Paul Churchland’s Neurophilosophy

Paul Churchland, Neurophilosophy at Work, CUP, 2007.

I just finished Paul Churchland’s latest book, a collection of essays published by him during the last 10 years.  They cover many topics: how to give a neurobiological account of consciousness, why functionalism is wrong, why (non-classical) connectionism is the way to go, how to build a cognitive neurobiology of the moral virtues, how to be an epistemological reliabilist without believing in the propositional attitudes, how to account for mental content in terms of neural maps, how to support the identity theory/reductionism (using “chimerical colors”, i.e., color experiences that do not correspond to any real color), and how to vindicate realism about colors.

On most of these topics, Churchland has plenty of insightful things to say.  He spends little time discussing traditional philosophical arguments and positions.  Instead, he tries to convince the reader that his views about the mind (semantic holism, empiricism, connectionism, etc.) are more or less mandated by his interpretation of contemporary neuroscience, which he presents as if it were uncontroversial.  All is done in Churchland’s characteristically charming and eloquent way.

Of course, you shouldn’t take Churchland at face value.  His views about the mind are often not, in fact, mandated by his own interpretation of neuroscience, and his own interpretation of neuroscience is often not uncontroversial.  Nevertheless, his account is original, powerful, and ably defended.

I do have one bone to pick.  Churchland is especially sloppy when it comes to matters of computation. 

For example, in Chap. 7 he commits the “Church-Turing fallacy”, which is Jack Copeland’s term for the belief that something demonstrated or defended by Church or Turing (such as the Church-Turing thesis) entails that the mind-brain is a computing system.  Committing the Church-Turing fallacy was somewhat excusable in the 1980s, when most philosophers of mind committed it (including Paul Churchland).  But in 2007, after ten years of hammering by people like Jack Copeland, people should be more careful about these matters.  (My own effort to combat fallacious arguments for computationalism from the Church-Turing thesis, such as Churchland’s argument, may be found here.)

Another problem is Churchland’s conflation of functionalism with classical computationalism (Chap. 2).  Churchland argues against a formulation of functionalism that is indistinguishable from an especially straw-manish version of classical computationalism.  The result is both unsound and confused.  (My own effort to clarify the distinction between functionalism and computationalism, and some related matters, is forthcoming in PPR.)

In short:  Churchland is one of my heroes, but he is not always as careful as he could be.



  1. Thanks for the review.

    In short: Churchland is one of my heroes, but he is not always as careful as he could be.

    Well put.

    One problem I’d like to see addressed for his state-space semantics is what I call the ‘three point problem.’ Given a vector space with three prototypes and their metric relations to one another, it likely maps onto an infinite number of triads in the world. With his internalist-leaning semantic theory, where the metric structure of the prototypes is supposed to be sufficient for content-fixation, this becomes a problem.

    I’ve tried to convince him to supplement the bare-bones metric-space account with some information from the world, but he really didn’t like that idea. I think a marriage of Dretske and Churchland would be quite natural, and more “naturalistic”, than Churchland’s present perspective.

    When we (neuroscientists) study how neural assemblies represent the world, we don’t just study the neurons and how they relate to one another and then try to find things in the world to map onto them. We quantify the relationship using information-like quantities to more directly see how the world and brain hook together. Sure, we often look at the metric structure of the neuronal space and see how it maps onto the metric structure of the world-space, but this is typically done in conjunction with a more Dretsky-like approach.

    If the auditory cortex had a map that looked like the map of color space, we still wouldn’t say it was representing color space if we didn’t see responses to visual stimuli, but did see responses to auditory stimuli!

  2. gualtiero piccinini


    I couldn’t agree more.  This is one place where Churchland’s interpretation of neuroscience is at odds with mainstream scientific practices. 

    Do you have a sense of why Churchland rejected your suggestion?

  3. Eric Thomson

    If memory serves, I think he lays out his problems with informational semantics in the book. I never found them compelling. E.g., the problem of handling errors with information-based accounts. That’s one that Dretske and others have dealt with at length.

    It’s strange, as if you look at he and Pat’s wonderful little article ‘Stalking the wild epistemic engine’ it is a much more information-friendly, and more ethological, perspective. I think Old Paul and New Paul need to kiss and make up to make a Super Paul.

  4. Eric Thomson

    I’m not sure, as I don’t know anything about Clark and Chalmers’ active externalism. But clearly it needs to be supplemented with something, likely some sort of facts about the world rather than just the nervous system.

    Has this ‘three point problem’ been written up in a paper yet? It is an obvious problem that is pretty easy to fix. Quick and easy publication for any philosophers out there! Better hurry before I point my philosophically obtuse mind at it.

  5. I regret to say that although Clark´s and Chalmers active externalism deals with the long held problem of making sense of how our internal states have meaning pointing to the external world as a means of content ascription, it is not as sophisticated as Churchland´s neurosemantics.

    But as you say Clark´s and Chalmer´s active externalism complements Churchland´s neurosemantics, because they take into account the world rather than merely facts about the nervous system. So if we want a good neurosemantic theory, we need some sort of Clark´s active externalism on a par to Churchland´s view on representation.

    I have no idea about any paper in the literature on the “three point problem” so you are obligued to publish and advance the field!

  6. Eric Thomson

    Dretske does a good job with including the world, and since I already understand and have read Dretske it would be a much quicker study.


Comments are closed.