This paper implements an algorithm to infer. from the implied volatility tree, the risk-neutral density that is comparable to the one implicit in the kernel-regression volatility surface on which the tree is constructed. The equality between these two densities becomes an empirical issue on the validity of the tree model in option pricing. Three statistical tests, out-of-sample fit and hedging effectiveness are investigated for the S&P 500 index options from January 1990 to December 1995. The matches between these two densities for 30-day and 100-day options are found, while the inaccuracy for 300-day options is remarkably stable for different periods.