Purpose: Nonword repetition (NWR) has been proposed as a culturally and linguistically fair measure of children's language abilities that is useful for the identification of specific language impairment (SLI). However, Moyle, Heilmann, and Finneran (2014) suggested that the density of a child's nonmainstream forms also influences NWR in ways that could complicate its use. Using speakers of either African American English (AAE) or Southern White English (SWE), we asked if NWR performance differed in children with SLI and same dialect-speaking typically developing (TD) children and if nonmainstream form density impacted their scores. Method: The participants were 106 kindergartners (AAE: SLI n = 35; TD n = 35; SWE: SLI n = 18; TD n = 18; groups matched for age and IQ) who performed the NWR task of Dollaghan and Campbell (1998). Nonmainstream form density measures were gathered from listener judgments of conversational samples. Results: NWR performance differed between those with and without SLI, but the difference was smaller in AAE than in SWE, especially at the longest syllable length. Nonmainstream form density was found to further explain NWR performance beyond the children's SLI status for AAE speakers; density and SLI status were confounded for the SWE speakers, making it harder to disentangle the effects of each in that dialect. Conclusions: Results indicate the NWR may differ in diagnostic utility between speakers of different dialects. Results also support Moyle et al.'s (2014) finding that density affects NWR. Thus, NWR is more sensitive to dialectal differences than originally assumed.