While peer review is a common practice in college composition courses, there is little consistency in approach and effectiveness within the field, owing in part to the dearth of empirical research that investigates peer-review processes. This study is designed to shed light on what a peer reviewer actually reads and attends to while providing peer-review feedback. Fifteen participants peer reviewed a student's essay that had both holistic and surface-level errors. Using eye-tracking technology, we collected detailed and informative data about which parts of the text the peer reviewer looked at, how long the peer reviewer looked there, and where the peer reviewer looked next. These data were analyzed according to eye-movement research methodologies and juxtaposed with each peer reviewer's comments and suggestions about the essay being reviewed during a typical peer-review exercise. Findings include an unexpected mismatch between what peer reviewers focus on, spend time on, and examine multiple times when reading and peer reviewing an essay and what they choose to give feedback about during the peer-review session. Implications of this study include a rethinking of the composition field's widespread use of a global-to-local progression during peer-review activities. (Contains 2 notes and 1 table.)