In educational settings, continuous assessment of the child's level of understanding is necessary to effectively utilize the principles of scaffolding and to create contexts that can advance the scientific reasoning of the child. In this article, we argue that a child's performance is a dynamic notion that is created by all elements in an interaction, including the task. Therefore, we studied preschoolers' levels of scientific reasoning varying different properties of the assessment context. Young children were interviewed about four scientific tasks using one out of four different protocols (varying in the degree of flexibility and adaptiveness) by an adult. In the first study, different task contents resulted in different performance levels. The second study indicated that the most structured protocol elicited the highest maximum level of reasoning in children and the highest percentage of correct predictions. The third study showed differences between the protocols in the adult's verbal behavior. Adaptation in verbal behavior to different children by the adult did not result in higher scientific understanding by the children, whereas a higher degree of task structure did. Combined, the studies emphasize the importance of context, which has implications for assessment and teaching situations.