Occurrences of grasslands in regions that climatically can support forests are commonly attributed to previous human activity. In mid to high latitudes this assumption is often correct. For example, in North America and north-west Europe numerous palaeoecological analyses have revealed the role of pre-historic and historic humans in the opening up of forested landscapes and the development of grasslands (e.g. Motzkin & Foster, 2002). These `cultural landscapes' are now viewed as important for conservation and there are many initiatives to conserve them’ (e.g. Agnoletti, 2006). What is less well known, however, is the role that humans have played in the formation of such grasslands in low-latitude environments. In South Africa and Madagascar, for example, there has been a similar tendency to attribute grasslands in regions that climatically can support forests to former human activity. But here, in contrast to North America and north-west Europe, there has been very little work to determine their antiquity. Also, in comparison with their mid- to high-latitude counterparts, the ‘anthropogenic’ label attached to these grasslands deems them unnatural and therefore of less conservation value. Consequently they are ignored in conservation plans, or worse, have management techniques put in place to restore ‘natural forested vegetation’. But what if some of these grasslands are in fact natural? What processes are responsible for the formation of grasslands in an environment that is ‘typically’ forested? And what does this mean in terms of their conservation value? A recent study by Bond et al. (2008) on the antiquity of Madagascar’s grasslands provides an excellent example of a study to address these questions. Here it has long been assumed that Madagascar’s highlands were once completely forested and that all grasslands are indicative of a degraded landscape; this assumption is mainly based on the apparent poverty of species and endemism in the grasslands relative to the forested parts of Madagascar (e.g. Lowry et al., 1997). Bond et al. (2008) therefore compared modern grasslands in southern Africa with the present-day species composition of Madagascan grasslands, using evidence including: (1) the number of C4 grass genera shared with representative areas of the African mainland; (2) the presence of endemic plant species or genera in the grasslands; and (3) the presence of a grassland specialist fauna indicative of the antiquity of the grasslands. Their analyses show that Madagascar’s grass flora has c. 87% of the C4 grasses found in southern Africa, with a significant endemic component and a broad range of ecological preferences, suggesting that some of the grasslands are unlikely to be of recent, anthropogenic origin. Furthermore, the presence of endemic, shade-intolerant trees and of endemic vertebrate and invertebrate grassland specialists (Fisher & Robertson, 2002; Goodman & Benstead, 2003) also points to an ancient origin for some Madagascan grasslands. This study therefore supports a growing body of evidence against the hypotheses that Madagascar was once completely forested and that all grasslands are indicative of a degraded, anthropogenic landscape; rather, the extensive highland grasslands are part of the natural landscape, and are relatively old. An obvious question to arise from this, then, is how and why did these grasslands form? What natural processes could have resulted in the opening up of a landscape and the dominance of grasslands (covering more than 75% of Madagascar) in a climate that can support forests? Three obvious contenders are: (1) climate change; (2) burning; and (3) the grazing of large herbivores. In relation to contention (1), palaeorecords from the central highlands of Madagascar indicate a possible climatic link, with fluctuations between wooded/grassland pollen signatures apparent at 125, 100, 83, 60 and 10–5 kyr intervals (Gasse & Van Campo, 2001). All of these transitions are suggested to be related to intervals of cooling/aridity. Thus climate change (intervals of cooling/aridity) could well account for some of their expansion. Another climatic explanation for grassland in the highlands is that the cool periods resulted in eroded soils that could only support grasslands (de Wit, 2003). Concerning contention (2), fire is a major factor influencing the distribution of present-day mesic C4 grasslands in South Africa and Madagascar. It is probable, therefore, that this was also the case in the past. Also, the effects of fire could have had synergies with climate changes in atmospheric CO2 (trees would recover from burning more slowly in cool and low-CO2 conditions favouring grassland expansion; W. Bond, personal communication). Palaeostudies to date do not indicate whether these ancient grasslands in Madagascar were dominated by C3 or C4 plants and this would be a good test of the burning hypothesis; grasses maintained by natural fires would be dominated by C4 plants, whereas climatic cooling would favour C3 grasslands. Evidence from grass cuticles has the potential to resolve this question (Wooller et al., 2000). Also a new mass spectrometry technique enables the δ13C of individual fossil pollen grains to be determined (Nelson et al., 2008), suggesting another possible line of future research. For point (3), the extent of large herbivore density and grazing upon this landscape is very difficult to determine from the fossil record. Fossil evidence suggests that Madagascar had a diverse herbivore assemblage, which may have included grazers – and this hypothesis needs further investigation (Burney et al., 2004). However, if grazers were important for the formation of highland grasslands, then there is also the question of how this openness remains, despite the extinction of hippos, elephant birds and giant tortoises in Madagascar (Dewar, 1984). Again, it may be that fire is the major ‘herbivore’ in these highland grasslands, and maintains them today as in the past (W. Bond, personal communication). Bond et al.’s (2008) synthesis therefore adds to an increasing body of evidence suggesting that Madagascar would have been a mosaic of forest, shrubland and grasslands prior to human settlement c. 2000 years ago. Such evidence is central to conservation planning and ecosystem management in Madagascar, which is currently, and perhaps understandably, focused on forest ecosystems. This broadening of focus to include the ancient grassland ecosystems highlighted in Bond et al.’s (2008) study would protect endemic plants and animals, including ground orchids, birds and ants, which are dependent on periodic fires that maintain open habitat. Clearly, the success of this conservation strategy also depends on the ability to distinguish ancient grasslands from recent anthropogenic grasslands, which are less likely to enrich the island’s beta diversity. Similarly, ancient, stable forest fragments are more likely to harbour endemic species, and identifying these will also be key to maintaining beta diversity. Studies such as Bond et al.’s provide essential data for understanding the antiquity of Madagascar’s grasslands, and together with future palaeoecological studies will help elucidate the nature and extent of the vegetation mosaic in Madagascar. Together, such studies could help to prevent the loss of beta diversity due to oversimplified management policies. Editor: John Lambshead