For example, an animal is exposed to conditioned stimulus 1 (CS1), which predicts the occurrence of a reinforcer. After learning the paired association between CS1 and unconditioned stimulus (US), a compound stimulus composed of CS1 and another stimulus 2 (CS2) is presented with the US. Hence both CS1 and CS2 are stimuli that predict the US.
However, when tested, the animal shows little, if any, CS2-US association. This is because the occurrence of the US was fully predicted by CS1 alone, and hence no learning occurs when CS2 is presented simultaneously. In other words, CS2-US association is blocked because CS1-US association already exists.
This effect was most famously captured mathematically by the Rescorla-Wagner model. As this model learns based on the extent to which the US isn't predicted by the set of present CSs, it can account for the blocking effect. Blocking can also be considered a general feature of many models that learn based on an error signal.
While some argue that blocking demonstrates that the organism did not learn the association between the CS2 and the US, this is not necessarily the case. For instance, after a traditional blocking paradigm, the CS2 does not elicit a response (alone). However, if the response to the CS1 is extinguished, the organism will begin to respond to CS2 alone. This demonstrates that the association between the CS2 and the US was initially learned but in comparison to the stronger predictive value of the CS1 there is no conditioned response. However, after extinction, the CS2 does have more predictive value than the now extinguished CS1.
The reverse of blocking is often called backward blocking. In backward blocking, the subject is exposed to the compound stimulus (CS1 and CS2 together) first, and only later to CS1 alone. In some human and animal studies, subjects show a reduction in the association between CS2 and the US, though the effect is often weaker than the standard blocking effect, and vanishes under some conditions. This effect is not predicted by the Rescorla-Wagner model although other models have been proposed that capture this effect.
- ↑ Kamin, L. J. (1969). Predictability, surprise, attention and conditioning. In B. A. Campbell & R. M. Church (eds.), Punishment and aversive behavior, 279-296, New York: Appleton-Century-Crofts