Abstract: Triangles varied in size and projected flash frequency were used as compound stimuli. Larger size combined with higher frequency served as positive stimulus, whereas smaller size combined with lower frequeincy served as negative stimulus. This pair of training stimuli appeared simultaneously on two small doors, and the dog was required to respond to the positive stimulus by pushing the door with its nose, from which food was given.The results of the crucial tests showed, as a pair of test stimuli, which did not include the negative training stimulus, were used, generalization of relative response was obvious. But as the test stimuli included the negative training stimulus, and it transfered to positive, the number of relative responses nearly approached the level of chance. When two variables combined with each other conversely, namely, large size combined with lower frequency, the dogs responded to size only, ignoring the frequency. With the size of the triangles equated, but with the frequency changed, they responded by chance.The mechanism of relative reflex and the interactions of the temporary connections of both absolute and relative stimulus were discussed. When the positive training stimulus remained positive in the test, the temporary connections of the absolute stimulus facilitated the temporary connections of the relative stimulus; but when the former was negative, the temporary connections of the relative stimulus were interfered. Similarly, when the negative training stimulus played positive role in the test, the temporary connections of the absolute stimulus destroyed the temporary, connections of the relative stimulus; whereas if it retained its negative function, it facilitated the temporary connections of the relative stimulus in most cases.

Key words: NULL