On the concept of experimental error