The Atlas of Reality
Page 16
The moral of the story seems to be that a thing has the B/A disposition just in case it is the sort of thing that would normally B if A, absent any exogenous interference. We must build into the relevant counterfactual conditionals the further condition that the object in question not be altered in its intrinsic character. This means that the truth of an attribution of a B/A disposition cannot consist merely in the truth of the corresponding conditional ‘if A then B’. Instead, one must have in addition an account of the intrinsic character of the object. Either the disposition itself is part of that intrinsic character, in which case we cannot reduce having the disposition to satisfying the conditional without circularity, or the attribution of the disposition involves other facts about the properties of the object, beyond the bare fact about how it would respond to the relevant stimulus. In either case, the having of a disposition does not consist merely in satisfying the conditional that we ordinarily associate with the disposition.
In 1997, David Lewis proposed a slightly more sophisticated fix for the conditional analysis of dispositions (Lewis 1997):
Lewis's Analysis of Dispositions. x has the B/A disposition if and only if x has some intrinsic property C such that x has C, and, if x were to be made to have A and continued to be C for some finite period of time, the fact that x has both A and C would be sufficient to cause x to have B.
If we try to employ Lewis's “fix,” we have to ask ourselves: why must we hold the intrinsic character of the object fixed, when hypothetically testing for the appropriate reaction? Indeed, why must we suppose that the object has to have any intrinsic property, other than the disposition itself? The Lewisian Strong Hypotheticalist can offer us no explanation of this fact. Therefore, it seems that conditionals are grounded in dispositions, not the other way around. Moreover, Lewis's solution makes use of the idea of causation: the intrinsic property C must (together with A) have the power to produce B. Lewis's solution can help the Strong Hypotheticalist only if the notion of causation can be reduced to conditionals, an issue that we will discuss in Section 27.1. By way of preview, it seems unlikely that such a conditional analysis of causation can succeed.
Therefore, Power Fundamentalism (4.3T) is vindicated, and we have yet another reason (in addition to worries about how conditionals could be fundamental) for rejecting Strong Hypotheticalism.8
In addition to finkishness, there are two additional problems with the hypotheticalist account of dispositions: dispositions can be masked and mimicked. For example, suppose that a fragile vase can be filled with a certain kind of foam in such a way that the foam absorbs all shocks delivered by striking the vase. The vase remains fragile, and yet it doesn't break when struck because the foam masks or acts as an antidote to the manifestation of the disposition (Johnston 1992, Bird 1998). Alternatively, suppose that an illusionist has set up a sonic beam that would cause a sturdy rock to break whenever it is tapped. The rock isn't in fact fragile and never becomes really fragile, but the illusionist causes the rock to mimic the disposition (Smith 1977, Prior, Pargetter, and Jackson 1982, Lewis 1997, Armstrong 1997). Cases of masking and mimicking are in fact quite common (Fara 2005).
Notes
1. We will treat ‘counterfactual conditional’ and ‘subjunctive conditional’ as though they were synonyms. It is permissible to use a subjunctive conditional even when one is not sure that the antecedent is false (especially when it concerns the future), but it is always odd to use it when one believes the antecedent to be true.
2. There are two niceties omitted from this characterization of possible worlds semantics. First, we need the rules that take us from the truth-values of the atomic sentences to the truth-values of all sentences. Second, we need an accessibility relation on the domain of worlds. Neither of these is necessary for understanding what we're up to here, and thus the complications they would introduce are needless.
3. We can generalize this by making * a function from world-proposition pairs to worlds and changing the definition of counterfactual truth to:
Stalnaker′. ‘(p q)’ is true at w if and only if q is true according to *(w,p)
4. Again, there are niceties here that we ignore. For ease of exposition, we assume “strong centering,” that the actual world is the only world in the innermost sphere.
5. The result is similar to the way that modern logic (following Gottlob Frege) treats “vacuous” quantification. If there are no unicorns, then modern logic counts both ‘all unicorns are ugly’ and ‘all unicorns are beautiful’ as true.
6. Some conditionals involve more than one individual, such as ‘if the forbidden fruit had been a kumquat, Adam and Eve wouldn't have eaten it’. A Molinist would have to suppose that in such cases, the habitude is a special kind of relation holding between Adam and Eve. In general habitudes could involve any number of individuals, perhaps even infinitely many.
7. One important thing to bear in mind: the sense of ‘closeness’ or ‘similarity’ that is used in Lewis's semantics for the counterfactual conditional is a special, technical sense. The words ‘close to’ or ‘similar to’ are terms of art. We don't look for those worlds that are most similar, all things considered, to the actual world, as Lewis's response to Jonathan Bennett makes clear (Bennett 1984, 2003). Consider the proposition: if the President were to press the nuclear button, all of civilization would end. This proposition is probably true, even though worlds in which the consequent are true are very dissimilar (we hope) from the actual world.
8. Bonevac, Dever, and Sosa have proposed altering the semantics and logic of the conditional in such a way as to avoid the refutation of the conditional account by finkish dispositions (Bonevac, Dever, and Sosa 2006). Their new “normality” conditional does indeed seem to correspond in the right way to the predication of powers and dispositions, but it seems clear that, metaphysically speaking, their work only confirms the fact that powers and dispositions are fundamental.
5
Laws of Nature
5.1 Strong Nomism: The Dretske-Armstrong-Tooley (DAT) Theory of Laws
As we mentioned above, Fred Dretske (1977), David M. Armstrong (1983), and Michael Tooley (1977, 1987) have all proposed that the truths about the laws of nature are metaphysically fundamental, consisting in a primitive, unanalyzable relation of “necessitation” holding between two or more properties or universals. This is Strong Nomism (4.4A.2). According to Strong Nomism, the laws of nature determine which counterfactual conditionals are true, and they also determine which powers and tendencies particular things have. For example, if it is a law of nature that whenever an F encounters a G in relation R, the G becomes H, then we can say that the law of nature confers a corresponding active power on all F's and a corresponding passive power on all G's. Similarly, if it is a law of nature that all F's become H's within a certain span of time, then the law confers on all F's the immanent power of making themselves H's within that span of time.
We also noticed that laws of nature can be probabilistic and “oaken” or exception-permitting. Such probabilistic and oaken laws can be thought of as conferring tendencies of the appropriate kind to the relevant objects. Thus, it seems that laws can indeed ground the truths of both conditionals and the attributions of power.
Let's take a closer look at the connection between laws and powers. In particular, we need to be clear about the issue of priority: are the laws grounded in truths about powers or vice versa? If the laws are supposed to be fundamental, with the powers of particular things grounded in them, how exactly is this supposed to work? What is the logical form of the laws of nature, from which we are supposed to be able to derive the attributions of powers to particular things? One option has the laws of nature directly state that certain kinds of things have certain powers and tendencies. On this view, we can certainly derive the attribution of powers and tendencies to such things from those laws. But it would seem that the laws are then made true by the particular attributions, not vice versa. For example, if it were a law of nature that all fire
has the power to burn wood, then we could infer that this fire has the power to burn that wood. But if we ask what makes it true that all fire has this power, it would seem that this universal generalization (‘all fire…’) is made true by its instances or by the absence of counterexamples (fire without the power to burn wood). Either way, the attributions and non-attributions of powers to particular things would ground the truth of the law, not the other way around.
As a second option, suppose the laws of nature merely say that all instance of fire's being brought into proximity to wood are followed in time by the burning of the wood. This is logically independent of the attribution of any power to the fire, so the law of nature could be independent of and prior to the particular attributions. However, now we are no longer able to deduce that fire has the power to burn wood from the law. The law merely says that the placing of fire and wood in proximity to one another is always followed by the wood's being burned. It says nothing about any instance of fire's having the power to burn the wood, which is what is needed.
Dretske, Armstrong, and Tooley adopt a third option concerning the logical form of laws of nature. The laws of nature neither ascribe powers to particular things, nor do they merely describe the unvarying sequences of events. Instead, a law of nature attributes a special kind of relation, nomic necessitation, to an ensemble of properties or universals. This view, the Dretske/Armstrong/Tooley theory of laws, is the most prominent and plausible Strong Nomist position in the literature. Through the remainder of this chapter, we will treat Nomism as committed to the DAT theory.
Return to the fire and wood case. The supposed law of nature would state that the nomic-necessitation relation holds between the following pair of complex properties: (1) being some wood brought into proximity to some fire, and (2) being some wood that is burned. As van Fraassen (1987, 1989) has pointed out, however, this leaves it somewhat mysterious how we are supposed to derive any information about particular cases of fire and wood from the fact that this supposed relation holds between the two properties. We want there to be some connection between the attribution of powers to things and the actual sequence of events.
Strong Nomists must simply posit a brute necessity between the holding of a law and the corresponding universal generalization. Armstrong was reluctant to admit this, since he was attempting to build a metaphysical system with no brute necessities whatsoever, but this was a quixotic quest. As we've seen, it is a corollary of Ockham's Razor (PMeth 1.2) that we should prefer a theory with fewer ad hoc postulations of necessities. Where necessities must be posited, we should prefer theories that derive the necessity from the essential structure of the things involved. Nomism offers no such explanation of the necessary connection between laws and generalizations, so this must be counted as a cost of the theory.
The main compensating advantage is that Nomism provides a metaphysical explanation of the difference between lawful and accidental generalization. A lawful generalization is one that corresponds to a relation of nomic necessitation connecting the relevant properties, while accidental generalizations have no such counterpart. However, this advantage is not unique to Nomism, as we shall see. A Strong Powerist account can make a similar distinction between generalizations that are supported by powers and those that are not.1
A key question for Nomism is this: is the relation of nomic necessitation an internal relation among properties involved? Here again is the definition of an internal relation:
Def D2.2 Internal Relation. R is an internal relation if and only if necessarily, for every x and y, whether R holds between x and y depends only on the intrinsic properties of x and of y.
A non-internal relation between two things depends on more than the intrinsic qualities of the two things. Here is an example of each kind of relation. If one book is twice as long as another, then they are related by an internal relation, the relation of being twice as long as. As soon as we know the length of each book (where length is intrinsic to each book), we know whether or not they stand in that relation. In contrast, if one book is higher than another in the Amazon.com sales ranking, they are thereby not internally related. You could know all there is to know about the intrinsic features of the two books without knowing which ranks higher than the other on Amazon. We will discuss internal relations in more detail in Chapter 18, on the structure of space.
It seems clear that the Nomist must count the nomic relation as an non-internal relation. If it were internal, then, assuming that properties have their intrinsic features essentially, the properties involved would have active and passive powers in their very natures. Thus, powers would be among the fundamental features of the world. This would fit with a Strong Powerist picture and not with the Strong Nomist one.
If the nomic relation is not an internal relation, is it a contingent or necessary one? If it is necessary, then its holding with necessity in each case would seem to be brute necessity, with no possible explanation. Similarly, however, if it is a contingent relation, its holding contingently would also seem to be something for which there could be no further explanation. It would be a brute contingency.
As we do in the rest of science, we should prefer the simple metaphysical theories to ones that are overly complicated. This is Ockham's Razor (PMeth 1). An aspect of Ockham's Razor is that we should minimize the brute external relations posited by a theory, whether necessary or contingent. Such a brute relation is one that cannot be explained in any way, whether by the structures or essences or internal character of the relata or by means of some cause.
PMeth 1.3 Third Corollary to Ockham's Razor. Other things being equal, adopt the theory that posits the fewest inexplicable and uncaused non-internal relations between things.
On Nomism, the relation of nomic necessitation between properties is inexplicable and necessarily uncaused. A causal connection presupposes the existence of a corresponding causal law. Hence, it is impossible to give a causal explanation of all of the causal laws. For example, the Nomist cannot suppose that God has caused the laws of nature, since for God to have the power to do so, there would have to be a law of nature giving him that power, and God could not have caused that law to hold without vicious circularity. Since the relation of nomic necessitation is external, instances of it cannot be explained by appeal to the structure or internal character of the properties involved.
Why do Nomists reject the idea that laws are made true by an internal relation between properties? Their motivation seems to have involved a desire to avoid the supposed obscurities and mysteries of an Aristotelian or scholastic essentialism. Since the seventeenth century, prominent philosophers and scientists (including Descartes, Galileo, Boyle, Locke, and Hume) have rejected the idea of essential powers as “occult” and vacuous. The French comic dramatist Molière famously poked fun at scholastic essentialism when he has a scholastic alchemist (in his play “The Imaginary Invalid”) “explain” the power of narcotics to cause sleep by hypothesizing that narcotics have a “dormitive virtue” or “virtus dormitiva” (which is simply Latin for ‘the power to cause sleep’). However, it is far from clear that Nomism involves less, rather than greater, mystery.
To sum up, from the perspective of simplicity and economy, Nomism suffers from four defects, as compared with either Strong Powerism or Neo-Humeism:
Nomism must posit an additional, irreducible relation of nomic necessitation between properties.
Nomism must posit that whenever the nomic-necessitation relation holds between two properties, its holding between them is a brute, inexplicable, and uncaused fact.
As a consequence of (1), Nomism must treat properties as real entities, capable of entering into relations.
Nomism must posit a brute necessity between each law of nature and the corresponding generalization, whether universal (e.g., all F's are followed by G's) or statistical (e.g., in x% of the cases, F's are followed by G's).
There are three additional objections to be lodged against Nomism:
There are a varie
ty of logical forms that laws of nature can take. Few, if any, have the simple form of a universal generalization involving just two universals, like (x)(Fx -> Gx).
Laws might involve disjunctions or conjunctions in various places, like (1) or (2):
(1) (x)((Fx & Gx) -> Hx) Everything that is both F and G must be H.
(2) (x)(Fx -> (Hx or Jx)) Everything that is F must be H or J.
As Tooley (1977) recognized, Nomists need a distinct necessitation relation for each logical form. In the simple case, we have one relation, N1, holding between F and G and entailing (x)(Fx -> Gx). In the case of laws (1) and (2), we would need two additional necessitation relations, N2 and N3, with N2 holding between F, G and H and entailing (1), and N3 holding between F, H, and J and entailing (2). If there are a large number of laws with a wide variety of logical forms, Nomism could be saddled with positing a large number of distinct necessitation relations.
Nomism must explain why we tend to accept simple scientific theories about the laws of nature. Whenever we find the data fitting to a simple pattern (such as a smooth, mathematically simple curve), we come to be quite confident that the data are to be explained by a correspondingly simple law of nature. We believe that gravitational force varies inversely with the square (power 2) of the distance, not with the distance to the power 2.000003, for example. This preference for simple laws is hard to explain on Nomism. What reason would we have for thinking that the nomic-necessitation relation is more likely to hold between two simple properties than between two complex ones? As we shall see, the Neo-Humeist has a neat explanation of this fact.2