Evidence for mesothermy in dinosaurs

See allHide authors and affiliations

Science  13 Jun 2014:
Vol. 344, Issue 6189, pp. 1268-1272
DOI: 10.1126/science.1253143

Not too fast, not too slow, somewhere in between

In early depictions, dinosaurs lumbered slowly, dragging their tails. More recently, we have imagined them lifting their tails and running. The question boils down to whether dinosaurs had energetic systems closer to those of rapidly metabolizing mammals and birds, or to those of slower reptiles that do not internally regulate their body temperature. However, determining the metabolic rate of extinct organisms is no easy task. Grady et al. analyzed a huge data set on growth rate in both extinct and living species, using a method that considers body temperature and body size. Dinosaur metabolism seems to have been neither fast nor slow, but somewhere in the middle—so, dinosaurs did not fully regulate their internal temperature but they were also not entirely at the whim of the environment; neither slow goliaths nor supercharged reptiles.

Science, this issue p. 1268


Were dinosaurs ectotherms or fast-metabolizing endotherms whose activities were unconstrained by temperature? To date, some of the strongest evidence for endothermy comes from the rapid growth rates derived from the analysis of fossil bones. However, these studies are constrained by a lack of comparative data and an appropriate energetic framework. Here we compile data on ontogenetic growth for extant and fossil vertebrates, including all major dinosaur clades. Using a metabolic scaling approach, we find that growth and metabolic rates follow theoretical predictions across clades, although some groups deviate. Moreover, when the effects of size and temperature are considered, dinosaur metabolic rates were intermediate to those of endotherms and ectotherms and closest to those of extant mesotherms. Our results suggest that the modern dichotomy of endothermic versus ectothermic is overly simplistic.

Over the past few decades, the original characterization of dinosaurs by early paleontologists as lumbering, slow-metabolizing ectotherms has been challenged. Recent studies propose that dinosaurs were capable of an active lifestyle and were metabolically similar to endothermic mammals and birds (13). This debate is of more than heuristic interest; energy consumption is closely linked to life history, demographic, and ecological traits (4). Extant endothermic mammals and birds possess metabolic rates ~5 to 10 times higher than those of reptiles and fish (5, 6), but characterizing the metabolic rates of dinosaurs has been difficult.

A promising method for inferring paleoenergetics comes from studies of ontogenetic growth, in which age is determined from annual rings in bone cross sections and mass is determined from bone dimensions. Ultimately, growth is powered by metabolism, and rates of growth and energy use should correspond. Pioneering work by Erickson and others has led to a growing body of literature on dinosaur growth and generated important insights (7, 8). However, many analyses were hampered by small samples, an outdated comparative data set, and the lack of an appropriate energetic framework. Increasing data availability permits a reassessment of dinosaur growth against a broader spectrum of animals, standardized for environmental temperature. Further, recent advances in metabolic theory provide a theoretical framework for evaluating metabolic rate on the basis of growth.

We used a comparative approach to characterize the energetics of dinosaurs and other extinct taxa. We examined the empirical and theoretical relationship between growth and resting metabolic rate, using a broad database of major vertebrate clades (9), and used our results to examine the energetics of Mesozoic dinosaurs. From empirical studies, we constructed ontogenetic growth curves and determined a maximum rate of growth for each species. Environmental temperature was standardized by only considering growth rates in ectotherms from tropical and subtropical climates or from laboratory settings between 24° and 30°C, comparable to temperatures experienced by dinosaurs during the Mesozoic (10). Data for dinosaur growth were taken from published reports that provided a minimum of five measurements of size and age. All metabolic rates were converted to watts (W). Where multiple metabolic or maximum growth rates for a species were recorded, the geometric mean was determined. Overall, our data set includes ~30,000 values and was used to characterize growth for 381 species, including 21 species of Mesozoic dinosaurs, 6 extinct crocodilians, and a Cretaceous shark (table S1). Dinosaurs are well represented both temporally (late Triassic to end-Cretaceous) and taxonomically (Theropoda, Sauropodomorpha, Ornithopoda, and Ceratopsia). Values for resting metabolic rates were compiled from the literature and standardized to a common temperature of 27°C (table S1). We performed phylogenetic independent contrasts (PICs) in addition to conventional ordinary least-squares regression (OLS) and standardized major axis regression (table S2).

Data show, within and across species, that resting metabolic rate B scales with body mass m as a power function, B = B0mα, where B0 is a normalization constant representing mass-independent metabolic rate, and α is ~3/4 and ranges from 0.65 to 0.85 (11, 12). Growth rate varies over ontogeny, but use of the maximum growth rate (Gmax) standardizes growth and permits interspecific comparisons. Empirical evidence (13) indicates that Gmax scales similarly to B, where Gmax = G0Mα. This suggests that BGmax1 and thus that metabolic rate may be inferred from growth. However, the relationship between Gmax and B across major vertebrate taxa has received little attention, and many uncertainties exist. For instance, Case (13) reported that fish Gmax was an order of magnitude lower than that of reptiles, despite similarities in metabolic and thermoregulatory lifestyle (6).

Theoretical assessments of growth complement a strictly empirical approach and can strengthen paleontological inferences. An ontogenetic growth model based on metabolic scaling theory (MST) quantifies the linkages between Gmax and metabolic rate from first principles of allometry and conservation of energy (14, 15). According to MST (9), the relationship between B (W) and Gmax (g day−1) at final adult mass M isBM = cGmax1(1)where c ≈ 0.66 (W day1 g−1 day). To observe the mass-independent relationship and compare energetic groups, we divide both sides by Mα, yielding

B0 = cG0(2)

To calculate metabolic rate at any ontogenetic mass m from the observed maximum growth rate, we combine Eqs. 1 and 2

Bm = cG0m3/4(3)

MST makes the following theoretical predictions regarding growth and metabolic rate:

(1) Gmax scales as Mα, where α ~3/4.

(2) B scales isometrically with Gmax if masses are standardized (9). Regression of B against Gmax yields a slope of 1 and an intercept of ≈0.66.

(3) Plotting G0 against B0 will reveal distinct energetic clusters corresponding to endotherms and ectotherms. High-power endotherms will exhibit an elevated G0 and B0, and ectothermic organisms the converse. Thermally intermediate taxa, termed mesotherms, such as tuna and lamnid sharks (16), should fall between the upper and lower quadrats. The predicted slope and intercept are 1 and 1.52, respectively. Similar clustering is observed if Gmax and B residuals are plotted.

(4) Bpredicted = Bobserved in extant animals, where Bpredicted is calculated from Eq. 3.

Our analyses find broad support for all four predictions. First, growth scales with mass as ~3/4, although taxonomic variation is observed (Fig. 1 and fig. S1, mean αOLS = 0.73; mean αPIC = 0.69, table S2). This indicates that larger species acquire their bulk by accelerating their maximum growth rate proportionate to ~M3/4. Second, Gmax is a strong predictor of B, where BM = 0.56Gmax1.03, which is close to theoretical predictions [figs. S3 and S4; slope confidence interval (CI) = 0.97 to 1.10; intercept CI = 0.47 to 0.97; coefficient of determination (r2) = 0.90, n = 118]. Third, we find that the observed relationship between mass-independent growth and metabolic rates corresponds closely to predicted values (slope = 0.90, CI = 0.77 to 1.03; intercept = 1.10, CI = 0.59 to 2.06, r2 = 0.61, n = 124). Ectothermic species fall in the lower left quadrat; endotherms in the upper right; and thermally intermediate taxa, including tuna, a lamnid shark, the leatherback turtle, and a prototherian mammal, fall between values for endo- and ectotherms (Fig. 2 and figs. S1, S2, and S5). These results are robust; the inclusion of cold-water fish, with reduced growth and metabolic rates, simply extends the lower portion of the regression line. Furthermore, the ratio G0/B0 (g J−1), a measure of efficiency in converting energy to biomass, does not differ significantly between endo- and ectotherms, indicating that energy allocation to growth does not vary with thermoregulatory strategy (t statistic = 0.46, P = 0.64, fig. S6). Finally, regression of observed against calculated metabolic rates does not differ significantly from unity (Fig. 3A; slope CI = 0.97 to 1.10; intercept CI = –0.14 to 0.02). We can therefore predict dinosaur resting metabolic rates from growth rate, using either a theoretical model (Eq. 3) or an empirically determined equation (9)

Fig. 1 The scaling of maximum growth rate in vertebrates.

(A) Growth rates of thermoregularoty guilds. Red indicates endothermy; blue, ectothermy, gray, dinosaurs; and black, mesothermy. (B) Vertebrate taxa scaling with 95% confidence bands. The red dashed line indicates marsupials, and the black dashed line is tuna; all other taxa are labeled. See table S2 for regression parameters and statistics.

Fig. 2 Vertebrate growth energetics.

(A) Relationship between growth and resting metabolic rate for vertebrates. The dashed line is the theoretical prediction; the solid line represents an OLS fitted regression with 95% confidence bands. (B) Predicted energetics of dinosaurs. Dinosaur rates (open squares) from Eq. 2 are plotted on the theoretical line. The ranges in metabolic rates occupied by extant endotherms, mesotherms, and ectotherms are indicated by color.

Fig. 3 Resting metabolic rates in vertebrates.

(A) Predicted metabolic rates compared to observed rates. The solid line is the fitted regression, with shaded 95% confidence bands; the dashed line is the theoretical fit. (B) Metabolic scaling of vertebrates. Dinosaur resting metabolic rates are predicted from growth (dashed line); all other fits are predicted from empirical data. Endotherms: y = 0.019x0.75, r2 = 0.98, n = 89; Ectotherms (27°C): y = 0.00099x0.84, r2 = 0.95, n = 22; Dinosaurs: y = 0.0020x0.82, r2 = 0.96, n = 21. P < 0.001 for all regressions.

BM = 0.6 Gmax(4)

Our analyses are robust to variation in the scaling exponent, phylogenetic correction, inclusion of captive versus wild animals, critiques of dinosaur growth studies, and uncertainty in estimating M and metabolic temperature (9).

Our results find that mass-independent growth rates in dinosaurs were intermediate to, and significantly different from, those of endothermic and ectothermic taxa (table S2). Although some dinosaur growth rates overlap with high-power ectotherms or low-power endotherms, they cluster closest to energetically and thermally intermediate taxa, such as tuna (Fig. 2). Further, our analyses uphold the somewhat surprising finding that feathered dinosaurs, including protoavian Archaeopteryx (17), did not grow markedly differently from other dinosaurs (Fig. 4). It appears that modern avian energetics did not coincide with feathers or flight, which is consistent with fossil evidence that modern bone histology in birds did not appear until the late Cretaceous (18).

Fig. 4 Phylogeny of mass-independent growth rates (g1/4 day–1).

Color signifies thermoregulatory state; branch lengths are not standardized for divergence times. Green shading indicates feathered coelurosaurian dinosaurs.

At the largest body masses, the growth rates of the largest dinosaurs and mammals overlap (Fig. 1B). This pattern is driven by two factors. First, dinosaurs have a relatively high slope (αOLS = 0.82, but αPIC = 0.76). This value is consistent with suggestions of thermal inertia for larger taxa; the removal of sauropods yields a reduced OLS slope of 0.77. Second, significantly reduced growth rates are observed in several large mammalian taxa, particularly primates, elephants, and toothed whales, whereas small shrews and rodents have relatively high rates, leading to a low overall slope for placental mammals (αOLS = 0.64, αPIC = 0.63; table S2 and fig. S11). The slow growth of many large endothermic mammals is associated with large brain size and low juvenile mortality (19, 20); this is unlikely to be relevant to most dinosaurs.

Our results highlight important similarities and differences from previous studies. For example, our work agrees with assessments by Erickson (7, 17) that dinosaurs grew at rates intermediate to most endo- and ectotherms. However, we find considerably more similarity in ectothermic growth rates than reported by Case (13) and significantly higher growth rates for fish (~seven times higher), marsupials (~four times higher) and precocial birds (~two times higher; fig. S8). We attribute these differences to enhanced sampling and standardization of the thermal environment for ectotherms (e.g., Case included temperate fish). Moreover, our expansion of the comparative growth framework indicates that dinosaurs grew and metabolized at rates most similar to those of active sharks and tuna (Fig. 2 and fig. S1), rather than those of endothermic marsupials, as has been suggested (17).

Past work has often struggled to fit dinosaurs into a simple energetic dichotomy; our work suggests that an intermediate view (17, 21) is more likely. Although dinosaur growth rates vary, they cluster most closely to those of thermally intermediate taxa (Figs. 1 and 2), which we term mesotherms. Mesothermic tuna, lamnid sharks, and the leatherback turtle rely on metabolic heat to raise their body temperature (Tb) above the ambient temperature (Ta) but do not metabolically defend a thermal set point as endotherms do (16, 22). This reliance on metabolic heat distinguishes them from other large homeothermic reptiles, such as crocodiles (23), which bask to elevate Tb. The echidna, while maintaining a set point of ~31°C, shows remarkable lability, because Tb values can range over 10°C while it is active (24). Unlike hibernating mammals or torpid hummingbirds, this variability is externally imposed. Collectively, these animals are distinguished from endotherms and ectotherms by a weak or absent metabolic defense of a thermal set point but sufficient internal heat production to maintain Tb > Ta when Ta is low [see (9) for further discussion]. Although some feathered dinosaurs may have been endotherms, they would have been uniquely low-powered compared to extant birds and mammals. We suggest that mesothermy may have been common among dinosaurs, ranging from modest metabolic control of Tb, as seen in furred echidnas, to the absent metabolic defense observed in tuna and leatherback turtles. Analysis of fossil isotopes, which can shed light on body temperatures, will be useful in testing this hypothesis. In particular, attention to neonate and juvenile dinosaurs in seasonally cool environments, such as polar regions, may help distinguish among thermoregulatory states.

Dinosaurs dominated the flux of matter and energy in terrestrial ecosystems for more than 135 million years. Consequently, our results have important implications for understanding ancient Mesozoic ecosystems. We emphasize the primary importance of comparative energetics for integrating form, function, and diversity. Knowing only two facts from the fossil record—adult mass and maximum growth rate—we show that the metabolic rates of extinct clades can be predicted with accuracy. Such an approach will be useful in resolving the energetics of metabolically ambiguous taxa, such as pterosaurs, therapsids, and Mesozoic birds.

Supplementary Materials

Figs. S1 to S15

Tables S1 to S4

References (25396)

References and Notes

  1. See the supplementary materials.
  2. Acknowledgments: This work was supported by a fellowship from the Program in Interdisciplinary Biological and Biomedical Sciences at the University of New Mexico (grant no. T32EB009414 from the National Institute of Biomedical Imaging and Bioengineering to F.A.S. and J. H. Brown). B.J.E. was supported by an NSF CAREER and ATB Award (EF 0742800). We thank C. White and two anonymous reviewers for valuable feedback on our manuscript and N. Milan for assistance with the figures. Data are available in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article