Working memory theory proposed by Daneman and Carpenter in 1980 provided the theoretical basis for the measurement of working memory span. Based on it, various versions of reading span tests appeared. However, reliability and validity of these methods worth further discussion. Through an empirical study of 47 intermediate and high level Japanese learners, this research gives an insight into the key issue that whether semantic comprehension sector on material sentences have influence on test effects. The results demonstrate the necessity of mandatory constraints on the semantic comprehension of material
sentences in reading span test. Meanwhile, the experiment also proves that the credibility and validity of the reading span test developed by authors.