Conceived and designed the experiments: JAW. Performed the experiments: JAW. Analyzed the data: JAW. Contributed reagents/materials/analysis tools: JAW. Wrote the paper: JAW.
The authors have declared that no competing interests exist.
As gigantic herbivores, sauropod dinosaurs were among the most important members of Mesozoic communities. Understanding their ecology is fundamental to developing a complete picture of Jurassic and Cretaceous food webs. One group of sauropods in particular, Diplodocoidea, has long been a source of debate with regard to what and how they ate. Because of their long lineage duration (Late Jurassic-Late Cretaceous) and cosmopolitan distribution, diplodocoids formed important parts of multiple ecosystems. Additionally, fortuitous preservation of a large proportion of cranial elements makes them an ideal clade in which to examine feeding behavior.
Hypotheses of various browsing behaviors (selective and nonselective browsing at ground-height, mid-height, or in the upper canopy) were examined using snout shape (square vs. round) and dental microwear. The square snouts, large proportion of pits, and fine subparallel scratches in
These results reaffirm previous work suggesting the presence of diverse feeding strategies in sauropods and provide solid evidence for two different feeding behaviors in Diplodocoidea. These feeding behaviors can subsequently be tied to paleoecology, such that non-selective, ground-height behaviors are restricted to open, savanna-type environments. Selective browsing behaviors are known from multiple sauropod clades and were practiced in multiple environments.
Herbivory evolved multiple times within Archosauria, the group uniting crocodiles and birds and containing a variety of fossil forms such as dinosaurs, aetosaurs, phytosaurs, rauisuchians, and pterosaurs. Among living archosaurs, however, herbivory is restricted to birds and potentially turtles, which have been recovered as the sister clade to crown-group Archosauria
Of particular importance to the understanding of Mesozoic ecology are sauropod dinosaurs, which were the dominant megaherbivores during most of the Jurassic and Cretaceous, a span of approximately 135 million years. Sauropods were typically quite large, with the largest reaching over 30 m in total body length
Although sauropods lack the obvious adaptations for herbivory (e.g., beaks and cheeks) present in the other major clade of herbivorous dinosaurs, Ornithischia
Within Sauropoda, no clade has received as much attention with regard to diet as Diplodocoidea
The reconstruction of diet in fossil taxa has a long methodological history, and in sauropods dates back to the 19th century
The shape of the premaxilla, constituting the entirety of the anterior-most extremity of the skull, or ‘snout’, in most mammals, has long been suggested to be related to dietary preference or selectivity in herbivorous mammals
Janis and Erhardt
Solounias et al.
Dompierre and Churcher
The relationship between snout shape and dietary habit has also been inferred outside of ungulate mammals. Christiansen
Although sauropods were typically much larger than both large ungulates and large hadrosaurs, absolute skull size does not differ greatly between the two dinosaurian clades. Cranial volume estimates in sauropods range from ca. 0.05 m3 in
Above: A)
Dental microwear is the study of damage done to teeth by contact with other surfaces, in particular the interactions between a tooth and food, grit, and opposing teeth. Studies of microwear typically include the quantification of features falling into one of the following three categories (
A) Scratches, features at least 4× longer than wide. B) Gouges, large features with irregular margins. C) Pits, subcircular features, typically small. D) A tooth of
Such features preserve information about the last few meals an organism ate and are one of the few direct lines of evidence we have for interpreting the diets of extinct organisms, particularly when those organisms have no extant descendants for comparison. Although caution must be used when interpreting diets of extinct organisms, broad dietary categories are often assignable. By providing direct evidence of what organisms ate, microwear features can also indicate overlap in resource exploitation and behavior.
The majority of microwear studies to date involve the reconstruction of diets in extinct and extant mammals
This reliance is a potential concern, due to broad differences in shape and function between the majority of teeth in question and the molariform teeth examined in the mammalian studies. Molars are used for oral processing of food in mammals, using a combination of puncturing, slicing, and crushing to break down foodstuffs. Non-mammalian animals typically have no oral processing (with the notable inferred exceptions of marginocephalian and ornithopod dinosaurs) and use their teeth solely for food-acquisition behaviors involving puncturing or slicing. Incisiform teeth, therefore, may prove a more useful analogue to archosaur teeth than molariform teeth.
The majority of incisor microwear studies have been done on primates, although isolated studies of non-primate incisor microwear exist
Although the degree to which the manual manipulation of foodstuff by primates alters the character of wear is uncertain, some behaviors, such as leaf stripping, are strong candidates to broadly correspond with potential food acquisition behaviors in other organisms. Importantly, leaf stripping is known to leave characteristic microwear patterns on incisors in
The only study to examine both molar and incisor microwear in an ungulate (
As noted above, several studies have discussed microwear features in sauropod dinosaurs, most commonly in diplodocoid sauropods
The influence of jaw motion on microwear features in sauropod dinosaurs is also worthy of discussion. The tooth row in diplodocoids is restricted anteriorly, resulting in a dentition that is primarily oriented transversely
It is also important to note that much of the work on mammalian microwear surrounds the grazer-browser continuum, which necessarily cannot apply to Jurassic and Early Cretaceous dinosaurs due to the lack of grasses at that time
Caution should be exercised when drawing parallels between microwear on mammalian teeth (including incisors) and that found on sauropod teeth. The chewing motion of an ungulate, with its transverse power stroke, is highly derived and without parallel in Sauropoda. Certain aspects of foodstuffs, particularly their toughness and grit content, however, can be expected to influence microwear features with reasonable consistency across taxa. Although the relationship between incisor microwear and diet is uncertain, inferences can still be drawn about the character of foodstuffs ingested. In particular, feature size and texture are expected to correspond with browse height and certain intrinsic plant properties, such as woody vs. herbaceous stems.
To develop a testable hypothesis for diplodocoid feeding behavior, evidence from snout shape and microwear will be compared, not just with data from modern herbivores, but also with each other. In this way, an approximation of the “total evidence” approach will be brought to bear on this question.
Morphological data, in this case snout shape and microwear features, can be used to distinguish between several combinations of browse height (ground-height, mid-height, and upper canopy) and browse strategy (selective and nonselective) in diplodocoid sauropods. Here, ground-level feeding is defined as feeding on vegetation within 1 m of ground height, mid-height feeding is defined as feeding between ground height and 10 m, and upper canopy feeding is feeding at all heights above 10 m. Ten meters is chosen as the upper limit for mid-height browsing based on estimated maximum head height of diplodocoids in quadrupedal stance, using combined neck and forelimb height
Feeding strategy | Snout Shape | Microwear features |
Browsing: Ground-height | Square | high proportion of pits relative to other features, fine scratches |
Browsing: Mid-height (1–10 m) | Square/Round | fewer pits relative to other features |
Browsing: Upper canopy | Square/Round | few pits |
Browsing: Non-selective | Square | subparallel scratches, fine features (i.e. fine scratches, no pits) |
Browsing: Selective | Round | cross-scratches, large features (i.e. coarse scratches, gouges) |
Browse height was examined primarily through the examination of dental microwear features. The primary influence of browse height on microwear is through the creation of pits as a consequence of ingested grit. Grit, exogenous mineral particles, are suspended and transported in the air through aeolian processes and through the actions of animals (e.g., walking). These particles then fall out of suspension and are deposited on the ground or on plants. A proportionally larger amount of grit is deposited on plant surfaces at lower heights than those that greater heights
The incisors of ground-height feeding primates also display a slightly different character than those of taxa browsing at greater heights. Mean scratch breadth is lower in taxa browsing at or near ground height, as a function of mean particle size and the ratio of soil particles to phytoliths
Browsing strategy (selective vs. nonselective) was examined using both microwear and snout shape indices. Snout shapes in selective browsers were predicted to be narrower than those of nonselective browsers, as in herbivorous mammals. Two microwear features can indicate browse type: scratch orientation consistency and feature size. Consistency of scratch orientation has been related to food texture, such that softer (e.g., herbaceous) foods result in more unimodally distributed scratch orientations, whereas low consistency of orientation (i.e., cross-scratches) is related to eating harder or more brittle foods
Six genera of diplodocoid sauropod dinosaur (
These reconstructions (
Reconstructions of
The uAI is a modification of the AI introduced by Boué
Snout depicted based on
The PMI is also a modification of an older metric; in this case it is a modification of the premaxillary shape index (PSI)
Due to the often fragmentary or deformed nature of fossil material, the prior two metrics rely on reconstructions of skulls; measurements based on deformed materials would necessarily represent an unrealistic shape. Although every effort was made to accurately reconstruct morphology, this has resulted in a small sample size. One method to measure squareness directly from fossil material (reducing potential error inherent in reconstruction) and increase sample size involves comparing the angles of divergence on the anterior margin of the premaxilla (PMDA). This is measured here by orienting the specimen in strict dorsal or ventral view and measuring the angle formed between the anterolateral and anteromedial corners of the premaxilla and a line drawn perpendicular to symphysis (
Forty-seven teeth belonging to seven genera were examined for microwear features (
Although sauropod remains are abundant globally, skulls are a rare component of sauropod fossil assemblages
The distributions of three variables were examined statistically: snout shape (for browsing strategy), scratch breadth (for browse height), and pit/gouge size (for both height and strategy). For two-sample comparisons, non-parametric Mann-Whitney U tests were performed; sample sizes for snout shape were too small to meet normality criteria, and both scratch breadth and pit/gouge size failed a normality test (Shapiro-Wilk) for all samples. The Kruskal-Wallis test (an extension of the Mann-Whitney U) was used for instances where multiple comparisons were desirable. For significant results of Kruskal-Wallis tests, pairwise comparisons (following the method of
Coahuila, Saltillo, Coahuila, Mexico;
Results of the snout shape analyses (uAI, PMI, PMDA) are summarized in
uAI | PMI | PMDA | |
|
1.5 | 84% | 6.3° |
|
0.6 | 74% | 24.4° |
|
1.2 | 84% | 7.4° |
|
4.0 | 95% | 3.5° |
|
— | 74% | 25° |
|
— | 71% | 25.4° |
|
0.6 | 68% | 33° |
|
0.4 | 63% | 40° |
Containing clade | Family | Genus | Species | Specimen # | PMDA | Position |
Diplodocoidea | Diplodocidae |
|
|
AMNH 969 | 4.5 | L |
7 | R | |||||
USNM 2673 | 9.8 | L | ||||
|
CMNH 11161 | 8.3 | L | |||
|
|
CMNH 11162 | 12.4 | L | ||
4 | R | |||||
|
CMC VP 7800 | 3 | L | |||
5.8 | R | |||||
|
|
MB.R.2346 | 25.4 | L | ||
Dicraeosauridae |
|
|
ANS 21122 | 25 | L | |
|
|
MB.R.2339 | 24.4 | L | ||
Rebbachisauridae |
|
|
MNN GAD-512 | 3.5 | L | |
Macronaria |
|
|
34.5 | L | ||
30.2 | R | |||||
|
|
CM 11338 | 45.4 | R | ||
46.5 | L | |||||
|
UUVP 3999 | 27 | R |
Four taxa (
As above,
The initial sample included 16 premaxillae. Of those 16, 12 were deemed complete enough to measure (
The diplodocoids again appear to have segregated into two groups, with
Although 47 teeth were examined for microwear features, only 11 (23%) were found to have features that record diet. Of these, only seven teeth belonging to five taxa (
A,
Taxon | Specimen | Pit # | L∶B | Area (µm2) | Circ. (µm) | S. # | S. L. (µm) | S. B. (µm) | C.S. |
|
CMC VP7180 | 109 | 1.27∶1 | 73.37 | 46.43 | — | — | — | n/a |
|
MB.R. 2204D | 359 | 1.64∶1 | 99.74 | 31.32 | 228 | 54.93 | 3.73 | Y |
MB.R. 2204E | 227 | 1:75∶1 | 160.06 | 43.75 | 176 | 49.12 | 4.93 | Y | |
|
CM 11161 LP1 | 14 | 1.28∶1 | 33.43 | 24.90 | 19 | 94.35 | 2.65 | N |
CM 11161 RP2 | 37 | 1.38∶1 | 16.54 | 18.43 | 26 | 52.46 | 3.4 | N | |
USNM 2673* | 15 | 1.32∶1 | 423.26 | 69.17 | 37 | 236.17 | 4.33 | Y | |
|
G2 | 276 | 1.48∶1 | 34.64 | 26.80 | 138 | 678.49 | 2.69 | N |
G100* | 9 | 2.28∶1 | 208.04 | 70.19 | 30 | 137.53 | 3.84 | Y | |
|
MNHN 1512a | 14 | 1.2∶1 | 174.73 | 69.17 | 7 | 166.29 | 3.57 | Y |
|
MB.R. 2190 | 49 | 1.48∶1 | 520.57 | 51.12 | 72 | 57.52 | 3.86 | Y |
|
UUVP 1949 | 124 | 1.95∶1 | 80.03 | 31.83 | 99 | 47.32 | 3.76 | N |
UUVP 3986 | 203 | 1.74∶1 | 93.46 | 33.31 | 185 | 80.21 | 4.02 | Y |
Abbreviations: L∶B, average length∶breadth ratio of pits/gouges; S. #, number of scratches; S. L., average scratch length; S. B., average scratch breadth; C.S., cross-scratches.
The probable
Two teeth (MB.R. 2204 and 2197), tentatively assigned to
The second tooth, MB.R. 2197, does not preserve a large amount of microwear, although some features are present on the labial enamel edge. Here, two exceptionally large gouges are preserved. No scratches or smaller pits were observed, and it is probable (although not certain) that these gouges are not the result of tooth-food or tooth-tooth contact.
Three teeth from two specimens (CM 11161 and USNM 2673) preserve microwear features on the enamel.
Microwear features were recovered from small areas of the labial enamel margin of the facet on the second right premaxillary tooth and the first left premaxillary tooth of CM 11161. Identification of features was hampered here and on other diplodocid specimens by the application of a preservative lacquer, likely around the turn of the century. Attempts to remove this lacquer using alcohol-based solvents were generally unsuccessful.
As also seen in
Seven loose teeth are associated with the
The wear features recovered from this surface differ from those seen on the labial facet in CM 11161 in both size and character. Scratches are substantially longer and broader than those from the facet margin. They are generally sub-parallel with rare-cross scratching, but the orientation is nearly perpendicular to the long axis of the tooth, with only a slight apicobasal component. Pits are also larger than those observed on the facet, although they are proportionally identical to those from CM 11161. Because these features do not come from either a facet or from the apical surface of the crown, they cannot be confirmed as the result of a bite stroke.
Wear features were recovered from two crowns, one each from the the G2 and G100 assemblages.
Wear features were observed on the labial enamel margin of the labial facet on the G2 crown. As in
No wear features were recovered from the facet of the crown from G100, perhaps as a result of post-mortem wear. Features were observed on the lingual surface of the enamel basal to the actively worn surface. Of the features observed, the majority were scratches. All features were generally larger than those observed on the facet of G2. Similar to the wear recovered from USNM 2673, scratch orientation is more mesiodistal than apicobasal, although there is still a minor apicobasal component. Cross scratching occurs but is rare. Pit size was also substantially larger than those seen in the G2 tooth. Because these features are located away from surfaces of active wear, it is likely that they were caused by some contact outside of the bite stroke.
A single loose tooth of rebbachisaurid type (MNHN 1512a; see
Two teeth (UUVP 1949, UUVP 3986) preserved a substantial amount of quantifiable microwear features. In both cases, pits outnumber scratches. The pits are large, similar in size to those seen in
Wear features were recovered from a single tooth, MB.R. 2190. Scratches dominate the preserved features. The pits that are preserved are often quite large, larger than but most similar to those preserved in
The six ingroup taxa examined can be broadly divided into two categories: square and round. Square-snouted diplodocoids include the diplodocids
Testing this statistically proves somewhat problematic, however. Grouping the sauropods a priori into three groups—“square” diplodocoids (
This pattern of snout shapes is reminiscent of the pattern seen in modern ungulates. In light of that comparison, the broad-snoutedness seen in
The snout shapes of macronarian sauropods like
Examination of sauropod taxa outside the neosauropod radiation (
As discussed previously (see
Broad-snouted diplodocids group within the range of PMI scores seen in hadrosaurines; in both clades, scores cluster around 80–85% (
The vertical axis separates taxa into sauropods (top), hadrosaurines (middle) and lambeosaurines (bottom). Squares represent taxa considered to have been non-selective browsers, circles represent taxa considered to have been selective browsers. Blue tones indicate the range of square snouts and yellow tones indicate the range of round snouts; green tone indicates overlap; dark blue/dark yellow represent the limits of hadrosaurian snout shape diversity. Overlap in snout shape occurs between behavioral guilds in hadrosaurs, but not in sauropods, although sample size is limited for sauropods. Sauropod snout shapes are also more disparate than snout shapes in hadrosaurids. Inferences of hadrosaur diet based on
A Mann-Whitney U test comparing the PMI scores recorded for combined diplodocoid and hadrosaurid “square” (
Although the relationship between wear on molar and incisiform teeth is uncertain, the pattern of variation in wear features between round-snouted (
Square-snouted diplodocoids
|
|
|
|
|
|
|
— |
|
0.00 |
|
0.000 |
|
|
— | 0.00 | 0.000 | 0.000 |
|
0.00 | 0.00 | — | 0.000 | 0.000 |
|
|
0.000 | 0.000 | — |
|
|
0.000 | 0.000 | 0.000 |
|
— |
Results in bold indicate cases where the null hypothesis (distributions in each sample are the same) cannot be rejected. Putative high browsers (
The greater orientational consistency in
Pit/gouge size was also found to differ significantly between several of the taxa (H = 229.386, df = 2, P = 0.000). Pairwise comparisons found significant differences in pit/gouge area between
|
|
|
|
|
|
|
— |
|
|
0.000 | 0.000 |
|
|
— |
|
0.000 | 0.000 |
|
|
|
— | 0.000 | 0.000 |
|
0.000 | 0.000 | 0.000 | — | 0.005 |
|
0.000 | 0.000 | 0.000 | 0.005 | — |
Results in bold indicate cases where the null hypothesis (distributions in each sample are the same) cannot be rejected. None of the taxa hypothesized to be selective browsers can be distinguished statistically from each other.
In addition to their significantly larger size, the features recovered from
Fiorillo
The absence of pits/gouges in most previous studies is somewhat perplexing, given their relative abundance in all teeth examined in this study. Further confusion comes from the overlap in sample between a number of studies: this study, the study of Upchurch and Barrett
Sereno et al.
Ryan
Features recorded from the non-occlusal surfaces of presumptive ground-level browsing taxa such as
Snout shape and microwear indices suggest the presence of both a ground-level, nonselective browsing behavior and a mid-height (above 1 m), selective browsing behavior in diplodocoid sauropods. Here, feeding behavior is examined in relation to body size, phylogeny, and paleoecology, to determine the influence of each on behavior in diplodocoid sauropods.
Among flagellicaudatans (the group containing Diplodocidae and Dicraeosauridae), selectively browsing taxa had smaller skulls, and were smaller overall, than non-selective browsers
Within individual diplodocoid clades, feeding behavior was reasonably consistent, although snout shape can be determined for only one rebbachisaurid (
Data suggest that ground-height, non-selective browsing evolved in open, savanna-like environments, whereas selective, mid-height browsing was most common in diplodocoids living in closed environments dominated by mid- and upper-canopy browse. Blue tones indicate data suggestive of ground-height, non-selective browsing; yellow tones indicate data suggestive of mid-height, selective browsing. Inferences for which insufficient data exists are represented in 50% grey tones. Abbreviations: S, savanna type ecosystem; F, forested ecosystem; G, ground-height browser; M, mid-height browser; N, nonselective browser; Sl, selective browser.
The ancestral condition for snout shape in diplodocoids is difficult to determine. Although both diplodocids and rebbachisaurids appear to have square snouts, there is no overwhelming evidence to suggest that this is the original condition for the group as a whole. The ancestral state in Flagellicaudata is equivocal: basal diplodocids have square snouts and basal dicraeosaurids have round snouts. Furthermore, the only rebbachisaurid for which snout shape can be determined (
Similar morphological plasticity in other dinosaurian groups (e.g., hadrosaurids) has typically been interpreted in relation to behavior, although often behavior and morphological divergence are correlated with phylogeny
Herbivory imposes two major constraints on behavior: foraging time and digestive time
Variation in forage quality may also have had some influence on behavior. Two hypotheses in particular, the forage abundance hypothesis (FAH)
Although the ecological conditions of Niger during the Aptian-Albian are largely unknown (particularly with regard to flora), the paleoecology of the Late Jurassic Morrison and Tendaguru formations, where the remaining five taxa have been found, is better understood.
The Morrison Formation of North America is dominated by diplodocoid sauropods, in particular the square-snouted diplodocids. Diplodocids (primarily
Plant fossils from the Morrison include a wide diversity of conifers, ginkophytes, podocarpaceans, ferns, cheirolepidiaceans, and horsetails
Low browse in the southern Morrison, particularly the ferns and small trees, appears to have been highly nutritious and digestible, on par with extant browse
The northern end of the Morrison Formation (e.g., Montana) has been recognized for its unusual sauropod fauna, composed of smaller adults and more juveniles than is typical of more southern localities
Unlike that of the Morrison, the sauropod fauna of Tendaguru is made up exclusively of round-snouted taxa, including
The sediments preserved in the Tendaguru Formation encompass both strictly terrestrial uplands and tidal flats/coastal regions, but the latter appear to have been poorly vegetated
No ground-height, non-selectively browsing sauropod dinosaurs have been recovered from Tendaguru. The sauropods that have been found there are exclusively mid-height (
Forested habitats that are linked with riparian environments are also associated with round-snouted sauropods in both the Morrison (
Hypotheses of feeding behaviors typical of modern mammalian herbivores (e.g., non-selective and selective browsing) are supported for diplodocoid sauropods using evidence from snout shape and dental microwear. Snout shapes in diplodocoids include both rounded and square snouts, similar to those seen in hadrosaurid dinosaurs. Square snouts have been correlated with non-selective feeding behavior in modern and extinct mammals and in hadrosaurine dinosaurs, whereas round snouts correlate with selective browsing behaviors in those taxa. Dental microwear features indicative of ground-height browsing on herbaceous plants correspond with square snouts in diplodocoid sauropods; microwear features suggestive of mid-height browsing on brittle, potentially woody plants correspond with round snouts.
There is a potential correspondence between body size and feeding behavior in diplodocoids: above 12–15 m body length, diplodocoids are exclusively non-selective, ground-height browsers; small diplodocoids include both selective and non-selective browsers, however. There is probably not a strong phylogenetic signal to morphology and behavior, although most diplodocids (except
Material examined for cranial reconstructions.
(DOC)
Teeth examined for microwear features. Teeth without position indicated were found isolated.
(DOC)
Complete list of microwear features (including dimensions, area, and type) for all teeth examined.
(XLSX)
Snout shape scores (PMI) for the hadrosaurids examined.
(DOC)
Jeffrey Wilson, Catherine Badgley, William Fink, and Tomasz Baumiller (UM) provided helpful comments on earlier versions of this paper. Paul Barrett, Andrew Farke, and an anonymous reviewer improved a subsequent draft. Matt Carrano (USNM), Ted Daeschler (ANSP), Nicole Klein (MB), Matt Lamanna (CM), Carl Mehling (AMNH), Paul Sereno (UC), and Phillipe Taquet (MNHN) provided access to specimens. Jeffrey Wilson (UM) provided images of