UC San Diego

A fundamental difficulty with producing language is that non-linear conceptual information must be expressed as a linear order of words. Consequently, speakers must often temporarily maintain ideas and lexical representations until they can be grammatically produced, a process that likely relies on working memory (WM). One way speakers might reduce demand on WM is by using syntactic structures that allow early mention of more accessible information (accessibility effects), thereby avoiding having to buffer ready-to-produce material. Speakers also use syntactic structures in which previously mentioned (given) information precedes new information. These given-new ordering effects are generally assumed to be communicatively driven, but might actually reduce to accessibility effects because given information is usually also highly accessible. Five experiments investigated the role and the type of WM involved in syntactic production by examining WM influences on accessibility effects and on given-new ordering. Experiment 1 showed that accessibility effects are sensitive to standard WM factors, suggesting these effects do, in fact, reflect WM processes. Experiment 2 showed that given-new ordering, like accessibility effects, is sensitive to WM factors, suggesting that it also arises from demands on speakers' limited capacity WM. Although these experiments show that WM demands influence syntactic production, they used a sentence-internal WM load, and thus do not distinguish between a form of WM specific to syntactic processing and domain-general WM processes. Experiments 3 and 4 manipulated WM load sentence-externally, and showed that speakers' use of accessibility effects and of given-new ordering are reduced when under an external load, thereby implicating domain general verbal WM in these processes. A potential concern is that the effects in Experiments 1-4 reflect not WM mechanisms but rather increased task difficulty unrelated to WM. Experiment 5 addressed this possibility by relying on the established distinction between verbal and spatial WM, comparing speakers' use of given-new ordering when under verbal and spatial WM load. A concurrent verbal WM load reduced speakers' use of syntactic structures respecting given-new ordering, but a concurrent spatial WM load did not. Together, these experiments show that both accessibility effects and given -new ordering reflect limitations in domain-general verbal WM