This study has two aims: (a) to describe verbal extensions in Sumbwa and theirvalence, and (b) to contribute to Bantu comparative data that may be used in researchon Bantu morphosyntactic parametric variation as proposed in Guérois et al. (2017).The description covers all extensions that could be identified in the data at hand(e.g. Capus 1898, Kahigi 2008a,b), focusing on the forms and their various meanings, their valence possibilities, their productivity and co-occurrence constraints.The study reveals that most of the Proto-Bantu verb extensions reconstructed byMeeussen (1967) and Guthrie (1971) are still active in the language. Some extensionsare found to be highly productive (applicative, passive, causative (also instrumental), associative, stative, and frequentative), others moderately or semi-productive(persistive, reversive, impositive and denominative) and quite a number may beregarded as being unproductive (associative/reciprocal -an-, reiterative, static, contactive and other minor extensions). The second aim of the study is to considerhow Sumbwa compares to other Bantu languages by drawing on the parameterson verbal derivation identified in Guérois et al. (2017). Some of the findings agreewith what is found in the majority of Eastern Bantu languages, e.g. the verb deriva-tional strategies follow closely those mapped out by the PB reconstructions, exceptfor a few innovations among the minor extensions (e.g. -agan-, -agil-). However,Sumbwa does not have ba-passives found in Bemba, nor does it have the CARPorder as postulated by Hyman (2002) for Bantu. Some of the notable characteristicsof Sumbwa verb extensions include the fact that (a) the causative and instrumentalshare extensions, (b) the associative markers include the post-verbal -an- and thepre-verbal -i-, which is homophonous with the reflexive; (c) the applicative conveys benefactive, directive, location, and reason meanings; (d) there is no systematicfixed order of extensions, except that in all co-occurrences, the passive comes last.