"Willingness to Pay" studies have made significant in-roads into public decision-making, especially on environmental issues, as a means of estimating a dollar value for goods or services when no direct market value can be established. Policymakers in the U.S. and elsewhere use the resulting value estimates to guide their decisions on topics like water quality protection, air quality improvement, watershed and ecosystem protection, and reduced human health risk. Recently, the U.S. Environmental Protection Agency (EPA) has sought to determine whether a recruited Internet panel of respondents can be used reliably for that purpose. This presentation will report the results of a national, EPA-funded mode test comparing an Internet panel survey (conducted through Knowledge Networks, n=1162) on air quality in national parks with virtually identical versions of the same questionnaire administered by mail (n=904) and by phone (n=1273). Principles of universal questionnaire design were used along with extensive pre-testing so as to focus on mode effects. For the same reason, sampling frame was held as nearly constant as possible by using national Random Digit Dialing samples with reverse-lookup of addresses for all three surveys. Weighting and matching the respondents did not eliminate significant demographic, behavioral, and attitudinal differences across modes. Mode differences found when modeling Willingness to Pay with appropriate statistical controls seem to be best interpreted as resulting from the effects of social desirability on survey responses.