The present project investigates apparently paradoxical phenomena concerning quantificational expressions in Japanese sentence processing as follows. Given structurally ambiguous sentences, (i) when both the nominative and the accusative NPs are bare nouns (nouns without quantifier, interpreted existentially), the garden path (GP) effect is observed, while (ii) when either one of the NPs has a quantifier such as _subeteno_ 'all/every', the GP effect reduces, but (iii) when both are quantified, the GP effect reemerges. The question is: if the existence of one quantifier reduces the GP effect, why not two? We argue that in (ii) the scope interaction of the quantifier and the bare noun is computed, which retards the syntactic processing, resulting in the low GP effect. On the other hand, in (iii), because of its complexity, the scope interaction is not computed, and those quantifiers are treated as non- quantificational, group objects.