London dispersion forces (LDFs) are the most fundamental of the noncovalent interactions in that they are interactions that act between all molecular species and require only that students can use the "electron cloud" model of atomic structure. In this longitudinal study we investigate how students respond to prompts, designed to elicit causal mechanistic reasoning about the origins of LDFs, using a modified coding scheme from a previous study. We find that including explicit scaffolding improves students' representations of the mechanism compared to student responses to an unscaffolded prompt. When, on a later administration of a similar prompt, the scaffolding is removed, the sophistication of the students' representations decreases, but does not decline to the level of students who had never been exposed to a scaffolded prompt. We also explore the connection between student drawing and writing and find that there is a significant association between the reasoning level of the drawing and the written explanation. Finally, we show that there appears to be no significant difference between performance on ungraded (formative) and graded (summative) items that are administered in the second semester when LDFs are not explicitly covered. That is, students respond to ungraded homework activities similarly to examinations at time points where LDFs are not explicitly discussed in the course.