The literature on corrective feedback (CF) that second language writers receive in response to their grammatical and lexical errors is plagued by controversies and conflicting findings about the merits of feedback. Although more recent studies suggest that CF is valuable (e.g., Bitchener, 2008; Sheen, 2007), it is still not clear whether direct or indirect feedback is the most effective, or why. This study explored the efficacy of two different forms of CF. The investigation focused on the nature of the learners’ engagement with the feedback received to gain a better understanding of why some feedback is taken up and retained and some is not. The study was composed of three sessions. In session 1, learners worked in pairs to compose a text based on a graphic prompt. Feedback was provided either in the form of reformulations (direct feedback) or editing symbols (indirect feedback). In session 2 (day 5), the learners reviewed the feedback they received and rewrote their text. All pair talk was audio-recorded. In session 3 (day 28), each of the learners composed a text individually using the same prompt as in session 1. The texts produced by the pairs after feedback were analyzed for evidence of uptake of the feedback given and texts produced individually in session 3 for evidence of retention. The learners’ transcribed pair talk proved a very rich source of data that showed not only how learners processed the feedback received but also their attitudes toward the feedback and their beliefs about language conventions and use. Closer analysis of four case study pairs suggests that uptake and retention may be affected by a host of linguistic and affective factors, including the type of errors the learners make in their writing and, more importantly, learners’ attitudes, beliefs, and goals. The findings suggest that, although often ignored in research on CF, these affective factors play an important role in uptake and retention of feedback.