Sal (Shorea robusta) forest is found in an extensive array of conditions in lowland Nepal, and has been heavily used by both government and local people. Thus, we did a study to provide information on sal forests occupying a broad range of historical and environmental influences across an extensive range in the western Terai of Nepal and explore factors - both environmental and anthropogenic - that may be responsible for variation of forest structure. Trees, saplings, seedlings and shrubs were sampled along transects (2 km long) in two protected areas and two proposed community forests. The protected areas had three transects each, and a single transect covered two proposed community forests. Samplings were done every 200 in along the transects, a plotless technique sampled trees (> 5 cm dbh). With tree sampling point as the center, shrubs and saplings (1-5 cm dbh and > 1 in height) were sampled in 5 in radius circular plots and herbs and seedlings (< 1 in height) were sampled in 1 m(2) circular plots nested within shrub plots. Altogether 131 species were recorded: 28 trees, 10 shrubs, 6 climbers and 87 herbs. The mean density across all plots was 220 trees/ha and the average basal area was 13.2 m(2)/ha. Three different associations of sal forest were identified by cluster analysis. Community types distinguished in the classification analysis were clearly separated in the site ordination. None of the environmental variables measured (pH, percent organic matter, total nitrogen, available phosphorous, available potassium and soil texture) explained the distribution of plots in the site ordination. We concluded that rainfall and past disturbances (fire and anthropogenic use) are mainly responsible for different community types. Community types were different in structure and composition, thereby representing unique entities. The protection and maintenance of each of the different communities through forest management is important for biodiversity conservation. (c) 2007 Elsevier B.V. All rights reserved.