The research hypotheses for this study were that there would be no statistically significant differences among the groups on the mean score of the cloze posttest, the reading section of the TOEIC posttest, and on error free clauses per 100 words on the writing posttest. The alpha level was set at .01, as multiple ANOVAs were used for the analyses. Threats to validity such as maturation exist in longitudinal research, but this research was continued for three semesters because error correction feedback might need a long amount of time to have an effect on grammatical accuracy in writing (Franzen, 1995; Rod Ellis, personal communication, 1999).

3. RESULTS

3.1 Cloze test

Table 2 presents descriptive statistics. A one-way ANOVA showed that there was no significant difference among the groups on the pretest, F(2, 93) = 1.514, p = .225. A tests (pretest/posttest) by groups (three levels) repeated-measures two-way analysis of variance showed that there was a statistically significant difference within the participants, F(1, 91) = 359.274, p = .000, indicating that all groups improved significantly across the pretest and the posttest. No statistically significant differences were found among the groups, F(2, 91) = 1.909, p = .154. In addition, the interaction was not significant, F(2, 91) = .549, p = .580 (Table 3).

Table 2. Descriptive Statistics for Cloze Tests

JSG = Japanese summary group
ESG = English summary group

Table 3. Repeated-Measures Two-Way Analysis of Variance on the Cloze Test

8