How traits influence species persistence is a fundamental question in ecology, evolution and palaeontology. We test the relationship between dietary traits and both species duration and locality coverage over 40 million years in North American canids, a clade with considerable ecomorphological disparity and a dense fossil record. Because ecomorphological generalization—broad resource use—may enable species to withstand disturbance, we predicted that canids of average size and mesocarnivory would exhibit longer durations and wider distributions than specialized larger or smaller species. Second, because locality coverage might reflect dispersal ability and/or survivability in a range of habitats, we predicted that high coverage would correspond with longer durations. We find a nonlinear relationship between species duration and degree of carnivory: species at either end of the carnivory spectrum tend to have shorter durations than mesocarnivores. Locality coverage shows no relationship with size, diet or duration. To test whether generalization (medium size, mesocarnivory) corresponds to an adaptive optimum, we fit trait evolution models to previously generated canid phylogenies. Our analyses identify no single optimum in size or diet. Instead, the primary model of size evolution is a classic Cope's Rule increase over time, while dietary evolution does not conform to a single model.

Skeletal-injury frequency and distribution are likely to reflect hunting behaviour in predatory vertebrates and might therefore differ between species with distinct hunting modes. Two Pleistocene predators from the Rancho La Brea asphalt seeps, the sabre-tooth cat, Smilodon fatalis, and dire wolf, Canis dirus, represent ambush and pursuit predators, respectively. On the basis of a collection of over 1,900 pathological elements, the frequency of traumatic injury across skeletal elements in these two species was calculated. Here we show that the frequency of trauma in the sabre-tooth cat exceeds that of the dire wolf (4.3% compared to 2.8%), implying that the killing behaviour of S. fatalis entailed greater risk of injury. The distribution of traumatic injuries also differed between the two species. S. fatalis, an ambush predator, was injured more often than expected across the lumbar vertebrae and shoulders whereas C. dirus, a pursuit predator, had higher than expected levels of injury in the limbs and cervical vertebrae. Spatial analysis was used to quantify differences in the distribution of putative hunting injuries. Analysis of injury locations discriminated true hotspots from injury-dense areas and facilitated interpretation of predatory behaviour, demonstrating the use of spatial analyses in the study of vertebrate behaviour and evolution. These results suggest that differences in trauma distribution reflect distinct hazards of each species' hunting mode.

Borophagine canids have long been hypothesized to be North American ecological ‘avatars’ of living hyenas in Africa and Asia, but direct fossil evidence of hyena-like bone consumption is hitherto unknown. We report rare coprolites (fossilized feces) of Borophagus parvus from the late Miocene of California and, for the first time, describe unambiguous evidence that these predatory canids ingested large amounts of bone. Surface morphology, micro-CT analyses, and contextual information reveal (1) droppings in concentrations signifying scent-marking behavior, similar to latrines used by living social carnivorans; (2) routine consumption of skeletons; (3) undissolved bones inside coprolites indicating gastrointestinal similarity to modern striped and brown hyenas; (4) B. parvus body weight of ~24 kg, reaching sizes of obligatory large-prey hunters; and (5) prey size ranging ~35–100 kg. This combination of traits suggests that bone-crushing Borophagus potentially hunted in collaborative social groups and occupied a niche no longer present in North American ecosystems.