The notion of coherence


Quite recently I started to think the notion of coherence in the context of cognitive sciences. (The sun was shining and so on…) As well know, there are several ways to interpret philosophically the notion of (conceptual) coherence. 

For instance, one way is to say that the notion of coherence can be simply defined as the similarity of certain concepts. The basic idea is that objects fall into natural clusters of similar kinds, and our concepts map into these concepts. Of course, as was pointed out by Quine, using similarity as the basis of concepts will raise the very same question again. It namely well be that we see some concepts similar, because they are part of the very same category, and thus one needs an independent theory of similarity (Quine, 1977, also Murphy and Medin, 1985).

Another, interesting and theoretically deep analysis of the notion of coherence is proposed by Thagard, Chris Eliasmith (he is my personal idol!!!) and their colleagues (Thagard and Verbeugt, 1998, Thagard, Eliasmith & et al). In their model epistemic coherence is understood in terms of maximization of constrain satisfaction.

According to this model knowledge involves at least five different kinds of coherence. These kinds of coherence are explanatory, analogical, deductive, perceptual and conceptual coherence. Each of these require different sorts of elements and constraints.

The conceptual coherence is defined technically as follows: (i) Conceptual coherence is a symmetric relation between the pairs of concepts, (ii) a concept coheres with another concept if they are positively associated, if there are objects to which they both apply, (iii) the applicability of a concept to an object may be given perceptually or by some other reliable source, (iv) a concept incoheres with another concept, if they are negatively associated, i.e. if an object falling under one concept tends not to fall under the other concept. Finally (v) the applicability of a concept to an object depends on the applicability of other concepts.

I do not, folks, what should we think about this… First (i) does anybody disagree, and if so, why? And second, (ii) does anybody know any other suggestions?  

From sunny Helsinki,



  1. Greetings from fairly-but-not-as-sunny as-Helsinki Brooklyn!

    I am not familiar with the work of Thagard, Eliasmith, et. al., so my
    question might seem silly:

    What work is the concept of coherence intended to do?

    The account of coherence you sketch seems to make coherence supervene
    on some other things, like association and dependence. Association
    looks like co-instantiation (ears and eyes are associated, tails and
    humans are not associated). Dependence looks like explanation: having
    a tail depends, maybe, on being a vertebrate and hence having a spine
    (and you need other things if you are going to have a tail).

    So then I wonder whether the concept of coherence is something that
    applies *after* you have an organized theory of something. If you
    have, for example, a computational account of some part of auditory
    processing, you will find that the concepts invoked in the theory
    cohere conceptually (in the sense defined in your post), and do not
    cohere conceptually with the concepts of other theories (does hearing
    involve the causation of non-physical “sounds”?).

    Coherence is sometimes discussed as a standard of truth. Russell
    criticized this: you could have a set of sentences that is consistent
    yet all false. Presumably adding in constraints wouldn’t help, if the
    ultimate threat is First Meditation skepticism.

    So what do we want to do with the concept of coherence?

  2. anna-mari rusanen

    Hi Tony,

    and thanks for your post. I am in a bit rush, but briefly: In the context of developmental psychology or cognitive psychology (of learning) the notion of “conceptual coherence” is needed for several reasons. The notion is used frequently for several purposes. One of them is, say, this: In theories of science education the experts and novices of a certain domain are usually contrasted by saying that novice`s belief system and/or conceptual organization of that system are not as “coherent” as the systems of experts.  Another example can be found in the context of cognitive neurosciences, in which it is quite common to describe, for instance, the visual system by saying that the “3D- interpretation of the world is coherent” or whatever.

    So, in the first example the notion of coherence is intended to describe the degree/structure of individual`s conceptual coherence, and in the second the processing principle or a rule that certain neurocognitive system “follows”. This last sentence is sloppy, but I really have to run now… I´ll get back to this asap.

    From still sunny and warm Helsinki,





  3. tony

    Ok. From your definition, I’d be most interested in (v) — what strikes me as the most important difference between a less and a more coherent cognitive system is how many explanatory connections it makes (rather than how many associations of things going together or not going togeter).

  4. anna-mari rusanen

    Hi Tony,

    and sorry for delay. I´ll combine your earlier post with the present one and try to comment both of them at the same time.

    First, I´d like emphasize that the account of coherence we are talking about is quoted actually from a paper titled ” Knowledge and Coherence” (Thagard, Eliasmith and Company, 199something). I am sorry, I should have made this explicit. But, since I have not read any commentaries of that paper yet, your analysis of it is extremely interesting and much appreciated. Thanks.

    Where should I start? Ok, I am afraid that I may have not been very clear in my earlier posts, but the real point of (that) paper is to characterize/specify the notions of coherences that are needed for understanding the human knowledge. Thus T, E and Company distinguish the (1) explanatory, (2) deductive, (3) analogical, (4) perceptual and (5) conceptual coherence. Basically, these different kinds of coherence are distinguished from each other by the different kinds of elements and constraints they involve. 

    And now, you ask about this relationship between conceptual coherence (the list of principles i-v mentioned in my first post are principles of conceptual coherence)  and explanatory connections… Well, I do not know what to say, since it seems to me that relationships between conceptual and explanatory (“propositional”) connections are generally left more or less obscure almost in every theory I know.

    In their paper T,E and Co say only that “one possibility is that a deeper representational level, such as systems of vectors used in neural networks, may provided a common subsratum for propositional, perceptual and and conceptual coherence).  So, even if I am not sure, whether or not you meant this, I guess the point you made in your earlier post may actually be “dead on”.

    Namely you wrote: “So then I wonder whether the concept of coherence is something that applies *after* you have an organized theory of something. If you have, for example, a computational account of some part of auditory processing…” 

    However, what does it imply if the notion of coherence sketched by T,E and co just “supervenes on other things”? What would be in danger then? You are clearly thinking that it is a bad thing, and I am curious to know why.

    looks like co-instantiation (ears and eyes are associated, tails and
    humans are not associated).”

    Hmmm… I am not sure, whether I understand this (or the original paper for that matter) correctly, but as far as I can see (which is not very far) I guess association is understood here in the standard manner i.e. as it is usually understood in prototypebased (or hybrid) theories. I will dig this a bit, and as soon as I find something specific, I`ll tell you.  

    “Coherence is sometimes discussed as a standard of truth. Russell
    criticized this: you could have a set of sentences that is consistent
    yet all false. Presumably adding in constraints wouldn’t help, if the
    ultimate threat is First Meditation skepticism. “

    Yes, I know and agree. Also, I would argue that the notion of  coherence cannot be seen as a criterion for a correct (or “true”) representation either in context of cogsci either. I mean… well,  I am an old-fashioned girl who likes old-fashioned things such as classical cognitive science and veridical theories of cognition, and the truth conditions for representations cannot be defined on the basis of the degree of coherence…

  5. Daniel Weiskopf


    I’ve only taken a very quick scan through the Thagard, Eliasmith, et al. paper. I admit I can’t quite see what some of their notions of coherence are doing towards giving a theory of epistemic coherence (as in coherentist theories of justification). In particular I don’t see what conceptual coherence contributes to that, since justification accrues to propositions (or propositional attitudes), not concepts. I’m not actually sure they intend it to do this, though.

    Leaving that issue aside, one of the better critiques of the idea of conceptual coherence is:

    Margolis, E. 1999. What is conceptual glue? Minds and Machines, 9, 241-255.

    Like Margolis, I doubt there’s much to the notion of conceptual coherence. What problem is it supposed to solve exactly? But even on its own terms, the proposal by Thagard et al. looks problematic to me.

    For instance: in (ii) & (iv) coherence involves ‘positive association’, which is glossed as partial co-reference. But lots of things are associated but referentially disjoint (salt & pepper, cats & dogs, etc.). Either I don’t understand what association is here, or I don’t understand what ‘application’ of a concept means. In the same vein, I don’t understand (v) at all. Concepts apply to whatever is in their extension–presumably this just depends on what their extension is. I don’t see how other concepts enter into it.

    Those are my immediate thoughts. I’d look at Margolis for more questions about what conceptual coherence is supposed to do for a theory of concepts.

  6. Anna-Mari

    Daniel (may I call you Daniel?)

    In a way I guess the Thagard et co paper offers only a sort of formalism, and it allows many interpretations. So, I agree that it is not completely clear, what the final target or focus of the paper is. However, it seems to me that their perspective is broader than the standard epistemic approach to the notion of coherence. It seems to me that they are primarily trying to sketch such a notion of coherence that could be useful for various purposes in cogsci. But I may be wrong here – this is only my interpretation.

    Thank´s for the Margolis tip, and
    I´ll read it.

    Nice to cybermeet you,


Comments are closed.

Back to Top