agreement with 'yes'

[MalFet]

It is the problem.

Ok, I agree. But isn't it a natural one?

I'm not sure I understand the question. Why would treating yes/no as proforms be "natural"?

[Daniel]

They represent the sentence in some contexts and not others. Beyond that, just embedding them in general is odd syntactically, so I don't see a major additional step on working out the semantics including the varied contexts.

[MalFet]

They represent the sentence in some contexts and not others. Beyond that, just embedding them in general is odd syntactically, so I don't see a major additional step on working out the semantics including the varied contexts.

If you have some way to solve the type-level problems that come with treating yes/no as pro-forms, I'd encourage you to publish it. It will probably get you tenure.

[Daniel]

Avoid them? There appear to be embedding restrictions. Simple (though certainly not worthy of tenure, if that does indeed "solve" it).

[MalFet]

Avoid them? There appear to be embedding restrictions. Simple (though certainly not worthy of tenure, if that does indeed "solve" it).

Like I said, most of the work in formal semantics is very committed to the idea that pro-forms should not have distributional constraints substantially different than those of their antecedents. This goes all the way back to Frege and is tied in with some very important Big Ideas™.

If you don't feel bound by that notion, that's certainly your prerogative, but you should realize that you're scrapping pretty much everything that makes the notion of anaphora meaningful to most people. If that's what you're doing, why use the concept pro-forms at all? Calling this plain old deixis would be equally descriptive and far less misleading.

[Daniel]

I'm not convinced that your sentence is grammatical in the first place. That isn't about semantics.

[MalFet]

Of course it's not grammatical! That's the entire point!

[Daniel]

Binding principle D: sentential pro-forms may not be embedded.

This is true of lots of things, anyway, unrelated to semantics:

?It's true that the more you know the better you do on the test.
?It's true that drink one more beer and I'm leaving!

[MalFet]

I'm just repeating myself now, which is always a good sign that it's time to let go.

Formal semantics is, in general, very attached to the principle that pro-forms and antecedents should not have substantially different distributional constraints. You seem not to be. That's fine. I'm not here to sell you on analytic cognitivism. But, you asked for clarification on why treating yes/no as a pro-form is widely understood to be problematic, and the answer is that it leads to significant problems in type-level representations. Most intro to semantics courses cover this precisely because it's so basic to the discipline. To that end, the embedding problem is one among many. Others include polarity in negative questions ("You're not coming?" "No, I am./No, I'm not") and anaphora reassignment ("Are you Bill?" "Yes"[≠You are Bill]).

In other words, you can't have your cake and eat it, too. You can't scrap the principles of reference that undergird anaphora in formal semantics but then expect people to know what you mean when you call something a "pro-sentence". I'm sure it all feels intuitive, but the purpose of this kind of theory is to figure out which intuitions are sound and which are slights of hand. If you're looking for a good exposition on this — one that clarifies why anaphora is understood in such rigid terms — the argument between Gareth Evans and Howard Lasnik is a great place to start.

[Daniel]

I'd simply see a distinction between distribution and interpretation. But I see what you mean.

[MalFet]

I'd simply see a distinction between distribution and interpretation.

That's why I keep suggesting that you call this "deixis". Deixis (generally) provides a wide berth for interpretation. Anaphora (specifically) does not. This is not casual terminology. Pronominal reference is precisely defined because it needs to be in order for modern semantics to work. That's a big deal, and there's a lot more at stake in that than simply linguistics.

[jkpate]

I'd simply see a distinction between distribution and interpretation.

That's why I keep suggesting that you call this "deixis". Deixis (generally) provides a wide berth for interpretation. Anaphora (specifically) does not. This is not casual terminology. Pronominal reference is precisely defined because it needs to be in order for modern semantics to work. That's a big deal, and there's a lot more at stake in that than simply linguistics.

What else is at stake? Are you talking about proforms as a window into how more general cognition is structured?

[MalFet]

What else is at stake? Are you talking about proforms as a window into how more general cognition is structured?

Either that or it's opposite, depending on which direction you want to come at it from.

If we're going to treat human language as something bearing a formal semantics (as most-but-certainly-not-all contemporary linguists are inclined to do), the idea that pronominal reference plays by the same rules as variables in predicate logic is a big-big deal. If we start breaking those rules, we're either (a) drastically curtailing the descriptive power of semantic analyses, or (b) running up against some very fundamental principles in formal logic.

Either way, it stirs up a lot of (probably very unintentional) dust.