This study examines the effect of class discussion as a pre-writing activity on actual writing performance. The experiment was conducted with all the Level 3 and Level 4 students enrolled in the Intensive English Language Institute of North Texas State University in the Spring, 1986 semester. Cochrans C test was performed to test significant differences between groups at the beginning of this test. Multivariate analysis of variance tests were used to determine the treatment effect between and within groups, and a matched t-test was also utilized to analyze the difference within tests. Pearson product moment correlation coefficient was calculated to determine the relationship between the discussion activity score and the actual writing score. Analysis of covariance tests were used to determine which variance of discussion activities had greater effect on the actual writing score.