Currently, a discursive resource widely used is images, whose use can facilitate the explanation of concepts, constituting essential support for the communication of scientific ideas. This article reports the results of a study on the perception of 107 students of two High Schools located in São Paulo (Brazil) concerning representations of the physical state change processes at a sub-microscopic domain. After the development of a teaching-learning sequence (TLS) about the properties of matter, discussions were conducted both in small groups and with the whole class. At this point, students were asked to draw up images that represent their understanding of the phenomena of physical changes at the sub-microscopic level. In the second phase of the TLS, after some discussion about the drawings previously produced, each group of students was asked to create an audiovisual object (AO) concerning the phenomena. It was possible to perceive both the evolution of their expressed mental representations that present scientific features more consistently and the students’ more coherent and secure speech. These results suggest that the multimodal approach promotes opportunities both for the building of models by the students, and their expression is essential to chemistry language learning. Keywords: multimodal approach, media literacy, sub-microscopic domain, science education