Quantifier Domain Restriction

I am still thinking about Soames' article The Gap Between Meaning and Assertion. Soames makes some interesting remarks about quantifier domain restriction. He considers the possibility that quantificational domains are sets, and that quantified noun phrases contain hidden domain variables that are contextually completed. On the view he has in mind, "each quantifier occurrence is assigned a contextually determined range as all, or part, of its semantic content relative to the context" (p. 16). But Soames thinks that "a fatal flaw in this approach is that althought contextually provided subsets of the domain may be relevant to the extensions of quantifiers, they are not parts of their semantic contents" (p. 16)

He offers the following example. Consider:

(1) Everyone writes a senior thesis.
(2) Everyone who is a Princeton student writes a senior thesis.

Suppose I am asserting (1) in a conversation about students at Princeton University. Then, according to Soames, "the statement I make is one that is true when evaluated at any possible circumstance of evaluation w iff something like [2] is true at w" (p. 16).

I think Soames may be right about (1), but I think that is because (1) is a kind of generic statement--a law-like generalization. But I do not think his criticism applies to all cases. Consider the following case:

You: How did Amy, Bob, Carl, Dan, and Elizabeth do on the logic exam last semester?
Me: Everyone got an A.

Here it is quite plausible that the quantifier is assigned the set containing Amy, Bob, Carl, Dan and Elizabeth as part of its content.