Competing perspectives on the nature of emotion are illustrated with latent and emergent variable models. Latent variable models draw from classical test theory, assuming that the measured indicators of emotion covary by virtue of some common executive, organizing neural circuit or network in the brain. By contrast, emergent variable models draw from a theory-driven, operational definition tradition, positing that emotions do not cause, but rather are caused by, the measured indicators of emotion, assuming no executive neural circuit or network, and (...) requiring no covariation among indicators. I suggest that the emergent variable model may be more compatible with the extant literature on the nature of emotion, especially in the age of affective neuroscience. (shrink)