The use of a variation of linear programing (goal programing) to develop a test blueprint with multiple specification requirements, is the topic of this paper. The computer program STORM was used to develop a table of test specifications for a test that would measure achievement in introductory statistics in the subdomains of frequency distributions and graphs, central tendency, and variability. Specific goals were minimal length, reasonable testing time, and feasible development cost (in terms of time). Twenty-five specific constraints were specified with regard to cognitive processes and content measured. The developed blueprint could not achieve all the goals, but did allow a compromise reflecting the individually prioritized goals. Goal programing appeared to be a practical method for specifying the dimensions of a test blueprint. Five tables illustrate the goal programming process. (SLD)