Contingency Theory

Contingency theory is one approach to formalizing associative learning (Rescorla, 1967, 1968). According to Rescorla, the "American" view of Pavlovian conditioning focused upon the frequency of pairings between reinforcement (or more generally the unconditioned stimulus (US)) and the conditioned stimulus (CS). This approach emphasized, almost exclusively, excitatory mechanisms -- that is, the ability of the CS to signal an imminent US. In contrast, contingency theory looks at both this pairing as well as the trials in which the CS and the US are not paired. That is, learning is related to two conditional probabilities: the probability of the US given the CS (P(US|CS)) and the probability of the US in the absence of the CS (P(US|~CS)). Learning would only occur if there was a difference between these two conditional probabilities. Note that defining learning in terms of this difference permits inhibitory processes to be important: if P(US|~CS) is higher than P(US|CS), then an agent can learn that the CS is a signal that the US is not forthcoming. It has been argued (Shanks, 1995) that contingency theory is a computational level account of associative learning.

References:

1. Rescorla, R. A. (1967). Pavlovian conditioning and its proper control procedures. Psychological Review, 74(1), 71-80.
2. Rescorla, R. A. (1968). Probability of shock in presence and absence of CS in fear conditioning. Journal of Comparative and Physiological Psychology, 66(1), 1-5.
3. Shanks, D. R. (1995). The Psychology Of Associative Learning. Cambridge, UK: Cambridge University Press.

(Added January 2010)