In response to requests I have often got as to how one approaches a tone language, I present a personal view of the three stages involved, starting from scratch and arriving at an analysis: Stage I: Determining the tonal contrasts and their approximate phonetic allotones. Stage II: Discovering any tonal alternations ("morphotonemics"). Stage III: establishing the tonal analysis itself. While most emphasis in the literature concerns this last stage, I show how the analysis crucially depends on the first two. A detailed illustration is presented from Oku, a Grassfields Bantu language spoken in Cameroon on which I personally worked in the field. The paper concludes with discussion of issues arising in other tone languages, illustrated by Corejuage (Tukanoan, Colombia), Peñoles Mixtec (Otomanguean, Mexico), Villa Alta Yatzachi Zapotec (Otomanguean, Mexico), Luganda (Bantu, Uganda), Hakha Lai (Tibeto-Burman, Myanmar and Northeast India), and Haya (Bantu, Tanzania).