How Many Knowledge Relations are There?

Taking a cue from an interesting discussion between Jonathan Schaffer and David Sosa, let’s say that there are three positions to hold about knowledge: eliminativism, which maintains that there are no knowledge relations; monism, which holds that there is one knowledge relation; and pluralism, which holds that there is more than one knowledge relation.

There is a danger, however, that pluralism is a fairly trivial thesis (so long as eliminativism is rejected). It looks as if it is confirmed by the fact that people use the term ‘knows’ and its variants in a variety of ways. Sometimes the term takes a person or thing as its complement, as in “Bo knows jazz.” And, as is well-known, there are other uses as well.

So suppose we restrict our three theories to sentences involving that-clauses as complements of ‘knows’. That helps some, but everyone already knew that there are competent uses of ‘knows that’ that are not factive, for example. “I just knew the Yankees were going to lose two years ago, but they took a year longer than I expected.” Etc.

Here’s what we might try. Distinguish between that-clause examples that can be explained away in terms of other concepts, and those that can’t. In the case above, we explain away the use of ‘knew’ in terms of psychological certainty. Among the remaining cases, the pluralist holds, there will be more than one relation found.

A niggling worry remains, however. Consider Ernie Sosa’s distinction between animal and reflective knowledge. Is this a version of pluralism? Or just two different kinds of the same monistic relation? I suspect Ernie would insist on the latter, but I’m not sure how to sustain this view. Perhaps as follows? Don’t multiply senses beyond necessity, so if one can model different usages of a term without ambiguity, don’t posit it. But perhaps there are other ideas?


How Many Knowledge Relations are There? — 19 Comments

  1. So the question is: what substantive thesis could the monist and pluralist be reasonably arguing about, given the (I suppose?) obvious fact that people actually use “knows” to denote any number of different relations?

    Here’s one thing that a monist and a pluralist might say in response to this question: What we’re reasonably arguing about is the kind of factual knowledge (“knowledge that”) which inquiry, by its nature, aims to achieve.

    Here’s another way to think about the debate. Consider the question: what is inquiry? What is it that makes a certain process one of inquiry? Now, someone might think that part of the answer is: the process aims at the achievement of knowledge. (Maybe such knowledge is also, as it happens, achievable without inquiry, but that’s irrelevant.) Moreover, what makes a certain process one of inquiry concerning whether p is that it aims at the achievement of knowledge whether p. What makes a certain process one of inquiry concerning why p is that it aims at the achievement of knowledge why p. What makes a certain process one of inquiry concerning when p is that it aims at the achievement of knowledge when p. And so on for other wh- inquiries.

    So maybe the monist and the pluralist can be understood as disagreeing about the number of dimensions along which inquiries differ. If that’s what they’re disagreeing about, then their disagreement seems pretty substantive (at least as substantive as some other disagreements in pragmatics.)

  2. Ram, that’s an interesting idea. So, I take it, the pluralist says that knowledge when is a different relation from knowledge why is a different relation from… They’ll also need to argue that the reductionist view that each can be explained in terms of knowledge that fails, right?

    Your last comment puzzles me, though–the comment about “other disagreements in pragmatics”. Do you think of the pluralist/monist issue as a pragmatic one rather than a semantic one? I was assuming that the issue would be semantic, but didn’t really think about it.

  3. As far as setting up the debate between epistemic monism and pluralism, perhaps the following will help. First, we might recognize that English â??knowsâ?? is ambiguous between two conceptsâ??what French distinguishes with â??savoirâ?? and â??connaitre,â?? and what German distinguishes with â??wissenâ?? and â??kennen.â?? The distinction between monists and pluralists here is within â??knowsâ?? in the sense of â??savoirâ?? and â??wissenâ??â??what I will call informational â??knowsâ??. I would be happy to say, along with Ram, that such knowledge is the goal of inquiry.

    Second, we might recognize that informational â??knowsâ?? can take at least three sorts of objects (syntactically characterized). Compare:
    (1) I know that it is midnight
    (2) I know what time it is
    (3) I know the time
    (1) takes a that-clause, and is the sort of construction that traditional epistemologists are fixated on. (2) takes a wh-clause, but is still informational, and is actually the more common sort of construction in natural language. (3) takes a determiner phrase, but remains informational, and is sometimes called a â??concealed questionâ?? use, since it seems semantically equivalent to (2). Morale: it is a mistake to think that knowledge-that is special, or to restrict our attention to it.

    Third, Jon’s bit about non-factive uses of informational â??knowsâ?? is tricky. But perhaps the best way to think of this is via implicit scare-quotes. The non-factive uses (i) seem to express felt certainty, and (ii) seem to require a distinctive intonation on â??know.â?? Perhaps all that is happening here is that the intonation is signaling scare quotation, to deliver (Jonâ??s example): â??I just â??knewâ?? the Yankees would lose two years ago, but they took a year longer than expected.â?? If anything like this is right, the non-factive uses of â??knowsâ?? are not really uses at all!

    So: the debate what monists and pluralists might come out as a debate about whether there is a single epistemic relation denoted by informational â??knows.â?? None of this is meant to resolve the debateâ??only to try to set it up. The point of setting up such a distinction between monism and pluralism is to express what the contextualist is committed to, in the object language. Now we might say: the contextualist is committed to a plurality of epistemic relations, for â??knowsâ?? to shift across.

  4. A small but perhaps important point… I don’t think using the concept of inquiry to fix the concept of knowledge is going to help. For example: Is mystical contemplation a kind of inquiry? Some have thought it was the only way to some bits of knowledge. Or put it this way: what do geometry proofs and reading the prices on the McDonald’s menu (assuming both lead to knowledge) have in common, besides the name ‘inquiry’?

    In the background seems to be the assumption that a certain kind of process is the way to knowledge, but I think either that process gets characterized too broadly to name anything, or narrowly enough that it begs the question.

  5. Jonathan, I like your characterization of informational knowledge–it strikes me as a better statement of what epistemologists have been interested in than “knowledge that” (even if we’ve focused on the latter in an attempt to understand the former). Do you think there are good arguments for holding that the other uses are not reducible to “knowledge that”? At least, knowledge-wh seems naturally associated with whatever knowledge claims answer certain questions, and maybe that amounts to your contrastive knowledge theory. And knowing the time seems easy to treat as short for knowing what time it is.

    I’m wary of appeals to scare quotes here to conclude that the non-factive uses are not really uses at all. I take it that using scare quotes is a way of signalling syntactically that the standard semantical apparatus can’t be used to interpret the utterance, and the phenomenon of stress on a term may sometimes amount to the same thing, though of course, not always In the example I used, had I written “I was certain the Yanks would lose,” you’d read with the same stress). But scare quotes are not a device to mark off mentioning a term rather than using it. The term is still used, but with a marker that tells us a different interpretation is called for.

  6. In the example I used, had I written â??I was certain the Yanks would lose,â?? youâ??d read with the same stress.

    Jon: I think the test Jonathan (this is subject to correction by Jonathan) is suggesting is whether you have to use stress to be speaking appropriately. Where the Yankees in fact didn’t lose two years ago, your original example —

    â??I just knew the Yankees were going to lose two years ago, but they took a year longer than I expectedâ??

    — seems wrong if “knew” is not stressed. It’s less clear (at least to my ears) that “I was certain they would lose” is wrong if “certain” isn’t similarly stressed. [To the extent that does seem wrong, that may perhaps be taken as evidence that some fairly objective merit for one’s belief that P that is jeopardized by P’s being false is being claimed with “I was certain that P.”]

    “I felt certain that they’d lose” is perfectly fine without the relevant stress.

  7. Jon: I think there are good arguments for thinking that all three types of informational knowledge claims (knowledge-that, knowledge-wh, and knowledge-DP) involve one and the same knowledge relation. So I am happy to agree with you that they are all inter-translatable, in a sense. But that said, I donâ??t think the knowledge-that format is the most perspicuous format. The knowledge-wh format actually carries more information, since it embeds a question, and a question denotes a contrast space. Plus knowledge-wh is far more natural in ordinary language. So I think that all three types of informational knowledge claims express a relation between a proposition and a contrast space. The knowledge-that format is just confusingly inexplicit here. It can even make one think that knowledge is a non-contrastive relation!

    Heath: Perhaps it will help if we characterize inquiry as an attempt to answer a question. Or, more precisely, we might see inquiry as a series of questions and answers, and a stage of inquiry as an attempt to answer a particular question. If we have well-formed questions about geometry or about McDonaldâ??s prices (your examples), we might try to answer them. To answer a question is to select the true alternative from the contrast space. So–following up on Ram’s suggestion–thatâ??s the sort of process we might use to fix what can count as a knowledge relation. Does that help?

    Keith: yes, thatâ??s what I meant. Though you put it much better than I did. What we then need to say, I take it, is that the fact the example only works with this sort of stress patterns shows that there is some funny business afoot. Though no doubt it will be tricky to say exactly what. Do you have a line on what the stress is doing here?

  8. Jonathan, is your view, then, that the pluralist is one who denies some unifying account of knowledge-that, knowledge-wh, and knowledge-DP (and maybe any other informational uses that don’t fit these schema)? I’m happy with that characterization, but it doesn’t seem to be what contextualists are after, and so doesn’t give us the theoretical import that we thought the metatheory was aiming at. That’s a messy formulation, but do you get the idea?

  9. Jon: yes that’s basically what I had in mind. Though much depends on what counts as a ‘unifying’ account. For the classic contextualist, who treats “knows” as an indexical, all the informational uses will have a common character (e.g. elimination of all the relevant alternatives), but will not have common contents (e.g. the set of relevant alternatives may shift). So I would prefer to put things in terms of the existence of a single relation, rather than in terms of a unifying account. Then I would say that what the contextualists are after is the idea that there is no single relation–but rather a family of related relations–invoked in knowledge talk. Now does that sound like what the contextualists are after, or do you still think this is not an apt characterization?

  10. Uhm, does this mean that you’ve all settled this question and that the number of knowledge relations is: 1?

    That would be a good result….


  11. I’m a little disturbed by the suggestion that embedded questions uses of “know” denote a different relation than obviously propositional uses. There is a big literature in semantics on embedded questions, dating by back to the 1970s, that I’m sure many some people here are familiar with. At first, people analyzed “x knows wh-F” as true iff x knows all the true answers to the question “wh-F”. Now, I think a standard analysis is Groenendijk and Stokhof’s (although this is pretty ancient too, from the 1980s), that uses facts about the acceptability of conjoining the two together as evidence that they are of the same semantic type, e.g.:

    (1) John knows who came to the party and that there was a huge argument over the music.

    “know” takes embedded questions, like a number of other clausal complement verbs (“learn”, “tell”, “said”). I presume we wouldn’t want to say that there are two learning relations, just because one can both learn who came to the party and that the party-goers had fun.

    Interestingly, “believes” doesn’t take embedded questions, and this is cross-linguistic. Get a semanticist drunk at a party and she’ll ask you why philosophers haven’t discussed this interesting fact (you can get “I believe what Hannah said”, but this is the relative clause “what Hannah said”, as in “the thing(s) Hannah said”, not the embedded question.)

    I don’t myself think that knowledge-by-acquaintance is any kind of knowledge at all…I certainly don’t think that embedded questions (know-why, know-when, know-how, know-who, know-where) show that there are many different knowledge relations, any more than they show that the fact that you can tell Sam who went to the party and tell Sam that it was fun shows that there are two different telling relations, or different kinds of telling-pursuits. There is just our good friend, the knowledge relation.

  12. I was puzzled by Stanley’s post concerning the “suggestion” that embedded questions uses of ‘know’ denote a different relation than do obviously propositional uses. I was puzzled because I thought that everyone — monist and pluralist alike — agrees that embedded questions uses do NOT denote a different relation (or, for the pluralist, different relations) than do obviously propositional uses. So then I went back through the thread to find the remark that may have provoked his comment, and I suspect that it may have been my own remark. (In any case, I cannot find any other plausible candidates.)

    So I’ll clarify: pluralist and monist, as I understand them, disagree about how many degrees of freedom the knowledge relation has. But how is this a substantive dispute? What data do the monist and the pluralist have to answer to? Lots of epistemologists today seem to think that the disputants have to answer to semantic data. But I don’t see how semantic data are going to tell us about the knowledge relation. Should a theorist of justice consult semantic data in developing her theory of justice? Should a theorist of criminal punishment consult semantic data in developing her theory of punishment? Should a philosopher of economics consult semantic data in developing her theory of welfare? I assume that the answer to all of the last three questions is “no”. And I don’t see why knowledge should be any different in this respect from justice, criminal punishment, or welfare.

    So this leaves us with the question of how to understand the monist/pluralist dispute as substantive, i.e., as answerable to some data or other. I propose one way to do this: Let’s think of knowledge as the goal of inquiry (knowledge when as the goal of inquiry when, knowledge how as the goal of inquiry how, knowledge why as the goal of inquiry why, etc. — in each case, the same relation, differently expressed). If we think of knowledge in this way, then we can understand disagreements about the metaphysics of the knowledge relation to be disagreements about the goal(s) of inquiry. And disagreements about the goals of activities are familiar empirical disagreements, settlable in ways that are familiar to anthropologists. (What is the point of shaking hands when you meet someone? Of putting your hand on the Bible in a courtroom? etc. etc.). In that way, we can understand the dispute between the pluralist and the monist as a substantive dispute of normative significance.

  13. Ram, the move you offer against the appeal to semantics in epistemology only works if issues about semantics are not already completely entangled with the epistemological dispute. If we’re worrying about questions like contextualism-vs.-speaker-sensitive-invariantism, then we’re already up to our elbows in semantics, and it’s no good trying to de-semanticize the dialectic. The whole point of contextualism, after all, is find a semantics of “knows” that can respect our judgments about a number of cases. So it would be odd to think that contextualism can progress apace without any considerations of semantics whatsoever. (Which is not to say, however, that the contextualist has to hand the whole problem over to the linguists, either; there’s room, I hope, to ask for a sensitivity to scientific approaches to semantics that need not entail a complete outsourcing to it.) It may just be that folks in the other literatures you refer to just have not tried to solve any of their problems with any heavy-duty semantic machinery, and so can ignore linguistic results with an impunity not available to the folks working in this neck of the epistemological woods.

    Alternatively, when the dialectic is so thoroughly suffused with semantics, to argue that we should de-semanticize the dialectic may be to argue that we should do away with that particular dialectic altogether. I think that that’s the right consequence of your Craigian move, in fact. If we take the right approach to knowledge to be finding out what the proper goal of inquiry is, it’s not at all obvious that a good way to do that is to hold our theories hostage to ordinary speakers’ judgments. Our current practices with “know” might carry with them all sorts of confusions about the real goal of inquiry — confusions that are especially likely to be in play when trying to apply those practices to esoteric cases, but which would manifest in more quotidian ones as well — and a good methodology would detach us from those judgments, rather than encouraging us to be loyal to them. So it may be that if your offered move works at all, it will work rather too well for many of the participants here!

  14. Jonathan, I think you’re right that Ram’s view will be seen as discomfiting to lots of epistemologists, but it cheers my heart, at least. I view Ram’s suggestion in terms of letting value considerations drive theoretical epistemology more than they have in recent times, since I think the plausibility of Ram’s approach rests on inquiry being a valuable activity (surely we can grant this one :-)), and then trying to find out what it is good for. That puts the project more within the realm of the process of cognition than what I’d prefer to think about in terms of the valuable end state of such, but there’s surely room for both projects. My only concern about Ram’s particular project surrounds the claim that knowledge is the goal of inquiry. If knowledge doesn’t have the right sort of value, then even if knowledge is the goal of inquiry, we ought to clarify how inquiry with a more appropriate goal would proceed. So, besides the kind of empirically-sensitive epistemology you’re working on, there’s also value-driven approaches that don’t recommend attending to ordinary usage as much as recent epistemology has.

  15. Hi Jonathan and Jon,

    Jonathan: I certainly didn’t mean to say that the dispute between contextualists, SSI’ers, relativists, and Schiffer-Weiner-Eklund “incoherence” theorists is not important, or is not an empirical semantic dispute. I think it’s a very important dispute, and a dispute that’s answerable to empirical semantic data. But it’s not a dispute that — by itself — is going to tell us much about what knowledge is, or about what’s epistemically valuable.

    Of course, to the extent that our first-order epistemological theorizing is guided by our epistemic intuitions, the contextualism dispute can be indirectly very useful to such theorizing: it shows us how many pitfalls there are simply in recording the data, never mind explaining it. In just the same way, the Weinberg-Stich-Nichols work on cross-cultural variation of intuitions can be indirectly very useful to such first-order theorizing.

    Jon: I blush to confess that I haven’t yet finished reading your book on the value of knowledge (but soon, I hope!), and so when I spoke of knowledge as the goal of inquiry, I was just blithely assuming away all sorts of dissent that the thesis can encounter. Maybe knowledge has some value other than that of being the goal of inquiry, and then we can relocate the monist-pluralist dispute over there. Or maybe knowledge has no value, in which case I think we’d be wasting our time worrying about it. But I’m defeasibly supposing that this isn’t the case.

  16. Ram, nice to know you’re thinking about it! The consequences, even if I’m right, are not severe for the project of using the goal of inquiry to guide theorizing. I’m more attracted to thinking of understanding as what we are aiming for, and given the closeness between understanding and knowledge, the adjustments are not going to be major.

  17. If we are to have a substantive dispute between monists and pluralists (/contextualists), I think we might allow BOTH linguistic and value-driven considerations.

    On the linguistic side, we have the question of whether ‘knows’ is referentially shifty. If ‘knows’ is shifty, then (given the prima facie plausible assumption that we should regard the relations it denotes as knowledge relations) this is some defeasible evidence for pluralism. If ‘knows’ is invariant in denotation, then (given the prima facie plausible assumption that a knowledge relation should be denotable by ‘knows’) this is some defeasible evidence for monism. Here I would suggest, on behalf of the monist, that the contextualist cases may be understood in terms of shifts, not in the denotation of ‘knows’, but in the denotation of the implicit question under discussion.

    On the value-driven side, we have Ram’s question of whether there are many goals of inquiry. Here I would suggest, on behalf of the monist, that there is a single goal of inquiry–to answer the question.

    So perhaps both the linguistic and value-driven ways of addressing the monism/pluralism dispute will converge. They will converge on a form of monism, on which to know is to know the answer to a question.

Leave a Reply

Your email address will not be published. Required fields are marked *