Two Kinds of Concept: Implicit and Explicit

I’ve written Piccinini and Scott (2006), “Splitting Concepts”.

In the new paper, I argue that there are two kinds of concept: explicit concepts and implicit concepts.  Implicit concepts are an implicit version of what psychologists call prototypes, although implicit concepts (unlike prototypes according to most theories) may encode some causal information about categories.  Explicit concepts may encode statistical and causal information, but more importantly, they may encode syntactic information, definitional information, and whatever else is needed for the language faculty (in the narrow sense) to process them.  Explicit concepts are necessary for explicit cognition – the distinctively human ability to use language, represent unobservable, nonexistent, abstract, and ad hoc aspects of the world, and perform linguistic inferences.

As always, any comments would be welcome.

23 Comments

  1. Joshua Shepherd

    Gualtiero,

    Nice paper. I’m left wanting to hear more about the relationships between implicit and explicit concepts – how they interact to produce judgments, and how, maybe, the idea that they are of different kinds might explain some difficulties we have in defining certain concepts.
    Towards this end: I wonder if you could say a bit about concepts which, in one sense, are like the ‘arbitrarily abstract’ concepts you mention (justice, truth), and yet which also seem to contain implicit content – and which, perhaps because of this, are not easy to analyze. I’m thinking of concepts like ‘intentional action,’ the analysis of which has proven incredibly difficult, in part because people have conflicting reactions to cases of purported intentional action. It seems that certain implicit content related to action makes trouble for our attempted explicit definition of intentional action. Might we say, on your account, that people have something like an implicit concept of intentional action as well as an explicit concept? Or is there an (as yet incompletely formed) explicit concept that lives uneasily with implicit content related to action (but not necessarily with an implicit concept of intentional action)? Is part of the confusion here related to the distinct purposes for which these concepts are developed and recruited (perhaps people need an implicit concept of intentional action for certain social cognitive tasks, and when philosophers attempt to make the concept explicit, they depart from this need, and thus the uneasiness sets in)?

    -Josh

  2. gualtiero

    Josh,

    Thanks for your interest. Excellent questions. I’d like to hear more about the relationships between implicit and explicit concepts too! 🙂 All I’ve written on this is in the papers cited in the post, so there is a lot of additional work that could be done to develop my sketchy account. Any desire to contribute?

    That being said, I do believe that people have an implicit concept of intentional action, and later they acquire an explicit one. The implicit concept is presumably part of people’s folk psychological ability to interpret other people’s actions in terms of their mental states. And it may well be difficult to make the boundaries of the implicit concept explicit. So I find plausible your suggestion that the distinction between implicit and explicit concepts might explain some of the phenomena you mention.

    I hope you write on this topic, because I surely don’t have time to do it. If you do, I hope you send me your work.

  3. Joshua Shepherd

    Gualtiero,

    I do hope to follow up on these ideas, although this will require some more thought on my part. I like your suggestion that people acquire the explicit concept of I.A. after (and thus in some sense on top of) the implicit concept. If this is right, one would expect explicit concept competency in this case to be importantly related to the nature of the implicit concept. But I wonder if the explicit concept could come to constrain, in some sense, the use of the implicit concept, or whether the two concepts would in effect lead separate lives.

    Suffice it to say that I’m looking forward to the symposium on Doing Without Concepts, and when I get something written worth sharing, I will be happy to let you see it.

    And here’s hoping some others have thoughts on this interesting paper as well.

    -Josh

  4. Martin Roth

    Hi Gualtiero,

    Thanks for posting your paper! The following comments/questions came to mind as I was reading it:

    #1: I agree with your claim that dividing theories of concepts into “psychological” and “philosophical” may not be sufficiently motivated. However, I do think that folks working on concepts are often talking at cross-purposes and that those purposes often fall along disciplinary lines. For example (as you briefly mention), philosophers seem to be more interested in “semantic” issues than psychologists are, and that interest may explain why many philosophers don’t like “theory” accounts of concepts. Psychologists are often neutral, e.g., on the issue of whether a “theory” is supposed to be a determiner of reference, and of course it is quite popular among philosophers to deny that it is. So, if one leans towards a causal/historical account of content, then the content of a concept can (and maybe often does) diverge from one’s theory (I’m thinking here of cases where a person’s concept “of” water includes a lot of incorrect information about water, such that the referent of this concept (H2O) does not satisfy the “theory”). It will be thus tempting, for some, to deny that concepts are theories. Related to this point are concerns about compositionality. Fodor argues that concepts cannot be theories because theories do not semantically compose in the way that Fodor thinks thoughts do (I think he argues something like this…). But if you don’t start out by assuming that a theory of concepts is supposed to settle the issues of reference and extension, then it is not clear that these semantic worries are objections against the theory view. So maybe the psychology/philosophy divide could be replaced with a non-semantic/semantic divide on the issue of concepts (as “semantics” is often understood in these debates).

    #2: In response to an objection by Machery, you claim that one can possess an explicit and an implicit concept of a lexicalized category. The implication is that implicit and explicit concepts can have the same “topic”; however, later in the paper the implicit/explicit distinction is used to account for the ability of language users to think about unobservable, non-existent and/or arbitrarily abstract objects/properties. But what is it about an explicit concept that supposedly explains this greater representational power, and why should we think that implicit concepts cannot represent these sorts of things? For example, it does not seem far-fetched to me to suppose that young children develop an implicit concept of justice or living thing. Might the response be that such concepts in children are likely situational/specific in ways that won’t allow us to say that they are abstract or non-observable in the intended sense? (This seems connected to Josh’s observations/questions, I think).

  5. Martin Roth

    By the way, reading this paper put me in mind of Hubert Dreyfus’ address to the Pacific APA some years ago:

    https://socrates.berkeley.edu/~hdreyfus/pdf/Dreyfus%20APA%20Address%20%2010.22.05%20.pdf

    I wonder whether what you are calling an ‘implicit concept’ is importantly related to Dreyfus’ idea of ‘non-conceptual’ embodied coping (and its relation to expertise). Finally, how do you think your account of implicit concepts overlaps with/differs from Churchland’s recent accounts of concepts/representation (e.g., in Neurophilosophy at Work, specifically “The Portrayal of Worlds”)?

    Good stuff!

  6. edouard machery

    Martin,

    Re your first point. This was exactly the point of my distinction between the philosophical and the psychological notions of concept. There are two issues that people who call themselves “theorists of concepts” have been tackling: Explaining how our thoughts refer or are about what they are about and explaining how we categorize, induce, etc. People (by and large psychologists) who are concerned with the second issue do not attempt to answer the first issue. For instance, prototype theorists (who identify concepts with prototypes) do not hold that one can think about x in virtue of having a prototype of x. Inversely, people concerned with the first issue (by and large philosophers) do not attempt to explain how we categorize, etc. For instance, neither Fodor nor Peacocke have this goal in mind. 
    Being clear about this distinction matters because many arguments (typically made by philosophers) against the kind of theories of concepts developed by psychologists assume that these psychologists are trying to solve the first issue (the issue philosophers have usually been concerned with).  
    I am afraid that by not clearly acknowledging that these two projects are fundamentally distinct Gualtiero is allowing people concerned with distinct issues to behave as if they were trying to solve the same problem. 
    Edouard
  7. Martin Roth

    Hi Edouard,

    Thanks for the clarification; I had the suspicion that this was a distinction you were marking, but I couldn’t remember for sure (DWC in on my winter break reading list, and it’s been awhile since I read a synopsis of it). I’ve been doing some stuff with Rob Cummins in this general area, and we are inclined to see things your way (so I shouldn’t put off DWC any longer!).

    Cheers.

    MR

  8. gualtiero

    Hi Martin,

    Thanks for the comments and interesting questions.

    Re 1: I agree with you (and Edouard) that psychologists and philosophers who work on concepts sometimes talk at cross-purposes (for the reasons you give), but it doesn’t follow from this that their respective theories are about different things. Their theories are (sometimes) about different aspects of the same thing — namely, different aspects of concepts. A full theory of concepts must explain both how concepts get their semantic content and how cognitive processes work.

    Re 2: On second thought, I find it plausible that children develop an implicit concept of justice (and living thing) without drawing from explicit cognition. So justice may not be a good example of a concept that requires explicit cognition. But surely there are some good examples, such as judge, attorney, atom, parliament, etc. What is it about explicit concept that allows them to represent more than implicit concepts? The main thing is their ability to interact with the language faculty, which in turn opens up access to other explicit concepts and the thoughts/sentences that can be formed with them, and hence entire cultures. This is very quick, of course, and would deserve much more detailed elaboration (which I have no time to provide, so it would be great if you were interested in writing about it :-).

    Re: Churchland. I like Churchland’s account of concepts as patterns of connection weights within a neural network (or something like that; do I remember right?); I think of something along those lines as the way implicit concepts might be realized in the brain. But Churchland does not have much room for explicit concepts in his account.

    Re: Dreyfus: Yes, implicit concepts are related to Dreyfus’s “non-conceptual” embodied coping. The main difference is that my embodied coping skills are conceptual, of course. To some extent this is a merely terminological difference, but to another extent, it’s a substantive difference because my account is a cognitivist and representationalist account whereas Dreyfus makes a big deal of rejecting cognitivism and representationalism, and that is a big mistake.

    (BTW, Martin, where are you based now? I can’t find your email address online. Would you mind sending it to me?)

  9. Martin Roth

    Hi Gualtiero,
    Re Re 1: The problem is that there may be no one thing which satisfies both the processing story and the semantic story (this is Edouard’s argument, right?). As I am understanding Edouard here (really, I need to read DWC!), the semantic story has to get the combinatorics of the propositional attitudes right, but the stuff that gets the processing story right isn’t up to the semantic task. Now, maybe we can split things up here, but this isn’t going to show us the various sides of one thing (processing and semantic aspects of concepts); rather, it’s going to show us that there are two things, each of which is being called ‘concept.’ Furthermore, there is no use in arguing about what concepts “really” are, so let’s just drop the terminology and replace it with something that is going to make for better science (maybe in the spirit of Quine’s dictum that philosophy of science is philosophy enough).
    Re Re 2 & Churchland: One reason Churchland isn’t going to be down with explicit concepts is that has abandoned the semantic project that is, in part, driving the move to explicit concepts. If you don’t think there are any propositional attitudes (or you don’t think there is anything more to them than, say, what Dennett thinks there is to them), then you aren’t going to worry about the semantic building blocks of such attitudes. But Churchland does think we have concepts of atoms, judges, etc., so he is moved to say that there is no interesting distinction in kind between the concept of bird and the concept of atom. Language may be crucial to acquiring certain concepts, but that won’t make the concepts acquired in any important sense “linguistic.” This is how I read Churchland, at any rate (and though I am sympathetic, I’m not defending that view here).
    Re Re Dreyfus: Yes, I think the matter here may be largely terminological. He’s anti-cognitivist/anti-representationalist in the way that Churchland is, and if you squint just right, Dreyfus and Churchland appear almost indistinguishable (and so maybe the stuff he is calling “non-conceptual” is what you are calling “implicit” is what Churchland is calling…”conceptual”). There has been an unfortunate tendency, methinks, to link cognitivism-representationalism with concepts, and then link concepts with the propositional attitudes. Take all the recent discussion of know-how: the options seem to be neo-behaviorism or “intellectualism” (read: propositional attitudes). If you are a cognitivist a la Bechtel and Abrahamsen (circa 1991), you get lumped with Ryle (since B & A don’t fit nicely into the knowledge-that crowd). But B & A are a long way from Ryle!

  10. edouard machery

    Again, I agree with Martin. 

    The argument is simple:
    1. For psychologists, a concept about x is, roughly, a body of knowledge retrieved from long-term memory when we categorize, draw inductions, etc., related to x.
    2. For philosophers, a concept of x is that which allows us to have propositional attitudes about x as such.
    3. That which allows us to have pa about x as such is not a body of knowledge  retrieved from long-term memory when we categorize, draw inductions, etc., related to x. (Martin’s point above)
    4. Thus, concepts for psychologists and concepts for philosophers are not the same thing.
    Now, you have criticized 1 and 2 in print (e.g., in one of the first footnotes of your commentary), but your objections are, I think, misguided. I have given numerous quotations and citations in support of both 1 and 2 (in DwC as well as in How to split Concepts? A reply to Piccinini and Scott (Philosophy of Science, 2006)
    E
  11. edouard machery

    Again, I agree with Martin. 


    The argument is simple:
    1. For psychologists, a concept about x is, roughly, a body of knowledge retrieved from long-term memory when we categorize, draw inductions, etc., related to x.
    2. For philosophers, a concept of x is that which allows us to have propositional attitudes about x as such.
    3. That which allows us to have pa about x as such is not a body of knowledge  retrieved from long-term memory when we categorize, draw inductions, etc., related to x. (Martin’s point above)
    4. Thus, concepts for psychologists and concepts for philosophers are not the same thing.

    Now, you have criticized 1 and 2 in print (e.g., in one of the first footnotes of your commentary), but your objections are, I think, misguided. I have given numerous quotations and citations in support of both 1 and 2 (in DwC as well as in How to split Concepts? A reply to Piccinini and Scott (Philosophy of Science, 2006)

    E

  12. gualtiero

    Ok, but premise 3 is not an immediate deductive consequence of 1 and 2. So even if you accept premises 1 and 2 (which I don’t), what is still missing is an argument that establishes premise 3. That’s the part that is missing from DWC. Martin gives some suggestions on how to construct such an argument, but it doesn’t look like it’s going to be a deductive proof. So I’m probably not going to buy the conclusion even if I grant you your premises 1 and 2.

  13. gualtiero

    Martin,

    Thanks again. I agree about Churchland and Dreyfus. As to your first point, I agree that there are difficulties in reconciling semantic and processing stories, but I also think it can be done. (See also my response to Edouard.)

  14. edouard machery

    1. If you reject Pr 1 and 2, you need another account of concepts that meets 2 conditions: (1) it should not be an unexplained metaphor (thus building blocks of thought is out) and (2) it should make sense of what psychologists and philosophers say and (more important) do. 

    2. Pr 1-3 are meant to be independent premises. 4 is the conclusion.
    I think most people who write on concepts agree with Pr. 3, and for fairly clear reasons: To think about x’s as such, you do not need a prototype of x’s, nor a theory of x’s, etc. 
    e
  15. gualtiero

    concepts = representations posited to explain certain cognitive phenomena, including recognition, naming, inference, and language understanding.

    Whether Pr. 3 is right depends on how exactly you explain the various cognitive phenomena. But in any case, a complete theory of concepts needs to have an account of how they get their semantic content.

  16. Anne Jacobson

    Let me suggest that one way to get a sense of how distinct the phil and psych projects are is to read the first couple of chapters of Jesse Prinz’s “Furnishing the Mind.” He does put the two projects together, but his desiderata that one meets are quite distinct from those met by the other.

    However, I am concerned about the discipline that contains statements like “the two projects are fundamentally distinct.” This looks like philosophy, and the fact that it needs to be said appears then to bring the psych theories into philosophy’s ken. Perhaps Edouard would say that after reading him, the philosophy of concepts will be better, and actually I agree with that. However, I think that philosophy has other criticisms – to do with our needing an account of error – that may challenge the heterogeneity account.

    More of that later when I finish the BBS comments!

  17. Martin Roth

    Hi Gualtiero,

    Here’s an attempt to flesh out the argument for the conclusion that we cannot put together the processing with the semantics. It’s a bit quick and blunt, and I am not even saying I agree with it. But I’m tired, and maybe putting things starkly will help to put our finger on the crucial issues.

    Joe believes that Fido is a dog. It’s possible for Joe to believe this because Joe possesses the concept DOG. The semantic content of DOG is the property of being a dog (or the set containing all and only dogs, whatever), and Joe’s belief is true just in case Fido has the property of being a dog.
    Not only does Joe believe that Fido is a dog, but Joe also believes that Fido is a sick dog. Joe’s belief is true just in case Fido has the property of being sick and the property of being a dog.

    Sarah is a psychologist. She has studied Joe for a long time, and she tells us that when Joe reasons about dogs, perceives dogs, etc., he draws on such things as: (1) memories of specific dogs; (2) some statistical information about dogs, gathered through a variety of encounters with dogs; (3) some background causal knowledge about dogs. Sarah also tells us that when Joe reasons about being sick, etc., he draws on such things as: (1) memories of specific instance of being sick; (2) some statistical information about sickness, gathered through a variety of encounters with sick things (3) some background causal knowledge about sickness.

    Michelle is a philosopher, and she tells us that Sarah’s account of how Joe reasons about dogs and how he reasons about sick things isn’t going to provide an account of how Joe can believe what he does (with the specified truth conditions). For starters, the kind of stuff included in (1)-(3) for dogs likely isn’t going to specify something that is satisfied by all and only dogs (similar points apply to sickness). But even if we put that worry aside, simply conjoining (1)-(3) for dogs with (1)-(3) for sickness isn’t necessarily going to get you something that includes: (1) memories of specific sick dogs; (2) some statistical information about sick dogs (of the sort that likely would be gathered through encounters with sick dogs); (3) some background causal knowledge about sick dogs. So even if reasoning about sick dogs somehow draws on (1)-(3) for dogs and (1)-(3) for sickness, it is not at all clear how that story would provide an account of semantic compositionality, the sort that seems required to explain the truth conditions of Joe’s beliefs about Fido.

  18. gualtiero

    Martin,

    Thanks for fleshing out the argument. (BTW, is Michelle Fodor’s twin? :-)) Two things come to mind:

    1. As far as I remember, Edouard does not give this argument in DWC, or any other argument for the conclusion that “philosophical” and “psychological” theories are about different things.

    2. The argument is far from demonstrative, and in fact there are lots of ways around it, including some combination of the following:

    A. Deny that the content of concepts has sharp boundaries (“all and only dogs”)

    B. Maintain that thanks to appropriate processes, such as deference to experts on the part of cognizers, (1)-(3) manage to refer to the right entities.

    C. Deny compositionality at least for concepts (if not for sentences in a public language).

    D. Maintain that combining (1)-(3) in the right way can give you enough compositionality to get the job done.

    E. Maintain that you can take care of reference and compositionality by appealing to certain specific aspects of concepts, i.e., something like words in a language of thought, to which (1)-(3) are associated in a way that allows cognizers to use them when the concepts are deployed.

    (F) reject Sarah’s theory and replace it with a theory that fits with Michelle’s requirements.

    Bottom line: Maybe some philosophers and some psychologists are talking about different things. But some philosophers and some psychologists are explicitly about the same things. In any case, we are not going to make progress on concepts by arguing over who is talking about what. We are going to make progress by arguing about what concepts are and what they do (which Eduard does plenty of in DWC).

  19. edouard machery

    So?! I thought that the issue at hand in the thread was what different theories of concepts are about! And, If your characterize the notion of concept in a way that is incompatible with the way Fodor, Peacocke, Rey (add Evans, etc.) characterize concepts, you are plausibly changing the topic (If Fodor is just another guy writing about concepts, then David Lewis is just another guy among many other metaphysics of possible worlds!). 

    There might be good reasons to prefer the notion of concept you endorse or the one I proposed (which again are really variants of each other), but it is important to be clear about all this since lack of clarity about this question has resulted in a large number of misdirected arguments. 
    e
    PS: Oh one more thing, you do not neet much of an argument to support Pr 3 above – no cognitive structure that is sufficient for explaining how we think, categorize, draw inductions, etc., will be necessary for thinking about an object. 
  20. gualtiero

    I don’t know what you are getting at with your comment that no structure sufficient for cognitive processes is necessary for thinking about something. How is that supposed to support the conclusion that what allows you to think about something is different from what explains cognitive processes?

Comments are closed.

Back to Top
%d bloggers like this: