We apply statistical analysis based on pairs of earthquakes to study spatial and temporal patterns of earthquake distribution in the aftershock zone of the 1984 Morgan Hill, California, earthquake (M=6.2). We examine 6 years of preshock and 23 months of aftershock data, using frequency distibutions of earthquake pairs. The nonrandomness in the observed spatial and temporal distribution of pairs is studied by comparison with simulated expected distributions. The significance of departures from the expected distributions is appraised by using tolerance limits that account for the scatter of the single trial expected frequencies. We find that the earthquake spatial distribution is nonuniform in two major ways. (1) Short-distance clustering (distances less than 7 km) occurs in both the preshock and aftershock period. It is more pronounced and covers larger zones during the aftershock sequence. (2) A long-distance nonuniformity is also observed (distances greater than 11 km) caused by concentration of events in two areas separated by a gap in earthquake activity. This feature is present in the preshock and early aftershock period but disappears in the late aftershock sequence. The distribution of time intervals between preshocks does not indicate any substantial nonrandom features. In contrast, the temporal distribution of aftershocks, although dominated by the classic power law decay in time, suggests the possible existence of characteristic time intervals that increase with magnitude and time after the mainshock. ¿ American Geophysical Union 1988