INDIVIDUAL NATURES

©  Ronald de Sousa

draft of a paper now published in Philosophia: Philosophical Quarterly of Israel, 26:3-21 (1998)

There is an old battle about whether the human sciences should aspire to the condition of natural sciences. Those who reject such aspirations sometimes think the human sciences must have an entirely different status; that they must be construed, for example, as concerned with interpretation rather than causal explanation.1 From that position there is perhaps only a verbal step to the view that `human sciences' is a misnomer--that what we call by that name are not properly sciences at all--and that we must resist the imperialism of science. That resistance to the ideal of unified science has taken many forms; but the following have been especially salient:

(i) Objections from free-will. A major aim of science is general understanding; that understanding requires that we formulate laws of nature. But there can be no laws about human behaviour, because we have free-will. Whatever else that might mean, it entails that there is always a real possibility of a choice invalidating even an ideally well-grounded prediction. Human choices, in brief, are not governed by laws at all, and so are not properly the object of scientific study.²

(ii) Objections from the independence of levels of description. One well known objection in this class was formulated by Donald Davidson, and is based on the thesis of the anomalousness of the mental. There can be no psycho-physical laws, Davidson argues,³ because the principles of organization of mental events are wholly different from those that apply to physical events. Only physical events can be causes, including mental causes. The mode of explanation applicable to behaviour is couched in terms of rationales, which cannot be adequately correlated with the real causes of the events. For those must be specified in terms of underlying natural laws.

(iii) Objections from the social character of psychological traits: these go back to Wittgenstein and also assume many forms. Nowadays the most common versions involve claims that psychological predicates are not actually predicates of individuals at all, but are ascribed on the basis of social facts.⁴ That puts them at yet one more remove than the last set from any putative physicalist base.

I mention these lines of attack only to remind you of how familiar is the thesis that we should give up on the ``human sciences,'' unless we intend by that term a vacuous honorific. Rather than to pursue any of the debates spawned by the classes of objections just alluded to, I propose in this paper to look at a different cluster of problems which I shall associate with the label of individuality. These problems have nothing essentially to do with our mentality or even our humanity. Yet I think they may at least in part underlie the worries expressed in the traditional objections, and account for their persuasive force.

Let me begin, then, with a bare announcement of what is to come. The suggestion I want to explore is this:

(a) Human beings are not, as such, members of any natural kind about which there might be discoverable laws of ``human nature,''

(b) But the reasons for thinking this do not preclude our having natures.

(c) Our natures are individual natures. Each of us is, quite literally, the only member of its kind.

I shall refer to the conjunction of (a)-(c) as the thesis of individuality. If it is true, the thesis of individuality has remarkable consequences for the project of human sciences. For it means that the usual inductive methods--the refinement of generalizations about human beings to the point where we promote them to the honorific category of laws--is incapable in principle of yielding scientific knowledge. (I shall later suggest that the attraction of many of the views listed above about what is so special about the human sciences might turn out to be corollaries of the thesis of individuality.) On the other hand, there is, in principle, scientific knowledge to be had about individual humans--though the obstacles to attaining it may remain in practice unassailable.

Why humans are not a natural kind

On a widely held view, scientific laws express properties of kinds of objects--chemical elements, physical particles, or perhaps classes of those such as gases, or fluids--which such objects necessarily share in virtue of their nature as objects of that kind. Those laws describe the causal powers of the members of the kind, in virtue of which events involving things of that kind are impossible, possible, or probable.

The argument to show that humans are not, as such, members of any natural kind starts with a biological truism: no two human beings--apart from identical twins--share a genotype, and even twins do not share all formative experiences. Consequently each human individual is unique.

Indeed, the uniqueness of individual human beings is a doctrine often proclaimed. But are we unique in any unique sense? Much the same might be said of non-human organisms, and for that matter of rivers, rocks and stones as well. Just so: the comparison is apt, for indeed there are no natural laws governing stones and rivers as such. Stones are governed by the laws of gravity, as well as by all kinds of other chemical and physical laws; but that has nothing to do with their peculiar nature as stones; and river waters are subject to the laws of hydrodynamics, but not qua rivers. And so on. Similarly, while human beings are subject to various laws of nature, there are no laws that apply to them as a kind ``human being.''5

Biologists and philosophers of biology have come to a consensus in recent years that from the logical point of view species are not classes or universals but individuals or sets.⁶ From the logical point of view, there seems to be little difference between the claim that species are individuals and the claim that they are sets. The crucial insight, on either view, is that species names are best construed not as general terms (re)applicable to any particular meeting certain necessary and sufficient conditions such as `featherless biped' or `rational animal,' but as singular terms, each designating a determinate--though discontinuous--region of space-time. Sets don't have definitions but only extensions. They are therefore logically individual in the relevant sense, even though, like classes or kinds, they have members rather than parts. The reason is that once we adopt the Darwinian, as opposed to the (so-called) Aristotelian conception of species, membership of a species is no longer determined by any essential or defining properties. (Nor is the issue a matter of the difference between nominal and real essences: for species, as sets of actual organisms linked by actual and potential reproductive relations, have neither real nor nominal essences.)

Now it may seem that as we think of it more carefully the contrast dissolves. We might view the classes of objects studied by physics as equally concrete sets of things in time and space--an actual mass of stuff such as Earth's body of water or the total supply of gold in the universe. From that perspective, it's only by courtesy that anything can be called a law of nature, they are merely those universal statements that are decreed to be true in all possible worlds.⁷ (But we don't get to go there and look.) Alternatively, we might simply give up the requirement that laws of nature be universal.⁸

But neither of these views will do. Insofar as our aim is scientific understanding, we want to understand why just this happens and not something else; and if it were to happen differently elsewhere to something of just this kind we would require an explanation. The causal power of a mere change of place-time might do the trick, but then that causal power would have to be expressed in terms of a universal law governing the effects of changes in space-time, or of different locations. Understanding is intrinsically general: if we offer an explanatory law, we mean it to apply to stuff that shares with the stuff at hand a crucial set of causal powers. For that kind of stuff is defined in terms of those causal powers (with the implicit proviso: ``so long as we have got them right.'') To increase one's scientific understanding is to grasp increasingly basic causal powers in virtue of which things behave as they do.

My claim is not that biological organisms don't behave as they do in virtue of certain causal powers; but rather that the causal powers in virtue of which they behave as they do are not ones they have in virtue of being members of a certain species, any more than a stone has the causal powers it has in virtue of being a pretty medium sized polished pink stone with dark vein-like markings. Some of these properties, to be sure, may be brought into certain explanations; but they won't have the privileged scientific form, which allows us to say--implicitly: `This is what being a thing of that kind involves.' In biology, by contrast, there are probably perfectly good laws, but they apply to the processes that lie behind biological stability and change. The best known example of such a law is the Hardy-Weinberg law, which is often compared to Newton's first law of motion because it tells us, in effect, what is to count as staying the same: if not everything in the distribution of genes remains the same, and the changes cannot be imputed merely to random fluctuations, the Hardy-Weinberg law licenses the inference that some ``force'' (selective pressure) has come to make it change. Other laws might govern the rate of mutation; or the boundary conditions that have to be satisfied if we are to get group selection, or speciation, and so forth. But those laws have nothing directly to say about the properties of those relatively arbitrary spatio-temporal slices of life we call `species.'

I am not here endorsing a view of science as aiming at the unveiling of objective essences of interest-free natural kinds.⁹ The point is rather that we must perforce concentrate on the interesting properties of the objects of our study; and the interesting properties for the purposes of biology are determined by our sequences of DNA. And that's precisely the respect in which (with the exception of the case of twins) the odds against two sexually reproducing organisms being relevantly similar are astronomical. The core of the Darwinian revolution, as Ernst Mayr pointed out a number of years ago, is that variability is the fundamental biological fact--the default assumption. In Aristotelian biology constant types are the norm, barring developmental accidents, and deviations from them require special explanation. For Darwin, on the contrary, what biological science needs to explain is precisely the relative and temporary uniformities between organisms that characterize species. We can no longer set out to discover the universal causes and effects of the characteristics of a species, taken as defining a kind of universal.

Individuals as Natural Kinds

Given the nature of our understanding of biology, then, species must be construed as logically individual. Curiously, however, there is nothing to prevent an individual's properties from being definitive of a species in the Aristotelian sense--i.e. in the sense of a genuine natural kind, of which there just happens to be exactly one member. The individual would instantiate such a single-membered kind, if there is a truth of the matter about the causal powers in virtue of which certain events involving that individual are possible, probable, or impossible.

The idea of individual natures seems to be intuitively attractive. For consider how we already talk about them: `he realized the job didn't suit him'; `no-one could see what they saw in each other'; `a person of her temperament was bound to react that way'; `he is not himself today'; `his personality has changed over the years'. Such phrases presuppose that in speaking of an individual there are truths to be told that don't apply to just anyone, let alone to all humans as such, but are essentially relative to just that individual. Even existentialists, who don't believe in human essence, prize a certain concept of authenticity: something I do is authentic, even in the context of a purely undetermined existential choice, if it constitutes a genuine expression of my individual nature.¹⁰

A thought experiment will help to fix this notion of individual natures.

Imagine something like Nietzsche's eternal recurrence. Suppose time is infinite but space and matter are finite. If the number of the possible combinations of elementary particles are also finite--which doesn't, as Nietzsche may have mistakenly assumed, follow from the finiteness of their number--then every one of those possible combinations,¹¹ including you and me, must recur an infinite number of times. Each of us, in that case, would be but an instance of the universal which is also instantiated by every occurrence of that specific combination in space-time.

Suppose, then, that we were to take this idea of individual natures seriously. Could we have scientific access to it?

Although no two individual humans have all their biologically interesting properties in common, we can safely assume that every group of humans (or for that matter of primates and even of mammals) will present a great number of interesting correlations. Hence the use of statistics. But what exactly is it that statistics reveal? What specifically can it tell us about individual members of the groups it studies?

According to Ian Hacking, statistics were invented, in an "avalanche of numbers," in the first half of the nineteenth century, precisely in connection with the human sciences and their gradual emergence into autonomy.¹² But if statistics represent knowledge that is irreducibly about groups, can it be used in some way to retrieve information about individuals?

At first sight, two approaches seem possible. Call the first the vectorial approach. It will consist in viewing the individual as the intersection of as many groups as possible about which there is statistical information. One will then have to combine the information about each of these groups into a vectorial sum that applies to their intersection. Or one can narrow down as much as possible the class to which the individual in question belongs, and then look for statistical evidence directly bearing on that narrowest class. Call this the set-theoretic method.¹³ In fact, however, neither method is capable of telling us anything specifically about an individual.

The vectorial method may seem to promise estimates of risk so accurate that X's life expectancy can be pinned down within a year or two with virtual certainty. It was recently reported, for example, that several genes have been found to cause predispositions to non-genetic diseases. And much is coming to light about the effect of different environmental factors on one's chances of disease. One can easily imagine getting to the point where one might predict of a given individual whether or not she will be one of the people whom the plague will strike, or one of those that will remain unaffected.

There is an obstacle, however, which is the non-monotonic bearing of different items of statistical information. Suppose, for example, that an actuary is attempting to set a fair premium for life insurance in the case of some individual X. The actuary will typically collect statistical information bearing on the individual in question. Each new piece of information may affect the actuary's estimate. But the changes in the bearing of the information are ``non-monotonic'': the next thing I learn may reverse a trend.

Suppose, for example, that I am writing life insurance on client X. On learning that X is an American, I know that X's life expectancy (L) is 76, say. Add that X is female: L increases. Add that she is Black: L decreases. Add that she has a Ph.D: L goes up again. Add that she is a smoker: L goes down. Add that her parents were both alive at 95: L goes up. And so on and on. In this way the bearing of additional information may flip-flop, but in general more information is better. The smallest class, that is, the class of American female Black PhD smokers born of long-lived parents, is the maximally relevant class.

But there is a twist: the various statistical grounds do not add up as if they were simple causal factors. In the above example, whether being female is a causal factor in longevity is a debated matter. Being female is at least tightly enough correlated with longevity to be treated, for statistical purposes, as if it were a genuine causal factor. But no one would claim that blackness as such is a causal factor in itself. Part of the reason for this is that some laws are associative, while others are really causal. The causal laws are those in accord with which the probability of an effect always increases, other things being equal, if we can bring about the causal condition. In merely associative cases, A may be quite genuinely correlated with B, without it being the case that bringing about A will have the slightest effect on the probability of B. (That fact is at the heart of Newcomb's problem.) Nancy Cartwright quotes a letter from an insurance company:

It simply wouldn't be true to say,

It may not be possible, in short, to tell for certain which correlations represent genuinely causal factors, and which are merely associative. And in any case the ceteris paribus clause often means that we can't predict from the bearing of some information ceteris paribus what bearing the information will have in context. In general, for example, being female raises life expectancy. But in some cultures--where there is widespread female infanticide and bride murder, for example--it may lower it. The only way to find out is to have independent information about how the various factors work together.

In that case, one will resort to the second method, basing all statistical inferences separately on the data available for that group (e.g. in this case American female Black PhD smokers born of long-lived parents....).

The Actuarial Paradox

The set-theoretic method, however, gives rise to a different problem, which I shall refer to as the actuarial paradox. The actuarial paradox consists roughly in this. In order to arrive at the best estimate of fair premiums, an actuary must obtain the greatest possible amount of statistically relevant information on the person to be insured. But an actuary who fully succeeds defeats the whole purpose of insurance, both epistemically and morally. In this paper I concentrate on the epistemological problem, pausing only briefly, in a moment, to raise the moral problem of fairness.

To clarify this, let us define a notion of absolute risk R. Intuitively, R is a measure of uncertainty for a given situation. It measures not whether you are likely to die, say, but how chancy it would be to make any prediction about it one way or the other.¹⁵ R is 0 when p is either certainly true or certainly false; it reaches its maximum when p and ~p are equiprobable. Now suppose we are interested in the risk of death at age 69 for a particular individual. As we acquire more information, absolute risk diminishes--but only up to a point. For as the target class grows narrower, the size of available samples will shrink. But the smaller the sample, the larger the margin of error. Finally, R will reach a non-zero minimum before rising again, as the margin of error grows larger and any available evidence becomes virtually useless. There will be no interesting statistics about American female Black PhD smokers born of long-lived parents with factor X in her genes, if there is only one of them.

This reasoning supposes, of course, that we are not dealing with a natural kind containing an indefinite number of instances. For in that case there will be no principled limit to the size of possible samples. But if each of us is, as I have argued, a single-membered class, then the paradox has all its virulence.

Here someone might offer the following objection. Suppose you are trying to make the best possible decision in relation to an individual X. Assume that whether X has character C is crucial to the decision, and suppose that X is member of a group G of which 90% have C. Suppose further that you have no other source of information about X. In that case, if you act as if you knew that X had C, the probability of your taking the right decision is .9; if you act on the opposite assumption you will simply guarantee that your chances of making the right decision are .1 instead of .9. Clearly, then, it must be rational to treat individuals on the basis of statistical information about the groups they belong to.

There is something right about this argument. When acting in uncertainty, one must use probability to maximize expected utility. But there are two issues that practical dilemmas force us to ignore. One is a moral question: whether it is fair to treat X in this way. The other is an epistemic point: whether such a method can ever yield knowledge about the individual in question.

Expediency doesn't settle the issue of fairness. To make this obvious, one has only to suppose that there is something punitive about the treatment based on the presence of C. Suppose, for example, that the characteristic in question, for a certain group, is being guilty of a certain crime, and that it can be solidly established that 90% of the group in question are guilty of that crime. Then it is true that if you punish everyone the number of people not getting what they deserve is smaller than if you punish no one. Still, that doesn't change the fact that some people, on either supposition, are not getting what they deserve. Fairness requires that any treatment meted out to X on the basis of S be based on X's actually possessing the trait.

Much the same goes for the epistemic question. When we get the statistic about the presence of C among members of G, this doesn't actually tell us about its distribution. The trait C might be one that an individual may possess to a greater or lesser degree. At once extreme, every member of G might have C to degree .9. At the opposite extreme, exactly 90% of G might have C to degree 1. Or there might be any of an indefinite number of possibilities in between. Statistics, almost by definition, are powerless to give us knowledge of the individual.

Now an objector might argue as follows:¹⁶ "why can't scientists who study humans find a relevant property of a subclass of humans, and make some useful statistical generalizations accordingly, for example, that only 30% of adult male smokers who take a certain anti-smoking program continue to smoke?"

They can indeed. And as far as the probability of a prediction's being right if based on this "law", one can do no better. But this has no bearing whatever on whether one's prediction actually concerns the individual member of the class. It would do so only if the statistical law in question were true in virtue of the possession, by each member of the class, of an equal propensity to the behaviour or characteristic in question. A generalization might be useful for the purposes of prediction over group in question. But given that human populations are not homogeneous, no specific propensity, and therefore no actual categorical property, can be attributed to any of the individuals in question on the basis of the statistical generalization.

In practice, of course, the actuarial paradox never threatens real actuaries, for both theoretical and political reasons.

The theoretical reason is that even if there were relevant causal laws we could never tell for sure whether the associative generalizations we have uncovered represent such laws. The intricacy of interacting factors is just too large. Many correlations, such as those between driving safety and age or sex, are treated on a par with causal factors. It makes no practical difference to the purely economic purpose of fixing premiums whether this is right or wrong.

Still, it might matter a good deal from the point of view of fairness. That is the political point. Insurance companies tread a fine line between actuarially justified and politically unacceptable discrimination. It long seemed acceptable, for example, to penalize women, because they live longer, by charging them higher pension premiums, while for the same reason charging them lower life insurance premiums. (This practice has now been made illegal in many jurisdictions.) On the other hand, it has never been the practice of insurance companies to give Blacks a bonus on pension premiums, on the ground that their life expectancy is lower than that of whites. Debate rages on, at the moment, about whether it is or ought to be legal to ``discriminate against'' young males or carriers of the AIDS virus. In a free market economy, insurance companies are bound to seek out low-risk prospects and lure them in with low premiums, so that for high-risk groups insurance will indeed become pointlessly unaffordable. But in a regulated economy it will probably remain politically unacceptable to push that process to the limit.

The epistemic moral remains: if we approach the task of predicting the course of events for any individual's life on the basis of statistical analysis, there will always remain a core of unpredictability, governed by causes inaccessible to the usual correlational methods.

Constructing Individual Natures

What then are we left with if we want to take seriously the idea that human behavior has causes?

In effect, we shall have to treat human beings like the stones and the rivers mentioned earlier: we shall have to look for causal factors which affect them in virtue of their nature as objects with a certain material constitution, but not in virtue of their status as human beings. The material constitution of a person can give us access to the relevant course of events, but only insofar as there is no unpredictable residue. What we would need, in other words, is reduction without emergence. But reduction without emergence is a chimaera.

Let me explain. I define property emergence as follows:

The properties peculiar to humans are very likely to be among those. And this has interesting implications. Since you can't find out about emergent properties without establishing empirically their correlation with their underlying causal conditions, the only way you can find out what effects this or that configuration will have is to experiment with it. But you could only do that in the orthodox manner if you could run statistical experiments; and for the reasons I have discussed this is not a realistic prospect. That a certain drug has this or that effect on this or that sample of subjects, for example, can give us plenty of good reasons to adopt this or that policy in respect of that drug, but it is powerless to reveal what precise effect the drug has on any particular individual. It cannot, therefore, yield any genuine causal law. For a causal law would have to link that particular chemical either to effects in humans in general, or to effects in some particular person. The first it cannot do since there is no such law-governed natural kind as humans. The second it cannot do because of the difficulties I have explained in terms of the actuarial paradox: we can never be sure enough that the property in question applies to this particular human, who strictly speaking is the only thing of its kind. In the first case, there is no scientific property to be discovered (where a property is scientific, roughly, only if it applies to some kind of entity members of which can be picked out in terms of necessary and sufficient conditions making no essential reference to particulars.) In the second, there are such scientific properties, since, on my hypothesis, the individual has a nature;¹⁷ but those properties are in principle inaccessible.

The Humanistic Objections again

I suggested at the beginning that some of the traditional objections to the application of science to human beings might in part be related to the individuality of human natures. Let me close with a few hunches about the nature of those connections.

The fact of individual natures makes trouble for three sorts of judgments about human beings: (a) judgments about what someone can do; (b) judgments about what someone will do; and (c) judgments about the appropriateness of someone's reactions. These, I believe, bear respectively on objections (i)-(iii) mentioned above to the program of human science.

(a) In scientific domains, there are two ways of supporting a claim that something can happen. The simplest is to establish by observation that it does (sometimes) happen under the relevant circumstances. The second is to deduce from laws of nature and the specification of some relevant initial conditions that it would happen under those conditions. But neither of these methods is available if we want to find out whether someone can do something. For the fact that I have never done something doesn't establish that I can't do it, and obviously we are not in possession of facts from which we could deduce, say, that Ben Johnson can beat the 100 metre record before he actually does so. Yet in an informal way we constantly make judgments about what people can and can't do. Such judgments are essentially speculations about those individuals' natures. But I suspect that the inevitable gap between such judgments and any facts about what people actually do partly encourages the notion--whether or not it is coherent--that people are blessed with free-will.

(b) If judgments about what people will do could be based on a definite knowledge of the underlying determinants of their behaviour, predicting what a person will do would not be such a chancy business. But part of the reason we can't do that is that those determinants are specific to each individual's kind: even if we know the micro-structure of each brain, we could not--given the emergent character of its consequences--predict those consequences in advance of observing them. What we observe of mentality, Davidson claims, is organized on entirely different principles than the physical. For mentality is rationale-giving, emotion expressing, decision making, perception interpreting. The properties of that category are not recoverable from their underlying causes, nor are their causes deducible from them. My hunch here can be expressed in terms of a purely utopian thought experiment: if we were, per impossibile, to acquire full knowledge of an individual's nature, the anomalousness of the mental would no longer be an obstacle to full prediction. For we should then know just what token identities held, in the case of that particular person, between particular physical states and particular mental states: the token identities posited by Davidson¹⁸ would have become type identities. We might find, in other words, that if a single individual is treated as a natural kind, type identities will underlie the token identities posited by the thesis of anomalous mentality.

(c) If it is indeed true that individual mental predicates have socially distributed truth conditions, then the individuality of each person is partly created by, and vested in the community.¹⁹ And indeed, the "uniqueness" of each individual is something that depends essentially on the social matrix in which it comes into being. As a result, in labeling and assessing a particular person's reaction to some event, we use mainly general precepts having to do with social conventions, norms of acceptable behaviour, expectations of civility in emotional expression, and so forth. Whether we classify some reaction as anger, for example, depends in part on whether its target can reasonably be claimed to have acted in a willfully harmful way. And whether we classify it as excessive anger depends on our assumptions about the grievousness of the harm suffered and the expectations we have in comparable situations. But those general expectations and norms do duty for accurate knowledge we don't have about the norms of that particular individual's nature.²⁰ If there were a species nature, we might expect the repertoire of mental predicates available to each individual to be just what it is in virtue of that individual's membership of humankind. As things are, however, individuals are chaotic effects of environmental forces, including social ones, on the causal powers inherent in the unique genotype of each biological organism. We must resort to social reality in order to define individuals, because there is no species nature from which a set of essential properties might be safely derived.

Concluding Fantasy

And what if there were Nietzschean eternal recurrence? We might then imagine a sort of super-universal observer, a godlike Practitioner of the Human Sciences Transcending Universes (PHUSTU). For PHUSTU, the possibility of an orthodox science of individual natures would exist, as it cannot for us, because there would be an infinite supply of individual members of the kind, say, `Sparshott'--a kind to which in this universe only one particular belongs. So all it would have to do is discover the causal laws in accordance with which Sparshotts behave.

But PHUSTU would now face another problem. Eternal recurrence is based on the idea--ironically, an Aristotelian idea--that given infinite time all possibilities must be realized. So PHUSTU's class of potential samples of Sparshotts will contain not only genuine Sparshotts, but also all possible kinds differing arbitrarily little from Sparshotts. PHUSTU will then be rather in the same position as the librarians of Borges's Great Library of Babel: first they were inordinately elated at the thought that since the great library contains every possible book, it must somewhere contain its own true catalogue, the answer to every possible question, and the true story of each of our lives. But soon they were equally depressed, on realizing that for the same reason the library must also contain all possible false catalogues, all possible false answers, and an indefinite number of possible descriptions of lives almost but not quite like our own. In the end, then, even PHUSTU would not be in a position to study scientifically--that is, by investigating the properties of representative samples true to their kind--any of our separate, ultimately unknowable individual natures.
 
 

NOTES:

<sup>1.  For one influential argument to this effect (among many others) see Charles Taylor, ``Interpretation and the Sciences of Man,'' in Human Agency and Language., vol II:15-57 (Cambridge University Press, 1985).

2..  Versions of this objection are found in Hampshire, Freedom of the Individual, (Harper and Row, 1965) and perhaps in Sartre.

3.  Donald Davidson, Essays on Actions and Events (Oxford: Clarendon Press, 1980). See also J. Fodor, "Special Sciences," in Representations (Cambridge: MIT-Bradford, 1981)

4.  See Tyler Burge, ``Individualism and the Mental'', Midwest Studies in Philosophy, 4:73-121 (1979) ; also (for a similar anti-individualist thesis about the emotions) Robert Kraut. ``Love de Re,'' in Midwest Studies in Philosophy, 10:413-430.

5.  Another characteristic that is sometimes cited to explain the difficulty of finding laws that describe human behaviour is their complexity. But as far as I can see complexity is not a priori an obstacle to being subject to natural law. In some cases, indeed, complexity is a necessary condition of the applicability of natural law. This can be illustrated by any of the laws, such as Boyle's laws, or the laws of thermodynamics, that are in effect stable statistical consequences of large numbers of elementary events.

6.  See the papers in Part VII of Elliott Sober, ed., Conceptual Issues in Evolutionary Biology (Cambridge: Bradford-MIT, 1984); also the contributions by Ghiselin, Mayr, Hull, Rosenberg and (especially) Kitcher in Biology and Philosophy, vol. 2 no. 2 (April 1987). The thesis that there can be no laws about humans as such, because species are individuals, is argued in some detail by Alexander Rosenberg in "Human Science and Biological Science: Defects and Prospects," in Scientific Explanation and Understanding: Essays on Reasoning and Rationality in Science, ed. Nicholas Rescher, (Lanham, MD: University Press of America, 1983). See also Alexander Rosenberg, Sociobiology and the Preemption of Social Science, Oxford: Blackwell's, and Baltimore: Johns Hopkins (1980) My own contribution to this debate is in "Kinds of Kinds: Individuality and Biological Species." International Studies in the Philosophy of Science 3.2 (The Dubrovnik Papers: Philosophy of Biology):119-135.

7.  This seems to be roughly the strategy followed by Kripke and Putnam. See Saul Kripke, Naming and Necessity. (Cambridge, MA: Harvard Univ. Press, 1980,) Hilary Putnam, "The Meaning of Meaning." In Mind, Language, and Reality: Philosophical Papers. (Cambridge: Cambridge Univ. Press, 1975).

8.  This is the stance rather casually adopted in Ronald Giere's Understanding Scientific Reasoning (Holt, Rinehart and Winston, 1984.)

9.   I've gone on record elsewhere as arguing that there are no such kinds. See ``The Natural Shiftiness of Natural Kinds,'' Canadian Journal of Philosophy 14:561-580 (1984).

10.  It is a separate question, about which I shall have nothing to say here, to what extent our individual nature is necessarily constant and to what extent we can have a succession of such natures. On a Parfitean view of the self, only a relatively short temporal slice of a person could strictly have a nature. On more traditional views, our nature might be precisely that which once formed could not be changed, though perhaps it could be perverted by traumatic experience.

11.  How to interpret possible combination? Strictly speaking, given infinite time only that is guaranteed to happen which has non-zero probability; but in an infinite domain that is not equivalent to possibility. Pick a point on a line: the measure of the probability of picking a particular point is zero, yet picking it is not impossible. Given that we have already posited a finite number of possible combinations, however, possibility and non-zero probability can be considered equivalent, for Nietzsche's intents and purposes.

12.  See Ian Hacking, "The Autonomy of Statistical Law," in Scientific Explanation and Understanding: Essays on Reasoning and Rationality in Science, ed. Nicholas Rescher, op. cit.

13. The set-theoretic method is essentially the traditional requirement of total evidence for statistical reasoning. See Carl G. Hempel, Aspects of Scientific Explanation (New York: Free Press, 1965) pp. 397 ff.

14.  Nancy Cartwright, ``Causal Laws and Effective Strategies,'' in Why the Laws of Physics Lie (New York: Oxford University Press, 1983); see also Ronald Giere, op. cit. on establishing causal connections on the basis of comparisons of hypothetical populations.

15.  We might express absolute risk R with respect to any proposition p with a number normalized between 0 and 1: R=4 prob(p) * prob(~p).

16.  As did an anonymous referee for an earlier version of this paper.

17.  Could individual natures be discovered by performing repeated experiments on a single individual? A more generally shared characteristic of biological organisms makes this difficult: any repeated experiment is likely to change the properties of the organism so tested. Still, some properties having essentially to do with that very fact might be explored in this way: an individual's characteristic learning or habituation curves, for example.

18.  Donald Davidson, "Mental Events," in Essays on Actions and Events. Oxford: Oxford University Press (1980).

19.  That fact gives rise to the tragedy of solidarity, which I have discussed in "Emotions and the Conduct of Life," chap. 12 of The Rationality of Emotion (Cambridge, MA: MIT-Bradford, 1987)

20.  Rough and ready approaches to such knowledge are exemplified by psychoanalysis, in which the chief source of knowledge is supposed to be the clinical experience of particular cases. Such clinical experience is, of course, open to easy objections: solid knowledge about individual natures is, for the reasons I have given, impossible. But if we take the notion of individual natures seriously, we'll think somewhat differently about the issue of clinical evidence. The way it's usually discussed (see, e.g., Adolph Grünbaum, The Foundations of Psychoanalysis: A Philosophical Critique, and discussions by 38 commentators in ``open peer commentary'' in Behavioral and Brain Sciences 9:228-284, 1986) the question concerns whether it's any good to get evidence from the clinical situation. But one could profitably turn the question about and ask, is there anything in the individual case (not just in psychoanalysis but in medecine generally) that might at best show up in the clinical situation but is inevitably inaccessible otherwise? If there are individual natures, clinical experience might be one partial and unsatisfactory method of access to properties otherwise quite unknowable.</sup>