This study explores three understudied facets--quadratic effects, change over time, and gender as a moderator--of the otherwise well-documented relationships between statistics anxiety and academic performance. Using pre- and post- course survey data among a sample of 111 undergraduate students in Social Statistics courses at a U.S. Midwestern university, we employ hierarchical linear modeling (HLM) to test for relationships between change in the six dimensions of the Statistics Anxiety Rating Scale (STARS) and exam grades over the course of the semester. We find that exam grades decreased over time, but at different rates across gender and the six STARS dimensions. We also identify a quadratic relationship between self-concept and final exam grades, as well as several gender interactions. This study highlights the multifaceted and dynamic nature of statistics anxiety/attitudes, as its relationship with academic performance is not always negative, linear, stable over time, or uniform across gender. Supplementary materials for this article are available online.