Previous studies of fertility in Bangladesh have mostly used crosstabular and multivariate analyses revealing the direct effect of selected socioeconomic and background variables. The indirect effects are relatively unknown. Analysis of data from the 1975-1976 Bangladesh Fertility Survey seeks to identify variables that significantly affect the cumulative fertility of both urban and rural women and to construct causal models to clarify the complex set of relationships. Path analysis is used on 8 variables: fertility index (number of children ever born to ever married women aged 9-49, age of wife (respondent), age at marriage, religion, childhood background of wife and husband, and educational attainment of both wife and husband. Data obtained from a sample of 4875 rural and 1403 urban women are used. Causal ordering of the variables began with the 4 exogenous variables: age, religion, and childhood backgrounds of the wife and husband. Education of husband and wife are placed next in sequence, followed by age at marriage and the fertility index. Path coefficients are estimated by applying ordinary least squares regression to the equations. "Children ever born" is regressed on all other variables for urban and rural areas. The 2 sets of coefficients are not equivalent and urban and rural data are analyzed separately. 13 paths are significant in the urban model and 11 in the rural model. The coefficients are small suggesting that other variables not included in the study effect fertility. In the urban model, 5 variables were found to have a statistically significant effect on fertility: age, religion, education for both husband and wife, and age at marriage. The indirect effect of religion operating through a network of compound paths suggests that Muslim affiliation tends to depress wife's and husband's education and age at marriage. Holding other variables constant, Muslim women tend to have more children than non-Muslim women. Wife's age at marriage negatively effects fertility. Wife's education has a negative effect on cumulative fertility, husband's education has a positive effect. Childhood backgrounds have insignificant effects, probably due to few childhoods being spent in urban areas; however, wife's childhood in an urban residence has a noticable effect through education and age at marriage. 3 variables in the rural model have significant direct effects: age, religon, and age at marriage. Religion's indirect effect through wife's education and age at marriage indicates that for Muslim women, education is lower, and hence age at marriage is lower and ultimately number of children ever born is higher. Wife's age at marriage has a negative direct effect. Neither wife's nor husband's education have a significant direct effect, unlike the urban model, probalby due to a relative lack of education in rural areas. The effects of religion in both models operate indirectly through age at marriage. Religion may affect fertility through other variables as well, e.g., perception of sexuality, preference for sons, contraceptives, and ritual abstinence. The positive effect of spouse's education in urban areas may be due to the effects of permanent household income and may reflect a lowering of the perceived financial burden of children, increased fecundity of his wife through improved nutrition and early weaning of children, decrease duration and frequency of occupational absences, and prevention of his wife from working outside the home. The negative effect of wife's education may be due to later marriage, increased independence, ability to accumulate resources for old age, and a desire to provide medical care for her children. These characteristics are likely to induce limited fertility through contraceptive usage. Given the universality of marriage, the tendency to marry at younger ages, and ineffective contraceptive practices, raising the age of marriage through governmental laws can lower fertility.