A cautionary note on gamma-ray burst nearest neighbour statistics

In this Letter we explore the suggestion of Quashnock & Lamb that nearest neighbour correlations among gamma-ray burst positions indicate the possibility of burst repetitions within various burst subclasses. We compare the sensitivity of these results to both the definitions of the assumed burst subclasses and the burst positional errors. Of the two, the positional errors are more significant and indicate that the results of Quashnock & Lamb may be due to systematic errors, rather than to any intrinsic correlation among the burst positions. Monte Carlo simulations also show that, with the current systematic errors, the nearest neighbour statistic is not very sensitive to moderate repetition rates. Furthermore, these Monte Carlo runs show that simulated data with repetitions are less sensitive to positional errors than are the observations, strengthening the argument for systematic errors producing the signal. Until the BATSE statistical and systematic errors are fully understood, the burst nearest neighbour correlations cannot be claimed to be significant evidence for burst repetitions.