Epistemic counterpart semantics

I have decided to write a series of posts on epistemic applications of counterpart semantics, mostly to organise my own thoughts.

Let's start with a motivating example, from Sæbø 2015.

On September 14 2006, Mary Beth Harshbarger shot her husband, whom she had mistaken for a bear. At the trial, she "steadily maintained that she thought her husband was a black bear", as you can read on Wikipedia.

We're interested in this sentence: "Mary thought her husband was a bear."

The sentence has two readings. On its de dicto reading, it says that Mary thought that she was married to a bear. It is safe to assume that this is not what Mary Harshbarger meant at the trial. The intended reading is de re. On this reading, the sentence says that Mary thought of her husband that he was a bear.

But what, exactly, does this mean?

A simple and popular answer is that the de re reading attributes to Mary a thought whose content is a singular proposition directly involving her husband (Mark).

Let's call this the direct interpretation of the (de re) attitude report. It leads to some well-known problems, of which I'll mention two.

First, Mary Harshbarger presumably knew a lot about her husband Mark. Among other things, she knew that he has a brother, and that he was human. So she believed of her husband that he is human, and that he is not a bear. But if what Mary said at the court is true, then the direct interpretation implies that she had inconsistent beliefs. She believed a singular proposition as well as its negation. Intuitively, however, there was nothing inconsistent about her state of belief when she saw a dark creature approaching in the woods and thought it was a bear.

This problem generalises. Assuming our attitude reports are by and large correct, the direct interpretation implies that people are belief-related to singular propositions about all kinds of things. How do these beliefs rationally interact? How do they respond to evidence? How do they guide behaviour? It's really hard to spell out a satisfactory answer.

Second, let's have a closer look at what Mary knew and didn't know about her partner. Even in long-term relationships, couples often have mistaken beliefs about one another. Perhaps Mary mistakenly believed that Mark was attracted to her sister. [Imagine we go on like this for a while.] But she did not mistakenly believe that he was a God, or a beetle, or a bear. – See? Now I said the opposite of what Mary said at the trial. I said she did not believe he was a bear. The truth-value of "Mary thought her husband was a bear" (on its de re reading) seems to depend on the context of attribution.

Friends of the direct interpretation might infer that what Mary Harshbarger believed at 7:55 pm on September 14 2006 is relative to a context of attribution. That is highly counter-intuitive. Alternatively, they might say that one of the two judgements is mistaken. But which one? And why?

Unlike the first problem, this one is really a linguistic puzzle. We are prepared to affirm and deny the very same belief report, depending on the linguistic context, without gaining or losing relevant information in the switch. One would like to understand what is going on here, and the direct interpretation of the report does not seem to offer an explanation. We are going to meet a lot of other linguistic puzzles later on.

For now, what I said should suffice to motivate considering an alternative to the direct interpretation.

The alternative I have in mind is often associated with Quine (1956) and Kaplan (1968), but might also be attributed to Russell (1911).

On this view, ordinary people are not belief-related to singular propositions involving ordinary objects at all. The beliefs we have about objects in our environment always represent the objects under a particular "mode of presentation". This is a form of descriptivism, but it is not a claim about the meaning of proper names or other linguistic expressions. I will call it doxastic descriptivism.

According to doxastic descriptivism, what Mary Harshbarger really believed at the fateful moment was something like this: that creature approaching through the woods is a bear. She also believed: the person to whom I am married is not a bear. The two beliefs are obviously consistent.

One problem for doxastic descriptivism is that it does not seem to fit the way we talk. If people don't have singular beliefs about ordinary objects, how should we interpret de re reports that appear to attribute just such beliefs?

Here's how Hawthorne and Manley (2014: 52) put the problem:

Those who reject [the direct interpretation] need an alternative account of how the truth-conditions for [de re] belief reports are compositionally generated. Are we to deny that 'believes' expresses the relation believing, or are we to deny that 'Bill is happy' expresses a singular content? Pending some worked-out semantics of belief ascriptions or some novel criterion for when sentences express singular propositions, [the direct interpretation] is an attractive principle.

Return to "Mary thought her husband was a bear". On the Quine-Kaplan model, what makes this true (on its de re reading) is that (1) Mary thought that the creature approaching through the woods was a bear, and (2) in fact, the creature approaching through the woods was her husband. More generally, "S believes that x is F" is true iff there is a suitable role R such that (1) S believes that whatever plays R is F, and (2) in fact, x plays R.

The challenge is to explain how these truth-conditions come about. It is clear that the relevant role is not part of the meaning of the corresponding singular term. In English, "my husband" does not mean the creature approaching through the woods.

If we assume that belief reports expresses an interesting relation between the subject and some kind of content expressed by the embedded sentence, it is really mysterious how the sentence could have the suggested truth-conditions.

It helps to think of belief reports on the model of modal constructions, with 'believes' functioning like a necessity operator. The Quine-Kaplan truth-conditions then fall out naturally from an attractive semantics for quantified modal logic.

The semantics I have in mind is a form of counterpart semantics – very roughly the kind of semantics outlined in Lewis's 1968 paper on counterpart theory. But we'll need to make some changes and fill in some details.


# on 13 May 2020, 17:30

Looking forward to this series, as I've been thinking a lot about related issues recently! I know the arguments here aren't meant to be decisive against the "direct interpretation" but I'm curious what you make of what I take to be the standard replies. (Jeremy Goodman and I are fans of the direct interpretation of this kind of sentence in "Perspectivism", and it's natural to endorse the same view on one way of spelling out the models I present in "Fine Grained Semantics".)

To the first argument, fans of the direct reading usually say that having contradictory beliefs isn't what makes for coherence/incoherence, but that underlying the beliefs we report in natural language is a belief-relation that is somehow relativized to representations, and that we can explain incoherence in terms of a person's relation to these representations. For instance, Lois stands in the belief-relation to a mental representation corresponding to "Superman flies" and also in the same relation to one corresponding to "Clark does not fly", but she's not incoherent provided she doesn't also stand in the relation to "Superman does not fly". Of course she has contradictory beliefs as a result of this, but that's not so bad.

To the second argument, why is it so counterintuitive that it depends on context? A sentence and its negation are both true -- isn't the usual diagnosis of such a case that there must be some kind of context-sensitivity?

# on 13 May 2020, 19:56

Thanks Harvey! I'm afraid I haven't yet read the papers you mentioned -- I've put them on the reading pile.

It definitely helps to assume that beliefs in singular propositions are somehow relative to modes of presentation. I wonder how much work will be left for singular content on this approach. Arguably, when we consider whether an agent is rational, whether her beliefs are supported by evidence, whether her acts make sense in light of her beliefs and desires, etc., the relevant beliefs are always non-singular or a combination of singular belief with non-singular mode of presentation, and then the mode of presentation is doing most of the work. I suspect there's a way of spelling out this idea on which it is equivalent to the Russell/Quine/Kaplan view I want to defend. But I definitely need to think more about all this.

Thanks for pushing me on the second argument. The intuition is that there's an important sense in which what an agent believes -- how she represents the world as being -- is not relative to a context of attribution. If Mary had a certain belief at the fateful moment, then there's no context in which one can truly deny that she had that belief. Does that not sound intuitive to you?

# on 19 May 2020, 01:32

Hi Wolfgang, I'm so sorry for being so slow! I assumed it would send me an email if there was a reply but then I didn't get one (maybe it went to spam?) and it slipped my mind.

Yes I agree that on the first way much more work will be done by the "mode of presentation". There may still be a lot done by "content" as well. It might be true that in terms of what's going on in the mind at some level of description the two views are quite similar. But they offer starkly different views of how we talk about the mind, and this will give different validities and entailment patterns in English.

On the second point: I agree that, at first sight, that's quite intuitive. But when you realize that Peter knows that Paderewski was in the concert hall last night, and also doesn't know that Paderewski was in the concert hall last night, there's a lot of pressure on the intuitive thought. I see that there's some way in which the letter of the intuitive thought can be preserved even in the face of this example, but I feel that once you see this example it's no longer clear that the intuition is something we want to respect. Peter knew that Paderewski was in the concert hall, and yet we we were happy to say he didn't.

# on 19 May 2020, 15:18

Hi Harvey, no problem. You're right, I should send an email notification when there's a follow up. I'll try to add this feature soon.

I'm not sure I agree that the sort of account I want to develop and the "direct" alternative necessarily make different predictions about validities and entailment. But I suspect the version I will defend does. It'll be good to keep this point in mind.

About the Paderewski case: I do want to respect the intuition. For me, the intuition isn't really that belief reports in ordinary language are context-independent. Rather, it is that there's a fact of the matter about how an agent represents the world to be, and that this is not relative to a context of attribution.

# on 21 May 2020, 12:54

On Paderewski: hm, I see, but as I understood what you call the second problem above, it was really one about attitude-*ascriptions* not attitudes themselves. Maybe we're just talking past each other, though. I'll stay tuned for the next posts!

Add a comment

Please leave these fields blank (spam trap):

No HTML please.
You can edit this comment until 30 minutes after posting.