English verbs selecting Bare forms

[Laurence Horn]

At 5:32 PM -0400 4/5/01, Robert Levine wrote:
>I've encountered `durst' as well, exclusively in pre-Shakespearean
>literature. If I recall correctly, there were a few tokens that didn't
>occur before negatives but did select the bare stem form; if so, this
>suggests that the odd confinement of the `dare' that selects the bare
>form to negative contexts when no inversion is involved is fairly
>recent.
>
>Interestingly, there's some kind of polarity sensitivity that these
>quasiauxiliary forms show; they don't need to actually precede a
>negative adverb:
>
>(1)a. I don't imagine that anyone need worry too much about Leslie's
>	complaints.
>    b. I can't imagine that Robin would dare say something like that
>	without a lawyer being present.
>
>I don't know of any other auxiliary(-like) elements that display this
>property. All in all, the need/dare property set strikes me as somewhat
>mysterious.
>
Well, it's somewhat mysterious how they got that way (although as
noted below the English modals are not unique in this respect), but
the standard description (going way back to Haj Ross, Fritz Newmeyer,
and me in the mists of time) is that "need" and "dare" AS MODALS are
both negative polarity items.  I have some discussion in my
dissertation, but the upshot is that they're NPIs of intermediate
strictness, with some wiggle room for dialect differences.   It's
quite easy to get such items in "neg-raising" contexts like (1a)
above, and (1b) is fine with the negative entailment; for me,
occurring in questions and with inherently negative predicates (like
"doubt" and "deny") is possible for "need", somewhat less so for
"dare".  Neither occurs in tags for me, but that's more a fact about
their restricted distribution than their status as NPIs.  The main
effect of the NPI-hood of "need" that I was discussing back then
(1972) was that unlike freer modal-neg sequences, a negation
associated with "need" can only be interpreted as taking scope over
it (cf.  "He need not go" vs. "He needs to not go"), with the result
that "needn't" is a counterexample to an otherwise robust
generalization that modal-neg sequences lexicalize only when the
result is what Aristotelians would assign to the E-vertex of the
square of opposition (necessary/obligatory not, not
possible/permitted), not when it's an O-vertex value (not
necessary/obligatory, possible/permitted not).  Thus "I couldn't go"
has unambiguously wide scope negation while "I could not go" has
either scope, and even the "orthographic lexicalization" of "cannot"
is only possible with neg scoping over "can".  But "needn't" is
O-vertex.  Given its NPI status, though, it couldn't be otherwise.

By the way, Ton van der Wouden has done some interesting
cross-linguistic work on negative polarity modals (e.g. Dutch
hoeven), and Johan van der Auwera has written on this as well.

larry