A previously unreported feature of the coalescence of water drops was observed for an isolated pair of drops (R = 275-mu-m and r = 200-mu-m) falling at terminal velocity through nearly saturated air. This new feature, grazing coalescence, occurred when the horizontal offset between centers of the colliding drops is near the sum of their radii, and was observed because of the use of orthogonally placed cameras that permitted the determination of the true horizontal offset between the drops just prior to collision. This discovery has prompted a reexamination of a study by Ochs et al. (1986) in which a single camera system was used to make observations of drop collisions at similar drop sizes. In the previous work it was assumed that more central collisions would result in coalescence and that collisions outside of the central region would result in bounce. The maximum observed horizontal offset resulting in coalescence was believed to be an observation from which the coalescence efficiency could be derived. Since this geometrical method proved unsatisfactory, probably due to grazing coalescences, they used a statistical approach to evaluate the data. In light of the present observations the older work is reexamined and found to have reported coalescence efficiencies that are probably too high, most likely from excess charge on the drops in the earlier study.