In this paper, I provide arguments in favor of subtonal features, with data from Laal, an endangered, three-tone isolate language of southern Chad. I show that a system using the two features [upper] and [raised] (Yip 1980; Pulleyblank 1986) offers a straightforward account of properties of the Mid tone that are otherwise impossible to account for using the three tonal primitives H, M, L. I show that Laal provides evidence for parallels between tonal and segmental behavior which justify a featural analysis of tone. In particular, Laal displays a tonal assimilation process involving natural classes of subtonal features, which works just like vowel harmony, and could be described as tone harmony. This goes against arguments frequently put forward against subtonal features (e.g. Hyman 2010; Clements, Michaud & Patin 2010). Additionally, the Laal case shows that the featural specification of the M tone in three-height tone systems is not necessarily ambiguous, making subtonal features potentially well-suited for the analysis of tone systems of all shapes.