A computational and behavioral study of the precision of visuo-spatial working-memory for several items
CBN (Computational Biology and Neurocomputing) seminars
Friday 18 November 2011
to 11:00 at
Rita Almeida (Karolinska Institutet)
A putative neuronal correlate of working-memory (WM) is persistent activity, that is stimulus-selective, elevated neuronal firing observed long after stimulus offset. One proposed computational model accounts both for electrophysiologically measured persistent activity in monkeys (Compte et al. Cereb Cortex 2000) and for behavior and neuroimaging measurements from humans (Edin et al. PNAS 2009) acquired during a visuo-spatial WM (vsWM) task that requires memorizing non-foveal positions located on a circle. This computational model consists of a network of integrate-and-fire excitatory and inhibitory neurons organized according to a ring topography in terms of internal connectivity and external inputs received. The topography enables the model to sustain a line attractor mechanism for vsWM. This assumption of a continuum vsWM is essential for the model but there is not much experimental evidence supporting it. We addressed this issue by formulating two specific predictions from the computational model that were then confirmed in behavioral experiments. The first prediction is that the efficiency with which different simultaneously presented items are memorized depends on their relative locations. The second prediction is that the observed trade-off between WM precision and capacity (Bays and Husain Science 2008) also depends on the relative locations of the items.