This study reports on a case study about multimodal work in a primary physics classroom focusing on forces. Previous research reports that students benefit from multimodal work in science classrooms. Yet, few systematic studies have been performed to reveal how students represent their experiences and ideas of science phenomena over time within and across different semiotic modes (e.g., through action, speech, writing, and image, including multimodal ensembles). The design of the lessons was built around a number of experimental activities, starting with a puzzling phenomenon and where students for each experiment predicted, observed, described, and tried to come up with explanations. Based on social-semiotics, we combined analysis at an overall level (long timescale and large grain size) and a detailed level (short timescale and small grain size) to shed light on how the science content was represented within and between modes over the teaching and learning period, and what content was expressed in these representations. Our findings reveal how students moved from focusing on attributes such as a "heavy" regarding artifacts used in the experiments, towards the central physics processes, such as one object exerting force on another object. Furthermore, we were able to detect that such "signs of learning" were shown in students' small-group discussions and multimodal texts following a carefully orchestrated multimodal exposition by the teacher. Hence, such a careful multimodal orchestration appeared to be critical for the students' meaning-making about the science content.