Int. J. Plant Sci. 169(7):863–870. 2008. Ó 2008 by The University of Chicago. All rights reserved. 1058-5893/2008/16907-0004$15.00 DOI: 10.1086/589888 EVOLUTION OF SEED SIZE AND BIOTIC SEED DISPERSAL IN ANGIOSPERMS: PALEOECOLOGICAL AND NEOECOLOGICAL EVIDENCE Ove Eriksson1 Department of Botany, Stockholm University, SE-106 91 Stockholm, Sweden Origins of biotic seed dispersal by vertebrates and evolution of different seed sizes are conspicuous features in angiosperm evolution. In the Cretaceous, angiosperms remained small seeded, and biotic seed dispersal was rare. In the Early Tertiary, average seed size increased drastically, and biotic seed dispersal by vertebrates became common. Later in the Tertiary, along with climate cooling and the opening of vegetation, average seed size dropped. Fossil data suggest a positive correlation between seed size and biotic seed dispersal. This article examines three hypotheses: (1) the coevolution hypothesis, which suggests that evolution of large seeds and biotic dispersal were driven by interactions between plants and frugivorous vertebrates; (2) the recruitment hypothesis, which suggests that vegetation change altered recruitment conditions that favored large seeds and in turn promoted biotic dispersal; and (3) the life-form hypothesis, which suggests that large seeds and biotic dispersal evolved as coadapted traits along with evolution of large plant life-forms. The hypotheses are complementary rather than mutually exclusive. Evidence suggest that changes in vegetation structure (open vs. closed) are probably a primary driving mechanism of seed size and dispersal evolution, but life-forms, seed sizes, and biotic dispersal systems have evolved as coadapted sets of traits in response to these environmental changes. Keywords: coevolution, fleshy fruits, plant size, seed size evolution, seedling recruitment. Introduction The Problem: Seed Size and Biotic Seed Dispersal from the Cretaceous to the Present Angiosperms have evolved a tremendous variation in seed size and seed dispersal systems. Seed sizes range from minute dust seeds, such as in orchids and weighing a few micrograms, to giants such as the nuts of some palms, weighing more than 10 kg (Leishman et al. 2000). Although seed dispersal is often unassisted, that is, with seeds lacking structures adapted to specific dispersal agents, many plants have evolved means to promote seed transport by external agents—by water, by wind or animals, and most commonly, by mammals, birds, and ants. This article synthesizes paleoecological and neoecological evidence relevant to the evolution of seed size and its association with vertebrate-mediated seed dispersal, that is, dispersal of seeds after animals have collected or ingested fruits (henceforth, ‘‘biotic dispersal’’). Most of the vertebrate-dispersed fruits are fleshy, that is, containing a juicy fruit pulp, but there are also biotically dispersed fruits that are leathery or dry, such as nuts. The term ‘‘fleshy fruits’’ also incorporates seed dispersal systems where the attractive fleshy tissue is not morphologically a part of the fruit (e.g., strawberries). More specifically, I address the question of how to explain a broad pattern of evolutionary changes in seed size and in biotic seed dispersal by vertebrates from observations of paleofloras sampled from the Early Cretaceous (age ;124 Myr) to the Pliocene (age ;2 Myr). 1 Angiosperms first appeared in the fossil record in the Early Cretaceous around 130 Ma (Crane et al. 1995). In contrast to the extraordinary variation in contemporary angiosperm seed size and dispersal systems, there is no substantial variation in seed size and dispersal systems during approximately the first half of angiosperm existence. Tiffney (1984), later corroborated by Eriksson et al. (2000a), concluded that during most of the Cretaceous, angiosperm seeds were generally small (typically ;1 mm3), and seed dispersal was mostly unassisted. One exception is from one of the oldest known floras from Famalicão, Portugal (Eriksson et al. 2000b), where ;25% of the species had small drupes or berries. However, overall, there is little evidence of biotic seed dispersal in angiosperms before the Late Cretaceous (Wing and Tiffney 1987b; Wing and Boucher 1998; Tiffney 2004). From the Late Cretaceous, and especially after the Cretaceous-Tertiary (K/T) boundary (65 Ma) in the Paleocene (;65–55 Ma) and the Eocene (;55–38 Ma), there was a dramatic increase in both seed size and range of seed sizes (fig. 1). The peak occurred in the Early to Middle Eocene (;55–42 Ma). In the Oligocene (;38–25 Ma), the Miocene (;25–5 Ma), and the Pliocene (;5–2 Ma), average seed size dropped, but the seed size range remained more or less the same (varying between three and six orders of magnitude). As a comparison, the seed size range in extant floras typically is around five to six orders of magnitude (Leishman et al. 1995). Seed dispersal systems are not randomly distributed among plants with different seed size. In species with biotic seed dispersal, excluding seed dispersal by ants and adhesive seed dispersal externally on animal fur and feathers, seeds are typically in the upper E-mail: ove.eriksson@botan.su.se. Manuscript received December 2007; revised manuscript received April 2008. 863 864 INTERNATIONAL JOURNAL OF PLANT SCIENCES (3) Why did seed size drop after the Eocene? and (4) Why is there a positive relationship between seed size and biotic seed dispersal? Wing and Tiffney (1987b) made a thorough summary of the changing interactions between angiosperms and herbivores from the Cretaceous to the Tertiary. They proposed a model in which the varying impact of large generalized herbivores was seen as a key driver of the interaction and the resulting changing patterns of angiosperm seed and fruit traits and vegetation. After the demise of large dinosaur herbivores at the K/T boundary and the subsequent radiation of small, arboreal, frugivorous birds and mammals, there was a major shift in the selection regimes in angiosperms, favoring larger seeds, biotic dispersal, and development of closed vegetation. Wing and Tiffney (1987b) suggested that several mechanisms were involved in this process. Without the effect of large generalized herbivores, competition among plants increased, leading to denser vegetation and in turn promoting larger plants with larger seeds. The interactions between plants and small animals resulted in a ‘‘coevolutionary spiral’’ in which increasing abundance of seeds and fruits as resources for frugivores stimulated frugivory that favored plant dispersal. The coevolution hypothesis, the recruitment hypothesis, and the life-form hypothesis will be examined in this article. These focus on different specific mechanisms that are all included in (and compatible with) the general model by Wing and Tiffney (1987b). The first two hypotheses have been inferred post hoc from the paleoecological information that they were intended to explain, whereas the third hypothesis derives from a combination of comparative studies of seed size and predictions from life-history theory. Possible Solutions: Three Hypotheses Fig. 1 Median seed size (A) and seed size range (orders of magnitude; B) in 25 paleofloras from the Early Cretaceous to the Pliocene (redrawn from Eriksson et al. 2000a). The ideas behind the coevolution hypothesis were discussed by Tiffney (1984) and Wing and Tiffney (1987b) but were suggested explicitly by Tiffney (2004). Although the model developed part of the seed size range, most common in plants with seeds heavier than 0.1 g, and biotic dispersal is dominating in the seed size range above 1 g (Hughes et al. 1994). However, there are examples of biotically dispersed plants with smaller seeds, particularly in berries (e.g., Vaccinium, Ericaceae). Biotic seed dispersal is generally most frequent in forest communities, especially in the tropics, where it is sometimes a totally dominating seed dispersal system among woody plant species (Fleming et al. 1987; Willson et al. 1990). On the basis of 25 paleofloras examined by Eriksson et al. (2000a), there is evidence of a strong positive relationship between seed size and the fraction of species with biotic seed dispersal (fig. 2). Such a strong relationship between seed size and the occurrence of biotic dispersal has been questioned by other studies, and I will return to this later; however, viewed from the paleoecological evidence, the changes in angiosperm seed size over the past 130 Myr are mirrored by changes in the frequency of biotic seed dispersal. On the basis of figures 1 and 2, I addressed four questions about the evolution of angiosperm seed size and dispersal systems: (1) Why were seeds generally small and biotic seed dispersal rare during most of the Cretaceous? (2) What caused the dramatic increase in seed size during the Early Tertiary? Fig. 2 Relationship between the fraction of species with biotic seed dispersal (mainly ‘‘fleshy fruits’’) and the median seed size in 25 paleofloras from the Early Cretaceous to the Pliocene (data from Eriksson et al. 2000a; r ¼ 0.844, P < 0.001). ERIKSSON—EVOLUTION OF SEED SIZE AND BIOTIC SEED DISPERSAL by Wing and Tiffney (1987b) was based on several mechanisms, Tiffney (2004, p. 15) emphasized the key role of the interactions between plants and frugivores: ‘‘A clear upsurge occurred in the importance of biotic dispersal in the Tertiary. Central to this are birds and mammals that, because of their small size, have organ-level interactions with plants . . . Wing and Tiffney (1987a, 1987b) and Tiffney (1992) postulated that the radiation of small vertebrates led to the rise in dispersal and the subsequent establishment of closed-canopy forest.’’ Tiffney (2004) argued that the key event that triggered the radiation of angiosperm seed sizes and dispersal systems was the simultaneous radiation of frugivores during the Early Tertiary. Mammals such as primates, rodents, and bats, common presentday frugivores, and several groups of potentially frugivorous birds radiated during the Paleocene and the Eocene. According to Tiffney (2004), this new arena for plant-animal coevolution allowed plants to evolve larger seeds, which were suitable for establishing offspring under closed canopies. This process promoted the establishment of closed forests, known to have developed during the Paleocene and the Eocene. The coevolution hypothesis implies a causal chain: the radiation of frugivores!coevolution with plants!evolution of larger seeds and biotic seed dispersal!development of closed vegetation. The paucity of suitable seed-dispersing vertebrates is the reason seed size remained small and biotic seed dispersal was rare during the Cretaceous. The radiation of frugivorous mammals and birds caused the increase in seed size and the association between large seeds and biotic dispersal. However, on the basis of the coevolution hypothesis, it is difficult to explain why seed size dropped after the Eocene. All the prominent groups of frugivores were present during that time and are still present in extant faunas. The recruitment hypothesis was also discussed by Tiffney (1984) and Wing and Tiffney (1987b) but was suggested explicitly by Eriksson et al. (2000a). According to this hypothesis, the key event was the change in vegetation composition during the Early Tertiary. In contrast to the more open vegetation dominating during the Cretaceous, an ecologically new kind of environment was established in the Tertiary: closed, probably multistratal forests dominated by angiosperms. According to Wing and Sues (1992, p. 389), ‘‘This appears to be the most dramatic structural shift to take place in terrestrial ecosystems from their full establishment in the mid-Permian to the present day.’’ The main drivers of this change were most likely climate, increasing temperature, and precipitation, but the expansion of forests may also have been promoted by a general lack of large herbivores after the demise of the dinosaurs at the CretaceousTertiary boundary (Wing and Tiffney 1987b; Collinson and Hooker 1991). A closed vegetation implies a strong competition for light at ground level, favoring evolution of increasing seed size. Because of a trade-off between seed size and seed number (for a plant of a given size allocating a given amount of resources to seeds), an increase in seed size means fewer seeds. With a given seed dispersal shadow, fewer seeds from the mother plant means that fewer offspring can be established some distance away from the mother plant (Eriksson and Jakobsson 1999). Thus, increasing seed size implies a cost in terms of reduced dispersal because the departure-related benefits of dispersal (Herrera 2002) are reduced. Vertebrates are the only vectors capable of transporting large seeds in terrestrial envi- 865 ronments, and biotic seed dispersal was therefore favored. Evolution of specialized frugivores, according to this hypothesis, is not a cause but an effect of evolution of large seeds. Presence of fruit-eating animals is a prerequisite, but they may initially not have been specialized frugivores. When vegetation became more open again during the Oligocene, the Miocene, and the Pliocene, selection favored small-seeded plants under conditions where recruitment was not light limited. Tiffney (2004, p. 19) summarized the difference between the coevolution and recruitment hypotheses as follows: ‘‘Tiffney (1984) and Wing and Tiffney (1987a, 1987b) suggested that the radiation of birds and mammals established dispersal agents scaled to move organs, which, in turn, allowed larger disseminule size and, ultimately, the establishment of closed forests with low-light germination regimes. Eriksson et al. (2000a) countered that these relations were inverted and that a warmer climate and increased precipitation allowed the establishment of closed forest, which shifted the ecological constraints on seed recruitment . . . and led to larger seeds. The dispersal agents therefore tracked, not led, the increase in seed size.’’ The recruitment hypothesis thus implies a causal chain that is the reverse of the coevolution hypothesis: closed vegetation! evolution of larger seeds!evolution of biotic dispersal!radiation of frugivores. The open vegetation explains why seed sizes remained small during the Cretaceous and biotic dispersal was not promoted. Vegetation changes caused both the increase and the following drop in seed size during the Tertiary, and biotic dispersal is positively related to seed size because of its function in dispersal of large seeds. The third life-form hypothesis was also discussed by Wing and Tiffney (1987b), but it was suggested explicitly by Moles et al. (2005a), and it also incorporates aspects developed in Moles and Westoby (2004, 2006) and Moles et al. (2004, 2005b). This hypothesis differs from the previous hypotheses because it was not formulated as a specific solution to the pattern of seed size evolution depicted in figures 1 and 2. Instead, it was suggested as an explanation for evolution of seed size in general. The life-form hypothesis does not consider why there is a positive relationship between seed size and biotic dispersal, but it is fully compatible with the explanation for such a relationship that is given by the recruitment hypothesis. Moles et al. (2005a) analyzed relationships between seed size, biotic dispersal, and plant growth form (vegetation), and in that context, they commented on Tiffney (1984) and Eriksson et al. (2000a), that is, on the coevolution and the recruitment hypotheses, and suggested an alternative hypothesis. Basically, the life-form hypothesis is similar to the recruitment hypothesis in that vegetation change, from open to closed forest systems, may be the initial driver of seed size evolution. The key difference in the life-form hypothesis, compared to the recruitment hypothesis, is that evolution of larger seeds is not caused by a process in which larger seeds increase the likelihood of recruitment under closed canopies. Instead, evolution of increasing seed size is a consequence of larger and more longlived plants having a longer juvenile period. To improve survival during this longer period to maturity, seeds must contain more resources, and therefore selection favors larger seeds in large and long-lived plants. Inspired by a theory by Charnov (1993) that predicted a positive relationship between offspring and adult size in mammals, Moles et al. (2005a) suggested 866 INTERNATIONAL JOURNAL OF PLANT SCIENCES that any factor that causes increasing plant size (and longer life span) also drives evolution of larger seeds. According to the life-form hypothesis, seeds were small during the Cretaceous because angiosperms were generally small. The drastic increase in seed size in the Early Tertiary occurred because plants generally increased in size along with the development of closed subtropical and tropical forests, and the drop in seed size from the Oligocene occurred because the average plant size decreased. The curve depicting changes in seed size (fig. 1) would, according to the life-form hypothesis, mirror the average (and range) of angiosperm plant size and life span. Examining the Hypotheses: Paleoecological and Neoecological Evidence The Coevolution Hypothesis Interactions between plants and their seed dispersers are generally considered as ‘‘diffuse,’’ meaning that the interaction occurs between a whole group of plants (e.g., several species with a certain type of fleshy fruits) and a whole group of animals (e.g., frugivorous primates; Howe and Smallwood 1982; Wheelwright and Orians 1982; Herrera 1985, 2002). Evidence suggests that the morphology of biotically dispersed fruits in many lineages has changed little since they originated in the Early Tertiary (Collinson and Hooker 1991; Bremer and Eriksson 1992; Jordano 1995), despite that the composition of the disperser fauna has changed much since that time. For example, Eriksson et al. (2000a) found that on average 91% of the lineages included in a data set from 22 paleofloras had not changed fruit type since the time the floras were deposited (which ranged between 2 and 87 Ma). The coevolution hypothesis implies that the interactions between animal dispersers and fruit traits (the arena for selection on fruit traits) resulted in directional selection and adaptive change in the Paleocene and the Eocene when a major phase of evolution of biotic dispersal took place. In contrast, the interactions during the epochs following the Eocene resulted in stabilizing selection (and thus trait stasis) or were subjected to constraints imposed by the morphological structure of biotically dispersed fruits. An important question is whether there were no (or few) potential seed-dispersing vertebrates during the Cretaceous, and, if this were the case, whether a paucity of vertebrate seed dispersers constituted a constraint for the evolution of large seeds. Although there is very limited evidence available on potential vertebrate seed dispersers before the Late Cretaceous, two lines of evidence suggest that it is unlikely that a general lack of seed dispersers constrained evolution of large angiosperm seeds. First, although it is controversial, molecular evidence suggests that several groups of mammals that became important frugivores during the Tertiary, for example, primates and rodents, originated in the Cretaceous (Adkins et al. 2001; Archibald et al. 2001; Douzery et al. 2003; Soligo and Martin 2006; Steiper and Young 2006; Bininda-Emonds et al. 2007; Martin et al. 2007). One of the most abundant groups of Cretaceous mammals, the multituberculates, is known from the Late Cretaceous and the Early Tertiary to have had seeds and fruits in its diet (Krause 1982; Del Tredici 1989; Collinson and Hooker 1991), and some multituberculates were arboreal (Jenkins and Krause 1983). It is possible that multituberculates evolved this feeding habit earlier during the Cretaceous. Second, early seed plants (seed ferns, gymnosperms) evolved fleshy structures apparently suitable for attracting seed dispersers, long before angiosperms originated (Tiffney 2004). It therefore seems likely that potential plant-seed disperser interactions existed during the Cretaceous, when angiosperms evolved and diversified. There is also evidence for interactions between seed-dispersing animals and angiosperms from one of the earliest angiosperm floras (age ;124 Myr), in which about a quarter of the species possessed small but fleshy fruits (Eriksson et al. 2000b). It has been suggested, however, that the initial function of the fleshy tissue in angiosperm fruits was not disperser attraction but protection from seed predators (Mack 2000). A second important question is the timing of evolution of known vertebrate seed dispersers in relation to the origin of biotic seed dispersal during the Paleocene and the Eocene. Major groups of frugivorous birds, for example, fruit pigeons, toucans, hornbills, and passerines such as thrushes, manakins, and cotingids, are not found in the fossil record before the Miocene or are not known at all from fossils (e.g., toucans; Feduccia 1996). An earlier origin of the passerines in the Southern Hemisphere has been suggested (Ericson et al. 2003), and molecular data indicate that modern birds originated in Cretaceous, but fossil evidence is lacking (Feduccia 1996; Chiappe and Dyke 2002). The most common group of Cretaceous birds is the extinct enantiornithes (Chiappe and Dyke 2002). Most of these birds were toothed, and many species lived in nearshore environments, suggesting that they foraged on aquatic animals. Some enantiornithes have been found in inland deposits, but their feeding habits are not known. Thus, although birds are important frugivores in the extant fauna, there is not much evidence suggesting that they were a significant part of the plantdisperser interactions in the Early Tertiary. In contrast to the lack of evidence for frugivorous birds, several Late Cretaceous and Early Paleocene mammals included seeds and fruit in their diet. During the Late Cretaceous, the dominating group of seed and fruit-eating mammals was the multituberculates, small rodentlike mammals, many of which were arboreal (Collinson and Hooker 1991). Multituberculates remained important seed and fruit eaters during the Paleocene, but they were accompanied by two other groups of fruit-eating mammals, the primate-like plesiadapiforms and condylarths (primitive ungulates; Bininda-Emonds et al. 2007; Bloch et al. 2007). A major radiation of primates and rodents occurred around the Paleocene-Eocene boundary (Collinson and Hooker 1991; Bowen et al. 2002; Gingerich 2006), and both groups developed to become important seed dispersers. Rodents were feeding mainly on hard, dry fruit types; nut size in the families Betulaceae, Fagaceae, and Juglandaceae increased from the Late Paleocene, and large nuts became abundant by the late Eocene (Collinson and Hooker 1991). Collinson and Hooker suggested that this interaction is a plausible case of diffuse coevolution between plants and mammals. Nuts are one-seeded fruits, so an increase in fruit size is likely to indicate a simultaneous increase in seed size. There are several other putative examples of diffuse coevolution between plants and seed-dispersing mammals during Paleocene and Eocene. Bloch et al. (2007) suggested that dental features of plesiadapiforms adapted to a changing diet from ERIKSSON—EVOLUTION OF SEED SIZE AND BIOTIC SEED DISPERSAL insectivory to granivory and frugivory, along with an increasing abundance of seeds and fruits as a food source. Soligo and Martin (2006) and Martin et al. (2007) argued that primates originated in the Cretaceous (but see Silcox et al. 2007; Soligo and Martin 2007) and shifted to a vegetarian diet, mainly fleshy fruits, when this new food source become available. Changes in the visual apparatus (forward-facing eyes) was suggested as a result of adaptations to frugivory. Since the first fossil evidence of primates are from the Eocene (Gingerich 2006; Martin et al. 2007), it is reasonable that the abundance of fruit as a food source promoted the radiation of primates, even if the primate lineage is older. Thus, interactions between plants and several groups of mammals (e.g., multituberculates, plesiadapiforms, primates, and rodents) may, at least from the perspective of timing, have been instrumental in the evolution of biotic dispersal systems. In contrast, fossil evidence suggests that one of the most important groups of extant frugivorous mammals, fruit bats, did not become abundant until the Miocene (although molecular evidence indicates an earlier origin of bats; Teeling et al. 2005). The evidence reviewed suggests that diffuse coevolution with mammals is likely to have played an important role for the origin and early evolution of biotic dispersal, although this role may not have been as a trigger of the selection process for larger seeds. The coevolution hypothesis does not explain why evolving larger seeds was favorable in the first place. Even if potential seed-dispersing vertebrates became increasingly abundant during the Early Tertiary, thereby enhancing seed dispersal if plants developed fleshy fruits, there is no reason why seeds would not have remained small (and thus more numerous). Even if coevolution with seed-dispersing vertebrates was a mechanism contributing to the radiation of seed sizes and biotic dispersal systems in the Early Tertiary, this process must have included some component of advantage of larger seeds, as suggested by the recruitment and the life-form hypotheses. It seems reasonable to conclude that the coevolution hypothesis is not sufficient to explain the patterns depicted in figures 1 and 2. The coevolution hypothesis is not compatible with the (admittedly limited) evidence for the existence of potential seed dispersers before the Late Cretaceous, and it provides no explanation of why seed size increased in the first place or why seed size dropped after the Eocene. 867 plants with large seeds (Salisbury 1942; Hughes et al. 1994; Butler et al. 2007). The recruitment hypothesis predicts that vegetation shifts from open to closed vegetation, which leads to selection for larger seeds with biotic dispersal or, if viewed in a shorter ecological timescale, leads to species sorting that favors those species in the species pool that possess these traits. Apart from being congruent with the patterns of change in seed size and biotic dispersal depicted in figures 1 and 2, this prediction is also supported by comparative analyses of changes within angiosperm lineages in fruit types in relation to vegetation shifts, which recurred during the Tertiary (Bolmgren and Eriksson 2005). Indirect support is also gained from studies of foraging strategies of Miocene mammals in which the proportion of frugivores was lower in savanna vegetation than in forests (Fortelius et al. 1996). Large-scale vegetation changes during the Tertiary, shifting between closed forests, open woodland, and open grasslands (which developed during the Miocene and the Pliocene; e.g., Jacobs et al. 1999; Strömberg 2005; Strömberg et al. 2007), were driven by climate change. Periods of cooling, for example, at the Eocene-Oligocene transition (Prothero 1994) were associated with development of more open vegetation. Indeed, average seed size from the paleofloras examined by Eriksson et al. (2000a) was correlated with a paleotemperature curve derived by Zachos et al. (2001) for the Late Cretaceous to the Pliocene (fig. 3). A similar result was obtained by Moles et al. (2005a) when correlating seed size and temperature across extant species. They found an even stronger correlation between seed size and precipitation. There is also evidence that does not support the recruitment hypothesis, or at least not the direct connection between evolution of larger seeds and biotic dispersal. First, in the flora of Famalicão from the Early Cretaceous, ;25% of the species had drupes or berries, but seed sizes averaged 0.78 mm3 (Eriksson The Recruitment Hypothesis The recruitment hypothesis rests on a number of assumptions that are supported by evidence. (1) Recruitment under shaded conditions is generally improved by increasing seed size (Foster 1986; Leishman and Westoby 1994; Moles and Westoby 2004). (2) There is a trade-off between seed size and seed number, implying that an increasing seed size for a plant of a given size and resource allocation to seeds leads to a reduced seed production (Henery and Westoby 2001; Moles et al. 2004). (3) Biotic seed dispersal is associated with an increase in seed dispersal distance, as compared with unassisted dispersal (Portnoy and Willson 1993), which means that a cost of reduced dispersal due to lower seed production may be offset by biotic dispersal. Biotic dispersal may also result in directed dispersal to suitable recruitment sites (Wenny and Levey 1998). (4) Seed dispersal by vertebrates is overrepresented among Fig. 3 Relationship between median seed size in 19 paleofloras from the Late Cretaceous to the Pliocene (from Eriksson et al. 2000) and deep sea temperature, based on fig. 2 in Zachos et al. (2001). Floras examined by Eriksson et al. (2000a) with an age outside the temporal range covered by Zachos et al. were excluded, and the temperature for the time of the youngest flora from the Pliocene, with a temperature below the scale shown by Zachos et al., was set to 0°C (r ¼ 0.485; P ¼ 0.02). 868 INTERNATIONAL JOURNAL OF PLANT SCIENCES et al. 2000b). Assuming that fleshy tissue of the fruits functioned to attract dispersers (but see Mack 2000), this suggests that biotic dispersal may evolve without being associated with particularly large seeds. Second, in subtropical forest in Australia, Butler et al. (2007) found a generally strong association between possession of fleshy fruits and large seeds across all studied sites. In contrast, the seed sizes within sites did not differ between species with and without fleshy fruits, indicating a weak causal relationship between seed size and biotic seed dispersal. Third, in a comparative phylogenetic analysis, Moles et al. (2005a) found that dispersal syndrome (biotic vs. abiotic) accounted for only 3% of the variation in seed size contrasts. The largest effect on seed size was found for growth form (herb vs. shrub vs. tree), where larger growth forms are associated with increasing seed size. The strongest argument for the recruitment hypothesis is that it provides an explanation for the initial evolution of increasing seed size. However, when it comes to explaining the evolution of biotic dispersal, the evidence is weaker, and it seems likely that other mechanisms than large seeds per se, must be invoked for explaining evolution of biotic dispersal. The Life-Form Hypothesis The weak relationship between dispersal syndrome and seed size contrasts led Moles et al. (2005b) to suggest the life-form hypothesis. In the context of the pattern depicted in figures 1 and 2, this hypothesis is critically dependent on how the changes in seed size and biotic dispersal are related to changes in plant size and life span. For the Tertiary, the trends in seed size and biotic dispersal, with a peak in the Eocene, a decline during the Oligocene, and a leveling off during the Miocene and the Pliocene, would fit well with the life-form hypothesis if we assume that average plant size and life span decreased as vegetation became more open after the forest cover maximum in the Eocene. The life-form hypothesis implies that angiosperms must have been generally small during the Cretaceous. Several studies suggest that early angiosperms were shrubs or small trees, or even herbs, and large angiosperm trees were less common during most of the Cretaceous than they were during the Tertiary (Wing and Tiffney 1987b; Wing and Boucher 1998). Although open canopy, broad-leaved evergreen woodland existed during Late Cretaceous (Upchurch and Wolfe 1987), closed forests during the Cretaceous were composed mainly of conifers (Wing and Sues 1992). A complicating issue when comparing the recruitment and the life-form hypotheses is that these hypotheses are difficult to distinguish based on interpretations of vegetation. Both would make the same prediction, since tree cover both creates an environment where recruitment occurs in shade and where plants are generally large and long lived. Both hypotheses focus on the requirements for successful recruitment in systems dominated by large plants. The key issue is whether the increasing competition for light during seedling establishment or the necessity of survivorship during an increasing length of time until maturity was the main driver of selection for larger seeds. This issue has to be resolved by observational and experimental studies in extant plants and by further development of life-history theory. The suggestion by Moles et al. (2004, 2005a, 2005b) that increasing seed size is associated with selection for large plant size and longer life span has been questioned. Grubb et al. (2005) claimed that a simple relationship between plant size and seed size is not supported by data, a claim that was disputed by Moles et al. (2005c), who argued that Grubb et al. had based their conclusion on a toolimited data set. Furthermore, Rees and Venable (2007) questioned whether the theory by Charnov (1993) actually supports the suggestion by Moles et al. that seed size evolution is driven by selection for juvenile time to maturity. In a reanalysis of this theory, Rees and Venable (2007, p. 926) found that a positive relationship between seed size and plant size is predicted only ‘‘if unrealistic assumptions are made about the effects of seed mass on survival, such as the effect of seed mass on instantaneous survival persisting to adulthood.’’ Irrespective of the differences in the exact mechanisms promoting evolution of larger seeds, it seems reasonable that the dramatic increase in seed size during the Early Tertiary and the following drop in seed size after the Eocene were driven by the changing composition of the vegetation, which affected conditions during recruitment and juvenile development. However, there seems to be no conclusive evidence available for distinguishing between these two mechanisms. Conclusions: A Synthesis of Coevolution, Recruitment, and Life-Form In the ‘‘Introduction,’’ four questions were asked: (1) Why were seeds generally small and biotic seed dispersal rare during most of the Cretaceous? (2) What caused the dramatic increase in seed size during the Early Tertiary? (3) Why did seed size drop after the Eocene? (4) Why is there a positive relationship between seed size and biotic seed dispersal? After an examination of three proposed hypotheses, mainly based on Eriksson et al. (2000a), Tiffney (2004), and Moles et al. (2005a), I conclude that none of these hypotheses, viewed individually, is fully compatible with paleo- and neoecological evidence but that they all are partly supported and do contribute to answering the questions. This implies indirect support for the model suggested by Wing and Tiffney (1987b), which incorporates all three mechanisms. Summarizing the evidence, it seems likely that the drastically changing environmental conditions that occurred in the Early Tertiary because of climate change and a paucity of large herbivores created a new arena for selection, one that was fundamentally different from what angiosperms experienced in the Cretaceous. The changing conditions caused a mismatch between angiosperm life-history traits and environment (Ackerly 2003), and in this context, recruitment processes and coevolution between plants and mammals were both involved in promoting changes in seed size and biotic dispersal systems in angiosperms. It is likely that large and long-lived plant lifeforms, large seed sizes, and biotic dispersal systems evolved as sets of coadapted traits. A tentative suggestion is that the change in seed size was mainly driven by the changing composition of vegetation. Undisturbed and closed forest vegetation, as developed during the global warming period in the Early Tertiary and with a peak in the Eocene, when the fauna was devoid of any large herbivores after the demise of the dinosaurs, created an envi- ERIKSSON—EVOLUTION OF SEED SIZE AND BIOTIC SEED DISPERSAL ronment where plants with large seeds were favored. The exact mechanism causing this advantage, whether competition for light or long juvenile periods, cannot be identified. Seed size dropped as vegetation became more open after the Eocene and as grasslands developed during the Miocene and the Pliocene. Interestingly, a similar drop in seed size is currently apparent along a latitudinal gradient from the equator poleward, at ;25° from the equator (Moles et al. 2007). This result indicates that a more complete sampling of seed sizes in paleofloras, allowing for an examination of seed size variation along paleolatitudes in the Tertiary, would be needed to gain a deeper insight in the evolutionary trends in seed size. Most of the floras used in Eriksson et al. (2000a) were from Europe, that is, within a relatively limited range of latitudes. It might well be that there was no drop in seed size in tropical floras after the Eocene. Although interactions with seed-dispersing vertebrates are not likely to be a primary mechanism for evolution of larger seeds during the Early Tertiary, coevolution between plants and mammals was probably an essential component of the evolution of biotic dispersal systems. Specialized frugivory evolved along with the changing features of seeds and fruits, indicating that during a formative phase from the Late Paleocene to the Eocene, both plant and mammal traits were reciprocally altered. Apart from feeding habits in general, there are several suggestive examples of mammal traits that evolved as adaptations to frugivory, such as the dentition in plesiadapiforms (Bloch et al. 2007) and the visual apparatus in primates (Soligo and Martin 2006). The abundance of large seeds and 869 fleshy fruits provided a new food niche dimension, and it may be that these new food resources promoted the radiation of frugivorous mammals in the Eocene. It is suggestive that BinindaEmonds et al. (2007) found that the diversification rate of mammals was low after the K/T boundary and did not start to increase until the Eocene, coinciding with the peak of average seed size and the fraction of biotic dispersal in plants. The observed stasis in fruit traits since the origin of fleshy fruits and large, dry fruits (Collinson and Hooker 1991; Eriksson et al. 2000a) reflects a pattern of evolutionary conservatism. That shifts in fruit type (biotic vs. abiotic) nevertheless have occurred along with vegetation shifts (open vs. closed) later in the Tertiary suggests that fruit-type shifts are not constrained if vegetation structure changes (Bolmgren and Eriksson 2005). Thus, stabilizing selection rather than evolutionary constraints is the most likely cause for trait conservatism in those lineages where fruit types have remained stable. As long as plant species continue in their original environments, there seems to be no strong selection to alter those functioning biotic seed dispersal systems that originated more than 40 million years ago in the forests of the Eocene. Acknowledgments I am grateful to A. T. Moles, J. Ehrlén, P. R. Crane, S. A. O. Cousins, and two anonymous reviewers for valuable comments on the manuscript. This study was supported by the Swedish Research Council. Literature Cited Ackerley DD 2003 Community assembly, niche conservatism, and adaptive evolution in changing environments. Int J Plant Sci 164(suppl): S165–S184. Adkins RM, EL Gelke, D Rowe, RL Honeycutt 2001 Molecular phylogeny and divergence time estimates for major rodent groups: evidence from multiple genes. Mol Biol Evol 18:777–791. Archibald JD, AO Averianov, EG Ekdale 2001 Late Cretaceous relatives of rabbits, rodents, and other extant eutherian mammals. Nature 414:62–65. Bininda-Emonds ORP, M Cardillo, KE Jones, RDE MacPhee, RMD Beck, R Grenyer, SA Price, RA Vos, JL Gittleman, A Purvis 2007 The delayed rise of present-day mammals. Nature 446:507–512. Bloch JI, MT Silcox, DM Boyer, EJ Sargis 2007 New Paleocene skeletons and the relationship of plesiadapiforms to crown-clade primates. Proc Natl Acad Sci USA 104:1159–1164. Bolmgren K, O Eriksson 2005 Fleshy fruits: origins, niche shifts, and diversification. Oikos 109:255–272. Bowen GJ, WC Clyde, PL Koch, S Ting, J Alroy, T Tsubamoto, Y Wang, Y Wang 2002 Mammalian dispersal at the Paleocene/Eocene boundary. Science 295:2062–2065. Bremer B, O Eriksson 1992 Evolution of fruit characters and dispersal modes in the tropical family Rubiaceae. Biol J Linn Soc 47:79–95. Butler DW, RJ Green, D Lamb, WJF McDonald, PI Forster 2007 Biogeography of seed-dispersal syndromes, life-forms and seed sizes among woody rain-forest plants in Australia’s subtropics. J Biogeogr 34:1736–1750. Charnov EL 1993 Life history invariants: some explorations of symmetry in evolutionary ecology. Oxford University Press, Oxford. Chiappe LM, GJ Dyke 2002 The mesozoic radiation of birds. Annu Rev Ecol Syst 33:91–124. Collinson ME, JJ Hooker 1991 Fossil evidence of interactions between plants and plant-eating mammals. Philos Trans R Soc B 333: 197–208. Crane PR, EM Friis, KR Pedersen 1995 The origin and early diversification of angiosperms. Nature 374:27–33. Del Tredici P 1989 Ginkgos and multituberculates: evolutionary interactions in the Tertiary. Biosystems 22:327–339. Douzery EJP, F Delsuc, MJ Stanhope, D Huchon 2003 Local molecular clocks in three nuclear genes: divergence times for rodents and other mammals and incompatibility among fossil calibrations. J Mol Evol 57:S201–S213. Ericson PGP, M Irestedt, US Johansson 2003 Evolution, biogeography, and patterns of diversification in passerine birds. J Avian Biol 34:3–15. Eriksson O, EM Friis, P Löfgren 2000a Seed size, fruit size, and dispersal systems in angiosperms from the Early Cretaceous to the Late Tertiary. Am Nat 156:47–58. Eriksson O, EM Friis, KR Pedersen, PR Crane 2000b Seed size and dispersal systems of Early Cretaceous angiosperms from Famalicão, Portugal. Int J Plant Sci 161:319–329. Eriksson O, A Jakobsson 1999 Recruitment trade-offs and the evolution of dispersal mechanisms in plants. Evol Ecol 13:411–423. Feduccia A 1996 The origin and evolution of birds. Yale University Press, New Haven, CT. Fleming TH, R Breitwisch, GH Whitesides 1987 Patterns of tropical vertebrate frugivore diversity. Annu Rev Ecol Syst 18:91–109. Fortelius M, LWerdelin, P Andrews, RL Bernor, A Gentry, L Humphrey, HW Mittmann, S Viratana 1996 Provinciality, diversity, turnover, and paleoecology in land mammal faunas of the later Miocene of western Eurasia. Pages 414–448 in RL Bernor, V Fahlbusch, H-W 870 INTERNATIONAL JOURNAL OF PLANT SCIENCES Mittmann, eds. The evolution of western Eurasian Neogene mammal faunas. Columbia University Press, New York. Foster SA 1986 On the adaptive value of large seeds for tropical moist forest trees: a review and synthesis. Bot Rev 52:260–299. Gingerich PD 2006 Environment and the evolution through the Paleocene-Eocene thermal maximum. Trends Ecol Evol 21:246–253. Grubb PJ, DA Coomes, DJ Metcalfe 2005 Comment on ‘‘A brief history of seed size.’’ Science 310:783a. Henery ML, M Westoby 2001 Seed mass and seed nutrient content as predictors of seed output variation between species. Oikos 92: 479–490. Herrera CM 1985 Determinants of plant-animal coevolution: the case of mutualistic dispersal of seeds by vertebrates. Oikos 44:132–141. ——— 2002 Seed dispersal by vertebrates. Pages 185–208 in CM Herrera and O Pellmyr, eds. Plant-animal interactions: an evolutionary approach. Blackwell, Oxford. Howe HF, J Smallwood 1982 Ecology of seed dispersal. Annu Rev Ecol Syst 13:201–228. Hughes L, M Dunlop, K French, MR Leishman, B Rice, L Rodgerson, M Westoby 1994 Predicting dispersal spectra: a minimal set of hypotheses based on plant attributes. J Ecol 82:933–950. Jacobs BF, JD Kingston, LL Jacobs 1999 The origin of grassdominated ecosystems. Ann Mo Bot Gard 86:590–643. Jenkins FA, Krause DW 1983 Adaptations for climbing in North American multituberculates (Mammalia). Science 220:712–715. Jordano P 1995 Angiosperm fleshy fruits and seed dispersers: a comparative analysis of adaptation and constraints on plant-animal interactions. Am Nat 145:163–191. Krause DW 1982 Jaw movement, dental function, and diet in the Paleocene multituberculate Ptilodus. Paleobiology 8:265–281. Leishman MR, M Westoby 1994 The role of large seeds in shaded conditions: experimental evidence. Funct Ecol 8:205–214. Leishman MR, M Westoby, E Jurado 1995 Correlates of seed size variation: a comparison among five temperate floras. J Ecol 83:517–530. Leishman MR, IJ Wright, AT Moles, M Westoby 2000 The evolutionary ecology of seed size. Pages 31–57 in ME Fenner, ed. Seeds: the ecology of regeneration in plant communities. 2nd ed. CAB International, Wallingford. Mack AL 2000 Did fleshy fruit pulp evolve as a defence against seed loss rather than as a dispersal mechanism? J Biosci 25:93–97. Martin RD, C Soligo, S Tavare 2007 Primate origins: implications of a Cretaceous ancestry. Folia Primatol 78:277–296. Moles AT, DD Ackerly, JC Tweddle, JB Dickie, R Smith, MR Leishman, MM Mayfield, A Pitman, JT Wood, M Westoby 2007 Global patterns in seed size. Global Ecol Biogeogr 16:109–116. Moles AT, DD Ackerly, CO Webb, JC Tweddle, JB Dickie, AJ Pitman, M Westoby 2005a Factors that shape seed mass evolution. Proc Natl Acad Sci USA 102:10540–10544. Moles AT, DD Ackerly, CO Webb, JC Tweddle, JB Dickie, M Westoby 2005b A brief history of seed size. Science 307:576–580. ——— 2005c Response to comment on ‘‘A brief history of seed size.’’ Science 310:783b. Moles AT, DS Falster, MR Leishman, M Westoby 2004 Small-seeded species produce more seeds per square-meter of canopy per year, but not per individual per lifetime. J Ecol 92:384–396. Moles AT, M Westoby 2004 Seedling survival and seed size: a synthesis of the literature. J Ecol 92:372–383. ——— 2006 Seed size and plant strategy across the whole life cycle. Oikos 113:91–105. Portnoy S, MF Willson 1993 Seed dispersal curves, behaviour of the tail of the distribution. Evol Ecol 7:25–44. Prothero DR 1994 The Eocene-Oligocene transition: paradise lost. Columbia University Press, New York. Rees M, L Venable 2007 Why do big plants make big seeds? J Ecol 95:926–936. Salisbury EJ 1942 The reproductive capacity of plants. Bell & Sons, London. Silcox MT, DM Boyer, JI Bloch, EJ Sargis 2007 Revisiting the adaptive origins of primates (again). J Hum Evol 53:321–324. Soligo C, RD Martin 2006 Adaptive origins of primates revisited. J Hum Evol 50:414–430. ——— 2007 The first primates: a reply to Silcox et al. (2007). J Hum Evol 53:325–328. Steiper ME, NM Young 2006 Primate molecular divergence rates. Mol Phylogenet Evol 41:384–394. Strömberg CAE 2005 Decoupled taxonomic radiation and ecological expansion of open-habitat grasses in the Cenozoic North America. Proc Natl Acad Sci USA 102:11980–11984. Strömberg CAE, L Werdelin, EM Friis, G Sara 2007 The spread of grass-dominated habitats in Turkey and surrounding areas during the Cenozoic: phytolith evidence. Palaeogeogr Palaeoclimatol Palaeoecol 250:18–49. Teeling EM, MS Springer, O Madsen, P Bates, J O’Brien, WJ Murphy 2005 A molecular phylogeny for bats illuminates biogeography and the fossil record. Science 307:580–584. Tiffney BH 1984 Seed size, dispersal syndromes, and the rise of the angiosperms: evidence and hypothesis. Ann Mo Bot Gard 71:551–576. ——— 1992 The role of vertebrate herbivory in the evolution of land plants. Palaeobotanist 41:87–97. ——— 2004 Vertebrate dispersal of seed plants through time. Annu Rev Ecol Evol Syst 35:1–29. Upchurch GR, JA Wolfe 1987 Mid-Cretaceous to Early Tertiary vegetation and climate: evidence from fossil leaves and woods. Pages 75–105 in EM Friis, WG Chaloner, PR Crane, eds. The origins of angiosperms and their biological consequences. Cambridge University Press, Cambridge. Wenny DG, DJ Levey 1998 Directed seed dispersal by bellbirds in a tropical cloud forest. Proc Natl Acad Sci USA 95:6204–6207. Wheelwright NT, GH Orians 1982 Seed dispersal by animals: contrasts with pollen dispersal, problems of terminology, and constraints on coevolution. Am Nat 119:402–413. Willson MF, BL Rice, M Westoby 1990 Seed dispersal spectra: a comparison of temperate plant communities. J Veget Sci 1:547–562. Wing SL, LD Boucher 1998 Ecological aspects of the Cretaceous flowering plant radiation. Annu Rev Earth Planet Sci 26:379–421. Wing SL, HD Sues 1992 Mesozoic and Early Cenozoic terrestrial ecosystems. Pages 327–416 in AK Behrensmeyer, JD Damuth, WA DiMichele, R Potts, H-D Sues, SL Wing, eds. Terrestrial ecosystems through time: evolutionary paleoecology of terrestrial plants and animals. University of Chicago Press, Chicago. Wing SL, BH Tiffney 1987a Interactions of angiosperms and herbivorous tetrapods through time. Pages 203–224 in EM Friis, WG Chaloner, PR Crane, eds. The origins of angiosperms and their biological consequences. Cambridge University Press, Cambridge. ——— 1987b The reciprocal interaction of angiosperm evolution and tetrapod herbivory. Rev Palaeobot Palynol 50:179–210. Zachos J, M Pagani, L Sloan, E Thomas, K Billups 2001 Trends, rhythms, and aberration in global climate 65 Ma to present. Science 292:686–693.