A Rasch-Validation Study of a Novel Speaking Span Task

Abstract

Working Memory refers to the capacity to temporarily retain a limited amount of information that is available for manipulation by higher-order cognitive processes. Several assessment instruments, such as the speaking span task, have been associated with the measurement of working memory span. However, despite the widespread use of the speaking span task, no study, to the best of my knowledge, has attempted to validate it using Rasch Measurement Theory. Rasch analysis can potentially shed light on the dimensionality of a complex construct such as working memory as well as examine whether a collection of items is working together to construct a coherent and reliable measure of a targeted population. This pilot study reports a Rasch analysis of a novel speaking span task, which was administered individually to 31 Japanese junior high school students and scored using a newly developed scoring system. Two separate analyses were conducted on the task: an analysis of the individual items using the Rasch dichotomous model and an analysis of the super items (sets) using the partial credit model. The results indicate that the task measures a coherent unidimensional latent variable and is thus a useful tool for measuring the construct. Moreover, Rasch analysis was shown to be suitable method for evaluating working memory tests.

The Construction and Validation of a New Listening Span Task

“…Second, although there are advantages to selecting the target words based on a corpus, such as a stricter control for word frequency and the elimination of a possible confound (i.e., the recall errors may be due to difficult word recognition rather than WM capacity), the intuitive approach used in the current study produced data that largely conform to the predictions of the Rasch model. Third, the results of the current pilot study lend support to the use of the scoring system corroborating the findings obtained by Bazan (2020). It is important to note, however, that this scoring system has the disadvantage that it requires participants to attempt all the trials (i.e., from Set 1 to Set 6), which may cause frustration once the task advances beyond the ability of the participants.…”

“…Each participant's performance was audio-recorded for scoring. The scoring procedure was adopted from Bazan (2020) and consisted of giving a credit for each word recalled in a string in the correct order of appearance until memory failure to recall in order. For example, if on a set of five utterances, participants correctly recalled the last word of the first and second utterances, failed to recall the last word of the third utterance (i.e., sanbanmewo oboete nai [I don't remember the third word]), but succeeded in recalling the last word of the fourth and fifth utterances, they were given two points (one for utterance 1 and another for utterance 2) and the rest of their responses did not count.…”