Surrogate endpoints frequently substitute for rare outcomes in research. The ability to learn about completed suicides by investigating more readily available and proximate outcomes, such as suicide attempts, has obvious appeal. However, concerns with surrogates from the statistical science perspective exist, and mounting evidence from psychometric, neurochemical, genetic, and neuroimaging studies suggests that surrogates may be particularly problematic in suicide research. The need for greater phenotypic refinement of suicide-related behaviors, development of and adherence to a shared suicide nomenclature, and conservative interpretation of investigational results that are limited to the precise population and suicide-related behavior under examination are discussed.