Estimating invisible target speed from neuronal activity in monkey frontal eye field.

Working memory involves transient storage of information and the ability to manipulate that information for short-range planning and prediction. The computational aspect of working memory can be probed using dynamic sensorimotor behavior requiring complex stimulus-response mappings. Such a transformation occurs when extrapolating the future location of a moving target that is rendered temporarily invisible. Estimating the trajectory of an invisible moving target requires encoding and storing several target features, including the direction and speed of motion. We trained monkeys to make saccades to the estimated position of invisible targets moving at various speeds. The activity of neurons in the frontal eye field (FEF) was consistently modulated according to the speed of target motion. A reconstruction algorithm showed that estimates of target speed based on FEF activity were similar to behavioral speed estimates. FEF may therefore be involved in updating an internal representation of target trajectory for predictive saccades.