Objectives: To study androgen treatment effects on arithmetic performance in girls with Turner syndrome. Study design: Forty-four girls, ages 10 to 14 years at baseline, completed 4 years of treatment with oxandrolone (Ox) or placebo (Pl). All received growth hormone and estrogen replacement therapy. We assessed the number of girls with severe learning disability (LD, standard score ≤5th percentile) on measures of academic arithmetic and reading achievement (WRAT-3, arithmetic and reading), given yearly, and the WIAT numerical operations (NOS) and reading subtests, given at year 4. Results: On the WRAT-3 arithmetic, the frequency of severe arithmetic LD was similar in the Ox and Pl groups at baseline and at years 1 and 2. At years 3 and 4, fewer girls in the Ox than Pl group had a severe arithmetic LD (year 4: 0/22 vs 5/21, P = .02). On the WIAT NOS (year 4), fewer girls in the Ox than Pl groups had a severe arithmetic LD (3/21 vs 8/20, P = .09). WIAT NOS error analysis suggested that the improved performance in the Ox group was associated with better performance on multiplication and division (P < .01). The frequency of severe LD for the WRAT-3 reading was similar for the Ox and Pl groups (all years) and for the WIAT reading subtest (year 4). Conclusions: Androgen treatment for 4 years in girls with Turner syndrome resulted in a small decrease in frequency of severe arithmetic LD, with no effect on reading LD.