Knowledge of Things
In chapter 5 of The Problems of Philosophy (“Knowledge by Acquaintance and Knowledge by Description”) Russell writes: “Knowledge of things, when it is of the kind we call knowledge by acquaintance, is essentially simpler than any knowledge of truths, and logically independent of knowledge of truths, though it would be rash to assume that human beings ever, in fact, have acquaintance with things without at the same time knowing some truth about them. Knowledge of things by description, on the contrary, always involves, as we shall find in the course of the present chapter, some knowledge of truths as its source and ground.”  He goes on: “We shall say that we have acquaintance with anything of which we are directly aware, without the intermediary of any process of inference or any knowledge of truths.” Thus Russell proposes a sharp duality in our knowledge of things: knowledge we have by bringing things under descriptions and knowledge we have by means of direct unmediated experience. As illustrations he cites knowledge of sense-data and knowledge of the external object that causes sense-data: we are “immediately conscious” of such sense-data as color and smoothness, he says, whereas we have no such direct awareness of the table itself, which we know only by means of the description “the physical object which causes such-and-such sense-data”. So our knowledge of things divides into two subspecies that act quite differently and are made possible by quite distinct mental faculties. The phrase “knowing X” is capable of two types of analysis depending on whether the knowledge relies on acquaintance or description. We might have thought we have a unitary concept here but in fact an irrefragable dualism confronts us: there are two completely different ways a thing may be known (or known of or known about).
The distinction appears robust when we follow Russell’s way of articulating it. We do know about sense-data by the faculty of introspection, which is different from the faculties of external perception and associated inference; and description is a linguistic matter while immediate awareness is not. There is also certainty concerning our acquaintance with our own minds, whereas we have no such certainty with respect to the external world. Further, it sounds intuitively correct to say that we have “direct” knowledge of some things and “indirect” knowledge of others. So Russell seems to be barking up the right tree when he announces a clear-cut distinction between two types of knowledge of things—as clear as the distinction between a priori and a posteriori knowledge, say. We are accustomed to such dualisms in philosophy and we can add this one to our stock. Certainly the distinction has been warmly received and smoothly transmitted down the philosophical generations. But is it as sharp and principled as it is easy to suppose that it is? Are there really two completely different ways of knowing about things with nothing to unify them? First, let us note two features of so-called knowledge by description as characterized by Russell: it is linguistically mediated, and it is dependent on knowledge of truths (propositions). These would certainly mark a key difference, but both are questionable, and indeed clearly wrong. It is not necessary to have and deploy a language in order to have so-called knowledge by description, or else animals and small children would have no such knowledge; what is necessary is the possession of concepts—ways of thinking about things. I can think of an external object as the cause of my sense-data without putting this thought into words—I just need to deploy the appropriate concepts. We can call these concepts “descriptive” if we like but that does not imply that they take a linguistic form (not in a public natural language anyway). Knowledge by description is therefore no more language-dependent than knowledge by acquaintance. Second, it is not part of the idea of knowledge by description that any knowledge of truths is essential: for a description is not a sentence; it does not express a proposition. Of course, if we accept Russell’s theory of descriptions, we get that result; but it is no part of the idea of knowledge by description itself. If we thought that some knowledge necessarily depends on names (“knowledge by naming”) that would not imply that such knowledge requires knowledge of truths (propositions). So again, that mark of distinction lapses. We are left with the claim that some knowledge of things is based on concepts—ways of conceiving of things, constituents of thought, elements of the reasoning faculty.
Attention now shifts to knowledge by acquaintance: is it not dependent on concepts? Russell appears to think so, but the question is debatable. He seems to think that our knowledge of sense-data has no conceptual component—that sense-data just loom up and slap us in the face, so to speak. But surely introspection involves bringing sense-data under concepts: I think of my sensation of red as a sensation of red when I introspect it. This means that I conceptualize sense-data according to their intrinsic character: I classify them, relate it to other sense-data, and have truth-bearing thoughts about them. I don’t just gaze at them blankly in uncomprehending wonder, as I might gaze at some strange animal quite unknown to me (imagine seeing an octopus for the first time). So when I am acquainted by introspection with my sense-data I apply concepts to them—whatever exactly concepts turn out to be. I am not cognitively void with respect to them. So isn’t knowledge by acquaintance a species of knowledge by description, i.e. concept-mediated knowledge? Consider your knowledge of the color red: you are acquainted with red and know what it is—is this knowledge completely independent of any means of mental representation? No, because you have to perceive the color in order to know what it is—it has to come before your mind. There must be some sort of intentionality involved: no cognition without mental representation. It isn’t that red just grabs your mind; your mind has to grab red, i.e. mentally represent it. There must be a mental act of grasping or perceiving or apprehending—some sort of intentionality. But then knowledge by acquaintance is like knowledge by description in that both involve a faculty of mental representation. It isn’t that knowledge of external objects requires intentionality while knowledge of internal objects doesn’t (which appears to be Russell’s view); both require it. The dualism totters; the distinction collapses; the division disappears. There is really only one type of knowledge of things—the type that builds in a system of mental representation. To be sure, we can distinguish different objects of knowledge, and record differences of epistemic status, and observe variations in the type of mental representation involved: but there is no deep distinction of the kind Russell claims—no difference of fundamental structure.
Interestingly, Russell himself essentially gives up on the contrast when he notes that all knowledge by description depends on knowledge by acquaintance. As he famously states: “Every proposition which we can understand must be composed wholly of constituents with which we are acquainted” (his italics). This means that all knowledge by description is reducible to knowledge by acquaintance—just not always acquaintance with the thing known. He should really have distinguished two types of knowledge by acquaintance: acquaintance with the thing itself and acquaintance with other things connected to that thing in various ways (as sense-data are connected causally to external objects). He could have called these “knowledge by direct acquaintance” and “knowledge by indirect acquaintance” or “knowledge by acquaintance with connected things”. If he had done that he wouldn’t have overdrawn the contrast with simple acquaintance by supposing that somehow language is essential to indirect knowledge. What he didn’t see is that acquaintance itself is also a matter of mental representation, and so is always a mode of “description”. All knowledge by description is knowledge by acquaintance and all knowledge by acquaintance is knowledge by description (in the wide sense). To put the point differently, some knowledge is by denotation (as applied to concepts as well as words) and some is by perception or introspection, but both of these are semantic relations in a broad sense. For example, I may know of a person merely by being told about him and I may know of a color by seeing it, but both of these relations involve a structure of intentionality: the distinction between methods of knowing is real, but it doesn’t entail the kind of cognitive duality postulated by Russell. Denoting and perceiving are indeed different relations, but both fall under a broader concept—and that concept unites the different instances of knowledge of things.
The same applies to knowledge of meanings: there are not two types of semantic knowledge, description-based and acquaintance-based. It isn’t that names are grasped “by description” and demonstratives for sense-data are grasped “by acquaintance”. In both cases we have mixture of direct acquaintance (possibly with universals) and indirect description (i.e. conceptualization); even when I grasp the meaning of “this pain” I invoke my introspective conception of pain, not just the pain itself. Nor are there two types of meaning corresponding to the different ways meanings are known—descriptive meanings and acquaintance meanings. A description theory of names will require a bedrock of acquaintance for the elements of the description; and a direct reference theory of mental demonstratives (“this pain”) will require a mode of conceptualization. Linguistic understanding is fundamentally uniform not divided into two completely different forms—the aloof inferential verbal form and the slap-in-the-face confrontational nonverbal form. This bears on the question of whether there are two types of definition, often labeled “verbal” and “ostensive”. I can give you some words to define a given word (e.g. “unmarried male” to define “bachelor”) or I can direct your attention to something in the world by pointing at it (e.g. I can point to something red in order to define “red”). So it is supposed that something fundamentally different is going on in the two cases: the word “definition” denotes two completely different procedures. But this distinction blurs when we examine things more closely: verbal definition depends on a prior grasp of the defining words, ultimately tracing back to something non-verbal; and so-called ostensive definition (as Wittgenstein taught us) is enmeshed in a sophisticated matrix of pre-existing conceptualization. The former is not so purely verbal as it might appear, and the latter is not so purely confrontational as we are inclined to suppose. We must already categorize the world in order to benefit from acts of ostensive definition, as we must already know the meaning of words non-verbally in order to benefit from acts of verbal definition. And doesn’t a verbal definition tacitly point to words in order to convey a meaning (“bachelor” means this: “unmarried male”)–as an ostensive definition converts the world into samples in order to convey a meaning, where samples act like elements of language? There is pointing in verbal definition (the “paratactic theory of definition”) and there is symbolism in ostensive definition (the “semantic theory of samples”).  These activities are not as distinct as we might casually suppose. There is no deep dualism of definition of the kind we tend to think. Distinctions can doubtless be drawn but they don’t add up to the sharp division by which philosophers are prone to be enchanted. Definitions are not cleanly divided into “definitions by description” and “definitions by acquaintance”, though words and perceptions can play different roles in different ways of getting meaning across in different contexts. It isn’t that some definitions involve simply pairing words with other words while others involve pairing words with chunks of reality, as if neither activity presupposes anything of the other. That would be an untenable dualism, a typically philosophical piece of distinction mongering. Acquaintance and description (so-called) bleed into each other. The terminology itself should be abandoned as misleading. 
 The book was first published in 1912 and has, I believe, exercised an enormous subliminal influence on the course of subsequent analytical philosophy, so much so that Russell’s distinction is taken for granted (I myself took it for granted until very recently). It is beautifully, and deceptively, seductive, largely owing to its confident and limpid style.
 On samples as parts of language see Wittgenstein, Philosophical Investigations, sections 16 and 50. The paratactic theory of definition would mirror Davidson’s paratactic theory of indirect discourse in “On Saying That”: “Oscar said that: We are all in the gutter but some of us are looking at the stars”. That is, that-clauses are analyzable as involving a demonstrative directed at a subsequent piece of discourse, as in “’Bachelor’ means that: ‘Unmarried male’”. In general, indexicality and description are closely intertwined.
 The terminology applies quite reasonably to another distinction, not Russell’s distinction–that between knowing about a thing by testimony and knowing about it by sense perception. Of course, it is true to say that testimony knowledge is acquired by means of “description”, i.e. verbal report, and also true that you can be “acquainted” with a person or place by actually seeing him, her or it, thereby acquiring knowledge: but that has nothing to do with Russell’s distinction. This is one of those cases in which a perfectly legitimate distinction drawn in ordinary language is used to confer spurious legitimacy on a quite different philosophical distinction. One always needs to study the fine print carefully and not be carried away by putative vernacular precursors.