This study investigates how collaborative knowledge construction within an inquiry learning environment can be assisted with scaffolds that aim to support students' hypothesis generation process. Sixty-six students on a university preparatory track worked collaboratively on a kinematics task. The instructional goal was to develop students' understanding of one-dimensional kinematics. All students completed a proposition test in which they indicated their individual opinions about the truth-value of specific propositions. Subsequently, students were coupled into dyads and assigned to one of three conditions: (a) a shared proposition scratchpad (expression builder), (b) a shared proposition table, and (c) a control condition. Students in the scratchpad condition were given an expression builder consisting of dropdown menus with pre-defined variables and relations. Students in the shared proposition table condition could combine individual opinions about the truth-value of a proposition into one shared proposition table that highlighted differences in opinion. Students in the control condition received no extra support related to propositions. Learning outcomes were assessed using a variety of pre- and posttests. The findings indicate that students supported with the shared proposition table showed significant gains for qualitative knowledge about relations. The number of unique propositions students discussed during the learning session was a significant predictor of learning gains. A more detailed analysis of students' interaction protocols suggests that students differed in their task perceptions, their interaction patterns, and their use of prior knowledge. (Contains 1 footnote, 11 tables, and 6 figures.)