Figure 2

Mean score across all 14 examiner assessments by candidate ethnicity (solid circles, white candidates; open circles, non-white candidates), in relation to the total number of assessments of a candidate that were made by non-white examiners. Error bars indicate ± one standard error of the mean, and are wider in the right-hand points because there were fewer candidates. The solid and dashed lines indicate fitted regression lines for white and non-white candidates, weighted for different numbers of candidates.