Précis of:
Causing Actions

(Oxford University Press, 2000)

by Paul M. Pietroski e-mail


Reasons are causes, but that's where the philosophical work begins.

I assume that there are mental events, and that they typically have effects. Indeed, mental events often cause nonmental events, like bodily motions. If a person raised her arm because she noticed a friend, and wanted to attract the friend's attention, the arm-rising had a mental cause. If someone decides to turn on the stereo, and so pushes a certain button with her right hand, the motion of her hand was caused by a mental event. But whenever we humans move our limbs, our bodily motions are caused by certain muscle contractions, which in turn have various biochemical causes. So there is pressure to say that mental events are biochemical events, even if we cannot say much about the details. For how could a single effect, like the rising of an arm, have both its biochemical causes and distinct mental causes? I think we can answer this rhetorical question, which exerts a strong grip on contemporary philosophy of mind, without lapsing into an objectionable form of dualism.


Here's the idea: when a person moves their body for a reason, the bodily motion is caused by an action, which in turn has a mental cause; actions are themselves mental events; and our concept of causation is such that there is no insoluble overdetermination problem if a bodily motion B is caused by a (mental) action that is distinct from any biochemical cause of B. For example, when a person intentionally raises her arm, her action is an event of trying to raise her arm. This mental event causes the rising of her arm. And this can be so, even if the action/trying is distinct from any biochemical cause of the arm-rising, because we have a concept of causation such that: an event C is a cause of event E if the occurrence of C explains-in accordance with a ceteris paribus law--the occurrence of E; and an arm-rising is suitably explained by the occurrence of an action, whether or not the action is a biochemical event, if the arm in question rose because the person in question raised her arm. A closely related semantic point is that 'x raised her arm' is true iff x was the agent of a process that began with some action by x and ended with a rising of her arm.

So if a person raised her arm because she noticed a friend, the causal sequence is as follows: the onslaught of some belief (about the friend), then the action/trying, then the arm-rising. The occurence of the first event explains the occurrence of the second, which explains the occurrence of the third; and this can be so, even if the first and second events are distinct from any biochemical causes of the third. This makes room for a kind of event dualism that allows for the causal closure of the nonmental, and the supervenience of the mental on the nonmental. Or so I claim--and try to argue--in Causing Actions. Here, I'll stick mainly to exposition of the claims. But let me also say a little by way of motivating the project.

In my view, Descartes was basically right about the specialness of persons, their thoughts, and their (free) actions; he (and Kant) just had the wrong metaphysics. In "Mental Events," Davidson tried to reconcile autonomy and causation via his version of token physicalism. While I find this proposal attractive, I just don't believe it's true--and not for lack of trying. Perhaps every mental event really is a physical event, in some nontrivializing sense of' 'physical'. But I don't see any evidence for this strong claim; and the various "what else" or metaphysical arguments evaporate given a little imagination. Causing Actions is, primarily, an attempt to provide a non-Cartesian alternative to token physicalism. I assume that if token physicalism is true, every human mental event is a biochemical event; and talk of non-human mental events, if such there be, just complicates matters. So my focus is on saying how one can avoid both Cartesian dualism and the "naturalistic" thesis that each human mental event is a biochemical event.

Whether one wants to avoid this thesis will depend, in part, on how one reacts to another rhetorical question: how could any biochemical event have all the properties of a mental event?

I suspect that this question is unanswerable. But even if you don't share this suspicion, even a little, you might want to know if some non-Cartesian form of dualism is tenable. For if so, the fact that our mental events often have bodily effects doesn't show that our mental events are biochemical. If I'm right, one shouldn't reason as follows: mental events are person-internal events that often cause bodily motions; and we have discovered, by doing science, that the person-internal causes of bodily motions are biochemical; so mental events--or at least those with bodily effects-are biochemical. I grant the first premise but not the second. If we can make sense of the idea that mental events are causes distinct from any biochemical events, then we haven't (yet) discovered that the person-internal causes of bodily motions are biochemical requires. Sellars spoke of us having "manifest" and "scientific" images of ourselves and our place in the world. In the manifest image, people act for reasons; in the scientific image, members of a certain species exhibit various behaviors caused by intricately structured brains. My hunch is that the (rationalizing) causes visible in the manifest image are not locatable in the (impersonal) scientific image. This is a familiar thought: our best scientific accounts of the world and our place in it will not make reference to reasons or people who do things for reasons; at least not if 'scientific' has any real bite--suggesting the kind of theoretical structure and unity exhibited by our best current theories in physics, chemistry, and biology. And while my mental events are located inside my body, it doesn't follow that you can open my head and witness my mental events; though perhaps you can see me raise my arm, even if the action that causes the arm-rising occurs beneath my skin, much as you can see a plant whose roots lie beneath the surface.

In the book, I sketch some reasons for taking this possibility seriously. But it's hard to even debate the issue with specters of eliminativism and Cartesian dualism still haunting. Those who follow Sellars in taking science to be "the measure of all things" will endorse the following conditional: if the familiar hunch is correct, there are no mental events. (Debate will then ensue about whether modus ponens or modus tollens is the correct response.) But I think the hunch itself is more plausible than the conditional; though I also think the conditional is more plausible than eliminativism, which may well be maximally implausible. Still, it's not enough to just declare that there are mental events, whether or not any identity thesis is correct. We need to see how there could be causes that somehow lie outside the scope of natural science. And the answer isn't that people are Cartesian souls (or transcendental egos) whose reasons lie "outside" of ordinary spacetime: even if this is comprehensible, it makes a mess of the idea that reasons are often causes of events that occur "within" ordinary spacetime. The question is whether any plausible alternative answer is available.


In trying to provide one, I cover a lot of terrain: the semantics of action sentences and 'that'-clauses; the nature of actions and propositions; the relation between events and facts; our concepts of causation and explanation; ceteris paribus laws; the essentially corporeal nature of persons; and why the mental supervenes on the physical. But this way of listing the topics may already suggest how they might be connected. And like Sellars, I think philosophers are in the business of saying how things "hang together."

In the introduction, I say a little about why the distinction(s) between identity and constitution don't matter much with respect to the "how is it possible" questions that interest me. However you describe the situation, there is work to do: either we show how it's possible for certain biochemical events to have all the properties that our mental events have; or we show how it's possible for bodily motions to have mental causes and distinct biochemical causes.

The first step in pursuing the latter strategy is to get clear about what is involved in paradigm cases of mental causation: cases in which a person does something for a reason. For a philosopher in the analytic tradition, this is an invitation to think about the meaning of action sentences. I argue that the logical form of a sentence like (1) involves quantification over events as in (1L), which is an elaborated variant on Davidson's semantics for action sentences (using '#' for an existential quantifier):

(1) Nora raised her arm

(1L) #e{Agent(e, Nora) & #f[Ends-in(e, f) & Rise(f)] & Theme(e, her arm)};

where 'Ends-In' expresses a whole-to-part relation beween (i) an "accordion-style" event e, whose first part is an action by the Agent (i.e., the doer) that causes subsequent parts of e, and (ii) an event f involving some change in the Theme (i.e., the object saliently affected object at the end of the process). On this view, (1) is true iff Nora performed an action that started a process that ended with a rising of her arm. In defending this proposal, I draw on the work of many theorists: Hornsby, O'Shaughnessy, Parsons, Thomson, etc. It's a familiar point that Nora's arm rose if (1) is true, and (1) is true if Nora raised her arm at noon. In my view, (1L) delivers the best explanation such entailments, and the gloss of (1L)-in terms of a process with an action and an arm-rising as (causally related) parts-delivers the best overall account of how actions are related to action sentences. (In forthcoming work, Events and Semantic Architecture, I'll discuss the issues about logical form in much greater detail.) Some philosophers are inclined to dismiss the idea that actions are internal causes of bodily motions. But there is much to be said in favor of this old idea, and little to be said against it, once it is detached from the mistaken claim that every action is caused by some act of willing.

Chapter two offers a Fregean semantics of 'that'-clauses: 'that P' is a device for referring (in a given context) to the Fregean sense of 'P' (in that context); so the phrases 'that Fido barked' and 'that Rex barked' can have different referents, even if Fido is Rex. But this result is achieved without abandoning semantic innocence. Even in 'that Fido barked' and 'that Rex barked', 'Fido' and 'Rex' have their customary referents; so pace Frege, if the dog Fido is the dog Rex, both names (always) refer to the same dog. In my view, the complementizer 'that' plays a significant semantic role; it serves as a kind of indexical that picks out the sense of the sentence it introduces. I show how this can account for the standard range of facts (involving Paderewski, Mates-sentences, etc.), and how it lets us apply to 'that'-clauses the independently motivated distinction between compositionality and substitutivity.

The main interest (for these purposes) of the neo-Fregean semantics is that lets us distinguish the fact that Fido barked from the fact that Rex barked, while granting that the embedded sentences 'Fido barked' and 'Rex barked' have the same "truth-maker"--viz., the event of Fido/Rex barking. Let a singular event thought be the Fregean sense of a sentence that is true (according to a correct eventish semantics) iff an event of a certain sort occurs, where (as a matter of a fact) exactly one event of that sort has occurred. For example, the sense of 'Socrates died' is a singular event thought. In chapter three, I use this notion to capture a familiar and attractive idea: we have a concept of explanation that is a concept of an intentional relation between facts, and an intimately related concept of causation that is a concept of an extensional relation between events. Indeed, I think we have concepts of explanation and causation such that: the latter is a concept of the transitive closure of the extensionalization of the intransitive intentional relation that the former concept is a concept of.

Where singular event thoughts are concerned: if the fact that P explains the fact that Q, then the event e that verifies 'P' is a cause of the event f that verifies 'Q'; and if the fact that Q explains the fact that S, then the event g that verifies 'Q' is a cause of the event h that verifies 'S'; and if f = g--that is, if 'Q' and 'R' are verified by the same event-then e is a cause of h. For example, if the fact that Garfield howled (at time t) explains why Fido barked (at t*) , then the event of Garfield's howling caused the event of Fido's barking. And if the fact that Rex barked (at t*) explains why the neighbor complained, then the event of Rex's barking caused the complaint. And since the event of Fido's barking was the event of Rex's barking, the event of Garfield's howling caused (without explaining) the neighbor's complaint. Of course, this conceptual connection between causation and explanation goes both ways: if the event that verifies 'P' is not a cause of the event that verifies 'Q', then (the singular event thought) that P does not explain (the singular event thought) that Q. To get an interesting sufficient condition for causation, which is what we need if we want to explain how a bodily motion can have a mental cause, we need an interesting sufficient condition for explanation.

I suggest a version of the traditional covering-law idea. A singular event thought 1 explains another singular event thought 2 if: 1 and 2 are (respectively) the senses of the antecedent and consequent of a ceteris paribus law; the events that verify 1 and 2 instantiate the law; and the instantiation is "normal" in a sense that flows from the proposed account of ceteris paribus laws. To a first approximation, if it's a ceteris paribus law that the temperature of a gas rises when its pressure rises (holding volume constant), then: if the pressure rises and the temperature goes up, the former is a cause of the latter. But there are lots of ancillary issues to deal with, from the alleged vacuity of ceteris paribus laws to the asymmetry of explanation. The burden of chapter four is to show that: ceteris paribus laws are just fine; and if you stick to singular event thoughts covered by ceteris paribus laws, the covering-law model is just fine as a sufficient condition for explanation. The details initially appear to be quite removed from the philosophy of mind; but such is life. Indeed, part of the suggestion in Causing Actions is that much of the work needed for an adequate philosophy of mind lies in getting straight about our concepts of causation and explanation; so in thinking about thought, we are led to think about (inter alia) the semantics of causative verbs and the relation of shadows to flagpoles.

In chapter five, however, the focus is squarely on traditional questions about persons and their mental properties. The aim here is to sketch and motivate the promised alternative to token physicalism, deferring until chapters six and seven the needed replies to objections. After some initial discussion of Descartes, and why his (disastrous) picture of mental causation is seperable from his (much better) intuitions about thinkers, I suggest replacing the picture of mental causation indicated in (2) with the picture indicated in (3):

(2) S --> BE --> ... --> BE --> ME --> ... --> ME --> BE --> ... --> BE --> R

         / --> ME --> ... --> ME --> \
(3)  S                                              R
         \ --> BE -->  ... --> BE  --> /

where arrows indicate causal relations; 'S' and 'R' stand for an environmental stimulus and behavioral response; 'BE' and 'ME' indicate biochemical and mental events.

According to (2), mental events lie "causally between" certain biochemical effects of the stimulus and other biochemical effects of the response. Descartes adds the claim that mental causes are different in kind from other causes. But token physicalists can adopt the picture, shorn of this claim, by saying that the mental events are biochemical events as shown in (2*):

(2*) S --> BE --> ... --> BE --> BE/ME --> ... --> BE/ME --> BE --> ... --> BE --> R

where the dual designation 'BE/ME' encodes the idea that some biochemical events have the properties of mental events-and thus differ from mere biochemical events.

According to (3), rationalizing causes do not intrude into biochemical causal chains--or threaten the causal closure of the physical; but neither are mental events a species of biochemical events. Rather, the rationalizing causes of a bodily response-say, the motion of Nora's arm-are ontologically distinct from any biochemical casues of the response. And if Nora raised her arm intentionally, then the last mental event in the "upper" causal chain will be an action: Nora's trying (successfully, as it turned out) to raise her arm. Why are we entitled to say that the action causes the bodily motion? The answer is long and requires chapter seven. But let e1 be the event of Nora's trying to raise her arm, and let e2 be the event of her arm rising. Then the core idea is that <e1, e2> instantiates a (psychophysical) ceteris paribus law, and the proposed sufficient condition for causation applies. I argue that this picture is fully compatible with Strawson's view that our concept of a person is a concept of something that is (pace Descartes) essentially corporeal. Chapter five also offers some arguments, developing work by Hornsby, for distinguishing actions/tryings from any biochemical events. But the initial point of these is not so much to convince as to flesh out the kind of non-Cartesian event dualism on offer. Once the position is clearly available, one can revisit the arguments.

An obvious worry is that bodily motions would be objectionably overdetermined if (3) were the right picture of mental causation. But I don't think one can criticize this picture simply by pointing to counterfactual intuitions (about what would have happened if the biochemical causes of the arm motion hadn't occurred), or analogies about pairs of assassins sent to kill the same person, or any of the usual stuff that is supposed to make (3) look like a nonstarter. The context-sensitivity of counterfactual claims, and disanologies between events and people, provide ample resources for keeping quick objections to (3) at bay. The first part of chapter six works out some details. The more serious issue is whether event dualists can allow that the mental supervenes on the nonmental. If so, the overdetermination problems go away; if not, the proposal threatens to be (disastrously) Cartesian after all.

So the bulk of chapter six is devoted to saying how you can have your cake and eat it: yes, the mental supervenes on the nonmental; and yes, the mental supervenes on that which can be described from within the scientific image; but no, mental events can't be identified with biochemical events (or any events describable from within the scientific image). Persons and their mental properties are, in an important sense, primitive; nonetheless, they are also supervenient--and not ontologically basic. It's not obvious that this view is sustainable. But I think it is a view we have of ourselves and our place in the world. And an independently plausible (if somewhat deflationary) conception of supervenience can help us sustain this view.

Here's the idea: global supervenience--the thesis that possible world w1 differs from possible world w2 only if w1 differs physically from w2--is best seen as a thesis about the individuation of possible worlds; it's not a modal thesis (about how all properties depend on physical properties) that calls for explanation (by showing how all properties depend on physical properties). If w1 does not differ physically from w2, then w1 is w2; and if w1 = w2, then it's hardly surprising that w1 is like w2 in all nonphysical respects. By analogy, suppose someone wanted to know why the following generalization is (necessarily) true: set S1 differs from set S2 only if S1 differs in membership from S2. We can reply that this is how we count sets; if S1 has the same members as S2, S1 = S2. Possible worlds aren't sets; but global supervenience may well be a reflection of how we (ought to) count possible worlds. Following Kripke, I take possible worlds to possible histories of the universe, where the space of possible worlds is constrained by the natures of the objects in the universe (in which we find ourselves). On the assumption that these objects-including us-are in fact corporeal, possible worlds can be identified with physically possible arrangements of the stuff around here. In short, the only possible worlds are the physically possible worlds; and that's because possible worlds are possible arrangements of the (basic) objects, all of which are physical.

That was all very fast. In the book, I try to spell it out more slowly, by way of comparing and contrasting the conceptions of possibility associated with Kripke, Lewis, and Wittgenstein's Tractatus. But the bottom line is this: there are no possible worlds where Hesperus isn't Phosphorus; nor are there any possible worlds where I am a Cartesian soul. I am a (Strawsonian) person with a corporeal nature; hence, there is no possible world in which some disembodied soul is me. That said, it's logically possible that Hesperus isn't Phosphorus; and it's logically possible that I'm a Cartesian soul. In general, many logical possibilities are (meta)physically impossible. So we have to drop the following Ludwigian idea: if it's logically possible that P, there is a possible world where it is true that P. Lewis (198x) shows how to maintain this idea--along with the contingency of supervenience--by adopting his kind of modal realism. But we can also drop this idea, and be a little deflationary about supervenience, given a Kripkean conception of possible worlds and Fregean conception of 'that'-clauses. The details, unsurprisingly, resist easy summary.

Still, I hope the basic line of thought is tolerably clear: if you don't think minds are actually Cartesian thingies that somehow float free of the physical world, you needn't think that there could have been such thingies. Kripke offers us a more restrained--and independently more attractive--conception of modality; and given this conception, one can endorse global supervenience (as a thesis about how possible worlds are individuated) without identifying the mental events of people with any biochemical events in human bodies. People and their mental events can be primitive, in the sense of being irreducible to anything describable within the scientific image; yet they can also be supervenient, because that which is describable within the scientific image determines the space of all possible worlds-and not just the contingent properties of the world as we find it.

Another large issue remains. Details aside, I say that event C is a cause of an event E if a fact about C explains a fact about E. (Child has a similar view.) But if 'explains' is short for 'causally explains', the proposed sufficient condition is trivial-and presumably not a basis for a substantive thesis in the philosophy of mind. On the other hand, if we don't pack the idea of causation into the relevant idea of explanation, why think the proposed sufficient condition is true? My answer involves appeal to ceteris paribus laws. Such appeal, however, is bound to raise the concern that I'm trying to smuggle in a regularity conception of causation. While these themes are in the background in chapters three and four, chapter seven is where they are addressed explicitly. And just as I develop Strawson's view of (our concept of) persons in explicitly distancting myself from Cartesian dualism, I develop Strawson's view (of our concept of) causation in explicitly distancting myself Humean views about causation-leaving it open whether Hume was a Humean.

The idea is to show that: we have a concept of explanation that is a concept of an objective mind-independent relation R that holds between facts, even though facts are intentional (and in that sense mind-dependent) objects; at least where R holds between singular event thoughts, we can speak of a corresponding relation R* that holds between events; and since events are mind-independent spatiotemporal particulars, R* is a fully natural relation in every sense that causation is a natural relation. We can say that R* reflects an objective and intrinsic feature of each causal sequence. We can-and should-deny that one event causes another by virtue of some relation (say, pattern instantiation) that the cause and effect bear to other similar events. And we can still say (without lapsing into triviality) that C is a cause of an event E if a fact about C explains a fact about E.

My defense of this position relies on the proposed semantics of causative verbs; for I suggest that such verbs play an important role in our thinking about causation and explanation. Unsurprisingly, the proposed account of ceteris paribus laws also comes into play. (And while there are mentalistic ceteris paribus laws, it doesn't follow that these will be incorporated into anything that deserves to be called natural science.) Indeed, for better or worse, a great deal of the preceeding material is interwoven--or at least juxtaposed--in chapter seven.

Finally, an appendix addresses some issues concerning mental content. The idea is to offer an indirect argument for non-Cartesian event dualism: if you adopt this view, it becomes a little easier to answer certain questions about mental content, since you're no longer obliged to provide a "naturalistic" theory of content--i.e., a sufficient condition stated in a nonintentional idiom for having a mental content. The main chapters do not, however, rely on this argument.

top | back to symposium index | Introduction Chapter (pdf)