Understanding and accommodating heterogeneity in variance and taste has become a major area of research within discrete choice analysis. This paper reviews and compares di erent approaches for specifying scale and preference heterogeneity. The scale factor can be specified as continuous or discrete, the latter can be associated with a parameter (i.e. scale derived deterministically) or it can be derived probabilistically. Unobserved preference heterogeneity can be represented by a continuous function, a discrete mixture or using a combination of both. This paper uses data from two recreational site choice studies (one based on stated preference elicitation methods and another based on revealed preference methods) to compare various model specifications for accommodating both types of heterogeneity. Results show that model fit and welfare estimates are sensitive to the manner in which scale and preference heterogeneity is addressed.