Introduction: the nature of singularist theories of causation.

Few topics in modem philosophy have been as centrally focused on the work of a single philosopher as has the theory of causation been focused on David Hume. Discussions of causation since Kant do not start with Aristotle or Plato as is common for other topics; rather, they proceed from Hume's theory. (1) Given this tradition in philosophy, one of the striking characteristics of work on causation in the last three or four decades has been the willingness to depart from fundamental aspects of Hume's views. Indeed, such work often explicitly puts the causal 'glue' back in just where Hume took it out. While this reversal of Hume is evident in many generalist, counterfactual, or probabilistic theories of causation, it is most evident in singularist theories. For these are theories that would return our concept of causation to being about a connection between particular events or particular states of affairs, a 'cement' (in Hume's celebrated wording) holding together our universe. It thus seemed an apt time to invite the proponents and critics of singularist theories to update the state of the discussion on these theories.

In this Introduction I seek to introduce the topic without entering the debates about the correctness of singularist theories. I do this by addressing three preliminary concerns. First, I seek to clarify what is commonly meant by calling a theory of causation a singularist theory. Second, a taxonomy of different types of singularist theories is laid out. Third, a very general overview is given of the kinds of arguments, pro and con, that have been addressed to singularist theories of causation.

I. WHAT Is A SINGULARIST THEORY?

The natural contrast case for singularist theories of causation are generalist theories of the relation. Before coming to the kinds or types of singularist theories on offer (the second topic of this introduction) and the arguments pro and con such theories (the third), we do well to revisit the singularist/generalist distinction. Doing so will clarify what we mean when we classify a theory of causation as singularist.

We should distinguish two brands of singularism, an extreme form and a moderate form. The axis of difference between these two forms is the degree of commitment to there being causal laws whenever there are singular causal relations. The extreme view, championed by Elizabeth Anscombe, (2) holds that singular causal relations could (and even do) exist even if there were no true causal laws that connected types of events of which these causes and effects were instances. A moderate singularism, by contrast, concedes that no singular causal relation can exist where there is not some true causal law connecting events of these types.

Both of these are recognizable forms of singularism about causation. Singularism can be, but does not have to be, the view that there are or can be lawless causal relations. What makes a theory singularist is not to be found along this axis of (either necessary or contingent) accompaniment. A theory can hold that every singular causal relation can (and even must) be accompanied by some casual law, and still be a singularist causal theory by my lights. This is because the more pertinent axis here is not accompaniment, but reduction.

In taxonomizing theories of causation, the introduction to a recent collection rightly observes that "the most useful distinction to make at the outset is that between accounts that do and accounts that do not attempt to reduce causal facts to facts about what happens, together with facts about what the laws are that governs what happens." (3) Put in my language, this is the distinction between accounts that do, and accounts that do not, reduce statements of singular causal relations to statements of instantiated laws (where the latter are taken to be statements of laws together with statements that particular events exist when those events are instances of the types of events connected by such laws). Singularism can be best seen as the account that does not reduce singular causal statements to statements of causal laws. Such a non-reductionist-defined singularism can then be agnostic about whether such laws are always present when there exist singular relations.

The generalist about causation who reduces singular causal relations to laws is not a two-way reductionist. Generalist theories regard the laws as basic, and singular relations as derived from these. A natural anti-reductionism is thus to reverse what is basic and what is derived. This view regards singular causal relations as basic, and laws simply as inductively derived generalizations from these. Chris Hitchcock calls this the "generalization strategy" about laws, (4) which is a strategy pursued by one kind of singularist theory of causation.

Another kind of singularist theory, however, is to grant parity between singular causal relations and causal laws. According to this second kind of non-reductionist singularist theory, there are two different kinds of causal relationships, one between types (the relation of laws) and the other between tokens (singular relations). (5) To my mind, this position too is singularist: it admits that singular causal relations are not to be identified simply as instantiated causal laws (even if it also admits that causal laws are not to be identified simply as inductively derived generalizations of singular relations). Each, on this view, is equally basic.

The difference between these two kinds of singularist theories does not lie in their ontological commitment to the existence of singular causal relations. (It is this common commitment that makes them both singularist.) Rather, the only difference lies in how they regard causal laws (as either equally basic in their own right, or as derivative of the more basic singular relations.)

A difficult test for this reductionist (to laws) criterion of singularism is provided by theories like that of David Armstrong. Early on, Armstrong defended the view that causal laws are essentially primitive relations between universals. (6) Much later, however, Armstrong speculated what an overt reductionism (of singular relations to laws) would look like when married to his views of laws as primitive relations between universals. (7) The upshot, he concluded, was a kind of singularism, even though there is a reduction to laws. As Armstrong cryptically put it, "the law will be present completely in each instantiation," and hence, "singular causation will be a completely intrinsic relation." (8)

A metaphorical way of putting Armstrong's thought would be to say that the 'glue' of causation, its necessitating power, resides as much in the singular relation as in the relation between types (the subject of a law). (9) Indeed it is the same 'glue.' It is as completely present in singular causal relations as in causal laws. One is not more basic than the other, not because these are different but equally basic relations; rather, the mysterious ('primitive') element is the same thing in each relation.

Even though reductionist, this by my lights counts as a singularist theory of causation. As we shall see, Armstrong's view here differs little from the primitivist singularism of Michael Tooley. (10) It too should count as a singularist theory of causation, just as Armstrong himself so classifies it.

II. TYPES OF SINGULARIST THEORIES

In clarifying singularist theories generically, we have along the way taxonomized singularist theories along two axes of differentiation. One was the axis of accompaniment, the extreme singularist denying that causal laws necessarily must be present if singular causal relations are present (and even denying that laws in fact are present for all existing singular relations), the moderate singularist admitting the omnipresence of laws with singular relations and perhaps even admitting the impossibility of there being singular relations with no accompanying laws.

The second axis of differentiation between singularist theories was in terms of reduction. Admitting (as does the moderate singularist) that causal laws always do or must accompany singular causal relations, it remains to ask which is more basic. To be a singularist at all, one must reject the generalist view which reduces the singular to the general. But two kinds of singularism can make this rejection: the generalizing singularist (who regards the singular relations as basic and the laws as derived) and the dualist singularist (who regards singular relations and laws as distinct, but equally basic, relations).

Now we need a third axis of differentiation, along another variable also (unfortunately) called 'reductionist/anti-reductionist.' (11) Here the singularist who is also an anti-reductionist in this second sense is a kind of primitivist. His slogan might be that 'each (at least metaphysically primitive) thing is what it is and no other thing.' The denial is not the denial of any reduction of singular relations to instantiated laws. Rather, the denial is much broader: singular causal relations cannot be reduced to anything else.

We shall come back shortly to primitivist kinds of singularism. But at this juncture it is more enlightening to examine the reductionist views the primitivist is denying. These are views that seek to reduce singular causal relations to singular relationships of other kinds. The most familiar of these reductionist theories are physicalist in the sense that the processes, qualities, or quantities to which causation is to be reduced, are all physical processes, qualities, or quantities. There is quite a variety here. One of the simplest is Humean in its inspiration: with Hume, identify causation as the succession of one event by another, but then, contra Hume, insist that simple spatio-temporally contiguous succession of event-tokens (rather than types) is all that causation amounts to. (12) Left out of this account, of course, is anything that looks like causal glue. The only connection between a cause and an effect, on this view, is that of immediate temporal succession and proximate spatial juxtaposition. This singularist version of Hume is subject to many of the objections that have been made to Hume's regularity account. Indeed, the 'accidentally true generalization' objection to Hume's theory is heightened against such a singularist Hume, for now 'post hoc, ergo propter hoc' is true on this view for single successions in time, with no requirement of regularity of succession.

Physicalist singularisms that are more discriminating in their reduction bases prominently include transference theories, according to which causation is the transfer of some quantity from one object to another. The quantity transferred is a physical quantity--either energy and/or momentum, (13) or (more vaguely) different kinds of quantities of which momentum, velocity, kinetic energy, and heat are examples, (14) or the even vaguer notions of physical force and matter in motion. (15)

The idea of such transfers of physical quantities fits quite well with familiar paradigms of causation, such as one billiard ball hitting another; where one change immediately begets another. Such transfers look more mysterious if they are across larger spatio-temporal regions. For such extended causal relations, causal chains need to be created, made up of a series of events or states of affairs, each pair of which is spatio-temporally contiguous. As Bertrand Russell put this assumption, "when there is a causal connection between two events that are not contiguous, there must be intermediate links in the casual chain such that each is contiguous to the next, or (alternatively) such that there is a process which is continuous." (16)

This Russellian denial of action at a distance poses a problem for any theory according to which causation consists only of these sorts of transfers, whether direct or indirect across a chain of direct transfers. The dangerous condition cases, for example, long bedeviled force/energy theorists like Beale and Epstein in law, for no matter how dangerous a condition may be, and no matter how culpable may be a defendant for allowing it to continue, it is difficult to see how such persistence (sameness of state) can be a transfer of anything across long spatiotemporal intervals. (17) More broadly, all transfer theories are bedeviled by their difficulty in accommodating the 'quiet times' in causal chains, i.e., the times during which objects simply persist. As Phil Dowe notes, "there appears to be a type of causation ... that is ruled out by the transference theory ... [which is] persistence as causation." (18)

Other physicalist theories respond directly to this worry. One is the causal process theory, a theory mostly associated with the work of Wesley Salmon. (19) According to Salmon, causal influence is propagated by causal processes, and causal processes in turn are conceived of as spatio-temporally continuous sequences. The inertia of a bullet, for example, is such a process, as is the persistence of a physical object over time. For processes, according to Salmon, are those things that display consistency of structure over time, and causal processes are things displaying such consistency of structure when the structure in question is some local modification (what Salmon calls a 'mark').

Notice that the process theory is tailor-made to deal with the action at a distance worry because causal processes, so conceived, allow causal influence to be propagated with ease across the 'quiet times' (such as stable but dangerous conditions) of causal chains. Yet causal processes (as persisting structures), by themselves, do poorly in accounting for changes that introduce modifications in structures. The idea of a causal process thus has to be supplemented with some other idea to account for production and initiation of such changes, and this Salmon does with his idea of single case propensities (of which, more below).

A well-known variation in the process family of theories is the conserved quantity theory of Phil Dowe. (20) On this theory, a causal process is the space-time points making up the history of a physical object through which some quantity is conserved according to the laws of science, and a causal initiation or production is where two objects exchange such a conserved quantity. Like Salmon's theory, the conserved quantity theory sees the need to accommodate both change and persistence to maintain spatio-temporally continuous links in its causal chains.

A third sort of physically reductionist singularist theory also emphasizes persistence, but the persistence is not that of structure (Salmon) or of the conserved quantity of some object. (Dowe). Rather, what persists are those concrete universals that are often called 'tropes', 'abstract particulars', or 'concrete universals'. (21) A trope is the instance of a property possessed by a particular object, such as the whiteness of some particular white car. The idea is that a trope literally persists across temporal intervals; a trope of a car's whiteness at h is the very same thing as the car's whiteness at h. Causation, on this view, is a kind of property persistence (as it is for other reductionist singularisms), but for tropists this means trope persistence. Thus, the whiteness of some car at [t.sub.1] persists as the very same whiteness trope at [t.sub.2], and the propagation of causation consists just in this trope persistence. As Douglas Elating notes of this form of causal process,

   Unchange as well as change falls within the causal structure of the
   world. Unchanging persisting tropes are no less important as causal
   processes because of their simplicity. Indeed, this form of causal
   process is pervasive. (22)

If one steps back from the trees to the forest, it is plain that common to all such theories is the conception of causation as involving spatio-temporally continuous events or states of affairs. It is this idea that drives such theories both to think that 'causing is physical producing' (23) and to think that causing is also the persistence of something, be it a structure, a property, an object, or a trope. (24) It bears repeating that on these physically reductionist accounts both the producing and the persisting involve physical quantities, objects, properties, and tropes.

It is important to see that there are other reductionist singularisms that are not physicalist in the way these theories are. Indeed, these other reductionist singularisms are close cousins of generalist versions of the counterfactual theory and of the probabalist theory. Despite their counterfactual/probabilist natures, they, too, should be considered singularist.

Consider first the counterfactual theory of causation. On one well-known view of conditionals generally, (25) the sentence, 'if p, then q' does not typically assert a proposition that is conditional in form; rather, it makes a conditional assertion of an unconditional proposition, namely, q. As Stalnaker/Lewis applied this treatment of conditionals generally to counterfactual conditionals (where we know the supposition of the statement is false), this results in an assertion that q is true in some possible world that is close to the actual world (save that p is there true). (26) Conjoin this account of the truth conditions of counterfactuals with David Lewis's trenchant modal realism about possible worlds (27)--they are as actual to their inhabitants as our world is to us--and the counterfactual theory of causation can be seen as a reductionist, singularist theory.

What makes such a theory reductionist is that causal relations are reduced to counterfactual dependencies. What makes such a theory singularist is that causal laws play no essential role in the truth conditions of the relevant counterfactuals. (28) "The short circuit caused the fire" is made true by the non-existence of the fire in that close possible world where the short circuit also is absent. Again, this is not a matter of extending the laws of our world to what would happen in this close possible world; rather, we use Lewis's similarity metric to isolate the relevant possible world and then we simply 'look' to see whether the fire is occurring in this close possible world.

What makes such a counterfactualist/singularist theory hard to see is the difficulty people have in accepting a full-blown modal realism, one that accepts the reality of possible worlds that exist in a sense more robust than as mere projections of the laws of our actual world. David Armstrong, for example, settled on an interpretation of Lewis according to which "in his theory of causation the possible worlds enter as mere calculational devices." (29) The singularist/counterfactualist I am imagining is much bolder: for him, the truth makers for counterfactual statements (and thus for causation on this theory) are the possibilia, those real particulars that are actual in close possible worlds even though they are not actual in ours.

Another reductionist, singularist theory is a probabilistic one. Unlike generalist probabilistic theories, this theory does not reduce causation to probabilistic laws. Rather, causation is identified as single-case chance-raising. The truth maker for 'c caused e' does thus not involve the probabilistic law, 'events of type C raise the chances of events of type E.' Rather, it is the chance of this particular event e occurring that was raised by c, another particular. The reduction base here is wholly singular.

Of course, like the singularist counterfactualist theory just imagined, such a singularist/probabilistic theory is ontologically very expensive. A holder of such a theory is committed to but one of the going interpretations of probability, (30) the propensity theory (for if he used the more standard relative frequency theory he would be using probabilistic laws ultimately for his reduction base). Moreover, propensities would be real properties of token events; a horse would have four hooves, a long nose, and a good chance of winning the Derby, all equally as its properties. As David Armstrong remarked about Hugh Mellor's version of such an ontology of chances, "those who live in such ontological glass houses should throw few rocks at others." (31)

We have thus far said little about anti-reductionist, singularist theories. These are theories of causation that take it to be in some sense 'primitive' or 'basic.' There are three sorts of 'primitiveness' we should distinguish: (1) Some concept will be analytically primitive if it cannot be analyzed in terms of some other concepts. To say that causation is primitive in this sense is to say that the concept of causation cannot be defined or otherwise analyzed in terms of other concepts. This is because causation is so basic within our conceptual scheme. As John Carroll put this view, "with regard to our total conceptual apparatus, causation is at the center of the center." (32) Somewhat earlier Michael Scriven put it this way: "The concept of cause is fundamental to our conception of the world in much the same way as the concept of number: we cannot define it in terms of other notions without ... circularity." (33)

(2) Some object, quality, or relation will be epistemically primitive if it is known non-inferentially; for empiricists this means known by direct observation, either by the ordinary senses or by our acquaintance with inner processes such as the pressure we feel on our skin when an object presses against us, or our willing of changes in the world which we then bring about. For Hume, of course, causation was not basic epistemically; for Hume famously held that all we ever observe is the occurrence of one event followed by another. Many contemporary philosophers have disputed this, holding either that in certain special cases such as willing we experience causation directly (34) or that in ordinary cases (such as watching a stone shatter a glass) we see the causing as much as the objects and their motions. (35)

(3) Some object, quality or relation will be metaphysically primitive if the type of which it is an instance is not identical to any (putatively different) type of thing. Causal relations will thus be metaphysically basic if there can be no reduction (or even any non-reductionist supervenience, if that is possible) to any non-causal properties or relations.

It is this third sense of 'basicness' that characterizes the primitivism about causation that opposes reductionism. Although arguments have been made from the supposed indefinability or non-observability of causation to the metaphysical conclusion, (36) it is the latter conclusion that defines primitivism for our purposes. (37)

The advantages of a metaphysical primitivism about causation should be obvious. In a nutshell, primitivism frees us from having to deal with all the counterexamples adduced against generalist theories, counterfactual theories, and reductionist singularisms. Such counterexamples juxtapose our pre-theoretical causal intuitions against the seeming dictates of the preferred reduction bases for causation--counterfactual dependency, chance-raising, lawful sufficiency, regular concurrence, energy-transfer, trope-persistence, etc. By contrast, primitivism offers up no non-causal facts with which to compare our causal intuitions. Causation can be just what those intuitions say that it is, on the primitivist view.

Such flexibility comes with an equally obvious price, however, and that is the aura of mystery that surrounds all metaphysically basic things. It may be that 'a cause is a cause' is the most that can be said, but that does not make such a truth very informative. Consider in this regard the reception of such causal primitivism within legal circles. Early in the last century Jeremiah Smith despaired of finding any helpful legal test for causation. (38) He did this explicitly on grounds of analytic and epistemic primitivism, for he thought causation to be indefinable but that jurors would 'know it when they saw it.' (He may also have thought causation to be metaphysically primitive, but he did not make his metaphysics dear, probably even to himself.) He therefore proposed the intentionally circular 'substantial factor' test, according to which some act e is a cause of some harm e if and only if c is 'a substantial factor in the bringing about of e.'

This test, picked up by the first two iterations of the American Law Institute's Restatement of Torts, has long been criticized for its circularity and its vagueness. (39) Yet such criticism ignores the explicit primitivism that underlay the test. 'A cause is a cause' was all the jury could be told, and all they needed to be told, on such primitivist views. (The only thing the test adds is a measure of amount: only substantial causes make one liable.) The reaction to such primitivism in legal circles is symptomatic of the 'not informative' charge leveled at all primitivist theories.

III. THE PROSPECTS FOR A SINGULARIST THEORY OF CAUSATION

Whether singularist theories of causation are viable depends initially on which brand of singularism one has in mind. I find either form of extreme singularism--according to which singular causal relations either do or at least can exist in the absence of any pertinent causal laws--implausible. Properties play too important a role in both events and states of affairs as causal relata to countenance extreme singularism. (40)

The choice between the generalizing and the dualist singularist is closer, depending only on how the singularist wishes to judge causal laws. My own sympathies are with the generalizing singularist, according to which causal laws are mere constructions inductively arrived at by generalizing over what is more basic, viz, singular causal relations. The dualist, by contrast, faces an extra task at explicating the nature of laws even after he is done with explicating singular causal relations. Moreover, this extra task must overcome the puzzle of how the supposedly different causal-law relation between types of states of affairs relates to the singular causal relation that exists between token states of affairs. Watching Davidson and Davidsonians undertake similar extra tasks (for explanations versus relations) does not give one confidence that it is doable without residual mystery. (41)

The choice between the reductionist and the primitivist--our third taxonomical axis above discussed--is closer still. My suspicions are that even primitivists start with a presumption in favor of reductionism of some kind. We all prefer more informative analyses to less. But if no reductionist analysis ends up being plausible, yet singularism is to be preferred to alternative theories of causation, then primitivist singularism becomes one's theory by default. One might call this a reluctant primitivism. Yet one's backing into such a position can be comforted with the thought, common to primitivists, that if anything is primitive, causation is.

Wherever one comes out on these debates inter se between singularists about causation, there remains the generic question of whether singularism itself competes favorably with the generalist (nomic sufficiency, probabalist, Humean regularist) and counterfactualist theories of causation. Those latter theories have had a familiar battery of arguments directed against them. (42) Vis-a-vis singularism, there are two ways of viewing the (roughly seven) arguments against these competing theories of causation. The first would be to say that these are merely negative arguments against these other theories, so that the fact that singularism survives such objections itself constitutes no positive reason to believe in singularist theories. On this view, surviving such arguments produces at most a kind of 'back-door singularism,' a singularism that has only this to be said in its favor: since we have to adopt some theory of causation, and since the taxonomy of possible theories is exhausted by generalist, counterfactualist, and singularist theories, singularism wins by default.

Yet such a 'back door' construal of what can be said in favor of singularism has to make good on both of the conditions just expressed: is the taxonomy of theories complete? (43) And, perhaps various forms of skepticism--such as having no single theory of causation, (44) or having no theories of causation at all (no matter how pluralistic), (45) or having no use for any concept of causation at all (46)--are to be preferred to singularism.

My own take on the standard arguments advanced against non-singularist theories is more positive for singularism. For those arguments have proceeded from firm intuitions about causation that generalist and counterfactual theories have difficulties in meeting. Just one example: the arguments developed from cases of pre-emptive causation. We all know (with a certainty that borders on the a priori) that in pre-emptive cause cases (such as where two fires advance on the same house, each sufficient to destroy it, but one gets there first), the second fire did not cause the destruction of the house because the first fire did. Any theory that cannot accommodate this conclusion is a bad theory of causation. But more than that, any theory that generates such a conclusion is supported by that fact. It is, in the old language, confirmed.

The seven arguments I have in mind are these:

1. General overbreadth in the proposed reduction base for causation

Uniformities in nature that are 'accidental generalizations,' raises in conditional probability, nomically sufficient conditions, and counterfactual dependencies, all seemingly exist where there is no causation. This is not, on its face, a problem for singularist theories. Primitivists of course have the license to shape causation to their pre-theoretical intuitions, so their theory cannot suffer from overbreadth. Physicalist-reductionist singularisms, by contrast, are hostage to the nature of their various reduction bases; yet such singularist theories are quite narrow in the reduction bases to which they would reduce causation. On their face, they do not appear overbroad. Indeed, one of the worries about such theories is that they are too narrow, excluding the causal relations asserted to exist in history and the social sciences generally.

2. Promiscuity and de minumus causal contributions

The remote and the insignificant can all be causes on generalist and counterfactualist theories. Whereas a scalar primitivism or a quantitative reductionist singularism can handle this problem easily. Causation can 'peter out' over extended chains, and some co-present causes can be much smaller than others, on any scalar version of singularism.

3. Negative Causation

Absences can be causes, they can be effects, and they can be causal intermediaries (where they are both an effect of an earlier cause and cause of a later effect), on generalist and counterfactual theories. By my lights, that counts against such theories. (47) Failing to prevent (omitting), preventing, and preventing preventions, are not causal, and the fact that generalist and counterfactual theories count them as causal counts against such theories. One of the great strengths of most singularisms--namely, all forms of singularism that deny action at a distance--is that they get it right in how they classify omissions, preventions, and double preventions as non-causal.

To be sure, some critics regard double preventions as the Achilles Heel of all forms of physically reductionist, and most forms of primitivist, singnlarism. Common intuition indeed treats double preventions as causal, except in that subclass of double preventions commonly called 'allowings' or 'letting die.' Yet a properly regimented common intuition--one recognizing that omissions and preventions cannot be non-causal unless double preventions are also non-causal, and which recognizes that there is no causal distinction between double preventions which are allowings and double preventions which are not--must at the end of the day treat double preventions as non-causal.

4. Epiphenomena

No theory of causation can afford to collapse the distinction between epiphenomenal relations between events (which are co-effects of a common cause), and causal relations between events. Yet generalist and counterfactual theories have well-known difficulties in doing this. Singularists of all stripes have no such problem. There is no casual chain between co-effects of a common cause on any of these theories, making them invulnerable to objections along these lines.

5. The Asymmetry and Directionality of Causation

Relations that permit symmetry, such as nomic sufficiency, chance-raising, regularity of concurrence, or counterfactual dependence, have a difficult time accounting for the necessary asymmetry of the causal relation. Relations that are not temporally ordered (the same group) also have difficulty in being the reduction base for a relation that seems to follow the direction of time. On its face, the temporally ordered, asymmetrical relation of causation looks a poor fit with relations that are neither. Singularists do not face these problems. Primitivists, of course, can simply endow their primitive relation with temporal direction and asymmetry and be done with it. Physicalist reductionist singularists have to do a bit more work, here as elsewhere. The reduction bases need to have these two features, but that seems unproblematic. Transfer, for example, occurs in one direction, and time-forward to boot; and persistence can be defined the same way.

6. The Transitivity of the Causal Relation

Moving to fine-grained states of affairs (rather than Davidsonian whole events) as causal relata alleviates much of the pressure transitivity puts on counterfactual and generalist theories. But not all. Generalists and counterfactualists have residual difficulties in accommodating the seeming transitivity of causation. (48) One can of course deny that causation is a transitive relation, and in a harmless sense this is true: causation can peter out over extended chains so that a cause of an early link is too de minimus a contributor to be a cause of a later link. But this scalarity-based limitation on transitivity is not what is needed by the generalist and the counterfactualist, (49) so they still have problems in accounting for causation's transitivity, even so limited.

Singularist theories might also seem to have a problem here, given their common (although not universal) denial of action at a distance. One might construe such a denial to also be a denial of any causal relation between spatio-temporally remote events even when such events are connected by a causal chain. In some sense of 'cause,' this will be true: when c causes d, and d causes e, then c cannot be a cause of e in the same sense that c is a cause of d, on these theories. Singularists need the distinction (ancient in Anglo-American law) between direct causal relations, and indirect causal relations. Events like c and d are related directly, but c causes e only indirectly. Causation should thus be defined, Lewis-style, disjunctively: either as direct causation or as the end points of a causal chain, each link of which is a link of direct causation. This will suffice for preserving transitivity for singularists in the way that a like move by Lewis helps in the preservation of transitivity for counterfactualists.

7. Overdetermination Problems

The overdetermination cases bedevil both generalist and counterfactual theories. In the symmetrically concurrent cause cases, the counterfactualist theory gives the wrong answer (neither is a cause), and the probabilistic theory gives the wrong answer when the combined causes lower the probability of the effect (vis-a-vis that probability given only one of the causes). In the mixed concurrent cause cases, where each cause is neither necessary nor sufficient, both the nomic sufficiency and the counterfactual theory give the wrong answer. And in the pre-emptive cause cases, the counterfactual theory holds neither to be the cause, the nomic sufficiency theorist regards both as causes, and the probabilistic theory seems committed either to both being causes (if they are each chance-raising), or neither being causes (if they are chance-lowering), where in truth only the pre-emptive factor is a cause.

Singularist theories slice through these cases like the proverbial hot knife through butter. There are continuous causal chains for each factor in all of these cases, save for the pre-empted factor in pre-emptive cause cases (which strong causal intuition does not regard as causal). Indeed, it would be fair to say that these cases motivate singularism as strongly as they demotivate generalist and counterfactualist theories. We know what the answers are here and it confirms a theory of causation if it yields these answers in these cases by implication.

The case for singularism is thus a strong one. What might be said against it? Several items are prominent in the literature. One stems from a different view on negative causation. If one thinks that omissions, preventions, and double preventions are all causal in nature, (50) then one will reject all forms of singularism (which is most of them) that require continuous causal chains, each link of which are pairs of spatio-temporally contiguous events. For such singularisms cannot 'skip over' absences and treat the more remote ends of the chain as standing in the cause/effect relation.

A second worry focuses on physically reductionist forms of singularism. The worry is that this seems to commit one to a rather grand vision of unified science. On their face, for example, the causal relations discussed in the social sciences do not seem to involve continuous physical processes; they relate things like the Great Depression to things like World War I. (51) One has to think that economics, history, sociology, etc., all can be reduced to physics to hold the view that causation is nothing but continuous physical processes and yet that social science discovers genuinely causal truths. Such singularists, in other words, have to abandon the caution of those like Phil Dowe who offer up only a theory of physical causation; if theirs is to be a truly general theory of causation as such, they must embrace an old ideal, what used to be called the ideal of unified science.

A third worry is ontological. Some forms of singularism are committed to 'queer' entities, qualities, or relations. Singularist counterfactualists-- those who embrace Lewis's full blown modal realism so that possibilia exist as robustly as 'actualia'--are one example. Another are those singularist probabilists who embrace ontologically primitive, single-case chance-raisings. To many, Armstrong's primitive relation between universals and Tooley's primitive singular relation look ontologically excessive as well. Physically reductionist singularisms, of course, are not queer in this way; their whole point is to reduce causation to some more familiar physical items. Yet such ontological respectability is bought at a price: often such reductionists give up the ambition to analyze causation (or even physical causation) as such. Rather, their stated ambition is to give an account of causation as it exists in our world, with our laws. (52) To critics of such singularisms, this is to give up the quest that a theory of causation should be undertaking.

Fourthly, there is Hume's old worry, a worry we might dub that of 'epistemological (as opposed to ontological) queerness.' The worry is how one could know of the existence of singular causal relations when our evidence is so limited. We can observe the existence of events, their spatiotemporal location, and thus, their order of succession in time; we can't observe any glue-like relation between them. On the verificationist view that our concepts should never exceed our evidence, Hume would thus conclude that causation is nothing other than the succession of events that we can observe.

None of these worries comes without replies by singularists. Negative causation in all its forms is regarded as an anathema by many. That most forms of singularism cannot accommodate it is thus no objection. Action at a distance is also an anathema, so that my own bets are indeed with the reductionist vision of a unified science that that implies (as long as that vision makes due allowance for possibilities of emergence at the levels of laws for social sciences, and so long as one is not overly optimistic about the scientific status of some of the shibboleths of contemporary social science). Some ontologies are strange enough to wish to avoid them, although many doubt that primitive causal relations are among them; and if the form that is taken by a theorist's ontological modesty is that of metaphysically necessary truths, that is not too lacking in ambition to be called a theory of causation. Finally, epistemologically speaking, singular causal relations are not obviously queer. Not only does the theory-ladeness of perception (so studied these last fifty years) make what is observed versus what is inferred an unclear distinction, and not only are there good arguments for why we do observe causal relations, but the entire verificationist impulse behind Hume-style objections is now commonly rejected. There is a 'gap' between the evidence we possess, on the one hand, and the thing evidenced on the other, in all areas of knowledge--phenomenal properties and physical objects, behavioral dispositions and the psychological states of other people, the facts of the present and the facts of the past. The verificationist finds the gap intolerable and so identifies the thing evidenced with the evidence for its existence. Yet this impoverishes science, as is now widely recognized. If we can live with such 'gaps' about induction, other minds, the past, and external objects, seemingly we can do the same with causation.

It is hoped that the essays collected here advance the discussion on singularist theories of causation. Authors were invited to refine already extant singularist theories, articulate new ones, or critique the correctness of any such theories or even their possibility. The essays that follow amply fulfill that hope.

Michael Moore

University of Illinois College of Law

Champaign

NOTES

(1.) This is true of all of the papers in this volume, although explicit only in Jessica Wilson's "Resemblance-based Resources for Reductive Singularism."

(2.) G.E.M. Anscombe, Causality and Determination (Cambridge: Cambridge University Press, 1971). Anscombe was arguing explicitly against Donald Davidson's view that singular causal relations presupposed causal laws, even if we did not know on a given occasion what the law might be. Bence Nanay, "The Properties of Singular Causation," this volume, interprets Davidson not to hold the view that it is only in virtue of an event's properties that it can cause anything (even though the latter view is a common basis for Davidson's moderate singularism about laws).

(3.) John Collins, Ned Hall, and L.A. Paul, eds., Causation and Counterfactuals (Cambridge, MA: 2004), 12. This is also the dimension relied on in Michael Rota's "An Anti-reductionist Account of Singular Causation," this volume, in his discussion of the character of singularist theories.

(4.) Christopher Hitchcock, "The Mishap at Reichenbach Fall: Singular vs. General Causation," Philosophical Studies 78 (1995), 258.

(5.) See id. for citations to examples in the literature.

(6.) David Armstrong, What Is a Law of Nature? (Cambridge: Cambridge University Press, 1984).

(7.) David Armstrong, "Going Through the Open Door Again: Counterfactual versus Singularist Theories of Causation," in Collins, Hall, and Paul, eds., Causation and Counterfactuals, 456.

(8.) Ibid.

(9.) Put orally by me to Armstrong at the University of Illinois Conference on Causation and Responsibility, Mt. Hood, Oregon, November, 2006.

(10.) Michael Tooley, Causation: A Realist Approach (Oxford: Clarendon Press, 1987). See also Tooley, "The Nature of Causation: A Singularist Account," in David Copp, ed., Causation Philosophers: Celebrating Twenty Years of the CJP, Canadian Journal of Philosophy Supp. 16 (1990), 271-322.

(11.) Collins, Hall, and Paul lump this sort of reductionism together with a reduction to laws in their classificatory scheme. They nonetheless see that certain singularist theories may be non-reductionist about laws while reductionist about certain facts of nature, facts which "appeal to fundamental laws only indirectly." Causation and Counterfactuals, 13.

(12.) A reading of C.J. Ducasse, "On the Nature and the Observability of the Causal Relation," Journal of Philosophy 23 (1926), 57-68.

(13.) David Fair, "Causation and the Flow of Energy," Erkenntuis 14 (1979), 219- 50.

(14.) Jerrold Aronson, "On the Grammar of 'Cause'," Synthese 22 (1971), 135-56.

(15.) The early speculations in these directions by legal theorists, notably Joseph Beale and Richard Epstein. See Beale, "Recovery for Consequences of an Act," Harvard Law Review 9 (1895), 80-89; Beale "The Proximate Consequences of an Act," Harvard Law Review 33 (1920), 633-58; Epstein, "A Theory of Strict Liability," Journal of Legal Studies 2 (1973), 151-204.

(16.) Bertrand Russell, Human Knowledge (New York: Simon and Schuster, 1948), 491.

(17.) Thus Beale made much of objects coming to rest, but such resting ends causal influence only when they come to rest "in positions of complete safety." Beale, "The Proximate Consequences of an Act." Similarly, Epstein added to his notions of force and energy an idea of energy stored in unstable positions making it a continuant of causal influence via such "dangerous conditions." Epstein, "A Theory of Strict Liability."

(18.) Phil Dowe, Physical Causation (Cambridge: Cambridge University Press, 2000), 61.

(19.) Wesley Salmon, Scientific Explanation and the Causal Structure of the World (Princeton, NJ: Princeton University Press, 1984); Salmon, Causality and Explanation (Oxford: Oxford University Press, 1998).

(20.) Dowe, Physical Causation. See also Dowe, "Absences, Possible Causation, and the Problem of Non-locality," this volume. A critical review of Dowe's conserved quantity theory is Christopher Hitchcock, "Problems for Conserved Quantity Theory: Counterexamples, Circularity, and Redundancy," this volume.

(21.) See, e.g., Keith Campbell, Abstract Particulars (Oxford: Basil Blackwell, 1990). As David Armstrong cheerily notes, "each happy discoverer, it seems, names them anew." Armstrong, A World of States of Affairs (Cambridge: Cambridge University Press, 1997), 21.

(22.) Douglas Ehring, Causation and Persistence (Oxford: Oxford University Press, 1997), 122. See also Ehring, "Abstracting Away from Preemption," this volume.

(23.) Jonathan Schaffer, "The Metaphysics of Causation," Stanford Encyclopedia of Philosophy, http://plato.stanford.edu. (August 13, 2007), 15.

(24.) Jonathan Schaffer, "Causes Need Not Be Physically Connected to Their Effects: The Case for Negative Causation," in Christopher Hitchcock ed., Contemporary Debates in Philosophy of Science (Oxford: Blackwell, 2003), 203-04.

(25.) See Quine, Methods of Logic, 4th ed. (Cambridge, MA: Harvard University Press, 1982), 21.

(26.) Robert Stalnaker, "A Theory of Conditionals," in N. Rescher, ed., Studies in Logical Theory (Oxford: Blackwells, 1968); David Lewis, Counterfactuals (Oxford: Blackwell, 1973).

(27.) Lewis, On the Plurality of Worlds (Oxford: Basil Blackwell, 1986).

(28.) A query that Jaegwon Kim introduced to discussions of Lewis, in Kim's early reaction to Lewis's theory. See Kim, "Causes and Counterfactuals," Journal of Philosophy 70 (1973), 570-72.

(29.) Armstrong, "Going Through the Open Door," 445.

(30.) Donald Gilles helpfully summarizes the going theories as to the semantics of the probability calculus in his Philosophical Theories of Probability (London: Routledge, 2000).

(31.) David Armstrong, "The Open Door: Counterfactual versus Singularist Theories of Causation," in Harvey Sankey, ed., Causation and Laws of Nature (Dordrecht: Kluwer Academic Publishers, 1999).

(32.) John Carroll, Laws of Nature (Cambridge: Cambridge University Press, 1994), 118.

(33.) Michael Scriven, "Defects of the Necessary Condition Analysis of Causation," in W. Dray, ed., Philosophical Analysis and History (New York: Harper and Row, 1966), 358, reprinted in part in E. Sosa and M. Tooley, eds., Causation (Oxford: Oxford University Press, 1993), 56.

(34.) E.g., Armstrong, "Going Through the Open Door," 454; Evan Fales, Causation and Universals (London: Routledge, 1990), Chapter 1.

(35.) E.g., Anscombe, Causality and Determination. See also Ducasse, "On the Nature and the Observability of the Causal Relation."

(36.) One might argue, for example, that because we can observe causation directly, therefore the concept of causation is indefinable, and from indefinability we move to lack of identity. Anscombe, for example, probably wished to make both of these moves.

(37.) Metaphysical primitivists about causation include David Armstrong, "Going Through the Open Door;" Michael Tooley, Causation: A Realist Approach; John Carroll, Laws of Nature, and James Woodward, "Supervenience and Singular Causal Statements," in Dudley Knowles, ed., Explanation and Its Limits (Cambridge: Cambridge University Press, 1990), 211-46.

(38.) Jeremiah Smith, "Legal Cause in Actions in Tort," Harvard Law Review 25 (1911-1912), 103-28, 223-52, 303-27.

(39.) See, e.g., Dan B. Dobbs, The Law of Torts (St. Paul, MN: West Publishing, 2000), 416 ("The substantial factor test is not so much a test as an incantation. It points neither to any reasoning nor to any facts that will assist courts or lawyers in resolving the question of causation.") See also Richard Wright, "Once More into the Bramble Bush: Duty, Causal Contribution, and the Extent of Legal Responsibility," Vanderbilt Law Review 54 (2001), 1080 ("As a test for determining ... causal contribution ... the substantial factor formulation is completely useless.")

(40.) I argue this in Moore, "Causal Relata," Sharon Byrd and J.C. Joerden, eds,. Philosophia Practica Universalis: Festschrift for Joachim Hruschka, Annual Review of Law and Ethics 13 (2005), 589-641, revised and reprinted as chapter 15 of Moore, Causation and Responsibility (Oxford: Oxford University Press, 2009).

(41.) For doubts about the Davidsonian project of regarding generalizing explanations as very different than singular relations, see Peter Menzies, "A Unified Account of Causal Relata," Australasian Journal of Philosophy 67 (1989), 64-67. Notice that the view of Armstrong and Tooley--that causal laws are the same primitive relation between universals as relates token states of affairs--does not involve one in this extra task.

(42.) These are deployed against counterfactualist and generalist theories in Moore, Causation and Responsibility, chapters 17 and 19, respectively.

(43.) One might object, for example, to the exclusion of the powers/disposition theory of causation. See George Molnar (as edited by Stephen Mumford), Powers: A Study in Metaphysics (Oxford: Oxford University Press, 2003); Stephen Mumford, Laws in Nature (London: Routledge, 2004). On the other hand, some powers theorists such as Stephen Mumford ("Passing Powers Around," this volume) regard such accounts as primitivist.

(44.) See Chris Hitchcock, "Of Humean Bondage," British Journal for the Philosophy of Science 54 (2003): 1-25.

(45.) Brian Skyrms regards our concepts of causation to be an "amiable jumble" of different ideas that resist unifying theory. "EPR: Lessons for Metaphysics," in P. French, T. Uehling Jr., and H. Wettstein, eds., Midwest Studies in Philosophy Vol. IX (Minneapolis, MN: University of Minnesota Press, 1984).

(46.) Bertrand Russell, "On the Notion of Cause," in J. Slater, ed., The Collected Papers of Bertrand Russell: Vol. 6, Logical and Philosophical Papers 1909-1913 (London: Routledge, 1992).

(47.) Moore, Causation and Responsibility, chapter 18.

(48.) Explored in Ned Hall, "Causation and the Price of Transitivity," Journal of Philosophy 97 (2000), 205; Jerome Schaffer, "The Metaphysics of Causation," 9; and Christopher Hitchcock, "The Intransitivity of Causation Revealed in Equations and Graphs," Journal of Philosophy 98 (2001), 273-99.

(49.) As Ned Hall points out, in his "Causation and the Price of Transitivity."

(50.) As does, for example, Jonathan Schaffer. See his "The Case for Negative Causation."

(51.) A worry fleetingly expressed in Collins, Hall, and Paul, Causation and Counterfactuals, 14.

(52.) As in Phil Dowe, Physical Causation.

Gale Document Number:A198472478