The literature on variation in endangered languages has largely focused on the structural and stylistic simplification that occurs when languages contract or on cases where variation is introduced through interference in the form of a non-native variant that exists alongside a native form (Dorian 1981, Mougeon and Beniak 1981; Aikhenvald 2002; Campbell and Muntzel 1989; among others). Some studies have considered the effect of language shift or moribundity on variation that existed in the healthy language. King’s (1989) study of Newfoundland French, for instance, concludes that variation can be maintained in dying languages, but the variation no longer carries the social meaning that it did in a healthier speech community. This paper, on variable /r/-deletion in Garifuna, presents an apparent change in progress in an endangered language and explores the effect of incipient language shift on previously existing variation. Garifuna is an Arawak language spoken in disparate communities throughout Central America. It is a moribund language in most of the Garifuna communities in Belize, where speakers are variably trilingual in Garifuna, English, and Belizean Creole (BC). In the village of Hopkins intergenerational transmission of Garifuna still occurs but locals and non-locals alike believe that Hopkins will follow the path of nearby Garifuna communities in shifting to English and BC, abandoning the use of Garifuna entirely. The /r/-deletion data in this paper come from speech samples of 26 Hopkins speakers (16F, 10M), ranging in age from 6 to 65. All speakers were asked to tell a story based on a children’s picture book (Mayer 1967), and the recordings were coded for deletion of the variable /r/. This variable (previously reported on by Hagiwara, ms) was chosen as an example of a sound change that is not likely to be a result of language contact. Post-vocalic /r/-vocalization exists in BC, but /r/ in Garifuna only occurs intervocalically and thus the environment is different from that of /r/-vocalization in BC, where it occurs in preconsonantal or word-final positions. An apparent time (Labov 1963) analysis of the data suggests that deletion of /r/ is a female-led change in progress: each successive age group shows increased deletion of /r/, and women lead in rate of deletion within each age group. This report is the first to describe this variation as a change in progress, and it concludes that the progression of a change akin to that in a healthy language may occur even while the language community is undergoing shift.