Anatomy informs geology: Hydrodynamic dispersal of alligator bones, with implications for taphonomic interpretations of fossil deposits of crocodylians, dinosaurs, and other morphologically novel taxa

Distinctive anatomical features of bones can influence not only how these structures perform in living animals but also the tendency of elements to be transported by flowing water after death. Such transport can be critical in the concentration of fossils from animals that live near freshwater habitats, providing important context for interpreting the composition of paleocommunities. Measurements of the tendency of flowing water to disperse skeletal elements have been collected for diverse taxa, including mammals, turtles, and birds. However, these extant models may not be entirely appropriate for many morphologically distinct extinct lineages, such as non‐avian dinosaurs. To expand the range of models available for evaluating the influence of hydrodynamic transport on the assembly of fossil deposits, we used a flow tank to measure the water speeds that disperse bones from a subadult American alligator (Alligator mississippiensis), with the skull and mandible tested in multiple starting orientations. Alligator bones are sorted into three main dispersal groups: early (vertebrae, most girdle elements), intermediate (ribs, most limb bones), and late (pubis, femur), with the skull and mandible varying between intermediate and late depending on orientation. Late dispersing elements tended to be heavy or very flat. These results can refine interpretations of the taphonomic context for deposits of fossil crocodylians and morphologically similar taxa (e.g., choristoderes, phytosaurs) and provide an additional comparative model for deposits of non‐avian dinosaurs. Moreover, variation in hydrodynamic sorting across lineages highlights how distinctive anatomical features can influence the concentration of fossils, shaping understanding of assemblage composition and paleofaunal evolution.


| INTRODUCTION
Anatomical diversity is compelling. The varied shapes and sizes of organisms-"endless forms most beautiful" (Darwin, 1872)-inspire widespread wonder and have led generations of scientists to search for explanatory mechanisms. Paleontology opens a wider window on this diversity, revealing fossil species with forms and structures unknown in the modern world. Much attention has been devoted to interpreting how such structures may have been used in life (Alexander, 1989;Hutchinson, 2012;McInroe et al., 2016;Witmer, 1995). However, distinctive anatomy can remain important to the understanding of organisms even after they have died. The rock beds from which fossils are recovered can help indicate the environments in which species once lived (Behrensmeyer & Hook, 1992). But as the remains of organisms decompose into separate elements and transform into fossils they can become sedimentary particles themselves, subject to transport by factors ranging from floods to scavengers (Behrensmeyer, 1975;Behrensmeyer, 1991;Dodson, 1973;Voorhies, 1969;Weigelt, 1989). Such transport could dramatically influence the composition of fossil assemblages and, thereby, reconstructions of paleocommunities (Blob & Fiorillo, 1996;Hanson, 1980;Shotwell, 1955;Wolff, 1973). For example, an accumulation of elements with a low tendency to disperse might reflect species buried where they lived, whereas an accumulation of elements with a high tendency to disperse might indicate species whose remains were transported to a site from a different life habitat (Fiorillo, 1991;Smith, 1993;Wolff, 1973). In systems influenced by flowing water, particle transport can depend heavily on particle shape (Sneed & Folk, 1958), and studies of vertebrate skeletal remains associated with fluvial systems have highlighted how the anatomical features of skeletal elements can influence their susceptibility to hydrodynamic transport (Voorhies, 1969;Dodson, 1973;Behrensmeyer, 1975Behrensmeyer, , 1988Blob, 1997;Trapani, 1998;Peterson & Bigalke, 2013). Thus, by influencing the potential of an element to be moved by flowing water, anatomical features of skeletal elements can also influence our ability to reconstruct the environments in which species lived, as well as our ability to understand the composition of ancient animal communities.
Early efforts to evaluate the impact of hydrodynamic transport on the assembly of vertebrate fossil deposits focused mainly on how dispersal tendencies varied across the different elements of the skeleton. Using a flume to measure the strength of currents required to move sheep and coyote bones, Voorhies (1969) found that bones from these species sorted in a predictable order, with ribs and vertebrae belonging to an early dispersing Group I, the pelvis and most long bones of the limbs comprising a later dispersing Group II, and the skull and mandible dispersing last as a final Group III. Behrensmeyer (1975) recognized these three broad categories as "Voorhies Groups" and found similar patterns to hold for a wide diversity of medium to large mammals, ranging from pigs and hippos to antelope and zebras. However, studies of a wider diversity of species showed results that diverged from these patterns. For example, in flume trials with mouse bones, Dodson (1973) found that the mandible was an early (rather than late) dispersing element; moreover, mouse humeri dispersed much later than those from toads and frogs. In further examples of patterns differing from the Voorhies Groups of large mammals, the skull dispersed before vertebrae in human skeletons subjected to flow (Boaz & Behrensmeyer, 1976), and was also found to be an early dispersing element in pigeons (Trapani, 1998) and, in some initial orientations, softshell turtles (Blob, 1997). An implication from these comparisons is that, while many species show characteristic gradations from early to late sorting elements, these patterns are often taxon specific. Such specificity is even more likely for taxa with distinctive anatomical elements or features like the shell bones of turtles (Blob, 1997), or the furculum of birds (Trapani, 1998).
The fossil record of vertebrates includes many lineages with distinctive skeletal features that may not readily conform to previous taphonomic models of hydrodynamic dispersal. Dinosaurs are a prominent example, including taxa with an enormous range of body sizes (Wilson & Carrano, 1999;Carrano, 2006) and remarkable structures (Farlow & Dodson, 1974;Hopson, 1975;Main et al., 2005;Molnar, 1977;Padian & Horner, 2011). Though concerns have been raised about uniformly applying mammalian Voorhies Groups to dinosaurs (Dodson, 1971;Blob, 1997;Gates, 2005;Peterson & Bigalke, 2013), limited data are available for considering the hydrodynamic sorting tendencies of dinosaur bones. Flume trials like those applied to the bones of modern species (Blob, 1997;Dodson, 1973;Trapani, 1998;Voorhies, 1969) are challenging to apply directly to dinosaur specimens given the diagenetic transformation of their elements into fossils, which can be fragile and, made of stone, might transport in different ways than the original bone (Peterson & Bigalke, 2013). One study of pachycephalosaur skulls conducted flume trials using resin casts to compare dispersal tendencies across species and between complete skulls versus separate frontoparietal domes, which are often found as isolated elements (Peterson & Bigalke, 2013). These trials had the advantage of standardizing the composition of items that were tested, allowing evaluations of the impact of different shapes on hydrodynamic sorting; however, they were focused on a very specific set of elements. Birds, as living dinosaurs (Makovicky & Zanno, 2011;Smith et al., 2015), might provide an alternative model for predicting hydrodynamic sorting patterns for the bones of non-avian dinosaurs, and data already exist for extant pigeons (Trapani, 1998). However, the derived anatomy of birds related to flight (Heers, 2016;Heers & Dial, 2015) as well as the small body size of most living species (Dunning, 1993), make it difficult for birds to provide straightforward models for skeletal dispersal in most non-avian dinosaurs. In the context of these limited options for evaluating how hydrodynamic sorting might affect dinosaur bones, additional actualistic models could be helpful for interpreting concentrations of dinosaur fossils.
Among non-avian lineages, crocodylians are the closest relatives to non-avian dinosaurs (Benton & Clark, 1988;Brusatte et al., 2010;Gauthier, 1986;Nesbitt, 2011;Nesbitt et al., 2017) and might provide an alternative model for the hydrodynamic sorting of their skeletal elements. Several aspects of the crocodylian body plan, such as the structure of the pelvis and tail, bear a closer resemblance to the elements of many non-avian dinosaurs than homologous elements from most mammals and birds. Crocodylian skulls, with diapsid fenestration, abundant small teeth, and a small cranial cavity for the brain, might also provide instructive models for the skulls of many non-avian dinosaurs. But beyond applications to dinosaurian fossils, the crocodylian lineage itself has a rich fossil record of numerous taxa from fluvial and lacustrine environments that could have been subjected to flowing water and transport prior to burial (e.g., Brochu, 2001;Erickson, 1976;Iijima et al., 2021;Mannion et al., 2019;Markwick, 1998;Molnar et al., 2015;Rogers II., 2003;Stubbs et al., 2021;Wilberg, 2017). Moreover, a variety of other extinct reptilian lineages common in fluvial and lacustrine deposits, such as choristoderans (Erickson, 1972;Evans & Hecht, 1993;Gao & Fox, 1998;Matsumoto & Evans, 2010;Vandermark et al., 2007) and phytosaurs (Datta et al., 2020;Stocker et al., 2017) exhibit skeletal features of the skull (e.g., elongate snouts) and postcrania (e.g., long, robust tails and short limbs) which more closely resemble those of modern crocodylians than other extant taxa. Although sequences of decomposition and disarticulation have been examined for crocodylians (Beardmore et al., 2012;McClain et al., 2019;Syme & Salisbury, 2014), most have focused on marine examples, and flume studies of bone dispersal have not been conducted. Thus, experimental data on the hydrodynamic sorting of crocodylian bones could be broadly informative for interpreting a diverse range of fossil assemblages.
In this study, we used a laboratory flow tank (Vogel & LaBarbera, 1978) to measure the water speeds required to transport diverse skeletal elements from a representative extant crocodylian, the American alligator (Alligator mississippiensis). With these data, we sought to evaluate whether broad categories of relative dispersal potentials could be established for the bones of crocodylians, and the extent to which the composition of such categories might differ from those of Voorhies Groups derived from bones of medium to large mammals (Behrensmeyer, 1975;Voorhies, 1969). In particular, we predicted that the skulls of alligators might disperse earlier relative to other elements than in ungulate mammals, due to the generally smaller size of highly dense teeth in crocodylians. We also predicted that many girdle elements of alligators might disperse relatively early compared to other elements because they have curved shapes that prevent them from lying flat along a substrate, promoting their instability in flow. By establishing this new framework of hydrodynamic dispersal potentials, we hope ultimately to provide a reference specifically appropriate to apply to fossil accumulations of crocodylians and morphologically similar taxa. In addition, these data will provide an additional option that could inform taphonomic interpretations for fossil deposits for a wider range of morphologically diverse vertebrates, including dinosaurs.

| MATERIALS AND METHODS
Competent velocities (Blob, 1997: the flow velocity at which downstream movement of an element commenced) were measured for bones from specimens of subadult Alligator mississippiensis (Daudin, 1802) from the osteological collections of Clemson University's Bob and Betsy Campbell Museum of Natural History ( Figure 1). Postcranial elements were from the same individual (CUSC 1891); however, because of damage to the skull of this individual, we used a skull from a second individual (CUSC 2369) of nearly identical size (length 22.8 cm, width 14.0 cm), corresponding to an individual of 1.48-1.75 m total length (Dodson, 1975;O'Brien et al., 2019).
To better simulate fresh skeletal elements, all bones were rehydrated prior to testing (Behrensmeyer, 1975;Blob, 1997) for 20 hours. This duration was based on limitations to facility access and is shorter than in some previous studies (e.g., 30 hours for turtles: Blob, 1997). However, Behrensmeyer (1975) has noted that most water absorption occurs during the first 5 minutes in which bones are submerged, and the submergence duration used in this study should allow complete rehydration for bones of the size we tested (all <100 g). The use of isolated, rehydrated specimens in our trials provided the best opportunity to produce results as comparable as possible to a wide range of similar, previous studies of other taxa (Behrensmeyer, 1975;Blob, 1997;Trapani, 1998;Voorhies, 1969), and reflected a portion of what Payne (1965) classified as the "dry" and "remains" stages of decay, when skin and flesh are largely removed and skeletal elements are not necessarily dehydrated from exposure but are considerably disarticulated, raising their likelihood of independent transport. As several studies of decomposition sequences have noted (Beardmore et al., 2012;Dodson, 1973;Payne, 1965;Syme & Salisbury, 2014;Weigelt, 1989), decomposition is a continuous process in which varying portions of carcasses may retain soft tissue and remain at least partially articulated. Hydrodynamic dispersal patterns observed for isolated elements might not apply straightforwardly to fossil deposits comprised of remains that show signs of having been more intact (Haglund & Sorg, 2006). Hydrodynamic dispersal data from partially articulated specimens, or specimens retaining flesh, could provide useful insights for understanding such fossil accumulations. Nonetheless, competent velocity data from rehydrated, isolated alligator elements add to an existing comparative framework and could aid in interpretations of deposits showing concentrations of disarticulated elements (e.g., Datta et al., 2020;Hunt & Downs, 2002;Lucas et al., 2016;Mukherjee & Ray, 2012).
Competent velocities were evaluated in a custom built, recirculating flow tank with a holding capacity of 400 L (Figure 2), built following designs described by Vogel and LaBarbera (1978). To generate a continuous range of flow speeds, two boat propellers connected to a drive shaft were powered by a calibrated, dial-controlled, variable speed electric motor. The tank had a working section 90 cm long by 33 cm wide, fitted with glass for viewing on one side and the bottom, and with collimators F I G U R E 2 Each bone was tested four times in a calibrated flow tank at Clemson University (example of the alligator mandible indicated by the red arrow), with the skull and mandible tested in multiple orientations. The flow speed at which the bone moved (i.e., its competent velocity) was recorded for each trial. Average competent velocity was calculated for each set of four trials of a bone in a specific orientation and used for comparisons of hydrodynamic sorting across all of the elements to improve laminarity of the flow at both ends of the working section. Trials were conducted with water filling the tank to depth of 20 cm in the working section, much deeper than the height of the tallest element we tested (the skull, 51.35 mm). Although the glass surface of the flow tank bottom is not directly comparable to natural substrates, this surface provided a repeatable arena for generating internally consistent results that could still reflect relative dispersal tendencies for the bones we compared (Blob, 1997).
Each bone was tested four times for at least one stable orientation relative to the direction of flow (i.e., an orientation that did not rock or tip prior to the start of flow). Limb bones were oriented with their distal end upstream, the dorsal rib with its head upstream, dorsal vertebrae with the distal, convex articular condyle contacting the substrate in a triangle of support with the posterior zygapophyses, and the caudal vertebra lying on its side. For girdle elements, the external surface was placed in contact with the substrate, and articular facets were directed upstream. Some elements (e.g., the skull and mandible) were tested in more than one initial orientation, to evaluate the effects of different starting positions on dispersal potential. For each trial, bones were placed with their upstream end at a consistent, marked point near the midpoint of the working area, and aligned along the midline of the tank with a long axis parallel to the direction of flow. Flow speed was then gradually increased until the element moved consistently downstream. Competent velocities measured from all four trials were averaged to calculate a mean and standard deviation for the element in a specific orientation, which then was incorporated into comparisons of hydrodynamic sorting potential across all of the elements.
After the completion of all trials, the heights of tested bones were measured with digital calipers and wet weights were measured on a counterweight balance. Height was measured as the greatest distance of the element from the substrate as it rested along the bottom of the flow tank. These measurements were used to calculate the strengths of alternative explanatory linear models for variation in competent velocity as a function of element height, element mass, both element height and mass, and both of these variables plus their interaction. Model selection was performed using Akaike's information criterion (AIC), with all calculations performed in R (http://www.R-project.org). The distributions of height and mass were significantly right skewed, and thus log 10 transformed prior to analysis.

| RESULTS
Competent velocities ranged from averages of just over 10 cm/s to nearly 50 cm/s across the 21 different cases we tested: four orientations for the skull, two orientations for the mandible, and single orientations for each of 15 additional elements (Table 1, Figure 3). There is substantial continuity and overlap in competent velocities across this range, with an average difference between consecutively ranking elements of only 1.86 cm/s. However, two major gaps of more than 5 cm/s (i.e., > two times the average difference) stand out: the first between the coracoid and ulna, and the second between the radius and mandible (oriented so that the symphysis pointed upstream). Nonparametric Mann-Whitney U tests between the competent velocities for each of these element pairs show that both of these gaps represent significant differences (p < 0.02 for both comparisons). For heuristic purposes, these gaps provide useful demarcations between groups of elements with broadly different hydrodynamic sorting tendencies, analogous to the original descriptions of Voorhies Groups I, II, and III for medium to large modern mammals (Behrensmeyer, 1975;Voorhies, 1969). For bones tested in single orientations, early dispersing ("Group I") bones for alligators include several girdle elements such as the ilium, scapula, ischium, and coracoid, as well as dorsal and caudal vertebrae. More resistant elements ("Group II") for alligators include the dorsal rib and most limb bones (ulna, tibia, humerus, fibula, and radius). The most resistant ("Group III") elements for alligators include the pubis and femur. Competent velocities for most of these elements were highly repeatable, with coefficients of variation under 10% (adjusted for the small sample size (N) of four trials per element: CV* = [1+ (1/4 N)] CV; Sokal & Rohlf, 1995). A few elements (ilium, fibula, and femur) showed CV* values over 15% for single orientations (Table S1), but the magnitude of average competent velocity for each of these elements still placed them clearly within Groups I, II, and III, respectively.
Greater variability was found for elements that were tested in multiple orientations. For example, the mandible (Table S1, Figure 3) dispersed earlier when the symphysis pointed downstream (35.29 ± 5.66 cm/s), compared to when it pointed upstream (42.31 ± 1.77 cm/ s). Although this difference narrowly missed the threshold for a significant difference in a Mann-Whitney U test (p = 0.076), this seems largely due to the highly variable competent velocity of the mandible when the symphysis was directed downstream. In this orientation, the retroarticular processes of the lower jaw were directed upstream; however, these processes curve upward off of the substrate, allowing currents to induce rocking of the lower jaw that produced varying levels of instability across trials. Nonetheless, even the lowest competent velocity measured for the mandible in this orientation (29.32 cm/s) was greater than the highest value for any Group I element. The highest value for the mandible in this orientation (41.96 cm/s), however, matched competent velocities observed when the symphysis was directed upstream, a more stable orientation in which the mandible would fall among the late dispersing, Group III elements ( Figure 3). Thus, in terms of classifying the tendency of the mandible to disperse, although a single assignment to Group II versus Group III might not be appropriate, it might still be viewed as one of the later dispersing elements, and at least "not early" or "not Group I." The skull also showed variation in competent velocity across the orientations in which we tested it. By far, the most resistant orientation was when the skull was positioned with the snout upstream and the teeth along the substrate (Table 1, Figure 3). In this position, the skull required the highest competent velocity of any element (48.63 ± 1.77 cm/s) and was clearly among the late dispersing, Group III elements. Non-parametric Mann-Whitney U tests showed that competent velocities for all three other skull orientations were significantly lower than when the snout pointed upstream with the teeth contacting the substrate (p < 0.02 for all comparisons). However, competent velocities for the other three orientations (snout upstream, teeth up; snout downstream, teeth up; and snout downstream, teeth down) did not differ significantly from each other (Mann-Whitney U tests, p > 0.234 for all comparisons), and fell solidly within the intermediate, Group II elements. Noticeably, CV* values were much greater for orientations where the teeth pointed upward, regardless of snout orientation (~10%-11%), compared to orientations where the teeth contacted the substrate (~2%-4%) (Table S1). Thus, similar to the mandible, the alligator skull has the potential to be among the latest dispersing elements, but even with its potential variability it can still be regarded as at least "not early" or "not Group I." Using only the maximum competent velocities of elements that we tested in multiple orientations (thus generating a single average competent velocity value for each of the 17 elements), the model that provided the strongest explanation of variation in competent velocity (i.e., with the lowest AIC) included element height, element mass, and the interaction between these terms (Table 2). In other words, the values of element height that would lead to high competent velocities depended on the mass of the T A B L E 1 Average competent velocities, heights, and masses of alligator skeletal elements evaluated in flow tank trials (N = 4 trials/ element)

Element
Average competent velocity ± 1 SD (cm/s) Height (cm) Mass (g) Note: Data from which averages were calculated are compiled in Table S1. Abbreviation: SD, standard deviation. element. For example, the skull was both the tallest and heaviest element, and had the greatest competent velocity (Table 1). However, across all elements, heavy bones that were also flat (e.g., the femur) tended to have high competent velocities and disperse last. Similarly, the pubis and caudal vertebra were nearly identical in height, but the pubis was more than 30% heavier and dispersed much later (Table 1, Figure 3).

| Comparative patterns of hydrodynamic transport across taxa
The skeletal elements of alligators show clear patterns of variation in their tendencies to be dispersed by flowing water. The presence of distinct discontinuities in the distribution of competent velocities allows a fairly straightforward demarcation of alligator elements into early, intermediate, and late dispersing groups (Figure 3). Although there are some commonalities in the sorting patterns observed in alligators compared to other taxa, the composition of elements in these groups shows several distinct differences between alligators and patterns previously noted for both large (Behrensmeyer, 1975;Voorhies, 1969) and small (Dodson, 1973) mammals, as well as from other taxa in which patterns have been studied such as turtles (Blob, 1997) and pigeons (Trapani, 1998). One consistent pattern across all of the taxa from which hydrodynamic dispersal data have been collected is the early transport of vertebrae from all regions of the spinal column. Beyond this similarity, however, patterns across other elements diverge across taxa. For example, a particular novelty in the patterns observed for alligators is that the disarticulation of the different elements of the pectoral and pelvic girdles allowed data to be collected separately from each bone. The pelves are fused elements in mammals and birds, and the pelvic bones tend to be strongly attached turtles; in these taxa, the pelvis has a complicated shape that projects into the current, promoting instability and, typically, early to intermediate dispersal (Blob, 1997;Dodson, 1973;Trapani, 1998;Voorhies, 1969). However, alligators show a typical pattern for many saurians (Romer, 1956) in which the pelvis is easily disarticulated into its components (ilium, ischium, and pubis), all of which can disperse separately. Once disarticulated, the ilium and ischium are among the earliest dispersing elements, whereas the pubis is one of the latest (Figure 3). With respect to the pectoral girdle, in mammals and turtles this typically forms a single articulated element, with reduction and fusion of the coracoid onto the scapula in mammals, and tight attachment of the elements into a triradiate girdle in turtles (Blob, 1997;Voorhies, 1969); however, the scapula and coracoid are easily separable and can sort independently in birds (Trapani, 1998) and alligators. In birds, these bones sort very differently, with the coracoid dispersing early and the scapula late (Trapani, 1998); in contrast, the scapula and coracoid are both Group I elements subject to early transport in alligators (Figure 3). Elongated bones show moderately different transport patterns across taxa. Ribs tend to be among the earliest dispersing elements in mammals (Voorhies, 1969) and birds (Trapani, 1998), but disperse somewhat later as Group II elements in alligators (Figure 3). The dispersal of limb bones is also not homogenous either within, or across taxa. For example, in contrast to mammals (Dodson, 1973;Voorhies, 1969) and birds (Trapani, 1998), the femur is a later dispersing element in both alligators ( Figure 3) and turtles (Blob, 1997). Multiple factors could contribute to these differences, including specific aspects of limb bone morphology; however, some aspects of limb function may also bear consideration in this context. Both alligators and turtles are hindlimb-dominated in their propulsion, with hindlimb elements that can exceed the sizes of serially homologous forelimb elements Blob et al., 2014;Blob & Biewener, 2001;. In addition, as animals that spend abundant time in water, there may be little cost for these species to maintain relatively robust limb elements, as reflected in the typically high margins of safety against fracture found in both turtles and alligators (Blob et al., 2014;Blob & Biewener, 1999;. In contrast, the limbs of both mammals and birds must commonly support the weight of the body against gravity which, with high metabolic rates and levels of activity, could select against excess bone robusticity as energetically costly, and as reflected in lower margins of safety against bone fracture (Biewener, 1993;Blob et al., 2014). In this context, the elevated resistance to hydrodynamic transport observed in the femora of alligators and turtles might at least partly reflect distinctive aspects of their locomotor function compared to other taxa.
One of the more striking contrasts in sorting patterns observed across taxa in previous studies is the difference in dispersal tendencies for the skull and mandible. In large and small mammals, the mandible is a late dispersing element, while the skull disperses late in large mammals but somewhat earlier in small mammals (Behrensmeyer, 1975;Dodson, 1973;Voorhies, 1969). In contrast, the skull is among the earliest dispersing elements in birds (Trapani, 1998) and, depending on its initial orientation, is also an early (or intermediate) dispersing element in turtles, with the mandible intermediate as well (Blob, 1997). In alligators, the skull and mandible are always at least in the intermediate transport group and, depending on initial orientation, can be among the latest elements to disperse (Figure 3). The flattened, drag-reducing shape of alligator skulls (McHenry et al., 2006;Pierce et al., 2008) likely reduces their susceptibility to transport. Resistance to transport would likely be amplified among even larger individuals than the one tested in this study, as the skull increases in size (Dodson, 1975;O'Brien et al., 2019). However, at the body sizes evaluated in this study, one likely factor contributing to the later dispersal of the skull and mandible in both mammals and alligators is their possession of teeth, which are absent in both birds and turtles. Teeth represent some of the highest density elements of vertebrate skeletons (Korth, 1979), so the presence of teeth likely elevates resistance to transport in elements that contain them (Behrensmeyer, 1975). Teeth can, however, be easily lost during processes of decomposition, disarticulation, and weathering (Behrensmeyer, 1978;Dodson, 1973). Thus, taphonomic interpretations of fossil deposits drawing on data from skulls and mandibles could be refined by evaluating whether teeth were preserved in recovered elements.

| Correlates of hydrodynamic transport tendencies
Our data show a complicated relationship between the morphology of skeletal elements and their susceptibility to hydrodynamic transport. Whereas flat elements might be substantially entrained within lower flow velocities near the substrate, and experience high levels of friction against the substrate that would resist transport (relative to drag that might promote dislodgement), tall elements can project far into the free-stream velocity of flow, with a large surface area exposed to drag that could detach them from the substrate (Blob, 1997;Vogel, 1994). Heavier elements would also be expected to require higher speeds of current flow before being transported. However, our results showed that neither element height nor weight provided strong explanations of transport tendencies on their own. Instead, much of the variation in competent velocities for alligator elements was best explained by the interaction of these two variables (Table 2). In general, light and tall elements are most easily transported, whereas heavy, flat elements are most resistant to transport. Although the physical principles that lead to these tendencies would be expected to hold generally across fluvial environments, which elements fall into each category could vary considerably across species. For example, in alligators, the two flattest elements (with nearly identical heights) were the caudal vertebra and the pubis; however, the pubis was among the latest dispersing elements, but the caudal vertebra, with lower mass, was among the earliest (Table 1). Similarly, the skull was by far the tallest element, but as the heaviest element it was also (in some orientations) among the last to disperse (Table 1). Such patterns do not necessarily apply to other taxa: species with taller pelvic elements or skulls might experience earlier transports of these elements than alligators. In this context, when applying actualistic dispersal data to understand the taphonomy of fossil deposits, it is advantageous for the extant species used as models for fossils to bear as close an anatomical resemblance as possible.
Although competent velocity was highly repeatable for most alligator bones, our tests of elements in multiple orientations showed that single anatomical structures still have the potential to vary considerably in their dispersal, depending on their initial orientation relative to flow. For example, the orientation of the alligator skull that best resisted transport had the snout directed into flow, with the teeth contacting the substrate-an orientation that minimizes the frontal area of the element that encounters flow, while also providing the broadest and most stable base to resist rolling or flipping. The skull dispersed earlier in other orientations, in which either the taller occipital portion of the skull faced upstream and exposed the skull to higher drag, or the roof of the skull formed a less stable contact with the substrate than teeth across the breadth of the palate, or both ( Figure 3). The distinctive anatomy of the crocodylian mandible, with the dorsally curving retroarticular process (Figures 1 and  2), also seems likely to have increased instability and susceptibility to transport when these processes were oriented upstream, compared to when they were directed downstream (Table 1). Other elements might also have the potential to show dispersal variation related to their initial orientation to flow, a possibility that could be addressed with additional testing. Anatomical effects on hydrodynamic instability were also evident in elements we tested only in single orientations, potentially contributing to early dispersal and, in some cases, variability of most girdle elements in alligators (Table 1, Figure 3). The ilium, for instance, tends to settle in an orientation with a single point of substrate contact and curvature that projects the margins of the bone into the flow; other girdle elements, like the scapula, coracoid, and ischium, have curvature or projections from the articular facets of the glenoid or acetabulum that project asymmetrically into flow, potentially increasing the chance for fluid drag to induce torques that rotate the element out of stable contact with the surface.

| Implications of hydrodynamic sorting data from alligators for taphonomic interpretations of fossil deposits
Data on hydrodynamic transport tendencies from the skeleton of another vertebrate lineage with distinct anatomical features provide an opportunity to clarify and refine the interpretation of whether elements recovered from fossil deposits likely came from animals that lived near where they were buried, or whether they were transported into a locality. This perspective could be especially useful for concentrations of fossils from crocodylians or other morphologically similar taxa (e.g., Datta et al., 2020;Rogers II., 2003) and also provides a wider foundation of comparisons for non-mammalian lineages (e.g., Lucas et al., 2010;Lucas et al., 2016;Mukherjee & Ray, 2012), potentially including dinosaurs (Fiorillo, 1991;Rogers, 1990;Ryan et al., 2001). Largely consistent with previous data from medium to large mammals (Behrensmeyer, 1975;Voorhies, 1969), data from alligators indicate for an even wider range of taxa that the presence of skulls at a locality would support a conclusion that deposition occurred near where the animals lived. Moreover, the presence of vertebrae with a wide diversity of other skeletal elements would suggest limited exposure of remains to hydrodynamic dispersal, whereas the presence of vertebrae as the primary or only preserved elements at a locality would suggest a high likelihood that those bones had been transported to the deposit. However, the implications of transport data from alligator girdle and limb material show more contrasts to mammalian models. Whereas the unified mammalian pelvis tended to have intermediate dispersal tendencies (Voorhies, 1969), the individual bones of the alligator pelvis (Romer, 1956) had widely divergent competent velocities, with the ilium and ischium dispersing early but the pubis dispersing late (Table 1; Figure 3). Thus, representation of all three elements, or an overrepresentation of pubes, could provide stronger evidence for preservation near the site of death than models based on mammalian transport data would suggest. Similarly, whereas limb bones tend to exhibit intermediate dispersal in mammals (Voorhies, 1969), in alligators (as in turtles: Blob, 1997) the femur dispersed considerably later than other limb bones. Thus, in non-mammalian taxa, consideration of which specific limb bones are preserved at a site, rather than a general category of limb bones, would provide stronger insight into whether remains were from animals that died near a site or were transported into it. Transport data from the alligator pectoral girdle provide a further novel perspective for non-mammalian taxa. The only previous dispersal data for independent coracoid and scapula elements were from pigeons (Trapani, 1998), which indicated early dispersal for the coracoid, but late dispersal for the scapula. In contrast, both the coracoid and scapula were early dispersing, Group I elements in alligators (Table 1, Figure 3). Because their pectoral anatomy does not exhibit derived specializations for flight, alligators likely provide a more appropriate model for interpreting the recovery of pectoral girdles in deposits of many non-flying reptiles, with isolated preservation of these elements likely the result of transport, but co-occurrence with more resistant elements suggesting less exposure to dispersal.
In several cases, re-evaluation of the taphonomy of fossil concentrations in the context of transport data from alligator bones could generate new insights, or increase confidence in previous interpretations. For example, a range of morphological similarities between crocodylians and choristoderes has led to evaluations of the co-occurrence of these taxa, in order to assess differences in habits or potential exclusion between these groups (Matsumoto & Evans, 2010). However, in many localities, one or the other of these taxa is represented only by vertebral elements (Skutschas & Vitenko, 2017). Such localities might not allow strong inferences about taxonomic co-occurrence, since isolated vertebrae likely arrived at deposits via hydrodynamic transport, and taxa represented solely by vertebrae could likely have lived elsewhere. In another example, Datta et al. (2020) applied mammalian Voorhies groups to a phytosaur bonebed from the Upper Triassic Tiki Formation of India, and concluded that the assemblage was autochthonous based (among other factors) on a relative overabundance of transport resistant elements (e.g., skulls) and a lower-than-expected proportion of more easily transported elements, such as vertebrae. In this framework, it seemed anomalous that ribs, classified as easily transported Group I elements based on mammalian models, were present in greater numbers than expected. However, data from alligators showed ribs as more resistant, Group II elements (Table 1, Figure 3). In this context, the proportion of ribs from this phytosaur bonebed shows even greater support for its interpretation as an autochthonous deposit, as they likely represent a more resistant element, rather than one that is easily transportable.
Applications of crocodylian hydrodynamic sorting patterns may be more appropriate for some dinosaur bonebeds (and taxa) than others. For example, the large size of most sauropod bones, and the relatively small size of sauropod skulls compared to their other elements, would suggest limited applicability of the transport data collected from alligators in this study. However, Dodson (1971) noted a curious pattern of dinosaur element preservation in two Upper Cretaceous quarries from Dinosaur Provincial Park in Alberta (62 and 111), in which ribs were found associated with skull elements. Based on mammalian models of hydrodynamic sorting (Voorhies, 1969), the association of early dispersing ribs and late dispersing skulls seemed surprising. Although any disarticulation of a skull might make it less resistant to transport and lead to competent velocities closer to those of ribs, the frequently similar dispersal of ribs and skulls in alligators indicates that cooccurrence of these elements may be less unusual in some taxa than mammalian patterns would suggest. Empirical data from partially articulated skulls and disarticulated skull elements could provide useful comparisons.
In another consideration of skeletal sorting in dinosaurs, analyses of ceratopsian remains from Bonebed 43 in Dinosaur Provincial Park by Ryan et al. (2001) followed an approach from Lehman (1982), and modified the dispersal group assignments of several elements compared to patterns established from mammals (Voorhies, 1969). The data from alligator bone dispersal collected in this study might not specifically inform the assignment of ceratopsian ilia as late dispersing elements, or the assignment of coracoids as earlier dispersing elements than scapulae (Ryan et al., 2001). However, the different skeletal sorting patterns recovered across taxa ranging from mammals to birds, turtles, and crocodylians do give additional support for considering lineage specific patterns of skeletal dispersal in taphonomic analyses. Moreover, variation across these patterns highlights how distinctive anatomical features can influence the concentration and preservation of fossils and, thereby, shape our understanding of assemblage composition and paleofaunal diversity and evolution.

ACKNOWLEDGMENTS
We would like to thank Tony Fiorillo, Cathy Forster, and Dave Weishampel for the invitation to contribute to this special issue of The Anatomical Record honoring Peter Dodson on his retirement. We also thank Melissa Fuentes, Curator of Vertebrates for Clemson University's Bob and Betsy Campbell Museum of Natural History, for loaning and permitting the use of the alligator specimens on which we conducted trials. Michael Ryan and an anonymous reviewer provided helpful comments on an earlier draft of this manuscript. Finally, RWB thanks Peter Dodson for his support and mentorship that helped start a career in science, and for his friendship ever since. DATA AVAILABILITY STATEMENT Competent velocity data from individual trials of each skeletal element in each of its tested orientations are reported in Table S1. All other relevant data are within the body of the article.