R-Computational Accounts of Decision-Making.pdf

Reading 13 January 2024 14:25 Source Notes The Neurobiology of Visual-Saccadic Decision-making Introduction: (Glimcher, 2003) How does the brain choose which movements to make from its repertoire? Understanding decision making, the connection between sensory data and action, has been a complex quest...

Reading 13 January 2024 14:25 Source Notes The Neurobiology of Visual-Saccadic Decision-making Introduction: (Glimcher, 2003) How does the brain choose which movements to make from its repertoire? Understanding decision making, the connection between sensory data and action, has been a complex question since ancient time s. Aristotle believed the nonmaterial soul controlled decisions. Descartes proposed two classes of behavior: deterministic sensory-motor links and unpredictable decisions relying on the soul. Sherrington focused on the first class, studying simple reflexes governed by the spinal cord. Recent progress has expanded studies to three classes of visual-saccadic decision making in primates. Historical roots of modern studies of decision making: Ancient scholars saw sensation and action as connected through the nonphysical soul. Galen proposed two systems, sensory and motor, connected by the soul in the heart or brain ventricles. Descartes divided behavior into reflexes and more complex decisions involving the soul. Sherrington emphasized studying reflexes as the simplest kind of decision making. Modern studies of simple decision making: Sensory systems were found to have highly specific receptors responsive to particular stimulus patterns. Motor systems, like the frontal eye fields and superior colliculus, control eye movements through topographically mapped neurons. These findings suggested visual-saccadic decisions could be studied using the Sherringtonian approach. Saccadic decisions about perceptual motion: Newsome's group studied how monkeys perceived direction of visual motion in moving dot displays. They identified the role of area MT neurons, each preferring motion in a specific direction. The probability of a rightward saccade matched the firing rate of rightward motion-preferring neurons in MT. Electrical stimulation of these neurons could alter the monkeys' saccade choices. Modelling the integrative element: Shadlen & Newsome's model described how MT signals might be transformed into saccade decisions. It proposed two sets of neurons integrating rightward and leftward motion information over time. Mutual inhibition ensured only one movement could be chosen. Area LIP was identified as a candidate for the hypothesized integrative element. Testing the model predictions: LIP neurons' activity matched the model's predictions: ○ Gradual increase or decrease during the stimulus period depending on the chosen saccade direction. ○ Firing rate growth proportional to the stimulus strength. ○ Activity even for random stimuli, reflecting prestimulus biases The Frontal Eye Fields and the Decision-Making Process Task: Monkeys make saccades to a single "oddball" target among seven distractors. Goal: Understand how the frontal eye fields trigger saccades based on sensory information. Method: Record neuronal activity in frontal eye fields while monkeys perform the oddball task. Findings: ○ Neuronal firing rate increases rapidly after target onset, then dips before gradually increasing again. ○ Firing rate divergence between oddball and distractor trials occurs around 80ms. ○ Firing rate reaches a common threshold before saccade onset, regardless of reaction time. ○ Stop-signal experiments suggest this threshold triggers saccades. A Model of Sensorimotor Decision Making Proposed model: ○ Visual cortices act as receptors, sending information to parietal cortex. ○ Parietal cortex integrates information across time, potentially via a competitive network. ○ Frontal eye fields receive this integrated signal and build activity towards a threshold. ○ Crossing the threshold triggers a saccade. Variability in Reaction Times How do different reaction times arise in simple tasks like the oddball task? Carpenter's experiments: Reaction times decrease when the target is more likely to appear in a specific location. Hypothesis: Neuronal activity in the decision-making circuit encodes the likelihood of different movements. Testing the Likelihood Hypothesis Gold & Shadlen's experiment: Monkeys perform a reaction-time version of the moving dot task. Prediction: Neuronal firing rate in frontal eye fields should reflect the log-likelihood of different saccades. Results: Stimulation-induced saccades shift towards the planned movement as the delay period increases. Conclusion: Neuronal activity in the frontal eye fields appears to encode log-likelihood ratios. Summary of Sensorimotor Decisions Sensory information is gathered by detectors and passed to integrative elements. Integrative elements, like LIP, encode the likelihood of different movements. Frontal eye fields build activity towards a threshold, triggering saccades when crossed. This model explains simple decisions based on accumulating sensory data. PSYC0032 The Brain in Action Page 1