近幾年來有愈來愈多的學者致力於三人行醫病對話的言談互動(例如Aronsson and Rundstrom 1988, Baker 1996, Rosenfeld 1996)。其共同的焦點為探討醫師是否因有第三者(例如陪同的家屬)在場而影響他對病人應有的注意力。藉由分析對話中的代名詞、稱呼語等言談成份，研究人員釐清互動中的參與者結構，例如當醫師對病人發問則參與者結構為”醫師病人之二人對話)，對陪同家屬發問則為“醫師家屬之二人對話”，研究人員計算出此兩種參與者結構之比例，進而得出在場的家屬是否影響了醫師的問診。在本研究中，我們試圖運用西方學者所建立的研究架構檢視我們於台灣南部某教學醫院所蒐集的三人行問診對話(即醫師、老年病人、病人之隨行子女)，發現文獻記載的研究架構並無法提供一套有效釐清參與者結構的方法。本文之目的即在於檢視三組研究架構(Aronsson and Rundstrom 1988, Baker 1996, Rosenfeld 1996)運用於中文言談對話之困難，尤其是當家屬與病人皆積極發言參與時，對話的參與者結構也就更難釐清。基於方法學之困難，我們認為以分析參與者結構為出發點探討隨行家屬對醫病互動影響，並非值得嘗試之角度，最後我們並提出其他可行的研究方向，提供學界參考。

This paper explicates the intimate relation between semantics and pragmatics as well as the significant role that pragmatic inferencing plays in semantic change and grammaticalization. It is pointed out, mainly based on Levinson's (1995, 2000) Neo-Gricean Theory of Implicature and Traugott's (1999) Invited Inferencing Theory of Semantic Changes, that the complicated polysemy often observed in synchronic data is best captured if new meaning is viewed as first arising from invited inferences (or specific conversational implicatures), later developed into generalized invited inferences (or generalized conversational implicature) through frequent use, and finally conventionalized as coded (or semantic) meaning. This path of semantic change is proposed to characterize the multiple functions of the polysemous adverbs koh(閣) and pheng/peng(並) in Taiwanese Southern Min, both cases exhibiting an obvious escalation of the speaker's subjective belief and attitude.