social bonds have been traditionally conceived to rely on two main streams of processing (Eslinger 1998): one more cognitively ori- ented and dependent on the dorsolateral prefrontal cortex, and the other more emotion-oriented and mediated by anterior tem- poral, limbic, and orbitomedial frontal cortical systems. In a neu- ropsychological study of humans with acquired lesions of the frontal lobe (Grattan et al. 1994), comparisons between quantita- tive measures of cognitive flexibility and empathy indicated that left and right dorsolateral frontal lesions caused deficits in both cognitive flexibility and empathy (with correlations ranging as high as r = .81). In contrast, orbitofrontal lesions profoundly affected empathy but did not alter cognitive flexibility, suggesting different mechanisms of pathophysiology for empathic change that may have its basis in deficient automatic tagging of salient stimuli or evocation of autonomic-visceral states. The latter pattern has also been reported in patients with lesions of the insula and deep white matter of the frontal lobe, likely disconnecting orbitofrontal, amygdala, and insula interconnections. Despite the lack of auto- matic emotional responsiveness, these individuals can be trained to use cognitive mechanisms for recognition of salient facial, vo- cal, and body expressions, and for the verbal expression of their concern. Thus, frontal regulation of social conduct favors the view that the prefrontal cortex may need further functional fractiona- tion, with polar, orbital, medial, dorsolateral, and deep white mat- ter sectors mediating distinct, but complementary roles in the emotional-cognitive regulation of social behavior. In our experi- ence, early damage to polar, mesial, and dorsolateral regions dis- rupts the developmental acquisition of automatic and learned em- pathy, moral judgment, and interpersonal conduct (e.g., Eslinger et al. 1992; 1997). The cerebral correlates of specific dimensions of moral and em- pathic behaviors need further study (Grattan & Eslinger 1989). P&deW’s article offers a solid conceptual frame against which spe- cific hypotheses and models can be put to empirical testing in nor- mal individuals, as well as in patients with a variety of neuropsy- chiatric disorders which express themselves as disorders of empathy, social, and moral behavior. The mirror matching system: A shared manifold for intersubjectivity Vittorio Gallese, Pier Francesco Ferrari, and Maria Alessandra Umiltà Istituto di Fisiologia Umana, Università di Parma, I-43100 Parma, Italy. vittorio.gallese@unipr.it ferrari@biol.unipr.it maumilta@unipr.it Abstract: Empathy is the phenomenal experience of mirroring ourselves into others. It can be explained in terms of simulations of actions, sensa- tions, and emotions which constitute a shared manifold for intersubjectiv- ity. Simulation, in turn, can be sustained at the subpersonal level by a se- ries of neural mirror matching systems. The article by Preston & de Waal (P&deW) has the merit of rais- ing the issue on how different and multifarious social behaviors such as emotional contagion, helping behaviors, imitation, and cognitive empathy may be reconciled within a unified explanatory framework. Their proposal is that empathy can constitute such a framework, provided that it is “construed broadly to include all processes that rely on the perception-action mechanism” (see tar- get article, sect. 1.1.3, Fig. 1 caption). Unfortunately, the discussion of these interesting issues is, in our view, most of the time too vague, and sometimes even con- founding, so that it is difficult to draw any firm conclusion on the validity of the proposed model. The definition of empathy given by P&deW seems to oscillate between a “superordinate category” of behaviors, and a process underpinning the same different types of behavior. They intro- duce the Perception-Action Model (PAM), which is presented as a “superordinate class,” capable of including and subsuming dif- ferent types of behaviors and effects, among which empathy is also listed (see their Fig. 1). This makes it almost impossible to under- stand whether in the authors’ view empathy and PAM are the same thing or not. Furthermore, the term PAM is used inter- changeably to denote either a model (as the acronym seems to suggest) or a basic (neurophysiological?) level of description. Such a use of the term makes several statements hard to understand. How can a model possibly interact with specific brain functions? (e.g., see the Abstract where it is stated that: “The interaction be- tween the PAM [Perception Action Model] and prefrontal func- tioning can also explain different levels of empathy.”) Models should explain, or interpret functions, not interact with them. The broad and general sense attributed by P&deW to the term empathy fails to provide a coherent picture of the mechanism that is at the basis of their model. There is confusion here – if not even a category mistake – about the chosen level of explanation. In our opinion, the term “empathy” makes sense only if used to denote a phenomenological level of description: the one respon- sible for the sense of similarity that we experience anytime we con- front ourselves with other human beings, and sometimes even with animals. Empathy is deeply grounded in the experience of our lived-body, and it is this experience that enables us to directly recognize others, not as bodies endowed with a mind, but as per- sons like us. Actions, sensations, and emotions experienced by oth- ers become meaningful to us because we can share them with oth- ers. How can such a sharing of experiences be possible? We need to introduce a second level of description, one pertaining to the functional mechanism that enables such an experience to occur. This functional mechanism may be constituted by simulation. Simulation mechanisms are a possible way for a given organism to control and model its performances. By modeling a given process, our brain provides a simulated representation of the same process that can be used to produce it, on the one hand, and to decode it when performed by someone else, on the other. We suggest that these “as if ” simulation mechanisms may subsume a wide range of processes as diverse as action perception and imitation (as simu- lation of the observed action; see Gallese 2000a; 2000b; 2001; see also Rizzolatti et al. 2001), emotion perception (as simulation of the perceived emotion; see Adolphs 1999; Adolphs et al. 2000; Commentary/ Preston & deWaal: Empathy: Its ultimate and proximate bases BEHAVIORAL AND BRAIN SCIENCES (2002) 25:1 35 Figure 1 (Eslinger et al.). Results of fMRI study of normal vol- unteers making explicit moral judgments, showing activation pri- marily of frontopolar cortex, medial frontal gyrus and related re- gions