In Bangladesh, use of modern contraceptive methods has increased dramatically in the last decade. However, the level of use has not been the same between rural clusters. This dissertation investigates the plausible reasons for areal variations in use of modern contraception in rural Bangladesh. This study randomly selects sixteen rural clusters as sub-samples from the 1991 Contraceptive Prevalence Survey of Bangladesh. Information from female respondents of eligible age, as well as family heads and family planning workers are collected along with cluster-level community data, and family level information. The data are analyzed using simple bivariate tables as well as more sophisticated multilevel analysis using the software 'MLn'. Analysis of respondents using modern reversible methods finds that apart from the woman's age and number of living children, education of the family planning worker is the most influential predictor of use. Other significant predictors of ever use of modern reversible methods are the woman's education, religion, socio-economic status and her position within the family. Bari-level random effects were significant, indicating that a woman with `typical' characteristics but residing in the most favourable bari would have a higher probability of use than a woman with very favourable characteristics residing in a `typical' bari. Simultaneous confidence intervals of probabilities of use, for all clusters, found that most clusters were not significantly different in their ever use, although pockets of low use were observed. Analysis of acceptors of sterilization finds that apart from age and number of living sons, religion is the most influential predictor of use. Non-Muslims are more likely to be acceptors. Other significant predictors are the woman's education, position within the family, the participation of bari members with non-government organisations, and sanction of bari girls' education by bari head. Bari-level random effects showed in the same directions as those in the modern reversible methods model. Heterogeneity in the bari-level effect could not be explained by the recorded explanatory variables. After controlling for these explanatory variables, between-cluster variance was very small for users of sterilization. However, educated women had more between-cluster variance compared to non-educated women. Religiosity, attitude and beliefs of the respondents are vitally important in shaping attitudes towards contraceptive use. Replacement of some family planning workers with those having appropriate education and training are recommended. Family Planning programmes are strongly recommended to target `bari heads' and encourage them to look favourably upon modern contraception. Attempts should be made at removing ill-conceived religious barriers. Further recommendations are to encourage non-government organisations in areas which have low contraceptive prevalence, and encourage female education and emancipation generally.