Summary: We derive confidence regions for the evolutionary trajectories derived from a model of \textit{S. Via} and \textit{R. Lande} [Evolution 39, 505-522 (1985)]. We utilize a nested, parametric bootstrap to calculate the confidence regions and show that a likelihood-based approach provides an unsatisfactory solution. We calculate the predicted evolutionary trajectories and confidence regions using data from an experiment on pupal weight of Tribolium beetles in two different flour types. With realistic sample sizes for experiments that estimate genetic parameters in multiple environments, the confidence regions are likely to be large, potentially limiting the ability to test hypotheses about evolutionary trajectories. The problem may be less acute for trajectories of characters expressed in only a single environment, because sample sizes can generally be larger than in multienvironmental studies.