The earliest undisputed vertebrate remains are represented by isolated dermal bone fragments from the base of the Horn Valley Siltstone, and probably also from the top of the Pacoota Sandstone, of the Amadeus Basin in central Australia [46]. They are dated as Early Ordovician (Arenig) and referred to a group of armoured jawless vertebrates, the Arandaspida, known from articulated specimens in younger (Caradoc) formations of Australia and Bolivia [15] and [33]. All other earlier fossils referred to the vertebrates remain more or less heatedly debated [3], [25] and [43], with the exception – to date – of the recently discovered Myllokunmingia and Haikouichthys, from the Lower Cambrian of Chengjiang, China, which have not yet been met with reservations to date [21], [36] and [38]. The present article is a brief review of the most likely Cambrian vertebrates – at any rate, those which have not been demonstrated to be anything else –, with special reference to the anatomy of Myllokunmingia and Haikouichthys.

The debates about the nature of many of the Early Palaeozoic isolated remains occasionally referred to vertebrates illustrate the difficulty of recognizing vertebrate characters in fragmentary fossil material. Currently, there is no better diagnose of the vertebrates than that based on two developmental characters; namely, the presence of neural crests and epidermal placodes, which are unique among metazoans [2], [29] and [45], although putative, non-migrating, neural crest cell precursors may be present in cephalochordates [19] and [20]. Neural crest cells are the source for many unique vertebrate characters, such as odontogenic tissues of the dermal skeleton and the branchial skeleton, and play a major role in the development of the skull. Epidermal placodes are involved in the formation of the major vertebrate sensory organs, such as the olfactory, optic and otic capsules, and the lateral line system. Although these embryonic structures cannot be directly observed in fossils, their adult derivatives can be regarded as good evidence for their presence. Even a small dermal bone fragment from the Ordovician, may display dentinous tubercles, which are evidence for neural crest cells, and lateral line grooves, which are evidence for placodes. This inference is nevertheless difficult to make in certain fossils, such as euconodonts or fossil lampreys and hagfish, when there are doubts as to their hard tissue homology, or when the neural crest and placodal derivative are only preserved as carbonaceous imprints.

Since the discovery of articulated specimens of the euconodonts Clydagnathus in the Carboniferous of Scotland and Promissum from the Ordovician of South Africa [5] and [14], the affinities of this long-enigmatic taxon has been regarded as settled by many authors. The large, paired eyes, chevron-shaped myomeres, and caudal fin radials, suggested that euconodonts are either crown-group vertebrates (i.e. any descendant of the youngest common ancestor of all extant vertebrates; Fig. 1 ) [1], [11], [12], [23] and [43], sister-group to the crown-group vertebrates, or more closely related to cephalochordates than to any other taxon [27] and [28]. Yet attempts at homologizing the hard tissues of the euconodont denticles with typical vertebrate hard tissues (enamel, dentine, bone) [10], [11] and [12] have failed to convince the entire community of vertebrate palaeontologists [27] and [35]. Nevertheless, the few other characters that can be observed as soft tissue imprints in the head and body of the two, best preserved articulated euconodonts rather accord with vertebrate anatomy [11] and [24]. The position of euconodonts as the sister-group to ‘ostracoderms’ and gnathostomes (i.e. gnathostomes sensu lato, or ‘total group gnathostomes’; Fig. 1) in most current phylogenies is largely imposed by assumptions about the presence of certain characters such as extinsic eye muscles or gills [11]. Since euconodonts are known from the Late Cambrian to the Triassic, they are thus regarded – yet not unanimously so – as the most likely representatives of the Cambrian vertebrates [43].

Anatolepis is represented by minute, three-layered phosphatic fragments from the Late Cambrian and Early Ordovician of North America and Greenland, which were first tentatively referred to the vertebrates [4]. This assigment has been debated, but recent histological studies of this taxon apparently support its vertebrate affinity to some extent. In particular, the tubercles of its ornamentation are made up by a canalicula-bearing hard tissue that agrees somewhat with the structure of vertebrate dentinous tissues [42], [43] and [44]. In contrast, none of the Anatolepis fragments recorded to date shows any trace of a sensory-line system, nor do they show any morphological feature which would agree with the dermal plates of ‘ostracoderms’ and gnathostomes in general. Tube-shaped Anatolepis fragments have also been compared to gill denticles of certain living sharks [43]. The question of the affinity of Anatolepis may only be settled by the discovery of articulated specimens.

During the past ten years, a number of phylogenetic analyses of the major living and fossil vertebrate groups have been carried out, and all of them resulted in a tree where the living hagfish and lampreys show a more basal divergence than the ‘ostracoderms’ (a grade of fossil, armoured jawless vertebrates, currently included in the gnathostomes sensu lato; Fig. 1) [11], [12], [16], [23], [24], [25], [36], [38], [41] and [43]. Contrary to recent molecular sequence-based trees, which show a strong support for the cyclostome clade (hagfish + lampreys) [9], [18] and [31], morphology-based trees always support a closer relationship of lampreys to the gnathostomes, than to hagfish (paraphyletic cyclostomes) [11], [23], [24] and [41]. Since there is no evidence that hagfish and lampreys have ever possessed an extensively calcified skeleton, the implication of cyclostome paraphyly is that the common ancestor to all vertebrates had no calcified skeleton, which thus appeared after the divergence of lampreys [12]. This suggests that the pre-Ordovician vertebrate fossil record of the vertebrates would essentially occur under exceptional conditions of fossilisation, which allow soft-tissue preservation, such as in the Burgess Shale or the Chengjiang area (see below). However, this, alone, cannot explain the scarcity of the typical, undisputable calcified vertebrate remains before the mid-Ordovician and their absence from the Cambrian. There are probably other, biogeographical or environmental reasons for this lack of record [41].

Presumed euchordates (i.e. cephalochordates + vertebrates; Fig. 1), or even vertebrates, all preserved as imprints, have been recorded from the Middle Cambrian Konservat-Lagerstätte of the Burgess Shale in Canada. The most often-cited of them is Pikaia gracilens, which has been widely popularized as either a cephalochordate, or some kind of vertebrate ‘precursor’ [8] and [17]. However, Pikaia shows no unambiguous character of either vertebrates or even euchordates. Its overall body shape recalls that of the cephalochordates, but its myomeres are not clearly chevron-shaped and may have looked more like those of the yunnanozoans [30]; its paired tentacles vaguely recall those of hagfish [25], but hagfish tentacles show no basal swelling, contrary to those of Pikaia. In addition, there is no evidence for any gill-like structure, and the presumed, oval-shaped pharyngeal cavity seems to be dorsal in position. Another, unnamed Burgess Shale fossil has been tentatively referred to the vertebrates and compared to a larval lamprey [22], [39] and [43], but the examination of this single specimen also provides no clear evidence for its euchordate affinity, except perhaps for its chevron-shaped myomeres, which, however, seem to point backwards.

The Lower Cambrian Konservat-Lagerstätte of Chengjiang, China, yields a number of well-preserved fossils which have been regarded as either crown-group vertebrates (Haikouichthys, Myllokunmingia) [21], [26], [36] and [38], or possible stem-vertebrates, stem-euchordates, and even stem-chordates (Yunnanozoon, Haikouella) [6], [7], [13], [19], [30] and [40]. The yunnanozoans Yunnanozoon and Haikouella are very similar in aspect and have recently been regarded as ‘naked’, close relatives of vetulicolans, and stem deuterostomes [37]. Both Yunnanozoon and Haikouella possess a large, laterally compressed body, an elongate branchial region with six gill arches, and possibly sensory capsules. However, their myomeres are not chevron-shaped, and the structures referred to as gill arches and ‘gill rays’ display an unusual, segmented structure [6] and [30]. Morever, the ‘gill rays’ are lateral to the gill arches, and possibly external [30] and [37]; that is, surprizingly gnathostome-like. The recent, detailed re-examination of Haikouella nevertheless provides convincing arguments in favour of its position as a basal vertebrate [30].

Haikouichthys ercaicunensis and Myllokunmingia fengjiaoa were initially described on the basis of two specimens which only differ by a few, ambiguous characters [38]. These two, monospecific genera have recently been synonymized as Myllokunmingia [21], but this is not unanimously accepted [36]. Both species display an elongate body, which shows imprints of chevron-shaped myomeres, and a head, with a relatively short branchial region composed of six filamentous gills (Fig. 2D–F ). The recent discovery of about 500 specimens referred to Haikouichthys provided additional evidence for vertebrate characters [36]. Several of these specimens are dorsoventrally compressed and show the head in dorsal view (Fig. 2A–C). It distinctly shows paired eyes and a dorsomedially placed, probably paired, olfactory organ. The presence of otic capsules is still unclear, but probable. The anterior margin of the head displays paired lobes which recall the endoskeleton of the upper lip of larval lampreys. The surroundings of the sensory capsules form a smooth surface, which suggests the presence of some kind of either cartilaginous or fibrous braincase, like that of hagfish, and show a conspicuous constriction behind the optic capsules. Behind the sensory capsules, and dorsally to the gill impressions, there is a series of irregularly-shaped, often sinuous stains which have been interpreted as possible arcualia of the axial skeleton [36] (Fig. 2C, E). These stains seem to be connected by a faint strand-shaped stain, which may be the trace of the notochord. Although cartilaginous arcualia (basidorsals and interdorsals) are present in lampreys, these stains in Haikouichthys are surprisingly large, widely spaced, and send off processes far ventrally, thereby recalling the cartilaginous branchial arches of lampreys. Moreover, these stains do not clearly extend into the posterior, myomere-bearing part of the body. Another, peculiar feature has been described in a specimen referred to Myllokunmingia, and consists of loop-shaped imprints extending from the branchial region into the body [21], which may be imprints of the segmental blood vessels within the myocommata. In both taxa, a large stain behind the gill series may be regarded as the trace of either the heart or a pericardic cartilage (Fig. 2E). As a whole, Haikouichthys (and probably Myllokunmingia) displays acceptable evidence for placodes in the form of optic (yet a lens is still unknown), olfactory, and possibly otic capsules, but less obvious evidence for neural crest tissues, except perhaps for a cartilaginous branchial skeleton. Three characters, however, may be at odds with what we know of crown-group vertebrates. The presence of serially arranged gonads has been suggested in both Haikouichthys and Myllokunmingia (Fig. 2E) [21] and [36], and in the yunnanozoans [6] and [30]. Although this might also be an artefact due to the compression of the digestive tract by the overlying myomeres, this retention of this cephalochordate-like condition, if confirmed by further discovery, would exclude these Chengjiang taxa from the crown-group vertebrates (assuming that serially arranged gonads are a general euchordate character, lost only once in the crown-group vertebrates). Ventrally to the gill series, there is a series of possibly calcified, oblique bars, the orientation of which recalls that of the ceratobranchials of the gnathostomes (Fig. 2E) [36] and [38]. These may be part of the branchial skeleton (but certainly not ceratobranchials) and have no clear homologue in the branchial skeleton of extant jawless vertebrates. Finally, the structure interpreted as a dorsal fin is at odds with that of the dorsal fin in all other known vertebrates, by its numerous, rostrally tilted radials (Fig. 2D) [36] and [38], that are seen in only a few specimens, but recently regarded as a possible fabric of the sediment [21]. The presence of other fins remains questionable. There is no clear evidence for paired fins, but a median ventral, preanal skin fold may be present (Fig. 2D), like in hagfish and larval lampreys. The tail is currently unknown.

As a whole, and provided that we are not misled by an extraordinary case of convergence, the overall aspect of Haikouichthys (and probably Myllokunmingia) may have been somewhat similar in aspect to that of a larval lamprey (Fig. 2F), yet many characters remain too uncertain to decide whether it is a stem vertebrate, or a crown-group vertebrate.

To date, undisputable crown-group vertebrates remain unknown before the Early Ordovician. However, some Cambrian taxa are more and more likely to be either actual crown-group vertebrates, or stem-vertebrates. During the past decade, new articulated euconodont material has provided acceptable evidence for the vertebrate affinities of this long-enigmatic group, although the homology of euconodont hard tissues and the conventional vertebrate hard tissues remains debated, as it is also the case for the hard tissues of the Late Cambrian–Early Ordovician skeleton fragments referred to Anatolepis. Currently, the best documented and most informative, putative Cambrian vertebrates are Haikouichthys and Myllokunmingia, from the Early Cambrian of Chengjiang, China, whose overall anatomy is quite suggestive of that of a larval lamprey, yet certain of their characters remain at odds with the structure of crown-group vertebrates. Although they show evidence for placode- and neural crest-derived organs (olfactory organ, eyes, ?otic capsules, branchial arches) which suggest vertebrate affinities, it is still uncertain whether these two taxa are stem-vertebrates or crown-group vertebrates. If their presumed, serially arranged gonads are confirmed, then would rather be stem-vertebrates and thus fill the morphological gap between the cephalochordates and vertebrates (Fig. 1). At any rate, these Early Cambrian forms strongly suggest that the divergence between cephalochordates and vertebrates occurred before the Cambrian; that is, in better agreement with recent molecular clock data (notwithstanding the numerous biases in molecular clock inferences [34] and [41]), which indicate a date of about 750 Myr for this divergence [18], and about 550 Myr for the cyclostome-gnathostome divergence [18] and [32].