Morphological convergence and the thylacine
Previous studies have viewed the thylacine from a starting point of convergence with the gray wolf/dog species complex [19,20,21,22,23,24,25]. We find little support for morphological convergence of the thylacine cranium within these species. Rather, we find repeated and substantial support for convergence with a specific group of canids, the African jackals and South American ‘foxes’, that share a distinct feeding ecology separate from that of the gray wolf/dog species complex (Canis lupus sensu amplo) (Fig. 7). These convergent species are, broadly speaking, mid-sized (5–25 kg) carnivores with an average prey size < 45% of their own body mass. Our results show little to suggest that the thylacine was morphologically convergent with the gray wolf/dog species complex, and by extension little to suggest ecological similarity.
Outside of the wolf/dox complex, the thylacine has previously been suggested to be morphologically similar to V. vulpes [24, 28], and phenotypic convergence between the two species has previously been assessed and interpreted as highly significant [41]. While we do find some support for convergence with V. vulpes, echoing these previous findings, we find much stronger repeated support outside of the true fox group. The previous phenotypic convergence study [41] fundamentally differed in purpose and design from the current study, and was tailored to assessing a concept of general convergence with canids within a broad grouping of mammals, not with the intent to identify ecological analogues. All aspects of that study design reflect this, from the taxa selection (inclusive of non-ecologically-meaningful species, e.g., koalas and wallabies) to the source of the landmark protocol/dataset [42], which avoided functionally-relevant areas of the cranium, such as the rostrum. Although that study was not designed to find informative convergent analogues with nor to infer the ecology of the thylacine, it did find that the thylacine was closer in phenotype to canids than expected from phylogeny [41]. That being said, the study was not attempting to, and did not, show with any precision which canids were most like the thylacine, why they were, or what that similarity may mean functionally or ecologically.
Prey size, morphology, and the thylacine
Cranial and facial shape place the thylacine with other predators that routinely take prey < 45% their own body mass, based on both morphospace occupation and canonical variate discrimination. This supports research showing that the cranium and mandible of the thylacine would perform poorly under the stresses encountered in taking large-bodied prey [35, 36, 43]. This is also consistent with interpretations following recent estimates of thylacine average body mass at ~ 16.7 kg [37]. Most mammalian predators take prey substantially smaller than themselves, in part due to the energy expenditure vs. intake costs brought about by locating, capturing, and killing generally uncooperative prey [40, 44,45,46]. This is especially true for carnivores under 21 kg in body mass, where foraging costs do not outweigh the metabolic demands of the predator, and are less than the costs and associated dangers of capturing and killing large-bodied prey. Within large-bodied predators over 21 kg (e.g., Canis lupus), there is a tendency to switch to prey larger than 45% of their own mass, due to the need to increase the net gain per hunting effort as their absolute metabolic rate scales with body mass—it becomes too costly to find, catch, and consume enough small meals, so they tend to switch to larger, and more difficult, prey.
Despite the thylacine being commonly considered as a ‘marsupial wolf’ [e.g., 27,47,48], some authors have been sceptical regarding such predatory capabilities in the thylacine [19, 29]. Two general strategies for procuring large and potentially dangerous prey are seen in extant mammalian carnivores. Felids tend to be ambush hunters, with powerful forelimbs capable of supination for the capture and restraint of prey, and robust, shortened rostra to deliver crushing or locking bites to the head, muzzle, and neck. Large hypercarnivorous canids (e.g., Ca. lupus, Cu. alpinus, and Lycaon pictus) tend to be highly social group-living pursuit hunters. These canids possess elongate and reduced distal limbs enabling efficient locomotion and use their pack size to overcome large prey with numerous shallow or tearing bites. The cranium of the thylacine is not cat-like, and the elongate and narrow rostrum precludes the muzzle, throat, or nape/back of skull bite used by large felids. Additionally, there is little indication of locomotor specialisation in the postcranial skeleton, and studies have shown that neither the limb proportions nor the forelimb morphology support a specialised, cursorial habit, nor do they support the ambush of large-prey [31,32,33]. While the thylacine may share similar cranial morphology to the African jackals and South American ‘foxes’, it shows none of the specialised limb morphology that these canids possess, which all show varying degrees of the cursorial specialisations (e.g., limb elongation, distal element reduction and compression) shared by Ca. lupus. Rather, the forelimb of the thylacine seems to be that of a relatively generalised ambush or pounce predator lacking the anatomical specialisations required to handle large prey.
The morphospace occupation and canonical variate discrimination results are echoed in the feeding habits of the canids found to be strongly convergent with the thylacine. This group of canids (Ch. brachyurus, Lu. adustus, Lu. mesomelas, and Ly. gymnocercus) as a whole focuses on prey far below their own body mass, mostly small vertebrates such as rodents and lagomorphs. Two species recovered as significantly convergent in both total cranial analyses but not recovered as such across both facial patch analyses are the ~ 6 kg V. vulpes and the ~ 15 kg Ca. latrans. Like the above canids, the diet of the red fox is also comprised largely of rodents, though it is a flexible and opportunistic predator that will occasionally take small mammals up to ~ 3.5 kg, roughly 50% of their body mass [49,50,51]. The coyote primarily consumes roughly similar-sized prey to the above canids, with lagomorphs making up the majority of its diet across much of its range. However, Ca. latrans has a highly flexible social structure, and in packs are capable of predation on relatively large-bodied prey, such as juvenile cervids [51,52,53]. The ~ 9 kg Ly. culpaeus, found to be significantly convergent with the thylacine in the neurocranial dataset only by the distance-based analysis, also has a dietary regime roughly similar to that of the red fox and coyote. Considered the most carnivorous of the South American ‘foxes’, the diet of Ly. culpaeus comprised mostly of rodent and lagomorph prey, though it is noted to often prey on the largest of the small mammals available, e.g., hares and occasionally newborn–juvenile domestic sheep [51, 54]. None of these three small prey-focused canids that are able to take larger-bodied prey (the culpeo, red fox, and coyote) are found to be significantly convergent with the thylacine across both facial patch analyses, with the culpeo not found to be significantly convergent in either facial patch analyses.
The neurocranial patch CVA grouping of the thylacine with large-prey specialists echoes a similar result found in marsupials by using muscle cross-sectional area to estimate bite force [34]. Within that study, the estimated bite forces for marsupials, and the thylacine in particular, were found to be exceptionally high, both relatively and absolutely. These results are not supported by biomechanical analyses of the thylacine cranium, which find it particularly unsuited to handle the stress of either producing such high bite forces or of handling large prey items, nor by the feeding ecology of some of the marsupials examined [34,35,36, 43]. The cross-sectional area available for muscle tissue is negatively affected by brain expansion, which limits the area available for musculature between the neurocranium and zygomatic arch. Marsupial carnivores have endocranial volumes that are approximately 40% of the volume in a placental carnivore of similar body mass, creating a much larger cross-sectional muscle area available for a given body size, and seemingly regardless of average prey size [34, 55]. We find that while 3D neurocranial shape does correlate with prey size in placental carnivores, it does not seem to be strongly correlated with prey size within marsupials, though the small sample size prevents any firm conclusion. Cross-sectional muscle area, and by extension neurocranial shape, may not be a good predictor of in vivo bite force in marsupials, as previously noted [35], and may not correlate with prey size in marsupials.
Morphology and diet
Surprisingly, we find no correlation between diet and cranial shape, a result in contrast to that of previous studies [e.g., 21,42]. This is possibly due to our focus on carnivorous species; we avoided including herbivorous carnivorans and those trending towards frugivory or omnivory, restricting the phenotypic range. Furthermore, relative size of the food object consumed may be a larger constraint on the cranium of faunivores than the material properties of the food [e.g., see 56]. A diet-based pattern might emerge if we included more disparate, herbivorous species, or sampled data from the dentition, as it actively engages with the food [57].
Concepts of convergence
The concept of phenotypic convergence is necessarily broad, and itself has many different possible interpretations. As a classic example of convergence, the ichthyosaurs are noted to have strongly converged on the same general body plan as fish [58], and the degree of convergence between thunnosaurian ichthyosaurs and lamnid sharks is striking [59]. However, ichthyosaurs displayed a range of body plans [60], from eel-like to tuna-like, and fish themselves display an incredibly vast array of body plans and ecologies. The issue then becomes, like so many others in science, one of scale or resolution: at what level are you invoking the concept of convergence? It is true that ichthyosaurs were convergent with fish, and that some were convergent with lamnid sharks. However, it is equally true that those ichthyosaurs were not convergent with all fish, and that some were not convergent with lamnid sharks. If meaningful inferences about the functional ecology of an extinct animal is the intent of the convergence study, then the data (comparative taxa, phenotype, etc.) should also meaningfully reflect the question.
When trying to understand the functional ecology of an extinct animal by comparison with modern analogues, broad scale concepts of convergence can be helpful and evocative. Such examples of broad scale convergence as the ichthyosaurs above, or as between ceratopsians and bovids [61], crocodilians and odontocetes [62], borophagine canids and hyaenids [63], or toxodontids and hippopotamids [64] can offer support for broad ecological similarities. However, this broad scale interpretation of convergence can fail to recover a higher fidelity view of the functional ecology of an extinct animal, since bovids, odontocetes, hyaenids, canids etc., each encompass animals with widely disparate ecologies. By framing both the concept of convergence and interpreting the resulting data within a high-resolution functional ecology study design, we can start to form much more precise hypotheses of the ecologies of extinct animals.
This is not to say, however, that a precise one-to-one matching between extinct and extant comparatives is necessarily the goal of such a study, or even possible. The thylacine here does not actually fall within the morphospace of any living canid comparatives, strongly suggesting that it is not directly comparable with any of these canids, whether jackal or wolf (Fig. 4a–c). A lack of direct correspondence here is unsurprising, as many previous studies have noted that the phenotypic similarities are largely superficial [31,32,33], and a concept of the thylacine as strongly ecologically convergent with African jackals is probably rather wrong. But, when trying to reconstruct the functional ecology of an extinct animal it is better to be less wrong, and identifying more precise analogues supports better and more informative reconstructions.