This study investigated the impact on chemistry learning of the degree to which students embedded or integrated multiple modes of representation in end of unit writing-to-learn activities. A multi-case study approach utilizing quasi-experimental methodology involving intact high school chemistry classes taught by two different teachers was employed. Approximately half of the classes for each teacher were designated treatment classes and students in these classes participated in specific classroom activities designed to encourage the use of strategies to embed multiple modes within text in student writing. Control classes did not participate in these activities. All classes with the same teacher participated in identical end of unit writing tasks followed by identical end of unit assessments. Writing tasks and end of unit assessments were teacher designed and were therefore unique to each setting. Data from each teacher was initially analyzed independently to explore characteristics of student writing and student performance on end of unit assessments. This was followed by cross case analysis. Analysis of quantitative data indicated that for the first teacher (n = 70 students), treatment classes significantly outperformed control classes on two different measures of writing characteristics during a first unit of study, two measures of writing for the second unit, and three categories of end of unit instruction for the second unit. For the second teacher (n = 95), treatment classes outperformed control classes on two writing characteristics and three end of unit assessment categories during the only unit of study assessed. In addition, at both sites, significant positive correlations were found between all writing characteristic measurements and end of unit assessment performance. These results not only support the use of multimodal writing-to-learn tasks as a pedagogical tool to improve chemistry learning, but specifically suggest benefit when multiple modes are effectively linked within these tasks.