Considerable evidence has accumulated which shows that the choice behavior of individuals exhibits systematic departures from expected utility maximization. The focus of the paper is to develop some measures of the extent to which utility maximization nevertheless remains a useful approximation. We do this by considering the extent to which individual choice behavior can be controlled, in the manner predicted by expected utility theory, by experimental designs which employ binary lottery payoffs in the manner of Roth and Malouf (1979) and Berg et al. (1986). The results of this study suggest that the gross features of risk preference can be reliably implemented, albeit with a non-negligible amount of error. Some errors were found to be systematic and can be attributed to subjects who did not know how to calculate the expected probbility of winning the prize in a compound binary lottery. The knowledge of compound lotteries also played the role in determinint which funcitonal forms are easier to induce.