The purpose of this study was to compare the reliabilities of two-, three-, four-, and five-choice tests using an incremental option paradigm. Test forms were created incrementally, a method approximating actual test construction procedures. Participants were 154 12th-grade students from the Portland (Oregon) area. A 45-item test with two options per item was developed; and three-, four-, and five-option test forms were constructed by adding options to the two-option per item test. Reliability coefficients were calculated from the different forms of the test before and after implementing the Tversky condition, which assumes that testing time is proportional to test length. Despite significant differences in the reliability coefficients before invoking the Tversky condition, the magnitudes of the internal consistency reliability estimates for the three-, four-, and five-option formats after implementing the Tversky condition were similar, suggesting that time and energy might have been saved by constructing three-option items without loss in reliability. One table presents data about the test forms, and an 11-item list of references is included. (SLD)