This talk looks at the conundrum on the dual responsibilities of the vocal folds in singing of songs which lyrics carry lexical tone. If tone is reproduced faithfully in the singing, then singing is speaking, and there would be no such thing as Chinese (or Dinka, Tibetan, Kukuya or Yoruba) songs. If not, then singing necessarily creates distortions and misinterpretations, yet many are moved by the lyrics heard sung even if for the first time. Any lyricist would probably respond to this query by saying they find the words must sound right, but no one has been able to say definitively what “sound right” means. Even if one concentrates only on Chinese songs, one would still need to contend with there being many different genres in singing (e.g. pop, jazz, ditties, operatic arias, and other narrative arts) across many different Chinese languages (e.g. Beijing Mandarin, Chaozhou, Cantonese, among many others). The issues are further complicated by the fact that there are tone sandhi patterns in many Chinese languages, so that given a particular lexical item, its tone varies depending on what position it is in the phrase or sentence. I am unaware of any Artificial Intelligence system that have tried to tackle this problem, nor are there any research publications that have spelled out fully the relationship between lexical tone and musical melody in Chinese. Available research have suggested that such a system of correspondence must exist, and in this talk we shall take a look at what little we know about Chinese songs. Perhaps this under-researched area might offer clues on this cross-media act of creativity that is song-writing.