This study intended to examine classroom English proficiency of pre-service teachers' and explore how to improve their competence. The instruments adopted here to collect data were a written test on classroom English and transcription of teaching demonstration. The data were analyzed with Corder(1981)'s error analysis criteria. The results were the following. The most frequent error on both the written test and teaching demonstration was omission. They omitted often articles in both cases. Next frequnt error was selection. The students used words that were not consistent in terms of tense, third person singular, and plural. The differences found in the written language(written test) and oral language(teaching demonstration) were the following. First, they repeated the same expression in oral language whenever they were not sure of the language or the class seemed not to understand what the teacher said. Also they corrected their errors themselves in teaching demonstration. This was probablly possible becauses they typed out their detailed lesson plan and had a time to practice it. Also more pragmatic errors were found in oral language than the written languae. This is probably due to the impromptuness of teaching. Lastly, the students seemed to read a script when they explained the rule of a game or an activity. This made the class hard to follow the steps of activities. From the results above, some teaching implications were derived. First, more focused classroom language teaching will be needed in the curriculum. Especially, students need to have time to practice classroom language during English classes. Second, language errors should be corrected and the speaker needs to be aware them and try to correct themselves. This will prevent facilization of incorrect expression. Third, partial use of mother tongue should be allowed for more effective class management. When explaining the rules of a game or a complicated activity will prevent any misunderstanding and help learners learn during the activity.