The Individuation of Causal Powers by Events (and Consequences of the Approach)
- First Online:
- Cite this article as:
- Towl, B.N. Int Ontology Metaphysics (2010) 11: 49. doi:10.1007/s12133-010-0058-y
- 62 Views
In this paper, I explore the notion of a “causal power,” particularly as it is relevant to a theory of properties whereby properties are individuated by the causal powers they bestow on the objects that instantiate them. I take as my target certain eliminativist positions that argue that certain kinds of properties (or relations) do not exist because they fail to bestow unique causal powers on objects. In reply, I argue that the notion of causal powers is inextricably bound up with our notion of what an event is, and not only is there disagreement as to which theory of events is appropriate, but on the three prevailing theories, it can be shown that the eliminativists arguments do not follow.
In this paper, I explore the notion of a “causal power,” particularly as it is relevant to a theory of properties whereby properties are individuated by the causal powers they bestow on the objects that instantiate them. Not only are such theories of properties popular these days (see Shoemaker 1984/2003; Gillett 2002, 2003; Heil 2003; Mellor and Oliver 1997; Mellor 2000; Unger 2006, and others), but such a theory has been used to motivate some substantial metaphysical arguments.1 For example, Jaegwon Kim uses such a theory to argue against the independent existence of second-order and multiply realized properties (Kim 1993, 1998), as does John Heil (1999, 2003). Carl Gillett and Bradley Rives have used the theory to argue against the existence of determinable properties (Gillett and Rives 2005), and in a recent piece, Chase Wrenn (forthcoming) has argued that relations such as the realization relation between a physical and a special-science property is not a “real” relation. Thus, thinking about properties along the lines of causal powers has lead these authors to a “sparse” ontology: properties that are determinable, or multiply realizable, or second order, fail to be “genuine” properties on this kind of theory. Let us call such arguments “eliminative” arguments, for ease of exposition. Eliminative arguments, it will turn out, depend crucially on just what causal powers turn out to be. The main result of this paper will be mostly negative: Not only is there no unequivocal notion of a causal power that can be culled from the literature, but a serious consideration of what causal powers are will show that many eliminative arguments do not work.
My strategy will be as follows. First, I will present a brief sketch of how causal powers are represented in the literature and why we should think that causal powers are individuated (at least in part) by the events they are powers to manifest. I will then discuss how the causal powers view of properties is used in eliminative arguments. I will then argue that our theory of causal powers depends deeply on what we take our theory of events to be and go on to draw out some consequences that certain theories of events have for a theory of causal powers (and the eliminative arguments that rely on them). Although I will not develop a fully worked out theory of causal powers, my arguments should be enough to show that the conclusions of the authors mentioned might indeed be hasty and in need of re-examination against the background of a more developed theory of causal powers.
1 Causal Powers: An Overview
Discussion of a theory of properties based on causal powers runs throughout much work by Shoemaker (1979, 1980, 1998, 2001), Armstrong (1978, 1983, 1997), and Heil (1999, 2003). Earlier, such a theory was sketched by Harre and Madden (1975), Swoyer (1982), and Mellor (Mellor and Oliver 1997; Mellor 2000). A general metaphysics based on powers is detailed in Molnar (1993). Obviously, space prevents me from discussing this vast literature here, but a few generally accepted remarks will be enough to present a sketch of causal powers and their relation to properties.
In the words of Shoemaker (2003), “For something to have a power... is for it to be such that its presence in circumstances of a particular sort will have certain effects” (1984/2003, p. 211). Consider a ball with several properties: spherical, blue, bouncy, etc. In virtue of the ball being spherical, it has a number of powers: the power to cause a round impression when pressed into clay, the power to roll down a smooth incline, the power to fit through a round hole..., etc. The powers granted by the ball’s other properties, the ball’s blueness say, will be different: It will appear blue to normal observers under normal conditions, it will stand-out against a yellow background, etc.
... what makes a property the property that it is, what determines its identity, is its potential for contributing to the causal powers of the things that have it. This means, among other things, that if under all possible circumstances properties X and Y make the same contribution to the causal powers of the things that have them, X and Y are the same property. (1984/2003, p.212)
Objective: Powers are features of the world, and not the result of our way of categorizing or conceptualizing it.
Actual: Powers are not dispositions cashed out in terms of conditionals or counterfactuals but exist fully in the actual world.
Intrinsic and independent: Powers are intrinsic to objects and independent from their manifestations. A crystal of salt has the power to dissolve in water, even if it never makes contact with water, or there is a defeater present, or there simply is no water.
Directed: Powers are powers to bring about some sort of manifestation; thus, they are directed toward certain kinds of occurrences.
Of course, any of these features can be argued for or against, but together, they do provide us with a good working theory of what causal powers are. The feature I wish to scrutinize most closely is the last feature: the directedness of causal powers. Although causal powers are intrinsic to the objects that possess them, they are individuated by the events (“occurrences” in Molnar’s terms) that they tend to bring about.
Directedness seems to be an essential feature of causal powers. If indeed directedness is a necessary feature of causal powers, one might suppose that causal powers are individuated, at least in part, by the events (Molnar’s “occurrences”) that they are the causal powers to bring about. (For ease of exposition, I will speak as if causal powers “possessed” events, but no actual possession is implied). Thus, our theory of what causal powers are (or, if one prefers, our theory about which causal powers exist) is deeply dependent on the concept an event. How we individuate (types of) events becomes relevant to how we individuate causal powers, and although I will not defend this more radical claim here, it is indeed plausible that every event type defines a unique causal power: the power to bring about events of that type.
Why might one adopt the view that causal powers are individuated by the events they possess? There seem to be at least three good reasons for doing so. First, the claim that causal powers are individuated by their events makes sense of the language we use to pick out causal powers. If one were to look at the examples of causal powers as they appear here and in the literature, we see that causal powers are usually described as “the power to x” or “the power to bring about y”, and it seems appropriate to say that, if there is a causal power that is “the power to bring about x” and a causal power that is “the power to bring about y”, and x and y are not type identical, then the causal powers being considered are of distinct types. Thus, a genuine difference in the events being described by x and y also amounts to a difference in the causal powers being picked out by these locutions.
Second, there are good epistemic reasons for adopting this view. Shoemaker, in describing the causal theory of properties, uses a similar consideration (Shoemaker 1984/2003, p. 214). He points out that properties are capable of engaging our knowledge and our language and believes that this fact is explained by the further fact that properties are individuated by their causal powers. He goes on to say that “We know and recognize properties by their effects, or, more precisely by the effects of the events which are the activations of the causal powers which things have in virtue of having the properties.” This claim is heavy with event-talk, and for good reason, rarely do we directly perceive the causal powers of objects. Rather, insofar as we do perceive causal powers, we do so via the manifestations of those powers. For example, I can see and hear a glass shattering, but I do not directly perceive the fragility of the glass until it does, in fact shatter. We would be hard pressed to come up with a case in which we perceived a causal power in the absence of any of its manifestations.
Our knowledge of causal powers, then, stems either from our perception of events that are the manifestations of causal powers, or from direct acquaintance with the effects of causal powers on our perceptual systems. If a bottle has the causal power to appear green to normal viewers under normal conditions, I have direct epistemic access to the appearance, not the causal power.
Shoemaker’s thesis thus rests on the view that, ultimately, it is events that make contact with us, and if Shoemaker’s epistemic argument works for individuating properties by their causal powers, it must also work for the claim that causal powers are individuated by their events.
Third, there is the related argument that if causal powers were not individuated by their events, then it should be possible to have two distinct types of causal power that gives rise to exactly the same types of events (i.e., that have the same causal profile). If that were the case, how would we be able to tell when one causal power is present and not the other?2 There would be no occurrence to perceive, nor any value to measure, that would provide definitive evidence one way or the other.3
Lastly, we should ask: if causal powers are not individuated by the events they possess, how are they individuated? What criteria can we use to make distinctions between various sorts of causal powers, and how do we build a taxonomy of them? I do not know how to answer these questions, and while ignorance of the answers is not a good reason for holding this view, such ignorance can be a good reason to hold the view conditionally until such issues can be worked out.
In summary, our language and our epistemology suggest that causal powers are individuated, at least in part, by the events they possess. Our taxonomy of causal powers will thus match our taxonomy of events, or at least fail to crosscut our event taxonomy.
One can accept these arguments and yet leave open the question of how finely we individuate causal powers. For example, is it the case that there is a distinct causal power for each event type? Or can a single causal power be the causal power for a number of different types of events? If one answers affirmative to the first question, one would be adopting a rather fine-grained taxonomy of causal powers. If, on the other hand, one answers affirmative to the second question, one would be adopting a coarse-grained taxonomy with fewer causal powers. Though I believe we should adopt a more fine-grained taxonomy, arguments that speak directly this debate are outside the scope of this paper. Fortunately, the arguments that follow do not hang critically on this issue.
2 Eliminativist Arguments Based on Causal Powers
Just how is a theory of properties based on causal powers used to motivate eliminative positions? Space prevents presenting the individual arguments in detail, but they all follow a typical pattern. The first step is to show that the target property to be eliminated either (1) bestows no causal powers, or (2) bestows only causal powers that are also bestowed by a co-occurring property (a realizer property in the case of multiply realized properties, for example). If (1) above is the case, then the target property is not a genuine property, by the theory. If (2) is the case, then there is a tension as to which property is really responsible for bestowing the causal powers in question, at the risk of “double counting” causal powers. This tension is usually resolved in favor of the co-occurring property, again showing that the target property contributes no new causal powers and is therefore not a genuine property. (To aid exposition, I will continue to speak of “target” properties as a blanket term for multiply realized, determinable, etc. properties; I will continue to use “co-occurring” properties as a blanket term for realizer, determinate, etc. properties).
Working out an example will help to illustrate the strategy. Let us take two purported properties: a multiply realized property M and a physical property P that realizes M. If we individuate properties by their causal powers, it must be the case that M “bestows” causal powers on the objects that instantiate M, but, if P is to realize M, it cannot be the case that M bestows causal powers not also bestowed by P. For, if M did bestow such new causal powers, P would not be sufficient for M and thus would be unfit to be a realizer for M, contrary to our assumption.
Although, if M does bestow causal powers, and P realizes M, both properties overlap in the causal powers they contribute—that is, some causal powers are “bestowed twice.” Thus, we have a “double counting” of causal powers..., and M does not bestow any causal powers not already bestowed by P. M here does not seem to be an addition of being above and beyond P: any contribution M makes to the world is already accounted for by P’s being present. Why, then, should scientists (or philosophers, for that matter) posit a separate property M?
Avoiding eliminative arguments, then, involves either side-stepping the dichotomy here or showing that there are reasons for resolving the tension in favor of the target property. A serious consideration of events will show how such moves may be possible.
3 Causal Powers and Theories of Events
I have argued above that causal powers are individuated, at least in part, by the events that they are powers to manifest. It is events that causal powers bring about, and the causal powers that we discuss are causal powers to do x or bring about y. We need not assume from the start that causal powers need to bring about a certain category of event. The relevant events might be states, or they might be processes or achievements, or they might be any of these, depending on the property under consideration. What I say here should apply to all of these categories regardless.
So what are events and how do we build a taxonomy of them? I will consider three popular theories of events here: events as descriptions at a time, events as property instantiations at a time (or states of affairs), and events as maximally determinate concrete particulars. What I hope to show is that, no matter which theory we choose, there are problems for the eliminative arguments that some philosophers use to clear-cut our ontology.
3.1 Events as Descriptions at a Time
One theory of events is that events are just (true) descriptions at a time: <φ, t > (see Van Benthem 1983). If it is true at time t that Brutus stabbed Caesar in the forum, then this is an event; and in fact, it is an event distinct from Caesar’s dying at time t and from Brutus’s stabbing Caesar violently at time t. On this view, events will be as plentiful as our descriptions, although we can make some allowances for descriptions that are synonymous as referring to the same event.4
Though this theory is currently not a very popular theory of events, seeing how it fails to entail the elimination of certain properties will be useful. The eliminative argument fails with this theory of events because descriptions can differ easily and in ways similar to properties.
Take, for example, the determinable/determinate relation. Although Gillett and Rives (2005) argue that there are no properties that are not maximally determinate properties (and thus that there are no determinable properties), there surely are determinable and determinate descriptions—and such descriptions are distinct descriptions. If that is so, then there are distinct events for each description’s being true and distinct event types for each description generally. It then follows that there are distinct causal powers that manifest in these different events, and these different causal powers can be bestowed by different properties: a determinate property, and a determinable property. The determinate/determinable distinction carries through.
Consider a more concrete example: If a ball’s rolling down an incline is a different event from the ball’s rolling down an incline slowly, then it is possible to have a different sort of causal power that results in each. The former seems to be the sort causal power we would expect from a determinable property, while the latter would be the sort of causal power we would expect from a more determinate property.5 Likewise, appearing red to a normal viewer in normal conditions would be a different event from appearing scarlet, and so the causal power to appear red will be different from the causal power to appear scarlet. The former will be due to a determinable property, being red, while the latter is due to a determinate property, being scarlet.
We can run similar arguments for multiply realized properties. A property such as desiring to hail a cab will contribute causal powers that seem appropriate to such mental states: namely, moving toward a street and hailing a cab. The physical realizer for this state will contribute its usual causal powers, which will have to do with causing certain goings-on in the brain, certain leg and arm movements, and certain articulations, but if my hailing a cab and my moving my arm in a certain way are distinct events because of their distinct descriptions, then the causal powers are distinct, and there is room for both the multiply realizable and the realizing property.
There may be problems with a theory that says that events are (true) descriptions at a time anyway. For example, events are usually considered to be concrete particulars, but descriptions are not. Descriptions have truth values, while events do not, and some authors would find it odd to think that Brutus’s stabbing Caesar is an event distinct from Brutus’s stabbing Caesar violently: Rather, it is more natural to regard them as two different descriptions of the same event. In the vocabulary of truthmakers, events are truthmakers for descriptions, while descriptions are truthbearers. The two are distinct sorts of entities.
Still, considering this view of events has served an important purpose: It has shown that distinctions between different sorts of properties (determinable/determinate, realized/realizer, etc.) might also be distinctions between sorts of causal powers as well. If that is so, it is unclear whether there is the sort of “double counting” going on that we see in eliminative arguments.
3.2 Events as Property Instantiations at a Time
So the above theory of events yields far too many events for most philosophers. A more popular alternative is to construe events as an object’s (or set of objects) instantiating a property (or a relation) at a time: < [x, P], t>. This theory is advanced by Kim (1993, essay 1), Golman (1970), and Taylor (1985). It has the advantage that events can co-occur at the same place and time, but not every true description will yield a unique event.
Although there is much to consider in this theory of events, it threatens to make our treatment of causal powers circular. Properties, recall, are supposed to be individuated by their causal powers, but causal powers are powers to bring about certain events in certain circumstances, and so, if events are property instantiations at a time, then causal powers turn out to be powers to cause certain other properties to be instantiated, and so the notion of properties has crept back into our theory of properties.
The circularity might not be vicious—after all, we are individuating properties by their power to bring about other property instantiations. We can speak of a “network” of properties tied to each by causal relations, mediated by the surrounding context. With such a network, we can make use of the Lewis–Ramsey method (cf. Lewis 1972) for defining mental properties but apply it more widely to properties in general.6 What I have in mind is something like the following.
The Ramsey sentence says that there exists a set of entities, x1, ... , xn, which exhibit just those relations which the properties named by the term p1, ... , pn exhibit (again, see Lewis 1972 for full details of the method). This method allows us to define properties in terms of causal powers which themselves bring about property instantiations, without evoking other properties in the definition of any one given property.
I think this is just the way to think about defining properties, and since we are individuating properties by causal powers, it makes sense to apply the Lewis–Ramsey method, since it was developed by Lewis to handle functionally defined mental properties, but its application to properties at large still threatens circularity. Consider: when we apply the method to mental properties, we eliminate all reference to mental properties while preserving the theory based upon them, but this theory ultimately must make reference to something in order to ground the reference of the terms in the theory (Lewis calls these “O terms”). For Ramsey sentences about mental properties, this “grounding” will be done by reference to the causes and effects of the mental states: the “inputs and outputs” to the system. Ultimately, some mental states must be the result of stimulation from the outside, and some states must give rise to publicly observable behavior. Thus, although we eliminate any mention of mental properties in our Ramsey theory, the theory is ultimately “pegged down” at the periphery.
If we apply Ramsey’s theory more broadly to cover all properties, what is there to “peg down” the theory? The Ramsey theory for a given set of properties will replace all those property terms with existentially quantified terms, but there are no other properties that will appear in the theory to ground it: any outside property, itself, will be part of a Ramsey theory. Putting differently, the Ramsey theory for mental properties will fix the reference of its terms by including descriptions of nonmental events and nonmental properties, but if we have a Ramsey theory for all properties, there are no “non-property” properties to fix the reference of our terms. Thus, a Ramsified theory of properties still leads to a regress.
For these reasons, I doubt whether a causal power theory of properties can be usefully combined with a states-of-affairs view of events, but for argument sake, let us assume that such a combination is workable. It is still unclear that eliminative arguments will work. For on this view, properties appear as parts of events. If we allow things like target properties into our ontology to begin with, then these properties can become constitutes of events just as any others, and then there can be causal powers that give arise to these events.
For example, suppose we are committed to there being a multiply realized property M. There would be no good reason to rule out the possibility of an event, m, which was an object o instantiating M at time t: <[o, M], t>. Naturally, there would be some causal power or other the manifestation of which is such an event. That causal power will be different from causal powers that give rise to other events that have, as parts, the realizers of M (<[o, P], t > and < [o, Q], t>, say). My suspicion is that there is a symmetry here: Determinable properties will have powers giving rise to other determinable properties; multiply realizable properties will have powers giving rise to other multiply realizable properties, etc. In fact, this seems to be the picture that Yablo has in mind when he talks about causation (see Yablo 1992) and that Fodor has in mind when he discusses the special sciences (see Fodor 1974, 1997).
The eliminativist can, of course, still deny that there is such a symmetry simply by allowing that the realizer property has at least two causal powers: One causal power will be the power to bring about p (where p is the event < [o, P], t>, and P is a given realizer property in its own right), and one causal power will be the power to bring about m. (Note that this commits the eliminativist to the theses that properties can have multiple causal powers and that different properties can have at least some causal powers that are of the same type). Since P is sufficient for M (on most ways of construing the realization relation), there should be no problem with this move.
This reply from the eliminativists shifts the burden of the argument, however. The problem for target properties is usually stated as one of “double counting”—that there is nothing for the suspect property to cause that is not already caused by the co-occurring property. The response the anti-eliminativist can give is that such properties do have unique causal powers: multiply realized properties will have causal powers to bring about other multiply realized properties, etc. The eliminativist now needs to provide a reason for assigning the causal power to the realizer property as well, but why would we accept causal powers to bring about multiply realized properties, only to turn around and assign them to realizer properties?
In short, we again have a formula for overcoming the sort of “double counting” used in eliminative arguments. Multiply realized properties and realizer properties, to use the going example, might not bestow exactly the same causal powers after all. Furthermore, eliminative arguments that use this theory of events to explain causal powers are question-begging: They cannot be used to argue against the existence of a certain kind of property without assuming that that sort of property does not exist to begin with. In order for the double counting part of the argument to work, it must be the case that the causal powers of the to-be-eliminated property are also causal powers of the co-occurring property, but for this to be the case, one must either (1) rule out causal powers that manifest in events with the eliminated properties as parts or (2) give a principled reason why such powers can only be the powers of the co-occurring properties. Without either of these steps being spelled out, the eliminativist risks a vicious circle.
My argument here can cut both ways, of course: It is equally question-begging for someone to assume that there are multiply realized properties (or determinable properties, etc.) and thus events of a certain kind. What puts the burden of proof on the shoulders of the eliminativist is the idea from above: that there could very well be causal powers that stand in these relations—determinable/determinate, realized/realizer, etc.—and that these causal powers could very well define properties that stand in the same relations. If this is a genuine possibility, the eliminativist must say something against this possibility without recourse to an argument that assumes causal powers will match.
3.3 Events as Maximally Determinate Particulars
Some philosophers will be uneasy with the preceding discussions of events. For some, it just seems ridiculous to think, for example, that Brutus’s stabbing Caesar and Brutus’s stabbing Caesar violently are actually two distinct events (Davidson, for example, holds this view). We may, after all, have many different ways of describing an event or determining event types, but events in and of themselves are maximally determinate (concrete) particulars, and so there is one event that we sometimes describe as Brutus’s stabbing Caesar and that we sometimes describe as Bruttus’s violently stabbing Caesar, or perhaps even Brutus’s stabbing Caesar violently with a force of 20 N using an arching motion but, different descriptions, same event.
If this is the way we should think of events, we end up with a “sparse” conception of events, and if real events in the world are sparse in just this way, then causal powers will be sparse as well. A sparse conception of causal powers may indeed lead to eliminative arguments in metaphysics, but problems arise when we try to reconcile this view with the ideas about causal powers set out above.
First, this account of events makes mysterious just how causal powers are supposed to work. After all, having a causal power is not to have a power to bring about a particular dated event. It is the power to bring about a certain event type—a type tokens of which will appear whenever the circumstances are right, and so, to make sense of causal powers, we need a way of talking about event types, but now it appears we are back to our earlier theories: events fall under event types either in virtue of certain descriptions being true at a time, or in virtue of a certain property or relation being instantiated at a time, but then the arguments of the preceding sections can be used to show that there are many different sorts of causal powers. Eliminative arguments again run into trouble.8
Second, it is doubtful that many of the normal sorts of things we consider properties—being a certain shade of color, having a certain shape, etc.—will be the sorts of things that bring about such maximally determinate events, even when we consider clusters of properties. While it would be natural to describe the property of being scarlet as bestowing a causal power to appear scarlet to normal viewers in normal conditions and while it would be natural to describe the property of being spherical as bestowing the causal power to roll down an incline, we are hard pressed to devise properties that bestow causal powers to give rise to singular, maximally determinate events.
Perhaps, we do get such a thing in physics: a molecule of a certain configuration and shape, and with a certain charge x, attracts a particle of a certain shape with a charge y, with a resulting force of z, but can we identify all of the properties of Brutus and Caesar that, in the circumstances, lead to Caesar’s death as maximally described? I doubt we can, since it is likely that properties beyond these two will also be relevant to the event, as maximally described. That the murder took place in the forum at a certain time and not just outside the forum will depend on the route that Caesar took and who he talked to; if he was delayed, his murder might have occurred elsewhere, and in a different manner entirely. A ball’s taking a certain path down an incline depends not only on the ball and the incline, but on the air currents present, other objects on the incline, etc. When we consider events as maximally precise, dated particulars, we quickly see that there are many properties that contribute to their occurrence—and many of these properties are outside of the objects being considered. Thus, it seems that these kinds of events are not the right sorts of events to use in individuating the causal powers of circumscribed objects.
Perhaps, all of these factors are simply subsumed under the “circumstances” part of the theory: Caesar’s possible routes and the possible air currents in the room form different sets of circumstances in which causal powers manifest themselves. Fair enough, but now the theory loses touch with what was so enticing about it in the first place: the simplicity and predictive power that we wanted for a theory of properties to begin with. If events are this finely individuated, and causal powers are likewise this finely individuated, it is amazing that we manage to say or discover anything general about anything at all.
So, on three popular theories of events, eliminative arguments against suspect properties—determinable properties, multiply realizable properties, etc.—are premature. At the very least, the arguments are question begging: They need to assume that the properties being eliminated are not themselves part of the events (or event types) that define causal powers. Likewise, we need to take seriously the idea that causal powers can fall under different categories in just the same way that properties can—some might be determinable, some might be multiply realizable, and some might be relational.
More importantly, it is unclear that there is one settled-upon theory of events to hand, and this is exactly what we need to talk about causal powers productively. Clarifying these issues will go a great way to validating or invalidating arguments based upon them.
Though note that Heil thinks that properties are particular (i.e., modes) and, thus, that properties are individuated by their particular instantiation as well.
Schaffer (2005) argues that we can have such knowledge, if knowledge is possible at all, since skepticism about which possible world we are in with regards to properties parallels skeptical arguments about the external world (for example), but I take it that Schaffer’s offered solutions will not help us further with the individuation of causal powers; but this is exactly the burden that one must meet if denying that causal powers are individuated by events.
From hereon out, I will eliminate reference to time for ease of exposition; it should be assumed unless stated otherwise.
Of course, being a determinable or determinate property is a matter of degree; but we can safely ignore this fact for the sake of simplicity.
Keep in mind that I am dealing here with properties, and not predicates.