I think the point is that (integral) numbers stored in hardware naturally wrap, and that this behaviour is not restricted to two's complement. For that matter it's not even restricted to binary - the mechanical adding machines based on odometer-like gear wheels, operating in decimal, would wrap around much the same way, from the largest value back to the smallest... and these were around for several centuries before computers: http://en.wikipedia.org/wiki/Pascal%27s_calculator