The aim of this study is to examine what students in grade 3, half of them multilingual,stress as important in multimodal text production while performing a task that isgoing to be assessed by their teacher. Sociosemiotic multimodal theories, and researchon multiliteracies, form the theoretical basis for the study. The empirical data consistsof fifteen individual interviews with the students. Through the interviews theirmultimodal text production appears to be complex. Also it becomes clear that images,rather than written text, play an important role for conveying messages in theirtexts. Furthermore, the interviews show that the students find it difficult to verbalizethoughts on qualities in their text productions. To sum up, the study stresses the riskfor discrepancy between what students achieve, by the use of several resources, andthe teachers’ notions of the students´ achievements. Additionally, the study highlightsthe role of task formulation and the need for competence when it comes to assessmentof multimodal texts.