Philosophy 730                                                                                                                                                                                 W. Lycan
Fall, 2007

On Marc's New Incompatibilist Argument

    I'll represent Marcs argument as follows.  CS different means, If my conative state were quite different from what it actually is.  A is my action.  S is glwv’s total efficient cause of A existing before I was born.  > is the subjunctive connective; I'm too lazy to find or construct a box-arrow.  (I also can't find a horseshoe.)
                                       (1)  CS different  >  S
                                       (2)  CS different  >  S
horseshoe A

                        \   (3)  CS differs  >  A  

  <>The conclusion (3) favors incompatibilism, though exactly what it says depends on some choices of interpretation.

A is supposed to be a future action of mine.  So (3)’s obvious counterfactual reading is, “If my CS were quite different from what it actually is, I would do A.”  Considered as an ordinary subjunctive, the latter seems truth-valueless; assuming only that my CS is quite different but saying nothing about any specific differences or even types of difference, we can (even under determinism) project nothing about my doing or my not doing A.  But in order to bring out the conflict between (3) and compatibilism, we must read it, rather, as a semifactual: “(Even) if my CS were quite different from what it actually is, I would (still) do A,” the idea being loosely that A would happen regardless of my CS and hence is out of my control.
            (Note that semifactuals differ logically from ordinary counterfactuals.  E.g., all known counterexamples to Contraposition for subjunctives are semifactuals.<1>)

What if we now take A to be a past action of mine instead of a future one?  Then it would read, “(Even) if my CS had been quite different from what it actually is, I would (still) have done A.”  This seems to pose the same threat to compatibilism, whatever threat that is, so apparently the tense doesn’t matter.

            Several issues ensue.


It’s a good thing for the argument that (3) needs to be read semifactually, because each of the premises is obviously semifactual too.  If (3) were an ordinary counterfactual while (1) and (2) were semifactuals, we could well doubt the argument’s validity on the grounds that a different similarity relation was mobilized as between premises and conclusion. 

But as things are, the argument is intuively valid.  And any standard similarity semantics would make the argument valid, so long as there was no other reason why the similarity relation might shift between the premises and the conclusion.  For simplicity, I assume Stalnaker semantics (Lewis’ semantics differs only in that he abandons the Limit Assumption, and I see no relevance of that to the free-will issue):  At the closest CS-differing world, S; at the closest CS-differing world, S É A; therefore, at the closest CS-differing world, A (duh).


Premise (1) says that even if my CS were quite different, S would still have obtained.  But under determinism, that is clearly false:  Had anything that is a partial cause of A not obtained, then the initial conditions that were the actual causes of that thing would have had to differ all the way back to the Big Bang.  (Though (2) remains true, since it holds by law of nature alone.)  Thus, the argument does not show that free will is incompatible with determinism, which was its primary objective.


What about (1) under indeterminism?  (1) is no longer clearly false, because the laws no longer logically force initial conditions to have been different given that CS was.  But I see no reason to accept (1), even so.  (1) would now seem either truth-valueless, because of the vagueness of its antecedent, or very probably false given the (always problematic) view that a “quasi-” or “near”-determinism holds “at the macro- level.”

(2) no longer holds by law of nature alone, for  S & ~A  is now logically compatible with the laws.  One may say that (2) is still very probable, given the macro-quasi-determinism assumption.  (Objection: That assumption would also apply to the relation between my actual CS and S.  If, however improbably, S would have been the same even if my CS had differed, can we then say that A would still have been the same?  Reply: Very probably, yes, because  S horseshoe A  is still overwhelmingly probabilified by the laws regardless of my CS.)

 Verdict:  Even assuming indeterminism, we still have no reason to grant (1).  As it stands, the argument fails.


1.  The distinction between semifactuals and ordinary counterfactuals was made by Goodman (immortal 1947).  But Stalnaker (ditto 1968) shrugged off the distinction in passing, saying that a uniform treatment should be preferred.  Lewis (ditto 1973) bought that, with barely a comment.  And Bennett's authoritative 2003 text (Chs. 10-13 of which we shall read in two weeks' time) doesn't even mention semifactuals.  Bennett should have at least mentioned them, because I'd pointed out in my 2001 book that semifactuals differ logically from ordinary counterfactuals.