I grew up in the shadow of the Bakkerian Renaissance, a time when dinosaurs were no longer considered to be giant versions of modern lizards, but actually much more mammal-like in their behavior and physiology. Also, bird origins became a hot topic, and John Ostrom's 1969 discovery of Deinonychus seemed to seal the "live fast" lifestyle that Bob Bakker dreamed for the Dinosauria. However, in Ostrom and Bakker's minds, the avian anatomy of Deinonychus, plus the fact that a Deinonychus/Tenantosaurus bonebed existed, pointed to a pack-hunting lifestyle for the newly described theropod. This view became quickly accepted by scientists and the mainstream press (and thus, the public) with barely any criticism. So in 1993, when Jurassic Park featured "Velociraptor" (actually Deinonychus) hunting in a group, nobody blinked an eye.
This "pack-hunting" behavior in one theropod, however, began to influence others. Bonebeds of Albertosaurus, Daspletosaurus, and Allosaurus have been interpreted as death sites of a "pack" or "group" of animals. Even the enormous bonebed of Falcarius has been seen as an enormous group of animals, traveling together. While perhaps not hunting as a group, these examples show cooperative and voluntary gatherings of these animals. The Deinonychus mindset has become so ingrained these days that any multi-individual theropod dig site is immediately thought of as a potential "pack" site.
A new paper by Roach & Brinkman (2007) questions this long-held dogma. The authors begin by point out that true cooperative pack-hunting behavior is incredibly rare among any animal. The only two examples are wolves and African hunting dogs, who cooperate on a hunt thanks to their highly social lifestyles. Think about what they're saying here: Only one living lineage of one major family of one group of an incredibly diverse amniote sample cooperatively hunts in packs. Well, how does that fare for Deinonychus' relatives? Poorly, it turns out. Crocodilians will "cooperate" to bring down large game in Africa, but after the carcass is dragged onto the shore, it's every croc for himself. Same with vultures, who will descend en masse onto a carcass, but it's every bird for himself. There are rare examples of raptorial birds hunting in mated pairs, but nowhere else in the (extant) amniote lineage do two or more members of the same species hunt in a cooperative manner.
In fact, fossil evidence may point to a new modern analogue for theropod hunting behavior: the Komodo Dragon. These largest living squamates live on a few islands in Indonesia and exhibit rather brutal feeding strategies. First of all, they hunt alone. Second, a single dragon can kill an animal much larger than itself by inflicting a deadly wound, then backing off until the prey drops from shock or blood loss (or both). At that point, drawn by the smell of a fresh kill, dragons from all over converge on the corpse, fighting amongst themselves for chunks of meat. Dragons will scratch and bite each other on the head while feasting, and amazingly often, kill each other. The dead dragon becomes a new feeding frenzy among the ravenous group.
The implications here for, chiefly, Deinonychus, are huge. The most common prey of the big dromaeosaur, according to shed teeth around corpses, is still Tenantosaurus. Amazingly, Deinonychus seems to have a size preference for its favored dinner--one that pretty much hits the favored komodo dragon prey size ratio right on the head. With its large recurved claws and teeth, it's doubtful that an individual Deinonychus would have much trouble killing Tenantosaurus, an ornithopod with no visible defense mechanism. In all liklihood, the dromaeosaur probably backed off upon inflicting a major wound--risking injury from an angered ornithopod that weighs several times its attacker's size would be silly.
After the tenantosaur died of wounds sustained, or at least became immobile due to shock, the killer would, presumably, begin feasting. Theropods having the nasal cavities that they do, it's doubtless that other resident deinonychi would converge on the scene, at which time a Komodo dragon-like battle royale may take place. The fossil evidence for such a scenario is downright strong, according to Roach & Brinkman.
First, that famous first kill site that Ostrom found (YMP 64-75), where the remains of a Tenantosaurus is surrounded by three Deinonychus individuals is a lot more like a Komodo kill site than anything else. Whereas the majority of the ornithopod is gone, the three Deinonychus individuals consist almost entirely of feet and tails. These are areas on a theropod's body where virtually no meat would be found. However, like Komodo dragons, it's probably not unreasonable to assume that an individual Deinonychus was capable of swallowing sections of bone whole, thus the incompleteness of the three raptors and the near disappearance of the ornithopod (by the by, the only real piece of the Tenantosaurus that's left is the tail). If we are to buy the dragon analogue, the dead individuals would be those killed by conspecifics and subsequently cannibalized. Interestingly, among Komodo dragons, a strict social heirarchy based chiefly on size is enforced. The largest individuals are also the most dangerous killers, and frequently attack and kill other Komodos with barely a second thought. Immature dragons stay far from any kill site until the larger adults are gone completely, as younger dragons are generally the first to go.
This may explain the lack of young theropods at fossil "kill sites." An immature animal would be less likely to fossilize (and in some cases more likely to be eaten entirely) than adults. Youngsters would be wise to stay away from kill sites lest they be consumed by their older cousins!
Another kill site (OMNH V706), interpreted as a mass death via flood, shows several tenantosaur specimens in various stages of disarticulation surrounding (more or less) a single subadult Deinonychus specimen. Roach & Brinkman interpret the lack of more cannibalized dromaeosaurs to the fact that the amount of tenantosaur meat was far more abundant.
Other non-dromaeosaur theropod kill sites provide more evidence toward a frenzied feeding strategy. The famous Cleveland-Lloyd Quarry, in which allosaurs outnumber the herbivorous dinosaurs 3:1, is usually attributed as a "muddy death trap" where "packs" of allosaurs entered the pits to attack herbivores that were temporarily trapped but perhaps more able to eventually escape. Interestingly, 82% of the allosaurs represented juveniles and subadults. In Roach & Brinkman's view, all the evidence points toward a "feeding frenzy" strategy, whereupon perhaps dozens of individual allosaurs, drawn to the stench of rotting meat after a mire had dried up, converged upon the mass of corpses, only to fight amongst themselves for a spot at the table. This intraspecific aggression resulted in the mortality of several juvenile and subadult individuals, just as in Komodo dragon kill sites today.
Another famous quarry, Dry Messa in Alberta, consists of several dozen skeletons of Albertosaurus sarcophagus which seemingly died en masse. The bonebed has been used to calculate tyrannosaur survival rates by Erickson et al., who concluded that, given the near absence of juvenile tyrannosaurs in the bonebed (and in the fossil record in general), tyrannosaurs must enjoy a fantastic survival rate until adulthood, at which point their chances of survival drop every year they live past physical maturity. However, Erickson et al. forget that their survival rates are based on fossilized animals, which by definition are exceptions to the rule. Roach & Brinkman suggest, logically, that the absence of juvenile animals in the fossil record may be thanks to 1) a preservation bias against smaller animals with less ossified skeletons, and 2) the theory that the juvenile mortality rate may actually be incredibly high, but that instead of being fossilized, juvenile animals are actually being consumed by adults.
The final bonebed I will go into is the Two Medicine Formation in Montana. At that site, at least three individual Daspletosaurus are intermingled with at least five hadrosaurs. Currie et al. of course concluded that the site provided evidence of gregariousness in tyrannosaurs and that the big theropods may have adopted cooperative attack strategies to "break through the defenses" of herds of ceratopsians and duckbills. Interestingly, the authors fail to cite a reason for the presence of the three Daspletosaurus individuals at the kill site. It's hard for me to think that a duckbill could actually kill a tyrannosaur. Applying Roach & Brinkman's Komodo analogue, and everything makes sense. Just like at the Cleveland-Lloyd Quarry, a small group of dead hadrosaurs drew a large group of tyrannosaurs, who fought and killed each other in an effort to stuff their bellies. It's quite telling, in fact, that the daspletosaur skeletons were disarticulated and surrounded with "dozens" of shed conspecific teeth.
There's even trackway and behavioral evidence for Komodo-like feeding behavior, but this post is already starting to run long, so I'll just give you the paper citation. It's really fascinating, and I actually have a PDF of it, so if anyone wants it, feel free to email me. But what's most interesting, perhaps, is that the Bakkerian Rennaisance pulled dinosaurs more toward the mammalian ideal, but what we're finding is that while these great archosaurs may have looked like birds, their behavior was more more like that of big lizards.