We examine the effects of parameter uncertainty and Bayesian learning on equilibrium asset prices when all the structural parameters of the aggregate consumption and dividend growth rate processes are unknown. With realistic calibration of a parsimonious set of prior parameters, the model generates a sizeable equity premium and a low risk-free rate even with a power utility function, low risk aversion, and absence of persistence in growth rates. Raising the prior uncertainty on consumption growth induces a "flight to safety" that results in lower risk-free rates, higher equity premium, and greater risk-free rate volatility. But raising the prior uncertainty on dividend growth rates has ambiguous effects on the equity premium. Learning does not produce a monotonically declining equity premium and shocks to growth rates can induce sharp fluctuations in the market returns even after one hundred years of data. Finally, the effects of persistence in growth rates on the market returns are highly sensitive to the prior estimates of persistence.