The effect on students' scores of different item formats used in standardized reading achievement tests was investigated. Thirty-one bilingual Mexican-American second grade students were given a phonics test using four different formats derived from three standardized reading achievement tests. The formats sampled identical content. A two-way analysis of variance was performed and a Newman Keuls' Multiple Range Test for Differences Between Means was calculated, since main effect for format was statistically significant. Type of format accounted for almost three quarters a standard deviation in students' test scores. Results indicated that what a student appears to know is substantially influenced by the format of the particular test used in measuring achievement. On this basis, it is recommended that greater attention should be given to the effect of test item format in selecting and administering achievement tests. (Author/AEF)