This paper considers prediction in language processing by examining the role of the visual context, and specifically, the role of speaker referent gaze on cognitive load. We inspect the anticipatory visual attention during sentence processing together with the cognitive load induced at the points of the gaze cue, and the linguistic referent. Employing a novel measurement of cognitive load - the Index of Cognitive Activity (Marschall, 2000) allowed us to simultaneously consider both anticipatory eye-movements and cognitive load. Our results show that the gaze cue is being followed, and considered as a relevant piece of information, which subsequently reduces the cognitive load on the linguistic referent. In addition, we found that considering the gaze cue is in itself not costly, unless it cues an object mismatching with the previous linguistic context.