As of 2017, only four of the world’s languages are known to present opacity effects in coronal harmony: Sanskrit, Slovenian, Imdlawn Tashlhiyt and the Bantu language Kinyarwanda which is the focus of this paper (Hansson, to appear). In Kinyarwanda, coronal harmony is triggered by the retroflex fricatives [ʂ] [ʐ] and targets the alveolar fricatives [s] and [z]. The process operates regressively and is blocked by coronal stops [t] [d], palatal consonants [ɲ] [j] and the alveolar affricate [ts]. The harmony is obligatory in adjacent syllables but optional across non-adjacent syllables. The rarity and complexity of this phenomenon presents challenges to its theoretical representation. Past literature argues primarily for two different forms of analysis: (1) feature spreading (Mpiranya and Walker, 2005, Walker and Mpiranya, 2006) and (2) constraint-based Analysis by Correspondence (Walker and Rose 2004, Hansson 2010). After an analysis of both approaches, this paper suggests that for the facts of Kinyarwanda coronal harmony specifically, feature spreading is more economical and accurate in its predictions than feature agreement. In addition, past feature spreading approaches (Mpiranya and Walker, 2006) use Optimality Theory constraints to make predictions, avoiding a “true” autosegmental approach. An autosegmental approach to Kinyarwanda coronal harmony is not known. Therefore, in an adaptation of Sagey (1986), this paper proposes an autosegmental spreading approach that accounts for the obligatory harmony (a), optional harmony (b), and opaque segments (c) by virtue of spreading the COR node to the left to the Place node. The spread is only successful if the target Place node is dominated by [+cont] and dominates COR, resulting in its inability to spread across another COR node. Therefore, this paper suggests that opacity is present in Kinyarwanda coronal harmony due to the opaque segment’s intervening COR node.