Fish Physiology: Primitive Fishes

Fish Physiology

Primitive Fishes

David J. McKenzie, Anthony P. Farrell and Colin J. Brauner

ISSN  1546-5098

Volume 26 • Suppl. (C) • 2007

Table of Contents

Cover image

Title page

Living Primitive Fishes and Fishes From Deep Time

1 Introduction

2 Primitive Characters, Primitive Taxa, and Ancient Taxa

3 Living Fossils

4 Living Primitive Fishes in Vertebrate Phylogeny

5 Living Primitive Fishes and Their Fossil Relatives: Naming and Dating Taxa

6 Extinct Major Fish Taxa and Their Position in Vertebrate Phylogeny

7 How Stable is Vertebrate Phylogeny?

8 Fossils and Physiology

9 The Environment of Early Fishes: Marine Versus Freshwater Vertebrates

10 Conclusions

Cardiovascular Systems in Primitive Fishes

1 Introduction

2 An Overview of Evolutionary Progressions

3 Details of the Cyclostome Circulatory Systems

4 Details of the Sarcopterygii (Lobe‐Finned Fishes) Circulatory Systems

5 Details of the Circulatory Systems in Polypterids, Gars, and Bowfins

6 Details of the Sturgeon Circulatory Systems

7 Conclusions

Nervous and Sensory Systems

1 Introduction

2 Development of the CNS

3 The Brains of Primitive Fishes

4 Functional Classification of Cranial Nerves in Fishes

5 The Visual System

6 Chemoreceptive Systems

7 Octavolateralis System

8 Electroreception

9 Concluding Remarks

Ventilatory Systems

1 Introduction

2 Respiratory Strategies

3 Respiratory Organs

4 Ventilatory Mechanisms

5 Respiratory Control

6 Conclusions

Gas Transport and Exchange

1 Introduction

2 Partitioning of O2 and CO2 Exchange Across the Respiratory Surfaces

3 Blood O2 Transport

4 Transport and Elimination of CO2

5 Synthesis

Ionic, Osmotic, and Nitrogenous Waste Regulation

1 Introduction

2 Ionic and Osmotic Regulation

3 Nitrogen Excretion

4 Concluding Remarks

Locomotion in Primitive Fishes

1 Introduction

2 Swimming Modes and Associated Morphological Adaptations

3 Locomotor Muscles

4 Neuromotor Coordination

5 Locomotor Performance and Physiology

6 Conclusions

Peripheral Endocrine Glands. I. The Gastroenteropancreatic Endocrine System and the Thyroid Gland

1 Introduction

2 Endocrine Pancreas and Related Gastrointestinal Endocrine System

3 Thyroid Gland

4 Summary and Conclusions

Peripheral Endocrine Glands. II. The Adrenal Glands and the Corpuscles of Stannius

1 Introduction

2 Adrenal Glands

3 Corpuscles of Stannius

4 Summary and Conclusions

Why Have Primitive Fishes Survived?

1 Introduction

2 Life During the Early Phanerozoic

3 The Teleosts

4 Primitive Fishes: Relationships Between Groups

5 Why Have These Primitive Fishes Survived?

6 Conclusions

Series Editors

Contributors to Volume 26

Preface

Index

Contents of Previous Volumes

Living Primitive Fishes and Fishes From Deep Time

Philippe Janvier

Abstract

The notion of primitive living taxon, or living fossil, largely stems from the evolutionary concepts that have pervaded systematics for nearly a century, notably the view that paraphyletic taxa are real and ancestral. Certain living taxa are regarded as primitive because some of their characters remain in a plesiomorphic state relative to their homologues in other, closely related living taxa, and this assessment rests on both out‐group comparison and fossil data. The biology of primitive living taxa is thus supposed to mirror that of the related fossil taxa they resemble. Physiologists, therefore, bet that the physiological functions of a reputedly primitive living fish are the same as those of its fossil anatomical proxies, but paleontologists often infer those of the latter on the basis of primitive living models. In some cases, such circular reasoning can be avoided by considering paleoenvironmental data that are inferred preferably from geochemical parameters. An overview of living and fossil vertebrate phylogeny, however stable it may seem, shows that there are several ways of defining and naming taxa, and that shared physiological characters of a crown group may not be extrapolated to its stem group, the divergence of which may be much earlier. Physiological characters are probably no more and no less homoplastic than morphological characters. Like the latter, they can be decomposed into series of states that can be included in fractioned and combined parsimony analyses, and can contribute to patterning the trees, instead of being interpreted a priori as adaptive and mapped as attributes on trees based on other kinds of characters.

1 Introduction

By comparison to that of the morphological or molecular characters, the question of the homology of physiological characters has been little debated during the last three decades of the twentieth century, which roughly correspond to the time of the cladistic revolution in comparative biology. The reason for this neglect is that physiology was long regarded as a discipline of general biology: the biology of processes, as opposed to comparative biology, that is, the biology of patterns, as outlined by Nelson (1970, 1994). There is indeed an old tradition of considering physiological characters as highly adaptive; that is, they are assumed to be commonly subject to homoplasy and thus their distribution, however hierarchical it may sometimes look, tells us little about their evolutionary history. There are multiple historical reasons for this deep‐rooted belief, some of which date back to the nineteenth century, possibly with a scent of Lamarckism, but there is no clear evidence that physiological characters are more adaptive than such anatomical structures, as the pattern of the skull bones or tail skeleton morphology. Nevertheless, already in the early twentieth century, some physiologists pointed out congruences between the laws based on morphological character distributions (or evolution) and the presumed history of physiological characters. For example, Needham (1938) came to the conclusion that, following the morphology‐based Dollo's Law, losses of physiological functions are irreversible.

During the past two or three decades, there have been some attempts at exploiting the phylogenetic message of physiological characters, notably by Løvtrup (1977), who was the first to suggest cyclostome paraphyly on this ground (see below). More recently, Cunchillos and Lecointre (2005) demonstrated that metabolic pathways could be described as nested sets of character states and coded like any other kind of characters in a data matrix aimed at reconstructing phylogenetic relationships between taxa. Thus, there are multiple ways of including physiological traits in phylogenetic analyses, by considering either the distribution of a particular function (coded as absent/present) or that of different states of a function (as a hierarchy of more and more complex pathways). A classical bias in physiology‐based phylogenies is perhaps that physiologists readily know (or think they can readily know) the selective advantage of a physiological character, notably by means of experiments. Therefore, they are tempted to first make inferences about the evolution of physiological characters on the basis of the supposedly known history of the environment or behavior of an organism. In contrast, morphologists generally can make only vague inferences about the selective advantage of morphological characters, not to speak of molecular phylogeneticists, whose nucleotide sequences tell little about their impact on the phenotype. Yet comparative functional genomics may soon provide information in this field.

It is thus time to restore the consideration of physiological characters as a source of potential shared homologies, irrespective of the morphological characters they are inferred from, and stop considering that their interest essentially lies in their adaptive plasticity, that is homoplasy. Physiological characters are no worse, no better than any other legacy of evolution: they provide examples of both phylogenetic messages (synapomorphies) and adaptive convergences (homoplasies), but their assessment is always relative, in the light of parsimony. There are nevertheless certainly some very robust physiological signatures in phylogeny (e.g., uric acid excretion in sauropsid amniotes), which can be regarded as being just as good node supports as, for example, gnathostome jaws or tetrapod limbs.

In this introduction, I should like first to make clear that the relationships between organisms (and thus the criteria on the basis of which we decide whether the latter are primitive) are exclusively based on assumptions about homology relationships between parts of these organisms, be they anatomical or physiological characters, or even nucleotide sequences. Therefore, the phylogenetic trees from which physiologists may infer evolutionary patterns are mere theories based on most parsimonious character distributions, and in which fossils provide additional character combinations, as well as information about the minimum age of characters and taxa. Then, I shall briefly depict the lost world of the ancient fossil fishes, on which rests the notion of living primitive fishes, or, more generally, living fossils.

2 Primitive Characters, Primitive Taxa, and Ancient Taxa

Taxa reflect relationships between organisms, which are inferred from homology relationships between parts of these organisms (Nelson, 1994). Consequently, a theory of phylogenetic relationships (and thus a phylogenetic classification) reflects the most parsimonious distribution of congruent homology relationships at one particular time and for one particular sample of characters and terminal taxa. The structure of the vertebrate tree at the level of the higher terminal taxa has been relatively stable during the past two decades, despite some divergent theories that essentially arose from conflicts between morphological and molecular sequence data (see below). However, in detail, the numerous trees of each of the higher terminal taxa, expressed in the same graphic way (a branching diagram), are derived from several, entirely different, conceptual backgrounds, such as genetic distance (phenetics), parsimony, or model‐based approaches such as maximum likelihood or Bayesian approaches. Although phylogeneticists now tend to provide, for the same data set, the trees that are yielded by these respective methods and generally consider that the differences are minimal, this often generates confusion because subsequent authors often compare trees that are not, in fact, comparable. Phylogenetic trees are doomed to remain theories forever, some of which will be less and less frequently refuted. Therefore, the unending quest of phylogenetics (i.e., to progressively tend toward a more and more stable pattern of relationship) requires ever more data, and, above all, different kinds of data of approximately equal quality. Physiological characters are certainly underexploited to this end. Perhaps genomics (i.e., parsimony‐based analysis of the organization of the genome as a whole, rather than mere sequences of particular genes) may also provide one of these new sources of data.

Vertebrates are currently regarded as a taxon because they are the only living beings that share such diverse characters as migrating neural crest cells, massive gene duplication, or a labyrinth with at least one semicircular canal with ampullae. The congruence of these character distributions is, to date, not contradicted by conflicting distributions of other characters, apart from minor examples currently regarded as homoplasies. Although phylogeneticists focus their interest on the search for shared derived characters (synapomorphies) and thus regard shared primitive characters (symplesiomorphies) as uninformative, other biologists whose main interest lies in biological or evolutionary processes consider living organisms as a functional assemblage of characters that can tell us something about the biology of a hypothetical ancestor at one particular node of the tree of life. Thus, the closer to that node, the more interesting the real organisms! This is particularly true for physiologists, who pay much attention to the so‐called primitive living taxa, that is, extant taxa that retain a large number of symplesiomorphies. For example, hagfishes have a number of characters, notably physiological ones, that are lacking in all other vertebrates (lampreys and gnathostomes, or jawed vertebrates), but shared with other chordates (cephalochordates, tunicates), and other deuterostomes (Jørgensen et al., 1998). These plesiomorphous characters are thus regarded as primitive and inherited from invertebrates, and tell us nothing about the particular relationships of hagfishes but their absence (or presumed modification) in lampreys and gnathostomes does as they exclude hagfishes. However, hagfishes also share some unique anatomical characters with lampreys (Yalden, 1985), and these homology relationships might also suggest that hagfishes and lampreys are a taxon, the cyclostomes, unless these characters are lost in jawed vertebrates. Relationships between parts of organisms almost inevitably conflict, and relationships between organisms are, in principle, never stable. They depend on the number and quality of characters that biological research progressively and endlessly pours into the bag of comparative biology. The principle of parsimony is currently considered the best way to choose one theory of relationships between organisms: the one that is supported by the largest number of congruent derived character distributions beyond the bounds of chance.

However, this way of considering homology relationships (and thus relationships between taxa) developed only in the 1960s, with the rise of Hennig's (1950) phylogenetic systematics. Before that time (and occasionally still now), evolutionary systematics defined taxa on the basis of the overall resemblance of the organisms they include, and a demarcation was arbitrarily drawn between groups of organism that possessed a character and others that did not, but both groups were named and thus were equally regarded as taxa. Therefore, we still find such names as protozoans (versus metazoans), invertebrates (versus vertebrates), fishes (versus tetrapods), anamniotes (versus amniotes), and so on, used by some biologists, not in a colloquial context but as real groups of organisms whose history has a beginning and an end. The beginning of fishes is that of vertebrates, but do fishes end with the first tetrapod? In all these cases, the group that lacks the character that defines the other group is generally said to be primitive or ancestral. The former is what Hennig (1950) called a paraphyletic group (or a grade), the latter is what he called a monophyletic group (or a clade). In a historical context, these notions are admittedly relative, and fishes were monophyletic before the rise of limbs (and thus tetrapods).

Grades were very convenient to paleontologists for accommodating fossils whose relationships were unclear and whose anatomy displayed an overwhelming number of general (i.e., plesiomorphous) characters. Among vertebrates, agnathans (jawless fishes), crossopterygians (lobe‐finned fishes), or paleoniscoids (primitive ray‐finned fishes) were such groups that are only defined by the lack of the characters of their presumed descendants. Primitive living fishes are thus such extant taxa that have once been classified in a grade along with many extinct taxa.

3 Living Fossils

The term living fossil, coined by Darwin in his Origin of Species, became widespread in the literature by the end of the nineteenth century. It has been applied to a wide range of modern taxa for a variety of reasons. The concept of living fossil is in general linked to evolutionary classifications and thus to the notion that grades are taxa. The case of the living actinistian (coelacanth) Latimeria is a good example of how a taxon became regarded as a living fossil. The status of living fossil assigned to Latimeria is essentially due to the fact that actinistians (which were known only as fossils for almost a century) have long been classified in a taxon Crossopterygii (crossopterygians), which also included a number of Paleozoic taxa and is now considered to be paraphyletic because it includes stem tetrapods (rhizodontids, osteolepiforms, elpistostegalians) and stem dipnoans (porolepiforms, youngolepidids; Figure 1.1). All the crossopterygian characters that Latimeria shares with these taxa are merely general sarcopterygian or even osteichthyan characters (e.g., monobasal‐paired fin skeleton, intracranial articulation, large notochord, Figure 1.1A), which have been modified or lost in most of the living descendants of crossopterygians (Figure 1.1C and D). Although actinistians in general show a relatively stable morphology during the past 380 million years (Myr), Latimeria is no more a living fossil than many other fish taxa, such as lungfishes, and probably less so, unless by the virtue of its retaining a few plesiomorphous characters, such as the intracranial joint (Figure 1.1B). The survival of living fossils is thus rather a survival of some characters that were once general, but have largely been lost in related extant taxa (Janvier, 1984).

Fig. 1.1 Latimeria as a living fossil. In typical evolutionary classifications, Latimeria is often referred to as the only living crossopterygian, an essentially Paleozoic, paraphyletic group that includes stem lungfishes and stem tetrapods (in gray). Crossopterygians are in fact diagnosed on the basis of characters that have been lost or modified in their descendants, lungfishes and tetrapods. One of these characters is the intracranial joint (arrowheads), a general (plesiomorphous) character of the sarcopterygian braincase (A), conserved in Latimeria (B), but lost in lungfishes (C) and tetrapods (D). (A) Braincases of Nesides, (B) a late Devonian actinistian, Latimeria, (C) Neoceratodus, and (D) a primitive Carboniferous tetrapod. [Based on Janvier (1984, 1996a) by permission of Oxford University Press.]

The concept of living fossil can also be seen in a different way, that is when conspicuous specializations appear very early in phylogeny, and are retained during a long period of morphological stability. This refers to the panchronic taxa which, among fishes, are relatively rare. Example of such panchronic taxa are hagfishes and lampreys (Figure 1.2). The earliest fossil hagfishes are 305 Myr old, and the earliest lamprey 360 Myr, and, as far as it can be seen from fossils preserved as soft‐tissue imprints, looked practically identical to the living hagfishes and lampreys (Figure 1.2A, C, E, and G). Hagfishes already possessed the characteristic tentacles (Figure 1.2B and D) and lampreys already possessed a piston cartilage (hence a protractible and retractable tongue), an annular cartilage surrounding the mouth (Figure 1.2F and H), and a sucker (Figure 1.2I and J; see also Gess et al., 2006 for new data). Such cases of panchronic taxa are perhaps the only ones that can actually be referred to as living fossils since it does not rest on the retention of a few plesiomorphous characters, but the conservation (or survival) of a large number of apomorphous characters that appeared early in time and have remained unmodified (Janvier, 1984). Large gaps in the fossil record, like in the case of Latimeria (for which there is no Tertiary record), are also regarded as a criterion to assess a living taxon as a living fossil.

Fig. 1.2 Hagfishes and lampreys are regarded as living fossils, not because they retain characters that have once been general to a larger group but because they have gained some highly derived characters that remained extremely stable throughout time. The overall morphology (A) and skull structure (B) of living hagfishes strikingly resemble those of the late Carboniferous hagfish Myxinikela (C and D). Although Myxinikela is somewhat stouter in body shape than the living forms (C), it already displays the main unique characters of the living hagfish head skeleton (D), such as the tentacles and long prenasal duct. Similarly, the overall morphology (E) and head skeleton of the living lampreys (F) are almost identical to those of the late Carboniferous Mayomyzon (G and H). Both share the characteristic annular and piston cartilages. The late Carboniferous Pipiscius (I), regarded as a fossil lamprey, also possessed an oral funnel (J) armed with horny plates, which strikingly resembles that of living lampreys (K). Scale bars = 10 mm. [A–C, E–G, J, and K, from Janvier (1984, 1996a by permission of Oxford University Press); H, redrawn and modified from Bardack and Zangerl (1968); D, redrawn and modified from Bardack (1991); and I, redrawn and modified from Bardack and Richardson (1977).]

The notion of living fossil, in common with evolutionary systematics, remains rather vague and subjective. This is reflected in Eldredge and Stanley's (1984) review of the different taxa that may be regarded as such and which includes elopomorph teleosts but, strangely, not lampreys. Whatever the ground for regarding a taxon as a living fossil, it always rests on morphology since this is essentially what is known in fossils. It is, of course, tempting to consider that if the morphology of an extant taxon has undergone little change through time, the same may apply to its physiological characters that leave no fossil record. This postulate, which cannot be tested by direct evidence, is nevertheless often regarded as the only means for attempting a reconstruction of the history of physiological characters. Thus, the only possible test for theories about physiological character phylogeny is their congruence with molecular sequence‐based phylogenies of living taxa and/or morphological character distributions in living and fossil taxa.

4 Living Primitive Fishes in Vertebrate Phylogeny

Living fish taxa that are traditionally regarded as primitive are essentially the jawless vertebrates or agnathans [hagfishes (Hyperotreti) and lampreys (Hyperoartia)], and, among the gnathostomes, some chondrichthyans [batomorphs (sawfishes, torpedoes, skate, and rays), hexanchiform sharks, Chlamydoselachus, and chimaeriforms], cladistians (i.e., polypteriformes or bichirs), acipenceriforms (sturgeons and paddlefishes), ginglymods (or lepisosteiforms, i.e., gars), Amia calva (bowfin), actinistians (Latimeria), and lungfishes (or dipnoans). In addition, the osteoglossomorphs (e.g., bony‐tongues) and elopomorphs (e.g., tarpons, eels) are sometimes regarded as examples of primitive teleosts. The trees in Figure 1.3 show the relationships of the major living vertebrate taxa, with special reference to the so‐called primitive fishes. The fully resolved tree on the left‐hand side shows the phylogeny that is most widely accepted by morphologists and paleontologists. The tree on the right‐hand side shows a consensus of the various trees based on either morphological or molecular sequence data, which may yield slightly different topologies (hence, the debated nodes A–D for which different data sets provide different topologies for the more crownward taxa).

Fig. 1.3 Interrelationships of the living vertebrates, with particular reference to the reputedly primitive fishes. The tree on the right‐hand side shows the four major polytomies (A–D), which are essentially due to conflicts between morphological and molecular sequence data. The question of the relationships of sharks and batomorphs is regarded here as resolved as there is increasingly strong support for their sister‐group relationships (thus shark monophyly). The tree on the left‐hand side depicts the topology that is most widely accepted by morphologists and paleontologists. Main clades and selected characters applying to living taxa: 1, craniates (migrating neural crest cells, epidermal placodes, skull, olfactory, optic, and otic capsules); 2, vertebrates (arcualia, extrinsic eye muscles, radial muscles in unpaired fins); 3, gnathostomes (jaws, horizontal semicircular canal, paired fins, calcified or ossified endo‐ and exoskeleton, epicercal tail); 4, chondrichthyans (prismatic calcified cartilage, pelvic claspers); 5, elasmobranchs (posteriorly directed basibranchials, paired occipital condyles); 6, osteichthyans (endochondral bone, large dermal bones covering the head and shoulder girdle, lepidotrichs); 7, actinopterygians (only one dorsal fin, primitively ganoid scales, acrodin cap on teeth, everted telencephalon); 8, actinopterans (fringing fulcla on the leading edge of fins); 9, acipenseriforms (anterior symphysis of palatoquadrate); 10, neopterygians (unpaired fin lepidotrichs equal in number to their supports); 11, halecostomes (mobile maxillae); 12, teleosts (median tooth plate on basihyal, mobile premaxillae); 13, elopocephalans (two uroneurals); 14, sarcopterygians (monobasal paired fins, pulmonary vein, veina cava); 15, lungfishes (massive entopterygoid and prearticular tooth plates); 16, lepidosirenidae (reduced paired fins); and 17, rhipidistians, or choanates (alveolae in lungs, partially divided and sigmoid arterial cone in heart, incipient atrial septum). [Illustrations for terminal taxa from Janvier (1996a) by permission of Oxford University Press.]

There are, however, some widely different, molecular sequence‐based phylogenies, notably that proposed by Arnason et al. (2001, 2004) and Rasmussen and Arnason (1999a,b), one of which is shown here in Figure 1.4. To date, these phylogenies are difficult to reconcile with most of the currently available morphological data as they support fish monophyly but not osteichthyan, sarcopterygian, and actinopterygian monophyly. However, provided that they are merely noise due to inappropriate gene sequences, these odd phylogenies (Janvier, 1998) are not to be ignored and should stimulate a new look at certain conflicting character combinations displayed by some early fossil chondrichthyans and osteichthyans (Janvier, 1998; Zhu et al., 1999, 2006; Maisey, 2001), which, at any rate, suggest that some of the classical osteichthyan and sarcopterygian characters may be more general than currently believed.

Fig. 1.4 Mitochondrial DNA‐based vertebrate tree (hagfishes not considered) proposed by Arnason et al. (2004). This tree is strongly at odds with the current consensus (compare to Figure 1.3), notably by the breakdown of the osteichthyans, sarcopterygians, actinopterygians, and neopterygians. [Illustrations for terminal taxa from Janvier (1996a) by permission of Oxford University Press.]

Current morphology‐based gnathostome phylogenies (basically that in Figure 1.3, left‐hand tree), notably the assumption that chondrichthyans are the plesiomorphous sister group of osteichthyans, stem from (or is consistent with) Huxley's (1880) conception that chondrichthyans were ancestral to osteichthyans. Most molecular sequence‐based phylogenies have proven to be consistent with the monophyly of these two respective groups (Hedges, 2001; Zardoya and Meyer, 2001). Arnason et al. (2004) pointed out, however, that molecular (mitogenomic) sequence‐based osteichthyan phylogenies may be biased by the fact that they are generally rooted with chondrichthyans, rather than being rooted with either lampreys or hagfishes.

When considering the current (or conventional) consensus tree for higher vertebrate taxa, only three major piscine nodes remain controversial: the hagfish‐lamprey‐gnathostome node, the gar‐bowfin‐teleost node, and the coelacanth‐lungfish‐tetrapod node. However, one must concede, in agreement with Arnason et al. (2004), that the morphological support to the osteichthyan clade is relatively low and even more so when early fossil taxa are considered.

4.1 The Hagfish‐Lamprey‐Gnathostome Node

Since the early nineteenth century, lampreys and hagfishes have long been gathered in a group, the Cyclostomi (cyclostomes) characterized by horny teeth covering a protractible and retractable tongue, and pouch‐shaped gills. All other cyclostome features are either absence of gnathostome characters or characters whose state cannot be assessed on the basis of out‐group comparison (i.e., not applicable to nonvertebrate chordate taxa). Since the mid‐twentieth century, morphologists and physiologists had long been aware that lampreys were in many respects more similar to gnathostomes than to hagfish, but the apparently more simple or invertebrate‐like characters of hagfishes were regarded as a consequence of degeneracy, due to their legendary endoparasitic habits (already alluded to by Linnaeus; Jørgensen et al., 1998). Løvtrup (1977) was the first to suggest clearly that this appeal to degeneracy was groundless and that lampreys display an overwhelming number of morphological and physiological characters shared only with gnathostomes, which suggest sister‐group relationships between these two taxa (2, Figure 1.3). Then, cyclostome paraphyly became progressively accepted by many comparative biologists (Hardisty, 1982; Maisey, 1986; Janvier, 1996a, b). In contrast, molecular sequence‐based phylogenies tended to support cyclostome monophyly (Stock and Whitt, 1992; Delarbre et al., 2000; Hedges, 2001; Mallatt et al., 2001), although analysis of data sets partitioned into small and large subunit components provided conflicting support for both monophyly and paraphyly (Zrvav'y et al., 1998). Cyclostome monophyly in molecular sequence‐based phylogenies was first regarded as an artifact of long‐branch attraction, but it is still strongly supported by Bayesian methods, which are supposed to minimize this bias (Furlong and Holland, 2002). Some molecular sequence‐based data, however, rather support the position of lampreys as sister group to gnathostomes (cyclostome paraphyly) such as those based on small RNA units (Gursoy et al., 2000). Currently, the problem is still unresolved. To morphologists and physiologists, cyclostome monophyly would imply an impressive number of either reversions in hagfishes or convergences in lampreys and gnathostomes.

4.2 The Gar‐Bowfin‐Teleosts Node

Gars and bowfins have long been grouped in a group called Holostei (holosteans), along with a number of fossil neopterygian actinopterygian taxa (e.g., semionotids, parasemionotids, macrosemiiforms). However, Patterson (1973) regarded bowfins as more closely related to teleosts than gars. Bowfins and teleosts were thus gathered in the clade Halecostomi (halecostomes), characterized notably by mobile maxillae (11, Figure 1.3). Apart from a paleontology‐based phylogeny proposed by Olsen and McCune (1991), very few anatomical data support a sister‐group relationship between gars and teleosts. Similarly, no anatomical data supports the topology proposed by Inoue et al. (2003) and Arnason et al. (2004), in which gars, bowfins, and acipenseriforms form an unresolved clade, sister to the teleosts (Figure 1.4). Conversely, a reconsideration of fossil neopterygians anatomy (notably for early gars, bowfins, semionotids, and parasemionotids) as well as molecular sequence‐based neopterygian phylogenies (Kikugawa et al., 2004) now strongly supports holostean monophyly again; that is, gars and bowfins would be sister groups.

4.3 The Coelacanth‐Lungfish‐Tetrapod Node

Piscine sarcopterygians are only represented by two strongly depauperized living taxa: actinistians (coelacanths) and dipnoans (lungfishes). The living coelacanths are represented by Latimeria chalumnae from the Strait of Mozambique, and possibly a second species Latimerai menadoensis from Sulawezi (Indonesia). Lungfishes fall into three genera, the Australian Neoceratodus (one species), the African Protopterus (four species), and the South American Lepidosiren (one species). Neoceratodus is the sister group of the clade Lepidosirenidae, which includes the other two genera (15, Figure 1.3).

Although long debated, the question of actinistian‐lungfish‐tetrapod relationships is now generally regarded as settled by morphologists and paleontologists (Rosen et al., 1981; Cloutier and Ahlberg, 1996; Zhu et al., 2006), with coelacanths being sister to lungfishes and tetrapods, despite a few anatomical data that may support an actinistian–tetrapod sister‐group relationship. In contrast, molecular data remain ambiguous on this issue (Zardoya and Meyer, 1997, 2001; Zardoya et al., 1998; Brinkmann et al., 2004; Takezaki et al., 2004). Molecular sequence‐based trees sometimes show coelacanths as sister to tetrapods and sometimes as sister to lungfishes. Only the molecular sequence‐based phylogeny proposed by Arnason et al. (2004) is strongly at odds with the current consensus, as it shows monophyletic living fishes, with coelacanths, chondrichthyans, lungfishes, and actinopterans (i.e., actinopterygians minus cladistians) as forming an unresolved clade, sister to cladistians (Figure 1.4).

4.4 Other Problematic Nodes

The in‐group relationships of some other higher piscine gnathostome taxa are also the subject of controversies, yet to a lesser degree. Within elasmobranch chondrichthyans, the relationships of batomorphs (pristiophoriforms, sawfish, torpedoes, skates, and rays) have long been debated, but, during the past 20 years, most morphologists agreed that batomorphs were nested within the squalomorph sharks as sister group to pristiophoroids (Maisey, 1984; Shirai, 1996). Molecular sequence‐based phylogenies do not clearly support this relationship and suggest, rather, that batomorphs are the sister group of all living sharks (Arnason et al., 2001). Recent consideration of this question, involving collaboration between morphologists, paleontologists, and molecular phylogeneticists, has provided much stronger support for this theory, which entails living shark monophyly (Maisey et al., 2004; Figures 1.3 and 1.4). It now appears that living shark monophyly is actually supported by a number of morphological characters, but previous parsimony analyses considered too many doubtful characters, which now turn out to be homoplastic.

Cladistians (bichirs) are yet another taxon whose phylogenetic position has been much debated since the nineteenth century. Cladistians share some unique skeletal characters with actinopterans (i.e., acipenseriformes and neopterygians), such as the ganoid scale structure or the acrodin cap on teeth, and numerous soft‐tissue characters, notably in brain development (everted telencephalon) and muscles of the jaw and gill arches. Most other characters seen in cladistians are unique to them (pectoral fin and dorsal fin structures), but a few characters were once regarded as either shared only with sarcopterygians, notably actinistians, or supposedly primitive for osteichthyans in general (Jarvik, 1980). Cladistians are almost unanimously regarded as the sister group to actinopterans (7, Figure 1.3), but molecular sequence data remain ambiguous in this respect. Early sequence‐based gnathostome phylogenies generally ignored cladistians because they made trees either collapse or display very odd topologies, as in Arnason et al.'s (2004) tree (Figure 1.4), where they appear as the sister group of all other living fishes. Other molecular trees show a more conventional position for cladistians as sister to actinopterans (Noack et al., 1996). The question may not be entirely settled as long as the early phase of cladistian history, possibly in early Devonian times, remains undocumented by fossils.

Within teleosts, osteoglossomorphs (osteoglossids, arapaimids, mormyrids, and hiodontids) and elopomorphs (elopids, megalopids, albuloids, nothacanthids, anguilloids, and saccopharyngids) are currently considered as the most inclusive living teleost taxa. However, the relationships of these two taxa to other teleosts (i.e., clupeocephalans) remain debated. A classical theory is that osteoglossomorphs are the sister group of elopomorphs and all other living teleosts (Patterson, 1977; Patterson and Rosen, 1977; 12 and 13, Figure 1.3), but this has been challenged over the last decade, notably by Arratia (2004), who favors the theory that elopomorphs are the sister group of osteoglossomorphs and all other living teleosts. The latter result is a consequence of the new inclusion of a large number of extinct Cretaceous and Jurassic teleost taxa in data matrices. Other phylogenies, either molecular and morphological, suggest that either the two groups form a clade, sister to all other living teleosts (Lêet al., 1993), or the relationships of osteoglossomorphs, elopomorphs, and the ensemble of all other living teleosts are unresolved (Li and Wilson, 1996) (D, Figure 1.3). Elopomorph monophyly has long been regarded as strongly supported by at least the leptocephalous larva, supposed to be a uniquely derived condition. However, by combining sequence‐based and morphology‐based analyses, Filleul and Lavoué (2001) have challenged elopomorph monophyly. Surprisingly, they suggest that the leptocephalous larva may be a primitive condition for teleosts and that eels are the sister group of osteoglossomorphs.

5 Living Primitive Fishes and Their Fossil Relatives: Naming and Dating Taxa

Most of the reputedly primitive fish taxa referred to above have a large number of fossil relatives. However, the fossil record for some of these taxa remains desperately poor, as is notably the case for hagfishes and lampreys because they lack an extensively mineralized skeleton and can only be fossilized under particular conditions.

The principle of a molecular clock (which followed in the wake of phenetics) was based on the assumption of a constant mutation rate (now regarded as unfounded; Maisey et al., 2004) and required accurate, paleontology‐based calibrations of divergence times for taxa that have extant representatives. Bracketing divergence times thus became a raison d'être for paleontologists working on early vertebrates (see Donoghue et al., 2003 for a review of the question). The phylogenetic tree of any taxon that includes living and fossil representatives comprises a crown group and a stem group (Figure 1.5). The crown group includes the youngest common ancestor to all the living representatives of the taxon under consideration and their respective fossil relatives (A1, Figure 1.5). The stem group includes all the taxa that have diverged before the common ancestor of the crown group and after the youngest common ancestor it shares with its living sister group (A2, Figure 1.5). For example, assuming that lampreys are the sister group of gnathostomes, crown‐group gnathostomes include the youngest common ancestor to a shark and a tetrapod and all its other living and fossil descendants such as actinopterygians, piscine tetrapodomorphs, or extinct chondrichthyan taxa (A1, Figure 1.5). Stem gnathostomes include all the extinct taxa that diverged before the divergence between any crown‐group gnathostome but after the divergence between lampreys and the latter (A2, Figure 1.5). Some of these stem gnathostomes have jaws (e.g., placoderms), whereas others do not (e.g., heterostracans and osteostracans), but they all share at least one uniquely derived character with crown‐group gnathostomes (the most general one being the ability to produce bone). The total group includes the crown group and the stem group. The crown group and the total group are clades (monophyletic taxa), the stem group is a grade (paraphyletic taxon). The next question is how to name these different segments of a tree. This is currently a much debated matter among certain systematicists, within the framework of the Phylocode debate (for information about the Phylocode, see http://www.ohiou.edu/phylocode). In brief, there are three ways of considering a clade: node‐based (e.g., a crown group), stem‐based (i.e., a total group), or apomorphy‐based (i.e., all organisms that display a particular, uniquely derived character, e.g. pectoral fins or jaws; Figure 1.5). Consequently and depending on which type of clade one considers for a taxon name, the range of this taxon in time (thus its earliest occurrence) may vary considerably. Biologists who are interested in the minimum age of characters that cannot be directly documented by fossils (e.g., gene sequences or physiological functions) generally prefer to use node‐based crown groups. However, one must keep in mind that a node is a matter of data set, methodological procedure, and options in phylogenetic softwares. Therefore, it may not always refer to any conspicuous character and is often labile. Examples are the node‐based definition of tetrapods (Laurin and Anderson, 2004), which does not refer to the presence of digits and limbs, or the node‐based definition of gnathostomes, which does not refer to jaws (Figure 1.5). Apomorphy‐based clades may be of interest to ecophysiologists, as certain fossil taxa that are members of a stem group may provide indirect evidence for a particular function in the form of a unique combination of anatomical structures. An apomorphy‐based definition of gnathostomes would include the presence of jaws (Figure 1.5), which actually are ecologically important structures. In contrast, an apomorphy‐based definition of tetrapods would include limbs with digits, supposedly important for the conquest of land, but the taxon would also comprise such stem tetrapods as Acanthostega (see tree in Figure 1.12) in which limbs and digits had no role in terrestrial locomotion. The interest of stem‐based taxa is that they may provide information about a ghost range, that is a segment of the phylogeny of a taxon that is not documented by fossils but which must have existed because its sister group has earlier representatives (Figure 1.6). For example, assuming that the most general character of the total‐group gnathostomes is the presence of bone, which occurs first in the early Ordovician, the minimum age of this total group is about 475 Myr (Figure 1.6A). In contrast, the earliest record of lampreys is only 360 Myr (Figure 1.6). Now, assuming that lampreys alone are the sister group of the gnathostomes, one may infer that lampreys and/or hagfishes have a ghost range of at least 115 Myr. This ghost range may even be increased by 25 Myr if the character bone is extended to mineralized dermal skeleton in general, and if one considers that the denticles of euconodonts (conodonts; see Figure 1.15) actually are evidence for dermal skeleton and thus that euconodonts are stem gnathostomes (Figure 1.6B) (Donoghue and Sansom, 2002).

Fig. 1.5 The structure of taxa. When considering the living (above t0) and fossil (below t0) representatives of a taxon, there are various ways of considering its composition and definition, as exemplified here by some fossil and living gnathostomes. The crown‐group gnathostomes (gray) include the most recent common ancestor of the three major living gnathostome groups, namely chondrichthyans (e. g., Squalus), actinopterygians (e.g., Elops), and sarcopterygians (e.g., Salamandra), as well as all their respective extinct fossil relatives (e. g., Xenacanthus, Cheirolepis, Eusthenopteron) (A1). The stem‐group gnathostomes (light gray) include all the extinct taxa (e.g., heterostracans, osteostracans, and placoderms) that have diverged before the common ancestor to crown‐group vertebrates (A1) and after the most recent common to crown‐group gnathostomes and lampreys (A2; assuming that lampreys are the sister group of gnathostomes). Some stem‐group gnathostomes are jawless (heterostracans, osteostracans), whereas others are jawed (placoderms). The total‐group gnathostomes (dark gray) include both the stem‐ and crown‐group gnathostomes, that is, all fossil or living vertebrates that are more closely related to any member of the crown‐group than to lampreys. An apomorphy‐based taxon is defined on the basis of at least one particular derived character that supports a node of the tree. Apomorphy‐based gnathostomes defined on the basis of the presence of jaws would include both the crown‐group gnathostomes and some stem gnathostomes (placoderms). [Illustrations for terminal taxa from Janvier (1996a), by permission of Oxford University Press.]

Fig. 1.6 The length of the ghost ranges that are inferred from phylogenetic trees may vary considerably, depending on the topology of the latter. Assuming that lampreys are the sister group of the total‐group gnathostomes and that the earliest evidence for the latter are early Ordovician arandaspid remains, the minimum age for the divergence between the two taxa is about 475 Myr (A). The ghost range of the lampreys (in gray) is thus about 115 Myr. However, if euconodonts are regarded as stem gnathostomes, the minimum age for the lamprey‐gnathostome divergence would be about 500 Myr (B), and the ghost range of lampreys about 140 Myr.

5.1 Hagfishes and Lampreys

Only two fossils, Myxinikela and Myxineides, both late Carboniferous (305 Myr) in age, are referred to hagfishes. Myxinikela (Figure 1.2C and D) is perhaps the most convincing fossil hagfish as it shows traces of the typical nasal basket, prenasal sinus, and tentacles, but its body is stouter than that of living hagfishes (Bardack, 1991). Myxineides shows no clear evidence for tentacles, but the internal cast of its oral cavity clearly shows the imprint of the two V‐shaped rows of horny teeth. Its body is eel‐shaped, like in modern hagfishes (Poplin et al., 2001). Myxinikela occurs in marine sediments, but Myxineides poses a problem, as it occurs in reputedly lacustrine sediments, whereas hagfish physiology supposedly precludes freshwater habits.

Two fossil lampreys, Mayomyzon and Hardistiella, are known from the late Carboniferous (Bardack and Zangerl, 1968; Janvier and Lund, 1983). Mayomyzon (Figure 1.2G and H) is from the same locality as the hagfish Myxinikela, and its excellent preservation leaves little doubt about its assignment. Certain specimens, preserved in lateral view, display imprints of exactly the same cartilages as in modern lampreys, notably a piston cartilage, large tectal cartilages, and an annular cartilage (Figure 1.2F and H). Another form from the same locality, Pipiscius (Figure 1.2I and J), may also be a lamprey and possesses an oral funnel armed with horny plates that recall those of modern lampreys (Figure 1.2K; Bardack and Richardson, 1977). Hardistiella is about 20‐Myr older than Mayomyzon. It is quite similar to the latter but displays less distinct cartilage imprints. The recent discovery of the first Devonian (360 Myr) lamprey, Priscomyzon (Gess et al., 2006), also confirmed the remarkable conservatism of lamprey morphology through time. These fossil lampreys only differ from the living ones by their lack of separate dorsal fins and their shorter branchial basket.

It is difficult to assign these fossil hagfishes and lampreys a particular position relative to the members of their respective crown groups, all the more so because there are no reliable phylogenies for the living representatives of these two taxa. Nevertheless, it is likely that the Carboniferous forms are stem hagfishes and stem lampreys, respectively. The apomorphy‐based minimum age of these two taxa is thus ∼300 Myr for hagfishes and 360 Myr for lampreys, and the stem‐based minimum age for both is from 475 to about 500 Myr (see above).

5.2 Chondrichthyans

The morphological disparity of the living chondrichthyans is significantly lower than the late Paleozoic ones. The trees in Figures 1.7 and 1.8 show reconstructions of some representatives of the major Paleozoic elasmobranch‐ and chimaeriform‐related extinct chondrichthyan taxa. The overall morphology of some of them is grossly sharklike (e.g., ctenacanthiforms, eugeneodontids), but others, in particular stem chimaeriformes, display a most unusual aspect (e.g., petalodontids, iniopterygians, chondrencheliids).

Fig. 1.7 Elasmobranch phylogeny (the out‐group being euchondrocephalans). Geologic timescale in Myr, with dates referring to the limits between periods (Cam, Cambrian; Carb, Carbonifère; Cen, Cenozoic; Cret, Cretaceous; Dev, Devonian; Jur, Jurassic; Ord, Ordovician; Perm, Permian; Sil, Silurian; Tr, Triassic). Paleozoic in white, Mesozoic in light gray, Cenozoic in dark gray. Horizontal bars in the timescale indicate the distributions of the taxa through time. Interrelationships of batomorphs not detailed. Carcharhinoids include scylliorhinids. Symmoriids and stethacanthids are regarded as single clade, but their relationships remain debated, and it is not inconceivable that they are in fact basal euchondrocephalans (Janvier, 1996a). The clade that includes xenacanthiforms and all the more crownward taxa is better supported, and is therefore regarded as the most reliable total‐group elasmobranchs. 1, elasmobranchs; 2, euselachians; and 3, neoselachians. [Illustrations for terminal taxa from Janvier (1996a), by permission of Oxford University Press.]

Fig. 1.8 Euchondrocephalan phylogeny (the out‐group being elasmobranchs). For abbreviations to the geologic timescale, see Figure 1.7. 1, euchondrocephalans; 2, holocephalans; and 3, chimaeriforms. [Illustrations for terminal taxa from Janvier (1996a), by permission of Oxford University Press.]

All living elasmobranchs (sharks and rays) are neoselachians, whose earliest undisputable representatives are only early Jurassic (199 Myr) in age (3, Figure 1.7). However, on the basis of their structure, some isolated scales and teeth suggest, albeit with great reservations, a Carboniferous and even Devonian neoselachian record. By the early Jurassic, several major living neoselachian higher taxa were already represented, notably batomorphs, hexanchoids, heterodontoids, and orectoloboids, and in the late Jurassic appear the earliest carcharinoids, lamnoids, and squatinoids (Maisey et al., 2004). Thus, the minimum age for crown‐group neoselachians is about 200 Myr, but a number of Triassic and even Permian taxa are regarded as stem neoselachians on the basis of various anatomical characters. Major extinct elasmobranch clades are the Carboniferous to Cretaceous hybodontiforms, the sister group of neoselachians, and the essentially Paleozoic ctenacanthiforms and xenacanthiforms (Figure 1.7). The earliest evidence for elasmobranch based on articulated material is from the early Devonian (400–410 Myr) (Figures 1.7 and 1.15).

Living holocephalans, or chimaeriforms, are the relicts of a very diverse ensemble of chondrichthyans, referred to as euchondrocephalans (Grogan and Lund, 2004) (1, Figure 1.8). Living chimaeriform overall morphology is certainly much derived, relative to the general chondrichthyan morphology, but it appeared early in time, probably in the late Paleozoic (e.g., Echinochimaera, Figure 1.8). The living chimaeriforms only consists of six genera distributed in three families: the Callorhinchidae, Rhinochimaeridae, and Chimaeridae, the former being sister group to the latter two (Didier, 2004) (3, Figure 1.8). The phylogeny of the crown‐group chimaeriforms is thus quite clear, but informative fossils referred to these three families are relatively rare, hence the difficulty to infer their minimum age. Nevertheless the earliest representatives of the crown‐group chimaeriforms are late Jurassic callorhinchids, and egg capsules referred to rhinochimaerids are known since the early Triassic (245 Myr) (Stahl, 1999). Stem chimaeriforms are known since the late Devonian (370 Myr) and include notably the bradyodonts (e.g., chondrenchelyids), a paraphyletic ensemble of Paleozoic taxa which display chimaeriform‐like features such as holostyly and tubular tooth structure.

Apart from the in‐group phylogeny of elasmobranchs and chimaeriforms, chondrichthyan phylogeny remains tenuously supported, and the phylogenetic position of a number of fossil higher taxa, such as cladoselachids, edestids, petalodontids, symmoriids (including stethacanthids), inopterygians, or Pucapampella, remain extremely labile (Figures 1.7, 1.8, and 1.15).

5.3 Actinopterygians

The fossil record of cladistians is very poor and the earliest evidence for this taxon are isolated scales, dorsal fin pinnules, and vertebrae from the early Cretaceous of Africa. Serenoichthys, from the late Cretaceous (98 Myr) freshwater deposits of Morocco (Dutheil, 1999), is the only fossil cladistian known from articulated specimens (Figure 1.9). It is remarkably similar to modern cladistians, except for its stouter body shape, but certainly represents a stem cladistian as it is not the sister group of one or the other living cladistian genera, Polypterus and Erpetoichthys. Considering that (according to current morphology‐based actinopterygian phylogenies) cladistians are supposed to have diverged from actinopterans before such late Devonian taxa as Moythomasia (Coates, 1999; Cloutier and Arratia, 2004) and, at any rate, before the earliest crown‐group actinopterans (2 and 4, Figure 1.9), the rather late occurrence of the earliest cladistians remains a riddle. It is possible that earlier (Paleozoic) stem cladistians are in fact known, but we do not recognize them because they lack conspicuous cladistian characters. Lund (2000) suggested that the Carboniferous guildayichthyiform actinopterygians are the fossil sister group of cladistians, but this rests on a very small number of characters, which are frequently homoplastic among early actinopterygians (Cloutier and Arratia, 2004).

Fig. 1.9 Actinopterygian phylogeny. For abbreviations to the geologic timescale, see Figure 1.7. The assignment of Dialipina to the actinopterygians still rests on tenuous characters, notably scale histology. In this tree, we regard as undoubted actinopterygians Cheirolepis and all more crownward taxa. 1, actinopterygians; 2, crown‐group actinopterygians; 3, cladistians (Polypteriformes); 4, actinopterans; 5, crown‐group actinopterans; 6, chondrosteans; 7, Acipenseriformes; 8, neopterygians; 9, crown‐group neopterygians; 10, Semionotiformes; 11, halecomorphs; 12, Amiiformes; and 13, teleosts. [Illustrations for terminal taxa from Janvier (1996a), by permission of Oxford University Press, except for the reconstruction of Dialipina, which is based on photographs in Schultze and Cumbaa (2001).]

The earliest crown‐group acipenseriforms are late Jurassic polyodontids and late Cretaceous acipenserids (Figure 1.9). The minimum age for the crown‐group acipenseriforms is thus about 150 Myr (Bemis et al., 1997). However, stem acipenseriformes, such as Chondrosteus (Figure 1.9), are known since the early Jurassic, and certain extinct taxa, such as the early Triassic (250 Myr) Birgeria (Figure 1.9) and saurichthyids, are possibly more closely related to the acipenseriforms than to any other actinopterygian group (6 and 7, Figure 1.9).

Living gars (ginglymods) fall into two genera, Lepisosteus (four species) and Atractosteus (three species), both of which are known since the late Cretaceous (65–100 Myr) (Figure 1.9). The stem lepisosteid Obaichthys is early Cretaceous in age, and the minimum age of the crown‐group ginglymods is thus about 112 Myr. The earliest known gars display much the same morphology as the extant ones. Stem gars are represented by the early Triassic to late Cretaceous (250–65 Myr) semionotids, a probably paraphyletic taxon that includes large, ubiquitous fishes with thick ganoid scales (10, Figure 1.9).

The bowfin (A. calva) is represented by a single living species, which is thus the crown group, and is recorded since the early Pleistocene (about 1 Myr). However, the more and more inclusive taxa Amiinae, Amiidae, Amiiformes, and Halecomorphi comprise a large number of fossil marine and freshwater species. The earliest Amiinae are known since the early Cretaceous and were already exclusively freshwater, but the earliest halecomorphs, the middle Triassic (230 Myr) paraseminotids, were exclusively marine (Grande and Bemis, 1999) (11 and 12, Figure 1.9).

Osteoglossomorphs and elopomorphs are known since the late Jurassic (elopomorphs having a slightly older first occurrence than osteoglossomorphs) and include a large number of fossil species. The debate about their relationships to other teleosts arose with the interpretation of certain fossil representatives of those respective taxa, as well as that of some stem teleosts (Arratia, 2004). The minimum age for crown‐group teleosts is thus indicated by the earliest elopomorphs, that is, about 150 Myr. However, there is a large number of stem teleost taxa the earliest of which, such as Pholidophorus (Figure 1.9), are late Triassic (215 Myr) in age.

5.4 Sarcopterygians

Apart from the two living species of Latimeria, all other actinistians are fossils and thus stem actinistians (Figure 1.10). The earliest known actinistians were long believed to date from the middle Devonian only, but their position within sarcopterygian phylogeny entailed a rather long ghost range because their living sister group, the rhipidistians (i.e., dipnomorphs and tetrapodomorphs), are known since the base of the Devonian (Figure 1.15). However, there is now some evidence for actinistian remains in the early Devonian (400 Myr) (Johanson et al., 2006). Although the group superficially shows relatively little morphological disparity (in particular since the beginning of the Mesozoic), some Paleozoic coelacanths display strongly divergent morphologies such as the eel‐shaped middle Devonian Holopterygius or the deep‐bodied Carboniferous Allenypterus (Friedman and Coates, 2005) (Figure 1.10). There is no fossil actinistian record between the late Cretaceous (70 Myr) and present.

Fig. 1.10 Actinistian (coelacanth) phylogeny. For abbreviations to the geologic timescale, see Figure 1.7. 1, actinistians; 2, coelacanthiformes. [Illustrations for terminal taxa from Janvier (1996a) by permission of Oxford University Press, except for that of Holopterygius, redrawn and modified after Friedman and Coates (2005).]

Lungfishes are among the earliest known osteichthyans (2, Figure 1.11), as they are recorded since the beginning of the Devonian (415 Myr) and show a spectacular radiation during the Devonian, with a majority of marine forms, often associated with coral reef environments. Lungfish diversity declines progressively during the Carboniferous, when most species occur in reputedly freshwater environments. After the Paleozoic, lungfishes undergo a considerable reduction of their dermal skeleton and are essentially known by poorly informative isolated tooth plates. Therefore, the relationships of the Mesozoic taxa to the living Neoceratodus and lepidosirenids remain unclear, except for some species. It is generally assumed that the earliest tooth plates of Neoceratodus type occur in the early Cretaceous, and tooth plates of lepidosirenid type are known in the late Cretaceous. The minimum age for crown‐group lungfishes is thus about 140 Myr. The earliest lungfish estivation burrows are from the late Permian.

Fig. 1.11 Dipnomorph phylogeny. For abbreviations to the geologic timescale, see Figure 1.7.1, Dipnomorphs; 2, dipnoiforms; 3, dipnoans; and 4, crown‐group dipnoans. [Illustrations for terminal taxa from Janvier (1996a, by permission of Oxford University Press, 2004b).]

Although this book is about fishes, it must be kept in mind that tetrapods are only part of a large total group called tetrapodomorphs and which includes a number of piscine taxa (Figure 1.12). The paleobiology of the piscine tetrapodomorphs has little bearing on the biology of the living tetrapods, but these extinct taxa will be dealt with briefly in the next section.

Fig. 1.12 Tetrapodomorph phylogeny. For geologic timescale, see Figure 1.7. 1, Tetrapodomorphs; and 2, apomorphy‐based tetrapods. [Illustrations for terminal taxa from Janvier (1996a, by permission of Oxford University Press, and 2004b), except for Tiktaalik, redrawn and modified from Daeschleret al. (2006).]

6 Extinct Major Fish Taxa and Their Position in Vertebrate Phylogeny

Besides these living primitive taxa and their extinct closest relatives, vertebrates include a number of major extinct clades, which are shown in the trees in (Figures 1.13 and 1.15). Although these will not be discussed in detail here, they deserve some comment because they illustrate the most interesting property of fossils, that is they provide character combinations that no longer exist in present‐day nature and thus often help in resolving conflicting relationships between the major living taxa.

6.1 Yunnanozoans and Myllokunmingiids

Strangely, there are very few stem vertebrates (i.e., fossil vertebrates that have diverged earlier than the most recent common ancestor to all living vertebrates, and after the divergence between vertebrates and either cephalochordates or tunicates). There is no fossil cephalochordate, despite certain claims (Blieck, 1992; Janvier, 1997), but there are undoubted tunicates in the lower Cambrian Chengjiang marine Lagerstätte of Yunnan, China (535 Myr) (Shu et al., 2001). Two taxa from the same fossil locality, Yunnanozoa and Myllokunmingiida, have been regarded as possible stem vertebrates that filled the morphological gap between cephalochordates and vertebrates. Yunnanozoans possess a distinct branchial apparatus with six pairs of filamentous gills, followed posteriorly by a series of vertical body muscle blocks. Despite Mallatt and Chen's (2003) attempts to interpret other structures of the yunnanozoan's head with regard to larval lamprey anatomy, there is no clear evidence that this taxon is more closely related to the vertebrates than to any other chordate or deuterostome taxon. Yunnanozoans have been assigned to a wide range of phylogenetic positions, notably stem vertebrates, stem cephalochordates, stem hemichordates, and finally stem deuterostostomes (see review in Janvier, 2003). It is possible that Pikaia, from the late Cambrian Burgess Shale Lagertstätte, long popularized as a the earliest chordate (Conway Morris, 1998), is in fact a close relative of yunnanozoans (Janvier, 2003). In contrast, myllokunmingiids (Figure 1.13) are more likely to be stem vertebrates. Although preserved as imprints, they display most of the vertebrate characters such as olfactory organs, eyes,and possibly otic capsules, which are acceptable evidence for neurogenic placodes, as well as six gill arches, which may be evidence for migrating neural crest cells. They also display W‐shaped myomeres that resemble more those of vertebrates than those of cephalochordates (Shu et al., 2003). Although myllokunmingiids were initially regarded as a paraphyletic group (Shu et al., 1999), including stem lampreys (Myllokunmingia) and the sister group of lampreys and all other vertebrates except for hagfishes (Haikouichthys), they are now regarded as probably a monophyletic group; they lack endoskeletal fin radials, which are a character of crown‐group vertebrates (Janvier, 2003, in press; Zhang and Hou, 2004).

Fig. 1.13 Vertebrate phylogeny. For abbreviations to the geologic timescale, see Figure 7. 1, vertebrates/craniates; 2, crown‐group vertebrates/craniates; and 3, pteraspidomorphs. The name ostracoderms, a grade of basal, jawless gnathostome‐related vertebrates, generally refers to anaspids, pteraspidomorphs, thelodonts, galeaspids, pituriaspids, and osteostracans. The monophyly of the thelodonts remains poorly supported. [Illustrations for terminal taxa from Janvier (1996a by permission of Oxford University press, and in press).]

6.2 Ostracoderms

The name Ostracodermi was coined by Cope (1889) for lower Paleozoic vertebrates (essentially heterostracans and osteostracans) that lack jaws but possess a heavily ossified dermal skeleton, like gnathostomes. These armored jawless vertebrates are regarded as stem gnathostomes (Figures 1.5 and 1.13), but the name ostracoderm is still used informally by some authors. Eight major taxa are commonly referred to as ostracoderms. In addition to heterostracans and osteostracans, ostracoderms include anaspids, galeaspids, arandaspids, eriptychiids, pituriaspids, and thelodonts. Even some partly or entirely soft‐bodied jawless vertebrate taxa (such as euconodonts and euphaneropids; Figure 1.13) are now sometimes referred to as ostracoderms as they are regarded as stem gnathostomes. At the beginning of the twentieth century, ostracoderms were thought to be ancestral to hagfishes and lampreys essentially because they were jawless and ancient. Stensiö (1927) was the first to unravel the details of the internal anatomy of osteostracans and pointed out some unique characters they (and also anaspids) shared with lampreys, notably the presence of a median, dorsal nasohypophysial opening. Stensiö also proposed that hagfishes were derived from yet another ostracoderm group, heterostracans, although this idea is now discarded. Stensiö's conception of vertebrate phylogeny was that agnathans are monophyletic and that ostracoderms are ancestral to the modern cyclostomes as a whole. This theory was also coherent with the old idea that hagfishes and lampreys were degenerate and had lost numerous characters such as the paired fins and mineralized skeleton. Some authors, however, suggested that certain ostracoderms, notably heterostracans, could be ancestral to the gnathostomes because they possessed paired olfactory organs (Halstead, 1973). During the last two decades, evidence has accumulated to indicate that ostracoderms, though undoubtedly lacking jaws, are in fact more closely related to the jawed vertebrates than to either hagfishes or lampreys (Gagnier, 1991; Forey and Janvier, 1993; Forey, 1995; Janvier, 1996b; Donoghue et al., 2000; Donoghue and Smith, 2001). As a consequence, the classical assumption that hagfishes and lampreys had lost many characters became pointless, as the presence of bone, along with other characters uniquely shared by ostracoderms and gnathostomes, came to be considered as synapomorphies. Although some characters, such as the strikingly similar condition of the nasohypophysial complex in lampreys and osteostracans (and probably anaspids), have to be regarded as homoplastic, the current tree of fossil and living vertebrates is, if not robust, at any rate clearly more parsimonious than any tree assuming agnathan monophyly (Donoghue et al., 2000). It shows ostracoderms as a grade of stem gnathostomes that fills the morphological gap between the living cyclostomes (or the living lampreys, if cyclostomes are not a clade) and the living gnathostomes (Figures 1.5 and 1.13).

The most interesting implication of the current vertebrate tree is that what is often referred to as the so‐called gnathostome body plan is not the result of a burst of anatomical innovations, as frequently alleged, but a progressive accumulation of new characters that can be inferred at several nodes of the tree (Mazan et al., 2000; Janvier, 2001; Donoghue and Sansom, 2002). For example, this tree tells us that the dermal skeleton became ossified before the endoskeleton or that paired fins appeared before jaws. Osteostracans, which now appear as the closest jawless fossil relatives of the jawed vertebrates, show that such characters as cellular bone, perichondral bone, sclerotic ring, pectoral fins, shoulder girdle, and epicercal tail have also preceded jaws in vertebrate evolution (Figure 1.13). Of course, we know nothing of the evolution of physiology along this long segment of vertebrate phylogeny that extends between the divergence of lampreys (or the cyclostomes as whole) and crown‐group gnathostomes, and that is only documented by fossil taxa (apart from functions linked to calcified tissue development), but one might expect that some anatomical or histological data