This article focuses on two important issues faced by researchers working on choice experiments: how to find the optimal number of alternatives in each choice set and the number of choice sets in each survey. The authors first develop survey instruments with different numbers of choice sets and alternatives. Then, a heteroscedastic logit model is developed to relate the varying number of alternatives and choice sets to changes in the error term by parameterizing the scale factor of the heteroscedastic logit model. The authors study the effects of two simultaneous forms of complexity on the consistency of respondents' choices, i.e., the number of choice sets per questionnaire and the number of options per choice set. Their findings suggest that respondents' choices do vary with the amount of information given resulting in an optimal five options and six choice sets per survey. Results from the marginal effects and willingness to pay (WTP) estimates indicate that varying the number of alternatives and choice sets can also affect consumers' marginal WTP estimates. [EconLit citations: Q110; Q130; Q180]. (C) 2010 Wiley Periodicals, Inc.