Most of my seminar classes are small, especially if you only count the students who are actually enrolled. By that standard, some of them are very small.

Forget for now about the truth values of those claims (they are truer than I’d normally care to admit) and just focus on their semantic content. I think _very small_ in the last sentence is vague. But it’s a tricky kind of vagueness. In particular, it doesn’t seem susceptible to (normal kinds) of Sorites arguments. Conditionals of the following form don’t seem particularly compelling.

(S) If a seminar class with 3 enrolled students is very small, then a seminar class with 4 enrolled students is very small.

In degree-of-truth-speak, there will be a sizable gap between the truth values of any two sentences on this list (unless they both get truth value 0 or 1).

(0) A seminar class with 0 enrolled students is very small.
(1) A seminar class with 1 enrolled students is very small.
(2) A seminar class with 2 enrolled students is very small.
(3) A seminar class with 3 enrolled students is very small.
(4) A seminar class with 4 enrolled students is very small.
(5) A seminar class with 5 enrolled students is very small.

This seems to me just fatal for definitions of vagueness in terms of either Sorites-susceptibility or smooth variation between applicability and non-applicability. And it strongly suggests (to me at least!) that we’re well off returning to something like the definition in terms of borderline cases.

In case it isn’t completely obvious, I should note that this argument leans _very heavily_ on the ideas presented in Delia Graff’s “Gap Principles, Penumbral Consequence, and Infinitely Higher-Order Vagueness”:http://instruct1.cit.cornell.edu/research/graff/papers/gps.pdf, though of course Delia wouldn’t like the conclusions.