Rotation 1: The Computational Mechanism of Visual Working Memory Retrieval

BBSRC strategic theme: Biosciences for an integrated understanding of health

The proposed project will focus on computational modelling of visual working memory (VWM) retrieval. The analyses will utilise existing datasets, which employed the cued-recall task. The subjects were required to remember colour and location features of simple visual items. Following a retention interval, one of the features probed a single target item, and the subjects reported a value of the second feature. The key variables of interests include retrieval times, response precision, and components of error distributions. The behavioural patterns will be modelled using the multiple-alternative drift diffusion model, in which sequential sampling of evidence occurs in the location and colour feature spaces. Furthermore, framing the retrieval mechanism as an accumulation-to-bound process allows to draw a parallel between VWM retrieval and visual decision making. For example, altered numbers of visual items may be tested as equivalent to varied task difficulties in visual perception. Moreover, manipulating the speed-accuracy tradeoff can be directly compared to well-evidenced effects in visual decision making. Additionally, utilising opposite conditions in terms of cue and report features will enable us to dissociate the effects of cue-to-item mapping and report-feature readout processes onto the behavioural trends. Taken together, this project will formulate competing mechanistic models to investigate whether the cue matching and report selection processes are parallel versus serial and to explore further important aspects of the sampling mechanism (e.g., using shared versus independent sources of information across the visual feature spaces). Lastly, the project will explore whether the effects of experimental manipulations onto drift diffusion parameters are analogous to canonical findings in visual perception.