The present study was designed to examine validity of House's chain model and Bowers'rule model in identity and oddity learnings. In Experiment 1, 320 young children (mean age:5;2) were run either identity or oddity learnings either to criterion or under 12 overtraining trial conditions. Completing a given overtraining, they were run shift learning. Group S-D, in which the rule was the same but the dimension different, took fewer trials to criterion than Group D-S did, in which the rule was different but of same dimension. Overtraining facilitated shift learnings of both Group S-D and Group S-S, in which both the rule and the dimension were the same. In Experiment 2, 90 young children (mean age:5;1) were run either identity or oddity learnings under a given verbaling condition. In both identity and oddity learnings, Group RV, in which children were compulsorily verbalized the rule, took fewer trials to criterion than Group DV did, in which they were compulsorily verbalized the dimension. These results in both Experiments 1 and 2 are in line with Bowers'rule model.