Survival is a key vital rate for projecting the viability of wild populations. Estimating survival is difficult for many rare or elusive species because recapture rates of marked individuals are low, and the ultimate fate of individuals is unknown. Low recapture rates for many species have made it difficult to accurately estimate survival, and to evaluate the importance of individual and environmental covariates for survival. Individual covariates such as size are particularly difficult to include in capture–mark–recapture models for elusive species because the state of the individual is unknown during periods when it is not captured. Here, we integrate a von Bertalanffy growth model with a multi‐state robust‐design Cormack‐Jolly‐Seber model to test for a relationship between body size and survival in the elusive, threatened giant gartersnake, Thamnophis gigas. We take a Bayesian approach to model the size of an individual during periods when it was not captured and measured, which fully propagates uncertainty in this unobserved covariate. We found strong support for a positive relationship between snake size and annual survival, with survival increasing with size up to a peak for adult snakes, after which survival either declines slightly or plateaus for the largest individuals. Few captures of very small and very large individuals led to high uncertainty in the survival rates of these sizes. Survival of giant gartersnakes was also positively related to the amount of precipitation and the cover of emergent and floating vegetation at a site. To our knowledge, our study is the first to estimate a size–survival relationship in a snake while fully accounting for uncertainty in the size of unobserved individuals. Our results have implications for the management of this threatened species and illustrate the utility of integrating hierarchical Bayesian models to the study of survival in elusive species.