A great issue of concern in valuation studies is whether respondents provide trustworthy and reliable answers conditional on the perceived information. Respondent may report either a higher than the true Willingness-To-Pay (WTP) due to warm glow or embedding effects or zero WTP which is lower than the true WTP due to protest behavior. We conduct a contingent valuation study to estimate the WTP for conserving a Natura 2000 wetland area in Greece. We find that 54% of the positive bidders exert warm glow motivations while 29% of all responses can be classified as protest zero bids. We employ three different models to test for the potential impacts of how these positive warm glow and protest zero bidders are treated. We first exclude the warm glow cases, secondly we include them, and, finally, we correct for selection bias by using the Full Information Maximum Likelihood method for grouped data model. Our findings show that removal of warm glow positive bidders does not distort the WTP estimate in any significant way. However, using the same approach for protest zero bidders, we find strong evidence of selection bias associated with removal of protest zero responses. Specifically, WTP estimates obtained after removal of protest responses are found to be biased downwards and the aggregated welfare measures would be significantly underestimated in our case. These results suggest that there could be serious consequences associated with the common approach of removing protest zero bidders in CVM.