Writing cohesive texts is a crucial but challenging skill to master. Recently, cognitive tools that provide students with a graphical representation of their texts in the form of concept-maps have been shown to support students' writing. Despite its beneficial effects, the addition of a graphical representation may have the disadvantage that students have to process multiple isolated representations (i.e., text, graphic), which may increase cognitive load. By applying principles of multimedia learning, in two experiments, we investigated whether interrepresentational signaling and spatial contiguity would have differential effects on students' subsequent writing performance and on the processing of the graphical feedback. In Experiment 1, students wrote an expository text and either received conventional concept-map feedback, correspondence-enhanced concept-map feedback with interrepresentational signaling, spatially contiguous feedback, or no feedback during text revision. Regarding local cohesion, we found that students profited most when they received spatially contiguous feedback. Contrarily, correspondence-enhanced concept-map feedback was most effective for improving global cohesion. In Experiment 2, we examined the attentional processes while using correspondence-enhanced concept-map feedback versus conventional concept-map feedback by means of eye-tracking. Students receiving correspondence-enhanced concept-map feedback had longer fixation times on the concept-maps, more transitions between their text and the concept-map and were more efficient in improving their text for global cohesion than students receiving concept-maps without signaling. The findings suggest that interrepresentational signaling and spatial contiguity differentially contributed to students' writing. Therefore, choosing the adequate format of instructional support plays a critical role in scaffolding students' writing.