1: \begin{abstract}

2: A key property of linguistic conventions is that they hold over an

3: entire community of speakers, allowing us to communicate efficiently

4: even with people we have never met before. At the same time, much of our

5: language use is partner-specific: we know that words may be understood

6: differently by different people based on our shared history. This poses

7: a challenge for accounts of convention formation. Exactly how do agents

8: make the inferential leap to community-wide expectations while

9: maintaining partner-specific knowledge? We propose a hierarchical

10: Bayesian model to explain how speakers and listeners solve this

11: inductive problem. To evaluate our model's predictions, we conducted an

12: experiment where participants played an extended natural-language

13: communication game with different partners in a small community. We

14: examine several measures of generalization and find key signatures of

15: both partner-specificity and community convergence that distinguish our

16: model from alternatives. These results suggest that partner-specificity

17: is not only compatible with the formation of community-wide conventions,

18: but may facilitate it when coupled with a powerful inductive mechanism.

19: 

20: \textbf{Keywords:}

21: learning; communication; coordination

22: \end{abstract}

23: