NOTES ON WHAT CAUSES ASSOCIATIONS DURING CLASSICAL CONDITIONING [Sheet 17 of paper handout]

Pavlovian conditioning is interpreted as the formation of an association between two stimuli.

Usually the “unconditioned stimulus” (US or UCS) reliably elicits a response before the experiment (this is known as the unconditioned response: UCR or UR) and the association is measured by the degree to a new event (the “conditioned stimulus” or CS) comes to elicit the same response during the course of the experiment.

The simplest rule to explain association formation is that the association is increased whenever the conditioned stimulus precedes and overlaps with the unconditioned stimulus — the rule of contiguity.

The simple rule of contiguity is inadequate: current theories relate strength of association to the predictive properties of a signalling stimulus (Rescorla, 1988; Lieberman, 2000).

Three experimental results support this —

Phenomena showing that stimulus-pairing (contiguity) is not the only factor in Pavlovian conditioning. [bottom of page 17]

1. Contingency. If a bell is paired with food 10 times during one hour, the contiguity rule would imply that the bell should elicited conditioned responses (e.g. salivation). But if, during the same hour, food has also been given 20 times without the bell, or the bell has sounded 20 times without being followed by food, then the bell will not be a reliable predictor of food, and experiments suggest that the strength of conditioned responses usually varies according to the reliability of the conditioned stimulus as a predictor of the unconditioned stimulus. (See Gleitman, 1999, pp. 141-3 or Gleitman, 2004, pp. 144-146; Lieberman, 2000, pp107-115; a recent paper confirming this point for rats' anticipations of food pellets is Murphy and Baker, 2004 gif pdf )

[CONTIGUITY = just the number of pairings matters.

CONTINGENCY = the meaningfullness of the pairings matters, not just the number of pairings]

See Gleitman, 1999/2004, pp. 141-3/144-6; Lieberman, 2000, pp. 110-12

2. Blocking. Even if a signal predicts food reliably, there is little conditioning effect if it is redundant, because an alternative reliable signal has already been used. E.g. if a light has been the signal for food 100 times and then the light plus a sound is used as a compound signal 100 times, testing with the bell alone shows little conditioning (but it does if the initial trials with the light are omitted. (See Kamin, 1969)

3. Relevance of the stimulus. The main experimental phenomenon here is taste-aversion learning: animals (and people) readily associate tastes of particular foods with illness, even if the taste has been experienced hours before the illness (i.e. there is no contiguity: Garcia and Koehling, 1966).

This leads to a more cognitive “information processing view” of conditioning in animals than the “knee-jerk” idea of conditioned reflexes. However when classical conditioned associations occur in people they are distinctive in being involuntary, and not verbally accessible (e.g. Brewin, 1989, see week 7).