It is claimed in the literature (e.g., Suh 1990, Sohn 1995, etc.) that an NPI freezes the scope relation between a quantifier (QP) and negation: a QP to the left of an NPI scopes over negation and a QP to the right of an NPI scopes under negation. (A QP scopes over or under negation when no NPI is around.) The scope freezing effect induced by NPIs has been accounted for by (often tacitly made) assumptions that (i) NPIs are in the scope of negation; (ii) the negation scope is NegP; and (iii) NPIs are licensed at the SPEC of NegP. This paper first provides sets of data showing that the so-called scope freezing effect is not a well established generalization. Once the scope freezing effect turns out to be an illusion, theories based on the generalization should be reconsidered. Thus, this paper discusses some related issues and leads to the following conclusions: (i) NegP does not necessarily indicate the negation scope: (ii) NPIs may be located outside of negation scope at least in some environments, where they have a Free Choice reading; and (iii) contra Sells (2001), the NPI distribution in Korean cannot be captured in terms of the `command` relation but in terms of the `government` relation based on Chung`s (1995) typology of negation licensing.