IF-AT (also called answer until correct) is a revised form of traditional multiple-choice testing. It was created to provide instant feedback during testing and to tap into partial knowledge of learners. The purposes of this meta-analytic study were to compare the effectiveness of IF-AT with the traditional multiple-choice test; to assess effectiveness of the computer-based IF-AT with the scratch card based IF-AT; and to identify modifies of IFAT effects. A total of 35 IFAT studies were examined. Data were analyzed using traditional meta-analysis and FAT-PET-PEESE-MRA. The results indicated: 1) The overall effect size of IF-AT over the traditional multiple choice was 0.581(fixed) and 0.618(random). By FAT-PET-PEESE MRA, the net overall effect size of IF-AT was 0.522, which was close to a previous meta-analysis result (Bangert-Drowns et.al., 1991) based on 4 answer-until-correct studies; 2) In terms of IF-AT forms, effectiveness of IFAT interacted with testing materials. Scratch card-based IF-AT was significantly more effective than computer-based IF-AT in language arts and social sciences (LS); while they worked equally well in mathematics and science (MS); and 3) Test materials was identified as a significant moderator to explain a large part of heterogeneity between IF-AT studies. In terms of test materials, students taking IF-AT tests always performed better in LS than MS regardless whether it was scratch cardbased IF-AT or computer-based IFAT. The evidence suggests that IF-AT is very effective in enhancing students' learning within LS by memory retention, and it is less effective in MS of which requires more conceptual understanding than memory retention.