Summary. Analysis of self-assessments may yield biased results when respondents use the response categories of survey questions in different ways. An anchoring vignette is a short description of relevant aspects of a hypothetical person's life and can be used to correct self-assessments for heterogeneity in reporting behaviour. The paper studies the validity of the parametric model for the anchoring vignette method. Rather than testing separately the assumptions of this model, we take the following two approaches. First, we study whether different vignettes within a certain domain lead to similar adjusted self-assessments. Second, we study whether different adjusted self-assessments, each adjusted by using a single vignette, are closer to a measure of the actual situation than the unadjusted self-assessments. We use data from the 2004 and 2007 waves of the Survey of Health, Ageing and Retirement in Europe for three domains of health: cognition, breathing and mobility. Our results indicate that the anchoring vignette method is sensitive to the choice of the vignette for cognition and breathing, whereas the results for mobility are more encouraging.