False Knowledge

Is it true that all knowledge is knowledge of truths? Does the concept of knowledge entail that the proposition known is a true proposition? Certainly, we have been schooled to think so; and the idea is far from preposterous. But is the propositional content of the knowledge literally, universally, and necessarily true? We have to concede that not all knowledge involves justified belief: sometimes a creature can know without believing (many animals), and sometimes knowers have no justification for their knowledge (they know directly or intuitively or subconsciously). These requirements are too strong if taken strictly. However, it has generally been held that truth is non-negotiable: you can’t know what is false. But maybe this holds only for some or most knowledge; maybe some cases of knowledge are not beholden to truth. Perhaps we have been misled into a rash generalization by (allegedly) paradigm cases. In our thirst for generality, we have neglected certain peripheral or statistically rare cases. So, let’s explore some of the hinterlands of our cognitive life; we might turn up some unusual specimens (like black butterflies or crimson swans).[1]

First, we should pay attention to the concept of truth, which is not exactly uncontested. Truth, we say, is correspondence to fact—reality, actuality, existence. Truth is denoting an actually existing state of affairs, a real fact. There is no truth but reality makes it so, as Quine once said. All truth is literal truth. There is no your truth and my truth, just the truth. All truth is objective truth. So-called approximate truth is not really truth; it is falsehood that is close to the truth. Metaphors are not true, though they may allude to truth. To be true a proposition must describe a state of affairs in which certain objects really do have certain properties. We must not be sloppy with the concept of truth; truth is a strict concept. If someone uses the word “true” loosely, we might introduce the concept of “strict truth”, which obeys the principles just laid down (like “strict laws”). Then we could ask whether all knowledge requires strict truth. In any case, we have a notion of truth that meets the conditions laid down: good old-fashioned no-nonsense truth—realist truth, we might say (not imaginary truth, whatever that may be).

Now we must test our intuitions. Do I know that the Sun rises in the east (not the west or south)? If you ask me where the Sun rises, will I hesitate to answer or say “In the south, I believe”, or will I promptly and confidently reply “In the east, of course”? Surely, we would say that I have knowledge of where the Sun rises. I have observed that fact innumerable times. But is it true that the Sun rises in the east? No, because the Sun doesn’t rise at all—the Earth rotates. It is not a real objective fact that the Sun rises in the east; this is a kind of illusion. Do I know the color of my coffee cup? Do I know that it is blue not red or some other color? We would surely say so. But is it true that my cup is blue? Not if it has no color at all—that is, if color does not belong to physical things. Do I know that Great Britain is triangular (not square or circular)? Yes, I do know that, if I know anything. And yet that land mass is not really and truly triangular; this is a false statement that only approximates to a true statement. Do I know that my car didn’t move all day? Yes, but of course it is not true that it didn’t move, because the Earth moves. Do I know that the eyes are the windows to the soul? Yes, I do know that—but it is not literally true that the eyes are the windows to the soul. Granted, this metaphor is related to a true proposition (“the eyes are sources of information about someone’s state of mind”), but it isn’t really true; yet I can be said to know it. Do I know that my friend is bald? Yeah, I know this, but it isn’t really true—he has some hair on his head, quite a lot in fact. Do I know that Smith is a damned fool? Yes, but it isn’t true that he is damned. Do I know that Hamlet was prince of Denmark? Sure, but it is not true that he was (look at the history books). Truth is stricter than knowledge, more demanding. In addition to this, not all knowledge is propositional; there is also knowledge of things (as Russell insisted). Knowledge of things by acquaintance is not knowledge of propositional truths; it is not propositional at all. Nor is knowing-how a bearer of truth. The concept of knowledge, in its full generality, is not necessarily tied to the concept of truth; that is the case only in certain cases not in all. To know that p is not necessarily to know that it is true that p, since in many cases p is not true. You can be said to know what is false, though it may have to be tied to something true; but that doesn’t make it true. The connection to truth is indirect, if it exists at all. If we wanted to retain the old style of definition, we would need to say something like, “X knows that p only if p is connected to some proposition q such that q is true”.

What is going on here? Why doesn’t knowledge precisely track truth? The answer lies in the function of both concepts. The concept of knowledge is used to assess someone’s epistemic credentials; the concept of truth is used to characterize the objective facts of reality. You can be said to know if you are a reliable indicator of reality, if only a rough indicator; a proposition is true if (but only if) it corresponds exactly to how things objectively are. Truth requires strict isomorphism (to borrow from Wittgenstein); knowledge requires a useful degree of fit. The latter is pragmatic, but the former is metaphysical (mathematical almost). Truth is formalizable; knowledge is humanistic. Truth is strict; knowledge is lenient. Knowledge is about passing the exam; truth is about how things really and genuinely are. You are not going to make any practical errors by believing the Sun rises in the east (even literally), but it is quite false to assert that the Sun rises in the east. Eyes aren’t truly windows, though talking this way shows you know how eyes function in human interactions. That is why we aren’t too pedantic in our attributions of knowledge, but we can become quite schoolmasterly if pressed about the truth. It really wouldn’t matter if all knowledge were of literal falsehoods, so long as the corresponding beliefs didn’t land us in too much trouble; but we would still insist that truth is truth and falsehood is falsehood. To take a classic example, you can know that the Battle of Hastings took place in 1066 even if, strictly speaking, it began in 1065 on New Year’s Eve. True, we can’t get too lax about knowledge, but we are laxer than we are about truth, more forgiving.

This bears on the question of ethical knowledge. The possibility of ethical knowledge is not hostage to a thick notion of ethical truth. You can be said to know ethical propositions without those propositions being true. They might have no truth-value, being logically imperatival, and yet they could still be known: we can say “he knows that stealing is wrong” without being committed to the (literal) truth of “stealing is wrong”. For knowledge is not necessarily truth-entailing, though it may be reliability-entailing. You can make inferences about the epistemic credentials of the person in question, but there is no requirement to infer the truth of “stealing is wrong”. You can even hold an error theory about ethical statements while accepting that people have ethical knowledge. A person can know it’s wrong to steal even if “stealing is wrong” is false as a statement of fact or a pseudo-statement. Ethical knowledge does not imply ethical truth (though there may well be such a thing). If we put this together with other criticisms of the true-justified-belief account of knowledge, we can say that “X knows that p” is consistent with “it is false that p, X does not believe that p, and X has no justification for believing thatp”.  These conditions are all too strong, though they may apply in many or most cases. Broadly speaking, they are too intellectualist. Not all knowledge is like scientific knowledge; some knowledge is more rough and tumble than that. Often all that is required is acquaintance with a suitable fact, not grasp of a literally true proposition. Propositional truth is strictly irrelevant to knowledge, broadly understood, as is language. Birds and bees know nothing of truth and propositions, yet they know.[2]

[1] I know, reader, you are skeptical—have I gone mad? False knowledge! But bear with me; the Earth once seemed self-evidently stationary.

[2] They perceive facts, record them in memory, and act on them, which is the essence of knowledge. Believing true propositions is strictly separate. Thus, you can know without believing truths. Humans see the Sun appearing in the east and remember what they have seen, thereby knowing where the Sun appears in the morning. They express this knowledge in the sentence “the Sun rises in the east”, with its accompanying proposition. This proposition is false, but that doesn‘t undermine their status as knowers of the relevant fact. Knowing facts is one thing, believing true propositions is another: see my “Perceptual Knowledge” and “Non-Perceptual Knowledge”.

Descriptions and Names

The distinction between names and descriptions is not as sharp as we tend to suppose. We are prone to think that names are purely denotative (tags, labels) and descriptions are purely connotative (attributional, predicational), but actually the two overlap. If you favor a description theory of names, you are still up to your neck in names; and if you prefer to take names as primitive, you are still wallowing in descriptive content. There is no such thing as pure description or pure denotation. Pure description theories of descriptions are false, and pure denotation theories of names are false. Both are both. They are not mutually exclusive.

In the case of names, there are several kinds of descriptive content that they carry: a certain natural kind, male or female, family lineage, nationality, language spoken, social class, historical period, being called by the name in question, etc. A person named “Susan Smith” is thereby classified as a human being, female, of the Smith family, likely to be of British descent, a speaker of English, probably not upper class, born at a certain period of history, not nameless, and called “Susan Smith”. There are no names that are purely denotational (“logically proper names”), having no meaning apart from their bearer; that idea is a myth. And we all tend, as users of names, to know these descriptions, or someone in the linguistic community does. They also play a role in fixing the reference of the name—they tell you who or what is in question. They are not purely decorative; they have a semantic function. No one has ever spoken a language in which the names lack such descriptive content. We might even say that names necessarily have such content; it is essential to their being names. The whole institution of names, from baptism to burial, womb to tomb, is steeped in these connotations. Names have sense as well as reference; they contain information. It is not possible to analyze them by means of descriptions that don’t express this sense. Their meaning is not exhausted by their bearer. They are not semantically simple.

Why do I say that descriptions are always name-involving? Two reasons: one, they often contain proper names, and two, the predicates they contain are themselves name-like. Thus “the capital of France” and “the cat in the corner”.  They need to contain names because we don’t typically have purely general descriptions to hand (like “the first dog born at sea”), since we often lack that kind of individuating knowledge; and terms like “cat” are precisely names of natural zoological kinds. They are not, on their face, definite descriptions of the denoted natural kind. Also, adjectives and artifact terms are name-like: “square”, “blue”, “table”, “television”. These all stand for (denote) some attribute or kind of thing: they are classified as “common names”. Naming is not restricted to names of concrete individuals; we name qualities and kinds as well as particulars. So, descriptions don’t dispense with names; they depend on names. You don’t get rid of names by replacing them with descriptions, because the description will contain further names, particular or general. The idea of a purely descriptive name-free language is also a myth. Actually, all language depends on names; we can’t eliminate names and replace them with descriptions, because descriptions are made of names. If we can’t make sense of names, we can’t make sense of language. Language consists of names (“and” is the English name for logical conjunction).

Thus, names are inherently descriptive and descriptions are inherently nominative. The distinction is bogus. Every word of a natural language is both descriptive and nominative—a description and a name. Sense and reference, denotation and connotation, intension and extension. Meaning is inherently connotative and denotative simultaneously. There can’t be theories that are connotational but not denotational or vice versa. Language relies on both working together. The quest for a theory that is purely one or the other is quixotic. The words “Susan Smith” and “navy blue” work in much the same way—both are descriptive names of something (a person or a quality). The name-description distinction, as commonly understood, is an untenable dualism.

Descriptions in Disguise

One might have thought that the description theory of names had been bashed enough; I propose to bash it some more. It hasn’t been bashed from every angle; indeed, it positively invites further bashing, given its audacity. The theory is quite disrespectful to names, insinuating that descriptions are the superior semantic citizen. We can break the question down into three parts, corresponding to three locutions that crop up in these investigations: disguise, synonymy, and abbreviation. Names are said to be disguised descriptions, synonymous with descriptions, and abbreviations of descriptions—are any of these statements true? Each locution denotes a two-place relation: are names and descriptions really related by these relations? I am not discussing the usual questions of speaker knowledge, referential rigidity, and scope distinctions; I am focusing on the correct characterization of the relation between names and descriptions—is it one of disguise, synonymy, and abbreviation?

First, disguise: is it true that a name is a description in disguise? This implies that you might be fooled by a name into thinking it is not a description when it really is one. It doesn’t look like a description, but we are told that it is one. But how is that possible given the theory? For synonymy is not open to the possibility of disguise: if “a” is synonymous with “b”, then anyone competent with both expressions will know that they are synonymous; it will be quite self-evident. If you know what “bachelor” means and you know what “unmarried man” means, then you know a priori and with certainty that they are synonyms; you can’t wonder whether they are. But you can wonder whether the name “Plato” is synonymous with the description “the teacher of Aristotle”—that is why it is a philosophical question whether it is. So, the name can’t be a description in disguise, because then it would be an open question whether the name is synonymous with the description; but it can’t be an open question if they are really synonyms. The two ideas are inconsistent. And how is the disguise supposed to work—is it like some sort of mask? How does a description disguise itself as a name—that is, how does the meaning of a description disguise itself as the meaning of a name? Does it pretend to belong to a different semantic category by curling up into a ball or by wearing a different outfit? The whole notion of disguise here is pure metaphor. How can meanings come in disguises? If they do, they can escape detection and pass for another kind of meaning; but that is incompatible with the description theory, which insists on synonymy. Synonymy is transparent, but disguise is opaque. And don’t say the disguise takes the form of a different word, because that is true of ordinary synonyms, which are semantically transparent. Theorists have opted for the word “disguise” because names are plainly not ordinary synonyms of descriptions, so they need to introduce some machinery to explain this fact. Are there any bona fide cases of semantic disguise in language—cases in which we would all agree that one kind of expression is disguising itself as another? Not that I can see; the idea is manifestly ad hoc. The word “bachelor” is not a simple noun that acts as a disguise for the adjective-noun combination “unmarried male”; there is nothing misleading or deceptive about it. We don’t have to rip off the disguise to reveal the synonymy.

Second, synonymy itself: are names and descriptions synonymous? We know about the counterexamples that have been produced against this claim (Kripke et al), but there is another way to question it. The relation of synonymy is symmetrical, so descriptions must be synonymous with names, actual or possible. So, why don’t we have name theories of descriptions? The description means the same as the name (suppose we started with a direct reference theory of names). Then we would have a non-descriptive purely referential theory of the meaning of definite descriptions! Dialectically, we have begged the question against the direct reference theory. But there is this further point: why aren’t all descriptions synonymous with names not just some? For every name there is a corresponding description (allegedly), but why not conversely? There could be if speakers wanted it that way—just introduce one! The name is really always loitering in latent form, waiting to be recognized; the semantic category of names is implicit in the semantic category of descriptions, in virtue of the alleged synonymy. But that seems implausible: names look like a genuine enrichment of the semantic resources of a language. And why bother with the category of descriptions if you already have the category of names that are synonymous with them? Surely, we are dealing with two distinct categories of expression. Descriptions are not disguised names! Why are some descriptions synonymous with (existing) names while some are not? That is puzzling under the description theory.

Now we reach the nub of the description theory of names: names are said to abbreviate descriptions. The disguise comes from this because an abbreviation doesn’t look or sound like what it abbreviates; and the reason names don’t exist for every description is that speakers don’t need convenient abbreviations for all descriptions. Names exist in the presence of descriptions because they are a handy condensation of their lengthier brethren. It is a pragmatic matter. Names are not semantically useful—there is nothing they can do semantically that descriptions can’t do better–but they are pragmatically useful. That is the reason for the apparent difference, and the excuse for the metaphor of disguise. But this is a terrible idea—absolutely pathetic, in fact. There are two points, both perfectly obvious, yet devastating. First, names are not syntactically or orthographically or aurally abbreviations of descriptions—they look and sound nothing like descriptions. The name “Mike” is an abbreviation of “Michael”, but it is not an abbreviation of “the guy who owns the local pawn shop”; it is a distinct expression entirely. Second, names are not always shorter than descriptions, and certainly not necessarily shorter: people sometimes have long names! Even when they are relatively short people sometimes prefer to use a description, because it is more informative. Names are not short forms of descriptions in any meaningful sense. This is just a lazy label cooked up to paper over a glaring weakness in the description theory, namely that names are nothing like descriptions syntactically. We concede too much to the description theory if we let this patch-up job pass.

There is a more general point to be made: if a description theory of demonstratives is incorrect and seldom advocated, why should a description theory of names be thought compulsory? Names are naturally akin to demonstratives, so why not group them with demonstratives instead of descriptions? Neither type of expression requires extensive knowledge of the referent; names are commonly introduced via a demonstrative (“this child”); and context plays an important role in both cases—so why not align these two categories semantically? This is a semantic natural kind, unlike the natural kind of definite descriptions, which belongs with indefinite descriptions and other descriptive devices. It would surely be highly implausible to suggest that a demonstrative like “that man” is synonymous with some description referring to the person’s famous deeds or some such. Does “he” as uttered on an occasion mean “the guy drinking a martini”? So, it is a reasonable challenge to the description theorist to explain why he insists on that theory for names but not for demonstratives. There are clearly more referential devices in natural languages than those provided by the definite description, which anyway seems destined for a Russellian paraphrase and doesn’t refer at all. Enough bashing.[1]

[1] Why is that in philosophy all the best theories are false? The description theory is a nice theory, but it is riddled with problems. Likewise, materialism is an excellent theory, if it weren’t for the falsity. Ditto for emotivism, nominalism, panpsychism, phenomenalism, and all the rest.