Emergence of accuracy is a first step toward acquiring adult-like phonological abilities. Mandarin Chinese is a tone language where speakers employ both tonal and segmental properties to code lexical meanings. Study of this dual-level complexity of tone and segment enables a broader view of how phonology is acquired than the view afforded by study of Indo-European languages. Understanding the interaction between phonatory properties for tone and articulatory adjustments for segments in emergence of early words helps to understand more generally the first steps toward the complex system embodied in phonology. The present study investigated tone acquisition in relation to segmental and lexical development in Mandarin-speaking children in the earliest word stages. Spontaneous speech samples were collected longitudinally from 12 to 24 months from four Mandarin-speaking children. The relationships between tone accuracy, segmental accuracy, and word-level variables were examined quantitatively over time. Results indicated that tone accuracy is not always higher than segmental accuracy. The relationship between these two seems to be influenced by the physiological complexity of tonal shapes and children’s developmental age. Autonomy of control over phonatory adjustments for tone and articulatory adjustments for segments was already apparent. Children were not sensitive to the contrastivity (characterized by Productive Tone Neighborhood Density) involved in tonal categories with a vocabulary of less than 50 words. Associations between production accuracy and word-level variables (articulatory complexity, neighborhood density and word frequency) established based on later developmental periods were not found in younger Mandarin-speaking children. Findings suggest that tonal acquisition at the onset of speech development is not a passive process where innate phonological knowledge is revealed solely through children’s maturation. Rather, phonological knowledge is established on the basis of children’s pre-linguistic motor capacities in concert with cognitive learning occurring via the expansion of their lexicon. Tones and segments may be produced as holistic entities in early words. Tone acquisition at the onset of word learning is more child-centered in that availability of tonal forms to the child’s production system underlies accuracy. Influences from lexical properties of word would only be apparent when phonological knowledge of tonal categories is established with vocabulary expansion. / text

正常人的語誤中可以取得在心理語言上何為音韻單位的證據，近來有許多針對失語症的病人語誤的研究，發現失語症病人的語誤也可當做測試語言表徵的重要來源資料。 　　本研究針對失語症病人在聲調方面表現，做聲調錯誤頻率分佈、語境影響錯誤、以及錯誤方向種類之分析 然後將整失語症病人聲調表現的分析跟心理語言學語言產出機制作連結。 　　本研究結果顯示，漢語失語症病人聲調表現上第四生錯誤雖然最多，但第四聲在漢語裡原本出現的頻率就最高，所以第四聲不應被認為對失語症病人最為困難。而第三聲雖然在漢語裡出現頻率最低，卻佔了失語症病人聲調錯誤頻率第二高位，因此我們判斷第三聲對失語症病人來說是最為困難的聲調。第二，我們發現漢語失語症病人聲調錯誤受語境影響比正常人低。第三，漢語失語症病人聲調錯誤裡，語境影響語誤之方向性中，錯誤來源在語誤出現前類型較多。第四，我們發現Dell的連續模型(connectionist model)，可以解釋失語症病人在聲調錯誤類型偏好原因。 / Naturally occurring speech errors have been used as evidence for testing the psychological reality of phonological units. A number of researchers start to look at evidence from aphasic speech for testing linguistic representation so the phenomena of aphasia provide tests of the validity of certain aspects of linguistic theory. 　　In this thesis, tone errors made by aphasics were examined and related overall analyses to psycholinguistic models of speech production. In light of the analysis of tone frequency in Mandarin written texts and oral data and tone errors produced by aphasics, relating issues concerning the research questions we put forward were elucidated. First, it seems that tone 4 is the most problematic tone for the aphasics in terms of error rate since tone 4 occurs the most often in errors. However, tone 4 errors intrinsically will occur more that other three tones because of its highest incidence in Mandarin. Thus, tone 4 may not be taken as the most difficult tone for aphasics merely from its high error rate. Second, the results indicated that non-contextual errors constitute most of the aphasic errors. Third, it has been shown that perseveration is a cross-linguistic feature of aphasia. Fourth, Dell’s connectionist model is proved to be able to account for and make predictions about the error patterns found in both aphasic and normal speech.