This paper addresses the following criticisms of task analysis: its requirement for an existing system; its focus on analysis rather than design; its limited scope within the design cycle; its underspecification of the application domain; and its inadequate documentation of design outputs. Such criticisms constitute problems for the application of task analysis. Solutions to these problems have emerged indirectly from the development of a structured method intended to rectify the typically late involvement of human factors in system development, e.g. only at the evaluation stage. This paper describes how task analysis has been extended and incorporated throughout the design cycle.