This study examined multidimensional item response theory (MIRT) modeling to assess social desirability (SocD) influences on self-reported physical activity self-efficacy (PASE) and fruit and vegetable self-efficacy (FVSE). The observed sample included 473 Houston-area adolescent males (10-14 years). SocD (nine items), PASE (19 items) and FVSE (21 items) were measured with previously validated self-report instruments containing Likert-type responses. Physical activity was objectively measured using the Computer Science Application Incorporated/Manufacturing Technology Incorporated (CSA/MTI) accelerometer. Total fruit, juice and vegetable consumption was measured with a food frequency questionnaire. Correlations between self-efficacy and behaviors were minimal, regardless of controlling for SocD. However, in a simulated sample derived to demonstrate the utility of MIRT when relationships exist, the pseudo-relationships between self-efficacy and behaviors were substantially weaker after controlling for SocD. MIRT provided disattenuated correlations between SocD and self-efficacy, thereby providing more precise estimates of the real influence of SocD on the relationship between self-efficacy and behavior. However, as shown in the observed sample, more research is needed to understand the influence of SocD on the relationship between self-efficacy and behaviors for different populations and for different degrees of SocD response bias. (Contains 4 figures and 1 table.)