This paper applies a new procedure for text structure analysis of explanatory texts, relying on well-defined use of linguistic knowledge, to a corpus of child essays. The findings illuminate language users' cognitive representations and affect research on text processing, writing proficiency and text quality.