The subject matter of this review is the seasonally dry tropical and subtropical woodlands and forests of the world. We follow Specht (1981) in differentiating woodland (typically <30 percent canopy cover) from forest (typically> 30 percent cover) onthe basis of percentage canopy cover but clearly discrete boundaries do not exist. The definition of a dry season is more problematic. However, we propose that a dry season has two attributes. First, it is a predictable annual event, and second, inthedriest 3 months of the year , less than 10 percent of annual precipitation occurs and typically less than 25 mm per month falls. In few sites does absolutely no rainfall occur in the dry season, if a sufficiently long (a century or more) time average istaken. Seasonally dry tropical areas are occupied largely by savanna, but also include areas of dry forest in which grasses are mostly absent. Savannas are here defined as: tropical or sub-tropical biomes with a tall (> 30cm) continuous grass understoreyduring the wet season and a discontinuous tree canopy . Grasses are usually dominated by those with the C4 photosynthelic pathway. Rainfall is highly seasonal, with the wettest period occurring during the warmest months. Dry tropical forests and savanna vegetation types occupy far greater total areas worldwide than does wet tropical rainforest (Olivares and Medina , 1992; Murphy and Lugo, 1995). These seasonally dry tropical areas are distinguished from wet tropical areas by a marked dry period that occurs each year. This means that closed forest canopies do not develop, except in small areas where topographic factors lead to permanently high soil moisture content, and which are thus not truly dry. Conversely, shallow and freely draining soils can also support savannas in the middle of areas otherwise covered in rainforest. This clearly illustrates the importance of soil water availability over the year, rather than rainfall alone, as a determinant of vegetation structure in seasonally dry systems. Seasonally dry tropical areas do, however, have a reliable and pronounced wet season, which distinguishes them from deserts. Rather than become embroiled in semantic discussion as to the definition of savannas, we must accept thatsavannas represent a regionalong a continuum extending from desert at one extreme to tall closed rainforest at the other, with subtle shifts and mergers occurring at boundaries. Vegetation units occur as segments of continua, not as distinct separate units (Murphy and Lugo, 1995). The co-existence of trees and grasses constitutes perhaps the most striking feature of savannas; this has been reviewed by Scholes and Archer (1997). These authors conclude that traditional simple models based on separation in rooting depth of trees and grasses are inadequate, and that the effects of browsers, grazers and fires on tree recruitment must be considered. In many parts of the world, savanna reverts to woodland in the absence of fire or other disturbance. Thus the distinction between savanna and dry tropical forest or woodland is somewhat arbitrary. In both biomes, trees must contend with seasonal drought and high temperatures, to which conditions they can be expected to share similar physiological responses.