How to Get an Ought from a Biological Is.^(*)

Revised and Expanded 12/22/99. There is now a Reply to Harman, 4/7/00. Comments welcome!

[Go to individualism, below, for its role, and that of supervenience, in the is-ought problem.]

In Vaulting Ambition, Philip Kitcher roundly criticizes E. O. Wilson for failing to explain just how ethics and evolutionary biology are supposed to be related. Kitcher lists four types of project one might have in mind. Evolutionary biology might

A. Explain how people come to acquire ethical concepts, to make ethical judgments about themselves and others, and to formulate systems of ethical principles.

B. Teach us facts about human beings that, in conjunction with moral principles that we already accept, can be used to derive normative principles that we had not yet appreciated.

C. Explain what ethics is all about and settle traditional questions about the objectivity of ethics. In short, evolutionary theory is the key to meta-ethics.

D. Lead us to revise our system of ethical principles, not simply by leading us to accept new derivative statements -- as in (B) -- but by teaching us new fundamental normative principles. In short evolutionary biology is not just a source of facts but a source of norms.⁽¹⁾

As Kitcher says, "Wilson appears to accept all four projects, (A) through (D)." He also says, and again I agree, that (A) and (B) are legitimate tasks, that Wilson is thoroughly confused about (C) and (D), and that "the ambitious projects (C) and (D) end in failure." We may wish to add to project (A) that evolutionary biology can legitimately venture explanations of how people came to have what Hume called sympathy and moral sentiment. But this would be a relatively minor revision of Kitcher's list.

A less minor revision concerns (C). I do agree of course that by itself evolutionary biology could never hope to explain what ethics is all about, or hope to settle traditional questions about the objectivity of ethics. But while I think evolutionary biology is not the key to meta-ethics, I also think we can borrow something from it which enables us to give an account of an important kind of objective normativity, one that is significantly related to moral normativity, indeed appears to be a primitive kind of it. If one can show how this kind of objective normativity is not only possible but actual, that will undermine the widespread irrealist presumption that there can be no sort of objective normativity in the world, or at least none that is remotely like a moral ought.

So this paper is an attempt -- or rather a theory-sketch of an attempt -- to naturalize a kind of objective normativity, conducted in the spirit not of elimination but of vindication. Before you roll your eyes so far they get stuck, I hasten to add that this naturalization differs fundamentally from its ill-fated predecessors. It differs mainly by way of satisfying conditions they are typically faulted for violating. Among the conditions any such theory must satisfy, there is one many philosophers are blind to, or at least those philosophers -- and they are legion -- who shun what Strawson calls the revisionary in favor of the descriptive.⁽²⁾ We see this condition at work perhaps most obviously in the sciences. Thus in physics we learn from Steven Weinberg that "Bohr...doubted the [Heisenberg-Pauli] theory would be the great new revolution in physics because it was not sufficiently 'crazy."'⁽³⁾ I call this the craziness condition, and I'm confident you'll come to agree that my theory satisfies it.

Fortunately, there are further adequacy conditions on any naturalistic theory of objective normativity. The first is this. As Kitcher rightly emphasizes, one must somehow get around what he calls "Hume's identification of the 'naturalistic fallacy,"' or what others call Hume's Law: One cannot deduce an ought from an is, or indeed any normative conclusions from purely factual premises. So any adequate naturalistic theory must conform to Hume's Law. But as Kitcher also says,

Pop sociobiology makes no serious attempt to face up to [Hume's Law] -- to pinpoint the conditions under which normative assertions can be garnered from biological premises and to show that the moral principles ... really do stand in the proper relation to the biological findings.

Since I will be trying to "garner" a kind of objective normativity from biological facts, non-deductively, I need to explain what this non-deductive "garnering" comes to, as does any naturalist.

Another adequacy condition concerns Moore's open-question argument (OQA). According to OQA, we can easily imagine ourselves both recognizing that some descriptive condition C obtains (say, that an act or policy x would conduce to the greatest happiness of the greatest number) and nonetheless asking meaningfully -- or, as he says, "with significance" -- whether x has normative property N (say whether x is good); this "shows clearly that we have two different notions before our minds"; therefore, N and C cannot be identical or equivalent, nor can their predicates be synonymous;⁽⁴⁾ N cannot be reduced to C. An adequate naturalistic theory of objective normativity must either explain why OQA is unsound or why, though possibly sound, OQA does not apply to the theory's account of how the normativity is determined by purely natural affairs.

Another condition concerns J. L. Mackie's argument from queerness (AQ). According to AQ, objective normativity would have to be a very queer sort of thing, since its relation to what clearly is objectively the case would be highly mysterious. No allegedly objective normative property N of x is deducible from x's non-normative or natural properties, as Hume taught us. Nor can N be reduced to natural properties of x, as Moore taught us. Furthermore, talk of supervenience of N on the facts is too watery to do the job. Since no other relation has been spelled out that works, the allegedly objective normativity must be a very queer thing indeed. "How much simpler and more comprehensible the situation would be," therefore, "if we could replace the [normative] quality with some sort of subjective response which could be causally related to the detection of the natural features on which the supposed quality is said to be consequential."⁽⁵⁾ What we call objective normativity is just our subjective attitudes projected onto the value-neutral real world. Thus an adequate naturalistic account of objective normativity must show that AQ is unsound, by way of explaining just what the needed non-mysterious relation is between the objective normative properties and the relevant natural features.

It is true that Mackie construes (or seems to construe) the would-be objective norms as intrinsically motivating, and, Mackie aside, it might seem that according to any adequate theory of the relevant objective normative properties, anyone aware of something's having such a property must be motivated to act in accord with it. But this issue of whether the normative properties should be intrinsically motivating is one of the most contested in metaethics; to impose one answer or the other as an adequacy condition would be question-begging. So too for the idea that an objective normative property is such that anyone aware of it will necessarily have a reason to act in accord with it. And in any case, the kind of normative property targeted here is not meant to be the sort of fully moral property, applicable to us modern humans, which might plausibly be construed as intrinsically motivating or reason-affording.⁽⁶⁾

However, there is another feature which the targeted kind of normativity does share with the fully moral kind. This is the distinction -- or gap -- between what someone or something actually does or is disposed to do and what would be normatively better or worse. We see the distinction at work in our reaction to cynical remarks like "An impeachable offense is whatever the House says it is" (as then Representative Gerald Ford said with regard to Chief Justice Earl Warren). What the House ought to do cannot be reduced to what it does or is disposed to do. We see the distinction at work as well when we speak of a defective heart, meaning one that does not or is not disposed to do what it is for, or perform as it is supposed to or as designed. Let a token heart be so diseased, damaged or deformed as to be totally incapable of pumping blood; nonetheless, pumping blood is what it is for, what it is supposed to do.

This gap between how a thing x actually behaves or is disposed to behave and how it should behave is essential for normativity, or at least for the kind of normativity at issue here.⁽⁷⁾ Any adequate theory of such normativity must preserve the gap. More precisely, where N is such a normative property -- a property to the effect that x should behave thus-and-so -- how an individual item x actually behaves in this respect, or is disposed so to behave, does not affect whether x has N, just as the actual behavior or disposition of your heart does not affect whether it has the property of being supposed to pump blood.

Another adequacy condition for a naturalistic theory of objective normativity is that the matter of whether an item x has a normative property N is to be determined solely by purely natural affairs. After all, one of naturalism's core theses is that all properties whatever are thus determined. But another reason for imposing the condition is that a property is clearly an objective one only if it is determined by objective affairs, and the theory had better show that it is. Note also that if N is determined by natural conditions, it follows that there can be no difference between individuals x and y as regards having N without some relevant difference in natural conditions. This amounts to a kind of universalizability principle that even non-naturalists insist on.⁽⁸⁾

The final adequacy condition -- unless others turn up -- is that the normative properties play some appropriate explanatory role. Or at least this is a condition I will impose on myself. To require it of any adequate naturalistic theory may well be excessive, in the sense that it is problematic even for naturalism whether to be is to be capable of figuring in a causal explanation of something else (or: "To be real is to have causal powers").⁽⁹⁾ Even for naturalists it may be enough to require only that a genuine property either play an explanatory role or be explainable or at least determined by affairs that do.⁽¹⁰⁾ Such a property would be "ontologically grounded," in an important sense, and its relation to the causal order would neither be mysterious nor land us in substance dualism or a dualism of matter and spooky entelechies or pulls from the future. On the other hand, many naturalists and others seem united in believing that the result would still be unacceptably epiphenomenal from the point of view of naturalism. So to placate them I will require, for the duration of this paper, that the objective normative properties play an appropriate explanatory role.

Summing up, an adequate naturalistic theory of the targeted normativity must (i) conform to Hume's Law; (ii) escape OQA; (iii) escape AQ; (iv) preserve the gap between how x actually behaves or is disposed to behave and how it should behave; (v) explain how the matter of whether x has a normative property N is determined solely by objective affairs; and, at least in the present case, (vi) accord N an appropriate explanatory role. Oh, and yes, (vii) the theory must be sufficiently crazy.

Now how can we possibly make our way through this minefield? Start with condition (iv). The gap between how something actually behaves or is disposed to behave and how it should behave is difficult to preserve, perhaps impossible, if you are not only a naturalist but an individualist. According to individualism, all a thing's properties are determined by a subset of its own base properties or those of its parts. This individualism pervades the mechanism present at the creation of modern science and philosophy, according to which all of a thing's genuine properties are determined by kinematic properties of its parts. Individualism survives in the reductivism according to which all of a thing's genuine properties must be reducible to its own base properties, in the sense of being identical or at least (nomologically) equivalent to some compound or (first-order) construction of them. When the base properties are allowed to include relational properties, the individualism is relational rather than non-relational or intrinsic.

Note too how individualist OQA and AQ are in their focus on -- obsession with? -- the connection between an individual's normative properties and its own natural properties and relations. Individualism persists also in so-called "supervenience theories" of the moral, according to which, in Robert Audi's rendition,

first, no two things, whether acts or persons, can share all their natural properties and differ in their moral ones (if they have any); and second, any entity having moral properties possesses those properties in virtue of its natural properties (or certain of them)....⁽⁸⁾

where the entity's having the properties "in virtue of" its natural properties is meant to imply that what determines a thing's moral properties are its own natural properties.⁽¹¹⁾

The difficulty for individualist naturalism is twofold. First, suppose a normative property of x is determined by some natural property of x -- say the normative property of being supposed to pump blood is determined by some natural property of the heart. Which one? The most likely candidate would be the heart's disposition to pump blood. But this means there would be no gap between the heart's disposition to pump blood and its being supposed to do so, contrary to adequacy condition (iv). There may be ways around this difficulty, but they would come at a high price, and in any case, rather than struggle with them here, we do better to turn to the second difficulty, which appears fatal anyway.

Individualism's starting assumption -- that all a thing's properties are determined by its own base properties or those of its parts -- appears to be false even for many non-normative properties. In biology, social science, perhaps even physics itself, one encounters properties that on the evidence are not determined by their bearers' own base properties, relational properties included, but only by these together with the base properties of other things at some distance in space and time, as we'll see in a moment in the case of biology.⁽¹²⁾ Furthermore, non-individualism offers a far smoother way of preserving the gap between what something does or is disposed to do and what would be normatively better or worse. According to the non-individualist, what determines whether x has normative property N could well be natural conditions that include no natural properties or relations of x, or at least none entailed by how x itself actually behaves or is disposed to behave; in such cases, obviously, what x itself does or is disposed to do cannot affect whether x has N.

Now back to biology. The story I'm about to tell first occurred to me while wondering what Ruth Millikan's theory of proper function and the teleosemantics it drives might imply for metaethics.⁽¹³⁾ But it proves useful to reformulate the story in terms other than Millikan's. Instead of talking about a thing's proper function, let's talk about what an adaptation is for, in the sense of 'what for' unpacked by Robert Brandon.⁽¹⁴⁾

Consider a teleological "what-for" question, say "What is the heart for?" Of course a given trait might not be for anything, in which case a what-for question is out of order. But "Whenever we hypothesize that some trait is an adaptation, it makes sense to inquire about its function," what it's for. I would add only that an assertion of what something is for is a normative assertion (as may be implicit in Brandon's account). When we say, in the intended sense, that the heart is for pumping blood, we imply that even when a token heart cannot possibly pump blood, nonetheless pumping blood is what it is for, what it is supposed to do; we call it defective, distinguishing between what it actually does or is disposed to do and what it should do. The notion of what x is for is a member of a family of interlocking notions, such as the notion of what x is supposed to do, what its proper function is, its purpose, what it should or ought to do; x is for C iff x is supposed to C iff C is its proper function, its purpose, what it should or ought or is supposed to do. At least that is how I use the terms here; in context, to explicate one is to explicate them all.

One important context of use of such notions is in connection with adaptation. As Brandon says, "Adaptations in nature seem to call for teleological explanations" -- explanations of what the adaptation is for, of its function. Black color in the peppered moth is an adaptation, and as such it calls for an explanation of what the color is for (it is for camouflage). Explanations of adaptation "are teleological in the sense that they are answers to what-for questions." They tell us specifically that an adaptation A is for C (the black color is for camouflage), and they explain how it came about that A is for C. But even though adaptation explanations are answers to what-for questions, "they are also perfectly good causal/mechanical explanations." For the "adaptation is the direct product of the process of evolution by natural selection," and we explain the process in terms of "the ecological consequences of the adaptation, or its precursors, that explain its adaptive advantage over its alternatives."⁽¹⁵⁾

As Brandon argues, adaptation explanations fit not only the causal/mechanical or bottom-up model of explanation. They fit the top-down or unification model as well, in that "All adaptation explanations, no matter how diverse the adaptations, subsume the explained adaptation under the [Principle of Natural Selection]."⁽¹⁶⁾ The two models are not in conflict; rather, "a complete adaptation explanation has both unification and causal/mechanical aspects." In any event, we may conclude that "(1) adaptation explanations are thoroughly mechanistic; but (2) they serve to answer teleological questions," so that "the sense in which what-for questions and their answers are teleological" is explained in purely mechanistic terms.

Note also that a normative property N to the effect that adaptation A is for C plays a suitable explanatory role, so that adequacy condition (vi) is satisfied. Where A is an adaptation, we can explain A's existence by reference to what A is for: A exists because A is for C; the normative property N of A's being for C explains why A exists. For example, the black color of the peppered moth in soot-polluted environments exists because it is for camouflage. More fully, the black color exists because of the camouflaging effect of past instances of the black color (the first instances being explained by mutation). More fully still, and generalizing, "adaptation A's existence is explained by a complete adaptation explanation that includes not only the ecological account of the function of the adaptation, but also the other four [kinds of information needed for a complete adaptation explanation]."⁽¹⁷⁾

What about the other adequacy conditions? Clearly, the normative matter of A's being for C is determined by objective affairs; that A is for C is determined by physical affairs in the relevant natural-selective history (as explained by a complete adaptation explanation). So condition (v) is satisfied. Further, it is clear from the theory that given an instance x of an adaptation A -- that is, given a token x of A -- x's having the normative property N of being for C is determined by causal/mechanical affairs in a natural-selective history prior to the appearance of x, which affairs clearly do not include those of x's own causal properties or relations that are entailed by how x itself actually behaves or is disposed to behave.⁽¹⁸⁾ Hence this kind of non-individualist determination of N makes it clear that x's being for C is not affected by what x actually does or is disposed to do. Condition (iv) is satisfied.

What about conditions (i)-(iii)? Condition (i) -- Hume's Law -- is satisfied, since the normative matter of adaptation A's being for C is not deduced from a description of the causal/mechanical affairs in the relevant natural-selective history in which (often enough) ancestors of A had the causal effect C in environment E. Rather, the theory proposes -- hypothesizes, tries on for size -- that A's being for C is a matter of past instances of A having been selected for in virtue of their causal effect C in E. Here we are basically following Darwin, who did not deduce what seemed to many to be a real feature of the biological world -- design, what something is for -- from a description of certain effects of past instances. Rather, he proposed that this is what design is. Then he justified the proposal, in effect, by detailing its fruits within his surrounding theory.

This strategy, though routine in science and congenial to naturalism, is alien to conventional philosophical methodology -- even to some naturalists. Hence we need to be clear what it comes to. Consider first an example from physics (oversimplified for purposes of illustration). From only the bare physical fact that the molecules in my coffee have a certain mean kinetic energy, one cannot deduce that it is piping hot or any other temperature. What then is temperature, and can it be explained in terms of mean molecular kinetic energy (mmke)? The strategy adopted by kinetic theorists of temperature (near enough) is to propose or provisionally assume the identity or at least the (nomological) equivalence of temperature and mmke; they propose treating the two phenomena as the same (in standard conditions), on the ground that doing so would enable the resulting theory to unify the phenomena, and to do so with greater adequacy way than alternative theories. The assumption of identity or at least equivalence is then justified by its consequences under the containing theory. Provided the theory does enjoy the greater adequacy, we are warranted in claiming to have discovered that the temperature of liquids is identical or at least (nomologically) equivalent to the mmke of their molecules (in standard conditions).

This sort of strategy is so widespread and useful that it deserves a name -- say, "inference to the best identity," or, where the situation warrants only equivalence, "inference to the best (nomological) equivalence," though the name could mislead. By whatever name, the strategy is revisionary, not descriptive, again by contrast with conventional philosophical methodology. The aim is not to analyze or capture or preserve our vernacular concept, whether of temperature or design or whatever. Nor is it to analyze or capture or preserve ordinary usage or even philosophers' usage of, or intuitions about, the predicates 'temperature' or 'design', say by way of advancing necessary and sufficient conditions for such usage or for whatever property the vernacular/philosophical concept or usage might pick out if any.

Similarly, Brandon's account of what an adaptation is for, like Millikan's of proper function, is not an analysis, either of the vernacular concept or of ordinary usage or of philosophers' usage -- or even biologists' usage. Nor does it propose necessary and sufficient conditions for any of these or for whatever property they might pick out if any. The theory no more attempts this than Darwin attempted to capture what philosophers, theologians and ordinary folk might have meant -- or even fellow biologists -- when they spoke of design; most of them thought of design as entailing a knowing designer, whereas Darwin emphatically did not; the watchmaker is blind. Indeed we folk can easily imagine ourselves both recognizing that past instances of A had the effect C and nonetheless asking meaningfully -- or, as Moore says, "with significance" -- whether A was designed to do C; this "shows clearly that we have two different notions before our minds." So too can we imagine ourselves both recognizing that the coffee's molecules have a high mean kinetic energy and nonetheless asking "with significance" whether the coffee is hot.

But to conclude from this, as Moore's OQA would require, that the pair "A was designed to do C" and "A's past instances had the effect C" cannot be equivalent, or that the affairs they denote cannot be identical, would be to misconstrue Darwin's strategy, which is inference to the best identity or equivalence. Objecting to the resulting notion of design that it is not what the folk meant -- or the philosophers or even the biologists of the day -- would be as beside the point as objecting to Einstein's theory of mass that it fails to accord with what the folk or the philosophers or even the physicists of the day meant by `mass', on the ground that for him mass is not conserved in all interactions (which is true) whereas for the philosophers and the Newtonians it is (also true). Thus OQA is unsound, committed as it is to the non-sequitur from non-equivalence of entrenched notions to non-equivalence of the affairs in the world targeted by inference to the best identity or equivalence.

As to the present theory, it too is revisionary, along the lines of Darwin's. It proposes that what has seemed to many to be a real feature of the biological world -- the normative matter of what an adaptation is for, what it is supposed to do, its proper function -- be understood in terms of certain effects of past instances. Provided the proposal figures in a theory that is more adequate in relevant ways than competing theories, it amounts to a discovery. What we've discovered is that a feature or phenomenon we saw but darkly -- indeed often doubted was objectively there at all, as many still doubt -- turns out to be a matter of the ecological consequences of an adaptation's precursors which explain its advantage over alternatives. In evolutionary biology, the answer to the normative question "What is adaptation A for?" is, "A is for producing the consequences in virtue of which past instances of A were selected for," and the strategy used to get to this conclusion is basically inference to the best identity or equivalence (though as I've said, the name could mislead). Open Question Arguments are as unsound here as they are against the kinetic theorists or Darwin or Einstein or anywhere else a theory uses inference to the best identity or equivalence. Thus condition (ii) is satisfied.

Finally, according to condition (iii), an adequate naturalistic account of objective normativity must show that the Argument from Queerness is unsound, by way of explaining just what the needed non-mysterious relation is between the objective normative properties and the relevant natural features. Clearly, AQ is out of place here insofar as it relies on Hume's Law and OQA, and for the same reasons. But in addition, contrary to AQ, we can explain what the needed non-mysterious relation is, namely the relation of nonreductive determination (also called "global supervenience," often pejoratively).⁽¹⁹⁾ As seen, that a token x of an adaptation A has the normative property N of being for C is determined by causal/mechanical affairs in a natural-selective history prior to the appearance of x. Since many of these affairs do not amount to properties or relations of x (contrary to the conventional wisdom of chasing propertyhood up the tree of syntax), the normative property N of A's being for C is not being reduced to properties or relations of x (though this is not by itself what makes the determination relation nonreductive).

So all seven adequacy conditions appear to be satisfied (the satisfaction of the craziness condition being so obvious as to require no argument); we seem to have crossed the minefield. I regard this, in conjunction with Brandon's theory, as a vindication, if one is needed, of Millikan's theory of proper function (strictly, of what she calls direct proper function, but the rest follow immediately given her recursive definition of the general notion of proper function). Furthermore, it is a vindication of the objective normativity she assumes proper function has but nowhere argues.⁽²⁰⁾ Call this kind of objective normativity primitive or minimal normativity. Where adaptation A is for C, A's property of being for C is a primitive normative property.

This way of negotiating the minefield can be generalized, in at least two ways. One is by extending it so as to apply to items that are not adaptations. The imprinting mechanism in a newly-hatched chick is an adaptation for imprinting Junior on his mother. But in the event that Junior's mother is Congreva, it follows (by substitutivity of identity) that the mechanism is for imprinting Junior on Congreva. It follows further that Congreva is what Junior is supposed to be imprinted on; Congreva has the normative property of being what Junior should be imprinted on. Yet Congreva is not an adaptation, even though she contains many adaptations. She is a particular individual organism, new under the sun. The mechanism does not imprint Junior on a type (motherhood) or on any other denizen of Plato's heaven, but on a concrete token; nor was the imprinting mechanism selected for, or its ancestors selected for, in virtue of the effect of imprinting past hatchlings on Congreva. We might call this normativity a kind of "derived" normativity, derived as it is from that of what the mechanism is for. It is significantly less primitive than the normativity involve in an adaptation A's being for C.⁽²¹⁾

Here is another example of this less primitive normativity. There are mechanisms in the rat that are an adaptation for preventing the rat from eating what tastes or smells like the stuff it had when it got sick (or its fellows had when they got sick or died). Suppose that the stuff King Rat had when King Rat got sick is the children's silly putty -- a substance nowhere encountered in the evolutionary history of rats. Then King Rat's mechanisms are for preventing King Rat from eating silly putty; silly putty is something King Rat (and his fellows) should not eat. "In this manner, animals that learn can acquire biological purposes that are peculiar to them as individuals, tailored to their own peculiar circumstances or peculiar history"⁽²²⁾ An animal can come to have an objective normative property that is not only novel but instantaneously so, and so can things in its environment, including silly putty.

A second generalization is this. What amounts to formally the same strategy for negotiating the minefield might well be made to work for kinds of normativity that presumably are not determined by affairs in a natural selective history (and not even by such affairs conjoined with derivations from what an adaptation is for). Instead, some not-at-all-primitive kind of normativity might arguably be determined by, say, what the parties to the original position would choose by way of first principles, or by what would contribute to the greatest happiness of the greatest number, or whatever. Suppose so. Then the strategy would be to (a) use inference to the best identity or equivalence in order to equate the targeted normativity with the relevant non-reductively determining descriptive affairs in such a way as to avoid individualism, thus preserving the gap between how an individual behaves or is disposed to behave and what would be better or worse, as well as undermining objections from Hume's Law, OQA and AQ; and (b) show how the targeted normative properties play an appropriate explanatory role. Provided the resulting theory of the targeted normativity is successful at each step of the way, the result would be a naturalistic theory of objective normativity that satisfies the key adequacy conditions for any such theory, and thereby deflects the usual-suspect objections to any such naturalism.

In the present paper the aim is less ambitious: to give an account of an important kind of objective normativity that is significantly related to moral normativity, indeed may be a primitive kind of it, and thereby to undermine the widespread irrealist presumption that there can be no sort of objective normativity in the world, or at least none that is remotely like a moral ought. On the other hand, what could the primitive normativity discussed so far possibly have to do with ethics?

Consider the guppy.⁽²³⁾ When a large fish nears a school of guppies, some of the guppies will approach the stranger to see whether it is a predator. If it is, then even though the scouts can warn the school, they will likely be eaten unless they stick together long enough to confuse the predator by scattering. As they approach the stranger, they will hold back if they see that certain others are holding back. Indeed they keep track of which ones held back on the previous occasion, and will themselves hold back when accompanied by such individuals. When accompanied by individuals that did not hold back on the previous occasion, they will themselves keep up on this occasion (often enough). That is, they will keep up or hold back according to what those others did on the previous occasion. If a new individual swims out, the other scouts will keep up the first time, but thereafter do what that individual did the time before.

Clearly the guppies are able to identify individual guppies and to remember which ones did what on previous occasions. In addition, the behavior of the scout guppies (often enough) can be described as Tit for Tat: on the first occasion, keep up, thereafter do whatever the other did on the previous occasion. Evidently there are guppy mechanisms responsible for, among other things, identifying other guppies, remembering what they did on the previous occasion, and (often enough) getting the guppy to keep up on the first occasion while thereafter doing what the other did on the previous occasion. Furthermore, such reciprocal cooperative behavior significantly increases each individual guppy's chances of survival by contrast to always keeping up, never keeping up, and so on. The extra risk to the individual of keeping up on any given occasion is outweighed by the benefit to it of Tit-for-Tat behavior in the longer run. And of course the whole school benefits as well.

Presumably these mechanisms are an adaptation. If so, among the things they are for is producing the Tit-for-Tat behavior. More precisely, they are for getting the guppy to keep up on the first occasion and thereafter do what the other did on the previous occasion. Equivalently, past instances of the mechanisms were causally responsible, often enough, for getting guppies to conform to the Tit-for-Tat rule, "Keep up on the first occasion and thereafter do what the other did on the previous occasion." In general, let us say that where adaptation A is for getting organism x to C (keep up . . .), then A is for getting x to conform to rule "C" ('keep up . . .'); getting x to conform to rule "C" = getting x to C = the effect in virtue of which A's ancestors were selected for. Clearly the Tit-for-Tat rule is unexpressed and unexpressible by the guppy, and presumably not otherwise represented by the guppy to itself; nor do we find the rule by peering into the guppy's head. Nonetheless it is one of the guppy's biological purposes to conform to the Tit-for-Tat rule, which is to say only that the guppy contains mechanisms that are for producing conformity to the rule.⁽²⁴⁾

Tit for Tat is a rule that bears on interactions among individual organisms, in this case among conspecifics. As in many other examples, the encounters can be dangerous for one or both parties. The history of encounter amounts to an iterated prisoner's dilemma (for which the element of danger is not necessary, only a certain ordering of payoffs). In many cases, the rule of encounter conformity to which best promotes the organism's fitness is Tit for Tat.⁽²⁵⁾ For certain kinds of organism in the relevant encounter situations (environments included), what the relevant mechanisms in the organism are for -- what they should or ought to do -- is to cause it to behave in conformity to the Tit-for-Tat rule. And to say that this is what the organism should or ought to do is simply to say that it contains a mechanism of a kind that has historically proliferated in virtue of causing behavior, often enough, that conforms to the rule.

Now I want to suggest that this is an instance of a primitive or minimal kind of moral normativity, a protomoral ought. Not only does conformity to Tit for Tat bear on interactions among individual organisms, which is an essential characteristic of a moral rule. Tit for Tat resembles moral rules in a respect so fundamental that we may call it protomoral. According to most philosophers and ordinary folk alike, the point of morality is primarily, if not entirely, to overrule acting in one's own self-interest when so acting would harm others.⁽²⁶⁾ In the guppy case, the point of Tit for Tat is to overrule the individual scout's acting in its own immediate self interest -- namely, acting to minimize its chances of being eaten, by holding back even when others do not. Further, by so holding back the scout would harm those who do not, by increasing their chances of being eaten. So in the guppy case, Tit for Tat resembles moral rules in a quite fundamental respect: it is to overrule following self-interest when following self-interest would harm others. This respect is so central to our concept of morality that we may call Tit for Tat a protomoral rule. We may say further, in line with this usage, that for the relevant kinds of organism in the relevant kinds of encounter situations, behavior in conformity to the Tit-for-Tat rule has the property of being protomorally obligatory; behavior contrary to the rule is protomorally wrong. (Hence the proposition that keeping up on such-and-such occasion is protomorally obligatory would be "intrinsically action-guiding" or have "to-be-doneness" built into it, in the sense that it entails a rule or standard or norm, namely "Keep up on the first occasion, thereafter do what the other did on the previous occasion.")⁽²⁷⁾

Can we go further? Here is a tempting possibility. Kitcher advances an explanation of how human altruism could possibly have evolved under natural selection.⁽²⁸⁾ Suppose he is right that for our savannah-dwelling hominid ancestors, "selection will favor 'golden-rule' altruism of a discriminating kind (treat the other as oneself so long as one has no basis for thinking that the other will not do the same)."⁽²⁹⁾ Then by the present account, conformity to this rule is protomorally obligatory, and objectively so, for relevantly like organisms in the relevant kinds of encounters (which certainly need not include us modern humans). To say that conformity to this rule is protomorally obligatory is to say that the organism contains a mechanism or mechanisms -- an adaptation -- of a kind that historically proliferated at least partly in virtue of getting ancestor organisms, often enough, to conform to the rule. It may be possible to reduce the reliance on genetically determined mechanisms if the organism is able to learn the appropriate rule-conforming competence and pass it on, often enough, to others, including offspring, and to refuse to play with those who either don't learn it or don't conform to it. In this case, we would be dealing with a special case of the less primitive derived objective normativity described above. (And in either case the proposition of protomoral obligation would again be intrinsically motivating in the sense of entailing a rule or standard or norm.)

Of course this falls well short of fully human altruism, as Kitcher emphasizes; you can't get here from there. That is, you can't infer that we modern humans should conform to golden-rule altruism of a discriminating kind from the fact that our savannah-dwelling hominid ancestors had a protomoral obligation to do so, at least not without helping yourself to further and highly contestable assumptions. Nonetheless, we would at least be entitled not only to say with Kitcher that "a recognizable, if rather minimal, type of human altruism" might evolve under natural selection. We could also conclude that conformity to this altruism is objectively protomorally obligatory for relevantly similar creatures in the relevant kinds of environment. If nothing else, this should disturb the slumber -- or rattle the cage -- of those convinced that there can be no sort of objective normativity in the world, and certainly none that is remotely like a moral ought.

* To be presented at the Invited Symposium on the Evolution of Norms, APA Pacific Division meeting, Albuquerque, April 7, 2000, Gilbert Harman commenting; see Reply to Harman. For comments on ancestor drafts, I am indebted to audiences at Duke, Virginia Polytechnic and Western Michigan universities and the North Carolina Philosophical Society, and especially to Michael Ferejohn for detailed and persistent critique in follow-up correspondence over some weeks.

1. Kitcher (1985), 417-418.

2. Strawson, xiii.

3. Weinberg (1976), 13.

4. Moore (1980), 15ff. Cf. Ball (1988).

5. Mackie (1977), 41.

6. Also the account developed below may prove neutral between the internalist and externalist positions here, insofar as one could combine it with the standard-based approach in Copp (1995).

7. I am indebted to Ümit Yalçin for getting me to see that there may be some esthetic normative properties, perhaps among others, to which this distinction does not apply.

8. For details, see Post (1987), Ch. 6; and, with regard to whether the natural conditions must all be properties or relations of x, Post (1995), 80-85.

9. This is what Kim (1992), 135 endorses as Alexander's Dictum. Post (1999a) argues that both the dictum and Kim's closely related principle of the Causal Individuation of Kinds suffer numerous solid empirical counter-examples, especially from biology.

10. On the relations between determination and explanation, see Post (1999b).

11. Audi (1993), 96. Nearly all the supervenience relations currently on offer are individualist. Indeed Kim (1993), 58, 79, 84, goes so far as to say that it is "highly plausible to regard ...Weak Supervenience ... as minimally necessary for any claim of determination or dependency between sets of properties,'' where WS is the thesis that necessarily, for any x and any y, if x and y have the same base properties P, then they have the same higher-level properties N -- which is individualist if anything is. The notable exception to these individualist relations is "global" supervenience, otherwise known as nonreductive determination, which is the relation invoked in the present paper and is defended at length in Post (1987), (1995), and (1999a).

12. Cf. Post (1995), §§2-3, which contains further references, and where the objection that there must be a (compound) physical relational property of the bearer in virtue of which the determination holds is met in detail in by showing how the candidate relations all fail to do so in critical cases; the culprit is the uncritical tendency of many philosophers to chase propertyhood up the tree of syntax. Naturalists, at least, would be well advised to avoid supervenience and determination relations that entail individualism, which includes any relation that entails Weak Supervenience, such as Strong Supervenience. According to Railton (1995), 100n18,"It is widely held that the supervenience of the moral upon the natural is (what has been called) strong supervenience."

13. Why would anyone wonder about that? Well, during most of this century each new development in philosophy of language, from positivism through sprachspielism and causal theories to post-structuralism, has driven some new program in metaethics, from emotivism through prescriptivism and realism to textualist genealogy; what about hers? But the present paper is not meant as offering a general program in metaethics, and even if it should prove to anticipate one, there are other ways one might try constructing a metaethics from teleosemantics or its underpinnings.

14. Brandon (1990); all quoted passages below are from pp. 139, 165, 185-89 of this book.

15. More fully, an ideally complete adaptation explanation requires supplying five kinds of information: "(1) Evidence that selection has occurred, that is, that some types are better adapted than others in the relevant selective environment (and that this has resulted in differential reproduction); (2) an ecological explanation of the fact that some types are better adapted than others; (3) evidence that the traits in question are heritable; (4) information about the structures of the population from both a genetic and a selective point of view, that is, information about patterns of gene flow and patterns of selective environments; and (5) phylogenetic information concerning what has evolved from what, that is, which character states are primitive and which are derived."

16. Where PNS is "If a is better adapted than b in environment E, then (probably) a will have greater reproductive success than b in E." Brandon (1990), 11.

17. See note 15, above.

18. Again see Post (1995), §§2-3, for a reply to the objection that there must be a physical relational property of the bearer in virtue of which the determination holds.

19. Cf. note 11, above, as well as Post (1984) and Post (1987), Ch. 6.

20. It is also an illustration, if not a vindication, of an important part of the nonreductive determinationist metaethics floated in Post (1987), Ch. 6.

21. Cp. Millikan (1984), (1993), on derived proper function. Whether derived normativity satisfies the adequacy conditions may be left as an exercise for the reader. But condition (vi) -- that N must play an appropriate explanatory role -- is a bit tricky, so here is a hint: Why did Jr. imprint on Congreva? Because Congreva is what Jr. should imprint on. More fully, Jr. contains an adaptation -- the imprinting mechanism -- which exists because it is for imprinting Juniors on their Moms, and Jr.'s Mom is Congreva. More fully still, Jr. contains an adaptation A which exists because of the effect of past instances of A in imprinting past juniors on their Moms in E, and Jr. is in E and his Mom is Congreva.

22. Millikan (1993), 226.

23. And the stickleback, among others. Cf. Pool (1995), from which the following is drawn.

24. Cf. Millikan (1993), 210-239, on purposefully conforming to an unexpressed rule.

25. There are important cases where other rules do better. Cf. Axelrod and Hamilton (1981); Axelrod (1984); Axelrod and Dion (1988); Kitcher (1993), which contains further references.

26. Cp. Baier (1958), 309: "The very raison d'etre of a morality is to yield reasons which overrule the reasons of self-interest in those cases when everyone's following self-interest would be harmful to everyone."

27. If Copp (1995), 199-200, is right and we apply his account here, then the present position is compatible with both internalism and externalism.

28. As opposed to how it actually evolved. Kitcher (1993), note 6.

29. Kitcher (1993), 513. He labels this rule DA_0.5R. An Altruist rule is Discriminating if it requires the organism to play with any organism that has never defected on it, and, when playing, always cooperate. DA_0.5R is also unforgiving: the organism is never to play with anyone who has ever defected on it. Both more and less unforgiving strategies are possible, though probably dominated by DA_0.5R.

References: