Writing self-efficacy has been found to be associated with writing achievement, and can even be improved with the help of an intervention program. However, with regards to writers who are college students, there is still ambiguity or partial information. First, unlike the case with primary and secondary school students, findings on the relationship between self-efficacy and achievement in college students are inconsistent, and second, when findings from intervention programs designed to improve selfefficacy, for the most part do not refer to differences in improvement between students with varying levels of language proficiency. Thus, the aim of the present study is to offer a possible explanation for the inconsistency in the findings and also to analyze the differences between more and less skilled writers following an intervention program. 81 female college students participated in the study during a compulsory course in language skills. Self-efficacy was measured before and after the writing program by having the students fill out a questionnaire. The first major finding was that only in the measurement after the program was there a correlation between self-efficacy and achievement, and this correlation seemed to depend on the students knowing what writing requirements they were expected to meet. In writing courses in college, unlike in schools, there are no standardized tests and therefore expectations tend to be unclear. Reports about correlation are then inconsistent. However, by the second measurement of the program, the expectations were clear and therefore a correlation was shown. The second key finding concerns the intervention program and the improvement in self-efficacy. It was found that most of the change was among the more skilled writers, both in the ability to create ideas and in selfregulation, whereas those less skilled reported a moderate improvement only in self-regulation.