Character separation indices are useful for determining which binary characters best discriminate between two populations. To date these indices are unable to cope sufficiently with binary data that contain equivocal responses. A character separation index is described which can be applied to data containing equivocal responses. Furthermore, for larger numbers of group members this index can be used to derive the significance that a given character would have given a particular separation value by chance. The BASIC implementation of a program to calculate character separation indices is presented.