This chapter aims to explore the cohesive devices in the layout of multimodal discourses.On the basis of a review of related works and an analysis of the corpus of the English learning advertisements，I propose a working framework consisting of two aspects：structure of the discourse and importance in meaning.The first is realized through grouping devices，including visual similarity，framing and spatial nearness.The second is realized through devices of adjusting visual salience，such as color，relative size，and spatial placement.The specific devices through which the two principles are materialized are analyzed with samples from the corpus.

Three implications are obtained from the analysis and worth pointing out.

First of all，as shown in the analysis in Section 6.3，these devices are usually employed simultaneously to achieve the intended layout design.That is，the grouping devices work to signal the structure of the discourse while the devices for adjusting salience work at the same time to signal the relations and the importance of concerning elements.The simultaneous operation of all these cohesive devices creates an organized，coherent，and hierarchical representation of the discourse through layout.

Secondly，the cohesive devices in layout often align with the coherence strategies in content.It means that the three strata of coherence construction in multimodal discourses—namely global coherence，local coherence and layout cohesion—are not independent levels separated from each other.Rather，they are separated just for the convenience of analysis.In actual cases coherence is constructed by the collaboration of the resources at all the three levels of coherence.

Thirdly，some resources have multiple functions.For example，color can do many things：similarity in color can create visual similarity between two or more elements;contrast in color can form framing effects;brightness and saturation of color can adjust the visual salience of an element.As Kress and van Leeuwen（2002）hold，color is a semiotic mode which，like other modes，is multifunctional in its uses in the culturally located making of signs.It has its own grammar.The implication is that，seen from the perspective of social semiotic，meaning-making is dynamic，no matter what kind of semiotic resources are deployed.

As Mancini and Scott（2006）admit，the research on the structure of multimodal document is still in its infancy and needs further development.Thus the work is at this stage more concerned with identifying the right questions than with presenting the right answers.This chapter is an attempt in this direction.Although the proposal in this chapter is put forward on the basis of psychological principles and can generally explain the samples，it is far from perfect.Hopefully more insights from editing and design will be useful.More efforts need to be made in the future in this regard.