Credible findings from well-crafted research studies are essential in assessing the impact of child work on children's health. Researchers, however, encounter significant challenges in defining the relevant group of workers for a study and identifying an appropriate comparison group. This article describes some of those challenges and explains how choices about study and comparison groups can lead to biased research results. When selecting study groups, researchers should be aware that the impact of work on health may depend on the type and intensity of the work, and on the context in which it occurs. They should avoid drawing conclusions about the health effects of particular work situations from studies of very heterogeneous groups of workers and should not overgeneralize from studies of more homogenous groups. When choosing comparison groups, researchers should select children whose health outcomes are likely to be comparable to the outcomes working children would experience if they did not work. In particular, researchers should attempt to find children who are similar to the workers of interest on relevant non-work characteristics, including socioeconomic status and levels of parental education. In addition, they should consider the extent to which healthier children are more likely to select into the labor force as a result of decisions by parents or employers, or due to their own greater fitness. Ideally, studies of the health effects of child work should use multiple comparison groups, including children who work in relatively safe, non-strenuous occupations.