Nowadays, writers rarely ever write from scratch. They integrate information from multiple online sources (e.g., reports, articles, blogs, tweets) into a new text that is coherent and relevant. Source-based writing is a complex mental process: writers must compare, contrast and evaluate sources, plan the text, select the relevant information from the sources and add new data, and write the text. The academic research into source-based writing is limited, yet in full development due to its growing professional importance. The existing research is focused primarily on the writing product and writing in the L1 (i.e., dominant language). This research project aims to create a theoretical model that describes source-based writing in the L1 (Dutch), L2 (English) and FL (French, Spanish). We will explore (1) how writers consult digital sources during the writing process, (2) how they integrate input (e.g., content, structure and wording) from those sources in their writing product, and (3) how these processes relate to the quality of the writing product and the writer's working memory and linguistic proficiency. We will do so by: (1) analysing keystroke logging data of about 600 texts written by master's students in Multilingual Professional Communication; (2) analyzing the students' writing products and processes by various plagiarism and linguistic annotation tools; (3) investigating the effect of process feedback on source-based writing via peer-based examples (modelling) in an experimental study.