Simulation techniques were used to generate conventional test responses and track the proportion of alternatives examinees could classify independently before and after taking the test. Finite-state scores were compared with these actual values and with number-correct and formula scores. Finite-state scores proved useful. (TJH)