ABSTRACT
Computer-based tests (CBTs) take many forms, with designs that range from simple to quite elaborate. At the simple end of the spectrum are “linear” tests that use the computer only as a means of administering what is in most respects identical to a conventional paper-and-pencil test. In contrast, computerized adaptive testing (CAT) assembles different tests for different test takers, with the goal of optimizing measurement. Designs may include those that select items individually or those that select items in blocks or sets. This range of variation among adaptive test designs requires this chapter to narrow its focus in order to present information at a depth greater than a superficial overview. We will therefore focus on designs that work at the level of individual items as opposed to larger blocks or sets. 1 The exception to this will be the allowance of groups or sets comprising items that are naturally and inextricably linked; an example would be a text passage or other stimulus material to which several items are attached and upon which they collectively draw. We will also concentrate on adaptive tests that provide continuous scores rather than classification decisions. A related chapter in The Handbook of Test Development by Luecht (this volume) discusses the alternative case, where the test is designed to make accurate classification decisions.
