Dennett, Daniel C. (1991) Lovely and Suspect Qualities. In: E. Villanueva (ed.) Consciousness. Atascadero, CA: Ridgeview.

Lovely and Suspect Qualities

Commentary on David Rosenthal, "The Independence of Consciousness and Sensory Quality" in E. Villanueva, ed., Consciousness, (SOFIA Conference, Buenos Aires), Atascadero, CA: Ridgeview 1991.

Daniel C. Dennett

A family of compelling intuitions work to keep "the problem of consciousness" systematically insoluble, and David Rosenthal, in a series of papers including the one under discussion, has been resolutely driving these intuitions apart, exposing them individually to the light, and proposing alternatives. In this instance the intuition that has seemed sacrosanct, but falls to his analysis, is the intuition that "sensory quality" and consciousness are necessarily united: that, for instance, there could not be unconscious pains, or unconscious subjective shades of blue, or unconscious aromas of freshly roasted coffee beans. The particular airborne polymers that are the vehicles of freshly roasted coffee beans could exist, of course, in the absence of any observer, and hence of any consciousness, but the sensory quality of that aroma requires--according to well-entrenched intuition--not only an observer but a conscious observer. Such properties have no esse except as percipi.

Rosenthal argues that this utterly familiar idea is just wrong; the connection between what he calls sensory quality and consciousness is only contingent. Moreover--and this is the most important contribution of his paper--it is only by denying this traditional link that progress can be made. I have no substantive quarrels with Rosenthal's arguments, so instead of raising obstacles, I will try to push his claims a little further, and support them with further considerations.

Hume pointed to what he thought was an important difference between our "internal impressions" and our sensations. The latter require (or at least invite) us to posit continuously existing bodies and properties in order to preserve the coherence and constancy of our discontinuous impressions of sensation. His example was the fire in his study fireplace, which gradually burns down, during the intervals between his different sensations of it (Treatise I.iv.2). He thought it was otherwise with "internal impressions": "on no occasion is it necessary to suppose that they have existed and operated, when they were not perceived, in order to preserve the same dependence and connexion, of which we have had experience." But Hume's claim is by no means obvious, however well supported by tradition. Unfelt pains make a convenient entering wedge. As Rosenthal points out, it is natural, contrary to Hume's supposition, to speak of having had a single, persistent pain or ache during the longish stretch of time when one is intermittently distracted by--conscious of--a headache. If all the other psychological effects of pain are present--the lowered effectiveness, the inability to concentrate, the irritability--as well as the physiological effects that are responsible for them, such as the characteristic changes in neuromodulator balances, there is good reason to treat one's pain as persisting unobserved, just like the fire in the hearth. And if there can be unfelt pains, there can be unconscious sensations of red.

Suppose, to make the idea vivid, that we train human subjects to respond to color-coded signals in a video game, and suppose they have accustomed themselves to expect some sort of disagreeable event in the aftermath of a flashing red spot; we then arrange to present them with a flashing red spot under conditions in which their attention is concentrated on other matters; we observe that they do not, and cannot, report the occurrence of the flashing red spot, but their galvanic skin response (as a measure of their anxiety, their expectation of a disagreeable sequel) indicates that they have distinguished the redness of the spot. Experiments exhibiting similar effects are legion, and while I cannot think of any offhand that have exactly the effect described, I have no doubt that such effects could be produced. Suppose, in any case, that we did have evidence that the effects of "seeing red" (and seeing it as red, of course) could occur in the absence of consciousness of red (as revealed in the normal experimental way--by the subjects' reports). Would this not give us good reason to admit that sensory qualities can exist independently of consciousness?

Why not? What arguments have been offered, by Hume or others, to oppose this idea? If it seems repugnant to reason, just why is this? Perhaps the most persuasive ground is the idea--so persuasive, apparently, that it need not be made into an explicit argument!--that once science has established that secondary properties are just powers in objects to produce certain ideas in observers, these ideas-in-observers must "intrinsically" involve consciousness. Rosenthal says: "It is often pointed out that we cannot repeat [the Lockean] move; there is no place to relocate the qualitative character that the distinctive properties of sensory states seem to exhibit." [ms p.16] Rosenthal questions this claim, but I will go further. We can repeat the Lockean move, and in fact we must.

The common wisdom is that modern science has removed the color from the physical world, replacing it with colorless electromagnetic radiation of various wavelengths, bouncing off surfaces that variably reflect and absorb that radiation. It may look as if the color is out there, but it isn't. It's in here. It seems to follow that what is "in here" is both necessarily conscious (otherwise it isn't all the way "in") and necessarily "qualitative" (otherwise color would be utterly missing in the world). This reasoning is confused. What science has actually shown us is just that the light-reflecting properties of objects cause creatures to go into various discriminative states, underlying a host of innate dispositions and learned habits of varying complexity. And what are their properties? Here we can indeed play Locke's card a second time: these discriminative states of observers' brains have various "primary" properties (their mechanistic properties due to their connections, the excitation states of their elements, etc.), and in virtue of these primary properties, they have various secondary, merely dispositional properties. In human creatures with language, for instance, these discriminative states often eventually dispose the creatures to express verbal judgments alluding to the "color" of various things. The semantics of these statements makes it clear what colors supposedly are: reflective properties of the surfaces of objects, or of transparent volumes (the pink ice cube, the shaft of limelight). And that is just what colors are in fact--though saying just which reflective properties they are is tricky.

Don't our internal discriminative states also have some special "intrinsic" properties, the subjective, private, ineffable, properties that constitute the way things look to us (sound to us, smell to us, etc.)? No. The dispositional properties of those discriminative states already suffice to explain all the effects: the effects on both peripheral behavior (saying "Red!", stepping the brake, etc.) and "internal" behavior (judging "Red!", seeing something as red, reacting with uneasiness or displeasure if, say, red things upset one). Any additional "qualitative" properties or qualia would thus have no positive role to play in any explanations, nor are they somehow vouchsafed to us "directly" in intuition. Qualitative properties that are intrinsically conscious are a myth, an artifact of misguided theorizing, not anything given pretheoretically.

We do have a need, as Rosenthal shows, for properties of discriminative states that are in one sense independent of consciousness, and that can be for that very reason informatively cited in explanations of particular contents of our consciousness. These properties are partially, but not entirely, independent of consciousness. We may call such properties lovely properties as contrasted with suspect properties. Someone could be lovely who had never yet, as it happened, been observed by any observer of the sort who would find her lovely, but she could not--as a matter of logic--be a suspect until someone actually suspected her of something. Particular instances of lovely qualities (such as the quality of loveliness) can be said to exist as Lockean dispositions prior to the moment (if any) where they exercise their power over an observer, producing the defining effect therein. Thus some unseen woman (self-raised on a desert island, I guess) could be genuinely lovely, having the dispositional power to affect normal observers of a certain class in a certain way, in spite of never having the opportunity to do so. But lovely qualities cannot be defined independently of the proclivities, susceptibilities, or dispositions of a class of observers. Actually, that is a bit too strong. Lovely qualities would not be defined--there would be no point in defining them, in contrast to all the other logically possible gerrymandered properties--independently of such a class of observers. So while it might be logically possible ("in retrospect" one might say) to gather color property instances together by something like brute force enumeration, the reasons for singling out such properties (for instance, in order to explain certain causal regularities in a set of curiously complicated objects) depend on the existence of the class of observers.

Are sea elephants lovely? Not to us. It is hard to imagine an uglier creature. What makes a sea elephant lovely to another sea elephant is not what makes a woman lovely to another human being, and to call some as-yet-unobserved woman lovely who, as it happens, would mightily appeal to sea elephants would be to abuse both her and the term. It is only by reference to human tastes, which are contingent and indeed idiosyncratic features of the world, that the property of loveliness (-to-a-human-being) can be identified.

On the other hand, suspect qualities (such as the property of being a suspect) are understood in such a way as to presuppose that any instance of the property has already had its defining effect on at least one observer. You may be eminently worthy of suspicion--you may even be obviously guilty--but you can't be a suspect until someone actually suspects you. The tradition that Rosenthal is denying would have it that "sensory qualities" are suspect properties--their esse is in every instance percipi. Just as an unsuspected suspect is no suspect at all, so an unfelt pain is supposedly no pain at all. But, for the reasons Rosenthal adduces, this is exactly as unreasonable as the claim that an unseen object cannot be colored. He claims, in effect, that sensory qualities should rather be considered lovely properties--like Lockean secondary qualities generally. Our intuition that the as-yet-unobserved emerald in the middle of the clump of ore is already green does not have to be denied, even though its being green is not a property it can be said to have "intrinsically." This is easier to accept for some secondary qualities than for others. That the sulphurous fumes spewed forth by primordial volcanos were yellow seems somehow more objective than that they stank, but so long as what we mean by "yellow" is what we mean by "yellow," the claims are parallel. For suppose some primordial earthquake cast up a cliff face exposing the stripes of hundreds of chemically different layers to the atmosphere. Were those stripes visible? We must ask to whom. Perhaps some of them would be visible to us and others not. Perhaps some of the invisible stripes would be visible to pigeons (with their tetrachromat color vision), or to creatures who saw in the infra-red or ultraviolet part of the electromagnetic spectrum. For the same reason one cannot meaningfully ask whether the difference between emeralds and rubies is a visible difference without specifying the vision system in question.

The same moral should be drawn about the sensory qualities Rosenthal attributes to mental (or cerebral) states. Like Lockean secondary qualities in general, they are equivalence classes of complexes of primary qualities of those states, and thus can exist independently of any observer, but since the equivalence classes of different complexes that compose the property are gathered by their characteristic effect on normal observers, it makes no sense to single them out as properties in the absence of the class of observers. There wouldn't be colors at all if there weren't observers with color vision, and there wouldn't be pains at all if there weren't subjects capable of conscious experience of pains, but that does not make either colors or pains into suspect properties.

Rosenthal (in a personal communication) asks whether this is not too strong. Why should the existence of pains require subjects capable of conscious experience of pains, as opposed simply to subjects capable of having nonconscious pains? Fair question, and his implied point is a good one--except for what amounts, in the end, to a lexical quandary, which can be brought out by considering the parallel with color. There is nothing except the specific effects on normal human beings that demarcates the boundaries of the "visible spectrum". Infra-red and ultra-violet radiation does not count as subserving color vision (at least according to a sort of purist definitional taste) even in creatures who respond to it in the ways we respond to the (humanly) visible spectrum. "Yes, it's like color vision, but it isn't color vision," someone might insist. "Color vision is vision whose proper objects are (only) red through violet." Now imagine that we confront a set of primary property complexes as candidates for the secondary property of pain, and suppose it is a somewhat enlarged set (it includes infra-pain and ultra-pain, in effect), including out-lying cases of which we human beings would never be conscious (but which have the sorts of effects on variant human beings that paradigmatic pains have on us, etc). Would those be pains? There would certainly be a property which was the property picked out by that set, but would it be pain? (Not a terribly interesting question.)

I claim, then, that sensory qualities are nothing other than the dispositional properties of cerebral states to produce certain further effects in the very observers whose states they are.Endnote 1 It is no objection to declare that it just seems obvious that our mental states really do have intrinsic properties over and above their dispositional properties. (If this were a good argument, it would be a good argument against the original distinction, by Locke, Boyle and others, between primary and secondary qualities, for it certainly "seems obvious" that physical objects have their color properties "intrinsically"--just look at 'em!) It does indeed appear to us as if we somehow enjoy, in our minds, some sort of direct and intimate access to "intrinsic" properties of our conscious states, but as Rosenthal observes, "We need not preserve the 'element of truth' in erroneous commonsense intuitions when we become convinced that these intuitions reflect how things appear, rather than how they really are." (ms. p.16)


1. The claims advanced briefly here are drawn from a longer discussion and defense of them in my forthcoming book, Consciousness Explained, which has been much influenced by Rosenthal's articles, and from many discussions with him.