A search for patterns in grade retention rates across grades 1-12 to construct simulation research models is described in this paper. Aggregate data from Shepard and Smith (1989) on grade retention rates from 12 American states for the year 1979-80 and from 11 states for 1985-86 were examined for across-grade patterns, supplemented by Dade County School District data for 4 consecutive years, 1982-83 through 1985-86. First, the elementary level pattern is explored. A process to explain the pattern is described and a mathematical function derived from the explanation is proposed to fit the data. Next, a pattern is sought for the highest magnitudes of retention rates, and the elementary grade function is extended to provide a proposed comprehensive fit. Fluctuations in the states' data are accounted for, enhancing the model's credibility. Finally, median summaries of the data by grade are presented and compared to an across-grade model. Findings are as follows: (1) the pattern of the rate magnitude distribution in grades 1-6 suggests a negative exponential function as the target population is depleted across the grades; (2) the retention rate magnitudes show a pronounced tendency to peak at the beginning of each grade level; (3) this finding is supported by shifts in the magnitude of the peaks following the move from a junior high to middle school configuration; and (4) simple equations fit the patterns logically and empirically. A conclusion is that the simulation approach offers an unobtrusive way to investigate patterns of different populations in the retention pool. Five figures are included. (6 references) (LMI)