Twin Causes 3

8/14/2019 Twin Causes 3

1/21

Twin Causes

0 Introduction

In the traditional rubric of two-place causal relations, let me say that aMultiple Antecedent Causal Relation (henceforth MACR) has more than

one event in the causal role.1 The received view is that there are at mosttwo varieties of MACR: joint-causes and (for some philosophers) overdeter-miners. This paper contends that there is a third, twin causes.

Failure to recognize this third MACR has inspired a number of implausi-ble views, many of them skepticalabout non-living macroscopic objects2,about dispositions3, about mental causes4. In general, these arguments pur-port to establish skepticism about higher-level causes, and they all sharea common form5:

(P1) a is a sufficient cause of e

(P2) b is a sufficient cause of e

(P3) a does not cause b and b does not cause a

(P4) e is not overdetermined

Hence:

(C) a = b

1Throughout this paper, I write as though events are the causal relata. If it is insteadthe case that tropes, facts, or some other entities are the causal relata, it would not muchaffect the kernel of my argument; examples and many details, though, would demandamendment.

2

[Mer03]3 [Mum98,PPJ82]4 [Kim93b,Kim95,Kim98,Sos95]5See [Low00,Rit05], [Sos95,Kim97,Pea79,Mum98,PPJ82,Kim89,Stu05,Kim93b,Kim95,

Kim98, LL87], [Pap93], [Dre89], [Mal68], [Mer03]

1


2/21

But this deduction of C falsely presupposes that there are no twin causes.

For twin causes a and b are both sufficient for an effect, e, and neither is acause of the other, but they do not overdetermine e. And twin causes arenot identical.

In all cases in which an effect e is not overdetermined by antecedentevents a and b while a and b are sufficient causes, I say that a and b arecausal equivalents. Where it is nonetheless the case that a b, I say thata and b are twin causes. The crux of my argument is that causal equivalentsa and b act as a single cause of an effect, whether a = b or not. 6 Thereis thus a hidden assumption in P1-C: if a and b are causal equivalents withrespect to a particular effect, then a = b; i.e., there are no twin causes. Callit the Principle of the Identity of Causal Equivalents, ICE:

ICE If a and b are causal equivalents with respect to a particular effect,then a = b.

I believe that ICE is implausible on its face. In conjunction with LeibnizsLaw, it claims that if (i) a and b share one actual causal propertybeing asufficient cause of eand if (ii) e is not overdetermined, then a and b shareall of their properties, including their other causal properties, and includingtheir modal properties. If ICE is false, this papers central claim is justified:there is a third variety of MACR, twin causation.

Here is the plan for the paper. First, in order to clarify the sort ofcase at hand, I give a plausible example of twin causes: an event involving a

macroscopic object at a time is a twin cause with an event involving its partsat that time. Second, I distinguish twin causes from (i) overdeterminers,(ii) joint-causes, and (iii) the case in which a genuine cause is necessarilyaccompanied by an epiphenomenon. In 3, I rehearse and refute JaegwonKims arguments in support of the reasoning expressed in P1-C above.

1 An Example

a and b are causal equivalents if and only if there is some effect e which aand b bring about by acting as a single cause. Let me now define acting asa single cause for an effect e:

6I intend the phrase act as a single cause to be a convenient shorthand for thepurposes of discussion; it is not a metaphysical commitment: I am not committed to theclaim that events perform actions, and I am not committed to the claim that there is onlyone cause in the relevant cases, whatever that might mean beyond the claim that twincauses do not overdetermine their effects.

2


3/21

SC a and b act as a single cause of e iff (i) both a and b are sufficient to

cause e and (ii) a occurs iff b occurs.

Where a = b and a (b) is a cause of e, this is uncontroversial. But if, inaddition to acting as a single cause, a b, then a and b are twin causes.

Consider a weight, o, and its left and right halves, o1 and o2, respec-tively.7 Suppose that putting o on a scale caused it to tip at time t. Callthis effect event e, and let P be the property or set of properties involved inbringing e about, lets say exerting force x on the scales pan.

Finally consider two Kim events. Kim events are triples of an individ-ual, a constitutive property instantiated by the individual(s), and the timeat which this instantiation occurs: i, P, t. And event1 i,P,t = event2i1,P1,t1 iff i = i1, P = P1, and t = t1. [Kim93a] Using the individuals and

defined above and P, then, we are interested in the two distinct events: a= o,P,t and b = o1 + o2, P, t. a is os exerting force x at t and b is osparts exerting force x at t. I use halves o1 and o2 here for convenience; theexample may be rewritten for subatomic particles o3 on or for whateverindividuals are in fact os parts.

Clearly, a occurs iff b occurs. And a tips the pan of the scale iff b does,so each is sufficient to cause e and necessary for the other. That is, a andb act as a single cause of e. It is dubious, however, that a = b. If a wereidentical to b, then o would be identical to o1 + o2; and if this were true,then it would be true that a composite individual is identical to its parts.That is, the thesis known as composition as identity would hold. At least

one advocate of the arguments given above, Trenton Merricks, explicitlydenies composition as identity. He provides a reductio: if o = o1 + o2, thenidentity sometimes holds one-to-many. But this is absurd. Merricks deniescomposition as identity on these grounds, and thus he would believe that a b.8 On Merrickss own view, then, a and b should be twin causes.

One neednt agree with Merricks to agree that a and b are twin causes,though. The relation between an object and its parts is undoubtedly inti-mate, and many philosophers have claimed that it is loose identity; andmany have denied that the relation is strict identity while affirming thatcomposites are nothing over and above their parts. These phrases arefrustratingly elusive, but perhaps the concepts were developing here are il-

luminating. Ted Sider has set out eleven desiderata for the relation of parts7I am grateful to XX not only for supplying this example but also for showing very

clearly that it is apt for my purposes here.8 [Mer03]: 21.

3


4/21

to wholes9; I argue elsewhere that the relation obtaining between the indi-

viduals in twin causes meets Siders demands. We might say by extensionthen that the relation between a and b is loose identity.10 The relation be-tween twin causes may be loose identity, but we shall see that it needntbe.

1.1 Merricks v. Macroscopic Objects

Assuming that e is not overdetermined, P1 - P4 are true of the aforemen-tioned events a, b, and e. If adding ICE to this argument makes it valid,and if ICE is true, then a = b, contra the conclusion we drew above. I takethis as evidence that ICE is false.

Trenton Merricks has proposed a similar but far more shocking argu-

ment as regards macroscopic objects and the microscopic particles of whichtheyre composed. [Mer03] His idea is that if macroscopic objects exist, theycause the same effects as do their microscopic components and they are notidentical to those components. In our terms, an event involving a macro-scopic object at a time is a twin cause with an event involving that objectscomponent microscopica at that time. But Merricks assumes that if an ef-fect has more than one cause, then it is overdetermined. He also assumesthat the effects of macroscopic causes are not overdetermined. And so heconcludes that macroscopic objects do not exist.

Merrickss premises are similar to P1 - P4. Where a is an event involvingany non-living macroscopic object and b involves the microscopic particles

of which that object is composed, Merricks replaces P1 with P1*: if a (b) exists, then a causes e. On this basis, he concludes that the antecedentof P1 is false; that is, no event involving a non-living macroscopic objectexists, and no macroscopic objects exist.

(P1*) (x)(x = a) (a b & a is a sufficient cause of e))

(P2) b is a sufficient cause of e

(P3) a is not a cause of b and vice versa

(P4) e is not overdetermined

9

[Sid07]: 7110In a separate paper, Causal Indiscernibles, I specify a relation somewhat strongerthan causal equivalence called causal indiscernibility. All causal indiscernibles arecausally equivalent, but it may be that not all causal equivalents are causally indiscernible(depending on the nature of causation). My exact view is that composition is causalindiscernibility.

4


5/21

Hence:

(C*) (x)(x = a)11

I shall prove below that events acting as a single cause do not overde-termine an effect, and we have said that where two non-identical events actas a single cause, they are twin causes. Thus, twin causes are two suffi-cient causes that dont overdetermine an effect. But Merricks assumes thatif an effect has two or more sufficient causes, then it is overdetermined.Thus he assumes that there are no twin (or triplet or quadruplet . . . ) causes,and the dubious elimination of macroscopicaor the events involving them,anywayseems to follow.

We can see the same point by considering ICEs contrapositive: if a b,

then a and b are not causal equivalents. That is to say, a and b do not actas a single cause if a b, which is to say that there are no twin (or tripletor quadruplet. . . ) causes. And hence, if a and b are sufficient for the sameeffect, then that effect is overdetermined. I propose, then, that ICE is thehidden, perfidious premise here too.12

Perhaps C* is not absurd, but this argument strikes me as a reductioagainst the conjunction of ICE and P4. On my view, the proposition that aand b are twin causes of e is plausible on its own, and in this case it has theadditional virtue of entailing that ICE is false. Thus, I propose that P1* -P4 are true, but C* does not follow. Rather, a and b do not overdeterminee because they are causal equivalents.

The concept introduced here, twin causation, has given us (i) a plausibleobjection to Merrickss argument against the existence of non-living macro-scopic objects and (ii) a plausible view of a relation between these and theirmicroscopic constituents.

11For brevitys sake, I have replaced Merrickss premise that the two purported causesare causally irrelevant to one another with the simpler and weaker P3; twin causesalso satisfy Merrickss original premisetwin causes are indeed causally irrelevant to oneanother on Merrickss account of causal irrelevance. (57) I have also elided Merrickss rea-soning from the (alleged) epiphenomenalism of macroscopica to their elimination. Thesedifferences have no bearing on the present paper.

12See [Mer03]: 58. The Causal Principle endorses the inference from an effect es havingtwo mutually causally irrelevant causes to es being overdetermined. Twin causes indeedmeet the criteria for overdetermination that Merricks delineates on that same page.

5


6/21

2 Comparison

We now distinguish the case in which a and b are twin causes from thosecases in which (1) f and g jointly cause an effect, (2) i and j overdeterminean effect, and (3) l causes an effect and m is an epiphenomenon of l . Briefly,

joint causes are both insufficient to cause an effect; overdetermining causesare both sufficient for an effect and at least one of them is unnecessary for theother; where one of two apparent causes is an epiphenomenon with respectto an effect, one of the two is merely an apparent cause of the relevant effect.Where a and b are twin causes, by contrast, each is necessary for the otherand sufficient to cause the effect, and both actually do cause it (by actingas a single cause).

Let us now articulate these distinctions in more detail.

2.1 Joint Causes

Consider the case in which f and g jointly cause an effect. Let us definethe compound event h as follows: h = (f & g). That is, h is the event ofboth f and g occurring. And let h be a cause of e: for whatever theory ofcausation is true, let h and e satisfy the criteria for hs being a cause of e onthis theory.

Now consider two of the ways in which this situation might obtain. (i) fand g are each sufficient to cause e; (ii) neither f nor g is sufficient to causee. In case (ii), f and g are joint causes of e. In case (i), the two causes may

be twin causes or they may overdetermine the effect. Let us suppose fornow that they are twin causes.Consider again events a and b. The former is a weight, os, instantiating

property P at a time; the latter is os parts instantiating P at the same time.These suggest another tidy example. Suppose the scale is set up so that ittips only if force x is applied to the pan. Let the scales tipping at a timebe effect e. As defined above, event a is the weights exerting force x on thepan at time t. As we saw above a and b act as a single cause of e. Eachis sufficient for the effect and necessary for its twin. By contrast, an eventinvolving either half without the other is insufficient to bring about e; twosuch events would be joint causes.

2.2 Overdeterminers

If e has two sufficient causes, however, then e seems to be overdetermined.But this is not without exception: twin causes do not satisfy a necessary

6


7/21

condition for overdetermination.13 Where i and j overdetermine an effect e,

it is necessary that either i or j is unnecessary for the other.Consider the familiar firing squad case of overdetermination. The con-

demned man, call him Garcin, is brought before the line of executioners:and then Ready. . . Aim. . . Fire! Suppose that two of the shots strike himsuch that each would, under normal circumstances, be fatal. It seems thatGarcins death is overdetermined.

The germane point is as follows: of the two shots, if either had failedto occur, then the other would still have been fatal. The telltale signs ofoverdetermination are the situations modal properties. In particular, it isthat the absence of one or the other of the causes would not have changedthe outcome. I take it that this is what being overdetermined consists in.

Compare two other events sufficient for Garcins death. Dub one of thebullets o; then call its left half o1, and its right half o2. Suppose o struckthe bridge of Garcins nose with such and such a force at time t, and it wasfatal. Call this event a, and let b be the event of o1 + o2 striking Garcin attime t with the same force. a is sufficient for Garcins death iff b is. Do aand b overdetermine Garcins death?

My intuition is: of course they dont. This intuition is justified by aplausible account of the nearest worlds in which either a or b fails to occur.The nearby worlds without a are without b and vice versa. Each is necessaryfor the other in the nearby worlds, and so if this shot had been the solesufficient cause of death, then Garcins death wouldnt have occurred inthe nearby worlds without a or in the nearby worlds without b. My intuition

is that this death would not have been overdetermined. And I submit thata and b act as a single cause of this death, and since a b for reasonssimilar to those rehearsed above for the weight and its parts, a and b aretwin causes.

Karen Bennett proposes an ostensibly lower bar for overdetermination:

e is overdetermined by c1 and c2 only if:

(O1) ifc1 were to occur without c2, e would still have occurred:(c1 & c2) e, and

(O2) ifc2 were to occur without c1, e would still have occurred:(c2 & c1) e ([Ben03]: 476)

Bennett justifies O1 and O2 as necessary conditions by the following rea-soning:

13Karen Bennett ([Ben03]) and Jesper Kallestrup ([Kal06]) have recently argued similarpoints about overdetermination in addressing the problem of causal exclusion.

7


8/21

The main reason is simply that they capture the reasoning we

engage in when we want to distinguish cases of genuine overde-termination from cases of joint causation, or from cases in whichone of the putative causes is not really a cause at all. Let c1 andc2 be the shots fired by two members of a firing squad, and e bethe victims death. If we needed to decide whether or not thedeath was overdetermined, we would ask precisely whether thesetwo counterfactuals are true. . .If the answer to both questions isnoif both counterfactuals are falsethen the death was notoverdetermined, for it was jointly caused by the two gunshots.If only one of the counterfactuals is false, at most only one ofthe gunmen is guilty. . .It is hard to see how [these counterfactu-

als] couldfail

to be true when the relevante

is overdetermined.([Ben03]: 477)

Finally, Bennett adds that not only must O1 and O2 be true in actuality,both must be true nonvacuously. If there is no causally possible world inwhich, say, the antecedent of O1 is satisfied, then O1 is true only vacuously,and thus c1 and c2 do not overdetermine e. This is the case of course if, say,c1 cannot occur without c2 (c1 c2); similar considerations apply for thereverse (c2 c1).

Thus, Bennetts necessary condition fails to hold for twin causes if eithera b or b a. Plausibly, both of these are true for the bullet and its halves,and so both O1 and O2 come out vacuous; and so e is not overdetermined;

and so a and b are twin causes.But I also think that Bennetts condition is over-cautious. Its true

that if its impossible for one twin cause to occur without the other, thentheir effect is not overdetermined; but weaker relations between twin causespreclude overdetermination as well.

Bennetts insight is that two causes a and b overdetermine an effect eonly if each of a and b would have caused e in the absence of the other. Ifits simply impossible for a to occur without b or impossible that b occurwithout a, then, yes, it is not possible that each of a and b would have causede in the absence of the other. But we neednt concern ourselves with thefull range of possibilities between a and b, for were interested specifically

in whether a and b overdetermine e in actuality. And so were interestedonly in the possibilities pertinent to this particular causal relation in ourparticular world. It neednt be, then, that a always occurs with b or viceversa throughout all possible worlds, but only in those worlds that reflectas, bs, and es actual causal properties.

8


9/21

So which are the possibilities pertinent to a particular, actual causal

relation? We dont know, since we dont yet know what causation is. Buton the best-known theories, the sum is far short ofall the possibilities. On aLewisian analysis of causation, they are simply the nearest relevant worlds,and so if a b holds in actuality and in the nearest relevantly dissimilarworlds, then neither a nor b shall cause e in the absence of the other. Andthus e wont be overdetermined. On a nomological account of causation, thepertinent worlds are all the nomically possible worlds; and so if a b holdsby nomic necessity, then a and b are twin causes, and their effects arentoverdetermined.14

2.3 Twin Causes

Above, we defined a compound event h: h = (f& g). We added that h is acause of e, and we said that if f and g are both insufficient to bring aboute, then f and g are joint causes. But if f and g are each sufficient for e,then either e is overdetermined or f and g are twin causes. We now knowwhat separates these two latter possibilities: whether or not each cause isnecessary for the other.

So let us define another compound event k: k = (a & b). k is a cause ofe; unlike joint causes f and g, a and b are each sufficient for this same effect,e. And unlike overdetermining causes i and j, a and b are both necessaryfor one another in actualityby which I just mean that if either a or b failsto occur in the relevant worlds, the other fails to occur as well. So a and

b are twin causes. And since a and b are the causes of e in actuality, esoccurrence depends on them, and so e is also absent in the relevant nearbyworlds.

None of this is new. But let us attend to the strange relationship betweentwin causes a and b and the cause of e into which they compose, k. Sinceweve said that both a and b are sufficient causes of e, it appears thatthe occurrence of either a or b suffices for the occurrence of k. It may bemysterious, then, how a could be sufficient for ks occurrence if b is necessary

14Suppose then that a b holds by nomic necessity, and this suffices for a and bto be causal equivalents. Does it follow that a = b? Is ICE true? As long as the nomicpossibilities arent the only ones pertinent to as identity and bs identity, then, the answer

is clearly no and no. For in this case, though a and b co-occur throughout the nomicpossibilities, they may diverge wildly in, say, their metaphysical possibilities. But evenif only nomic possibilities determine as identity and bs identity, as some have proposed(Most notably: [Sho04]), it may still be that a b. To take an example from Ned Block,the singleton set {Socrates} exists in a world iff Socrates exists in that world, and yet thetwo are not identical.

9


10/21

for the same. For one might assume that as being sufficient for k means that

as occurring in the absence of any other events suffices for k; but surely acannot suffice for k in the absence of other events if k cannot occur withoutb.

a does not need to suffice for k in isolation, however, in order to sufficefor k. Rather, it need only be the case that a k holds in actuality andrelevant nearby possible worlds, as is the case.

Similarly, it may be puzzling how b can be necessary for k if a is sufficientfor the samehow can k depend on the presence of b if as presence sufficesfor k? But b need not prevent any other occurrences from sufficing for k inorder to be necessary for k. It need be only that k b holds in actualityand relevant nearby worlds; and this is the case.

Let us be explicit about the principle that would seem to preclude suchrelations between sufficiency and necessity: if A suffices for E, then no B isnecessary for E. First, note that if A = B = E, then the principle carriesan implausible entailment. Assuming that A (= B = E) suffices for itself(E), it would follow that it (B) is not necessary for itself (E). Indeed, ifthis principle is true and if every E suffices for itself, it follows that B isnot necessary for E no matter what Bs value. That is, it follows that noindividual is necessary for itself or for any other: there are no necessarydependencies.

But suppose we stipulate that the principle applies only if A, B, and Eare all distinct from one another. Still it is implausible: (A E) clearlydoesnt establish that (E B). Consider an example:

A ABC is an equilateral triangle

B ABC has at least one 60 degree angle

E ABC has at least two 60 degree angles

And yet A does suffice for E while B is also necessary for it. As regardscausal sufficiency and necessity, let A, B, and E be the events from thesecond appearance of our earlier example, in which the scape tips only ifforce x is applied to the pain:

a The weight os exerting force x on the scales pan at t

b os parts exerting force x on the pan at t

e The pans tipping at t + n

10


11/21

This is perhaps less intuitive than the previous case, but it strikes me as

clear that a suffices for e and b is necessary for the same.Furthermore, where events a and b are twin causes, the relation between

them shall be symmetrical. a suffices for e and b is necessary for the same;and b suffices for e while a is necessary for the same. The principle thatwould support worries about twin causes is thus false. And, as we haveseen, distinct events a and b may each be sufficient and necessary for thesame effect e; the requirement is that a b holds in actuality and therelevant nearby worlds. The apparatus of possible worlds is powerful anduseful partly because it reveals and clarifies such distinctions. Indeed, ithas permitted us to discover an arrangement between cause and effect thathas hitherto been obscured. Events that jointly cause an effect are both

insufficient for the presence of a cause. Events that overdetermine an effectare both sufficient but at least one is unnecessary for the presence of acause. And, finally, twin causes are each both necessary and sufficient forthe presence of a cause.

The following is a list of the salient relations that hold in actuality andnearby worlds between twin causes a, b, and the cause k ( = a & b).

1. a k

2. b k

3. a b

4. (a b) k

2.4 . . . and Epiphenomena

But now there arises the difficult issue of a necessary epiphenomenon. Sup-pose l causes e, and l m holds with causal necessity, while m is an epiphe-nomenon of l .15 Now let us define n: n = ( l & m). And let us add that nis a cause of e. Thanks to l m, then, its true that ( l m) n. Andthus, l n and m n. Of course, l , m, and n now stand in the relationsdelineated just above for a, b, and k. Does nothing, then, distinguish twincauses a and b from a cause and epiphenomenon like l and m?

The answer is qualified: nothing in the relation between a, b, and k

suffices to distinguish them from l , m, and n. But there is nonetheless a15Or, if you are dubious about the existence of completely impotent events, as I am,

then suppose simply that m is not a cause of e. Perhaps the individual involved in both

l and m is the candy that made you sick, and l has its glucose properties while m has itscolor properties.

11


12/21

difference: in the relation between a or b and e on the one hand and m and

e on the other. Both a and b are causes of e while m is not.But of course this incites the rejoinder: but nothing in what Ive said so

far justifies this distinction. This is true, but its not my problem. So far asms causal relation to e depends on the pattern of actual and counterfactualoccurrences they cut in the relevant worlds, so far shall cases like a and bremain indiscernible from cases like l and m. In other words, we have justas much reason to think that m is a cause of e as to think that a or b is.

And this reveals the pressing question in the epiphenomenon case onoffer: on what grounds is it supposed that m is an epiphenomenon withrespect to e? By hypothesis, m occurs in the same pattern of actual andcounterfactual occurrences as l does. And, by our assumption, this same

pattern of occurrences establishes thatl

is a cause of e. Why, then, shouldwe suppose that m is not a cause of e? Whatever reasons there may be, theyare ancillary to an account of twin causes. Our analysis is general: whateverconditions establish that a cause is sufficient for an effect, twin causes satisfythem; epiphenomena dont.

In addition, where there should be tension between these conditions andwhatever evidence there might be that some event is an epiphenomenon,the conflict has nothing to do with the account of twin causes. Rather, itobtains between the evidence for the pertinent events being a cause and theevidence for its being an epiphenomenon.

This then suggests a second rejoinder: is it not evidence for an eventsbeing epiphenomenal that its putative effect also appears to be effected by

another event? This sentiment is very similar to Kims challenge to AlvinGoldmans account of nomic equivalentsevents that co-occur by nomicnecessity and thus seem to act as a single cause of all of their nomicallypossible effects. And conveniently, Merricks simply endorses Kims argumentin his brief response to a counter-proposal like twin causes. 16 To decidethe issue, then, we should evaluate Kims case.

3 Kims Argument Against Nomic Equivalents

So far, we have distinguished twin causes from those that act as joint causes,those that overdetermine their effects, and those related as cause to epiphe-nomenon. Twin causes are each sufficient to cause an effect and necessaryfor one another. Overdetermining causes are both sufficient for an effect,

16 [Mer03]: 71

12


13/21

and at least one is unnecessary for the other. Joint causes are both in-

sufficient for an effect; and, where one event is a cause and another is itsabiding epiphenomenon, the epiphenomenon does not cause the given effect.In addition, we have made explicit the relation between twin causes.

Still, this wouldnt satisfy the most thorough of skeptics. If causes areassumed to suffice for their effects, then one might ask: given that theoccurrence of a suffices to bring about e, what causal work is there for bto do?17 The implication is a trilemma. Given that there is no causal workfor b to do, b must either do causal work already being done, in which case eis overdetermined, or it must do no work, in which case b is epiphenomenal.If we deny both of these, one is tempted to say that a didnt really sufficefor e in the first place; hence, a and b jointly cause e.

To stick to the metaphor, the mistake is to assume that every causal jobis single occupancy. This may be the case, but it isnt obvious. If two eventsmay work the same job, the trilemma is only apparent. Since the eventsact as a single cause, the effect is not overdetermined. Since each event is acause of e, neither is epiphenomenal. And since each of a and b is sufficientto bring about e (again assuming that causes are sufficient for their effects),it is not the case that they are joint causes.

Now let us compare remarks found in [Kim89]. Kim addresses simulta-neous nomic equivalents, as found in [Gol69]. Simultaneous nomic equiv-alents are each necessary and sufficient for one another by nomic necessity.Kim alleges that the relation between nomic equivalents is unstable.

Since nomic equivalents are necessary and sufficient for one another by

nomic necessity, they act as a single cause of all of their nomically possibleeffects; thus, all nomic equivalents are causal equivalents. And, since Gold-man proposes that nomic equivalents are distinct events, all nomic equiv-alents are twin causes. But not all twin causes are nomic equivalents. aand b may be twin causes with respect to some events but not others, orthey may not be caused by all of the same events. Thus, if Kims argumentagainst nomic equivalents holds, it may not impugn twin causes; but if itfails against nomic equivalents, it fails against twin causes as well. We shallsee that it fails for both.

In what follows, I rehearse Kims many arguments, refuting each in turn.

17See ([Ben08]: 280-1):The basic idea is that if everything that happens can be ac-counted for in purely physical terms, then the mental seems to be left with nothing todo.

13


14/21

3.1 The Principle of Explanatory Exclusion

Kim sets out to defend the following principle:

EE No event can be given more than one complete and independent [causal]explanation. ([Kim89]: 79)

Kims defense of EE is an attempt to show that where there are two pur-ported causes of a single effect, both are complete causes only if they arenot independent of one another, and they are independent of one anotheronly if one or the other is not a complete cause of the effect.

But Goldman does not dispute this, and it is consistent with all wevesaid so far as regards twin causes. Each nomic equivalent depends on theother by nomic necessity, and each twin cause depends on its mate in all therelevant nearby worlds. Twin causes are not independent.

But notice that (i) these dependency relations are symmetric, and that(ii) Kim doesnt have symmetric dependence in mind. Rather, he claimsthat there must be an asymmetric dependence between nomic equivalents.

The kind of situation Goldman describes. . . is an inherentlyunstable situation. . . The instability of the situation generates astrong pressure to find an acceptable account of the relationshipbetween C and C*, and, by extension, that between the two sys-tems to which they belong; the instability is dissipated and acognitive equilibrium restored when we come to see a more spe-

cific relationship between the two explanations. As we shall see,in cases of interest, the specific relationship replacing equivalencewill be either identity or some asymmetric dependency relation.([Kim89]: 85-6)

As it is stated, then, EE is irrelevant. All parties agree that twin causesare dependent on one another. The disagreement, rather, is over anotherprinciple that Kim does not explicitly defend; lets call it EE*.

EE* No event has more than one cause unless one of the causes is asym-metrically dependent on the other.

Notice that EE* and EE suggest that overdetermination is impossible. Thisis a tendentious claim; more likely, overdetermination is improbable or sim-ply coincidental. This is prima facie evidence against EE*. Still, again, letus see how Kim attempts to establish EE*.

14


15/21

3.2 Two arguments for dependence

Kim provides two arguments for EE that he apparently believes to supportEE* as well. But they do not: they establish that there is a dependencerelation between twin causes, but they do not establish that this dependenceis asymmetric.

1. Let us consider the more succinct argument first.

These considerations suggest the following simple argumentfor explanatory exclusion for causal explanations: Suppose thatC and C* are invoked as each giving a complete explanation ofE. Consider the two questions: (1) Would E have occurred if Chad not occurred? and (2) Would E have occurred if C* had

not occurred? If the answer is yes to both questions, this is aclassic case of overdetermination. . . If the answer is a no to atleast one of the questions, say the first, that must be because ifC had not occurred, C* would not have either. And this meansthat C and C* are not independent, and hence that the twoexplanations are not independent explanations of E. ([Kim89]:92)

As we have seen, a b holds in actuality and the nearby worlds for twincauses a and b; so we answer no to both questions (provided that all else isheld equal and that there are no other causes of E in the context, of course).Thus, there is, as Kim says, a dependence relation between them. But we

have no reason to believe that this dependence is asymmetric. The samegoes for nomic equivalents. On to the other argument.

2. The argument were addressing second in fact comes first in Kims ar-ticle (it is the referent of these considerations in the preceding quotation).Kims strategy here is to give an exhaustive list of the possible relationsbetween nomic equivalents, showing that unless E is overdetermined, therelation between the causes is a dependence relation. If the relation is iden-tity, then E has only one cause, and so EE stands; if it is a dependencerelation weaker than identity, then EE still stands. EE* remains unsup-ported, though. Here is Kims list:

1. C = C* (89)

2. C is distinct from C*, but is in some clear sense reducible to, orsupervenient on C*. This sort of situation will arise in a reductivecontext of the sort just considered provided that for whatever reason

15


16/21

we stop short of identifying the reductively related events or states.

(90)

3. Neither C nor C* is in itself a sufficient cause of E, though eachis an indispensable component of a sufficient cause. (90) (Kim adds:Case 3a: C is a proper part of C*; the difference is immaterial tous.)

4. C and C* are different links in the same causal chain leading, say,from C to C* and then to E. In this case again we do not have twoindependent causal explanations; the explanans of one, C*, is causallydependent on the explanans of the other, C. (91)

5. C and C* overdetermine E. (91)We have distinguished single causes from 3, joint causes, and 5, overde-terminers, above. 1 is true only if ICE obtains, and ICE is implausible;moreover, Goldman proposes nomic equivalents as incompatible with bothICE and 1. If we assume that links in a causal chain are not simultaneous,4 is also inconsistent with Goldmans simultaneous nomic equivalents. Andof course the twin causes weve discussed are not different links in the samecausal chain (though I have left it open that there are such twin causes).

Thus, only 2 remains. As stated, it does not entail asymmetric depen-dence: supervenience is sometimes symmetric, and Kim admits as much ina later work: . . . mind-body supervenience as stated isnt symmetric; in

general, the supervenience of A on B does not exclude the supervenience ofB on A.18

Supervenience is a relation between sets of properties: a supervenienceset, call it set A, and a base set, call it set B. The popular slogan for therelation is: A supervenes on B if and only if no two individuals can differ intheir A properties without also differing with respect to their B properties.19

If {P, P} is the base set and {Q, Q} is the supervenience set, then for allx and y, if x and y are both P, then x and y must either both be Q or bothbe Q. They must both have the same properties from the supervenienceset since they have the same properties from the base set.

And this relation is sometimes symmetric. For instance, a set of posi-

tive properties {A, B, C...} supervenes on the set of its complements {A,18 [Kim98]: 1119There are in fact a number of supervenience relations. For simplicity, I am reviewing

only local, strong supervenience, which is Kims preferred relation. See, for example,[Kim98]: 9.

16


17/21

B, C...}, and vice versa: x and y cannot differ in their positive prop-

erties without differing in their negative properties. Similarly, notice thatevery set of properties supervenes on itself. For all x and y, if x and yhave the same A-properties, x and y must have the same A-properties. Thedependence herein is of course symmetrical.

Moreover, if 2 did state that there must be an asymmetric relation be-tween C and C*, it would be implausible that the list is exhaustive: theforegoing accounts of twin causes and nomic equivalents demonstrate asmuch. Insofar as these accounts are plausible, it is thus far implausible thatsuch a revised edition of Kims list would be exhaustive.

3.3 An argument for asymmetric dependence

But Kim does attempt to show that reduction is an asymmetric relation,and he does attempt to show that nomic equivalents satisfy the conditionsfor reduction. Let us grant that each psychological event has a physicalnomic equivalent; this licenses the inference to P1.

(P1) Every psychological event depends on a physical event by nomic ne-cessity.

(P2) If psychological events are dependent on physical events by nomicnecessity, then psychological events are reducible to physical events.

(P3) If psychological events are reducible to physical events, then psycho-

logical events asymmetrically depend on physical events.

(C1) Psychological events asymmetrically depend on physical events.

([Kim89]: 88)

This argument is valid, but either P2 or P3 is false. I am sympatheticwith P2, but if P3 is true as well, then the one-way dependence of psycho-logical events on physical events logically precludes dependence in the otherdirection.

The general principle is that as depending on b precludes bs dependingon a. Thus, there are no symmetric dependence relations short of identity.This is implausible on its face. In addition to the examples provided inthe foregoing text, symmetric dependencies short of identity abound. Themovements of a baseball bats heavy end depend on those of the light endand vice versa. Where x = y2, the value of x depends on y and vice versa.

17


18/21

And, given the way the rear wheels of a car are connected to the drive shaft,

the left wheel can turn only if the right wheel does and vice versa.20

Given that the principle fails in general, Kim must convince us that itholds in the special case of nomic equivalents or in the case of psychologicaland physical nomic equivalents. Independently, I find it very plausible thatthe dependence relation between macroscopic objects and their microscopicconstituents is symmetric; if they are not identical, then Kims principlefails, and there is no instability between nomic equivalents or twin causes.

I take it, then, that Kims arguments against twin causes and nomicequivalents are impotent.

4 Conclusion

1. We have described a heretofore overlooked species of Multiple AntecedentCausal Relation, contrasting twin causes with joint causes and overdetermin-ers. Where k is a cause of e, a and b are twin causes of e if each is bothnecessary and sufficient for k. Twin causation is inconsistent with the viewthat any effect with more than one sufficient cause is overdetermined.2. There are in fact several other MACRs that have received little atten-tion, if any at all. Discussions of MACRs have tended to focus on (i) suf-ficiency and insufficiency, and (ii) symmetric cases. Thus, joint causes areboth insufficient, overdeterminers are both sufficient. Many have pointedout that overdetermination may be more complex than this, though. If

one or the other antecedent is necessary for the other, then perhaps the ef-fect is not overdetermined; at any rate, the result differs from the standardfiring-squad cases of overdetermination. Lets say such effects are overde-termined*. Twin causes are then an additional deviation from the standardcase: each antecedent is necessary for the other.

In the table below, the Xs below indicate the number of antecedentsthat are sufficient for the effect; below , Xs indicate the number ofantecedents that are necessary for the occurrence of a compound causalevent the simples of which are the given antecedents. For example, if thecausal compound is k = (a & b) and there are two Xs under , then eachof ks simplesi.e. a and bis necessary for ks occurrence.

effect is:XX OverdeterminedXX X Overdetermined*XX XX Twin Caused

20Thanks again to XX for this last example.

18


19/21

As has been shown in the foregoing, these are indeed distinct varieties of

MACR: whether the antecedents are necessary makes a difference to theeffects status as overdetermined, overdetermined*, or not overdeterminedat all. The same may be true for cases in which only one of the antecedentsis sufficient for the effect or neither is, though I am agnostic on the matterat present. Schematically at least, these are all distinct MACRs:

antecedents are:X A CauseX X Cause w/ 1 Nec. ConditionX XX Cause w/ 2 Nec. Conditions Joint Causes X Joint Cause w/ 1 Nec. Condition

XX Joint Cause w/ 2 Nec. Conditions

Again, I do not claim that these differences are substantive.3. We have also shown that insofar as a and b may act as a single causeof a single effect without acting as a single cause of all of their effects orwithout themselves having the same causes, it is implausible that if a and bare causal equivalents, then a = b. That is, ICE is implausible.

ICE If a and b are causal equivalents, then a = b.

Finally, ICE is a hidden premise in the arguments we reviewed in 0 - 1.Since ICE is false, those arguments fail.

References

[Ben03] Karen Bennett. Why the exclusion problem seems intractable,and how, just maybe, to tract it. Nous, 37(3):471497, July 2003.

[Ben08] Karen Bennett. Exclusion again. In Jakob Hohwy and JesperKallestrup, editors, Being Reduced: New Essays on Reduction,Explanation, and Causation, chapter 14, pages 280306. OxfordUniversity Press, 2008.

[Dre89] Fred Dretske. Reasons and causes. Philosophical Perspectives,

3:115, 1989.

[Gol69] Alvin I. Goldman. The compatibility of mechanism and purpose.The Philosophical Review, 78(4):468482, October 1969.

19


20/21

[Kal06] Jesper Kallestrup. The causal exclusion argument. Philosophical

Studies: An International Journal for Philosophy in the AnalyticTradition, 131(2):459485, November 2006.

[Kim89] Jaegwon Kim. Mechanism, purpose, and explanatory exclusion.Philosophical Perspectives, 3:77108, 1989.

[Kim93a] Jaegwon Kim. Events as property exemplifications. In Superve-nience and Mind, chapter 3, pages 3353. Cambridge UniversityPress, 1993.

[Kim93b] Jaegwon Kim. Mental causation in a physical world. Science andKnowledge, 3:157176, 1993.

[Kim95] Jaegwon Kim. Mental causation: What? me worry? Philosophi-cal Issues, 6:123151, 1995.

[Kim97] Jaegwon Kim. Does the problem of mental causation generalize?Proceedings of the Aristotelian Society, 97:281297, 1997.

[Kim98] Jaegwon Kim. Mind in a Physical World. Representation andMind. The MIT Press, 1998.

[LL87] Ernest Lepore and Barry Loewer. Mind matters. The Journal ofPhilosophy, 84(11):630642, November 1987.

[Low00] E.J. Lowe. Causal closure principles and emergentism. Philosophy,75(294):571585, October 2000.

[Mal68] Norman Malcolm. The conceivability of mechanism. PhilosophicalReview, LXXVII, 1968.

[Mer03] Trenton Merricks. Objects and Persons. Oxford University Press,2003.

[Mum98] Stephen Mumford. Dispositions. Oxford University Press, 1998.

[Pap93] David Papineau. Philosophical Naturalism. Blackwell, 1993.

[Pea79] Christopher Peacocke. Holistic Explanation: Action, Space, In-terpretation. Clarendon Press, 1979.

[PPJ82] Elizabeth W. Prior, Robert Pargetter, and Frank Jackson. Threetheses about dispositions. American Philosophical Quarterly,19(3):251257, 1982.

20


21/21

[Rit05] Jack Ritchie. Causal compatibilism: What chance? Erkenntnis,

63(1):119132, July 2005.

[Sho04] Sydney Shoemaker. Identity, Cause, and Mind: Expanded Edi-tion. Clarendon Press, 2004.

[Sid07] Theodore Sider. Parthood. Philosophical Review, 116:5191, 2007.

[Sos95] Ernest Sosa. Davidsons thinking causes. In John Heil and AlfredMele, editors, Mental Causation, chapter 4, pages 4150. Claren-don Press, 1995.

[Stu05] Karsten R. Stueber. Mental causation and the paradoxes of ex-planation. Philosophical Studies: An International Journal for

Philosophy in the Analytic Tradition, 122(3):243277, February2005.

21

Twin Causes 3

Documents

Transcript of Twin Causes 3