The use of Gickling's model of Curriculum-Based Assessment (CBA) has been shown to increase student achievement. However, data are lacking to support the psychometric adequacy of the procedure. This study examined the interscorer, alternate-form, internal consistency, and test-retest reliability of CBA for reading using 93 elementary school aged children. The data were expressed as a percentage of words known and as conversions to the independent, instructional, and frustrational level based on Gickling and Rosenfield (1995). Results suggested reliability coefficients that exceeded recommended levels for screening instruments and were consistent with the fluency model of CBA. Recommendations for practice and further investigation are included.