Different speakers sometimes convey similar meanings differently. This study examined whether listeners could learn to associate a specific linguistic ‘style’ with a certain speaker, with no apparent difference in meaning, and the role of unnatural linguistic choices (or unexpectedness) in such learning. We created an inter-speaker variation in ‘style’ using the weak adjective ordering preferences in Hebrew. Participants were exposed to two different speakers, each producing a different adjective order, consistently. We manipulated the combinations of order pairings, based on their naturalness (with two natural orders, a natural and an unnatural order, and two unnatural orders), and examined participants’ ability to associate a unique order with a specific speaker. In two experiments, using different statistical analyses, we show that listeners can learn speaker-specific language use when it is irrelevant for meaning inferences, when deviance from natural or expected language use is involved. We further discuss whether learning may be facilitated by differences in naturalness or structural form. Our findings suggest that listeners are sensitive to inter-speaker variability in ‘style’, mostly when this ‘style’ is unexpected. This is in line with the predictions of Surprisal theory, and may suggest that surprisal plays a major role in learning speaker-specific language use.

不同的说话者有时会以不同的方式传达相似的含义。这项研究考察了听众是否能够学会将特定的语言“风格”与特定的说话者联系起来，而在含义上没有明显的差异，以及非自然的语言选择（或意外性）在这种学习中的作用。我们使用希伯来语中的弱形容词排序偏好创建了“风格”的不同说话者之间的变化。参与者接触到两个不同的说话者，每个人都一致地发出不同的形容词顺序。我们根据顺序配对的自然性（两个自然顺序，一个自然顺序和一个非自然顺序，以及两个非自然顺序）操纵顺序配对的组合，并检查参与者将独特顺序与特定说话者关联起来的能力。在两个实验中，使用不同的统计分析，我们表明，当涉及与自然或预期语言使用的偏差时，听众可以学习特定于说话者的语言使用，当它与意义推断无关时。我们进一步讨论自然性或结构形式的差异是否可以促进学习。我们的研究结果表明，听众对说话者之间“风格”的变化很敏感，尤其是当这种“风格”出乎意料时。这与惊奇理论的预测一致，并且可能表明惊奇在学习特定说话者的语言使用中发挥着重要作用。