Ever: polarity sensitivity and polysemy

This paper offers a semantic account of the use and distribution of the
English adverbial quantifier *ever*. The analysis draws on a corpus of
over 1000 examples from the Wall Street Journal supplemented by
introspective data on grammaticality. It will be shown that the
various uses of *ever* form a network of distinct but tightly related
senses with significant parallels to the uses of the indefinite
determiner *any*. These parallels include an almost isomorphic
indefinite semantics for the polarity sensitive uses of both forms,
similar liberal licensing constraints for both as polarity items, and
a similar (though by no means identical) pattern of polysemy for the
non-polarity sensitive uses of both.
These parallels offer strong evidence for the lexical semantic basis of
polarity sensitivity and for the nature and importance of grammatical
polysemy in general.

*Ever* functions in the temporal domain much like *any* in the nominal
domain. In the paradigm of temporal adverbs, *ever* occupies the same
position as does *any* in the paradigm of nominal quantifiers: as *always*
is to *all*, *usually* to *most*, *often* to *many* and *rarely* to
*few*, so is *ever* to *any*. In particular it is argued that *ever*
quantifies over moments in the same way that *any* quantifies over
individuals, and that both exhibit a similar split between usages in
which they resemble existentials and usages in which they resemble
universals. Moreover, while the existential and universal usages of
both items represent arguably distinct senses of a single form, in both
the two usages share the basic function, as Kadmon and Landman (1993)
suggest for *any*, of "indicating reduced tolerance of exceptions." In addition
to these semantic parallels, the two forms exhibit strikingly similar
distributions as robust polarity items which are subject to unusually
liberal licensing constraints.

While debate continues on the viability of a unified analysis for *any*, it is
clear that a unified and monosemous analysis will not work for *ever*. The
present account distinguishes three basic constructional variants: a negative
polarity sense in which *ever* functions as an adverbial indefinite, a positive
polarity, universalizing sense in which *ever* indicates the constant
validity of a predication, and a non-quantificational sense in which
*ever* functions as an emphatic particle. These 3 senses are illustrated in order below.

1.      If you ever go to Milan, you'll love the fashions.
2.      As the night wore on, the stories of my ever faithful friend grew
        ever more lurid.
3.      Did we ever have a good time!

Brief analyses are provided for each of the senses and it is argued that
all three are distinct though tightly related synchronically. Partly
on the basis of corpus
distributions it is further suggested that each of these major senses may
include a number of special variants which should be accorded some independent
recognition within the grammar. Thus, for example, the two instances of *ever*
in (2) above represent slightly distinct variants of the universalizing sense.

The polysemy of *ever* and its strong parallels with *any have important
consequences for a theory of polarity in general and for the analysis of any in
particular. In general, it suggests that otherwise normal polarity items
may have closely related variants which are not polarity
sensitive. Indeed, this only reflects the fact that in grammar, as in
lexical semantics, polysemy is normal. In
particular, the analysis of *ever* suggests that even if one should accept
distinct free choice and polarity sensitive variants of *any*, the
insights of a unified analysis may be preserved by recognizing the two
senses as elements of a regular and highly motivated polysemy network.