In determining the average numbern¯of tracks per bubble chamber picture, for purposes of determining cross sections, one usually cannot count all tracks, but counts tracks in only some of the pictures, selected at random. An alternative procedure is described here in which the pictures are not selected at random, but are those which contain the ``interesting events.'' The average of 1/nover this ``linearly biased'' sample equals 1/n¯, wheren¯is the desired average over arandomsample.The linear‐bias method has some advantages that are complementary to those of the usual random method.