Linguists tend to believe that total complexity of human languages is invariable. In order to test this hypothesis empirically, we need to calculate the complexity in different domains of language structure: phonology, morphology, syntax, etc. In this paper I provide some guidelines for documenting tonal systems and evaluating their complexity. I then apply my methodology to the Mande languages of West Africa and test a tonal equi-complexity hypothesis which says that languages with more tonal contrasts tend to have fewer tonal rules and vice versa. The data presented do not support such a concept of tonal equi-complexity in the domain of phonology, but there is a strong positive correlation between the number of tonal contrasts and the number of tonal morphemes. My explanation is that tonal contrasts and tonal morphemes tend to appear as a result of segmental loss, so the two phenomena are likely to co-occur.