Poster F47, Tuesday, March 28, 8:00 – 10:00 am, Pacific Concourse
Rapid synaptic plasticity as a substrate for working memory maintenance
Eelke Spaak1, Christos Constantinidis2, John Duncan1,5, Timothy Buschman3, Earl Miller4, Mark Stokes1; 1University of Oxford, 2Wake Forest University, 3Princeton University, 4Massachusetts Institute of Technology, 5University of Cambridge
Recent theoretical models suggest that rapid, temporary changes in effective synaptic connectivity could play a crucial role in maintaining information in working memory. Short-term synaptic plasticity could thus provide a neurophysiological basis for WM maintenance in an ‘activity-silent’ form, which is energy-efficient and robust to interference. Despite theoretical appeal, directly testing such models remains difficult, as it requires the simultaneous recording of large numbers of neurons and their connectivity strengths. The latter is particularly hard to estimate in populations of sparsely firing PFC neurons. We now overcome these challenges by pooling data from a large archive of experiments involving simultaneous recordings from multiple neurons of non-human primate PFC. We have developed a novel, powerful method for the detection of synaptic connectivity from such large data sets, based on Bayesian inference for stochastic point processes. We demonstrate that our method has both a high sensitivity and specificity in recovering effective connectivity patterns. We leverage this novel method and the large amount of data to directly test the hypothesis of synaptic WM maintenance. Our results indicate an important role for rapid synaptic plasticity in the maintenance of WM, thus extending classical persistent-firing based models.
Topic Area: EXECUTIVE PROCESSES: Working memory