Notes on Russell’s principle of acquaintance

Bertrand Russell famously defended a principle of acquaintance taken to govern some or all forms of propositional cognition—for example judging that p or supposing that p. One of the central statements and defences of the principle occurs in ‘Knowledge by Acquaintance and Knowledge by Description’ (Proceedings of the Aristotelian Society 1910–11; reprinted in Mysticism and Logic (1917; references to the 1986 reissue)). The statement of the principle, in italics, and defence occur in the following brief passage:

Whenever a relation of supposing or judging occurs, the terms to which the supposing or judging mind is related by the relation of supposing or judging must be terms with which the mind in question is acquainted. This is merely to say that we cannot make a judgement or a supposition without knowing what it is that we are making our judgement or supposition about. It seems to me that the truth of this principle is evident as soon as the principle is understood…” (211, Russell’s italics)

Two preliminary comments about this passage are in order. The first is that Russell seems to suppose that the subjects of judging and supposing are minds rather than things with minds—i.e. thinkers. I won’t comment further on that supposition here. The second is that Russell appears to take instances of judging or supposing to be occurrences—plausibly, episodes or events, rather than e.g. states or processes. (209) A question therefore arises about the generality of the principle of acquaintance. Is the principle intended to apply to all forms of propositional cognition? First, does Russell think that all instances of propositional cognition—including for example instances of believing or knowing—are episodic, or does he allow that some such instances are states? Second, if Russell were willing to allow that some forms of propositional cognition are not episodic—because, for example, their instances are states—, would he take the principle of acquaintance to apply also to the states? In the first instance, it would be natural to assume that Russell intended the principle to apply to all forms of propositional cognition. Hence, since Russell says nothing explicitly to restrict the principle’s application, it would be natural to take Russell to be committed to holding either that all forms of propositional cognition are episodic or that the principle applies to cognitive states as well as episodes.

The principle then claims the following:

(PA) In any case in which a subject, S, cognizes that p, then for all constituents C of the proposition that p, S is acquainted with each of C.

The defence of the principle takes the following form (ignoring issues about unrestricted generalisation over propositions):

(P1) For all S, p, if S cognizes that p, then S knows what it is that they are making their cognizing about.

(P2) For all S, p, if S knows what it is that they are making their cognizing about, then for all constituents C of the proposition that p, S knows each of C.

(P3) For all S, p, and constituents C of the proposition that p, if S knows each of C, then S is acquainted with each of C.

(C4) For all S, p, and constituents C of the proposition that p, if S cognizes that p, then S is acquainted with each of C. [≈ (PA)]

Russell takes (P1) to be obvious. (P2) and (P3) are the most natural ways of attempting to mediate a straightforward transition to Russell’s conclusion, (PA), via its plausible equivalent, (C4).

One serious problem with this version of Russell’s argument is that (P2) is undefended, and looks suspiciously close to an application of (PA). For knowing what it is that one is making one’s cognizing about is not obviously a matter of knowing, in the sense of being acquainted with, the elements that one’s cognizing is about. Rather, where it is a fact that one is making one’s cognizing about the elements C, knowing what it is that one is making one’s cognizing about is knowing that fact: it is knowing that one is making one’s cognizing about the elements C. In that case, (P2) looks to depend on a principle to the effect that propositional knowledge is subject to an analogue of (PA), so that knowing that one is making one’s cognizing about the elements C requires knowing (in the acquaintance sense) each of C.

Now a fairly standard manoeuvre at this point would be to suggest that Russell really meant for his appeal to the subject’s knowing what it is that they are cognizing about to be an appeal to the subject’s knowing that which they are cognizing about. Although Russell appears to be wrong that there is a natural reading of the “knowing what” construction that takes the required reading, he may have been mislead by the availability of an analogous reading of the “seeing what” construction. For although there are readings on which seeing what is before one is a matter of seeing that such-and-such, there are also readings on which it is a matter of seeing that which is before one. For instance, one might see what is before one either in seeing that a computer screen is before one, or in seeing the computer screen. Thus, what Russell should have exploited is the following replacement for (P1):

(P1’) For all S, p, if S cognizes that p, then S knows that which that they are making their cognizing about.

A natural concern about (P1’) is that it is too close to (PA) to serve in argumentative defence of the latter principle. It may be that Russell nonetheless took (P1’) to be obvious. However, that won’t help others to get themselves into a position from which they also take it to be obvious.

At that stage, one might be tempted to give up, as hopeless, the attempt to reconstruct Russell’s argument. Instead, one might accept that Russell’s attempt to defend (PA) fails, and then try to diagnose the source of his error. Given the materials we have, a natural diagnosis would be the following. The premise Russell needs, (P1’), is not independently plausible. However, it might be argued that the premise to which he appeals, (P1), is independently plausible. If that were right, then we could understand why Russell might take that premise to be obviously correct. We would then be in a position to explain his presentation of the erroneous argument as due to his slipping unconsciously between appeals to knowing what and knowing that which.

That diagnosis might be correct. However, it appeals, in passing, to the claim that (P1) is plausible. Now it might be that people are in fact inclined to find (P1) plausible. But further reflection on that premise raises a number of delicate issues.

Let’s continue on the assumption that (P1) is supposed to hold with respect to all forms of propositional cognition, including propositional knowing. Then we have the following.

(P1) For all S, p, if S cognizes that p, then S knows what it is that they are making their cognizing about.

(P2’) For all S, if S knows what it is that they are making their cognizing that p about, then for some appropriate true proposition, that q, that answers the question, what is S making their cognizing about?, S knows that q. [Plausible approximate account of knowing what.]

(P3’) For all S, p, if S knows that p, then S cognizes that p. [From the opening assumption.]

The apparent availability of this derivation presents the proponent of (P1) with two potential difficulties for (P1). The first difficulty is that the derivation is recursive, so regressive. Suppose

(P4) S cognizes that that patch is red.

From (P1), we have:

(P5) S knows what it is that they are making their cognizing about.

(P6) An appropriate answer to the question, what is S making their cognizing about?, is that S is making their cognizing about that patch’s being red. [Plausible account.]

(P7) S knows that S is making their cognizing about that patch’s being red. [From (P2’)]

(P8) S cognizes that S is making their cognizing about that patch’s being red. [From (P3’)]

(P5) S knows what it is that they are making their cognizing about. [Reiteration.]

(P9) An appropriate answer to the question, what is S making their cognizing about?, is that S is making their cognizing about S’s making their cognizing about that patch’s being red.  [Plausible account.]

(P10) S knows that S is making their cognizing about S’s making their cognizing about that patch’s being red. [From (P2’)]

(P11) S cognizes that S is making their cognizing about S’s making their cognizing about that patch’s being red. [From (P3’)]

And so it goes. In combination with (P2’) and (P3’), (P1) is regressive, and imposes cognitive conditions on cognizers that seem on their face highly implausible. In order to cognize that p, one must engage in cognition at the next level in the hierarchy that is directed onto elements that may be absent from the proposition that p, including cognition directed at the cognition taking place at the previous level in the hierarchy. Moreover, a subject’s cognizing that p requires that they suffer indefinitely many pieces of cognition, each about about pieces of cognition that are lower in the hierarchy. Perhaps those consequences are defensible. If they are not, the defender of (P1) will need to reject either (P2’) or (P3’). I take no stand here on which of those options constitutes their optimal response.

Advertisements
10 comments
  1. deontologistics said:

    Have you ever looked at Hegel’s critique of ‘Sense Certainty’ at the beginning of the Phenomenology of Spirit? I’ve always though that it’s the greatest example of someone anticipating and critiquing a philosophical position (namely, Russell on acquaintance) before it appears (though there is the possibility took up this position to spite Hegelians). It’s a somewhat different critique to what you present here, but I think you’d find it interesting. Either consult Stephen Houlgate or check out Brandom’s ‘Mediating the Immediate’ here: http://www.pitt.edu/~brandom/hegel/

  2. Thanks very much. I don’t know that bit of Hegel, but will look at that and the Brandom, and also consult with Stephen Houlgate. I think there’s certainly an element of engagement with Neo-Hegelians in this part of Russell’s thinking. It would be interesting to see how they pre-sponded.

  3. Karl Bentley said:

    Thanks for posting this as it links to some thinking I’ve been doing around Susan Haack’s ‘Foundherentism’. Still early stages, but the fog will clear eventually, I hope!

  4. Walter Horn said:

    Hi, Guy. I’m a little skeptical about the move from P1 to P2′. I mean the point of Russell’s piece is to explain propositional knowledge in terms of non-propositional knowledge. He took there to be a requirement that we have some OTHER type of knowledge of all of the constituents of any proposition P for it to be possible for us to understand P.

    But your move from P1 to P2′ is precisely to require that there be some sort of propositional knowledge around that can explain propositional knowledge. That’s why it seems to me unsurprising that you end up with a regress. Russell’s goal was to explain the very possibility of propositional knowledge–that can’t be done if we use propositional knowledge in our explanation.

    I do agree, though, that he doesn’t give much of an argument. He simply makes an appeal: “Hey, if I can understand ‘Smith is short’ then in some sense or other I must know who Smith is and what it is to be short, right?” And, while I do think there’s something appealing about that, as you point out, it’s not exactly a clear deduction. Suppose someone just responds (as I guess Wittgenstein eventually did), “Well, no. Propositions don’t actually resolve into parts in the way you’re imagining. It’s a misleading picture to take Fregean thoughts and pretend they’re Leggos.”

    Best,

    W

  5. Thanks for your interesting comment.

    You’re right, of course, that it’s part of Russell’s aim to (in some sense) explain propositional cognition, including knowledge, in terms of a form of non-propositional cognition (perhaps a form of knowledge). The note discusses whether Russell presents a cogent argument for his principle of acquaintance, a principle in effect connecting propositional cognition with acquaintance (a form of non-propositional cognition). I raise some worries about cogency.

    I think the move you suggest with respect to the (P1)–(P2′) transition is really dealt with earlier, in discussing the possibility of replacing (P1) with (P1′). That would perhaps serve the ends you mention, though in this context would, I think, beg the question in favour of Russell’s principle.

    What you suggest is that Russell would seek to resist the move from (P1)–(P2′). But then he would need an argument that that move can be resisted. One can’t simply stipulate that entailments fail to hold. Again, Russell might try to stipulate an understanding of (P1) on which the entailment fails. In that case, again, he would in effect be seeking to replace the superficially plausible claim in (P1) with something like (P1′), which lacks the patina of plausibility, and begs the question. At that stage, Russell would require serious argument for (P1′). As far as I can see, he doesn’t give such an argument, but my focus here was on one particular argument that Russell does give.

    The move you suggest is from something like

    (W1) S understands “Smith is short”

    to something like

    (W2) S knows who Smith is

    As you in effect suggest, one would like the argument for that move spelled out. (So, pursuit of the sort of reconstructive analysis that I aim for in the post.) It’s not clear exactly how that would go, and I shan’t attempt to pursue it here. But there will anyway be issues in mediating a transition to Russell’s conclusion, put here in terms of non-propositional cognition:

    (W3) S has non-propositional cognition (perhaps knowledge) of Smith

    For knowing-who, like other knowing-wh forms, is naturally contrued as a non-specific form of propositional knowledge, roughly a matter of knowing the answer to the question “Who is Smith?”. So, anyway, two tasks remain: (1) the task of spelling out the transition from (W1) to (W2); (2) the task of spelling out the transition from (W2) to (W3). (If one thinks that (W2) has a straightforward non-propositional reading, then (2) can be skipped, but the relevant reading must figure in performing task (1).)

    Finally, I don’t think this part of Russellian doctrine depends upon a building block conception of propositions. Wittgenstein ought to accept that there is Smith, in addition to the proposition that Smith is short. The question at issue here concerns broadly logical relations between cognition of the proposition and cognition of Smith. Such relations might obtain even if the proposition does not factor into Smith (or, say, a representation of Smith) plus other elements.

  6. Walter Horn said:

    Hi. I take up a few of your points below.

    You may be right that P’ already has this feature that concerns me (and I think would concern Russell), but I’m not certain because the sense of the expression “P knows that which they are making their cognizing about” isn’t clear to me. It seems ambiguous as between propositional and non-propositional knowledge. If I am “acquainted” in Russell’s sense with something, must I know anything about it? And, if not, do I know “that which I am cognizing about”? I simply don’t know the answer to that. P2′, on the other hand, is clearly a form of propositional knowledge.

    The question of whether P1 entails P2′ seems to be a “one man’s modus ponens….” thing. To me (and I think to Russell at that time) there is no entailment, precisely because P1 (understood correctly, at any rate) cannot require any propositional knowledge. That is, it must be construed in a way that allows somebody to “know” (be acquainted with) a patch without knowing (in the sense of propositional knowledge) a single thing about it. Thus, answers to the question “What is this patch?” or “Who is Smith?” would always seem to provide more than acquaintance in Russell’s sense.

    I’d think that there’s a sense of W2 in which one can’t get to it from W1, the sense in which one can (Russellian acquaintance) will not allow anything like P2′. That is, there’s a clear sense in which W1 will NOT allow us to answer the question “Who is Smith”–that is a sense in which Smith can be “known” without any proposition of the form ‘Smith is____’ is even believed (never mind known).

    I don’t like to go too far defending Russell’s concept of acquaintance, because it seems confused to me, and maybe wrong in just the ways that you are setting forth. But I would put it, rather, that there IS no such thing as Russellian acquaintance, instead of suggesting that if one looks at it closely enough it devolves into propositional knowledge. Maybe, it’s that I take this to be an empirico-psychological error of Russell’s rather than the conceptual mistake with which I take you to be accusing him. In the end, though, the criticism comes to pretty much the same, I think. Knowledge does, as you say, imply propositional acceptances of some kind, and Russell misses that.

    Best,

    W

    • (P1′) is to be read as generalising over ordinary forms of acquaintance knowledge, as in “Jill knows Peter”. Must one know anything about the object in order to have such knowledge? With respect to demands imposed by the semantics, as opposed to perhaps contingent demands on meeting the former demands, it’s not clear to me either way. (For example, it may be that it’s possible to know Peter without knowing anything about Peter, except for the contingent face that knowing humans typically involves seeing them, and seeing them is typically a way of coming to know stuff about them.) Anyway, it seems to be the available form that most closely approximates Russell’s needs. Note that he can’t just stipulate a better form without much further work, on pain of the input to his argument lacking intuitive plausibility. Finally, whatever stands here in Russell’s argument needs to be plausible independently of his conclusion, on pain of begging the question. The last point was the main objection at that point in my discussion. Note that the point about not begging the question holds also for (P1).

      With respect to whether (P1) entails propositional knowledge, this is in some sense a partly empirical question about the proper reading of the sentence (in particular, a proper reading on which it has some independent plausibility). One can’t simply decide that it does or doesn’t. As it happens, all the analyses of which I’m aware that have any plausibility carry the consequence, in part because the entailment seems so plausible to most speakers. (“Knows what” seems to differ in this respect from “sees what”, as the latter appears to some,at least, to allow a “sees that which” reading.) As it happens, though, this doesn’t really matter. For even if the “knows that which” reading were available, that would simply trigger the charge of begging the question. The point of discussing it here is just to trace out problematic consequences of generalised appeal to a “knowing what” condition on propositional cognition.

      Note that the point here isn’t that Russell’s position is mistaken. It’s that the argument presented in the passage I quote from Russell has superficial plausibility, but is in fact flawed. For present purposes, I’m neutral both about the position itself and about whether better arguments than this one are available, either in the rest of Russell’s corpus, or elsewhere.

  7. Walter Horn said:

    Yikes. Sorry about all the typos and other grammatical failures! I’m very bad about copy-editing. So, if any of my nonsense-disguised-as-sentences can’t be deciphered, please let me know, and I’ll try to clarify (making new messes as I go, no doubt).

    Please don’t take my sloppiness personally. FWIW, my motto at work (for about 30 years) has been “Nothing shall ever leave my desk without at least one mistake in it.” It’s like a special scent I have.

    W

  8. Walter Horn said:

    You write:

    “With respect to whether (P1) entails propositional knowledge, this is in some sense a partly empirical question about the proper reading of the sentence (in particular, a proper reading on which it has some independent plausibility). One can’t simply decide that it does or doesn’t.”

    I think that’s where we differ. IMO anything that is Russell’s knowledge by acquaintance cannot support any reading of P1 that entails propositional knowledge. Because, again, if a reading would support such an inference it is explicitly not what Russell was talking about. So, for me, the question is whether or not there IS anything like Russellian P1 stuff. I believe there isn’t–but, as you correctly point out, it’s because I AM just deciding that P1 does not entail propositional knowledge. I make this choice in an attempt to capture what I take Russell to have meant–not what the average Englishman might mean by a sentence of the same form.

    W

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: