This paper extends the literature on grader variation and the role of moderation and socialisation processes in reducing this variation. It offers a fresh categorisation of academics' assessment beliefs and expectations, and uses this categorisation to analyse the interaction between implicit and explicit expectations in relation to grader variation and socialisation processes. A combination of quantitative and qualitative data from a large class grading moderation process was used to identify the existence and character of residual implicit grader expectations, the impact of these implicit expectations in causing grader variation, and the effectiveness of socialisation in reducing the extent of variation. An effective strategy for designing and focusing socialisation processes is also outlined. (Contains 8 tables.)