1. Conditional learning Charlotte Bonardi
C82NAB Neuroscience and Behaviour
Conditional learning
Charlotte Bonardi

3. A red light means....

4. A red light means....

6. JAM!!

7. JAM!!

9. newspaper

10. newspaper
journal

11. periódico
shinbun

12. Conditional control of associations
The same stimulus can be associated with two different outcomes.
Which association is retrieved is conditional on the context in which
the stimulus is presented.
This allows the associations to represent knowledge in a versatile way.
The context appears to control access to the CS-->US association.
spanish
“periódico”
“newspaper”
english

13. Conditional control of associations
Standard associative theories can't explain this -- the association forms and that’s it.
... maybe associative theory cannot explain all animal learning?
Questions:
Is conditional control really independent of associative learning?
If so, how does it work?

14. Conditional learning in animals
Feature-positive discrimination:
light+tone ---> food tone --> nothing

15. Conditional learning in animals
Feature-positive discrimination:
light+tone ---> food tone --> nothing
Associative theory predicts that the light is strongly associated with food ... and that the tone is not
On reinforced trials the light and the tone both acquire some strength;
on nonreinforced trials the tone loses it again.
As there is only limited associative strength available, eventually the light gets it all, and the tone is left with nothing.

16. Conditional learning in animals
Feature-positive discrimination:
light+tone ---> food tone --> nothing
Associative theory predicts that the light is strongly associated with food ... and that the tone is not
On reinforced trials the light and the tone both acquire some strength;
on nonreinforced trials the tone loses it again.
As there is only limited associative strength available, eventually the light gets it all, and the tone is left with nothing.
But how do you know this is what is happening?

17. Ross and Holland 1981
Used the fact that auditory and visual stimuli elicit different behaviours in the rat:
auditory: "headjerk" visual: "rearing"
They examined responding in two groups of animals:
1 light + tone --> food tone --> nothing
2 light + tone --> food light --> nothing

18. Ross and Holland 1981
Used the fact that auditory and visual stimuli elicit different behaviours in the rat:
auditory: "headjerk" visual: "rearing"
They examined responding in two groups of animals:
1 light* + tone --> food tone --> nothing
2 light + tone*--> food light --> nothing

20. The rats learned exactly according to the Rescorla Wagner model:
light* + tone --> food tone --> nothing
rearing to light
light + tone* --> food light --> nothing
headjerking to tone
But if you don't have the light and the tone simultaneous something else happens...

21. Ross and Holland, 1981
simultaneous

22. Ross and Holland, 1981
simultaneous

23. Ross and Holland, 1981
simultaneous
serial

24. Ross and Holland, 1981
simultaneous
rear
serial

25. Ross and Holland, 1981 simultaneous rear serial headjerk headjerk
In serial case hard for light to associate with food – so tone can

26. headjerk
headjerk
Headjerk response suggests responding is based on the tonefoood association and it seems to be stronger after the light!
tone
food
so is the light acting as a switch???
light

27. headjerk
headjerk
...or is it just associated with food too??
tone
food
Although the light is not closely followed by food, it is always followed by food delivery
Maybe the rats headjerk more to the tone
when the light has been on because they
are already expecting food, and this boosts
the headjerk CR elicited by the tone.
light

28. grub
up!

29. grub
up!

30. tone
food
light
We can test this by extinguishing
the light-->food association.
Will the rats headjerk to the tone when
it is followed by the light?
If they do, we cannot explain the results in terms of associative learning.

31. Holland, 1989 Two groups of animals:
Group FP: Lighttonefood, tone no food
Group PP: Lighttonefood, tone no food...

32. Holland, 1989 Two groups of animals:
Group FP: Lighttonefood, tone no food
Group PP: Lighttonefood, tone no food, light no food

33. Holland, 1989 Two groups of animals:
Group FP: Lighttonefood, tone no food
Group PP: Lighttonefood, tone no food, light no food
If the light-->food association is responsible for discrimination in Group FP, then Group PP should not show any discrimination.
This type of discrimination is called a positive patterning (PP) discrimination.

36. The light is still controlling responding to the tone, despite the fact that it’s not associated with anything.
In these cases standard associative theory cannot explain how the
light is controlling responding. The light may be called a conditional cue, a modulator, or an occasion setter.
Can get negative occasion setting too – turn OFF an association

37. SIM light & tone  nothing tone  shock noise shock
SER lighttone  nothing tone  shock noise shock
A normal Pavlovian inhibitor will suppress responding to the noise – a summation test
will this also happen in the serial case?
Holland & Lamarre, 1984

38. tone
shock
so is the light acting as a negative switch???
light
tone
food
or is the light just a Pavlovian inhibitor?
light
if you can explain the discrimination in terms of regular classically conditioned associations, associative theory can explain it – don’t need any other theory
but if associative theory can’t explain result, need something else --
then light is an occasion setter

39. suppression ratio

40. Rescorla’s modulation theory (Rescorla, 1985)
tone
food
light
Whenever a CS is presented, it must activate the US representation to get a conditioned response.
If the light is a positive occasion setter, it lowers its activation threshold -- making it easier for the CS to activate.

41. Holland’s and-gate theory (1983)
tone
food
light
The light acts as an and-gate, allowing activation to flow from the CS to the US, and so elicit a conditioned response.

42. These theories make different
predictions about transfer
Suppose you also pair a clicker
with food; would you expect the
light to elevate responding to the
clicker as well?
click
tone
food

43. click
tone
food
Rescorla’s theory says YES.
The light is altering the
activation threshold of the food, and will boost responding to any stimulus that is associated with it.
light

44. click
tone
food
Holland’s theory says NO
The light is acting as an and-gate for
the tone-->food association
-- not the click-->food association
light

45. Hundreds of experiments later...
Occasion setters do sometimes transfer their effects to other CSs paired with the same US
(ii) But this transfer effect is seldom as big as with the original CS
Many people have done this experiment (e.g., Holland, 1986; Rescorla,
1985). Here is one example, using rats:

46. TRAIN
light.....tone-->food tone--> no food
click-->food

47. TRAIN light.....tone-->food tone--> no food click-->food TEST
light.....tone? tone?
light.....click? click?

49. Holland might say...
the animals could confuse (“generalise between”) the two CSs
(the tone and the click). This might give a false and weak) transfer effect
click
light
food
tone
food
light

50. Rescorla might say...
animals are “upset” by experiencing novel CS combinations (such as the click and the light) -- this could disrupt responding, and transfer to a different CS.
click
light
food
tone
food
light

51. So get some transfer to the click (supporting Rescorla) but less effect than with the tone (supporting Holland)
results are not conclusive...

52. a complication!
(iii) Transfer is better if the transfer CS has its own occasion setter
click
tone
food
light

53. a complication!
(iii) Transfer is better if the transfer CS has its own occasion setter
click
dark
tone
food
light

54. light.....tone-->food,
tone--> no food
dark....click-->food
click-->no food

55. light.....tone?
tone?
dark...click?
click?
SAME
light.....tone-->food,
tone--> no food
dark....click-->food
click-->no food

56. light.....tone?
tone?
dark...click?
click?
light.....click?
dark...tone?
SAME
light.....tone-->food,
tone--> no food
dark....click-->food
click-->no food
DIFFERENT

57. Again, many people have done this experiment (e.g., Holland, 1986,
Rescorla, 1985). This is one example (it actually used pigeons).

58. Conclusions:
Everyone (including Rescorla!) agrees Rescorla’s theory is too simple; occasion setters specific to training CS
Holland’s theory a possibility -- but why better transfer to a CS from another conditional discrimination?
So Holland (1989) suggested an occasion setter will transfer perfectly to a CS that has its own occasion setter – not to a CS that hasn’t.
Circular...!!

59. Bouton & Nelson (1998):
occasion setters form best when tone can acquire strength more easily than feature...
light  tone  food tone  nothing
but tone always losing strength too.. both excitatory and inhibitory links with US:
tone
food
light
maybe light inhibits the inhibitory link!

60. Bouton & Nelson (1998):
problem: can get occasion setting even without tonenothing trials
light  tone  food tone  nothing
tone
food
maybe light inhibits the inhibitory link!
light

61. Bouton & Nelson (1998):
problem: can get occasion setting even without tonenothing trials
light  tone  food
tone
food ?
no inhibitory link to inhibit!
light

62. An alternative account ... configural learning
although they are given this....
they actually experience something like this...
tone after light

63. ..when the tone and the light are presented together, the animal experiences a “configural” blend of the stimuli that is different from either
This tone/light configure is associated with food in the normal way -- and that is what produces discrimination performance.
(and extinction of the light doesn’t affect discrimination – it’s a different
stimulus!)

64. It is difficult to discriminate between the and-gate and configural accounts of occasion setting -- but it seems the configural account is not enough to explain all occasion-setting effects.
Occasion setters can be contexts (places)
smells
drug states (state dependence)
maybe even apply to tasks like Stroop:
red red
green green
blue blue
task is occasion setter telling you which association to use – the one linked with the name of the word, or that linked with the colour

65. General references
Bouton, M.E. (2007). Learning and Behavior. Sinauer Associates.
Domjan, M. The principles of learning and behaviour. (1998). Brooks/Cole.
Advanced: Swartzentruber, D. (1995). Modulatory mechanisms in Pavlovian conditioning.
Animal Learning and Behavior, 23,
Specific references
Bouton, M.E., & Nelson, J.B. (1998). Mechanisms of feature-positive and feature-negative discrimination learning in an appetitive conditioning paradigm. In P.C. Holland and N.A. Schmajuk (Eds.) Associative learning and cognition in animals: Occasion setting. APA, Washington, D.C.
Holland, P.C. (1983). Occasion-setting in Pavlovian feature positive discriminations. In M.L. Commons, R.J. Herrnstein, & A.R. Wagner (Eds.), Quantitative analyses of behavior (Vol. 4, pp ). New York: Ballinger.
Holland, P.C. (1986) Transfer after serial feature-positive occasion setting. Learning and Motivation, 17,
Holland, P.C. (1989b). Feature extinction enhances transfer of occasion setting. Animal Learning and Behavior, 17,
Holland, P.C., & Lamarre, J. (1984). Transfer after serial and simultaneous feature negative discrimination training. Learning and Motivation, 15,
Rescorla, R.A. (1985). Conditioned inhibition and facilitation. In R.R. Miller & N.E. Spear (Eds.), Information processing in animals: Conditioned Inhibition.(pp ).
Ross, R.T., & Holland, P.C. (1981). Conditioning of simultaneous and serial feature-positive discriminations. Animal Learning and Behavior, 9,