Front | Back |
What is the learning-performance distinction?
|
Performance (behavior of an animal) depends on learning, but is not the same as learning.
|
What are the two characteristics of predictability in the Rescorla-Wagner model
|
1) how surprising or unpredictable the UCS is
2) the informativeness of the CS; how much the CS acts as a good signal or predictor for a UCS
|
What are the different features of CS informativeness?
|
1) Relative signal value; in the presence of several CSs, is one more likely to be selected than another? A CS that has become redundant with (or similar to) another already attended CS will not be likely to condition.
2) Contingency - Does the presence or absence of the CS predict the likelihood of the UCS?
|
1. What is the formula for measuring the change in the strength of an association? 2. Who is responsible for this formula? 3. What claim does this formula make?
|
1) DVi = ai bi (l - VPresentCues)
2) Rescorla and Wagner
3) how surprisingness and CS informativeness would predict the formation of an association
|
What is (triangle/delta) D Vi refer to?
What do the positive, negative in front of the symbol, or the zero in place of it indicate?
|
Change in strength of the CR. (change in strength of the association between the CSi and UCSj
The positive sign (+) indicates excitatation. The negative sign (-) indicates inhibition. The zero (0) indicates no association.
|
What do the subscripts i and j refer to?
|
I = CS
j = UCS
|
Why is this called the delta rule?
|
Because the formula is attempting to inform us about how much change there will be in the CR (or the association) on any given trial.
|
What does the VPresentCues refer to?
What is inferred from this?
|
It is the extent to which all the cues present on a given trial are already predicting the UCS.
We need to know how surprising the UCS is. If a UCS is adequately predicted, it is not surprising at all, and no learning should occur.
|
If we are actually conditioning multiple CRs when multiple Cues or CSs present, what will our formula have to be used for?
|
It will be used to determine the CR for each CS separately, and the individual CRs will be summed to find the overall reaction to the compound cues.
|
What is lambda? What could this value range from? Why can it be 0? What results in higher maximum potentials?
|
Lambda (l) is the value representing a UCS's maximum potential. The value ranges from 0 to the maximum potential. It is 0 during extinction trials, when the expected UCS is not there. Higher UCS intensity should result in higher maximum potentials.
|
What can the maximum potential be spoken of as?
What can this maximum level be regarded as identical to?
|
As the maximum association a UCS can form with a given CS. It can be identical to the asymptote of a learning curve.
|
What does this refer to? : (lj - VPresentCues)
|
The degree of UCS surprisingness = maximum level - the extent to which the UCS is already predicted.
|
What does ai and bj refer to? What are these values referred to, and why?
|
Alphai = noticability of the CS
Betaj = noticability of the UCS
Alphai and Betaj are referred to as rate parameters because they determine the rate that asymptote is approached, therefore, the speed of learning.
|
What does the graph in Fig. 1 predict?
|
It predicts the correct general shape of a learning curve (diminishing returns), and the CS salience effect. The more salient CSs exhibit stronger CRs than less salient CSs.
|
What happens to lambda during extinction?
|
Lambda becomes 0, because the CS is presented without the UCS.
|