Aligned neuronal encoding of sensory information, biases and choices in perceptual decision making
Rubén Moreno-Bote, Serra Hunter Associate Professor, UPF
Identifying what aspects of neuronal population activity are relevant for the encoding of information and choices is a crucial step toward understanding the neural code. Several statistical features of the neuronal population responses, such as tuning, synchronization and global activity could affect the amount of information encoded and modulate behavioral performance. Here we show, however, that only two of these features correlate wtih information: the length of the vector joining the mean responses across conditions and the inverse trial-by-trial variability of the responses projected along that vector. We find that modulations of the two statistical features are correlated with fluctuations of behavioral performance in various tasks. In contrast, modulations in mean correlations among neurons and global activity have negligible or no consistent effects on information encoding and behavioral performance. These results suggest that the neuronal representation of sensory information and choices are aligned. Interestingly, we also find that harmful, intrinsically generated behavioral biases are aligned with the choice representation in neuronal populations in the prefrontal cortex. I will describe a recently published sequential theory of decision making that could explain why these variables are represented along aligned axes in neuronal activity space.