Introduction A group of 317 HIV-1 infected pregnant women and 53 postpartum HIV-negative women were recruited for a two-year prospective descriptive study of psychosocial and other determinants of antenatally planned and actual postnatal feeding, associations between maternal status and infant feeding practices, and health outcomes. Methods The subjects were interviewed periodically for 2 years using structured research instruments. Anthropometric measurements, biomarkers of nutritional status and measurements of pysychosocial wellbeing were obtained from the mothers. Data was collected on infant feeding and outcomes for the babies. Results At recruitment, 74% of mothers planned to formula-feed. Significant differences between these women and those who planned to breastfeed emerged. After delivery, 25% of the women who antenatally planned to formula-feed changed their minds and actually breastfed. Conversely, half of the women who antenatally planned to breastfeed actually formula-fed. Some significant reasons emerged for these feeding changes. Most mothers were well-nourished or overweight. Breastfeeding mothers lost little weight between six weeks and six months after delivery. At the end of follow-up, 65% were obese. While there were differences between HIV-infected and uninfected women in respect of micronutrients, no deficiencies were observed. Vitamin A and selenium concentrations were higher in the HIV-infected women than uninfected women at six weeks. There were no significant micronutrient changes over time. Most mothers maintained an adequate immune status with only slow deterioration of CD4 counts. At two years postpartum, 60% had a CD4 cell count greater than 500cells/mm³, and only about 8% less than 200/mm3. HIV transmission was 15% by 24 months of follow-up. Among the 65 ever breastfed children, 16 (24.6%) were HIV-infected compared to 12.8% of never breastfed children. Most children were growing normally, suggesting that, overall, maternal HIV status did not interfere with feeding ability. Eight mothers (3%) and 33 children (11%) died. Only 12 of 33 children who had died had a positive HIV-PCR. By 2 years, 78% surviving HIV-infected children had been initiated onto ARV therapy. Maternal adherence to HAART was poor. Conclusion HIV and infant feeding counselling is inadequate in the routine PMTCT programme, with stigma and lack of disclosure continuing as major barriers to appropriate care. Whilst maternal obesity was common, most children were growing normally. Weaknesses in routine PMTCT services were identified, and compliance with HAART was poor.