A new turn in the research agenda of environmental valuation is under way. Rather than treating stated preference (SP) and revealed preference (RP) as competing valuation techniques, analysts have begun to view them as complementary, where the strengths of each approach can be used to provide more precise and possibly more accurate benefit estimates. In this paper, we reexamine the models and motives for combining revealed and stated preference data. First, we note that because the different kinds of SP data contain different amounts of information, they may indicate different degrees of consistency with RP data. We also reconsider the interpretation of "consistent" or "inconsistent" findings of RP and SP data. We argue that while the conventional approach of treating the RP data as true and testing whether the SP data is consistent with it is intuitively appealing, this approach is based on the tenuous premise that the RP data generates unbiased welfare estimates. In particular, we propose three hypotheses for why the two data sources might be exhibiting inconsistency: (1) SP respondents ignore their budget constraint, (2) analysts inaccurately measure the price of recreation in RP data, and (3) SP respondents do not accurately understand the contingent market proposed by the analyst. Using these hypotheses in conjunction with the jointly estimated models, we test for the presence of these effects (using the alternative model as the basis of comparison).