An aftershock interevent time series, initiated on February 18, 1996, in the eastern Pyrenees was analyzed. The threshold detection magnitude was set at 1.9, and the series was assumed to be complete for an interval of 77 days. The original time series does not fit Omori's law, probably because of sudden changes in the rate of occurrence, interpreted as an increase in the production rate. When the recorded interevent time series is classified in terms of leading aftershocks (those that satisfy a relaxation process) and cascades (those occurred at a nearly constant rate), the new time series of the leading aftershocks fits Omori's law quite well, with p=0.94. Interpreted in terms of Dietrich's model, the series of leading aftershocks correctly predicts a return time for the main shock of the order of 50 years. To interpret the series of cascades, a minimalist, self-organized critical model was used. Although it is very simple, the model correctly reproduces the two-level structure in the observed time series, that is, the sequence of leading aftershocks and a cascade sequence emerging from each aftershock. This model may be given physical justification in terms of the Cochard and Madariaga <1996> nucleation model. ¿ 1997 American Geophysical Union