Abstract
Neurons in the primate lateral prefrontal cortex (LPFC) encode and maintain WM representations in the absence of external stimuli. Neural computations underlying spatial WM in primates are traditionally studied using highly controlled tasks consisting of simple 2D visual stimuli and require a saccadic response. Therefore, there is little known about how populations of LPFC neurons may maintain and transform 3D representations of space for animals to navigate towards remembered object locations.
To explore this issue, we created a spatial WM task that takes place in a 3D virtual environment. In this task, a target is presented in one of nine locations in the virtual arena. The target disappears during a two-second delay period, and then the subject is required to navigate to the remembered location using a joystick. Neural recordings were conducted in two male rhesus macaques using two 10×10 Utah arrays located in the LPFC (area 8A), resulting in a total of 3847 neurons.
Using a novel high-efficiency classification technique, we decoded the target location on a single trial basis during different trial epochs. This method resulted in high decoding accuracy using a minimum number of neurons containing the greatest amount of target-specific information. Ensembles of 8-12 neurons resulted in 60%-90% decoding accuracy (chance= ~11). Importantly, we were also able to decode trial outcomes (correct or incorrect). Furthermore, to elucidate the underlying mechanism in which neural ensembles encode and maintain target locations in 3D space, we decoded each 3D location based on depth and direction (i.e. front, middle, back, right, center, and left). The results indicate that target direction and depth were retained with similar accuracy by neural ensembles. Moreover, we show that LPFC neurons vary in the amount of information they maintain regarding either depth or direction of a remembered target in 3D space.