In On the origin of species by natural selection, Charles Darwin (1859) famously fretted over the complexity of the oldest known animal fossils. To account for the stratigraphic pattern he observed, Darwin postulated a long prior history of metazoan evolution, during which the morphological complexity and diversity displayed by trilobites and other Cambrian animals slowly accumulated. The absence of confirming fossils was ascribed to massive failure of the pre-Cambrian stratigraphic record.

Today, we recognize a relative abundance of Proterozoic sedimentary rocks, distributed globally. Palaeontologists have also documented Proterozoic animal fossils, but only from the last 30–40Myr of the Eon (Xiao & Knoll 2000; Narbonne 2005). On the other hand, insights from cell biology, molecular phylogeny and developmental genetics have not only positioned animals within the greater diversity of eukaryotic organisms, but shown that the defining traits of animal life are built on a foundation of antecedent molecular features (e.g. Maynard Smith & Szathmáry 1995; King 2004).

The view that animal evolution was made possible, at least in part, by prior innovations in eukaryotic genetics and cell biology requires that eukaryotic life diversified before the advent of metazoans. Opinions on the antiquity of the Eucarya range widely (e.g. Brocks et al. 1999 versus Cavalier-Smith 2002a), but molecular clock estimates commonly suggest an earlier Proterozoic origin and later Proterozoic diversification of the clade (Doolittle et al. 1996; Douzery et al. 2004; Yoon et al. 2004; in contrast, Hedges et al. 2004 proposed Palaeoproterozoic origin and early crown group divergences). Such conjectures obviously make predictions about what palaeontologists should see in Proterozoic sedimentary rocks. In this paper, we review the early fossil record of eukaryotic organisms and use it to explore the cellular and functional evolution of protists in Proterozoic oceans.

How do we recognize ancient fossils as eukaryotic? A concrete example serves to address this question. Shuiyousphaeridium macroreticulatum is known from a large population of microfossils preserved in coastal marine shales of the Ruyang Group, northern China (Yan & Zhu 1992; Xiao et al. 1997). The fossils are spheroidal organic vesicles characterized by a reticulated surface and numerous regularly spaced cylindrical processes that flare outward (ca 80 processes are visible around the periphery of a single specimen; figure 1a,b). Vesicles range in diameter from 50 to 300μm (mean, 148μm; s.d., 38μm); processes are hollow, 10–15μm long and 2–3μm wide. Scanning electron microscopy (SEM) examination shows that the vesicle's outer surface is covered with ridges that delimit polygonal fields ca 2μm across (figure 1c,f). Inner wall surfaces show the same ornamentation, but in reverse relief—fields are visible as closely packed, beveled hexagonal plates (Javaux et al. 2004; figure 1e). Transmission electron microscopy (TEM) images further show that the ca 1.5μm wall is multilayered, with the electron-dense, homogeneous layer of organic plates lying above a thin electron-tenuous layer (Javaux et al. 2004; figure 1d).

Figure 1 Shuiyousphaeridium macroreticulatum from the Mesoproterozoic Ruyang Group, China. (a) Light microphotograph showing specimen with numerous regularly spaced cylindrical processes that flare outward; (b, c, e–f) SEM images showing (b) whole specimen, (more ...)

Prokaryotes can be large, they can have processes, and they can have preservable walls. But we do not know any prokaryote that combines these three characters, nor any that exhibit the complexity of form that light microscopy, SEM and TEM document in Shuiyousphaeridium. Many eukaryotes do exhibit these features in combination; therefore, we believe the most straightforward interpretation of these fossils is that they were made by a eukaryotic organism. According to Cavalier-Smith (2002a, p. 37), ‘cysts with spines or reticulate surface sculpturing would probably have required both an endomembrane system and a cytoskeleton, the most fundamental features of the eukaryotic cell, for their construction’. Shuiyousphaeridium clearly fits this description.

Phylogenetic conjectures for Shuiyousphaeridium range from dinoflagellates (Meng et al. 2005) to possible fungi (Butterfield 2004), but the fossils provide little support for any specific attribution. The population could represent either a stem or crown group eukaryote.

Radiometric dates indicate that Ruyang deposition occurred after 1600Myr ago but before 1000Myr. Carbon isotopic stratigraphy, in turn, suggests an age greater than 1250Myr (Xiao et al. 1997). Thus, Shuiyousphaeridium is a Mesoproterozoic protist.

Other fossils corroborate the observation that eukaryotes lived in Early Mesoproterozoic oceans. Microfossils in the Ruyang Group include a second process-bearing taxon, Tappania plana (Yin 1997; figure 2a). In contrast to Shuiyousphaeridium, Tappania displays broadly spheroidal vesicles of variable size (20–160μm) that bear 0–20 closed, heteromorphic processes distributed asymmetrically on the vesicle surface (Yin 1997; Javaux et al. 2001). Processes vary in length from 25 to 60μm and uncommonly branch. A single specimen shows what may be a septum between vesicle and process (Butterfield 2005a), but, in general, processes communicate freely with the vesicle interior. Vesicles may also exhibit up to three bulbous extensions, suggestive of budding. In addition to their Chinese occurrence, Tappania populations have been found in reliably dated (U–Pb zircon age of 1492±3Myr for a subtending ash bed) shales of the Roper Group, northern Australia (Javaux et al. 2001), and in Mesoproterozoic successions from India (Prasad & Asher 2001) and Russia (Nagovitsin 2001).

Figure 2 Diversity of Late Palaeoproterozoic to Early Mesoproterozoic eukaryotic fossils. (a) Tappania plana, from the Early Mesoproterozoic Roper Group, Australia; (b) Horodyskia moniliformis, from the Mesoproterozoic Bangemall Group, Western Australia; (c,f (more ...)

Its irregular morphology and asymmetric distribution of processes suggest that Tappania was an actively growing vegetative cell or germinating structure rather than a metabolically inert spore (Javaux et al. 2001). As in the case of Shuiyousphaeridium, Tappania's combination of large size, preservable walls, complex processes and possible budding structures finds no matches among known bacteria or archaea. (Readers interested in the proposal that Tappania was an actinomycete similar in overall structure to extant Kibdelosporangium (a soil-dwelling, hyphal bacterium that synthesizes no decay-resistant wall polymers) should consult Shearer et al. (1989).) Butterfield (2005a) has proposed that Tappania plana was fungal, based on putative similarities to complex Neoproterozoic microfossils from arctic Canada. Fungal affinity is possible, but given the limited number of systematically informative characters, we prefer to view this fossil as problematic (and distinct from the Neoproterozoic population to which it has been compared). Regardless of phylogenetic interpretation, however, we agree with Butterfield (2005a) that Tappania was a eukaryote with a complex cytoskeleton, probably preserved in an actively growing phase of its life cycle, and at least plausibly heterotrophic.

Other protists in the Roper assemblage include Valeria lophostriata (Jankauskas 1989; Xiao et al. 1997; Javaux et al. 2003; figure 2d,e) and Satka favosa (Jankauskas 1989; Javaux et al. 2003; figure 2c,f), distinguished by surface ornamentation of parallel ridges uniformly spaced on the internal surface of the vesicle and tessellated polygonal plates up to 15μm across, respectively. Roper microfossils additionally include three leiosphaerids (simple organic-walled spheres) with distinctively heterogeneous wall ultrastructures, as observed under TEM (figure 2g,h). Once again, such wall structure is common among extant protists that make preservable walls of comparable size and morphology, but distinct from the most likely prokaryotic candidates, envelope-forming cyanobacteria (Javaux et al. 2004).

Macroscopic compressions, impressions and casts occur globally in rocks of comparable age. Approximately 1450Myr shales of the Helena Formation, Montana contain a variety of carbonaceous compressions up to several centimetres long (Walter et al. 1976). Among these, coiled fossils assigned to Grypania spiralis are most confidently interpreted as eukaryotic. (Most other forms could be fortuitously shaped fragments of microbial mats.) Grypania fossils are narrow ribbons, originally cylindrical, up to 13mm long and 2mm wide, that commonly form a regular coil up to 24mm across (Walter et al. 1990; figure 2i). Indian populations illustrated by Kumar (1995) preserve a distinct millimetre-scale annulation that may reflect underlying cell structure. These fossils are very likely of eukaryotic origin, although phylogenetic relationships are not well constrained. Well-preserved Grypania populations occur, as well, in Mesoproterozoic rocks from China (Walter et al. 1990), and Palaeoproterozoic (ca 1850Myr) populations have been reported from Canada (Han & Runnegar 1992). Samuelsson & Butterfield (2001) regard these latter specimens as possible composites of much smaller prokaryotic filaments, but the pronounced morphological regularity of specimens within a large population examined by one of us (A. H. Knoll) convinces that the Canadian Grypania was an organism, not a colony or composite.

The other Early Mesoproterozoic macrofossil with a pursuable claim to eukaryotic status is Horodyskia moniliformis, known as casts and moulds in sandstones from Montana and Western Australia (Horodyski 1982a; Grey & Williams 1990; Yochelson & Fedonkin 2000; figure 2b). Horodyskia consisted of 1–4mm spheroidal (less commonly ovoid, rectangular or conoidal) bodies connected by thin cylindrical strings to form uniseriate, pearl-necklace-like structures up to 10cm long. Grey & Williams (1990) drew structural analogies to articulated seaweeds, whereas Yochelson & Fedonkin (2000) favoured comparisons with animals. In fact, archaeal–bacterial consortia in modern sulphur springs form millimetre-scale strings of beads (Rudolph et al. 2001), although it remains an open question whether such features would form or preserve in the environments inferred for Horodyskia fossils. We interpret Horodyskia as a problematic macrofossil whose eukaryotic affinities are probable, but not beyond debate.

As discussed by Summons (2005) in this issue, preserved steranes independently suggest that eukaryotic organisms inhabited mid-Proterozoic oceans. If eukaryotes diverged only 700–800Myr ago (Cavalier-Smith 2002a), then all putatively eukaryotic fossils and biomarkers in older rocks must be misinterpreted, and numerous mid-Proterozoic prokaryotes must have possessed attributes that were subsequently lost and re-evolved convergently by eukaryotes. If any one of these records is correctly interpreted, eukaryotes existed in mid-Proterozoic oceans.

Other fossils in mid-Proterozoic rocks could be eukaryotic, but their affinities are less clear. Large (greater than 50μm) spheroidal microfossils are widely distributed in Mesoproterozoic shales. As noted above, a few of these are known to have walls with complex ultrastructure of types best known from eukaryotes, but most have unknown wall structure and could, in principle, be preserved envelopes of colonial cyanobacteria (Knoll 1996). Similarly, longitudinally striated carbonaceous tubes up to 150μm in diameter and more than a millimetre long found in shales of the Roper Group could be eukaryotic, but this interpretation remains tentative.

Can we extend the eukaryotic fossil record backward into the Palaeoproterozoic or beyond? The oldest acritarchs (a collective term applied by palaeontologists to closed, organic-walled microfossils of uncertain systematic affinities) with regular ornamentation are populations of the corduroy-like Valeria preserved in shales of the Mallapunyah Formation, northern Australia, closely constrained by radiometric ages to about 1650Myr (Javaux et al. 2004), and in the Changcheng Group, northern China, constrained to be more than 1683±67Myr and less than 1780±20Myr (Yan & Liu 1993; Li et al. 1995; Wan et al. 2003). Changcheng assemblages also include large (up to 237μm) unornamented acritarchs, some of which display regular medial splits, similar to those formed during excystment of younger protists (Yan & Liu 1993). The splits are both common and regularly equatorial, favouring a eukaryotic interpretation. Cyanobacterial envelopes can split in regular patterns (Waterbury & Stanier 1978), however; so the medially split leiosphaerids in Changcheng rocks are conservatively interpreted as possibly but not unambiguously eukaryotic. Comparable uncertainty attends uniseriate filaments up to about 100μm wide, also found in Changcheng shales (Yan & Liu 1993).

Carbonaceous compressions in Late Palaeoproterozoic shales from China have been interpreted as seaweeds, with specimens from the 1600–1700Myr Tuanshanzi Formation specifically attributed to the Phaeophyta (Zhu & Chen 1995). Examination of Tuanshanzi structures in outcrop by one of us (A. H. Knoll) suggests that the features in question can alternatively be interpreted as rare, fortuitously shaped fragments deposited among many irregular mat shards. Lamb et al. (2005) have drawn similar conclusions about macroscopic compressions in the older Changzhougou Formation. Steranes in 2780Myr shales from Western Australia have been interpreted as evidence that stem eukaryotes diverged early in Earth history (Brocks et al. 1999), but a general paucity of well-preserved microfossils precludes independent morphological inferences about eukaryotic biology prior to about 1800Myr.

In summary, Late Palaeoproterozoic and Early Mesoproterozoic rocks preserve evidence for a moderate diversity of preservable eukaryotic organisms. This evidence includes cell walls without surface ornament (but with complex ultrastructure) and walls with regularly distributed surface ornamentation, with asymmetrically arranged processes that appear to reflect active cell growth, and with numerous symmetrically arranged processes. Collectively, these fossils suggest that eukaryotes not only existed in mid-Proterozoic oceans, but possessed flexible membranes and cytoskeletons capable of directing cell remodelling and surface morphology (Javaux et al. 2001; see below).

Many, perhaps most, preserved Proterozoic protists cannot be assigned with confidence to any specific branch of the eukaryotic tree (to the extent that we know it). A few, however, preserve diagnostic features of life cycle and morphology that support reasonable systematic interpretation. (The formation of preservable parts is, itself, a character of interest, as many protists synthesize no preservable walls or cysts during their life cycle.) Butterfield (2000) has marshalled a strong case for the close phylogenetic relationship of the fossil Bangiomorpha pubescens (figure 3c) to bangiophyte red algae. Bangiomorpha bears a superficial resemblance to uniseriate filamentous cyanobacteria, but as Butterfield (2000) points out, it differs in a number of key characters. Bangiomorpha filaments have cellularly differentiated holdfasts and zones of discoidal cells that expand and divide radially in several planes to produce distinctive wedge-shaped cells. Such features are unknown in cyanobacteria, but occur together along with other characters displayed by Bangiomorpha in extant bangiophyte red algae (Butterfield et al. 1990; Butterfield 2000). Moreover, taphonomic features displayed by Bangiomorpha are similar to those of other early eukaryotes, but different from those of filamentous cyanobacteria known from the Proterozoic fossil record. Bangiomorpha displays three-dimensionally competent preservation of outer and inner walls, with no cytoplasmic preservation, whereas cyanobacterial filaments with sheaths generally show competent three-dimensional preservation of the sheath, but partial or complete collapse of the cells inside (Bartley 1996). Published radiometric dates constrain Bangiomorpha only to the interval 1267±2 to 723±3Myr, but an unpublished Pb–Pb date of 1198±24Myr and physical stratigraphic relationships strongly suggest that the fossils' age lies close to the lower radiometric boundary (Butterfield 2000).

Figure 3 Late Mesoproterozoic and Neoproterozoic eukaryotic fossils: (a,b) ‘Tappania plana’ from the Neoproterozoic Wynniatt Formation, arctic Canada, a complex form with septate, anastomosing processes, shown in detail in (b); (c) Bangiomorpha (more ...)

Latest Mesoproterozoic (more than 1005±4Myr; Rainbird et al. 1998) microfossils from the Lakhanda Group, Siberia, contain several additional populations of coenocytic to multicellular filaments whose morphologies and dimensions suggest eukaryotic origin (Herman 1990; figure 3f). Principal among these are fossils assigned to Palaoevaucheria clavata and other form taxa that appear to preserve vegetative and reproductive phases of a heterokont protist comparable to the extant xanthophyte alga Vaucheria (Jankauskas 1989; Herman 1990). Vaucheria-like populations preserving several life-cycle stages also occur in the 750–800Myr Svanbergfjellet Formation, Spitsbergen (Butterfield 2004). Latest Meosoproterozoic and Early Neoproterozoic acritarchs (figure 3h) continue the record of moderate diversity established earlier, although some taxa characteristic of these younger assemblages have not, to date, been found in older rocks (Knoll 1996).

By 750–800Myr, the most diverse fossil assemblages contain an increased diversity of acritarch and other protistan morphotypes (e.g. Butterfield et al. 1994; Butterfield & Rainbird 1998), including small branched structures interpreted as siphonocladalean green algae (Butterfield et al. 1994); vase-shaped structures interpreted as both filose and lobose testate amoebae—and, therefore, as both amoebozoan and cercozoan protists (Porter et al. 2003; figure 3i); remarkable spheroidal fossils from which numerous anastomosing cellular filaments arise, interpreted as possible fungi (Butterfield 2005a; figure 3a,b); and a moderate diversity of other colonial to multicellular eukaryotes with less certain systematic affinities (Butterfield et al. 1994; Butterfield 2005b). Collectively, carefully studied microfossil assemblages support the hypothesis that the later Mesoproterozoic and Early Neoproterozoic was a time of major clade divergence within the Eucarya, although diversity within most major clades remained relatively low (see Porter 2004 for a recent review).

In studies of Phanerozoic diversity, the problem of sampling in compilations of total diversity has been addressed in part by comparison with records of taxonomic richness for individual assemblages, broad proxies for community diversity that are not subject to many of the biases thought to influence global samples (Bambach 1977). Figure 4 (see also table 1) shows a comparison of eukaryotic diversity for a selection of well-preserved and well-studied Proterozoic to Early Cambrian fossil assemblages. Consistent with discussions in the preceding section, we have included all non-metazoan fossils whose organization, ornamentation and/or ultrastructure suggest a likely eukaryotic origin. The most problematic fossils included in our compilation are leiosphaerids, spheroidal acritarchs without distinguishing ornament. In some cases (e.g. Butterfield et al. 1994; Javaux et al. 2004), TEM observations support a eukaryotic origin for these fossils, but ultrastructural data are lacking for most Proterozoic leiosphaerids. By including only leiosphaerids with a diameter greater than 50μm, we have likely excluded smaller protistan fossils and may, as well, have included cyanobacterial envelopes. Our compilation segregates leiosphaerids from other presumable eukaryotes, enabling readers to track different fossil morphotypes through time.

Figure 4 The composition and taxonomic richness of non-metazoan eukaryotes in Proterozoic to Early Cambrian fossil assemblages. (a) Total diversity of eukaryotic morphospecies for selected Proterozoic and Early Cambrian assemblages—in each column, a thin (more ...)

Table 1. Estimates of taxonomic richness used to construct figure 4. (Note that numbers provided here reflect the present authors' diversity estimates, which do not necessarily coincide with those in original publications.) age max and min, maximum and (more ...)

From these assemblages, we infer the following:

• In Late Palaeoproterozoic to Early Mesoproterozoic rocks (ca 1600–1300Myr), eukaryotic biology is recorded by a modest diversity of macroscopic fossils and preserved cell walls, including forms with complex ultrastructure, regular ornamentation and/or cylindrical processes. These fossils may, but need not, include crown group eukaryotes.
• Late Mesoproterozoic and Early Neoproterozoic (ca 1300–720Myr) assemblages continue the record of modest acritarch diversity, although many taxa found in rocks of this age differ from those in earlier assemblages. Unornamented and ornamented sphaeromorphs, some showing evidence of binary division or budding, are common. Acritarchs with asymmetrically placed processes (in some cases reflecting actively growing cells; Knoll et al. 1991; Butterfield et al. 1994; Butterfield 2004) form a taxonomically and numerically subordinate part of assemblages and acritarchs with symmetrically distributed processes are uncommon. Read literally, the published record suggests that acritarch diversity was a bit higher in the later part of this interval (ca 800–720Myr) than in the earlier part (figure 4b). Certainly, the later record includes a greater diversity of non-acritarchous eukaryotes, especially vase-shaped protists and microscopic multicellular protists (figure 4d). Fossils from this interval include the earliest recognizable representatives of extant eukaryotic clades, including red algae, green algae, heterokonts, amoebozoans, cercozoans and possibly fungi (Porter 2004). Macroscopic compressions document a small diversity of centimetre-scale blades and closed tubes (Du & Tian 1985; Hofmann 1992; figure 4c).
  The curtain drops on these assemblages with the onset of Sturtian glaciation, and it rises again on a substantially different biota only after the Marinoan glaciation ca 635Myr. At least some earlier Neoproterozoic morphotypes survived the Sturtian ice age (Allison & Awramik 1989; redated by Kaufman et al. 1992), but only a few microfossil assemblages unambiguously document marine life between the major glaciations. Whether this reflects persistently unfavourable environments (e.g. James et al. 2005) or bad luck in sampling is, at present, unclear.
• Early Ediacaran (635 to ca 580–550Myr) assemblages display a notable increase in observable diversity, and the composition of these assemblages also shifts markedly—for the first time, populations with symmetrically distributed processes dominate acritarch biotas (Zang & Walter 1992; Moczydlowska et al. 1993; Zhang et al. 1998; Grey 2005; figure 4a,b). Macroscopic compressions in coeval or slightly younger rocks similarly record a much higher diversity of form, including a variety of seaweeds and possible animals (Steiner 1994; Xiao et al. 2002; figure 4c). Macroscopic algae include the first dichotomously branching thalli (figure 3d), as well as anisotomously branching forms. Animals are independently known from Ediacaran phosphorites, carbonates and sandstones (Grotzinger et al. 2000; Xiao & Knoll 2000; Narbonne 2005). Interestingly, Late Ediacaran (ca 550–542Myr) acritarch assemblages known to date display none of the acanthomorphic diversity so obvious in preceding deposits (figure 4a). Unornamented spheroids, commonly abundant and including forms several hundred micrometres in diameter, dominate microfossil assemblages (Germs et al. 1986; Jankauskas 1989; Burzin et al. 1997).
• Cambrian and Ordovician rocks record renewed diversification of ornamented and symmetrical process-bearing acritarchs—now, we believe, properly ascribed to cyst-forming algae (and perhaps, heterotrophic dinoflagellates). This two-step increase parallels the Cambrian and Ordovician radiations of marine animals (Knoll 1994; Vidal & Moczydlowska 1997; figure 4a,b), as well as the diversification of skeletal protists, including both foraminiferans and radiolarians (see references in Knoll 2003a).

How reliable is this record as a chronicle of evolution? This question deserves fuller treatment than can be provided here, but the short answer is, ‘Reasonably, with a few caveats’. Acritarchs comprise the most abundant and widely distributed record of Proterozoic protists, and the assemblages noted for the five intervals outlined above each occur in at least four different basins on multiple continents, preserved as both compressions in shale and silica permineralizations (and, in Ediacaran successions, phosphorite). The basic trends outlined for acritarch morphology and diversity appear to be broadly predictive and, therefore, reflective of evolutionary history. The same is true of macroscopic fossils. Early Mesoproterozoic assemblages characterized by Grypania and Horodyskia, an earlier Neoproterozoic assemblage dominated by the morphotaxa Tawuia and Longfengshania, diverse Ediacaran macroalgal assemblages, and Ediacaran to Cambrian animal assemblages occur in stratigraphically predictable fashion. Moreover, some of the macroalgae documented in Early Ediacaran shales also left a conspicuous signature in sandstones, decreasing the probability that older records would be missed. For example, the centimetre-scale, originally fluid-filled macrosphere Beltinellaformis is preserved in Ediacaran but not much older sandstones as casts and moulds that commonly cover bedding planes (Xiao et al. 2002). Similarly, concentric grooves made by basally attached, semi-rigid algae or animals are well known from Ediacaran but not older bedding surfaces (Jensen et al. 2002). We accept the large diversity increase observed in macroscopic compressions as evidence for Early Ediacaran algal diversification.

A similar case can be made for the three-dimensionally preserved vase-shaped protists described from ca 800–700Myr rocks. Taxonomic diversity in this class of fossils is best gauged from exceptionally preserved assemblages in early diagenetic dolomite concretions within Grand Canyon shales (Porter et al. 2003), but because of the rigid nature of their tests, vase-shaped protists are commonly preserved as casts in mid-Neoproterozoic carbonates and cherts. Indeed, such casts can be among the most abundant fossils in these rocks (e.g. Knoll & Calder 1983), but they have not yet been reported from older successions.

The fossils that are most problematic in this regard—frustratingly so, as they provide some of our best information on phylogeny—are the multicellular fossils found as microscopic compressions in shales and, less commonly, as permineralizations in chert (Herman 1990; Butterfield 2000, 2004, 2005a,b). The best argument we can muster here is that such microfossil complexity has been found repeatedly in Late Mesoproterozoic and Neoproterozoic biotas but not in older assemblages that appear to be comparably well preserved. Once again, the record read literally may well provide a historical chronicle of Early eukaryotic divergence, but we stress that the record as currently known provides minimum estimates of character evolution and clade divergence that may well change with continuing discovery.

As summarized in the preceding section, Proterozoic protists included simple, unornamented unicells; morphologically complex and elaborately ornamented vesicles; clusters (colonies?) of uniform cells; uniseriate filaments, both branched and unbranched; and three-dimensionally complex multicellular organisms displaying cellular differentiation. For many of these fossils, phylogenetic placement is difficult because few characters ally fossil populations with extant clades. We can sidestep these issues by focusing on the characters themselves, and not on any implied phylogenetic affinity. For example, Tappania may or may not be related to fungi, but it inarguably displays an asymmetrical arrangement of processes. Similarly, one might debate the attribution of Bangiomorpha to the red algae, but it demonstrably exhibits cell differentiation. This suggests an independent, ‘taxonomy-free’ avenue of inquiry focused on function, development, and the cell biological processes that underlie these features.

Like the vegetative expansion of Tappania processes, growth of multicellular filaments requires a means of molecular transport to specific locations—i.e. polarized transport. It is possible that materials used for differential growth moved by simple diffusion, and that localized growth in older Proterozoic cells reflects anabolic enzymes attached to the cell cortex. However, the ubiquity of the proteins used for intracellular transport argues for the antiquity of this process within the Eucarya—the phylogenetic breadth of organisms known to contain genes coding for actin, myosin (Richards & Cavalier-Smith 2005), tubulin (Baldauf et al. 2000), kinesin (Lawrence et al. 2002) and dynein (Asai & Wilkes 2004) suggests the microtubule/kinesin/dynein and f-actin/myosin systems for intracellular transport evolved early in eukaryotic evolution.

Uniseriate filaments interpreted as eukaryotes occur in later Mesoproterozoic rocks (Herman 1990; Butterfield 2000) and may be represented by the Late Palaeoproterozoic (1780±20 to 1683±67Myr) filament Qanshania (Yan & Liu 1993). There is no reason to believe that any of the filamentous protists preserved in Proterozoic rocks were directly ancestral to animals. Rather, such fossils document the early evolution of a molecular capacity for simple multicellularity exploited through time by multiple clades.

Compared to simple filaments, branched structures allow more effective exploration for and exploitation of environmental resources. For example, fungi grow more densely in richer growth media (Ritz 1995) due to increased branching that allows more effective space filling and, therefore, increased absorptive capacity. Branched forms require a more sophisticated intracellular spatial organization because cellular locations must be mapped in a way that is geometrically more complex than simply ‘opposite prior site of growth or division’. Branched filaments evolved in the Eucarya no later than ca 1000Myr (Herman 1990) and were present in several clades by 800–720Myr (Butterfield et al. 1994; Butterfield 2005a). The oldest branching structures of any type observed to date in Proterozoic protists are the bifurcating processes of nearly 1500Myr Tappania (Javaux et al. 2001).

To build three-dimensionally organized structures that are mechanically stable (as opposed to loose tangles of filaments or consortia of cells), cells must be able to adhere to each other via cell wall fusion or cell membrane adhesion. Indeed, cells must adhere to neighbours that are not the immediate products of a given division, requiring a sophisticated level of cell–cell communication that enables individual cells to accept adherence to their kin. Mesoproterozoic and Early Neoproterozoic macrofossils, such as Grypania and Longfengshania, may reflect a molecular capacity for tissue formation, but in the absence of anatomical data, this is hard to demonstrate (or refute). The first clear example of three-dimensional organization and cell wall fusion are the complex ‘Tappania’ fossils reported by Butterfield (2005a) from 820–780Myr rocks in Canada (figure 3a,b). Diverse fossils in Ediacaran rocks document tissue-grade three-dimensionality in florideophyte red (Xiao et al. 2004) and other (Xiao et al. 2002) algae, discrete apical and intercalary meristems in diverse seaweeds (Xiao et al. 1998), and, of course, the complex ontogenies and astogenies of early macroscopic animals (Narbonne 2005).

In principle, genetics, ecology or environment could govern the pattern of protistan evolution inferred from Proterozoic fossils. Genetic facilitation is most difficult to assess, not least because the genetic determinants of morphology in most single-celled eukaryotes and algal thalli remain poorly understood. Nonetheless, observations reviewed in the preceding section suggest that the basic molecular mechanisms that control morphogenesis in unicellular and simple multicellular eukaryotes were in place fairly early in the Proterozoic Eon. Thus, without questioning the role of genetic innovation in the evolution of animals and complex, tissue-grade algae, it seems likely that Early Ediacaran diversification of acritarchs and macrophytes must reflect, at least in part, influences from ecology and/or environmental change.

Because eukaryotic phytoplankton and radiolarians diversified in step with marine invertebrates during the Cambrian and Ordovician periods, ecology can be justifiably invoked as a major driver of protistan evolution in the age of animals (Knoll 1994; Butterfield 1997). Butterfield (1997) proposed grazing by newly evolved mesozooplankton as a governing influence on the proliferation of ornamented and process-bearing acritarchs, although it must be remembered that the acritarch forms most common in Cambro-Ordovician rocks are cysts (Strother 1996), many of which may have rested on the sea floor like those of modern dinoflagellates (Dale 1983). Diversity increase and a pronounced coeval increase in rates of evolutionary turnover (Knoll 1994) could reflect the further influences of metazoan grazing and excretion on water mass heterogeneity at the appropriate spatial scale, or checks by grazers on the population growth of superior algal competitors. To the extent that nutrient depletion induces cyst differentiation in planktonic algae, increased patchiness of nutrients could also enhance the selective advantage of life cycles with resting stages—imparting a spatial heterogeneity to marine phytoplankton that might, itself, facilitate diversification. Ecologists continue to debate the physical and biological controls on phytoplankton diversity in modern environments (e.g. Grover & Chrzanowski 2004), so it is not surprising that the nature of animal–protist interactions in Cambro-Ordovician seas remains incompletely understood. Argued in reverse, the absence of macroscopic grazers might explain the low diversity and long stratigraphic ranges of most Proterozoic acritarchs (Peterson & Butterfield 2005).

Where is the line of demarcation between these two states of the marine ecosystem? Peterson & Butterfield (2005) drew it at the beginning of Early Ediacaran acritarch radiation, in the wake of global Marinoan glaciation. Post-Marinoan rocks contain the first acritarch assemblages dominated by forms with abundant, symmetrically arrayed spines, features interpreted by Peterson & Butterfield (2005) as mechanical defences against predation by newly evolved benthic bilaterians. This explanation is plausible, although spiny Ediacaran acritarchs antedate the first geological records of motile bilaterians by as much as 25–50Myr, and the entire assemblage of spiny acritarchs disappears by the time that bilaterian trace fossils enter the record. The systematic relationships of Early Ediacaran acritarchs remain unclear, but given their size, (in some cases) multicellular contents and strong morphological similarity to egg hulls made by living animals (Van Waveren & Marcus 1993), these distinctive fossils could well provide direct rather than exclusively indirect evidence for early animal diversification (see Yin et al. 2004). With less conviction, grazing arguments might also be applied to the Early Ediacaran radiation of macroscopic seaweeds.

Whatever the nature of specific interactions, there can be little doubt that emerging animals influenced the continuing evolution of protists in Late Neoproterozoic and Cambrian oceans. Does this obviate any role for environmental change? Of course not. Environmental and ecological facilitation are not mutually exclusive, and so ecological plausibility does not constitute evidence against environmental facilitation (pace Peterson & Butterfield 2005). Hypotheses of environmental influence may be correct or incorrect, but their tests lie in the precise stratigraphic resolution of geochemical and palaeobiological data.

The dawning age of animals is associated with several environmental events of limited duration and one long-term state change in the Earth system. Two global ice ages (and at least one additional, regionally important glaciation; Halverson et al. 2005) and a brief but pronounced C-isotopic event coupled to transient shallow water anoxia near the Proterozoic–Cambrian boundary (Amthor et al. 2003) could each have influenced evolution by removing a large proportion of pre-existing eukaryotic biomass. This would create permissive ecologies in which novel (and not necessarily fitter) mutants generated by surviving populations could persist under low selective pressure, providing raw materials for evolutionary innovation (Knoll 2003b). We have only limited knowledge of life after the Sturtian (ca 720–710Myr) ice age, but there is evidence of evolutionary innovation in the wakes of both the Marinoan (ca 650–635Myr) ice age and the Proterozoic–Cambrian boundary event.

There is also widespread agreement that oxygen levels rose in the surface ocean and atmosphere during the Neoproterozoic Eon (e.g. Canfield & Teske 1995; Catling & Claire 2005; Holland 2005). Evidence from S-isotopic fractionation (Shen et al. 2003), secular variation in S isotopes (Kah et al. 2004), and sulphate abundances in carbonate lattices (Gellatly & Lyons 2005) collectively indicate that sulphate levels in Mesoproterozoic oceans were lower than today's by about an order of magnitude, implying P_O2 lower than at present. Fe speciation chemistry (Shen et al. 2003), Mo isotopes (Arnold et al. 2003) and biomarker profiles that document anoxygenic phototrophs in open waters (Brocks et al. 2005) independently suggest that sulphidic conditions were easily induced in the oxygen minimum zones (if not all deep waters) of Mesoproterozoic oceans.

The unresolved question is when oceans transited to a more modern state. A single sample from ca 750Myr black shales in the Grand Canyon exhibits the increased S-isotopic fractionation associated with oxygenic water columns (Canfield & Teske 1995), but marine shales do not routinely preserve this biogeochemical signal until after 580Myr (Hurtgen et al. 2005), suggesting a potentially close temporal relationship between the rise of oxygen and the appearance of both macroscopic animals and seaweeds. Rising oxygen might have influenced algal evolution in two ways—indirectly via the origin and diversification of large bilaterian animals and directly by the effects of redox change on nutrient status (Anbar & Knoll 2002). Unfortunately, no known geochemical process trips a recordable mineralogical or biogeochemical switch at PO₂ levels that constitute physiological barriers to large size and motility in animals. Complicating the issue further, there is no reason to believe that oxygen levels trended monotonically through the Proterozoic Eon. The presence of iron formation in successions associated with Sturtian glaciation suggests to some (e.g. Canfield 1998; Holland 2005) that oxygen levels may have been especially low during times of Neoproterozoic climatic extremes.

In the end, all hypotheses to explain observed patterns of protistan evolution require better stratigraphic data: geochemical data that illuminate short-term climatic and biogeochemical events as well as long-term redox change, and fossil and biomarker data that document the first appearances of protists and ecologically important animals. Better data, in turn, will require improved models—oceanographic, palaeophysiological and ecological—to integrate palaeobiological and palaeoenvironmental observations. We predict that a mature understanding of protistan evolution will implicate genetics, ecology and geophysical change rather than single any one of them out as a solitary driver.

Much has been learned about Proterozoic life in the oceans, but, clearly, much remains to be learned. Straightforward discovery played a principal role in the past two decades of research, and it will likely continue as the main source of new understanding for some time to come. Particularly important will be the elucidation of well-preserved assemblages from post-Sturtian but pre-Marinoan rocks and from Palaeo-Mesoproterozoic successions.

New ways of looking at existing fossils will also loom large, however; TEM and microchemical techniques that can elaborate the physical and chemical composition of individual fossils will increasingly provide insights into phylogenetic relationships of early eukaryotes (e.g. Arouri et al. 1999; Talyzina 2000; Javaux et al. 2004; Marshall et al. 2005; Schopf & Kudryavtsev 2005). Finally, improved geochemical techniques for tracking the redox history of seawater, coupled with new and better insights into protistan genetics and ecology, will provide badly needed context for understanding the major changes recorded by eukaryotic fossils in Ediacaran and Early Cambrian rocks. The same questions that animated Proterozoic palaeontological research a decade ago animate it now, but the data are richer and the answers are getting better. The same will undoubtedly be true a decade hence.

Research leading to this paper was supported by NSF grant EAR-0420592 (A.H.K.), an NSF graduate student fellowship (P.C.), and a return grant from the Belgian Federal Science Policy (E.J.J.). We thank Nick Butterfield and Malcolm Walter for images used in figures 2 and 3, Jon Wilson for comments on a preliminary draft and Nick Butterfield and Susannah Porter for constructive reviews.