About 7.2 million years ago, African ungulate species began to appear and disappear faster. Ungulates are the hoofed herbivores, from antelopes to rhinoceroses, giraffes to hippopotamuses. Since then the giants, herbivores above a thousand kilograms, have dwindled.
Juan L. Cantalapiedra, of the National Museum of Natural Sciences in Madrid and the University of Alcalá, and ten colleagues set out to find why. The work appeared in Nature Communications on 31 July 2026. The team assembled 3,327 vetted fossil occurrences of 396 species spread over some 23 million years.
A species can drop out of the diversity count in two ways. Extinctions can become more frequent, or fewer new species can arise. The question is which mattered for the giants. To answer it the authors used a birth-death model of species, which works from the dates of first and last appearance in the sediments and estimates both rates through time. The model, called BDNN, uses a neural network so that animal traits and climate can change those rates, including in non-linear ways. It runs in PyRate, software written by Daniele Silvestro, one of the co-authors.
Two traits enter the model. Body mass, in six classes. And tooth crown height, low, medium or high. A high-crowned tooth withstands wear from grass and grit. The combinations give 18 ecological forms. Climate enters through two records from seabed cores off Africa. Wind-blown dust tracks aridity. The carbon isotope composition of leaf waxes tracks the spread of tropical grasses.
The overall picture runs as follows. After 7.2 million years ago speciation and extinction speed up together. After 3.6 million years ago speciation stops rising. At the start of the Pleistocene, 2.58 million years ago, extinctions jump and diversity falls across all families.
Aridity is the strongest driver of extinction. On its own it triples the rate at which species vanish. At peak aridity extinctions outnumber new species about two to one. Large species, though, held out better. At that peak a species under 45 kilograms went extinct only about 15 per cent more often than one over a tonne, while 2.2 times as many new small species arose. In 88 per cent of the model's posterior samples the extinction response was weaker in the larger size classes. High-crowned teeth, the abstract says, were also linked to fewer extinctions.
The giants' weakness lay in birth. In large herbivores speciation, the appearance of new species, is low to begin with. Environmental change sped speciation up by about 50 per cent in animals under 180 kilograms and slowed it by about 20 per cent in megaherbivores. In some lineages the aridity of the past five million years held it back further. The paper's title puts it plainly. Lower speciation, not higher extinction.
And humans? The model has no human factor. The authors write that African megaherbivore diversity and abundance were already falling by about four million years ago, long before hominins could hunt them or reshape the landscape. In the final phase, roughly the last 130,000 years, the model finds no size-selective extinction. The authors allow that hunting may have contributed to particular losses, but not to the disappearance of whole ecological niches. Elsewhere, in the Americas and Australia, the debate on the human role is a different one, and the paper does not address it.
The limitations are stated. The climate records are averaged over wide areas and miss regional differences. Environmental data are sparse before five million years ago. The fossils come mostly from savanna settings. The credible intervals, the Bayesian measure of uncertainty, are broad, and the authors warn against extending the size results to much older periods.
The version published on 31 July lacked the funding section. It was added later, and Crossref records an update on 9 September 2026. The work is supported by public money from Spain, the European Union, Germany, Sweden, Finland and Switzerland. The authors declare no competing interests. Data and code are openly deposited.
A rhinoceros lives long, has few young and needs a great deal of room. Those traits shielded it from extinction. Over millions of years the same traits made new species rare.
References
Cantalapiedra Juan L., Lazagabaster Ignacio A., Blanco Fernando, Hauffe Torsten, Bibi Faysal, Ríos María, Mennecart Bastien, Saarinen Juha, Silvestro Daniele, Hernández Fernández Manuel, Sanisidro Oscar, "Lower speciation, not higher extinction, drove African megaherbivore diversity collapse", Nature Communications, 17, 1, 9373, 31 July 2026 (received 4 August 2025, accepted 17 July 2026; updated 9 September 2026 to add the funding section), open access, CC BY 4.0 licence. doi: 10.1038/s41467-026-76105-2
Trait, phylogeny and occurrence data: figshare, doi: 10.6084/m9.figshare.29589368; PyRate code: github.com/dsilvestro/PyRate
