This goal is a bold one indeed, and Stroud's subtle and interesting arguments in its service have important implications concerning not only the metaphysics of color, but broader topics in metaphysics, epistemology, perception, and the philosophy of mind as well. Although I cannot do justice to the scope and detail of Stroud's argumentation here, I want to touch on some of his general themes and evaluate his argumentative strategies. In particular, I'll focus on Stroud's arguments against color irrealisms that turn on particular conceptions of reality (§1), his contention that, barring a sensationalist understanding of color perception, there is no way in which we could become convinced of the truth of color irrealism (§2), and his rejection of sensationalism about color perception and resulting forms of color subjectivism and color irrealism (§3).
The argument Stroud is considering, then, can be represented as follows:
Stroud responds that this argument turns on an equivocation on `anthropocentric'. For while he is prepared to admit that colors are anthropocentric in the sense that our acquaintance with them is mediated by the perceptual and conceptual faculties we happen to have, he denies that this sense of anthropocentricity legitimates (P2). After all, he insists, our acquaintance with numbers and shapes (if not all properties) is mediated by our conceptual and perceptual makeups; but since we're not tempted to infer from this that numbers and shapes are unreal, the corresponding inference about colors is equally unwarranted.
While I do not wish to disagree with Stroud's ultimate assessment of the argument, I think his way of responding to it is not as convincing as it might be, insofar as it invites an answer that should be addressed. The counterexamples he offers (35-36), which run together shape, number, smell, and color, seem to be aimed at denying that colors are anthropocentric in a way that these other sorts of properties are not. But there is a sense in which colors have been claimed to be anthropocentric in a way that shape and number, for example, are not, and it is worth wondering whether this disanalogy provides a foothold from which the color irrealist could rescue something like the argument for (C). Namely, many have found plausible the claim that color properties, unlike shape or number, are anthropocentric in the sense that they can only be characterized in terms of the reactions they produce in perceivers. Thus, McGinn writes that ``to grasp the concept of red it is necessary to know what it is for something to look red, since this latter constitutes the satisfaction condition for an object's being red. In contrast, to grasp what it is for something to be square it is not constitutively necessary to know how square things look or feel, since what it is to be square does not involve any such relation to experience'' ([McGinn, 1983], 8, emphasis in original). McGinn's suggestion, then, is not only that our access to colors is mediated by our perceptual and conceptual capacities, but that what constitutes a color is spelled out in terms of a particular way of affecting creatures with our perceptual endowment. Shape properties, it is said, are not anthropocentric in this latter sense because, although square things do (sometimes) look square, there could be squares even if nothing did or could look square. The root of the difference between color and shape, on this account, turns on whether each is characterizable in terms independent of how its instances could or would strike beings with our perceptual and conceptual makeup: the proposal has it that shapes are so characterizable (viz., in geometric terms), while colors are not.
If this is right, then there would seem to be a sense of `anthropocentric' that applies to colors but not shapes (and therefore that would evade Stroud's objection). Could this sense of `anthropocentricity' be used to resuscitate the irrealist argument from (P1) and (P2)?
It could not. The subjectivist proposal under consideration has it that colors are what they are in virtue of their relations to perceiving subjects. But, far from entailing the irrealist conclusion (C), this proposal is intended to tell us how something that does exist and is perfectly real (viz., color) is constituted.2 Therefore, even if a subjectivist account like the one I've mentioned could provide a notion of `anthropocentrism' (and a corresponding version of (P1)) that would apply to the colors but not shapes or numbers, the invalidity of the inference from color subjectivism to color irrealism prevents us from endorsing (P2), and therefore undermines the irrealist argument under review.
The problem with this argument for the unreality of color, as Stroud hammers home, is that it presupposes a robust notion of the physical the likes of which we simply don't have.3 That is, to arrive at the conclusion that colors (or anything else) don't count as physical, we need a substantive criterion for the physical that colors fail to satisfy. We cannot spell out this criterion in terms of fitting into the vocabulary of current physical science, since this would inappropriately close off the possibility of further discoveries of elements of (physical) reality. But, on the other hand, if we maintain a conception of the physical sufficiently flexible to allow for the discovery of new physical kinds, it is no longer clear what justifies our conclusion that colors lie outside of the physical.
The problem with unmasking explanations of color (or psychology, or economics), according to Stroud, is that their success depends on the achievement of a full explanatory reduction according to which all that is so (including claims about color, psychologies, and economies) can be explained in terms of an austere vocabulary that doesn't advert to color, psychological, or economic kinds. But Stroud doubts that such an explanatory reduction is in the cards. He cites [Davidson, 1980] in support of his contention that no amount of ``physical explanation'' and no amount of evidence of a correlation between physical and psychological (etc.) events would answer the persistent question - would dissolve what [Levine, 1984] called an ``explanatory gap'' (cf. the famous open-question argument of [Moore, 1903]) - of why the sequence of physical events eventuates in a psychological event; and since this question would persist, we would be wrong to claim that the occurrence of the psychological event had been explained by the unmasker. Nor, according to Stroud, could our questions be allayed by connecting high-level psychological (or color, ethical, economic, etc.) explanations to low-level physical explanations through bridge laws; for this move would merely label the problem rather than solve it - we would then be faced with the equally persistent question why the bridge laws in question hold.
Another worry about the prospects for reduction - one that Stroud does not bring up explicitly - is that the events in question (exemplifications of color properties) are plausibly realizable by tokens of multiple physical types (see [Nassau, 1980]). But if this is so, then, as urged in [Fodor, 1974], confining ourselves to stating generalizations in terms of the more austere physical vocabulary compromises our explanatory needs by inviting us to miss generalizations subsuming all the physically distinct realizers.4
For these reasons, Stroud concludes that none of the conceptions of reality he surveys could justify the conclusion that colors are unreal.
Assume, then, that color terms are univocal in all of their occurrences. On this assumption, Stroud finds the prospects for unmasking our beliefs about color dim. His argument for this conclusion, which is in some ways the most original and ambitious part of his case against color irrealism, is rooted in a thesis he attributes to Wittgenstein and Davidson to the effect that ``we interpreters and ascribers of beliefs and other psychological states must be engaged in the world, in the sense of taking certain nonpsychological things to be true of it, if we are ever going to attribute beliefs or perceptions to anyone'' (151).
Stroud argues for his conclusion by considering the unmasker who begins by attributing to another subject, S, the belief that a lemon is yellow. Of course, the unmasker holds that this belief is false, since she thinks that nothing is colored. The unmasker, then, attributes to S what she takes to be a false belief - the belief that a particular lemon is yellow. But, Stroud claims, the unmasker cannot correctly be said to attribute this belief (even to attribute it as a false belief) unless she understands its content. Thus, the unmasker must understand the content of the belief that a particular lemon is yellow, even if she takes that content to be false. However, Stroud claims, understanding the content of this belief requires ``knowing under what conditions it would be correct to judge or assert it to be true'' (154). What follows from the chain of reasoning so far, then, is that even an unmasker, who believes that nothing is colored, must know under what conditions it would be correct to judge or assert that something is colored in order to make the attributions of beliefs about color she in fact makes (attributions of false beliefs about color, according to her).5
Crucially, however, Stroud argues that having this knowledge presupposes a capacity to judge or assert that things are colored, even for someone who never exercises that capacity (because, for example, she is an unmasker who believes that nothing is colored). This is so because, according to Stroud, an unmasker who lacked the relevant capacity would lack the resources to understand just which properties she thinks others misattribute to objects when they say that objects are colored. She could not understand a color property such as yellow in terms of the colors of exemplar objects (e.g., as the color exemplified by ripe lemons), since she is attempting to explain away (and not presuppose) the claim that lemons and other objects exemplify colors. Other proposals seem equally unhelpful to the would-be unmasker:
Nor could we identify perceptions of yellow as perceptions of that property that people believe to belong to objects that are yellow. We would admit no objects that are yellow. It would get us no further to say that perceptions of yellow are perceptions of that property that people believe to belong to objects that they believe to be yellow. That specification is not unique. There are many properties that people believe to belong to objects they believe to be yellow. Even to say that perceptions of yellow are perceptions of that property that people believe to belong to an object in believing that it is yellow would be no help. That is true, but [on the current assumption that `yellow' is univocal in all its uses] it serves to identify perceptions of yellow only if the content of the belief that an objects is yellow can be specified independently (160).Finally, she could not understand perceptions of yellow by the application of a demonstrative - as perceptions of this property, said while demonstrating the property we hope to understand - because, Stroud thinks, successful demonstrative reference presupposes a prior mental grasp of the referent that is not itself a demonstrative grasp (on pain of regress), so the appeal to demonstration just pushes the problem back a step (here Stroud refers to the passages on ostensive definition in [Wittgenstein, 1953]).6
From the failure of these alternatives, Stroud concludes that a would-be unmasker would lack a way of understanding the properties she takes others to misattribute to objects, and therefore cannot be said to ascribe to them mistaken beliefs about those properties. Of course, the unmasker must begin by ascribing such mistaken beliefs to others - that is, after all, the target of her unmasking project. Consequently, no putative unmasker could ever convince herself of the conclusion that there are no colors (or that nothing is colored).7
This is a fascinating argument against color irrealism. By arguing that the attribution of (even false) beliefs about color demands an understanding of color properties, Stroud has saddled the unmasker with the need to analyze our understanding of colors in terms of some other ingredients. However, his assumption that sensational and predicational uses of color terms are univocal in meaning (the assumption that is in place for this horn of the dilemma) makes color sensations unsuitable to play the role of those other ingredients: on this assumption, an appeal to color sensation is an appeal to the very property whose analysis we initially sought, and therefore will not advance the unmasker's goal. On the other hand, many have thought that the only possible non-color ingredients suitable for explaining our understanding of colors are color sensations. To see this, consider a different property beliefs about which are easily unmasked: the property unicorn . Unmasking beliefs about unicorns is easy because our understanding of unicorn is plausibly composed in some way from our independently secured understanding of horse and horn (cf. Hume on the idea of a golden mountain); since we can construct an understanding of the property unicorn by appeal to these conceptual ingredients, we can understand the content of the belief (according to us, an erroneous belief) we ascribe to people who utter `unicorns are docile', and therefore have no difficulty unmasking beliefs about unicorns. But an analogous proposal won't save the unmasker of beliefs about colors unless our understanding of color properties is structured from non-color ingredients in the way that our understanding of unicorn plausibly is structured from non-unicorn ingredients, and this is something that Stroud doubts: ``It seems that our understanding and recognition of the colours of things cannot be built up out of noncolour building blocks in the way we can perhaps build up a thought of a unicorn from thoughts of a horselike body and a well-placed horn'' (147).8
Stroud's argument, then, demands that the unmasker construct our understanding of color from our understanding of other properties, and then makes unavailable precisely the materials with which that construction could be most plausibly carried out. It is little surprise that the unmasker who accepts the constructive task Stroud has set finds herself unable to complete it. It seems to me that there are two main avenues of reply left to the unmasker at this point. First, she might refuse to engage in Stroud's constructive task on the grounds that the understanding it aims to produce is not necessary for the unmasker's successful attribution of beliefs about color (see note 5). Alternatively, she should abandon the assumption that color terms have a univocal meaning in sensational and predicational uses.9
A proponent of the first such view moves to irrealism about color by beginning with, and then ammending, his sensationalism about color perception. To buoy this line of thought, the sensationalist points out that an analogous sensational story about pain leads us to withhold ascribing pains to extramental objects: when a pin divides my flesh, I have a pain sensation, but I do not think of pain as a property of any extramental bit of reality (e.g., the pin). Similarly, according to this kind of sensationalist, the yellow sensation produced in the observer by the lemon gives us no grounds for attributing a color property to any bit of extramental reality (e.g., the lemon). If we refrain from making such ascriptions generally, we will naturally be led to revise the initial claim that color terms have legitimate applications to both mental and extramental bits of reality. Instead we will claim that colors are not a part of reality - that, although occurrences of color terms in contexts like (1) are innocent because color sensations are real, claims like (7) and (8) are always false because colors, qua properties of extramental bits of reality, are unreal. Forms of this kind of color irrealism are defended by [Hardin, 1988] and [Maund, 1995], among others.10
The second form of sensationalism Stroud wants to reject is one that admits the reality of colors, but takes them to be subjective - constituted in terms of relations to sensations in perceiving subjects. Perhaps the most familiar theory of this sort is the dispositionalist view that colors are dispositions to produce color sensations in perceivers.11
Both kinds of sensationalist agree that color terms are ambiguous between a (``sensational'') sense in contexts like (1) to pick out sensation types and a (``predicational'') sense in contexts like (7) or (8) to predicate some (non-sensational) property to extramental objects. The two disagree, however, in their attitude toward the ostensible referents of predicational uses of these terms: sensationalist irrealists think predicational uses attempt but fail to pick out any properties of objects, while sensational dispositionalists think these uses successfully pick out dispositional properties of objects.
Stroud thinks both of these views are non-starters because he objects to the sensationalism on which they are grounded. For he thinks that the sensationalist's claim that color terms are ambiguous results in an unacceptable barrier between perception and thought.12 He objects that if, as per sensationalism, color terms are ambiguous between a sensation-reporting sense (as in (1)) and a predicational sense (as in (7)), it is hard to see how perception could be a source of belief - that is, it is hard to see how Jones's believing that something is yellow on the strength of (1) would be anything but a gross fallacy of equivocation.
How might sensationalists respond to Stroud's objection?
The sensational irrealist, then, responds to Stroud's attack on sensationalism by conceding that the inference in question is fallacious, and by offering an explanation of its appeal. This, it seems to me, should count as an answer to Stroud's challenge. Unfortunately, Stroud never considers this possible answer on behalf of the irrealist, and therefore never says why we shouldn't accept it. So far as I can see, then, the sensational irrealist escapes Stroud's arguments unscathed.
But Stroud is dissatisfied with this dispositionalist answer for other reasons. As he notes, dispositionalism is rooted in biconditionals of the following form:
(D) x is yellow if and only if normal human perceivers standing in certain relations R to x in certain kinds of perceptual circumstances C would get perceptions [sensations] of yellow (121).Note that, since (D) unpacks yellow in terms of sensations of yellow, the viability of dispositionalism demands an independent and prior way of singling out the relevant sensations. But, given this constraint, Stroud thinks there is no way of understanding (D) that will show colors to be subjective in a way that, for example, shape, size, and other so-called primary qualities are not. For, since it is plausible that a biconditional corresponding to (D) is true about a paradigm primary quality like round, the claim must be that yellow differs from round in that the biconditional corresponding to the former (but not the latter) is not only true but necessary. But, again assuming our grasp of yellow sensations is independent of our grasp of yellow, Stroud thinks (D) is contingent rather than necessary:
... perceptions of yellow, understood in [an independent] way, are the sort of thing that could have occurred in normal human beings standing in the stated relations to blue objects - or, for that matter, in the presence of no objects at all. And the perceptions normal perceivers standardly got in the presence of yellow objects could have been perceptions of blue, understood simply as ``sensations'' of a different kind. But in conceiving of perceivers' ``sensations'' having been different in those ways, we can continue to conceive of all the ripe lemons in that world as yellow (129, emphasis and shudder-quotes in original).Thus, Stroud concludes, the dispositional view cannot secure the necessity of the biconditionals that would make colors turn out to be subjective, and therefore dispositionalism cannot provide a successful route to subjectivism about color.
There is, however, an obvious answer to Stroud's objection. The dispositionalist will claim that non-actual worlds in which sensations of yellow are produced in human beings appropriately related to ripe blueberries (or any things that are blue in the actual world) cannot be correctly described as worlds in which sensations of yellow are produced by things that are blue. On the contrary, the dispositionalist believes that (D) is necessary, so she will insist that blueberries are yellow (and inappropriately named) in the worlds under consideration (since, for her, something is yellow just in case it gives rise to sensations of yellow). Similarly, biconditionals like (D) entail that, in worlds where ripe lemons are disposed to produce sensations of blue (/no sensations), those lemons are blue (/colorless). But if this is right, then the worlds Stroud imagines are not counterexamples to the necessity of the dispositionalist's biconditionals. Of course, one could block this response by insisting that worlds in which ripe lemons produce sensations of blue are, after all, worlds where yellow things are disposed to look blue; but this move requires an alternative (non-dispositional) theory of colors according to which the colors of things are independent of what sensations they are disposed to produce in perceivers. But to say this is precisely to reject the dispositionalist analysis of colors, and therefore the move under consideration cannot be used without begging the question against dispositionalism.14
Stroud anticipates this line of response, and objects that it presupposes a subjectivist account according to which the colors of things are linked (of necessity) to the reactions produced by those things in us. But, he suggests, it is question-begging to appeal to the subjectivism of color as a premise of an argument for color subjectivism:
To object that the relevant biconditionals will be seen to hold necessarily if we have adopted the subjectivist view is to no avail. To assess that view we need an independent judgment of whether the relevant biconditionals are necessarily true. We cannot simply insist on the necessity of a candidate biconditional to rule out what look like possibilities that show that it is not necessarily true (137).
Is Stroud right that the dispositionalist's answer must presuppose the subjectivist conclusion for which dispositionalists are trying to argue? I don't believe so. True, the dispositionalist enlists (D) as a way of answering Stroud's question about how her theory can explain the inference from (1) to (7); but there is no reason to think that she will appeal to (D) in the course of an argument for color subjectivism itself. If this is right, then Stroud is incorrect in regarding the dispositionalist's answer as begging questions in favor of subjectivism. Still, all of this invites us to ask whether there are any non-question-begging reasons to accept the kind of subjectivism about color entailed (/presupposed) by (D).
One important line of evidence for subjectivism that has appealed to a number of dispositionalists comes from considerations of interpersonal and intrapersonal variations in color perception.15 However, and crucially, the argument does not end by noting that there are such variations. It goes on to make the further point that there seems not to be any motivated way of choosing which of the variants is correct (at the expense of the other variants), as there must be if colors were constituted independently of their effects on perceivers. To see this point, contrast the property square, which presumably is not a subjective property. Of course, there is perceptual variation with respect to squares as well: x can look square to observer S1 and fail to look square to observer S2 (perhaps S2 is viewing x at an angle oblique to its face). However, we have objective, observer-independent, well-motivated criteria for deciding whether x is square (viz., it is square just in case it has four right interior angles and sides of equal length), and this justifies us in choosing which among the perceptual variants is correct. We can say, in this case, that S's perception that x is a square is correct just in case x meets the (independently certified - in this case, geometrically certified) conditions that are necessary and sufficient for being a square. By this test, we know that S2 (as described above) perceives x incorrectly (with respect to the property square): she takes x to be a non-square, even though x satisfies the independent (geometric) necessary and sufficient conditions for being a square. The argument from the perceptual variation of color to the subjectivity of color claims not only that there is variation in the perception of color, but that, unlike the case of variation in the perception of squares, there is no independent standard on the strength of which we can choose among the variants.
Stroud briefly turns to considerations about perceptual variation toward the end of the book (173-176), but remains unconvinced:
What colour you see when you see an object on a particular occasion does, of course, depend on the condition you are in and the circumstances in which you find yourself.... But that is true of the perception of all properties. Whether you get a perception of something ovoid from an ovoid object or a perception of an elephant from an elephant equally depends on your current state and the perceptual conditions (174).16
But these comments miss the force of the arguments for subjectivism based on perceptual variation. The point is not merely that this variation exists, but that, unlike the cases of ovoid and elephant perception, there is no well-motivated and non-arbitrary way of choosing between the variants. That is, while ovoid objects can look round and elephants can look like hippopotami under certain circumstances, we have criteria for ovoidness and elephantness (geometric and biological) that are independent of the way things look; with these criteria in hand, we have reason to say that the way objects look in such circumstances is not a definitive guide to whether they are ovoids or elephants. In stark contrast to these cases, the lack of perception-independent criteria for the colors of things prevents us from choosing among the perceptual variants when it comes to color, and shows that the colors of things are not independent of the way those things strike perceiving subjects.
Is it true, as the subjectivist claims, that there is no motivated, independent standard that we could use to select from among the variants in color perception? After all, one might suggest, there are several scientific and industrial recipes for deciding which among a range of observers is correct in her color judgments. Unfortunately, there are several reasons for thinking that these recipes won't serve our purposes.17
First, as noted in ([Hardin, 1988], 76-82), the scientific and industrial specifications that have been articulated are typically statistical constructs over a range of distinct actual individuals, but that differ significantly from most (perhaps as many as 90% of) human visual systems ([Evans, 1948], 196-197); therefore, fixing the colors by appeal to such standards would commit one to the unpalatable conclusion that the color discriminations of most (perhaps as many as 90% of) human visual systems are erroneous. Second, these definitions are stipulatively chosen for particular purposes (mathematical convenience, industrial standardization), and reflect the particular purposes of their users; therefore, there is no reason for hoping that such standards will reveal the objective nature of color. Third, the large number of distinct standards that are adopted for different purposes itself suggests that no one of these standards will serve to fix the colors authoritatively, but rather that each is only useful in certain delimited contexts.18
These empirically-motivated considerations give us good reason for supposing that there are no principled and non-arbitrary grounds for deciding that the way x looks (with respect to color) to one observer is correct while the way x looks (with respect to color) to another observer is incorrect. If there were such grounds, then we could appeal to those grounds to determine the true colors of objects without adverting to the way things look (with respect to color) to perceiving subjects. But the lack of such perceiver-independent grounds means that there is no way, over and above the way objects affect perceiving subjects, to attribute colors to those objects. And to say this, of course, is precisely to say that colors are subjective - that the colors of objects are not independent of truths about perceiving subjects. Moreover, the empirical considerations on which this argument depends are, as far as I can tell, free of subjectivist presuppositions.
In sum, I think Stroud's argument against dispositionalism and other ``indirect'' accounts of the relation between distinct senses of color terms amounts to not much more than registering his anti-subjectivist intuitions about color. He claims that subjectivists are guilty of presupposing the subjectivism for which they argue, and therefore thinks that questioning these presuppositions (by registering his anti-subjectivist intuitions) shifts the argumentative burden onto subjectivist shoulders. However, as I've indicated, I think that subjectivists have already met this burden by providing empirically grounded, theory-independent argumentation for their position. Moreover, Stroud's brief discussion of the relevant concerns suggests that he has failed to appreciate how and why they support color subjectivism. For all these reasons, I conclude that Stroud's case against sensationalist dispositionalism is unsuccessful, and therefore that his argumentation leaves room for at least one form of subjectivism about color that has seemed attractive to many.
1Stroud attributes this argument for the unreality of color, and the conception of reality on which it is grounded, to [Williams, 1978], [Smart, 1963], and [Mackie, 1976].
2Cf. [McDowell, 1985] for related points about dispositional accounts of value, and [Hilbert, 1987] on the compatibility of realism and anthropocentrism about color.
3This point has been made recently in a number of places, including [Chomsky, 1995], [Van Frassen, 1996], and [Montero, 1999].
4[Fodor, 1997] points out that this concern applies particularly to what he calls ``open'' properties - those not necessarily equivalent to a finite disjunction of realizers, since such properties extend to (not just infinitely many cases, but) new cases never before instantiated, such as those you might construct in your basement one afternoon.
See also [Broackes, 1992], where these and other points about multiple realizability are used to argue for the explanatory autonomy of color properties.
5 Stroud's argument so far depends on controversial assumptions, for which he does not argue in the book, about the preconditions for belief attribution: one might disagree that the unmasker's attribution of beliefs about color requires her to understand the content she attributes in the sense of ``knowing under what conditions it would be correct to judge or assert it to be true.'' While there are cases where our bona fide belief attributions meet this condition (e.g., my report of some of Einstein's beliefs about baseball), there seem to be other cases where they do not (e.g., my report of some of Einstein's beliefs about spacetime). At the very least, I suggest, Stroud owes us an argument to show that the unmasker's attributions of beliefs about colors must fall into the former category rather than the latter.
6 Stroud's appeal to Wittgenstein here seems to me a vulnerable point in his argument. For the Wittgensteinian premise he relies on - that ostension could not secure a determinate reference unless the thinker possessed a prior, descriptive way of singling out the demonstratum - has been the subject of a fair amount of criticism in recent work on demonstrative reference (see, for example, [Evans, 1982], [Recanati, 1993], [Bach, 1994], and [Larson and Segal, 1995]). On the other hand, I think Stroud is right to conclude that demonstrative reference won't help the unmasker understand color properties. After all, it is hard to see how an unmasker who wanted to understand color properties in terms of an ostensive demonstration could simultaneously insist that nothing is colored: how could one ostend to a property one takes to be uninstantiated? One's own belief that the property is uninstantiated would undercut the possibility of making a demonstrative reference to it. More on this in note 12.
7As Stroud goes on to note, this argument does not establish that the unmasker's sought-after conclusion - that nothing in the world is colored - is false. Rather, it aims to show that no unmasker could rationally convince herself of that conclusion. Stroud's aim is to show that the status of the unmasking conclusion is, in this sense, akin to that of Moore's paradox: he argues that while the unmasking conclusion is, like `I believe that it is not raining, and it is raining' or `I am not speaking now,' possibly true, the pragmatic preconditions of asserting or believing it cannot be met by a rational inquirer (204-209).
8Stroud is not alone in this doubt; it is for this reason, for example, that ideas of color are traditionally included among the primitives in empiricist theorizing from Hume's Treatise to Carnap's Aufbau.
That said, some have disagreed with Stroud on this point. Thus, for example, [Smart, 1975] and [Armstrong, 1987] maintain that color properties can be understood in terms of our patterns of discriminatory sorting behavior, which (they claim) can themselves be understood independently of any reference to colors. For this reason, I would have liked to see Stroud argue against the Smart-Armstrong proposal specifically at this point in his argument.
9In this regard, it is worth noting that most contemporary irrealists about color have denied this assumption, and have instead embraced a version of sensational irrealism (see §3.1).
10Hardin and Maund, however, do not argue for their view along these lines. Something like this argument for irrealism occurs in [Boghossian and Velleman, 1989].
11Such dispositionalist theories have often been ascribed to Galileo, Boyle, Newton, Locke, and Descartes, and have been defended more recently by many writers including [McGinn, 1983], [Peacocke, 1984], [Johnston, 1992], and [Wright, 1992]. In addition, there are a number of non-dispositional theories of color that understand colors as constituted in terms of a relation to perceiving subjects, including those of [Jackson, 1998], [McLaughlin, 2001], and [Cohen, 2000]. For brevity, I'll ignore these non-dispositionalist treatments here, although much of what follows will be applicable to them as well.
12 In addition to this point, Stroud registers broadly Wittgensteinian scepticism of the intelligibility of the sensationalist's conception of sensations (and our access to them). Echoing his criticism of the unmasker's appeal to demonstration, Stroud suggests that the impossibility of making a determinate ostensive reference without a prior, descriptive conception of the referent prevents the sensationalist dispositionalist from understanding color sensations as (as we would put it) states like this (here we mount an internal demonstration to our own mental state).
Unfortunately, as observed in note 6, Stroud's presuppositions about what is required for successful demonstration are questionable in the general case, and may be even more questionable in the particular case where the demonstrata are our own mental states (see, for example, the discussion in [Loar, 1997]). Moreover, the non-Wittgensteinian version of the objection about demonstratives contemplated in note 6 (that one cannot mount a successful demonstration to something one believes to be unavailable as a demonstratum) is inapplicable here, since the sensationalist is concerned with demonstrations to our own mental states (states that are, of course, instantiated), rather than demonstrations to uninstantiated properties.
13The brand of irrealism under discussion leaves open the possibility that there is a non-intrinsic property to which yellow sensations are selectively sensitive (for example, there is the disposition to evoke yellow sensations in perceivers). However, such irrealists typically take the view that these non-intrinsic properties cannot be colors, because it is an essential feature of colors, as we understand them, that they are intrinsic properties. On this line of thought, see especially [Maund, 1995].
14To see this point, it may be helpful to consider an analogous dispositionalist analysis of the property humorousness, according to which a joke is humorous if and only if it produces reactions of amusement in appropriately situated thinkers. Once again, this analysis will make the property it analyzes subjective in an interesting sense only if the biconditional relationship between that property and the reactions of thinkers is necessary. One might doubt that that relationship is necessary if one believes that a humorous joke might have failed to produce amusement reactions in thinkers. But, once again, notice that this accusation turns on claiming that whether jokes are humorous or not is independent of the responses they produce in thinkers, and this is precisely what the dispositionalist about humorousness denies.
15In addition to this sort of argument, versions of which were offered by Galileo, Locke, Berkeley, and, more recently, in [Bennett, 1971] and [McGinn, 1983], the literature contains a number of other arguments for subjectivist (dispositional) theories of color, including those based on structural relations among the colors and those based on the nature of our epistemic and metaphysical relationship with the colors (cf., [McGinn, 1983], [Johnston, 1992]). Whether these arguments are found convincing or not, it's unfair to accuse dispositionalists of having given no independent arguments for their view.
16Stroud also doubts that perceptual variation with respect to color is as extensive as has been supposed: ``It is at best an extreme exaggeration to say that there are great variations in the colours people see, especially if the colours in question are identified by words as tolerant as `yellow' and `red' '' (174).
Here I think Stroud is simply wrong about the facts. One way to see this is to note the significant interpersonal variation in unique hues (which are defined precisely in terms of the tolerant words Stroud considers): there is variation in the wavelengths subjects select as looking greenish without looking at all yellowish or at all bluish, as looking bluish without looking at all reddish or at all greenish, etc. (see [Hurvich, 1981], 222-223). It is unsurprising that there is this range of variation in color vision - indeed, it would be surprising if there were not a range of visual effects of an objectively specified stimulus, insofar as there are significant interpersonal variations in human visual systems. For example, even putting aside the non-negligible number of anomalous trichromats, dichromats, and (much more rarely) monochromats, achromatopes, and (possibly) some tetrachromats in the population, there are variations in the numbers and proportions of each type of photoreceptor and in their sensitivity curves in normal trichromats (see [Dartnall et al., 1983], [Pokorny et al., 1991], [Jacobs and Deegan, 1997], [Yamaguchi et al., 1998], and [Roorda and Williams, 1999]), and in the thickness and absorption spectra of the filtering pigment of the macula and the lens of the eye ([Hurvich, 1981], 113-116).
17Here and in what follows I'll concentrate only on interpersonal differences in respect of the color an object looks to different observers, although the same points can be made in terms of intrapersonal differences in respect of the color an object looks to one observer in different perceptual circumstances.
18As an alternative to relying on the scientific and industrial recipes, some have proposed deferring to the perceptual variant exemplified in the largest subgroup of subjects in the population. The by-now standard argument against this proposal, which I believe originates in ([Bennett, 1968], 105-107) involves phenol, which apparently tastes bitter to about seventy percent of the population and is tasteless to about thirty percent of the population. Is phenol bitter or tasteless? It may be tempting to answer that phenol is bitter because it tastes bitter to a larger sub-population. However, as Bennett points out, the numerical majority of one of the groups seems insufficient as a motivation for relativizing to that group in fixing the properties of phenol, since numerical majority could be achieved by either group through mass micro-surgery or selective breeding (cf. [Bennett, 1971], chapter IV, [McGinn, 1983], 9-10, and [Jackson and Pargetter, 1987], 71-72).
19Thanks to Jonathan Ellis, Melinda Hogan, Mohan Matthen, Brian McLaughlin, Patrick Rysiew, Catherine Wilson, and Brad Wray for helpful conversations about these matters and comments on earlier drafts.