Background: The assessment of physical activity for surveillance or population based studies is usually done with self-report questionnaires. However, bias in self-reported physical activity may be greater in lower educated than in higher educated populations. The aim of the present study is to describe educational differences in the validity of self-reported physical activity. Methods: We included 196 healthy adults (age 57 ± 15.4, of whom 17 % low, 24 % medium and 59 % high educated). Criterion validity of an adapted International Physical Activity Questionnaire was assessed against the ActiGraph GT3X+ accelerometer. Results: While criterion validity of self-reported physical activity was low to moderate in the total sample (Spearman rho ranged from 0.16 to 0.27, depending on the variables used), the validity in lower educated respondents was poor (-0.07 to 0.05). Conclusions: The results confirm the hypothesis that self-report physical activity questionnaires are less valid in lower educated populations.