The present study examined whether several student background characteristics (e.g., father's occupation) could act as moderators leading to the differential prediction of law school grades. A new procedure based on a modification of Ward's hierarchical grouping technique was used for this purpose on the validation sample (n = 201). This resulted in a four-group solution using five moderators. Separate regression equations were constructed for each group and applied to corresponding groups in the cross-validation sample (n = 201). The results of these analyses indicated that the new multiple moderator technique could reliably identify groups that differed in predictability.