<![CDATA[Abstract: The objectives of this research were to find out in each category self confidence (high, medium, and low) which one providing better mathematics learning achievement, GI AfL with a peer assessment, GI, or direct learning model. This research used the quasi experimental research method with the factorial design 3x3. Population of this research was all VIII graders of Junior High School of Surakarta Regency in the school year of 2015/2016. The samples of the research were taken by using the stratified cluster random sampling. The proposed hypothesis of the research were analyzed by using the two-way analysis of variance with unbalanced cell. The result of the research could be concluded as follows: in high and medium self confidence, GI based AfL with a peer assessment, GI, and direct learning model gave the same learning achievement; in low self confidence, GI based AfL with a peer assessment gave the same achievement as GI learning model, GI based AfL with a peer assessment and GI learning model gave better achievement than direct learning model; in GI based AfL with a peer assessment and GI, students with high, medium, and low self confidence had the same achievement; in direct learning model, students with high and medium self confidence had the same achievement, students with high and medium self confidence had better achievement than low self confidence.Keywords: GI Based AfL with a Peer Assessment, GI, Direct Learning, and Self Confidence]]>
