A fundamental question about speech is whether it is governed by rhythmic constraints. One phenomenon that may support the existence of such constraints is the rhythm rule, a phonological pattern hypothesized to resolve prominence clashes and enforce alternations of prominent and non-prominent syllables via shift/deletion of stress and/or pitch accents. We evaluated evidence for the rhythm rule by studying the acoustic correlates of clash in two experiments with speakers of Italian. We found that the first prominent syllable in a clash displays a durational increase and more extreme formant values, when compared to no clash. Thus, a clash is manifested as a localized decrease in speech rate, not as a change to the prominence profile of a word. Since durational increases have been reported for other languages, we argue that they are an online acoustic correlate of clash. We compare two dynamical models of the durational effects, rooted in the framework of Articulatory Phonology: a π-gesture model and a feedback modulation model. Based on our findings, we argue that the rhythm rule is best conceptualized as the result of contextual biases on lexical selection of prominence patterns.

关于语音的一个基本问题是它是否受节奏约束。可能支持这种限制存在的一种现象是节奏规则，这是一种语音模式，被假设为通过重音和/或音高重音的移位/删除来解决突出冲突并强制突出和非突出音节的交替。我们通过在两个讲意大利语的人的实验中研究冲突的声学相关性来评估节奏规则的证据。我们发现，与没有冲突相比，冲突中的第一个突出音节显示持续时间增加和更极端的共振峰值。因此，冲突表现为语速的局部降低，而不是单词突出轮廓的变化。由于报告了其他语言的持续时间增加，我们认为它们是冲突的在线声学关联。我们比较了两种持续时间效应的动态模型，植根于发音音系框架：π 手势模型和反馈调制模型。根据我们的发现，我们认为节奏规则最好被概念化为语境偏差对突出模式的词汇选择的结果。