Abstract: Introduction
There are mainly three models concerning cognitive mechanisms underlying the limitations of working memory span, namely, resource-sharing model, task-switching model and time-based resource-sharing model. The resource-sharing model indicagtes that working memory span is constrained by limited cognitive resources. The task-switching model claims that memory decay is the main factor affecting working memory span. Finally, the time-based resource-sharing model suggests that working memory span is affected not only by limited attentional resource and memory decay, but also by switching mechanism. This experiment was conducted to test the resource constraint hypothesis, memory decay hypothesis and switching mechanism hypothesis proposed according to the three models.
Method
First, to test the resource constraint hypothesis, we manipulated the cognitive load of the processing component (hereinafter referred to as processing load) by designing two types of working memory span tasks with different levels of difficulty, one being continuous operation span task and the other de-de-de span task. In the continuous operation span task, some participants were asked to maintain consonant English letters while performing simple arithmetic operations, whereas the others were required to maintain the same letters while reading aloud the syllable de presented continuously in de-de-de span task. Second, to test the memory decay hypothesis, the retention duration was also manipulated and this was achieved by varying the number of sign-operand pairs in continuous span task and the number of de syllables in de-de-de span task. Under long retention duration conditions, the participants were asked to solve 4 sign-operand pairs of arithmetic operations and to read 13 de syllables respectively in the two span tasks. While under short retention duration conditions, the corresponding number was 2 and 7. Finally, to test the switching mechanism hypothesis, all the participants were required to complete a switching efficiency task before performing working memory span tasks. Switching efficiency was denoted by switching cost resulting from switching trials and repetition trials, and high switching cost meant low switching efficiency, whereas low switching cost meant high switching efficiency. 80 undergraduates and graduates were randomly assigned to the four working memory span tasks.
Results
Two-way (processing load × retention duration) analyses of variance on the span data revealed that, there was a main significant effect of processing load, with de-de-de span being predominantly greater than continuous operation span, and that neither the retention duration nor the interaction between processing load and retention duration had a significant effect. Correlational analyses of switching cost and working memory span indicated that there was no systematic relation between them.
Conclusions
The results suggest that working memory span is constrained by limited cognitive resources, and that memory decay and switching mechanism have no effect upon working memory span. Although the results supported the resource constraint hypothesis, further research needs to be done, for the nature of resource constraint remains to be an open question

Key words: resource constraint, memory decay, switching mechanism, working memory span