October 23, 2005

Rarely better than de re

A week ago, I puzzled over the sentence "It is rare, Hamblin knows, for these kinds of situations to end better than they normally do." Kenny Easwaran wrote to suggest that this might be a sort of de dicto vs. de re ambiguity, citing as an analogous example

"What these people don’t understand is that He built the world to make us think the earth is older than it really is."

or the shorter, if less interesting, classroom favorites like "Kim thinks Leslie is taller than she is."

De dicto vs. de re is a classical distinction in the interaction of reference and modality. I'm not convinced that it works as an account of the Hamblin sentence, but back in February I went so far as to suggest that understanding the de re/de dicto distinction might help keep journalists out of trouble, so let's give it a shot.

I'll start with last Febuary's example. At the World Economic Forum in Davos, Eason Jordan, then CNN's chief news executive, said some things that got him into hot water. Rony Abovitz,who was at the session, blogged it in these terms on 1/28/2005:

During one of the discussions about the number of journalists killed in the Iraq War, Eason Jordan asserted that he knew of 12 journalists who had not only been killed by US troops in Iraq, but they had in fact been targeted. He repeated the assertion a few times, which seemed to win favor in parts of the audience (the anti-US crowd) and cause great strain on others.

Controversy ensued, and Jordan resigned (as he put it) to "prevent CNN from being unfairly tarnished by the controversy over conflicting accounts of my recent remarks regarding the alarming number of journalists killed in Iraq". The disagreements were not only over what Jordan said, and what the facts in Iraq were, but also about how to interpret whatever it was he said. And some of these interpretive differences hinged on the classic de re/de dicto ambiguity. Suppose that Jordan said something like

U.S. forces in Iraq have intentionally killed 12 journalists.

This isn't an exact quote -- some people believe that Jordan resigned to prevent the transcript from being released -- but whatever he said was apparently subject to the same ambiguity. Jay Rosen, from NYU's department of journalism, put it like this:

"The original account was too ambiguous for me. It had him saying United States soldiers targeted journalists, and then claiming that's not what he meant. He later explained it as: the soldiers were trying to kill these people, but did not know they were shooting at journalists."

That interpretation is the de re ("about the thing") reading. The soldiers were trying to kill certain people, without knowing that their targets would turn out to be journalists. And Oedipus wanted to marry Jocasta, without knowing that she was his mother. Here the belief or desire is all about the thing referred to -- the targeted person, the spouse, whatever -- and the description comes from outside.

The alternative is the de dicto ("about the saying" ) reading. Here the belief or desire is all about the description: the soldiers want to go out and kill some journalists; Oedipus is an adoptee who wants to find his mother and marry her.

According to the Stanford Encyclopedia of Philosophy,

The idea of the systematic distinction between the readings de dicto (in sensu composito) and de re (in sensu diviso) of modally qualified statements was introduced into medieval discussions in Abelard's investigations of modal statements (Super Periherm. 3-47, Dialectica 191.1-210.19), and was often mentioned, as in the Dialectica Monacensis, in discussions of the composition-division ambiguity of sentences.

Despite his understanding of the de re/de dicto distinction, Abelard came to a more troubled end than Eason Jordan did. All the same, his legacy includes the University of Paris, an enduring story of love in adversity -- and one of the few bits of Latin that remain in common philosophical use.

One example of the prevalence of the traditional use of modal notions can be found in the early medieval de dicto/de re analysis of examples such as ‘A standing man can sit’. It was commonly stated that the composite (de dicto) sense is ‘It is possible that a man sits and stands at the same time’ and that on this reading the sentence is false. The divided (de re) sense is ‘A man who is now standing can sit’ and on this reading the sentence is true.

In the middle of the 20th century, W.V. Quine re-analyzed this distinction as a matter of the scope of logical operators, and applied it to propositional attitude terms such as believe. As the Stanford Encyclopedia of Philosophy explains his analysis:

[1] Ortcutt believes that someone is a spy.
This could mean just that
[2] Ortcutt believes that there are spies
or that Ortcutt has more interesting information:
[3] Someone is an x such that Ortcutt believes that x is a spy.
The distinction here can be seen as a distinction of scope for the existential quantifier. In [2], the existential quantifier is interpreted as having small scope, within the propositional clause of the belief attribution.
[2*] Ortcutt believes: ∃x, x is a spy.
In [3], however, the existential quantifier has large scope, selecting an individual and then ascribing a belief that relates Ortcutt to that particular individual.
[3*] ∃x, Ortcutt believes that x is a spy.

(The backwards E is the "existential quantifier", so that the de re version ∃x, Ortcutt believes that x is a spy is read "there exists an x such that Ortcutt believes that x is a spy".)

OK, how could this help us with "It is rare, Hamblin knows, for these kinds of situations to end better than they normally do"?

Let's start with a scope ambiguity involving a comparative and a verb like want, which will make a good Quinian de re/de dicto example:

Kim wants to score higher than Leslie scored.

This could mean that there exists some score L (which we describe as the score Leslie got) such that Kim wants to score higher than L. That's the de re reading -- it's all about the numerical score. Kim doesn't care that it was Leslie's score, and maybe she doesn't even know who Leslie is.

Alternatively, the same sentence could mean that Kim wants her score to beat Leslie's, regardless of what it is. That's the de dicto reading -- it's all about Leslie's level of achievement.

We can get a similar scope ambiguity with a predicate like rare.

It was rare for Kim to score higher than Leslie did.

Translated into "heavy English", this might mean something like

"There was an L=Leslie's score, such that it was rare that there was a K=Kim's score and K was greater than L."

or it might mean something like

"It was rare that there was an L=Leslie's score and a K=Kim's score such that K was greater than L."

Here the de re/de dicto distinction doesn't arise from the interpretation of a traditional modal operator (like "necessarily") or a propositional attitude verb (like "wants"), but instead in reference to a statistical sampling process. We're not talking about whether necessity, belief or desire applies to a thing per se or to a thing under a given description. Instead, we've got a scope ambiguity having to do with a sampling process: do we fix Leslie's score and then look at the statistics of Kim's scores relative to it? or do we look at the statistics of the relationship between Kim's scores and Leslie's scores?

However, I'm not sure that the original example

"It is rare ... for these kinds of situations to end better than they normally do."

involves a coherent scope ambiguity of this type, because the reference point "how these situations normally [end]" comes out of the same sampling process referred to by the phrase "it is rare".

The idea seems to be that how these situations normally end is "badly", and if we substitute "badly" for "how they normally do" in

It's rare for these situations to end better than they normally do.

then we get

It's rare for these situations to end better than badly.

which is awkward but coherent. I agree that this is probably what the writer had in mind, and it does seem analogous to some of the medieval de re/de dicto examples, but I don't see a coherent reconstruction in terms of a Quinian scope difference. However, that's probably because I'm not a real semanticist, I just play one occasionally on Language Log.

Posted by Mark Liberman at October 23, 2005 07:44 AM