Working memory span tasks: A methodological review and user's guide

Abstract: Working memory (WM) span tasks-and in particular, counting span, operation span, and reading span tasksare widely used measures of WM capacity. Despite their popularity, however, there has never been a comprehensive analysis of the merits of WM span tasks as measurement tools. Here, we review the genesis of these tasks and discuss how and why they came to be so influential. In so doing, we address the reliability and validity of the tasks, and we consider more technical aspects of the tasks, such as optimal administration and scoring procedures. Finally, we discuss statistical and methodological techniques that have commonly been used in conjunction with WM span tasks, such as latent variable analysis and extreme-groups designs. Article: Other than standardized instruments, such as intelligence test batteries, working memory (WM) span tasks, such as the counting span, operation span, and reading span tasks, are among the most widely used measurement tools in cognitive psychology. These tasks have come to prominence not only for their methodological merit, but also because theoretical advances in the study of human behavior since the cognitive revolution have placed WM as a central construct in psychology. Methodologically, WM span tasks have proven to be both reliable and valid measures of WM capacity (WMC), which we will document below. However, the larger factor in accounting for their increased use is simply that WM has become a widely useful, scientifically fruitful construct. It plays an important role in con-temporary global models of cognition (e.g., J. R. A diverse set of researchers is now using WM as a construct in research programs, as well as measures of WMC in the arsenal of research tools. Within psychology, discussions of WM are now common in almost all branches of the discipline, including cognitive, clinical, social, developmental, and educational psychology. For example, clinical research has demonstrated that WM is related to depression Although the WM construct has been successfully and appropriately exported from cognitive psychology to other disciplines, WM tasks have, in our opinion, suffered a bit in translation. Presumably because of their documented reliability and validity, WM span tasks, among all the available measures of WMC, have been embraced most strongly by researchers outside of cognitive psychology. Inevitably, as WM is exported to other scientific disciplines and as a more varied pool of investigators use WM span tasks, misconceptions and misuses are bound to increase. In particular, the literature presents inconsistent information regarding the reliability of WM span tasks, as well as inconsistent and, in our opinion, problematic procedures for administration and scoring. We therefore believe that the time is right for a review of how and why WM span tasks came to dominate the measurement landscape of WM research and to provide guidelines for researchers who would like to use these tasks. That is, we perceive a need for a thorough review of all aspects of WM span tasks, from optimal administration and scoring procedures to assessment of reliability and validity. Another motivating factor is more self-serving. Each author of this article receives numerous inquiries about these tasks, including questions about administration, scoring, assessment of reliability, use of extreme-groups designs, and so forth, and so we devote considerable time to instructing other researchers about the proper use of these tasks. As such, an important purpose of the present article is to provide a -user's guide‖ for WM span tasks, such that any interested researcher can read this article, download the programs from our Web site (http://psychology.gatech.edu/ renglelab/tasks.htm), and use them most appropriately. We will begin with a historical overview of the development of WM span tasks, followed by guidelines for administration and scoring. The reliability and validity of the tasks will then be discussed. The tasks will then be contrasted with other empirical measures of WM function. Finally, we will discuss common research strategies that have been used in conjunction with these tasks, such as latent variable analysis and extreme-groups designs. The review is limited to considering just three WM span tasks; counting span Finally, we note that the review is as theory neutral as possible with respect to a particular model of WM and/or the nature of individual differences in WMC, because our goal is to review the merits of WM span tasks as research tools for any researcher, with any theoretical stance. However, it is simply impossible to review the history of these tasks, discuss their validity, or even suggest how to score them without revealing some theoretical bias. Therefore, at the outset, we will briefly outline our theoretical approach to WM, for the simple purpose of warning the reader of any bias that may reveal itself later. BRIEF THEORETICAL OVERVIEW OF WM AND WMC We view WM as a multicomponent system responsible for active maintenance of information in the face of ongoing processing and/or distraction. Active maintenance of information is the result of converging processes-most notably, domain-specific storage and rehearsal processes and domain-general executive attention. Furthermore, the extent to which maintenance depends on domain-specific skills versus domaingeneral executive attention varies as a function of individual ability, task context, and ability × context interactions. For instance, a novice chess player will rely more on domain-general executive attention to maintain game information (e.g., recent moves or future positions) than on domain-specific skills (e.g., learned strategies and position patterns). In contrast, an expert chess player typically will rely more on domain-specific processes and skills to maintain information. However, even the expert might need to call upon executive attention under some circumstances, such as playing the game in particularly demanding situations or under some sort of cognitive or emotional load. Similarly, performance on WM span tasks depends on multiple factors, with domain-specific skills, such as chunking and rehearsal, facilitating storage and a domain-general capability allowing for cognitive control and executive attention. A critical aspect of our view, however, is that WM span tasks predict complex cognitive behavior across domains, such as reading comprehension, problem solving, and reasoning, primarily because of the general, executive attention demands of the tasks, rather than the domain-specific demands of the tasks HISTORY OF WM SPAN TASKS AND OVERVIEW OF ADMINISTRATION PROCEDURES WM span tasks, such as counting span, operation span, and reading span, were designed from the perspective of Baddeley and Hitch's (1974) theory of WM, which stressed the functional importance of an immediate-memory system that could briefly store a limited amount of information in the service of ongoing mental activity. That is, a WM system would be unlikely to evolve for the sole purpose of allowing an organism to store or rehearse information (such as a phone number) while it was doing nothing else. A more adaptive system would allow the organism to keep task-relevant information active and accessible in memory during the execution of complex cognitive tasks. WMC measures were, therefore, created to require not only information storage and rehearsal (as do -simple‖ measures of STMC, such as digit span or word span), but also the simultaneous processing of additional information The reading span task was the first task developed with the purpose of jointly tapping the storage and processing functions of WM 1 Word recall is prompted at the completion of an item. In the original version of reading span Daneman and Carpenter (1980) added a true-false component to the task in their Experiment 2. The subjects indicated the veracity of each sentence they read by responding true or false within 1.5 sec of each sentence's presentation. Here, the sentences were drawn from general knowledge quiz books and covered multiple domains, including the biological and physical sciences, literature, geography, history, and current affairs, and were selected to be of moderate difficulty (e.g., -You can trace the languages English and German back to the same roots‖). Although The reading span task used by Engle and colleagues is somewhat different from The version of reading span that we have most commonly used is one in which the to-be-remembered word is different from the last word, or any word, in the sentences (see, e.g., Engle, Tuholski, et al., 1999), and so each sentence is followed by an unrelated word. Subjects still read the sentences aloud and verify whether the sentence is semantically or syntactically correct. Here, however, they are charged with remembering the unrelated word and not the last word of the sentence. We made this change because individual differences in reading ability could lead to differences in the ability to generate the words at test on the basis of the gist of the sentence (rather than on the basis of episodic recall). In our latest version of the task, subjects no longer remember words; instead, they remember isolated letters that follow each sentence Does the Secondary Task Need to Involve Reading? In a version of the task designed for adults by , the stimuli consisted of three items of each size from one to eight. The visual displays were made more complex by placing the target shapes among a field of distractors that shared either the same shape or the same color (and so counting required conjunctive search, à la Summary of Critical Task Components The reading span, operation span, and counting span tasks share an underlying structure and are implemented in much the same way. The tasks are designed to force WM storage in the face of processing (or distraction), in order to engage executive attention processes. As such, the following procedural recommendations should be kept in mind when these tasks are administered (again, note that these recommendations are most appropriate for studies involving healthy young adults). Immediate and vigilant stimulus presentation. A critical feature of the processing component of WM span tasks is that it interferes with rehearsal. Substantial delays between stimulus presentations may, therefore, permit rehearsal of the to-be-remembered stimuli, thereby making the task more a measure of STM storage than of WM/executive functioning. Indeed