本研究係探討三十名國小聽障學生讀話所見的國音聲母視素形成內涵，以及對於不同說話者所說出的聲母，聽障受試讀話所見的視素差異。本研究所用的研究工具為自編之「注音符號讀話能力評量表聲母篇」。所得資料以群落分析和卡方考驗進行處理。本研究的主要發現如下: 1.受試只能將二十一個聲母區分出兩群落：唇音[ㄅㄆㄇㄈ]與非唇音[ㄉㄊㄋㄌㄍㄎㄏㄐㄑㄒㄓㄔㄕㄖㄗㄘㄙ]。 2.各個聲母讀話正確率從94.7)%到5.29％不等，平均正確率是28.28%。而聲母視素讀話正確率為：[ㄅㄆㄇㄈ]視素96.24%；[ㄉㄊㄋㄌㄍㄎㄏㄐㄑㄒㄓㄔㄕㄖㄗㄘㄙ]視訴98.96％。兩個視素內個別的音案錯誤雖不少，但辨識錯誤大都發生在同視素群內，表示同一視素內的聲母會有視覺上易混淆的特性。 3.本研究結果依照聲母是否為唇音，可區分出二個視素，可部分支持發音部位不同會產生不同視素的假說。 4.觀看不同說話者的口形，並末導致讀者所見之視素有顯著差異。但三名說話者的口形動作仍有些微差異：(1)由群落凝聚的過程，可以看到某些音素與其他音凝聚的順序不同；(2)受試觀看三名說話者的口形動作，所得之音素辨識率以丙說話者最高，乙說話者最低；(3)三名說話者說出的聲母中，ㄅ、ㄌ、ㄏ、ㄓ四音的得分有顯著差異。

　　　　　The purposes of this study were to explore the contents of consonant visemes lipread by Grade Six students with hearing-impairment, and to examine whether talker differences existed by way of different visemes revealed. 30 subjects from 4 resource programs for the hearing-impaired in Taipei City, lipread a set of 259 Cy syllables (formed by 21 Manderine consonants plus /a/ or /ia/ vowels). Their written answers were analyzed via Agglomerative hierarchical cluster analysis. The results were as follows: 1. Two clusters were formed: labials [ㄅㄆㄇㄈ] and non-labials [ㄉㄊㄋㄌㄍㄎㄏㄐㄑㄒㄓㄔㄕㄖㄗㄘㄙ]. 2. Correct response percentage of each consonant varied: ranged from 94.71% to 5.29%, with average 28.28%. Correct response percentage of each cluster was high: 96.24% and 98.96%respectively. Most of the errors were located within the very cluster, which indicated that consonants within the same cluster were highly confused visually. 3. There seemed to be no significant differences in visemes formed across three talkers: minute discrepencies, however, did emerge in two areas: (1) the sequences of consonants clustered differed; and (2) scores of four consonants showed significant differences.