In literature, there are mainly two rival approaches for the FNQ-constructions in Korean andJapanese, the ‘Locality’ and the ‘Cyclic Linearization.’ And the former approach again has two distinct versions,a traditional ‘strong’ view (Haig, 1980; Kuroda, 1980; Miyagawa, 1989, 2001; Saito, 1985, etc.) and a ‘weaker’view proposed by Miyagawa and Arigawa (2007) and Miyagawa (2010), This paper examines all these approachesin a wide variety of scrambling contexts, and shows that the revised Locality (i.e., a weak Locality) isthe best; it explains many recalcitrant examples that pose significant problems under the other approaches. Inthis paper, I maintain that the revised locality still has some shortcomings. Crucially, it is a stipulation; it doesnot give an account as to why the subject-object asymmetry occurs in the distribution patterns of FNQs andwhy only subjects, but not objects, leave a licensing trace optionally. This paper claims that the real heart of theinvestigation to this area of research is; how we could explain the bewildering variety of trace patterns in a simpleand principled way.