Madagascar today is 65% grassland. Its grassy biomes were widely interpreted as 'anthropogenic' before Bond et al. [1] challenged that view, claiming that grasses colonized Madagascar during the Miocene. Through an analysis of grass clade endemism and specialization, Vorontsova et al. [2] have marshalled compelling support for this argument.Still, many questions about Madagascar's grasslands remain, and they are critical to debates regarding why so many forest-dwelling animals disappeared after human arrival. We know that forests declined and grasslands expanded, but we do not know precisely when, why or to what extent. Did significant pre-contact expansion occur [3]? Did post-contact aridification trigger this transformation [4]? Did forests disappear as people shifted their subsistence strategy from ephemeral foraging to dedicated agro-pastoralism [5]? Did early human colonizers severely reduce Madagascar's natural grazer community through hunting, radically changing the fire ecology and ultimately destroying forests [6,7]? Normally, grasslands are maintained by natural fire and by grazers, usually large-bodied herbivores. But Madagascar's modern grazers (e.g. cattle and goats) were recently introduced. If pre-contact grasslands were extensive, we would expect to see a significant palaeo-grazer guild. The prime candidates are pygmy hippopotamuses and giant tortoises-animals that, along with giant lemurs and elephant birds, comprised Madagascar's now-extinct megafauna.According to Vorontsova et al.[2], previously published stable carbon isotope (d 13 C) data suggest that (i) Madagascar's hippopotamuses consumed a 'high proportion of C 4 plants' (i.e. grasses) [8] and (ii) its giant tortoises had diverse diets [9]. We were motivated to write this comment for three reasons. First, Crowley et al. [8] never claimed that Madagascar's hippopotamuses ate a 'high proportion of C 4 plants'. We did note that subfossil hippos, tenrecs, carnivorans and the endemic rodent Hypogeomys australis had higher d 13 C values than subfossil primates (except Hadropithecus [10]), but we never defended a high proportion of C 4 plants in the diets of any of these animals. Second, Vorontsova et al.[2] omitted important published isotope data on hippos and tortoises from Madagascar [8,11,12]. Third, we believe that the literature would benefit from greater discussion of these data. C 4 grasses (including those measured in Madagascar [11]) typically have d 13 C values between 211 and 214‰, while C 3 plants have much lower values, averaging between 226.5 and 231.5‰ for different Madagascan forests [13]. Assuming an offset of þ5‰ between herbivore bone collagen and plant diet (reviewed in [14]), pure C 4 consumers should have d 13 C values more than 29‰, those with pure C 3 diets should have d 13 C values less than 221.5‰, and those that consumed a relatively even mixture of C 3 and C 4 plants should have values of 214 to 218‰. Thus, herbivores with values between 218 and 221.5‰ likely consumed more C 3 than C 4 plants.Carbon isotope values f...