Lecture 6 Flashcards
1
Q
What is stimulus generalisation?
A
- When we learn individual things, we respond to more than the thing we learned about - we generalise e.g pigeon autoshaping pecking light and food
- When lights of different colours are presented, the response drops the further you go from the wavelength in a lawful fashion as proportion of drop off is same regardless of what colour you first use
- Pattern of dropping and maximal response is the generalisation gradient, and the difference between two points on this is called generalisation decrement
2
Q
What is the counteract of generalisation?
A
- If you generalise too much, you cannot predict accurately
- This is called discrimination
- Generalisation is useful but not too far, and needs to be inhibited at some times
- Selectively learning what is and isn’t followed by reward = restricting generalisation
3
Q
What is a study looking at discrimination?
A
- Dog knows a bell means food and is drooling
- Banging a triangle = dog generalises = but food does not come so invalid cue
- Over time, dog will discriminate
- As conditioning occurs, we see an increase in response in both CS. But the CS- is lower = shows generalisation, but CS- drops off whereas CS+ increases = shows discrimination
4
Q
What can animals discriminate?
A
- Anything they can perceive a difference between two stimuli = can teach them discrimination
- Random squiggles were randomly set up as a CS+ and CS-, pigeons learned which squiggles to peck = as long as they can see and perceive
- Wide range of species and stimuli are vast for discrimination
5
Q
What is the elemental theory?
A
- CS+ is made of lots of elements, each of which has its own associative strength, same for CS-
- Red circle followed by food, black circle not followed by food
- Element is about colour and shape = could be reconceived red+circle/black+circle or AX and BX where X is the circle as they are common and A/B are unique elements
- Red circle would have an increase in associative strength = red and circle element gain associative strength
- Black circle = error term is negative but there is learning = black element will lose associative strength
- Keep doing trials until asymptote = red element will gain Associative strength, black will lose, circle will fluctuate but will stay at 0
6
Q
What is the problem with the elemental theory and Rescorla Wagner?
A
- Rescorla Wagner as an elemental model has implications that if you learn about two things separately, when you present them together, the sum of associative strength must be bigger than the separate responses
- This should be impossible and is called negative patterning - if animals learn A and B separately, they should not be able to eliminate AB
- Tone followed by food, click followed by food, tone and click together = no food, tone + click together gets smaller than the tone/click on its own
- When you put things together = creates a configuration cue which is a combination of other elements = playing tone and click at the same time = there is an extra element for the animals
- e.g orchestra creates something beyond each one, not just a summation
7
Q
What is a lack of catastrophic interference?
A
- Teaching red triangle followed by food, red+blue triangle no food, red triangle has associative strength, red/blue = neg associative strength, blue triangle should have negative strength
- Blue triangle by self is paired with food, when we go back to original discrimination, blue neg association should be interfered with by the extra training of B
- Stage 1: discrimination learned
- Stage 2: response acquired to B
- Stage 3: original discrimination not abolished = configural cue idea does not help (did damage slightly though)
8
Q
What is the similarity and discrimination learning?
A
- Discrimination should be hard when stimuli are similar and easy when different
- Combo of three colours no food, one colour and two colours = food
- Discrimination between one colour and 3 is easy, but two colours and three is harder as it is similar
- Rescorla Wagner says the opposite = easier when overlapping = WRONG, proved experimentally
9
Q
What is the evaluation of Rescorla Wagner Theory?
A
- Precise predictions
- Explains a wide range of phenomena
- Important step toward understanding discrimination learning
10
Q
What is the exemplar theory?
A
- In LTM = representation of everything
- Animal with very limited experience learns that tone, and stimulus C is followed by food, but stimulus D is not followed by food
- What you see now is in STM, and you look for the similarity in STM and LTM, and if STM overlaps with something/s in LTM = whatever it overlaps with will activate the relevant response
- If overlaps perfectly = strong conditioned response
- If overlaps a little bit, activates LTM but not very well = moderate response = proportional
- Explains overshadowing
11
Q
How to determine strength of conditioned response?
A
- LTM stores a set of associative strengths
- Overlap is a similarity term
- Conditioned response = Similarity between STM and LTM multiplied by associative strength stored in LTM
12
Q
How to determine similarity?
A
- Counting up no of elements that are in both LTM and STM, squaring it, dividing it by no. of elements in STM x No of elements in LTM
- Similarity is symmetrical
- Adding something to a stimulus causes a generalisation decrement
13
Q
What does configural theory do that elemental theory cannot?
A
- Rescorla Wagner = first trial associative strength is 0 = can explain overshadowing if you have multiple trials BUT Can do overshadowing in one trial via Pearce Model
- A+ AB- Discrimination: When present A on its own = overlaps perfectly so large CR, but only partially similar to AB so smaller CR, and vice versa when AB presented in STM
- Negative Patterning: A in STM perfectly activates US and only partially in AB (no US) so larger CR than no CR
- Requires a lot of summation = isn’t problem for Pearce theory = he predicts it wont but it sometimes does happen
