This paper aims to answer the two questions raised by Kubozono (1999); namely, is it true (i) that the voicing phenomenon in Japanese compounds occur mainly in the left-branching structure and rarely in the right-branching structure, and (ii) that the right-branching structure will prevent the voicing phenomenon from occurring in compounds? We reanalyze the relevant linguistic facts and propose alternative solutions different viewpoints from Kubozono's. Our conclusions are: (i) the parsing of compounds is easier in the right-branching structure, in which parsing starts with the left-most constituent and proceeds forward according to the linear order of the constituents, than in the left-branching structure, in which parsing starts with the right-most constituent and proceeds backward contrary to the linear order of the constituents, and (ii) the occurrence or nonoccurrence of the voicing phenomenon is better explained not in terms of right-branching versus left-branching structures, but in terms of the application of the voicing rule in a ”bottom-up” manner.