We investigate HST /ACS and WFPC2 images at the positions of five candidate microlensing events from a large survey of variability in M31 ( MEGA ) .
Three closely match unresolved sources , and two produce only flux upper limits .
All are confined to regions of the color-magnitude diagram where stellar variability is unlikely to be easily confused with microlensing .
Red variable stars can not explain these events ( although background supernova are possible for two ) .
If these lenses arise in M31 ’ s halo , they are due to masses 0.15 < m / M _ { \odot } < 0.49 ( 95 % certainty , for a \delta -function mass distribution ) , brown dwarfs for disk lenses , and stellar masses for bulge lenses .