Seemingly arbitrary cues—be they linguistic, sartorial, or behavioral—can come to serve as markers of group membership. However, such markers are only effective if the benefits of using them outweigh the costs of being recognized by hostile outgroups, as is the case often faced by oppressed minorities, political dissidents, and others in an increasingly polarized society. I will present mathematical and computational modeling, along with motivating empirical work, to show how stable identity signals can be disrupted, leading to cyclical lifespans or even the total suppression of effective group markers as the potential cost of being identified by a hostile outgroup increases. I will then discuss implications for our understanding of communication and the censorship of dissent in both on- and offline communities.