The overwhelming similarity of several molecular components of complex developmental pathways of the genetic model organisms (fruitfly, soil nematode, mouse and zebrafish) gives strong evidence that these pathways were already present in the most recent common ancestor of these animals. Examples include the anterior–posterior patterning mechanism using HOX genes, the sog/chordin dpp/BMP2/4 in dorsoventral patterning (Arendt & Nübler-Jung 1994; De Robertis & Sasai 1996) and PAX6/eyeless for eye development (Gehring & Ikeo 1999), and were extended through the discovery of NK-class gene function in ‘heart’ tissue of both Drosophila and mouse (Harvey 1996), finding common members of an immune system (Hoffmann et al. 1999) and potential similarities in the process of body segmentation (Balavoine & Adoutte 2003; Tautz 2004). Soon afterwards, specific anterior patterning genes (empty spiracles/EMX and orthodenticle/otx) responsible for the formation of a tripartite brain (Reichert 2005) and posterior patterning genes, even skipped/evx (Patel et al. 1992) and caudal/cdx (Wu & Lengyel 1998) were added to the list of shared features. Together these studies supported the notion of a molecular ‘zootype’ defining a common plan for the construction of all animals (Slack et al. 1993). The extension of gene expression studies to members of the third ‘superclade’, the Lophotrochozoa, primarily members of the Annelida, appears to confirm the notion of a morphologically complex urbilaterian (Shankland & Bruce 1998; Arendt et al. 2001, 2002; Prud'homme et al. 2003; Seaver et al. 2005; Fröbius & Seaver 2006; Seaver & Kaneshige 2006; Denes et al. 2007; Steinmetz et al. 2007). This rather complex bilaterian ancestor would have possessed serially reiterated body segments, a mesodermally lined coelomic space, a through gut with separate mouth and anus, lateral appendages, heart, eyes, epigenetic germ-cell determination (Extavour & Akam 2003; Rebscher et al. 2007), a tripartite brain with a centralized nervous system and a biphasic life cycle with a feeding larvae (Arendt et al. 2001; Carroll et al. 2001; see figure 2).

The subdivision of the Metazoa in to three major clades, the Ecdysozoa, Lophotrochozoa and Deuterostomia, focused the hunt for the ‘Urbilateria’; however, the lack of resolution within these three clades prevented rational statements about the characteristics of each clade's stem species and thus the direction of evolutionary change. With recent improvements in animal phylogeny, mainly by the increase of the number of genetic loci and improved taxon sampling (‘phylogenomics’), characters found in animals such as polychaete annelids, like segmentation, trochophore larvae, spiral cleavage and amphistomic gastrulation, might now be viewed as derived features of specific clades (Giribet 2008). The placement of the unsegmented Chaetognatha as the sister group to the Lophotrochozoa (Matus et al. 2006a) has a huge impact on our understanding of the lophotrochozoan stem species, and recent work placing the enigmatic Xenoturbella inside the Deuterostomia as sister to the Ambulacraria (Bourlat et al. 2003, 2006) indicates that both protostomes and deuterostomes have rather simple unsegmented worms with only one body opening and a non-centralized nervous system near their base. In particular, the placement of the Acoela and Nemertodermatida as the possible sister group of the remaining Bilateria (Ruiz-Trillo et al. 1999, 2002, 2004; Jondelius et al. 2002; Telford et al. 2003; Baguñá & Riutort 2004) allows a more critical evaluation of the statements made about the morphology of the urbilaterian. The first obvious result of this placement is that the stem species of the Bilateria does not correspond to the protostome–deuterostome ancestor, but to the last common ancestor of the Acoelomorpha and the remaining Bilateria (the Eubilateria or Nephrozoa; figure 1). If acoelomorphs do occupy this pivotal position, they will provide important insight into estimating the morphological complexity of the urbilaterian.

Despite initial optimism that an increase in genomic complexity and a cursory analysis of comparative gene expression of a handful of conserved genes in a few model systems might lead to a simple explanation of the evolution of organismal form (Carroll et al. 2001), recent evidence suggests that greater care needs to be exercised when performing these analyses. There are no such things as ‘segmentation’, ‘eye’, ‘heart’ or ‘limb’ genes. Genes that are supposed to specify ‘mesodermal’ cell types are present in cnidarians, which do not even possess muscle or a mesodermal germ layer (Spring et al. 2002; Martindale et al. 2004), and animals that possess several different kinds of eyes do not possess ‘the eye gene’ Pax6 (Matus et al. 2007). There are just molecules that can bind to DNA or interact with receptors, phosphorylate other molecules, etc. Furthermore, not even the most intimate understanding of a gene or genetic pathway can predict a priori its developmental/morphological outcome. The problems with homologizing structures using only the underlying regulatory developmental gene network have been discussed elsewhere and had led to the explosion of new ideas around the notion of ‘homology’ (Dickinson 1995; Bolker & Raff 1996; Nielsen & Martinez 2003; Scholtz 2005; Wagner 2007). Recent findings from the analysis of the genome of cnidarians show that conclusions based on the relationship of the morphological ‘complexity’ of an organism to its gene content are over simplified (Putnam et al. 2007).

Confusion reins regarding the role of gastrulation in body-plan evolution. Its importance in understanding the evolution of development hails back to the earliest days when animals were categorized as being either protostomes (the site of gastrulation becomes the mouth) or deuterostomes (the site of gastrulation becomes the anus; Grobben 1908). Gastrulation is a complicated process in all animals and is difficult to define and study. In its simplest sense, it is the formation of distinct germ layers (endoderm, endomesoderm and mesoderm) but the process is dynamic and can begin and end over highly variable periods (i.e. when embryos have 28 cells in soil nematodes to thousands of cells in vertebrate embryos), and precursors of these tissues can arise from multiple locations in the embryo (e.g. ectomesoderm and endomesoderm). In most embryos, the site of gastrulation has a definitive position relative to the primary egg axis, the animal–vegetal axis. Deuterostomes, for example, gastrulate at the vegetal pole and this site becomes the future anus with the mouth forming secondarily in the animal hemisphere. In most protostomes, the site of gastrulation (endomesoderm formation) also derives from the vegetal pole, but this location does not correspond to either the mouth or the anus. For example, in spiralian embryos (Boyer et al. 1998; Maslakova et al. 2004; Ackermann et al. 2005; Hejnol et al. 2007), intracellular fate-mapping experiments clearly show that endomesoderm forms from ‘macromeres’ at the vegetal pole, but the mouth and foregut form from ectodermal derivatives in the animal hemisphere posterior to the prototroch (first and second quartet micromere derivatives) and the anus (when present) forms de novo from cells that are pushed posteriorly and ventrally by asymmetric growth on the dorsal side of the embryo. The same pattern is seen in acoel flatworms (Henry et al. 2000) where endomesoderm forms from both third duet macromeres at the vegetal pole and the mouth forms anteriorly as 1a micromere descendants expand around the posterior pole (although an anus never forms).