hide
Free keywords:
-
Abstract:
Understanding how the brain represents, updates, and accommodates uncertainty is a key challenge for computational neuroscience. Neuromodulators such as acetylcholine (ACh) have been suggested as playing important roles over multiple timescales, regulating excitability and plasticity to mediate various effects of uncertainty on inference and learning. While these influences are consistent with ACh’s long-standing association with general functions of attention and arousal, recent studies using novel techniques to measure and manipulate this system with increasing exactitude have revealed intriguing patterns of activity at fast timescales. Notably, Sarter and colleagues used ACh amperometry (Howe et al., 2013, J.Neurosci., 33(20):8742-8752) and optogenetics (Gritton et al., 2016, PNAS, 113(8):E1089-E1097) as rodents performed a challenging signal detection task; they reported effects such as a serial dependency over multiple trials as to whether ACh is released, and increased false alarm rates when optogenetic stimulation is applied during non-signal trials. Inspired by their task and findings, we construct an abstract detection task, and consider how attentional state might be optimally controlled over each trial, assuming that more focused attention improves sensory information but incurs costs. We show similarities between the resulting attentional dynamics and task performance in the model and experimental results.
