The paper demonstrates how information on text structure can be used to improve the performance on the identification of topical words in texts, which is based on a probabilistic model of text categorization. We use texts which are not explicitly structured. A text structure is identified by measuring the similarity between segments comprising the text and its title. It is shown that a text structure thus identified gives a good clue to finding out parts of the text most relevant to its content. The significance of exploiting information on the structure for topic identification is demonstrated by a set of experiments conducted on the 19Mb of Japanese newspaper articles. The paper also brings concepts from the rhetorical structure theory (RST) to the statistical analysis of a text structure. Finally, it is shown that information on text structure is more effective for large documents than for small documents.