1. Most previous work on individual recognition of felids through their tracks has concentrated on tigers (Panthera tigris) [9-12], although other studies have looked at mountain lion (Felis concolor) [13-15] and snow leopard (Uncia uncia) . Karanth et al.  determined that 30 years of pugmark censuses to estimate tiger abundance in India failed because statistical assumptions for abundance estimates were not considered and even with 30 years of censuses, data on spatial distribution was still lacking. While Wilting et al.  adopted more rigorous methods, we maintain that the recognition of individual felids from pugmark sets is extremely difficult. Their principal component analysis (PCA) of tracks does indeed separate the six track sets into three or four groups, but this method is flawed because each of the 50 tracks in the six track sets is treated as an individual case, whereas a PCA assumes that cases are independent. It is therefore not surprising that the individual tracks group so clearly; they were part of a dependent set of tracks made by one individual. Wilting et al. should have averaged track measurements for each track set and conducted their PCA with six rather than 50 cases. This reduced sample size is too small to reliably extrapolate to a large area.
2. When investigating population sizes and space use of carnivores, metabolic requirements and energy acquisition, as influenced by prey availability, are also important aspects to take into account [16,17]. Prey diversity and abundance are features that vary between sites, for example logged and unlogged forests, or areas where hunting is permitted [18-20] and these features will certainly have an effect on the population of clouded leopards that can be supported in such areas. However, Witling et al.  have given no consideration to the effect of prey base when estimating clouded leopard population sizes, and have extrapolated information obtained in one site with a particular prey assemblage to other forest areas throughout Sabah. The close proximity of their study areato oil palm plantations and to a river would affect local prey densities and consequently could result in higher clouded leopard densities. Mammal densities derived from such locations are liabletooverestimation, and mistakes at the smaller scale will become more pronounced when extrapolating these numbers to forested areas across the entire state.
3. The main threat to clouded leopards apart from forest loss is hunting . Although hunting pressure on particular species in Borneo remains mostly unstudied, it varies considerably. Hunting in coastal, Muslim-dominated areas focuses on different species than in Borneo's interior. Also, local culture and food preferences determine whether certain species are targeted [, TNC unpubl. data]. Without knowing how hunting pressure varies within Sabah, extrapolation of densities from one protected area to the whole of the State is unjustified.
4. Wilting et al.  conclude that there are 8–17 individuals per 100 km2 in their study area. While the authors maintain that their density estimate is just that, an estimate, and make the point of referring to it as a rough figure, the variance in their estimate is considerable, especially as they have attempted to calculate populations for the whole Sabah region. They assume that all forested areas would have similar densities. In doing so, they have also ignored small isolated populations that fall outside larger protected areas and reserves. It has been shown that clouded leopards are present in areas of heavy logging and high human activity, with suggestions that home range size may be larger in such habitats  and consequently, densities lower.
As a result of the above factors, the available data do not justify the estimation of the number of clouded leopards in the study area and, even less the extrapolation, however careful, of such data regarding the size of populations in an entire region, especially when spatial variation in threats such as hunting, or ecological requirements, have not been quantified.