We apply Camerer and Ho's experience-weighted attraction (EWA) model of lea
rning to extensive-form signaling games. Since these games often have many
equilibria, logical 'refinements' have been used to predict which equilibri
um will occur. Brandts and Holt conjectured that belief formation could lea
d to less refined equilibria, and confirmed their conjecture experimentally
. Our adaptation of EWA to signaling games includes a formalization of the
Brandts-Holt belief formation idea as a special case. We find that the Bran
dts-Holt dynamic captures the direction of switching from one strategy to a
nother, but does not capture the rate at which switching occurs. EWA does b
etter at predicting the rate of switching land also forecasts better than r
einforcement models). Extensions of EWA which update unchosen signals by di
fferent functions of the set of unobserved foregone payoffs further improve
predictive accuracy.