Dr. Smith ate a banana

Suppose you receive this trustworthy report:

1. Dr. Smith ate a banana.

You are now in a position to learn this additional fact:

2. Someone whose last name is “Smith” ate a banana.

But (2) does not logically follow from (1). So how do we learn (2) from the report?

Knowledge of English tells us that “Dr. Smith ate a banana” has “Dr. Smith” as the subject and that this sentence attributes eating a banana to the subject of the sentence. Assuming defeasibly that the the use of English in the report is correct, we conclude that someone correctly styled “Dr. Smith” was reported to have eaten a banana. And assuming defeasibly that the report itself is factually correct, we conclude that:

3. Someone correctly styled “Dr. Smith” ate a banana.

Knowing English, we also know that anyone correctly styled “Dr. Smith” has the last name “Smith”, so we get (2). We also know that anyone correctly styled “Dr. Smith” has a doctorate, so:

4. Someone with a doctorate ate a banana.

These are instances of the familiar fact that what we learn from receiving a report goes beyond the propositional content of the report.

GPS signals, normativity and the morality of lying

I will argue that lying is never permissible. The argument is a curious argument, maybe Kantian in flavor, which attempts to establish the conclusion without actually adverting to any explanation of what is bad about lying.

GPS satellites constantly broadcast messages that precisely specify the time at which the message is sent together with precise data as to the satellite orbit. Comparing receipt times of message from multiple GPS satellites with the positions of the satellites, a GPS receiver can calculate its position.

A part of the current design specifications of US GPS satellites is apparently that they can regionally degrade the signal in wartime in order to prevent enemies from making use of the signal (US military receivers can presumably circumvent the degradation).

Now, let’s oversimplify the situation and make up some details (the actual GPS signal specifications are here and the points I am making don’t match the actual specifications), since my point is philosophy of language, not GPS engineering. So I’m really talking about GPS satellites in another possible world.

Suppose that normally the satellite is broadcasting the time n in picoseconds up to a precision of plus or minus ten picoseconds, and suppose that currently we receive a message of n in the time field from a satellite. What does that message mean?

First of all, the message does not mean that the current time is n picoseconds. For the design specifications, I have stipulated, are that there is a precision of plus or minus ten picoseconds. Thus, what it means is something more like:

1. The current time is n ± 10 ps, i.e., is within 10 ps of n ps.

But now suppose that it is a part of the design and operation specifications that in wartime the locally relevant satellites add a pseudorandom error of plus or minus up to a million picoseconds (remember that I’m making this up). Then what the message field means is something like:

2. Either (a) this is a satellite that is relevant to a war region, the current time is n ± 10⁶ ps and [extra information available to the military], or (b) the current time is n ± 10 ps.

In particular, when wartime signal degradation happens, the time field of the GPS message is (assuming the satellite is working properly) still conveying correct information—the satellite isn’t lying. For the semantic content of the time field supervenes on the norms in the design and operation specifications, and if these norms specify that wartime degradation occurs, then that possibility becomes a part of the content of the message.

Suppose lying is sometimes morally obligatory. Thus, there will be a sentence “s” and circumstances C_s in which it is both true that s and morally required to say that not s. Suppose Alice is uttering “Not s” in an assertoric way. Morality is part of Alice’s (and any other human being’s) “design and operation specifications”. Thus on the model of my analysis (2) of the semantic content of the (fictionalized) time field of the GPS message, what is being stated or asserted by Alice is not simply:

3. Not s

but rather:

4. Either (a) C_s obtains, or (b) not s.

But if that’s the content of Alice’s statement, then Alice is not actually lying when she says “Not s” in C_s. And the same point goes through even if Alice isn’t obligated but is merely permitted to say “Not s” in C_s. The norms in her design and operation specifications make (4) be the content of her statement rather than (3).

In other words:

5. If lying that s is obligatory or permissible in C_s, then lying is actually impossible in C_s.

But the consequent of (5) is clearly false. Thus, the antecedent is false. And hence:

6. Lying is never obligatory or permissible.

Note that a crucial ingredient in my GPS story is that the norms governing the degradation of GPS messages are in some way public. If these norms were secret, then the military would be making the GPS satellites do something akin to lying when they degraded their messages. But moral norms are essentially public.

Objection 1: The norms relevant to the determination of the content of a statement are not moral but linguistic norms. The moral norms require that Alice utter “Not s” in an assertoric way only when (4) obtains. But the linguistic norms require that Alice utter “Not s” in an assertoric way only when (3) obtains. And hence (3) is the content of “Not s”, not (4).

Response: This is a powerful objection. But compare the GPS case. We could try to distinguish narrowly technical norms of satellite operation from the larger norms on which GPS satellites are controlled by the US military in support of military aims. That would lead to the thought that the time field of the satellite (on my fictionalized version of the story) would mean (1). But I think it is pretty compelling that the time field of the satellite would mean (2). The meaning of the message needs to be determined according to the overall norms of design and operation, not some narrow technical subset of the specifications. Similarly, the meaning of a linguistic performance needs to be determined according to the overall norms of design and operation of the human being engaging in the performance. And it is precisely the moral norms that are such overall norms.

Second, linguistic norms are norms of voluntary behavior, since linguistic performance is a form of voluntary behavior. But a norm of voluntary behavior that conflicts with morality is null and void insofar as it conflicts, much as an illegal order is no order and an unconstitutional law is no law.

Third, on a view on which linguistic norms have the kind of independence from moral norms that the objection requires, it is difficult to specify what makes them linguistic. For we cannot simply say that they are the overall norms governing linguistic behavior. Moral norms do that, as well. A distinction like the one in the objection would make sense in the case of something where the rules are formalized. Thus, there are circumstances when the rules of chess require one to do something immoral. (For instance, suppose that a tyrant tells you she will kill an innocent unless you move a pawn forward by three squares. The rules of chess require you to refrain from doing that, but it is immoral for you to refrain from it.) But the rules of chess are simply a well-defined set of statements about what constitutes a game of chess, and it is relatively easy to tell if something is a rule of chess or not. But linguistic norms are just some among the many norms governing human behavior, and it is hard to specify which ones they are, if one can't do it by the subject matter of the norms. (I am also inclined to think that the rules of chess might not actually be norms; they are, rather, classificatory rules that specify what counts as a victory, loss, draw or forfeit; the norms governing play are moral.)

Objection 2: Content is not normatively determined.

Response: If that’s right, then my line of argument does fail. But I think a normative picture of content is the right one. In part it’s my Pittsburgh pedigree that makes me want to say that. :-)

Objection 3: Bite the bullet and say that when Alice utters “Not s”, she is in fact asserting (4) and not lying even if C_s obtains. While on this view, technically, lying is never permissible, in practice the view permits the same behaviors as a view on which lying is sometimes permissible.

Response: This just seems implausible. But I wish I had a better response.

Whatever content can be conveyed metaphorically can be said literally

Suppose you say something metaphorical, and by means of that you convey to me a content p. I now stipulate that "It's zinging" expresses precisely the content you conveyed. Technically, "It's zinging" is a zero-place predicate, like "It's raining." And now I say: "It's zinging." The literal content expressed by "It's zinging" is now equal to the metaphorical content conveyed by what you said. A third party can then pick up the phrase "It's zinging" from me without having heard the original metaphor, get a vague idea of its literal content from observing my use of it, and now a literal statement which has the same content as was conveyed by the metaphorical statement can start roaming the linguistic community.

Thus: If you cannot say something literally, you cannot whistle it either. For if you could convey it by whistling, you could stipulate a zero-place predicate to mean that which the whistling conveys.

Objection 1: My grasp of "It's zinging" is parasitic on your metaphor, while the third-party doesn't have any understanding.

Response: Yes, and so what what? I wasn't arguing that you can usefully get rid of metaphor. It may well be essential to understanding the content in question. My point was simply that there can be a statement whose literal semantic content is the same as the content conveyed by your metaphor. Understanding is something further. This is very familiar in cases of semantic deference. (I hear physicists talking about a new property of particles. I don't really understand what they're saying, but I make the suggestion that they call that property: "Zinginess." My suggestion catches on. I can say: "There are zingy particles", and what I say has the same content as the scientists' attribution of that property. But while the scientists understand what they're saying, I have very little understanding.) The third party who hasn't heard the original metaphor may not understand much of what he's saying with "It's zinging." But what he's saying nonetheless has the literal semantic content it does by deference to my use of the sentence, and my use of the sentence has the literal semantic content it does by stipulation. All this is quite compatible with the claim that any decent understanding of "It's zinging" will require getting back to the metaphor. But, nonetheless, "It's zinging" literally means what the metaphor metaphorically conveyed.

Objection 2: The stipulation does not succeed. (This is due to Mike Rea.)

Response: Why not? If I can refer to an entity, I can stipulate a name for it, no matter how little I know about it. I may have no idea who killed certain people, but I can stipulate "Jack the Ripper" names that individual. My stipulation will succeed if and only if exactly one individual killed those people. Similarly, if I can refer to a property, I can stipulate a one-place predicate that expresses that property. (If a certain kind of Platonism is true, this just follows from the name case: I name the property "Bob", and then I have the predicate "instantiates Bob".) In cases without vagueness, contents seem to be propositions, and zero-place predicates express propositions, so just as I can stipulate a one-place predicate to express a property, I should be able to stipulate a zero-place predicate to express a propositions. And in cases of vagueness, where maybe a set of propositions (or, better, a weighted set of propositions) is a content, I should be able to stipulate a similarly vague literal zero-place predicate as having as its content the same set of propositions.

There are many ways of introducing a new term into our language. One way is by stipulating it in terms of literal language. That's common in mathematics and the sciences, but rare in other cases. Another way is by ostension. Another is just by talking-around, hoping you'll get it. One way of doing this talking-around is to engage in metaphor: "I think we need a new word in English, 'shmet'. You know that butterflies in the stomach feeling? That's what I mean." We all understand what's going on when people do this kind of stipulation. For all we know, significant parts of our language came about this way.

Epiphenomenalism and causal theories of content

According to causal theories of content, what makes my beliefs be about, say, horses is that the beliefs have the right causal connection with horses. Of course, there are going to be harder cases: I also have beliefs about unicorns, namely that they do no exist, and these beliefs do not have a causal connection with unicorns. But the unicorn beliefs get their intentionality derivatively from other thoughts, like those of animals and horns, from which they are constituted, derived, etc.

Now, beliefs about qualia are not constituted, derived, etc. from thoughts about non-qualia. So by causal theories of content, my beliefs about qualia have to have the right causal connections with qualia. So, causal theories of content are incompatible with epiphenomenalism, since according to epiphenomenalism qualia aren't causes.

Causal theories of content are the materialist's best bet for a theory of content. Qualia are meant to be a very modest addition to the materialist's story. But they aren't a modest addition—they require a revision of the theory of content.

Causal theory of content, religious experience, numinousness and naturalism

1. If naturalism (of the non-Aristotelian sort) is true, the causal theory of content is true. (It's the only decent naturalistically acceptable theory of content.)
2. The content of some religious experience involves the property of numinousness.
3. Numinousness is not a natural property and cannot be reduced solely to natural properties.
4. If the causal theory of content is true, and the content of an experience E involves a property P, then some experience is caused either by something's having P or by a combination of entities' having the properties that P reduces to.
5. If some experience is at least partly caused by something's having a non-natural property, then naturalism is false.
6. So, if naturalism is true, some experience is caused by something's being numinous or by a combination of entities' having the properties that numinousness reduces to. (1, 2 and 4)
7. So, if naturalism is true, some experience is at least partly caused by something's having a non-natural property. (3, 4)
8. So, if naturalism is true, naturalism is false. (5, 7)
9. So, naturalism is false. (8)

A simple reductive theory of content

The simple normative theory of content says that:

1. A state S of x represents that p iff x ought not be in S if not-p.

The "ought not be in S if not-p" should be read as a conditional norm. In particular, it should not be simply analyzed as "not-p → ought(not-S)" where the → is a conditional.

Some illustrations. One ought not say "Snow is white" if it's not the case that snow is white. So saying "Snow is white" represents that snow is white. One ought not kill a person without appropriate authority. Therefore, killing a person represents having that authority. One ought not have an apparent perception of an elephant if it's not the case that an elephant is present. Hence, an apparent perception of an elephant represents that an elephant is present. A joey ought not venture far from the pouch if it is not mature enough. Hence, the joey's venturing far from the pouch represents that it is mature enough.

The joey example shows that representation is a wide phenomenon, and need not involve a mind (the joey has a mind of some sort, but the representation in question need not be mental). The speech example shows that the theory requires the truth account of the norm of assertion.

Objection 1: If that p entails that q, and x ought not to be in S if not-p, then x ought not to be in S if not-q. But then if a state represents a proposition, it represents everything which that proposition entails. And that is implausible.

Response: It does not seem to me that the conclusion is really so very bad. Bradwardine's account of assertion has a similar consequence, and that seems to be a strength of it. But I do not know that the conclusion follows. It depends on the logic of obligations. Suppose the entailment thesis is true and imagine that I validly promise not to scratch my head unless it's an odd numbered day of the month. Suppose that I scratch my head on the 8th of the month. Do I then violate an infinite number of obligations: the duty not to scratch if the day is even, the duty not to scratch if the day is divisible by four, the duty not to scratch if the day is even and less than 9, the duty not to scratch if the day is even and less than 10, and so on? This seems to be multiplying immorality without good reason. It seems better to say: I now have a duty not to scratch my head if it's not an odd numbered day, and if I scratch my head on a day that's divisible by four, then necessary I violate that duty.

Objection 2: It is very plausible that on this view every living thing represents something.

Response: What's wrong with that? We have simply reprised a part of Leibniz's system.

Objection 3: Often norms conflict. When they do, the same state can end up representing contradictory states of affairs. That seems counterintuitive.

Response: We should take the "ought" in (1) as an all-things-considered ought for x. We could deny that the prudential and moral oughts are potentially conflicting all-things-considered oughts. (Either by denying that they are potentially conflicting or denying of at least one of them that it is all-things-considred.) If we can't do that, then we do in fact get the consequence that sometimes the same state can represent contradictory states of affairs. But why is that so bad? After all, if I am speaking to two people, each of whom speaks a different dialect of English, I might duplicitously utter a sentence which to one speaker will represent p and to the other a conflicting q. In such a case, my one speaking does represent two things that conflict.

Objection 4: It is sometimes all-things-considered right to speak falsely.

Response: Lying is always all-things-considered wrong. It is a betrayal of trust which does not treat the interlocutor as an end in herself. And it may be that the right way to understand the moral duty correlate to the wrong of lying is as the duty to speak only the truth.

Final remark: Even though these objections can be overcome, I have a strong intuition that representation involves a mind, which (1) does not do justice to. For that reason, I reject the view.