A vegetation survey in four regions (Deltaic, Dry, Hilly and Plain) in Bangladesh was conducted with reference to marginal (> 0.02 - 0.08 ha), small (> 0.08 - 0.14 ha), medium (>0.14 - 0.20 ha) and large (> 0.20 ha) farm categories. Eighty homegardens (five from each farm category in each of the four regions) were assessed in terms of structure, species composition and diversity of the perennial species. Ordination of floristic data from the 80 homegardens showed a distinction between the Dry and the other regions due to a combination of lower species richness in the Dry region homegardens and several species exclusive to the regions. Floristic differences led to less marked but nevertheless important differences among the other regions, also. Most species were planted in the border of the homegardens irrespective of farm size and region. Food and fruit producing species dominated near the living quarter and working areas and small plots of annual vegetables and crops separated this part of the garden from the more distant parts favoured for timber species. Six vertical strata were recognised with higher plant density and species richness recorded in the lower three (<7 m). In total ninety two perennial species were recorded for the set of 80 homegardens surveyed. From gardens in the Deltaic region 67 species were recorded. Corresponding figures were 56 for the Plain region, 54 for the Hilly region and 46 for the Dry region. Diversity was highest among food and fruit producing species, followed by the timber species. Indigenous management techniques in homegardens were investigated using different PRA methods. Farmers' used all sorts of planting materials to regenerate homegardens plants. Homegarden provided more than three quarters of the required planting materials. Mother trees were selected for fruit species only. Farmers practised simple cultural operations (weeding and pruning) which were rarely intensive. The fertility of homegardens was maintained naturally from leaf litter, faeces of animals, kitchen waste and the mud of fish ponds. There was a clear separation of tasks between men and women for homegarden management. Farmers spent only 5- 12 % of their labour and 4-7.5% of their active time in homegarden management. Farmers have considerable knowledge about mother tree selection, silvicultural protection, as well as about positive and negative interactions. Farmers' had more knowledge about above-ground interaction than below-ground interactions. Women are more knowledgeable than men in many management aspects of homegardens, but their access to various resources is limited. Farmers are very much aware of different functional aspects of homegardens, but due to the lack of good planting materials, funds and extension supports do not utilise their full awareness and knowledge of these matters for homegarden development