While contingent valuation remains the only option available for measurement of total economic value of nonmarketed goods, the method has been criticized due to its hypothetical nature. We analyze field experimental data to evaluate two ex ante approaches to attenuating hypothetical bias, directly comparing value statements across four distinct referenda: hypothetical, "cheap talk," "consequential," and real. Our empirical evidence suggests two major findings: hypothetical responses are significantly different from real responses; and responses in the consequential and cheap talk treatments are statistically indistinguishable from real responses. We review the potential for each method to produce reliable results in the field.