The increasing availability of digital tools in science classrooms can provide students with more frequent and easier access to large amounts of data. Large data sets have considerable epistemological potential, as they enable, for instance, the observation of otherwise unobservable phenomena, but it must be assumed that handling them places greater demands on learners. However, it remains unclear how learners use large amounts of data when arguing from data, the challenges they encounter, and how they can be supported effectively. To address these gaps, we derived assumptions from a constructivist epistemological, sociocultural, and cognitive science perspective on how evaluating large amounts of data affects the quality of arguments in educational lab work. To investigate these assumptions, 664 high school students were randomly assigned a diagram containing a small, medium, or large amount of data from one experiment. The students were then asked to work on an argumentation task, which consisted of selecting a scientific claim and providing a written argument to support the selection. In addition to the impact of the different amounts of data, we investigated how the quality of the argument was related to the correctness of the claim the students selected as well as the impact of cognitive load (CL) and data competence on that quality. The results showed that the amount of data, CL, and data competence had no meaningful effects on the quality of arguments, and there was no relation between such quality and the claim the students supported. The study provides insights into how students deal with data in the learning of science. Implications for effectively supporting students when learning from large data sets are discussed.