Two chi square techniques were applied to identify culturally biased items on the Reading Vocabulary and Reading Comprehension sections of the California Achievement Tests. Subjects were 1,035 Mexican Americans and 2,474 Anglo Americans in grade 7. Eight items appeared biased against Mexican Americans, but not Anglo Americans; seven occurred in the Vocabulary section. Three of the eight items also showed significant cultural group differences in response to incorrect answers, as well as four items not previously identified. Items were conditionally biased; that is, the bias appeared significant only for students in one level of reading ability--those answering 26-50% correct on the total test. Removing the biased items did not increase the mean score for the Mexican Americans. Mexican American educators cited the inability to understand stem words as the main source of bias for at least six of the items, rather than language usage, cultural values, or test wiseness. In conclusion, the two methods detected biased items, but did not necessarily identify the same items. Used jointly, they were unable to identify a biased item by achievement level, or to specify the incorrect response that was the source of bias. (CP)