Please enable JavaScript.
Coggle requires JavaScript to display documents.
Learning and the Brain Synaptic Plasticity (SIMULATED THEORY OF MEMORY-…
Learning and the Brain Synaptic Plasticity
IS LTP the basis of learning. 4 links between LTP and Learning
Detectability
- do neurons change in synaptic strength after behavioural learning?
Studies of Aplysia- have reflex when tail touched withdraw their siphon.
Use classical conditioning of reflex. Pair stimulation of tail (US) with stimulation of mantle (CS). leads to potentiated motor response (withdrawing siphon to CS alone)
Found differences in synaptic efficacy before, during and after learning. Learning-related changes would be blocked by NMDA receptor antagonists.
Mimicry
- if it's possible to create synaptic strength change artifically animals should have memory of an event that didn't actually happen.
Not possible to program real networks with artifical memories by changing synaptic weights
Anterograde Alteration!!!
- does blocking synaptic change prevent learning?
Lesions of cerebellum impair
motor learning,
but don't impair mechanisms of plasticity
But
lesions studies
are not specific as they destroy local and surrounding areas.
Lesions of hippocampus severely impair
spatial learning
Does blocking NMDA function impair spatial learning
Normally in Morris Water Maze rats spontaneously search for surface. use memories to find one and takes less time each trial.
AP-5 is NMDA antagonist- blocks glutamate from binding to receptor.
Morries et al (1989) Group 1 given saline, Group 2 given AP-5. implanted in hippocampus. Found Group2 AP-5 impaired spatial learning task compared to control. AP-5 affected memory hence rats spent less time in correct area of maze
Classical eye blink conditioning
During learning air puff (US) paired with auditory tone (CS) . Over trials starts to shut eye earlier (eye blink reflex) before air puff, because learns to use tone (CS) to predict air puff.
Even lesion studies found eye blink conditioning can still be learned after damage /removal of cerebral cortex, hippocampus, and forebrain.
Therefore is engram (physical manifestation of memory in the cerebellum? . When drug infusions switch off function in small brain areas = learning doesn't occur. During
acquisition
When inactivated CRs not expressed but return to normal when activated. Also when cerebellum inactivated during extinction, CRs not expressed but as normal learning decreases
Found when drug wore off animals learned from scratch, therefore drug did BLOCK learning not just learning in background
Long-term depression
- is decrease in efficacy between two neurons. May underlie motor learning. LTD is mediated by AMPA. If LTD and eyeblink conditioning share common mechanisms. Therefore blocking AMPA receptor should block learning-> found blocking AMPA prevents learning and LTD in cerebrum
Attwell et al (2001) control group with no AMPA antagonist (CNQX) acquired CRs normally. Groups where HVI lobule of cerebellar cortex was blocked -> no learning in phase one but after drug wore off learned CRs. Also other CNQX group acquired CRs slower than control group. conclude olivo-cortico-nuclear loop and HVI lobule are critical in classical conditioning motor response.
Retrograde Alteration
- by changing the synaptic strength changes from a past learning experience, it should change memory of event
If synaptic strength acquired during previous learning are changed -> does this cause forgetting?
Evidence that AMPA receptors are essential for LTD , CR acquisition and CR retention. (Attwell, 1999)
SIMULATED THEORY OF MEMORY- ARTIFICIAL NETWORKS FROM ARTIFICIAL NEURONS
Hebb (1949) when axon of cell A excites neighbouring cell B and repeatedly participates in firing there is growth/change in one or both cells increases A's efficiency at firing B
Similarly McClelland & Rumelhard (1986) when A & B simultaneously excited , increases strength of connection between them
Connectionism- memory can be encoded by changes in neuron's strength of connection/ ability to excite neighbouring neurons and changes depend on experience
In theory a network can be trained to change its responses.
Before learning- the primary neuron (A) stimulates the secondary neuron (B).
During Learning- neuron A stimulates another (B) but another supporting neuron also stimulates B.
With each trial- the secondary neuron (B) produces a greater response due to
experience of stimulation
from previous neurons. Despite receiving same level of stimulation
After learning- stimulation from supporting neuron not needed, primary neuron (A) is enough to keep B at high level
Long-term Potentiation
- long lasting increase in efficacy of synaptic connection between 2 neurons
One neuron stimulates the next
When there's low frequency stimulation there''s stable activity in secondary neuron
When applying at tetanus (large amount of stimulation) the EPSPs increase in size and high response is maintained
Associative LTP- when tetanus applied to strong stimulus = LTP . when tetanus applied to weak input = doesn't elicit LTP. When tetanus applied to both weak and strong inputs it produced = LTP in both. Therefore activity in both neurons communicate
4 properties of Synapses-
Rapidity (LTP produced quickly)
Cooperatively (LTP produced by single strong stimulation or several weaker stimulations)
Associativity (weak stimulation alone doesn't produce LTP, but simultaneous strong stimulation will)
Input Specificity- LTP elicted at one synapse does not spread to other syanpses
Connectionism- experience encoded in strength of connections between neurons.
example of Parallel Distribution processing- knowledge is shared across network rather than 1 point
Study cytoarchitecture of networks by creating simulated neural network (on computer) and neurons. to understand input and output
Backpropagation- when output is wrong its sent back up hierarchy to fix system. system is trained using inputs and outputs like neurons from Hebb's theory. The connections between input and desired outputs represent the strength/ability of inputs to activate target outputs.
Advantages of Connectionism-
Has biological realism, theory can be applied to actual biological phenomena
Networks naturally learn through trial and error not artificially programmed.
Cannot get rid a memory (even if structures destroyed the memory is not)
Learns mapping on its own
Disadvantaged of Connectionism - networks can learn ANYTHING with enough training but brains can't
Information is too simple (1s and 0s) not semantic information
Training model is time consuming
Networks have retrograde interference- forget learned material when trained on other
MEMORY IN A PETRI DISH- ARTIFICIAL NETWORKS MADE FROM REAL NEURONS
Artificial Networks of real neurons are grown in cell culture to studied in controlled conditions
Real neurons taken from embryonic rat brains. Found neurons spontaneously branch out and form synaptic connections with each other
Pros of artificial networks- can monitor and manipulate behaviour easier than real brain.
Put network in robot and can be trained to learn from experience
REAL BIOLOGICAL MEMORY- REAL NETWORKS FROM REAL NEURONS
Brain slices of small networks kept in culture . Found hippocampus important tole in memory formation
Hippcampal LTP found in synapses connecting prefrontal pathways and granule cells of dentate gyrus
activity increases after tetanus and LTP continues aslong as brain slice is alive
Mechanisms of LTP- understanding mechanisms of LTP help understanding of cellular mechanisms of learning
LTP increases efficacy of synapses by: more neurotransmitters released from presynaptic neuron. Also more recpetors avalilable for binidng on post-syanpstic neuron. Also retrograde messengers from post-synaptic to presynaptic instruct release of more neurotransmitter
NMDA receptors- LTP depends on this glutamate receptor. receptor contains ion channel for CA+ of NA+ ions. But channel blocked by magnesium. Pre-synaptic APs trigger glutamate release but not Ca2+ can enter because of magnesium block. When unblocked post-synaptic terminal depolarises. Pre and post terminal activated simulataneously allows LTP and ion channel work effectively
LTP and LTD found in In artificially grown networks of real neuron, in slices of real brain tissue, in the nervous systems of simple organisms suggests neurons in brain areas use Hebbian Learning