Effects of Different Instructions on Consistency of Personality Inventory Responses

A novel feature of the Edwards Personality Inventory (EPI) is that the S responds to the items as he believes those who know him best would answer if asked to describe him. This study tested Edwards' hypothesis that reliability would be higher for these instructions vs. the more conventional ones to describe yourself directly as you see yourself. 180 Ss were tested with approximately one-third in each of three conditions: Booklet IA (test-retest), Booklet IB (test-retest), and Booklets IA and IB (alternate forms reliability). Within each condition, approximately half responded under Edwards' novel instructions, and half responded under conventional instructions. Directly contrary to Edwards' prediction, it was found that the proportion of conventional instruction correlations which were higher than the corresponding Edwards' instruction correlations was significantly greater than chance (p <.01).