The number of times, for each of the 40 pairs of doctors, that the doctor in each pair of doctors who was actually more satisfied, was identified as happier by each of the 48 assessors who considered that pair of doctors. The horizontal axis shows the number of assessors (0–48) who were correct for a particular pair of doctors, and the light and dark blue bars shows the distribution for each pair of doctors. Note that in this figure the bars correspond to pairs of doctors (and hence sum to 40), whereas in figure 2 they corresponded to assessors (and hence sum to 96). The yellow bars show the expected distribution under a binomial distribution (left-hand vertical axis). The dashed red line shows the expected cumulative binomial distribution, and the solid line the actual cumulative binomial distribution (right-hand vertical axis). The light blue bars show pairs of doctors within the 95% range of the binomial distribution, whereas the dark blue bars show pairs which are outside the 95% range, and hence are significant at the 5% level.