This paper presents an overview of the concepts underlying computer-assisted medical decision-making (CMD) systems. Alternative approaches to constructing CMD systems are reviewed, including “conventional programming techniques,” statistical pattern classification, rule-based deduction, modelling of diagnostic reasoning, and data base comparisons. Each of these methods is illustrated with examples taken from pulmonary medicine. Although much progress has been made over the last two decades, all existing methods for creating CMD systems are seen to be limited in their ability to represent and process medical knowledge. Other problems facing the field include the time and cost of CMD system development, physician resistance to using these systems, the question of transferability, and the lack of large, reliable data bases for many clinical problems. Recent trends that may overcome these problems are identified and discussed.