Speakers' lexical choices are affected by interpersonal-level influences, like a tendency to reuse an interlocutor's words. Here, we examined how those choices are additionally affected by community-level factors, like whether the interlocutor is from their own or another speech community (in-community vs. out-community partner), and how such interpersonal experiences contribute to the acquisition of community-level linguistic knowledge. Our three experiments tested (i) how speakers' lexical choices varied depending on their partner's choices and speech community, and (ii) how speakers' extrapolation of these choices to a subsequent partner was influenced by their partners' speech communities. In Experiment 1, Spanish participants played two sessions of an online picture-matching-and-naming task, encountering the same pictures but different confederates in each session. The first confederate was either an in-community partner (Spanish) or an out-community partner (Latin American); the second confederate was either from the same community as the first confederate or not. Participants' referential choices in Session 1 were influenced by their partner's choices, but not by their community. However, participants' likelihood to subsequently maintain these choices was affected by their partners' communities. Experiment 2 replicated this pattern in Mexicans, and Experiment 3 confirmed that these results were driven by confederates' communities, rather than perceived linguistic status. Our results suggest that speakers encode speech community information during dialogue and store it to inform future contexts of language use, even when it has not affected their choices during that particular encounter. Thus, speakers learn community-level knowledge by extrapolating linguistic information from interpersonal-level experiences.

说话者的词汇选择受人际交往的影响，例如重用对话者话语的倾向。在这里，我们研究如何将这些选择另外通过社区层面因素的影响，如是否对话者是从自己或他人的言语社区（在社区对出社区合作伙伴），以及这种人际关系的经验如何有助于获得社区一级的语言知识。我们的三个实验测试了（i）说话者的词汇选择如何根据其伴侣的选择和言语社区而变化，以及（ii）说话者将这些选择外推至后续伴侣的方式如何受到其伴侣的言语社区的影响。在实验1中，西班牙参与者参加了两次在线图片匹配和命名任务，每次遇到相同的图片，但联盟不同。第一个同盟者是社区内的伙伴（西班牙）或社区外的伙伴（拉丁美洲人）; 第二个同盟或与第一个同盟来自同一社区。参与者在会议1中的推荐选择受其伴侣选择的影响，但不受其社区的影响。但是，参与者随后保留这些选择的可能性受到其伙伴社区的影响。实验2在墨西哥人中复制了这种模式，实验3证实了这些结果是由同盟者的社区所驱动，而不是由感知的语言地位所驱动。我们的结果表明，即使对话在特定的遭遇中没有影响他们的选择，说话者也可以在对话过程中对语音社区信息进行编码，并将其存储以告知将来的语言使用情况。从而，