Error and the Growth of Experimental Knowledge