Should I use pairs, groups of 3, 4 or more? The answer to this often-asked question usually depends on the aims of the lesson but is often guided by teacher intuition and not research. This study aims to address this issue with respect to one common task: peer review of writing. Following earlier research by the author that found more feedback moves in triads over pairs, this study compared triads and groups of four. The study involved two classes of university undergraduates (n=36) studying English in Seoul. There were three peer review sessions for three essays during one semester. Class A was divided into 4s, then 3s, then mixed 3s and 4s; Class B was divided into 3s, 4s, then mixed 3s and 4s. The results show more feedback moves were discussed in 3s rather than 4s, but this conflicted with student reflections. Implications for teaching are discussed.