1 Greenland, New Guinea, and Borneo are all larger than Madagascar. However, Greenland is joined to North America by sea ice during winter, while during the Pleistocene New Guinea repeatedly merged with Australia and Tasmania to form the super-continent of Sahul, just as Borneo did with Java, Sumatra, and the Southeast Asian mainland. Separated from continental Africa by the 3292-m-deep Mozambique Channel and covering some 587,000 km², there is thus ample justification for deeming Madagascar to be the world’s largest (persistent) island.

2 Muldoon et al. (2012:25) note that, unusually for an open-air fossil-producing site in Madagascar, Christmas River “has the potential to yield a stratigraphic profile in the sense of vertical, chronological relations of sedimentary units”, but do not then provide such a section drawing. Their description of the site’s stratigraphy is confined to the observation that “the deepest and most prolific layer reached… is a grey-green clay” containing very early Holocene remains of Aepyornis, crocodiles, tortoises, and dwarf hippopotami and that “isolated remains” were found above this layer. Elsewhere, they note that the bone bed is “10–15 m deep” and that—while the sapphire miners who found the site “focused on the collection of large macrofossils, and the younger layers of the pit have not yet been fully explored”—sediment samples from the subsurface also confirmed the presence of small vertebrate and botanical remains (Muldoon et al. 2012 Muldoon, K. M., B. E. Crowley, L. R. Godfrey, A. Rasoamiaramanana, A. Aronson, J. Jernvall, P. C. Wright, and E. L. Simons. 2012. Early Holocene fauna from a new fossil site: a first assessment from Christmas River, south central Madagascar. Madagascar Conservation and Development 7:23–29. [Google Scholar]:24).

3 Lack of space precludes discussion of other potential paleoenvironmental proxies for human settlement. However, changes in the frequency of other pollen taxa, charcoal particles, and Sporormiella fungal spores that have been cited as evidence of human presence and landscape modification (e.g., Burney et al. 2003) present similar difficulties to those outlined here at Lake Tritrivakely: chronology depends upon interpolation from a small number of radiocarbon dates and the assumption that sedimentation rates have remained constant; and the changes observed are open to explanation by non-human agencies, including climate and taphonomy (see Perrotti and van Asperen 2019). They are not, in other words, based upon the presence of unimpeachable traces of human activity recovered in primary context and dated by unambiguously associated radiometric ages, the three criteria set out above.