A man in daily muddy contact with field experiments could not be expected to have much faith in any direct assumption of independently distributed normal errors.