Face-specific and domain-general characteristics of cortical responses during self-recognition

Abstract

The ability of visual self-recognition in animals and infants is considered a hallmark of the domain-general cognitive representation of the self, which also underpins higher social ability. Cortical regions activated during self-face recognition in human adults have been accordingly expected to play the domain-general role in self-processing. However, there is no evidence of the involvement of this network in non-face domains. We compared cortical responses during face and name recognition of self, a friend, and an unfamiliar person, using functional magnetic resonance imaging (fMRI). Recognition of the self-face activated the right inferior frontal, precentral, supramarginal, and bilateral ventral occipitotemporal regions, consistent with previous findings, whereas these regions did not show self-specific activation during name recognition. During both face and name recognitions, increased activation for the friend and unfamiliar person than for the self was observed in the bilateral temporoparietal regions, and higher activation for the self and friend than for the unfamiliar person was observed in the medial cortical structures. These results suggest that the role of the self-specific networks during face recognition is not domain-general, but rather face-specific, and that the medial cortical structures, which are also implicated in self-referential processes, are not relevant to self–other distinction during face or name recognition. Instead, the reduced temporoparietal activation is a domain-general characteristic of the cortical response during self-recognition, which may reflect suppression of an automatic preparatory process for social interaction, possibly paralleling the disappearance of social behavior to the mirrored self-image at the emergence of self-recognition in animals and infants.

Introduction

Since the observation of chimpanzees' ability to recognize themselves in a mirror (Gallup, 1970), visual self-recognition has been considered evidence of the domain-general representation of the self in animals (Gallup, 1982, Reiss and Marino, 2001, Plotnik et al., 2006) and human infants (Anderson, 1984). Given that the ability of visual self-recognition phylogenetically and ontogenetically co-emerged with the ability for empathy, a close association of the ability for visual self-recognition with higher social cognition has been assumed (Gallup, 1982, Plotnik et al., 2006). A neuroanatomical basis for this assumption has been provided by the fact that this ability is afforded only to animals with relatively large brains (Marino, 2002, Shoshani et al., 2006). It is then understandable that the neural substrate for visual self-recognition is expected to coincide with the key mechanism for domain-general self-processing and possibly higher social cognition (Gallup, 1982, Keenan et al., 2000b, Decety and Sommerville, 2003). In fact, early studies often demonstrated right hemispheric dominance in visual self-recognition (e.g., Preilowski, 1977, Keenan et al., 2000a, Keenan et al., 2001b; but see Turk et al., 2002, Brady et al., 2004), which was consistent with the proposed notion of right hemispheric dominance of domain-general self-processing or self-other distinction; this notion has been supported in other domains of self-processing, such as self-awareness related to bodily sensation and action (see Feinberg and Keenan, 2005 for review) and the retrieval of autobiographical memory (Fink et al., 1996).

However, the validity of assuming the cortical network for visual self-recognition as a domain-general self-processing mechanism has been questioned in light of information provided by recent functional imaging studies regarding the precise cortical organization of the visual self-recognition mechanism (Keenan et al., 2001a, Kircher et al., 2000, Kircher et al., 2001, Sugiura et al., 2000, Sugiura et al., 2005, Sugiura et al., 2006a, Platek et al., 2004a, Platek et al., 2006, Uddin et al., 2005, Devue et al., 2007, Morita et al., 2008). Although these studies reported different sets of cortical regions, using different experimental designs, several specific cortical regions have been frequently reported: the right mid-inferior frontal (Uddin et al., 2005, Sugiura et al., 2006a, Devue et al., 2007, Morita et al., 2008), precentral (Platek et al., 2004a, Sugiura et al., 2005, Sugiura et al., 2006a, Morita et al., 2008), supramarginal (Sugiura et al., 2000, Sugiura et al., 2006a, Uddin et al., 2005, Platek et al., 2006), and bilateral ventral occipitotemporal (Kircher et al., 2000, Kircher et al., 2001, Sugiura et al., 2000, Sugiura et al., 2005, Sugiura et al., 2006a, Uddin et al., 2005, Morita et al., 2008) regions (i.e., distinction between the gyrus and adjacent sulci is not made here). Advocates of right hemispheric dominance of domain-general self-processing took the right frontal and parietal activation as support for their argument (Keenan et al., 2000b, Decety and Sommerville, 2003, Feinberg and Keenan, 2005). On the other hand, Gillihan and Farah (2005) underscored the insufficiency of evidence for the neural basis of domain-general self-processing. Furthermore, Sugiura et al., 2005, Sugiura et al., 2006a and Uddin et al. (2007) took the view that activation during visual self-recognition may be domain-specific; that is, activation is specific to the recognition of self-face or representations related to the self-body, rather than a general module for self-processing.

The self-specific involvement of these cortical regions has so far been demonstrated only in recognition of the face and body. To our knowledge, nobody has examined whether recognition of a self-relevant stimulus other than the face or body activates the same areas that show self-face-specific activation. Thus, this study compared cortical activation during recognition of self-face and self-name, using functional magnetic resonance imaging (fMRI). We chose one's own name as the non-face self-relevant stimulus, because the self-name represents the self at a comparable level to the self-face in our social life, but is basically irrelevant to one's own face or body. If self-name recognition activates the self-face-specific areas, it supports the domain-general hypothesis and a role for these areas in a general module for self-processing. If it does not, a domain-specific role for these areas is suggested.

A few studies have reported activation during the hearing of a spoken self-name (Kampe et al., 2003, Perrin et al., 2005, Carmody and Lewis, 2006); however, it was not evident whether the observed activation was self-specific, because these studies compared the response to the self-name to that to unfamiliar names. It is difficult to draw self-specificity from this comparison alone, because the self is likely to carry attributes related to person familiarity, such as autobiographical memory and affective value, which would be shared by familiar people. In fact, cortical responses common to the self-face and personally familiar faces, relative to an unfamiliar face, were observed in the posterior cingulate cortex (Sugiura et al., 2005). Thus, to investigate a self-specific cortical response, it is essential to exclude the effect of person familiarity by comparison with responses to a familiar person.

With regard to the comparison between the self and the familiar person, it is also interesting to examine the reverse activation pattern; that is, the lower activation for the self than for the familiar person. Such an activation pattern may also be a characteristic of the cortical response during self-processing. This pattern was previously reported for the bilateral temporoparietal junction during face recognition (Uddin et al., 2005, Sugiura et al., 2005, Devue et al., 2007, Morita et al., 2008), although its relevance to the self-processing mechanism was not discussed by these authors.

Accordingly, in this study, we compared activation during recognition between the self and a friend, using face and name stimuli. A name was presented visually, rather than in a spoken voice, for compatibility with the face condition and to prevent the subject from perceiving the heard self-name as a request of social interaction from others (Kampe et al., 2003). Normal subjects were presented with a static picture of a face (f) or a written name (n) of the subject him/herself (S), the subject's friend (F), and an unfamiliar person (C: Control); a familiar/unfamiliar judgment was required in each trial. The total of six conditions (Sf, Sn, Ff, Fn, Cf, Cn) accordingly comprised a two-factorial design composed of factor person type (self, friend, unfamiliar) and factor stimulus type (face, name).