The transport of passive, dissolved and conservative matter is calculated with a three-dimensional Eulerian transport model in order to estimate the water exchange times for the ICES-Boxes. Daily flow fields calculated with a baroclinic circulation model (Pohlmann [1991]) are used to drive the transport model. The half-life time of water exchange in a box is defined analogously to the half-life time of radioactive substances. To determine the half-life time for every ICES-Box, the water in this specific box is marked by a constant concentration, whereas the concentration outside the box is set to zero. The calculation stops when the concentration of matter in the box reaches 50% of its initial value. In the classical approach the total exchange of water in a box is characterized by the time that is needed for the total box mass to flow through the open boundaries of the respective box (flushing-time approach). The advantage of the half-life time approach introduced in this study is the fact that it takes into account the structure of the underlying flow field as well as horizontal and vertical diffusion processes. Moreover, it is possible that matter leaves a box and returns, following a change in the flow field. In such a situation the concentration in the box may increase, inducing a longer exchange time compared to the flushing-time. Depending on the starting time there are significant inter-annual deviations between the half-life times for individual boxes. In comparison with the classical flushing-time approach used by Davies [1982], Backhaus [1984], Lenhart [1990] and Lenhart & Pohlmann [1996], the results of the method presented here show that in boxes with a predominantly inhomogeneous flow field (i.e. the ICES-Boxes 3a, 4, 5a, (6b), 7a and 7b) the time needed to reduce the concentration of a contaminant is longer than was to be expected from earlier studies. This is due to the fact that in these boxes the flushing depends not only on the strength of the flow but also on the structure of the flow field, which is not taken into account in the classical flushing-time approach.