This paper contributes to the literature examining the effects of stereotype threat on performance on Raven's Advanced Progressive Matrices (APM) task, a nonverbal task that has been described as "culture-reduced" and a relatively "pure" measure of cognitive ability and reasoning (see Mayer & Hanges, 2003; McKay, Doverspike, Bowen-Hilton, & Martin, 2002). Black and White undergraduates completed the APM after being given standard instructions that "the Advanced Progressive Matrices is a measure of observation and clear thinking" (control), that it "is an IQ test...frequently used to measure intelligence and ability (stereotype threat for Blacks), or that it "is a series of puzzles" (low stereotype threat). Performance on the APM was assessed after controlling for differences in ACT scores across experimental conditions. Results showed that AMP performance was better for Whites than for Blacks under high stereotype threat and standard instructions. However, the performance gap was eliminated (and even reversed, though not significantly) in the low threat condition, when the APM was described as a series of puzzles. These results show that stereotype threat can undermine performance on tests that are viewed as free of cultural bias and not reliant on reading skills or background knowledge. It also suggests that "standard instructions" used with many test instruments might invoke stereotype threat. Only when the APM was described as a set of "puzzles" were performance differences between Blacks and Whites eliminated.