Abstract
Rats were trained in conditioned suppression discriminations where shock at the beginning of a trial signaled either shock or no-shock at the end of the trial. In the shock-positive condition, shock at the beginning of a presentation of white noise signaled that noise would end with shock; noise that did not begin with shock did not end with shock. In the shock-negative discrimination, shock at the beginning of noise signaled that noise would not end with shock; presentations of noise that did not begin with shock ended with shock. In shock-random training, shock at the beginning of noise did not reliably signal whether the noise presentation would or would not end with shock. Most subjects in shock-negative training quickly developed a differential pattern of suppression on positive (shock reinforced) trials and no suppression on negative (nonreinforced) trials. The shock-positive discrimination was much more difficult to establish and was not acquired by the majority of the rats. This "feature-negative" effect is a clear exception to the general superiority of feature-positive learning commonly observed in discriminations based on a single distinguishing feature. The results are discussed in terms of Pavlovian stimulus-shock contingencies in the shock-positive and shock-negative paradigms, which appear to favor rapid development of the shock-negative discrimination.