The aim of this study was to examine the effects of discourse-based mathematics instruction (through classroom discussion) on female fourth graders’ math performance. To this end, students’ mathematics performance in three cognitive domains of TIMSS, i.e., understanding, applying and reasoning, was probed. The statistical population of the study included all fourth grade female students in Shahriyar studying in public schools in academic year 2014-2015. The sample included 80 students who were selected by random multi-stage cluster sampling method and assigned to experimental and control groups. The instrument comprised a performance assessment test which included items from TIMSS cognitive math domains of knowing, applying and reasoning. The reliability of the measure was computed to be 0.77. The experimental group underwent the intervention (discourse-based mathematics instruction through classroom discussion) in three stages of "action", "exploration" and "discussion and summarizing" within the framework of Cirillo’s (2013) strategies for facilitating class discussions in 16 sessions. The results of MANOVA indicated that students’ math performance in the experimental group was higher than that of the control group in two cognitive domains of knowing and applying; however, no significant difference was observed between the two groups in the reasoning domain. In general, discourse-based mathematics instruction improves students’ math performance in TIMSS cognitive domains.