Aumann has proved that common knowledge of substantive rationality implies
the backward induction solution in games of perfect information. StaInaker
has proved that it does not, Roughly speaking, a player is substantively ra
tional if for all vertices v, if the player were to reach vertex v, then he
would be rational at vertex v. It is shown here that the key difference be
tween Aumann and Stalnaker lies in how they interpret this counterfactual.
A formal model is presented that captures this difference, in which both Au
mann's result and Stalnaker's result are true (under appropriate assumption
s). (C) 2001 Academic Press.