This paper analyzes two price discovery processes: ordinary least squares learning from public information and a Bayesian learni ng made feasible by futures markets. The former tends to produce cobwe b behavior. In the latter, there is no cobweb, there is a faster convergence to Muth rational expectations, and the forecast errors are positively serially correlated. The evidence drawn from the Sydney Futures Exchange is consistent with the Bayesian learning process. Copyright 1992 by The Economic Society of Australia.
