In this research the effects of reciprocal peer tutoring on students' mathematics anxiety levels were examined. A pretest posttest with control group design was used at a public middle school in Spain. A total of 420 students in 7th, 8th, and 9th grades participated in the study, of which 215 were female and 205 were male. Students were randomly assigned and equally distributed by course grade (140 in each course grade) and experimental condition (210 in the experimental group and 210 in the control group). Quantitative data were gathered using the Mathematics Anxiety Scale developed by Chiu and Henry (1990). Qualitative information was gathered during eight focus group sessions that were held with students. Two main factors were analyzed using the quantitative and qualitative information: mathematics learning anxiety and mathematics evaluation anxiety. Results were analyzed by gender and course grade. Statistically significant improvements were reported for both male and female students in the experimental group and for each course grade for both factors. No statistically significant differences were reported for students in the control group in any case. A moderate effect size was reported for mathematics evaluation anxiety (Hedge's g = 0.42), and a large effect size was reported for mathematics learning anxiety (Hedge's g = 0.84). Information obtained from the focus groups was consistent with the reported quantitative results. The main conclusion is that peer tutoring may be very beneficial for reducing middle school students' mathematics anxiety, regardless of their gender or grade.