., 1999). That these common and differential mechanisms are truly associated with TOM and SELF, respectively, is suggested by the significance of these activations relative to the baseline, which employed no SELF and/or TOM capacities. The possibility that the baseline might have involved a higher working memory load (as it may have been more difficult to hold unlinked material for subsequent answering of the questions) does not confound the differential assessment of the main factors of TOM and SELF or their interaction (as these comparisons do not include the baseline). Thus the putative increase of working memory load in the baseline affects only the direct comparison of the story conditions with the baseline. The factorial design employed allows the assessment of differential activations of SELF and TOM relative to each other. In this context we note that the physical story condition involved no explicit TOM or SELF-like capacity for correct task performance. Although implicit processes of self-identification or role playing during this condition cannot be ruled out entirely, it seems unlikely that participants focussed explicitly during the physical event stories on SELF or TOM.


177

FIG. 2. Main effects of TOM and SELF and their interaction. (a) (T⫹S⫹ plus T⫹S⫺) relative to (T⫺S⫹ plus T⫺S⫺). This SPM shows the neural activations associated with the main factor TOM. There is significant activation in the right anterior cingulate cortex, and left superior temporal cortex. (b) (T⫹S⫹ plus T⫺S⫹) relative to (T⫹S⫺ plus T⫺S⫺). This SPM represents the neural activation under the main factor SELF. There is still considerable activation at the anterior cingulate cortex. In addition to this, there is significant activation in the right temporoparietal junction and the precuneus bilaterally. (c) (T⫺S⫹ vs T⫹S⫺). This SPM represents the neural activation under the direct comparison of self versus TOM. There is significant activation in the right temporoparietal junction and the precuneus bilaterally. (d) (T⫹S⫹ vs T⫹S⫺) relative to (T⫺S⫹ vs T⫺S⫺). This SPM represents the neural activation under the interaction of both factors TOM and SELF. There is only one single activation site in the area of the right prefrontal cortex. All images: P ⬍ 0.05, voxel-level corrected; extent threshold ⫽ 17 voxels.

A number of functional imaging studies using PET and fMRI have previously delineated brain regions involved in “reading other minds” (Fletcher et al., 1995; Happe´ et al., 1996; Gallagher et al., 2000). These studies have repeatedly demonstrated increased neural activity associated with TOM conditions in the anterior cingulate cortex. We here replicate these findings by showing activation of the anterior cingulate cortex

when TOM was present. We extend previous results by demonstrating that the activation of the anterior cingulate cortex reflects TOM irrespective of whether or not subjects also needed to assign SELF. Previous PET studies on TOM, however, associated the anterior cingulate cortex activation with the left hemisphere, while we observed right hemisphere dominance. There are several explanations for this discrep-

178

VOGELEY ET AL.

FIG. 3. Anatomical localization of main effects of TOM and SELF and their interaction. In these images the principally activated voxels associated with TOM and SELF, respectively, and their interaction are superimposed on the anatomical mean image of all subjects participating in this study (after normalization of single images at 1 ⫻ 1 ⫻ 1 mm). (a) (T⫹S⫹ plus T⫹S⫺) relative to (T⫺S⫹ plus T⫺S⫺) at voxel x ⫽ 6, y ⫽ 56, z ⫽ 2. Under the main factor TOM there is strong activation in the anterior cingulate cortex and the medial aspect of the superior frontal cortex, confined to the right hemisphere and left-sided temporopolar activation. (b) (T⫹S⫹ plus T⫺S⫹) relative to (T⫹S⫺ plus T⫺S⫺) at voxel x ⫽ 58, y ⫽ ⫺56, z ⫽ 12. This image represents the neural activation under the main factor SELF, under which activation the most prominent activation was found in the right temporoparietal junction. (c) (T⫹S⫹ vs T⫹S⫺) relative to (T⫺S⫹ vs T⫺S⫺) at voxel x ⫽ 56, y ⫽ 32, z ⫽ ⫺4. Only one single activation site is associated with the interaction of both factors, which is situated in the right prefrontal cortex. All images: P ⬍ 0.05, voxel-level corrected; extent threshold ⫽ 17 voxels.

ancy. A merely technical one is concerned with the differences in spatial resolution of PET and fMRI studies as the PET data were smoothed with larger filters (up to 16 mm). Thus, the correct assignment of an activation of a midline structure to either hemisphere might be more difficult with PET than with fMRI. Empirical support for our right-sided activation of the anterior cingulate cortex stems from another recent fMRI study which also observed right anterior cingu-

late cortex activation during TOM (Gallagher et al., 2000). In this context one should remember that such hemispheric side-to-side differences are relative rather than absolute. Furthermore, hemispheric lateralization in higher order cognitive tasks seems to be stimulus- (Fink et al., 1997) and context-sensitive (Macaluso and Frith, 2000) which may also account for the differences observed. Yet, we note that our finding of right hemisphere dominance for TOM is in good accor-


dance with right hemispheric activations observed during pragmatic language tasks which involved activations of the middle temporal gyrus, dorsolateral prefrontal cortex, and medial areas including the precuneus and the anterior cingulate cortex (Brownell et al., 1990; Bottini et al., 1994). Furthermore, such functional imaging data are complemented by neuropsychological data. Patients with right hemispheric lesions demonstrate difficulties with verbal and non-verbal communication, the understanding of metaphors, non-conventional or indirect meaning, indirect questions or the emotional-prosodic quality of expressions, and TOM (Ross, 1981; Brookshire and Nicholas, 1984; Hirst et al., 1984; Foldi, 1987; Bryan, 1988; Weylman et al., 1989; Brownell et al., 1994; Happe´ et al., 1999). Also, Stone et al. (1998) studied TOM of varying difficulty in patients with bilateral orbitofrontal and left dorsolateral prefrontal lesions. Patients with bilateral orbitofrontal lesions demonstrated difficulties in second order TOM tasks, whereas no specific TOM deficits were detected in patients with left dorsolateral prefrontal lesions. Taken together, our current imaging data, previous functional imaging data (Gallagher et al., 2000), and neuropsychological data strongly implicate right anterior cingulate cortex when subjects assign opinions, perceptions or attitudes to others, as assessed by TOM. While the anterior cingulate cortex seems to be the key structure for assigning a mental state to someone else (irrespective of whether SELF is involved in the situation), our results also imply that activation of this brain region is not sufficient when the ability to apply SELF is required. In addition to anterior cingulate activation, taking SELF draws upon right inferior temporoparietal cortex, irrespective of whether or not subjects need to assign TOM at the same time. Both the significant effect of the main factor SELF and the analysis of single condition effects involving SELF (relative to the baseline) implicate the right temporoparietal junction in taking SELF (Figs. 2b, 3b). This finding is of considerable interest as lesions to the right temporoparietal junction typically result in visuo-spatial neglect, a neuropsychological deficit which often involves a disturbed (egocentric) frame of reference (Vallar et al., 1999). The activation of the temporoparietal junction during SELF is also compatible with evidence for the implementation of our body image in this region (Berlucchi and Aglioti, 1997), suggesting, that taking SELF may draw on a body representation as the center of an ego-centric experiential space. This conjecture in turn is in good accordance with reports on increased neural activity of right inferior parietal cortex involving visuo-spatial attention, e.g., navigation through virtual reality scenes (Maguire et al., 1998) or assessment of the subjective mid-sagittal plane (Vallar et al., 1999).

179

Iacoboni et al. (1999) performed a study on cortical mechanisms of motor imitation. Contrasting an imitation task relative to two different observation tasks of specific finger movements, the authors found a significant activation in the left frontal operculum, the right parietal region and the right parietal operculum. These findings are consistent with our data, as it is suggested, that as soon as ones own behavior is involved, there is an increased activity in the area of the temporoparietal junction. In summary, both functional imaging and the neuropsychological data imply that the temporoparietal junction is involved in computing an egocentric reference frame. However, our data strongly suggest a more general role for this region which goes beyond visuo-spatial judgments per se: Increased neural activity in this region was also evoked when assigning SELF by use of personal pronouns in languagebased stimulus material compared to cartoons as stimuli. It is noteworthy that Gallagher et al. (2000) using the same classical TOM story collection without SELF evoking stimuli also observed activation in the inferior parietal lobule bilaterally. This activation pattern is in contrast to classical TOM studies (Fletcher et al., 1995; Happe´ et al., 1996). Although their stimulus material did not include any explicit SELF material or reference, it cannot be ruled out completely that the participants implicitly involved some SELF component (that was not explicitly monitored). Likewise, other unknown or uncontrolled factors may have been involved in that study. Our findings lead to the hypothesis that the temporoparietal activation indicates the involvement of SELF, whereas the medial frontal activation indicates the involvement of TOM. Further studies operationalizing SELF in different ways including other sensory modalities are needed to clarify this issue. Notably, there was an additional significant activation in the right premotor and motor cortex during conditions that required the involvement of SELF, both in the analyses of the simple effects of the single conditions relative to the baseline as well as in the analysis of the main effect of SELF (Figs. 1c, 1d, 2b, 2c, and 3b). One plausible explanation involves the hypothesis that the evocation of SELF might also involve planning and preparation of action. It seems likely that there is a closer coupling to planning and execution of action during the attribution of mental states to oneself as opposed to a situation in which mental states are attributed to someone else. Such an interpretation, although speculative, is in good accordance with a recent report by Weiss et al. (2000) who demonstrated that the neural activations underlying action are modulated by near (peripersonal) and far (extrapersonal) space, respectively. Only those experimental conditions that involved situations in which near space was experienced were associated with activations in the motor and premotor cortex areas. By contrast, the ex-

180

VOGELEY ET AL.

perimental conditions in which far space was experienced did not involve activation of motor and premotor cortex. The interaction of TOM capacity and SELF draws upon right prefrontal cortex. The nature of the interaction suggests that an area within right prefrontal cortex is specifically activated when an integration of TOM and SELF is demanded. Previous studies suggested an involvement of right prefrontal cortex in the segregation and integration of different cognitive capacities including situations with increased monitoring demand (Fink et al., 1999) and self-recognition (Keenan et al., 2000). Keenan et al. studied self-recognition paradigms in which the own face appeared as an object, that had to be identified, e.g., subjects viewed a photograph of their own face. This task is psychologically different from our approach, in which the self is examined as the subject of an experience, and not as the object. Other functions executive probably closely linked to SELF like working memory have also been associated with the prefrontal cortex (Goldman-Rakic, 1987, 1994; Vogeley et al., 1999). Finally, our data may also contribute to the debate on “simulation theory” (ST) and “theory theory” (TT). The “mirror neuron system” described by Gallese et al. (1996) responding both to observation of other animals performing goal-directed actions as well as to execution of the same movements of the experimental animal studied has been taken as evidence for the validity of ST. It has been hypothesized that the mirror neurons “represent a primitive version . . . of a simulation heuristic that might underlie mind-reading” (Gallese and Goldman, 1998, p. 498). Furthermore, it has been assumed that human subjects performing TOM use their own mental states to predict or explain mental processes of others (Gallese and Goldman, 1998, p. 496) as opposed to the TT concept, according to which subjects performing TOM use a specific commonsense psychological theory, also referred to as “folk psychology.” It must be emphasized that this particular finding of the mirror neuron system cannot prove the exclusive validity of ST and completely reject TT. In the case of exclusive validity of ST, all mental states requiring TOM, attributed to someone else or to oneself, should activate the same brain region. This prediction is in contrast to the results of our study, which show an additional specific activation associated with SELF which is not observed with TOM (besides the marked overlap with shared neural activations associated with both SELF and TOM). That TOM and SELF involve at least in part distinct neural mechanisms is further corroborated by the finding of a significant interaction between both factors. Thus, our data speak against both ST and TT in pure form and rather suggest a mixture of both concepts. We suggest that the TT component is based on the anterior cingulate cortex activation, whereas the ST component is primarily associ-

ated with increased brain activity in the area of the right temporoparietal junction. Our data thus support the view that “knowledge of another’s subjectivity is going to have to involve one’s own” (Bolton and Hill, 1996, p. 135). The increasing knowledge of mental states of others acquired during development could then be based at least in part on the knowledge of which perspective should be taken in a particular situation, for which our data strongly imply that the right prefrontal cortex is the critical locus. Allowing a differential induction of a self perspective involvement in cognitive processes, our study design and related paradigms to be developed may become useful as an experimental tool in cognitive sciences and especially in relation to possible disorders of TOM, e.g., autism and schizophrenia. In summary, this study provides experimental evidence for an at least in part independent cerebral implementation of self perspective in the context of theory of mind. Expansions of classical TOM paradigms could become useful tools for the further study of the interdependency of the firstperson and third-person perspective. ACKNOWLEDGMENTS We gratefully acknowledge the support of Chris Frith during development of the stimulus material and Oliver Zafiris and Paul Fletcher for their help in data analysis.

REFERENCES Baron-Cohen, S. 1995. Mindblindness. MIT Press, Cambridge, MA. Berlucchi, G., and Aglioti, S. 1997. The body in the brain: Neural bases of corporeal awareness. Trends Neurosci. 20: 560 –564. Bolton, D., and Hill, J. 1996. Mind, Meaning and Mental Disorder. The Nature of Causal Explanation in Psychology and Psychiatry. Oxford Univ. Press, Oxford. Bottini, G., Corcoran, R., Sterzi, R., Paulesu, E., Schenone, P., Scarpa, P., Frackowiak, R. S. J., and Frith, C. D. 1994. The role of the right hemisphere in the interpretation of figurative aspects of language. Brain 117: 1241–1253. Brookshire, R. H., and Nicholas, C. E. 1984. Comprehension of directly and indirectly stated main ideas and details in discourse by brain-damaged and non-brain-damaged listeners. Brain Lang. 21: 21–36. Brownell, H. H., Simpson, T. L., Bihrle, A. M., Potter, H. H., and Gardner, H. 1990. Appreciation of metaphorical alternative word meanings by left and right brain-damaged patients. Neuropsychologia 28: 375–383. Bryan, K. L. 1988. Assessment of language disorders after right hemisphere damage. Br. J. Dis. Commun. 23: 111–125. Buccino, G., Binkofski, F., Fink, G. R., Fadiga, L., Fogassi, L., Gallese, V., Seitz, R. J., Zilles, K., Rizzolatti, G., and Freund, H.-J. 2001. Action observation activates premotor and parietal areas in a somatotopic manner: An fMRI study. Eur. J. Neurosci. 13: 400 – 405. Carruthers, P. 1996. Simulation and self-knowledge: A defence of theory-theory. In Theories of Theories of Mind (P. Carruthers and P. K. Smith, Eds.), pp. 22–38. Cambridge Univ. Press, Cambridge. Fink, G. R., Marshall, J. C., Halligan, P. W., Frith, C. D., Frackowiak, R. S., and Dolan, R. J. 1997. Hemispheric specialization for

SELF PERSPECTIVE AND THEORY OF MIND global and local processing: The effect of stimulus category. Proc. R. Soc. Lond. B. Biol. Sci. 264: 487– 494. Fink, G. R., Marshall, J. C., Halligan, P. W., Frith, C. D., Driver, J., Frackowiak, R. S., and Dolan, R. J. 1999. The neural consequences of conflict between intention and the senses. Brain 122: 497–512. Fletcher, P., Happe´, F., Frith, U., Baker, S. C., Dolan, R. J., Frackowiak, R. S. J., and Frith, C. D. 1995. Other minds in the brain: A functional imaging study of “theory of mind” in story comprehension. Cognition 57: 109 –128. Foldi, N. S. 1987. Appreciation of pragmatic interpretation of indirect commands: Comparison of right and left hemisphere braindamaged patients. Brain Lang. 31: 88 –108. Friston, K. J., Homes, A. P., Worsley, K. J., Poline, J.-P., Frith, C. D., and Frackowiak, R. S. J. 1995. Statistical parametric maps in functional imaging: A general linear approach. Hum. Brain Mapp. 2: 189 –210. Gallagher, I. 2000. Philosophical conceptions of the self: Implications for cognitive science. Trends Cogn. Sci. 4: 14 –21. Gallagher, H. L., Happe´, F., Brunswick, N., Fletcher, P. C., Frith, U., and Frith, C. D. 2000. Reading the mind in cartoons and stories: An fMRI study of ‘theory of mind’ in verbal and nonverbal tasks. Neuropsychologia 38: 11–21. Gallese, V., Fadiga, L., Fogassi, L., and Rizzolatti, G. 1996. Action recognition in the premotor cortex. Brain 119: 593– 609. Gallese, V., and Goldman, A. 1998. Mirror neurons and the simulation theory of mind-reading. Trends Cogn. Sci. 2: 493–501. Goldman-Rakic, P. S. 1987. Circuitry of primate prefrontal cortex and regulation of behavior by representational memory. In Handbook of Physiology (F. Plum and U. Mountcastle, Eds.), Vol. 5, pp. 373– 417. Am. Physiol. Soc., Washington. Goldman-Rakic, P. S. 1994. Working memory dysfunction in schizophrenia. J. Neuropsych. Clin. Neurosci. 6: 348 –357. Gopnik, A., and Wellmann, H. 1992. Why the child’s theory of mind really is a theory. Mind Lang. 7: 145–151. Gopnik, A. 1993. How we know our minds: The illusion of firstperson-knowledge of intentionality. Behav. Brain Sci. 16: 1–14. Happe´, F., Ehlers, S., Fletcher, P., Frith, U., Johansson, M., Gillberg, C., Dolan, R., Frackowiak, R., and Frith, C. 1996. “Theory of mind” in the brain. Evidence from a PET scan study of Asperger syndrome. Neuroreport 8: 197–201. Happe´, F. G. E., Brownell, H., and Winner, E. 1999. Acquired “theory of mind” impairments following stroke. Cognition 70: 211–240. Harris, P. L. 1992. From simulation to folk psychology: The case for development. Mind Lang. 7: 120 –144.

181

Hirst, W., LeDoux, J., and Stein, S. 1984. Constraints on the processing of indirect speech acts: Evidence from aphasiology. Brain Lang. 23: 26 –33. Iacoboni, M., Woods, R. P., Brass, M., Bekkering, H., Mazziotta, J. C., and Rizzolatti, G. 1999. Cortical mechanisms of human imitation. Science 286: 2526 –2528. Keenan, J. P., Wheeler, M. A., Gallup, G. G., Jr., and Pascual-Leone, A. 2000. Self-recognition and the right prefrontal cotex. Trends Cogn. Sci. 4: 338 –344. Macaluso, E., and Frith, C. 2000. Interhemispheric differences in extrastriate areas during visuo-spatial selective attention. NeuroImage 12: 485– 494. Maguire, E. A., Burgess, N., Donnett, J. G., Frackowiak, R. S., Frith, C. D., and O’Keefe, J. 1998. Knowing where and getting there: A humn navigation network. Science 280: 921–924. Perner, J., and Howes, D. 1992. “He thinks he knows:” And more developmental evidence against simulation (role taking) theory. Mind Lang. 7: 72– 86. Premack, D., and Woodruff, G. 1978. Does the chimpanzee have a theory of mind? Behav. Sci. 4: 515–526. Ross, E. D. 1981. The aprosodias. Functional-anatomic organization of the affective components of language in the right hemisphere. Arch. Neurol. 38: 561–569. Shallice, T., and Burgess, P. 1996. The domain of supervisory processes and temporal organization of behaviour. Philos. Trans. R. Soc. Lond. B Biol. Sci. 351: 1405–1411. Stone, V. E., Baron-Cohen, S., and Knight, R. T. 1998. Frontal lobe contributions to theory of mind. J. Cogn. Neurosci. 10: 640 – 656. Talairach, J., and Tournoux, P. 1988. Coplanar Stereotactic Atlas of the Human Brain. Thieme Verlag, Stuttgart. Vallar, G., Lobel, E., Galati, G., Berthoz, A., Pizzamiglio, L., and Le Bihan, D. 1999. A fronto-parietal system for computing the egocentric spatial frame of reference in humans. Exp. Brain. Res. 124: 281–286. Vogeley, K., Kurthen, M., Falkai, P., and Maier, W. 1999. The prefrontal cortex generates the basic constituents of the self. Consc. Cogn. 8: 343–363. Weiss, P. H., Marshall, J. C., Wunderlich, G., Tellmann, L., Halligan, P. W., Freund, H. J., Zilles, K., and Fink, G. R. 2000. Neural consequences of acting in near versus far space: A physiological basis for clinical dissociations. Brain 123: 2531–2541. Weylman, S. T., Brownell, H. H., Roman, M., and Gardner, H. 1989. Appreciation of indirect requests by left- and right-brain-damaged patients: The effects of verbal context and conventionality of wording. Brain Lang. 36: 580 –591.