How reliable are dictionaries?

by Mark Israel
     [This is a fast-access FAQ excerpt.]
A former senior lexicographer at a major dictionary publisher has
told me by e-mail:  "An editor seldom sits down and composes new
text for any lemma out of whole cloth.  Even for a supposedly
thoroughgoing revision, what usually happens is that you take the
text from your previous edition, apply whatever mechanical
formatting changes have been decreed, and then check two or three of
your competitors' books to see if they've said anything different
from what you have.  (Right -- it's no accident that all the major
dictionaries look so much alike!)  This practice can lead to some
pretty awful results; the nautical terminology in [a dictionary that
I worked on] was based on 19th-century square-rigger stuff
originally copied out of OED -- and evidently *they* didn't have any
sailors on the staff either!
   "In any case, the citation files don't normally even get looked
at unless something in the entry raises a red flag -- it's a new
word, or a member of some class of words marked for special scrutiny
(e.g., gender-specific terms or personal pronouns), or has been
tagged for special attention as the result of someone's query
somewhere along the line."
   For more on the frightening extent to which dictionaries copy
from one another, see "The Genealogy of Dictionaries", in Robert
Burchfield's Unlocking the English Language (Hill and Wang,
1992, ISBN 0-374-52339-8), pp. 147-165.
   Thus we see that a consensus of dictionaries does not necessarily
indicate a consensus of actual research.  Nor does disagreement
among dictionaries necessarily indicate actual scholarly
controversy:  it may simply be that the lexicographers were too
overworked and deadline-pressed to copy from one another more
thoroughly.  Samuel Johnson's observation, "Dictionaries are like
watches; the worst is better than none, and the best cannot be
expected to go quite true", remains highly pertinent today, despite
the improvements in both products since Johnson's day.