Certain correlations between characteristic events and smaller events are studied in the context of a spatially homogeneous, mechanical model for a fault. The model satisfies a one-dimensional quasi-linear wave equation in which the key nonlinearity is associated with the slipstick friction force at the interface between tectonic plates. We find that the model gives rise to slipping events of all sizes; the smaller events are consistent with the Gutenberg-Richter statistical law, whereas the characteristic events have a separate statistical distribution. The crossover from smaller events to characteristic events coincides with a correlation length which sets the scale for clustering of smaller events before the characteristic rupture. Numerical results and scaling arguments relate the features of the distribution of characteristic events to the corresponding features of the distribution of smaller events as well as the length of the fault. The distribution of time intervals between characteristic events has the distinguishing features of asymmetry and a finite upper cutoff. The mean T and standard deviation &sgr; of the time interval distribution are found to be related to features of the magnitude versus frequency distribution of smaller events, and the coefficient of variation &sgr;/T is independent of the driving rate. Because this analysis relates properties of characteristic earthquakes to statistical features of smaller, more numerous events as well as geometric features of the fault, it may ultimately be useful for long-term earthquake prediction. ¿1991 American Geophysical Union