The purpose of the present study is to examine concurrent and predictive evidence used in the validation of ACCUPLACER, a purchased test used to place first-year students into writing courses at an urban, public research university devoted to science and technology education. Concurrent evidence was determined by correlations between ACCUPLACER scores and scores on two other tests designed to measure writing ability: the New Jersey Basic Skills Placement Test and the SAT Writing Section. Predictive evidence was determined by coefficients of determination between ACCUPLACER scores and end-of-semester performance measures. A longitudinal study was also conducted to investigate the grade history of students placed into first-year writing by established and new methods. When analyzed in terms of gender and ethnicity impact, ACCUPLACER failed to achieve statistically significant prediction rates for student performance. The study reveals some limits of placement testing and the problems related to it.