Identification of factors that predict a woman's infant feeding choice is important for breastfeeding promotion programmes. We analysed a subsample of children under 2 years of age from the most recent Sri Lanka Demographic and Health Survey (SLDHS) to assess breastfeeding practices and factors associated with suboptimal practices. SLDHS 2006–2007 used a stratified two-stage cluster sample of ever-married women aged 15–49 years. Breastfeeding indicators were estimated for the last-born children (n = 2735). Selected indicators were examined against independent variables through cross-tabulations and multivariate analyses. Of the sample, 83.3% initiated breastfeeding within 1 h of birth. Continuation rates declined from 92.6% in first year to 83.5% in second year. Exclusive breastfeeding (EBF) rate under 6 months of age was 75.8%, with median duration being 4.8 months. Delayed initiation of breastfeeding was associated with low birthweight [odds ratio (OR) = 2.24] and caesarean delivery (OR = 3.30), but less likely among female infants (OR = 0.75), mothers from ‘estate’ sector (OR = 0.61) or richer wealth quintile (OR = 0.60). Non-EBF was associated with children from urban areas (OR = 1.72) and estate sector (OR = 4.48) and absence of post-natal visits by a public health midwife (OR = 1.89). A child was at risk for not currently breastfeeding if born in a private hospital (OR = 3.73), delivered by caesarean section (OR = 1.46) or lived in urban areas (OR = 2.80) or estate sector (OR = 3.23). Those living in estates (OR = 11.4) and not receiving post-natal home visits (OR = 2.62) were more likely to discontinue breastfeeding by 1 year. Breastfeeding indicators in Sri Lanka were higher compared with many countries and determined by socio-economic and health care system factors