Non-cognitive skills are routinely measured using self-reports in the context of large-scale international assessments. However questions remain on the adequacy of self-reports to conduct comparisons. Measures that exploit test-taker's behaviour during the completion of questionnaires or of the cognitive tests have been proposed in the literature but little is known about their properties. We evaluate these measures using PISA data. High correlation with self-reported questionnaires and with test performance suggest that behavioural indicators measure what they are supposed to measure. Differences by gender and socio-economic status are in line with expectations. However, some measures are sensitive to the design of the test and questionnaire, and some predict better than others life outcomes of students ten years after the PISA test.