proach to understanding the properties of persons (their traits, desires, abilities, interests) which is not only very popular and historically important, but also intuitively plausible. It begins with a division of human properties into three categories. Natural properties are those persons have in virtue of being members of a natural kind, and they originate in the structures definitive of the species. Other properties are unnatural, in that they result from abnormal structures. And some properties are nonnatural (or social) in that they represent replacements, modifications, or extensions brought about by the social environment operating on the basic structures.1 Such is the ontology. It suggests immediately the epistemology for assigning observed properties to the three categories, in particular to the natural and the nonnatural. The central epistemological thesis is a counterfactual: natural properties are those that persons would exhibit were they never influenced by a social environment. John Stuart Mill, in his The Subjection of Women, asserts this view: “the artificial state superinduced by society disguises the natural tendencies of the thing which is the subject of observation. . . .” Suppose “all artificial causes of difference to be withdrawn,” the “natural character would be revealed.”2
The central epistemological thesis implies two methodological rules and a corollary for discovering which properties are natural and which are nonnatural. First, the natural properties are those which are common among persons who live in different social environments. Properties which are observed in all types of social environment are just those properties which are most resistent to social influences and which would be observed regardless of social influences. Similarly, if observed properties vary as social environments vary, this is evidence for their being nonnatural. (Mill does, I think, assert the first rule in the Logic.)3 Second, the properties of persons who live in environments containing relatively few social influences are more likely to be natural than those of persons who live in environments containing many. But most important for Mill is the corollary, which says that differences in natural properties between two sub-groups of the species are those differences which are observed even though people in both sub-groups are not only influenced by the same social environment but are also subjected to the least possible amount of social influence.4 The corollary is expressed in this passage from The Subjection of Women concerning “literary women”: “their sentiments are compounded of a small element of individual observation and consciousness, and a very large one of acquired associations. This . . . will remain true . . . as long as social institutions do not admit the same free development of originality in women which is possible to men. When that time comes . . . we shall see . . . as much as it is necessary to know of the nature of women.”5
Note that Mill speaks of the “free” development of women. In the history of philosophy a social doctrine has usually been attached to those ontological and epistemological theses, the doctrine that persons are free (or have freedom) to the extent that they are not constrained, either in their behavior or in their development, by social influences. The liberal doctrine of “equal maximal liberty,” then, is both a methodological rule for the discovery of the natural as well as a political principle limiting one sort of social influence. Mill elaborates and defends the political doctrine in On Liberty and employs it in The Subjection of Women, which also contains the epistemology. Another statement of the corollary from the latter work more clearly connects the discovery of the natural with freedom: “no one can safely pronounce that if women’s nature were left to choose its direction as freely as men’s, and if no artificial bent were attempted to be given to it except that required by the conditions of human society, and given to both sexes alike, there would be any material difference . . . in the character and capacities which would unfold themselves.”6 We can conclude, then, that observed differences (in ability, say) between the sexes can be identified as natural only when (1) both sexes undergo the same socialization and (2) women and men have equal and maximal freedom.7 It is testimony to the plausibility of this methodological rule that it has been appealed to not only by Freud8 but also by many contemporary feminists.9 I propose to provide a critique. The main point is that the application of the corollary presupposes that we already have what the rule is supposed to yield: knowledge of the natural.

The first point to be made is that the application of the corollary presents a moral problem. The corollary suggests that we attempt to discover whether men and women are naturally different by treating them in the same way and by allowing them equal maximal liberty. But it is reasonable to suppose that whether women and men (especially as children) ought to be treated identically or differently depends on whether their natural properties are the same or different. If in fact women and men differ naturally, they ought not to be treated identically, for doing so will have harmful effects on the people so treated and on their society. Thus for moral reasons we may be prevented from using the corollary to discover what it was designed to discover. Now, it is true that the corollary is not to be applied in an epistemic vacuum, but this fact confirms my point. Through the application of other methodological rules (primarily the first and its variants) there is already some evidence that women and men differ so little that treating them identically would be unlikely to produce harmful effects (or at least so little harm that it would be outweighed by the value of knowing what the natural differences between the sexes are). The moral permission to treat women and men in the same way to discover their natural differences can be granted only to the extent that we already know enough about their respective natures from the application of other rules.

The second criticism is derived from Mill’s saying that in applying the corollary, no influences ought to be brought to bear on either women or men except those “required by the conditions of human society.”10 He means that in allowing men and women equal maximal freedom from social influences, “maximal” does not stand for “absolutely no” influence, for some influences are required to maintain society, to prevent such a free appearance of unmodified human nature that civilized life is threatened. But in order to place this limit on the extent of equal maximal liberty, one must assume prior to the application of the corollary that a good deal about human nature is already known; it must be assumed that we already know that there are aspects of the nature of both women and men some of which are compatible with, and others of which are incompatible with, civilized society. In particular, the application of the corollary presupposes that with respect to potentially socially disruptive natural properties, there are no significant differences between the sexes. To say that the same minimally-necessary social restraints should be applied to women and men is to assume that the natures of women and men are sufficiently similar to require only the same socializing constraints. This, however, is one of the things that the application of the corollary is supposed to tell us is true or false.11
The third criticism is that the kind of indeterminacy generated by the application of the corollary can be resolved, if at all, only by the application of other methodological rules. Suppose that we believe that women and men have been raised and treated identically, and that they have had equal opportunity to achieve desirable positions in society. Yet we still observe that more men than women enter the professions. The persistence of observed differences could mean not that there are natural sexual differences in ability, but only that (1) women and men have not in fact been raised or educated in exactly the same way (or one sufficiently similar), and/or (2) they did not in fact have equality of opportunity (for example, some discriminatory practices–even morally innocent ones–have not been identified). Conclusions reached on the basis of the corollary, then, are only as dependable as the methods used for deciding that women and men have been raised identically and have had equal opportunity. In the face of this indeterminacy, we can use the information supplied by the corollary as evidence of natural differences only if there is evidence supplied by other rules pointing in the same direction. At best the corollary provides evidence only if the information it yields is consistent with what we already know, that is, if it supports or is supported by other information with which it coheres.12
The fourth point is that any conclusions reached by the application of the corollary have to be severely qualified. It is highly debatable what social arrangements count as establishing equal opportunity. There are narrow, moderate, and broad notions of equal opportunity, and to decide which notion is the “right” one is to take a stand on some very difficult normative (evaluative) issues.13 The normative nature of the concept of equal opportunity will necessarily infect the conclusions arrived at by using the corollary. Of any observed property difference that persists in the face of equal opportunity, we could say only that it is a natural property given this particular conception of equal opportunity (but that given a different conception, it might not be natural). This qualification considerably undermines the value of the corollary, for it implies that the natural-social distinction is not entirely empirical. Furthermore, to select one conception of equal opportunity as the right one requires that we make some assumptions about human nature.14 If so, the application of the corollary presupposes at least some of what the application of the corollary is supposed to discover.

There is perhaps a final point to be made. We seem to be able to specify the conditions of equal maximal freedom without knowing the social and natural sources of observed properties. We seem, that is, to be able to specify independently what the social influences on people are, the influences that need to be minimized if we wish to allow natural properties to express themselves. But I am not sure that we can do so. Ordinarily what we mean by minimizing social influences is making sure that children have a lot of freedom in which to grow, take chances, and explore the world and themselves. But choosing the exact things we do or refrain from doing, I think, presupposes that we already have some idea which sources of behavior are social and which are natural. Freud, for example, suggested that to discover whether there are any natural differences in intellectual ability between the sexes, both religious teaching and interferences with sexuality would have to be eliminated.15 But what criterion is being employed in specifying “interference with sexuality” as social influence that can be minimized? Regardless of what particular criterion is employed, it seems already to presuppose that we can identify and distinguish social sources from natural sources of observed properties. Maybe what we do to children is natural (social) even though we conceive of it to be social (natural).

The paradox of this ontological and epistemological approach is that the ontology both suggests and undermines the epistemology. Saying that natural properties are those exhibited in the absence of social influences suggests that we can discover the natural by equalizing and minimizing social influences. But as soon as it is also said that the social influences can work over, transform, and even eliminate the natural, it becomes precarious epistemologically to disentangle, with respect to observed properties, which ones are natural and which are nonnatural.16 In this regard the first methodological rule fares no better. I want to say a few words about this rule because the application of the corollary depends on information it supplies. I shall argue that the application of the first rule in effect presupposes what it was designed to discover. This means that theories of human nature (and therefore theories of natural sexual differences) are either largely a priori or merely express moral commitments (perhaps both).

According to the first rule, natural properties are those which are exhibited by people regardless of the sort of social influences or environment they have been exposed to. But the requirement of universality is too strong. (Why is this?) To justify the inference from near-universality to naturalness, it is necessary either to admit that the criteria for the identification of the causal elements of social environments are not as precise as the methodological rule requires, or to assume that the exceptions to universality are analyzable as cases of the unnatural or as natural sub-groups within the species. If we take the first route, a more precise categorization of the elements of social environments is required to allow the discovery of social causes previously unseen through imprecision, while if we take the second route we must look for structural abnormalities or differences that correlate with the exceptions. A decision to follow either the first route or the second is fraught with the typical dangers of research based on a Weltanschauung (for example, finding gender everywhere) but could, on the other hand, represent the elaboration of a potentially fruitful heuristic. For example, the failure to find structural abnormalities does not entail that none exist, and we already know that slight structural deviations abound even in unexceptional cases. On the other hand, a decision to follow the second route could very well yield significant structural discoveries. Similarly, a decision to follow the first route could yield extravagantly detailed intrasocial typologies (a proliferation of social classes, say) or it could result in a new and powerful classification that explains the exceptions quite neatly. But choosing between these two routes is already to decide, in effect, that certain phenomena are due either to natural or to social causes. To decide that the best place to look for explanations of the exceptions is in one place (or the other), is to decide, prior to further applications of the rule and in defense of previous applications, that certain properties are either natural or nonnatural. Thus the first methodological rule presupposes rather than yields that which we want to discover by its application.

The generalizations about human nature derived from the application of the first methodological rule, then, can have no other status within social science than that of the synthetic a priori or the pragmatic a priori. The generalizations stand as truths of human nature only if the exceptions can be explained. They serve as a protected set of propositions, in that subsequent research is dedicated to preserving their truth (this is a sure mark of the pragmatic a priori). If the ability of the corollary to provide further or complementary knowledge about human nature depends on the groundwork laid by the first methodological rule, then the entire approach builds upon, rather than provides a sufficient condition for establishing, an account of human nature.

There is an alterative view of the status of theories of human nature and of natural sexual differences. On this view a theory of human nature is held not as (or not only as) a guide to research, but as (or also as) a moral commitment. It might follow that social science is, after all, essentially normative. Whether that should worry us is, of course, an important and difficult question.


Notes
1. Mill writes that social influences operate “in two different ways: by extinguishing the nature, or by transforming it. In the one case there is but a starved residuum of nature remaining to be studied; in the other case there is much, but it may have expanded in any direction rather than that in which it would spontaneously grow” (The Subjection of Women London: Longmans, Green, Reader, and Dyer, 1869, 125).

2. Ibid.

3. Mill rather explicitly rejects the first methodological rule (for the social sciences) in book VI, chapter vii, section 2 of the Logic. But I would argue that Mill’s account of his “Deductive Method” in book VI, chapter ix, section 3, commits him to the first rule even in the social sciences. How else could he have formulated the “psychological law” that “a greater gain is preferred to a smaller”? In book VI, chapter v, section 1, Mill claims that arriving at a posteriori generalizations about human nature is insufficient; these generalizations must be supported by causal laws which explain them. But this claim allows that the first rule plays at least a preliminary role.

4. The demonstration that observed differences are natural is incomplete unless some foundation in structural differences is found for them. Therefore, another necessary condition must be added to this formulation of the corollary: the observed differences must correlate with systematic structural differences. But even this is not enough (see fn. 3). Ideally, the infrastructural mechanism, whereby structural differences become manifest as observed behavioral differences, must also be described.

5. Subjection, 47.

6. Ibid., 105. Partial statements of the corollary can also be found in Subjection, 23-4, 38-9, 48-9. Harriet Taylor also asserts the corollary in her essay “Enfranchisement of Women,” Westminster Review 55 (1851), 295-6.

7. In Subjection, Mill does assert that, strictly-speaking, it cannot be known that any of the observed differences between men and women are natural unless women have equal maximal freedom with men and are treated the same, but he also clearly believes that women are not naturally different from men, at least in intellectual ability and related (important) skills. His argument is that the observed differences between