Rodd Joe, Chen Aoju
Office of Education, Ministry of Education, Culture and Science, The Hague, Netherlands.
Institute for Language Sciences, Utrecht University, Utrecht, Netherlands.
Front Psychol. 2023 Jan 17;13:911349. doi: 10.3389/fpsyg.2022.911349. eCollection 2022.
The question of whether intonation events are speech categories like phonemes and lexical tones has long been a puzzle in prosodic research. In past work, researchers have studied categoricality of pitch accents and boundary tones by examining perceptual phenomena stemming from research on phoneme categories (i.e., intonation boundary effects-peaks in discrimination sensitivity at category boundaries, perceptual magnet effects-sensitivity minima near the best exemplar or prototype of a category). Both lines of research have yielded mixed results. However, boundary effects are not necessarily related to categoricality of speech. Using improved methodology, the present study examines whether pitch accents have domain-general internal structure of categories by testing the perceptual magnet effect. Perceived goodness and discriminability of re-synthesized productions of Dutch rising pitch accent (L*H) were evaluated by native speakers of Dutch in three experiments. The variation between these stimuli was quantified using a polynomial-parametric modeling approach. A perceptual magnet effect was detected: (1) rated "goodness" decreased as acoustic-perceptual distance relative to the prototype increased (Experiment 1), and (2) equally spaced items far from the prototype were more frequently discriminated than equally spaced items in the neighborhood of the prototype (Experiment 2). These results provide first evidence for internal structure of pitch accents, similar to that found in color and phoneme categories. However, the discrimination accuracy gathered here was lower than that reported for phonemes. The discrimination advantage in the neighborhood far from the prototype disappeared when participants were tested on a very large number of stimuli (Experiment 3), similar to findings on phonemes and different from findings for lexical tones in neutral network simulations of distributional learning. These results suggest a more transient nature of the perceptual magnet effect in the perception of pitch accents and arguably weaker categoricality of pitch accents, compared to that of phonemes and in particular of lexical tones.
语调事件是否像音素和词汇调一样属于语音范畴,这个问题长期以来一直是韵律研究中的一个难题。在过去的研究中,研究人员通过研究源于音素范畴研究的感知现象(即语调边界效应——范畴边界处辨别敏感性的峰值,感知磁体效应——在一个范畴的最佳范例或原型附近的敏感性最小值)来研究音高重音和边界调的范畴性。这两条研究路线都得出了混合的结果。然而,边界效应不一定与语音的范畴性相关。本研究采用改进的方法,通过测试感知磁体效应来检验音高重音是否具有范畴的领域通用内部结构。以荷兰语为母语的人在三个实验中对重新合成的荷兰语升调重音(L*H)的发音的感知优劣度和可辨别性进行了评估。使用多项式参数建模方法对这些刺激之间的差异进行了量化。检测到了一种感知磁体效应:(1)随着相对于原型的声学-感知距离增加,评定的“优劣度”降低(实验1),并且(2)远离原型的等距项目比原型附近的等距项目更频繁地被辨别出来(实验2)。这些结果为音高重音的内部结构提供了首个证据,类似于在颜色和音素范畴中发现的结构。然而,这里收集到的辨别准确率低于报道的音素辨别准确率。当对参与者进行大量刺激测试时(实验3),远离原型区域的辨别优势消失了,这类似于音素的研究结果,并且与分布学习的神经网络模拟中词汇调的研究结果不同。这些结果表明,与音素尤其是词汇调相比,音高重音感知中的感知磁体效应具有更短暂的性质,可以说音高重音的范畴性较弱。