Comparing pre/post item response curves to identify changes in misconceptions

Abstract

We use quantitative measures and visual inspection to compare the item response curves (IRCs) of matched pre-/post-instruction Force Concept Inventory (FCI) data. We find that the IRCs are not static; the pre-instruction IRCs differ from the post-instruction IRCs by more than can be explained by random chance. This result is also the case for a subpopulation consisting of students who make little or no gains on the FCI, suggesting that learning is taking place even when scores do not change appreciably. We consider three items where students make substantial progress (item 4) or little progress (items 14 and 21) compared to overall changes in FCI scores.