The melodic tone system of Kikamba, as described by Roberts-Kohno (2000; 2014), stands out as particularly complex within the context of recent crosslinguistic work on melodic tone in Bantu (Odden & Bickmore 2014; Bickmore 2015). It is unique, for example, in possessing a melody that assigns four distinct tones to three stem-internal positions simultaneously. The apparent existence of such com- plex melodies raises doubts as to whether there are any substantive restrictions on the possible form of a tonal melody. We argue, however, that these doubts are premature. We propose a new analysis of Kikamba in which (a) melodies refer to no more than two target positions at a time and (b) melodies target only two possible stem-internal positions, each of which occurs commonly within Bantu melodic tone systems. This simplification is achieved by (a) rejecting the existence of a melodic L tone assigned to the penult, and attributing its putative effects to in- teractions among other, more basic tones, and (b) distinguishing between melodic tones assigned early in the phonological derivation and other suffixal tones added later. In general, we argue that since core properties of melodic tone are often obscured in surface forms due to interactions with language-particular rules, the crosslinguistic comparison of melodic tone should proceed on the basis of a (more) underlying level in which these rules are controlled for. Once this is done, the ex- ceptional properties of Kikamba melodic tone largely disappear.