This study reports an investigation of real-time writing activity of Japanese (L1) and British (L2) writers. An overall aim of this study is to provide a complementary data to the conventional spelling error analysis (EA) that was conducted prior to the current study. A list of error categories, which is taken from L2 spelling errors in 43 word-processed essays in Japanese, and their distribution will be presented. However, as Nelson (1980) pointed out, EA can shed very little light on multiple type errors, because the readers have to rely on their subjective judgement to specify the main source of error among various possible underlying sources. In other words, it is almost impossible to distinguish each element in a heavily distorted multiple error word with any degree of objectivity or reliability. A typical solution for the reader is to ask the writer's original intention and the writing situation, but this ad hoc basis of analysis does not seem to produce any universal tendency of this error category.

This study is, therefore, designed to probe these possible sources of spelling errors in real-time writing performance. The database will be drawn from full-screen movements captured by a PC screen recording program. A naturalistic and observational analysis of recorded screen movements is expected to uncover writers' dynamic cognitive processes such as their procedures in error-repair sequences. Although the data contains a variety of grammatical, lexical, and orthographic (spelling) errors, the analysis is mainly concerned with spelling mistakes. A further analysis is carried out from a conversation-analytic perspective in order to isolate the underlying causes of identified area of spelling problems in writing with Japanese word processor.