"Morning sickness" is the common term for nausea and vomiting in early human pregnancy (NVP). Recent interest in why NVP occurs-that is, in the evolutionary costs and benefits of NVP-has spurred the development of two alternative hypotheses. The "prophylaxis," or "maternal and embryonic protection," hypothesis suggests that NVP serves a beneficial function by expelling foods that may contain harmful toxins and microorganisms and triggering aversions to such foods throughout pregnancy. The alternative "by-product" hypothesis suggests that NVP is a nonfunctional by-product of conflict--over resource allocation--between the pregnant woman and the embryo. The critical predictions of the prophylaxis hypothesis have been developed and tested, whereas the by-product hypothesis has not been subjected to similar scrutiny. To address this gap, we developed a graphical model and used it to derive predictions from the by-product hypothesis under two different assumptions, namely, that NVP is either (i) a by-product of current conflict between a pregnant woman and an embryo or (ii) a by-product of honest signals of viability produced by the embryo. Neither version of the by-product hypothesis is fully consistent with available data. By contrast, the timing of NVP, its variation among societies, and associated patterns of food cravings and aversions are consistent with the prophylaxis hypothesis.