The purpose of this paper is to identity children's internal process during production of writings. A new procedure was devised to deal with the problem of estimating the internal dynamics of children from 8 to 12 years of age. At first, a simple procedural model of discourse production was presented; then, where and how long pauses were generated during writing was recorded for each child subject. Each child was also interviewed for the introspective report of what he or she thought at each pause. Then, the model and independently analyzed introspective and behavioral data were studied together to find a new model of the writing process. This procedure succeeded in identifying15strategies that may work in production of writings. Furthermore, the results also suggest the following. First, children's writing plans function not merely for controlling the writing process itself, but also for a global monitoring on whether the ongoing process matched what the writer wanted to do. Second, combination of strategies results in four kinds of writing style, all distributed in the age bracket studied. Thus, writing style seemed to depend mainly on individuals and possibly on contexts, though some age trends were detected. Next, proceeding to more specific problems, two points were investigated by the second and complementary experiment; i. e., whether writing style could be changed externally, and how plan monitoring would work. The first point, flexibility of writing style, was examined in a similar way as the first experiment except that each child was allowed to plan ahead for five minutes before starting writing. The result showed that some (but not all) children were apt to change writing style to one that generated a global plan and work under it. The second point, plan monitoring, was investigated by using recall of subjects' own writings a week later. It was shown that the way to recall depended on the original writing style, and children who wrote in a style with a global plan generally provided, good performance of recall. It implies, at least indirectly, that making global plans were useful for monitoring the process of writing. From the analysis of the model and experimental data, it was suggested that the writing process was a process of adaptively organizing various writing strategies: it might be the cause of an apparent variety of children's writings showing many degrees of freedom of procedural organization in such an adaptive process. But it was also suggested that there existed some constancy in the way using strategies. The set of these two complementary aspects did not seem to be specific to children's writing, thongh. It seemed to reflect the general characteristics of a divergent, ill-structured problem-solving process.