We discuss the possibility of colour rearrangement in events, i.e.\ that the original colour singlets and may be transmuted, for instance, into new singlets and . The effects on event properties could be quite large if such a rearrangement would occur instantaneously, so that gluon emission would be restricted to each of the new singlets separately. We argue that such a scenario is unlikely for two reasons. Firstly, the and usually decay at separate times after the production, which leads to large relative phases for energetic radiation off the two constituents of a rearranged system, and a corresponding dampening of the QCD cascades. Secondly, within the perturbative scenario the colour transmutation appears only in order and is colour-suppressed. Colour reconnection at longer time scales is quite feasible, however, and may affect the fragmentation phase. If so, the nature of non-perturbative QCD can be probed in a new way. We formulate several alternative toy models and use these to estimate the colour reconnection probability as a function of the event kinematics. Possible consequences for LEP~2 events are illustrated, with special attention to systematic errors in mass determinations.