Objective To examine the adjusted and unadjusted effects of infant feeding practices on infant mortality in Southern Africa. Methods A merged dataset from the most recent Demographic and Health Surveys for Lesotho, Swaziland, Zambia and Zimbabwe was analysed using the Cox Proportional Hazard Model. A total number of 13,218 infants born in 5 years preceding all the surveys with information on infant feeding practices constituted the study population. Infant mortality was the outcome variable and infant feeding practices categorised into; no breastfeeding, partial breastfeeding and exclusive breastfeeding were the main explanatory variables. Maternal demographic and socio-economic characteristics and infants' bio-demographic characteristics were also studied. Results Although, exclusive breastfeeding was quite low (12 %), exclusively breastfed infants exhibited a 97 % lower risk of dying during infancy compared to infants not breastfed in the region. Variations existed by country in the levels and patterns of both infant mortality and infant feeding practices. Mother's country, highest level of education and marital status; child's sex, birth weight and preceding birth interval were the significant predictors of infant mortality in Southern Africa. Conclusions Any form of breastfeeding whether exclusive or partial breastfeeding greatly reduces the risk of infant mortality with the greatest mortality reduction effect observed among exclusively breastfed infants in Southern Africa. To reduce the upsurge of infant mortality, there is the need to step up the effectiveness of child nutrition programmes that promote breastfeeding and put emphasis on exclusive breastfeeding of infants in the region.