本篇論文旨在從優選理論（Optimality Theory）之框架背景來探討國、台語夾碼（Mandarin-Taiwanese Code-Mixing）時的連讀變調現象。由於此現象之探討涉及國語及台語之變調，本篇論文亦從優選理論之觀點來探討國語三聲變調（Mandarin Tone Sandhi）以及台語的變調（Taiwanese Tone Sandhi）。 在國語三聲變調方面，由於傳統派生（derivational）模式在處理國語三聲變調時，仍遺留了些許的問題。例如，無法以一致的方式來處理non-PP和PP的變調現象。本文主要針對這個問題，重新以優選理論的角度提出解決之道。文中分別提出了一組韻律制約（Prosodic Constraint）及一組聲調制約（Tonal Constraint）；根據這兩組制約，傳統派生模式所遺留下的缺失則得以獲得妥善的處理。 在台語變調方面，台語變調在傳統派生模式的探討中亦遺留下些許問題。例如，無法以一致的方式來處理non-adjunct和adjunct以及non-clitic和clitic的變調問題。本文根據優選理論，重新分析台語變調。文中分別提出了一組韻律制約及一組聲調制約；根據此兩組制約，則可以成功的免除傳統派生分析模式下的缺失。 國、台語夾碼時的變調是未曾被探討過的問題。本文先就國、台語夾碼時的變調，提出一個派生模式的分析。不過，以派生模式來處理國、台語夾碼變調時，遺留下必須限定國語變調規則和台語變調規則的運作次序這項缺點。因此，本文根據優選理論，重新分析了這個變調現象，並提出了一組聲調制約；這組聲調制約，配合國語及台語的韻律制約，得以成功的處理傳統派生模式所遺留下的問題。 / The present thesis is mainly devoted to the research issue of tone sandhi in Mandarin-Taiwanese code-mixing using an Optimality Theoretic (OT) Approach. Since the study of this code-mixed tone sandhi necessarily involves the tone sandhi rules of Mandarin and of Taiwanese, the tone sandhi phenomena in Mandarin and in Taiwanese are carefully discussed under OT as well. Tone sandhi in Mandarin has been widely studied by many Chinese linguistics under the derivational approach. This thesis argues that the previous analyses to Mandarin tone sandhi are inadequate and that an Optimality Theoretic approach is needed to account for tone sandhi in Mandarin. The inadequacy of the previous analyses comes from the common failure of offering a uniform solution to account for non-PP and PP word stings in Mandarin. Based on the Optimality Theory, this thesis proposes two sets of constraints to account for Mandarin tone sandhi; one is the prosodic constraints set and the other is the tonal constraints set. This thesis proves that these two constraints can successfully get rid of the problem left by the derivational tradition. Tone sandhi in Taiwanese is also widely studied by many Chinese linguistics under the derivational approach. This thesis argues that the previous analyses to Taiwanese tone sandhi are inadequate and that an Optimality Theoretic approach is needed to account for tone sandhi in Taiwanese. The inadequacy of the previous analyses comes from the fact that Taiwanese word strings with non-adjunct and adjunct structures as well as with non-clitic and clitic structures fail to be accounted for using a uniform solution. Based on the Optimality Theory, this thesis proposes two sets of constraints to account for Taiwanese tone sandhi; one is the prosodic constraints set and the other is the tonal constraints set. This thesis proves that these two constraints can successfully get rid of the problem left by the derivational tradition. Tone sandhi in Mandarin-Taiwanese code-mixing is a phenomenon that is very interesting but has not been studied before. This thesis thus offers a derivational analysis to this tone sandhi phenomenon first, where it is found that (1) in Mandarin-Taiwanese code-mixing, a Mandarin base tone can trigger Taiwanese tone sandhi and a Taiwanese low tone can trigger Mandarin tone sandhi, (2) in Mandarin-Taiwanese code-mixing, the Mandarin tone sandhi rule is sensitive to all and only the Mandarin tone sandhi domain and the Taiwanese tone sandhi rule is sensitive to all and only the Taiwanese tone sandhi domain, and (3) in Mandarin-Taiwanese code-mixing, the Taiwanese tone sandhi rule must apply before the Mandarin tone sandhi rule to derive the tonal output. However, since the demand for an extrinsic rule order between the tone sandhi rules of Mandarin and Taiwanese is inadequate, the tone sandhi phenomenon is reanalyzed under the framework of the Optimality Theory. A set of tonal constraints is proposed. According to the tonal constraints set, the inadequacy left by the derivational approach is successfully avoided.