Kim, Kyung-Sook. 2003. Textual Variability of Newspapers According to Topics. The Sociolinguistic Journal of Korea, 11(2). This study analyzes two major newspaper registers, editorials and news reports, to show that the composition of these registers is variable according to topics. Previous studies assume, a priori, that editorials and news reports belong to different text types or treat them as if they had a homogeneous textuality. The current study tries to show that there is a variability even within the same register of editorials or news reports.The database for the study consists of editorial and news articles from English version of the Dong-A Ilbo on the Internet and from the New York Times, published between 1999 and 2002. The articles were chosen on the basis that they attracted wide attention in Korea or in the U.S. The topic of political corruption was chosen in the case of Korea and the topic of the 9-11 terrorist attacks in the case of the U.S. The database is comprised of a total of 200 texts and these texts are subdivided into eight different groups. Based on this database, the frequencies of 59 linguistic features which have been found to be discourse-functionally relevant were calculated. Then, distribution of these features was analyzed through Multi-Dimensional Method (Biber, 1988).The findings are as follows. First, the eight text groups show a considerable textual variability on five discourse dimensions as there is a complex interactive effect. Second, a two-way ANOVA reveals a textual variability that arises from the topic difference. Especially in the case of the Times, the terrorism-related news reports were found to be much more 'involved' than normal news texts.