Considerable research has examined the contrasting predictions of configural and elemental associative accounts of learning. One of the simplest methods to distinguish between these approaches is the summation test, in which the associative strength of a novel compound (AB) made of two separately-trained cues (A+ and B+) is examined. The configural view predicts that the strength of the compound will approximate the average strength of its components, whereas the elemental approach predicts that the strength of the compound will be greater than the strength of either component. The summation test has lead to contradictory evidence in experiments with animals as well as with human adults. The purpose of this research was to examine summation in predictive learning of 5-9 years old children. The results provided evidence of summation (i.e., AB greater than A and B) after training with a "simple summation" procedure (A+ B+ test with AB; Experiment 1, n = 26); but no summation following a "preserved conditioned inhibition" procedure (i.e., A+ B- AB- followed by B+; and test with AB; Experiment 2, n = 26). In Experiment 3, both effects were simultaneously observed (n = 14). These results are consistent with reported findings in both Pavlovian conditioning with animals and predictive learning with adult humans. Theoretical alternatives within the elemental and configural approaches are discussed. (Contains 1 table and 4 figures.)