Since grade-appropriate levels of standardized achievement tests must frequently be used in Elementary and Secondary Education Act Title I evaluations, there may be large discrepancies between test difficulty and the achievement level of those being tested, with resultant inaccuracies in the scores. Procedures based on the Rasch scaling model were used to assess the degree to which scores on grade-appropriate reading achievement tests reflected true achievement, and to develop a subset of the item battery that reflected true achievement and that could be linked to national percentile ranks. Comprehensive Tests of Basic Skills reading vocabulary and reading comprehension subtest scores were obtained for 1,000 West Virginia third graders, and submitted to a Rasch scaling analysis using the BICAL computer program. BICAL performed the scaling, assigned item difficulty and student ability values, and generated item-fit statistics for each item and for six student ability groups. Results of the reliability analyses contradicted the original expectations. In conclusion, there was support for testing achievement with a subset of items that was most appropriate for low-achieving students. Similar procedures were recommended for assessing performance of a specific skill. (Author/GDC)