A conventional vocabulary test and two forms of a stradaptive vocabulary test were administered by a time-shared computer system to undergraduate college students. The two stradaptive tests differed in that one counted question mark responses (i.e., omitted items) as incorrect and the other ignored items responded to with question marks. Stradaptive test scores were more consistent with the hypothesized nature of the population distribution of verbal ability. When corrected for differing levels of item discrimination and memory effects, the test-retest stabilities of the two testing strategies were about equal. Scores on one form of the stradaptive test were found to be very stable for testees who had highly consistent response records on initial testing. Stability of "subject characteristic curve" data was high, suggesting the usefulness of these data for describing test-testee interactions. Of the 10 stradaptive ability scores studied, which grouped into four clusters, average difficulty scores had the highest stabilities. Analysis of difficulties of items associated with correct, incorrect, and question mark responses suggested that items with question mark responses should not be ignored, but should be treated as incorrect responses in branching decisions. Suggestions for future research on the stradaptive testing model are made. (Author)