The Forward Premium Puzzle is widely considered to indicate inefficiency in the foreign exchange market. This paper proposes a resolution of the puzzle using recursive least squares learning. Risk-neutral agents learn the unknown parameters, underlying the exchange rate generation process, using constant-gain recursive least squares. Simulations using plausible model parameter values replicate the anomaly along with other observed empirical features of the forward and spot exchange rate data. Estimates of parameter values from data support the model assumptions and justify the use of constant-gain learning. The conclusion is that the puzzle is not necessarily a reflection of inefficiency.