Evidence on the coherence–pieces debate from the force concept inventory

We use force concept inventory (FCI) data to probe the consistency of commonsense physics as a knowledge system. The source of this data is the administration of the FCI to first-year science university students. Data quality was checked using item response theory and studying answer distributions for each question. We find apparently paradoxical results: depending on how the data is analysed, answers seem highly systematic or almost random-like. These results are compatible with others found in the literature and can be construed as arising either from a coherent knowledge system or from knowledge in pieces. We hypothesise as a possible source of this apparent contradiction that predictions and explanations use different resources: the former would use reflex, low-cost cognitive resources while the latter would involve con-ceptualisations. We show that the articulation of both resources may be crucial for expert thinking productivity (the ability to apply a theory to novel situations). We sketch some consequences of the proposed structure of commonsense thinking for teaching and further research.