This paper presents the results of an acoustic investigation of /f, v, s, z/ in Greek, Serbian, Russian, and English. The study is motivated by phonological considerations, specifically the cross-linguistic phonological identity of /v/ as either an obstruent (Greek, English), a sonorant (Serbian), or as a segment that patterns with both obstruents and sonorants (Russian). The investigation is framed in two ways, reflecting different interpretations of what it means for /v/ to be classified as part of the obstruent or sonorant system of a language: (1) a cross-linguistic comparison of /v/ tokens tackles the question of whether /v/ tokens are realized with frication indicative of an obstruent or sonorant realization; (2) within-language investigations into the relationship between voicing and frication type probe whether /f, v/ are a voicing pair analogous to /z, s/. Four acoustic measures are considered: duration, harmonicity, spectral centroid, and spectral energy difference. Furthermore, devoicing rates of /v, z/ are examined, contributing to our understanding of how aerodynamic tensions in voiced fricatives are resolved on a language-specific basis. Results show that in all languages, /z/ devoices more than /v/, but that otherwise fidelity to underlying voicing differs across languages. The cross-linguistic comparison of /v/ tokens and the within-language investigations suggests a partial correlation between phonological identity and phonetic realization, but that there is not a one-to-one relationship between phonological patterning and phonetic realization.

本文介绍了希腊语、塞尔维亚语、俄语和英语中 / f、v、s、z / 的声学调查结果。这项研究的动机是语音考虑，特别是 / v /的跨语言语音识别作为一个阻塞音（希腊语、英语）、一个响音（塞尔维亚语），或者作为一个同时具有阻塞音和响音模式的片段（俄语）。调查以两种方式进行，反映了对 / v / 被归类为语言的阻塞或响亮系统的一部分意味着什么的不同解释：（1） / v / 标记的跨语言比较解决了这个问题是否/ v/ 标记是通过摩擦来实现的，表示一种阻碍或响亮的实现；(2) 语言内对浊音与摩擦类型关系的调查探讨 / f , v / 是否是类似于 / z , s / 的浊音对。考虑了四种声学测量：持续时间、谐波、频谱质心和频谱能量差。此外，还检查了 / v、z / 的去清音率，有助于我们了解浊擦音中的空气动力学张力是如何在特定语言的基础上解决的。结果表明，在所有语言中，/ z / 变音多于 / v/，但在其他方面对基础发声的保真度因语言而异。/ v / 标记的跨语言比较和语言内调查表明语音身份和语音实现之间存在部分相关性，但语音模式和语音实现之间没有一对一的关系。