The COMPTEL instrument on the Compton Gamma Ray Observatory is used to
measure the locations of gamma-ray bursts through direct imaging of MeV
photons. In a comprehensive search, we have detected and localized 29 bursts
observed between 1991 April 19 and 1995 May 31. The average location accuracy
of these events is 1.25$\arcdeg$ (1$\sigma$), including a systematic error of
$\sim0.5\arcdeg$, which is verified through comparison with Interplanetary
Network (IPN) timing annuli. The combination of COMPTEL and IPN measurements
results in locations for 26 of the bursts with an average ``error box'' area of
only $\sim$0.3 deg$^2$ (1$\sigma$). We find that the angular distribution of
COMPTEL burst locations is consistent with large-scale isotropy and that there
is no statistically significant evidence of small-angle auto-correlations. We
conclude that there is no compelling evidence for burst repetition since no
more than two of the events (or $\sim$7% of the 29 bursts) could possibly have
come from the same source. We also find that there is no significant
correlation between the burst locations and either Abell clusters of galaxies
or radio-quiet quasars. Agreement between individual COMPTEL locations and IPN
annuli places a lower limit of $\sim$100~AU (95% confidence) on the distance to
the stronger bursts.
