口譯研究領域裡對於翻譯單位的探討通常是以EVS來做表示，這是指訊息理解與訊息產出的單位、以及兩者的問liE落差。也可以說，EVS研究關切的是每個訊息單位的大小、內容、結構、切 分原則、切分效能，以及從切分至1J產出的時間落差，如何控制在訊息不流失的前提之下去進行。
本文首先將以口語理解研究的多項指標，如「靜默停頓」(pausing)、「可解度」 (comprehensibility)、「訊息處理單元」(processing units)等，做為觀察原語(source language)最小翻譯單位的主要線索。 其次，再以席演講、視譯、同步口譯等語言事實，對照前述自然語言的實證數據以及口譯與手語的EVS研究成果，例如靜默停頓的頻率、時間、訊息處理單元的句法結構與語意特徵、位置與談話形式的相關性等，進一步描述口譯理解與產出的單位訊息結構、以及單位辨識標誌等細部特微。通過以上觀察與被驗結果，針對同步口譯的訊息轉碼過程，提出條理化的訊息單位判定基準與操作原則。

Discussion about the unit oftranslation in interpretation research is normally centered on ear-voice span or EVS. EVS refers to the units of information comprehension and production and the span that exists between them. Itcan be said that EVS research is also concerned with the size of each information unit， the content and structure of each information unit， principles of segmentation， effects of the segmentation， the time lag that exists between segmentation and production and how this is all performed under the premise ofthe interpreter not losing any information. This paper will first of all utilize the multi index signs from oral comprehension studies as the main clues for inspecting the smallest translation units of the source language. The index signs that will be used include those such as pausing， comprehensibility and processing units. Secondly， impromptu speeches， sight translation and simultaneous interpretation will be used as linguistic proof that will be compared with the aforementioned natural language data and EVS research results from interpretation and sign language. Research results used include those such as the frequency of pausing， time， the syntactic structure and semantic characteristics of information processing units， as well as the relationship between position and discourse types. These research results will be used to describe in further detail the information structures of the comprehension and production units of interpretation， as well as information structure markers. The above-mentioned observations and test results will then be used to propose general judgment models for information units and operation principles that are focused on the transcoding process of simultaneous interpretation.