Yas- CM

Lecture 1 Overview

• Reminder of Pavlovian conditioning.

• Demonstrating Pavlovian conditioning.

• Prediction Error reminder.

• Inhibition: when it happens, what it is, how to test for it

• Second-order conditioning vs. inhibition.

• Pavlovian conditioning is often good, sometimes bad: omission schedules.

Applications of Associative Learning

• Learning cause-effect relationships (Lecture 1).

• Attention and addiction (Lecture 2).

• Anxiety, phobia, and exposure therapy (Lecture 2).

Pavlovian Conditioning Components

• NS: Neutral Stimulus

• CS: Conditioned Stimulus

• US: Unconditioned Stimulus

• CR: Conditioned Response

• UR: Unconditioned Response

Example of Pavlovian Conditioning

• Initial Stage:

  • Metronome (NS) -> No salivation

  • Food (US) -> Salivation (UR)

• Conditioning Process:

  • Metronome (NS) followed by Food (US) -> Salivation (UR)

• After Conditioning:

  • Metronome (CS) -> Salivation (CR)

Control Conditions for Pavlovian Conditioning

• Purpose: To determine if CS-US pairings are crucial.

• Control setups:

  • CS with no US

  • US then CS

  • CS/US presented randomly

• Key aspect: keep all conditions identical to the CS-US experimental condition, except for the CS-US pairings.

What is Learned in Pavlovian Conditioning?

• An 'association' between the CS and the UR: S-R link

• An association between the CS and US

• Often conceptualized as a link in memory between the CS and US 'representations'.

S-R Link

• Bell (CS) -> Salivation (CR)

• Food (US) -> Salivation (UR)

S-S Link

• Bell (CS) -> Food (US)

• Food (US) -> Salivation (UR)

• Bell (CS) -> Salivation (CR)

Learned Fears

• Teddy (CS) -> Stab (US)

• Stab (US) -> Pain (UR)

• Teddy (CS) -> Fear (CR)

Example: Pleyers et al (2007) Experiment

• Procedure:

  • Participants see a brand followed by a picture.

  • Participants say ‘ahh’ for nice pictures and ‘yuck’ for nasty pictures.

  • Participants reflect on their expectations during each trial.

Learning Curves

• A graph showing the strength of the conditioned response (CR) over time.

• The y-axis represents the strength of the CR.

• The x-axis represents the number of conditioning trials.

• Asymptote \, of \, conditioning: The plateau where further training yields minimal increase in CR strength.

Prediction Error Model of Learning

• Learning happens when predictions are incorrect.

• A discrepancy exists between expectation and outcome (surprise).

• Early Learning:

  • Predictions often incorrect due to unfamiliarity.

  • Results in fast learning early in the learning curve.

• Late Learning:

  • Predictions largely correct.

  • Minimal learning occurs late in the learning curve.

Five Effects of Prediction Error

• Learning curves: A+ (Lecture 1)

• Inhibition: A+/AX- (Lecture 1)

• Blocking: A+ then AB+ (Lecture 2)

• Extinction: A+ then A- (Lecture 2)

• Protection from extinction: A+/AX-/B+ then XB- (Lecture 2)

• They all follow the principle of ‘prediction error’.

Causal Learning Example

• Scenario: A doctor trying to identify which foods cause migraines in a patient.

• Method:

  • The patient consumes particular foods each day.

  • The patient predicts whether a migraine will occur ('yes' or 'no').

Example Days

• Day 17: Food eaten - Apple. Migraine? Yes.

• Day 18: Food eaten - Pineapple. Migraine? Yes.

• Day 19: Food eaten - Nectarine. Migraine? No.

• Day 20: Food eaten - Pineapple and Banana. Migraine? No.

Prediction Error Example

• Pineapple -> Migraine

• Pineapple + Banana -> No Migraine

• Expectation: Migraine

• Experience: No Migraine

• Result: Negative prediction error leading Banana to acquire negative associative strength.

Summation Test

• Testing the inhibitory properties of a cue by combining it with an excitatory cue.

• Banana and Apple - Migraine?

• Nectarine and Apple - Migraine?

Summation Test Explained

• Apple causes migraine.

• Pineapple causes migraine.

• Banana inhibits the migraine caused by Pineapple.

• Inhibition by Banana transfers to the Apple.

Inhibition in the Associative Model (1)

• Apple -> Migraine

• Banana -> No Migraine

Inhibition in the Associative Model (2)

• Apple -> Migraine

• Banana -> No Migraine

Terms for Similar Things

• Excitatory CS-US association (e.g., Pineapple -> migraine)

  • The Pineapple CS has positive associative strength

  • The Pineapple CS causes the migraine outcome (US)

• Inhibitory CS-US association (e.g., Banana)

  • The Banana CS has negative associative strength

  • The Banana CS prevents the migraine outcome (US)

Alternative Inhibition Training

• Migraines occur almost every day.

• Migraines do not occur on days when the patient eats Pears.

• Conclusion: Pears prevent migraines, leading to an inhibitory association (Pears have negative associative strength).

Summation and Retardation Tests

• Summation Test: Apple -> Migraine, but Bananas reduce US expectation.

• Retardation Test: Pairing Bananas with Migraine leads to slower learning compared to Nectarine-Migraine pairing.

• Nectarine is a neutral control cue.

Second Order Conditioning

• Conditioning where a CS is paired with another CS, rather than a US.

• Cue A = Pineapple (excitatory stimulus)

• Cue X = Banana (inhibitory stimulus)

Within-Compound Associations

• Two kinds of trial:

  • Pineapple = migraine (A+ trials)

  • Pineapple + Banana = no reaction (AX- trials).

Two Opposite Things Occur

• Inhibition

• Second-order conditioning

When Inhibition and When 2nd Order Conditioning

• Non-humans (Yin, Barnett & Miller, 1994):

  • Large number of training trials = inhibition.

  • Small number of training trials = 2nd order conditioning.

• Humans (Karazinov & Boakes, 2007):

  • Lots of time to think = inhibition.

  • Little time to think = 2nd order conditioning.

• Inhibition requires some time and thought.

• 2nd order conditioning is just remembering what went with what?

Karazinov & Boakes (2007) Results

• X is an inhibitor (PX-/P+).

• M followed by no outcome (M-).

• T followed by outcome (T+).

Summary of Inhibition and 2nd Order Conditioning

• Banana -> no migraine US when the US is expected (due to pineapple).

• Inhibition: Banana predicts 'no US'. An inhibitory association can form with the US (negative associative strength).

• 2nd order conditioning: banana can bring to mind pineapple, which brings to mind the migraine US.

• Both processes happen. Second-order conditioning requires less training (less thought). Inhibition requires more training (more thought).

Usefulness of Pavlovian Conditioning

• Fear conditioning:

  • CR is freezing – less likely to be spotted.

  • CR is flight (increased adrenalin) – more likely to be able to escape.

• Conditioning with nausea: CR is a taste aversion – less likely to eat this in the future.

• Predicting the delivery of food: CR is salivation and other preparation for digestion.

Pavlovian Conditioning Can Be Unhelpful (Hearst and Jenkins 1974)

• Pavlovian conditioning can be unhelpful

  • Light

  • 60+cm

  • Food hopper

Omission Schedule (Williams and Williams, 1969)

• Light -> food delivery

• If the pigeon pecks the light, the light goes off, and food is not delivered.

• What should the pigeon do? What does the pigeon do?

Lecture Summary

• Excitatory conditioning often results in links between representations: CS can activate US representation via this link.

• Inhibition training results in a CS-US inhibitory link or CS-no US link.

• Learning of both kinds requires a prediction error (positive or negative).

• Initially, before inhibition develops, 2nd order conditioning is seen in the A+/AX- design, possibly due to associations between cues A and X.

• Pavlovian conditioning allows the organism to prepare for the US.

• Sometimes, the CR is unhelpful.

Reading

• Pearce (2008) Animal Learning and Cognition: An introduction, Chapter 2. In particular pages: 39-43; 46-55.

• Additional Yin, H., Barnet, R. C., & Miller, R. R. (1994). Second-order conditioning and Pavlovian conditioned inhibition: operational similarities and differences. Journal of Experimental Psychology: Animal Behavior Processes, 20(4), 419.

• Karazinov, D. M., & Boakes, R. A. (2007). Second-order conditioning in human predictive judgements when there is little time to think. Quarterly Journal of Experimental Psychology, 60(3), 448-460.