Kindergartners and college students were taught the same novel word (e.g., torubi) to two objects (e.g., a dog and a cat) by one of the four instruction strategies and then were required to select one of the two objects and the objects which they thought fit to the novel word from eight probe objects (e.g., two dogs, two cats, a deer, a horse, a sparrow, and a goldfish). After that the subjects were taught the novel word three times and then were required to select the probe objects again. Based on the selection patterns for the novel word, the subjects were classified into four types. Most children were identified as the mutual exclusive (ME) and the subordinate subjects. The ME children decreased by labeling the objects in the instruction. Almost all adults were identified as the subordinate and the intermediate subjects but there were almost no superordinate adults. The intermediate adults increased by the inclusion strategy in the instruction. With instruction trials the ME children decreased and the subordinate children increased, while the subordinate adults decreased and the intermediate adults increased. The findings were explained by assuming the following cognitive processes. The ME children recognize weak associations of the objects and the novel word, and the subordinate children recognize strong ones. The subordinate adults recognize equivalence of the objects represented by the novel word and the intermediate adults recognize equivalence represented by the intermediate category term.