Recognising critical reasoning and problem-solving as one of the key skills for twenty-first century citizenship, various types of problem contexts have been practiced in science classrooms to enhance students' understandings and use of evidence-based thinking and justification. Good problems need to allow students to adapt and evaluate the effectiveness of their knowledge, reasoning and problem-solving strategies. When students are engaged in complex and open-ended problem tasks, it is assumed their reasoning and problem-solving paths become complex with creativity and evidence in order to justify their conclusion and solutions. This study investigated the levels of reasoning evident in student discourse when engaging in different types of problem-solving tasks and the role of teacher interactions on students' reasoning. Fifteen students and a classroom teacher in a Grade 5-6 classroom participated in this study. Through case analyses, the study findings suggest that (a) there was no clear co-relation between certain structures of problem tasks and the level of reasoning in students' problem-solving discourse, (b) students exhibited more data-based reasoning than evidence-based and rule-based justification in experiment-based problem-solving tasks, and (c) teacher intervention supported higher levels of student reasoning. Pedagogical reflections on the difficulties of constructing effective problem-solving tasks and the need for developing teacher scaffolding strategies are discussed.