New York University
Marcia stopped at a red light because she wanted to avoid getting a ticket and thought that stopping would help her avoid getting one. Greg also stopped at a red light because he wanted to avoid a ticket and thought that that would help him avoid one. According to our ordinary way of thinking and talking about human action, there's some sense in which Greg and Marcia act in the same way and do so for the same reasons. But saying that they act in the same way shouldn't lead us to say that they only act in one way. Marcia is riding her cousin's Harley and stops at the red light by squeezing a lever with her right hand. She also, without realizing it, casts a shadow in my direction. Greg is driving the family station wagon, and he stops by pressing a pedal with his foot. Assuming that Marcia's squeezing the lever is her stopping at the red light, it is somewhat peculiar to ask whether that event, which can be described in so many different ways, is the same kind as Greg's pressing the pedal. The two events are similar in important respects and different in important respects. The relevant questions are in what important respects are the two events similar and different, and, more importantly, what is it for a respect to be important.
Despite the difference in vehicles, the example of Greg and Marcia is a standard example of people acting in the same way. Even if they were both driving the same model car, it's quite likely that they would move their legs in slightly different ways. Most of the time, for any way in which people behave similarly, there's a way in which they behave differently. In many cases, bodily movement types as well as more basic action types are different. And these differences are reflected in differences in the mental states of the actors. Marcia has different beliefs from Greg about how to stop at the red light. Or, even if the sets of dispositional beliefs are the same, the sets of beliefs causally implicated in the action will differ. Even if Marcia knows how Greg should stop his car, this belief will play no role in the production or guidance of her behavior.
Not all differences in behavior are reflected in differences in mental states. Marcia casts a shadow in my direction, while Greg does not. But like Greg, she has no thoughts at all about me, my direction, or her shadow. Casting her shadow is not something she does intentionally. Doing something intentionally requires a mental representation involving the relevant action type. And giving a certain kind of explanation of an action under a particular description requires that the action is intentional under that description. If we are interested in discovering systematic or lawlike connections between mental states, typed in some way or another, and behavioral events, typed in some way, then if we individuate the mental states intentionally, we must individuate the behavioral events intentionally as well.
If we type behavior purely in terms of the end results, Marcia's stopping at a red light is completely on a par with her casting a shadow in my direction. But, it seems, there is some kind of systematic correlation between Marcia's mental states and her stopping but no such correlation between her mental states and her casting a shadow. We can explain the similarity between Greg and Marcia's behavior, their both stopping at red lights, in terms of similarities in their mental states. But we can't explain the difference, her casting a shadow but his failure to do so, in terms of some difference in their mental states. And even if he did also unintentionally cast a shadow in my direction, this similarity in their behavior would not be explained in terms of beliefs and desires.
So if we are interested in systematic correlations between mental states and behavioral events, the important respects in which behaviors may be similar and different are the intentional respects. This idea is just beneath the surface of our ordinary explanatory practice. We do not typically say "Marcia intentionally stopped at the red light because..." But this does not show that the relevant behavioral type is an end result type, stopping at a red light, rather than the corresponding intentional type. We do not typically say that the behavior was intentional under the relevant description because this is obvious in context.
If someone says that Marcia stopped because she didn't want a ticket, there is a clear and innocuous sense in which the explanation is incomplete. Only a philosopher would be tempted to say that we need to add the fact that she believed that stopping would help her avoid one. It is unnecessary to state the belief, since, given the action and the desire, it's obvious in context what the belief is.
Similarly, while no one uses the word "intentionally" in the description of Marcia's behavior, it seems that ordinary intentional explanations entail, or at least strongly suggest that the behavior was intentional under the relevant description. Consider a case of a deviant causal chain.1 A climber wants to rid himself of the danger of holding a rope attached to another climber and believes that letting go will rid him of the danger. The belief and desire so unnerve him that he unintentionally lets go. In this situation, there is something clearly wrong with the following explanation: he let go of the rope because he wanted to rid himself of the danger. "Because" is stronger than the inverse of "caused." While the climber's letting go of the rope may not have been an action (it wasn't caused in the right way by beliefs and desires), the description of the behavior still applies. If asked why he let go of the rope, he will not say, "I didn't do it" but rather "because I got nervous."
There may be another sense in which ordinary explanations are incomplete. These explanations neither give, nor purport to give nomologically sufficient conditions for the occurrence of the event in question. Marcia's having the relevant desire to avoid traffic tickets and her having a belief about a means to this end are consistent with her failing to stop. She may, for example, not know how to stop a motorcycle. I am not suggesting that this lack of sufficiency is a defect in ordinary explanations which we should go on to remedy. Typically, what we want to know when we ask for an explanation of behavior is the further end the agent had in mind when performing the action. This is part of the reason the intentionality of the behavior is typically presupposed. If our ordinary explanations give us the information we are interested in, there's nothing wrong with that.
Given this picture of explanation, it is clear that there will often be causally relevant features that do not figure in ordinary explanations. We can explain the fire in terms of the short circuit without mentioning the presence of oxygen. But we may still believe that the presence of oxygen was causally relevant to the production of the effect. If there hadn't been any oxygen, there wouldn't have been a fire. The truth of such counterfactuals is usually taken as a test for causal relevance, where causal relevance is supposed to be a relation between a property of the cause and a property of the effect. Unfortunately, a precise analysis of this relation has so far eluded our grasp.2 There are even those who are sceptical about the existence of such a relation.
Without some notion of causal relevance, a relation between properties and not merely events, it is very difficult to understand what Mill's methods are supposed to be methods for. To apply the method of difference, you construct or consider a situation in which you have A, B, and C, and you check the effects. Then you consider a similar situation in which you have B and C without A and check the effects. What could "A," "B," and "C" possibly refer to in these instructions? If the instructions are possible to follow, they can't refer to concrete, particular, nonrepeatable events. The expressions refer to properties, and the test determines the causal relevance of repeatable features.
In addition to scientific practice, there are familiar examples which show the intuitive difference between causally relevant and causally irrelevant features. With your indulgence, I will add my personal favorite to the list. When you play pool in a bar rather than a pool hall, you have to pay for each game. With one exception, when you pocket a ball, the ball remains inside the pool table until you put in more money. The exception, of course, is the white ball. Certain regularities and corresponding counterfactuals hold for a wide range of pool tables of various sizes, constructions, and designs. If the ball that goes in is white, it will come back out; if it's not, it won't.
There are various possible explanations for this interesting phenomenon. Here's one: the cue ball's going in doesn't mean the game is over and no one would want to play if they had to pay every time they scratched. This gives good reason for people to design the tables in such a way that they support the regularity. But we might also want to know how the table knows that it's the white ball and not the black ball. Or, to put the question less metaphorically and more metaphysically, what are the causally relevant features of the cue ball that account for the difference in its behavior?
So far we've been thinking of the regularities and counterfactuals in terms of the color of the ball. But this, by itself does not convince us that color is causally relevant. If there is a color video camera inside the pool table that plays an important role in the release of the white ball, then color is causally relevant. But we don't think that this is how it works. How does it work? Three plausible suggestions come to mind: size, weight, and mass. Through ordinary empirical research, I have determined that size is not the causally relevant feature. Though many cue balls are in fact larger than the other balls, many are not, and the same table will treat cue balls of different sizes the same way.
Suppose that there is some spring loaded mechanism inside the table, and when the cue ball rolls over it, it is heavy enough or massive enough to trigger the mechanism, while the other balls are not. How do we find out whether the causally relevant feature is mass or weight? Are there any facts that would make it the case one way or the other? Well, if you can consider or construct a situation in which the mass is the same but the weight is different, and if, in that situation, the relevant effects are different, this would suggest that weight rather than mass is the causally relevant property. But of course we can consider such a situation: take the pool table to the moon. On the moon, none of the balls are heavy enough to trigger the mechanism. So you get a relevantly different effect. The cue ball is not treated differently from the others. So it is the weight of the cue ball, rather than its mass that is causally relevant. This looks like a case where a relational property rather than some intrinsic correlate of that property is "doing the causal work."
So we have a conception, though not an analysis, of this relation of causal relevance. And this conception does not rule out the possibility that relational properties may be causally relevant. But when we talk about the relevance of some property of the cause, we need to know which properties of the effect are relevant. A property may be causally relevant to certain features of the effect but not others. I have argued that if we are concerned with systematic or lawlike connections between mental states and behavioral events, the relevant properties of the effects are the intentional behavioral properties. Now I claim that the truth value, not just truth conditions or content of your beliefs is causally relevant. Truth is a causal power.
Various kinds of things are true, for example, sentences, propositions, and beliefs. The property of being true, as applied to belief, passes our counterfactual test for causal relevance to the intentional properties of behavior. Marcia wants to stop at the red light and believes that by squeezing this lever, she will stop. If her belief is true, she will stop. If the belief had been false, she wouldn't have stopped. This suggests that truth is causally relevant to ordinary end result types like stopping at a red light as well as their intentional counterparts.
When we considered the cue ball problem, we talked in terms of counterfactuals involving the property of being white even though we agreed that color was not the causally relevant property. Consider a specific situation in which possible worlds where I scratch are always closer than worlds where the crowd at the bar allows you to repaint any of the pool balls. Most actual situations will do. In such a situation, the following counterfactual is true.
If the ball that went in had been white, it would have come back out.
This is one problem for a straightforward counterfactual analysis of causal relevance. Though passing the counterfactual test may give us some reason to believe the property is causally relevant, it is not a sufficient condition. Without an analysis, how can we argue that truth is causally relevant?
I will argue for the causal relevance of truth in much the same way that I argued for the causal relevance of weight. But first we need a competitor. We need a property or set of properties that might be doing the causal work that I claim truth is doing. I will consider a proposal motivated by the deflationary conception of truth.
Accepting the deflationary conception of truth is more than accepting instances of the following schema:
p is true iff p.
Everyone accepts instances of this schema. According to the deflationist, this is all there is to our notion of truth. There is, according to the deflationist, no interesting and important feature shared by the claims that snow is white and that money doesn't grow on trees. Similarly, there is no common feature shared by the claims that it's true that snow is white and that it's true that money doesn't grow on trees. What makes it the case that it's true that snow is white is simply snow's being white. What makes it the case that it's true that money doesn't grow on trees is something completely different: money's failure to grow on trees.
If you are tempted to say that there is something interesting and important shared by the claims that snow is white and that money doesn't grow on trees but not shared by the claim that penguins can fly, namely the first two are true and the third is false, and if you are willing to stand up to questions about what "interesting and important" means, then you are not a deflationist about truth. I think that being causally relevant to a range of interesting and important properties is sufficient for being interesting and important. And I assume without further ado that the intentional properties of behavior are interesting and important. In short, if truth is a causal power in the way I have suggested, the deflationary conception of truth is false.
What facts can a deflationist point to to do the causal work that I claim truth does? Easy. Instead of pointing to the single fact that Marcia's belief is true, point to the fact that Marcia believes that p and the fact that p. Of course, this can be done without making use of the notion of a fact. So on a deflationist's conception, a slightly fuller explanation of Marcia's stopping would include facts like these:
(1) Marcia wanted to stop at a red light.
(2) Marcia believes that if she squeezes the lever she will stop.
(3) If Marcia squeezes the lever she will stop.
Of course there will be other facts involved in a complete explanation, but they will all be facts of the same three sorts: facts about Marcia's desires, facts about her beliefs, and facts about the world. There is no need to mention further facts that involve the notion of truth.
I say you have to talk about the truth of Marcia's belief about how to stop. They say you only have to talk about Marcia's belief that by squeezing the lever she will stop, and the fact that by squeezing the lever she will stop. There's supposed to be a difference here? Yes. Our conception of causal relevance is intimately connected to Mill's methods. You use the methods to find out about causal relevance. Mill's methods are about looking at a wide variety of situations that are similar and different in certain important respects and seeing in what other ways those situations are similar and different. Our notion of a causally relevant property is a property of a cause that is responsible for a property of an effect in a wide range of situations. To determine the causal relevance of a feature, you need to look at more than just one case.
So let's look at another case. Greg, you remember, stopped at a red light. He, like Marcia, stopped in order to avoid traffic tickets. There are some interesting similarities between the two cases. Both causes include desires to avoid tickets, and both effects are of the type, intentionally stopping at a red light. Are there any further similarities? It seems so. They both stop partly because they both know how to stop, i.e., they both have true beliefs about how to stop. This similarity in the causes is missed by any explanation that makes reference to any particular belief about how to stop, since they have different beliefs about how to stop. He believes you stop by pressing a pedal, and she believes you stop by squeezing a lever.
It might seem that the deflationist has a reply to this. The problem with making reference to facts (1), (2), and (3) above is that they are too specific. They cover Marcia's case but not Greg's. But the deflationist can still talk about the similarities between the two cases without using the word "true." Instead of referring to the three facts above, the deflationist can refer to the following two.
(4) Both want to stop at a red light.
(5) Both satisfy the following condition:
For some p about stopping at red lights, S believes that p and p.
Now, (5) does not say that Greg and Marcia believe the same things. We do not, in this case, make reference to any particular mental fact, someone's believing such and such, or any particular non-mental fact that such and such. Instead, we seem to be referring to that peculiar relation between the mind and the world in virtue of which some, but not all beliefs are true.
According to the deflationist, the notion of having a true belief just is the notion of satisfying the condition that for some p, you believe that p and p. This is the notion that appears in (5). But that means that this notion, the notion of having a true belief, is playing a causal role. If, as I have claimed, truth's being a causal power is inconsistent with deflationism, then this strategy is not available to the deflationist. The deflationist does not use the word "true" in picking out the causal powers, but does use the notion of truth.
I have been arguing primarily for two things. First, I'm arguing for the interest and importance of a certain picture of causal relevance, one which is primarily concerned with similarities between different cases, and which allows for the relevance of relational properties. Second, I'm arguing for the causal relevance of truth. The argument against deflationism is part of the argument for this second claim. The first deflationist strategy tries to show that something other than truth, namely a particular belief and the corresponding non-mental fact, is the real causal power. This strategy is available to the deflationist but it does not cover all of the relevantly similar cases, e.g., Greg and Marcia. The second strategy quantifies over beliefs and non-mental facts, so it does capture the relevant similarities. But this is simply a way of reintroducing the notion of truth without using the word, and so it amounts to the claim that truth is, after all, a causal power.
1. Donald Davidson, "Freedom To Act," in Essays on Actions and Events (Oxford: Clarendon, 1980) .
2. See, for example, LePore and Loewer, "Mind Matters," Journal of Philosophy 84 (1987), pp. 630-641; Fodor, "Making Mind Matter More," Philosophical Topics 17, (1989), pp. 59-79; and Block, "Can the Mind Change the World?" in Meaning and Method: Essays in Honor of Hilary Putnam edited by George Boolos (Cambridge: Cambridge University Press, 1990 )