§1. The One and the Many. Since natures are to be part of the required ontology, there are some questions about them that call for an immediate answer. Chief among these is the question of the relation of entities to their natures. If the nature of an entity is numerically distinct from that entity, then how can it be the basis for what the entity must be? Moreover, since terms for natural kinds appear in universal necessity claims, it is important to understand the relation of natures to natural kinds. Do all individuals in a natural kind have similar natures, and if natures are not properties, what can it mean to say two entities have similar natures?
The question about the distinctness of entities from their natures is best approached with some general machinery in hand. As far as sameness is concerned, ontologies divide into two basic kinds. According to those of the one, individuals are simples. According to those of the other, individuals have components. For an “ontology of simples” a potentially divided individual cannot be a genuine individual. It is a complex of many individuals. If it were one, it would, in contradiction to an ontology of simples, have many components. For the parts which could be separated from it are components of it. In effect, potential dividedness, for an ontology of simples, is always only an appearance overlaid on actual dividedness.1 Moreover, for an ontology of simples, a property of an individual cannot be a component of it, for then the individual would not be simple. In the ontology of simples there are only two possibilities for a property.
First, the property might be something that is not different from the individual having it. Thus, Hume held that a cube of marble is “neither distinguishable nor different nor separable” from the shape of the cube.2 For such an extreme nominalism, the basic individuals, whether they are things, impressions, or events, are alone in the world. There are no entities that are their properties or parts.
Second, the property might be a totally distinct entity from the individual having it. Thus, for example, the quality in one of Russell's monadic atomic facts is distinct from the simple that is the particular in that fact.3 For such a Platonizing ontology of simples, the distinctness of particulars is precisely like that between particulars and universals.
In short, an ontology of simples rejects sameness without identity. By “identity” I mean not being discernibly different. Identical entities are such that if one has a property the other has it, too. But by “sameness” I mean not being numerically distinct. For an ontology of simples, no distinction is drawn between these two notions; for such an ontology, if a property and an individual are the same particular, they are identical. If the property were different from the individual while remaining the same particular –idem subjecto–then the individual would have an entity as a component, and would thus not be simple.
When a property is the same particular as an individual, one of them depends for its existence on the other. The very idea of a complex entity brings in the idea of existential dependence, which will be discussed in Chapter XI, §4. As will be pointed out there, existential dependence is only a necessary condition for sameness; some entities are existentially dependent on individuals while distinct from them. Here it is to be pointed out only that existential dependence is not dependence in the usual causal sense. Jones may bring it about that a book comes into being, but though he is the cause of its coming into existence, he is not the cause of the existence that it has. Similarly, Jones may bring a circular cookie into being, but he does not thereby cause the property circularity itself. If individuals are not dependent as regards their existence and if properties are the same particulars as their respective individuals, properties will exist only in the sense that individuals exist with them and hence will be existentially dependent on individuals. The conflation of sameness with identity by the ontologist of simples is a first step toward denying that properties are existentially dependent on individuals. For this conflation leads us to say that properties are distinct from individuals since they are different from individuals. And only distinct entities can be existentially independent. In this chapter and in Chapter VII, I shall state my misgivings about the thesis that properties, natures, and relations are not the same particulars as individuals with them.
An “ontology of components” neither collapses the difference between physical individuals and their properties–as extreme nominalism does–nor represents them as numerically distinct entities–as the Platonizing ontologist of simples does. Collapsing the difference puts grave problems in the way of understanding how there can be real dissimilarities between individuals. How are a red and a green patch dissimilar if there are patches but no colors, or no bases for what one takes to be colors? To say that their dissimilarity is merely their distinctness will not do, since two red patches are distinct without being dissimilar.
Suppose then we take the view that properties are distinct from particulars as the most promising form of the ontology of simples. The undeniable unity of a pale man then becomes the unity of a fact, not of an individual. For an encompassing fact is needed to unite the man and his paleness. On this view, such a fact is not made up of concepts since the individual and the property are parts of the fact. Moreover, such a fact is not an intentional entity, and hence is not about the pale man. For the unity of the fact is to be the unity of the pale man, and not the unity of some entity about him. So the unifying fact is not the fact that the man is pale, for facts-that are intentional. Rather the unifying fact will be the fact of the man's being pale, since facts־of are not intentional.
But how can unity be generated from distinct entities? On the one hand, suppose the property is held to be just the sort of entity that does join an individual together with it into the unitary entity I have called a fact without the assistance of a third entity. So the property joins together, it unifies, entities. This uniting is either an entity distinct from both the individual and the property, or an entity that is a component of the property that does the joining together, or finally an entity that is not different from the property itself. If the unifying is a distinct entity, then our supposition that the property acts to unify without the assistance of a third entity would be violated. If the unifying is a component, then we violate the conclusion that all basic entities are simples–a conclusion which follows from the premiss that for such entities sameness implies identity. If the unifying is not different from the property, then the fact is only the unification of an individual, but since individuals are simples, there is nothing in them to be unified.
So we are led to suppose, on the other hand, that the unity of a fact requires a third entity. Call it the copulation, the being in the fact of the man's being pale. As a third entity, distinct from the property and the individual, it will not be a component of either of them. One can say that its function is simply to unite,4 but how can it perform this function within an ontology of simples? If the uniting is distinct from the copulation, a fourth entity is introduced and a regress has been started. If the uniting is a component of the copulation, we no longer have an ontology of simples.
One is left with the alternative that the uniting is not different from the copulation itself. But this uniting is then included in the fact since the copulation is. The copulation must account not only for the uniting of the property and the individual in the fact but also for a uniting of them together with itself in the fact. This is just our original problem, but now with three rather than two entities to be united. A further entity is needed to unite the uniting to the other entities in the fact, and so on endlessly. In short neither a property and an individual nor a property, a copulation, and an individual account for the unity of a fact on the ontology of simples.
In order to skirt this difficulty one might demote individuals and properties to dependent entities that are mere components of facts. But this carries us over to an ontology of components in which individuals and properties are not distinct from one another or from facts, but are merely different from one another and from an encompassing fact. The advantage of this is that individual and property as non-distinct require no unification, and thus the copulation need not be a distinct entity. According to one version of this view, the copulation in a fact-of is a “tie” but this tie is not identical with any entity allowed in the ontology.5 It becomes possible to treat the copulation in this way when components are substituted for the distinct entities of the ontology of simples. Such an anomalous tie, like any non-entity, suffices when there is no work to do, and there is no uniting to be done when, as in this case, there are no distinct entities. Likewise, the view that since properties are “unsaturated” they can be united with particulars in facts without a third entity is an awkward version of the ontology of components.6 For, on this view, to say that the paleness is distinct from the man would be to treat the paleness as a particular and no longer as a property with its characteristic unsaturatedness. The awkwardness comes from not being able to say they are the same either.
So for an ontology of components there is no question as to how a particular and a property are united, since they are not distinct. The only question that could be raised is how, being one particular, they are different. An unacceptable way of answering this question is to say that difference is rooted in conceptual distinctness. But then difference is not real but only intentional, and this position reduces to Hume's form of the ontology of simples. Other means of accounting for differences also invert important ontological priorities. It is best then to ask, not how components are different, but how difference shows up. Though facts-of were unsuccessful in the role of unifying distinct entities, they can be employed as a criterion for differentiation. Thus entities that are the same but are associated with distinct facts-of are different. Consider Jones and his paleness. Jones is different from his paleness, even though he is not distinct from his paleness. This difference shows up as the distinctness of the fact of or, to revert now to our earlier expression, the distinctness of the condition of being Jones from the condition of being pale. For the property paleness the associated condition is the condition of being pale. As I shall argue in Chapter VIII, §4, conditions, unlike components, are distinct from the particulars with them. Only because of this distinctness of conditions from particulars having them can the difference between components show up as the distinctness of the conditions of having those components. Note also that by the above criterion, a property is different from a similar one that recurs in the same particular. Jones's hunger today is different from his hunger tomorrow, assuming he ate an ample meal in between. For his condition of being hungry tomorrow will be a new condition. None of this commits one to the view that components of individuals or individuals themselves are ontologically dependent on conditions. I shall argue the converse in Chapter XI, §4. Here I am saying only that difference is shown in distinctness of conditions.
A necessary equivalence between facts-that is neither a sufficient nor a necessary condition for the sameness of the corresponding facts־of. Where a is a plane figure, the fact that a is a tri-lateral necessarily implies and is implied by the fact that a is a tri-angle. But being a tri-lateral and being a tri-angle are distinct facts-of or conditions of a. This distinctness results from the obvious difference between the components tri-laterality and tri-angularity.
Further, the fact that a is a figure does not necessarily imply that a is a tri-lateral. But a s being a figure and a s being a tri-lateral are not distinct conditions. Of course, the terms ‘figure’ and ‘trilateral’ have different significations. However, they signify not just properties of a but of other entities as well. When one comes to consider what it is in a that makes these terms truly applicable, one is no longer concerned with the properties of other entities, which properties happen to be in the significations of the terms ‘figure’ and ‘tri-lateral’. There is no special component in a that makes it a figure beyond the component of tri-laterality. There is, that is to say, no “figureness” in a. Of course, a may have the property that is the disjunction of all the properties in the signification of ‘figure’. But that is clearly not the property one wants. For the property that makes a a figure is the same even when there are some variations in what properties are included in the signification of ‘figure’. In short, it is futile to go beyond facts-of to facts-that in order to get a grounding for difference.
It follows from the picture I have drawn of the contrast between an ontology of components and an ontology of simples that most fundamental issues will be handled differently by the two schools. In particular, they diverge on predication and on sameness.
In the ontology of components, a predication is true when, among the components of entities signified by the predicate, there is a component of the individual referred to by the subject. Since components are not distinct from their individuals, the truth of a monadic predication requires the sameness of the individual with its component. In the ontology of simples, however, a predication is true when the predicate signifies an entity that is exemplified by the individual referred to by the subject. Since the individual referred to is simple, it cannot exemplify a component of itself. So exemplification stands between it and a distinct entity.7
A true monadic proposition requires, for the ontology of components, only a single entity as its objective counterpart. Nonetheless, that entity has components that are different from it but not distinct from it. This difference in sameness is reflected by the propositional copula. But the ontology of simples requires a structure of distinct entities, if there is to be a true monadic proposition. Even exemplification itself must be a distinct entity in such a structure. If it were a non-entity, there would be no structure. If it were a component of the individual or of the property, they would no longer be simples. And, finally, if facts-of were the basic entities of which individuals, properties, and exemplifications were the non-distinct components, there would be no simples among the basic entities.
Identity is characterized by lack of a discernible difference. But is there for sameness an analogue of the Principle of the Indiscernibility of Identicals? It will not in general be the case that entities counting as the same particular lack a discernible difference. That is, it will not in general be true that the Principle of the Indiscernibility of Sameness holds. Indeed this principle holds only within the ontology of simples. Within the ontology of components entities that are the same need not have all their properties in common. From the standpoint of his own ontology of components, Aristotle noted that entities may be the same that do not have all their properties in common.8 The implicational form (x is the same particular as y) →(ø) (øx↔ øy) does not have all true instances in the ontology of components. For, though Jones and his paleness will be the same particular, Jones will be carnivorous, even though his paleness is not. (My primary reason for saying that the relation of an individual to one of its properties is that of sameness, rather than identity, is that if it were the relation of identity, the widely held indiscernibility of identicals would fail.)
The above implication (that the same particulars have all their properties in common) holds even in the ontology of components if the ranges of the variables x and y are restricted to individuals and exclude their components. But, without this restriction, if individuals are not simples, there are propositions claiming that one entity is the same particular as another for which the implication does not hold. Individuals are the same particulars as their properties; they are distinct from other individuals and distinct from the properties of other individuals.9 The properties of a single individual are the same between or among themselves10 and are distinct from those of another individual. These cross-categorial samenesses between individuals and properties do not hold for the ontology of simples. This has the consequence that an entity in one of these categories cannot be regarded as more than an external factor in accounting for the behavior of an entity in another category. Since simples have no make-up, they make no contribution to their own behavior. The chief factor in the behavior of simples is not the simples themselves but the entities they exemplify.
It is far from clear that the authority of Leibniz can be appealed to in favor of the Principle of the Indiscernibility of Sameness. It might be thought to follow from his principle of interchangeability salva veritate construed as a principle governing concepts. For, as applied to the same concepts of individual substances, it seems to imply that the corresponding same individual substances have the same properties.11 Nonetheless, Leibniz stated explicitly that distinct subjects could not have the same property, “it being impossible that the same individual accident should be in two subjects or pass from one subject to another.”12 This impossibility can stem only from the fact that, by being “in subjects” accidents are the same particulars as their subjects. Thus Leibniz would be forced to reject the indiscernibility of particulars that are the same. I shall then speak of the Principle of the Indiscernibility of Sameness as the pseudo-Leibnizian Principle.
It seems clear only that Leibniz did assert the converse principle, the Principle of the Sameness of Indiscernibles. Thus he held that (ø) (øx ↔ øy) → (x is the same particular as y) since it seems clear that if x and y are identical they are the same. This principle is logically true when the range of the property variable, ø, is extended to include the relational property of being the same as some specific individual. But when ø is restricted to physical properties, there are objections to this principle. Since electrons are described by antisymmetric state functions, two electrons cannot be distinguished by physical properties, even though their distinctness–their being two–leads to different observations from those that would be met with when there were only one electron.13
But is it not downright absurd to reject the Principle of the Indiscernibility of Sameness and thus to hold that Jones is the same particular as his paleness even though he is carnivorous and his paleness is not? Indeed it is absurd if Jones and his paleness are related as a simple to one of its properties. The absurdity disappears if Jones has components. Even so, one might try to avoid rejecting the pseudo-Leibnizian Principle. For one might say that the relation between Jones and his paleness is not true sameness but a sameness of composition. This sameness of composition is expressed with an ‘is’ of composition.14 On the one hand, this splitting of senses of sameness is, in the context, a merely ad hoc way of salvaging the pseudo-Leibnizian Principle. On the other hand, it is a failure to come to grips with the basic fact that the component of any real entity–which must be a unity–cannot be a distinct entity. So in the end, the rejection of the pseudo-Leibnizian Principle is unsatisfactory only if one is committed to an ontology of simples. It may be said that the pseudo-Leibnizian Sameness is the only clear kind of sameness and that for this reason one should in no event give up the ontology of simples; but Aristotle’s classical theory of sameness is surely clear if one grants the use of the notion of components. And if one refuses, then it is fair for the component ontologist to counter that pseudo-Leibnizian sameness is not clear since the notion of exemplification is not clear. For the pseudo-Leibnizian view of sameness requires that properties, as entities that are necessarily distinct from individuals, be relevant to individuals only by exemplification. If there is no advantage to be gained by polemics of this sort, is there anything in the offing that would genuinely support either side?
§2. Natures the Same as Things. The ontology of components receives rather strong confirmation from the fact that natures are part of the required ontology. We have seen that natures come to be part of the required ontology as bases for necessity. That is, if it is a necessity of a that it is ø’ then there is an entity that is the nature of a and a’s having this entity implies that a will be ø. But could natures be bases for necessity if they were distinct from the entities whose natures they were?15 If not, then physical individuals can no longer be regarded as simples, since they have natures as components. Once their simplicity has been compromised by natures, we are led, for the sake of coherence, to consider properties and parts also as components.
If natures are distinct entities, then one has to ask what the conditions of their association with individuals are. There is no comparable question for the component ontologist. For if a nature is a component and thus the same particular as the individual having it, there are not two entities whose association can be in question. Now if N is a nature distinct from Jones, then N is “associated” with Jones provided it is a necessity of N that if N supports (say) mobility, then Jones is mobile. If this condition were not satisfied, how could we say that N is the nature of Jones? Jones has whatever property flows from or is supported by the nature of Jones. Furthermore, this is not just a contingent fact; for otherwise Jones could fail to be mobile even though his nature supports mobility. Now this necessity required for association is a necessity of N, not a necessity of Jones. For, a necessity of Jones would be grounded on a nature associated with Jones. Our condition of association would then presuppose, rather than itself fix, the associated nature.
Since the necessity is of the nature N, it must be grounded on a further nature, M. This supposes, of course, that M is associated with N so as to be the nature of the nature N. And the condition of association is a necessity of M. There will be a necessity of M only if there is a further nature associated with M, and so on without stop. So for the distinct entity N to be the nature of Jones, N must be the beginning of an unceasing chain of natures of natures.
Before asking whether this regress is objectionable, let us see whether our argument for it is sound, by facing it with two questions. (I) The regress develops supposedly because at each stage the association of a nature with the entity that it is the nature of depends on the nature of the nature. But is there really such a dependency?
Consider the difference between the following propositions:
(1) □ a(øa → ψa),
2)) □ b(øa → ψa),
where ‘□a’ means ‘it is a necessity of a that’. So the difference is that (1) expresses a necessity of a and (2) a necessity of b. Suppose that a–being a piece of fresh litmus paper just dipped in a solution b–is of such a nature that if it is blue then b is alkaline. This amounts to supposing that:
(3) □ a(a is blue→b is alkaline).
But from this it does not follow that:
(4) □ b(a is blue b is alkaline),
for since we are concerned in (4) only with a’s blueness and not with a’ss nature, a could be a shard of blue pottery rather than a piece of litmus paper. Even when (3) is true, it is surely not in the nature of the solution b that the blueness of whatever might be a– including the shard–implies the alkalinity of b. Conversely, though it may be true that:
(5) □ a(b is alkaline→ a is blue),
it may still not be true that:
(6) □ b(b is alkaline→ a is blue).
For, again, it is not the nature of a but a’s blueness that is the concern of (6). So a may be any object dipped in b.
Applying this to the case of the association of natures with individuals, one sees that neither of the following implies the other:
(8) □ j(the nature N supports mobility→j is mobile).
This is relevant since there would be no regress set off by appealing to (7) if it were the case that (7) co-implied (8). For then one could say that (7) is grounded on the nature of Jones, which is simply N, rather than on the nature of the nature N. But since there is no implication between (7) and (8), the appeal to (7) sets off a regress.
(II) This leads directly to our second question. Why appeal to (7) in the first place? Would not an appeal to (8) suffice to establish an association of nature with individual? The problem was to give some sense to the notion of a nature’s being the nature of an individual when the nature is distinct from the individual. But when one uses ‘□j’ in (8), one presupposes that one has in hand a sense for the notion of a nature’s being the nature of Jones. Using (2) of Chapter II, §4, (8) above becomes:
(9) (Ǝ m)(j has the nature m • (j has the nature m→ (the natureN supports mobility → j is mobile))).
The trouble is that ‘j has the nature m’ in (9) is not supposed to be intelligible until we have explained the association of ; with its nature. But (9) is one of the conditions for association, that is, for a nature’s being the nature of Jones. Of course, our expansion of (7) will contain ‘N has the nature k’. But if this is susceptible of independent explanation, it can be used in explaining ’; has the naturem’.
Thus when entities are distinct from their natures, one must go in either of two directions. Either entities are associated with natures through natures of these natures (in which case association is obtained at the price of an endless sequence of natures of natures); or entities fail to have an association with natures because an appeal to natures of natures is found objectionable (in which case entities are contingently connected with their natures, and thus any real necessities for these entities are precluded). The price for an ontology of simples with real necessities seems to be merely a regress of natures of natures. Is there any objection to such a regress?
An observation about the logic of de re modalities of the kind being used here is relevant now. Given (1) or (2), it does not matter which, and the additional premiss that □ a(øa), it follows directly that □ a(ψb), but not that □ b(ψb). For example, suppose God’s nature is such that if He exists, then the World exists. So, if He exists by His nature, then it follows not that the World exists by its nature, but that it exists by His nature. Thus:
(10) □ God (God exists→ the World exists)→(□ God (God exists)→ □ God (the World exists)).
Even if the existence of the World is not the result of a free choice made by God–as Leibniz thought it was–but is determined, it does not follow that the World necessarily exists of itself–as Spinoza seems to have thought.16
This calls for a qualification on the view of de re necessity expressed in Chapter II, §5. De re necessities cannot be limited, as they were there, to those that were necessities of the entities that their non-modal components were about. For the proposition that □ God (the World exists) is a necessity of God about the World. We may, however, call such a necessity a “derivative” de re necessity. For □a(ψb) will follow from its being the case that □ a(øa), for some ø.
How does this conclusion apply to the regress problem? If natures are distinct entities, it will not be a necessity of Jones but at best of his nature that he is mobile. For it is false that:
(11) □N (the nature N supports mobility → j is mobile) →(N (the nature N supports mobility) → □j (j is mobile)).
(Clearly, (11) does not cease to be false when only the left-most □N is changed to ‘□j’.) Perhaps then one should simply settle for necessities about individuals in the world that are not necessities of those individuals, but of entities distinct from them. Such necessities would still be real. The difficulty is that such necessities are derivative and must then be derived from non-derivative necessities. But it turns out there are, on the ontology of simples, no non-derivative necessities. At first this might not seem to be the case. It seems as though □ N(j is mobile) derives from the nonderivative □ N (the nature N supports mobility) via the antecedent of (11). Yet there is no such non-derivative necessity. For, by all the above reasoning, the necessity that N support mobility is a necessity of the nature M of N, not a necessity of N despite the way we have written it. After all, N must be associated with M, and only on the basis of that association are there any necessities about N. But then such a necessity about N will be a necessity of M. However far back one goes, one never reaches a non-derivative necessity.
The need for non-derivative necessities to back up derivative necessities runs deeper than the mere fact that I have defined derivative necessities in terms of non-derivative ones. To reject the definition is to say that the way one individual affects another in no way depends on the components of the former. But it seems clear that a’s being ø will be a necessity for the distinct individual b only because of what b is itself by its nature. Otherwise, it would be possible for b and a third individual c to have natures that are exactly alike in regard to what they imply for themselves, but are dissimilar only in that they imply different properties for a. The most detailed probing of b and c would show no dissimilarities in them. Yet they would necessitate dissimiliar properties for a. On this basis, I judge the regress of natures of natures to be objectionable. Neither non-derivative nor derivative de re necessities for individuals are allowable on the view that entities are distinct from their natures.
The Platonic view that forms and physical individuals are distinct leads to just this view. There may well be necessary connections among the forms. But for an individual to be something necessarily it is not sufficient that there be a corresponding form. The form must be such that it is necessarily exemplified by the individual, and this is a necessity either of the form or of the individual. We have just seen, however, that neither of these conditions is true. It can then be no more than likely that the individual will exemplify the form. Only if the forms were indwelling natures, rather than distinct entities, would the world of becoming be a domain of necessity rather than one of the merely likely (eikos) and hence as well a domain of the real rather than of the mere image (eikon).17
Once natures are recognized to be components, reservations based on the pseudo-Leibnizian Principle can no longer stand in the way of recognizing properties and parts as components. For if natures are components, then Jones and his nature are the same even though Jones is pale and his nature is neither pale nor dark.
The consequences of avoiding the problem of association by taking natures to be components are theoretically quite satisfying. Consider, for example, how Jones does not fall short of what his nature requires. For the ontology of simples, Jones could be kept from falling short of his nature only by the nature of his nature, and the regress thus generated was found to be objectionable. But with natures as components there is no circularity to the claim that Jones’s nature grounds the necessity of his being mobile if it is his nature to be mobile. There is no circularity since this claim is not needed to say what nature it is that is associated with Jones. There is no need to pick a nature out of heaven and tie it to Jones, for he is the same as his nature. This sameness with his nature is, of course, a necessity of Jones. As a necessity, it is grounded by Jones’ nature. But also it is a necessity of Jones’ nature that it is the same as no other entity than the one it is actually of, that is, Jones. Jones cannot switch natures, and his nature cannot switch subjects. It might seem that the necessity of the nature could be grounded only in the nature of the nature. But since the original nature is the same as Jones, one is free to say that the necessity of the nature to be the same as its subject is just a necessity of the subject for its nature to be the same as its subject. Thus where j is the same as N, ‘□ j’ can replace ‘□N’, when the necessity is real. So with natures as components, an appeal to natures of natures is superfluous.
Despite this necessary sameness, natures are different from the entities that have them. The difference between individual and nature is not, as Aristotle thought, a difference between the indefinable and the definable.18 What is definable is not the nature but certain components that the individual has by nature. The difference between an individual and its nature is that between what has but is not a component and what is a component that cannot be expressed by a definition.
The view that natures are not different from properties had by nature comes to grief over the matter of necessity. A property had by nature is a property that something necessarily has. But if F is both a nature and by nature, what grounds the necessity with which an individual has F? Since natures ground necessities, F will ground the necessity with which the individual has F. Clearly this cannot mean, as our nature model for □ a(Fa) might suggest, that there is a nature, namely F, that a has and, if a has it, then Fa. For then natures fail to perform a function that other properties cannot perform, since any property implies itself. What it must mean is that F has a special power over individuals whereby each of them will have a property exactly similar to F. But it is simply false that the properties signified by a real definition of an entity are universally distributed. Otherwise, every individual would simultaneously be a man, a tree, and everything else.
Logical properties are alleged to apply universally. But this can hardly be evidence that these properties ground the necessity of things for having them. That is, it is no indication that the properties themselves have a special power over individuals whereby the individuals submit to being their instances. If this were the case for the property of being non-contradictory, why should it not be the case for the non-logical property, characteristic of men, of being capable of consciousness. The mere fact of wider applicability is not going to make the difference. It is not a property that grounds the necessity of an individual to have it. Rather it is the nature of an individual that determines whether a property, however widely or narrowly distributed, is had necessarily.
§3. Natural Kinds and the Primacy of Individuals. Natural kinds are in some way related to natures. If two individuals belong to the same natural kind, they are not just individuals that share some properties (or some components other than properties such as actions or parts). They are alike in sharing properties that they have by nature. Normally an even stronger interpretation is put on the concept of natural kind. Thus two individuals belong to the same natural kind not just when they share some properties that each has by nature but when the properties which one has by nature are just those that the other has by nature. A simpler way of putting this would be to say that two individuals belong to the same natural kind if they have exactly similar natures. Now since natures are the same as the individuals having them, two individuals cannot have precisely the same nature. They can nonetheless have exactly similar natures in the sense that those and only those properties that are by the nature of the one are by the nature of the other. Of course, this only pushes the problem back to properties. For, properties, too, are the same as the individuals having them. So the properties that one individual has by nature are exactly similar to but not the same as those that another has by nature. To say what natural kinds are I must, then, first say what similarity is.
There are some component ontologies that yield a ready answer to the question of the nature of (exact) similarity. Thus, suppose one admits, in addition to properties-in-individuals, universal properties as well, that is, those that are both distinct from individuals and capable of being exemplified by several individuals. Then one can say that property A in individual a is similar to B in b if there is a universal property ø such that A is an exemplification of ø in a and such that B is an exemplification of ø in b.19 If, however, there is an alternative explanation of similarity that dispenses with universal properties, there is no reason to add universai properties to the required ontology.
I think there is such an alternative. Parts, as well as properties, are components. Possibly then some interesting truths about parts can be deployed in respect to properties. One fact about parts is that they can often times be transferred from one individual to another. But can properties be transferred? The blackness of this newstype comes off onto my hand; the stickiness of this candy gets onto my fingers; the heat of this tea is transferred to my mouth. Protoza, which are single cells, divide to form new cells. The parent cell transfers, not just parts, but also properties to the daughter cells. There is then some basis for extending the notion of transferral from parts to properties. Now a part that is taken from one individual and inserted in another does not remain the same as the former individual since, successively, it is a component of two individuals. The familiar notion of transferral is, I am assuming, broad enough to allow for such a double sameness. The part that is transferred is similar to but not the same as itself at the end of the transfer. Likewise, blackness, stickiness, or heat is transferred despite the change of sameness that occurs as the property finds a new seat. Taking this notion of transferral as basic, one can account for similarity by means of it.
Thus, property A in individual a is (exactly) “similar” to B in b,when, if the result of transferring A of a to b is some component C of b, then C and B are not different. It will be recalled that C and B are not different only if b’s condition of being C is the same as its condition of being B. Of course, for this account of similarity to be adequate there need not be a way of transferring every property. Color is transferred from a worm to each of the two smaller worms resulting from splitting it. But communicating ideas is not a transferral, since there is an intermediate causal link here in which the ideas are not present. All that is required is that the general notion of property transfer be meaningful. This account of similarity does imply a regress, but, pace Russell, a harmless one.20 If A and B have the similarity property, then D and E could have a similar similarity property. But one can treat this similarity of similarities in a similar way, and so on. One can let the regress run on for, since it is harmless, there is no need to introduce universals to stop it.
In the strong sense of natural kind, two individuals belong to the same natural kind if they have natures supporting similar properties, that is, properties that would not differ if all of them in the one individual were transferred to the other. But, as far as the practice of action on prior experience goes, that is, as far as the required ontology goes, it will be unnecessary to give this stronger interpretation to the concept of natural kind. It will, that is, be unnecessary to think of individuals of a natural kind as having (exactly) similar natures. So it may be that no two individuals in the universe are such that every necessary property of one is matched by a similar necessary property of the other.
Moderate empiricism requires at most that there be classes of individuals that are similar in respect only to certain of their necessary properties. The requirement that the universe be segregated into a manageable number of classes of individuals with similar natures is seen to be excessive. This point will be developed in Chapter V,§2. For the present we shall say that in the weak sense of natural kind to be developed there, a natural kind is natural only in that the properties determining the kind exist in the individuals of that kind by virtue of their natures. Individual variations in a natural kind in the strong sense can be regarded only as accidents. But it is well to have the option, provided by the weak sense of natural kind, of thinking of important differences between men, say, in different cultures, not as mere accidents of culture, but as differences in the necessary properties of individual men.
Associated with a natural kind, such as the class of men, there may be a natural-kind noun, such as ‘man’. Like many other terms it both refers and signifies. This noun refers, distributively, to the members of the associated kind. It does not signify a universal, but like other terms, it signifies components of entities. ‘Red’ signifies the rednesses of red entities, and ‘frictionless surface’, though meaningful on the basis of its construction from terms with significations, signifies no entity. So the sentence ‘a is red’ is true when a is the same as one of the rednesses signified by ‘red’. The rednesses signified by ‘red’ are, of course, similar in the above sense.
Now if K is a natural kind, the following three conditions are to be satisfied by a natural-kind noun, W, for members of K. (1) ‘N’ must signify components. These components will be similar among themselves, and each will belong to a distinct member of K. Since natural kinds are to be non-empty, ‘N’ will never fail this requirement because of K’s being empty. (2) As a natural-kind noun, W will signify these components as ones that belong to members of K by the natures of these members. Thus when ‘a is an N’ is true and W signifies the component ø of a, a has ø by its nature. If a term signifies components as had by nature when in fact these components are contingent to some of the individuals referred to, it is at )est a “purported” natural-kind noun. Likewise, when K is empty, ‘N’ is at best a purported natural-kind term. (3) The components signified by ‘N’ must be “representative” of other components that are similar among members of K and had by the natures of its members. That is, though the component that ‘N’ signifies for any member of K need not be the only component that member needs to belong to K, still no entity outside K will have, by nature, a component similar to the one signified by ‘N’.
The component of a member of K signified by ‘N’ I shall call “the kind component” of that member in respect to ‘N’ and K. So if W signifies ø for a, then, assuming ‘N’ is a genuine naturalkind noun, ø is the kind component of a in respect to ‘N’ and K.
Suppose the kind component, ø of a in respect to ‘N’ and K is in fact a property. Thenøcan be in the signification not only of the natural-kind noun ‘N’ but also of the property predicate, say, ‘F’. Thus the kind component humanness is signified both by the natural-kind noun ‘man’ and the property predicate ‘human’. The two terms do not differ in what is signified. In particular, ‘man’ does not differ from ‘human’ by signifying a putative entity called a species. The difference is only that ‘man’ does, though ‘human’ does not, signify humanness as had by nature and as representative of other components that are both similar among and had by the natures of members of mankind. If the semantics of natural-kind nouns does not require species and genera as special components of individuals over and above their properties, parts, and actions, it is unlikely that there is any reason why such special components should be countenanced.
Aristotle made this point by saying that the last differentia reached in defining a species is the being of or the nature of any entity of that species.21 The generic term in the definition adds no new entity; the property it signifies is already implied by the property signified by the differentia term. And the species term introduces no new entity; it signifies exactly what the differentia term signifies, though only the former signifies the property as one had by nature. The term expressing what an entity is (ti estin)– the species or natural-kind term–does not signify a new entity0a species. This term signifies, rather, what makes the entity be of its kind (to ti ēn einai), and this is what the differentia term signifies, though in a different way.
The old dispute as to whether natural kinds are real or conventional has the following resolution here. Natural kinds are, at least in our weak sense, not real if by being real one means that within each of them all members have exactly similar natures.22 The members of a natural kind need be similar only in respect to some, not all, of the necessary properties of their members. This leaves room for an element of conventionality. For a natural kind in this weak sense will be overlapped by numerous closely related natural kinds. But one cannot think in terms of all these natural kinds. A selection is made from competing kinds, and this fact is recorded linguistically in the use of certain natural-kind nouns. Despite the fact that this selection is not imposed by natural divisions, and is to that extent conventional, natural kinds are real in the limited sense that an individual is bound by a real necessity to its kind component. So what a genuine natural-kind noun signifies for an individual is a component the individual has necessarily. Contrary to Locke, this necessity does not derive from the fact that the individual is brought under that noun.23
Even though a natural kind is not based on exactly similar natures or upon special components called species and genera, the concept of natural kind is still a useful one, as I shall try to illusträte. We know that when charges are accelerated through radiotransmitter towers, electromagnetic waves are radiated. We are convinced it is the nature of such things to emit radiation when charges are accelerated through them. But on this basis we are unwilling to say of an arbitrary individual that it emits radiation by its nature when a charge is accelerated through it. One feels safe in making an inductive generalization only to things within some restricted natural kind. I shall indicate the reason for this in Chapter V. But even here it is easy to recognize the pitfalls of ignoring the restriction to natural kinds. For according to early quantum theory at least, electrons revolve in atoms for extended periods without emitting radiation. They do not emit continuously and in their ground states there is no emission. It is only of systems of a certain kind that it can be said they emit energy when charges are accelerated through them. Of course, when an induction is made for a natural kind, the kind may not be restrictive enough, but without some restriction by kind, the field is so wide open that no induction is warranted.
In other words, even if it is true that:
(1) □ a(Fa),
where a is a metal bar and F is the property an object has of being an emitter of electromagnetic energy when a charge is sent back and forth through the object, this statement hardly constitutes support for the unrestricted claim that:
(2) (x) □ x(Fx).
(3) (x)(Nx→□ x(Fx))
is true. The restriction imposed by ‘N’ makes the difference. The idea is that if one is convinced that a is F by nature, then one is warranted in believing that other individuals are F by their natures only if one assumes that those other individuals are somewhat close to a in nature. How close will be specified in Chapter V.
I noted earlier (Chapter II, §2) that a necessary condition for induction is that the property to be projected have a chance of being possessed necessarily by the observed individuals. What I am saying here, in denying that the induction to (2) from (1) could be warranted, is that the chance of the observed individuals’ necessarily having the projectible property is not a sufficient condition for warranted induction. A further requirement is being sketched here, and it introduces the notion of natural kind. This additional requirement will ultimately be that the projected property be a conditional property in respect to which there is a chance that the antecedent will specify a natural kind.
Now (3) says that entities of a certain natural kind have a certain property by their natures. On its surface at least, this is not the same as saying that any entity is such by nature that if it is of a certain natural kind it has a certain property. That is, it seems different from:
(4) (x) □ x(Nx →Fx).
If we call (3) a “kind-specific” necessity, then inductively generalizing claims like (1) involves moving to a kind-specific necessity. But to generalize only in this way, and not to (2), does not mean we have lost faith in the original singular claim, (1), and wish to replace it with the more qualified claim that:
(5) Na→□ a(Fa).
For, an individual’s necessary properties are not conditioned by its assignment to a kind; rather, its kind is conditioned by the necessary properties it has.
Though (3) seems to say something quite different from (4), the two are in fact interderivable via a principle for natural kinds that follows from the way we have described natural kinds. This is the Principle of Essentialism, which says that whatever may be of a given natural kind must be of that kind; that is:
(6) (x) (◊ x(Nx) →□ x(Nx)).
The principle is clearly equivalent to the conjunction of:
(7) (x) ( ◊ x(Nx) □→ Nx),
(8) (x) (Nx →□ x(Nx)).
In discussions of essentialism, emphasis is ordinarily placed on (8) to the neglect of (7).24 From (3) with the aid of (7) one derives (4) by an indirect proof. The idea is that if an individual might be N but not F– that is, if (4) is false–then the individual will be N in view of (7) and thus by (3). it must be F. On the other hand, from (4) with the aid of (8) one derives (3) by distributing the modality in (4). So either (3) or (4) is an acceptable way of generalizing (1) inductively.
It remains to justify the Principle of Essentialism itself. I rely entirely on the notion of a natural kind, in the weak sense, and on the characteristic laws of S4 and S5, which are, as I noted in Chapter II, §4, true on the nature-interpretation of necessity. To prove (7), assume that an arbitrary individual might belong to a certain natural kind. Thus it is possible that this arbitrary individual has certain properties necessarily. Now that the possibly necessary implies the necessary is a direct result of the characteristic law of S5 that the possible is necessarily possible. So the individual necessarily has those properties and thus actually belongs to that natural kind. So (7) holds. To prove (8), assume an arbitrary individual belongs to a natural kind. It will then have certain properties necessarily. Now that the necessary is necessarily necessary is the characteristic law of S4. So the individual necessarily has those properties necessarily, and thereby necessarily belongs to that natural kind. So (8) holds.
§4. Ontological and Methodological Essentialism. Karl Popper has quite rightly objected to the view I shall call “methodological essentialism.”25 Among other things, the methodological essentialist holds that the scientist can ultimately acquire indubitable knowledge and that natures provide an end to the chain of seientific explanation. On the other hand, the view expounded here that natures are the grounds of the necessity of necessities and that every individual is subject to necessities and hence has a nature is a view I shall call “ontological essentialism.” Now, must ontological lead to methodological essentialism? If it must, then in view of Popper’s criticism, one should abandon real necessity altogether. One should settle for what philosophers have long been settling for–intentional necessity. The practice of action on prior experience would, however basic to human practice, have to be declared an illusion. What that practice would have us believe about the world, we simply should not believe.
As to the matter of indubitability, there is nothing in what has been said about a necessity to imply that it is indubitable. To assert that what entities are necessarily proceeds from their natures does not imply that the kind of evidence that one has for a necessity, or could have for it, puts any doubting of the necessity out of the question. Moreover, my position does not imply that a necessity is a priori in the technical sense of being supportable up to any degree without recourse to sense experience. Quite the contrary, it seems clear, in general, that, whether an entity has a character contingently or by nature, justifying the claim that the entity has the character can be accomplished by an empirical procedure.
It is also well to point out that the Principle of Essentialism–that what could belong to a natural kind must belong to it–implies no objectionable apriorism. It might be alleged that the principle commits us to holding a priori that men cannot change into bats and that lead cannot change into gold. Indeed, it commits us to holding that if ‘man’ and ‘bat’ signify properties that are necessary but incompatible, then a man cannot change into a bat. (I revert here to Platonizing talk about a property that many individuals have as short for talk about many exactly similar properties.) But assuming their incompatibility, inductive considerations are involved in determining whether the two properties should be treated as necessary properties and thus, possibly, as kind properties. Our judging that a man cannot change into a bat is not then automatic. It waits upon the empirically motivated decision to treat the properties in question as necessary, or even as kind properties. Once it is decided they are necessary properties, or kind properties, we must judge either that Count Dracula was an impossible being or that he was not both truly a bat and truly a man, even though on different occasions he resembled each.
As to the second matter of natures being ultimate scientific explainers, nothing I have said about natures implies that they enter scientific explanation as explainers at all. In fact, it is consistent with ontological essentialism that there be no terminus to scientific explanation. There might, that is, be a level below any level that scientific explanation will ever reach. Natures are not, however, at the bottom level of any scheme of scientific levels; they are altogether outside any such scheme of levels. The necessity of the explanatory principles at each level and between the levels is grounded by natures; what the principles say entities are flows from the natures of these entities. Even though a principle is explained by parts or properties at a more basic level, it is nonetheless a principle that holds by nature. Those entities–whether properties or parts–explaining it are simply some of the manitestations of the natures by which the principle holds.
The reductionist may complain that this requires that we treat the existence of complex macro-individuals as irreducible. For it seems that the explaining parts and properties are always to be manifestations of the natures of individuals that contain them. But even this is not a consequence of ontological essentialism. When we speak of what a proposition says about individuals, we always mean what it says about unreduced individuals. These may not be the individuals that the proposition seems to be about. So it is the natures of unreduced individuals that ground the necessities at the infinity of possible levels.
Suppose, for illustration, that certain individuals, taken to belong to the natural kind, the class of volumes of gases, are found to obey certain relations, Q, between temperature, pressure, and volume. Assuming this behavior is by nature, and letting ‘N’ be the appropriate natural-kind noun, we have the truth that (x)(Nx→ □ x(Qx)). It may be wrong to take volumes of gases as unreduced to molecules, but we start with that assumption. Now we wish to inquire what, if anything, about the natures of gases makes them Q by their natures. Are there properties flowing from their natures from which Q is derived? Is there, that is, some E such that both (x)(Nx→ □ x(Ex)) and (x)(Nx →□ x(Ex Qx))? In this case, E might be the molecular structure of gases.
Now suppose it is the case that there are no volumes of gases, but only the molecules said to compose them. The truth conditions for propositions apparently about gases will be the truth conditions for propositions actually about molecules. As a consequence of this and of the way we have described E, the proposition that (x)(Nx→ n x(Ex)) becomes a bare tautology, and the above step ceases to be an explanatory one in the inquiry. In order to start the inquiry again, a new natural-kind term is required that represents a kind of unreduced individuals. What matters here is not whether gases are unreduced or not. What matters is rather that, whatever the unreduced individuals, an explanatory step consists not in introducing a nature but in discovering behind the properties already deemed to be by the natures of the unreduced individuals properties that are “more fundamental” in respect to the same natures. The explainer of E’s implying Q might be G, where, assuming gases are unreduced individuals, (x)(Nx → □ x(Ex □→ Gx)) and (x)(Nx→□ x(Ex (Gx □→ Qx))). In this case, G would be a complex property combining a certain probability distribution for molecules of a gas with their mechanical behavior. On the one hand, the grounding of necessity in natures requires no end term for an inquiry of this sort. On the other hand, it is not natures but properties or parts had by nature that are introduced at each step. Ontological essentialism is then incompatible with methodological essentialism, since the former can be joined quite naturally with an account of scientific explanation that requires no end point.