The aim of this study is to propose some task and method which can clarify the psychological process of children's text production. Children(2nd, 4th and 6th graders)were given a stimulus sentence and were asked to make a story, writing one sentence after the stimulus sentence. Each child made six stories. Then 672 short texts were analysed in terms of(1-1)number of letters, (1-2)number of pausal units, (1-3)number of propositions(the first three are indices of textual quantity), (2-1)coherence forms, (2-2)number of new arguments(indices of texual coherence), (3-1)conceptual framework of children's sentences, (3-2)temporal relation of stimulus-children's sentence, (3-3)junctive relations of the two sentences(the last three are indices of semantic analysis). In evaluating the coherence forms and three semantic relations, texts were assorted into some categories and the frequencies were counted. Main results are(1)Sentences written by children increase in the three indices of textual quantity as grades. (2)As for the coherence forms, 2nd graders use ellipsis most frequently, but 4th and 6th graders use noun repetition. (3)The number of new arguments increase as grade and children's texts became more progressing. (4)The most frequent conceptual framework is action in all grades. But the dominance of action frame diminishes with development. From these results, it was suggested that planning of texts was likely to rely upon an action sequence in all the grades. And realization of the planning differs in choice of coherence forms. The youngers choose the more cognitively efforless form, e.g.ellipsis.