Subsidence due to longwall coal mining may conveniently be considered as taking place in two phases: active subsidence as the face advances, and residual subsidence after it has come to a halt. Published data show that the dynamic subsidence profile at low face advance rates is independent of the advance rate itself. This finding is incompatible with the common practice of modelling subsidence as a simple viscoelastic process. Qualitative consideration of the mechanical processes taking place during active and residual subsidence shows that different timedependent behaviour during the two phases of subsidence is to be expected. We infer from reported observations of timedependent subsidence that the relaxation times involved during active subsidence are of the order of days, whereas the relaxation times involved during residual subsidence are of the order of one year. As different mechanical processes are involved during each phase, one should not attempt to predict the duration of residual subsidence from observations of active subsidence.