Anabolic-androgenic steroid (AAS) abuse is prevalent not only among elite athletes, but is increasingly common in high school and collegiate sports. AAS are implicated in maladaptive behaviors such as increased aggression and risk taking, which may result from impaired cognition. Because they affect dopamine function in prefrontal cortical (PFC)-striatal circuitry, AAS may disrupt PFC-dependent processes such as behavioral flexibility. This was the focus of the present study. Adolescent male Long-Evans rats were treated chronically with high-dose testosterone (7.5 mg/kg in water with 13% cyclodextrin) or vehicle sc, and tested for set-shifting and reversal-learning. For set-shifting, rats were trained on a visual cue task (VCT), then were shifted to a direction cue task (DCT), or vice-versa. For reversal learning, rats were first trained on VCT and were then required to press the opposite lever. 2-cue set-shifting introduced a novel paradigm in which rats shifted from a 1-Light Visual Task (1LVT) to a tone cue task (TCT). Testosterone-treated rats were significantly impaired on the set-shift from DCT to VCT compared to vehicle-treated controls (trials to criterion: vehicle 240.9±29.9, testosterone 388.3±59.3, p<0.05). However, on the set-shift from VCT to DCT, testosterone did not affect performance. During reversal-learning, testosterone significantly increased trials to criterion (vehicle: 495.9±91.8 trials, testosterone: 793.7±96.7 trials, p<0.05). In 2-cue set-shifting, testosterone diminished performance and the difference showed borderline significance (vehicle: 443.2±84.4 trials, testosterone: 800.4±178.2 trials, p=0.09). Our results show that testosterone impairs behavioral flexibility and have implications for understanding cognitive and behavioral changes in human AAS users.