During face-to-face communication, people use visual cues about what their partners are attending to as they process language. An eyetracking experiment explored how addressees use speakers' eye gaze and head orientation while interpreting references to objects in a spatial task. Thirty-six naive director/matcher pairs seated face-to-face were separated by a low or high barrier that hid identical mirror-image arrangements of objects. We compared matchers' ability to disambiguate referring expressions when directors' eyes and heads were visible over the barrier, when only directors' heads were visible (with eyes obscured by mirrored sunglasses), or when directors were completely hidden. Seeing directors' head orientation helped matchers quickly restrict attention to the target side of the display. Seeing directors' eye gaze helped matchers disambiguate referring expressions earlier (before the linguistic point of disambiguation) than did seeing head orientation alone. Along with benefits, however, there were some costs to monitoring eye gaze in face-to-face communication.