Department of Psychological Sciences, University of Connecticut, Storrs, CT, 06269-1020, USA.
Department of Speech, Language and Hearing Sciences, University of Connecticut, Storrs, CT, USA.
Atten Percept Psychophys. 2021 Aug;83(6):2367-2376. doi: 10.3758/s13414-021-02317-x. Epub 2021 May 4.
Researchers have hypothesized that in order to accommodate variability in how talkers produce their speech sounds, listeners must perform a process of talker normalization. Consistent with this proposal, several studies have shown that spoken word recognition is slowed when speech is produced by multiple talkers compared with when all speech is produced by one talker (a multitalker processing cost). Nusbaum and colleagues have argued that talker normalization is modulated by attention (e.g., Nusbaum & Morin, 1992, Speech Perception, Production and Linguistic Structure, pp. 113-134). Some of the strongest evidence for this claim is from a speeded monitoring study where a group of participants who expected to hear two talkers showed a multitalker processing cost, but a separate group who expected one talker did not (Magnuson & Nusbaum, 2007, Journal of Experimental Psychology, 33[2], 391-409). In that study, however, the sample size was small and the crucial interaction was not significant. In this registered report, we present the results of a well-powered attempt to replicate those findings. In contrast to the previous study, we did not observe multitalker processing costs in either of our groups. To rule out the possibility that the null result was due to task constraints, we conducted a second experiment using a speeded classification task. As in Experiment 1, we found no influence of expectations on talker normalization, with no multitalker processing cost observed in either group. Our data suggest that the previous findings of Magnuson and Nusbaum (2007) be regarded with skepticism and that talker normalization may not be permeable to high-level expectations.
研究人员假设，为了适应说话者产生语音的变化，听者必须进行说话者归一化的过程。这一假设与几项研究一致，这些研究表明，与所有语音都由一个说话者产生相比，当语音由多个说话者产生时，口语识别会变慢（多说话者处理成本）。Nusbaum 和同事认为，说话者归一化受注意力的调节（例如，Nusbaum & Morin，1992，语音感知、产生和语言结构，第 113-134 页）。支持这一说法的最有力证据之一来自一项快速监控研究，该研究中，一组预期听到两个说话者的参与者表现出多说话者处理成本，但另一组预期听到一个说话者的参与者则没有（Magnuson & Nusbaum，2007，实验心理学杂志，33[2]，391-409）。然而，在这项研究中，样本量很小，关键的交互作用并不显著。在本注册报告中，我们展示了一项强有力的尝试，旨在复制这些发现。与之前的研究不同，我们在我们的两个组中都没有观察到多说话者处理成本。为了排除零结果是由于任务限制的可能性，我们进行了第二项实验，使用了快速分类任务。与实验 1 一样，我们发现期望对说话者归一化没有影响，两个组都没有观察到多说话者处理成本。我们的数据表明，Magnuson 和 Nusbaum（2007）的先前发现应持怀疑态度，并且说话者归一化可能不受高级期望的影响。
