Sunday, May 17, 2020

Making Sense of Alvarezsaurid Paleobiology—I Think We're Doing Pretty Well, Actually

Alvarezsaurids may not have much of a presence in popular culture, but these dinosaurs have been a source of fascination and befuddlement to just about anyone who is familiar with them. In particular, the function of their bizarre forelimbs has inspired a great deal of discussion and speculation.

The forelimbs of alvarezsaurids were incredibly small relative to their body size. If it were only for their size alone, it would be easy to write these forelimbs off as having been vestigial structures with little to no function. However, their forelimbs had several well-developed features that suggest they were capable of some very powerful movements (and not just "powerful" in the sense that the arms of tyrannosaurids are occasionally claimed to have been).

Skeletal reconstruction of the alvarezsaurid Mononykus, by Scott Hartman (used with permission).

For starters, the humerus (upper arm bone) of alvarezsaurids had a strongly projecting deltopectoral crest, an attachment point for musculature responsible for drawing the forelimbs towards the body. This action likely would have been further enhanced by the keeled, bony sternum (breastbone). Furthermore, the ulna (one of the forearm bones) of alvarezsaurids had a well-developed olecranon process, which projects backward from the elbow joint and serves as an attachment point for muscles that extend the elbow.

Ever since the discovery of Mononykus revealed this unusual forelimb anatomy, these features have been likened to those of many animals in which the forelimbs are specialized for digging (Perle et al., 1993). Both the enlarged deltopectoral crest and olecranon process support muscles that can help digging animals break up whatever substrate they are excavating and shift it aside.

There are other aspects of alvarezsaurid anatomy that appear to be consistent with some form of digging behavior as well. These include an elongated hip region (later alvarezsaurids had seven hip vertebrae, up from the five ancestrally present in maniraptorans) in which the hip vertebrae were fused with the pelvic girdle into a synsacrum, which could have potentially served a bracing function during digging activity. In contrast, the lower extremities of the pelvic girdle in digging animals generally experience less stress, and as a result the tips of their pubis bones have often reduced or lost their connection with one another. This, too, is seen in alvarezsaurids.

Diagram showing the unusual forelimb skeleton of Mononykus, from Perle et al. (1993). Later studies would show that most alvarezsaurids had three-fingered hands, though the other two fingers were much smaller than the large thumb (as can be seen in Hartman's skeletal in the previous figure).

To be clear, it is widely recognized that alvarezsaurids almost certainly weren't using their forelimbs to dig tunnels or construct burrows for themselves. It's quite evident that the forelimbs would have been too short for that to have been practical. Instead, they share some additional forelimb characteristics with animals that practice hook-and-pull digging, in which a pointed appendage is hooked into a hard substrate and then used to tear that material apart, usually during the process of finding food. Animals that regularly perform this behavior include anteaters, pangolins, and some armadillos, namely the giant armadillo (Priodontes maximus), naked-tailed armadillos (Cabassous spp.), and three-banded armadillos (Tolypeutes spp.).

Although many adaptations for scratch-digging burrows are also advantageous for hook-and-pull digging, and some species practice both, hook-and-pull diggers tend to have a few other features that are generally not found in species that only scratch-dig. For example, hook-and-pull diggers usually have a single claw on each hand that is massively enlarged compared to the other claws, which allows them to concentrate force into a single sharp point to fracture the hard substrates that they break into. As it happens, alvarezsaurids fit this pattern as well, seeing as their thumb and its corresponding claw were much, much larger and more robust than the other two fingers on their hand. Mononykus (meaning "one claw") was named for this trait. In one alvarezsaurid (Linhenykus), the two smaller fingers appear to have been lost entirely.

From left to right: diagrams depicting the hand skeletons of an aardvark (Orycteropus afer, a scratch digger), a silky anteater (Cyclopes sp., a hook-and-pull digger), and Mononykus, from Senter (2005). (In case you missed the memo: it has been proposed that there are seven species of silky anteater.)

The enlarged claw of hook-and-pull diggers exhibits a great capacity for flexing towards the palm, which is important for gaining purchase on the materials that they excavate. At the same time, the fingers and claws are typically restricted from performing side-to-side rotation, guarding them from the risk of dislocation during strenuous digging (Taylor, 1978). In 2005, Phil Senter presented a study investigating the range of motion in the forelimbs of Mononykus. He found that even though the shoulder and elbow joints were quite limited in their range of motion, the base of the thumb could extend and flex in a wide arc of 86° and the corresponding claw could do so in an arc of 162° while being restricted in side-to-side rotation, providing another line of evidence consistent with hook-and-pull digging. (The mobility of the wrist joint could not be reliably reconstructed.)

Inferred range of motion in the upper arm (A), forearm (B), base of the thumb (C), and thumb claw (D) of Mononykus, from Senter (2005).

For this blog post, the concept that alvarezsaurids used their forelimbs for hook-and-pull digging will be called the "pickaxe model". This emphasizes the core concept that, much like how one might use a pickaxe, hook-and-pull digging primarily involves fracturing and prying apart hard substrates, as opposed to other methods of digging. This term also avoids assumptions about what type of substrates alvarezsaurids might have been breaking apart, which I will discuss later in this post.

It's probably fair to say that the pickaxe model is the leading hypothesis regarding alvarezsaurid forelimb function (Perle et al., 1994; Chiappe et al., 2002; Senter, 2005; Longrich and Currie, 2009). However, in my experience it is not rare for it to be viewed with some incredulity in popular contexts. Even when it is not outright challenged, it is sometimes presented as being the best only out of a range of subpar options.

Before I consider some doubts regarding the pickaxe model that have been raised in informal settings though, I would like to examine responses to the hypothesis in technical publications. Has the pickaxe model been disputed in the scientific literature? The answer to this is yes, and the most extensive arguments that I'm aware of were put forth by Agnolin et al. (2012) in their description of Bonapartenykus.

Agnolin et al. argued that movement in the alvarezsaurid forelimb was severely limited, with the implication that this would have prevented alvarezsaurids from engaging in digging behavior. To start off, they cited the restricted range of motion at the shoulder joint and the absence of an olecranon fossa on the humerus (not to be confused with the olecranon process on the ulna) in support of this premise.

However, limited range of motion at the shoulder is actually quite common in digging animals (Sansalone et al., 2020), likely as an adaptation to buttress their joints against the stresses of excavation. As for the absence of an olecranon fossa, this fossa is often reduced or lost in animals that typically hold the elbow joint in a flexed posture (Seiffert, 2010), which many digging animals do, and is consistent with the results of Senter (2005). In fact, the olecranon fossa tends to be most strongly developed not in digging animals, but in quadrupedal runners that need to keep the forelimbs straightened most of the time (Heinrich and Houde, 2006; Salton and Sargis, 2008). Although some digging specialists do have a marked olecranon fossa (Puttick and Jarvis, 1977), these species habitually use methods of digging in which increased extension of the elbow is important (Gasc et al., 1986). In other diggers, like pangolins (Steyn et al., 2018), the olecranon fossa may be fairly modest. Thus, neither of these traits is in direct contradiction to the pickaxe model.

The humerus of a ground pangolin (Smutsia temminckii), from Steyn et al. (2018). Note the shallow olecranon fossa ("Fo").

Agnolin et al. went on to question Senter's finding that the thumb of Mononykus was capable of swinging in a wide arc, referencing a "detailed analysis" by Sereno (2001) that had found that the base of the thumb could only extend by 15°. Yet Senter's results also indicate relatively limited extension at this joint, with flexion accounting for most of the range of motion. Sereno likewise found that the joint was more capable of flexion than extension, with both extension and flexion combined permitting an arc of 50°. Although not as large as the 86° arc that Senter found, Sereno's results still indicate that the thumb of Mononykus could form an adequate-looking hook when flexed, especially when considered alongside the movement possible by the thumb claw.

As far as I can tell, Sereno did not explain exactly what methods he used to infer range of motion (his paper was primarily concerned with the phylogenetic position of alvarezsaurids and not their functional morphology), so assessing the cause of discrepancies between the two studies is difficult. Regardless, even if one accepts the more limited range of motion that Sereno reported, it could scarcely be interpreted as evidence that alvarezsaurids were incapable of digging. As a matter of fact, Sereno wrote within his article that the unusual features of alvarezsaurid forelimbs "probably indicate digging habits", so he himself evidently didn't view his range of motion study as an impediment to the pickaxe model.

Range of motion in the thumb of Mononykus as inferred by Sereno (2001). Compare to the results of Senter (2005) two figures above.

Lastly, Agnolin et al. expressed skepticism that alvarezsaurid forelimbs were capable of powerful movements at all, noting that some flightless birds with essentially useless forelimbs have an elbow joint and enlarged olecranon process similar to those of alvarezsaurids. However, as I will discuss shortly, the functional advantages of having a large olecranon process in diggers also requires that the rest of the forearm is relatively short. Although flightless birds may have quite reduced forelimbs, to my knowledge none of them exhibit proportions like those of alvarezsaurids, in which the olecranon process is so long compared to the rest of the forearm. Furthermore, flightless birds generally lack the other features suggestive of the pickaxe model, such as enlarged muscle attachments on the humerus or the strongly-developed hand claw.

The forearm bones of the extinct flightless rail Mundia elpenor, from Olson (1973). The ulna is on the right, with the olecranon process pointing downward. Agnolin et al. (2012) cited this paper in support of their argument that flightless birds may have well-developed olecranon processes comparable to those of alvarezsaurids. Pardon me if I'm not convinced of the similarity...
(Olson described a second species of flightless rail in the same paper, and Agnolin et al. also reference the phorusrhacid Paraphysornis, but the ulnae of those birds don't exhibit more than a passing similarity to those of alvarezsaurids either, in my estimation.)

Given all that, I can't say that I find Agnolin et al.'s objections to the pickaxe model very convincing. How about discussions outside of the technical literature? From what I've seen, by far the main factor that raises apprehension of the pickaxe model is just how short alvarezsaurid forelimbs were. It's a reasonable thought to have: intuitively, we expect appendages that were employed in such arduous activity to be large in absolute size. However, there's a reason I don't think that the extreme length reduction of alvarezsaurid forelimbs is in conflict with the pickaxe model at all.

That reason is this: short appendages are good for digging.

That may sound strange, but at a basic level (i.e.: the level of biomechanics that I understand), it comes down to lever mechanics. In a lever system, force is applied to an in-lever and exerted on an external load by an out-lever. The longer the in-lever is relative to the out-lever, the more force can be exerted by the out-lever for the same amount of force applied. (Conversely, a relatively shorter in-lever allows the lever system to act more quickly, but with less strength.)

Diagram of a lever system (specifically a class 1 lever, though the distinction is not particularly important for this post), by Pearson Scott Foresman, public domain. "Effort" represents the force applied to the in-lever, causing the out-lever to act against the resistance of a load.

The forearm of an animal that uses its forelimbs for digging can be seen as a lever system, with the olecranon process as the in-lever and the rest of the forearm as the out-lever. Seeing as digging generally demands great strength, it is usually advantageous for a digging animal to have a relatively long olecranon process (increasing the length of the in-lever) and a relatively short remainder of the forearm (decreasing the length of the out-lever). The same principle can explain why a digging lifestyle might favor pronounced muscle attachment points on the humerus, but reduction in the length of the humerus overall.

Lever mechanics applied to ulnae, from Longrich and Currie (2009). (A) shows how the ulna can be seen as a lever system and (B) compares the ulnae of various theropods and mammals. The ulnae on the left belong to theropods, including several alvarezsaurids and the presumably non-digging Deinonychus and Allosaurus. The ulnae on the right belong to digging mammals, including an aardvark (Orycteropus), a pangolin (Manis), a giant armadillo (Priodontes), and an eastern mole (Scalopus). The numbers represent mechanical advantage, a measure of how long the in-lever is compared to the out-lever. The ulnae of alvarezsaurids consistently have mechanical advantage values comparable to those of digging mammals.

And this is essentially what we see in nature (Coombs, 1983; Kley and Kearney, 2007). You can look at any specialized digging animal, and I'm willing to bet that the vast majority of the time, the appendages that it uses for digging will be shorter than those of a similarly-sized, non-digging species to which it is reasonably closely related. The short limbs are a feature, not a bug.

Skeleton of a nine-banded armadillo (Dasypus novemcinctus), a scratch digger, photographed by David F. Schmidt, public domain. This species is not even a particularly specialized digger as armadillos go, but the forelimbs (especially the forearms) are visibly short relative to its body size.

Skeleton of a rock hyrax (Procavia capensis), a non-digging atlantogenatan similar in size to the nine-banded armadillo, photographed by David F. Schmidt, public domain. Despite not being especially specialized for fast running (which tends to promote increased limb length), the hyrax has relatively longer forelimbs, even accounting for the more extended limb posture it has been mounted in.

We need not even look only at mammals or forelimbs. Below are two species of burrowing frogs, the black rain frog (Breviceps fuscus), which digs with its hindlimbs, and the marbled snout-burrower (Hemisus marmoratus), which digs mostly with its head. Can you guess which is which?

The marbled snout-burrower (left) digs mostly with its head, whereas the black rain frog (right) digs with its feet, from van Dijk (2001) and composited into a single figure by Darren Naish. The marbled snout-burrower has a more pointed snout that it uses to penetrate soil, but the overall length of its skull is shorter compared to the rest of its body. Meanwhile, the hindlimbs of the black rain frog are much shorter and stockier for its size.

All right. But if short appendages are so useful for digging, why don't any of the extant hook-and-pull diggers have forelimbs reduced to the extent seen in alvarezsaurids? I suspect that it has to do with the fact that, even though these animals evolved this behavior several times independently, all of them are mammals that inherited quadrupedal locomotion from their ancestors. Using the limbs in locomotion exerts selective pressures on their anatomy that may be at odds with their use in digging. Not only are the stresses experienced during walking different from those during digging (Toledo et al., 2020), limbs that are used in weight support must be long enough to raise an animal's body off the ground, constraining them in how short they can get.

Ah, comes the rejoinder: pangolins are bipedal hook-and-pull diggers, and they haven't shrunk their forelimbs down to alvarezsaurid proportions! The bipedal locomotion of pangolins is now fairly well known thanks to popular documentary footage of them trundling endearingly on two legs. Far be it from me to complain about any positive attention given to pangolins; they certainly could use as much of it as they can get. However, many viewers appear to have come away from these video clips with misconceptions about the prevalence and frequency of pangolin bipedality.

In reality, only one out of the eight extant pangolin species walks primarily on two legs (Kingdon and Hoffmann, 2013; Challender et al., 2020). That one is the ground pangolin, which happens to be the one most frequently featured in documentaries. Even the ground pangolin is not an obligate biped, however; it can and does use quadrupedal locomotion on occasion, especially to clamber over inclined surfaces. Perhaps more importantly, it also uses its forelimbs to excavate and enlarge burrows. Although burrowing involves digging, and thus would select for relatively short digging appendages, this must be balanced with pressures demanding that those appendages have sufficient reach, as would be required for an animal to construct a space large enough to fit its entire body into.

Diagram showing part of the burrowing cycle of a Grant's golden mole (Eremitalpa granti), from Gasc et al. (1986). Note how the forelimb skeleton is relatively short, but needs to be just long enough to reach past the head.

In contrast, alvarezsaurids evolved from fully bipedal ancestors that most likely did not use the forelimbs for any form of locomotion, which could have given their forelimbs the freedom to shrink to smaller relative sizes than in digging mammals. There may have been another factor at play here, too: limbs that are not involved in locomotion have a tendency to increase drag while an animal is moving about, and it is therefore beneficial for there to be a mechanism that can keep these limbs as close to the body as possible. As noted previously, the shoulder and elbow joints of alvarezsaurids had limited ranges of motion (potentially another adaptation to digging), which would have prevented the forelimbs from folding up against the body. An alternative to achieving a similar effect then would have been to reduce the size of the limbs themselves.

Thus, there are at least two plausible selective pressures (improved digging mechanics and reduced drag) that would have favored short forelimbs in alvarezsaurids if they were using their forelimbs for digging, and both of these pressures may have conceivably acted to a greater extent on alvarezsaurids than on digging mammals on account of their demonstrably different ancestral body plan. Are these explanations speculative? Certainly. However, they illustrate how the short length of alvarezsaurid forelimbs could have been a logical consequence of specialization for digging, rather than a contradiction of the pickaxe model.

Another argument I've sometimes seen leveled against the pickaxe model is that alvarezsaurid thumb claws were of the wrong shape to have served as digging implements; that they were not sharp enough, flat enough, or hooked enough compared to those of mammalian hook-and-pull diggers. I've found these allegations difficult to verify for several reasons, not least of which is because high-resolution photographs showing the hand claws of a range of extant hook-and-pull diggers from multiple angles and (especially critically) without the overlying keratin sheath are hard to come by.

I can still give it a try though using what resources I can access. The vast majority of alvarezsaurid thumb claw fossils that I know of have worn tips, making it hard to assess how sharp they may have been in life. The most completely preserved example that I'm aware of is from Linhenykus, and it tapers to what looks like a serviceable point to me. The curvature, though seemingly not as hooked as in some mammalian hook-and-pull diggers, looks comparable to that of the digging claws in certain pangolins, especially given that the keratin sheath is not preserved. I will grant that the underside of the claw tends to be rounded instead of flattened in alvarezsaurids (though I remain uncertain as to the status of this feature in the bony claw core of mammalian hook-and-pull diggers). Despite this, a flat-bottomed claw doesn't seem imperative if the claw was being used mostly for breaking (as opposed to shoveling) hard substrate. I don't see anything that clearly would have prevented Linhenykus from using its thumb claw for hook-and-pull digging, let alone Albertonykus, Mononykus, or MPD 100/120 (the "Tugriken Shireh alvarezsaur") with their more robust claws.

The thumb claw of Linhenykus from multiple angles (attached to the rest of the thumb in the leftmost image), from Xu et al. (2013).

As with the subject of limb length, we must also consider that extant hook-and-pull diggers additionally use their hand claws for other functions, like weight-bearing (armadillos), climbing (most pangolins and anteaters), and burrowing (armadillos and pangolins), exposing them to potentially competing selective pressures. Alvarezsaurids most likely weren't using their claws for any of these functions, and as a result I would not expect their claws to have been strictly identical to those of mammalian hook-and-pull diggers.

However, qualitative impressions only go so far. It would be nice if someone did a quantitative comparison between alvarezsaurid claw shape and those of other animals... well, as it turns out, someone has! Lautenschlager (2014) was mostly concerned with investigating the function of therizinosaur claws, but as part of this study he assembled a large dataset characterizing claw shape in a variety of mammals and theropods, including alvarezsaurids. When all the claw shapes were plotted out (using a principal components analysis), alvarezsaurids fell right within the field formed by digging mammals. What's more, of the three mammals that alvarezsaurids plotted closest to, two of them were hook-and-pull diggers: the giant armadillo and a pangolin of unspecified species. (The third mammal was the plains pocket gopher, Geomys bursarius, a scratch-digging rodent.) Although this does not demonstrate that alvarezsaurids used their claws for digging, it does suggest that their claw shapes were within the range of variation we'd expect of digging animals.

Morphospace graph depicting the range of claw shapes observed in theropods and mammals, from Lautenschlager (2014). The claws of alvarezsaurids fall entirely within the space formed by those of fossorial (digging) mammals.

So as far as I'm concerned, no aspect of alvarezsaurid morphology described to date is inconsistent with the pickaxe model. We can also approach this topic from a different perspective: what characteristics would we expect to see in an animal that practices hook-and-pull digging?

In a 1983 study on the functional morphology of clawed herbivorous mammals (e.g.: ground sloths, chalicotheres, etc.), Margery Chalifoux Coombs outlined a list of features that are commonly found in and likely adaptive for digging mammals, including hook-and-pull diggers. We can use this list as a guideline for predicting forelimb-dominated digging in extinct tetrapods. It's worthy of note that Coombs used these features to infer digging habits in mylodontid ground sloths, years before mylodontids were identified as probable tracemakers of large burrows from South America (Vizcaíno et al., 2001).

What follows is Coombs's list of digging-associated traits (not necessarily in the order she presented them in), with parenthetical comments assessing their presence in alvarezsaurids. I have modified the wording of each entry, partly to account for differences between the musculature of mammals and dinosaurs, but I've tried to preserve the functional significance of each feature:
  1. Relatively short forelimbs (yes)
  2. Enlarged attachment points for muscles that pull the forelimb towards the body (yes, large deltopectoral crest and keeled sternum)
  3. Enlarged attachment points for muscles that extend the elbow (yes, elongated olecranon process)
  4. Enlarged attachment points for muscles that extend the wrist and fingers (yes, large humeral ectepicondyle)
  5. Enlarged attachment points for muscles that flex the wrist and fingers (yes, large humeral entepicondyle in Patagonykus, merged or replaced with an enlarged distal condyle in later alvarezsaurids)
  6. Sequential reduction of the forelimb elements towards the tip, such that the wrist and palm are shorter than the forearm, which is in turn shorter than the upper arm (yes), including palm bones that are short and wide (yes)
  7. Hip region of the vertebral column long and fused to the pelvis (yes, synsacrum with up to seven vertebrae)
  8. Reduced contact between the pubis bones (yes)
(Funnily, several of these traits are convergent with modern birds, which once led some paleontologists to consider alvarezsaurids a type of avialan. It makes some sense why those traits would arise in both groups: forelimb-powered flight likely also favors the ability to draw the forelimbs forcefully towards the body. Of course, it selects in a rather different direction when it comes to forelimb length.)

Coombs additionally noted a long tail for balance as a common trait of large digging mammals, and one is certainly present in most mammalian hook-and-pull diggers. Like most other non-pygostylian theropods, alvarezsaurids had tails that would be considered long by mammalian standards. However, seeing as they would have inherited this tail from already long-tailed ancestors, I think it's fair to exclude it as an adaptation specifically for digging in this case, even if it was potentially of use during excavation.

Based on similarities among extant hook-and-pull diggers, we can probably add at least a few more anatomical predictions specific to this method of digging:
  1. Single massively enlarged claw on each hand (yes)
  2. Great capacity for flexion of the enlarged claw (yes)
As Coombs also pointed out, though we might expect digging species to possess many of these features, not all of them are found in all digging mammals. She specifically noted trade-offs with other forelimb functions (such as running and climbing) as potential reasons why the acquisition of these characteristics might be compromised. However, two of the digging taxa she examined did have all or nearly all of the listed features: pangolins and the giant armadillo. Beyond their ecological similarities, these two taxa share an interesting commonality: they may not be obligate bipeds, but they are the most bipedal of the taxa studied, habitually adopting bipedal locomotion for at least short distances (Vizcaíno and Milne, 2002). Perhaps, then, we can formulate one more prediction:
  1. A digging animal that is less dependent on its forelimbs for non-digging functions is likely to acquire more of the previously listed features
Do alvarezsaurids fit this pattern? They sure do!

In a conversation with science communicator Aron Ra, Viktor Radermacher commented (around 1:16:30 in the linked video), "If you have the arm of Shuvuuia or Mononykus, there's only one thing you can do with it." One might say that that's not strictly true; extant hook-and-pull diggers may also use their forelimbs for walking, burrowing, climbing, fighting, or self defense. However, the core of the matter is this: none of those functions on its own is known (from extant analogues and biomechanical principles) to have favored the specific combination of features that we observe in alvarezsaurid forelimbs. The only behavior I know of that selects for this suite of characteristics is hook-and-pull digging, and I've yet to see any convincing evidence demonstrating that alvarezsaurids were incapable of it.

Restoration of Nemegtonykus here to break up the wall of text, by Scott Reid (used with permission).

If alvarezsaurids were digging, what were they digging into? Extant hook-and-pull diggers are all myrmecophages (eaters of ants and termites) that use their digging skills to open up insect nests so that they can consume the occupants. As a consequence, this behavior has been inferred for alvarezsaurids as well. In an abstract for the Society of Vertebrate Paleontology conference in 2000, Nick Longrich called this concept the "aardraptor" hypothesis.

Besides their adaptations for hook-and-pull digging, alvarezsaurids do seem to have exhibited other traits that are common in myrmecophagous mammals (Longrich and Currie, 2009). These include a toothless gap at the tip of the lower jaw, a large number of small, simple teeth*, long, narrow jaws, weak mandibles, and simplified jaw joints. Additionally, myrmecophages often have a long, sticky tongue for capturing their prey. In alvarezsaurids, the bony hyoid that would have supported the tongue is known in Shuvuuia; it has not been described in detail, but has been said to be "well-developed" (Chiappe et al., 2002). (Unfortunately, Shuvuuia was not included in a recent study on the hyoid morphology of archosaurs and its relation to tongue mobility.)

*Some myrmecophages (such as anteaters and pangolins) have lost teeth altogether. Nonetheless, increase in tooth number is also a recurring trend in myrmecophages: the giant armadillo, the numbat (Myrmecobius fasciatus), and the bat-eared fox (Otocyon megalotis) all have some of the highest tooth counts among land mammals.

Skull of Shuvuuia, from Chiappe et al. (1998). Note the long, tubular snout and well-developed hyoid ("hyo").

Unlike hook-and-pull digging, myrmecophagy is widespread enough among living animals that we need not restrict ourselves to mammals in considering extant analogues. In fact, there are myrmecophagous theropods today: many species of woodpeckers feed primarily on ants (which can comprise 50-95% of their diet), particularly those in the genera Colaptes, Geocolaptes, Jynx, and Picus, as well as some species of Dryocopus and Mulleripicus (Billerman et al., 2020; Winkler and Christie, 2020), and they share some similarities with mammalian myrmecophages. As flying birds, woodpeckers naturally haven't modified their forelimbs into excavation tools, but their head is famously specialized for breaking through hard substrates. They also possess a long, sticky tongue for collecting prey. Myrmecophagous woodpeckers often forage on the ground, on rotting logs, and at the bases of trees, habitats that presumably would have been accessible to alvarezsaurids as well.

An immature European green woodpecker (Picus viridis) feeding on ants using its long tongue, photographed by Luis García, under CC BY-SA 3.0 ES.

What types of social insects might alvarezsaurids have eaten? Longrich and Currie (2009) suggested that their primary targets were wood-nesting termites, traces of which are commonly found in fossilized wood at some localities where alvarezsaurids are known from. Contrary to popular artistic depictions, it appears unlikely that alvarezsaurids fed on mound-building termites, which might not even have existed during the Mesozoic. Even though supposed fossil termite mounds have been reported from Jurassic and even Triassic deposits, these records have been viewed with much skepticism (Krishna et al., 2013; Genise, 2017). Instead, the termite body fossil record and molecular clock estimates suggest that mound-building termites only evolved during the Cenozoic.

Longrich and Currie also argued that Cretaceous ants probably weren't abundant enough to have been the main prey of alvarezsaurids, noting that ants only account for a small fraction of species in Cretaceous insect assemblages (up to 1.2% according to LaPolla et al., 2013). Although body fossils of Cretaceous termites account for similarly small fractions despite the presence of their trace fossils (1% or less according to Engel et al., 2009), Longrich and Currie suggested that their lower preservation potential could be attributed to the "cryptic lifestyle" of wood-nesting termites, likely referencing the fact that such species may spend all or most of their time inside their nests. In any event, the relatively limited insect fossil record seems to suggest that early termites were an available food source to alvarezsaurids, maybe supplemented by the possibly rarer ants.

Pacific dampwood termites (Zootermopsis angusticollis), photographed by Joe Kunkel, under CC BY-NC 4.0. These are an extant type of wood-nesting termite. Did alvarezsaurids feed on similar species during the Cretaceous? Maybe...

As with the pickaxe model, the aardraptor hypothesis has its critics. Agnolin et al. (2012) took exception to the evidence from cranial morphology that Longrich and Currie presented, pointing out that all of those characteristics are widespread in armadillos (only some of which are specialized myrmecophages). Regardless, the fact remains that the skull features mentioned have evolved independently in many myrmecophagous groups. Thus, the observable cranial anatomy of alvarezsaurids is at least consistent with a myrmecophagous diet.

A common concern I've seen directed towards the aardraptor hypothesis again centers around the short forelimbs of alvarezsaurids, namely that process of breaching a insect nest would put the rest of an alvarezsaurid's body too close to the bites and stings of its potential prey for much comfort. However, I have yet to come across any evidence that forelimb length in myrmecophages plays an important role in minimizing counterattacks from their prey, except possibly in anteaters (Navarrete and Ortega, 2011), and even they frequently take numerous bites and stings. Instead, both myrmecophagous mammals (Reiss, 2000) and woodpeckers (Winkler and Christie, 2020) tend to have thick, tough skin that allows them to tolerate retribution from the insects, at least long enough for a worthwhile feeding bout. Although most myrmecophages do eventually retreat from the onslaught, the skin of the aardvark is said to be such an effective defense that it is virtually unfazed by the bites of most termites (Kingdon et al., 2013). In addition, myrmecophages usually have behavioral adaptations that reduce their exposure to their prey's arsenal, such as targeting less aggressive insect species and limiting the amount of time spent at each nest (Redford, 1987; Reiss, 2000). Both of these certainly could have been options for alvarezsaurids.

In any case, it's not clear to me how longer forelimbs in myrmecophages are critical protection against biting and stinging insects. Once the insects start mounting a defense, they're likely to attack any point of contact with the nest, no matter how long the appendage it is attached to. And even though myrmecophages generally feed only for short periods at each nest, these feeding bouts often still allow ample time for defending insects to march up beyond the snouts and forelimbs of their predators. Not to mention, what exactly is so vulnerable about the rest of the body that it needs to be unreachable by insects? If one is willingly exposing themselves to a multitude of bites and stings, a few more on the back are unlikely to be more painful than, say, on the nose, which already needs to be put in close contact with the nest to begin with.

At most, longer limbs might limit the number of insects that can attack at once by placing a smaller surface area in contact with the nest, but this evidently isn't essential to most myrmecophages. In the video below, a feeding short-beaked echidna (Tachyglossus aculeatus) is shown climbing all over a termite mound with its belly held very close to the substrate, and at one point it shoves almost its entire head into the hole in the mound that it has dug out. Not exactly ideal actions to take if minimal contact with the nest was a top priority.

In this other video, a black-bellied pangolin (Phataginus tetradactylus) is shown eating ants. The ants climb up its body via its snout and limbs pretty quickly, but this doesn't prevent it from continuing to feed. Is this because its relatively long forelimbs (compared to those of an alvarezsaurid) bought it significantly more feeding time? I doubt it. (As an aside, notice how other than the strong flexion of the fingers, only a fairly small degree of motion by the rest of the forelimb is needed for the pangolin to tear off pieces of wood.)

Another video on the same channel also shows how quickly ants can start climbing all over a feeding pangolin without deterring it. If anything, the pangolin in that video (understandably) looks more irritated by the ants on its face, which would have been there irrespective of its forelimb length.

Though one might be tempted to attribute these feeding patterns to the protection offered by the pangolins' scales, the scales are more suited to defense against predators than against insects (Challender et al., 2020), and the more heavily-armored ground pangolin is not immune to counterattacks from ants (Swart et al., 1999). Besides, can we really exclude the possibility that alvarezsaurids had their own integumentary specializations that served as armor against insect weaponry, perhaps akin to the dense, scale-like feathering on the faces of honey buzzards (Sievwright and Higuchi, 2016)? We cannot.

Furthermore, there is no particular reason to assume that if alvarezsaurids were myrmecophagous, they must have kept their forelimbs and chest constantly pressed up against the substrate while they were feeding. It is at least as likely that they stepped back once they had broken into a nest and then brought only their narrow snout in contact with defending insects. Even some extant myrmecophages, like the Indian pangolin (Manis crassicaudata), normally cease digging once they have exposed their prey (Challender et al., 2020).

Later alvarezsaurids not only had unusual forelimbs, but highly specialized hindlimbs as well, and it has been proposed that this, too, poses a problem for the aardraptor hypothesis. Their hindlimb anatomy indicates that they were likely well adapted to cursoriality (running), having very long lower legs and feet relative to their body size. Their feet also exhibit a very extreme form of the arctometatarsalian condition, in which the second and fourth metatarsals (long bones of the foot) are tightly appressed to each other close to the base of the foot, "pinching" the third metatarsal between them. An arctometatarsus evolved independently in many groups of Mesozoic theropods, and may have been advantageous for agile locomotion (Snively and Russell, 2003).

The foot of Linhenykus, in front (left) and back (right) views, from Xu et al. (2013). Note how the two metatarsals on either side are tightly appressed to each other for much of the foot's length.

These cursorial adaptations may seem at odds with myrmecophagy, because many myrmecophagous mammals have unusually low metabolic rates, which are unlikely to support a highly cursorial lifestyle. Myrmecophages have the potential to attain much larger body sizes than other insectivorous animals, because ants and termites tend to live at high concentrations. However, even though ants and termites are rich in protein, much of this protein is locked up inside their hard-to-digest chitin exoskeletons, and it appears that this may require large myrmecophages to maintain low metabolic rates.

A direct investigation of whether myrmecophagy could support alvarezsaurids would be highly theoretical at best, given that we'd need to obtain values for many variables we currently lack solid knowledge of, such as the productivity of ecosystems in which alvarezsaurids lived, insect abundance in those same communities, and the energy budgets of alvarezsaurids. Even so, maybe extant myrmecophages can provide us with clues about how large a myrmecophage can get without substantially reducing its metabolism.

Although early studies suggested that only mammals less than 1 kg could live off a mainly myrmecophagous diet while maintaining a typical mammalian metabolism (McNab, 1984), more recent research has upped this threshold to 11-13 kg (McNab, 2000a), which happens to be close to the maximum size of the aardwolf (Proteles cristata; Hunter, 2019), a myrmecophagous hyena. The aardwolf's metabolic rate was once considered to be abnormally low, which was interpreted as evidence that myrmecophagy promotes low metabolism at its size. Subsequent studies though have found that it has a metabolism typical for a mammal of its size (McNab, 2000b) and not significantly lower than expected of a carnivoran (Cooper and Withers, 2006). Among myrmecophagous woodpeckers, even the very specialized ground woodpecker (Geocolaptes olivaceus) does not have a reduced metabolism for a bird of its size (Kemp et al., 2017), but this is perhaps unsurprising given that even the largest woodpeckers are well below 1 kg in mass.

An aardwolf, photographed by Dominik Käuferle, under CC BY-SA 3.0. Is this small hyena the largest an myrmecophage can get while maintaining a standard mammalian metabolism?

Interestingly, a threshold of 11-13 kg would exceed the body size of nearly all known alvarezsaurids (Benson et al., 2018), other than the emu-sized Patagonykus and Bonapartenykus. Although the forelimb of Patagonykus indicates that it had many of the digging-related traits mentioned previously, the hindlimbs of these larger taxa are incompletely known and their skulls remain undiscovered, so there exists a real possibility that they did not share the same ecology or physiology as the smaller alvarezsaurids. One of the most curious trends known in alvarezsaurid evolution is their tendency towards smaller body size later in their history, being among the few non-avialan dinosaurs known to have reached sizes smaller than 1 kg. Might this have been an adaptation to specialize in myrmecophagy without compromising metabolic rate? It would be hard to test, but it's not an unreasonable hypothesis.

It has also been suggested that the cursorial adaptations of alvarezsaurids could have been advantageous for a myrmecophagous lifestyle (Xu et al., 2010). Being speedy runners likely would have helped such small dinosaurs escape from predators, but cursorial limb proportions can allow for increased energy savings during locomotion as well (as has been recently proposed for large theropods by Dececchi et al., 2020). Due to the often patchy distribution of their prey, myrmecophages may need to travel long distances while foraging, and an efficient means of getting around certainly would have been conducive to doing so.

Even so, it's strange that there are no cursorial myrmecophages alive today, isn't it? Well... there are, actually: sengis!

Sengis are a group of small African mammals that are sometimes called "elephant shrews", though they are more closely related to elephants than to shrews. Many sengi species feed primarily on ants and termites, which they lick up with a long, sticky tongue (Rathbun, 2009; Kingdon et al., 2013). They are also the most cursorially specialized mammals of their size (Lovegrove and Mowoe, 2014), able to run very quickly by bounding in an antelope-like manner. The largest sengis are about the same size as some alvarezsaurids. In fact, at over 700 g (Rovero et al., 2008), they are more than four times the size of the smallest known alvarezsaurids!

A black and rufous sengi (Rhynchocyon petersi), photographed by Joey Makalintal, under CC BY 2.0.

Sengis provide some interesting insights into what it takes to be a cursorial insectivore. To sustain their cursorial lifestyle, sengis have higher body temperatures (averaging 37.2°C) than their closest living relatives, the tenrecs and golden moles (averaging 32.8°C) (Lovegrove and Mowoe, 2014). As it happens, the lower body temperatures of the latter groups are close to some of the body temperatures that have been estimated for non-avialan maniraptorans through eggshell geochemistry (Eagle et al., 2015; Dawson et al., 2020). Did alvarezsaurids exhibit elevated body temperatures compared to other non-avialan maniraptorans as well? It would be interesting to find out; we do have possible alvarezsaurid eggshell fossils (Agnolin et al., 2012; Yang et al., 2018).

"Wait, you combine cursoriality and insectivory at small body size? You're me!"
"... Well, I also have armadillo arms."
(Sengi head from a photograph by Joey Makalintal, figured previously, and Shuvuuia skull from Chiappe and Dyke, 2002.)

The parallels between sengis and potentially myrmecophagous alvarezsaurids are not perfect, I'll grant. The giant sengis (Rhynchocyon spp.) that overlap in size with alvarezsaurids are less myrmecophagous than smaller sengis, instead feeding on invertebrates largely according to their availability (Rathbun, 2009; Kingdon et al., 2013). Then again, sengis lack any means of penetrating hard insect nests, and therefore would likely have trouble being specialized myrmecophages in environments where ants and termites are less exposed, such as the forested habitats where Rhynchocyon live.

Maybe this is another discrepancy that can be explained by the bipedal ancestry of alvarezsaurids. As described before, limbs suited to digging are typically short and robust, unlike the long, gracile limbs demanded by running. As a result, it would probably be difficult for a quadruped to specialize in both activities. I can think of two more extant myrmecophages that descended from relatively cursorial ancestors: the aardwolf and the bat-eared fox. Both of them have retained similar limb proportions to their close relatives, but are also unable to break into hard insect nests. The aardwolf instead licks up termites from the surface of the ground, whereas the bat-eared fox scratches them up from the soil. In an obligate biped, however, the function of the front and hind limbs do not need to be interlinked, and it may thus have the potential to become proficient at both digging and running, as alvarezsaurids might have done.

I'm not saying that the existence of sengis or the decoupling of front and hind limb function in bipeds shows that alvarezsaurids must have been myrmecophages. It's easy to come up with plausible-sounding evolutionary scenarios and analogies, but it's much more difficult to test them. However, I do think that these factors indicate that the aardraptor hypothesis has yet to be falsified.

A bat-eared fox, photographed by Yathin S Krishnappa, under CC BY-SA 3.0. This species feeds mostly on termites, but it retains the limb proportions of a typical fox and cannot dig into hard termite mounds.

An alternative to the aardraptor hypothesis was put forth in the scientific literature by Lü Junchang and colleagues in their 2018 description of Qiupanykus. They hypothesized that alvarezsaurids might have fed on the eggs of other dinosaurs. I previously commented that their reasoning for this was "extremely unconvincing" (bold text included), and I stand by that. Simple association between an alvarezsaurid skeleton and an eggshell fragment does not make for direct evidence of alvarezsaurid dietary habits.

With that said... the concept of egg-eating alvarezsaurids may not be so unfathomable. Very few tetrapods today specialize in eating eggs, but eggs could have been a very abundant food source during the Mesozoic, during which most large terrestrial animals were probably egg layers. In addition, if recent estimates of months-long incubation periods for some Mesozoic dinosaur eggs are correct (Erickson et al., 2017), eggs may have been available for much of the year in many Mesozoic ecosystems. The powerful forelimbs of alvarezsaurids could have conceivably been used to break into eggs, as Lü et al. noted, and eating eggs might not have required a strong bite or large teeth, consistent with the delicately-built alvarezsaurid skull. I will say though that a long, narrow snout does not seem to me like a particularly appropriate tool for efficiently collecting the contents of a smashed egg. All in all, I still favor alvarezsaurid myrmecophagy on the basis of there being (more) suitable extant analogues, but I'd be interested in seeing the egg specialist hypothesis tested further in future studies.

At the very least, we probably can't rule out opportunistic egg eating in alvarezsaurids, even if they were primarily myrmecophagous. Many myrmecophages do feed almost entirely on social insects, but some regularly expand their diet (Redford, 1987) to include other types of invertebrates, small vertebrates, carrion, or fruit, as has been documented in the giant armadillo (Carter et al., 2016), the sloth bear (Melursus ursinus; Hunter, 2019), and the northern flicker (Colaptes auratus; Billerman et al., 2020), just to name a few examples.

An alvarezsaurid devours a juvenile sauropod, by Scott Reid (used with permission). Even if alvarezsaurids were myrmecophagous, might they have occasionally expanded their palates to include larger prey items like this? It's possible.

So what's my (non-expert) take on alvarezsaurid paleobiology? I consider the pickaxe model to be the most likely explanation of their forelimb anatomy, and a myrmecophagous diet at least seems plausible. In a discussion on the Dinosaur Mailing List, David Marjanović** remarked, "To me, the unique combination of traits in alvarezsaurs isn't more mysterious than Coombs's chimera [referencing a description of sauropods by Walter Coombs]." I agree. Both sauropods and alvarezsaurids had puzzling body plans not seen in any living animal, but that does not mean we cannot paint a broad picture of how they likely lived and functioned if we employ rigorous comparative anatomy, biomechanics, and multiple extant analogues.

**I should mention that Marjanović independently came up with many of the same points that I've brought up here, years before this blog even existed!

Is it possible we'll conclude one day that alvarezsaurids were not specialized for digging or myrmecophagy after all? Of course! However, I think we first need good reason to discard what seems to me like the most straightforward possibilities: that a forelimb that looks like that of a digging animal was used for digging, and that an animal with this specific type of digging limb was likely eating social insects. And though alvarezsaurids exhibited an odd combination of features for an animal with such a lifestyle, I have not yet encountered any compelling evidence for why this would be demonstrably inconsistent with the aardraptor hypothesis, especially when considered within their evolutionary context.

Does this mean that the behavior depicted in my profile picture is still plausible? We shall see...

What further studies could be applied to test our ideas about alvarezsaurid ecology? Plenty! More detailed descriptions of their skulls and teeth could reveal hints about how and what they ate, especially if relevant specimens could be CT-scanned. Perhaps direct evidence of their diet could be obtained through isotopic signatures in their teeth. Quantitative biomechanics might also help us narrow down what types of behaviors they could perform with their jaws and forelimbs. We still have much to learn about alvarezsaurids, some of the most delightfully bizarre of all dinosaurs.