Once believed uniquely human, the capacity to reason is now investigated in a wide range of species. One component of this ability, inference by exclusion, has been traditionally explored through the cups task, where 2 containers are presented but only 1 covers a food reward (if Cup A is empty, then choose Cup B). Often based on low-level cognitive mechanisms (learning), performance on this task can also reflect reasoning processes in some individuals. Limited taxonomic investigation, however, has hindered our understanding of the factors that contribute to the evolution of reasoning in animals. Using the cups task, we first investigated the ability of 14 semiwild donkeys to locate a food reward by using visual or acoustic cues alone (pretest phase). Although all subjects failed the task when provided with visual cues, 7 donkeys succeeded in the acoustic modality. We thus tested the capacity of donkeys to make choices by exclusion in the acoustic modality (test phase). Three conditions were presented, where subjects received information about either both containers (full information condition), solely the empty container (exclusion condition), or no information (control condition). Most subjects chose the correct container in the full information condition, whereas they chose randomly in the control condition. Except for 1 individual, donkeys improved their performance across trials in the exclusion condition. Our study contributes to the very small body of literature on exclusion based on acoustic cues in animals and paves the way to further experiments on the cognitive processes underlying exclusion performance in donkeys. (PsycInfo Database Record (c) 2022 APA, all rights reserved).