The investigation presented here explores the hypothesis that participants are less likely to respond in a socially desirable fashion on self-report questionnaires completed on the Web relative to those completed in the laboratory--the candor hypothesis. A battery of social desirability questionnaires (i.e., Balanced Inventory of Desirable Responding [Paulhaus, 1984], Marlowe-Crowne Social Desirability Scale [Crowne & Marlowe, 1964], Eysenck Personality Questionnaire-Revised Lie Scale [Eysenck & Eysenck, 1994]) was administered to 3 groups: 2 groups consisted of undergraduate participants who were randomly assigned to complete the measures either in the laboratory (n = 60) or on the Web (n = 60), and 1 group consisted of self-selected participants who visited our experimental Web page and completed the measures online (n = 284). This design allowed us to assess the role of Web administration while controlling for differences in sample type, an oft-neglected issue in the Web literature. Results do not support the claim that administering self-report measures over the Web results in a decrease in socially desirable responding. Furthermore, these findings highlight the problems associated with confounding sample and medium. Implications for the use of Web as a research tool are discussed.