A theory of associative learning- Rescorla-Wagner theory

need to predict and control world around them to survive

cause precedes effect in temporal relationship

• a stimulus CS to predict a stimulus US
• a repsonse R to predict a stimulus US

trained different groups of rats with a tone and shock in 4 conditions:

1. delay conditioning
2. trace conditioning (longer interval)
3. simultaneous conditioning
4. backward conditioning (shock first)

two conditions with shock afterwards had longest latency effect (most scared) so had learnt the most

1. postive - more shocks paired with tone than without
2. zero- same amount of shocks paired with tone as without
3. negative - more shocks without tone 

1. positive- evidence for excitatory learning
2. zero- no evidence of learning
3. negative - evidence for inhibition 

when US is shocking

1. light conditioned with pretrained noise (with shock)
2. light conditioned with novel noise

the group that had be pretrained with the noise blocked learning that the light caused the shock as they predicted the shock when they were presented together

change in associative strength between trials

salience of CS - intensity e.g. brightness of light

salience of US - e.g. how appetising food is 

size of US

sum of all the previous change in associative strength

how surprising the US is also known as prediction error - difference between what you expect and what actually happens

∆V = αβ ( λ - ΣV )

growth decreases between trials as it becomes more predictable

when change in growth of learning reaches asymptote, when US is no longer surprising

increase size of US e.g. more food

learning is faster as there is a faster association betweeen CS and US

as one of the preconditioned stimuli fully predicts US there is no prediction error when the second uncoditioned stimuli is presented

Equal numbers of pairings of light with shock, but in overshadowing groups a noise also present – either weedy, quiet n or loud N

• most learning in control and with quiet n
• When light low intensity did not really affect how much they learnt
• less with loud N as it overshadowed

the more salient one will have the higher strength but they will be equal if they are equally as salient

• change in associative strength is negative as US is taken away
• the omission of the US is what becomes surprising but in the opposite way

• CS presented in absence of US after it has become a strong association predictor

• one stimuli will predict US but nothing will happen when both present
• so both stimuli lose associative strength
• if one begins neutral it will end up negative 

• its negative associative strength subtracts from other predictors of the US
• And because it starts negative it takes longer to predict the US itself

the context can become associated more strongly than the specific stimuli

learning causal relationships