Laboratory assessment of verbal working memory
Modern empirical assessment of working memory can be traced to Daneman and Carpenter’s (1980) landmark study, which used a reading span test that assessed college students’ use of both processing and storage capacity. Sets of sentences of approximately the same length were printed on cards, which were then read aloud by the students at a normal reading pace. Students were unable to view their previously read sentences. Presentation of a blank card signaled to the students to recite the last word of each sentence that was shown on the cards. Longer sets of sentences were presented in ensuing sets, until students failed all trials within a given set. Reading span was calculated based on the largest list size of perfectly recalled final words.
Tirre and Pena (1992) expanded on Daneman and Carpenter’s test by presenting sentences on a computer screen. In their task, respondents, who were US Air Force personnel, were required to answer “True” or “False” to each sentence before proceeding; this ensured meaningful processing and provided a measure of knowledge which was also correlated with working memory. Unlike the previous studies, respondents in de Jong’s (1998) study were children who were required to read sentences ranging from four to seven words. All these studies directly related working memory to reading skills, but Friedman and Miyake (2004) concluded that reading span tasks may be mediated by variables other than working memory capacity, such as time of processing.
Digit, letter, and word span tasks
Digit span tasks typically require the respondent to listen to verbally presented digits, and then recall them either in the same order in which they are presented (digits forward) or in backwards order (digits backward). Ramsay and Reynolds (1995) reviewed 27 articles on the separate scaling of the Digits Forward and Digits Backward subtests of the Test of Memory and Learning. Using factor analysis, Ramsay and Reynolds concluded that, despite similarities, the two tasks reflect different abilities. Similarly, Reynolds (1997) confirmed that the two tasks differ in that digits backward span requires the use of transformation (i.e., working memory), whereas the digits forward span does not. Despite Reynolds’ warning that “separate scaled scores for forward and for backward memory span tasks should be provided routinely on any standardized assessment” (1997, p. 39), raw scores on the two tasks continue to be summed on some commercially available assessments of working memory to produce a single standard score.
There is a multitude of other span tasks that are derived from the digits for- ward/backward tasks, such as letter spans, number—letter combinations, and various word spans. All of these require participants to respond selectively according to certain criteria: for instance, the letters in alphabetical order or words according to prescribed categories (e.g., body parts followed by non-body parts).