Not the GP, but this is my take on their comment. Children learn counting before they learn an algorithm for addition. The "theory" to accept here is that counting a set of size x, then continuing to count, starting from x, a set of size y, yields the same result as the addition algorithm for x+y that they learn.
I think he means loosely that the theory here, is that they add up to three. Kids will grab objects and group and ungroup them while counting and so on, before outright accepting math is real.