A matching model is analysed in which firms imperfectly test workers prior to hiring them. If (some) firms hire only workers who pass the test, there is an informational externality; unemployment duration is a signal of productivity. In equilibrium, if it is profitable for a firm to test, it is also profitable for it to condition its hiring decision on duration, hiring those whose duration is less than a critical value. This testing equilibrium is inefficient, with too much testing and too low a critical duration value. Sensitivity analysis of the latter suggests explanations for the dependence of re-employment probabilities on duration and the instability of the U- V curve. © 1991 The Review of Economic Studies Limited.