Systems Neuroscience CISC 325 Memory Types of Memory Declarative Non-declarative Episodic Semantic Professor Daniel Leeds dleeds@fordham.edu JMH 328A Hippocampus (MTL) Cerebral cortex Basal ganglia Motor cortex Cerebellum Basal ganglia Amygdala Cerebellum Cerebral cortex 2 Declarative vs. non-declarative Declarative Spring break ended on March 24 Apples are edible, chairs are not edible Non-declarative Throwing a baseball Pattern completion (seeing the dog behind the fence) Short-term vs. long-term Short-term aka working Hold facts in for -, seconds Sometimes prolonged version of perception Associated with prefrontal cortex (PFC) Long-term Stores facts over years Associated with hippocampus (also, amygdala) 3 4 Working Move eyes to target Delayed saccade (move eyes) to target task Neural in dlpfc for delayedaction task a: stimulus display onset b: stimulus display offset c: performance of action Funahashi et al. 989 5 Banana picture from Fir2/Flagstaffotos Neural dynamics in cortical sheet Cortical sheet: group of s on same level of hierarchy interacting with lateral connections Balance between local cooperation and local inhibition determined from h= + + 6
Neural dynamics in action V/IT PFC Neuron activated Neuron exciting Neuron inhibiting t= t=2 t=3 t=4 t=5 Trappenberg 7.3.2 Memory activity:balance of mutual excitation and mutual inhibition produces maintained sparse distributed coding Growing activity:mutual excitation produces global, non-stop activity over time epilepsy Decaying activity:mutual inhibition suppresses continued neural activity V 7 8 Anatomy of long term Hippocampus ( sea horse ) In medial temporal lobe (MTL) Input: Entorhinal cortex EC Dentate gyrus DG Cornusammonis CA, CA3 Perforantpathway: EC -> CA3 9 connections in CA3 Broader loop: EC -> CA3 -> CA->EC Memory through associator network Learning locations Rats learn neural representations of locations within a maze Hippocampal place cells in CA, CA3 tail back legs head front legs dog tail back legs head front legs Samsonowich, J Neurosci 997 Neurons organized in 2D based on similarity of tuning curves Connections between hippocampal s before (A) and after (B,C) learning 2 2
Further hippocampal representations Learning/remembering Grid cells In dorsocaudal medial EC Represent multiple locations Learning: neurogenesis in DG Retrieval: pattern completion in CA3 Alternate between learning and retrieval phases DG granule cells enable learning Perforant pathway probes (CC) Some rights reserved, Torkel Hafting 3 5 Modeling limits of working How much can we hold in working? 7±2 things Things can be simple Things can be complex A Q R L G Why is our working limited? Binding hypothesis: distributed code with synchronous spiking errors with spurious synchronization 6 7 On an earlier note Horizontal cells lateral inhibition in the retina Photoreceptor (rod/cone granule) activates horizontal cell Horizontal cell inhibits nearby photoreceptors (with GABA NT) Memory review Long term Short term On-center, off- surround retinal response 8 9 3
Declarative : long-term Remembering (over years): Your childhood house Your high school friends Your family connections in CA3 enhance associative ice cream ice cream 2 family high school home childhood family high school home 2 connections in CA3 Broader loop: EC -> CA3 -> CA->EC Declarative : short-term Remembering (over -2 s): List of numbers Set of pictures Δ = Cells that fire together, wire together Loop repeatedly increases weight increasingly encourage simultaneous firing 22 23 In an interconnected cortical sheet, neural activity can continue after feedforwardinput is gone maintain Neural dynamics in action Neuron activated Neuron exciting Neuron inhibiting t= t=2 t=3 t=4 t=5 feedforward input 24 25 4
Balance of positive and negative weights 5 5 2 25 3 35 4 5 5 2 25 3 35 4 26 27 Binding problem Neurons firing at same time represent same thing Pink Blue Cat 5 5 2 25 3 35 4 Dog Cow 2 3 4 5 6ms 28 29 Spurious synchronization Pink Blue Cat Dog Cow 2 3 4 5 6ms If spikes occurring within msof each other are considered synchronous, hard to incorporate increasing number of spikes in fixed time 3 5