Neurocomputational models of sequential discrimination

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): This application seeks to gain insight into the mechanisms by which short-term memories are held and interact with sensations in order to produce decisions. Data from monkeys performing a somatosensory sequential discrimination task will be analyzed and modeled computationally. This task involves both a short-term memory component (subjects must remember the first stimulus in each trial) and a decision component (the second stimulus in each trial must be compared to the first, and a decision must be made based on that comparison). Three specific aims will be addressed: (1) We will build computational neural models of the sequential discrimination task. Using these models, we will explore biophysical mechanisms by which the first stimulus may be loaded into short-term memory; preserved in short-term memory; and combined with the second stimulus, so as to form a decision based on the comparison of the two stimuli. We will explore how all three task components (loading, memory, and comparison/decision-making) may be combined and understood in a single biophysically-based neural network. (2) Based on the idea that the short-term memory of continuous variables is held in networks with dynamics that approximate a continuous attractor, we propose a model of noise correlations between neurons. This model fundamentally reconciles the precision of the neurophysiological representation of the memory of the first stimulus with the behavioral precision of the monkey subjects, as inferred psychophysically. We will analyze correlations between simultaneously recorded PFC neurons and seek evidence for, or against, this noise model. (3) The hypothesis that a decision regarding the comparison between the first and second stimulus is made in the PFC and/or the secondary somatosensory cortex (S2), with the PFC leading S2 in time, will be tested by analyzing data recorded from these cortical areas during the task. Our long-term goal is to understand how memory context can affect decisions based on incoming sensory data; that is, we seek an understanding of the neurophysiological basis of context-dependent decisions. In the long term, this research could potentially have significant impact on the understanding of mental health.