Simulation trouble

Abstract

I present arguments against both explicit and implicit versions of the simulation theory for intersubjective understanding. Logical, developmental, and phenomenological evidence counts against the concept of explicit simulation if this is to be understood as the pervasive or default way that we understand others. The concept of implicit (subpersonal) simulation, identified with neural resonance systems (mirror systems or shared representations), fails to be the kind of simulation required by simulation theory, because it fails to explain how neuronal processes meet constraints that involve instrumentality and pretense. Implicit simulation theory also fails to explain how I can attribute a mental or emotion state that is different from my own to another person. I also provide a brief indication of an alternative interpretation of neural resonance systems.

Acknowledgements

Part of the research for this paper was done at the Center for Subjectivity Research at the University of Copenhagen where I was Visiting Professor and supported by the university's Research Priority Initiative on Body and Mind.

An earlier version of this paper, entitled “La perception d'autrui en action,” was presented as a public lecture in the series Fondements cognitifs de l'interaction avec autrui. Collège de France (February 2006). My thanks to Alain Berthoz, Jean-Luc Petit, and Alvin Goldman, and to three anonymous reviewers for help in clarifying some of these issues.

Notes

¹This concept is developed in phenomenologists like Scheler (1923/1954), who calls it “primary perception” (e.g., p. 10) and Merleau-Ponty (1964).

²Narratives that frame our social interaction are different from the kind of theories to which TT appeals. The latter involve a reference to unobservable entities, that is, mental states which are theoretically postulated entities comparable to the black holes of astrophysics. If Bruner (1986) is right, narrative is a particular mode of thinking that relates to the concrete and particular; it takes context to be primary in the determination of meaning. In contrast, the theoretical is concerned with the abstract and general, and, in this sense, abstracts away from the particular context in favor of propositional attitudes. See Gallagher and Zahavi, Chapter 9 (in press) for further discussion.

³I'm using Goldman as a representative of explicit simulation, and this was certainly his view in his earlier papers (e.g., 1989, 1992). In his most recent work, Goldman (2006) defends a hybrid version of ST. He distinguishes between high-level and low-level mind reading. I take this distinction to be compatible with the distinction between explicit and implicit versions, respectively. Low-level simulation is “simple, primitive, automatic, and largely below the level of consciousness” (p. 113). High-level simulation includes one or more of the following features: “(a) it targets mental states of a relatively complex nature, such as propositional attitudes; (b) some components of the mindreading process are subject to voluntary control; and (c) the process has some degree of accessibility to consciousness … . The low-level prototype is the mirroring type of simulation process. The high-level prototype is the kind that uses pretense or E[nactment]-imagination” (p. 147).

⁴I do not mean to rule out all isomorphisms. This only goes to the claim that there is a necessary isomorphism between neuronal, functional, and phenomenological levels.

⁵I think it is clear that, given the history of ST, the definition of simulation relevant to ST is first worked out in accounts of explicit ST, and then is uncritically used in accounts of implicit ST.

⁶It may seem contradictory to claim in the previous paragraphs that perception is enactive, or as Noë says, “perception is action,” and in this argument to claim that the activation of the resonance system is the result of a passive elicitation, so that the motor aspect of perception does not involve our action, but is a case of us being affected by the other. I think that a fuller account of enactive perception has to be able to accommodate this passive, affective aspect of perception (see Gallagher, 2005).

⁷My thanks to an anonymous referee for this objection.

⁸This motivates Goldman's move to a hybrid model. Implicit mirroring may be considered a minimal condition for simulation, but for the full simulation that constitutes mind reading (the attribution of mental states to another) one requires higher-level processes of “classification” (which he characterizes as introspective, although not necessarily conscious introspection; 2006, p. 245 ff.) and “projection.” Although pretense is not required for low-level mirroring, it is one possible way to realize high-level simulation and mind reading (p. 49).

⁹The requirement that the simulation has to be concretely similar also raises problems for the instrumental and pretense conditions even for the explicit version of ST. If our simulation has to be concretely similar to the simulated state for it to be considered a simulation, assuming explicit instrumental control of our simulation process, how will we know how to run or control our simulation unless we already know in some detail what the other's state is like. And how do we come by that knowledge? If the answer is through simulation, then we have an infinite regress. In regard to explicit pretense, Fisher (2006), who models simulation as a reasoning process, rejects this aspect as inconsistent with simulation being concretely similar, for if we simulate a reasoning process, we are really reasoning, and not just pretending to.

¹⁰Some theorists, however, have appealed to these motor emulation processes as possible mechanisms involved in the simulation of another's action (e.g., Gallese, 2001; Hurley, 2005, pp. 181–188; see Iacoboni, cited in Millikan, 2005, p. 188, note 2). Our own motor system comparators are activated to simulate and thereby anticipate the other's action. The brain could be said to predict the other person's actions in this way. On this account the perception of the other's action is automatically informed by a subpersonal simulation; perception of action involves a loop through the motor control comparator. Can ST adopt this model of simulation? The problem, again, is that the pretense condition is not met; there is no “as if it were I” involved, and in that regard it fails to be the kind of simulation required by ST. If indeed the subpersonal emulation is neutral in regard to whose action is at stake, then it can be only a representation of an intentional action in my motor system, but not a representation of my own motor action as if it were the other's. It is not at all clear that, as Gordon (2005, p. 96) suggests, the neurons respond “as if I were carrying out the [other's] behavior” in any sense in which the “as if” registers subpersonally. Even assuming a “Who system,” a specification in my motor system that the action belongs to another is not equivalent to the specification “as if I were carrying out the action.” If this is a simulation of intentional action, it is, nonetheless, not the kind of simulation that ST needs; it may be nothing more than motor priming or emulation, or what Hurley calls mirroring (2005, p. 184).