Our first ancestors were all aquatic. The oldest known vertebrates are about 500 Ma old, but the first potentially terrestrial vertebrates are less than 350 Ma old. For more than 150 Ma, our ancestors swam with their fins and breathed through their gills; on dry land, these structures were very inefficient. Their sensory organs worked poorly in air, if at all, and had to undergo various modifications to adapt to life on the continents. The eyes of our ancestors lacked eyelids and tear glands and could dry out rapidly; their ears did not enable them to hear most airborne sounds, such as the vocalizations of many frogs, birds, and mammals, such as the human voice. Yet all these problems were solved, and the few vertebrate species that succeeded in adapting to this new environment about 320 Ma ago diversified into the more than 25,000 extant species of land vertebrates.
To reconstruct this history, we need objective methods to use the indirect information on evolution provided by fossils or the extant biodiversity, as well as principles of nomenclature to produce classifications. These techniques and concepts are widely used in modern evolutionary biology. Thus, phylogenetics provides evolutionary trees that are the starting point of comparative or biodiversity analyses for a broad range of evolutionary problems or taxa. Biological nomenclature provides rules that enable systematists to present classifications (better called taxonomies) to summarize the evolutionary relationships between species and to sort our knowledge of the biosphere. Recent developments in phylogenetics and, to a lesser extent, in biological nomenclature have given new life to paleontology and evolutionary biology. Until approximately the 1970s, paleontologists reconstructed evolutionary trees by hand, using criteria that they did not always explain. Since then, the advent of cladistics, soon followed by software that enabled systematists to tap into the tremendous processing power of computers, introduced more objectivity into phylogenetics because the data used to produce the trees are generally published. This triggered a proliferation of phylogenetic studies and led to a re-examination of many long-held hypotheses on the phylogeny of life. As a result, we now have a much better resolved tree of life than a few decades ago, even though much of this tree will probably change as a result of future investigations. These methods are presented in a simplified manner in this chapter, and the bibliography provides an introduction to the most relevant papers where more technical information can be found.
Rank-Based Nomenclature
A form of classification is essential to sort information, whatever its nature. Man has classified animals since antiquity, as attested in the Bible (ESV, 2001), in which we can read: “So out of the ground the Lord God formed every beast of the field and every bird of the heavens and brought them to the man to see what he would call them. And whatever the man called every living creature, that was its name.” (Genesis, 2:19). Since Aristotle (384–322 BCE), many authors have proposed classifications of living beings. The subdiscipline of biology that consists of naming, defining, and delimiting the groups of living organisms (the taxa) is called “taxonomy,” like the product of this activity (the taxonomies). Thus, taxonomy harks back to antiquity (under a form substantially different from today’s), but, initially, only vernacular names were used. These were part of the standard vocabulary of a language, in contrast to formal names that are often known only by scientists.
The drawback of vernacular names is that their meaning can vary in space and time (this is typical of most words in any language), and there are often no exact synonyms among languages. Thus, the word “fish” once included whales (until the 19th century), although they are now excluded because we now know that whales are mammals that have returned to the seas. In English, this word has also included, at least in its broadest sense, aquatic animals that are no longer considered “fishes,” such as echinoderms (e.g., “starfish”), arthropods (e.g., “crayfish”), mollusks (e.g., “cuttlefish”), or even cnidarians (e.g., “jellyfish”); but this is not true of many other European languages, such as French, in which the equivalent word “poisson” has long had a narrower sense restricted to aquatic vertebrates. These two words (“fish” and “poisson”), often considered synonyms, have thus not always referred to the same groups of animals.
Vernacular words are not ideally suited to scientific use because of their variability in space and time, and because of the imperfect synonymy between names used in various languages (Minelli et al., 2005). Thus, scientists began to develop, as early as the 18th century, precise taxonomies based on names that would ideally have the same meaning for all scientists, no matter when or where they lived. Such developments were becoming increasingly important because of the exponential growth of our knowledge of the biodiversity that resulted from the scientific exploration (in which several biologists took part) of various continents in the 18th century. The Swedish botanist Linnaeus (1707–1778) was the first to propose a comprehensive taxonomy that was widely adopted among scientists. In his system, the names were based on Ancient Greek and Latin roots, an advantage because these dead languages were no longer changing, and because they were widely read by 18th-century scientists. (Most of Linnaeus’ works, and even his letters to foreign colleagues, are written in Latin.) To cope with the astronomical number of species to name, he proposed to form names consisting of two words, a genus name and a specific epithet. This constituted a great nomenclatural simplification because species names had grown to Latin descriptions sometimes spanning several lines of text. Thus, our species belongs to the genus Homo and bears the epithet sapiens. Furthermore, each genus belongs to an order, each order belongs to a class, and each class fits into a kingdom. For our species, the taxa of these ranks are Primates (order), Mammalia (class), and Animalia (kingdom). Linnaeus thus used the categories species, genus, order, class, and kingdom that encompass increasingly more inclusive groups. More recently, additional categories (ranks) were introduced, such as the family between the genus and the order. For some ranks, there are now standard endings. Thus, in zoology, taxa at the family rank end in -idae. The stem of the name of a family is always formed by the name of a genus that belongs to the family. Our family name (Hominidae) derives from our genus name (Homo) and the suffix -idae. For subfamilies, the suffix is -inae, and this explains why our subfamily is named Homininae. Using such rules, the following classification of our species can be given:
Figure 1.1. Delimitation of taxa in Linnaean (rank-based) nomenclature. Hypothetical phylogeny of a dozen species. The types are designated by asterisks (*). Type species define genera, and type genera define families.
Well after Linnaeus, taxonomists proposed rules to determine how to apply names; this forms what we call “nomenclature.” Such an explicit nomenclature became necessary in the 19th century because of the very rapid growth in our knowledge of biodiversity (we presently know about 2,000,000 species, and several thousand are added every year). The nomenclature used by most taxonomists is often called “Linnaean” because some of its principles were established by Linnaeus, but it differs by using types, a nomenclatural novelty introduced in the 19th century. Types are either individuals (an animal preserved in alcohol or a skeleton, for instance) used in defining species, or taxa of lower rank that are used to define taxa of higher rank. Thus, a genus is defined by a type species, and a family is defined by a type genus (Fig. 1.1). Because of these additions and the extensive use of ranks, some systematists now prefer the expression “rank-based nomenclature.”
DELIMITATION OF TAXA IN LINNAEAN (RANK-BASED) NOMENCLATURE
In rank-based nomenclature, taxa are delimited using their type and rank. A taxon of a given rank cannot be included in another taxon of the same or lower rank (a kingdom contains classes, and the latter contain orders, but not the reverse). This system rests largely on subjective absolute ranks, also called Linnaean categories. Indeed, no objective criterion has ever been used to determine the rank of taxa, except in a few studies (see Laurin, 2005). The geological age of taxa was used a few times to determine ranks, but this practice was soon abandoned because it resulted in drastic changes to the ranks traditionally attributed to most taxa. For instance, several arthropod families are as old as the class Mammalia (which includes all mammals); most mammalian families are no older than several arthropod tribes.
Evolution and Vertebrate Taxonomy:
There Are No Fishes Anymore!
When Linnaeus proposed his taxonomy, virtually no scientists accepted any sort of theory of biological evolution (Linnaeus was initially a creationist). The theory of evolution by natural selection was proposed, discussed, and accepted (at least by scientists) in the middle to late 19th century. This theory was further elaborated in the 20th century by the discovery of genetic mutations and genetic drift. However, biologists have only recently changed their taxonomies to reflect this scientific revolution, and the rules of rank-based nomenclature have not drastically changed for more than a century.
Acceptance of the idea of organic evolution has led biologists to include all descendants of an ancestor in the same taxon as that ancestor. A group thus delimited is objective because it includes species that share a common history and inherited similarities. Such a taxon is called “monophyletic,” as opposed to a “paraphyletic” group (Fig. 1.2A), which excludes part of the descendants, or a “polyphyletic” group, which contains species that are not closely related to each other (Fig. 1.2B). The taxon Pisces (the “fishes”) is no longer considered valid by most biologists because it excludes some descendants of “fishes,” namely, the tetrapods (Fig. 1.3). Paraphyletic taxa are artificial because their delimitation is arbitrary. They were erected long ago, when biologists classified organisms according to their similarities and along the gaps in biodiversity. Such gaps, found between “fishes” and tetrapods in the extant fauna, result from the extinction of intermediate forms. Fossils can fill these gaps, though only to an extent, because museum collections of fossils represent only a small proportion of the extinct species that once inhabited the Earth. Thus, these gaps are merely artifacts that reflect inadequacies in our knowledge of nature; they do not form justifiable borders between taxa. Indeed, why exclude only the tetrapods from the “fishes”? Why should we not also exclude the lungfishes and the coelacanth as well? The taxon Pisces is paraphyletic (Fig. 1.3), and it is preferable to replace it by a monophyletic taxon issued from the same ancestor, namely, Vertebrata, the taxon that includes all vertebrates. Monophyletic groups are considered natural and can be considered individuals in the philosophical sense: they have an origin (the appearance of the last common ancestor) and an end (the extinction of the last descendant of that ancestor). However, the rank-based codes (that presently rule the application of taxon names) do not require that taxa be monophyletic; paraphyletic and even polyphyletic taxa are allowed.
Figure 1.2. Monophyly, paraphyly, and polyphyly. Hypothetical example of monophyletic, paraphyletic (A), and polyphyletic (B) taxa. The last common ancestors of low-ranking taxa included in higher-ranking taxa 13 to 17 are identified by shaded squares. The content of taxa 13 to 17 is identified by brackets and by shades of gray.
Figure 1.3. Phylogeny and vertebrate taxonomy. The groups “fishes” and “agnathans” are paraphyletic, and as such invalid, according to most contemporary taxonomists. All the other taxa in this figure are monophyletic. The time axis on this diagram is vertical, the past being at the bottom and the present at the top. Thus, each node (bifurcation) of the tree represents the last common ancestor of two taxa.
Figure 1.4. Phylogeny and sarcopterygian taxonomy. The traditional taxonomy, including paraphyletic taxa (between quotation marks), is indicated by shades of gray; the more recent delimitation of the taxa Amphibia and Reptilia is indicated by brackets above the tree. The famous sarcopterygian Eusthenopteron is a tristichopterid; the gymnophionans (caecilians) are limbless tropical amphibians; the squamates include the “lizards” (a paraphyletic group) and the snakes. The time axis is vertical, with the past at the bottom and the present at the top. The nodes (bifurcations in the tree) represent the last common ancestors of pairs of taxa.
Taxonomy has been deeply transformed by the application of these principles. For instance, it was once customary to divide the limbed vertebrates (often called tetrapods, but called stegocephalians in this book) into amphibians, reptiles, birds, and mammals, but the first two of these groups are paraphyletic (Fig. 1.4). Today, some authors want to eliminate all names of paraphyletic taxa, such as Reptilia (which traditionally includes the turtles, snakes, and crocodilians, but not the birds, which are nevertheless the closest relatives of the crocodilians in the extant fauna), whereas others prefer to re-delimit these taxa to make them monophyletic. Under this latter approach, Amphibia no longer includes the first limbed vertebrates, such as the Devonian genus Ichthyostega (360 Ma old), because the latter are not more closely related to the extant amphibians (frogs, toads, salamanders, etc.) than to the mammals. The taxon Reptilia can be made monophyletic by excluding the “mammal-like reptiles” (stem synapsids) and including the birds (Fig. 1.4), as advocated by some authors.
Phylogenetic Nomenclature
Taxonomy is currently undergoing a revolution in an attempt to make biological classification less ambiguous. Use of rank-based nomenclature is increasingly unsatisfactory because taxa are delimited under that system through the use of a type and a rank (a Linnaean category, such as species, genus, or family). Since these ranks are subjective, taxonomists can—and often do—change them at will. This system also allows new taxa to be erected (to encompass the same species or individuals) and established taxa to be suppressed (by declaring them synonyms) without requiring an objective basis for any such decisions. Understandably, this results in great taxonomic instability, even if the evolutionary tree (the phylogeny) is stable. In other words, even if our ideas about the evolution of a taxon are stable (in the long term this is admittedly an idealized scenario), the classification of this taxon can be unstable, simply because taxonomists are free to expand or reduce the membership of taxa.
This problem can be illustrated by an example using the origin of mammals (Fig. 1.5). The formal name of the taxon that includes all mammals is Mammalia. This name was used by Linnaeus, who knew only placental mammals (Placentalia) and one marsupial (Marsupialia), the Virginia opossum (1). Later, we discovered monotremes, which lay eggs (unlike other mammals) but possess mammary glands. Mammalia was then expanded to encompass the monotremes (2). With the subsequent discovery of fossils similar to extant mammals, Mammalia was further expanded (3 to 6; most commonly 4 in recent times), although some authors advocate a return to an older meaning of this word (2, or occasionally even 1). Other authors extend the taxon Mammalia to encompass a much larger clade (7 to 10), although they fortunately represent a small minority.
Figure 1.5. Delimitations of the taxon Mammalia in rank-based nomenclature. The smallest clade that includes all extant mammals is shaded. The last common ancestor of the clades that have been called Mammalia in various studies are identified by numbered gray circles. The sign “+” designates extinct taxa. The reptiles, the mammals’ closest extant relatives, are highlighted in a gray rectangle. Modified from Rowe and Gauthier (1992).
This problem, far from affecting only the taxon Mammalia, results from the application of principles of rank-based nomenclature (see the box titled “Instability in rank-based nomenclature”), which were proposed in the 18th and 19th century, when most taxonomists considered taxa to be classes in the philosophical sense of the word. Classes can be defined by intrinsic properties that are both necessary and sufficient for an element to belong to this class. For instance, tetrapods possess four limbs, as suggested by their name, and this could be viewed as the defining property of the class Tetrapoda. However, many contemporary systematists think that taxa are individuals in the philosophical sense of the word, since they have a beginning (the appearance of a clade) and an end (the extinction of the clade) in time. Because taxa evolve, their members do not necessarily share intrinsic properties (Ereshefsky, 2007). Thus, snakes and caecilians are tetrapods, even though they have lost their limbs, because they are descended from tetrapods. For these reasons, it is preferable to define taxon names using types and the phylogeny. This is analogous to the delimitation of families (in humans) or breeds (of domestic animals), which depends on ancestry (genealogy).
INSTABILITY IN RANK-BASED NOMENCLATURE
The inherent instability in taxon delimitation under rank-based nomenclature can be illustrated by a hypothetical example (shown below) of a taxon including four species (a to d), initially attributed (by the first taxonomist who worked on this group) into two genera (E and F) and a single family (Eidae). Since determination of the rank (Linnaean category) of taxa is subjective, the rank of any taxon can be changed for subjective reasons (such as personal preference) and result in taxonomic changes that do not reflect objective discoveries. A taxonomist can declare that genus F is invalid because he considers the species that it includes not sufficiently distinct from those included in genus E; he then declares genus F a junior synonym of E, thus abolishing it. On the other hand, he may recognize additional genera, which results in yet other taxonomies for the same set of species under the same phylogeny. Finally, he may erect new families, which yields additional alternative taxonomies. All these changes are allowed by the codes of rank-based nomenclature (the zoological, botanical, and bacteriological codes). As a result, even in the absence of any objective reasons to reject the first proposed taxonomy, several alternatives can be proposed and coexist in the scientific literature; all are then simultaneously valid. This makes taxon delimitation ambiguous; for instance, genus E can contain species a, or species a and b, or species a to d. The examples of allowed alternative taxonomies figured here are not exhaustive; see Laurin (2008a) for a more exhaustive list.
Types, ranks, and instability in rank-based nomenclature. Rank-based nomenclature cannot stabilize taxonomy, because even without changes to our objective knowledge of nature (such as the discovery of new species or publication of a new phylogeny), taxonomists can suppress taxon names (by putting them into synonymy) or erect new ones. Taxa identified by an asterisk (*) are types.
Figure 1.6. Definitions of taxon names in phylogenetic nomenclature. The three main kinds of phylogenetic definitions of taxon names. A, B, and Z are species or specimens (individuals); M is an apomorphy (derived character state). Modified from Lee (1998).
To replace rank-based nomenclature (at least above the species level), several systematists have developed a phylogenetic nomenclature. In that system, the taxon name Mammalia could be defined (for instance) as the smallest clade that includes monotremes, marsupials, and placentals (Fig. 1.5, node 2). If we wish to discuss more or less inclusive clades than this one, we have to use other names for them (each name must have a single definition). Phylogenetic nomenclature should clarify the meaning of taxon names since each name will correspond to a single clade on a given tree. For this reason, it is adopted in this book. It differs from rank-based nomenclature by using at least two types, called “specifiers,” to define taxon names. In phylogenetic nomenclature, three main kinds of definitions can be given to these names (Fig. 1.6): 1, node-based (for instance, the smallest clade that includes species A and B, in pale gray); 2, apomorphy-based (an apomorphy is a new character state; such a definition can be, for instance, the clade delimited by apomorphy M shared with species A, shown in white); 3, branch-based (the largest clade that includes species A but not species Z in dark gray). (See the box titled “Nodes, apomorphies, and branches.”)
NODES, APOMORPHIES, AND BRANCHES
A node is a point on a tree or on a cladogram that gives rise to two distinct evolutionary lineages. Two nodes are shown in Figure 1.6: one gave rise to species Z and the clade (A, B), and the other is the last common ancestor of species A and B.
An apomorphy is a new character state; for instance, apomorphy M is shared by species A and B in Figure 1.6.
A branch is a segment located between two nodes (also called an “internode”), or between a node and a terminal taxon (such as species Z, A, and B). For example, a branch links the two nodes shown in Figure 1.6.
Contrary to rank-based nomenclature, in phylogenetic nomenclature, the taxonomy is stable if the assumed phylogeny does not change and if we do not discover new species. Only such changes to our objective knowledge of nature can result in changes in taxonomic content. A code of phylogenetic nomenclature, called the PhyloCode, has been developed (Cantino and de Queiroz, 2006). Its development is over-seen by the International Society for Phylogenetic Nomenclature (Laurin and Cantino, 2007; Laurin and Bryant, 2009).
Ancestors and Characters
From the 19th century to the 1980s, paleontologists searched for ancestors of extant species and higher taxa. This search has not been especially successful, for several reasons. First, only a small proportion of the species that once existed have left fossils. This is hardly surprising because, normally, only mineralized structures (especially skeletons) fossilize. Thus, some taxa, such as slugs and earthworms, normally leave no body fossils. Earthworm burrows may fossilize, and these may enable us to determine that earthworms were present in a given area at a certain time, but such trace fossils yield very little morphological data, so that the identity of the animal that left such burrows is often uncertain. Even taxa that develop a mineralized skeleton may leave no fossils if they live in a habitat unfavorable to fossilization. Most fossils form underwater, in environments characterized by a high sedimentation rate (i.e., where many particles suspended in water are deposited on the substrate), which ensures that carcasses are buried before they are entirely destroyed by carnivores and scavengers. Thus, desert or mountain dwellers are almost unknown in the fossil record. Given the low overall proportion of extinct species represented by fossils, it is likely that fossils of most ancestors of extant and extinct species will never be found.
The second reason why few ancestors are known is that most fossils do not represent ancestors of extant or extinct species; instead, they are simply relatives (which, like cousins, are not ancestors) that we call “sister groups” (Fig. 1.7). This is shown by the presence of derived character states in these fossils that should also occur in their descendants, if they were known, but that do not in fact occur in more recent species.
Characters describe attributes of taxa; they may be morphological (for instance, they may describe the presence or absence of limbs, their size, or their shape), physiological (e.g., metabolism, ecto- or endothermy, ability to hibernate), behavioral (solitary, gregarious, social, etc.), or molecular (presence or absence of certain genes, number of copies of a gene, insertions or deletions in the genome, substitutions of nucleotides, etc.). They constitute the data that phylogeneticists use to reconstruct the evolutionary tree of life.
Figure 1.7. Real ancestors, hypothetical ancestors, and sister groups. Taxa 1 to 9 are closely related to each other; thus, taxon 8 is the sister group of taxon 7; together, they form the sister group of taxon 6, and so on. The portions of the evolutionary tree represented by fossils are shown in black; the inferred portions are shown as thin gray lines. Some extinct species are totally unknown (no fossil is preserved, as in taxa 3 and 5). Only two real ancestors are known, but all hypothetical ancestors (gray squares) can be inferred.
Even if we had the extraordinary good luck of finding a fossil representing a direct ancestor of an extant taxon, it would be impossible to prove that it is indeed an ancestor, rather than the sister group, because ancestral status can only be shown by a lack of evidence: the absence of unique character states in the ancestor.
For all these reasons, most vertebrate paleontologists have stopped looking for direct ancestors and concentrate instead on determining the sister group relationships between extant and extinct taxa. Taxa are considered a priori not to be ancestral to each other, but even if some are, this does not invalidate the approach. The advantage of this approach is that it is testable and that it does not rest on absence of evidence, but instead on the discovery of new characters that unite sister groups. For instance, digits are recent structures (compared with fins), and this suggests that amphibians are closely related to amniotes (mammals and reptiles), because these taxa all possess digits (Fig. 1.8).
Conversely, primitive (old) character states cannot demonstrate relationships, because they have been inherited from an old, distant ancestor. Thus, the presence of fins in sharks, actinopterygians (sturgeons, salmons, etc.), the coelacanth, and dipnoans (lungfishes) does not indicate that these taxa are more closely related to each other than to tetrapods, because the fin is an older structure than the limb. Using such primitive characters to infer the phylogeny of gnathostomes (jawed vertebrates) would lead to erroneous results (Fig. 1.9).
The age of character states is their first appearance on an evolutionary tree, as shown by fossils or inferred by their distribution in extant taxa. A character state is primitive (and not to be used to infer clades) if its appearance precedes the divergence between the taxa whose phylogeny we want to study. Thus, the presence of fins cannot be used to study gnathostome phylogeny, because fossils show that the last common ancestor of extant gnathostomes lived at least 430 Ma ago, but fins appeared earlier, in jawless vertebrates, at least 490 Ma ago. However, the presence of limbs with digits in tetrapods (amphibians, mammals, and reptiles) suggests that they form a clade that excludes all finned vertebrates because digited limbs appeared about 365 Ma ago, well after gnathostomes. The presence of such limbs can thus be used to assess gnathostome phylogeny. In the context of phylogenetic analyses, we usually consider only the polarity of characters (relative age), rather than their absolute age, which simplifies discussions. Indeed, we often cannot determine the absolute age of character states, especially if they concern soft anatomy, physiology, or behavior and leave no fossils. Similarly, the age of many taxa cannot be determined by fossils, because they lack a mineralized skeleton; examples include slugs and earthworms. The relative age is much easier to determine, by looking at the distribution of character states on the tree (widespread characters are generally older than characters with restricted distributions). Old characters are called “primitive,” and recent ones are called “derived.” Determining the status of these character states (in other words, determining the polarity of the character) is called polarizing the character.
Figure 1.8. Phylogeny and character polarity. Phylogeny based partly on the appearance of the limb with digits and showing that the limb is more recent than the fin and results from a transformation of the latter.
Figure 1.9. Incorrect phylogeny based (in part) on a primitive character state. Incorrect gnathostome phylogeny that would result from using old (primitive) character states (here, fins) to infer relationships.
Character polarity is always relative. Thus, the presence of limbs with digits is derived in the context of an analysis of gnathostomes, but primitive in the context of an analysis of tetrapod phylogeny. In a study of reptilian phylogeny, we could not exclude snakes to form a clade uniting turtles, crocodilians, and birds, because the limb appeared well before the first reptile (which is barely 315 Ma old). In that context, the presence of a limb is a primitive character. Conversely, the presence of limbs is derived (hence phylogenetically useful) in the context of metazoan phylogeny because the first metazoans (multicellular animals) are older (550 Ma) than the first fins. Therefore, the chondrichthyans and the actinopterygians can be grouped into the gnathostomes, whereas the echinoderms, mollusks, and arthropods can be excluded, based on the presence or absence of fins in these taxa. Generally, character polarity is assessed through outgroup comparison. The outgroup must be fairly closely related to the taxon whose phylogeny we want to study, but it must not be part of it. Thus, to study sarcopterygian phylogeny, actinopterygians can be used as an outgroup (Fig. 1.3). The character state found in the outgroup and in part of the ingroup (the group whose phylogeny we want to study; sarcopterygians in this example) is generally primitive, whereas character states found only in part (or all) of the ingroup are generally derived. This criterion enables us to establish that, for the character “appendage type,” the state “fin” is primitive (because it is found in actinopterygians and in some sarcopterygians) whereas the state “limb” is derived (Fig. 1.8), because it occurs only in some sarcopterygians.
The phylogenetic relationships that can be established through parsimony specify only a topology, which is information about the relative kinship of taxa. Figures 1.8 and 1.9 represent topologies (cladograms). By itself, the topology does not yield the geological age of taxa. On the contrary, a phylogeny (Fig. 1.7) incorporates a topology and additional data represented by branch lengths (which usually represent evolutionary time).
Parsimony and Reconstruction of Character Evolution
It is impossible to read the history of characters directly, even when their evolution is documented in a fairly rich fossil record. This history must nearly always be inferred using various methods. Before the advent of cladistics in zoology in the 1970s and 1980s, paleontologists inferred these events mentally, each in their own way, and often without providing the data on which their reasoning was based. In 1950, the German entomologist Willi Hennig proposed a new method, cladistics. Cladistics can be used to infer character history based on parsimony, a principle that is used in various ways in all sciences, since it rests on the principle that as few hypotheses as possible must be made to explain data (and these hypotheses should be as simple as possible). For instance, if the relationship between two variables (between temperature of a gas, for example, and its volume at a given pressure) is tested for three values, and if the three data points seem to line up (Fig. 1.10, black circles), we always infer a linear relationship between these variables (Fig. 1.10, black line), even if other relationships can also explain these data (Fig. 1.10, gray lines). Why? Simply because the linear relationship is the simplest, since it requires estimating only two constants, “a” and “b” in the following equation, where X and Y represent the variables):
Y = a X + b
The other relationships (Fig. 1.10, gray lines) could be represented by more complex equations that require estimating more variables. The latter would be invoked only if further research yielded data that suggest that a more complex model is required (for example, the gray circle in Figure 1.10 suggests that the dark gray line is more adequate than the black straight line).
Figure 1.10. Principle of parsimony in science. If three data points (in black) document the relationship between two variables, and seem to line up, we always infer a linear relationship (black line). Other relationships are possible (gray lines), but they are chosen only if discovery of additional data require it. The gray circle suggests that the dark gray line fits the data better than the straight black line.
Parsimony, when used to study character evolution, simply stipulates that we must infer as few character transformations as possible to explain the distribution of its states. This does not imply that the most parsimonious pattern is the true one (one might say that Nature is not necessarily parsimonious), but if evolution has followed a more complex pattern, more data and further analyses can potentially demonstrate it. Thus, the distribution of the character on the tree of Figure 1.11 suggests that the state “dark gray” appeared twice and that the state “light gray” is older, because this history implies only two character transformations (two appearances of the state “dark gray,” as shown in Figure 1.11A). Other hypotheses could be made, but they are all more complex; they require at least three transformations (Fig. 1.11B). The procedure to infer the history requiring the lowest possible number of steps in a character is called “parsimony optimization.” Since we infer that the state “light gray” is older than the state “dark gray,” we conclude that the first one is primitive and the second derived. A derived state is what we call an apomorphy. When it is shared by at least two taxa that inherited it from a common ancestor (like the state “dark gray” in taxa 6 and 7 in Figure 1.11A, or the appearance of digits in amphibians, mammals, and reptiles in Figure 1.8), we call it a synapomorphy of those taxa. When it is present in a single taxon, such as feathers in birds in Figure 1.4, we call it an autapomorphy of that taxon. A primitive character is often called a plesiomorphy (such as the state “light gray” in Figure 1.11A), and it must not be used to infer the existence of clades, because this would often result in errors (Fig. 1.9).
Figure 1.11. Parsimony and character evolution. Given the character distribution and the phylogeny, hypothesis A is simpler because it requires only two transformations. Any other hypothesis, such as shown in B, requires at least three transformations, and will not be discussed further.
Evolutionary Trees:
Intuition, Parsimony, and Evolution
The evolutionary history of taxa cannot be read directly from the fossil record, even when fossils are abundant; just like for characters, the history of taxa must be inferred. For a long time, paleontologists and other systematists tried to reconstruct that history using global (unpolarized) similarity. This method went out of fashion (for good reasons) with the introduction of cladistics, which uses the principle of parsimony; it was initially developed by Hennig (1950, English translation 1965), and has been in use by other systematists since the 1970s (in some field). Cladistics can be used to infer both character history and phylogeny. The trees that imply the fewest character transformations must be preferred, just as we saw for character histories. For n taxa, at least n – 2 characters with at least two states are required to get a fully resolved phylogeny, since each node needs to be justified by at least one character-state transformation (Fig. 1.12). We also need to know, before we begin the analysis, which of the taxa is the most distantly related to the others (this is the outgroup, as opposed to the ingroup). The identification of the outgroup rests on prior knowledge. For instance, if we wanted to determine the relationships among amphibians, mammals, and reptiles among tetrapods, and if we already knew that lungfishes and the coelacanth are closely related to tetrapods, we could choose lungfishes, the coelacanth, or both, as the outgroup. It is generally advisable to choose the outgroup that is most closely related to the ingroup, although this is not strictly necessary. However, the outgroup must never be part of the ingroup (i.e., it must never be more closely related to part of the ingroup than to another part), or the results will be erroneous. Since most evidence suggests that the lungfishes are more closely related to the tetrapods than to the coelacanth, lungfishes would be the preferred outgroup. It would nevertheless be possible to choose the coelacanth, but under no circumstance could urodeles (salamanders, a group of amphibians) be selected, because they are tetrapods and therefore are more closely related to one part of the ingroup (in this case, frogs and other amphibians) than to another (mammals and reptiles).
Table 1.1. Data Matrix Character 1, limb type: a, fin; b, limb (with digits). Character 2, skin permeability: a, great; b, small.
Taxon |
Character 1 |
Character 2 |
Lungfishes |
a |
a |
Amphibians |
b |
a |
Mammals |
b |
b |
Reptiles |
b |
b |
Table 1.2. More Complex Data Matrix Containing Some Convergence
Taxon |
Character 1 |
Character 2 |
Character 3 |
Character 4 |
Lungfishes |
a |
a |
a |
a |
Amphibians |
b |
a |
a |
b |
Mammals |
b |
b |
b |
b |
Reptiles |
b |
b |
b |
a |
After having established the list of taxa (ingroup and outgroup) and characters, we have to code the data matrix, which could look like Table 1.1.
This matrix suggests the tree shown in Figure 1.12, because this tree requires a single transformation for each character. On the contrary, the tree in Figure 1.13 requires two transitions for one of the characters (skin permeability); it is less parsimonious and so we prefer the tree shown in Figure 1.12 for further testing.
This very simple demonstration hides a far more complex reality. Most analyses include dozens of taxa and many more characters; therefore, finding the most parsimonious tree requires computers and sophisticated software such as PAUP* (Swofford, 2003). The presence of convergence often makes analysis fairly difficult because no tree requires a single transformation for each character, but we must still minimize the number of assumed evolutionary steps. For instance, the matrix in Table 1.2 would result in the tree shown in Figure 1.12, but note that character 4 requires at least two steps on that tree.
So far, only morphological or physiological characters have been discussed, but, for more than a decade, phylogenetic analyses have been performed most frequently on molecular data, particularly on DNA and RNA sequences, and, less frequently, on amino acid sequences. Furthermore, to keep the discussion simple, only parsimony has been presented, but most recent phylogenetic analyses of molecular data use more complex methods (which do not, however, necessarily guarantee more accurate results), such as maximum likelihood and Bayesian analysis (Huelsenbeck et al., 2001).
Dating Taxa
PALEONTOLOGICAL DATING
For a long time, the age of taxa was determined almost exclusively using the fossil record. Indeed, what simpler method than using fossils to date the appearance of taxa could we devise?
Unfortunately, fossils cannot provide complete information about the age of taxa, because only a small proportion of species, and a small proportion of the temporal range of each species, are represented in the fossil record (Fig. 1.14). To determine if paleontological data are reliable for dating taxa, various indices of stratigraphic fit have been developed. Some of these indices use the congruence between the expected order of appearance of taxa (based on the topology of the tree) and the observed order of appearance of taxa as shown by fossils. For instance, according to Figure 1.14, taxon 2 must have appeared before taxon 3, and the fossil record is congruent. On the contrary, taxon 6, which should appear before taxa 7 and 8, appears after them, according to the fossil record, and this implies a long ghost range (a temporal extension of a taxon beneath the lowest stratigraphic level in which it is represented by fossils; it is usually inferred because its sister group is older). Other indices estimate the minimal length of branches unrepresented by fossils, the “Minimum Implied Gap” (MIG), which indicates the minimal proportion of missing data in the fossil record. In all cases, for these measures to be useful, they must be compared with a null distribution of the same index produced by a large number of randomized datasets. Such data sets are produced by randomly permuting the observed ages of taxa (assessed through the fossil record) over the tree a large number of times (typically, a thousand or ten thousand times). Thus, if the stratigraphic fit over the reference tree is better than in at least 95% of the randomized datasets, we can conclude, with a 5% probability of being wrong, that there is a genuine correlation between the predicted and observed order of appearance of taxa (we reject the null hypothesis that no relationship exists between both orders). This conclusion suggests that the paleontological data are reliable. If we cannot reject the null hypothesis (if at least 5% of the randomized datasets display as good a stratigraphic fit as the original data), paleontological data presumably do not provide a reliable estimate of the absolute and relative age of taxa. A computer program can compute these indices and create a null distribution based on randomized datasets through permutations (Wills, 1999). Of course, this method rests on the assumption that the age of the fossils was not used to infer the topology of the tree, and that the latter is reliable. This technique was recently used (Fig. 1.15) to show (by comparison with the results of other methods) that paleontological data provide a reliable estimate of the age of lissamphibian taxa (Lissamphibia is the smallest clade that includes extant amphibians).
Figure 1.12. Most parsimonious tree. Shortest tree for the data matrices (Tables 1.1 and 1.2). Ch. 4, ta. 2, Convergent appearance of character 4 from Table 1.2 on this tree.
Figure 1.13. Less parsimonious tree. Less parsimonious phylogeny for the same data matrices (Tables 1.1 and 1.2).
Figure 1.14. Paleontological dating. Fossils can be used to determine the minimal age of taxa. However, except if a mistake is made, this age will always underestimate the (generally unknown) true age of taxa. The true phylogeny (A) is imperfectly known because only a small fraction of its lineages is represented in the fossil record (bold black lines).
If we possess temporal data of the order of resolution of the fine scale (FS X to FS X + 8), and if we consider that each fossil occurs at the end of the time subdivision in which it occurs, we will underestimate the geological age of the taxa (B). Note that use of a finer temporal scale would not solve this problem. Even an error-free, extremely precise dating of fossils does not resolve the problem of underestimation of the age of taxa according to the fossil record.
If we consider that each species occupies an entire temporal subdivision (C), we no longer estimate a minimal age, but we may still underestimate the age of some taxa (n2, 5, 7), whereas the age of other taxa (n1, 3, 4, 6, 8) is overestimated. This dating method is less biased than using the minimal age of each fossil, and its precision increases with the resolution of the temporal scale used. The MIG (“Minimum Implied Gap,” or ghost range) of taxon 6 is shown in the true phylogeny (A); the MIG of a tree is the sum of MIGs of all the taxa (1 to 9, plus the clades that include these terminal taxa).
Figure 1.15. Time-calibrated supertree of Anura and its closest relatives. This phylogeny incorporates data on topology and on stratigraphy, based on several sources. It was compiled using new software that facilitates the compilation of paleontological trees incorporating stratigraphic information. In that tree, each taxon occupies at least a whole geological stage. For instance, Prosalirus bitis (to the left) occupies the Pliensbachian (from 189.6 to 183 Ma). Each internal branch had a minimal length of 3 Ma. Modified from Marjanović and Laurin (2007).
Paleontological data are fairly easy to use to estimate the age of taxa, but what can be done when a taxon has little or no fossil record? This fairly common situation prevails in soft-bodied taxa lacking a mineralized skeleton, such as annelids and nematodes. Even in taxa with a rich fossil record, gaps may result in serious underestimation of the age of various clades. Thus, even though vertebrates are well represented in paleontological collections, taxa restricted to mountainous areas have left virtually no fossils. This is why other sources of data may be very useful for dating taxa.
Since the 1990s, molecular data have been increasingly used to date the origins of taxa. The first methods rested on the hypothesis of a global molecular clock, which assumed that DNA substitutions accumulated at a steady, constant rate in all taxa and at all times. If we know at least one divergence date within a clade (usually through the fossil record, but occasionally using geological data, such as the separation of continental plates), we can then estimate other dates using molecular data, assuming that molecular distances (the proportion of nucleotides that differ) are proportional to divergence dates (Fig. 1.16). The known divergence date (which is not estimated using molecular data) is called a calibration date or calibration constraint. Two types of calibration dates are known: internal and external. Internal calibration dates occur within the clade that we want to date (Fig. 1.16, IC1–2). External calibration dates occur outside the taxon whose age is to be estimated (Fig. 1.16, EC). The hypothesis of a global molecular clock is occasionally useful, but this is mostly in noncoding portions of the genome (which do not code for proteins and do not regulate gene expression), so most phylogenetically informative data do not fit this model.
In the simplest possible case, if the data fit the molecular clock model, the simplest method is to estimate the rate of molecular evolution and to divide the molecular distances by this rate to obtain the ages of nodes. Thus, in the tree shown in Figure 1.16, and using the data in Table 1.3, the molecular evolutionary rate can be estimated at about 0.000341 substitutions/site/Ma. This is the average value of two evolutionary rates that can be computed: 0.09/250 for lissamphibians and 0.10/310 for amniotes. We could then infer the divergence date between lissamphibians and amniotes by dividing the divergence between each pair of taxa including a lissamphibian and an amniote (Anura/Aves, Anura/Mammalia, Gymnophiona/Aves, and Gymnophiona/Mammalia) by this rate, which yields an average age of 344 Ma. Since the molecular evolutionary rate differs slightly between branches, the estimated ages range from 322 to 381 Ma, but this hypothetical example fits the molecular clock hypothesis nicely (real applications are always more complex than this).
Figure 1.16. Molecular dating. We wish to estimate the divergence date between lissamphibians and amniotes; this is the date of origin of tetrapods. In this phylogeny, we know from the fossil record two internal calibration dates (IC1 and 2), which are the divergence between birds and mammals (at least 310 Ma) and the divergence between gymnophionans and anurans (at least about 250 Ma). We also know an external calibration date (EC), which is not used in this example but could be (even though its age is not as precisely known).
Table 1.3. Hypothetical Molecular Distance Matrices Showing the Proportion of Divergent Nucleotide Sites
Taxa |
Anura |
Gymnophiona |
Aves |
Mammalia |
Anura |
– |
0.09 |
0.12 |
0.11 |
Gymnophiona |
|
– |
0.11 |
0.13 |
Aves |
|
|
– |
0.10 |
Mammalia |
|
|
|
– |
NOTE: This proportion can range from 0 (no difference, which normally occurs only within a small part of the genome and often only within a given species) to 1 (no similarity, a value never reached since there are only four nucleotide types, so we always expect a distance inferior to 0.75). The distance between a taxon and itself is always 0, so the diagonal is represented by dashes (–). Since the matrix is symmetrical (the distance between A and B is the same as between B and A), only half the matrix (without the diagonal) needs to be shown. In this example, only internal calibration dates have been used, but similar calculations could be performed using the external calibration date.
The quartet dating method (Rambaut and Bromham, 1998) is more realistic and was widely used in the late 1990s. This method can use two calibration dates and estimates two evolutionary rates (one per date). Thus, in the example shown in Figure 1.16, the method estimates separately evolutionary rates for amniotes and for lissamphibians. These two evolutionary rates are then used for the three branches to the left, and the three branches to the right of the last common ancestor of tetrapods (identified by a question mark on the figure). This method allows moderate deviations from the global molecular work, since it only requires that the evolutionary rate be constant within each of the two clades on each side of the node whose date we want to estimate. It incorporates an evolutionary model estimated from the sequences (using different software), and this allows for a more precise branch-length estimate accounting for multiple substitutions. This is useful because successive substitutions at a site may result in a return to the initial state; for instance, a site initially occupied by an adenine can be occupied by a thymine, and then switch back to adenine. Without an evolutionary model, we would always conclude, in such cases, that no change has taken place, whereas in fact two changes took place but resulted in identical initial and terminal states. Quartet dating is now rarely used because several more sophisticated methods have been developed.
Among these methods, we find Penalized Likelihood (PL), which can use several calibration points simultaneously, and these can be both internal and external (Sanderson, 2002). Furthermore, uncertainty about these dates can be incorporated into the analysis. Indeed, quartet dating uses calibration dates as if they were known without error, but PL allows specification of lower and maximal values of calibration ages. Thus, the divergence date between birds and mammals is not really 310 Ma ago; instead, it is comprised within the interval between 310 and 345 Ma ago (Marjanović and Laurin, 2007). For some calibration points, it is possible to specify only a minimal age, for others, only a maximal age, or, finally, both can be specified for a given event. This method rests neither on the unrealistic hypothesis of a single, global evolutionary rate nor on two rates; it allows the evolutionary rate to be estimated for each branch, although this rate depends on the rate of neighboring branches (the degree of this dependency can be estimated from the data).
Other methods at least as sophisticated as PL exist. Some are based on a Bayesian approach, but they fall outside the scope of this introduction to molecular dating. A good review of these methods and relevant software was recently published (Rutschmann, 2006).
COMPARISON BETWEEN PALEONTOLOGICAL
AND MOLECULAR AGES
Several molecular studies have dated the diversification of life (e.g., Kumar and Hedges, 1998). A strange but widespread phenomenon is that estimated molecular ages are in most cases considerably older than minimal paleontological ages. Several molecular biologists explain this discrepancy by invoking gaps in the fossil record. On the contrary, several paleontologists blame simplifying assumptions made in the estimation of molecular ages, and important variations in molecular evolutionary rates between taxa and in time.
A partial explanation of this discrepancy lies in the choice of calibration dates. Brochu (2004) showed that molecular ages estimated by quartet dating are proportional to the age of the calibration points used. His demonstration was based on sequences of five mitochondrial genes of crocodilians, a group represented by a rich fossil record that gives reliable data on the true ages of various clades. Thus, using recent calibration points (less than 20 Ma) to estimate the divergence date between crocodilids (crocodiles in the strictest sense) and alligatorids (alligators and caimans) yielded estimates more recent than the age of the oldest fossils (78 Ma) that belong to those clades. The age of this divergence was estimated at less than 30 Ma in some cases, which clearly reveals a major problem in the method. Conversely, using ancient calibration points (more than 50 Ma old) overestimated the age of this divergence, which was sometimes estimated at more than 200 Ma; this is not plausible given our knowledge of the fossil record of crocodiles and the fauna that lived 200 Ma ago. These problems do not affect only quartet dating. It is probable that they affect all molecular dating methods, since Marjanović and Laurin (2007) showed a similar phenomenon using PL. This would explain the huge age difference between some molecular and paleontological estimates of the age of Lissamphibia (Fig. 1.17): the very old ages inferred by Zhang et al. (2005) were based only on two very old external calibration dates. Molecular ages compatible with the paleontological estimates can be obtained if ancient external and more recent internal calibration dates are used simultaneously, and if both minimum and maximum bounds are specified for at least some calibration dates. Unfortunately, many molecular biologists are reluctant to accept paleontological evidence about maximum ages for clades and only use minimum bounds for most calibration dates, which results in biased, inflated ages (Marjanović and Laurin, 2007).
Figure 1.17. Age of lissamphibian diversification. Age of Lissamphibia inferred from mitochondrial data by a Bayesian approach (A), and minimum age estimated by the fossil record (B). The molecular ages include credibility intervals (in parentheses). For paleontological ages, the numbers in parentheses represent the minimal ages implied by various hypotheses about minimal branch lengths. These are not true confidence intervals. Note that the molecular ages are systematically older than the paleontological ages, and that the latter fall outside the credibility intervals of the molecular ages (represented by rectangles). Modified from Marjanovic and Laurin (2007).
After an initial period of unrealistic euphoria, during which numerous scientists thought that molecular dating would enable us to easily date the whole tree of life, we have entered a period of greater realism. Some molecular biologists have strongly criticized the molecular approach and even argued that it should be dropped completely (Shaul and Graur, 2002; Graur and Martin, 2004). Others have discussed the difficulty in estimating time from branch lengths (Britton, 2005). Indeed, molecular distances yield branch lengths, but these lengths depend on both the rate of evolution (which is unknown) and time (also unknown). This means that if only a small proportion of taxa can be dated using fossils (or other geological data), even very long molecular sequences will not necessarily result in very reliable molecular ages. However, as the dating methods become more sophisticated, as the number of available sequences and sampled species increases, and as new calibration dates become available, these methods will become more reliable and lead to plausible results (e.g., Zhang et al., 2008).
Organs of animals often resemble each other in their structure, embryonic origin, and function, but these similarities can result from at least two very different processes. If the similarity results from a shared evolutionary origin, we call these organs homologous, whereas if it results from convergence (independent origin), the organs are analogous. Convergence often arises from the development of organs that perform similar functions in various taxa.
An example of homologous organs is provided by the lung and the swim bladder (Fig. 1.18). The latter exists in teleosts (a taxon that includes most actinopterygians, such as the trout, salmon, and swordfish, among tens of thousands of others) and is a median air-filled structure located against the dorsal wall of the thoracic cavity. By varying the amount of gas in this bladder, teleosts can regulate its volume and, hence, their body density, which enables them to move up or down the water column with minimal energy expenditure. This organ only remotely resembles the tetrapod lung, which is paired (there is a left and a right lung), ventral, and is mostly involved in breathing. Yet, the teleost swim bladder is homologous with the tetrapod lung, despite the differences in function, position, and morphology. In other words, the last common ancestor of tetrapods and teleosts already possessed a structure that gave rise to the tetrapod lung and the teleost swim bladder. We have evidence that this organ was a relatively simple lung (Fig. 1.18).
Figure 1.18. Lung and swim bladder. Lung and swim bladder in transverse section, seen from the front (to the left) and in lateral view (to the right). In all these taxa except in teleosts (A), the lung or swim bladder is involved in gas exchange. In most teleosts, the connection between the swim bladder and the esophagus has disappeared. The oxygen is brought to and removed from the swim bladder by the blood. The phylogeny is shown on the left. Redrawn from figure 3.4 (p. 71) of Graham (1997).
PARTS: A, typical teleost with swim bladder. B, the teleost Erythrinus. C, Lepisosteus and Amia. D, Polypterus and Erpetoichthys. E, the dipnoans (lungfishes) Lepidosiren and Protopterus. F, tetrapods.
As examples of analogous organs, the lung and gills may be mentioned. Gills of primitively aquatic vertebrates (sharks, trout, etc.) are the main organ for gas exchange (although they are also involved in osmoregulation), just like the tetrapod lung. Thus, gills and lungs perform the same function. However, their evolutionary origin differs, as shown by the presence of both lungs and gills in dipnoans (the presence of both organs in a species proves that they are not homologous), and by the numerous anatomical and developmental differences (for instance, gills are outgrowths of the aortic arches, whereas lungs are outgrowths of the pharynx). This is why lungs and gills are analogous structures.
Since the 19th century, geologists have been mapping geological strata on our planet, trying to establish the relative age of rocks. This chronology is based (at least for sedimentary rocks) on the principle of superposition (generally, the lowest, deepest strata are the oldest, because successive strata are laid on top of each other), on the fossil record, and, since the 20th century, on radiometric methods that can yield absolute ages (in millions of years) of certain rocks. We have thus established that the Earth is about 4.56 Ga (billion years) old, and that life probably appeared more than 3 Ga ago, even though fossils are extremely rare before the Cambrian, which started a measly 542 Ma ago according to the latest dating. The geological times were thus divided into the Cryptozoic (which means “hidden life”), also called the Precambrian (both terms are informal and refer to the formal eons Archean and Proterozoic), and the Phanerozoic (which means “visible life”). This means that only the last 12% of the Earth’s history, and at most 25% of the history of life, is represented by the rich Phanerozoic fossil record. Towards the end of the Precambrian, animals (metazoans) appeared, even though they are represented by few fossils, most of which are difficult to interpret.
The Phanerozoic is subdivided into three eras of unequal duration: the Paleozoic, which means “ancient life” (from 542 to 251 Ma ago), the Mesozoic, which means “middle life” (from 251 to 65.5 Ma ago), and the Cenozoic, which means “recent life” (from 65.5 Ma ago to the present). Vertebrates appeared and diversified in the Paleozoic. The conquest of land by plants and various animal groups, including vertebrates, likewise took place during the Paleozoic, which is the era on which this book focuses.
Figure 1.19. Simplified geological time scale. This geological time scale emphasizes the periods during which the oldest terrestrial vertebrates lived. The drawing is not to scale, since the Cryptozoic should be about eight times as long as the Phanerozoic. The limb with digits may have appeared in the Middle Devonian if the dating of the recently described trackway from Poland is correct (Niedzwiedzki et al., 2010).
The Paleozoic comprises six periods, which are, from the oldest to the most recent, the Cambrian, the Ordovician, the Silurian, the Devonian, the Carboniferous, and the Permian (Fig. 1.19). The first gnathostomes (jawed vertebrates) appeared in the Ordovician or Silurian, but the limb with digits appeared only in the Devonian. The first truly terrestrial vertebrates appeared in the Carboniferous, the period during which the oldest amniotes lived.
Fossils occur in many places; it is thus not possible to list all sites that have yielded early limbed vertebrates or their precursors. Only localities that have yielded many specimens or especially important taxa are described below (Figure 1.20).
Figure 1.20. Distribution of vertebrate fossiliferous sites from the Late Devonian and Carboniferous. This map shows the position of continents in mid-Early Carboniferous times (335 Ma ago). Sites from the Late Devonian (1 to 7), from the Early Carboniferous (8 and 9), and from the Late Carboniferous (10 and 11) are shown The Paleo-Tethys ocean occupied a position vaguely similar to that of the present-day Mediterranean and the Indian Ocean. The Variscan (or Armorican) orogeny was caused by the collision between Laurasia (a northern supercontinent that included North America, Baltica, and Siberia) and Avalonia (a small plate that included the Avalon peninsula of Newfoundland, Nova Scotia, England, the northern half of Germany, and part of Poland; it was located a bit south of numbers 10 and 11). It formed mountains whose traces remain in Brittany, in the French Massif Central, in the Ardennes, in central Germany, and in the northern part of the Appalachian Mountains. Later, this complex collided with Gondwana and Asia to form Pangea, which encompassed most continents from the Carboniferous to the Early Jurassic. Modified from Gradstein et al. (2004).
LOCALITIES: 1, Miguasha, Quebec, Canada. 2, Red Hill, Pennsylvania, USA. 3, eastern Greenland. 4, New South Wales, Australia. 5, Ningxia Hui Autonomous Region, China. 6, Tula, Russia. 7, Ellesmere island, Canada. 8, Delta, Iowa, USA. 9, East Kirkton, Scotland. 10, Joggins and Florence, Nova Scotia, Canada. 11, Nýřany, Czech Republic.
ABBREVIATIONS: BAL, Baltica (a continental plate that included a major portion of central and western Europe). GON, Gondwana (southern continent that included South America, Africa, Antarctica, Australia, India, and other fragments). NAM, North America. SIB, Siberia.
Fossils of our closest finned relatives have been found in several localities, including Miguasha, in the province of Quebec (Canada). That site is especially famous for the Escuminac Formation (rock stratum), which dates from the Late Devonian (about 380 Ma). It has yielded numerous fossils of the sarcopterygian Eusthenopteron, which is one of the best-known Devonian vertebrates thanks to the detailed anatomical descriptions of the Swedish paleontologist Erik Jarvik (1980). Elpistostege, which is more closely related to us, was also found in Miguasha. Tiktaalik is still closer to us, although it, too, retained paired fins; it was found in southern Ellesmere Island in the Canadian Arctic. Several sarcopterygians more distantly related to limbed vertebrates were found in the Gogo Formation (also Late Devonian) in western Australia.
The first limbed vertebrates come from eastern Greenland and date from the end of the Devonian, a little more than 360 Ma ago. They include Ichthyostega and Acanthostega. A similar, contemporary taxon called Tulerpeton was found near Tula, Russia. Fragmentary remains suggest that, in the Late Devonian, other limbed vertebrates or their near relatives also lived in the territories that now fall into the Baltic countries, Belgium, Australia, and China.
After the Devonian, “Romer’s gap” (360–345 Ma), which may have been caused by the low atmospheric concentration of oxygen that prevailed around that time (Ward et al., 2006), hampers research on Early Carboniferous continental vertebrate faunae. Slightly more recent Early Carboniferous tetrapods (345–326 Ma) are known mostly from the coal-rich formations of Great Britain. These may include the first truly terrestrial vertebrates, some of which had already lost their limbs (Germain, 2008). Similar, contemporary forms have been found in West Virginia and Iowa (USA).
In the Late Carboniferous, several fossiliferous sites now located in North America and Europe document the extensive diversification of limbed vertebrates. For instance, at least 26 species (Hook and Baird, 1986) were found in Linton (Ohio, USA). All these sites were then located close to the Equator, and North America then formed, along with much of Europe, a continent called Euramerica. It has not been possible to establish if the apparent absence of limbed vertebrates from higher paleolatitudes reflects their climatic preferences, or if this is simply caused by the absence of fossiliferous localities preserving the appropriate environments. The first amniotes appeared in the Late Carboniferous and are represented by fossils found at Joggins in Nova Scotia (Canada); these demonstrate that the conquest of land by vertebrates was completed by about 315 Ma ago. Joggins is unusual to the extent that most vertebrate fossils found there were preserved in the fossilized stumps of tree-sized lycopods (club mosses), such as Sigillaria and Lepidodendron. It is thought that after the death of these lycopods, the loose tissue in the pith at the base of the stumps decomposed, thus leaving a hollow stump into which small vertebrates fell. The nearby site of Florence, also in Nova Scotia, similarly preserves slightly more recent (310 Ma) limbed vertebrates in giant lycopod tree stumps.
To study the evolution of life, we need a precise system of nomenclature adapted to dealing with taxa, since we already know millions of species, and since we cannot classify evolving beings in the same way as universal entities such as atoms. For a long time, rank-based nomenclature was used to classify living beings, but this system is ill-adapted to biological classification, since taxa are individuals rather than classes. Thus, a phylogenetic nomenclature was developed to better delimit taxa with the help of the phylogeny (the tree of life).
Inferring phylogenetic relationships between taxa and character evolution requires relatively sophisticated methods. Thus, even when a rich fossil record is available, as is often the case with vertebrates, we cannot read the history of the group directly from it; this history must be reconstructed through various methods, such as parsimony (one of the simplest and most widely used methods, at least for morphological data), maximum likelihood, and the like. These methods, developed mostly in the 1960s, have triggered a true revolution in systematics and allow a tremendous gain in objectivity and, probably, in precision as well. We can thus study the transformation of homologous structures (such as the lung and swim bladder), and trace the origin of taxa in the distant past. The age of origin of taxa can be determined either through the fossil record, which yields a minimal age, or through various molecular dating techniques, which estimate the time of origin of taxa using molecular data (DNA or, more rarely, proteins). Molecular dating methods have progressed steadily in the last decade, but their use remains difficult; many studies, even recent ones, have obtained ages that seem unrealistic when compared with paleontological ages. However, recent progress in these methods, and the increasing number of available calibration dates, makes molecular dating a very useful tool.
The geological time scale is divided into the Cryptozoic (hidden life) and the Phanerozoic (visible life). The latter is subdivided into the Paleozoic, Mesozoic, and Cenozoic. The conquest of land occurred in the Paleozoic.
