We develop a mechanistically motivated von Bertalanffy growth model to estimate growth rate and its predictors from spatial–temporal data and compare this model’s performance with a suite of commonly used mixed-effects growth models. We test these models with simulated data and then apply them to test whether concerns that high density is causing growth suppression of walleye (Sander vitreus) in Alberta, Canada, are supported using data collected during 2000–2017. Simulation experiments demonstrated that models that failed to account for complex dependency structures often resulted in growth rate estimates that were less accurate and biased low as judged by median absolute relative error and median relative error, respectively. The magnitude of this bias depended on the parameter values used for simulation. For the case study, a spatial–temporal model was more parsimonious and had higher predictive performance relative to simpler models and did not support the slow-growing walleye hypothesis in Alberta. These findings demonstrate the importance of considering spatial–temporal correlation in analyses that rely on surveillance-style monitoring datasets, particularly when examining relationships between life-history traits and environmental characteristics.