A random sample of October, 1979, test sites was selected to assess the potential bias of the ACT Assessment. From this sample, sites which tested at least five Anglos and five Hispanics were included; the final sample contained a total of 876 students (403 Anglo and 373 Hispanic). Using the primary hypothesis, "There is no interaction between the items and ethnic groups," a three factor (items, sex, and ethnic group) analysis of variance design was used to separately analyze the four tests (English, mathematics, social studies, and natural science). The item by ethnic group interactions were statistically significant for two of the four content tests, English usage and social studies. Items significantly contributing to the item by ethnic group interactions were analyzed further. Within the English usage test, five reading passages (six items) contributed to the item by ethnic group interaction; three items favored Anglos and three favored Hispanics. Seven items were identified for the Social Studies Reading tests, four favored Anglos and three favored Hispanics. Over the two tests, there were seven items favoring Anglos and six favoring Hispanics; therefore, the tests could not reasonably be said to favor one group decisively more than the other. Four tables are appended. (BRR)