The purpose of this study was to examine students' experiences and perceptions of multiple interaction activities (self-directed, peer, and teacher feedback) implemented in a large multilevel EFL writing class in one private technological university in the southern part of Taiwan. Large size writing classes, quite common in private institutions of higher education in Taiwan, cannot be effectively operated to meet individual students' needs in improving their writing performance. Low achievers have difficulties in keeping up with competent writers in learning writing skills while advanced students complain of their learning too little from the class. This research, based on the activity system model proposed by Engestrom (1987), was a case study in which interviewing student participants, observing classroom activities, audiotaping peer response sessions, and examining students' drafts and feedback sheets were the methods to collect data. The qualitative software, ATLAS.ti, was employed to analyze interview and peer response data according to the code lists developed for this purpose. A rubric was developed to examine the changes students made after having incorporated the three types of feedback into their drafts. Major findings indicated that intermediate and low achievers, though making more efforts in conducting self-directed feedback, felt unsatisfied with this activity while high achievers, investing less energy and time, gave more positive opinions to this activity. However, intermediate and low achievers gave a higher percentage of satisfaction to peer response activities than high achievers because the former could obtain more constructive peer feedback than the latter. In addition, all students were in favor of modified teacher feedback but gave negative opinions to traditional teacher feedback. On the whole, intermediate and low achievers, based on their preference, ranked teacher feedback the most important, then peer feedback and finally self-directed feedback whereas high achievers placed teacher feedback first, self-directed feedback second, and peer feedback last. Student writers' responses to each type of feedback were closely related to the amount of constructive comments they received. The more helpful suggestions they obtained, the more positive opinions they gave to a certain type of feedback. In the end of the study, recommendations were made for curriculum designers, classroom practitioners, and further studies.