Kent Bach has for many years been advancing a line of argument the overall structure of which is the following: If we can show, for some expression e, that the determination of its content in context somehow depends on the speaker's intentions, then we have shown that the content-determination of e is not (purely) semantic. This stance has both supporters and adversaries. Among the former are pramgatists like Recanati (in Literal Meaning) and Carson (in Thoughts and Utterances). Among the adversaries are King & Stanley.
When it comes to pronouns (on their indexical uses), the consesus is roughly that they pretty much all of them rely on the speaker's referential intentions for their reference-determination (many persist in holding that 'I' is a pure indexical, but post-it notes and recorded messages are creeping their way in). In Thought and Reference (1987/1994), Bach applies the above strategy to pronouns in the following way:
Since the meaning of an indexical is supposed to specify the contextual parameters that constrain the referential intentions with which the indexical can be used (literally), the theory cannot allow the speaker's intention to count as a separate parameter of context. If it did allow this, it could not explain how that intention is constrained by the meaning. Instead, it would self-defeatingly treat that intention as just another contextual parameter, an independent variable on which meaning operates. (p. 178)
I've been trying for some time to figure out what Bach is getting at in this passage. Here's what I currently think it is:
(i) Meaning introduces parameters of indices. For instance, we have an agent-parameter because we have the expression 'I' in our language.
(ii) The parameter for a referential expression e is what e can refer to.
(iii) Assume that if you intend for x to be the parameter for e, then x is the parameter for e.
(iv) By (ii), (iii) entails: If you intend to refer to x with e, then you can refer to x with e.
(v) However, it is a fact that you can only intend to refer to x with e if meaning allows you to do so.
(vi) Therefore, (iii) is false.
Does that seem like a reasonable interpretation of Bach? One thing I'm worried about is that this way of laying out the argumet doesn't really mirror Bach's remarks about letting the intenions themselves be a parameter of contexts. Instead, (iii) interprets him as meaning that the person he is objecting to is proposing that we just let the intended referents be parameters. (Stefano Predelli has been pushing this line for years now.)
I guess one interpretation of what it would mean to let intentions themselves be a parameter is the following: One way of looking at indices (i.e. the n-tuples that formally represent contexts) is as constraints on variable assignments (given that we treat pronouns as variables). But if a parameter was the speaker's referential intentions, then the constraint on variable-assignments that indices are supposed to deliver would boil down to the instruction: Assign whatever the speaker intends. This is bad because we all believe in something like (v).
