Value-based decision-making involves multiple cortical and subcortical brain areas, but the exact nature of neural activity underlying choice behavior has been difficult to parse out in the human brain. Here, we use intracranial recordings from neurosurgical patients to decode trial-by-trial choice, showing that information related to expressed choices is contained in high frequency bands (gamma, high frequency activity) and is distributed across multiple brain regions, suggesting that distributed processes underlie human decisions under uncertainty. Furthermore, our results show that reward-based choices can be robustly decoded on a trial-by-trial basis from anatomically distributed iEEG signals.

18When learning to use a brain-machine interface (BMI), the brain modulates neuronal activity 19 patterns, exploring and exploiting the state space defined by their neural manifold. Neurons 20 directly involved in BMI control can display marked changes in their firing patterns during BMI 21 learning. However, whether these changes extend to neurons not directly involved in BMI 22 control remains unclear. To clarify this issue, we studied BMI learning in animals that were 23 required to control the position of a platform with their neural signals. Animals that learned to 24 control the platform and improved their performance in the task shifted from a global strategy, 25 where both direct and indirect neurons modified their firing patterns, to a local strategy, where 26 only direct neurons modified their firing rate, as animals became expert in the task. These results 27 provide important insights into what differentiates successful and unsuccessful BMI learning and 28 the computational mechanisms adopted by the neurons. 30Keywords: neural trajectory, posture control, brain computer interface, neuroprosthetic 31 65 type, the platform continues to tilt to an extreme position, a natural punishment. The task is 66 relatively natural and does not require pre-training, making it possible to study BMI control in 67 completely naïve animals. We used this task to study changes in performance of the BMI, which 68 was determined by the ability of our classifier to discriminate between four different tilt types on 69 a trial-by-trial basis. We studied the impact of learning, increases in performance, on the 70 underlying cellular and network properties of direct and indirect neurons. 71 We hypothesized that animals would use a global strategy when learning to control the platform 72 and, therefore, the learning mechanisms for direct and indirect neurons would be the same. 73While this was not what we observed, important differences between learners and nonlearners 74 and direct and indirect neurons suggest that the brain uses a global strategy early to initiate 75 learning but then switches to a local strategy, releasing indirect neurons from participation in 76 conveying information about the task 77 78 RESULTS 79 Animals improve performance by increasing differences between neural responses 80 The task involves postural control and bilaterally engages the cortex (Bridges et al., 2018) 81 (Deliagina et al., 2014; Jacobs and Horak, 2007). This allowed us to use one hemisphere for the 82 direct neurons and the other hemisphere for the indirect neurons. Direct neurons were used 83 control the platform as part of the BMI. The tilt was initiated and the neural activity within the 84 first 200 ms was used to decode the type of tilt (out of four possible tilts, see Figure 1A). If the 85 type of tilt was correctly classified, the animal was rewarded by having the platform return to its 86 neutral position; incorrectly classified tilts resulted in a punishment and the platform continued 87 to tilt to an extreme position. In...