This study was designed to find: a suitable effect size (ES) measure or common metric (CM) for comparing the results of a set of single subject research (SSR) studies; and an easy way to convert the published graphs back into raw data from which ESs could be calculated. To meet the first objective, three possible formulas for measuring treatment effect were evaluated and then compared using identical data sets. To meet the second objective, a computer method was developed. Data for the study were taken from two previously published meta-analyses of SSR studies. The larger of the two (n=23 articles) which used piece-wise regression (PR) to calculate an ES, was the Skiba, Casey, and Center (1985-86) study concerning the use of non-aversive procedures in the treatment of classroom behavior problems. The smaller study (n=13 articles) which used the percent of non-overlapping data (PND) as a CM, was conducted by Scruggs, Mastropieri, Cook, and Escobar (1986) and concerned early intervention for children with conduct disorders. The CM study included a total of 195 graphed experiments. A simple comparison of the baseline and treatment phases from each graph was used; and a third possible CM was developed from an adaptation of Glass' formula (AGF) for group studies. The computerized method study used a scanner and a "mouse" device with a microcomputer. Data points were estimated by finding the screen coordinates of each point and using the scale on the ordinate. The PR model calculated an ES based on the combined effect of a change in level and slope between the two phases; the PND model grossly measured the differences in level between the two phases; and the AGF ES measured a standardized difference between the means of the phases. In meta-analyses for SSR, when different ESs (based on different sets of assumptions) are calculated yielding similar results, the validity of the conclusions increases. Three graphs are included. (RLC)