Many subjects in early phase clinical trials expect to benefit in some way from the research intervention. It is understandable that people hope for improvement in their condition, no matter what the evidence. Yet unreasonable expectation of medical benefit may reflect problems with informed consent: Investigators may not disclose clearly that direct medical benefit from an early phase experimental intervention is unlikely or impossible, or subjects may not appreciate the differences between treatment and research. This paper presents findings from recent interviews with researchers and subjects and analysis of consent forms in early phase gene transfer research, a cutting-edge technology often called 'gene therapy'. We use three variables to construct a composite measure of therapeutic misconception TM, tapping misconceptions about the purposes of early phase research and the potential for direct medical benefit in these trials. Our multivariate model demonstrates the importance of both subject- and study-level factors as predictors of this TM index: education, disease type, and communication by study personnel about the likelihood of benefit. We hope that this work will deepen the discussion of how to define and measure TM, and refine the specification of factors that are related to subjects' TM.