Error and the growth of experimental knowledge