In 1999 Horner et al. described the bone histology of the type specimen of Hypacrosaurus stebingeri (Museum of the Rockies, Bozeman, Montana, MOR 549), a large hadrosaurid dinosaur from the Late Cretaceous of Montana. A variety of bones in the skeleton was sectioned. A principal focus of that paper was to determine the animal's age at death, which presented a challenge because the numbers of preserved LAGs (lines of arrested growth) varied among the sampled bones. In some bones, LAGs were obscured or erased by the formation of erosion spaces in the deep cortex, the expansion of the medullary cavity, and the proliferation of secondary osteons, which usually developed first in the deep cortex and increasingly expanded toward the periphery (Francillon-Vieillot et al., 1990). We concluded at the time that it was difficult to assess skeletal age (inferred from the number of LAGs) from one bone alone, and suggested the use of several in order to arrive at the most reasonable estimates. In the present note, we offer a refinement on this suggestion, and propose a hypothesis of process to explain some of these patterns.

Two other factors affect the assessment of age, however. One is growth rate, because in very rapidly growing bones, particularly at early growth stages, annual lines may be difficult to discern (Horner et al., 2000), as is notorious for very large, rapidly growing dinosaurs such as sauropods (Sander, 2000). There has been some question whether some larger dinosaurs even deposited a visible first-year LAG, even when later ones are visible (e.g., Horner et al., 2000). This pattern is understandable according to two general classes of observation: larger species of a clade tend to grow at higher rates than smaller species (Case, 1978), and individuals grow more rapidly at earlier stages than at later ones (Amprino, 1947).

The second factor that affects the assessment of age is the interpretation of the EFS, or External Fundamental System (Cormack, 1987), a relatively avascular and acellular bone tissue that forms sub-periosteally on the bones of individuals that have essentially reached a plateau in skeletal growth (i.e., the upper asymptote of their growth curve). Although this feature was first reported in mammals, it has been found in many archosaurs, including birds, non-avian dinosaurs, crocodiles, and other extinct pseudosuchians (e.g, de Erickson et al., 2004, Horner et al., 1999, Ricqlès et al., 2003, Sander, 2000, Turvey et al., 2005 and Woodward et al., 2011). It reflects asymptotic growth, which appears to be common to all vertebrates (more slowly growing taxa are not sampled or die before reaching maximum size, and grow very slowly as they approach it: Sebens, 1987).

It is now universally accepted that LAGs are annually deposited, although their interpretation in individual taxa with aberrant histological features (double and triple LAGs, polish lines, etc.) may be complicated, and environmental or physiological conditions may affect their production (Castanet, 1986-1987). In the EFS, growth marks may appear that are more or less continuous around the periphery of a bone, much like LAGs; but they are often difficult to trace for much of the periphery. There is also reason to think that the lines of the EFS may not be annual, or at least have not been shown to be annual. Horner et al. (1999; see Table 1) found that they varied in number more than cortical LAGs did in a single skeleton of the hadrosaurid dinosaur Hypacrosaurus. Castanet et al. (1988) found that the number of lines in the EFS in the tuatara, (Sphenodon) varied between the phalanges and the femur of the same individual. The femur had more lines overall, in both the cortex and EFS, and they inferred that the cause was the longer growing period of the femur. Martinez-Maza et al. (2014) treated EFS lines in the metapodials of the Miocene horse Hipparion as potentially annual, although they admitted that there is no experimental evidence that validates this assumption. Horner et al. (1999) found it particularly problematic to interpret the skeletochronology of the metapodials in Hypacrosaurus (see Table 1), and that throughout the skeleton the lines in the EFS provided no consistent signal.

We hypothesize that much of the variation in the expression of skeletochronological indicators (and potential indicators) is often a function of growth rates of individual bones, especially in large, rapidly growing animals. These growth rates are in turn underpinned by metabolic rates. The difficulty of counting growth lines in the cortex of the limb bones of giant sauropods presents an example.

In addition to the presence of LAGs and the EFS, another feature often considered a manifestation of age is the proliferation of secondary (Haversian) osteons in cortical bone. When these structures appear only in one region of a bone or along a single axis, this suggests an association with mechanical stress, cortical drift, or repair of microdamage, rather than a pure reflection of absolute age (Frost, 1994 and McFarlin et al., 2008). Secondary osteons that form as a result of age rather than biomechanical processes generally appear first in the deep cortex of long bones and spread centrifugally through ontogeny (Alquist and Damsten, 1969, Francillon-Vieillot et al., 1990 and Kerley, 1965). This general relationship between the extent of secondary remodeling and age has been noted in dinosaur growth series (e.g., Horner et al., 2000, Klein and Sander, 2008 and Werning, 2012), and has been used to infer relative ontogenetic stages among conspecifics (e.g., Scannella and Horner, 2010, although this involved cranial remodeling).

Notably, however, the amount of secondary remodeling often differs among elements. For example, Werning (2012) found that in the bipedal ornithopod dinosaur Tenontosaurus, the forearm elements always showed greater secondary remodeling than the femur or tibia of the same individual. Furthermore, the remodeling of the forelimb began at earlier ontogenetic stages, and thus these elements appeared to be histologically “older” than the larger elements. This is also the case in the ornithopods Hypacrosaurus (Horner et al., 1999; Fig. 1) and Maiasaura (Horner et al., 2000), and the prosauropod Plateosaurus (Klein and Sander, 2007).

But what causes these differences? Why do some bones remodel more extensively or earlier in ontogeny? Without casting doubt on the validity of the biomechanical processes discussed above, we suggest the hypothesis that in large animals with high growth rates, the onset and pace of secondary bone formation can be largely a function of the size of the bone, and of the growth strategies of individual bones.

Dinosaurs are among the largest animals for which we have a reasonable sense of growth rates (Padian and Lamm, 2013). Some, such as tyrannosaurs, are obligate bipeds; some, such as sauropods, are obligate quadrupeds; and some, such as hadrosaurs, may have been facultatively quadrupedal or bipedal (Coombs, 1978). In animals with serially homologous long bones of the same approximate size (e.g., humerus and femur, ulna and tibia), the development of secondary osteons should be similar in serially corresponding bones, unless there is a biomechanical reason that complicates this (for example, regional variation near a muscle attachment site). Such animals will normally be quadrupeds, and their long bones are sharing the weight-bearing and the locomotory impetus (Alexander et al., 1979, Anderson et al., 1985 and Campione and Evans, 2012). In quadrupeds with more disparity between the fore and hind limbs, the situation may be somewhat different: the hind limbs, being larger, are thought to bear greater responsibility for weight-bearing and locomotion (Anderson et al., 1985 and Campione et al., 2014). We further hypothesize that in bipedal (and possibly facultatively bipedal) animals, in which there is a particularly strong limb disparity, the development of secondary osteons will be greater in the smaller bones than in larger bones, even if they support little or no weight.

This disparity in growth begins with eclosion (hatching). When embryonic reptile skeletons are contained in eggs, their long bones are of more similar lengths than later in ontogeny, because their maximum lengths are constrained by the geometry of the egg. For example, pterosaur long bones are longer than axial skeletal elements in adult forms, and also much differently proportioned among themselves. But in the egg, the length differences among the bones are inconsiderable. Jenkins et al. (2001, Table 1) described a very young post-hatchling pterosaur and compared its proportions to those of other juvenile and adult Triassic pterosaurs (Eudimorphodon and Peteinosaurus). They showed that there was much less difference among the long bone elements in juveniles than in adult forms. Upon hatching, differential growth commences.

Our hypothesis centers on the problem of differential growth after eclosion. As many animals grow, different bones assume different functional roles, and their sizes and proportions grow accordingly. To achieve allometric growth, smaller elements must either grow more slowly than larger ones, or reach their adult (maximum) proportions earlier than larger ones. Each element experiences its own growth trajectory, which may differ from those of the other elements, and from the rate of accumulation of body mass. In the absence of other biomechanical pressures, larger bones (especially the main mass-bearing elements) may experience both higher growth rates and longer durations of growth than smaller bones. These elements should also record histological signals of higher growth rates than the more slowly growing elements (e.g., higher canal density, more disorganized collagen fibers, wider spacing between LAGs; Amprino, 1947 and Francillon-Vieillot et al., 1990).

It could be, as a reviewer suggested, that this pattern may be explained by Von Baer's Law. In this case, the long bones would be of more similar lengths early in ontogeny because general characteristics of the group develop before special ones, so isometric growth would prevail before allometric growth in the developing embryo.

We hypothesize that the degree of secondary remodeling in such animals is not mainly a function of biomechanical stress or the animal's age, but of different growth trajectories among elements. Under this hypothesis, the largest long bones would be growing too rapidly to deposit secondary tissues, because the nutrients delivered to them would largely be used in primary periosteal bone deposition. This reflects their role in support; these bones must pace their growth with that of the animal's increasing body mass. On the other hand, smaller bones, especially if they do not bear weight, will experience considerable secondary remodeling as a means of slowing their element-specific growth rates.

Secondary remodeling may serve as the mechanism of slowed growth in two ways. First, remodeling reduces the overall porosity of the element relative to the faster-growing primary tissues, which have greater canal density and connectivity. Reducing the total number of canals and their anastomoses in turn reduces nutrient supply from the blood. Second, remodeling diverts resources from periosteal expansion. The available nutrient content contained within the blood will be similar whether flowing through smaller or the larger bones (which still need additional nutrients to accommodate continued rapid growth). Rather than using the nutrients to increase circumference, smaller bones use them to rework existing tissues. This would account for the preponderance of secondary bone tissue in these smaller bones that do not bear weight or experience significant strain. This is a different situation from that of smaller weight-bearing bones such as the fibula and metatarsals, which may experience considerable reworking because they take on substantial biomechanical stress.

Our hypothesis suggests that high metabolic rates may be forcing bone tissue development, which is uncontroversial; but it further suggests that secondary bone tissue deposition is greater in the smaller long bones of animals with high metabolic rates than in the larger bones, because the smaller bones may not “refuse,” in a sense, the metabolites carried by the blood. Here, for convenience, we are using somewhat metaphorical language in order to make a point. Because they do not “want” to grow, in a vitalistic sense (mediated by genetic and epigenetic factors), these bones remodel their tissues in order to expend the energy and metabolites. In other words, our hypothesis reduces to this: in some taxa with disparate limb proportions, the smaller bones don’t “want” to get bigger, so they have to remodel their tissues because they can’t refuse the supply of metabolites that the vascular system brings to all bones in the body. We think that this largely explains the disproportionate distribution of secondary osteons in large and small bones in these animals, regardless of weight-bearing.

To use another metaphor, it is like constantly constructing a house: larger bones such as the femur and tibia make new tissue, like adding on rooms of a house, whereas smaller bones such as the radius and ulna are doing what amounts to interior remodeling of a house, in both cases using the same rate of flow of materials.

This hypothesis can be directly tested by examination of the histology of the long bones of a variety of bipedal and quadrupedal tetrapods with high growth rates. Among bipeds, we predict that the smaller bones of the forelimb (1) will show a different ontogenetic trajectory (i.e., growth will cease before it does in the larger bones, or we will see disproportionately slower growth with age); (2) will acquire greater development of secondary bone tissues, even though they bear little or no weight; and (3) will show the appearance of secondary osteons that will coincide with (earlier) growth cessation in smaller elements, or with their transition to slower growth than in larger elements. Animals that are facultative quadrupeds may be expected to show different results, but this will require careful testing.

This prediction is borne out histologically in obligate and facultative bipeds such as Hypacrosaurus, Maiasaura, Tenontosaurus, and Plateosaurus. In these taxa, the largest long bones are the femur and tibia, which typically experience little or no remodeling until growth slows considerably, in what would be called the “late juvenile” or “sub-adult” stage, depending on the taxon. However, in all cases the forelimb elements begin remodeling much earlier in terms of the animal's ontogeny (e.g., Horner et al., 1999, Horner et al., 2000, Klein and Sander, 2007 and Werning, 2012). However, this is not always the case. For example, the ornithopod Dysalotosaurus shows no difference in the onset or extent of secondary remodeling between forelimb and weight-bearing elements (Hübner, 2012).

Caution will need to be exercised. For example, this hypothesis may be less applicable depending on the adult size of the taxon (more remodeling often occurs in smaller taxa, because they have lower growth rates than larger related species: Padian et al., 2001). Additionally, in animals with high metabolic rates that secondarily grow slowly, secondary bone growth should be much more prevalent in their bones than in the bones of related animals that grow quickly. Humans, for example, have high metabolic rates like other mammals but they grow relatively slowly; as a result, the long bones of adults and sub-adults have extensive development of secondary osteons. Compare, for example, the secondary bone development recorded in the humerus and femur of the green monkey, the gibbon, and the chimp–all slowly growing primates of small to medium size (McFarlin et al., 2008) – with that of Hypacrosaurus (Horner et al., 1999). The latter taxon has less Haversian remodeling, especially in its larger elements, largely a function of growth rates (even though all these taxa have relatively higher metabolic rates than most reptiles and amphibians).

Another example may be provided by dwarf sauropods (e.g., Benton et al., 2010, D’Emic and Wilson, 2012 and Stein et al., 2010) whose long bones show extensive secondary remodeling. Biomechanical stresses could have been involved, but because all bones tend to be substantially affected by Haversian remodeling, and the long bones of much larger related sauropods (with much higher proportional stresses) are less so, we find no support for a biomechanical hypothesis. On the contrary, we suspect that secondary dwarfing of these small sauropods resulted in bones that grew to adult size at somewhat lower rates than their larger relatives as a consequence of phylogenetic legacy, but continued to be remodeled extensively because they retained more or less their relatively high ancestral metabolic rates (Fig. 2).

We predict that in many large animals with rapid growth trajectories and some disparity in size in the long bones and other skeletal elements, the largest bones will show less secondary remodeling than smaller ones. The reason is that, whereas the largest bones are increasing their dimensions too rapidly to accommodate much secondary reworking (until they approach full size), the smaller bones, freed from the weight-bearing constraints of pacing body size, must slow or cease growth earlier to achieve allometry despite high metabolic rates and a continued flow of metabolites. Secondary remodeling forces slower growth both by reducing overall porosity (i.e., the available blood supply) and by diverting resources from growth to replacement of existing bone tissue.

We stress that it is the smaller bones, and the non-weight-bearing bones, that are at issue here, because they can test patterns of remodeling seen in larger, weight-bearing bones in both the forelimb and hindlimb. Explicit tests of this hypothesis will require sampling the same series of several elements in individuals comprising a growth series, in order to confirm different growth rates and trajectories, and the exact timing of secondary osteon appearance among elements. As Horner et al. (2000) noted, sampling different elements from the same skeleton may result in different estimates of the animal's overall ontogenetic stage. Notably, if smaller elements had much different ontogenetic trajectories compared to larger ones or to the animal's overall growth, it would reduce their utility in reconstructing the animal's overall ontogeny. In these cases, we would recommend using larger, weight-bearing elements whose growth paced mass to estimate the animal's age-mass trajectory.

This hypothesis is about a process, and it attempts to predict some of the patterns outlined in the paragraphs above. It does not attempt to undermine or replace other circumstances in which secondary bone tissue is produced. In each case, hypotheses have to be tested. Ours can be wrong in a given case for a variety of reasons related to functional load or growth and metabolic rates, or to the sequestration of tissues through the life of the animal, as outlined by the alternative hypotheses discussed above. This hypothesis is mostly confined to the explanation of differential development of tissue in animals of high rates of growth and metabolism, in which some long bones show differential size and mechanical load.

Sauropod dinosaurs can be particularly instructive here because all their limb bones bear weight. According to our hypothesis, the zeugopodial bones should be relatively more reworked than the propodials. (To our knowledge, the histology of these bones has not been assessed in a single individual sauropod) In contrast, bipedal dinosaurs would be differently instructive. The smaller, non-weight bearing bones of the forelimb (propodials and zeugopodials) might be more reworked than the corresponding larger, weight-bearing bones of the hindlimb. (This could be productively tested in theropods and bipedal ornithopods, for example.) Many other taxa could be recruited to test this hypothesis.

Our general aim is to put forward a hypothesis that might help to discriminate between phylogenetic or mechanical effects and those related to the size of the bone and the rate of its growth, given its underlying metabolic regime. If our hypothesis stimulates further productive work, then it will have done its job.