This page contains a provisional statement of my epistemological viewpoint, which may or may not be of interest to the philosophically inclined. I wrote it in response to an interesting article defending functionalism and the use of psychological constructs, which is available here (though I first read it in preprint):
Practically speaking, my epistemology suggests the utility of psychological research and theory that is guided by psychological constructs but constrained by neuroscience, with the understanding that the two domains describe the same phenomena at different levels of resolution or abstraction.
The following requires some understanding of the terms "role" and "realizer." "Role" refers to something's function, what it does (e.g., money). "Realizer" refers to the specific entity that plays a given role (e.g., a dollar, a peso). Basic physical entities (e.g., subatomic particles), are often considered to be the real or ultimate realizers of any role, but I argue that it's not so simple:
Deep Functionalism: A Response to Ross and Spurrett
A deeper functionalism might assert that because all of our explanations are partial and imperfect (which I think we can accept as a pragmatic necessity, even if the question of whether this must be so in principle remains open), even the explanations of physics are essentially functionalist. This is not incompatible with reductionism, in one sense: forms of less abstract, more general, explanation (e.g., physics) could still be used to explain the same phenomena that are explained in more abstract, more specialized, ways (e.g., psychology), though usually with a loss of efficiency so massive as to render them pragmatically useless (imagine having to specify the interaction of every particle directly involved in any act of perception, in both the organism and its environment).* In another sense, however, it illegitimates the chauvinism of reductionism, which claims hegemony for more basic (typically physical) causal explanations. From my perspective, it makes as much sense to talk about a “role” causing an effect as it does to talk about a “realizer” causing an effect because there can be no hard and fast line between roles and realizers. All of our explanatory constructs (even those of physics) are ultimately roles rather than realizers because of their partialness and their imperfection. The actual complexity of the world (i.e., that which exists ontologically) is always infinite relative to our capacity to model it (again, perhaps in principle, but at least from a pragmatic perspective). Unless we naively thought that our model of a photon captured every possible thing there is to know about the phenomena we describe with this model, we could not avoid the question of what it is that realizes the role of photon (string theory anyone?). Further, an answer to this question would not settle the matter, but would simply extend the infinite regress one step further.
All of this entails that reductionism is a concept only applicable in practice epistemologically – that is, only applicable to our explanations and understandings. The question of ontological reductionism is another one entirely, and one that is perhaps ultimately meaningless. What could ontological reduction mean? Given that all of our explanations are partial and imperfect, none of them picks out any portion of the universe in all its detail. This does not necessarily mean, however, that we have no access to some Kantian thing-in-itself; rather it means that whatever access we may have must always be limited. Nonetheless, this limitation has important consequences. If we cannot grant the ontological sanctity of basic physical constructs, do we have grounds for saying that particles x, y, and z are what “really” exist, and all other phenomena, ontologically speaking, are merely the interaction of these particles? It seems to me that we do not. Rather, we must grant equal potential ontological status to descriptions of phenomena at any level. Any of them may be picking out genuine patterns of regularity in the structure of the universe (cf. Dennett, 1991, "Real Patterns").
Applying the idea of epistemological reduction (perhaps “epistemological unification” would be a less objectionable phrase) – which refers to the harmony possible, in principle, between different levels of explanation – to the relation between psychology and neuroscience, we can assume that the mental description M is a description of the same phenomenon as the physical description P, but at a different level. The brain state, P, doesn’t cause you to think the thoughts, M; the brain state is you thinking the thoughts (assuming M and P are coextensive in time). Inasmuch as there is some hierarchy of levels of explanation, in which things seem intuitively to become more realizer-like as we move down through biology and chemistry toward physics, this hierarchy does not indicate ontological primacy, but rather spatio-temporal specificity. Brain states may change without changing mental states, but mental states may not change without changing brain states (this state of affairs is often called “supervenience”; the mental supervenes on the physical). Does this indicate that the brain states are more real? No, just that they describe more specifically localized patterns. Again, remember that, in practice, neither mental nor physical description will capture the phenomenon in all its detail.
Causal assertions are thus equally valid at either level (mental or physical), but what about across levels? M1 may be said to cause M2 and P1 may be said to cause P2, but can M1 cause P2 or P1 cause M2? Whether we allow the latter kind of statement appears to be merely an aesthetic matter. It may be less tidy to describe causes across levels, but if epistemological reductionism holds, nothing makes such a crossing inherently unintelligible. An example: if I am anxious, my being anxious is equivalent to a particular brain state (more accurately, a set of brain processes). Those brain processes do not cause my anxiety; rather, they are my anxiety. Being anxious may then cause me to think less clearly. My muddled thinking is another particular set of brain processes. This chain of events can be validly described in several other ways, in addition to (1) "Being anxious caused me to think less clearly." These include: (2) "The first set of brain processes caused the second"; (3) "Being anxious caused the second set of brain processes"; and (4) "The first set of brain processes caused me to think less clearly." This illustrates my contention that a description of patterns at any level, or across any levels, may validly discuss causes. Further a description of causes across levels does not necessarily deny the closedness of physics, unless we assume that a mental description is somehow fundamentally different from a physical description. Rather, let me suggest that all descriptions are physical descriptions (though “physical” could perhaps fruitfully be replaced with some more general term), in the sense that all descriptions attempt to describe patterns, of equal ontological status, in the same universe.
The example in the previous paragraph raises an important issue, which I want to acknowledge without going into too much detail. One might argue that a crucial difference between anxiety and its associated brain processes is that anxiety can be subjectively experienced, whereas the brain processes can be detected only by various instruments. I would agree, but I don't think that this difference renders these two things fundamentally distinct, ontologically -- rather it indicates two different methods for learning about the same events.
C. G. DeYoung, 2003, Christchurch, New Zealand
*A later update: I now think that even the possibility of reductionism in this first sense must be qualified because of ideas that I first encountered in Jeffrey Gray's (2004) remarkable book, Consciousness: Creeping up on the Hard Problem. The only sort of low-level physical explanation that could be used to "explain" biological phenomena would be historical (i.e., recounting the positions of molecules and their constituent particles and the trajectories and interactions that led them through some portion of their history to their present state). Such historical explanations would be not only impractical but also unsatisfying, relative to the kinds of explanations of non-biological phenomena that physics offers, which are based on laws that describe the behavior of a particular system at the current point in time without reference to its history. A historical explanation of biological phenomena would involve recounting evolutionary history in low-level physical terms that would be meaningless at the more abstract evolutionary and biological levels. The laws of evolution and the cybernetic laws that govern the behavior of organisms are not reducible to physical laws for a very simple reason: there is no explanation based on the laws of physics that explains the order of molecules in a given strand of DNA. In fact, the very reason that DNA can function as the basis of life is the fact that its base pairs are physically equally capable of being sequenced in any order. Hence, laws of biology and information processing are not reducible to physical laws, though of course the phenomena they describe supervene on their physical substrate and obey the laws of physics.