In the study of child's thinking in a “20 Questions” game, Mosher and Hornsby found out two typical strategies, constraint seeking (CS) and hypothetical scanning (HS). They attributed these strategic variations to representative ability differences, and assumed that CS required symbolic representation. On the other hand, this CS type strategy was the process of sequential classification, so that it could be considered to correspond to Piaget's additive classificatory operation. On this framework the socalled concrete period children must have a repertory of such a strategy, but there had been no evidence of that correspondence in the studies on the “20 Questions”. The present study was aimed at examining the problem of such correspondence by two cross-sectional studies. In Experiment 1, the usual “20 Questions” task was used with modified stimulus and procedure. The results showed that most of the 2nd or 3rd-grade children could exhibit CS (2nd graders=80%, 3rd graders=90%) as well as some of the 1st-grade or pre-school children (1st graders=45%, pre-school= 30%). This developmental tendency was similarly observed in Experiment 2 in which Ss were assigned “20 Questions” listening task. These findings show that the so-called concrete period children could employ CS type problem solving and such CS type strategy corresponded more to classificatory operation than to representative ability. Further, the developmental process of classificatory operation was discussed at length.