In patients with chronic diseases education should improve knowledge about the disease and increase certainty in knowledge. We present here a technique to measure changes in certainty after an educational intervention. For this purpose, before and after a course, patients answer a questionnaire in which answers are accompanied by an estimate of the degree of certainty. Answers are then assigned to areas of knowledge defined a priori: mastered (certainty > or = 90%, correctness > or = 90%), hazardous (certainty > or = 90%, correctness < or = 50%), uncertain (certainty < or = 50%, correctness > or = 90%) and residual. Finally differences in the distribution of answers among different areas are analysed statistically. Using this technique in a group of patients with type I diabetes who followed a course on insulin use, we found significant changes in the distribution of answers among different areas of knowledge. Thus changes in certainty can be analysed quantitatively and used to evaluate better the effect of therapeutic education.