Mundale, J. (2026). Multiple Realizability: Still Something After All These Years. Biological Theory. doi:10.1007/s13752-026-00538-7
[Abstract]Arguing about the types and tokens of neuroscience and psychology has provided philosophical sport for decades. Type, or strong identity theory, was the target of Putnam’s original multiple realizability (MR) argument and asserts that because mental types are multiply realized by several different physical types, no lawlike identity relation between mental and physical types exists. Both psychology and neuroscience have made rapid progress in the intervening years, allowing for more empirically-guided ways of exploring their mutual explanatory value to each other. I briefly discuss the historical impact of MR, offer a diagnosis of its endurance, and revisit an earlier position I co-authored with William Bechtel. I then consider MR in light of neuroplasticity and localization, both commonly taken to support the MR argument. I argue that these not only fail in this support, but instead, show the advantage of a more empirically-grounded approach based on structure–function relationships. In this light, MR itself may even be regarded as a kind of neuro-psychological discovery principle; discovering a function that appears to be subserved by multiple neuromechanisms may point us toward interesting commonalities, rather than differences, of neurological structure.
[Citing Place (1967)]  
Citing Place (1967) in context (citations start with an asterisk *):
Section Introduction
* Both Jerry Fodor and Hilary Putnam had achieved notoriety in philosophy of language before turning to philosophy of mind. Perhaps it was this background that inspired Fodor to apply, and Putnam to adopt, an old linguistic distinction between types and tokens into new domains; namely, to the identity relations between psychology and neuroscience. Although different mind–brain identity theories had already been advanced in the 1950s and 1960s by philosophers such as U.T. Place, Herbert Feigel, J. J. C. Smart, D. M. Armstrong, David K. Lewis, and others, Fodor applied the notion of types and tokens to distinguish between two broad kinds of identity theory: strong, or type identity theory, and weak, or token identity theory. Though Putnam (1967) first made the multiple realizability (MR) argument without the use of type–token terminology, MR was quickly recast by Fodor (1974), then widely adopted by Putnam and others as an argument against type identity theory, which left token identity theory as the (apparently) plausible alternative.
This is not to claim that the earlier conceptions of identity theory were themselves all the same, nor that they were equivalent to either Putnam’s working conception of strong identity theory or to Fodor’s newly coined type and token identity theories; they were not. Place, for example, explicitly criticized Putnam on this point (Place 1967, p. 56). Putnam’s 1967 article, where the MR argument originated, is first motivated by an analysis of whether pain is best conceived of as a brain state, functional state, or behavioral disposition. Place’s response is that these are all false, because pain is a process, not a state. He writes: “The theory as I understand it is a theory about mental processes, not a theory about mental states, and having a pain on this view is a mental process, not a mental state. And if it is not a mental state it cannot be a brain state.” (Place 1967, p. 56). Therefore, Putnam’s conception of identity theory (brain state theory) already differs from a well-known, earlier version of identity theory (Place’s), and exchanging such terms as states, processes, or predicates in favor of types or tokens is yet a greater, and more ambiguous kind of change. Nonetheless, the type–token language stuck.
Section MR and Linguistic Confusion
* [Owen Flanagan (1991)] writes: “… you cannot possibly have your type–type identity statements in place if you do not have the typologies of both the science to be reduced and the reducing science down pat. For example, if you have not clearly worked out, within your psychology, what love is, and if you have not clearly worked out, within your neuroscience, what qrxt-firings are, then you don’t stand a chance of credibly asserting that love is a qrxt-firing.”(1991, p. 61) This would seem to be a mark against reduction, and contextually, that was Flanagan’s intent. For the present purposes, however, he illustrates a problem with the lack of taxonomy. We still don’t have our typologies worked out in either psychology or neuroscience, and it’s not clear what it would look like if we did. Suppose, though, for the sake of argument, that the typologies have been completed. We still have terminological difficulties, however, and one way to illustrate this is to recall Durso’s critique of autonomy, mentioned above. He points to several micro-level mechanisms that are widely conserved in the nervous systems across many biological taxa. To be clear, his point is not to criticise terminology, but rather to challenge the autonomy of psychology. Nonetheless, his examples are useful toward several ends. With regard to conserved pain mechanisms, he observes that nociceptors are “present in nearly all species whose pain mechanisms have been studied. Nociceptors are sensory neurons that detect noxious stimuli” (Durso 2025, p. 16). As philosophers of mind doubtless know, there are also different kinds of nociceptors that detect different kinds of pain stimuli and have different speeds of neural transmission, as Durso points out (C-fibers, A-delta fibers, and others). So, in a time of completed taxonomies, what might pain typologies look like? If John has foot pain because someone stepped on it, and his dog, Rover, has tail pain because someone stepped on that, are these of the same type at the neurological level? Assuming Rover and John both have similar nociceptors that carry out the same function, using a similar mechanism with regard to pain physiology (this having all been worked out in a completed taxonomy), we might say, “yes”. Foot-stomped and tail-stomped pain are of the same type in virtue of being implemented by the same nociceptive mechanisms. On the other hand, Rover doesn’t have a foot (he has a paw), and John doesn’t have a tail. These realizers are necessary in order to have foot pain or tail pain, and they are different. Similarly, Rover’s and John’s spinal cords and brains will be different, at the gross level, and similar, at the neurochemical level. C-fibers may carry pain detection to the brain, but once there, other mechanisms will be involved in the realization of the pain state (or process, as Place would have preferred to say), and there will be further kinds of sameness and difference. In this particular instance, the issue is not about whether pain is multiply realized, but instead, how do we determine if pain mechanisms are of the same type? John’s foot pain and Rover’s tail pain could be said to be token instances of the more general type pain, or mammalian pain, or maybe pain should be a supertype, with numbered subtypes all the way down to John and Rover themselves. Alternatively, we could have nested type–token relationships, as suggested with the backpain and fastener examples, above, where what is a token in one instance is a type in another.