This paper presents an experimental investigation of optimal learning in re
peated coordination games. We find evidence for such learning when we limit
both the cognitive demands on players and the information available to the
m. We also find that uniqueness of the optimal strategy is no guarantee tha
t it will be used. Optimal learning can be impeded by both irrelevant infor
mation and the complexity of the coordination task. Classification Numbers:
C72, C92. 2000 Academic Press.