Demuth and Harford (1999) contend that in Bantu relatives, the verb raises from I-C if the relative morpheme is a bound morpheme while the subject remains in spec-IP resulting in subject –verb inversion. Ikalanga, a Bantu language spoken in Botswana has no subject verb inversion in relatives although the relative morpheme appears to be a bound morpheme. This observation challenges the conclusion reached in Demuth and Harford (1999). This raises the question, What then is the structure of the relative clause in languages like Ikalanga and Luganda? This paper argues that Ikalanga relative clauses differ from other Bantu relative clauses in that the projection that houses the relative feature (RelP) projects below TP while in Bantu languages where subject verb inversion is observed such as Shona it projects higher than TP. Thus, the variation in the structures of Bantu relative clauses can be accounted for if we understand that there is a parametric variation in the position in which RelP projects; lower than TP or higher than TP.