ABSTRACT The concept of collaboration has been valued in second language pedagogy. However, it was found that the use of pair/group work in an L2 writing class has always been limited to pre-writing and re-writing stages, not yet applicable to the writing stage itself. From the limited number of empirical studies conducted concerning collaborative writing, collaboratively produced texts are found consistently more grammatically accurate. In addition, during collaborative writing process, students are found to engage in activities such as co-constructing knowledge, scaffolding, and pooling resources to each other, which are believed to be effective in improving the quality of their produced text and in facilitating their learning. However, due to the limitation of the experimental design in previous studies, participants could only choose to write collaboratively or individually. Very little research allowed participants to experience both writing modes besides their perceptions were investigated. The present study attempts to fill such a gap in addition to have a more comprehensive understanding of collaborative writing. Among various web 2.0 technologies, wikis have drawn many researchers’ attention as a good platform for collaborative-writing projects. With the assistance of wikis, the current study aims to compare collaboratively and individually produced texts, explore the nature of collaboratively composed process, and investigate students’ perspectives on wiki-based collaborative writing. Eighteen EFL college English-major junior students as nine pairs took part in this study. This study contained three phases. In the first phase, students received training sessions about the use of wikis and peer-reviewing skills. The second phase consisted of two writing cycles. The nine pairs of students were divided into two groups: collaborative (CG) and individual (IG) ones. In the first writing cycle, five pairs volunteered to be in the collaborative group (CG) while the other four pairs were in the individual group (IG). While students in the CG discussed with their partners to compose drafts, students in the IG worked by themselves. After the first drafts were done, students in the two groups invited their classmates to peer-review the drafts and give them comments. Again, students in the CG made revisions together while those in the IG did it by themselves. The writing cycle ended with students’ revised drafts. In the second writing cycle, the two groups of students reversed their writing modes, and they followed the same procedures the other group did in the first writing cycle to complete another essay. In the final phase, an overall evaluation questionnaire was administered to participants to elicit their responses regarding wikis-based writing experience. Triangulation of different data sources was adopted in this study, including copies of students’ writing products in the two writing cycles, MSN discussion logs, time logs, revision history as recorded by wikis, and students’ responses to a self/peer evaluation questionnaire and an overall evaluation questionnaire. The texts are measured quantitatively in terms of fluency, accuracy, and complexity. Analysis of the data from MSN discussion logs, time logs, and revision history presented a qualitative picture of how students engaged in collaborative work. The self/peer and overall evaluation questionnaires were designed with the aim to understand students’ attitudes toward wiki-based writing experience, and what they think they learnt from each other during the collaborative process. The findings revealed that firstly, when students were allowed more time to work on the collaborative writing task as a take-home assignment, pairs consistently produced longer and more accurate texts than individuals. Exploration of students’ collaborative process indicated that they were willing to work more collaboratively to enhance the quality of writing products once their awareness was raised. In addition, pairs were found to make fewer revisions but higher percentage of text-base changes, large-scale meaning changes, at the re-writing phase than individuals, which indicated that with their partners’ assistance, pairs seemed to be confident in making more text-base changes. Thirdly, students’ positive attitudes toward product, process, and effects of the wiki-based collaborative writing task showed that it is feasible to implement such a task in an EFL college class. This study provided significant evidence that the wiki-based collaborative writing task, combining the advantages of wikis and collaborative writing, could not only be implemented successfully as an after-class assignment but also offered students great opportunities to learn from each other and improve their writing. Implications and future research are discussed.