We challenge the recent claim that mispricing in the experimental asset markets introduced by Smith, Suchanek, and Williams (1988) is merely an artefact of confusion over declining fundamental value, and can be eliminated through appropriate training. We instead propose that when training is public knowledge, it reduces uncertainty over the behavior of others and facilitates the formation of common expectations. We disentangle the effect of training from the effect of its public knowledge, and find that when all subjects are trained to understand fundamental value, but this is not public knowledge, mispricing is as great as when training is absent.