SOME HAVE DOUBTED that, when we appropriately sift through our catalog of what there is, moral properties will have what it takes to remain (e.g., Mackie 1977; Blackburn 1984; Gibbard 1990; Joyce 2001). In one well-known iteration of this challenge, Gilbert Harman (1977) questions the explanatory utility of moral properties. It seems, he suggested, that we may be able to slim down our ontological commitments without explanatory loss by doing away with moral properties.
This challenge has given rise to a compelling form of nonreductive moral naturalism--sometimes called Cornell realism. Cornell realists accept Harman's challenge and respond by arguing that moral properties have earned a place in our ontology in virtue of the indispensable role they play in the best explanation of various phenomena (Sturgeon 1988, 2006; Brink 1989; Sayre-McCord 1988). (1)
This inventive line of response has considerable appeal. However, its viability depends importantly on a particular thesis about the semantics of thick concepts. Unfortunately for Cornell realists, Vayrynen's (2013) Pragmatic View of thick concepts enjoys considerable support and is hostile to the strongest moral explanations currently on offer. If concepts like COURAGEOUS, KIND, LEWD, OPPRESSIVE and so on are not inherently evaluative, then the most plausible moral explanations Cornell realists have offered cannot do the work they are meant to. Thus, in the absence of strong reasons to reject the Pragmatic View, Cornell realism is significantly weakened. In particular, I will argue, if Vayrynen's Pragmatic View of thick concepts is correct, then Cornell realists must either endorse reductive naturalism in order to retain their most credible moral explanations, or they must restrict moral explanations to their least plausible examples--namely those that employ thin concepts such as GOOD and RIGHT. Neither choice bodes well for their view.
Ontological Rights for Moral Properties?
In his original presentation of the explanatory challenge, Harman (1977) asks us to contrast two cases. In the first, Jane sees Albert pouring gasoline on a cat and then lighting it on fire for fun. This observation leads Jane to noninferentially form the belief that Albert has done something wrong. (2) In the second case, a physicist observes a vapor trail in a cloud chamber and from this observation noninferentially forms the belief that a proton has gone by. According to Harman, it appears that, while we need to appeal to protons in order to explain the observation of the physicist, we need not appeal to the wrongness of igniting cats in order to explain Jane's observation about Albert's action. It seems that facts about Jane's ethical sensibility and her beliefs about what Albert was doing are all that is required to explain her observation. Indeed, given those facts, it looks like Jane would make the same judgment whether or not Albert's act was wrong. So perhaps our moral judgments do not constitute evidence for moral facts.
Sturgeon (1988) objects that this argument relies on an illicit assumption. only one who already accepts moral skepticism will find Harman's contrast convincing, because the dissimilarity between the two cases can only be motivated by assuming that the physicist's theory is roughly true and that Jane's moral views are roughly false. For, if the physicist's theory of particles is mistaken, then we would not need to appeal to protons in order to account for her observation (cf. Brink 1989: 185-86). Thus, in order to formulate a challenge that is convincing to the realist, Harman would need to show that moral facts are explanatorily irrelevant even if commonsense moral beliefs are roughly true. And, Sturgeon maintains, this simply is not plausible. If Jane's moral views are roughly true, then lighting cats on fire for fun is wrong. Since the moral supervenes on the natural, Albert would have needed to do something different to the cat (e.g., pour water on it to clean its fur) in order for his act not to be wrong. And if Jane observed that happening, then she would not have judged that what Albert did was wrong. So it is not the case that she would have judged his act wrong regardless of whether or not it was wrong. And, in general, it seems that our moral judgments counterfactually depend on the moral facts in this way, which suggests that such facts are indeed relevant to the explanation of our judgments. (3)
Now, it is important to remember that Harman himself is not a moral skeptic. (4) He answers his explanatory challenge by motivating a version of reductive naturalism according to which moral properties reduce to certain kinds of descriptive properties (e.g., Harman 1977, ch. 11). Thus he thinks that moral properties in fact do figure in our best explanations, because they are identical to certain explanatorily relevant properties that initially appear less problematic. Sturgeon and the other Cornell realists, however, wish to maintain a version of naturalism according to which moral properties are explanatorily relevant and constituted by nonmoral properties, but cannot be reduced to or identified with nonmoral properties.
A central point of contention between these positions revolves around the distinction between descriptive and evaluative vocabulary. What is the difference between these vocabularies? There is no straightforward answer and I do not intend to propose an analysis of evaluative language here, but it will be useful to have the following criterion from Vayrynen (2013: 55) in the background:
[A] thick term or a concept T is inherently evaluative in meaning if literal uses of sentences of the form x is T in normal contexts entail, as a conceptual matter or in virtue of a semantic rule, that x is good or bad (depending on T) in a certain way--where "a semantic rule" covers also such further semantic properties as conventional implicature and semantic presupposition. Reductive naturalists claim that the properties ascribed by evaluative predicates--call them evaluative properties--are identical to some of the properties ascribed by descriptive predicates--call them descriptive properties. (5) Cornell realists are characterized by the denial of this claim: According to them, the properties ascribed by our evaluative vocabulary are not reducible (analytically or synthetically) to the properties ascribed by some subset of our descriptive vocabulary. (6) Thus, the present debate is appropriately conceived of as centering on explanations that employ evaluative terms and concepts. The key question is whether or not the properties ascribed by those kinds of terms and concepts can be reduced to the properties ascribed by descriptive terms and concepts. (7)
In the sense relevant here, reductions identify one kind of thing with another kind of thing. Such an identification is appropriate when it turns out that one and the same thing (or property) is denoted by two different predicates, e.g., water and [H.sub.2]O. However sometimes it is inappropriate to pursue reductive identifications, such as reducing types of psychological states to types of neurophysiological states (Fodor 1974). For physicalists, instances of e.g., pain are (typically) identical to some instance of a neurophysiological state. However, the fact that the same psychological state can be realized in a variety of different ways makes it improper to reduce the type of psychological state ascribed by "pain" to some particular type of state ascribable using terms available in the vocabulary of neurophysiology. Pain is a natural kind about which interesting counterfactual-supporting generalizations can be made. However, its various instances are not homogeneous at the level of neurophysiology and thus cannot be identified with the natural kinds posited by neurophysiology, despite the fact that such states supervene on and are constituted by neurophysiological states.
In what follows, i want to concede a lot of ground to the cornell realists for the sake of argument. So while it may sound question begging against e.g., compositional nihilists and psychological reductionists, I will accept the antireductionist attitude just outlined. Toward that end, I hope I will be permitted here to use the label reductive mistakes for cases in which a particular view entails a reductive error, such as type-type reductionism about psychological and neurophysiological states. In contrast, I will use the label objectionable reductionism for views that imply that everything--from DNA to beliefs to water--reduces to (or can be eliminated in favor of) the fundamental constituents of the universe (e.g., particles, strings or whatever). With this in place, we return to the exchange between Harman and the Cornell realists.
Cornell Realism and Moral Explanations
In reply to Harman's challenge of explanatory impotence, Cornell realists have tried to motivate nonreductive naturalism about moral properties by arguing that we are familiar with a range of robust, irreducibly moral explanations of various phenomena. Here are some putative examples:
(1) Honesty engenders trust.
(2) Kindness encourages friendship.
(3) Mother Teresa's goodness won her a Nobel Prize.
(4) People are starving unnecessarily because of the selfishness of others.
(5) Racial oppression caused political instability and social protest in apartheid South Africa.
(6) Hitler instigated and oversaw the deaths of millions of persons because he was morally depraved.
(7) Woodworth spent his time and energy arranging his own conveniences rather than organizing an effective search for the missing people because he was no damn good. (8)
These moral explanations, it is claimed, appeal to irreducible moral properties and thus motivate a nonreductive view about moral properties that is analogous to the case for nonreductionism about psychological states. The multiple realizability of moral properties like kindness, oppression and goodness suggests that they cannot be reduced...