In the real world, listeners seem to implicitly learn talkers’ vocal identities during interactions that prioritize attending to the content of talkers’ speech. In contrast, most laboratory experiments of talker identification employ training paradigms that require listeners to explicitly practice identifying voices. Here, we investigated whether listeners become familiar with talkers’ vocal identities during initial exposures that do not involve explicit talker identification. Participants were assigned to one of three exposure tasks, in which they heard identical stimuli but were differentially required to attend to the talkers’ vocal identity or to the verbal content of their speech: (1) matching the talker to a concurrent visual cue (talker-matching); (2) discriminating whether the talker was the same as the prior trial (talker 1-back); or (3) discriminating whether speech content matched the previous trial (verbal 1-back). All participants were then tested on their ability to learn to identify talkers from novel speech content. Critically, we manipulated whether the talkers during this post-test differed from those heard during training. Compared to learning to identify novel talkers, listeners were significantly more accurate learning to identify the talkers they had previously been exposed to in the talker-matching and verbal 1-back tasks, but not the talker 1-back task. The correlation between talker identification test performance and exposure task performance was also greater when the talkers were the same in both tasks. These results suggest that listeners learn talkers’ vocal identity implicitly during speech perception, even if they are not explicitly attending to the talkers’ identity.

在现实世界中，听众似乎在优先关注说话者语音内容的交互过程中隐式地学习了说话者的声音身份。相比之下，大多数说话者识别的实验室实验都采用要求听众明确练习识别声音的训练范式。在这里，我们调查了在不涉及明确说话者识别的初始暴露期间，听众是否熟悉说话者的声音身份。参与者被分配到三个暴露任务中的一个，在这些任务中，他们听到相同的刺激，但需要不同地注意说话者的声音身份或他们演讲的语言内容：（1）将说话者与同时出现的视觉提示（说话者-匹配); (2) 判别说话者是否与之前的试验相同( talker 1-back )；或（3）判别语音内容是否与之前的试验匹配（verbal 1-back). 然后测试所有参与者学习从新颖的语音内容中识别说话者的能力。至关重要的是，我们操纵了此后测试期间的说话者是否与训练期间听到的不同。与学习识别新说话者相比，听众在说话者匹配和语言 1-back 任务中更准确地学习识别他们之前接触过的说话者，但不是说话者 1-back 任务。当说话者在两个任务中相同时，说话者识别测试性能与暴露任务性能之间的相关性也更大。这些结果表明，听众在语音感知过程中隐式地学习了说话者的声音身份，即使他们没有明确注意说话者的身份。