Assessing students’ understanding about levers: better test instruments are not enough

This study was designed to investigate variations in students’ responses to lever problems across multiple assessment contexts and formats. Prior to and at the end of a four‐month science unit on simple machines, grade 6/7 students prepared semantic maps, wrote responses to application questions, discussed their written answers, and modelled solutions to practical problems using physical materials. We present data from pre‐ and post‐instruction interviews to show that students’ responses varied in kind and extent across assessment contexts and formats. Such variations in performance are consistent with situated cognition theories which assert that competence is heterogeneous across situations and is a function of interactions between individuals and the contexts in which they perform. Based on these findings, implications are drawn regarding more productive means of assessing students’ understanding in classrooms.