Abstract
Five groups of humans were employed to test the hypothesis that noncontingent punishment (electric shock) may suppress or facilitate a goal response, and these effects are determined by the particular response-punishment pairings which happen to occur. After a goal response was trained for money, this reward was withheld. Three groups then received noncontingent, variable-interval punishment and in each group punishment was paired with different behaviour. A fourth group received consistent punishment contingent only on goal responses, and the fifth group served as a no-punishment control. The evidence supported the hypothesis and also indicated that consistency of punishment has an important effect which is distinct from sheer frequency of response-punishment pairing. These results, plus observations of goal responses during an aftereffects period when treatments had terminated, were interpreted to suggest that, although punishment may have emotion-provoking properties, it also supplies informative feedback to S about his responses and this factor also must be considered in order to predict human behaviour.