This study investigates empirical methods of generating prediction intervals for WASDE forecasts of corn, soybean, and wheat prices over the 1980/81 through 2006/07 marketing years. Empirical methods use historical forecast errors to estimate forecast error distributions, which are then used to predict confidence limits of forecasts. Five procedures were used to estimate empirical confidence limits, including histograms, kernel density estimation, logistic distribution, quantile regression, and quantile regression with stocks-to-use ratios. The procedures were compared based on out-of-sample performance, where the first 15 observations (1980/81- 1994/95) were used to generate confidence limits for the 16th year (1995/96); the first 16 observations were used to generate confidence limits for the 17th year (1996/97) and so on. Based on the results of accuracy tests for empirical confidence intervals over 1995/96 through 2006/07, all five empirical procedures included in this study generated confidence intervals that were not significantly different from the target confidence levels (80% pre-harvest and 90% post harvest). When monthly hit rates were averaged pre- and post-harvest across all three commodities, the kernel density-based method appears most accurate prior to harvest with an average hit rate of 82%, followed by the logistic distribution (76%), quantile regression-based methods (71-72%), and histogram (71%). After harvest, the kernel density-based method and the quantile regression-based method were the most accurate with average hit rates of 95%, followed by the logistic distribution based methods (92%), the histogram-based methods (89%), and the quantile regression methods with stocks/use ratio (88%). Overall, this study demonstrates that empirical approaches may be used to construct more accurate confidence intervals for WASDE corn, soybean, and wheat price forecasts.