Lahiri A, Gewirth L, Blumstein S E
J Acoust Soc Am. 1984 Aug;76(2):391-404. doi: 10.1121/1.391580.
This study explored the claim that invariant acoustic properties corresponding to phonetic features generalize across languages. Experiment I examined whether the same invariant properties can characterize diffuse stop consonants in Malayalam, French, and English. Results showed that, contrary to theoretical predictions, we could not distinguish labials from dentals, nor could we classify dentals and alveolars together in terms of the same invariant properties. We developed an alternative metric based on the change in the distribution of spectral energy from the burst onset to the onset of voicing. This metric classified over 91% of the stops in Malayalam, French, and English. In experiment II, we investigated whether the invariant properties defined by the metric are used by English-speaking listeners in making phonetic decisions for place of articulation. Prototype CV syllables--[b d] in the context of [i e a o u]--were synthesized. The gross shape of the spectrum was manipulated first at the burst onset, then at the onset of voicing, such that the stimulus configuration had the spectral properties prescribed by our metric for labial and dental consonants, while the formant frequencies and transitions were appropriate to the contrasting place of articulation. Results of identification tests showed that listeners were able to perceive place of articulation as a function of the relative distribution of spectral energy specified by the metric.
本研究探讨了与语音特征相对应的不变声学特性是否能跨语言推广这一说法。实验一考察了相同的不变特性是否能表征马拉雅拉姆语、法语和英语中的浊塞音。结果表明,与理论预测相反,我们无法区分唇音和齿音,也无法根据相同的不变特性将齿音和齿龈音归为一类。我们基于从爆发开始到发声开始的频谱能量分布变化开发了一种替代度量。该度量对马拉雅拉姆语、法语和英语中超过91%的塞音进行了分类。在实验二中,我们研究了以英语为母语的听众在做出发音部位的语音决策时是否使用了该度量所定义的不变特性。合成了原型CV音节——在[i e a o u]语境中的[b d]。首先在爆发开始时,然后在发声开始时操纵频谱的总体形状,使得刺激配置具有我们的度量为唇音和齿音规定的频谱特性,同时共振峰频率和过渡适合于对比的发音部位。识别测试结果表明,听众能够将发音部位感知为该度量所指定的频谱能量相对分布的函数。