In this paper we examine the "some + n" construction, where the apparent quantifier "some" is used with a numerical expression, as in "some 27 students were arrested." Contrary to previous claims in the literature, we show that while many speakers prefer an approximative interpretation for some + n, it is untenable to analyze “some" as an approximator akin to "about" or "roughly." We survey some constraints on the distribution of some + n, and propose a semantic analysis based on recent theories of indefinite determiners (e.g. Alonso-Ovalle and Menendez-Benito 2010) which is able to explain these constraints. On our account, "some" introduces a manipulation of the domain of quantification, either restricting it to contextually relevant pluralities or widening it to include pluralities whose cardinality is approximately that of the associated numeral. In contexts where approximation is disfavored and there is no obvious restriction on the domain, the meaning contribution of "some" is essentially vacuous. We claim that this vacuousness, in conjunction with Horn's (1984) division of pragmatic labor, explains why some + n is most felicitous in emphatic contexts, e.g., "some 17 Republicans ran in the primary!" (where 17 is higher than expected) as opposed to "?Some 5 Democrats ran in the primary" (where 4 is average).