Splitting Concepts

Gualtiero Piccinini and Sam Scott’s paper “Splitting Concepts” together with my reply
“How to Split Concepts: A Reply to Piccinini and Scott” is now published in Philosophy of Science (73, 4).

This is a very nice exchange and I hope it will get other people interested in this issue.

Edouard Machery


  1. Mike Wiest


    I have only skimmed the Machery paper, but I have a basic question. The debate seems to be about whether we need multiple independent brain processes to account for conceptual knowledge in each task domain, though I thought the central question was whether the various kinds of concepts have any nontrivial properties in common.

    And I thought that the traditional idea of concepts was that they are a form of representation that can be easily generalized and compared to other representations and abstracted and so on–in a word, I thought concepts were conceived as the format for conscious reasoning (although they can be used implicitly too).

    So, regardless of how many different representational strategies we find implemented in the brain, if they are accessible to conscious manipulation, doesn’t that make them “conceptual” representations?

    I don’t know how people define “natural kind” these days, but might not “consciousness-accessibility” be the distinguishing mark of “concepts”?

  2. Mike,

    Thanks for your interesting comment.

    1. You are right that the central issue is, as you put it, “the various kinds of concepts have any nontrivial properties in common” (more on this in point 3 below). But, this issue ties in two ways with whether we have several categorization processes, several induction processes, etc. (I assume that this is what you mean by “whether we need multiple independent brain processes to account for conceptual knowledge in each task domain”.) For the sake of simplicity, I will focus on categorization.

    First, suppose that, as I believe is in fact the case, we have several categorization processes (say a prototype-based and a rule-based categorization process), these categorization processes are markedly different from each other, and each kind of concepts is used in a distinct categorization process (prototypes in the prototype-based process and rules in the rule-based process). Then, each kind of concepts will have distinct functional properties–that is, roughly, they will be used differently by cognitive processes. For instance, prototypes and rules will be used differently by categorization processes. So, functional properties won’t be generalizable across all concepts.

    Second, the issue of whether we have several categorization processes ties in complex ways with the issue of whether one should think of the different kinds of concepts (say, a prototype of dog and a set of exemplars of dogs) as *parts* of concepts (say, two parts of the concept of dog rather than being two concepts of dog). I discuss this issue in the third chapter of my forthcoming book.

    2. Some people have indeed tied concepts and accessibility to consciousness. In some respects, Prinz and Dennett come to mind. But I don’t think psychologists want to tie concepts to consciousness-availability. Indeed, they are happy with concepts that are acquired implicitly and used implicitly. I discuss this issue in the first chapter of my forthcoming book.

    3. It is very important to distinguish having “a distinguishing mark”, as you put it, and being a natural kind. Many cases are well-defined, but are not natural kinds. Consider, for example, the class of supernatural objects or the class of objects that weigh more than 20 pounds. Perfectly defined kinds (they have a distinguishing mark), but they are not natural kinds. Similarly, concepts have a mark, since they form a non-empty functional kind. But they are not a natural kind, or so I argue.

    For more on natural kinds, see my 2005 paper in Philosophy of Science, section 1 and the chapter 8 of my book.


  3. Mike Wiest

    Hi Edouard,

    Thanks for your reply. I’ll refer to your numbered points:

    1. Ties between categorization processes and concepts:

    (i) I can certainly accept that we have multiple categorization processes with different functional properties, so I can accept that corresponding concepts would also have different functional properties. Otherwise there wouldn’t be much point in having multiple concepts. I suppose they could also have certain functional properties in common, though, that would justify our calling them all “concepts.” (For example, maybe any pair of concepts can in principle be voluntarily compared explicitly.) Do I understand correctly that you ultimately reject this possibility?

    (ii) This issue about parts of concepts sounds intriguing…what’s your book called? Is it summarized on this blog somewhere?

    2. Consciousness:

    I’m surprised to hear that Dennet tied anything to consciousness! Anyway, I’m currently reading “The Foundations of Mind” by Jean Matter Mandler and she uses the word “concept” explicitly in connection with some kind of conscious reasoning faculty. (Wasn’t this the original meaning of the word as used by Kant, Leibniz & co?) She makes a pretty compelling case that infant cognitive development data support a theory that distinguishes “percepts” (which relate according to perceptual similarity) from “concepts” (which relate according to semantic or functional similarity).
    Do you make a distinction between conceptual representations and others, like maybe perceptual representations? Or are all mental representations concepts? (If this is in the posted paper forgive me and perhaps you could steer me to the relevant part of the paper?)

    3. Natural kindness:

    I can see that a mere distinguishing mark would not suffice to establish a natural kind, which I basically understand to mean an objective type. Intuitively though (and perhaps naively), it seems to me that we may not have any stronger candidate for a natural kind than consciousness itself. Would your formulation (which I hope to read) allow the possibility that conscious states or the conscious “aspect” of a state may form a natural kind? Or do you feel that experience’s ineliminable subjectivity is incompatible with an objective ontological status?

    Best regards,
    Mike Wiest

  4. Mike,

    1(i) I doubt that concepts used in different categorization processes have *non-trivial* properties in common. By non-trivial properties, I mean the kind of properties psychologists are interested in and that are discovered empirically.

    (ii) The book is called Doing Without Concepts. I will send the complete draft to Oxford UP very soon.

    2. Dennett tied concepts to self-consciousness in his short 1996 book (Kinds of Minds).

    Not all mental representations are concepts. Concepts are used in the processes that underlie our higher cognitive competences, such as categorization, induction, deduction, and so on.

    3. The nature of natural kinds is of course a controversial topic. I have my own views, loosely based on Boyd’s account. I propose roughly that natural kinds are those kinds that yield numerous generalizations that are underwritten by causal mechanisms. so, having one common property, even an important one such as being consciousness-accessible, would not make concepts a natural kind. What matters is whether concepts yield numerous causally grounded, scientifically relevant generalizations.



Comments are closed.