THE MEASUREMENT OF VERBAL WORKING-MEMORY CAPACITY AND ITS RELATION TOREADING-COMPREHENSION

Ninety-four subjects were tested on the Daneman and Carpenter (1980) r eading span task, four versions of a related sentence span task in whi ch reaction times and accuracy on sentence processing were measured al ong with sentence-final word recall, two number generation tasks desig ned to test working memory, digit span, and two shape-generation tasks designed to measure visual-spatial working memory. Forty-four subject s were retested on a subset of these measures at a 3-month interval. A ll subjects were tested on standard vocabulary and reading tests. Corr elational analyses showed better internal consistency and test-retest reliability of the sentence span tasks than of the Daneman-Carpenter r eading span task. Factor analysis showed no factor that could be relat ed to a central verbal working memory; rotated factors suggested group ings of tests into factors that correspond to digit-related tasks, spa tial tasks, sentence processing in sentence span tasks, and recall in sentence span tasks. Correlational analyses and regression analyses sh owed that the sentence processing component of the sentence span tasks was the best predictor of performance on the reading test, with a sma ll independent contribution of the recall component. The results sugge st that sentence span tasks are unreliable unless measurements are mad e of both their sentence processing and recall components, and that th e predictive value of these tasks for reading comprehension abilities lies in the overlap of operations rather than in limitations in verbal working memory that apply to both.