Rats trained in simultaneous brightness discrimination problems varying in difficulty in comparison to a following test discrimination condition, and given equal numbers of reinforced trials were compared for errors during a given number of trials on the test discrimination. The findings were that: (1) "learning was more efficient when animals were first trained on an easy and then shifted to the test discrimination than if all the training was given directly on the latter," and (2) "learning was most efficient when the animals approached the test discrimination through a series of graduated discriminations." (PsycINFO Database Record (c) 2006 APA, all rights reserved)