How the Red Queen Drives Terrestrial Mammals to Extinction

See allHide authors and affiliations

Science  19 Jul 2013:
Vol. 341, Issue 6143, pp. 290-292
DOI: 10.1126/science.1239431

Background Extinction

Diversity results through both the processes of species origination and extinction. However, studies of extinction have tended to focus on mass extinctions, despite the fact that the background extinction represents a greater loss in terms of the absolute number of extinct taxa. In order to identify what factors affect this rate of background extinction, Quental and Marshall (p. 290, published online 20 June) explored the dynamics of 19 mammalian clades and compared the rates of expansions and declines among taxa to expected models assuming random processes. Most clades decline to extinction in a “driven” manner—that is, faster than expected by chance alone.


Most species disappear by the processes of background extinction, yet those processes are poorly understood. We analyzed the evolutionary dynamics of 19 Cenozoic terrestrial mammalian clades with rich fossil records that are now fully extinct or in diversity decline. We find their diversity loss was not just a consequence of “gamblers ruin” but resulted from the evolutionary loss to the Red Queen, a failure to keep pace with a deteriorating environment. Diversity loss is driven equally by both depressed origination rates and elevated extinction rates. Although we find diversity-dependent origination and extinction rates, the diversity of each clade only transiently equaled the implied equilibrium diversity. Thus, the processes that drove diversity loss in terrestrial mammal clades were fundamentally nonequilibrial and overwhelmed diversity-dependent processes.

The majority of all species that have ever lived are now extinct (1), yet we know little about the dynamics of extinction. Most prior work has examined the mechanisms and selectivity of mass extinctions (2), although mass extinctions only account for a minority of extinctions in the history of life. To examine the dynamics of background extinction, we analyzed the generic diversity trajectories of 19 Cenozoic terrestrial mammalian families with excellent fossil records. To qualify for analysis, each family had to be monophyletic and have at least 100 genus occurrences (average = 419), a total diversity of at least five genera, a longevity of at least eight stratigraphic stages, and an average preservation potential of at least 0.6 per genus per stage (average = 0.89) (see supplementary materials). We only analyzed Cenozoic clades to avoid any complicating factors that might have been introduced by the end-Cretaceous mass extinction event.

We first tested the hypothesis that the observed diversity trajectories were a consequence of the loss to the Red Queen, that is, an evolutionary loss driven by the deterioration of the environment (36), against the alternative hypothesis that the waxing and waning in the clades’ diversity was random, the evolutionary equivalent of “gamblers ruin” (79). If the rise and fall in diversity was deterministic, then we would expect the longevities of the clades to be shorter than if their diversity trajectories were due to stochastic fluctuations in intrinsically constant rates of origination and extinction, where diversity would have simply drifted up and then down. Indeed, this is what we find: On average, the longevities of the clades are too short to simply be the result of stochastic processes [figs. S3 and S4; see also (10)], suggesting a deterministic component to the diversity dynamics.

Although the exact causes of the decline are hard to determine, we were able to characterize the dynamics responsible for the diversity trajectories. Quantitative analysis revealed four generalities. First, on average the duration of the rise phase is statistically indistinguishable from the duration of the decline phase—the diversity trajectories are temporally symmetrical [see (10) for how we dealt with clades that are not extinct]. A similar pattern has also been observed in Paleozoic marine invertebrates (11).

Second, both the per-genus origination rates and the extinction rates exhibit diversity dependence [fig. S8 and table S2; see also (10)]: When diversity increased, origination rates dropped and extinction rates increased. Thus, these mammalian clades exhibit the macroevolutionary equivalent of MacArthur and Wilson’s (12) model for diversity change during island colonization. In both scenarios, the existence of diversity-dependent rates implies that each island (in MacArthur and Wilson’s model) or clade (in the macroevolutionary equivalent of their model) has an equilibrium diversity, the diversity at which the origination rate equals the extinction rate. Diversity dependence in origination rates, but not in extinction rates, has also been reported in Cenozoic North American mammals [(13), but see (14)].

Third, we unexpectedly find that, during the decline phase, decreases in the per-genus origination rate are just as important as increases in the per-genus extinction rate in driving the observed diversity losses (Fig. 1). In fact, on average the initial origination rate is of a similar magnitude to the final extinction rate, and the final origination rate is as low as the initial extinction rate (Fig. 2). Most discussions of clade extinction focus only on the processes and rates of extinction and seldom consider the possibility that diversity can also be lost because of a failure to replace extinct taxa. However, Bambach et al. (15) showed that the loss in generic diversity in the end-Devonian and end-Triassic mass extinctions was primarily driven by a lack of origination. Similarly, Van Valen (3) noted that the decline in generic diversity of perissodactyl mammals was largely due to a drop in origination rate. The causes of a failure to originate, the evolutionary sterility that we call the Entwives effect (16), are not understood and require more attention.

Fig. 1 Changes in per-genus origination and extinction rates for events of increasing and decreasing diversity within the rise and decline phases of mammalian diversity trajectories.

Increases in diversity during the rise phases (A) are mostly controlled by increases in the per-genus origination rate (the change in the origination rate is significantly larger than the change in the extinction rate, N = 32, V (test statistic) = 438, P = 0.00071, Wilcoxon rank paired test). In contrast, increases in the per-genus origination rates and decreases in the per-genus extinction rates contribute equally to increases in diversity during the decline phases (B) (the magnitude of these changes is not significantly different, N = 11, V = 44, P = 0.3652, Wilcoxon rank paired test). Decreases in the per-genus origination rates and increases in the per-genus extinction rates also contribute equally to decreases in diversity during both rise (C) and decline (D) phases (the magnitude of these changes is not significantly different, N = 9, V = 18, P = 0.6523, Wilcoxon rank paired test, for the rise phases, and N = 42, V = 512, P = 0.4571, Wilcoxon rank paired test, for the decline phases). Changes in rates and diversity were measured between adjacent stages (fig. S5). The changes in the per-genus origination and extinction rates for each stage are connected by a line, and each pair constituted a replicate in the Wilcoxon rank paired test. Solid lines indicate that change in origination rate is more important than change in extinction rate in driving diversity change; dashed lines, that change in extinction is more important. The boxplots show the median and first and third quartiles of the data. The whiskers indicate the datum still within 1.5 interquartile range defined by the first and third quartiles. n.s., not significant.

Fig. 2 Origination rates decrease and extinction rates increase as the mammalian clades age.

There is a significant and roughly equal change in the average per-genus origination (A) and the average per-genus extinction (B) rates between the diversity-rise and the diversity-decline phases across the 19 families analyzed. Each pair of values corresponds to one of the analyzed families. For origination, N = 19, V = 184, P = 0.000053, Wilcoxon rank paired test; for extinction, N = 19, V = 25, P = 0.0033, Wilcoxon rank paired test. The crosses in the legend identify extinct clades. LMY, lineage million years.

Last, on average the overall diversity trajectories were more influenced by changes in origination rate than by changes in extinction rate (fig. S6). This disparity is due to the fact that changes in origination rate dominated the diversification phases of the diversity trajectories (Fig. 1), whereas changes in origination and extinction rates contributed equally during the decline phases. Similarly, Gilinsky and Bambach (17) found that family diversity within marine orders and suborders was largely driven by changes (decreases) in family origination rates.

The simplest way of modeling these observed diversity dynamics is with the macroevolutionary equivalent of MacArthur and Wilson’s model (12). However, this model leads to logistic diversification with a stable equilibrium diversity and thus requires modification to accommodate diversity loss. Whereas Whittaker et al. (18) provides a qualitative modification of the model that incorporates the formation and ultimate demise of oceanic islands with the extinction of their terrestrial biotas, we quantitatively extended MacArthur and Wilson’s framework to incorporate loss to the Red Queen by adding in a temporal decay in the intrinsic diversification rate, the diversification rate at the inception of the clade. We achieved this by decreasing the intrinsic origination rate and increasing the intrinsic extinction rate at constant rates with time (10), which translates into a constant rate of decay in the expected equilibrium diversity (Fig. 3 and eq. S13). Thus, under this model the expected equilibrium diversity steadily decays to zero and then becomes increasingly negative, driving the clade to extinction.

Fig. 3 How a clade loses to the Red Queen via a decay in its intrinsic per-genus rate of diversification.

(A to C) The change of the intrinsic origination and extinction rates (shown by the arrows); the decay of the equilibrium diversity (shown by the moving position of the dashed orange line); and the realized per-genus origination (blue points) and per-genus extinction (red points) rates at different times in its history. (D) The diversity trajectory generated by the diversity dynamics shown in (A) to (C). Light blue points show the diversity for the time points shown in (A) to (C). The graphs depict solution to eqs. S10, S11, and S17 (10). The running Red Queen symbolizes the deterioration of the environment. Myr, million years.

We began with a slow rate of decay in the intrinsic diversification rate (from 0.03% to 0.3% per million years) but could not generate diversity trajectories with temporally symmetrical waxing and waning phases nor the switch in the magnitudes of the initial origination and extinction rates with their final rates (10). Instead, clades quickly reached equilibrium diversity and then slowly rode the decaying equilibrium diversity down to extinction—the decline phase was longer than the diversification phase, and the final origination and extinction rates remained at intermediate values between the high initial origination rate and the low initial extinction rate, rather than switching in value.

The only way to accommodate the observed diversity dynamics is if the intrinsic diversification rate (and thus the equilibrium diversity) deteriorated sufficiently fast. For example, when we modeled the decay in the intrinsic diversification at a rate of ~3% per million years, the clade was left with a standing diversity that increasingly lagged behind the equilibrium diversity as the clade went extinct (Fig. 3C). This resulted in the switching in the values of the initial and final per-genus origination and extinction rates and led to a sufficiently negative diversification rate during the clade’s decline to produce the temporally symmetric waxing and waning phases of diversity change.

An unexpected consequence of the rapid decline in the per-genus rate of diversification is that a clade’s diversity only transiently equals the equilibrium diversity. In contrast, in typical diversity-dependent models, species diversity remains at or close to the equilibrium diversity after the initial radiation, even when the equilibrium diversity decays with time, for example, as in Whittaker et al.’s modeling of the disappearance of islands through erosion (18, 19). Under our model, the diversification phase involves a gain toward an equilibrium diversity, as in standard logistic growth. However, as diversity increases, the equilibrium diversity is decaying in response to an already deteriorating environment, and the clade reaches its peak diversity at an equilibrium value less than the initial equilibrium diversity. Then, as the clade moves into the decline phase, the decay in its intrinsic rate of diversification leads to a sufficiently rapid decrease in its equilibrium diversity that the clade’s realized diversity increasingly lags behind the decaying equilibrium diversity (Fig. 3). Thus, although diversity dependence in the per-genus origination and extinction rates plays a role in determining the duration of the clade’s history, the diversity dynamics is dominated by the decay in the intrinsic diversification rates, not by the diversity-dependent equilibrium processes.

The secondary role that diversity-dependent rates of origination and extinction play in the diversity dynamics of the mammalian clades in decline offers a resolution to a debate in the paleontological literature, where diversity dependence has been proposed (13, 20) but where the evidence of equilibrium is scarce (2123). In our model, the mechanism of diversity dependence is decoupled from the ultimate factors that determine the clades’ fates: the deterioration of their environment. Our results suggest that diversity dependence plays a role in diversity dynamics similar to the role that friction plays in the dynamics of motion—although it must be accounted for in the dynamics of diversity change, the dominant forces of diversity change lie beyond the existence of diversity dependence.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

Tables S1 and S2

References (2435)

References and Notes

  1. We used Van Valen’s original definition of the Red Queen as a measure of environmental deterioration regardless of the role that biotic and abiotic factors might have played in that deterioration (3). More recently, some have restricted the meaning of the Red Queen to biotic factors (5, 6), using the term Court Jester for abiotic factors (5, 6).
  2. The idea that the major patterns in Phanerozoic diversity change could be attributed to purely stochastic process was later rejected (9).
  3. Material and methods and supplementary materials can be found on Science Online.
  4. In J. R. R. Tolkein’s Middle Earth [J. R. R. Tolkien, The Lord of the Rings (Mariner Books, Boston, 2012)], the Ents lost their wives and thus had no means of regenerating their race, hence the term the Entwives effect.
  5. In Whittaker et al.’s model (18), their time axis is on a log scale.
  6. Acknowledgments: We thank all those who generated the mammal data as well as those who entered the data into the Paleobiology Database, especially J. Alroy, K. Behrensmeyer, A. Turner, M. Uhen, and M. Carrano. This is the Paleobiology Database publication number 178. We thank S. Finnegan, H. Morlon, and S. P. Quek for discussion. T.B.Q. thanks Fundação de Amparo à Pesquisa do Estado de São Paulo (2012/04072-3) and USP for funding. All of the data are available from the Paleobiology Database (
View Abstract

Navigate This Article