From a base in military models, computer simulation has evolved to provide a wide variety of applications in social science. General purpose simulation packages and languages such as FIRM, DYNAMO, and others have made significant contributions toward policy discussion in the social sciences and have well-documented efficacy in instructional settings compared to traditional settings. With the microcomputer revolution has come a revival of computer simulation, first in pre-collegiate education, then in a variety of other areas including urban policy, social processes, and research methodology. In spite of the significant body of research and software in social science it is striking that (1) this research and teaching tradition exists almost entirely outside the "mainstream" of social science journals and textbooks, and (2) the impact of computer simulation has had a relatively greater impact on teaching than research. These findings reflect the state of the social sciences. Attempts at simulation in research-poor environments inevitably lead to models characterized by low reliability, oversimplification, and neglect of causal factors. These three obstacles are the reason why computer simulation has remained a specialized area within social science and is removed from the mainstream practice of social scientists. It is concluded that computer simulation in social science, though a positive way of accumulating knowledge, is apt to remain the step-child it has always been. Over 90 related references conclude the publication. (LH)