A dynamic random effects probit model is estimated on the first six waves of the German Socio-Economic Panel to test for state dependence effects in male unemployment behaviour. Estimation of the model is based on the marginal likelihood approach. In the model an individual's unemployment probability at a given point in time within the period 1985 - 1989 depends on his labour force status in the previous period and on the cumulated duration of past unemployment. Controlling for observed and unobserved population heterogeneity, we show that there are strong state dependence effects in individual unemployment dynamics with respect to both the incidence and the duration of an individual's past unemployment. These results are compatible with the 'scar theory' of unemployment which holds that an individual's previous unemployment experience may have long-term effects because it induces a depreciation of human capital and/or acts as a screening device in employers hiring decisions.