Functionalism and the Metaphysics of Causal Exclusion

volume12,no.13 july2012

Functionalism and

the Metaphysics of

Causal Exclusion

David YatesKings College, London

© 2012 DavidYatesThis work is licensed under a Creative Commons

Attribution-NonCommercial-NoDerivatives 3.0 License. <www.philosophersimprint.org/012013/>

1. Introduction

Take functionalism to be the thesis that mental property M isthe property of having some other property that plays a certaincharacteristic causal roleR.1 Functionalists are usually physicalists,and so takemental properties to be physically realized, such thatfor anymental propertyM, there’s a physical property P that fillsR. Causal exclusion looms. Functionalism takesmental propertiestobecharacterisedbycausalrolesthatarefilledbysomethingelse.As M’s realizer, P must do all the causal work (whatever that is)thatR involves, apparently leavingnothing forM to do. Ifmentalpropertiesarecausallyredundant,andifcausalnoveltyisnecessaryfor robust ontological commitment, then mental properties aren’treallyreal.2Thiswouldbenosmallironyiftrue,becausehistorically,functionalismwasmotivatedbytheneedtoexplainhowphysicallydifferentcreaturescouldbeinthesamementalstate;smallcomfortto be told that it’s by dint of falling under predicates that don’tpick out genuine properties. Label properties that do causalwork“causally efficacious”, setting aside for now the question of whatcausal work is. Those who argue that functionalism is consistentwiththecausalefficacyofmentalpropertiestypicallyadoptoneofthefollowingstrategies:(i)denythatcausalnoveltyisnecessaryforontological commitment and argue that functional properties caninherit the efficacy of their realizers, resulting in a kind of causalover-determination;3 (ii) defend a difference-making theory ofcausation that entails that functional properties are causally novelafterall.4Proponentsof(i)defend,oratleastrecognisetheneedtodefend,theoriesofcausationthatvindicatetheirclaimthatfunctionalproperties inherit efficacy from their realizers. The task then is toarguethatthekindofover-determinationinvolvedisn’tproblematic.

1. Functionalismsoconstruedisn’tlimitedtomentalproperties.Myargumentsinthispaperdependonlyonthegeneralmetaphysicoutlinedabove.

2. Kim(1992a,b,1998).Iwillfillintheprecedingsketchinduecourse.

3. SegalandSober(1992);Bennett(2003);Witmer(2003);Kallestrup(2006).

4. Yablo(1992);ListandMenzies(2009).

ImprintPhilosophers’

davidyates Functionalism and the Metaphysics of Causal Exclusion

philosophers’imprint –2– vol.12,no.13(july2012)

ofproperties toSection3,where I argue that themere fact that,byHumeanlights,aproperty-instancecausessomeeffect,doesn’tentailthatit’saninstanceofapropertythatdoescausalworkinrelationtothat effect. Kim treats causation as a relation between fine-grainedevents,construedasproperty-instances.Onthisviewmentaleventsare numerically distinct from the physical events that realize them.Those who prefer coarse-grained events may recast the argumentsthatfollowintermsofsingleeventshavingbothmentalandphysicalproperties,withtheformersuperveningonthelatter.

2.1 Kim’s supervenience argument.Themost completepresentationofKim’s supervenience argumentoccurs inhis(2005).6Kimofferstwoversionsof theargument; forbrevity I focus on the simplest, which concludes that no propertythat’s not identical to some physical property could cause aninstance of a supervenient property. It’s the supervenience of theeffect property-instance on a physical base that drives the simpleversionofKim’sargument.7Kimbeginsby focusingon “mental-to-mental” causation,which I take to involve instancesof intentionalmental properties such as thirst causing instances of behaviouralproperties such as drinking. The argument also applies to causalchains of mental property-instances of the first kind; I focus oncausationofbehaviourforreasonsofexposition.FollowingKim,letthementalproperty-instancesbeMandM*,and letM*’sphysicalbase property-instance be P*. For reductio, suppose thatM causesM*.Kimdefendsaprincipleofdownwardscausation,accordingtowhichtheonlywaytocauseM*istocauseP*.TheideaisthatsinceP*realizesM*,tosupposeaneventcouldcauseM*withoutcausingP*islikesupposingapillcouldalleviateaheadachewithoutcausinganybrainevents.

6. ch.2,pp.39–52.

7. Themorecomplexversiondependsonthesupervenienceofbothcauseandeffectproperties.

Proponentsof(ii)needn’tworryaboutover-determination,becauseaccording to difference-making theories of causation, functionalpropertiesmakeadifferencetheirrealizersdon’t.

Both strategies suffer, I argue, froma failure to properly analysethenotionofcausalworkasappliedtoproperties.If,asiscommonlysupposed, the causal work of a property consists in grounding thedispositions of its bearers, then the nature of functional propertiesrenders it impossible for them todo the same causalwork as theirrealizers. For related reasons, I argue, whatever the merits of thedifference-makingaccountofcausation,itcan’tbeanaccountofthecausalwork thatpropertiesdo.Clarificationof thenotionof causalwork reveals a novel solution to the exclusion problem based onthe relations between dispositional properties at different levels ofmechanism,whichinvolvesthreecentralclaims:(i)thecausalworkofpropertiesconsistsingroundingdispositions,(ii)functionalpropertiesaredispositions,and(iii)thedispositionsofmechanismsaregroundedin the dispositions of their components. Treating functionalmentalpropertiesasdispositionsofcomponentsinpsychologicalmechanisms,Iargue that suchpropertiesdo thecausalworkofgroundingagent-level dispositions.Thesedispositions,whileultimately grounded inthephysicalrealizersofmentalproperties,areindirectlysogrounded,throughahierarchyofgroundingrelationsthatextendsupwards,ofnecessity,throughthementaldomain.

2. The Causal Exclusion Problem and the Humean Backlash

ProponentsofHumeanapproachestocausalexclusionsupposethatthecausalefficacyofpropertiescanbecapturedbycounterfactualornomicrelationsbetweenevents,typically—butnotnecessarily—construedas instances of the target properties. In Section 2.1, I discuss Kim’s“supervenienceargument” againstmental causation,5 and inSection2.2,IshowhowtheHumeanstrategiesoutlinedabovearesupposedto block that argument. I postpone discussion of the causal work

5. Kim(1998),ch.2;(2005)ch.2.



of the other.As he notes, that doesn’t seem to be the rightway todescribethewayinwhich—putatively—MandPbothcauseP*.12Incases such as executionbyfiring squad, thebullets of thedifferentmarksmen over-determine the victim’s death by causing instancesofdistinctproperties in thevictim.Themannerof thevictim’sdeathisaltered by its being over-determined in thisway. This isn’t to saythat the victimdies a different death by virtue of this kind of over-determination—as Lewis argues, in ordinary discourse we speakof events as if robustwith respect to small enough changes in themanneroftheiroccurrence13—butwecanatleastfindforeachcausea different effect. A death is an instance of a highly determinableproperty,andonemight suggest that causingadeath isamatterofcausingoneofitsdeterminates.Ifwethinkofdeterminablepropertiesas supervenient properties, then this much follows from (DC). It’spossible,then,fortwomarksmentoover-determineadeathbyeachcausingoneofitsdeterminatesatthesametime,andthisisconsistentwithholdingthatthevictimwouldhavediedthesamedeathhadoneofthemarksmenmissed.Over-determiningcausesofthiskindhavedifferentcausalroles.Notsointhecaseofmentalcausation,itseems:MandPputativelyover-determineP*bydoingexactly the same thing, viz. causing P*. Call this “redundant over-determination”. The onlyremainingoptionistodenytheassumptionwestartedwith,viz.thatMcausesP*.Butby(DC),ifMdoesn’tcauseP*,thenitdoesn’tcauseM*either,andthere’snomentalcausation.

Kimdoesn’thaveaknock-downargumentagainstredundantover-determination, but in earlierwork, he argues, based on a principlehecalls“Alexander’sdictum”,thatifmentalpropertieshavenonovelcausalworktodo,weshouldeliminatethem:14

(AD) Toberealistohavecausalpowers.

12. Op. cit.pp.46–52.

13. Lewis(1986).

14. Kim(1992a,b).

(DC) M causesM* by causing its physical superveniencebaseP*.8

Kimneeds two furtherprinciples toshowthatMdoesn’t causeM*.First,thecausalclosureofthephysical:

(CC) Ifaphysicaleventhasa [complete, sufficient]causethatoccursatt,ithasa[complete,sufficient]physicalcausethatoccursatt.9

OntheassumptionthatMcausesM*,itfollowsfrom(DC)thatMcausesP*.Butfrom(CC),P*musthaveaphysicalcauseatthetimetwhenMoccurs.Thisevent,P,ismostnaturallyregardedasM’ssuperveniencebase, but the simple version doesn’t depend on this. One furtherprincipleisrequired,theoft-cited“causalexclusionprinciple”:

(CX) No single event can have more than one sufficientcause occurring at any given time—unless this is agenuinecaseofcausalover-determination.10

AssumingthatMandPbothcauseP*,itfollowsfrom(CX)thateitherM=P, or this is a genuine case of causal over-determination. Sincefunctionalism is non-reductive, identifyingM andP isn’t an option.Identifying property instances entails identifying the propertiesthey are instances of, since a property instance (x,P,Δt) is identicaltoaproperty instance (y,Q,Δt’) if andonly ifx=y,P=QandΔt=Δt’.11 What about over-determination? Kim takes “genuine causal over-determination” to involvetwoindependentcausalchains leadingtothesameeffect,whereeachwouldhavebeensufficientintheabsence

8. Kim(2005)p.44.

9. Op. cit. p. 43. Ihaveadded theparenthetical ‘complete, sufficient’ toKim’sformulationbecause(a)‘sufficient’isclearlyimplicitinthatformulation,and(b)without‘complete’(CC)wouldbeconsistentwithcertainformsofemer-gentism,accordingtowhichmentalandphysicalcausescombinetocauseP*.Ireturntothecompletenessofphysicalcausesin(4).

10. Op. cit.p.41.

11. (x,P,Δt)shouldbeinterpretedas:x’shavingPduringintervalΔt.



property of being a neuron doesn’t, unless there’s some incrediblycomplex structural property definable in terms of properties offundamentalphysicsthatallneuronshaveincommon.Moreobviously,mentalpropertiesareeliminated,becauseit’shugelyimplausiblethatall those who like roast chestnuts share a physical property. Kim’scausal exclusion argument, then, is a combination of two separatearguments: (i) an argument that mental property-instances are atbest redundant over-determining causes; (ii) an argument that real,irreduciblepropertiesaren’tredundant,sothatmentalpropertiesareeitherunrealorreducible.Thosewhoendorseover-determinationasa response to theexclusionargumentaccept (i)but take issuewith(ii); thosewho endorse difference-making causation can ignore (ii),becausetheyreject(i).

2.2. Humean solutions to the exclusion problemHumean solutions are Humean because they appeal to nomic orcounterfactual relations between mental property-instances andbehaviouralproperty-instancestoshowthatmentalpropertiesdocausalwork.Suchstrategiestypicallyinvolve(a)definingcausationintermsoflawsorcounterfactuals,andshowingthataccordingtothedefinition,mentalproperty-instancesarecausesofcertainbehaviouraleffects;and(b)a tacit assumption that ifmentalproperty-instancesare causesofbehaviouraleffects,thenmentalpropertiesdocausal workinbringingthoseeffectsabout.Igrantboth(a)and(b)fornow,butinSection3,Iarguethat(b)isdemonstrablyfalseonanindependentlymotivatedaccountof causalwork.First, I’lloutline theover-determinationanddifference-makingsolutions,beginningwiththeformer.

Those who think that M and P over-determine P* take this tomean thatMdoescausalwork, inheriting someorallof thecausalpower that Phas to causeP*. Someoffer criteria of causal efficacyaccordingtowhichwecanshowthatifPcausesP*,then—giventhenature of the supervenience relation betweenM and P—M causesP*.SegalandSober,forinstance,thinkthatthe(possiblynon-strict)lawofnaturethatM-instancesarefollowedbyM*-instances,together

Accordingto(AD),thereisnosuchthingasanentitythatdoesn’tcause(orcontributetocausing)anything.Theprinciplemayseemtoostrong,since it rulesout theexistenceofabstractentities suchasnumbers,sets,andsoforth.Iaminclinedtothinkareasonablycircumscribed(AD)canavoidsuchproblems:wemight,forinstance,limititsscopetonaturalentities.It’sonethingtoclaimthatsomeentitieshavenocausal powers, quite another to claim that mental properties havenone.Ineedn’tworryaboutpreciselyhowtoformulate(AD),however,sincemyaiminthispaperistoshowthatmentalpropertieshavenovelcausalpowers.Mystrategywillbetograntthe(perhapsimplausibly)strongversionof(AD)aboveandshowthateventhisprincipledoesnotthreatentheexistenceofmentalproperties.Kimthinksthatgiven(AD), their irreducibility implies that “mental properties bringwiththem … causal powers … that no underlying physical-biologicalprocessescandeliver….[T]obereal,newandirreducible…mustbetohavenew,irreduciblecausalpowers.”15

Ifcogent,thislineofreasoningrulesoutontologicalcommitmentto properties that are nevermore than over-determining causes. Ifmentalpropertiesare irreducibleandreal, then theircausalpowersmustbeirreducibleaswell,whichistosaytheymustdosomethingtheirrealizersdon’t.Accordingtofunctionalism,mentalpropertiesaredefinedbyrolesthatarefilledbyphysicalproperties,soanovelcausalroleseemsoutofthequestion.

The upshot of the argument is that there are no functionalproperties.Theonlypropertiesthatsurvivethecullarethepropertiesof fundamental physics, and any properties that are identical toparticularconfigurationsofthoseproperties.16Thepropertyofbeingahydrogenatomsurvives,asdoesthepropertyofbeingH2O;butthe

15. Kim(1992a).

16. SeeKim(2003),wherehearguesagainstBlock (2003) that causalpowersdon’t“drainaway”ifthere’snofundamentallevel.ThisisbecauseKim’star-getismultiplyrealizableproperties,andKimthinksthatwhenwegetdownsufficientlydeep,therewon’tbemultiplerealizationanymore,butapoten-tiallyinfinitedescentthroughmereologicallevelswiththepropertiesofeachLidenticaltostructuralpropertiesofL-1.



base,itisplausiblymetaphysicallynecessarythatifanythinghasP,ithasM.19 If therearenoP-occurs-without-Mworlds, conjunct (iii) isvacuouslytrue.

The fact that M supervenes with metaphysical necessity on Palso—ontheassumptionthatMinheritsP’scausalefficacy—rendersit unmysterious that there are two causes of P*. If widespreadcoincidence is the reason over-determination is problematic, notonlyisthekindofover-determinationwehavehereunproblematic;we also have a test—non-vacuous truth of all of (i)–(iv)—for thepotentially problematic kind. Bennett takes a slightly different view,arguing that thenon-vacuous truthof (i)–(iv) is necessary for over-determinationsimpliciter,meaningthatMandPdon’tover-determineP*atall.20Themainimportofthisdifferencerelatesto(CX).Bennettmust reject (CX)—M and P are distinct causes of, but don’t over-determine,P*.Kallestrup,ontheotherhand,canholdthatMandPdoover-determineP*,butinawayweshouldn’tworryabout.Itmatterslittleformypurposespreciselywhat‘over-determination’means.

Assuming M inherits its causal powers from P, it remains onlyfor over-determinationists to rebut Kim’s redundancy argument.Accordingto(AD),nothinglackscausalpowers.Functionalpropertiesareirreducibletotheirbases.Crucially,Kimtakes(AD)toimply:

AD*: Toberealandirreducibleistohaveirreduciblecausalpowers.

Butasothershavepointedout, (AD)doesnot imply(AD*).21From(AD),itfollowsonlythatifmentalpropertiesarerealandirreducible,thentheyhavecausalpowersandareirreducible.AsKallestrupargues,there’snoreasontotake ‘irreducible’ toqualify thecausalpowersof

19. Kim(2005)attemptstoblocktheargumentthat followsat thisstage,argu-ingthatM’ssupervenienceonPisnomologicallynecessaryandbackedbybridge-laws.Idon’tfindthispersuasive,buthaven’tthespacetotakeuptheissuehere.

20.Bennett(2003).

21. Stephan(1997);Kallestrup(2006).

withsufficientlytightsuperveniencerelationsbetweenMandP,andbetweenM*andP*,issufficientforMtocauseM*.17Thistheoryissupposedtoaccountforthecausalefficacyofmentalproperties,andsoisintendedtobeasufficientconditionforapropertytodocausalwork.MinheritscausalefficacywithrespecttoM*fromP’sefficacywithrespecttoP*,viathesupervenienceofMandM*onPandP*,respectively,togetherwithalawrelatingMtoM*.Whileitisdesirableto defend a theory of causation according to which M causes M*,mostover-determinationistsfocusonshowingthatthekindofover-determination involved inM causingM* wouldn’t be problematic.Kallestrupthinksofover-determinationasfollows:

(OD) Eisover-determinedbyC1andC2iff(i)C1issufficientforE,(ii)C2issufficientforE,(iii)ifC1hadoccurredwithoutC2,Ewouldhaveoccurred,and(iv)ifC2hadoccurredwithoutC1,Ewouldhaveoccurred.18

In firing-squad cases, all four conjuncts are non-vacuously true.There’snothingproblematicaboutthis,becauseweknowwhytheseconjunctsaretrueinsuchcases:thoseresponsibleforfiringsquadsactpreciselysoastomakethemtrue.Buttheright-handsideof(OD)is also non-vacuously satisfied in some cases without there beingany explanation of why this is so. It’s possible that two assassinsindependently decide to assassinate the same person at the sametime without a common cause that might explain this. This too isunproblematic,provideditdoesn’thappenveryoften.Ifitdid,therewould be widespread inexplicable coincidence of forward-lookingcausalpowers,which—grantforthesakeofargument—wouldbeaverybadthing.NowMandPsatisfy(OD)withrespecttoP*,butinadifferentway.LetP=C1,M=C2,P*=E.SincePisM’ssupervenience

17. SegalandSober(1992).SegalandSoberthinkintermsofDavidson-eventshavingbothmentalandphysicalproperties,andsupplyasufficientconditionforittobeinvirtueofMthattheP∧MeventcausestheP*∧M*event.IhaverecasttheirtheoryintermsofKim-eventsforconsistency.SeealsoWitmer(2003),esp.pp.205ff.

18. Kallestrup(2006),p.471.Adjustedfortypographicconsistency.



difference-making. Following List and Menzies, define difference-makingasfollows,whereFandGareproperty-instances:25

DM: FmakesadifferencetoGiff:(i)Foccurs*→Goccurs,and(ii)¬(Foccurs)*→ ¬(Goccurs)

According to (DM), Fmakesadifference toG if andonly if the setof closest Fworlds to actuality areGworlds, and the set of closestnon-F worlds are non-G worlds. Now let’s evaluate the followingcounterfactuals:

(i) Moccurs*→M*occurs

(ii) ¬(Moccurs)*→ ¬(M*occurs)

(iii) Poccurs*→M*occurs

(iv)¬(Poccurs)*→ ¬(M*occurs)

It’s easy to see that (i), (ii), and (iii) are true.The closestMworldsto actualitywillbeworlds inwhichM is realizedby somephysicalpropertythatcausesarealizerofM*;theclosestnon-Mworldswillbeworldswhereithasnophysicalrealizerproperty,andatwhichnorealizerofM*occurs;andtheclosestPworldswillbeworldsatwhichP* occurs.However, the closest non-Pworlds to actualitywill—sothe argument goes—be worlds at which some other realizer ofMoccurs.Intuitively,thisseemscorrect:aworldatwhichanalternativerealizerP’ofMoccurs,whichissimilarbutnotidenticaltoP,andatwhichanalternativerealizerP*’ofM*occurs,iscloserthanworldsatwhichneitherMnorM*occurs.Butifthisisso,then(iv)isfalse,andPdoesn’tmakeadifferencetoM*.

Thiskindofargumentholdswherevercausationofasupervenientproperty-instance is involved, provided the occurrence of thatproperty-instanceisinsensitivetotheprecisemannerofitsrealization.Interestingly,suchcasesdon’t falsify (CX),but that’sbecauseP failstobe a causeofM*.This is a caseofM’sdifference-making causal

25. ListandMenzies(2009);Menzies(2008).

propertiesratherthanthepropertiesthemselves,unlessofcoursewetake the view that properties are exhausted by their causal powers.In that case, theonly thing that couldmakea supervenientpropertyirreducibleisbestowingcausalpowersitsbasepropertydoesn’t.22Over-determinationistscanreplythatthereareotherwaysforasupervenientpropertytosecureirreducibilitytoitsbase.ProvidedMcaninheritthecausalpowersofP,andMhasakindofnoveltythatdoesn’trequireittobestownovelcausalpowers,thenthere’snopressuretoeliminate.Wemight,forinstance,endorseShoemaker’ssubsettheoryofrealization,accordingtowhichMsupervenesonPbecauseitscausalpowersareapropersubsetofthepowersofP.ThiswouldseemtoprecludeouridentifyingM and P, but whether it’s enough to secure the kind ofnoveltyover-determinationistsneedisamootpoint.23

Those who endorse difference-making causation as a responsetotheexclusionproblemfollowYablo inthinkingthatcausesmustbe proportional to their effects.24 M* has many distinct possiblesuperveniencebases,ofwhichP*isone;similarly,mutatis mutandis, forMandP.Theidea, informally, isthatPisn’tproportionaltoM*,becauseP is causally sufficient for a specific realizationofM*, viz.P*. SinceM*might have occurred differently, P causally explainswhy M* happened in a particular way, but not why it happenedsimpliciter. Similarly, there will be property-instances that fail toexplainM*duetonotbeingspecificenough:thepropertyofhavingsomemental property, for instance. The cause of M*, by contrast,ought tobeaproperty-instance that’s just right, in the sense that itcausesM*howeveritorM*arerealized,andMfitsthebill.Yablo’snotion of proportionality is typically now described in terms of

22. Kallestrup(2006)pp.468–470

23. Shoemaker(2001).Kim(2010)suggeststhatShoemaker’stheoryisinfactaformoftype-identitytheory,butthesemattersarebeyondthescopeofthispaper.Iwillnote,however,thatifPhasasaconstituentaphysicalpropertyP’whosepowersarethesamepropersubsetofP’spowersthatMinherits,thenthere’snothingtopreventidentificationofMwithP’.

24. Yablo(1992).



house—pushing,lugging,chopping—thenhowcantherebeanyleftforMtodo,givenP’ssufficiencyforP*?Butthat,sayHumeans,isthewrongwaytothinkaboutcausalwork.Builderscertainlydothiskindofworkwhen theybuildhouses,but thatdoesn’tmean thateventsdoitincausingotherevents.AsSiderpointsout,it’ssurelywrongtothinkofcausalworkas“akindoffluiddividedamongthepotentialcausesofaneffect”,suchthat“[i]fonepotentialcauseactstoproduceaneffect,thatfluidisusedup,andnootherpotentialcausecanact”.28 If that’showKimthinksofcausalwork, thentheburdenofproof isonhimtoshowthatthiswayofthinkingismoreplausiblethanthedifference-makingtheory—noeasytask.

Let’stakestock.Therearetwosolutionsunderconsideration.ThefirstsaysthatMandPbothcauseM*,andgoesontosayeitherthatthisisn’tagenuineformofover-determination,sothat(CX)isfalse,orthatitis,butofanon-problematickind.Still,Mshouldn’tbeeliminated,becauseitcanearnitsontologicalkeepwithoutbeingcausallynovel.ThesecondsaysthatMiscausallynovel,becauseonanindependentlymotivated theoryofcausation,MandnotPcausesM*,and itdoessowithoutcausingP*,so(DC)isfalse.BothstrategiesassumethatifwecanshowthatanappropriateHumeanrelationholdsbetweenMandM*,thenwewillhaveshownthatMdoescausalwork.Andthat’swheretheygowrong.

3. Causal work and the inadequacy of the Humean response

Humeans assume that if an instance of mental property M causesan instanceof behavioural propertyB, thenM is causally efficacious with respect toB, in the sense that it does the kindof causalworkrelatingtoB’soccurrencethatexclusionistssayitdoesn’tdo.I’llnowargue that this seemingly innocuous assumption is inconsistentwith an independently plausible and widely held view concerningthe kindof causalwork that propertiesdo.Mostphilosopherswhothink properties do causal work think that this work consists in

28.Sider(2003),p.721.

role with respect toM* excluding a similar difference-making rolefor P. What’s gone wrong with the exclusion argument, then? It’snotobviouswhatweshouldsay.Ifwesubstitute ‘difference-makingcause’ for ‘cause’ in (CC), then if we treat M* as a physical event(qua behavioural) then (CC) comes out false, because M* doesn’thaveaphysicaldifference-maker. Ifwesubstitutedifference-makingcausationin(DC),ittoocomesoutfalse.Misn’tadifferencemakingcauseofP*,since¬(Moccurs*→P*occurs).Perhapswecouldthenkeepholdof(CC)byemployingamoreflexiblenotionofcausation.Thereis,afterall,asenseinwhichM*doeshaveasufficientphysicalcause.PissufficientforP*,andthesuperveniencerelationbetweenP*andM*issynchronicandnon-causal,soitwouldseemfoolishtodenythatsomekindofcausalrelationholdsbetweenPandM*,andIseenoreasonwhythiscausalrelationshouldn’tsatisfyproponentsof(CC).26Let’ssaythatdifference-makingcausationinvolvesarejectionof(DC):McausesM*,butnotbycausingP*,sothereisn’tanyover-determination,andM’scausalroleissecure.

But doesn’t P still do all the causal work involved in M*’soccurrence? Well, not if causal work is understood as making adifference!It’safactabouttheworldweliveinthateventhoughP*hasasufficientphysicalcauseP,itdoesn’tfollowthatPdoesallthecausalwork involved inM*’s occurrence, even though the relationbetweenP*andM*isnon-causal.SincePisn’tadifference-makingcauseofM*,Mhasnovelcausalpower: itmakesadifferencetoM*thatnothingelsedoes.Atthispointonemightsuspect,withKim,thatdifference-making causation is a cheat:27 the theory proposes twoindependentlevelsofcausalwork,MmakingadifferencetoM*andPmakingadifferencetoP*.What’soddaboutthisisthatPcausallynecessitatesM*,inthesensethatitcausessomething,P*,thatnon-causallynecessitatesM*.IfwethinkofthecausalworkrequiredforM*tohappenintermsofthekindofworkabuilderhastodotobuilda

26.Yablocallsthisrelation“causalsufficiency”,anddistinguishesthisfromcausa-tion,withthelatterrequiringproportionalityofcauseandeffect.

27. Kim(1998),ch.3.



3.1. DispositionsFamously,sentencesoftheform‘xisdisposedtoMwhenC’,can’tbeanalysedintermsofsubjunctivesoftheform‘wereCtooccur,xwouldM’.31Supposeavaseisdisposedtoshatterwhenstruckbyanobjectwithmomentum≥ m,andlabelthisdispositionDS.(HereafterIsometimesomit‘byanobjectwithmomentum≥ m’forbrevity.)SupposethevasehasDSiff,werethevasetobestruck,itwouldshatter.Finksfalsifytheputativeanalysisinbothdirections.ThisvaseisGandalf’sfavourite,andhe’llintervene,wheneverit’sstruckbyanobjectwhoseimpactwouldotherwisehaveshatteredit,castingaspelltoalteritsatomicstructuresothatitdoesn’tshatter.Assumingdispositionsareintrinsicproperties,thevasehasDSpriortobeingstruck,buttheanalysansisn’tsatisfied.Lewisproposesthatthesimpleconditionalanalysiscanbefixedbymodifyingtheanalysansto‘thevasehasanintrinsicpropertyPsuchthatwereittobestruckandretainP,itwouldshatterbecauseofitshavingPandbeing struck’.32Gandalf’s favourite vase satisfies this analysans, so isdisposedtoshatterwhenstruck,eventhoughitdoesn’t.However,masksandantidotesfalsifytherevisedanalysisintheleft-to-rightdirection.33 Fragilevasescanbewrappedinprotectivepackagingsothattheydon’tshatterwhenstruck.Suchvasesretainalltheirintrinsicproperties,andoughtthereforetoretaintheirdispositions,butLewis’analysisentailsthatsuchvasesdon’thaveDS.Poisonoussubstancesaredisposedtokillwheningested,butdon’tcausedeathwhentakenwiththeirantidotes.Butantidotesdon’taltertheintrinsicpropertiesofpoisons,whichoughtthereforetoretaintheirdispositiontokillwheningested.

It’s difficult to see how to rule out finks and antidotes in aprincipledway.Amaskorantidoteforonedispositionneedn’tworkon another. For this reason,modifications that appeal tonormal oridealconditions,or(ifdifferent)ceteris paribusclauses, tendtowardsvacuity:‘thevasehasDSiffwereitstruckinidealconditions,thenit

31. Martin(1994).

32. Lewis(1997);simplifiedforexposition.

33. Johnston(1992);Bird(1998).

groundingthedispositionsoftheirbearers.29Thisisacommonplace,althoughsometimesstatedastheclaimthatpropertiesbestowcausalpowers.30Thecentralideaisthatthingshavedispositionstoissueincertaintypesofeffectswhenappropriatelystimulated,andthatthesedispositionsaregroundedintheirintrinsicproperties.Bygroundingthedispositionsofparticulars,efficaciouspropertiestherebygroundcausalrelationsinvolvingthoseparticulars.

Thereadermayworrythatthisunderstandingofcausalworkisn’tonethatHumeanswouldaccept,soalittleclarificationofmyaimsisinorderbeforeproceeding.Ithinktherearegoodindependentgroundsfor understanding causal work in terms of grounding dispositions,and that thismakesgoodsenseofwhatexclusionistshave inmindwhentheysaythereisn’tanycausalworkleftforphysicallyrealizedfunctionalpropertiestodo.Thinkingofcausalworkthiswaymakesfor persuasive arguments that neither the over-determination nordifference-making strategies work as responses to the exclusionproblem. However, I don’t appeal to causal work as grounding toargue against broadlyHumean approaches tomental causation.Onthe contrary,what I propose to do is: (i) grant the exclusionists anindependentlyplausiblenotionofcausalworkthatshowswhyneitherthe over-determination nor difference-making strategy solves theexclusionproblem;(ii)grantHumeansthedifference-makingtheoryofeventcausation;and(iii)showthatgivendifference-makingcausation,there’splentyofnovelcausalworkforfunctionalpropertiestodo,sothatHumeansneedn’trejecttheviewthatpropertiesdocausalworkbygroundingcausalpowers.

29.Theclaimthatdispositionsaregroundeddoesn’tentailthattheyhavecategor-icalgrounds.Mypositionisconsistentwithdispositionalessentialismaboutfundamentalproperties,whichIcanunderstandastheclaimthatsuchprop-erties essentially ground certain dispositions. See Shoemaker (1980); Bird(2007).Icanalsotreatfundamentalpropertiesaspure,ungroundedpowers,asinMolnar(2003),providedsuchpowersaren’tidentifiedwithdispositionsinmysense,andsocanbethoughtofasgroundingthem.

30.Shoemaker(1980);Wilson(2002);McLaughlin(2006).



hasDS.Similarly,a carefullypackagedvaseneedn’tbesopackaged,soaccordingto(DISP)willhaveDSeventhoughtheclosestworldswhereit’sstruckareoneswhereitdoesn’tbreak.

A prima faciedifficultyariseswith“reversefinks”.Consideracubicblockofgranite,andsupposeGandalfhatestheblock’sshapesomuchthatwheneverit’sstruck,hechangesitsatomicstructuresothattheimpactissufficienttobreakit.Theblockhasanintrinsicproperty—itsshape—in virtue ofwhich there’s a rangeof conditions inwhich itwouldshatterwhenstruck,sothestonehasDSafterall.Italso,however,hasintrinsicpropertiesinvirtueofwhichthere’sarangeofconditionsin which it wouldn’t shatter when struck—a proper subset of thepossibleconditionsinwhichnowizardhatescubes.Whatthismeansisthattheblockisbothdisposedtobreakwhenstruckbyanobjectofmomentum≥ m,anddisposedtoremainintactwhensostruck.ThedispositiontoMwhenCisnot,accordingto(DISP),acontradictoryof thedisposition tonot-MwhenC—provided there’sno overlap inthepossiblebackgroundconditionsinwhichitwouldMwhenCandthoseinwhichitwouldn’t.Notethatthisisn’tthecontext-sensitivityothershavepointedoutinthesatisfactionofpredicateslike‘fragile’.37 A chair regarded by its Lilliputian designers as robustwill be seenas fragilebyGulliver,butbothcanagree that it’sdisposed tobreakwhenGulliversitsonit.Theywillsimplydisagreeaboutwhetherthisdispositionoughttobecountedasacaseoffragility.Thatthingscanbedisposedtoshatterwhenstruckanddisposedtonot-shatterwhenstruckhasnothingtodowiththiskindofcontext-sensitivity,butdoespointtoanotherkind.Ifthebackgroundconditionsinwhichgraniteblockswould shatterwhenstruckwere likely toobtain, rather thandistantnomicpossibilities,wemightwellregardsuchthingsasfragile.

3.2. Causal work as grounding dispositionsI take grounding to be a transitive, irreflexive, and asymmetricrelation which holds between facts—construed as things having

37.Mumford(1998);HawthorneandManley(2005).

wouldbreak’isn’tterriblyinformativeifidealconditionscanonlybespecifiedintermsoftheirenablingthemanifestationofDS.

34Aplanetwould shatter when struck in those conditions in which it wouldshatterwhenstruck—evenifnosuchconditionsarepossible.Allofwhichisbothersomeforme,sincethecentralargumentofthispaperdependsonconstruingcausalworkasthegroundingofdispositions.Withoutsayingsomethingaboutdispositions, itwon’tbeclearhowpropertiesgroundthem,orwhyweshouldcall itcausalworkwhentheydo.Soheregoes:35

(DISP) x isdisposed toMwhenC iffx has an intrinsicproperty P in virtue of which [there is a set ofnomically possible background conditions {B1, …,Bk} such that ifxwere inanyof theBi andCoccurred,xwouldM.]

(DISP) says nothing about the nature of the conditions in whichx would M if C, except that they are nomically possible. Treat x’sbeing insuchconditionsasa relationalpropertyofx. I say that theintrinsic properties of things determine the range of conditions inwhich theywouldexhibitcertainresponses tocertainstimuli.Moreformally,onmyaccount,thegroundingpropertyPofthedispositionto M when C is an intrinsic physical property in virtue of which:$B1,… ,Bk∀i[{Bi(x)∧C(x)}□→M(x)], where 1 ≤ i ≤ k. (DISP) handlesfinksandmasks.Assume for thesakeofargument thatwizardsarenomically possible.36 No vase is necessarily Gandalf’s favourite, soevenavaseprotectedbyGandalfhasintrinsicpropertiesinvirtueofwhichthere’sarangeofnomicallypossibleconditionssuchthatifitwerestruckinthoseconditions,itwouldshatter.Suchavasetherefore34.Martin(1994);Fara(2005).

35. Mycentralargumentswillgothroughonotheraccountsofdispositions,forinstanceLewis(1997);Fara(2005).Ingeneral,myaccountwillworkforanyaccountofdispositionsthatexplains(i)theirgrounding,(ii)therelationshipbetweengroundingdispositionsandcausality.

36. Itdoesn’tmatteriftheactuallawsofnatureruleoutwizards,becauseIwant(DISP)tobetrueatworldswherethelawsofnaturethatholdtheredon’t.



thatthereareconditionsinwhichitcanshatterobjectsincapableofabsorbingmorethanacertainamountofenergywithoutshattering.It’s because these two sets of conditions overlap that hammers arebestkeptintheshed,wheretherearenovases.Let’sturnnowtothequestionofwhat’scausalaboutcausalworkasIunderstandit.

Ordinarycausaltalkenablesustoidentifypropertiesthatgrounddispositions, and the stimulus conditions of those dispositions, ascauses.41 In ‘the vase broke because of its atomic structure’ we dothe former; in ‘thehammer’s impact caused thevase to shatter’, thelatter.Thinkingoftherelataofcausationasproperty-instancesblursthisdistinction:thelocution‘thevase’shavingthepropertyofbeingstruckbyahammerwithmomentum≥ m’refersnotonlytopropertiesthatdocausalworkingroundingthepowerofthehammertoshatterthevase,butalsotothestimulusconditionofthatdisposition—thestrikingofthevasebythehammer.Whetherornotthisconditionismetmakesadifferencetowhetherornotthevaseshatters.Sincestimulusconditionsdon’tgroundthedispositionswhosestimulusconditionstheyare,butdomakeadifferencetotheirmanifestations,itfollowsrightawaythattherearedifference-makingcausesthatdonocausalwork, which should give us grounds for doubting the difference-makingsolutiontotheexclusionproblem.Idon’tdependonthisinwhat follows, however, and for now will focus on how groundingdispositionsalsogroundsdifference-makingcausalrelationsbetweenstimulusconditionsandmanifestations.

A vase’s intrinsic properties determine a range of possibleconditionssuchthatwereitstruckinanyofthoseconditions,itwouldshatter.Supposesuchaconditionobtainsattheactualworld,andthatavaseisstruckandshatters.Theclosestpossibleworldstoactualityatwhichthevaseisstruckwillbeworldsatwhichthesamebackgroundconditionshold:thevaseisn’tbubble-wrapped,anditisn’tGandalf’sfavourite.Clearly thevasewillalsohaveDSat theseworlds,andsoshatters. Hence, at ourworld the vase is such that (a) had it been

41. Wecanalsoidentifythedispositionsthemselvesascauses.MoreonthisinSection3.4.

properties—andreflectsontologicalpriority.38IfanentityxhavingadispositionDisgroundedin itshavingsomepropertyP, thenxhasDinvirtueofhavingP,andx’shavingPismorefundamentalthanitshavingD.Sincegroundingentitiesaremorefundamentalthanthosetheyground,nothingcanbeitsownground.(Fundamentalentitiesareungrounded,notself-grounding.)Further,ifPgroundsD,itcan’tbethecasethatDgroundsP;forotherwiseeachwouldbemorefundamentalthan the other. So understood, groundingmust be both irreflexiveandasymmetric.IholdthatgroundingistransitivebecauseItakeX’sgroundtobeametaphysicallyexplanatoryreasonforX.Ifpsychologyisgroundedinneuroscience,forinstance,thenneuroscienceexplainswhy we have psychological properties. But if psychology, in thissense,groundseconomics,thenItakethattoimply—atleastinsome sense—thatneuroscienceexplainseconomics.39

Intrinsic properties determine how things would respond tovariousstimuliinarangeofnomicallypossiblecircumstances.Looselyspeaking, the vase’s atomic structure groundsDS bybeing the kindofstructurethatcan’tabsorbmorethanacertainamountofenergywithout undergoing the kind of rearrangement that counts as ashattering.Havingthisstructuredeterminesthatthereareconditionsinwhichthevasewouldbreakwhenstruckbyobjectsthathaveatleastthatmuchenergytogive.Theseconditionswillincludethatnothing(suchasbubblewrap)preventsthevasefromabsorbingtheenergyofthestrikingobject;thatnowizardlovesthevasesomuchhe’llchangeits structure so that it can absorb the energy without shattering;andsoon.Sincedispositionsmanifest reciprocally,40 the samegoesmutatis mutandis for the striking object—a hammer, say—whosepropertieswilldeterminehowmuchenergyitcantransfer,inarangeof conditions, to objects it strikes. Its properties thereby determine

38. I’lloftenspeakofpropertiesgroundingotherproperties,butthisisshorthand.

39. Isaymoreaboutthedifferenceinthesenseof‘grounds’between(e. g.)‘neu-rosciencegroundspsychology’and‘neurosciencegroundseconomics’inSec-tion4.2,whereIdistinguishproximalfromdistalgrounding.

40.Heil(2005);Martin(2007).



ionizing them by ejecting electrons, resulting in a positive charge.Mypointhere is that it isn’t the energy of thephoton that transfersenergy toelectrons in the tissues, causing theirejection.Havinganenergy greater than 1.022MeVmerely disposes the photon to bringaboutpairproductionwhenappropriateconditionsobtain,whichinturnproduceselectrons that aredisposed tobringabout ionizationoftissueswhenfurthersuchconditionsobtain.Propertiesgroundthedispositions of particulars, thereby determining the range of causalinteractionsinwhichtheycouldbeinvolved.

3.3. Functional realization as causal workFunctional properties are physically realized, and their realizersdo all the causal work associated with their defining roles. Giventhat causalwork is grounding dispositions, it follows that realizersgroundallthedispositionsparticularsneedinordertoenterintorole-definingcausalrelations.Wemightsaythatphysicalpropertiesrealizefunctional properties by grounding a certain characteristic set ofdispositions,withrealizationontologicallydependentongrounding.Inthissection,Iarguethatthisisdouble-counting,becauserealizationand grounding are the same relation. I first argue that functionalproperties,astypicallyconstrued,aresetsofdispositions,andproceedtoshowthatgiventhis,realizinga functionalproperty isgroundingit.Functionalpropertiesdon’tjustdependuponthecausalworktheirrealizersdo, they are (at least someof) that work.44Aswe’ll see, thisidentityofrealizationandcausalworkrenderstheover-determinationand difference-making strategies ineffectual against the exclusionargument. It also paves the way for my positive theory of mentalcausation,which I present in part (4). The arguments I give belowdependon(DISP),althoughtheycould,Ithink,beadaptedtootherbroadlyHumeantheoriesofdispositions.(DISP)suggests(althoughitdoesn’tentail)thatdispositionsaresecond-orderproperties;Itreatthemassuchinwhatfollows.

44.Manyarepreparedtoacceptwithoutargumentthatfunctionalpropertiesarecomplexdispositions.SeeforinstanceMcLaughlin(2006).

struck, it would have shattered. The vase’s intrinsic properties alsodetermineotherdispositions,suchasthedispositiontoremain intact, DI,whenstruckbyanobjectofmomentum<m,ornoobjectatall.TherewillbesignificantoverlapinthebackgroundconditionsforDS andDI; in fact,DIwasmanifesting rightbefore thevasewas struck.GiventhattheshatteredvasewasinbackgroundconditionsforbothDIandDS,itfollowsstraightaway(assumingalsothatit’sinthesameconditionsattheclosestworldstoactualitywhereitisn’tstruck)that(b)haditnotbeenstruck,itwouldn’thaveshattered.Nowgiventhedifference-making theory of causation, it’s clear how grounding anobject’s dispositions also grounds its causal potentialities, because(a)and(b)arejointlynecessaryandsufficientforthestrikingtobeadifference-makingcauseoftheshattering.42

One might object at this point that it is things, not properties,thatdocausalwork.Thehammerexertsaforceonthevase,therebytransferringenergytoit—aninformalversionofthephysicist’snotionofwork.Theexclusionproblem,however,dependsontherebeingasenseof‘causalwork’thatmakesittruethatpropertiesdocausalwork,and transferring energy isn’t it. To see this, consider that energy isitselfacausallyefficaciouspropertyparexcellence.Gamma-rayscauseskinburnsandradiationsicknessinvirtueoftheirhighenergy,andthecausalprocessesbywhich theydoso involves transferof someof that very energy to the unfortunate victim.One such process ispairproduction,wherebyaphotonofenergygreater than 1.022MeV interactingwithaheavynucleusproducesanelectron-positronpair.43 The electron, for instance, may then interact with organic tissues,

42. TherewillbeotherHumeantheoriesofcausationanddispositionsthathavethesameexplanatoryvirtues,somycentralargumentsdon’tdependonthepresentonesbeingcorrect.Perhaps the reader isprepared toaccept,with-outaparticulartheoryofcausationorofdispositions,thatwhatgroundsx’sdispositiontoMwhenCtherebygroundsthecausalrelationthatobtainsonsomeoccasionwhenCoccursandxMs.

43. Gammaphotonsofenergygreaterthan2mec2arerequired(wherem is the

electronicmass,andcthespeedoflight),withtheenergyofthephotonbeingconverted,accordingtoEinstein’senergy-massequation,intotherestmassesoftheelectronandpositron.Thedetailsneedn’tconcernushere.



weonlyneedtoRamsifyawaythementalpredicates,andthecausal-nomiconesaren’tmental.45Sofar,Iclaimonlythatif(2)istodefineafunctionalproperty,R(T)mustspecifythecausalstructurethatthepropertiesP1,… ,Pnmust satisfy inorder to count as realizersof F1, … ,Fn. But now if, as I amassuming, causation is difference-making,thenR(T)mustcontainaconjunctionofsubjunctivesspecifyingthedifferences physical properties need to make in order to count asrealizersofmentalproperties.SupposeforthesakeofargumentthatTisfolk-psychology,andthatpartofwhatTclaimsaboutthepropertyWofwantingsomeX isthatgivenaspecifiedpsychologicalstateFT, wantingXcausessoactingastoacquireit.46Givendifference-makingcausation,thiscomponentofTcanbewritten:

(S) ∀x[{[FT(x)∧W(x)]□→A(x)}∧{[FT(x)∧ ¬W(x)]□→¬A(x)}]

whereAisthepropertyofsoactingastoacquireX.HoldingFTfixed,wantingXmakesadifferencetowhetheryoutry toget it.R(T)willpreserve this subjunctive structure as a constraint on the Pi , so (2)defines functional properties bymeans of subjunctive conditionals,which is how (1) defines dispositions. My argument assumes thatRamsey sentences specify causal structure using causal predicates.Readers persuaded that psychological theories have an explicitlysubjunctiveformwon’tneedtocommittothisclaim,orthedifference-makingtheoryofcausation,toseethissimilaritybetween(1)and(2).

45. ThereaderwillberemindedhereofthefamiliarNewmanobjectiontostruc-tural realism in thephilosophyof science. Structural realists often employRamseysentencestodefinetheoreticalterms,butifthenon-theoreticaltermsleftun-Ramsifiedconsistsolelyoflogicalandobservational(inthepositivists’sense) vocabulary, then it’s provable that the Ramsey sentence constrainsonlythecardinalityofthetheoreticaldomain.Thesemattersarebeyondthescopeofthispaper,butnotethatstructuralrealistswantingtoavoidtheNew-manobjectioncanarguethatcausalvocabularybelongsintheO-languageandsoshouldn’tbeRamsified. Ineedn’targueanythingsostrong: formypurposesitsufficesthatthecausalvocabularyisn’tintheF-language,i. e.,thatit’snon-mental.SeePapineau(2010),pp.381–2fordiscussion.

46. FTwillbeaconjunctionofothermentalpropertiessuchasnothavingstron-gerdesires fornot-X,believing that theactionsnecessary toacquireX aremorallypermissible,etc.

Using‘λx.p’todenotethepropertyofbeinganxsuchthatp,lettingthe existential quantifier range over intrinsic physical properties,andwhere1≤ i ≤ k,wecanrepresentthedispositionDtoMwhenCasfollows:

(1) D=λx.$P[P(x)∧{$B1,…,Bk∀i[{Bi(x)∧C(x)}□→M(x)]invirtueofP}]

Given(1),DisthepropertyofhavinganintrinsicpropertyPinvirtueofwhichthereexistconditionssuchthat if thebearerwereinthemandsubjecttoC,itwouldM.Functionalismaboutthemindistypicallycharacterisedbyappealingtopsychologicaltheories:agivenfunctionalpropertyisthepropertyofhavingsomepropertythatoccupiesacausalrole specified (in a way to be clarified) by that theory. In the nowstandardway,let’swritethetheory(whateveritmaybe)asT(F1,…,Fn;O1,…,Om),wheretheF-termsarethepredicatesthatdenotementalproperties,andtheO-termsareeverythingelse.ReplacingtheF-termswith appropriately indexed variables, and prefixing the resultingformulawith an existential quantifier, we get T’s Ramsey sentence,R(T):$P1,…,Pn[T(P1,…,Pn;O1,…,Om)].FunctionalistsdefinetheithmentalpropertyFiasfollows:

(2) Fi=λx.[$P1,…,Pn[T(P1,…,Pn;O1,…,Om)∧Pi(x)]

Given (2), Fi is the property of having some other property thatoccupiesaspecificplaceinanetworkofstatesthatjointlysatisfyT.Itisn’timmediatelyobviousfrom(1)and(2)thatfunctionalpropertiesare dispositions, but I’ll now argue that on reasonable assumption,theyare.Idon’thaveanythinglikeaproof;rather,Isuggestaseriesof further constraints on (2) if it is to adequately define functionalproperties,witheachonebringingitcloserto(1).

InordertodefineaspecificfunctionalpropertyFi ,theright-handsideof(2)mustspecifyacausal-nomicstructureandthepartofitoccupiedbyFi ’srealizerPi .Forthistobethecase,theO-languagemustcontaincausal-nomic predicates. This isn’t a problematic requirement here,becauseforpresentpurposesthedefiniendaarementalproperties,so



are properties of the same agent—then T encodes such stimulusconditions in virtue of quantifying over agents. Perhapswe can dobetter.SupposeIwantsomefood,butdon’tyetknowit—distractedbyphilosophicaltheorising,Ihaven’treflectedonmycurrentdesires.Partof thepsychologicalstateFT relativetowhichmywantingfoodmakesadifferencetomyhavingsomewillbemyentertainingboththedesireandtherestofFT.Perhapsjointlyentertainingbeliefsinadeliberative process is, in a sense, bringing them into contactwitheachother.Ifso,thenTwillcontaintermsforconditionsanalogoustothestimulisuchascontacts,impacts,andsoforth,towhichweappealwhendefiningdispositionslikefragility.

Andfinally,(c).ItreattherealizersofthedispositiontoMwhenCasdeterminingarangeofconditionssuchthatiftheirbearerswereinthoseconditionsandChappened,theywouldM.AswesawinSection3.1,thisenablesmyapproachtodealwithfinksandmasks.Iarguedabove that ifR(T) is todefineacausalstructure, thenTmustmakegeneralcausalclaimsaboutmentalproperties,andhence,assumingdifference-making causation, must imply subjunctive claims of theformgiven in (S). Iomittedceteris paribus (cp)clauses for simplicity,butit’snowtimetoputthembackin.Sincethepsychologicaldomainisn’t causally closed,Tmust contain implicit cp clauses—there isn’t aconjunctionofpsychologicalpropertiessuchthatifyouhaveallofthem,thenwantingfoodmakestheappropriatedifferencetoeating.Suppose,forinstance,thatthere’sadrug—backwards powder—whoseprimaryeffectisthathumanswhohaveFTandingestitaresuchthatiftheywantedfood,theywouldn’thaveany,andiftheydidn’t,theywould.We might deal with the implicit cp clauses in T by simplyincorporating them into (2), qualifying the truth of the relevantsubjunctives.Or,asIprefer,wecoulddealwiththembysayingthatthe realizerpropertiesP1,… ,Pn aren’tproperties invirtueofwhichthosesubjunctiveshold,butpropertiesinvirtueofwhichthereexistarangeofconditions inwhichtheyhold.Onesuchcondition, forus,istheabsenceofbackwardspowder,butthisisn’tsoforallpossiblerealizersofourmentalstates.Martiansareacaseinpoint:onMars,

Severaldifferencesremain:(a)(2)doesn’tsayanythingaboutitsbeingin virtue ofthePithatthespecifiedcausalstructureobtains,andmakesno explicit mention of (b) stimulus conditions, or (c) backgroundconditions.I’llnowbrieflyaddressthesepointsinturn.

First, (a). It’s widely supposed in the literature that Ramsifiedtheoriescanfullydefinephysicallyrealizedfunctionalproperties,butifcausationisdifference-making(ormoregenerally,Humean), thenthis isn’t so. The reason is that implicit in the notion of realizationis the claim that the physical realizers of a functional property doallthecausalworkassociatedwiththerolethatdefinesit.That,asIsaid,iswhythere’sacausalexclusionproblem.InthenextsectionI’llpresentanargumentthatdoesn’tdependontheviewthatfunctionalpropertiesaredispositions(andonethatdoes)totheeffectthatmakingadifference(ormoregenerally,standinginaHumeancausalrelation)tosomeeffectisn’tsufficientfordoinganyofthecausalworkinvolvedin itsoccurrence.Thisbeingso, if theRamseysentenceonlyplacesHumeancausalconstraintsontherealizers,itstruthwillbeconsistentwiththoserealizersnotdoinganyofthecausalworkassociatedwiththerolestheyrealize.Inadditiontosuchcausalclaims,then,weneedtoincludein(2)thefurtherstipulationthatthePistandintheirvariousrole-definingcausalrelationsin virtue ofthePi .

Let’smoveonto(b).(1)characterisesadispositionDintermsofastimulusconditionCandmanifestationM,whereas(2)characterisesa functional property Fi in terms of a network of physical states,causally related (in virtue of their physical properties) as specifiedin T. I claim that the Fi , so understood, are dispositions, whichwill sometimes manifest as behaviour, sometimes as other Fi ; butwhere are their stimulus conditions? What makes for the prima facie disanalogy between (1) and (2) is that it’s natural to think ofhammersashavingdispositionstobreakvases,whichhavereciprocaldispositions to be brokenbyhammers; and in addition to supposethatthestimulusconditionsofthesedispositionsarethattheirbearerscomeintocontact.Where’sthepsychologicalanalogue?Supposethestimulusconditionofpsychologicaldispositionsaresimplythatthey



for the dispositions itself to do. All the causal work isbeingdonebythebondingPtogetherwiththedropping.

Why, goes the standard over-determinationist response, can’t thefragility (F)do the same causalworkasP? Jacksondoesn’t say,butgiventhatthecausalworkofPconsistsingroundingF,theanswerisclear:because grounding is irreflexive. It isn’t justthatthere’snocausalworkleftforFtodo;theredoesn’tseemtobeanyworkitcouldpossibly do.WecannowseewhyHumeantheoriesofcausationdon’tcapturethenotionofcausalworkatstakeintheexclusionargument.Suppose,forinstance,thatthenomic-subsumptiontheoryofcausationistrue.Dispositionssuperveneon theirphysical realizersandareplausiblylawfullyrelatedtotheirownmanifestations,buttheyarepatentlynotself-grounding.Thesamegoesmutatis mutandisforanycounterfactualtheories, such as the difference-making theory, which entail thatdispositionscausetheirownmanifestations.48

Difference-makingtheoriesofcausation,whentakenassufficientconditions on causal work, imply that dispositions do causal worktheirrealizersdon’t.ConsiderthefragilityofthevaseFanditsphysicalrealizer P. Assuming the vase’s shattering, S, is realizer-invariant, FmakesadifferencetoSthatPdoesn’t.Theclosestnon-Fworldsarenon-S worlds, and the closest F-worlds are S-worlds, which is allthat’srequiredforFtomakeadifferencetoS.Bycontrast,theclosestnon-P worlds are worlds where some other realizer P’ of F occurs,togetherwithadifferentrealizerofS.Sincetheclosestnon-PworldsareS-worlds,Pdoesn’tmakeadifferencetoS.FindexesthevasetoadifferentportionofmodalrealitythanP,andsomakesadifferencetoS thatPdoesn’t.49Thatdispositionsaredifference-makingcausesoftheirmanifestationsshowsthatmakingadifferenceisn’tsufficient

48. Thereadermayworrythatdispositionsaren’tcausallyrelevanttotheirmani-festations,butasMcKitrick(2005)argues,independentlyplausibletheoriesofcausalrelevanceimplythattheyare.Ifwetreatdifference-makingasanaccountofcausalrelevance,it’sacaseinpoint.

49. ThiswillproveimportantinSection4.4.

backwardspowder is anaturallyoccurringmineral the ingestionornon-ingestionofwhichmakesnodifferenceatalltowhetherornottheysatisfyT.

Functionalist mental properties are, I conclude, a complex kindofdispositionalproperty.Not,itmustbesaid,thepropertyofhavingaproperty invirtueofwhichthereareconditions inwhicha certain subjunctive is true; rather, the property of being in one of a range of states in virtue of which, collectively, there exist conditions inwhichalargeconjunctionofsubjunctivesistrue.Buttheyarenolessdispositionalfortheiraddedcomplexity.

3.4. Why Humean causation isn’t sufficient for causal workAlthough I endorse the difference-making theory of causation,findingcausalworkforapropertytodoisharderthanshowingthatits instancesare causes. I arguedabove that (A) thecausalworkofproperties consists in grounding dispositions; and (B) functionalproperties are complex, physically realized dispositions. It followsfrom(A)and(B)thatfunctionalrealizationisgrounding:functionalpropertiesare(atleastsomeof)thecausalworkthattheirrealizersdo.Thislicensesasimpleargumentfortheviewthat(C)beingaHumeancauseofsomeeffectisn’tsufficientfordoinganyofthecausalworkinvolvedinitsoccurrence.Thosewhoargueagainstthecausalefficacyof dispositions, construed as second-order functional properties,simplyemploytheexclusionargument.Here’sJackson: 47

Consider…afragileglassthatshattersonbeingdroppedbecause it is fragile… .Therewill be… a certain kindof bonding P between the glass’s molecules which isresponsiblefortheglassbeingsuchthat[itbreakswhendropped]….ButthenitisbondingPtogetherwiththedropping thatcauses thebreaking; there isnothing left

47. Jackson(1996),p.393.Adjustedfortypographicconsistency.SeealsoPrior,PargetterandJackson(1982);McLaughlin(2006).



in groundingDi , and again, since grounding is asymmetric andDi

groundsF,thisisworkthatFsimplycan’tdo.51

Atthispointonemightsuggest thatexpectingmentalpropertiestodo causalwork inmy sense is expecting toomuch.52 If this is toprovideforareplytotheexclusionargument,thenifweunderstandaproperty’shavingcausalpowersintermsofitsdoingcausalwork,andunderstandcausalworkintermsofgroundingdispositions,thenwe’llhavetoreformulate(AD),becauseasthingsstand(AD)requiresreal,irreduciblepropertiestobeirreducibleand to have causal powers.Thisamountstoacceptingdifferentstandardsforfunctionalpropertiesandat leastsomephysicalproperties.Physicists, for instance,commit topropertiesonthebasisofgenuinelynovelcausalwork,butperhapsthese standards aren’t appropriate in psychology. This approachattempts to solve the exclusion problem by admitting that mentalpropertiesdonocausalwork,thenarguingthatthisdoesn’tresultintheireliminationbecausetheconditionsforontologicalcommitmenttosuchpropertiesarelooserthanthoseforphysicalproperties.Settingpsychology aside, science isn’t all physics, and there are plenty offunctional properties in biology and neuroscience that don’t earntheirontologicalkeepinthesamewaytheirrealizersdo,buttowhoseexistencebiologistsandneuroscientistsarenonethelesscommitted.

Ihaveacertainsympathywith thiswayof thinking,but itwon’tdoanythingtoconvincethedeterminedexclusionist,whowillsimplyreply that when it gets down to brass tacks, there aren’t really anybiologicalorneuroscientificpropertieseither.Retreatingtothepositionthat functional properties earn their keep by making a difference,despitethefactthattheydonocausalwork,reducesthedisagreementbetweenexclusionistsandHumeans,whichusedtobeaboutwhetherthere’sanycausalworkleftforfunctionalpropertiestodo,tooneaboutwhatourstandardsforontologicalcommitmenttopropertiesoughtto

51. Itshouldbenotedthatmyargumentsagainstboththeover-determinationanddifference-makingstrategiestellonlyagainsttheuseofthesestrategiestoaccountfortheefficacyoffunctional(i. e.,dispositional)properties.

52. LePoreandLoewer(1987).

for doing causal work. The property F is the glass’s disposition toshatter when dropped, and so, since grounding is irreflexive, can’tground that disposition. Difference-making causation identifies thecausalworkdoneasacause,whichseemsright.Causalworkmakesadifferencetowhathappens;ifitdidn’t,whatwouldbethepointofdoingit?Butthismeansweshouldn’texpectalldifference-makerstodothekindofcausalworktheexclusionargumentdeniestomentalproperties.50Conversely,notallpropertiesthatgroundadispositionmakeadifferencetoitsmanifestations.

ItwillbenotedthatinSection3.3,Ipresupposed(C)aboveaspartofmyargumentfor(B).Buttheargumentfor(C)justgivenpresupposes(B),so Istillowethereadera furtherargument for (C) thatdoesn’tdependon(B).I’llnowarguethat(C)followsfrom(A)alone,togetherwiththereasonableclaimthatrealizationisaformofgrounding,evenif functional properties aren’t dispositions. Functional property F isthepropertyofhavingsomepropertyPwithcausalroleR.ForP tohavecausalroleR,given(A),isforittogroundthedispositionsthatcharacteriseR:labelthissetofdispositionsDi .GiventhatPrealizesFbydoing thecausalworkassociatedwith roleR, it follows thatPrealizesFbygroundingDi .Itdoesn’tmatterwhichPgroundsDi—it’sDithatreallymatterstoF’srealization.SinceFontologicallydependsonDi ,andhavingDi(groundedbysomeP)issufficientforhavingF,itfollowsthatDigroundsF.ThereasonwhyFcan’tinheritthecausalwork of P is simple: because grounding is asymmetric. F ontologicallydepends on the causalwork that P does, so the supposition that FinheritsthisworkfromPisincoherent.Similarconsiderationsapplymutatis mutandistothedifference-makingstrategy.Forfamiliarreasons,FwillmakedifferencestothemanifestationsofthedispositionsinDi

thatPdoesn’t.Butfromthiswecan’tinferthatFdoesanyofthecausalwork required for those manifestations, since that work consists

50.Cranemakesasimilarpointconcerningcounterfactualsandthecausaleffi-cacyofpropertiesinhis(2008).Crane’sargumentsforthisconclusionaren’trelatedtomine.



Now if we interpret Alexander’s Dictum (AD) as the claim thatpropertiesmust do at least some causalwork (novel or otherwise)in order to earn their ontological keep, it seems to follow straightaway that functional properties ought to be eliminated, since: (i)supervenientpropertiesingeneralcan’tdocausalworktheirphysicalbasesdon’tdo,sincethatwouldrenderthemontologicallyemergent,violatingcausal closure; (ii) functionalproperties can’tdo the samecausalwork as their realizers, since theyare (at least someof) thatwork. True, functional properties make a difference their realizersdon’t, but that doesn’t matter, because making a difference isn’tsufficientfordoingcausalwork,andit’spreciselybecausetheyseemto do no causal work that functional properties face the threat ofelimination. I now present a theory according to which functionalpropertiesdonovelcausalworkwithoutgroundinganydispositionsthataren’tgroundedinthephysical,andsowithoutdoinganycausalworkthatphysicalpropertiesdon’tdo.Accordingtothistheory,thecausalnoveltyoffunctionalpropertiesconsistsnotinthedispositionstheyground,butinthelevel,withinahierarchyofgroundingrelations,fromwhichtheygroundthem.Icallitupwards causation.

4.1. Upwards Causation54

IwillsaythatamechanismXisindividuatedby(i)possessionofasetDofdispositions,and(ii)asetofcomponents{x1,…,xn},eachhavingfurthersetsofdispositions{D1,…,Dn},suchthat{x1,…,xn}having{D1, …,Dn}constitutivelyexplainswhyXhasD.SupposeIwishtoconstructa mousetrap using the following items: (a) a 1,000,000V battery,(b) conductingwireof resistance 10Ω, (c) a thin sheetof copperofnegligibleresistance.Ifirstcutthewireintwo,andattachoneendofonepiecetothepositiveterminalofthebattery,oneendoftheothertothenegative.Ithenattachtheotherendofthepositivewiretothe

sufficiency)andsynchroniccausalwork(completeness).SeeYates(2009)fordetaileddiscussion.

54. TheaccountIgivehereowesmuchtoCraver’s(2007),inparticularchapters4and5,butIdon’tattributewhatfollowstohim.

be.Farbetter forHumeans tograntexclusionists themorestringentcondition and then show that functional properties meet it. In theremainderofthispaper,Iarguethathigher-orderfunctionalpropertiescandothesamekindofcausalwork—butnot,astheargumentsofthissectionshow,thesamework—astheirphysicalrealizers.

4. Mental Causation for Functionalists

Thecausalworkofpropertiesconsistsinsynchronicallygroundingthedispositionsoftheirbearers;findingsuchworkforfunctionalpropertiestodoisharderthanshowingthattheirinstancesarediachroniccauses.I’llnowbrieflyrecasttheexclusionargumentintermsofcausalwork.InSection2.1,Idefinedclosure(CC)astheclaimthateveryphysicaleffecthasa complete, sufficientphysical cause, so that I could laterdefine‘complete’intermsofmypreferredconceptionofcausalwork:put simply, a complete, sufficientphysical causeof someevent is asufficientcauseofitwhosephysicalpropertiesgroundalltherelevantdispositions.Thinkagainofavasebrokenbytheimpactofahammer.Sincedispositionsmanifestreciprocally,wemustincludeinthecausetheintrinsicphysicalpropertiesofthevaseitself,aswellasthoseofthehammer,sincetheformerdosomeofthecausalworkrequiredifthehammeristobreakit.Closureimpliesthatthephysicalpropertiesofhammerandvasegroundboththehammer’sdispositiontobreakthatkindofvase,andthevase’sdispositiontobebrokenbythatkindofhammer.Aphysicalcausecanbesufficientforsomeeffectwithoutbeingcomplete.Supposeemergentdownwardscausationispossible,whereby a supervenient property-instance emerges from a physicalbase property-instance, but contributes novel causal powers notgroundedby itsphysicalbase.Given the sufficiencyof thephysicalbasefortheemergentproperty,theeffectsofemergentpowershavesufficientphysicalcauses,bytransitivityofsufficiency;buttheydon’thavecompletephysicalcauses.53

53. AsIunderstandit,(CC)claimsnotonlythatphysicaleffectshavesufficientphysicalcauses,butalsothatthepropertiesthatgroundsuchcausalrelationsarephysical.Sounderstood,(CC)isaboutbothdiachroniccausation(causal



beingasetofpossibleconditions{B1’,…,Bn’}suchthatwerethesheetinanyoftheBi’andaforceof1Nwereapplied,itwouldflexby1cm.If the intersectionof these twosets isempty, themechanismwon’twork.Whendesigningamechanism,wemaximise the intersectionsetandmakesureitcorrespondstotheconditionsunderwhichwewant themechanism towork. This iswhy itmatters not only that, but also how, the dispositions of the components are realized.MymousetraphasDMbecause:(1)thecomponentshaveD1–D3;(2)thecomponentshaveacertainspatiotemporalstructure;(3)D1–D3haveanon-emptyintersectionsetofbackgroundconditions.Call(1)–(3)thedispositional structureof themousetrap.Suchstructurestypicallyinvolve components which are themselves mechanisms, and havetheir dispositional properties in virtue of their own dispositionalstructures. As in the example of the hammer and vase in Section3.2, the properties that groundDM thereby groundpotential causalrelations.Themousetrap’shavingDMconsistsinitshavingintrinsicproperties—its dispositional structure—in virtue of which there’sarangeofpossibleconditionsunderwhich,wereamousetositonthe copper sheet, it would be shocked.What grounds DM therebygroundspotentialdifference-makingcausalrelations:whetherornotthemousestepsonthesheetmakesadifferencetowhetherornotit’sshocked.

An obvious rejoinder:Why isn’t it the physical properties of itscomponents that do the causalwork of grounding the dispositionsofamechanism?Reply: It is!Thecausalpowersofamechanismaregrounded in the fundamentalphysicalpropertiesof itscomponents.However,Idenythatthisleavesnocausalworkforfunctionalpropertiestodo.Infact,Iclaimthatthegroundingofamechanism’sdispositionsbythebasicphysicalpropertiesofitscomponentsissecuredonlybythe transitivityofgrounding,and the fact thatsuchdispositionsaregrounded in the dispositional properties of its components. Thinkagainofthemousetrap,andfocusonthecoppersheet,andthephysicalpropertiesofthesheetinvirtueofwhichithasD2.SupposethecoppersheetbreaksandIneedtoreplaceitbuthaveonlyathick,inflexible

coppersheet,andset thesheet inahousingso that theunattachedendofthenegativewireis1cmawayfromthesheet.Thecomponentsofmymechanismhavethefollowingdispositions:(D1)thebatteryisdisposedtosendacurrentof(1,000,000/R)AthroughaconductorofresistanceRconnectedacross itsterminals;(D2)thecoppersheet isdisposedtoflexby1cmwhenaforceof1Norgreaterisapplied;(D3)thewireisdisposedtoconductacurrentof(V/10)Awhenapotentialdifference of V is applied to it. The mechanism as a whole—mymousetrap—is disposed to shock mice of mass greater than 100gwhentheysitonthecoppersheet.LabelthisdispositionDM.Amouseofmass200gsitsonthecoppersheet,meetingthestimulusconditionofthedispositionofthesheettoflexby1cm,bringingitintocontactwiththewire.Thisinturnstimulatesthereciprocaldispositionsofthebattery andwire, sending a current of 100,000A through the sheet,whichshocksthemouse.Theimportanceofstructureshouldbeclear.Werethenegativewirenotlocated1cmbelowthecoppersheet,thesheet’smanifesting itsdisposition toflexby 1cmwhenanobjectofmassgreaterthan100grestsonitwouldnottriggerthedispositionsof the other components. The way the components are structuredenables themanifestation of one disposition to be the stimulus ofanother.Ofcoursenoteverymechanismisstructuredinsuchalinearway, but I needn’t considermore complex cases tomakemy point.The entity that hasDM is themousetrap itself:DM isn’t identical toanydispositionthatcanbeattributedto the individualcomponents,eventhoughitsmanifestationonsomeoccasionisconstitutedbythestructuredmanifestationsofthedispositionsofthecomponents.

Nowconsider:whichpropertiesdothecausalworkofgroundingDM? A key element of my proposal is that DM is grounded in thedispositions of the mousetrap’s components, together with theirspatiotemporalstructure.Accordingto(DISP),thebattery’shavingD1 consistsintherebeingasetofpossiblebackgroundconditions{B1,…,Bn},suchthatifthebatterywereinanyoftheBiandaconductorofresistanceRwereconnectedtoitsterminals,itwouldconductacurrentof (1,000,000/R)A. The copper sheet’s having D2 consists in there



overparticulars(includingmechanisms),andDoverdispositions,wemaydefineproximalgroundingasfollows:

(PG) X’s having P proximally groundsX’s havingD iff (i)X’shavingPgroundsX’shavingD, (ii) there isnoP’suchthat:(a)XhasP’,and(b)X’shavingPgroundsX’shavingP’,and(c)X’shavingP’groundsX’shavingD.

Intuitively,proximalgrounding isgroundingwithout intermediaries.My mousetrap has many levels of mechanism: the battery, for instance, is a mechanism, which is disposed to send a current of(1,000,000/R)AthroughaconductorofresistanceRconnectedtoitsterminals,invirtueofitsdispositionalstructure—beingcomposedofananodeandcathode,separatedbyanelectrolyte,say.Thesamegoesforthecoppersheet,whichisdisposedtoflexby1cmwhenaforceof1Nisappliedinvirtueofthewayitsatomsarearranged,theirdispositionstoexertcertainforcesontheirneighbours,andsoforth.MymousetraphasdispositionalstructuresatlowerlevelsofmechanismthantheoneI’vebeendiscussingsofar,whichdon’tproximally,butdistally,grounditspowertoshockmice:

(DG) X’s having P distally groundsX’s havingD iff (i)X’shavingPgroundsX’shavingD,and(ii)¬(X’shavingPproximallygroundsX’shavingD).

Grounding simpliciter, construed as either proximally or distallygroundingadisposition,istransitive.Proximalgroundingisintransitive,because by definition if a proximally grounds b and b proximallygroundsc,thenadistallygroundsc;distalgroundingistransitive.Thebasic physical properties of amechanism at bestdistally ground itscharacteristicdispositions.Theroleoffundamentalphysicalpropertiesistoproximallygroundthepowersoffundamentalphysicalparticles.Suchparticlescombineintomechanisms(atoms)whichhavecertaincharacteristicpowersinvirtueoftheirdispositionalstructures.Theseinturncombineintomolecules,chemicalcompounds,cells,andsoonallthewayup.Thefundamentalpropertiesofmymousetrapdistally

aluminiumsheettohand.WhatImustdo,ifIwantmymousetraptocontinueworking, is tofile thesheet toa thicknesssuch that it tooflexesby1cmwhenaforceof1Nisapplied.Grantingforargument’ssake that a suitably thin sheet of aluminium also has negligibleresistance,thisexampleshowsthatwhat’simportantwhenitcomestogroundingDMisthatthecomponentsareappropriatelystructuredandhave D1–D3 (realized so that their background conditions overlap).Holdingthemousetrap’sothercomponentsfixed,andremovingthecopper sheet, whatwemust do to get it working again is install acomponentofnegligibleresistancewithD2.Putdifferently:invirtueofitsintrinsicnature,thecoppersheethasD2,andinvirtueofhavingD2 it’s capable,whenplaced inappropriate structural relationswiththeothercomponents,ofcompletingthemechanism.Thecausalworkthatthephysicalpropertiesofourreplacementsheetmustdoinorderto partially groundDM is to groundD2, thehavingofwhichmakesitpossibleforthesheettoformpartofadispositionalstructurethatconstitutivelyexplainswhythemousetraphasDM.

4.2. Proximal and distal groundingThephysicalpropertiesofcomponentsinamechanismdothecausalworkofgroundingthedispositionsofthosecomponents,whichthenground thoseof themechanism.Thesame is trueevenat the leveloffundamentalphysics.Anelectronorbitsaprotoninthehydrogenatomdue to thedispositional structureof electronandproton.Thephysicalpropertiesofelectronsandprotonsgroundthedispositionsof the hydrogen atom, by grounding the reciprocal dispositionsof electron and proton to attract each other. It’s unlikely that thedispositions grounded by fundamental properties such as chargearemultiply realizable. It remains true, however, that fundamentalparticles combine into more complex mechanisms in virtue of dispositions that theyhave in virtueof their fundamental properties.Assumingthenotionofgroundingtobesufficientlywellunderstood,wecandistinguishtwokindsofgrounding:proximalanddistal.WherePandP’rangeoverproperties(includingdispositionalstructures),X



grounded in both functional properties and their realizers; what’sincoherentisthestrongerclaimthattheformerinheritthecausalworkof the latter. Indeed, given upwards causation,whenbasic physicalproperties distally ground a power via functional intermediaries,itmakesmore sense to say that thephysical properties inherit thiscomponentof their causalwork from the functionalproperties thatproximallygroundit.

4.3. Upwards Causation and the MentalIn order to save functionalist mental properties from the threat ofelimination,weneedtofindnovelcausalworkforthemtodo.Theycan’tdothesamecausalworkastheirrealizers,becausetheyarethecausalworktheirrealizersdo.Wearenowinapositiontofindsuchworkforfunctionalmentalpropertiestodo.55Mentalpropertiesproximallyground agent-level dispositions not proximally grounded in agents’physicalproperties.Theyaredistallysogrounded,but thisdependsonthetransitivityofgroundingandthefactthatsuchdispositionsareproximallygroundedinthemental.Supposeforthesakeofargumentthattheambienttemperaturecomesintwoflavours:coldandwarm.I have the followingmental properties: (i) Iwant tobeneither toohotnortoocold,(ii)Iwanttogooutside,(iii)Ibelievethatifit’scoldoutside,thenifIgooutsideI’llbetoocoldunlessIdresswarmly,(iv)Ibelievethat if it’swarmoutside, then if Igooutside I’llbe toohotunlessIdon’tdresswarmly.Givenfunctionalism,(i)–(iv)are(clustersof) dispositions. For instance, (i) is—inter alia—the disposition todresswarmlyifIhavestates(ii)and(iii)andbelieveit’scoldoutside.WhatIdon’tyetknowiswhetherit’scoldorwarm:I’veonlyjustgotupandhaven’tyethadmycoffee,whichcomesbeforeopeningthedoortocheckthetemperature.Beingacomparativelynormal,rationalagentwithproperlyfunctioningsenses,I’mdisposedtobelievethatit’scoldoutsidewhenit iscold;dittowarm.Thesestatescombinetogroundthefollowingagent-leveldispositions:(a)I’mdisposedtodresswarmly

55.Whatfollowsisn’tintendedasanexhaustiveaccountofthecausalnoveltyofsuchproperties.

grounditspowertoshockmice,whichleavesplentyofcausalworkleftforthefunctionalpropertiesofitscomponentstodo—proximally grounding thispower.Proximalanddistalgroundsoccupydifferentplacesinahierarchyofgroundingrelationsthroughwhichthecausalinfluenceoffundamentalphysicsextendsupwardstomedium-sizeddry goods. This doesn’t involve causal closure violations, becausethere’snopressureatalltoreadthenotionofcausalworkimplicitin(CC)intermsofproximalgrounding.Indeed,readinthisway,(CC)is false, for there are many dispositional structures between basicphysicsand thepowersofordinaryphysicalparticulars.Conversely,ifwetake(CC)toimplythatthepowersofordinarythings—agents,engines, aeroplanes, batteries, bananas—to bring about certainphysicaleffectsaregroundedintheirfundamentalphysicalproperties,wemust read ‘grounded’ as “distally grounded”. Far from precluding the causal novelty of functional properties, (CC) actually requires it.Functional properties at a specific level of mechanism are causallynovelbecausetheyoccupyauniqueplaceinhierarchyofdispositionalstructures,withoutwhichbasicphysicalpropertiescouldgroundonlythepowersofbasicphysicalparticulars.

Upwardscausationmayinitiallystrikethereaderassimilartothekind of over-determination strategy I rejected in Section 3.4. There,I argued that it’s incoherent to suppose that functional propertiesinheritthecausalworktheirrealizersdo,becausetheyareatleastpartofthatwork.SinceIholdthatfunctionalpropertiesproximallygrounddispositions that are distally grounded in the physical, however, Imust accept that some causalwork is, in a sense, over-determined.Forupwards causation towork, theremustbedispositions that aregroundedsimpliciter(butnotproximally)inbothfunctionalpropertiesandtheirbasicphysicalrealizers,sogroundingsimpliciterissometimesover-determined. This doesn’t undermine my previous argumentsagainstover-determination,becauseitremainsthecaseiffunctionalproperties simply inherited the causal work of their realizers, theywould be self-grounding dispositions, which doesn’t make sense.There’snothingwrongwith supposing that therearecausalpowers



oftheircomponents;thatsuchmechanismsareultimatelygroundedin fundamentalphysics isonlypossibleatallbecauseof theway inwhich dispositions compose. Basic physical properties ground thepowers of complex mechanisms by grounding those of their mostbasiccomponents;fromthereonit’sfunctionalallthewayup.

The theory detailed above involves treating mental propertiesaspropertiesofcomponentsinpsychologicalmechanisms.Itmightbe objected that mental properties are properties of agents, andare therefore at the same level of mechanism as the agent-leveldispositionsIclaimtheyground.Afirstrejoinder:beingpropertiesofagentsdoesn’tprecludementalpropertiesfromgroundingfurtherdispositionsattheagentiallevel.Iseenoreasonwhythereshouldn’tbedispositionalstructuresinwhichasinglecomponenthasasetofdispositionswhichconstitutivelyexplainwhythatverycomponenthassomefurtherdisposition.Still,thatisn’thowIseefunctionalistmental properties, nor, arguably, is itwhat functionalists ought tosay. Functional properties are physically grounded dispositions. Ifthephysicalrealizersoffunctionalpropertiesareneuralproperties,then it seems the brain states that bear them will also bear anydispositions theyground.Onmyaccount, functionalpropertyFi isthepropertyofbeinganxsuchthattherearepropertiesP1,…,Pninvirtueofwhich,collectively,acertainconjunctionofsubjunctivesistrue,andxhasPi .Treatingfunctionalrealizationasakindofsame-subject necessitation is commonplace, and onewhich entails thatifPi isabrainproperty,so isFi .Butonreflection,whyshouldn’tapropertyofmybrainalsobeapropertyofme?IfIgrowmyfingernailslong,isn’tittrueboththatmyfingershavelongnails,andthatIdo?Onemight wish to insist that mental properties are properties ofwholeagents,andnotofanyproperpartsthereof,but inthatcasefunctionalistsmusteithersaythesameabouttheirrealizers,orelsefindanalternativeto thestandardquantificationalwayofdefiningsecond-orderproperties.Whatfunctionalistsoughttosay,Isubmit,isthatagentshavementalpropertiesinthesamederivativewaythatagents have fingernails. Perhaps there’s some reason for thinking

whenit’scoldoutside,(b)I’mdisposednottodresswarmlywhenit’swarmoutside.LabeltheseDCandDWrespectively.

MyhavingDCconsistsinmyhavingintrinsicpropertiesinvirtueofwhichthere’sarangeofnomicallypossibleconditions inwhichIwoulddresswarmlyifitwerecoldoutside;similarlymutatis mutandis forDW.Theseconditions,asbefore,dependonthoseofthegroundingdispositions—including,forexample,thatI’mnotundertheinfluenceofdrugs such that if itwerecoldoutside, Iwouldn’tbelieve it.TheintrinsicpropertiesthatgroundDCandDWtherebygroundpotentialcausal relations. Suppose it’s cold, that some relevant backgroundconditions forDC andDWobtain, and I dresswarmly. The fact thatit’scoldisadifference-makingcauseofmysodressing.Thenearestworldsatwhichit’scoldwillbeworldsatwhichIhaveDCandrelevantbackgroundconditionsobtains,henceworldsatwhichIdresswarmly.Thenearestworldsatwhichit’snotcoldwillbeworldsatwhichIhaveDWandrelevantbackgroundconditionsobtain,henceworldsatwhichI don’t dresswarmly. The psychological dispositional structure thatgroundsDCandDWtherebygroundsarangeofpotentialdifference-making causal relations between the ambient temperature andmyattire. This is no philosopher’s invention: I really do have DC andDW, in virtue of something like the dispositional structure outlined,andmy attire really is counterfactually correlatedwith the ambienttemperature.56Psychologyproximallygroundscausalrelationswhich,since physics grounds psychology, are distally grounded in physics.Since,by anyone’s lights, anovel causal role is sufficient for robustontologicalcommitment,functionalpropertiesareasnon-redundantasthepropertiesoffundamentalphysics.Mechanismsatanylevelhavetheir defining dispositions proximally grounded in the dispositions56.Notethatthisstructurealsogroundspsychologicaldifference-makingcauses.

The ambient temperaturemakes a difference to theway I dress because Ihavereliablebelief-formingmechanismsthatenablemetodetectit.AswellasDW,forinstance,Ialsohave—invirtueofthedispositionalstructuregivenby(i)–(iv)—thedispositiontodresswarmlyifIbelievethatit’scoldoutside.Mybeliefthatit’scoldoutsidenowwilllaterbeadifference-makingcauseofmydressingwarmlybefore leavingwork,aswill thevariousotherdisposi-tionsinthestructure.



dispositionsof fundamental physical particulars are theonly causalpowers toground.57Clearly, it’snousearguing thatdispositionsdonovelcausalworkinproximallygroundinghigher-orderdispositionsiftherearen’tanyhigher-orderdispositionstoground.Itseemsthatinorder toshowthat there’scausalwork foranydispositions todo,I need to assume the reality of higher-order dispositions, but that’sexactlywhat’satissue.

It’simportanttogetclearaboutthedialecticbeforeproceeding.Onthe table is theclaimthatanyeventwecanexplain in termsof themanifestationofahigher-orderdispositioncanbeexplainedintermsofthestructuredmanifestationsoffundamentalphysicaldispositions.I assume the reality of agent-level dispositions, and argue thatfunctionalistmentalpropertiesarecausallynovelinvirtueofproximallygroundingsuchdispositions. Ifmentalpropertiesarecausallynovel,there’s no question of their being eliminated, so my strategy aimstosecurementalpropertiesvianoveltyofsynchronic causal work.Buttheobjectordoubts therealityofagent-leveldispositions,andsoofcourse won’t grant me that mental properties ground them. Non-fundamentaldispositionslackdiachronic causal-explanatory novelty,andsoshouldthemselvesbeeliminated.Withoutsuchdispositions,there’snothingforfunctionalpropertiestoground,andthereforenoupwardscausation.Fundamentaldispositionssecuretheirontologicalstatus,ifatall,notbydoingthecausalworkofgroundingfurtherdispositions,butbydintoftheirdiachroniccausal-explanatoryrole.

IreplythatHumeanshavealreadyshownhowtorebutthisobjection:whetherornotavaseisfragilemakesadifferencetowhetherornotitbreaks,butwhetherornot ithas thisbasicdispositionalstructuredoesn’t. Ifanykindofcausalexplanationiscontrastive,thenhigher-orderdispositionsaren’texplanatorilyredundant,becausetheyhaveacontrastiveexplanatoryrolethatfundamentaldispositionsdon’t.Ifwe

57. SomethinglikethisunderpinsMerricks’(2003)eliminativismaboutordinaryobjects.Anyirreduciblecausalpowerswemightattributetosuchobjectsarerenderedotiosebythecausalpowersoftheirultimateconstituents.Lackingnovelcausalpowers,theobjectsthemselvesshouldbeeliminated.

thatmentalpropertiescan’tbothbepropertiesofagentsand theirbrains,butIdon’tseewhatitcouldbe.

4.4. Causal exclusion bites back?Iclaimthatthecausalworkdispositionsdoconsistsintheirgroundingdispositions, and thereby causal relations, at higher levels ofmechanism.ButthusfarI’vesaidnothingtoexplainwhyweshouldbelieve there are such dispositions to ground. Given physicalism,it follows that every mechanism has a fundamental dispositionalstructure, consisting in itsbeingcomposedof fundamentalphysicalcomponentshavingcertaindispositions,andacertainspatiotemporalstructure.Thebasiccomponentsofeachmechanismwillhave theirdispositions in virtue of their fundamental physical properties, andthesedispositions,togetherwiththewaytheirbearersarestructured,aresufficienttofullyexplaineverythingthemechanismdoes.Supposea proton and an electron combine, in virtue of the dispositionsgroundedintheirrespectivecharges,toformahydrogenatom.Suchanatomisamechanism,inmysense,whosedispositionalstructure—itsbeing composedof a suitablydisposedand structuredelectronandproton pair—grounds dispositions such as its being combustible.Suppose a sample of hydrogen combusts on some occasion, undercircumstancesC.Thedispositionsoftheelectronandproton,togetherwith their spatiotemporal relations, are sufficient to explain thesample’scombustioninC.Thesamewillbetrueallthewayup.Givenanymechanism, however complex, we will in principle be able toexplainwhatitdoesonsomeoccasionintermsofthemanifestationofthedispositionsofitsfundamentalphysicalcomponents,andtheirspatiotemporalrelations.

The explanatory adequacy of fundamental dispositions licensesthe following objection: (A) we shouldn’t posit any dispositionswedon’t need inorder to explainwhy thingsbehave theway theydo, and (B) we’ll never need to posit higher-order dispositions forexplanatorypurposes. If this objection is correct, then fundamentalphysicalpropertiesaretheonlyonesthatdocausalwork,becausethe



objectiongetstheburdenofproofallwrong:thefunctionalpropertiesthe exclusionargument targetsare higher-orderdispositions.ThosewhorunKim’sexclusionargumentagainstfunctionalpropertiescan’tassumethatfunctionalpropertiesdon’texist,becausethat’swhattheargument is supposed to show. And since functional properties aredefined in terms ofmacro-events, the exclusion argument can’t bepremisedontheirnon-existenceeither.

Wecanthinkoftheexclusionargumentasareductio.First,assumethattherearefunctionalproperties.Thenshowthatsuchpropertiescan’t do causal work, and conclude that they don’t exist after all,because if they did, there would be causally redundant properties,which(AD)rulesout.Theargumentthereforedependsonthetruthofthefollowingsubjunctive: if functional properties existed, there wouldn’t be any causal work for them to do. And that’s exactly what upwardscausationrefutes: if thereare functionalproperties, there’splentyofcausalworkleftforthemtodo,viz.,proximallygroundingfunctionalproperties at the next level up. If I were arguing that functional properties exist,thiswouldbebootstrapping,butI’mnot,soitisn’t.Thediachronic explanatory novelty of functional properties is sufficienttojustifyourbeliefthattheyexist,sotheburdenofproofrestswiththosewhowouldarguethattheydon’t.Withouttacitlyassumingthattherearenofunctionalproperties,proponentsofthecausalexclusionargumentcan’tshowthatfunctionalpropertieshavenocausalworktodo.But if theyareprepared toassume that functionalpropertiesaren’t real, what’s the exclusion argument for? Perhaps the readerhasanaggingsuspicionthatmentalpropertiesstillaren’tdoinganyessentialcausalwork.Couldn’ttheagent-leveldispositionsofagentsbegroundedsolelybypropertiesofbasicphysics?Whatifoneagreeswith me that agents have agent-level dispositions, but denies therealityofmentalproperties?Well,mentalproperties are dispositions(of components in psychological mechanisms, which contribute togrounding thedispositionsofagents).There’snoobvious reason toallow that there are agent-level functional properties, such as thedispositiontodresswarmlywhenit’scoldoutside,butdenythatthere

wanttoexplainwhythevasebrokeratherthannot,it’snousecitingitsbasicdispositionalstructureifwealsothinkthatattheclosestpossibleworldswhereitlacksthatspecificstructure,itstillbreaks.Thevase’sbasic dispositional structurewill be required to explain the precisemanner of its breaking, but unless contrastive causal explanationis itself dispensable, higher-order dispositions like fragility aren’tdiachronically redundant. This is of course the “dual explanandumstrategy”, which can also be employed as a direct response to thecausal exclusion problem.58 I don’t employ the dual explanandumstrategy in this way, because Humean causation isn’t sufficient forcausalwork, and theexclusionproblem isprecisely theproblemoffindingsuchworkforfunctionalpropertiestodo.Myaimhereisto(i) allow for the sake of argument that if higher-order dispositionswere diachronically redundant, then we’d have grounds for theirelimination,and(ii)employthedualexplanandumstrategytoshowthat they aren’t redundant in this sense.Having a novel difference-makingroleisn’tsufficientfordoingcausalwork(novelorotherwise),butitissufficientforthekindofcausal-explanatoryrelevanceweneedtoblockexplanatoryredundancyarguments.Similarargumentscanberuntoshowthatmacro-propertiesingeneral—whetherdispositionalor not—aren’t explanatorily redundant: striking a vase makes adifferencetoitsshatteringthataparticularmicrophysical realizationofstrikingdoesn’tmake.

Exclusion, however, still isn’t done biting. All I’ve done so faris block a redundancy argument to the effect that higher-orderdispositionsaren’t real. If suchpropertiesaren’t real, I claim, it isn’tbecausetheylackdiachronicexplanatorynovelty.It’sanothermatter,however,toshowthattheyarereal,anditremainsthecasethatmytheorypresupposesthem.Worsethanthat,Ineedmacro-eventssuchasitsbeingcold,dressingwarmly,etc.astherelataofthedifference-makingcausalrelationsthatlicensetheclaimthatgroundinghigher-order dispositions counts as causal work in the first place. This

58.Marras(1998).



culminating in the psychological mechanisms whose functionalpropertiesproximallygroundthosedispositions.Fodorsays: 59

So,then,why is there anything except physics?That,Ithink,iswhatisreallybuggingKim.Well,IadmitthatIdon’tknowwhy.Idon’tevenknowhowtothink aboutwhy.Iexpecttofigureoutwhythereisanythingexceptphysicsthedaybefore Ifigureoutwhythere isanythingatall,another(and,presumably,related)metaphysicalconundrumthatIfindperplexing.

IagreewithFodorthatthisiswhat’sbuggingKim—andMerricks,andHeil60—andwhile I shareFodor’spessimismabout theprospects forananswer,Iseenocompellingargumentfortheconclusionthatthereisn’tanythingbutphysics.ExtantHumeanresponsestotheexclusionproblemare,toregistermyagreementwithKim,afreelunch:61makingadifferenceisn’tsufficientfordoingthekindofcausalworkthatcausalclosureappearstorendertheprovinceoffundamentalphysicsalone,and the thought that functionalpropertiesdo the samecausalworkas their realizers is incoherent. However, the distinction betweenproximalanddistalgrounding,togetherwiththerealityofhigher-ordermechanisms and their characteristic dispositions, enables functionalpropertiestopayfortheirlunchthesamewayphysicalpropertiesdo.The causal structureof theworld, inmyview, is irreducibly layered.There’sno causal exclusionproblembecause there’s farmore causalworktodo,inconstructingsuchaworld,thaniscommonlysupposed.62

59. Fodor(1997)p.161.

60.Merricks(2003);Heil(2003).

61. Kim(1998),ch.3.

62.ThispapergrewoutofaseminaratKing’sCollegeLondon,inwhichItriedtopersuadeJimHopkinsoftheintractabilityoftheexclusionproblem.Mythanks to Jim for lively and thought-provokingopposition.Thanks also toMahradAlmotahari,PhillipGoff,ChrisHughes,NickJones,ShalomLappin,MarkTextor,RaphaelWoolf,andtwoanonymousreferees.BasedonresearchfundedbyaBritishAcademyPostdoctoralFellowship.

are functional mental properties. Whatever the current objector’sreasonfordoubtingtheexistenceofmentalproperties,ithadbetternotbe theirdispositionalnature. If I’m right thatmentalpropertiesform part of a hierarchy of dispositional structures through whichbasic physical properties ground the powers of agents, then theirgroundingrolesareeverybitasimportantasthoseoftheirultimatephysicalgrounds.

AsIseeit,thereareasmanylevelsofdispositionalpropertiesastherearelevelsofmechanism.ProtonsandelectronsaredisposedtoformHydrogen,invirtueoftheirbasicphysicalproperties.Hydrogenhas the further disposition to combust under certain circumstances,releasing water and energy. Now suppose we make a combustionengine inwhichHydrogen is a component—the fuel. The engine’shavingthepowertomakethevehiclemovewilldependinter aliaonthedispositionalpropertiesofHydrogen,butthat’sjusttosaythatthebasicphysicalpropertiesoftheelectronsandprotonsthatcomposetheHydrogeninthemechanismdistallygroundthepoweroftheenginetomakethevehiclemove.Causalclosureentailsthatalldispositionsaregroundedinpropertiesofbasicphysics,butnotthattherearenointermediatedispositions.Mentalproperties,Iclaim,areamongthoseintermediates,andsoareas important to theworkingsof theagentasthecombustibilityofhydrogenistoanenginethatburnsitasfuel.

4.5. ConclusionWeliveinaworldwhereallcausalpowers,henceallcausalrelations,aregroundedinfundamentalphysics,andthereforedependentuponit. This is apt tomake it seemas though there isn’t really anythingbutphysics. If thephysical isdoingall thecausalwork,whybotherwithanythingelse?Thecentralcontentionofthispaperisthatiftherearehigher-ordercausalpowerstoground,thenalthoughsuchpowersare distally grounded in the physical, distal grounding isn’t all thecausalworkthereis.Agent-leveldispositions,forinstance,aredistallygroundedinthephysicalthroughahierarchyofnestedmechanisms,



Physicalism”,inA.Beckermann,H.Flohr,andJ.Kim(eds.)(1992).Emergence or Reduction? Essays on the Prospects of Nonreductive Physicalism.Berlin:DeGruyter.

———(1998).Mind in a Physical World.MITPress.——— (2003). “Blocking Causal Drainage and Other Maintenance

Chores with Mental Causation”, Philosophy and Phenomenological Research67,pp.151–176.

——— (2005). Physicalism, Or Something Near Enough. PrincetonUniversityPress.

LePore,E.,andB.Loewer(1987).“MindMatters”,Journal of Philosophy 84,pp.630–642.

Lewis, D. (1986). “Events”, in Philosophical Papers Vol. II. OxfordUniversityPress,pp.241–269.

——— (1997). “Finkish Dispositions”, Philosophical Quarterly 47, pp.143–158.

List, C., and P. Menzies. (2009). “Non-Reductive Physicalism andtheLimitsoftheExclusionPrinciple”,Journal of Philosophy106,pp.475–502.

Marras, A. (1998). “Kim’s Principle of Explanatory Exclusion”,Australasian Journal of Philosophy76,pp.439–451.

Martin, C. (1994). “Dispositions and Conditionals”, Philosophical Quarterly44,pp.1–8.

———(2007).The Mind in Nature.OxfordUniversityPress.McKitrick, J. (2005). “AreDispositionsCausally Relevant?”,Synthese

144,pp.357–371.McLaughlin, B. (2006). “Is Role-Functionalism Committed to

Epiphenomenalism?”,Journal of Consciousness Studies13,pp.39–66.Menzies, P. (2008). “The Exclusion Problem, the Determination

Relation, and Contrastive Causation”, in Hohwy and Kallestrup(2008),pp.196–217.

Merricks,T.(2003).Objects and Persons.OxfordUniversityPress.Molnar,G.(2003).Powers: A Study in Metaphysics.OxfordUniversity

Press.Mumford,S.(1998).Dispositions.OxfordUniversityPress.

Bibliography

Bennett, K. (2003). “Why the Exclusion Problem Seems IntractableandHow,JustMaybe,toTractIt”,Noûs37,pp.471–497.

Bird,A.(1998).“DispositionsandAntidotes”,Philosophical Quarterly48,pp.227–234.

———(2007).Nature’s Metaphysics.OxfordUniversityPress.Block, N. (2003). “Do Causal Powers Drain Away?”, Philosophy and

Phenomenological Research67,pp.133–150.Crane, T. (2008). “Causation and Determinable Properties: On the

Efficacy of Colour, Shape and Size”, in Hohwy and Kallestrup(2008),pp.176–195.

Craver,C.(2007).Explaining the Brain.OxfordUniversityPress.Fara,M.(2005).“DispositionsandHabituals”,Noûs39,pp.43–82.Fodor, J.(1997).“SpecialSciences:StillAutonomousAfterAllThese

Years”, Noûs 31, Supplement: Philosophical Perspectives 11, Mind, Causation and World,pp.149–163.

Hawthorne,J.,andD.Manley(2005).“Mumford’sDispositions”, Noûs 39,pp.179–105.

Heil, J. (2003). From an Ontological Point of View. Oxford UniversityPress.

———(2005).“Dispositions”,Synthese144,pp.343–356.Hohwy,J.,andJ.Kallestrup,eds.(2008).Being Reduced: New Essays on

Reduction, Explanation and Causation.OxfordUniversityPress.Jackson,F.(1996).“MentalCausation”,Mind105,pp.377–413.Johnston, M. (1992). “How to Speak of the Colours”, Philosophical

Studies68,pp.221–263.Kallestrup,J.(2006).“TheCausalExclusionArgument”,Philosophical

Studies131,pp.459–485.Kim,J.(1992a).“TheNonreductivist’sTroubleswithMentalCausation”,

inJ.HeilandA.Mele,A.(eds.),Mental Causation.OxfordUniversityPress.ReprintedinKim(1993).Supervenience and Mind.CambridgeUniversityPress,pp.336–357.

———(1992b).“’DownwardCausation’inEmergenceandNonreductive



Papineau,D.(2010).“Realism,RamseySentencesandthePessimisticMeta-induction”,Studies in History and Philosophy of Science 41,pp.375–85.

Prior, E., R. Pargetter, and F. Jackson (1982). “Three Theses AboutDispositions”,American Philosophical Quarterly19,pp.251–257.

Segal, G., and E. Sober (1992). “The Causal Efficacy of Content”,Philosophical Studies63,pp.1–30.

Shoemaker, S. (1980). “Causality andProperties”, inP. van Inwagen(ed.),Time and Cause.Dordrecht:Reidel.

——— (2001). “Realization and Mental Causation”, in C. Gillett andB. Loewer (eds.), Physicalism and its Discontents (Cambridge:CambridgeUniversityPress),pp.74–98

Sider,T.(2003).“What’sSoBadAboutOver-determination?”Philosophy and Phenomenological Research 67,pp.719–726.

Stephan, A. (1997). “Armchair Argument Against Emergentism”,Erkenntnis46,pp.305–314.

Wilson, J. (2002). “Causal Powers, Forces and Superdupervenience”,Grazer Philosophische Studien63,pp.53–78.

Witmer, G. (2003). “Functionalism and Causal Exclusion”, Pacific Philosophical Quarterly84,pp.198–214.

Yablo, S. (1992). “Mental Causation”, Philosophical Review 101, pp.245–280.

Yates, D. (2009). “Emergence, Downwards Causation and theCompletenessofPhysics”,Philosophical Quarterly59,pp.110–131.

Functionalism and the Metaphysics of Causal Exclusion

Documents