Human social groups are central to social organization and pervasively impact interpersonal interactions. Although immensely varied, all social groups can be considered specific instantiations of a common and abstract ingroup–outgroup structure. How much of the power of human social groups stems from learned variation versus abstract commonality? I review evidence demonstrating that from early in development a wide range of intergroup phenomena, most prominently many ingroup biases, follow solely from simple membership in an abstract social collective. Such effects cannot be attributed to rich social learning, and thus (i) constrain theories seeking to explain or intervene on ingroup bias, and (ii) provide reason to think that our species is powerfully predisposed towards ingroup favoritism from early in development. A wide range of ingroup biases emerge spontaneously following assignment to previously unfamiliar and otherwise meaningless ‘minimal’ social groups. Because they emerge in the absence of richer forms of social learning, these findings challenge explanations for ingroup bias that appeal to the role of environmental messages. Ingroup bias in the minimal groups setting extends far beyond mere preferences in favor of the ingroup, and affects many aspects of learning and memory. Because this emerges early in development, it constitutes a powerful learning gradient favoring the further entrenchment of ingroup bias. Continued advances will require the careful classification of ingroup biases that emerge solely as the result of group membership versus those that depend on additional forms of environmental input. The minimal groups paradigm can be used to perform this sorting, because effects that emerge in that setting by definition do not require additional learning.