S. Newaz¹ (shah.newaz-1@ou.edu), M. P. Dungan², M. J. Wenger¹; ¹University of Oklahoma, ²Florida International University

When we temporarily store information about a multidimensional stimulus and keep it in an active state, are the features of that stimulus bound together or are they maintained independently? Most approaches to addressing this question have relied on operational definitions of working memory binding, allowing for the possibility of circular reasoning, rather than relying on theoretically-driven approaches. General recognition theory (GRT), a formal framework that explains multidimensional feature binding using three core constructs, provides a means to address this limitation. The violation of these three constructs---informational independence (II), informational separability (IS), and decisional separability (DS)---can be used to test for varieties of binding in working memory representations. In our previous study, we observed violations of II, inferences regarding which were limited by high marginal accuracy in our sample. To address this limitation, we used a modified n-back task to examine feature binding between color and pitch, with multiple levels determined by assessing each participant’s discrimination thresholds. This resulted in higher but interpretable error rates with marginal accuracy rates off ceiling and above chance. The analysis of marginal response invariance and the marginal signal detection measures revealed few if any violations of either IS or DS. In addition, analysis of report independence showed limited violations of II. These results indicate that the features of color and tone were generally not bound together in this n-back task and raises questions about the extent to which strong evidence for feature binding can be obtained across vs. within sensory modalities.

Acknowledgements: OU Undergraduate Research Opportunities Program