Purchase this article with an account.
Hyung-Bum Park, Weiwei Zhang; Hierarchical Bayesian Modeling for Testing Representational Shift in Visual Working Memory. Journal of Vision 2019;19(10):80a. doi: https://doi.org/10.1167/19.10.80a.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The recent visual working memory (VWM) research assessing the quality of retained mnemonic representations independently from the probability of remembering has primarily focused on the variance of internal representations. However, overall memory quality (i.e., the consistency in correspondence between internal representations and stimuli) could also manifest as the accuracy of mnemonic representation. The present study has thus assessed representational shifts, manifested as changes in the third parameter μ in Zhang and Luck mixture model of VWM representing the central tendency of noisy mnemonic representations, due to various experimental manipulations (e.g. attraction/repulsion between memory items, changes in inter-item context, representational momentum, etc.). Although experimental effects on representation shifts in these experiments are central for some hypotheses testing (e.g., serial versus parallel VWM consolidations), the numerical values of these shifts tend to be small. To adequately capture these effects against various sources of noise, we have developed a hierarchical Bayesian modeling (HBM) method simultaneously accounting for multiple sources of variance in recall error and reaction time. Formal model comparison is also implemented to test competing models generated from different theoretical features. The resulting posterior distributions of group-level parameters in HBM provide some strong evidence for representational shifts in these experiments. The advantages of HBM over some conventional methods will be further discussed.
This PDF is available to Subscribers Only