For nearly three decades, autosegmental phonology has provided an elegant and natural means of describing phenomena typical of tone. The problem with this simple approach is that it may be too simple. The innocuous character of this framework has made it easy to find parallels between tone and segmental features like voicing, nasality, and so on—though without revealing any deep similarity between them. As an alternative, this paper examines the Optimal Domains Theory of Cassimjee and Kisseberth. For a variety of Bantu cases, these authors show that significant generalizations follow directly from Optimal Domains Theory yet elude a standard autosegmental approach. By expanding the set of languages to include West African ones, this paper finds new support for the tonal domain, one of the key elements of Optimal Domains Theory, while suggesting an alternative for the tenet that a domain can refer at the same time to underlying and surface tones.