Fictitious play and "gradient" learning are examined in the context of a large population where agents are repeatedly randomly matched. We show that the aggregation of this learning behaviour can be qualitatively di®erent from learning at the level of the individual. This aggregate dynamic belongs to the same class of simply de