The focus of this study was on the visual analysis of time series data for evaluating educational programs. Two characteristics of the data--changes in slope and variability--and two characteristics of evaluation--training in data utilization and the use of aimlines/decision rules--were manipulated. A total of 51 students and/or teachers in education evaluated a set of 28 graphs on two dimensions: (1) Was the program depicted on the graph an effective program? and (2) What about the data supported such a conclusion? Findings of the study indicated that visual analysis is not very reliable for evaluating educational programs, and is influenced considerably by the characteristics of the data array (specifically, slope and variability). Training in data utilization or the use of aimlines did not appear to be particularly powerful procedures for improving visual analysis. At the same time, the findings indicated evaluation consistent with established data analysis paradigms. Implications for training in visual analysis are discussed. (Author)