Evidence that gamma-ray burst sources repeat

We investigate clustering in the angular distribution of the 260 γ-ray bursts in the publicly available BATSE catalog, using a nearest neighbor analysis and the measures of burst brightness B and short-time-scale variability V that we introduced earlier. We find that, while all 260 bursts are only modestly clustered $$(Q-\text {value} = 1.8 \times {10}^{-2})$$, the 202 bursts in this sample for which the statistical error in location is $$\lt 9^\circ$$ are significantly clustered on an angular scale $$\approx 5^{\circ} (Q- \text {value} = 2.5 {\times } {10}^{-4}$$, taking into account having chosen the cut-off in the statistical error). We also find a significant correlation between bright type I bursts and faint type I and II bursts on an angular scale $$\approx {5}^{\circ} (Q- \text {value} = 4.0{\times } {10}^{-3})$$. This angular scale is smaller than the typical (statistical plus systematic) error in burst location of $$6.^{\circ}8$$, suggesting multiple recurrences from individual sources. We conclude that ‘classical’ γ-ray burst sources repeat on a time-scale of months, and that many faint type I and II bursts come from the sources of bright type I bursts.