Against NTV for Subjunctives
(1) The “where” and “when” parallel: If we maintain that a conditional like (1a) lacks truth-value even though the corresponding temporal sentence (1b) has one, we shall have to attribute that vast semantic difference solely to the lexical difference between “if” and “when.”<> (1) a. I would leave if you did.>
b. I would leave when you did.<>(Also “where”: “I would live where you did.”) NTV advocates are going to have to explain how it is that those two words could have virtually the same syntax but differ so drastically in their semantic treatment.>
Moreover and more to the point, “if” and “when” exhibit dramatic semantic similarities as well, especially in regard to modification, pronominalization, and relativization (the second and third of these both requiring semantical coreference). It would be amazing if (1a) and (1b) were semantically as similar as they are, and yet (1a) did not even have a truth condition at all. And consider the locution “if and when,” as in “If and when she were to submit a paper, we’d read it within a month”; does NTV award that sentence a truth-value, or not? Further, NTV would have to be extended to “unless” sentences (cf. “unless and until”) and to all variants of the “in the event that” construction (“I’d leave in the event that she did,” “In that case I’d leave too,” etc.), as well as “on the condition that” and the like.
The crass unlinguisticness of NTV becomes all the more evident when one applies it to the numerous languages that employ the same word for our “if” and “when.” (Hittite, Swahili, Tagalog, Mandarin.) NTV would have it that a temporal/conditional sentence of such a language has a truth condition when understood temporally, but entirely lacks one when understood conditionally. Yet speakers of such languages usually do not trouble to notice the distinction, treating “when” as simply a factive variant of “if;” the distinction is virtually pragmatic, resolved automatically by context without anyone’s noticing. If you tried to tell a linguist who is a native speaker of such a language that actually the ambiguity is very dramatic, not just semantic but hyper- or metasemantic in that one of the two readings has an entirely different kind of meaning from the other’s, the linguist would think you were crazy.
(2) Embedding: “If this vase would break if I dropped it on the driveway, then I’d be careful not to drop it on the driveway” asserts, on the condition that the vase would break if s/he dropped it on the driveway, that s/he would be careful not to drop the vase on the driveway. “If it were true that you’d fail unless you studied hard, I’d suggest you hit the books.” And, paraphrased from Edgington (1995), “If John were to be punished if he had taken the money, then Mary would be punished if she had taken the money.” But according to NTV, the “conditions” specified are not matters of fact, and so these antecedents cannot be true. (So much the worse for conditional-assertion and suppositional theories.) We would need a 3-valued logic to project assertibility values.
Maybe that’s no big problem for NTV views that are not also suppositional. But notice also that subjunctives embed in propositional-attitude ascriptions. “Angie believes that Bob believes that Cindy would attend if she did, and Dave dislikes Cindy so much that he thinks if it’s true that she would attend if Angie did, he will try to persuade Angie that if Cindy does attend, Angie should try to convince her that if she were to get anywhere near him she’d catch something”; “Dave is afraid that if Cindy were to look at him his left cheek would tic visibly”; “Angie is embarrassed that if Dave were to see her he would send her what he thinks are subtle secret signals”; “Bob hopes that if Cindy and Dave had both attended, no one would have sung ‘You Must Have Been a Beautiful Baby’”; “Cindy is sad that Bob would have left the room if she had sung anything at all”; “Bob is ashamed that he would run crying from the room if someone were to sing ‘You Must Have Been a Beautiful Baby’.” I do not see how such attitudes can be explicated in terms of conditional probability without reference to truth.
(3) Nomologicals: An NTV defender is forced to deny that indicative conditionals follow logically from the corresponding nomologicals. Thus (2a), a nomological truth, cannot entail (2b).
(2) a. Every piece of iron heated to 200°C turns red,
b. If this piece of iron had been heated to 200°C it would have turned red.
Also, the falsity of the corresponding material conditional would strangely fail to make the subjunctive conditional false.
(4) Finally, it weighs with me that Stalnaker-Lewis semantics has seemed right to nearly everyone (though of course not to Hájek). There’s been no such agreement on semantics for indicatives. (The latter lack of agreement is perhaps a point in favor of NTV for indicatives.) Why, then, the agreement regarding subjunctives, if S-L semantics is not only wrong but radically misguided in that it assigns counterfactuals the wrong kind of meaning?