The popularity of growth modeling has increased in psychological and cognitive development research as a means to investigate patterns of changes and differences between observation units over time. Random coefficient modeling, such as multilevel modeling and latent growth curve modeling as a special application of structural equation modeling are two widely used frameworks for traditional growth modeling. Although random coefficient and latent growth curve modeling approaches emerged from the literature independently, both approaches provide identical sets of estimates for the average level of proficiency (such as in mathematics and reading) at a given point in time and illustrate that proficiency's trajectory of growth for the population of interest. Also, both approaches allow the incorporation of time-invariant and time-varying covariates into the model. Random coefficient and latent growth curve modeling approaches typically assume that the parameters of interest are sampled from a single, finite population. Growth mixture modeling (GMM) has been applied in some studies, such as those investigating adolescent alcohol use, to identify latent classes that are characterized by different initial status and developmental trajectories. However, the popularity of this approach is still in its infancy in educational research. The goal of this study is, therefore, to demonstrate an application of GMM using a longitudinal, large-scale reading test data set collected over 4 years. Specifically, this study demonstrates: (a) detection of latent classes and decision making on the appropriate number of latent classes; (b) examination of the initial proficiency and growth trajectory for reading ability in distinct classes of the study population; and (c) interpretation of the identified classes based on the characteristics of their growth patterns. (Contains 8 tables and 4 figures.)