Argumentation is an important component of scientific education (Osborne, 2010). However, how students create and evaluate competing arguments in scientific investigations is a complex construct, which presents significant challenges for assessment. We engaged 349 middle and high school students in a virtual scientific investigation based on an authentic problem involving the ecosystem of Yellowstone National Park. Students evaluated three possible causes for vegetation loss in Lamar Valley: tourism, global warming, and killing wolves. Students compared their responses to individual claims to formulate a final decision. The authors developed a coding system to assess student arguments and analysed how students changed their evaluations across the investigation. Four distinct levels of responses reveal the range of student decision making and patterns that have implications for instruction and assessment. Few student evaluations explained both merits and weaknesses of claims.