3.1 From Why-qustiosn to Reasons why
A theory of “because” answers to why-questions (that do not give an agent’s reason for acting) fills in the right-hand side of the schema
(1) Q because R if and only If... .
I said at the end of section. I would give a theory like this, but that was false advertising. I will give a theory, not of answers of the form “Q because R,” but of answers of the form “that R is a reason why Q,” or “one reason why Q is that R.”
It wasn’t really false advertising, though, since “reasons why” answers and corresponding “because” answers are often equivalent, in some sense. In response to the question “Why is it raining?” one might answer
(2) It is raining because a cold front has moved in
Alternatively, one might answer
(3) The reason why it is raining is that a cold front has moved in.
There may certainly be stylistic reasons for preferring (2) to (3). But the answers one gives by uttering (2) and by uttering (3) seem to be the same, the two sentences just being different means for giving it. I’m not saying that (2) and (3) are synonymous sentences; if sentence-meanings are fine-grained enough, the fact that (2) and (3) do not have the same surface syntactic structure is good evidence that they are not synonymous.
When there is only one relevant reason why Q, and it is that R, then (and only then) is it the case that Q because R.
The reasons for the above is: Better evidence might come from contemplating a scenario with more than one relevant reason why something happened.
One this reason fot the above formulation is:
- It is the best precisification of the vague idea that “explanations cite causes.”
- Another is that the notion of a complete answer, and of a partial answer, to a why-question, can be analyzed in terms of the notion of a reason why.
- Less important to me, but also worth noting, is that speaking of the reasons why such-and-such happened is a more flexible way of talking. It lets us make more fine-grained distinctions. “Reason,” as used in “reason why Q,” is a count noun, so we can speak of a reason why Q, or the reason why Q; we can discuss some or all or just three of the reasons why Q. This flexibility is absent if we are limited to saying things of the form “Q because R.
Declarative sentences express propositions, interrogative sentences “express” (have as their semantic value, for the purposes of compositional semantics) sets of propositions. For example, the #semantic_value of “Who came to the party?” at a world W is the set of propositions P satisfying (i) there is a person X such that P is the proposition that X came to the party, and (ii) P is true at W.
Jason Stanley suggests that the natural extension of his theory to why questions treats interrogative “why,” like “who,” as an existential quantifier. But instead of ranging over people, like “who” does, “why,” Stanley suggests, ranges over reasons. “Why” contributes the “there is a reason” to this specification of the semantic value of an interrogative “why Q”:
• The semantic value of “why Q” at W is the set of propositions X satisfying (i) there is a reason R such that X is the proposition that R is a reason why Q, and (ii) X is true.
3.2 A Simplish Theory
. It is a theory of reasons why Q when the fact that Q “corresponds to the occurrence of a concrete event.” The fact that a certain firecracker exploded corresponds to a concrete event, namely the explosion. The fact that 2 + 2 = 4 does not. I will not try to give a precise definition of “concrete event,” or much of a theory of which concrete events there are, or a rule for determining which event corresponds to a given fact. It is never going to matter which event corresponds to a given fact; and only some very general claims about which events exist, and are concrete, will be important.
I’m a liberal about what events there are. Events don’t have to be changes, like the explosion of a firecracker, or the swerving of an atom. If a room is completely empty for an hour, there is an #event that occurs in that room during that hour. I am not, however, an extreme liberal about events. I don’t think that every fact corresponds to an event; I also doubt that every fact “about some particular things” corresponds to an event.
Concrete events include, among other kinds, physical, biological, sociological, geo-political, and mental events. So my theory makes claims, not just about the reasons why a given rock accelerated when dropped but also about the reasons why the economy collapsed in , and even the reasons why a given non-physical mind (if there are any such things) thought “I think.” What my use of “concrete event” excludes are “events” concerning, for example, only mathematical objects, like— to use my earlier example—the event consisting in the numbers and being related such that 2 + 2 = 4. I imagine most philosophers would deny that there is any such event; even if they’re wrong, my theory says nothing about the reasons why 2 + 2 = 4.
A first draft of my theory consists of instances of the following schema, in which “Q” is replaced by a sentence expressing a fact that corresponds to the occurrence of a concrete event:
(T0) That R is a reason why Q if and only if the fact that R is a cause of the fact that Q.
Many philosophers have believed that “explanations of events cite, or describe, causes.” They do not all construe this vague claim in the same way. Theory (T0) is, I think, the best way to make it precise. I will say something about the ways in which it is superior to another well-known precisification.
For example, (T0) has a lot going for it. The car skidded off the road. One reason why is that the tire blew out. The fact that the tire blew out is also a cause. Again, the main reason why the dinosaurs went extinct is that a comet (or asteroid) hit the earth. The comet impact also caused the extinction.
Nevertheless, I have to acknowledge that (T0) is false, in fact quite obviously false. For example, one reason why this ball is either red or green is that it is red. But the fact that the ball is red is not a cause of the fact that it is either red or green. If this reason why is not a cause, what kind of thing is it? It is a ground. The fact that the ball is red grounds the fact that the ball is either red or green. In general, “grounding explanations” of events are counterexamples to (T0).
Is there really such an event as the ball’s being either red or green? I’m liberal enough to say yes, but this question can, as Judith Thomson likes to say, be bypassed. Even if there is no such event as this ball’s being either red or green, there are other examples of reasons why that are grounds that are uncontroversially reasons why some event occurred. The temperature of the air in this room just increased from X degrees Fahrenheit to X+1 degrees Fahrenheit. Why? Because the average kinetic energy of the molecules that constitute the air in this room just increased from X Joules to X+1 Joules. The increase in kinetic energy did not cause the increase in temperature; it is the ground for the increase in temperature.
So there exist reasons that are grounds as well as reasons that are causes. But that is where I draw the line. Here is a preliminary statement of my theory:
(T1p) That R is a reason why Q if and only if the fact that R is a cause of the fact that Q, or a (partial) ground of the fact that Q.
Carl Hempel entertained, but then rejected, the idea that “all explanations of events are causal.
cosndier this: why the period of some given pendulum is two seconds: this pendulum has a period of two seconds because it is one hundred centimeters long. But “surely” the length does not cause the period.
Now “this pendulum has a period of two seconds because it is one hundred centimeters long” is true only if we interpret the talk of the pendulum’s period as talk of a disposition. Phrased in terms of reasons why the claim then becomes: one reason why the pendulum is disposed to take two seconds to complete a full swing is that it is one hundred centimeters long. Hempel is right, the length does not cause the dispositional period. Once again we have a counterexample to (T), but once again it is compatible with (Tp). While the length may not cause the period, it does ground it.
David Lewis entertained, and then accepted, the idea that “all explanations of events are causal. Reasons that are grounds refute Lewis’s theory.
Walt’s immunity to smallpox cannot be caused? Go to the press and announce that philosophers have discovered that no one at the Centers for Disease Control has ever immunized anyone. We’ll be laughed out of town. certainly whenever someone immunizes someone against smallpox, he causes that person to be immune to smallpox. Walt’s immunity can be caused; so it is an event.
What we should do is abandon (T0) and move to (T1p). Walt’s immunity is grounded in his possession of antibodies. That’s why his having the antibodies is a reason why he is immune.
Theory (T1p) says that these are both correct answers. Still, it might seem bad to give both of them at once. “The temperature increased for several reasons, among them: the fact that someone turned up the thermostat, and the fact that the mean molecular kinetic energy went up.” I don’t find this particularly odd, but I imagine someone might. Even if it is odd, and my judgment that it is fine is off, I don’t see a threat to (T1p). Maybe the right thing to say is that in some contexts we use “why” to ask for reasons why that are causes, and in other contexts we use “why” to ask for reasons why that are grounds, and we never use “why” to ask for both kinds of reasons simultaneously.
It’s not crazy to think that we make this restriction, since causation and grounding are different things. All of this is compatible with the claim that, speaking unrestrictedly, causes and grounds, and only causes or grounds, are reasons why.
Theory (T1p) raises the question: what is grounding? I do not have a lot to say about this question. I have given examples: the temperature of a gas is grounded in the average kinetic energy of its constituent molecules, a disjunction is grounded in its true disjunct(s). If these sound right then you already sort of know what grounding is. Another way into the notion of grounding is to trace its connections to other notions. For example, fact G grounds fact H iff G is a metaphysically more basic fact in virtue of which H obtains. That is not much help since “H obtains in virtue of the fact that G obtains” is close to synonymous with “H is grounded in G.” But few of those who theorize about grounding give reductive definitions of grounding. They do not try to say in more basic terms what grounding is.
in theory (T1p) causation is treated as a relation between facts rather than events. I prefer “the fact that Suzy threw a rock at the window was a cause of the fact that the window broke” to “the throw was a cause of the breaking”—at least when it comes to choosing a way of talking to build a theory of causation around. But I’m not going to defend here the thesis that a #theory_of_causation should treat causation as fundamentally a relation between facts.
but “causation is not a relation” in the first place (for the author). That is, I like the idea that the proper way to regiment causal talk, for philosophical purposes, is to use, not a predicate “X is a cause of Y” that is true of two things only if both are facts (or both are events), but to use a sentential connective that takes two sentences and makes a sentence. If we use “C” for this connective, then “Suzy’s throw caused the breaking” and “The fact that Suzy threw a rock is a cause of the fact that the window broke” get replaced by “C(Suzy threw a rock; the window broke).”There seem to be expressions in English the function of which is close to that of “C”: for example, “Suzy threw a rock, and as a result, the window broke.” In fact, I think that “The fact that ... is a cause of the fact that ... ” is another such expression. (I am saying that these expressions have similar functions, not that they are synonymous; they have syntactic, and therefore presumably semantic, complexity missing from “C(... ; ...).”) One objection to treating causation as a relation between facts, rather than events, is that it makes causation too “unworldly”: surely it is things in space and time that cause other things in space and time, but it is events that happen in space and time; facts are “out there” with the other abstract objects, the numbers for example, and numbers are the wrong kind of thing to do any causing. The thesis that causal talk is best regimented using a sentential connective, while in many respects very similar to the thesis that causation relates facts, avoids this objection. If asked to identify the thing doing the causing, and the thing being affected, in the sentence “Suzy threw a rock, and as a result, the window broke,” the best answer is that it is Suzy and the rock (both worldly things in time and space) that do the causing, and the window that is affected.
Anyway, (Tp) can be reformulated to treat causation as a relation between events. It will say: that R is a reason why Q iff the fact that R is a ground of the fact that Q, or the event corresponding to the fact that R is a cause of the event corresponding to the fact that Q.
(Tp), reformulated in this way, is most plausible if we assume a plenitudinous theory of events, one that, roughly speaking, denies that distinct facts ever correspond to the same event. If one held, instead, that, for example, the fact that Jones walked home, and the fact that Jones walked home slowly, correspond to the same event, an event that is both a walking home by Jones and a slow walking home by Jones, then there is trouble; I’ll be stuck saying that, if one reason why Jones arrived late is that he walked home slowly, then also one reason why Jones arrived late is that he walked home, even if it is true that had he walked home without walking home slowly, he wouldn’t have been late. (As is well-known, Davidson had a “non-plenitudinous” theory of events; see for example his paper “Causal Relations.” I guess I should mention that Davidson also objected to speaking of “the” event that corresponds to a given fact. I’m not going to dwell on this, though, since in my official theory it never matters whether a fact corresponds to a single unique event.)
I will sometimes speak of causation as a relation between events, and also sometimes speak as if it is events that are reasons why
While I am discussing what I am taking causes to be, I should say something about what I am taking reasons to be. It is probably clear that I am taking reasons to be facts (even if, as I just said, I will sometimes write as if reasons are events). It is worth emphasizing, however, that officially, grammatically, the “X” in “X is a reason why Q” and in “one reason why Q is X” holds the place, not for a term denoting a fact, but for a thatclause that expresses a fact. The officially-true thing to say is: “that the car had a flat tire is a reason why it skidded off the road,” not “the fact that the car had a flat tire is a reason why it skidded off the road.” This is not super-important, and I will sometimes put terms for facts in for “X” when I think it aids comprehension.The difference is still worth noting, though, because it makes my theory immune to a certain kind of objection. If you could put a term for a fact in for “X,” then you could answer the question why the car skidded off the road by saying “Fred’s favorite fact is a reason why the car skidded off the road,” or (worse) “Xvvb is a reason why the car skidded off the road,” where “Xvvb” is a name one has introduced for the fact that the car had a flat tire. But these do not seem to be answers. (Saying that Fred’s favorite fact is a reason seems instead to be a way to inform someone of how to find out the answer—ask Fred what his favorite fact is.)
One choice point in the theory of causation is between facts and events; another is between treating the causal relation as a two-place relation and treating it as a four-place relation. “Contrastivists” about causation think it is a four-place relation: a relation between a cause, a contrasting fact/event (or set of contrasts), an effect, and another contrasting fact/event (or set of contrasts).- Although I will speak exclusively in binary terms, this is just for convenience; I do not take a stand in this book on whether contrastivism is correct. (If contrastivism about causation turns out to be correct, then I will advocate contrastivism about reasons why as well. On that combination of views, the number of relata of the causal relation will continue to line up with the number of relata of the reasons-why relation.
Theory (Tp), again, applies only when the sentence that goes in for “Q” describes the occurrence of a concrete event. But lots of why-questions are not about events: why is Galileo’s law of free-fall true? Why does every complex polynomial have a root? Presumably there is some entirely general theory of reasons why, from which it will follow that, for example, certain facts about gravity are among the reasons why Galileo’s law is true.
Oh, how I wish I had a completely general theory of reasons why to show you. But I do not. I have part of a general theory: I certainly think that, in general, if G grounds H, then G is a reason why H obtains. Why not go all the way? Why not take (Tp) and remove the restriction on what goes in for “Q,” to get the theory that the schema is true no matter what goes in for “Q” and “R”:
(T2) That R is a reason why Q if and only if the fact that R is a cause of the fact that Q, or a ground of the fact that Q.
When the sentence that goes in for “Q” is true but does not describe the occurrence of a concrete event, then the fact that Q is not the right sort of thing to have causes. (T2) says that, in this case, the reasons why Q are all and only the grounds of Q. I do not endorse (T2) because I doubt that, in this kind of case, the only way a fact can get to be a reason why Q is by grounding the fact that Q. One general reason for doubting this comes from looking at answers to why-questions about mathematical facts. There are mathematical facts F with the property that the question why F obtains has an answer. But I doubt that every answer to a question like this cites facts that ground that mathematical fact. (I will give another reason for rejecting (T2) in chapter 4, and float a completely general theory of reasons why, one that does better than (T2) but which I still hesitate to fully endorse, in Appendix B, in chapter 4.)
Theory (Tp) connects reasons why to causes and grounds using the biconditional “if and only if.” It is common to state theories this way, but read literally the theory is too weak. I don’t just hold that, as it turned out, reasons why Q are always either causes or grounds of the fact that Q. I hold that this is necessary. More precisely, every instance of the following is true, when “Q” is replaced by a sentence that purports to describe the occurence of an event, and “R” is replaced by any sentence:.
- Necessarily, if it is a fact that Q and it is a fact that R, then that R is a reason why Q if and only if the fact that R is a cause of, or a ground of, the fact that Q.
But there is another way in which (T1p) is too weak, that adding “necessarily” does not address. Suppose that it is a fact that Q, and a fact that R, and that the fact that R is a cause of the fact that Q. According to (T1p), that R is a reason why Q. This is a book about why-questions. It is thus natural in this context to ask, about this kind of example, why it is that one reason why Q is that R. Although my restriction to why-questions about events entitles me to remain silent about this question, it goes against the spirit of (T1p) to say nothing. It is, instead, in the spirit of (T1p) to answer that it is not just a coincidence that (i) the fact that R is a cause of the fact that Q, and also that (ii) the fact that R is a reason why Q. It is in the spirit of (T1p) to assert further that the fact that R is a reason why Q because the fact that R is a cause of the fact that Q. Put into reasons-why talk the assertion is that the reason why one reason why Q is that R is that the fact that R is a cause of the fact that Q.
This iterated reasons-why claim is in the spirit of my theory but does not follow from (Tp). I want it to follow. So a closer-to-official statement of my theory is this:
(T1) Necessarily, if it is a fact that R and it is a fact that Q, then: if the fact that R is a cause of the fact that Q, then one reason why Q is that R, and the reason why one reason why Q is that R is that the fact that R is a cause of the fact that Q. Similarly if the fact that R is a ground of the fact that Q. And every reason why Q is (i) either a cause of the fact that Q or a ground the fact that Q, and (ii) satisfies the relevant one of these conditionals.
Necessarily, if it is a fact that R and it is a fact that Q, then: if the fact that R is a cause of the fact that Q, then one reason why Q is that R, and the reason why one reason why Q is that R is that the fact that R is a cause of the fact that Q. Similarly if the fact that R is a ground of the fact that Q. And every reason why Q is (i) either a cause of the fact that Q or a ground the fact that Q, and (ii) satisfies the relevant one of these conditionals.
But I think it is a mistake to suggest, as Fine does, that “... because ... ” is a close English equivalent of this two-place sentential connective. For if “the fact that the ball is red grounds the fact that the ball is either red or green” means pretty much the same as “the ball is red or green because it is red,” then it cannot be right to say that:
4: ((the ball is red or green because it is red)) because ((the fact that the ball is red grounds the fact that it is either red or green))
For if “because” is just a sentential-connective way of expressing grounding claims, then sentence () has the form “P because P,” and so cannot be true. But () is true. Although every fact that is a ground is also a reason why, it is false that being a ground and being a reason why (or, better, that being a cause or a ground and being a reason why) are the same thing.
Partial and Complete Answers
“Explanations of events cite, or describe, causes”—many philosophers have accepted this sentence, without agreeing on what it means. Theory (T0) is, I have said, the best way to make it precise. So what is wrong with the other ways? The answer will emerge from looking closely at one alternative precisification.
Even though Lewis formulated his theory using, first, “explain,” and later “explanatory information,” I think that Lewis’s theory is best interpreted as a theory of answers to why-questions. As a theory of answers to why-questions, Lewis’s theory, like (T0) and (T1), is limited to why-questions about events. The most direct way to translate Lewis’s statement of his theory into a theory of answers to why questions about events gives us this: X is an answer to the question why E happened iff X is a proposition about E’s causal history. But putting the theory this way leaves it unclear. There are, as we are about to see, several different kinds of answers to why-questions. Just what kind of answer to a why-question is this theory a theory of? It is certainly not a theory of reasons-why answers. Lewis never uses the term “reason why.” Neither does it appear to be a theory of because-answers. Instead it is most naturally understood to be a theory of partial answers to why-questions
The distinction between complete and partial answers applies to any question. The complete answer to the question who came to the party says, of all the people who came to the party, that they came to the party (and perhaps add that no one else came to the party).
#partial_answer #Partial_Explanation #Complete_Answer #complete_Explanation
As for partial answers, in general a proposition is a partial answer to a question if it “rules out” (entails the falsity of) at least one possibly-true-but-actually false complete answer to that question. So “Two people whose names begin with ‘J’, ” if true, is a partial answer to the question who came to the party, since it entails that “Aaron and Moses and no one else came to the party” is not the complete answer.
I take Lewis’s view to be that the complete answer to the question why E happened lists all of E’s causes (and perhaps adds that they are all of the causes). The part of his theory that he is most explicit about, his theory of partial answers, is what you get by taking the general characterization of partial answers to questions that I just gave, and applying it to this case. Since, in general, a partial answer is something that rules out a possible complete answer, a partial answer to the question why E happened is a proposition that rules out some possible specification of the causes of E.
I accept these claims... for the most part. I hold that the complete answer to the question why E happened will list all of the reasons why E happened (and perhaps add that they are all of the reasons why E happened). A canonical statement of the complete answer will have the form “one reason why E happened is that A, another reason why E happened is that B, another reason why E happened is that C, ... , and these are all the reasons why E happened.” Since I think that every cause is a reason why, I agree that the complete answer specifies all the causes. But I do not accept that the complete answer specifies only causes. I hold that grounds of E are also reasons why E happened.
Like Lewis, I accept that a partial answer to the question why E happened is a proposition that rules out a possible complete answer. But since I disagree with him about what a complete answer looks like, I disagree with him about which propositions are partial answers
These places where I disagree with (my reconstruction of) Lewis’s theory, however, are not where I locate the theory’s biggest flaw. Its biggest flaw is that it is incomplete. There is more to a theory of answers to why-questions than a theory of complete answers and a theory of partial answers. A theory of answers to why-questions needs to employ the notion of a reason why, and it needs to contain a theory of reasons why.
The place for disagreement is over what it takes to be a reason why. We recover Lewis’s theory of complete answers to why-questions if we combine this neutral characterization of complete answers with (T), the claim that the (only) reasons why E happened are its causes. If you think that Lewis’s theory is wrong then this claim is the one you should put pressure on. Presenting the theory as a theory of complete and partial answers and nothing else makes the target its opponents should aim at invisible.
This is the same objection again: the hidden structure is the complete answer (it is “hidden” because people can exchange explanatory information, in Lewis’s sense, without knowing much about the complete answer); Lewis, when presented with explanations that don’t identify any causes of the event being explained, maintained that those explanations do convey some information, even if only a little bit, about the complete answer (about the causes of the event being explained); given how easy it is for a proposition to do this, Woodward thinks that this strategy for dealing with apparent counterexamples is, somehow, cheating
This line of opposition to Lewis’s theory is aided and abetted by phrasing his theory of partial answers to why-questions as a theory of explanation, as a theory that fills in the right-hand side of () Fact X explains fact Y if and only if ... It certainly sounds wrong to say that the fact that Jones was at school explains the fact that the window broke. So if one thinks that Lewis’s theory of partial answers is an attempt to complete schema (), one will think that it is false
But the example of Jones and the window does not show Lewis’s theory of partial answers, or his theory of complete answers, to why-questions to be false. The example does show that it was a public-relations mistake for Lewis to use “explanatory information” as a name for the information contained in a partial answer. Once again, blinkered focus on what is or is not an “explanation” can blind you to what is really going on.
The moral is that there is something missing from Lewis’s theory. What is missing is a theory of reasons why.The proposition that Jones was at school may be a partial answer to the question why the window broke, but (if (T) is true) that Jones was at school is not a reason why the window broke. That’s how to reconcile the thought that this proposition does say something (very weak) about the causes of the breaking with the thought that it does not seem to explain the breaking.
The fact that it is very easy for a proposition to be a partial answer to the question why E happened does not make (T) and (T) easy (or “too easy”) to defend. It just shows that trying to refute these theories by producing a proposition that (i) is in fact a partial answer to the question why E happened, but (ii) is not a partial answer according to (T) or (T), is a bad strategy. The good strategy is to produce a (true)
proposition that (i) is a reason why E happened, but (ii) is not a reason why E happened according to one of these theories. For while it is easy for a proposition to be a partial answer to the question why E happened, it is quite hard for a proposition to be a reason why E happened. If one successfully executes this strategy, the proposition one produces will probably be, by the lights of (T) and (T), a partial answer to the question why E happened. It doesn’t matter. If the proposition is not, according to those theories, a reason why E happened, those theories still stand refuted. Lewis cannot distinguish between these two strategies.
Railton offered Lewis the stellar collapse example as a counterexample to his theory. A star is collapsing, then stops. Why? Lewis accepted the answer “Because it’s gone as far as it can go ...It’s not that anything caused it to stop”. He shouldn’t have accepted this answer—something does cause the star to stop. But what Lewis wanted to say about this example applies to uncaused events generally. He was happy to say that uncaused events can be explained by the fact that they are uncaused. To keep things simple let’s suppose that stellar collapse really is this kind of example. Here is how Lewis puts his response:
I reply that information about the causal history of the stopping has been provided, but it was information of an unexpectedly negative sort. It was the information that the stopping had no causes at all, except for all the causes of the collapse which was a precondition of the stopping. Negative information is still information. If you request information about arctic penguins, the best information I can give you is that there aren’t any.
Lewis’s reply is a good one if the fact that the stopping was uncaused was offered (merely) as a partial answer to the question why the collapse stopped. But his reply is irrelevant, as a defense of (T0) anyway, if the fact that the stopping was uncaused was offered as a reason why the collapse stopped. The fact that the stopping was uncaused is not itself a cause of the stopping. So if this fact is a reason why the collapse stopped, we have a counterexample to (T0)—even though Lewis is right that the fact that the stopping was uncaused constitutes a partial answer to the question why the collapse stopped.
But uncaused events are not, in fact, counterexamples to (T)—for I see no reason to accept that the fact that E was uncaused is ever a reason why E happened. To whatever extent “E was uncaused” is a good thing to say in response to the question why E happened, it is only because it constitutes a partial answer to that question.
“E was uncaused” is a way to convey that there are no reasons why E happened, not a way to convey that the fact that E was uncaused is itself a reason why E happened.
Without the distinction between merely partial answers, and partial answers that also put forward reasons why E happened, Lewis is unable to distinguish the stronger reading of the example (where the fact that E was uncaused is offered as a reason why) from the weaker reading, or to disarm the stronger reading
it is Lange’s claim that some explanations that cite causes are not causal explanations. How can this be? If proposition P describes some causes of E, then according to Lewis’s theory, and according to my theory, P is a partial answer to the question why E happened. In what sense, then, is P not a “causal explanation”? Does Lange reject our theories of partial answers? It is hard to say; Lange does not organize his discussion around the notion of an answer to a why-question. Instead Lange’s focus is on “explanations.” He says that an explanation that cites causes may not “derive its explanatory power” from the fact that it cites causes. Such an explanation “just happens” to cite causes, and “[the fact that] a distinctively mathematical explanation happens to cite facts about the explanandum’s causes [causes of the event being explained] does not mean that it works by virtue of describing the explanandum’s causes”. Lange wants to say that citing causes only makes an explanation into a causal explanation if the fact that it cites causes is the “source” of that explanation’s “explanatory power.”
It is much easier to appreciate what Lange is trying to get at if his claims are expressed as claims about answers to why-questions, specifically as claims about reasons why
An explanation of E can happen to cite causes of E without working by virtue of citing those causes
as
An answer to the question why E happened can mention some of E’s causes, without being a correct answer in virtue of the fact that it mentions some of E’s causes.
Does this make sense? It always helps to think about questions other than why-questions. Is there any sense to be made of this claim?
An answer to the question who came to the party can mention Saturn’s rings, without being a correct answer in virtue of the fact that it mentions Saturn’s rings
It is not obvious what to think about this claim. But this much is certainly true: “Tim came to the party, and James Maxwell first figured out the nature of Saturn’s rings” cannot be a correct answer to the question who came to the party in virtue of mentioning Saturn’s rings—because it cannot be a correct answer to the question in the first place! Instead it is a body of fact consisting of two parts, an answer to the question, and also an irrelevant fact
One might think that it is always like this for why-questions about events. Any attempt to present a body of fact that (i) is an answer to the question why E happened, (ii) mentions a cause C of E, but (iii) is not correct in virtue of mentioning C, is really not an answer at all; it is at best the conjunction of an answer with the (irrelevant) fact that C happened. But in fact this conclusion is wrong. For it could be like this: C is a cause of E; and, moreover, C is a reason why E happened—and so the occurrence of C may be mentioned in an answer to the question why E happened; but the fact that C is a cause of E is not the reason why C is a reason why E happened. This, I think, is the distinction Lange was reaching for. Speaking in terms of reasons why, rather than “sources of explanatory power,” provides a clearer way to draw it.
I said that the biggest flaw in Lewis’s theory is that it is incomplete. The biggest flaw lies not in what the theory says but in what it doesn’t say: it says nothing about reasons why.
Lewis begins with an eloquent statement of the idea that we never traffic in complete answers to why-questions, that it is partial answers that we really care about. I remain skeptical. If “why” is a quantifier, if “Why did E happen?” is close in meaning to “Tell me all the reasons why E happened, then in typical contexts the domain of quantification will be restricted, and “tell me all the reasons why E happened” will mean something like, tell me all the reasons of kind K why E happened. So in a restricted context the complete answer to the question expressed by “Why did E happen?” might well be quite short, so short that we do “know enough,” and have enough breath, to state it.
But this isn’t the main point I want to make about what Lewis says. The moral he draws at the end, about the unit of explanation: Lewis was wrong. Yes, it was a mistake to think that the DN argument was the basic unit of explanation. Yes, the assumption that “explanations are arguments” did distort the search for a theory of answers to why-questions. But it is false to say that this mistake and distortion stemmed from a deeper mistake, that of thinking that there is a “basic unit of explanatory information.” There is a basic unit. Hempel just had a false view about what it is. A complete answer to the question why Q is a conjunction of facts expressed by sentences of the form “that R is a reason why Q.” The basic unit out of which answers to why-questions are ultimately built is the reason why.
The word “partial” contains “part,” but there is a sense in which being a partial answer to a why-question is not sufficient for being part of the complete answer. The complete answer is a long conjunction, so one natural sense to give “part of the complete answer” has it that the parts of the complete answer are its sub-conjunctions: a (proper) part of the complete answer to the question why Q is a conjunction of some but not all of the facts expressed by sentences of the form “that R is a reason why Q. The atomic parts of complete answers are individual reasons why. “That Jones was at school is not a reason why the window broke” is a partial answer to the question why the window broke, but it is merely a partial answer, a partial answer that is not part of the answer. After learning it you know a little bit more about the complete answer, but still cannot identify a single one of the reasons why the window broke.
—I claim that the reason why is the unit—and what we care about is not just learning some propositions that rule out one or another hypothesis about what the reasons why Q are, as Lewis maintained—for that would be like having only the note but no treasure at all—, but actually learning what some of those reasons are.
3.4 Preliminary Objections
Some “counterexamples” to (T) are easier to handle than others. This section is for a few easy ones.
- A body exists alone in a Newtonian universe, moving along a straight line at constant speed. Why is it doing that, instead of accelerating? Because there are no forces acting on it. That is, the reason why it is not accelerating is that no forces are acting on it. But the absence of forces does not appear to be a cause; surely absences cannot be causes. If this is right, then (T) is false. Presumably the example, if good, refutes (T) as well, since the absence of forces does not ground the body’s inertial motion.
This argument rests on controversial premises about causation. Two of them are: (i) fundamentally speaking, the causal relation relates events, and (ii) there are no such events as “absences.” I reject (i). Since facts can be causes, the fact that there are no forces is available. But the argument would not convince me if someone convinced me that (i) is wrong. If at bottom causation is a relation between events, I am willing to embrace the existence of absences. There’s a lot of metaphysics that needs to be done to defend one or another of these claims, but others have defended them well, I think; this is not the place.
Other examples rest on different controversial theses about causation. Many philosophers hold that “dispositions explain their manifestations.” For example: a wine glass is struck, and breaks; one reason why it broke is that it was fragile. But, the argument goes, dispositions do not cause their manifestations. That the glass was fragile was not a cause of its breaking. This is a controversial claim about causation, one that I do not find plausible.
These two examples—one about inertial motion, the other about dispositions—were not intended by their authors to be counterexamples to (T0), they were used to hat “causal explanations” need not cite causes. Was he (Lang) at least right in his intended conclusion?
So far I have studiously avoided, as far as I could, using the terms “causal explanation” and “non-causal explanation.” The time has come for me to say something about them. This game, of saying what it takes to be a “causal explanation,” while popular, baffles me. I have discussed several kinds of answers to why-questions: partial answers, complete answers, and reasons-why answers. When a philosopher urges that causal explanations are explanations that meet conditions X, I never quite know what to think.
Bas Van Fraassen held that why-interrogatives are context-dependent: one and the same interrogative sentence can be used to ask different why questions in different contexts of utterance. I think he is absolutely right about this, and I think it is now a pretty common view. But Van Fraassen also traced the context-sensitivity of why-interrogatives to a particular source. He held that the question a why-interrogative asks in a context depends in part on what “relation of explanatory relevance” is salient in the context. Causation is, perhaps, one relation of explanatory relevance, but it is not the only one, and Van Fraassen seemed to think that there were lots and lots of relations of explanatory relevance, where “lots and lots” is certainly greater than two.
If Van Fraassen is right then (T) is false, for (T) is at best compatible with the thesis that causation and grounding are the only two relations of explanatory relevance. But the examples Van Fraassen used to motivate his view do not convince me. Here is part of his central discussion
Aristotle’s lantern example (Posterior Analytics II) shows that he recognized that in different contexts, verbally the same why question [that is, the same interrogative sentence] may be a request for different types of explanatory factors. In modern dress the example would run as follows. Suppose a father asks his teenage son, ‘Why is the porch light on?’ and the son replies ‘The porch switch is closed and the electricity is reaching the bulb through that switch.’ At this point you are most likely to feel that the son is being impudent. This is because you are most likely to think that the sort of answer the father needed was something like: ‘Because we are expecting company.’ But it is easy to imagine a less likely question context: the father and son are re-wiring the house and the father, unexpectedly seeing the porch light on, fears that he has caused a short circuit that bypasses the porch light switch. In the second case, he is not interested in the human expectations or desires that led to the depressing of the switch.
Let us grant that “A reason why the porch light is on is that the switch is closed” is true in one context and false in the other (and “a reason why the light is on is that we are expecting company” is true in the other and false in the one). Van Fraassen’s thesis is that this difference in truthvalue is due to a difference in which relation of explanatory relevance is operative in the contexts; I interpret this to be the thesis that the word “reason” is context-sensitive, and has a different semantic value in the two contexts.
I think his example supports a different thesis about the source of the context-sensitivity: contextual variation in the domain of quantification. “A reason why Q is that R” can be paraphrased roughly as “There is an X such that X is a reason why Q and X is identical to the fact that R.” Our interests can influence what, in a given context, the quantifier “There is” has as its domain. Just as “There are five things in the fridge” can be true even though the refrigerator has millions of air molecules in it, because we have restricted our quantifier to edible things, “There is a reason why the porch light is on that is identical to the fact that the switch is closed” can be true in one context and false in another, even if “reason why the light is on” applies to the same set of causes of the light’s being on in both contexts; all that is required is for the father through his interests to restrict the quantifier to certain causes, causes other than human intentions. Since domain restriction can account for the example, and since domain restriction is a well-attested phenomenon,there is no pressure, from this example anyway, to also say that “reason” is itself context-sensitive
Van Fraassen cares about the context-sensitivity of “the porch light is on because the switch is closed” because its context-sensitivity shows “Why is the porch light on?” to be context-sensitive. I hold the same view about the source of this interrogative’s context-sensitivity. “Why is the porch light on?” is close in meaning to “Tell me all the reasons why the porch light is on,” so “the porch light is on because the switch is closed” will be true (and so an answer) only in contexts where the causes/reasons in the domain quantified over by “all the reasons” includes the fact that the switch is closed.
One common way of thinking about the semantic values of interrogatives has it that the semantic value of an interrogative is a partition of the set of all possible worlds (a partition in turn can be taken to be a set of propositions that are exhaustive and pairwise incompatible). Each cell of the partition is a possible complete answer to the question, and the true complete answer is the cell in the partition that includes the actual world.