OBJECTIVE: The contingent valuation (CV) method is increasingly employed within the health care sector (HCS); however, the extent to which hypothetical choices mimic real economic choices remains unclear. A few previous experiments have studied the correspondence between hypothetical and real willingness to pay (WTP) within the HCS. The findings suggest that hypothetical bias (HB) (i.e. overestimation of real WTP) exists in the HCS, just as it does outside of the HCS. An important area of research is whether or not HB can be "calibrated" so that hypothetical choices more closely mimic real economic choices. We conducted an experiment directly comparing responses to a dichotomous choice CV question with real purchase decisions using a pharmacist provided diabetes management service as the item valued. We examine whether HB exists and we evaluate the usefulness of two HB mitigation techniques: the 'certainty approach' and the 'cheap-talk approach'. METHODS: A CV survey using 267 subjects with diabetes recruited from 9 Kentucky community pharmacies was conducted. Subjects fell into one of three groups: 1) hypothetical offer followed with certainty calibration; 2) hypothetical offer preceded by cheap talk; 3) real offer to purchase the service. All surveys were face-to-face with a trained interviewer. RESULTS: Before calibration, 45% of subjects in the hypothetical group stated that they would purchase the service compared to 26% of subjects in the real group (p = 0.006). After calibration with the 'certainty approach' the difference between the calibrated (24%) and real response rate was no longer significant (p = 0.83). The cheap talk group had a purchase rate of 45%, which was significantly different from the real group (p = 0.006). CONCLUSIONS: This experiment confirms the existence of HB in CV applications within the HCS. Results suggest that a 'certainty-approach' calibration technique is successful in mitigating HB, while 'cheap talk' script is not.