According to the encoding flexibility model, stereotypes are efficient because they facilitate, in different ways, the encoding of both stereotype-consistent and -inconsistent information when capacity is low. Whereas inconsistent information receives more extensive attention and perceptual encoding, the conceptual meaning of consistent information is extracted to a greater degree. Experiments 1-2 demonstrated that, under low capacity conditions, perceivers attend more carefully to inconsistent than consistent information. Experiment 3 showed that participants forced to attend selectively to either consistent or inconsistent information encoded inconsistent information more extensively when capacity was depleted. Experiment 4 demonstrated that perceptual encoding favors inconsistent over consistent information. Finally, Experiment 5 showed that, despite the attentional and perceptual encoding advantages for inconsistent information, conceptual encoding favors consistent information when capacity is low