Week 10 Notes: What is actually learned during Learning?

stimulus Reponse Theory: Pavlovian Conditiong

  • when we present a CS there will be a cascade in the neural activity when learning is happening

    • before learning happens the CS will not evoke any conditioned response and only the US will evoke an unconditioned reponse

    • we expect no response

  • When we present a US, the US will evoke a cascade of neural/psychological activity

    • After the cascade we expect the ripple effect to leads to a UR

  • Both events cause changes in neural activity

  • We might be creating a link when learning is happening to result in a CR

  • After learning an association the CS will lead to a CR

    • perhaps learning will establish a new brain link between the CS and the UR

  • Stimulus-Response (S-R) Pavlovian Learning: During learning, a new neural pathway is created which between a stimulus (CS) and the neurons which control for the unconditioned response

  • S-R Theory Prediction:

    • CR should always be the same as the UR (since the links of the CR will leads to the pathway of the UR) * we know that this isn’t true

      • Food US: UR→ salivation CR → saliviation

      • SHOCK US: UR → jump CR→ freezing

    • No learning should ever occur without a UR

    • we know that this is also false due to examples such as sensory pre-conditioning where learning is still happening even though the animal is exhibiting no conditioned response

  • There are instance where the S-R association can exist!

  • Stimulus-Stimulus Pavlovian Learning

    • Definition: During learning, a new neurla pathway is formed between neurons directly activated by a stimulus (CS) and neurons normally activated by the event that follows a US. Thought of as being S-O (stimulus - outcome) association.

  • When the dog hears the bell then it will lead the subject to think about food

    • the bell doesn’t directly activate the salivary glands

  • Some associations are formed by S-S learning and S-R learning

S-R and S-S Learning

  • S-S Learning: mentally evoke the representation food which will lead to the activation of the salivary glands

    • The bell will leads to subject to think about the US which leads to the UR

  • S-R Learning: bell representation in the brain will lead to the activation of the salivary glands due to the connection of the neural pathways

    • The bell itself leads to the salivation (UR)

    • Direct short circuit

Associative Content of S-S and S-R Pavlovian Learning

  • S-R Association: Association is stored to act as an instruction, association is less flexible

    • Will not know why the response is taking place

    • No real thinking is require here

    • Faster to lead to a change in behavior

  • S-S Association: association conveys knowledge about what is going to happen next

    • Subjects learns that the CS is a predictor

    • More flexible to use

    • Provides knowledge of what is happening in the world

    • Contains more informations of what the actual outcome is going to be

    • More thinking and processing is going to happen

US Revaluation as an Index of What is Learned

  • q: how can we tell the associations apart?

  • US Revaluation Procedure: Post-learning changes to some property of the US ared used to probe the content of associations by testing for changes in conditioned responding

    • Animals could be showing the same type of response but the causes can be different for each of them

    • We can devalue the US (devaluation)

    • We can increase the value of the US as well

  • Test: we would give animals food → bad taste

    • If an animal shows a S-R response

      • We would expect the level of responding to remain the same since the US is activating the salivary glands themselves

      • Food is not activating the salivary glands

      • No change in the level of responding

    • If an animal shows S-S response

      • knows that tone → food and food→ illness

      • We would expect a decreasing in responding

      • CS is causing the animals to think about food and the response itself was caused by the mental representation of the US

      • There would be a change in responding

  • Instrumental : lever press → food

    • tone → shock pairings are given

    • Revaluation occurs : one group is habituated to the shock (can ignore) and another group gets no training to the shock

    • Test: measuring conditioned suppression

      • S-S: shock doesn’t both the group, no fearful response

      • S-R: hearing the tone will directly activate the fear response, the fear response should be happening anyway

  • we are changing the value of the US during both of these examples

  • Rats will get the same learning and some rats will learn S-R or S-S evaluations

    • We usually can’t guess what type of association is going to be learned

    • tools let us understand what type of associations are going to be learned

What is Learned during Instrumental Conditioning?

  • Stimulus → Response

    • Responding is independent of the identity of the reinforcer

    • Law of Effect

    • Habitual Behavior → environment will elicit the behavior

    • No sense in which the outcome is what is motivating the behavior

    • Consistent with Thorndike

  • Response → Outcome

    • Responding is based on beliefs and desires

    • Cognitive Theories of Learning

    • Goal directed behavior

    • Are what we typically think of when we think of instrumental learning

    • Subjects are aware of the potential outcomes and will be responding in way where they are purposely working toward the reinforcer

    • Goal-Directed Behavior (subjects are working towards a specific outcome)

    • Drive more complicated behaviors

  • Both types of associate are difficult to dissociated when seeing a subject act a certain way

  • Procedures used to identify instrumental associations

    • Plus-Maze Task

    • Reinforcer Devaluation 

      • CTA

      • Sensory-Specific Satiety

    • Contingency Degradation

    • Pavlovian- Instrumental Transfer Effects

Dynamics of Goal-Directed & Habitual Behavior: Plus Maze Task

  • Procedure: The rat will be placed in a maze where it will have the option to turn left or right and will only be awards with a food pellet if it turns left

    • Rat needs to learn that the location is associated with food

  • Q: What type of association is learned? S-O or S-R association? How can we test this?

  • We are able to do probe training, unexpectedly start the rat on the opposite end of the maze

    • Interpretation:

      • If the rat goes to the place where food was previously delivered then it will show a S-O association as it will associate the location with the reward

      • If the rat just turns left then the association is going to be S-R as it will automatically turn left

  • Early in training, behavior is goal directed (R-O system)

  • After extensive training behavior transitions to habitual control (S-R system)

    • Makes sense because with the more practice you have the more automatic a behavior becomes

  • The timing during which a probe test is conduct will can affect which type of association is being shown by the subject as the association can change over time

  • Animals will frequently learn both S-R and R-O associations

    • after responding becomes habitual, if we lesion the dorsal striatum the place strategy (R-O) will re-emerge

      • Dorsal striatum is critical for showing habitual learning

      • Shows that the previous R-O strategy still exists in the brain and is used by the subject when necessary

  • If we lesion the hippocampus without any training have happened then the subject will only learn habitual strategy

  • There is competition between brain regions for the control of behavior in which there's a transfer between goal-direct to habitual learning processes

  • Both Goal-Directed and Habitual Behaviors are useful

    • When there’s place strategy (goal-directed) the subject will be able to have more flexible behavior

      • ex. being able to readjust your way to work when there’s a roadblock

    • When the subject shows more of the habitual behavior, it allows for faster and more efficient behavior that is rigid and less flexible

      • ex. automatically driving your usual route to work

  • The transition between goal directed behavior to habitual behavior shows the best of both worlds

  • Good to have both of the association to lean on

Distinguishing Goal-Directed & Habitual Behavior: Reinforcer Devaluation

  • Three main steps

    • Training

    • Devlauation

    • Test

  • 1st: Training

    • Teaching two pavlovian associations

      • Two actions lead to two outcomes (two different types of food pellets)

        • One lever leads to grape food

        • one lever leads to cherry food

  • Devaluation Test:

    • Sensory Specific Satiety: feed rat a lottt of food pellets (have the animal full on one of the food pellets)

    • Conditioned Taste Aversion: have the grape pellets be associated with illness

  • Test:

    • Look at how frequently the rat will show both of the trained outcomes

    • Habitual: we seen that the frequency at which a rat presses both of the lever stays the same

      • Just seeing the lever leads to the rat to press them

    • Goal Directed: we see that the rat will decrease the frequency of the food outcome that was devalued while the the non-devalued lever will have a higher lever pressing rate

      • Shows that the subject will work to earn the outcome it desires

  • The amount of instrumental training will affect whether behavior is under habitual or goal-directed control

    • We can see that when we conduct this type of test when the subject has more experience it will show more results of equal amount of responding to both types of reinforcers since it will show more amounts of habitual learning behaviors

Distinguishing Goal-Directed & Habitual Behavior: Contingency Degradation

  • If the goal is goal directed

  • Same type of training procedure as the reinforcer devaluation

    • A reinforcer will only be delievered when the subject presses a specific type of lever

  • Contingency degradation:

    • For one of the reinforcer: The frequency of getting the reinforcer is the same whether or not the rat will press or not press the lever

    • Pellet is delivered during their lever press breaks

  • Test:

    • The rat will have the option to press a lever

  • Goal-Directed: We expect that the rat will dectrease their lever presses for the reinforcer that is always being delievered since it should know that the food is going to be delivered anyway

  • Habitual: we would see the rat continue to press both of the levers regardless of the contingency degradation

  • We see similar results for

    • test of habitual control

    • Pre-exposure to drugs

    • Pavlovian devaluation

  • Suggests long-term exposure to drug of abuse strengthen habitual control of behavior

  • contingency degradation is the instrumental version of truly random control 

    • changing how tightly a subject’s actions are equal to the probability of the subject getting the outcome

Pavlovian-Instrumental Transfer

  • Wtn to see if there are any moments where a pavlovain association will change the way an instrumental baehvior si carrried out

  • ex. Pavlovian association (with an aversive stimulus, shock) will terminate the instrumental behavioral outcomes

  • A CS associated with an aversive outcome causes a decrease in instrumental behavior

  • Pavlovian-to-Instrumental Transfer (PIT): A pavlovian CS can increase or decrease ongoing instrumental behavior

  • A CS+ increased instrumental behavior

  • A CS- will not affect instrumental behavior and can at time decrease instrumental behavior

  • Pavlovian Instrumental Transfer success will shows that Pavlovian associations and instrumental associations were not S-R

    • we wouldn’t expect an increase/change in the behavior if the associations were S-R since it wouldn’t cause the subject to think about the reinforcer when the pavlovian association is shown

General Pit: a pavlovan association between a CS and the motivation properties of the US ( appetitive/aversive) energizes or de-energizes instrumental responding

  • only established one pavlovian association so we wouldn’t be able to determine specificity to instrumental learning

Specific PIT: A pavlovain association between a CS and specific representation of the US eneegizes instrumental responding reinforced by the same US

  • ex. training three different associations (three separate pavlovian associations), two instrumental learning responses

  • We pair both of the pavlovian associations with each of the instrumental learning response in which we will see that each of the pavlovian associations will have specific effects on the response rate of the lever presses for a specific reinforcers

  • There is selective enhancement of the specific instrumental learning

  • We have the largest amount of evidence that the Pavlovian association is S-S and the instrumental learning association is R-O

  • For the general behaviors we see that the change to the instrumental responding is not specific and the pavlovain association will cause the same amount of change

  • Specifc PIT depend on an R-O (goal-directed) instrumental association and an S-S pavlovian association

  • General PIT: A pavlovian association between a CS and the motivational properties of the US energiezes instrumental responding

    • good leads to good

    • easy to obtain these types of results

    • animals aren’t purely using S-R association

    • points in the direction of not using S-R

  • Specific PIT: Pavlovian association between a CS and a specific representation of the US energizes instrumental responding reinforced by the same US

    • more energy is allocated a specific response

  • PIT provides a useful tool for analyzing the content of instrumental and Pavlovian associaiton