This paper examines the current state of attempts to incorporate expert test construction practices into a computerized adaptive testing paradigm. A new methodology for incorporating large numbers of constraints on adaptive test item selection is described. This methodology builds on the conventional, that is, linear, test construction practices employed by expert test specialists. The incorporation of such practices is necessary if computerized adaptive testing is to become a serious competitor to conventional paper-and- pencil testing in the future. The successful application of this new methodology to the design of an adaptive test in a real-life setting is demonstrated by an example. (36pp.)