The present study explored the psychometrie properties ofTurner and Engle's (1989) operation span task, a widely used measure of working memory capacity. We administered the task three tirnes to 33 college students, using equivalent test materials. The interval between the first and second administrations was 3 weeks, with 6-7 weeks between the second and third administrations. Alpha coefficients were all .75 or more. RecalI accuracy decreased as operation set size increased. Raw test-retest correlations ranged from .67 to .81, the corrected reliability was .88, and stability scores ranged from .76 to .92.Performance irnproved from the first to the second test. Relative to reported reliabilities of other tasks used to assess individual differences in working memory capacity, the operation span task appears to have several statistical advantages.In recent years, individual differences in working memory (WM) capacity have received a great deal of attention from cognitive psychologists. WM capacity is considered to be a relatively stable individual difference that is related to performance on a variety of other cognitive tasks (Kyllonen & Christal, 1990). WM differences are also of interest to investigators exploring other basic cognitive processes such as the role of attention in storage and retrieval (Rosen & Engle, 1997).There are a number of measures of WM capacity, all requiring participants to process and store information simultaneously. Individuals must read a sentence (see, e.g., Daneman & Carpenter, 1980), make a judgment about a sentence's acceptability (Waters & Caplan, 1996), or solve a simple arithmetic problem (La Pointe & Engle, 1990) while remembering a word across varying numbers of sentences or arithmetic operations. Scores on these complex WM span tasks are used as both continuous and categorical measures. Engle, Cantor, and Carullo 's (1992) work testing the relationship between complex memory span scores and verbal Scholastic Aptitude Test (SAT) scores illustrates the correlational approach. Verbal tluency differences in participants categorized as high versus low span on the basis oftheir operation-span scores are an example of the second approach (Rosen & Engle, 1997). Both these methodologies, correlational and quasi-experimental, rest on a single administration of the task with the presumption that WM measures are reliable. There is some evidence regarding the internal consistency of WM measures, but little attention has been devoted to estimating test-retest reliability or stability. AIthough Turner and Engle (1989) found internal consisWe thank Adriel Boals, who programmed the working memory task, and Randall W Engle, who provided many helpful comments during the preparation of this report. Correspondence should be addressed to K. Klein, Department ofPsychology, Box 7801, North Carolina State University, Raleigh, NC 27695 (e-mail: kitty_klein@ncsu.edu).tency to range between .74 and .81 among operations of the same set size across trials of their operation word span task, they did not r...