Pigeons were exposed to two keys, a main key and a changeover key. Initially non‐differential training was given in which pecking the main key was reinforced on a variable‐interval 2‐min schedule when the key displayed the first stimulus, a black line on a blue background, and was reinforced on an identical but independent variable‐interval 2‐min schedule when the key displayed a plain blue stimulus. Later, differential training was given in which pecking the main key was reinforced on a variable‐interval 2‐min schedule when the first stimulus was displayed; and was reinforced on a variable‐interval 10‐min schedule when a second stimulus, a black line of another orientation on a blue background, was displayed. During non‐differential and differential training, each peck on the changeover key changed the stimulus on the main key. Generalization tests were given before and after the differential training. These consisted of presentations on the main key of seven orientations of the black line on the blue background, including the first and second stimuli, with no reinforcements being given. Changeover‐key pecks changed the stimuli on the main key. Generalization gradients were obtained using three measures: time spent, responses, and response rate in the presence of each test stimulus. Typically, maximum values on these measures occurred to stimuli away from the first in a direction opposite the second stimulus, and minimum values occurred to stimuli away from the second in a directi