This dissertation is intended as a contribution to the study and understanding of how language is used among the Akawaio peoples of Guyana, South America. It is a first attempt, limited to one Amerindian village, Waramadong, a community that has not been studied very much. The study begins by analysing four Akawaio speech genres, tareng 'ritual healing chant', mire aburobodi 'praising rhymes for children', pandong 'story', and zegareme'no 'personal narratives', at the levels of both content and grammar. The first entails making anthropological, ethnographical, sociolinguistic and general cultural commentaries on the content and social context within which speech is performed. This is to investigate how the Akawaio speech genres are categorised and classified and in what ways they are performed, interrelated and connected to the wider domain of speaking in Akawaio society. The second is committed to analysing the observed linguistic variation in the speech genres at the levels of both dialectal variation across speakers and stylistic variation across the genres. The analysis takes a multidisciplinary approach, transcending traditional linguistic boundaries, and invoking both social and linguistic theories, especially in the analysis of Akawaio spirituality as a crucial component to understanding the native Akawaio view of speech genres. Thus, this study offers a primary description of a wider, extended view of what is known about language and culture in Akawaio society. This dissertation also seeks to rectify a serious situation, to provide an emphatic counterexample to the common image in linguistics and anthropological literature, where Amerindian communities are treated as largely homogenous groups. The premise behind most anthropological and linguistic studies is that everyone acts and speaks alike within these societies. One aim of this study is to replace this homogenized image of the Amerindian with a richer, more complex and internally diverse picture, of the kind shown here for Waramadong. Appendix B presents a small, but representatively diverse, selection of transcribed, translated, and linguistically annotated texts, representing a small subset of the overall collection of texts recorded for this study.