The phenomenon which this thesis explores is known as washback - is generally defined as the influence of testing on teaching and learning (Bailey, 1996:259). The test explored in my study is the English oral test component of the Senior Secondary School Entrance Examination (the SEE Oral Test) in Guangzhou, China. To date most washback research has focused on the kinds of changes that have been brought about by the introduction of a new examination (e.g. teaching behaviour, learner behaviour and materials) rather than how the change has taken place. It is important to understand the mechanisms by which washback is achieved, to help to improve the likelihood of positive washback. My study seeks to explore and explain why washback has taken the form it has amongst local Chinese teachers in China, using a framework adapted from the discipline of social psychology - the Theory of Planned Behaviour (Ajzen, 1985,1991,2006). I first investigate the types of washback that were intended by the SEE Oral Test designers through interviewers with four of the design team. I then compare the beliefs and behaviour of four teachers, using diary data and lesson observations, and I assess the extent to which the intended washback has been realised. Lastly, I attempt to explore the reasons for the different forms of wash back achieved, mainly through interviews with the four teachers. The findings indicate that the washback from the SEE Oral Test manifested itself in complicated and inconsistent ways. One of the four types of wash back intended by the test designers appears for all the participants, two types do not exist at all and one type appears for only two participants. The findings also show that teacher beliefs play an important role in shaping washback effects. I claim that a test does not influence teacher's teaching behaviour in a direct way but operates through exerting influence on teachers' beliefs and attitudes, their perception of social pressures and their perceptions of their own ability to perform the desired teaching behaviour. I also claim that the forms in which washback appears might be influenced by the appropriateness of the washback intended by the test designers. I propose a process-oriented model of wash back on teaching, which contributes to an understanding of the mechanisms by which washback works. I also propose a working scheme including a set of successive intervention procedures to promote positive washback of tests.