A two-stage adaptive test and a conventional peaked test were constructed and administered on a time-shared computer system to students in undergraduate psychology courses. (The two-stage adaptive test consisted of a routing test followed by one of a series of measurement tests.) Comparison of the score distributions showed that the two-stage test scores were more variable than the linear test scores; the distribution of two-stage scores was normal, whereas that of the linear test scores tended toward flatness. When considering the memory of the items, the two-stage test was found to have higher test-retest stability than the conventional. The relationship between the two-stage and conventional test scores was relatively high and primarily linear, but about 20% of the reliable variance in the conventional test scores was left unaccounted for. Further analyses of the two-stage test showed that the difficulty levels of the measurement tests were not optimal, and that 4 to 5% of the examinees were routed to inappropriate measurement tests. The poor internal consistency of the measurement tests in comparison with that of the routing and conventional tests was apparently due to the extreme homogeneity of ability within the measurement test sub-groups. The findings of the study were interpreted as favorable to continue exploration of two-stage testing. (Author/NE)