Accurate marking is crucial to the reliability and validity of public examinations, in England and internationally. Factors contributing to accuracy have been conceptualised as affecting either marking task demands or markers' personal expertise. The aim of this empirical study was to develop this conceptualisation through investigating the relative effects on accuracy of some key contributory factors. This was done in the context of an international biology examination designed for 16-year-olds. Forty-two markers participated, comprising five groups: (1) experienced examiners, (2) biology teachers, (3) graduates in biology, (4) graduates in other subjects, and (5) non-graduates. The design enabled the relative effects on accuracy of the following factors to be elicited: marking experience, teaching experience, highest education in a relevant subject, highest education in any subject and gender. The significance of markers' accuracy on standardisation samples was also investigated. Twenty-three examination questions were explored, varying in format, number of marks, difficulty for candidates and cognitive marking strategy complexity. All markers marked identical response samples for each question. The data were modelled with logistic regression and AN(C)OVA, generating a refined model of marking accuracy. This may prove useful for evidence-based decisions surrounding marker selection and training. (Contains 7 tables, 7 figures and 4 notes.)