Abstract

In an attempt to investigate that individual differences in working memory capacity, as measured by the reading span test (Daneman & Carpenter, 1980; Lee, 1995), would result in different patterns of language processing, three experiments were carried out. In experiment 1, our high span and low span subjects were required to perform the dual task in which spatial inference from referentially continuous and discontinuous descriptions were performed concurrently with the memory task with different loads. In experiment 2, self-paced reading times were measured for the sentences with different relative clause constructions of SOV/OO and OSV/SO. In experiment 3 adopting Gernsbacher, Varner and Faust's (1990) paradigm to our subjects, reaction times were measured for accepting conceptually appropriate meaning of an ambiguous word as well as for rejecting inappropriate meaning. In all these experiments significant differences in the various tasks were observed between the high-span and low span readers. More important, putting the results of experiment 1 and 2 together provides the evidence that low span readers showed the marked processing cost whenever there was an additional load while high span readers did not. And putting the results of experiment 2 and 3 together provides the evidence that low span readers showed the spill-over effects during processing in both tasks of the sentence reading and the decision on accepting the appropriate meaning while the immediacy of integration occurred for high span readers. We discussed that these data patterns seriously question Just and Carpenter's (1992) assumption that high span and low span readers perform the same operations and that individual differences are inherent in the total capacity of working memory. Instead, it may be the case that high span readers have the function of working memory which is assumed to be capable of governing and operating the processing efficiency and the sophisticated processing strategies that low span readers lack.