The Knowledge Argument Adam Vinueza Department of Philosophy, University of Colorado vinueza@colorado.edu Keywords: acquaintance, fact, physicalism, proposition, qualia. The Knowledge Argument and Its Significance The Knowledge Argument in Historical Context Responses to the Knowledge Argument Conclusion The knowledge argument purports to show that there are non-physical facts facts that cannot be expressed in physical terms on the ground that one can know all the physical facts without knowing facts about what it is like to have an experience. Introduction The knowledge argument is against physicalism, the doctrine that all facts are physical facts. Physicalism is a notoriously difficult doctrine to state clearly and plausibly, but we can explicate its main idea by taking a physical fact to be a fact that can be expressed in the language of physics. Thus, to say that physicalism is true is to say that all facts can be expressed in the language of physics. The knowledge argument purports to show that there are facts about what it is like to have an experience, and that these facts cannot be expressed in the language of physics. What does it mean to say that there is something it is like to have an experience? If you have never tasted asparagus (or anything relevantly like it), you do not know what it is like to taste asparagus: there is a qualitative aspect to tasting asparagus that you have never experienced. Likewise, if you are an achromat, you have never had color experiences, so you do not know what it is like to have such experiences. These aspects of experience are called qualia (singular quale): there is a unique quale to tasting 1
asparagus, a distinct unique quale to seeing vermilion, and so on. Thus, the knowledge argument purports to show that certain facts about qualia cannot be expressed in the language of physics. The Knowledge Argument and Its Significance The knowledge argument (Jackson, 1982, 1986) is standardly introduced with the following thought experiment. Imagine that some person we shall call Mary has been confined from birth to a black-and-white room, and has been educated through black-and-white books and through lectures and discussions relayed on black-and-white television. Let us also imagine that Mary, in this room, has been taught all the physical facts: she has learned all the facts one can learn through studying physics, chemistry, neurophysiology, and so on, and that these theories are all complete and correct. Then if physicalism is true, Mary knows everything, because the physical facts are all the facts. But according to Jackson, Mary does not know everything, because if she were to leave her black-and-white environment she would see her first color: she would have her first experiences with color qualia, and would thereby come to know that is, learn what it is like to have color experiences. But if she knows all the physical facts before leaving the black-and-white room and learns what it is like to have color experiences upon leaving it, then what it is like to have color experiences is something Mary learns that is not a physical fact. So, according to Jackson, there are things one can know, some facts, that are not physical facts. If the knowledge argument is sound, physicalism is false. But physicalism is a methodological assumption that guides all scientific inquiry. We reject explanations that are inconsistent with physical explanations such as explanations of events by appeal to miracles on the ground that they are physically impossible, and it is arguable that the only way we can make sense of this dependence of all explanations on physical explanations is by supposing that all the facts, ultimately, are physical facts. Physicalism is of course also foundational for cognitive science, which is guided by the hypothesis that all features of our mental lives can be explained by appeal to functional (in particular, computational) processes. So if the knowledge argument is sound, the central hypothesis guiding cognitive science is simply false, and cognitive science is arguably misguided. That the knowledge argument has such powerful consequences can easily lead one to reject it out of hand: one might think that if it entails that physicalism is false, then the argument simply has to be unsound. But it is one thing to be convinced that it is unsound, and another to explain precisely where it goes wrong. In this article, we 2
will see the main ways philosophers have tried to show that the knowledge argument is unsound, and examine their virtues and faults. The Knowledge Argument in Historical Context The intuition motivating the knowledge argument goes back at least to Leibniz, who argued that mentality cannot be explained by appeal to physical principles. Leibniz imagined a machine that thinks and perceives, and claimed that if we looked inside, we would never see anything that could explain perceiving. (See Leibniz, 1981, pp. 66-7.) Leibniz clearly had an intuition akin to the intuition that Mary would learn something upon seeing her first color: knowing all the physical properties of a thing, for Leibniz, does not tell one anything about what perceptions are like. Leibniz s intuition was almost universally accepted by philosophers from the seventeenth century until the middle of the twentieth century, when philosophical orthodoxy shifted from Cartesian dualism to materialism. While materialism is still the orthodox view, challenges to it arose in the past thirty years, mainly because of the sense many philosophers had that materialist explanations of mentality were deeply unsatisfying. This dissatisfaction had the same source that led Leibniz and others to reject materialism: materialist explanations seemed to them to be unable to capture the nature of qualia. A recent expression of this dissatisfaction, which just preceded the knowledge argument, is from Nagel (1974). Nagel argued that knowing all there is to know physically about the mechanisms underlying the echolocatory abilities of bats would not enable one to know what it is like for a bat to echolocate, because no one has any idea of what it is like for bats to echolocate. Nagel concluded that our abilities to explain qualia are limited by our abilities to conceive qualia, hence that the project of explaining mentality in physical terms is doomed to failure. Nagel s argument is often cited as a version of the knowledge argument, but it is important to note that its conclusion is weaker: Nagel concludes only that we can never know whether physicalism is true (because we cannot form a conception of all the phenomena we wish to explain), not that physicalism is false. Nevertheless, Nagel s argument is an important precursor to the knowledge argument, and is close kin to it in spirit. 3
Responses to the Knowledge Argument Let us see the knowledge argument more formally, so as to more easily classify responses to it: (1) Mary comes to know a fact about her experiences when she leaves her black-and-white environment viz., what it is like for her to see colors. (2) Mary s knowledge of what it is like to see colors is something she did not know before leaving her black-and-white environment. (3) If physicalism is true, then Mary s knowledge of what it is like to see colors is not a fact she did not know before leaving her black-and-white environment. (4) Therefore, physicalism is false. Premise (3) follows directly from physicalism plus the assumption that Mary knows all the physical facts, so the only premises that can be questioned are (1) and (2). Those who challenge the knowledge argument, then, do so in one of two ways: they deny that what Mary comes to know is a fact (premise (1)), or they deny that it is a fact she did not know before leaving her black-and-white environment (premise (2)). Let us call a response of the former sort a no-fact response; let us call one of the latter sort an old-fact response. Before looking at these responses, though, let us take a brief look at two initial responses many have been tempted to make, to see why they fail. The assumption that Mary knows all the physical facts is of course an implausible one: not only is it unlikely that we will ever be in a position to know all the physical facts (such as facts about physically inaccessible regions of the universe), but it is also unlikely that any one person could have the cognitive capacity to know even a tiny fraction of these facts. Thus, one tempting response is to conclude that it is this assumption that what is wrong with the knowledge argument. We should resist this temptation, however. No doubt the physical facts that constitute what it is like for Mary to have color experiences, if they exist, are knowable; so we may suppose that Mary knows only those facts (as well as whatever facts are necessary in order to know them), and the argument would still go through. The assumption that Mary knows all the physical facts is a harmless simplification. Another tempting response involves pointing out that we are often unable to draw conclusions that follow from what we know, either because doing so would involve too much time, or because the inferences are too complex for us to figure out. Thus, we might think that Mary really does know what it is like to have color experiences in virtue of knowing all the physical facts, but has simply been unable to draw the relevant inferences. 4
Unfortunately for this response, however, we could give Mary as much time and computational resources she would ever need to draw inferences from her knowledge, and she arguably still would not be able to infer what it is like to have color experiences. Intuitively, the only thing that would give her such knowledge is her having color experiences. So we cannot plausibly blame Mary s ignorance on her inability to draw inferences. No-Fact Responses No-fact responses themselves come in two forms. The most popular form is known as the ability hypothesis: the hypothesis that knowing what it is like to have an experience is not propositional knowledge (knowledge that something is the case), but instead merely know-how, the possession of an ability (Nemirow 1990; Lewis 1983, 1988; Churchland, 1985). The other form of no-fact response holds that knowing what it is like to have an experience is what Russell (1910-11) called knowledge by acquaintance, an intuitively immediate, nonpropositional awareness of a property (Conee 1990). The ability hypothesis holds that knowing what it is like to have an experience is knowing how to imagine and/or re-identify that experience. No doubt, knowing how to do many things involves factual knowledge: knowing how to drive a car requires knowing many facts about how the car works, for example. But one can know all the facts about cars without knowing how to drive a car. Given that Mary surely gains abilities when she leaves the black-and-white room among other things, she gains the ability to imagine and re-identify various new color experiences it is arguable that Mary s knowing what it is like to have an experience simply consists in her having those abilities. And because it is plausible that one cannot imagine or re-identify experiences one has never had, ability hypothesists can explain why Mary learns something upon leaving her black-and-white environment without having to deny physicalism, because what Mary acquires is not factual knowledge at all, but know-how. The ability hypothesis is controversial, not least because our best cognitive theories of know-how in general explain it as the possession of tacit factual knowledge: for example, our ability to produce and understand sentences of a natural language is paradigmatic know-how, but is arguably best explained in terms of tacit knowledge of a grammar for that language, a set of rules determining the grammatical structures of its sentences (Chomsky, 1965). Another argument against the ability hypothesis is a semantic one: the semantics of knows wh- constructions in general treats them as being true in virtue of the knower having factual knowledge. For example, to know where Tanzania is to know that it is at such-and-such a place, and to know why Smith murdered Jones is to 5
know that Smith murdered Jones for such-and-such a reason. By parity of reasoning, to know what it is like to have an experience is arguably to know that having that experience is like such-and-such (Lycan, 1996). And if semantics dictates that sentences of the form X knows what it is like to have experience E are true in virtue of the knower knowing that something is the case, then the ability hypothesis cannot be true. The acquaintance hypothesis holds that knowing what it is like to have an experience is to be acquainted with it, where knowledge by acquaintance is an intuitively immediate (viz., non-inferential) and non-propositional form of knowledge. Paradigms of acquaintance include the awareness we have of our own conscious mental states, and our perceptual awareness of colored patches in the visual field. Nemirow (1990) and Churchland (1985) offered versions of the acquaintance hypothesis, but analyzed acquiantance in terms of the possession of abilities; for this reason, many identify the acquaintance hypothesis with the ability hypothesis. Conee (1990) does not analyze acquaintance in this way, but does not say precisely what knowledge by acquaintance is. Unfortunately, the acquaintance hypothesis faces precisely objections analogous to those just given to the ability hypothesis. One can argue that acquaintance understood as perceptual awareness, at least is best characterized as awareness that something is the case, because perceptual awareness is typically characterized by cognitive scientists in computational terms in terms of rule-governed manipulations of mental representations and mental representations are typically characterized in this way. And Lycan s argument about the semantics of knows what it is like applies with equal force to the acquaintance hypothesis. Note that these objections to the ability and acquaintance hypotheses are not objections against either the view that knowing what it is like is an ability or the view that it is a kind of acquaintance; they are that knowing what it is like, whatever it is, is nevertheless a kind of factual knowledge. Thus, it is open to these theorists to revise their positions slightly, by accepting that knowing what it is like is a kind of factual knowledge, and that it is, under an unfamiliar guise, knowledge Mary already has. If they do so, however, they are offering a kind of old-fact response, not a no-fact response. It is to this sort of response that I now turn. Old-Fact Responses Old-fact responses hold that while Mary does come to know a fact when she leaves her black-and-white environment, this fact is just one of the facts she already knew. On certain views of facts, one can know the same fact in various ways: for example, one might think that the fact that Mark Twain is Mark Twain is identical to the 6
fact that Mark Twain is Samuel Clemens, even though one can know that Mark Twain is Mark Twain without knowing that Mark Twain is Samuel Clemens. Knowledge, arguably, is a relation between a knower and a true proposition, and true propositions are on certain views more plentiful than facts: different propositions can be true in virtue of the same facts holding. On this view, then, to learn a fact is to learn a proposition that is true in virtue of that fact holding; hence, one can learn the same fact in as many different ways as there are propositions corresponding to that fact. (Some hold that propositions and facts are the same sort of thing, but also that knowledge is a relation between knowers and guises of propositions, where guises bear a many-one relation to propositions. Because there is arguably only a verbal difference for our purposes between the view just described and this one, I will discuss only the former.) Old-fact responses are by far the most popular sort of response to the knowledge argument. (See, for example: Churchland, 1985; Bigelow and Pargetter, 1990; Loar, 1990; Lycan, 1990, 1995, and 1996; Pereboom 1994; and Tye, 1986.) The burden on those who make this sort of response is to make it plausible that what Mary learns is an old fact under the guise of a different proposition. This is standardly done by drawing analogies between Mary s situation and situations in which it is uncontroversial that someone comes to know a new proposition without learning any new facts. For example, suppose some person Bill knows that Mark Twain wrote Huckleberry Finn, but does not know that Mark Twain is a pseudonym for Samuel Clemens. When he learns that Samuel Clemens wrote Huckleberry Finn, it is plausible that this is no new fact for Bill, but simply one of the facts Bill already knows, under a different guise. (Bill may thereby learn that Mark Twain and Samuel Clemens are the same person, but the point is that while the two propositions about who wrote Huckleberry Finn are distinct, they correspond to a single fact.) If what Mary learns in learning what it is like (say) to see red is factual knowledge, then we may suppose that what Mary learns is that seeing red is like Q, where Q is the name of the quale Mary has when she sees red things. Therefore, on the old-fact view, Mary already knows that seeing red is like Q, only in virtue of her knowing a different proposition. How plausible is this? The holder of the old-fact view argues in the following way. Plausibly, the propositions that Mark Twain is Samuel Clemens and that Mark Twain is Mark Twain are distinct, yet the fact that Mark Twain is Samuel Clemens is the same fact as the fact that Mark Twain is Mark Twain; analogously, the propositions that seeing red is being in a certain physical state P and that seeing red is like Q are distinct, yet correspond to the same facts, because being in P just is being in a state like Q. 7
The main challenge for this sort of response derives from the standard explanation of not knowing a proposition while knowing the fact corresponding to it, viz., that one knows some but not all of a thing s properties. If someone knows that Mark Twain wrote Huckleberry Finn but not that Samuel Clemens did, that is arguably because there are some properties of Twain she does not realize he has viz., being named Samuel Clemens. But Mary is not in this position: if she knows all the physical facts and physicalism is true, she knows all the properties of every thing. This means that one who maintains an old-fact response must offer some other explanation of not knowing a proposition while knowing the fact corresponding to it. Appropriately old-fact theorists have offered a variety of alternative explanations. I will mention only the most prominent alternative, from Loar, 1990: that there are concepts whose possession requires having certain experiences, that some concepts of color qualia are like this, and that Mary has not had the experiences requisite for having such concepts of color qualia. These concepts, called phenomenal concepts, are concepts that enable one to recognize one s own qualia when one has them: they are those concepts in virtue of which one can self-ascribe sensations. (For related closely views see Lycan, 1995 and 1996, and Tye, 1986). If knowing what it is like to have an experience with quale Q is knowing that the experience is like Q, and the concepts of Q required for having this knowledge are phenomenal concepts, then old-fact theorists can explain why Mary does not know what it is like to have color experiences by pointing out that she lacks the requisite concepts, while denying that knowing what it is like to have experiences is knowledge of distinct facts. Interestingly, this sort of old-fact response leads to a different, though related worry. According to Leibniz s Law, if any thing X is identical to a thing Y, then whatever is true of X is true of Y. Leibniz s Law has implications for modal claims about identity claims about necessity and possibility. Because X is necessarily identical to X, it follows from Leibniz s Law that X is necessarily identical to Y. Nevertheless, many identities can seem to hold only contingently, because one can coherently conceive them to be false, and we need to do empirical investigation to find out whether they hold: the standard example is the identity of water and H 2 O. This would suggest that what we can coherently conceive is not a reliable guide to what is possible: water is necessarily identical to H 2 O, but we can coherently conceive that water is not H 2 O. But an influential argument from Kripke seems to show that these sorts of examples do not show that conceivability is no guide to possibility, and Kripke s argument for this leads to a real problem for old-fact responses. (Kripke, 1980; for further developments of this argument, see Chalmers, 1996 and Jackson, 1997.) 8
Kripke s argument is as follows. Because water is necessarily identical to H 2 O, there is no possible situation in which water is not H 2 O; therefore, when we claim to be able to conceive that water is not H 2 O, we are either not conceiving any situation at all, or what we are conceiving is something that we might legitimately confuse with this impossible situation. When we claim to be able to conceive that water is not H 2 O, Kripke argues, we are conceiving a situation in which something which has the properties we normally associate with water say, being a colorless, odorless potable liquid that flows in our rivers and streams is not H 2 O. This something would surely seem to be water, although it would not be (given that water is H 2 O). No doubt such a situation is possible, but it is not a situation in which water is not H 2 O, because the latter situation is impossible. Plausibly, we mistake that situation for the latter one; and this explains why it can seem possible that water is not H 2 O. We mistake a possibility in which something that seems to be water is not H 2 O for a possibility in which water is not H 2 O. Now, physicalists are committed to the view that certain physical properties are identical to qualia, and oldfact theorists claim that phenomenal concepts pick out the very same properties as certain physical concepts. Because these properties are identical, they are necessarily so; but because one can fail to realize that phenomenal concepts and physical concepts pick out the same properties, it may well seem to one that the properties picked out are not the same. Hence, it can seem possible that some physical property P is not identical to a quale Q, even if they are identical. Can we explain this seeming in the same way? Kripke argues that we cannot. The reason is that while there are situations in which something that seems to be water is not water, there is no situation in which something that seems to be a certain quale is not that quale. Whenever it seems to one, say, that one is in pain, Kripke argues, one is in pain; so we cannot say in this case that we are mistaking a possibility in which something that seems to be pain is not such-and-such a physical property for a possibility in which pain is not that physical property. If this is the correct explanation for why identities can seem contingent, old-fact theorists are left with being unable to explain why the identities of qualia with physical properties can seem contingent. And if they cannot explain this, one may legitimately suppose that the fact that we can coherently deny these identities means that they simply do not hold that is, that qualia are not physical properties, and physicalism is false. Of course, this argument is not Jackson s knowledge argument, but an important related one. Nevertheless, anyone who claims that Mary simply learns an old fact under a new guise when she comes to learn what it is like to have color experiences must have some response to it. The most current discussions of the knowledge argument tend focus on how such a response must go, and the main thrust has been to argue that we have independent reasons 9
for thinking that conceivability is no reliable guide to possibility: a good representative argument is from Yablo, 1999. Conclusion As yet, there is no decisive refutation of the knowledge argument. That there is no decisive refutation is of course no reason for theorists in cognitive science to reassess the import of their own research; but it should make them pause to think about the philosophical issues their research raises. It would be nice if we could simply dismiss the knowledge argument as a piece of fallacious reasoning, but we cannot. That it has survived all current attempts at refutation show that solving the mind-body problem calls not only for a good empirical theory of mind, but also for clear, careful thinking about such philosophical issues as the nature of facts and the epistemology of modal truths. References Bigelow, J and Pargetter, R (1994) Acquaintance with Qualia. Theoria 61: 129-47. Chalmers, D (1996) The Conscious Mind. Oxford: Oxford University Press. Chomsky, N (1965) Aspects of the Theory of Syntax. Cambridge, MA: The MIT Press. Churchland, P (1985) Reduction, Qualia, and the Direct Introspection of Brain States. Journal of Philosophy 82: 8-28. Conee, E (1990) Phenomenal Knowledge. Australasian Journal of Philosophy 72: 136-50. Jackson, F (1982) Epiphenomenal Qualia. Philosophical Quarterly 32: 127-36. Jackson, F (1986) What Mary Didn t Know. Journal of Philosophy 83: 291-95. Kripke, S (1980) Naming and Necessity. Cambridge, MA: Harvard University Press. Leibniz, G (1981) New Essays on Human Understanding, Remnant, P and Bennett, J (trs). Cambridge: Cambridge University Press. Lewis, D (1983) Postscript to Mad Pain and Martian Pain, In: Lewis, D, Philosophical Papers, Volume I. Oxford: Oxford University Press. Lewis, D (1988) What Experience Teaches. Proceedings of the Russellian Society. Sydney: University of Sydney Press. 10
Loar, B (1990) Phenomenal States. In: Tomberlin, J (ed) Philosophical Perspectives IV: Action Theory and the Philosophy of Mind: 81-108. Atascadero: Ridgeview Publishing Company. Lycan, W (1990) What is the Subjectivity of the Mental? In: Tomberlin, J (ed) Philosophical Perspectives IV: Action Theory and the Philosophy of Mind: 109-130. Atascadero: Ridgeview Publishing Company. Lycan, W (1995) A Limited Defense of Phenomenal Information. In: Metzinger, T (ed) Conscious Experience. Tucson: University of Arizona Press. Lycan, W (1996) Consciousness and Experience. Cambridge, MA: The MIT Press. Nagel, T (1974) What Is It Like to Be a Bat? Philosophical Review 83: 435-50. Nemirow, L (1990) Physicalism and the Cognitive Role of Acquaintance. In: Lycan, W (ed) Mind and Cognition. Oxford: Basil Blackwell. Pereboom, D (1994) Bats, Brain Scientists, and the Limitations of Introspection. Philosophy and Phenomenological Research 54: 315-29. Russell, B (1910-11) Knowledge by Acquaintance and Knowledge by Description. Aristotelian Society Proceedings 11: 108-28. Tye, M (1986) The Subjectivity of Experience. Mind 95: 1-17. Yablo, S (1999) Textbook Kripkeanism and the Open Texture of Concepts. Pacific Philosophical Quarterly 00: 000-000. 11