The E>100 MeV map for each viewing period, as well as the summed maps of Phase 1, Phase 2, and Phase 1+2, were examined in the course of routine EGRET data processing with the maximum likelihood program. Any excess above the diffuse model which showed a 
3 was initially retained as a candidate source. A cross-comparison of the candidate sources was made in order to eliminate cases in which the same source was assigned different positions in different observations. From this comparison, a list of candidate sources was derived. Excesses with 
4 were retained only if they were close enough to another candidate source that the point spread function would overlap substantially.