Introduction It has been argued often and convincingly that requirements determination is the most critical phase of information systems development (Byrd, Cossick, and Zmud, 1992; Dalal and Yadav, 1992). System misuse or disuse can often be traced to an inadequate requirements determination process. The majority of information systems developed for organizations require post-implementation revisions to meet users' needs (Wetherbe, 1991). Although information systems are expensive to develop initially, changes made after a system has been completed are much more expensive than making the same changes during the design process (Boehm, 1981). Consequently, research that can enlighten and improve the requirements elicitation process can make an important contribution to the quality and cost-effectiveness of software development efforts. This paper proposes a task behavior-oriented approach to the determination of information requirements for the design of decision support systems (DSS). It is argued that the behavioral nature of DSS requires a shift from the data-driven requirements determination approaches used to construct transaction processing and similar systems. The current trend toward task performance-centered DSS in industry (Gery, 1995) also supports a shift in requirements determination focus. Typical informationrequirements determination (IRD) methods include structured interviews, questionnaires, observation, and joint application design, among others (Whitten, Bentley, and Barlow, 1994). In most of these techniques, users are asked questions relating to goals,data, problems, and critical success factors, for example, and the answers are used to infer system requirements. However, these methods may not adequately specify the actual task performance behaviors in which users engage, as the methods are generally data-focused. Further, the level of requirements elicited through such methods is often too general to be of significant use to analysts attempting to understand behavior. Tools designed to elicit more specific task behaviors are likely to be of greater benefit, since such behaviors are better descriptors of user needs and arguably can lead to more complete system requirements (Keen, 1980). Because DSS are developed to support organizational tasks that decision makers perform, it is arguable that the requirements determination for such systems should more directly address users' performance of those tasks