forecasting futures prices is an integral component of a profitable futures trading strategy. Current research suggests the standard random walk assumption of futures prices may actually be only a veil of randomness that shrouds a noisy nonlinear process [see, for example, Savit (1990) and Tvede (1992) for an explanation of the apparent randomness in these price time series, and Frank and Stenzos (1989), Blank (1991), and DeCoster et al. (1992) for evidence of the existence of nonlinear structures of rates of return]. What is called for to remove this veil and thereby make futures price forecasting possible is the application of nonlinear models. Further, some believe that nonlinear prediction can be enhanced by incorporating information about the sentiments of major trading groups of these futures [e.g., Leuthold et al. (1989) and Turner et al. (1992)l. Accordingly, this research examines the feasibility of employing neural networks to forecast price changes of Standard Poor’s 500 Stock Index and gold futures based on past price changes’ and historical open interest patterns that are held to represent the beliefs of a majority of