The aim of this thesis is to propose that mimicry of signals in human communication is an effective behavioural strategy for exploiting social systems. Detection of mimicry may consequently have evolved as a counterstrategy, creating a competitive co-evolutionary relationship between signaller and receiver, potentially driving signal complexity and diversity. This perspective relies on previous work on social selection, kin recognition, and tag-based cooperation, which together suggest that mimicry, when successful, is likely to be an evolutionarily beneficial strategy for signallers. Insofar as receivers confer cooperation to those signallers with whom they perceive a shared social or kinship origin, free riding on signals indicating such a relationship will be evolutionarily favoured.

The aim of this thesis is to defend these claims, both through analysis of previously published work and through the presentation of empirical research and agent-based modelling. In chapter 1, the introduction, I introduce the general questions I aim to address and introduce the project’s overall theoretical framework. Chapter 2 is a review of strategies for exploiting relationships, both interspecific and conspecific, across taxa, which suggests that, both in human and non-human relationships, barriers — for example, policing in hymenopterans and social norms in humans — evolve to prevent exploitation. I present evidence in several species, with a particular focus on non-human animal and human social relationships, in support of this view, and conclude that, in competitive relationships where barriers exist, mimicry is likely to be a common strategy for exploitation.

In chapters 3 and 4, I explore, in two empirical studies, human signal mimicry detection in the context of accents, which have previously been proposed as likely tags that direct cooperative behaviours. The aim of these chapters is to suggest, within the context of sociolinguistic and evolutionary literature, that humans are effective at discerning whenspeakers are faking signals of the listeners’ shared social identity, with potential consequences for group boundaries and linguistic diversity.

Chapter 5, an agent-based model, aims to show that detection of mimicry is, itself, an effective barrier against free riding by mimics in social systems. In chapter 6, I develop a novel agent-based modelling paradigm that specifically compares the results in a series of one-shot Prisoner’s Dilemmas when individuals can mimic a cooperative tag. This chapter aims to bring the preceding four together, and proposes that covert mimicry allows free- riding individuals to proliferate successfully in groups of conditional cooperators.

In a final discussion chapter, I evaluate these findings in the context of previous work on strong reciprocity and cultural group selection, and suggest that barriers to exploitation, such as social norms, preclude the possibility of exploitation only insofar as individual actions are detectable.

Mimicry is, I propose, a potentially common strategy for evading these barriers, and insofar as a small subset of individuals mimic effectively, an otherwise cooperative system will be open to exploitation. I discuss this proposal in the context of the preceding chapters, and suggest that human signal mimicry detection is not, on average, effective enough to entirely prevent free riding — with implications for human social interactions in modern life.