Tupían languages typically mark evidentiality through freestanding particles located in a predicate- or clause-peripheral position. In Tuparí, however, non-witnessed evidentiality is realized by a bound verbal suffix (-pnẽ/psira). This article draws upon original fieldwork to offer a detailed description and analysis of -pnẽ/psira. I argue that using -pnẽ/psira presupposes commitment to the proposition p on the part of the speaker. This analysis explains the incompatibility between -pnẽ/psira and those clause-typing particles that signal doubt or uncertainty, and it accounts for the projection of the witnessed/nonwitnessed contrast out of embedded clauses. This article also puts forth an explanation for the historical origin of -pnẽ/psira. A separate suffix, resultative -psẽ/pnẽ/psira, is partially homophonous with the evidential but differs from it on several diagnostics. I propose that resultative constructions in the perfect aspect (‘the snake is in the state of having died’) were reinterpreted as non-firsthand statements in the past tense (‘the snake died [non-witnessed]’).

中文翻译：

---

图帕里的非见证证据及其与完美方面的结果结构的联系

图皮亚语言通常通过位于谓词或从句外围位置的独立粒子来标记证据。然而，在图帕里，非目击者的证据是通过绑定的口头后缀 (-pnẽ/psira) 实现的。本文借鉴了原始的实地考察，对 -pnẽ/psira 进行了详细的描述和分析。我认为使用 -pnẽ/psira 的前提是说话者对命题 p 的承诺。该分析解释了 -pnẽ/psira 与那些表示怀疑或不确定的子句类型粒子之间的不兼容性，并解释了从嵌入的子句中投射出的见证/非见证对比。本文还对-pnẽ/psira的历史渊源进行了说明。一个单独的后缀，结果为 -psẽ/pnẽ/psira，与证据部分同音，但在几个诊断方面与证据不同。我建议将完美方面的结果结构（“蛇处于死亡状态”）重新解释为过去时的非第一手陈述（“蛇死了[非目击者]”）。