Undergraduates participated in 3 experiments related to self-adapted t
esting. Experiment 1 demonstrated that, in comparison with computerize
d adaptive testing, self-adapted testing reduced the influence of anxi
ety on performance but took longer and was less efficient. Experiment
2 indicated that benefits of self-adapted testing cannot be attributed
solely to item ordering. Instead, active choice of item difficulty se
ems to be necessary. Experiment 3 demonstrated that the provision of f
eedback increased the efficiency of the test but had no effect on esti
mates of ability derived. The potential of self-adapted testing to red
uce the influence of extraneous sources of variation in test performan
ce is discussed.