This paper sets out to identify the role of computers in the architectural design process from a historical perspective. In the last thirty years, computer systems have changed almost entirely the process of design. Computer-Aided Design (CAD) systems have been introduced which vary from simple graphic tools for exploring alternative formal solutions to sophisticated automated programs which can substitute many design activities. In this paper CAD systems have been grouped into three major categories: automated, augmented and formalistic. Automated systems will be discussed first. In automated systems, the design process consists of an initial state which is the input data, a series of rules which transform data and a definition of desired or allowed output states. Each rule has attached to it a number of logical preconditions which test the current state of the data to see if the rule can be applied. The works of Alexander, Newell, and Simon will be briefly presented. A particular emphasis will be placed on the contribution of expert systems and artificial intelligence. Augmented systems will be discussed next. In augmented systems, the machine is introduced as a computer-aided instructional system, as a mediator for the goals and desires of the architects. Together with the advantage of being an information processing mechanism, the computer can communicate with architects by accepting information, manipulating it and providing output helpful to them. The works of Eastman, Kalay and Negroponte will be briefly presented. The inherent difficulties of these systems will be pointed out in order to identify the reasons for their marginal contribution in the architectural design process. Thirdly, a new category, that of formalistic systems will be presented. In these systems, design is viewed as a formalistic activity, as an invention and exploration of new forms and relationships between forms. The formalistic approach to computer-aided architectural design has been only recently considered as an interesting direction and therefore research work is quite limited. Three approaches in this direction will be discussed: shape grammars, generative systems and transformations. The works of Stiny, Mitchell, Yessios and Eisenmann will be briefly presented. Finally, some speculations on the future of computer-aided architectural design will be discussed.