Moral realism is the doctrine that there are true sentences that contain moral predicates that accurately describe the real world. Perhaps the strongest argument for investigating the possibility of moral realism is that the alternative is not pleasant to contemplate. Isn’t it true that it is morally wrong that people torture babies for the fun of it? Apparently not, if moral realism is false. Indeed, the very predicate ‘is morally wrong’ becomes meaningless if there are no moral truths. At best, ‘it is wrong that p’ becomes little more than a grunt that roughly means that ‘I bemoan that p’. Without moral realism, there is little incentive to argue for a particular moral position. At best, ethics would amount to a subgenre of literary criticism. The belief in moral realism seems practically indispensable, and not something to be given up lightly. The main problem for moral realism is epistemological. How can we justify our belief that some sentences containing moral predicates are true? One strategy is to model moral realism after scientific realism. However, scientific realism is not without its own problems. In particular, there is the pervasive underdetermination of theory by the evidence. Nevertheless, the scientific method, as fallible as it is, remains the best epistemological technique for forming theories, however uncertain, incomplete, and logically inconsistent though they may be, about the real world “out there”, because of the link between scientific theory and empirical observation. Hence, it would seem that modeling moral realism after scientific realism would entail uncertainty, incompleteness, and contradiction; but at least there would be a link between moral theory and observation. Quine doubted that there was such an empirical link. The empirical foothold of scientific theory is in the predicted observable event; that of a moral code is in the observable moral act. But whereas we can test a prediction against the independent course of observable nature, we can judge the morality of an act only by our moral standards themselves. Science, thanks to its links with observation, retains some title to a correspondence theory of truth; but a coherence theory is evidently the lot of ethics. (Quine 1981, 63) Contra Quine, I shall make a case for moral theory having an empirical foothold. To begin, any theory consists of a body of sentences linked by implication. Additionally, an ethical theory will contain sentences that make use of moral predicates of the form ‘It is morally right that p’, or ‘It is morally wrong that p’, where p is a content clause. I choose to write ‘It is right that p’ or ‘It is wrong that p’, rather than ‘x is right’ or ‘x is wrong’, because the latter forms tend to reify what is mere behavior. For example, the sentence ‘Murder is wrong’ apparently implies that Ex(x is murder and x is wrong). (‘Ex’ means the same as ‘there exist an x such that’.) Murder is not itself an object that exists and that may be described; ‘murder’ is a verb that describes what some objects (moral agents) do to other objects. That’s why it is better to write something like ‘it is wrong that Ex(x is murdering)’; the moral blame is where it belongs. For the sake of clarity, moral descriptions that apply to physical bodies (e.g., ‘good’ and ‘evil’) should be reserved for describing (human) moral agents (see below). Since the words ‘right’ and ‘wrong’ are used in various senses, and are often taken as being more or less synonymous with other words such as ‘good’, ‘bad’, ‘evil’, ‘true’, ‘false’, ‘correct’, or ‘incorrect’ we must distinguish: 1. the moral predicates ‘right’ and ‘wrong’ as they function in an ethical theory (that is the subject of the current discussion); from 2. the moral predicates ‘good’ and ‘evil’ that are used as descriptions of actual moral agents (as in ‘Hitler was evil’); 3. good for and bad for (as in ‘Arsenic was bad for Napoleon’); 4. good and bad of a thing’s kind (as in ‘A dog that hunts is a good dog’); and 5. right and wrong answers (as in ‘5 is the wrong answer to “2 + 2 = ?”’). The first distinction is essential for constructing ethical theory: a body of observational and theoretical sentences, among some of which are sentences of the form ‘It is right that p’ or ‘It is wrong that p’, linked by a network of logical implications. The second distinction is predicated only of moral agents (i.e., humans) or, possibly, events caused by moral agents. Analysis of the distinction between good and evil is important for determining what moral agents desiring to be virtuous should strive to be. It is right that we strive that we be good and not be evil. The third distinction is important for assessing the consequences of various courses of action. Part of the reason it is wrong that someone tortures a baby is that torturing is bad for babies. An important question for environmental ethics is determining the range of entities for which some human actions are bad for the entities in question. The fourth distinction relates to what Rolston (1988, 101) calls a “good-of-its-kind”. In this sense, ‘good’ refers to an excellence at whatever it is that a thing does. Thus, a good blue spruce is an excellent blue spruce. A good can opener is an excellent can opener. Note that ‘good’ in this sense, when applied to humans, may denote morally reprehensible excellences. As Rolston (1988, 101) points out, “Jack the Ripper was a good murderer in the sense that he was clever and never caught, but being a murderer is reprehensible.” Nonhuman organisms, on the other hand, are never evil. To the extent that an organism is a good organism in the excellence sense of ‘good’, that is evidence of the intrinsic value of such organisms, and hence of their moral considerability. (Intrinsic value and its relation to moral consideration will be discussed at length in § 1.2) The fifth distinction is trivially relevant, in that any theory should generate true, rather than false, answers to questions. If we are to model ethical theory after scientific theory, then ethical theory should be built upon a foundation of moral observation sentences (cf. Quine 1992a, 2-9, 1993) that describe recalcitrant empirical experiences. A moral observation sentence would be an occasion sentence of the form ‘It is right [or wrong] that p’, where p is a content clause that states what is the case on a particular occasion. As noted above, Quine himself opposed modeling ethics after his own model of the scientific process by denying that sentences of the form ‘It is wrong that p’ rarely, if ever, count as observation sentences. The only critique of Quine’s metaethics that Quine actually responded to was Morton White’s (1986). White argued that in addition to the usual five senses we use to perceive the world, we also use feeling. According to White, we can tell that an observation sentence of the form ‘it is wrong that p’ is true based on our feelings upon witnessing the situation recounted by p. Thus, if we were to witness a man beating a cripple (Quine’s [1986, 664] example), we know that ‘It is wrong that that man is beating that cripple’ is true because of our feelings of revulsion and outrage. The trouble with this argument, according to Quine’s (1986) response, is that a wide range of possible sensory intake may prompt similar emotional feelings of outrage—including many circumstances where there is little reason otherwise to suppose that something immoral has occurred (as when one is caught in a traffic jam). However, the question of whether the event described by the content clause evokes feelings in a witness is irrelevant to the problem of moral observation sentences. The question is rather whether sentences of the form ‘It is wrong that p’ ever “command a verdict [of true or false] outright” among native speakers of a language who witness the event (Quine 1992a, 5). Thus, a moral observation sentence “commands the same verdict from all witnesses who know the language”—that the described event be “condemned on sight without collateral information” (Quine 1986, 664). Of course, the superfluousness of collateral information for observation sentences in contrast to theoretical sentences is relative and the two types of sentences grade into one another (Quine 1970, 1992a, 6-9; 1993, 108-109). We know a sentence like ‘It is raining’ is true because of a characteristic suite of surface irritations of the outer nerve endings that prompt concurrent observations of drops falling through air, wet skin, characteristic sound and smell. Such characteristic sensory intake are what Quine (1992a, 2-4) called the stimulus meaning of an observation sentence such as ‘It is raining’. The sentence is taken in its simplest form, holophrastically as one word (i.e., as ‘Rain’). The extra “[c]omponent words [(‘it is ____ing’)] are there as component syllables, theory-free” (Quine 1992a, 7). To be sure, people can be hesitant and even make mistakes. I might say that it is raining but someone from Seattle might quibble and assert that it is misting instead. Rather than allow the English language to fission, however, the larger society imposes uniform standards that take into account meteorological theory that in turn govern the conditioning that determines correct usage. Thus, theory infuses even simple observations like ‘It is raining’ or ‘It is misting’. Indeed, merely to posit the existence of a physical body—any body—represents a huge leap of faith (Quine 1970, 16; 1993, 113). Nevertheless, at the most primitive stages of language acquisition, there are examples of entirely theory-free observation sentences. There are pure cases, and they prevail at the early stages of language acquisition. Observation sentences in this pristine purity are the child’s port of entry to cognitive language, for it is just these which he can acquire without the aid of previously acquired language. (Quine 1993, 109) There are, therefore, grades of theoreticity of observation sentences. A dog can learn the stimulus meaning of ‘Squirrel’ (personal observation); a chemist, that of ‘There is some copper in the solution’ (Quine 1993, 108). To assert ‘It is raining’ requires less collateral information than ‘That is a bachelor’. There might be a few people who can reliably pick out bachelors from a crowd based on such bachelorisms as a certain glint in the eye or an inordinate tidiness of appearance; but such bachelor-detectors are rare among English speakers compared to the vast majority qualified to utter ‘It is raining’ upon the appropriate receptor stimulation. So, it is neither the case that ‘It is raining’ is entirely observational, nor that ‘That is a bachelor’ is entirely theoretical. Similarly, observation sentences of the form ‘It is wrong that p’ also contain a non-negligible theoretical component; nevertheless, such sentences may also contain a recalcitrant observational component. For instance, imagine a man extinguishing a lit cigarette on the skin of a baby; it is my claim that the holophrastic sentence ‘Wrong’ uttered by a competent English-speaking witness would be affirmed outright by other competent English speaking witnesses, thus qualifying the sentence as observational on Quine’s criteria. No doubt, the scene would prompt other observations of cringing, wincing, screaming, physical size differences. And so, a foreigner not knowing English who wanted to write a translation manual for his radically different jungle language might be confused at first. But then the person observes some teenagers set fire to a cat, and is again told ‘Wrong’. After a few such incidents, the radical translator begins to see the common thread. In general, moral observation sentences tend to depict glaring examples of gratuitous harm caused by entities having a human-equivalent intelligence or better. Of course, not all harm is directly observable, and not all directly observable harm is gratuitous. However, for moral observation sentences, the harm should be blatant and there should be no obvious reason for causing the harm. Thus, we know that Quine’s (1986, 664) example of a possible moral observation sentence, ‘It is wrong that that man is beating that cripple’, is true not because of the feelings of outrage generated by the observation (although these are present), but because the harm being done is brazenly obvious, and there are no obvious reasons for such behavior. The collateral information required to utter such a moral observation sentence is perhaps more than is required to utter ‘it is raining’, but probably less than ‘that is a rabbit’—since most English speakers cannot tell at a glance the difference between a rabbit (Sylvilagus sp.) and a hare (Lepus sp.), and certainly much less than ‘that is a bachelor’. From a foundation of moral observation sentences, one conducts moral reasoning by considering how the implications generated by these moral observation sentences weave into the overall fabric of implications that constitutes one’s theory of the world. First, one begins to make generalizations based on the initial observations. For example, after witnessing an incident of baby torture, it would be reasonable for a budding ethicist to infer the generalization ‘It is wrong that anyone torture any baby’. Eventually, one will be able to utter justified theoretical moral sentences of the form ‘It is wrong that p’ that apply to circumstances far removed from the original moral observation sentences, even in cases where there is neither harm nor gratuitousness. For example, we probably want to argue that it is wrong that someone steals a ten-dollar item from a Super WalMart store, other things being equal. Such a theft would have the effect of reducing WalMart’s annual earnings by about 0.00000015%. Nevertheless, even though no practical harm was done, we still want to say that it is wrong that someone shoplifts. Eventually, ethical theory generates specific recommendations for particular cases that then offer an opportunity to test the theory. Of course, one contemplating the construction of an ethical experiment must be cautious. Probably, it is usually the case that it is morally wrong that someone actually tests an ethical hypothesis of the form ‘It is wrong that p’. However, this methodological problem is not peculiar to ethics. Many scientific questions cannot be ethically addressed through concrete experiments. For example, it would be an evil neurologist’s dream to have a steady supply of human experimental subjects from which he or she could selectively excise parts of their brains in order to test various neurological hypotheses. But it would be morally wrong that scientists conducted such experiments. In practice, the testing of ethical hypotheses must usually be conducted through case histories and thought experiments. One looks at actual past choices that people have made and then observes the present consequences in order to see what difference a choice recommended by a particular ethical hypothesis would have made; alternatively, one imagines the likely result of an ethical hypothesis if it were universally taken seriously, and then imagines the observation sentences that that result would likely prompt. However, this methodological problem is also not peculiar to ethics. The main data points in a science such as paleontology are primarily historical, and the use of thought experiments plays an important role in science, as in Einstein’s famous thought experiment about what it would be like to ride on a photon of light. In addition, case histories are about the only ethically permissible method of data gathering in sciences like neurology or anthropology. In point of testability, ethics and science differ only in degree, and not in kind. Given the pervasive underdetermination of theory by observational evidence that plagues both science and ethics, what happens is that people soberly reflect on the observational facts that are known, and then reason to the best of their ability until a stopping point is reached where no more empirical evidence is available and no more persuasion can settle an honest disagreement. Scientists will probably never agree whether quantum events are inherently probabilistic or are caused by unseen “hidden variables”—or even whether the universe splits in two every time a quantum event happens so that both alternative events actually happen. In the face of such irresolvable honest disagreements in both science and ethics, decisions about what to do practically must be hammered out democratically. References Quine, Willard Van Orman. 1970. Grades of theoreticity. In Experience and Theory, eds. Lawrence Foster and J. W. Swanson, 1-17. Amherst: University of Massachusetts Press. ———————. 1981. Theories and Things. Cambridge: Harvard University Press. ———————. 1986. Reply to Morton White. In The Philosophy of W.V. Quine, ed. Lewis Edwin Hahn, 663-665. Peru, Illinois: Open Court Publishing. ———————. 1992a. Pursuit of Truth, revised edition. Cambridge: Harvard University Press. ———————. 1993. In praise of observation sentences. Journal of Philosophy 90:107-116. Rolston, Holmes, III. 1988. Environmental Ethics: Duties to and Values in The Natural World. Philadelphia: Temple University Press. White, Morton, 1986. Normative ethics, normative epistemology, and Quine’s holism. In The Philosophy of W.V. Quine, ed. Lewis Edwin Hahn, 649-662. Peru, Illinois: Open Court Publishing.