Topic 5: Rescorla-Wagner Model

Question 1

General idea

Answer 1

Learning is proportional to surprise

Question 2

Rescorla-Wagner model

Answer 2

Change is associative strength between US and CS depends on how strongly the US is predicted on that trial.

Question 3

W

Answer 3

Maximum strength of association between US-CS

Question 4

A

Answer 4

Current strength of association between US-CS

Question 5

s

Answer 5

Salience. How noticeable/important is the relationship between US-CS

Question 6

W is set to 100/0 when…

Answer 6

Acquisition, US is present/Extinction, US is not present.

Question 7

Blocking definition

Answer 7

One US-CS association obstructs the formation of a second US-CS association.

Question 8

Why does blocking occur?

Answer 8

There isn’t much surprise left if another CS is added, doesn’t add much to the predictive power of CR.

Question 9

Overexpectation definition

Answer 9

When two CS are paired together, W decreases for each of the US-CS pairs.

Question 10

What happens to the R-W graph with overexpectation?

Answer 10

Drops off once they’re paired.

Question 11

Overshadowing definition

Answer 11

One US-CS association is going to be stronger than a second US-CS association.

Question 12

Why does overshadowing occur?

Answer 12

Higher salience for first pair, W will be higher for first pair as well.

Question 13

Limitation of R-W model

Answer 13

1. Some learning phenomena cannot be modelled
2. Each phase is discrete, single-step
3. No sense of time/history

Question 14

Examples

Answer 14

1) Acquisition: One CS/US pair, W = 100
2) Extinction: One CS/US pair, W = 0
3) Blocking: CS1 + CS2 + US, CS2 has a much smaller A then CS1
4) Overexpectation: CS1 + CS2 –> Diminishing W for both. Both start at A = 100 and same salience.
5) Overshadowing: S is higher for CS1/US pair than CS2/US pair, W will never be max for CS1/US pair