The Adaptable Mind: What Neuroplasticity and Neural Reuse Tell Us about Language and Cognition - John Zerilli 2021
New Directions in Faculty Psychology
Saving Faculty Psychology: Debunking the Argument from Multiple Realization
The discovery that traditional psychological faculties are implemented by neural systems consisting of shared domain-general components does not make traditional faculty psychology go away. On the contrary, as I argued in § 4.2.3, these high-level cognitive systems—I decline to call them modules—have an important role to play in our ongoing quest to understand the mind. The ramifications of reuse will in fact serve to sharpen our understanding of what makes these systems tick and reveal the extent of functional and semantic inheritance between traditional tasks and task categories. A mature twenty-first-century faculty psychology, therefore, has a lot to look forward to—and essentially nothing to fear. But if the work required to understand behavior in the light of neural reuse is going to get done by anyone, it will have to get done (one would think, in the first instance at least) by twenty-first-century faculty psychologists! All those with a wealth of experience investigating the higher faculties will simply have to get on board if the endeavor is to have any chance of success. In other words, the future of faculty psychology depends in no small part on the productive collaboration between neuroscience and psychology.
Unfortunately, there is a potential obstruction in the way of just this sort of intertheoretic collaboration. Inasmuch as high-level cognitive systems are understood to be software systems, pitched at the level of algorithmic or computational psychology, the thought arises that these systems can be investigated with a kind of disciplinary immunity from neuroscience, since (it is supposed) “[n]o amount of knowledge about the hardware of a computer will tell you anything serious about the nature of the software that computer runs” (Coltheart 2004, p. 22). The idea proceeds at least in part from the assumption that psychological processes are multiply realized. The multiple realization (MR) hypothesis asserts, at its baldest, that the same psychological state may be realized in neuroscientifically distinct substrates (Polger 2009). Hilary Putnam’s (1967) ingenious suggestion that “pain” is likely to be a multiply realized kind (MR kind) rather neatly captures the thought here—while presumably both mammals and mollusks experience pain, they pretty obviously lack the same neurobiological basis. MR was played against a popular philosophical theory of mind in the 1960s that attempted to identify mental states with neural states. Since MR implies a many-to-one mapping from neural states to mental states, if it is in fact true that mental states are multiply realized, it follows that no clear identity relationship can hold between them. Thus it was that many of those who advanced MR rejected mind—brain identity as a viable philosophical theory. As Bechtel and Mundale (1999, p. 176) framed the issue, “[o]ne corollary of this rejection of the identity thesis is the contention that information about the brain is of little or no relevance to understanding psychological processes.” Fodor et al. (1974, p. xiv) are blunt: “it is not only far from clear that neurology provides a useful theoretical vocabulary for psychology, it is even unclear that neurological theories will ever have much practical implication for the development of psychological models.”
Another development of the argument asserts that even if MR does not hold true among existing biological systems, the possibility that cognitive states might be shared by built artifacts or alien life forms having very different physical structures in itself establishes the salience of the MR thesis. This version of the argument led some philosophers of artificial intelligence (AI) to embrace the further metaphysical claim that “mental processes are the operations themselves, and are not identified with whatever biological or other substances realize them” (Bechtel & Mundale 1999, p. 176). Following Bechtel and Mundale’s lead, and the contours of the recent debate surrounding MR, I shall not address this version of the argument or the metaphysical claim it inspired here. The possibility that artifacts could have mental states is just the possibility that the identity theory is wrong, which is precisely the claim in dispute (Polger 2009, p. 459). Instead, I shall use this chapter to consider, admittedly quite briefly, the empirical claim—that the MR hypothesis can be verified having regard to existing organisms—since no doubt it is this claim that has played the lion’s share in encouraging a downbeat attitude to evidence from neuroscience in some quarters.
Daniel Dennett (1991, pp. 254, 270, n. 2), for instance, laments the functionalist’s penchant for “boxology”; i.e., drawing diagrams that install component functions in separate boxes, “while explicitly denying that these boxes have anatomical significance.” While he concedes that “in principle” it may be a good tactic, and one he himself has employed, “it does tend to blind the functionalist to alternative decompositions of function, and particularly to the prospect of [neural reuse].” Small wonder, then, that he calls for a “better vision, anchored in a positive acceptance—as opposed to a hysterical dismissal—of the foundational facts of functional neuroanatomy” (see also McGeer 2007; Hardcastle & Stewart 2009, p. 194; Karmiloff-Smith 1994, p. 702; Bechtel 2008a, p. 990; Gerrans 2014, pp. 22—23). In § 8.2, I put the two primary empirical arguments in favor of MR under pressure, as well as offer a survey of recent arguments skeptical of the MR hypothesis.
But let me be clear at the outset. My position is in no way hostage to the fortunes of MR, however one cares to define it. There are other ways to argue for the pertinence of neuroscience to psychology that need not presuppose type identity among cognitive and neural states. I can think of at least three. For one thing, it does not pay to have an uncompromisingly rigid understanding of the reduction relationship. John Bickle (2010, pp. 250—251; 1998, p. 30) appeals to examples of successful reduction from the history of science that happened to involve MR kinds as among the posits of the reduced theories, including examples of reduction involving radical MR—that over distinct physical states occurring within the same token physical system at different times. The reduction of temperature to mean molecular kinetic energy could be considered to involve radical MR, since classical thermodynamic kinds like temperature are in truth macroscopic states multiply realized over different microstates of the same macroscopic system over time. Yet the reduction of classical thermodynamics to the kinetic/corpuscular theory of matter is “the textbook example of scientific intertheoretic reduction.” In another vein, Jaegwon Kim (1992), Larry Shapiro (2000), and Colin Klein (2008) have each drawn attention to a significant dilemma confronting the MR advocate. If a given functional kind is not multiply realized, the traditional argument for its autonomy and irreducibility falls away. If, on the other hand, the kind is multiply realized, the kind will not be a proper scientific kind; i.e., of the sort that can enter into laws. “Brittleness” might be a multiply realized property, but glass, steel, and biscuits are each brittle in their own way: there can be no general science of brittle things. Ranging over such genuinely diverse physical realizations means the kind will not enter into laws (i.e., exhibit lawlike or projectable properties), except for those that are true analytically—such as All mousetraps catch mice, and All eyes see—and this in turn “undercut[s] the traditional motivation for admitting functional kinds into the ontologies of the special sciences” (Shapiro 2000, p. 637). Thus, if mental states really are multiply realized, neuroscience will matter very much indeed—not so much because psychology could not do without it, but because psychology’s claim to be a traditional science would be open to question. This is only the most skeptical conclusion one could draw, but the point is well taken (see Couch 2009, pp. 262—264, for criticisms, however). Finally, and somewhat trivially, multiply realizable does not mean infinitely realizable. Cognitive hypotheses will always have implications for realizers—mousetraps cannot be made of paper after all. A functionalist psychology, this is to say, can proceed only within the biophysical limits that its own constructs impose, and it ignores evidence of implementation to its peril. Take modularity as an example. A commitment to modularity standardly entails a belief in the functional dissociability of at least some cognitive capacities; but this, as we saw in Chapters 4, 5, and 7, is just not the sort of feature that the evidence of reuse makes available, at any rate in a straightforward manner. Of course, neural hypotheses may themselves be disconfirmed by evidence coming from other branches of the cognitive sciences, including psychology, so the constraints here are genuinely bidirectional and intertheoretic. But MR or not, there is simply no way of getting around the neuroscience (McGeer 2007).
Be that as it may, given the recent tide of empirical challenges to the MR thesis, and because MR has proven itself to be an occasional stumbling block in the path of those committed to the autonomy of the special sciences, I have considered it worthwhile saying at least something on the subject. While the more austere school of functionalism admittedly no longer enjoys the following it once had—and mainstream functionalists today would hardly dismiss neuroscience on the basis of psychology’s autonomy of neuroscience—pockets of the austere school do survive under the guise of cognitive neuropsychology and related fields (see McGeer 2007 for detailed analysis and criticism).