This paper considers a set of research and assessment strategies used to evaluate programme outcomes in the Stanford Teacher Education Programme (STEP) during a period of programme redesign over 10 years. These strategies include surveys and interviews of students' perceptions of programme elements and their own preparedness, observations of their practice during and after teacher education, evaluations of their practice on a structured portfolio of practice (the Performance Assessment for California Teachers (PACT)) and analyses of the effects of a sample of graduates of STEP and other programmes on student outcomes, including value-added measures. While the studies were able to ascertain that the students of STEP graduates had strong value-added learning gains, the paper concludes that the use of student learning data alone as a measure of teacher effectiveness does not help guide decisions related to programme improvement, and a range of approaches is required. In addition it suggests that there will be continuing concerns about the narrowness of the learning measured by standardised tests, and about the many challenges of collecting and analysing such data in ways that overcome the technical and practical problems associated with their use. (Contains 3 figures and 1 note.)