March 3 2014 V V t 1 V associative strength 1 to 1 0 0 means the animal knows nothing 1 means a perfect excitatory relationship has been learned and 1 means a perfect inhibitory relationship has been learned V Sum of associative strength How much you already know What you already learned change p l CS Salience 0 to 1 US Salience 0 to 1 f Perfect learning or asymptote is usually going to be 1 but in certain cases it could be 0 Rescorla model will attempt to address blocking overshadowing contingency cue to consequences latent inhibition practice salience and asymptote o What s wrong with the model Facilitated reacquisition and latent inhibition 1 Model assumes that learning increases until you understand what it is you re trying to learn 1 Discrepancy Resolution you learn more until there s nothing left to understand As you learn more and more there s less discrepancy between what you know and what you need to know Think of it as being surprised when you re surprised you learn and when you re not surprised you don t learn 2 The model also assumes that there s going to be positive and negative increment of associative strength it figured out 1 Associative strength strength between two stimuli Stronger association means you have 2 When you get more than you expect you have a positive increment of associative strength 3 When you get less than you expect you have a negative increment of associative strength 2 Learning is a limited resource process You can t learn about everything 1 Talking about overshadowing Let s say you go out to eat and eat a strong piece of fish and bland rice If you were to get sick you would blame it on the fish because the fish is a more relevant stimuli than the bland rice Stimuli compete with each other for associative strength ex the bland rice is competing with the fish for associative strength 2 Trials 16 trials in this experiment o o 1 8 were a tone paired with a shock 9 16 tested a tone by itself this phase is called extinction Acquisition Curve it s going to get shocked The top of the graph at 1 0 will be that it will be an expert at predicting shock Big difference between expert and novice Associative Strength surprised The curve will slowly reach the top In the first trial the rat gets shocked and learns a lot because it is Rat experienced shock and was surprised The rat wants to know when o Why does it look this way Is it going to be more or less surprised Less surprised because it now knows what to expect The next time it hears the tone it is going to know what is coming SO it is going to learn less on the second trial because there is less to know The tone will now predict nothing so the rat will have to change its Extinction Curve opinion of the tone The rat expects a shock and did not get it which is a big discrepancy because it is receiving less than expected Since there is no shock anymore it will need to start at the bottom again and this process will go on until the last trial Eventually the rat will disassociate tone with a shock He needs to go back to the bottom That goes on until the last trial and disassociates tone with shock 1 T S AS 5 2 T S AS 7 3 T S AS 8 4 T S AS 9 5 T S AS 95 6 T S AS 97 7 T S AS 99 8 T S AS 1 0 9 T AS 1 0 10 T AS 5 11 T AS 3 12 T 13 T 14 T 15 T 16 T The formula is designed to describe this process All of these variables in the model have exact correspondence that happens psychologically It s a formula that describes the curve V t 1 V Acquisition curve In order to do math we need to know that the salience of the CS 5 which is moderate strength Salience of US Beta 1 0 V 5 25 875 94 V 5 1 1 0 5 V 5 1 1 5 25 V 5 1 1 75 125 V 5 1 1 875 063 1 2 3 4 5 6 7 8 9 10 11 V 5 1 0 1 5 V 5 1 0 5 25 V 5 1 0 25 125 You won t ever get quite to 1 but you ll get really close For extinction change lambda to 0 because it knew nothing Change V to 1 because it expected the shock and was surprised V 5 25 125 In reality with extinction it doesn t happen as fast We re reluctant to give up the meaning The model fails to predicts what happens next Blocking Overshadowing Contingency Cue to Consequence Latent Inhibition Practice i e incremental acquisition We get better with practice Salience Intense stimuli are learned about better than weak stimuli Asymptote we stop learning March 5 2014 Cue to consequence certain stimuli are easy to learn such as food poison as a strong conditioning Rescorla and Wagner can accommodate that naturally Instead of light and shock let s use grape and poison In 1 trial asymptote is reached o o Grape CS Salience is 1 and Salience of US poison is also 1 So change of associative strength is equal to 1 cue to consequence effect Blocking vs no blocking Blocking Trial 1 Light shock 5 Trial 2 Light shock 75 Trial 3 Light shock 875 Trial 4 Light shock 94 o These are all the sums of associative change Trial 5 Light tone and shock No blocking First 4 trials no stimuli Trial 5 light tone and shock CS light salience is 5 US shock salience is 1 Change in associative strength for light is 5 so is change in associative strength in tone 5 for no blocking Blocking Trial 5 Change in associative strength for light is 03 while change in associative strength for tone is also 03 animal knows that the light causes shock this happens even though it never learned about tone simply because shock is associated with light Since it s known that light predicts shock adding another stimuli in this case tone doesn t tell you anything new about the shock that is why you only learn 0 3 about the tone Trial 6 of Blocking will show no additional learning occurs perfect learning has been achieved This is US processing model Intensity Overshadowing No overshadowing Trial 1 light shock 0 5 Trial 2 light shock 0 75 Trial 3 light shock 0 875 Trial 4 light shock 0 94 CS salience is 5 US salience is 1 Overshadowing Trial 1 light tone shock 0 5 Trial 2 light tone shock 0 5 Trial 3 light tone shock 0 5 Change in associative strength for light 5 1 1 0 5 same for tone Sum of associative strength is then equal to …
View Full Document