This study aimed to investigate the effects of geometry instruction based on Dienes' principles (constructivity, dynamic, mathematical variability and conceptual variability principles) on 4th graders' geometry success and retention of learning. It was a pretest posttest control group quasi-experimental study. The study group comprised three classes of 4th graders from three different elementary schools in Nevsehir who had been shown to be identical via a "Geometry Level Identification Test". The study was conducted with two experimental groups and a single control group. The study took 39 class hours. While instruction in the experimental groups was based on Dienes' principles, the researcher did not intervene in the instructional process in the control group. The data collection tool used in the study was the "Geometry Level Identification Test". In comparing the pretest, posttest and retention test mean scores, Covariance (ANCOVA) and paired-sample t-test analyses were used. The results revealed that the experimental groups, which experienced learning activities based on Dienes' principles, had better geometry success than the control group where instruction was not manipulated. While no significant difference emerged between the "Geometry Level Identification" posttest mean scores of the two experimental groups, both groups differed significantly from the control group. The retention test given three weeks following the study showed that some information had been forgotten by all three groups. The retention test scores of the groups were 5-7 points lower than their posttest scores. Even though a significant difference was not found between the retention scores of the groups, it may be stated that the decline in the control group was noteworthy.