Purchase this article with an account.
Garrett Swan, Brad Wyble; The Binding Pool model of VWM: A model for storing individuated objects in a shared resource pool. Journal of Vision 2014;14(10):160. doi: 10.1167/14.10.160.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Two prevalent models that describe Visual Working Memory (VWM) assume that information is either stored in discrete slots or within a shared resource pool. To develop the theoretical landscape further, we propose a hybrid model called the Binding Pool model. This model details how multiple items can be encoded and retrieved individually yet interact with one another in a distributed binding pool using a Type/Token architecture. These processes use simple neural mechanisms that can rapidly encode arbitrary connections between different features (types), a location, and an object-file (token). These connections are stored by accumulating and storing simulated neural activity in a set of neurons called the binding pool. The model provides a unified framework for understanding VWM capabilities as measured by change detection and continuous report tasks. The Binding Pool model also provides a mechanism for explaining simple ensemble effects, such as the shifting of a stored representation towards another (Huang & Sekular, 2010). This arises because tokens share representational space in the binding pool, creating crosstalk between two stored items. The Binding Pool model can also generate predictions, which simultaneously test the validity of the model and may help to drive further research. One prediction of the model that was recently confirmed is increased precision in a directed forgetting paradigm in which participants are instructed to forget a specific stimulus. In a forgetting trial, the precision of the remaining stimulus is higher relative to a non-forgetting trial, but this precision is still lower than precision of a representation in a single item trial. (Williams, Hong, Kang, Carlisle, & Woodman, 2013). In the model, reducing the activity of binding pool neurons connected to the forgotten item, reduces interference during retrieval, which enhances precision of the remaining items. The model also predicts that encoding more features per item reduces precision.
Meeting abstract presented at VSS 2014
This PDF is available to Subscribers Only