This paper discusses alternative criteria for defining the choice set in the analysis of recreational site choice decisions. Four definitions are considered: 1) all available sites compiled by the researcher, 2) sites known by individuals, 3) sites previously visited by individuals, and 4) sites grouped by distance from individual residence. A survey of randomly sampled visitors was conducted at 19 wetland sites in Taiwan. The recreational trip choices to 19 wetland sites are analyzed using the conditional logit (CL) and mixed logit (MXL) models under each of the four choice set definitions. The estimation results show that regardless of choice model, both parameter estimates and welfare measures are significantly affected by the definition of choice sets and in general a smaller measure of welfare results from a larger choice set. Despite the relaxation of the independence of irrelevant alternative (I.I.A.) hypothesis and individual specific parameter estimates, the flexible mixed logit model does not alleviate the impact of choice set definition on welfare measures.