The development of a short domain-general measure of working memory capacity

• Published in: Behavior research methods Vol. 47; no. 4; pp. 1343 - 1355
• Main Authors: Oswald, Frederick L., McAbee, Samuel T., Redick, Thomas S., Hambrick, David Z.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01-12-2015; Springer Nature B.V
• Subjects: Adolescent; Adult; Archives & records; Behavioral Science and Psychology; Cognitive Psychology; Discriminant analysis; Factor Analysis, Statistical; Female; Humans; Individuality; Male; Memory; Memory, Short-Term; Mental Processes - physiology; Models, Theoretical; Neuropsychological Tests; Psychology; Psychomotor Performance - physiology; Reading; Sampling; Symmetry; Young Adult; Measurement; Working memory; Complex span tasks
• DOI: 10.3758/s13428-014-0543-2
• EISSN: 1554-3528

Working memory capacity is one of the most frequently measured individual difference constructs in cognitive psychology and related fields. However, implementation of complex span and other working memory measures is generally time-consuming for administrators and examinees alike. Because researchers often must manage the tension between limited testing time and measuring numerous constructs reliably, a short and effective measure of working memory capacity would often be a major practical benefit in future research efforts. The current study developed a shortened computerized domain-general measure of working memory capacity by representatively sampling items from three existing complex working memory span tasks: operation span, reading span, and symmetry span. Using a large archival data set (Study 1,
N
= 4,845), we developed and applied a principled strategy for developing the reduced measure, based on testing a series of confirmatory factor analysis models. Adequate fit indices from these models lent support to this strategy. The resulting shortened measure was then administered to a second independent sample (Study 2,
N
= 172), demonstrating that the new measure saves roughly 15 min (30 %) of testing time on average, and even up to 25 min depending on the test-taker. On the basis of these initial promising findings, several directions for future research are discussed.