Making morphemes within a word distinct while keeping the word's coherence might be considered opposites. In this paper, I deal with four types of compounds in ancient and modern Japanese, and argue that the two tasks function effectively and are entirely compatible. The compounds dealt with in this paper are (i) /Q/-inserting compounds in modern Japanese: ma-Q-siro, huto-Q-para, kore-Q-kiri, and (ii) Rendaku (or sequential voicing): yama-bato, tani-gawa, sato-bito, (iii) compounds attested to be pronounced as two independent accent phrases in ancient Japanese, and (iv) vowel elision in consecutive vowels in compounds in ancient Japanese.