De Dicto vs. De Re Modalities

The terminology is medieval, and it is almost always applied as if the main issue is modality, but, in fact, the issue is about how descriptions are, or can be used.

Quine uses the following example: Ralph Ortcutt, who is really short, works at a defense plant. You are a fellow employee, and know Ralph, but you do not know that he is a spy. In fact, he is the shortest spy.

Leaving the plant late one night, you see someone hidden in the bushes having, it appears, just climbed out of a window that was supposed to be locked. You suspect (correctly) that the person you saw is a spy. In fact, though you don't know it, it was Ralph.

Consider the following sentences:

• You suspect that the person you saw is a spy. True.
• You think, "the person I saw is a spy." True.
• You suspect that Ralph is a spy. False de dicto (as said), but true de re (in reality).
• You think, "Ralph is a spy." False.
• You suspect that the shortest spy is a spy. True de dicto, and true de re, but not related in the way you may have expected.
• You think, "the shortest spy is a spy." Of course, everyone knows that, but it has nothing to do with your belief about the guy you saw.

Most of us, while we know that the shortest spy is a spy (de dicto), do not believe, of the shortest spy that he or she is a spy (de re). Ralph is in even worse shape: he knows, de dicto, that the shortest spy is a spy, and he believes, de re, that the shortest spy is a spy, though he doesn't think (de re) "the shortest spy is a spy."

Why did I need to use modalities (thinks, believes, suspects) in my examples? The distinction is just about how the names are used, not at all about anything modal. If what I have been saying makes sense, then there are two distinct readings of the following nonmodal sentence:

• The guy in the bushes is a spy.

Computer languages have a distinction between call by name and call by value which is the same distinction:

In a call by value language, to compute f(g(2)) the computer would first find the value of g(2) and then feed that to f. So (2+2)+1 is computed by finding that 2+2=4 then computing 4+1.

In a call by name language, to compute f(g(2)) the computer would first try to compute f(...), discover that what was in there g(2) (the name, not the value) needs to be computed, send out for the answer, and continue when the answer is returned.

Both yield the same answer, and so we don't ordinarily notice. The same is true for the nonmodal "de re" and "de dicto" readings. We can see the difference in computer languages through "side effects," like printing, and we can see the difference in ordinary language when we stick modalities on the front.

When a name is used descriptively (that is, in the way it is used in de dicto claims), it denotes, stands for, whatever object has the description. When a name is used referentially (de re), it stands for the very object picked out, not in virtue of the description. (Think of , "Lincoln might not have been named 'Lincoln.'") A name used referentially tags the thing it names without requiring or making ineliminable use of any description. A name used referentially is pretty much what Kripke is taking to be a rigid designator. ("Pretty much" because Kripke's criterion is modal, and one can make sense of the idea of a referential name even if there is no such thing as a modality.)

Geach doesn't think there is such a thing as a thing—only things of some kind or other, and so he won't think that there is a referential use of names, or that there are de dicto readings of sentences. Geach wouldn't think there is a de re use of "Lumpl": if there were, it would have to be the same as the de re use of "Goliath." What he would allow instead is de re uses of "Lumpl" when pieces of clay are the kind of thing we are discussing and de re uses of "Goliath" when statues are the kind of thing we are discussing. He would deny that the question whether Lumpl is the same thing as Goliath makes any sense.

Gibbard thinks something related: he thinks you can't make modal claims about things, only about things with reidentification conditions. So, he doesn't think there is a de re use of "Lumpl": if there were, it would have to be the same as the de re use of "Goliath." What he allows instead is de re uses of "the concept of Lumpl" and "the concept of Goliath" and a way of talking about whether the same thing falls under two concepts in a possible world.

Plantinga rehearses standard arguments against the coherence of de re modal statements. The discussion is always about either, whether certain claims are meant de dicto or de re, or whether it is coherent to think that we can tell which thing an object is in some counterfactual circumstance, that is, whether it is possible to refer in a purely referential way.

-- ShaughanLavine - 15 Feb 2008