A total of 262 women in the USA (161 breast cancer survivors and 101 controls) were exposed to a video vignette using modeling in which a physician discussed the concept of a clinical trial (CT) with a woman who was in the process of making a treatment decision. A pretest-post-test design was used and improvements in clinical trial knowledge and beliefs were assessed. Results indicate that video modeling is a powerful tool for increasing CT knowledge (pretest mean=41.5% correct, post-test mean=77.5% correct) but not for improving CT beliefs. Increased clinical trial knowledge, as measured by change scores, was associated with white race, lower levels of education and pretest breast cancer knowledge, more negative pretest CT beliefs, and a higher estimate of the lifetime probability that a woman will have breast cancer. When pretest CT knowledge was added to the analysis using hierarchical multiple regression, all variables except white race became nonsignificant; an increase in CT knowledge was associated with having lower pretest CT knowledge. Results indicate that the effects of low education, low breast cancer knowledge, and biased probability assessment were mediated through the pretest score. An increase in post-test positive CT beliefs was associated with older age, thinking about breast cancer less often, and having lower pretest CT knowledge in the total sample. When pretest CT beliefs was added to the analysis using hierarchical multiple regression, all other variables became nonsignificant; an increase in CT beliefs was associated with having lower pretest CT beliefs, again indicating mediation of the effects of other variables.