Scientific workflows are structured sequences of computational or data processing steps commonly used to automate complex scientific analyses. Abstract workflows, which provide high-level representations of these processes, play a crucial role in designing, understanding, and communicating complex workflows before delving into implementation details. In this paper, we explore the modeling patterns emerging from a collection of 172 workflow sketches created by students in a beginners’ course on scientific workflow management. Our aim is to understand how novice users intuitively model workflows when given the freedom to choose their own modeling styles. The study specifically focuses on graphical abstract workflow languages, which are generally perceived as more intuitive compared to their textual counterparts. The findings suggest that extending classical control-flow modeling with optionally usable data nodes can form an effective foundation for an abstract workflow specification language. These insights offer valuable guidance for designing a workflow language that aligns with the natural inclinations of users in scientific computing, enhancing both usability and functionality.