1 The Project

Can we know the world? The Cartesian intuition is that this is problematical, to put it mildly. We might be deceived as to every aspect of the external world, even as to its very existence. But luckily we do know our own minds and we know we ourselves exist. At least that much is certain. Isn’t that what Descartes’ cogito is all about?

In his Die Unerkennbarkeit des Geistes. Phänomenale Erfahrung und menschliche Erkenntnis, Steen Olaf Welding, philosophy professor at the Technical University of Braunschweig, tries to show that the Cartesian tradition has it all backwards: we do not and cannot know the mind, neither our own nor that of others, at all. Knowledge of the external world, however, is no problem. Of course we cannot be absolutely certain of anything, but justified knowledge of reality is perfectly possible.

More in particular the author wants to argue that naturalistic epistemologies are doomed to fail because they neglect fundamental epistemological problems. Moreover, the failure to acknowledge these problems undermines the whole project of a naturalistic theory of mind. The unknowability of the mind must be assumed, not only in all cognitive science, but in the very phenomenal experience of one’s own existence and in our (human) knowledge of reality.

The main tool Welding uses to upset the Cartesian applecart and to topple the project of naturalistic epistemology is conceptual analysis. Defining knowledge as something that can be justified, and distinguishing between mental states and mental facts, he tries to clarify the nature of both knowledge and phenomenal experience. The book stands clearly in the tradition of Wittgensteinian analytical philosophy.

Reading the book is hard work: the author does not make any concessions to the reader, and the book certainly does not wear its structure on its sleeve. It is not easy to reconstruct the argument from the rambling meditations and digressions on the densely printed pages.

2 Knowledge, Doubt and Certainty

Starting with Descartes’ methodical doubt, Welding first analyses the nature of doubt itself. “The distinction between the assertion of a doubt and of a possible doubt is such that on the one hand I have a reason for the assumption that p could be false, and that on the other hand I could have a reason for the assumption that p could be false”.Footnote 1 When no such ground could exist, doubt is impossible. Doubt is not the same as uncertainty: whereas doubt implies the ground to assume that something could be false, uncertainty simply assumes that something might be false, without any concrete grounds for this assumption. Descartes’ method of systematic doubt ends in a logical contradiction, when he tries to doubt something that is in fact indubitable: his own existence. Thus Descartes’ dualism of substances is based on a dualism of knowledge: on the one hand there is the indubitable knowledge of one’s own mind, on the other hand the dubitable knowledge of the material world.

But then Welding rejects this dualism of knowledge. In his analysis, the assertion of knowledge implies the capacity to justify one’s assertion, to give grounds for it. Though of course these grounds need justification in their turn, this regress does not mean that a knowledge claim can never be sufficiently justified. A knowledge claim is sufficiently justified when we have no concrete grounds to doubt its justification. The mere fact that something could be doubted does not mean in itself that it is doubtful.

Now what about an assertion that is indubitable—no grounds for doubt could be possible? In that case it does not make sense to ask for a justification. But if it does not make sense to ask for a justification, there can be no question of knowledge. One may be certain of some bit of knowledge, but that in itself does not preclude the possibility of justification. However, when one is immediately certain of something, the very immediacy means that there are no grounds for it. And when there are no grounds, there can be no justification so there is no knowledge, precisely because knowledge is defined as essentially involving justification. Therefore it is impossible to speak of knowledge of one’s own mental states: one is immediately certain of one’s mental states, one simply has them. There is no possibility of justification, and therefore no knowledge. One can direct one’s attention to one’s own mental states, but one cannot find out whether one has them or not. True, one may be uncertain whether one is, for instance, really in pain. But this is a case of immediate uncertainty; no further investigation is possible.

The consequence of this conception of the impossibility to know one’s own mental states is that the Nagelian reflection on what it is like to have mental states (Nagel 1974) cannot be about knowledge. Therefore also Jackson’s knowledge-argument (Jackson 1982, 1986) is off key. There simply is no knowledge involved here. When someone says “I know what it is like to see colours” she simply means that she has had phenomenal colour experiences. Just as it is impossible to know that one is in a certain mental state, it is impossible to know what it is like to be in such a state.

Mental states are immediately certain, yet nothing can with certainty be inferred from them. Although knowledge of conscious or mental states is impossible, one can have knowledge about mental facts. Mental facts are just that: facts. One can find out about them, one can be mistaken in them etc. My fear is a mental state; I have it, but I have no knowledge of it. Perhaps my fear is a case of ailurophobia: it may be that it is cats that I am afraid of. This ailurophobia is a mental fact: I can find out about it, experiment about it and I can also be mistaken about it: it may turn out that it was not cats that I am afraid of, but closed spaces. Yet the fear itself is a mental state, and no knowledge of it is possible, nor can any mental fact be inferred from it with certainty.

In a perceptual context, the distinction between mental states and mental facts boils down to the distinction between sense impressions and sense perceptions. A perception is a fact; an impression is a mental state. The distinction between sense impressions and perceptions is independent from the distinction between appearance and reality. Appearances are not the same as sense impressions; they are certain visual properties that the object simply may or may not have. We can find out which visual properties the object really has. But we cannot find out anything about our sense impressions.

We can cognitively elucidate our mental states and find out with which mental facts they correspond. This is, however, not a causal explanation of our mental state, but a psychological explanation. We can also give psychological explanations of the mental states of others; in that case we simply assume from the beginning that they do have mental states. We cannot doubt this, as we have no concrete grounds for doubt. But neither can we justify this conviction that others have mental states. We are inescapably committed to the existence of other minds. We are certain of this, though we have no right to be certain. As we have neither grounds to doubt that others have mental states, nor grounds to justify the conviction that they do have them, we cannot have knowledge of the mental states of others, just as we have no knowledge of our own.

Our own mental states are immediately certain, but not cognitively certain. A perception, on the other hand, has a cognitive indication. That is to say, it is based on grounds, however certain or uncertain, and so we have some cognitive attitude towards it. In the elucidation of mental states by mental facts we express a cognitive indication. We can have different cognitive attitudes to the ascriptions of mental facts to ourselves and others.

3 Cognitive Science and Naturalistic Epistemology

Having reached this point in his analysis, Welding goes on to criticise cognitive science. According to him we cannot ascribe mental facts to computers, as they do not have any cognitive attitude towards their own mental facts. Turing’s imitation game is better named a simulation game. The computer does not have any intention to imitate. It would be absurd to try and find out whether the computer knows or doubts or hopes that something is green. It has no cognitive attitudes; it can only simulate them. We can know or doubt or hope something is green on the basis of a computer outcome; we are the ones who interpret the computer’s outcome.

But neither can we ascribe mental states to the brain. Physical events in the brain are not mental states; we cannot explain what mental states are by describing brain events. There may be psychophysical connections, but we cannot explain them scientifically. Physical events are observable, but mental states are not. Nor can they be described as any event. They cannot be identical to physical events.

And just as we cannot claim that a mental state is identical to a physical event in the brain, we cannot claim that a mental state is the result of some physical event. Again: we cannot observe this mental state and we cannot have knowledge of it. It is dubious to try and find nomological relations between the observable and the unobservable. We may causally explain why a mental state arises, but we cannot causally explain the mental state itself.

Neuroscientists do not differentiate between mental states and their coming into being: they only explain the coming into being, but claim to have explained the mental states themselves. They have to learn to accept the fact that mental states are unobservable and inexplicable. They can only explain how, when or why mental states come into being, but they cannot explain what it is that has come into being. We can observe empirical objects only when we presuppose our own subject, and that is only possible when the subject is not an element in the class of empirical objects.

Naturalistic epistemology (in the tradition of Quine 1969) seeks to base observation statements by a causal theory of sensory receptors. The same sensory stimulus leads in different people to the same observation statement, according to this program. The stimulus is sufficient for the truth of the statement. In the end we can substitute ascriptions of brain events for ascriptions of mental states. But the interpretation of a brain event as a mental state cannot be verified. Moreover, it is the person that has mental states, not the brain. We can only mentalistically interpret brain events and states, when we have the assent of the person.

Things get even more problematical when the neurobiological thesis is considered that the objects of perception are constructs of the brain. For when we then look at our own brain, insurmountable difficulties arise. It is impossible that we see our own perception, that we catch our own perception red handed, so to speak, in an autocerebroscope. For that which we perceive must exist independently and temporally prior to the act of perceiving it. We are forever too late. What we see is a brain event, an empirical object, but our own act of perceiving that object is unobservable. Our perceptual state, our sense impression, is already presupposed in our perception of the brain.

In the end our own mental states are presupposed in all our knowledge. There is no duality of knowing: immediate knowledge of our own mental states and mediate knowledge of the world. We do not know our mental states; but we are immediately certain of them. They form the presupposition for all that we do know. This difference between immediate certainty and knowledge is more fundamental than the presumed difference between the mental and the physical. When the mental is unknowable, there is no mind–body problem left. For the realist this changes little: all we can know is the material world.

But without perceptions we could not arrive at empirical knowledge. The realists have it backwards: they try to explain phenomenal experience by empirical knowledge. Our knowledge of reality is dependent on our perceptual capacities. And our observations themselves, with their concomitant sensory impressions and cognitive attitudes are consciousness-dependent. We have no glassy brain: we cannot observe our own observations as objective happenings in our brains. We have no knowledge of our mental states. But we do have knowledge of reality.

Our knowledge, however, is limited: it is dependent on our perceptual capacities. Both materialism and idealism are therefore out of the question, and so is the assumption of the existence of mind and matter. It makes no sense to ask whether anything of a mental nature exists, because we cannot describe the mental by its properties. Though our phenomenal experience is presupposed in all our knowledge, we have no knowledge of this experience itself. The mind is unknowable.

4 Evaluation

Welding’s work is very analytical: it is like an axiomatic system. From the axiom that there is only one kind of knowledge and that it implies the capacity to give grounds, the whole structure follows. Welding claims that the assumption of the unknowability of the mind forms the basis of all our knowledge. But this assumption itself is based on an a priori definition of what knowledge is. So it is this definition that bears the burden of the whole argument.

In itself the argumentation is correct and at times ingenious, though the somewhat rambling style makes rather high demands on the reader. Given the initial definition of knowledge, the claims about the unknowability of the mind and the impossibility of any naturalistic study of it follow inexorably. But one wonders what grounds this definition itself. To be sure, it definitely has its merits. It bears the clear stamp of Wittgensteinian philosophy and it certainly solves problems. Or rather, it dissolves them. There is no duality of knowledge, there is no mind-body problem, there is not even any metaphysical problem of dualism or monism. It almost seems as if there is no metaphysics at all, only epistemology. Such Wittgensteinian therapeutics liberates us from a whole plethora of philosophical problems.

But at the same time this therapeutic philosophy has a very reactionary impact. The whole enterprise of naturalistic epistemology is deemed to be totally misguided, and so is most of cognitive science. The definition of what knowledge is fixes what we can know. It is simply no use to study the mind, as it is ruled out of the court of the knowable by definition. We know in advance that we cannot know it.

But should we thus cower for the power of concepts? And do we want to be liberated from philosophical problems in this way? The philosopher Thomas Nagel puts it like this:

“The history of [philosophy] is a continual discovery of problems that baffle existing concepts ... At every point it faces us with the question of how far beyond the relative safety of our present language we can afford to go without risking complete loss of touch with reality. ... Historicist interpretation doesn’t make philosophical problems go away, any more than the earlier diagnoses of the logical positivists or the linguistic analysts did. To the extent that such no-nonsense theories have an effect, they merely threaten to impoverish the intellectual landscape for a while by inhibiting the serious expression of certain questions. In the name of liberation, these movements have offered us intellectual repression” (Nagel 1986, 11).

Likewise the naturalistically inclined scientist will not be impressed. She thinks it always unwise to try and stem the tide of science with too stern and static methodological prohibitions. Welding says that “On methodological grounds it must appear dubious to link observable statements with unobservable statements with a nomological hypothesis”.Footnote 2 Apart from the fact that this is unclear—surely it is not the statements themselves that are observable or unobservable—this seems unnecessarily restricting. Of course there are all kinds of difficulties about the ontological status of the unobservable. But is science not precisely in the business of trying to find hypotheses linking the observable with the unobservable? What else is physics about? If there is anything we can learn from the history and philosophy of science, it is that whenever a philosopher claims that some kind of investigation is impossible, there is a scientist who is already doing just that.

Wittgenstein, for instance, claimed to know we could never get to the moon:

“... there might be people who believe that [it] is possible [to get to the moon] ... they are wrong and we know it” (Wittgenstein 1969, 286).

The most poignant example of such a fruitless resistance to the advance of science must be Mach’s phenomenalistic opposition to atomism (Mach 1897). We have abandoned a Machian attitude to natural science—why should we keep it on towards a science of the mental? According to Welding all knowledge of phenomenal experience is impossible, but there already exists scientific data on experience. Take, for example, Koenderink’s investigations of the experience of pictorial depth (Koenderink et al. 1995, 1996). The investigations clearly deal with sense impressions, not with the perception of something real, according to Welding’s own definitions. There is no real depth in a picture; the picture is two-dimensional. Neither is it the judgement of depth that is measured. The research investigates the individual differences in the subjective experience of depth, by ingeniously making the subjects reproduce the amount of depth they experience. It turns out that there are startlingly large differences in the amount of depth experienced by different subjects in the very same picture. One would say that this was new knowledge about the mental states of people, about their phenomenal experience. It would be interesting to know what Welding would make of these investigations.

But we do not know this, as Welding makes no mention of this kind of investigation. In fact the whole work is also rather conservative in another sense. The author condemns naturalistic epistemology and cognitive science, but he addresses himself almost exclusively to analytic philosophers. Much of the literature he discusses stems from the 1970s and 1980s. The arguments of naturalistic philosophers and cognitive scientists themselves are not dealt with in any detail. Quine, surely the main target of the whole argument, gets a meagre six pages, and no major cognitive scientist is to be found in the index (or the book) at all.

It is undeniable that the whole domain of the sciences of the mental is rife with conceptual confusions, and disciplined conceptual analyses of the problems involved are badly needed. But concepts are not immutably fixed for all eternity. When they forbid us to formulate certain intuitively felt problems, or to pursue certain inquiries, their liberating force becomes oppressive. Both philosophers and scientists of mind want to get on. Often they are too impatient with conceptual analyses. But though they certainly are in need of the discipline of conceptual clarification, they will not and should not be stymied by an analysis that tries to rule their whole enterprise out of court.