more on most

earlier i raised the following challenge to the generalization that only universals admit of exceptions:

most of the students helped out, except the tenth graders

the challenge is that exceptive phrases are pretty well established as applying to universals, and most is not a universal.

i suggested a solution: that there should be understood a presupposed partition on the set of students, based on grade. so that the sentence can be restated as follows:

given a partition of the students into equivalence classes based on grades, every equivalence class except the one consisting of the tenth graders is such that most of its members helped out.

this solves the problem of the universal by introducing one, albeit an invisible one, to license the exceptive. it also gets the truth conditions of the sentence right. but it is a challenging approach, because in the absence of independent justification for an invisible presupposed partition and universal quantifier, it is a rather ad hoc solution. in this post i try to support the analysis by bringing one source of independent evidence for an invisible universal, from negative polarity item licensing.

i noted in my previous post that most differs from other quantifiers, such as some.

*some of the students helped out, except the tenth graders

recently i remembered the fact, discussed in the literature on polarity items*, that some and most have different properties as far as licensing any. most licenses it sort of, while some doesn't.

most (of the) people with any self-respect abhor imperialism
*some (of the) people with any self-respect abhor imperialism

i'm generally a believer in semantic and pragmatic licensing of polarity items (as opposed to, say, syntactic licensing), though not a firm believer. i don't see pragmatics as being a licenser in this case, so let's assume it's semantics. usually, negative polarity items (which we'll assume any is here. why? cause fuck free choice any, that's why) are assumed to be licensed in downward-entailing contexts. the first argument of most is not generally seen as a downward-entailing context, because of facts like the following.

most high school students are evil
-/->
most tenth-graders are evil

but what if these sentences were interpreted as i've suggested, as universal statements ranging over equivalence classes of a presupposed partition? and what if the set expressed in the consequent was one of the equivalence classes presupposed in the antecedent? in that case, there would be an entailment.

given a partition of the set of high schoolers by grade, for every equivalence class E, for most x, x a high school student in E, x is evil.

-->

given a partition of the set of high schoolers by grade, for every equivalence class E, for most x, x a tenth-grader in E, x is evil.

if this universal interpretation is available, it helps us solve the puzzle of why most is a sort-of licenser of any. It's because most is sort of downward-entailing in its first argument, because it is sort of a universal quantifier.

now, the same would be true of some. If it was susceptible to a universal interpretation, it would be as downward-entailing as most is in the example just given. We would then expect it to license both an exceptive and any in its first argument. but it does neither, which we can attribute to its inability to accommodate to a universal interpretation.

the larger mystery, of why most can do what other quantifiers can't, still exists. but at least we've suggested a connection between two properties: licensing of any and support for exceptives.

this approach is surely wrong. and you know what? that doesn't bother me, because i don't have to publish in linguistics anymore.

-=-=-=-=-=-=-=-=--=-
* i would guess this is in myriam uribe-echebarria's dissertation, for instance.