This investigation considers the effects of learner-generated headings on memory. Participants (N = 63) completed a computer-based lesson with or without learner-generated text topic headings. Posttests included a cued recall test of factual knowledge and a sorting task measure of structural knowledge. A significant disordinal interaction was observed with the no headings group performing better on the cued recall test and worse on the structural knowledge sorting task. In addition, a measure of structural knowledge using a list-wise rating task was field trialed. The two structural knowledge tasks were analyzed using a Pathfinder network scoring technique that measured both linear and nonlinear text subtopic organization. The sorting and rating tasks were differentially sensitive to participants' linear and nonlinear knowledge organization. Unexpectedly, the sorting and rating posttest tasks improved cued recall posttest scores. The implications of the results for measuring structural knowledge are considered and suggestions for further studies are presented.Generative learning approaches are strategies that encourage learners to actively consider the relationships among lesson information elements and between lesson information and personal knowledge (Jonassen,