I have recently encountered, in several quite different contexts, the idea that all modal thought experiments are a priori. In particular, the claim shows up in defense of the idea that genuine philosophy is entirely a priori. (Aside: I have never met anyone with the list inkling of the history of philosophy who embraces the latter take on philosophy’s methodology. This is not a coincidence.) Or anyway, I seem to have encountered it: the idea is so outlandish that I suspect I must be misunderstanding.
Consider an example. Suppose I make a fist-sized snowball. To make the situation not just counterfactual but outside ordinary experience, suppose I go through a ceremony and dub the snowball ‘Mrs. Roberts’. Now the thought experiment: What would happen if I put Mrs. Roberts into a 450 degree F oven on a cookie sheet for an hour? We all know the answer… but our knowledge is patently not a priori. Both in terms of cause and justification, it rests on: knowing the enormous causal powers of a hot oven vis-a-vis a smallish frozen body; knowing the exceedingly negligible causal powers of a pointless dubbing ceremony.
A related mistake is to suppose that, because a conclusion is arrived at by “thinking about it in my armchair”, it’s therefore a priori. To the contrary, insofar as one’s armchair philosophizing is based in memory of relevant experiences, its deliverances cannot be justified a priori. And much of our thinking about modal properties is so based: if asked to reflect in your armchair about whether a rubber ball will still bounce, even if a tiny pink dot is painted onto it, you know the answer because of experience.
An interesting connection here is the idea, found in Jerry Katz, that linguistic semantics and syntax are not empirical because their usual data are intuitions about well-formedness, entailment relations, etc. We have here another confusion. If those linguistic intuitions in their turn trace (partly) to experience, and to discovery of form/meaning/context patterns, then the fact that we can now arrive at a conclusion without additional observation isn’t probative at all — because we are relying on memory. (To drive the point home, were it really a priori whether ‘Yo quiero comprar una naranja’ is grammatical, and whether it entails the existence of oranges and/or mental states, then one ought to be able to figure it out with no experience of Spanish. QED.)