ABSTRACT This thesis aimed to investigate how visual working memory takes advantage of long-term knowledge in order to allow semantic elaboration in the form of chunking and the role of the central executive in this process. Two leading theoretical frameworks of working memory which both emphasise the role of long-term memory are discussed. One of which views working memory as consisting of multiple discrete, modality specific subsystems (Baddeley, 2000) and one which views working memory as an activated subset of long term memory (Cowan, 2005). Both of these models propose the integration of short- and long-term representations to be attentionally demanding. To investigate this assumption, two forms of visual matrix pattern were generated; a high semantic set which lends itself to long-term memory support and a low semantic set which does so to a lesser extent. The initial block of empirical work aimed to establish the characteristics of the patterns sets. Superiority for the high semantic patterns was observed in terms of greater stability across increasing maintenance intervals. The benefit of increased presentation time was also shown to be greater for the high semantic pattern set indicating the importance of time in the semantic elaboration process. A second block of studies was then conducted to identify the implications of the two patterns sets for the functional architecture of working memory. In a secondary interference paradigm the pattern sets were shown to be differentially dependent on visual and verbal interference with low semantic patterns negatively affected by visual and not verbal interference and the opposite pattern observed for high semantic patterns. The use of executive and attentional interference paradigms demonstrated two levels of binding. Firstly, when attentional resources were continually captured by a secondary task, a degree of chunking was observed for both pattern sets, this is discussed in terms of passive binding on the basis of long-term knowledge in the absence of executive resources. In the absence of interference, effortful elaboration of the pattern sets is observed and this is greater for the high semantic patterns. This is discussed in terms of active binding with the involvement of executive resources. These findings are then discussed in terms of both the Cowan (2005) and Baddeley (2000) models and recent observations made by Baddeley et al (in press) to support a modification of the episodic buffer in Baddeley?s model to allow for both passive and active binding and how this leads to striking similarities between the two theoretical perspectives.
|Publication status||Accepted/In press - 29 Oct 2009|