Classical Conditioning: Mechanisms

advertisement
Classical Conditioning: Mechanisms
The general outline for this section:
I. What makes for an effective CS and/or US?
II. What is learned in classical conditioning?
III. Blocking and surprisingness
IV. Formal Models of Learning
I. What makes for an effective CS and/or US?
A. Novelty
CS-preexposure effect (or, latent inhibition)
Various explanations, depending on who you talk to
B. Intensity
C. Salience
D. Relevence
Remember Garcia’s studies?
Here’s a more elaborate version of it:
Phase 1
Phase 2
Group 1
Bright/noisy/tasty water
SHOCK
Group 2
Bright/noisy/tasty water
X-RAY
Test
X
A/V? 
Taste?

A/V?X
Taste?
Neither CS nor US salience could account for the idiosyncratic
results obtained….
Flies in the face of the “arbitrariness” of learning associations...
CS/US distinctions: The concept of “biological strength”
Pavlov was the first to propose a distinction between CS’s and US’s
e.g.: a light or tone does not initially possess much biological strength, whereas food or shock does
Low Biological Strength = CS’s; High = US’s
Also: higher strength of the US “energizes” learning
Implications of Pavlov’s notion:
1) Higher-order conditioning
-once a CS--US association is formed, the CS now has more biological strength
2) Strong—weak ordering should result in no learning
e.g.: food--light
Problems with Pavlov’s notion of biological strength:
1) CS-preexposure effect
-since no US, no learning should take place, according to Pavlov
2) Sensory Preconditioning
-learning does appear to occur with two “weak” stimuli
e.g.: light + tone
tone + food
light?
(notice that it is really higher-order conditioning in reverse)
So, while the existence of stimuli possessing biological strength is not debatable, where
it fits into the big picture of how learning takes place is still in question
II. What is learned in classical conditioning? (representations)
S-S
CS
US representation
representation
S-R
Response representation
The Evidence:
-Browne (1976): vicarious learning
-devaluation studies (e.g. Rescorla, 1973)
A typical example of a devaluation study (from Rescorla, 1973):
Phase 1
Phase 2
Test
E group
Light--loud noise
Habituate noise
Suppression to light?
C group
Light--loud noise
Don’t habituate
Suppression to light?
More suppression in C group than in E group
Serves as a test between S-S, S-R: if devaluation occurs, S-S supported
S-S
CS
US representation
Devalued response
to US
representation
S-R
Response representation
More devaluation studies:
Phase 1
gp. E
tone + food
gp. C
tone + food
Phase 2
Test
food + rotate
Tone?
rotate
Tone?
Found evidence of devaluation: rotation contingent on food showed less activity than uncorrelated rotation
Devaluation not restricted to rats, nor to illness as the devaluing technique
Phase 1
Phase 2
Phase 3
Test
E
Red--food
Green--Red
Red--000
Green?
C
Red--food
Green--Red
000
Green?
Evidence of devaluation (S-S): E-group pecks less than C group
However, not all devaluation studies support an S-S representation:
Phase 1
Phase 2
Phase 3
E
Light + food
Tone + Light
food + rotate
C
Light + food
Tone + Light
rotate
Test
Tone?
Tone?
No devaluation here: nonsignificant differences between the two groups
Phase 1
Phase 2
Phase 3
Light----000
E
Light + shock
Tone + Light
C
Light + shock
Tone + Light
000
Test
Tone?
Tone?
Again, no devaluation; tone is equally suppressive for the two groups
So, sometimes devaluation does occur (supporting S-S), sometimes it does not
(supporting S-R)
WHY?!?!
Potential explanation: Konorski’s distinction
Stimuli can be internally represented in more than one way:
1) Sensory properties ---- consummatory response
2) Affective/Motivational properties ---- preparatory response
light---food = a sensory code (“food!”),
tone---light = a motivational code (“something good”)
If you devalue the food, it will not change the representation pointed at by the tone.
(the light means different things in the two cases)
Reasons S-S is considered correct:
1) Konorski’s ideas provide an explanation for why devaluation did not occur,
but leave intact the idea of an S-S representation taking place
2) S-R proponents have no good explanation for when devaluation studies work!
3) S-R “support” = no devaluation. In other words, it is asserting the null hypothesis!
4) Browne’s vicarious learning study
Conclusions: S-R “support” is really a lack of evidence at all,
S-R cannot explain Browne, nor when devaluation works
S-S can explain Browne, when devaluation works, and even
when it doesn’t work (thanks to Konorski)
So, it appears S-S representations are the clear winners
Download