§1. The Stimulus-Response and Fine-Structure Models for Capacities. It might seem that capacities should be part of the required ontology. If the existence of entities of a given sort follows from the existence of physical necessities, then entities of that sort are part of the required ontology. And it would seem that there are capacities if there are physical necessities, for an entity has a capacity to be whatever it must become in certain possible circumstances.
After admitting–over and above individuals–natures, properties, actions, and conditions, it looks as though there are no restrictions at all on admittance to the required ontology. Nonetheless, the entities thus far admitted are perfectly sufficient, when taken in combination, to play the role played by capacities. Our policy of admission is then restrictive enough to eliminate capacities, since they are not in fact required by physical necessities. The conclusion to the above argument must then be changed. It is, in fact, merely an argument from necessities to either capacities or whatever entities that in combination have the role of capacities.
Not only do necessities require either capacities or combinations with their role, but, conversely, unless there are necessities, the observational datum that individuals change will not be sufficient to warrant the conclusion that there are either capacities or combinations with their role. As I shall show, liquid water lacks the capacity to change into ice, even though it does change into ice, if there are no necessities. This result is incompatible with the Aristotelian view that any action, such as the change from liquid to solid, is to be identified with an exercise of a capacity.1For, in this view, there could not be actions unless there were necessities, since to have a capacity involves the necessity of an action in certain circumstances. Though I introduced actions to allow for necessities across time, it does not follow that actions would be impossible in a world without necessities. Thus, in my view, it need not be the case that to act is to exercise a capacity. In a world without necessities there might still be actions, but they would not be exercises of capacities.
Return now to the question of the superfluousness of capacities. There are many arguments for denying capacities entitative status, but as they stand they are insufficient. On the one hand, there is the ‘stimulus-response’ model for capacities.2 The logical form of a capacity proposition is, on this model, a modal conditional proposition; its antecedent expresses an operation in certain circumstances and its consequent expresses a realization of the putative capacity. Unless the conditional is explicitly modal, it cannot be relied upon to support a counterfactual conditional. But a capacity proposition does support some associated counterfactual conditional. The modality must be that of necessity.
That is, associated with the capacity proposition that a is ønable is the counterfactual that if a were ψed it would ø□‹ To support this conditional, the modal stimulus-response conditional □ a(ψa→øa) is not required. For the counterfactual does not say that the consequent holds in all possible circumstances in which the antecedent holds, but only that it holds in certain of the possible circumstances in which the antecedent holds. Thus when a is ønable, there will be circumstances K such that □ a(ψa in K →øa On the other hand, the non-modal conditional (ψa in K →øa does not suffice to imply the counterfactual, for the latter explicitly requires that a ø’s in possible situations in which a ψ’s and implicitly restricts these possible situations to those in which K holds. So the stimulusresponse model requires a restricted modal conditional.
A further refinement is still needed. If capacity claims were equivalent to restricted modal conditionals, then, as a simple exereise in modal logic would make clear, an entity would have each of its capacities necessarily. But surely capacities are often contingent; this poker is not always able to sear wood but can do so only when it is hot. This suggests treating the logical form of capacity claims as conjunctive. One conjunct will be a modal conditional, whereas the other will not be modal and may be contingent. A full elaboration of the non-modal conjunct introduces considerations of the structure of entities, and these do not properly belong to the stimulus-response model.
On the other hand, there is the ‘fine-structure’ model for capacities.3 To have a capacity is to have components that are the causal basis for what is spoken of as the realization of the capacity in a certain stimulus situation. These components may be properties, parts, or actions of the entity with the capacity. They need not be components introduced by theoretical science; they may be quite familiar components. In this model, the entity with a capacity is not a black box of which a certain conditional is mysteriously true. Rather, the entity has a fine structure of components that are the ontological roots of the capacity.
The motivation for the fine-structure model was undoubtedly the conviction that there ought to be something about the entity with a capacity that implies the stimulus-response conditional. In the very concept of a capacity is the notion that an entity with it shapes that entity’s conditional properties. But to be satisfactory, the fine-structure model has to account not just for a conditional, but also for the modality of the conditional. We just saw that, in view of the connection with counterfactuals, the stimulus-response conditional must be a necessary one. The stimulus-response model is incomplete, not just because it does not explain why a certain stimulus is followed by a certain response, but also because it does not explain why its being so followed is a necessity.4 To take account of this additional requirement, it is natural to suggest that fine structure is also the basis for necessity. In other words, an entity will obey a given stimulus-response conditional of necessity provided the entity has a certain fine structure and having this fine structure implies that conditional.
It soon becomes evident that the difficulties in respect to modality remain for the fine-structure model. First, if a modal conditional of the above kind is equivalent to the existential claim that there is some component of fine structure the entity has the having of which implies the corresponding non-modal conditional, then this existential claim will itself be necessary. There is thus the problem of accounting for the necessity of there being some such component. Notice that it is not necessary that the entity have a specific component. The same entity in different circumstances may have the same stimulus-response conditional hold true of it because of quite different fine structures.
Second, if a given component of fine structure does account for a conditional, then there is not a mere contingent relation between the fine structure and the conditional. If it only happened that the fine structure implied the conditional, then the fine structure could not be the basis for the conditional. This will be argued in more detail in §5. The upshot of this second difficulty is that, independently of the attempt to account for the modality of the conditional by fine structure, modality appears in the fine-structure account of the unmodalized stimulus-response conditional.
The fine-structure model seems, then, to raise more problems of modality than it resolves. Is it possible, within that model, to make a frontal assault on the general problem of the basis for necessity? The prospect seems encouraging, for below any given level of fine structure there are always more basic levels. Necessities at the operational level are based on first-level fine structure; necessities of the kind we have just discovered about first-level fine structure are to be based on second-level fine structure; and so on. The general account of necessities in Chapter II in terms of natures then becomes gratuitous.
I shall argue that fine structure does not provide a solution to the problem of necessity that is raised by both the stimulusresponse and the fine-structure approaches to capacities. But first I should like to refine the picture given so far of the fine-structure model to the point where it is not objected to beyond its failure to deal with the problem of modality. When the account of necessity by natures is integrated with this version of the fine-structure model, the result is what I shall call the ‘nature’ model of capacities. The nature model will be crucial in the final chapter in the argument that there must be things as well as conditions.
§2. A Regress Problem for the Fine-Structure Model. Locke outlined one kind of fine-structure model. He made the common-sense observation that our idea of a substance of a given kind contains numerous ideas of powers of that substance. But he asserted a view of considerable metaphysical interest when he went on to say that the real essence of a substance of a given kind is not made up of those powers.5 Rather, the real essence is that ‘constitution of the parts of matter’ on which the powers of the substance depend.6 It is such a constitution of material parts, ‘texture of parts,’ ‘internal structure,’ or ‘fine contrivance’ that makes gold soluble in aqua regia, that makes antimony fusible, and that makes lead malleable. Even though he believed powers had a foundation in parts, Locke believed it would be impossible for us to know these parts in view of the limits of our senses. But his epistemological reservations in no way qualified his ontological conviction that capacities depend on unobserved parts.
Notice that in Locke’s view only material parts are the components included in fine structure; properties, actions, and natures are not included. Now since material parts could be separated from the individuals with them and exist by themselves, as properties could not, it is plausible to ascribe capacities to them in the same way we ascribe capacities to individuals of which they are the parts. Atoms of lead, for example, are capable of combining, one by one, with atoms of sulfur to form the substance known as galena. The fine-structure view must then be applied to the atoms of lead and sulfur as well as to quantities of the stuff, the galena, they compose.
So, by Locke’s logic, the atoms must–because of their having capacities–have material parts. These again will move and interact with other parts in such a way that we will wish to attribute capacities to them. Indeed, it is hard to imagine any point in an analysis by material parts at which the parts do not have some rudimentary capacities. But if there were to be parts without parts, then Locke’s version of the fine-structure model of capacities would have reached a dead end. There would then be capacities without a basis in parts. So if any individual has a capacity, it has no simple material parts, that is, no material parts not composed of other material parts.
Such a strong conclusion must make one suspicious of the path by which it is arrived at. It seems clear that individuals ought to be able to have capacities whether or not there are simple material parts.
Our suspicions lead us to ask why Locke excluded components other than material parts from the foundations of capacities. The reason is that for him an individual simply has no components other than its powers and its parts. That is, the components of an individual are: (1) powers for affecting minds–its secondary qualities; (2) powers for affecting matter; and (3) parts in configuration –its primary qualities–which are the foundations of all these powers.7 Solidity and extension are both a cohesion of parts; figure is a relation of parts; motion is presumably a multiplicity of parts of extension; and the fifth primary quality, number, is the unity of a part.
Since for Locke, “relation is not contained in the real existence of things,”8 primary qualities are not both parts and the various configurational relations, but are simply parts. My expression ‘parts in configuration’ is designed to indicate that it is parts, not both parts and configurational relations, that Locke sees as sufficient. If the parts of a thing are themselves ascribed primary qualities, this can only mean that these parts are composed of parts.9
Certainly there are times when such a mereological ontology pays handsome dividends. Newton defined inertia as ‘a power of resisting’ change of state, whether motion or rest, and attributed this power to all material bodies.10 What is the basis of the power in question? Should we say simply that it is based on the property of mass? If we were thinking in Lockean terms, this alternative would not appeal to us. We would hold out for an account of the power of resisting in terms of parts. In view of the relation between mass and energy in the theory of special relativity, ‘we now recognize that the potential energy contained in material bodies is the cause of inertia."11 But if potential energy is that capacity to do work that ‘depends on the configuration’ of a system–be it a watch spring, a charge of gunpowder, or a system consisting of the earth and a stone lifted above its surface–then Newton’s power of resisting is founded on parts.12 We would then withdraw the suggestion that inertia as a capacity depends on a property, mass.
There is, however, a difficulty with Locke’s ontology of parts that has never been satisfactorily resolved. As we have seen, since parts themselves will have capacities, there must be parts of parts. There is no escaping the fact that if there are no components of individuals other than capacities and parts, then there are no simple material parts. Now we must ask whether, first, what is composed of parts is ever a genuine entity or whether, second, the parts composing anything are entities though the thing is never itself a genuine entity.
For Locke, the second alternative–that composites are heaps and hence not unities or entities–leads to what might be called an ontological vacuum. For, since by Locke’s logic parts are themselves composed of parts, no part is an entity. No matter how far the analysis into parts is pushed, one fails to reach entities, since the parts one reaches are all composites, and by hypothesis composites are not entities.
This forces us to consider the first alternative–that there are entities composed merely of material parts, even though those parts are divisible. So some composites are unities or entities and not just heaps or multiplicities. There are two difficulties with this alternative.
First, if the only components are material parts, there is the problem of how an entity with material parts can differ at all from a mere multiplicity of these parts. Surely there is a difference between a genuine entity with parts a and b and a non-entitative composite of a and b. But this difference cannot be in the entity, since in it there is only a and b. And any difference not in the entity will not serve to differentiate it from the non-entitative composite. Second, if the only components are material parts, there is the problem of how an entity can be the same as itself. An entity’s sameness with itself is a component of it–its unity–as was noted in Chapter VII, §5. This is a component any entity must have, yet it is not a material part. In view of these two difficulties, one is thrown back to the alternative leading to an ontological vacuum. It is then clear that an ontology of parts is untenable.
But Locke allowed for capacities as well as the parts they depended on. A capacity might then be the element needed, over and above parts, if there are to be genuine entities. But if a capacity provides what is needed, then the parts would provide what is needed. For to have the capacity, it suffices, on the Lockean view, to have the parts. We have, however, just seen that parts by themselves are insufficient for an ontology. Hence capacities based solely on parts will not provide what is needed for there to be genuine entities.
Looking back, we find that the Lockean version of the finestructure account of capacities–one that limits fine structure to material parts–is warranted only if an ontology is warranted in which entities have parts and powers but no other components. We have just found that such an orttology is unwarranted in that it leads straight to the result that there is nothing. In formulating a fine-structure model for capacities, we might then wish to allow for the possibility that properties and actions, in addition to parts, might be elements of a fine structure. Once having made this allowance, it is no longer the case that the fine-structure model commits us to the view that there are no simple material parts.
It might seem that, even with properties and actions, there is still a regress. For suppose the entity a has the capacity to ø since it has the component ψ. Will it not then be true that ψ itself has the capacity to make a to ø?13 If so, this capacity of ψ will itself be based either on parts or other components of a, and whatever it is based on will have the capacity to make ψ to make a to ø and so on. For example, the kinetic energy of a point mass is the capacity it has to do work ‘in virtue of being in motion.’ The motion will itself have the capacity to make what has it do work. And we must then look still further for the grounding of this capacity.*
However, I do not think the fine-structure model is committed to such a regress. Suppose the property or action on which a capacity of an individual depends is not a component of that individual but is a distinct entity. Then, indeed, it would be appropriate to ascribe capacities to such properties or actions. Whitehead, who treats eternal objects as distinct entities from events or actual entities, is then able to say that ‘an eternal object can be described only in terms of its potentiality for ‘ingression’ into the becoming of actual entities."14 But properties and actions are merely components of individuals. They are hence not distinct from the individuals of which they are components. So to ascribe a capacity to a property or action can be viewed only as a way of ascribing a capacity to the individual with that property or action. When I say ψ has the capacity to make a ø, I am saying exactly the same thing as when I say that a, since it is ψ has the capacity to ø.
However, a’s condition of being ψ is, unlike the property ψ that a has, distinct from a. It may be implausible to say that the component ψ has a capacity to make a to ø. since components do not have the status of causes at all, as I shall try to show in Chapter XI, §3. But a condition can be a cause. It seems then that conditions do genuinely have capacities. It is easy enough, though, to account for their grounding without a regress. The condition of being a ψ has the capacity to make a to ø precisely because it is a condition of an entity, of which one component is ψ itself. The capacity of a condition is grounded in a corresponding component that has no capacity. Thus the regress is stopped. This is not quite the same as saying that properties ground their own capacities.15 It is rather to say that properties, which do not have capacities, ground the capacities of the conditions that are the havings of these properties.
Parts of entities were treated as components in our discussion of Locke. From what I have just said it appears that capacities cannot be attributed to parts. But can Locke still be faced with the consequence that there are no simple material parts if parts as components do not themselves have capacities? At this point, a difference between parts and other components is crucial. A part may be separated from the entity of which it is the part and made an independent entity. It then becomes distinct from the entity of which it was the part. Though a property or an action might become a component of a new individual, neither thereby becomes an independent entity. Strictly speaking, the part has capacities only when it becomes a distinct entity. When the part is a component, its supposed capacities are, in actuality, capacities of the whole. Still, the capacities it has once it is distinct are, in the Lockean view, based on its parts. These parts of the separated part did not come from nothing. They were themselves parts of the whole from which the original part was separated. (They were not strictly parts of the unseparated part, for a component has no components.) It does not matter, then, that parts as components do not have capacities. For all that is important is that the capacities a part has, once it has been separated from the whole, require that this part have parts and hence require that the whole have a yet deeper internal structure of parts. So if capacities are based solely on material parts, there will be, in the entity with the capacity, no parts that, when separated from this entity, are material simples.
§3. Conditional Causality. The proposition that lead is malleable is undoubtedly true. But I am contending that its truth does not require that there be a component called malleability. In fact, my ontology is Megaric16 in that it excludes such a ‘dispositional’ property as a component of any entity.17 If it is argued that lead must have the dispositional property of malleability for the proposition to be true, it should be pointed out that the proposition admits of reformulation as the modal proposition that any chunk of lead is capable of having its shape changed. The problem, then, is one of describing the truth conditions for a modal proposition. The general form of this problem is that of finding in what contexts the truth conditions for the constituent non-modal proposition are to be found satisfied.
Some thinkers find it hard to draw a line between ‘dispositional’ and other predicates. For the proposition expressed by almost any atomic sentence implies a capacity proposition. So, they reason, the predicate in such a sentence must be dispositional.18 That is, if (ψa →Cøa), where ‘Cøa’ is an abbreviation for a is capable of øing’, then ‘ψ’ is to be deemed a dispositional predicate.
On the one hand, if being a dispositional predicate means only that there is such an implication, then the designation is merely misleading. Analogously, we would not say that ‘ø’ is a modal predicate just because øa implies that a possibly ø’s. Moreover, even one of the positivist’s ‘occurrent’ predicates, say ‘looks orange’, is generally taken to imply, when it is true of some individual, that something has the capacity to appear orange. But it would be misleading to call this a dispositional predicate when looking orange is so clearly not a capacity.
On the other hand, if ‘ψ’ s being a dispositional predicate means that there is some θ such that the truth condition for 1ψa is identical to that for Cθa, then it is patently false that the above implication–or even a corresponding coimplication–is a sufficient condition for ‘ψ’ s being a dispositional predicate. This wheel’s being round implies, under normal suppositions, its having the capacity to roll. But the truth condition for its being round is its having the ‘actualized’ as opposed to the dispositional property roundness. The mere fact that having the condition of being round implies, or is also implied by, having the capacity to roll does not show that these conditions are identical.
In fact, there is a general difference between having an actualized property, an action, or a part and having a capacity. It is that having one of the former is often a cause, whereas having a capacity causes nothing, even though having it may be essential if a certain effect is to be realized.՝ The wheel’s having circularity may be the cause of my having a visual impression of it as circular, though its having the capacity for so impressing me is not the cause of the impression.
The reason for this is plain if the view of capacities presented in this chapter is accepted. If having a capacity were having a specific fine structure, then indeed having a capacity could be a cause. But, in fact, for a to have a capacity is for a to be such that there is some unspecified fine structure that a has. Such an existential condition is not itself causal. So the individual’s condition of having one or more capacities is not the truth condition for the proposition that it has a certain actualized property.19
The moral of all this for our Megaric program is that when we treat certain unspecified components as bases for capacities it does not follow that these components are themselves capacities. Indeed, insofar as such a component is an actualized property, an action, or a part, it is not a capacity.
In giving an account of capacities, it is unavoidable that we shall have to emphasize two seemingly opposed factors. On the one hand, the components on which a capacity is founded are such that the conditions of having these components are causally responsible for the behavior envisaged by the capacity. Having the components of fine structure causes, in appropriate circumstances, what one calls the actualization of the capacity. On the other hand, capacities are often such that they can be exercised freely. While providing for causation, we must insert restrictions that do not preclude freedom.
If the causation is what I shall call ‘conditional,’ then one can allow for freedom without eliminating causation. Now condition a unconditionally causes condition b only if the occurrence of a in the given circumstances is sufficient for the occurrence of b. However, a conditionally causes b only if there is not this unconditional sufficiency. Rather, a conditionally causes b only if, given that b occurs, the occurrence of a in the circumstances involved implies that of b. The lack of unconditional sufficiency allows for autonomy in the exercise of capacities. Nonetheless, where there is conditional causality, the individual’s having exercised the capacity implies that the fine structure together with the circumstances are sufficient for the exercise. The fine structure’s sufficiency for the exercise is conditional upon the exercise itself.
A productive skill often requires special training. The training modifies the human organism; it changes the fine structure. The conditions brought about through training are involved when the skill is exercised.20 Yet in free-enterprise capitalism, as opposed to an economy involving slavery, the trained laborer contracts freely with the entrepreneur for the use of his or her special skill. The contract is free in at least the respect that the laborer might have contracted for the use of some aspect of his or her labor power that did not involve special training. (There are other ways in which the contract is free that are not specific to the specially trained laborer.)
The point is that the exercise of the power of specially trained labor in the context of a labor market is not, for the individual laborer in the given circumstances, unconditionally caused by the modification of the laborer’s organism due to training. Still, it is conditionally caused by this modification. If the labor power is exercised after its use is contracted for, then the components of fine structure induced by the training are–upon this condition–sufficient for this exercise in the given circumstances.
The following schema reflects the considerations adduced to this point:
The θ’s are components of a–components of a’s fine structure– that are the basis for the causal conditions for realizing the capacity.* Note that dissimilar θ’s may be involved when a θ’s in circumstances of dissimilar K’s. No specific components of the fine structure are selected for the role of basing the capacity; making a capacity claim amounts only to claiming that there be some components in that role. Of course, to say that (Ǝ θ)θa can only mean that some component θ of some individual is such that a has a component exactly similar to this θ. A related remark applies to quantification over K.
This schema tells us that the modal claim Cøa is true provided that in at least some cases when øa would become true it would do so by virtue of the causality of a condition of having some component of a. There is no question of its being true on the basis of a’s having a dispositional property. So the existence of necessities does not require that there be capacities–dispositional properties –but requires that entities have components of fine structure, the having of which accounts causally for some of their behavior.
§4. Causality and Necessity. The notion of causality has not been an important one in the development of the required ontology. It is the more general notion of action that has been important. Actions are paired with results, which as such are not effects (Chapter IX, §3). To be an effect, a result must not only depend on an action but also be necessary in view of what has preceded it. Recall that dependence of results was explained by the factive conditional, which does not imply a necessary conditional even when the context for which the factive conditional holds is built into its antecedent (Chapter IV, §3). It may well be that all results are effects, but this will not be a consequence of our concept of result. In a world that includes action without causation there would be only mere togetherness, whereas in a world with causation there is also necessary togetherness (Chapter VII, §4). The reason, then, that the notion of causation has thus far not been examined directly is that the notions of action and of necessity are sufficient to account for it.
When a condition causes another we need to consider two factors. First, there is an action. It need not be an action of the entity with the causal condition; it may be an action of some entity in the circumstances. But, of course, the having of the action may be the causal condition itself. The knife’s condition of being sharp caused the cut when someone pressed it against the skin. Here the action of pressing belongs to an entity other than the entity with the causal condition of sharpness. But one can also say that the pressing caused the cut and thereby relegate the sharpness to the circumstances. Second, there is a necessary connection. It stands between the causal condition in its circumstances and the result of the action. Putting these together we have:
(2) a’s having θ causes b to have ø if and only if there is a type K of circumstances such that: (i) a’s having θ in circumstances of type K and b’s having ø both obtain; (ii) either circumstances of type K contain an action, or θ is itself an action, that has b’s having θ as a result; and (iii) □ a(θa in circumstances of type K → øb).
When θ is a causal basis for a response øand K resolves into stimulus S on a in K’, then (iii) can be written as □ a(θa→ (Sa in circumstances of type K’ → øb)) Thus by emphasizing the causal role of fine structure in (1), we have not completely lost sight of the role of fine structure in explaining stimulus-response conditionals. However, the stimulus-response conditional here need not be necessary unless a has θ of necessity. In the case of many important capacities, the fine structure is indeed a necessary feature of the individual. But what of cases where this condition is not satisfied? Clearly, in many such cases the modal conditional needed for the stimulus-response model must have concealed in its expression for circumstances a reference to fine structure. The poker must sear the wood when pressed against it provided circumstances obtain in which the poker is hot. A stimulus-response model unmixed with elements from a fine-structure model is, then, utterly implausible.
Schema (2) is for singular causal claims, but the causal component in (1) is generalized for various circumstances of the same unspecified type. A causal claim that is general in respect to the occasions on which the circumstances occur can be dealt with as follows, where (ii) and (iii) are clauses of (2):
(3) In circumstances of type K, a’s having θ causes b to have ø if and only if (ii) and (iii).
Finally, with a modification of (iii), an equivalent can be expressed for the notion of conditional causality:
(4) In circumstances of type K, a’s having θ conditionally causes b to have ø if and only if (ii) and (iii׳) □ a(øb ־֊> (θa in circumstances of type K → øb)).
So the conditionality of causality affects not the factor of action, expressed here in (ii), but it has to do exclusively with the necessary-connection factor.
If the conditional under the sign of necessity in (iii׳) were a ‘material’ conditional throughout, then it would be logically valid. For anything of the form A כ (B כ A) is logically valid. The material conditional cannot be adequate in interpreting (iii׳) since it is not to be merely a logical addition to the other clause in (4). But neither is a ‘strict’ conditional adequate, for it is a necessary conditional. If the →,s in (iii׳) signified necessity, then what would most likely be a contingent fact–the actualization of the effect– would imply a necessary connection between the causal condition and the effect. Yet this seems implausible. In addition, since the first → in (iii׳) is under a modal operator, this operator would be redundant if the→ were itself modal.
But at least if the conditionals were modal, permutation would be blocked. We could not then infer (θa in K→(øb→øb)) from (iii׳). And this might at first seem an advantage, for does not the causal factor θ do more in K than lay the basis for an implication to a tautology? This might not seem much for θ to do if we are thinking in terms of material or strict implication, for any proposition materially or strictly implies a tautology. But we have moved beyond these forms of implication. We can then view the permuted form as the significant claim that the self-implication property is something b has as a result of a having θ in K. It might have it for other reasons as well, but it is informative to have this source of the self-implication property pointed out. Similarly, inertia is perhaps a necessary property of an atom, but this does not mean that an atom’s having it is not due to anything. It may well be due to its configuration of physical parts. Thus, allowing permutation does not change (iii׳) into a mere logical truth. Since ‘relevant’ implication allows permutation and hence lacks the difficulties of a modal implication, it is again a plausible choice for interpreting if-then in the context of the required ontology.
In light of (4), (1) can now be rewritten to emphasize the factors of action and necessity. In doing so, let ‘B’ abbreviate ‘the condition that is a’s having ø’:
(5) Cøa if and only if (Ǝ θ,K)(θa • (either circumstances of type K contain an action or θ is itself an action that has B as a result) • □a (øa in circumstances of type K→ B obtains))).
Only the third conjunct here is modal. So capacity claims may well depend, as was noted in §1, on merely contingent features of entities and thus themselves may be contingent.
To see why ‘B obtains’ has been introduced in the last clause where we would expect simply ‘øa’, let us first look back at (1). There the causal condition was said to cause conditionally the condition that is a’s having ø not to cause conditionally a’s havingø• In rewriting (1) as (5) I want to allow for this distinction. But why is the distinction made at all? Everyone would agree that a die has a capacity to turn up a six, even though it is assumed that there is no feature of the die that causes it, or even conditionally causes it, to turn up a six. But everyone would agree to this only if, when the die turns up a six, there is some feature of it that at least makes it turn up a face. And when it turns up a six, the event of turning up a face is the same as the event of turning up a six. So since there is a feature that causes it to turn up a face, there is, in this case, a feature that causes an event that is the turning up of a six. There can then be a capacity to turn up a six since there is causation of an event that is the same as that envisaged by the capacity. In order to indicate that such an event must be caused and that there need be no cause of six’s being the side that turned up, (1) was written to say that the condition that happened to be a’s having ø is caused.
(a) X causes the condition of b’s having ø.
It can be interpreted as an ‘opaque’ context for the description of the condition. This means that substituting another description of the same condition need not preserve the truth value of (a). Under this interpretation, we agree that (a) is equivalent to:
(b) X causes b to have ø
However, interpreting (a) as a ‘transparent’ context does allow, without a change of truth value, the substitution of another description of the condition of b’s having ø Under this interprétation, we agree that (a) is equivalent to:
(c) X causes the condition that is b’s having ø
If getting a six is a matter of chance, then ‘Jones causes the condition of the die’s having six up’ is true–when a six is actually thrown–under interpretation (c), but not under interpretation (b). In logical terms, (b) interprets the scope of the description for the effect in (a) narrowly, whereas (c) interprets it broadly to include the entire sentence.
It is of utmost importance that we recognize this distinction if microscopic systems are to have capacities for features that are subject to quantum indeterminacies.21 It seems reasonable to say that an electron has a capacity to arrive at a certain spot on a screen at which it is shot. Yet when it hits that spot, it was not, according to present conceptions, determined to do so. Still it may well be that it must hit the screen. And when it hits a specific spot on the screen, its hitting this spot is the same condition as hitting the screen. So one can still use the causal interpretation of capacities in the midst of indeterminacy. This interpretation requires only that the event that is a hitting of the screen at a certain spot be caused. It does not require that the given spot’s being where the screen is hit be caused.
The justification for the sameness-of-condition claims made here is as follows. Where ‘F’ and ‘G’ are related as generic to specific, ‘Fb’ and ‘Gb’ are not made true by different components of b (Chapter III, §1). So ‘b’s having F’ and ‘b’s having G’ refer to the same condition. But if the same component of b is signified by both ‘F’ and ‘G’, how can anything that causes b to have F then fail to cause b to have G? To answer this one must consider the necessary-connection factor in the analyses of causation–(iii) and (iii׳) in (3) and (4). The conditionals (θa in K→ Fb) and (θa in K→ Gb)–to consider (iii) alone–need not both be true. Such conditionals are true or false whether or not it is true that Fb or that Gb, and hence whether or not ‘F’ or ‘G’ signifies any component of b. To decide the truth values of such conditionals, one must, then, consider what components the terms ‘F’ and ‘G’ might signify in b. Since the one is generic and the other specific, ‘F’ might well signify a component in b that ‘G’ does not signify. That is, b might belong to a species other than G in genus F. So what causes F to be a component of b need not cause G to be a component of b. But how then can what causes F to be a component of b cause the condition that is b’s having G to obtain? In this case, we are concerned with the causation of a previously fixed־on particular–the condition that is b’s having G. If ‘F’ signified a property in b such that b could not be G, then there would be no such particular. But since the causal claim here concerns just this particular, such a possibility for ‘F’ is irrelevant for this causal claim.
Similarly, if a is necessarily shaped and if a’s shapedness is identical with its circularity, it need not follow that a is necessarily circular. For otherwise both (N is the nature of a → a is shaped) and (N is the nature of a → a is circular) would be true. But since these conditionals could be true when their antecedents are false, a does not even have to be shaped for them to be true. To decide their truth values, one must consider what ’shaped’ and ‘circular’ might signify in a. Since ’shaped’ might signify squareness in a, the first conditional could be true and the second false.
Though the above refinement is needed in order to apply our view of capacities to indeterministic domains, it does not follow that quantum states are to be interpreted as capacities, in the manner of the Copenhagen school.22 Indeed, since for us capacities are not entities, the two views are fundamentally opposed. The rationale for the interpretation of quantum states as capacities is that the interference phenomena of micro-entities require treating their states as superpositions of incompatible states. Being incompatible, the superposed states cannot all be actual. There is even difficulty with saying that one of them is actual. For how could an actual state interfere with non-actual ones? The conclusion that seems to be the most plausible is that the superposed states are capacities for certain values, such as those of position and momentum, under appropriate measurements. But to insist, as I have been insisting, that there are no entities that are capacities seems, from the Copenhagen perspective, tantamount to insisting that quantum theory could do without quantum states. This would seem to involve abandoning quantum theory.
But is it really plausible to interpret quantum states as capacities? If they are capacities, they are quite different from the ones we have been talking about, for they must have the strange property of interference. One quantum potentiality “may involve or overlap other potentialities.”23 But there is no reasonable extension of the idea of capacity, which has been seen to function as a modality, that involves the idea of interference. The interpretation of quantum states in a superposition as capacities is, then, hampered by the fact that the idea of interference coming from the theory of quantum states must be grafted onto the idea of capacity. Yet the latter idea positively excludes any aspect of interference. So it need not be thought that a Megaric ontology deprives quantum theory of quantum states.
It is, nonetheless, undeniable that a distinctive capacity is associated with each state in a superposition. The coefficient of each state in the superposition gives the probability of actualizing that capacity under a certain kind of measurement. Are these capacities based on fine structure? Does (5) apply to them as it does to other capacities? It applies to them through the use of the transparent causal context which, as we have seen, is needed to deal with indeterminacy. To see that it applies, note that the state function of the system describes the superposition of states that together make up the state of the system. So the capacities associated with the states in the superposition are based on just those features of the system that are referred to in calculating the state function. These features are actual, and hence quantum capacities are grounded in actual components.
Whether an alpha particle in a nucleus has the capacity to peneträte the energy barrier provided by the nucleus when it strikes that barrier depends on the width of the nucleus, the Coulomb repulsion between the alpha particle and the protons, and the short-range nuclear forces. Precisely these features of fine structure are referred to in computing the state function of the particle. This function describes a superposition of the two relevant states–one associated with the particle’s capacity to be turned back at the nuclear wall and the other associated with its capacity to penetrate it. But to avoid the hopeless confusion resulting from holding that capacities interfere, one must distinguish these capacities from the corresponding states in the superposition.
§5. Fine Structures and Natures. Our job is not finished until an account is given of the necessity that appears in our fine-structure analysis. The appearance of this necessity is not a peculiarity of the analysis of capacities. For the problem of necessity arises not just when we consider the scientific realist’s fine-structure analysis of capacities, but also when we consider his account of the explanation of regularities.
The scientific realist explains how it is that entities obey regularities relating operations to responses by reference to their properties, parts, and actions, that is, to their fine structure.24 Entities that obey operational regularities are not, then, empty black boxes. Whether the scientific realist recognizes it or not, there is a consideration of modality involved here. If one explains how entities obey a given regularity by reference to their fine structure, then one must be prepared to hold that entities with such a fine structure must obey the given regularity, at least in the circumstances in which the regularity holds. It cannot be a mere happenstance that entities with this fine structure obey the regularity. Otherwise, one would have to say that the basis for the regularity might well be compatible with the regularity’s being broken in the absence of extenuating circumstances. In short, if one thinks that having the fine structure is only contingently related to the behavior described in the regularity, then one would not explain the regularity by the fine structure. What one would do would be merely to deduce the operational regularity from the propositions that entities satisfying the conditions of the operation have the stated fine structure and that entities with this fine structure manifest the given response. But this purely intentional notion of explanation through propositions, rather than through fine structure, is characteristic of the instrumentalist rather than the scientific realist.
So in accounting for regularities by fine structure, it is supposed that entities with the fine structure must obey these regularities in the given circumstances.25 These regularities are characteristic of entities with such fine structures. Granting that the problem of how entities obey operational regularities is solved by fine structure, the problem arises as to how entities with a certain fine structure must behave in a certain way.
The scientific realist will be forced to reflect that unless one forges beyond fine structure–beyond properties, parts, and actions –one’s advantage over the instrumentalist is negligible. As regards capacities, the instrumentalist is, in the manner of §1, forced to recognize that he cannot account for the modality of the stimulusresponse conditional needed by him to analyze capacities. The scientific realist might appear able to account for this modality by appeal to fine structure. But now he is faced with the unaccountedfor modality of the conditional relating fine structure to behavior. He is faced with this problem both with regard to explaining regularities and to grounding capacities. This time the problem of modality arises at the level of fine structure itself.26
Originally, scientific realism appeared more attractive to us precisely because it showed promise of solving the problem of the necessity of conditionals. But, in view of the seeming epistemological advantages of instrumentalism, it is inevitable that a certain disillusionment should set in over scientific realism once it is realized that it merely creates the same problem over again at the level of fine structure. Scientific realism and the fine-structure model for capacities–(5)–are only half-way houses.
Within the framework of scientific realism, one can at least approach the problem of modal connections in an ontological way. To say an entity has a necessity because of some component of it is not foreign to the framework of the scientific realist as it would have been to his positivist forebearers, who could not think beyond the notion that a necessity is based on intentions rather than on entities the necessity is about.27 The question I am pressing is whether the components of fine structure–properties, parts, and actions–are the right ones for accounting for modality. Is the ontology of scientific realism rich enough? I have suggested that it is not. But I wish to give it one more try before going beyond it.
Perhaps a necessity like the one in the account we have given of capacities can be accounted for by going to a deeper level of fine structure. Of course, at the deeper level there will be further necessities; having the deeper level fine structure will necessarily imply that the surface level fine structure implies the observable behavior. This means going to a still deeper level. There is no one necessity left unaccounted for, even though not all necessities are accounted for at once. The regress appears then to be benign. The problem of modality is solved by not admitting that any level is ultimate in the analysis of fine structure.28
It is surely an inconvenience that, to account for modal connections, we must commit ourselves to the view that there is no end to the complexity of entities. I tried earlier to show that the fine-structure model for capacities is not committed to the view that there is no end to the complexity of entities by expanding the notion of fine structure to cover properties as well as parts (see §2). But that was before we discovered the problem of necessity lurking behind the notion of causality in the fine-structure model. Yet the scientific realist might simply reconcile himself to the infinite complexity of fine structure as the inevitable price for a universe with temporal necessities and hence capacities.
There is, though, another weakness, one to which we cannot reconcile ourselves short of abandoning all standards for explanation. The regress is simply not benign. Suppose we are set the task of explaining the Fness any entity might have. We are then to provide a general account of Fness. Imagine that all entities are ordered in a series without beginning. Each entity, we suppose, is both F and G. There will be two cases regarding the explanation of the component F of any entity in the series.
(i) The F of any given entity in the series can be explained by the fact that the preceding member in the series is G only if it is also the case that this member is itself F. That is, the G in the preceding member explains the F in the original one, not of itself but only if the G is accompanied by F. Thus the F of the preceding member is essential to the account of the F of the original member.
(ii) The F of any given member of the series can be explained by the fact that the preceding member is G. Even if the preceding members having G depends on its having F, the explanation is, this time, possible without reliance on this fact. The F that the preceding member will, by hypothesis, have is accidental to the account of the F of the original member.
I shall, then, call explanations of type (i) ‘essentially repetitive/’ and those of type (ii) ‘accidentally repetitive.’ Both are repetitive since the account of the F of any member brings in the preceding member that, by hypothesis, is itself F. Whether its F is essential or accidental to the explanation, the attempt to explain F generally commits one to explaining F in an infinity of entities.
In earlier jargon, the explanations proceed to infinity per se in the type (i) case, and they proceed to infinity per accidens in the type (ii) case.29 Consider a series, with no beginning, of objects, each of which is heated by a preceding one. The task is to explain the condition of heat in these objects. The explanation would proceed to infinity per se if a preceding object’s condition of having heat were an essential part of the explanation of the succeeding object’s having heat. In fact, there is no reason why an object that heats must itself have heat. So the explanation goes to infinity only per accidens *
An essentially repetitive explanation leads to a vicious regress, though an accidentally repetitive one does not. Suppose that a is F, that the F of b is essential to an account of the F of a, and that the F of c is essential to an account of the F of b, and so on. Can one argue that eventually each F gets accounted for and hence that the regress is benign? One could claim both of these things if the repetition of F were accidental to the explanations. All we would then need be concerned about is that, taken one by one, each F comes to be accounted for.
But with the essentially repetitive explanation there is another factor. This is the dependence of the explanations on one another. There is no explanation of the F of a unless, as part of this explanation, there is an explanation of some component of b–the F of b. For if an explanation of this component were not part of the original explanation, the original explanation would explain one F by another F. It would not then be an instance of a general explanation of F, which we have supposed it must be. So an explanation of the F of b must include an explanation of some component of c–the F of c. And so on. An explanation of the F of a must contain explanations of the infinity of F’s in the series.
Here is where the rub comes in. We could easily say that each F in the infinite series had an explanation if the explanations were not essentially repetitive. In each case, the explanation would be • complete even though each entity referred to had Fג independently of the explanation. Yet in the case of an essentially repetitive explanation, there is no assurance that there is an explanation until it contains an explanation of the F of the entity referred to. Until it contains this, the explanation is merely putative. By going back in the series one fails to get more assurance that the F of a is explained. Rather, by going back, one merely gathers up more putative explanations.
By pursuing the matter through the entire series–assuming this were possible–one would not have come closer to satisfaction. The reason is that a completed infinity of putative explanations does not turn those putative explanations into genuine ones. At any given stage, an explanation depends on prior explanations, and by completing the series one does not tie things off, for one does not come to an explanation that does not depend on prior ones. By completing the series, one has merely given all the putative explanations. One has not eliminated reference to F in the explanation of the F of a. Thus, in the explanation of the F of a, one has failed to instantiate a general account of F. So an essentially repetitive explanation is not an explanation at all in the sense of a general account.
By contrast, in the case of an accidentally repetitive series, the explanation of the F of b is not part of the explanation of the F of a. The explanation of the F of a is, in this case, not merely putative, prior to that of the F of b. Of course, even in this case, if F were inexplicable somewhere in the series, the explanation of the F of a would not be an instance of a general account of F. Still, the one is not part of the other, as in the case where F is essential to an explanation of F. Since essentially repetitive explanations require explanations within explanations and since none of the contained explanations can ever be more than putative since they too require explanations within them, essentially repetitive explanations are viciously regressive. Therefore, they are not explanations at all.
I wish now to show that the scientific realist’s account of necessity by fine structure is essentially repetitive. It thus involves a vicious regress. Suppose □ a(Fa) is true. The scientific realist accounts for the modality involved by the fine structure of a. The relevant component of fine structure is, say, G. The following condition must be met. If G accounts for the modality, then the connection between G and F must itself be modal. So since necessity is an essential feature in the account of necessity, the explanation is essentially repetitive. The explanation of necessity in the series that leads down through deeper layers of fine structure requires at each level a necessary connection between fine structure and some condition that ultimately involves surface level behavior.
At the first level, the modal connection is □ a(Ga→ Fa). If the conditional were only contingent, it would be possible for a to have G but not F. Hence it would be possible that a is not F, which contradicts our assumption that a is necessarily F. The necessary connection, then, does not just happen to accompany the appeal to fine structure, as would be the case in an accidentally repetitive explanation. The repetition of necessity is essential, and thus the account is viciously regressive.
The scientific realist’s position lacks any advantage over the instrumentalist’s in regard to the crucial issue of modal connections. The former fails to provide a general account of necessity that can explain even the modality of the latter’s stimulus-response conditional. Moreover, since capacities involve necessities, this failure sounds the knell for Quine’s optimistic neo-Lockean view that ‘one can redefine water-solubility by simply describing the structural conditions’ of the mechanism of solution.30 Structural conditions–elements of fine structure–fail to account for the necessities involved.
How are we to get beyond the fine structure model of capacities, which is only a half-way house since it fails to deal with modalities? The scientific realist’s ontology needs to be supplemented with natures. There must be a source, which is not itself another property, part, or action, of the modal conditional properties needed in both the explanation of regularities and in the grounding of capacities. Having these conditional properties is to be implied by having natures, in the manner of (2) of Chapter II, §4. Even if there is no end to the levels of fine structure of an entity, it is the nature of the entity that is the basis for the modal connections that the fine structure is involved in at any level.
Salt has an ionic crystal structure with a crystal energy of 183 kcal./mole. It is soluble in water precisely because the attraction between the negative and positive poles of the dipolar water molecule and, respectively, the positive sodium and negative chlorine ions releases enough energy to overcome the 183 kcal./mole ionic attraction in the crystal.31 Salt’s ionic structure is a causal feature in its dissolution, but it is not the ground of the necessity of the causal connection just stated between this feature and its dissolving. Indeed, its ionic structure may be by nature, but it is its nature that gives it the properties and connections it has by nature.
Yet is not an account of necessity by natures essentially repetitive also? We saw in Chapter II, §4, that an entity will have its nature of necessity. Otherwise, it could have different necessities than it has. So natures account for necessities only because they are themselves necessary. This seems to raise the suspicion of an essentially repetitive account. However, the account for the necessity of a nature need not send us in search of a nature at a deeper level. For, as we saw, the nature of an entity accounts for all of its necessities, including the necessity with which it has its nature. There is no series to accommodate an essentially repetitive account, for we stop with the nature of an entity and do not go further to the nature of its nature, and so on.
Perhaps we can stop the series for fine structure as well as for natures. Then the fine structure account of necessity would not be regressive. When G accounts for the modality of □ a(Fa), there always remains the modality of □ a(Ga → Fa) to account for. Instead of going to a deeper level of fine structure, might one not stop the series altogether and account for this second modality by G itself? On the one hand, to employ this device would be to give up the idea of scientific realism according to which one proceeds to deeper and deeper levels of fine structure to explain necessary connections at more superficial levels.
On the other hand, there is a fatal internal flaw in this view. Component G may, unlike a nature, belong to a contingently, even though it implies the necessary property F. For in different possible circumstances, a could have F on the basis of different fine structures. Now, for G to explain the necessity of F, the conditional (Ga → Fa) must be necessary, and for this to be the case Ga must itself imply this conditional. In this way the series is stopped. But then, since being implied by fine structure means here being necessary, Ga will itself be.necessary since it obviously implies itself. This contradicts the evident fact that a may have the component of fine structure G contingently. A non-repetitive account of necessity by fine structure is then excluded.
In blending the fine-structure model with the view that natures are needed in the truth conditions for necessary truths, we get the nature model for capacities:
(6) Cøa if and only if (Ǝ θ,K)(θa • (either circumstances of type K contain an action or θ is itself an action that has B as a result)• (Ǝn)(a has the nature n • (a has the nature n → (øa→(θa in circumstances of type K→ B obtains ))))).
Here as before ‘B’ abbreviates ‘the condition that is a’ s having ø The fine-structure component θ need not be by nature. C. D. Broad calls those capacities that belong to a thing’s nature ‘supreme’ capacities. If an iron bar acquires the component θ that it needs to attract another iron bar only after being placed in a magnetic field, then its capacity to attract iron is not supreme. Still it might acquire this capacity to attract iron when placed in a magnetic field only if it has some other capacity that is supreme.32
However, Broad’s general position that the nature of an entity is a collection of certain of its capacities–the supreme capacities–is not one that can be accepted here. For, according to (6), the required ontology contains no capacities. The role of capacities is played by a complex coordination of natures and other components of entities. The common error of supposing that natures are collections of capacities33 is readily explained. It results from supposing that the content of our concept of a natural kind is the nature of things of that kind. Indeed, we do include, as Locke noted, the concepts of many capacities in the concept we have of a natural kind. But this provides us only with the ‘nominal essence’ of any entity falling under that kind. The nature or ‘real essence’ of any such entity is the basis for any of those features of the nominal essence that are genuinely–not just nominally– necessary.
The basic limitation resulting from such a confusion of nominal essence with nature is that it makes it impossible to offer any account of necessity. How does one account for the necessity of those capacities in the nominal essence that are in fact supreme? Since these capacities constitute the nature, we cannot appeal to anything more fundamental. It would appear then that these capacities would account for their own necessity. Yet to allow this, without further distinguishing supreme capacities, is to allow that any capacity is the basis for its own necessity, even though we have just emphasized that not all capacities are necessary. Thus, if collections of capacities are natures, the ontological basis for the distinction between supreme and contingent capacities, and indeed between the necessary and the contingent generally, disappears.
* For the purpose of the example, I assume local motion is a component. However, genuine components are not relative to coordinate systems in the way local motions are. Since space and time are grounded in actions, and since mere local motion as between individuals is defined by space and time, local motion between individuals must be based on actions that are not simply locomotive. Two simples in relative motion is ontologically impossible.
* By emphasizing the causality of fine structure, our analysis temporarily loses sight of the stimulus-response relation. Here the stimulus is submerged in the circumstances of type K.
* In this example, the explanation was of a condition, not of a component. The explanation of heat, the component, would be quite another matter. Though having heat is, in many cases, caused by another object’s having heat, the component had— the heat—is explained by fine structure—the motions of parts.