In simultaneous interpretation, the interpreter’s version may depart from the original version in three general ways: the interpreter (henceforth abbreviated T for ‘translator’, since it cannot conveniently be abbreviated ‘I’) may omit some material uttered by the speaker (abbreviated S), he may add some material to the text, or he may substitute material, resulting in his saying not quite the same thing as the S. If the substitution is at considerable variance with the original version, we may speak of an ‘error’ of translation. This paper presents a very general classification or coding scheme developed in relation to these three types of events, along with some relevant data. The work was carried out in the context of an investigation of simultaneous interpretation in which Ts of varying proficiency levels (professional TS, ‘student’ Ts and ‘amateurs’) were required to simultaneously interpret several passages on tape, the languages involved being English and French.