This paper asks how people treat groups that sit between a clear “us” and a clear “them.” Real life often has such middle-groups: they share some traits with the in-group but do not fully belong. The authors test whether people favour, ignore, or harm these middle-groups when deciding how to split resources.

The researchers ran a multi-lab online experiment with 376 participants. They built a formal “identity marker” setup that varied how many and which markers the middle-group shared with the in-group. Groups were defined by four criteria, for example a minimal painting preference, actual university affiliation, being in the same lab-in-zoom session, or a random assignment. In a private, 60-round die-rolling allocation task, participants could break an unenforceable rule to shift tokens to one group or another. Because choices were private and had no feedback, the design aimed to remove motives like reputation, reciprocity, or demand effects and treat rule-breaking as a measure of intrinsic altruism.

The main findings refine the usual in-group/out-group story. The in-group was consistently favoured and the out-group consistently lost out. A middle-group could be treated neutrally — that is, neither favoured nor harmed — but only in specific cases. When the middle-group shared two identity markers with the in-group, and one of those was the university affiliation, participants produced a clear hierarchy: in-group favoured, middle-group treated according to the allocation rule, out-group harmed. When the middle-group shared only one marker, it was treated like the out-group regardless of which marker was shared. Also, differences in treatment did not appear slowly across rounds; they emerged only after participants saw the instructions and group setup again in a second phase of the experiment.