Learning and memory- Task 4- Classical conditioning and our dearest cerebellum ЁЯТХ
Unconditioned response
Natural response
Unconditioned stimulus
Natural stimulus that lead to a unconditioned response
Conditioned stimulus
Neutral stimulus that is constantly associated with US
Where is the pathway when pairing loud tone with the former reflexive action?
Now it's not only regulated in lower cognitive areas, it's also regulated by the cerebellum
If тАОс╡ж= 0.2 ==>
On each trial, the weight will change by 20 % EXPL: Trial 1 (no learning): ╬ФVlight = тАОс╡ж X prediction error 0.2 X 100 = 20 ----> Next time, the animal will have an expectation of the US of 20%
The Resc.- Wagner model to explain: Blocking
Once the animal learned the association V light= 100 --> Prediction error= 100-100=0 If prediction error is 0 --> ╬ФVlight= тАОс╡ж X prediction error= 0 YOU CAN T LEARN ANYMORE: there is no relevance for the tone to be added
Spontaneous recovery
Tendency for a previously learned association to reappear after a perfiod of extinction (it never disappeared inside)
Prediction error zero means that climbing fibers
can no longer lead to depression of synapses between parallel fibers and PC
3 assumptions of the mathematical model of Rescorla-Wagner
1) A cue has a weight that predicts the US 2)In compound cues, the weight is spread over cues ┬╡ 3) Learning corresponds to an increase in cue weight (accumulated over trials), and a corresponding reduction in prediction error. Without prediction error, no learning.
2 equations
1) Prediction Error = Occurrence of US - Expectation of US 2) Increase in cue weight = Learning Rate (тАОс╡ж) X Prediction Error
Rescorla- Wagner Model & Error- Correction learning
Amount of change that occurs in the association of a CS and a US depends on a prediction error
The pathway: pons ---> Purkinje cells:
carries a copy of motor commands, visual info...
Climbing fibers
innervate just a few PCs --> Very precise
The aripuff experiment and the conditioning effect, is simple the.... phenomenon
long term depression: just the opposite of long term potentiation: Ca triggers a decrease of neural connectivity (Purkinje cells)
Granule cells, with their parallel fibers make contact with
with hundreds of PCs (= Purkinje cells)
Maybe the inferior olive is where...
you learn when you make mistakes to correct them
Association weight
How strongly the CS predicts the US = How strongly do I know that if I hear a tone, food will come
Cerebellum: structural division
* Anterior lobe * Posterior lobe * Tonsil * Arbor vitae * 2 hemispheres around the vernis * Vernis= Spinocerebellum (in the middle) * Flocculus
Where do mossy fibers and climbing fibers interact?
* At the same Purkinje cells's dendrites (see above) * At the interpositus !!!!
Alternatives to the Resc. and Wagner model:
* Attentional approach to stimulus selection: CS modulation
Can people live without a cerebellum?
* Chinese woman who lived without a cerebellum (diziness, late to learn to speak...) * Young boy who can't speak/walk ---> Yes you can live without it but big imparement
Granular cell layer contains...
*Granule cells *Unipolar brush cells *Golgi cells *Mossy fibers *Climbing fibers
Requirement for an association to be made
*Usefulness *Nonredundant
Mossy and climbing fibers: Respective roles: Mossy fibers (left side)
*carry a plan *bring sensory information
Olivary nucleus receives input from
- the red nucleus (extrapyramidal motor system for involuntary movements, or modulation of voluntary movements) - Corticospinal tract (pyramidal motor system for voluntary movement) - Somatotopic proprioceptive and skin receptors
Mossy and climbing fibers: Respective roles: Climing fibers (right side)
...carry info on execution and error (feedback)
Cerebellar syndroms= Cerebellar ataxia
4 types of incoordination 1) Walking: lesion of the spinocerebellum 2) Limb movements:lesions of the hemispheres of cerebellum (if right hemisphere -> right part of the body SURPRISING) 3)... JUST TO GIVE ME AN IDEA ????
In inferior Olive: Reduction of response
= reduction of prediction error. When the prediction error decreases (the association becomes more and more certain) --> Inf. ol. nuclei inhibit the airpuff signals --> the climbing fibers are useless
Latent inhibition
A familiar stimulus takes longer to acquire meaning than a new one When a group has already been prexposed to the tone (CS), they learn way slowlier than a control group. ----> The model doesn't explain this phenomena because it should be that both prediction error are the same since no association was taught in either cases-> Learning rate shoudl be the same CONDITIONING CAN NOT BE ONLY EXPLAINED BY THE RESCURLA WAGNER MODEL
Extinction
A previously acquired association diminishes throught repeated presentation of the CS in the absence of the US (If the truck tone is heard several times but there is no ice-cream truck coming)
Kamin's blocking effect
Association is impaired if, during the conditioning process, the CS is presented together with a second CS that has already been associated with the unconditioned stimulus
Eyeblink experiment
Aversive conditioning US: Airpuff UR: Blink CS: Tone CR: Blink --> The rabbit blinks at the tone because predicts airpuff is coming
Example conditioned stimulus
Bell
Super important structure for classical conditioning
CEREBELLUM
Non redundant
CR can't be competing with another cue for the same US because its usefulness is blocked by the co-occuring cue
Usefulness
CS provides valuable new info that helps the animal predict the future
Cerebellum for language:
Cerebellum's role is to correct motor movement, with feedback from sensory info ---> Learning to talk really needs this
╬Фcue weight =
Change in cue weight
Where does the feedback that will lead to the inhibition of the olivary comes from?
Climbing fibers have the sensory info --> Inhibition of the Purkinje Cells -> Inhibition at interpositus -> Inhibition at the olive LOOOP
The Resc.- Wagner model to explain: Why removing one of two compound cues leads to weaker CR
Compound conditioning: Expected US is the sum of the weights of all the cues presented in that trial. So let's say you are trying to associate light and tone, with some food. TRIAL 1: It's the same ╬ФV light = тАОс╡ж X prediction error= 0.20 X 100= 20 ╬ФV tone = тАОс╡ж X prediction error= 0.20 X 100= 20 --> Expected US= 20 + 20 = 40 --> Prediction error= 100- 40= 60 TRIAL 2: It's different... ╬ФVlight = тАОс╡ж X prediction error= 0.20 X 60= 12 ╬ФV tone = тАОс╡ж X prediction error= 0.20 X 60= 12 ---> ╬ФVlight = тАО╬ФV tone= 20 + 12 = 32 If it would have been only light: ╬ФVlight = 36 (don't ask me why) BUT THAT DOES NOT MEAN THAT YOU LEARN FASTER IN WITH A SINGLE CUE: The individual weight rise more slowly in compound conditioning but since the overal expected US is calculated as the sum of them --> it raises way more quickly Even is V light = V tone = 50 ---> Prediction error= 100 - (50+50)= 0 It's completly learned
Prediction error
Difference between whether the animal expects the US and whether the US actually occurs
Ionotropic
Excite the activation of PCs on glutamate
Exposure approach to alcohol abuse
Exposure to the particular triggers that set off the problematic behavior. HOW? * We unlearn the association by presenting the cue without alcohol following (BUT IT IS HARD FOR HUMANS BECAUSE COMPLEX HISTORY OF ABUSE) *We reinforce alternative beh (cue -> drink water)
Posterior part of the cerebellum=
Flocculus: vestibular functions
Example unconditioned stimulus
Food
When you present the tone and then the airpuff; inside the PCs
From Pontine nucleus (CS), -> Mossy fibers -> glutamate you have increased activity in PCs and some ionotropic are tagged but not internalized Then from Olivary nucleus (US): parallel fibers activate these sleeping ionotropic receptors -> they don't work anymore -> only the metabotropic receptors are left, which results in decrease of activity in PCs (long term depression for PCs)
Output pathway
From Purkinje cells --> Interpositus nuclei (in deep nuclei) -> Red nuclei
Canonical circuit of the cerebellum=
General make up of the circuit that is abundant throughout the cerebellum
Prediction error=Occurence of US - Expectation of US
High prediction error if the US followed and you failed to predict it
Example of the assumption 2
If in trial 1: only light -> Vlight = Expectation of US If in trial 2: light and tone -> Vlight + Vtone = Expectation of US
Where is the cerebellum located in the cranium?
In the posterior cranial fossa
2nd time PCs receive input: US (airpuff) input pathway to the cerebellum
Inferior olivary nucleus (medulla obligada of brainstem) {VIA climbing fibers }: * -> Cerebellar cortex { VIA fibers connecting with Purkinje cells' dendrites} --> Aspartate: open Aspartate channels are open--> Calcium gets in -> increase in PKC and -> Glu ionotropic receptor that were tagged are removed from the membrane * -> Interpositus nucleus (in deep nuclei region) -> Red nucleus (above in brainstem) -> Down the spinal cord for motor info
Metabotropic
Inhibit the activation of PCs on glutamate Activate an enzyme cascade that internalize ionotropic receptors-> for the metabotropic receptors to be left alone and have complete inhibitory effect
Mossy fibers and climbing fibers
Interact at the site of Purkinje cells to... * modify or adapt ongoing movements (NOT MUSCLES) * on a more permanent basis particular movements (= learning)
LIMITATION OF THE MODEL
Latent inhibition ---> Need for alternatives: The Resc. & Wagner model is based on the US modulation
The different layers in cerebellum
Molecular layer Purkinje cells layer Granular cells layer White matter layer
Lateral part of the cerebellum hemisphere=
Motor planing, higher cog. functions
CS modulation theories
The way the potential CS is processed determines which of the CSs will be associated with the US TO EXPLAIN LATENT INHIBITION
How does the cerebellum contributes to motor learning?
Plasticity of the synapse between parallel fibers and Purkinje cells: When Purkinje cells are activated by the climbing fibers, all the parallel fibers that were recently active go under long-term depression --> Next time the specific parallel fiber is active, it will have a lesser excitatory effect on Purkinje cells (Because the AMPA receptors have been re-absorbed and so --> less power)
1st time PCs receive input: CS (light/tone) input pathway to the cerebellum
Pontine nuclei (in pons of brainstem) {VIA mossy fibers -> granules cells -> parallel fibers} to the Purkinje cells --> At parallel fibers: Glutamate is released -> Activation of ionotropic (AMPA) (+)and metabotropic receptors (-). + Metabotropic activated --> Protein kinases in cell is activated -> Enzyme cascade that tag (=phospholorize) the AMPA receptors that were just activated
The airpuff experiment at first: when no conditioning yet
Puff air --> Blink: Reflex Not regulated by higher cognitive areas (NOT IN THE CEREBELLUM)
Purkinje layer contains...
Purkinje cells
Interoreceptive
Receptors inside the body
Increase in cue weight =
Reduction of prediction error
Conditioned response
Response of the conditioned stimulus, anticipatory
Example conditioned response
Salivation
Example unconditioned response
Salivation
Extroreceptive
Skin receptors, outside the body
Molecular layer contains...
Stellate and basket cells
Blocking in humans: Experiment
Students have to categorize figures in class A, B,C... They make the association that all circles are class A... THEN another cue appears: a little dot that also indicates which class the figure belongs to by its position THEN they don't have the same shapes, only the little dot ))> They can't categorize the figures anymore because they never paid attention to this second cue
Error-correction rule
The Rescorla- Wagner Model and error- Correction learning says that we learn trial-by-trial, by making mistakes, to adjust/ correct the association between CS and US (= to reduce the prediction erro)
THE BIG DIFFERENCE BETWEEN CR AND UR
The action is the same but the CR arrives before the US because it's an anticipatory action
╬Фcue weight = Learning rate (тАОс╡ж) X prediction error
The change in cue weight (╬ФVcue) is equal to a constant (тАОс╡ж) multiplied by the prediction error
Withdrawal syndrom
The cues are present -> Compensatory responses (CRs) but the drug is not coming -> Imballance in homeostasis -> Craving to reach the balance again
Where are the Purkinje cells located?
Top surface layers of the cerebellum: Dendritic tree spreading in the outside part Soma in Purkinje cell layer Axon that projects to the inside of the cerebellum
Karmin's experiment
Trains rats that a light predicts a shock and then a light + a tone predict a shock --> Rats don't associate well with the tone alone to predict the shock because it didn't improve the ability to predict--> BLOCKING: the first association (light -> shock) was enough
IF ╬ФV cue < 0 -->
V cue decreases (you don't learn anything)
iF ╬ФV cue >0 -->тАО
V cue increases (you learn a positive association)
Central part of the cerebellum=
Vernis: for motor control of the limbs...
Mackintosh
We have limited capacity to process incoming info --> You can only pay attention to one stimulus. ---> If the "choosed" stimulus was not followed by any US repeatedly, you just ignore this stimulus--> You don't give a chance to any other stimulus
Appetitive conditioning
When US is pleasant
Aversive conditioning
When US is unpleasant
What happens between mossy fibers and climbing fibers?
When conditioning worked: At the olivary nucleus, you have an inhibition that matches the input ---> they cancel each other out --> no signals coming up from climbing fibers BUT IF THEY ARE NOT IDENTICAL --> Signals from climbing fibers -> Update of the plans -> Descending signal from Purkinje cells down to the spinal cord to update the movement LOOOOP
Arbor vitae
White matter/Axons that spread around to the different parts of the cerebellum
If weight of light (Vlight) = 100 --->
You believe that light will predict an airpuff in 100% of cases ---> You will close your eye --> No error, always close your eye
If тАОс╡ж is a big value (still between 0 and 1) -->
You learn fast --> There is a big gap between V cue between 2 trials => ╬ФV cue is bigтАО
If тАОс╡ж is a small value (always between 0 and 1) -->
You learn slowly
The higher the prediction error -->
You seem to not learn the association (but you have more opportunity to learn though)
Internalization of receptors=
You take them out of the postsynaptic membrane (they are eaten up)
Conditioned compensatory response
Your body knows that you are in a situation where you will probably receive adrenaline (which will increase your heartbeat)--> To keep homeostatis; anticipatory decrease in heartbeat --> THATS WHY YOU WILL NEED LARGER DOSIS NEXT TIME
A CS light would travel via......... than a CS tone
another subregion of the pontine nuclei
the interpositus always wants to...
fire;; always want to eyeblink!
Purkinje cells characteristics
flat --> so that the dendritic tree of neightboring PCs are in parallel planes
The salivation arrives before the actual...
food is seen
Interpositus wants to fire all the time (so would lead to blinking all the time), but is
inhibited by high activity in Purkinje cell (PC)
Unlearning an association (CS doesn't predict US anymore) takes
more time than learning an association in the beginning
Somatotopic maps in the...
pontine nuclei, inferior olive and in spinocerebellum (you can see like a homunculus in the spinocerebellum (=center): FOR THE INPUT Depending on where the input comes from (=which part of the spinal cord), different subregions of the olivary nucleus (but always stays inside the olivary nucleus) BUT FOR THE OUTPUT Depending on where the output needs to go (which part of the body), you will use different part of the deep cerebellar nuclei
The conditioned response...... the....
preceeds the unconditioned stimulus
If there is calcium inside the cell
protein kinases --> change inside the cell
Microzones in different regions of cerebellar cortex
that are associated with the same body site, are grouped together in a single multizonal microcomplex
Granule cells extend up to the Purkinje cells layer, to form...
the parallel fibers (axons split and can enervate many Purkinje cells)
The climbing fibers need to be activated around...
the same time that the mossy fibers were activated. Like that, the tagged AMPA receptors will be removed --> More metabotropic receptors -> For the same amount of glutamate: hyperpolarization of the PC s
The Rescorla-Wagner model illustrates that classical conditioning is a...
way of reducing error
тАОс╡ж
тАОLearning rate; constant between 0 and 1
Types of cells in cerebellum
тАв Purkinje cells тАв Granular cells тАв Interneurons - Golgi cells - Basket cells - Stellate cells
Cerebellum is ideally situated to exert motor control
тАв Receives sensori-motor input from association cortex, spinal cord, vestibular apparatus тАв Monitors from that input the appropriateness of posture, smooth goal directed movements, and contributes to error-correction тАв Motor learning in development тАв Skill learning in adults