Accumulating evidences suggest that the cortico-basal ganglia-thalamic loops are crucial to action selection. Thalamic centromedian (CM) nucleus receives major inputs from the basal ganglia participating in internal drive-based action bias, brainstem processing sensorimotor signals, and the medial frontal cortices. Here we show that distinct classes of CM neurons exhibit activity which may monitor internal- and external-driven action bias and external demand for action.
We recorded activity of 106 CM neurons when two monkeys (Macaca fuscata) performed instructed choice (IC) task where pressing instructed buttons (left or right) were rewarded either a large or a small amount of water. Recorded CM neurons were classified into 3 types: those responded to visual and auditory stimuli at a long latency (LLF, n=29), those at a short-latency (SLF, n=5), and those without facilitatory responses to sensory stimuli (NSF, n=72). In the IC task, because of action-reward associations of two options were fixed for a block of 60-80 trials, monkeys could expect and bias toward an action with large reward in advance of instruction. During the pre-instruction period, some NSF (27/72, 38%) showed sustained activity when a particular direction (i.e., preferred direction) of action was associated with large reward, and negatively correlated with behavioral reaction times to the preferred direction, indicating action bias. In addition, action-reward instruction evoked phasic responses in both LLF (17/29, 59%) and NSF (27/72, 38%). In the instructed delayed choice (IDC) task in which action-reward instruction (CUE) was presented separately from succeeding action-trigger (GO) signal, many of task-related LLF exhibited strong phasic activation after CUE but only weak responses to GO signal (IC > IDC task GO, 13/17, 76%). On the other, NSF gradually increased their activity selectively after the CUE instructing the preferred direction that was consistent in the IC task.
These results suggest that NSF and LLF may monitor action bias and external demands under that bias, respectively, which could make a critical contribution to action monitoring and selection.