The Pasat (Paced auditory serial addition task) is a test requiring addition of simple digits presented auditorily in several series of a successively higher pace of presentation. The task reflects the capacity for divided attention, is a measure of information processing speed and has appeared to be sensitive to minor attention deficits. We made a Dutch version characterized by longer time intervals between digits, for use in a study among (altogether 130) middle-aged and elderly persons. This modified version was evaluated with respect to feasibility. It appeared to be fairly feasible, provided sufficient opportunity for practice is given. Two series with inter-stimulus intervals of 4 and 3 seconds respectively were then evaluated with respect to validity and reliability in subjects of 48 to 74 years. Both Pasat series correlated well with two tests of attention: Digit Symbol (WAIS) and d2-test (r = 0.53-0.67), the relation with reaction decision speed being generally weak (r = 0.29-0.33) and weak to absent with a number of memory tests (r = -0.04-0.29). The Pasat performance shows a positive relationship with educational level and intelligence (inter-stimulus interval 4 seconds: r = 0.28 and 0.33 resp.; inter-stimulus interval 3 seconds: r = 0.40, 0.57 resp.). The stability of test performance over a period of about two months is moderate (r = 0.54, 0.61 resp.). In view of the excellent inter-observer reliability (r = 0.99) and the good parallel-test reliability (r = 0.79, 0.68 resp.), this might be explained by fluctuations in the capacity measured, rather than measurement error.(ABSTRACT TRUNCATED AT 250 WORDS)