The search for intermediate units in discourse construction starts with sentences taken as elementary discourse units, assuming a constituent structure of complex discourse units built by recursive aggregation or integration of lower units into higher ones. Multimodal texts include image representations that provide information used in the discourse construction process. A configurational approach to discourse constituent structure is developed, where discourse relations resulting from a cartographic approach are considered as subordinating or coordinating relations in the configurational discourse constituent structure. Texts are data structures that package discourses in specific ways adapted to the transmission processes that take place within communities of practices following diverse textualization traditions. The interplay between text structure, which is top-down and overt, and discourse structures, which are bottom-up and covert, is shown in a political cartoon and in a newspaper column.