Abstract
Working memory span assesses the maximum number of items that can be remembered in the face of concurrent processing. Models of working memory differ on several dimensions, yet many rely exclusively on this span procedure for their evidence. Three experiments consider an alternative paradigm that attempts to capture the endurance limits for remembering a fixed number of items during concurrent processing. Eight-year-old children performed two versions of this working memory period measure—operation period and reading period. Period scores show healthy test–retest reliability and external validity for scholastic attainment, comparing well with span scores in these respects. In addition, period is highly correlated with span and shows similar effects of varying the order in which stimuli are presented. We conclude that the durability of representations is an important factor in both span and period.
Acknowledgments
We are grateful to the staff and pupils at the following schools; Emmer Green, Semley First, Shaftesbury, Abbey First, Gillingham Wyke, & Dallas Road. Matthew Hall helped to cross-check the accuracy of data, and anonymous reviewers provided useful comments. We acknowledge the support of the ESRC (Grant R000222789) for some of the empirical work. The Macintosh-based software used in this paper is available from the web at the following URL: http://www.lancs.ac.uk/staff/towse/wmperiod.html
Notes
In all the reported statistics, η2 values for multivariate F statistics are themselves multivariate, while values for univariate tests form partial η2values.
Individual scoring methods revealed general consistency over the two sessions for operations span; tolerant criteria, r(37) = .48, p < .01, medium criteria, r(37) = .51, p < .01, strict criteria, r(37) = .22, ns. Operation period showed consistency too; tolerant criteria, r(38) = .56, p < .01, medium criteria, r(38) = .47, p < .01, strict criteria, r(38) = .65, p < .01.
Considering specific scoring procedures, the highest overall correlation between operation period and scholastic attainment involved the strict scoring criterion, r(38) = .52, p < .01, contrasting with the marginal correlation under tolerant scoring, r(38) = .29, p < .10 (a comparison of the size of these correlations was marginally significant, z = 1.91, p < .10). With span scores, there was a nonsignificant (z = .49) differential among scoring methods between, at one extreme, attainment and medium scoring, r(38) = .50, p < .01, and at the other attainment and strict scoring, r(38) = .45, p < .01. The impact of scoring method is intriguing, but interpretation is constrained because models of working memory are generally not sufficiently well specified to account for why some criteria possess more predictive power than others. Understanding this effect, therefore, remains an important issue for future research.
Due to a publisher's production error, the test book omitted Question 14, which only came to light after testing commenced. Consequently, all assessments omitted this item, lowering the overall values but making little difference to the relative distribution of scores.
For period, scores improved significantly when measured by a strict criterion, t(58) = 3.96, p < .01, η2 = .213, and medium criterion, t(58) = 3.05, p < .05, η2 = .132, but there was no reliable change under a tolerant criterion, t(58) = 1.46, p > .10, η2 = .032. For span, specific criteria also showed equivalent performance across assessment (all ts < .070, η2 < .009).