There is increasing attention on the need to identify new immune markers for the evaluation of existing and new influenza vaccines. Immune markers that could predict individual protection against infection and disease, commonly called "correlates of protection" (CoPs), play an important role in vaccine development and licensing. Here, we discuss the epidemiological considerations when evaluating immune markers as potential CoPs for influenza vaccines and emphasize the distinction between correlation and causation. While an immune marker that correlates well with protection from infection may be used as a predictor of vaccine efficacy, it should be distinguished from an immune marker that plays a mechanistic role in conferring protection against a clinical endpoint, as the latter may be a more reliable predictor of vaccine efficacy and a more appropriate target for rational vaccine design. To clearly distinguish mechanistic and non-mechanistic CoPs, we suggest using the term "correlates of protection" for non-mechanistic CoPs, and "mediators of protection" for mechanistic CoPs. Furthermore, as the interactions among and relative importance of correlates or mediators of protection may vary according to age or prior vaccine experience, the effect sizes and thresholds for protective effects for CoPs may also vary in different segments of the population.