Gyselinck, V. &Tardieu, H. (1999). The role of illustrations in text comprehension: What, when, for whom, and why? In H. van Oostendorp & S. R. Goldman (Eds.), The Construction of Mental Operations During Reading. NJ: Erlbaum.

SUMMARY AND CONCLUSION. This chapter presented a number of recent studies concerned with the role of text illustrations on comprehension, and more specifically their role on inference making. We attempted to address questions about the conditions for illustrations to improve comprehension. We argued that to understand the role of illustrations in text processing, it is more useful to focus on the processes by which the beneficial effects of illustrations come about than to focus on the description of illustrations per se, given that this description is in most cases highly dependent on the processes involves in the formation of the reader's representation. Various types of illustrations may be designed. The results reported indicate that pictures that highlight the relationships between the objects being described in the text are the most beneficial for readers, allowing them to build connections in order to draw inferences. Moreover, the presentation of illustrations seems to be most effective when pictures are presented simultaneously with the text, thus permitting readers to navigate between the two sources of information. In addition, results indicate that the beneficial effect of illustrations can be observed during the course of reading itself, although methodological problems have been raised regarding the technique used to assess online comprehension. It can also be noted that this beneficial effect lasts even after a long delay. Finally, an important issue that bears emphasizing relates to what we named the internal conditions for illustrations to improve comprehension. It appears that mainly low-knowledge subjects benefit from the presentation of illustrations. The use of illustrations could in addition be a skill that has to be learned, and it could also depend on some specific abilities of the readers, such as spatial ability. We think that further research should be devoted to individual differences and to cognitive constraints imposed by the processing of illustrations, especially as regards working memory capacities. The last question we examined was why illustrations improve comprehension. The dual coding theory was presented, and we argued that it could not fully explain the role of illustrations in inference making. We suggested that mental model theory provides a better account of the beneficial effect of illustrations in text comprehension. The main advantage of iconic modes of representations is that they make structural relations more transparent, thus allowing readers to build a mental model that can be manipulated afterward to draw inferences. However, many questions about the involvement of pictures in the representation and about the form of the representations remain to be answered. Whether readers encode the picture, which then becomes a mental model, or whether they use the picture as a frame would allow us to comprehend the nature and the content of the representations built. Do readers separately build the linguistic representation, the pictorial representation, and then the mental model as a combination? This would amount to a kind of multiple codes representation. In addition, do readers keep traces of both the linguistic and the pictorial representations as separate entities and access them separately? If not, do readers construct a unique representation right from the beginning of the comprehension process? Research such as that conducted by Hegarty and Just (1993) gives us some of the elements of answers. We believe that research using concurrent methodology tasks could prove very useful to a more precise understanding of how pictures and texts are integrated by readers in working memory. We also need to know whether readers, even when they are not presented with pictures, make use of their visuospatial working memory. This could tell us if an analogical representation is built during understanding. In addition, more research should be devoted to the use of representations built with text and pictures after long delays, which could give us an idea of the characteristics of the representations held in long-term memory.