Describes an experiment in which ECSS, RAND's specialized language for simulating computer systems (RM-6132), was compared with PL/I and FORTRAN for programming a simulation of a multiprogrammed computer. Using a previously published problem specification, programming time was recorded in four categories: reviewing the language, designing the program, coding, and checkout. Quantitative data were also recorded on characteristics of the source and object code, errors made, machine usage, and run-time. A factor of three reduction in programming time and a factor of two reduction in machine time for model development using ECSS were obtained, but the number of statements required was not significantly reduced, more errors were made, and object program size was much larger for the ECSS version. Although a number of experimental difficulties were found to limit the generality of the results, they have been valuable in guiding further development of ECSS. 27 pp. Bibliog.