Self-efficacy is an affective learning outcome that has been associated with academic performance and retention in STEM. Self-efficacy has been defined as students' beliefs about their ability to complete a given task, and it can be affected by a student's positive or negative experience in a course. In this study, students' chemistry self-efficacy in an organic chemistry course and its reciprocal causation relationship with performance in the course was examined using structural equation modeling (SEM). Organic chemistry self-efficacy (OCSE) was measured five times during the semester and students' scores in four term exams and a final exam were used as measures of performance. Five different models were proposed and tested to explain the relationship between OCSE and performance. A reciprocal causation model with an exam snowball effect was found to best explain the relationship between OCSE and performance compared to other alternative models including a reciprocal causation model without the snowball effect and an autoregressive model. This model suggests a significant positive relationship between OCSE and performance throughout the semester, providing empirical support for the reciprocal relationship of these two measures. Implications for research and instruction are discussed in terms of the reciprocal relationship between self-efficacy and performance.