It’s well-known that there is a shift in how children reason about false beliefs around four years of age. By developing and investigating a processing model of belief-desire reasoning, we discovered two further shifts.  A second shift occurs around five to six years and a third occurs much later, perhaps around puberty. I will describe two versions of a model of how the contents of attributed beliefs are selected, the different biases they predict, and results from tasks designed to test their predictions. The famous first shift at four is not all that special and in particular does not signal the emergence of the concept of belief -- indeed, this has emerged long before.