Arboreality has allowed for the evolution of increased longevity in mammals

Overall, in agreement with previous studies (27, 28), body mass accounts for 60% of the variance in maximal lifespan in nonvolant, nonaquatic mammals. Longevity, like many life-history traits, is negatively allometric, so that lifespan increases at between one-fourth and one-third the rate of body mass (Fig. 2). An initial analysis on the entire dataset indicates that the slope for semiarboreal species is significantly different from the arboreal and terrestrial slopes. A separate analysis of Eutheria alone results in common slopes for all three groups (see below; Table 1), suggesting that the departure of the semiarboreal slope results from the influence of marsupials. Because of this result, semiarboreal species were removed for analysis on the overall dataset, and an ANCOVA was conducted on the remaining habitat groups (arboreal and terrestrial). This analysis demonstrates that, at common body sizes, arboreal mammals are characterized by greater longevities than terrestrial mammals (P < 0.001) (Table 1 and Fig. 2). Results of the PIC analysis on the mammalian dataset also are significant (P < 0.001), indicating the effect of habitat type on longevity remains even when the effects of phylogeny are removed.

Results of subsequent ANCOVA are listed in Table 1. In 8 of the 10 subclades included in the analysis, arboreal mammals are characterized by greater longevities than terrestrial mammals (Table 1). Semiarboreal eutherian mammals are either intermediately long-lived or are not significantly different from one or the other habitat type (Table 1 and Fig. 3). Results for two subclades, Metatheria (marsupials) and Euarchonta (primates and their close relatives), differ from the overall results. Within Metatheria, although arboreal and terrestrial marsupials are not significantly different from each other, both are significantly longer-lived than semiarboreal taxa (Table 1). Our results therefore confirm a previous finding that arboreality is not associated with increased longevity in marsupials (29). In Euarchonta, our analyses did not reveal significant differences between any of the habitat types (Table 1). However, taken together, the overall results demonstrate that there is a general, significant relationship between habitat type and longevity in mammals. That is, arboreal mammals have evolved greater longevities than terrestrial mammals, and semiarboreal mammals tend to fall intermediately between these two groups. Even in the most specific clade we analyzed, the superfamily Sciuroidea, arboreal squirrels are longer-lived than terrestrial squirrels at common body sizes (P < 0.001; Fig. 3).

Previous authors have suggested that primates are particularly long-lived among mammals, in part because they are largely arboreal (25, 26). However, among both euarchontans and marsupials, arboreal species do not possess greater longevities than terrestrial species. Arboreality is the primitive condition for primates, initially evolved in an ancestral euarchontan (30 –35). Marsupials also probably are primitively arboreal (34, 36 –38), with subsequent and repeated events of terrestriality (38). In contrast, many other eutherian orders and placental mammals in general are likely characterized by terrestrial evolutionary histories, with subsequent events of arboreality derived multiple times in these lineages (34, 39). (See ref. 34 and sources therein for a recent and thorough review of this topic.) The nature of these dichotomies may play out in the evolution of longevity. We propose that the lack of significant differences between arboreal and terrestrial euarchontans, and possibly marsupials, may be the result in part of the long history and persistence of arboreality in their evolutionary histories (30 –38). Unlike euarchontans, however, marsupials are not characterized by increased longevity in general (Fig. 4). Compared with marsupials, the extremely high degree of arboreality throughout the evolution of Euarchonta, and specifically within primates, may explain this discrepancy. A highly arboreal evolutionary history may have allowed increased longevity in all primates.

In addition to a long arboreal evolutionary history, it has been suggested that terrestrial primates may not experience increased extrinsic mortality because they have evolved physiological, social, and behavioral defenses against predation (40, 41). First, ground-dwelling primates have increased body size upon transitioning to a terrestrial lifestyle. Unlike other groups of mammals (e.g., carnivores; ref. 42), body mass is significantly correlated with habitat type in primates (43). It is likely that the larger body size of terrestrial primates decreases their susceptibility to predation, thus lowering the risk of spending time on the ground (1, 40, 41, 43, 44). Predation on small-bodied, arboreal primates by birds of prey and arboreal carnivores and leopard predation on primates both on the ground and in the trees may render arboreal and terrestrial primates equally susceptible to predation (40, 41). Second, primates are highly social, and the advantages of group living provide social defenses such as large group size, vigilance, and alarm calls (40). Terrestrial primates live in larger social groups than arboreal primates (40, 41, 43, 44), thus supporting this hypothesis. Humans, the most terrestrial of all primates, have reduced extrinsic mortality and increased longevity resulting from the obvious advantages provided by sociality and culture.

Similar arguments may be made for some marsupial species. Terrestrial marsupials that occupy open habitats are generally large and capable of fast speed (e.g., macropodids). In addition, they live in large social groups and demonstrate vigilance behavior, contrasting with the lack sociality of marsupials in general (45 –47). These factors might contribute to decreased predation risks associated with terrestriality and explain the lack of significant difference in longevity between arboreal and terrestrial marsupial species. However, as noted previously, marsupials are not long-lived in general, regardless of habitat type. This observation suggests that other confounding factors, such as development, brain size, or various aspects of physiology may significantly influence longevity in these species. We therefore are more cautious in our interpretation of marsupial longevity.

Many evolutionary transitions to arboreality among mammals have resulted in increased longevity, demonstrated by longer-lived arboreal taxa in the majority of mammalian clades. As with flight and gliding behavior, arboreality appears to have reduced extrinsic mortality and allowed increased longevity in many groups of placental mammals. Future studies should test this hypothesis by comparing mortality rates among closely related species that occupy different habitats. The highly arboreal evolutionary history of Euarchonta, and especially primates, may have allowed increased longevity in the group as a whole. This hypothesis will require further testing, as will other mechanisms that might influence extrinsic mortality, and therefore longevity, including body size, sociality, behavioral and morphological defenses, and adaptation to other habitat types (e.g., fossoriality). In summary, results of this study support the hypothesis that primates and other arboreal mammals are characterized by greater longevities than terrestrial mammals and confirm the fundamental predictions of the evolutionary theory of aging.

Analyses were conducted on published data on 776 species, representing 24 orders from all major divisions of Mammalia (Fig. 1). The dataset for longevity was compiled from published maximal lifespan records (48, 49). Both captive and wild records are included; in fact, captive records are preferred, because maximum potential lifespan is of interest. Average adult body masses and habitat categories (arboreal, semiarboreal, or terrestrial) were obtained from various sources (48, 50 –53). Although brain mass was once argued to be a better predictor of longevity than body mass in mammals (27), it has been demonstrated that other body organ masses are equally or more highly correlated with longevity, suggesting that organ mass may be a better proxy for body size than body mass (25, 28, 54). However, data for body mass are more widely available than data for various organ masses; therefore, we use body mass as a covariate in our study. Where quantitative data are available, the following criteria are used to determine habitat type: 75–100% time spent in trees is considered arboreal, 25–75% time spent in trees is considered semiarboreal, and 0–25% time spent in trees is considered terrestrial. However, for most species examined, this type of information is not available (although many cases are obvious, such as many “ungulates,” which are strictly terrestrial). For these species, categories are determined based on qualitative descriptions. Those species described as “arboreal/primarily arboreal” and “terrestrial/primarily terrestrial” are classified arboreal and terrestrial, respectively; those described as “semiarboreal,” “semiterrestrial,” ”occasionally arboreal,” or ”arboreal and terrestrial” are classified semiarboreal.

We constructed a mammalian supertree based on recently published molecular phylogenies (Fig. 1 and Fig. S1). To control for the effects of shared phylogenetic history, two methods were employed. First, several taxonomic groups were analyzed independently using ANCOVA to determine whether overall results were significantly influenced by one or a few subgroups. Taxonomic groups were chosen so that they contained sufficient numbers of arboreal and terrestrial species to conduct an ANCOVA. In some cases (e.g., Xenarthra), assumptions of ANCOVA are not met (slopes are not different from zero), so there is no linear relationship between body mass and longevity, presumably because of inadequate sample size for one or more habitat categories. Minimally, this analysis required at least three taxa in each category, although in most cases the number was much higher (Table 1). Furthermore, taxonomic groups were chosen to minimize redundancy. For example, the group primates is not analyzed separately from Euarchonta because only three nonprimate euarchontans are included in this dataset, and their inclusion (or exclusion) does not affect the results. Likewise, Carnivora and Rodentia are not analyzed separately but rather are subsumed within the larger clades Ferae (Carnivora + Pholidota) and Glires (Rodentia + Lagamorpha). In our dataset, 10 subclades are nonredundant and meet minimal sample size requirements (Table 1). Second, PIC (55, 56) were calculated and analyzed on the entire dataset using the AOT function in Phylocom (57) to determine the significance of habitat category on longevity and body mass after the influence of phylogeny is removed (29).

Ordinary least squares (OLS) regression lines were fitted to the total dataset and to the 10 relevant subgroups. ANCOVA tests were conducted on the total dataset and on the subclades to determine whether intercepts for habitat type are significantly different from each other using Bonferroni posthoc tests. Because phylogenetic relationships can obscure the independence of data points (55, 56), the potential effects of phylogeny were tested using the AOT function in Phylocom (57) to calculate and analyze PIC. The AOT function is used to create PIC between continuous and categorical traits. Because it is designed to work with a binary categorical variable, only arboreal and terrestrial taxa are used in this portion of the analysis. Following Fisher and colleagues (29), we first calculated a set of PIC using only the two continuous variables (body mass and longevity). Using these contrasts, we calculate the slope of the linear regression, m, with the intercept forced through the origin. The equation for the regression line is then y = mx. Next, a second set of PIC are calculated by contrasting body mass and longevity against habitat type, creating body mass contrasts relative to habitat (M_R) and longevity contrasts relative to habitat (L_R). The original regression line obtained using only continuous variables (y = mx) then is applied to the second set of contrasts to calculate residuals. This calculation is accomplished by subtracting the longevity contrasts relative to habitat (L_R) from the product of the original slope (m) and the body mass contrasts relative to habitat (M_R). That is, residuals are calculated as mM_R − L_R. These residuals then are subjected to a one-way t test to determine if they are significantly different from zero. A significant result (P < 0.05) indicates that habitat type has a significant effect on longevity relative to body mass. Because two aspects of our phylogeny are controversial, namely Atlantogenata (58, 59) and Pegasoferae (60), we analyze independent contrasts using modified phylogenies that do not recognize sister relationships between Afrotheria and Xenarthra or Perrisodactyla and Ferae, in addition to the main phylogeny presented in this analysis (Fig. 1 and Fig. S1).

We thank Steve Leigh, who encouraged, and whose courses inspired, this project. We also thank Greg Blomquist, John Polk, and Charles Roseman for valuable advice and discussions. Comments provided by two anonymous reviewers greatly improved the manuscript. This work was funded by a Cognitive Science/Artificial Intelligence Grant from the Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana–Champaign.