Learning in experimental 2×2 games

► Two new learning models (impulse-matching learning and action-sampling learning) are introduced. ► The aggregate behavior of experimental subjects is best described in simulations with impulse-matching and action-sampling learning. ► Individual round-by-round behavior is best described by self-tuning EWA. ► Only impulse-matching learning and self-tuning EWA perform better in describing subjectsʼ round-by-round behavior than the aggregated empirical frequencies.