Research on interpreting has traditionally focused on 'simultaneous conference interpreting', and only more recently on 'community interpreting' (also referred to as 'public service interpreting' or 'dialogue interpreting' elsewhere) (e.g. Edmondson 1986, Jones 1998, Diriker 2004, Wadensjö 1998, Mason 1999). The government press conference, a special genre of interpreter-mediated events, has so far received little scholarly attention from the discipline of translation and interpreting studies - with the exception of one recently completed doctoral thesis which focuses on the impact of audience reception on interpreters' strategies in press conferences (Liu 2010) and a few publications on related areas such as interpreters' performance in question and answer sessions in conference interpreting (Chang and Wu 2009) and interpreting in political interviews (Baker 1997; Wadensjö 2000, 2009). As is the case with interpreting in political interviews, government press conference interpreting is likely to have far reaching consequences for the lives of very large numbers of people across the globe, and to play a major role in constructing cultural images and aiding or obstructing world peace (Baker 1997: 124). Drawing on the work of Erving Goffman (1972, 1981a) as a source of theoretical insight, this study attempts to explore the way in which interpreters, with their background as civil servants, position themselves in government press conferences in China through choices that effect changes in footing and participation framework and reveal their involvement in face-work, based on video-transcribed data collected from a series of 6 government press conferences held in 2003 in response to the outbreak of the SARS (Severe Acute Respiratory Syndrome) epidemic. This study reveals that interpreters in Chinese government press conferences position themselves mainly through their negotiation of institutional alignment and protection of 'face', primarily for their institutional superiors. This finding supports the argument that interpreters do not function as a language "voicebox" (Davidson 2002: 1275), or "an impartial, self-effacing conduit" (Cronin 2006: 90); they are rather proactively engaged in the construction of interactional meaning, and may even function as "institutional insiders and ally themselves as such" in certain circumstances (Davidson 2010: 152). In the present study, the analysis reveals that interpreters adopt a distinctive position as 'semantically neutral' but 'emotionally/pragmatically partial'.