Why won't Prism or InStat compute Bartlett's test for equal homogeneity when N=4?

In the Prism 4 statisticsc guide, we mistakenly say that Bartlett's test requires four or more values. In fact it requires at least five values in each group.