Chalmers on De Re Epistemic Ascriptions

In "The Foundations of Two-Dimensional Semantics" David Chalmers mentions that it is difficult to make sense of de re epistemic ascriptions. To see what the problem is, consider:

(1) It is a priori that Phosphorus has been visible in the morning.

The proper name in (1) can, via property abstraction, move out of the scope of the a priori operator. Then we get:

(2) Phosphorus is an x such that it is a priori that x has been visible in the morning.

This is the de re reading of (1). The a priori operator operates on some aspect of a sentence that has a variable in it.

The reason it is difficult to make sense of this is that the a priori operator cannot be taken to operate on singular propositions. If it could, then we should expect 'Hesperus is Phosphorus' to be a priori (which, of course, some thinkers do want to argue that it is). It is better to think of the a priori operator as operating on what Chalmers calls the 'primary intension'. The primary intension is (roughly) the possible-world equivalent of a Fregean thought. The problem now is that it is difficult to see what sort of Fregean thought/primary intension could possibly be expressed by the operand sentence in (2), viz. 'x has been visible in the morning'.

It might be suggested that the Fregean sense/primary intension of 'x' is something that can be glossed as 'the bright object in the morning sky'. But that does not seem right. For what is special about the de re reading of a sentence like (1) is that it "gets at" the object directly, as Russell would put it.

An alternative approach would be to say that the primary intension of 'x has been visible in the morning' is contingent and hence not a priori (as 'apriority' is defined in terms of the necessity of the primary intension). This would follow if the primary intension of 'x' is similar to the primary intension of a name whose referent one cannot really describe, even though one can use the name proficiently (by being a link in a causal chain that leads back to the referent). It would follow, then, that even though (1) is true on the de dicto reading, (2) is false. For just like the Fregean element expressed by 'some x has been visible in the morning', the Fregean element expessed by 'x has been visible in the morning' is true at some conceivable scenarios and false at others.

If the latter is correct, then there is an independently interesting lesson to be learned, namely that property abstraction on sentences involving proper names is not simply a trivial move, as sometimes suggested. Property abstraction can make a difference to truth-conditions even for sentences containing no quantified noun phrases.