How does the c4.5 algorithm of decision trees work with numbers?

Question

data mining

With verbal meanings, it's more/less clear, but how does it work with numbers? how does it determine that a given attribute can be divided into >50 and <50? thought average, but no