【摘要】 本文结合语料库标注和跨语言细节分析，讨论一种新的特征方案——主动发音器官模型(Active Articulator Model, AM),并检测其工作能力。AM通过引入动态、动作导向的视角，统一描写元音和辅音，为语音学和音系学提供了一种新的分析工具。大型音位数据库P-base包含了600多个语言的音位系统和900多个语音，收录比较齐全，符号比较统一。我们首先把非音段层面的语音排除，然后使用AM为余下的语音赋值，接下来核查赋值完全相同的语音出现的音位系统和音系说明，分析是否为变体关系。结果表明，AM基本上可以有效地区分P-base中的所有辅音。颤音类多为变体，文献已有广泛报道，因此本文暂不处理。同时，本文通过对P-base中若干语言的细节分析，讨论了传统方法在音系归纳中存在的分歧。AM不仅有助于统一和简化语音描写，还可以帮助研究人员更好地理解和分析语音系统，减少传统方法中的分歧。更多还原

【Abstract】 Active Articulator Model(AM) is a developing feature model, which proposes six active articulators and their actions to describe all the speech sounds, introducing a dynamic, gesture-oriented perspective to phonetics and phonology to create a unified framework for describing both vowels and consonants. This article, focusing on consonants, aims to discuss AM and test whether the features based on the actions and possible combinations of actions of these active articultors are sufficient to distinguish all the segmental contrasts in human languages, by checking AM against the International Phonetic Alphabet and P-base, a database containing phonemic inventories from over 600 languages and more than 900 phonemes. Firstly, all the segments in P-base are discussed and the sounds bearing non-segmental features, such as diphthongs, consonantal clusters, sounds with tone contrast or length contrast, are excluded. Next the remaining sounds are specified in terms of a set of phonetic features provided by AM. It is found that some phonemes share the same specification, which indicates that they represent the same sound from the perspective of AM, even though transcribed with different symbols. This hypothesis needs to be verified by checking whether these phonemes with the same specification co-occur in the same phoneme inventory in P-base. For the co-occurring phonemes with the same specification, the related phonological processes provided in these inventories are further checked. The results show that these co-occurring phonemes are in complementary distribution, except for trills and flaps/taps, which present a much higher frequency of co-occurrence with stops, fricatives, or other r-sounds. Trills seem not to be specified with the features provided by AM, since they are not generated by the active actions of active articulators. However, considering it is widely reported that all the r-sounds including trills serve as variants across languages, and it is challenging to describe their articulation, it may be best not to analyze trills or even all the r-sounds as contrasting phonemes, and leave the description of their articulatory physiology for future. The results point to the conclusion that AM is powerful enough to distinguish almost all the consonantal contrasts and hence can serve as an alternative tool in phonemic analysis. Furthermore, the detailed cross-linguistic analysis of phonemic inventories in P-base shows that AM demonstrates the potential in reducing discrepancies in phonemic analysis, especially in transcription and variant processing, by establishing a clear, unified standard that is expected to help researchers better understand and analyze phonological processes.更多还原