Emotions in foreign language (FL) learning have become an increasingly popular research avenue in applied linguistics. Such research is often carried out using linear correlational or regression methods, which are limited in the extent to which they can depict interrelations. In this study, we aimed to reexamine the relationships between the emotion variables of foreign language classroom anxiety (FLCA) and foreign language enjoyment (FLE) and the outcome variable of self-perceived proficiency (SPP) through nonlinear methods. Specifically, we examined how different levels of FLCA and FLE influenced SPP in concert in a sample of N = 1,039 FL learners by utilizing polynomial regression with response surface analysis. The nonlinear method generated a response model that visually depicted a complex SPP pattern as a result of differing levels of FLCA and FLE. This three-dimensional model provides considerable insight into the interaction of positive and negative emotions above and beyond what could be derived from "classical" methods of correlation and linear regression.