I have commented several times on the early evolution of the animals, especially in regard to their ability to handle information. Here I want to look at the progression from a plausible common ancestor of the latest-branching clade of sponges and the Eumetazoans to the point of a plausible last common ancestor of the Ctenophores, the Cnidarians, and the Bilaterians.
Symmetry and Development
Since, in my view, all (or most) of the animals in the sequence we're looking at had biradial symmetry, I'll start with a discussion of that. (Especially since Wiki has only a stub on the subject).
First, as I've mentioned before, symmetry in all animals that move in response to external directional cues must be superficial. Even if the animal is only a sphere with sensory cells and ciliated swimming cells distributed evenly over its surface, the communications needed to coordinate its swimming with the direction of threats and food will require each cell to know its location relative to the whole. There can be no confusion between left and right, front and back, top and bottom. This requires a coordinate system of some sort, even if the physical morphology doesn't show it. For a spherical surface, there are two obvious options: radial (or angular) and multipolar.
The radial system is like the system of latitude and longitude: it starts with a pair of poles on opposite sides of the sphere, which we will call "oral and "aboral", named after the blastopore or (single) gut opening in these early animals. (Sponge larvae also demonstrate this polarity in their distribution of cell types, although they lack a blastopore.) A line between the poles (along the surface -- one through the center of the sphere is the axis) is then defined as the "zero longitude" line, after which any spot on the spherical surface may be uniquely located by its angular location relative to the poles (latitude), and its angular location relative to the "zero longitude" line. Of course, it will still have to know clockwise from counter-clockwise: 30° longitude is very different from -30°.
The multipolar system also starts out with a polar axis, but then goes on to define another (which might roughly correspond to 0° and 180° longitude) and then yet another (which might correspond to 90° and 270°). any location on the spherical surface is uniquely defined by three numbers: how far it is between each of the three pairs of poles.
Controlling the Development of Symmetry
Now, location in animals is almost always defined in terms of the diffusion of signaling molecules, a system well supported by a good epithelium. The tight junctions among all the superficial cells prevents these molecules from diffusing out into the surrounding medium, and the basal layer(s) can be specialized to allow much faster diffusion along than across it. The first set of poles, then, can define "latitude" with a cluster of cells at one end releasing a diffusible signal: latitude is known by the concentration of the signal. ...
A more robust mechanism would have cells at the other end also releasing a signaling chemical, a different one. Now a cell on the surface could get its latitude by sampling the relative concentrations of two chemical signals. This is a typical system in some developing embryos, although the oral signal usually comes from a ring of cells surrounding the blastopore.
If we try to use such a mechanism in the radial coordinate system, the "zero longitude" line would have to emit its signal in only one direction (or two signals, each in one direction), or else it would be impossible to distinguish clockwise from counter-clockwise. (Alternatively, a separate signal could be released at a point some angle away from the "zero longitude" line, but this is becoming a hybrid with the multipolar system.)
The multipolar system has this problem solved, at the cost of three signals (or pairs of signal chemicals). this is, in effect, what bilaterians all use, and IMO it's likely that research will eventually discover that the Ctenophores and Cnidarians do as well.
With all this in hand, let's look at symmetry. Since the Cnidarian medusa forms have a well-defined radial symmetry around their oral/aboral axis, we'll start with that. If we imagine a glove with poles, latitude, longitude, and an equator, let's break the equator into four equal sections, broken at 0°, 90°, 180°, and 270°. We'll define a primary mirror image at each break line, so each section looks identical to its reflection in each mirror. We'll also define a second set of mirrors intermediate to the the first, at 45°-225° and 135°-315°. Each section, then, is identical to every other, and to its reflection across the primary mirror line. Each section's halves are also identical but inverted across the secondary mirror line running down its middle. This is radial symmetry with an index of four. A simple square has this sort of symmetry. It can be divided into four identical sections, reflected in the two mirror axes, with each section having identical but inverted "left" and "right" halves. Note that if you divide a square into eight sections, four of them will be inverted relative to the others. Note also that it doesn't matter which set of axes you define as primary, and which secondary: a square can be divided into four sides (sharing corners) or four corners (each with two half-sides).
Now, many Cnidarians have a higher index than four, such as eight, or sixteen. It's easy to see how these body plans can be achieved developmentally. for instance, if we start with an index of four, at each point where one of the primary planes of symmetry (reflection) intersects the equator (or some other circle of equal "latitude"), there's a small patch of cells emitting a chemical signal. For robustness, there should be another patch at each point where a secondary plane intersects, emitting another chemical. By sampling this signal, any cell can know where it is within the section it occupies. Note that for coordinated control, each arm, at least, will have to know its location relative to the entire organism's coordinate system.
How does a developing organism double its index? At a specific point in development, a signal is released telling the cells at the secondary points to begin emitting the primary signal (instead of the secondary), and a group of cells halfway between each set (who already knew their position from the earlier signals) begin emitting the secondary signal. Now the index is eight. this process can be repeated as many times as needed, and it's a simple mutation to increase (or decrease) the number of repeats.
I've described how the multipolar system could produce four poles equally distributed around the equator, if a patch of cells at each of these poles started emitting the primary signal, and cells that found themselves at the midpoints began emitting the secondary signal, you've got radial symmetry with an index of four. But more interesting is how those four poles came to be.
Let's start with a simple sphere with no coordinate system except a single patch of cells (somewhere, anywhere on the surface). If these cells begin emitting a signal (different from the signals used in radial symmetry), it will diffuse around the sphere while breaking down and diffusing away (at right angles to the surface, which is presumably an epithelium capable of controlling the diffusion rates in different directions). Thus the concentration will reduce from a maximum at the emitting pole to a minimum at its antipode. Cells are capable of detecting a concentration gradient, the fact that there is none is enough to let cells at the antipode know that they should begin emitting the opposite signal.
More problematical is the next pole. Some mechanism is needed to select this pole (somewhere on the equator), various species use various methods, some of which are essentially random. Once this pole is selected, cells here may begin emitting a new signal (still probably different from those used in radial symmetry), which diffuses around the equator and allows its antipode to form in the same way as the first. The third pole also requires some external mechanism to distinguish between left and right, although cells equidistant between the second poles will know it. In many species this selection is random, although mammals and everybody's favorite flatworm Caenorhabditis elegans have fairly reliable mechanisms for determining their left side.
When it comes to radial symmetry, we've seen how four-fold symmetry can come about, but what if only two poles begin emitting the primary signal, while those at right angles to their axis begin emitting the secondary signal? This is radial symmetry with an index of two: biradial symmetry. Ctenophores have a fundamental biradial symmetry, as do most Cnidarian polyps.
(Ctenophores have imposed an eightfold symmetry on their oral end and the major expansions of their gut, but the aboral end is biradial. (The location of the "anal pores" breaks this symmetry, creating a rotational symmetry with an index of 2, but this is a minor feature, and I'm going to class it with the giant left claws of some crabs: a minor break asymmetry built on the basic symmetry.) Cnidarian Medusae also have a basal biradial symmetry underlying their 4-, 8-, or 16-fold symmetry. And remember that all symmetry is ultimately superficial: there is always an underlying coordinate system that uniquely defines each point.)
Symmetry and Controlling the Body
When it comes to controlling the body, symmetry has an advantage in that the same behavioral program can be used for each unit of symmetry. This applies both to the behavior of cells in the developing organism, and the behavior of units (limbs, tentacles) in the active animal. Even when differences arise, they can be overlaid on the repeated pattern: there will usually be many more similarities than differences, which means much less information is needed to define the entirety when symmetry is used. It also means that mutation to the underlying pattern can change all the units at once, which has advantages and disadvantages when it comes to overlaid differences. (But selection can take advantage of mutations that play to the advantages, while weeding out those that run into disadvantages.)
This makes much more difference to an animal that uses muscles to make coordinated changes to its body shape than to blobs swimming around with cilia, so for the origins of Eumetazoan symmetry we need to look to the ultimate origins of muscles.
Symmetry in the Earliest Animals
I've suggested that the common ancestor of Eumetazoans and the latest-branching clade of sponges, the homoscleromorpha, was a small globular animal like a swimming sponge: containing internal water canals, with the porocytes spread across the front, cilia for swimming in a broad belt around the middle, and a fairly sophisticated system for controlling its activity. What I didn't make explicit was that the osculum, the exit for water flowing through the body, would have been in the rear, so that the flows of water from swimming and feeding could complement one another, and so that oxygen-depleted water from either would not become input tot eh other. Being roughly spheroidal, this creature's swimming actions would have been identical in any direction around its "oral/aboral" axis.
It might seem, in these early days, that there wasn't much need for swimming, given the general lack of predators, but we need to consider the early environment: most models of the pre-Cambrian biosphere suggest that the deeper oceans were anoxic, with only a shallow layer of oxygenated water overlying it. With much of the productivity confined to the coastal areas, where tides and wave actions continually stirred things up, there would have been vast opportunities for collecting food in regions with substantial risk of being stranded by wave action. this means that some ability to predict wave action and position itself so as to be washed back out in a breaking wave rather than getting stranded on a sandy beach would have given an animal a selective advantage, even for an animal that could only swim with cilia. More predictive ability would have conferred more benefit, so we have a situation where the evolution of intelligence has a nice steep "fitness slope" to climb, even without changes to the body structure.
In this regard, I want to discuss a particularly important paper, which is unfortunately behind a paywall: Sponge Genes Provide New Insight into the Evolutionary Origin of the Neurogenic Circuit (by Gemma S. Richards, Elena Simionato, Muriel Perron, Maja Adamska, Michel Vervoort, and Bernard M. Degnan). What these researchers did was identify a gene in an early-branching sponge, Amphimedon queenslandica which is a demosponge, that appears to be descended from an ancestor of many types of genes involved in controlling the development of nerve cells in the bilaterians, and presumably also the other eumetazoans. Not only that, but they injected mRNA for the protein coded by this gene into frog and fruit fly embryos, and determined that it performed many of the neurogenic functions of its purported relative genes native to these animals.
Now in Amphimedon queenslandica this gene, called AmqbHLH1 is expressed along with a similar relative of the notch receptor genes of eumetazoans and a notch ligand, AmqDelta1, in the outer layer of cells of the early embryo. As it develops, the embryo separates into three layers, the inner cell, subepithelial, and epithelial.
Prior to the formation of the subepithelial layer (“spot stage”), all three of the conserved neurogenic genes are expressed in the outer layer (Figures 1A–1D), with AmqNotch expressed broadly and AmqDelta1 and AmqbHLH1 restricted to subpopulations of cells. A few hours later at the “early ring stage” (Figures 1E–1H), all three genes are coexpressed in cells of the subepithelial layer. In “late ring embryos” (Figures 1I–1L), AmqbHLH1 and AmqNotch remain expressed in the subepithelial layer, but AmqDelta1 is now highly expressed in globular cells that are located beyond the subepithelial layer, within the outer layer. We propose that the expression of AmqDelta1 in the globular cell population is continuous and that it tracks a specific migratory event from its initiation, when cells leave the subepithelial layer, until its conclusion, when the cells reach their final position scattered around the outer margin of the mature larva (Figures 1M–1P).These Globular Cells appear to function as sensory cells, helping the sponge "larva" find a spot to settle and metamorphose into a small version of the adult form.
Now, I've previously suggested that when looking at larvae, we shouldn't consider them to be representative of the immediate ancestor of the adult form, but rather a specialized developmental stage, simplified in many respects for the purpose of extremely rapid growth, that may in some respects recall ancestral forms, often very distant ancestors. I've also pointed out that in the classic definition of "larva", this form is specialized for growth, and we shouldn't assume that a form that does little but swim around and settle is similar to classic larvae. But it's reasonable that the "larvae" of sponges are descended from a more classic version of larvae, in which case their ancestors spent some time swimming around finding and eating large amounts of food before settling and metamorphosing.
So it's plausible that these larvae used an ancestral method of controlling their swimming, harking back to a "swimming sponge" ancestor that possessed a sophisticated system of controlling its movements, combining sensation and intelligence. The fact that remnants, quite likely far simpler than what the ancestor had, may be found in ancestral larvae makes it very plausible that the ancestor itself had something much better. It may not have been strictly divided into nerve cells and sensory cells as is often the case in more developed eumetazoans, but that didn't mean it wasn't capable of considerable intelligence.
And these weren't just the common ancestors of the latest branch prior to the Eumetazoans: Amphimedon queenslandica is a demosponge, a group that branched prior to the homoscleromorphs, showing that intelligence was already evolving prior to the latest branching.
What, precisely, were the active resources available to this evolving intelligence? We know that some sponges can propagate an electro-chemical (actually electro-osmoidal) wave along their epithelia, stimulating closure of the porocytes, the latter function being present in all sponge lineages. We also know of sponges that can expel a large proportion of the water in their canal system, although the process takes too long to be useful in emergency swimming. (But then, sponges don't swim, so it's plausible that this function had degenerated from a faster one in swimming ancestors.)
But what if our early "swimming sponges" had a greatly enlarged osculum, such that they resembled a hollow sphere with an opening in the back and their systems of water canals and their bodies in a layer surrounding the hollow full of water? In this case, a contraction around the equator, coupled with a relaxation at a right angle, along the "oral/aboral" lines, could have produced a large reduction in the volume of the enclose space, without reducing the surface area. Indeed, a reduction in surface area would have been no problem, as long as a thickening of the body layer could compensate to keep the overall body volume unchanged, since the major speed impediment is squeezing the water out of all those tiny channels.
We know, from observations of Cnidarians, that it's possible for epithelial cells to produce substantial contractions, and other changes in shape. What makes muscles so powerful is their organization of myosin, actin, and other proteins into specialized linear structures that can shorten quickly in response to action potentials.   All Eukaryote cells possess forms of myosin and actin, which they use for a variety of movements within the cell, including changing its shape and moving organelles around. The organization of many overlapping fibers of actin and myosin into something like muscles probably constituted a smooth "fitness slope" which could be easily climbed through progressive small mutations to the organization and structure of myosins and actins in the cell.
We can start with the epithelial cells creating bundles of these overlapping fibers connecting membranes on opposing sides of the cell, with some sort of trans-membrane structures providing mechanical linkage across the pair of cell walls. If the cells were connected by gap junctions that could propagate the electrical waves of action potentials from cell to cell, the entire skin could have contracted quickly to expel water in the oversized osculum for escape swimming. The development from this point to the simple myotubes of Ctenophores is probably a progression of simple refinements.
A key obstacle to this sort of swimming is that there are limits to the internal flexibility of the water-canal filled sponge, which in turn produce limits to how much water could be expelled. Although it's hard to be sure, I suspect that no more than 30% to 50% of the total volume could be made up of expellable water. To enlarge the body cavity further, allowing for greater emergency swimming, the creature would have to give up its circular symmetry around the "oral/aboral" axis, and collapse in some way that produced a lesser symmetry.
An obvious candidate is biradial symmetry. A spheroid can flatten into a shape roughly like a pumpkin seed, reducing its volume to a tiny fraction of the original. But to do this, the internal skeleton (presumably much like that of small modern sponges without spicules), would have to become specialized in a biradial pattern. The part that becomes the flat part of the "pumpkin seed" would have to be resistant to further bending, while being able to flatten out from its spherical curvature. And the part that becomes the edge would have to be much more conducive to bending, becoming something of a hinge.
Similar biradial specializations would be necessary for the contractile cells. Indeed, while "shrinking" a sphere could be accomplished by contracting the outer skin, flattening it would require tension from more internal cells. If such a system evolved gradually, the end result would be lines of cells penetrating the internal structure (between the water canals) which could contract, pulling on one another, to pull the appropriate parts of the outer skin into the "pumpkin seed" configuration. These structures would be different in different areas around the equator, in a biradial configuration.
Note that this simple system is only possible with an index of two. Radial systems with higher indices would require not only a convex bend in the outer skin, but concave bends. The structure of tensile elements (primitive myotubes) would also be extremely complex. It is highly unlikely then that a higher index of symmetry would have arisen under these circumstances.
What we have, then, is a small swimming sponge in the form of a hollow sphere, with a fairly sophisticated "brain", capable of maneuvering through the wave-stirred coastal waters where large amounts of edible detritus was available and the constant stirring of the water provided a good supply of oxygen from the air. It probably didn't have nerve cells (in the sense of having axons carrying action potentials), rather the cells that performed calculations and communications may have looked much more like the modern mammalian astrocytes: cells with multiple processes reaching out to all (or many of) the cells in its vicinity, having an ability to perform localized calculations and carry signals by means of "calcium waves" that could transfer from one cell to another over the electrical connections made by gap junctions. They may well have also performed sensory functions, lacking the division into specialized cell types found in Eumetazoans.
Building the First Eumetazoans
The transition from this to the Eumetazoans almost certainly required the key transition from internal water canals and chambers lined with choanocytes to a system of mucociliary feeding similar to that seen in a huge variety of Eumetazoan larvae and many adults. In this system, mucus is secreted from parts of the skin that are supplied with large numbers of cilia of two types: one type stays within the mucus and keeps it moving, eventually collecting into channels which feed it into the mouth. The other type sticks up out of the mucus, and sweeps water along the mucus covered skin, moving in a way that tends to force suspended particles (and life-forms) down into the mucus where they stick and are drawn along with it into the mouth where they are digested.
Mucus is a rather complex secretion, present in sponges as well as most Eumetazoans. It performs important protective functions by trapping would-be invaders where they can be isolated from the underlying skin. It is made up primarily of glycoproteins and complex sugars, well supplied with other materials secreted by the underlying cells. The composition can easily vary by species, and many of the various mucins can co-evolve with digestive enzymes, allowing a species to secrete mucus it can easily digest but that most of the living creatures trapped in it can't break down.
The easiest path by which a swimming sponge might evolve into a mucociliary feeder is to begin with using mucus as a defensive mechanism. Many sponges have cilia, so it's an easy adaptation to using them to move secreted mucus along the surface to somewhere that trapped would-be invaders wouldn't be a threat. Most sponges have amoeboid cells that can feed using phagocytosis, which actually isn't that different from the process used by choanocytes when one of the microvilli around their flagellum traps a food particle. (Indeed, there are sponges that have multinucleated aboeboids that feed on small crabs and other crustaceans with a similar process.) It's an easy adaptation for these cells to develop the ability to secrete proteins (or keep them on the outer surface of their cell membranes) that would cause the mucus to become thinner right around them, allowing them to "swim" through it and chase down, trap, and eat whatever they find caught in it. They would also have the ability to collect large amounts of this mucus, digest it back into its basic components, and return packages of these materials into the internal matrix of the animal, where they could be transported back to the cells that secrete mucus for reuse.
The next adaptation would be to modify part of the inner surface of the osculum as a "trap" for this mucus, with the cells lining it being specialized to be resistant to attack from whatever invaders are caught in the mucus. Alternatively, this trap might be lined with a syncytium of cells capable of surrounding and absorbing such invaders. A further adaptation would be for the "trap" to develop the ability to close itself away from the outside intermittently, then some of the cells lining it could secrete acids or other chemicals that would alter the environment to one much more hostile to invaders adapted to the open ocean (or whatever). This would also allow the secretion of digestive enzymes, both to break down the mucus for reuse of its components (without the effort of absorption by the amoeboid cells, and to break down the tissues of the various trapped invaders. At this point we have a rudimentary gastric system, although it has originated as part of a protective mechanism rather than the primary method of feeding.
However, given a creature that already uses cilia to move water around for swimming, and has now developed their use to move mucus over the skin, it's another easy adaptation to modify some of the ciliary action to sweep food particles down onto the film of mucus, enhancing the ability to trap food. At this point, it's reasonable to assume that some lineages would specialize in this process, abandoning the use of choanocytes and collared flagella. The developmental mechanisms they had used to control the creation of water canals and choanocyte chambers would then be free to elaborate the shape of their ectoderm, providing specialized improvements for increased trapping of food with their now fully-developed muco-ciliary feeding system.
Here, then, is our fully emerged Eumetazoan. It has biradial symmetry, uses cilia to swim, posses a blastopore, into which mucus full of suspended food is periodically swept, after which it closes so acids and digestive enzymes can be secreted into it and digestion can take place. The lining of the digestive tract is probably a syncytium, as it is in the most basally-branching bilaterians, which may well preserve this feature from the earliest Eumetazoans. (As I recently pointed out, both the Cnidarians and the Ctenophores are probably highly specialized, and cannot be expected to represent the early common ancestors.)
Where does it go from here? When we look at the ctenophores, we see a creature with a rudimentary sensory organ at the aboral end, two tentacles near it, and a system of (typically) eight lines of cilia running up from the oral opening towards the equator. The Cnidarian medusa, which I've argued is the ancestral adult form of this group, has four-, eight-, or sixteen-fold radial symmetry, which as shown above could easily develop from a biradial ancestral condition (something regressed to in most of the (IMO) larval form represented by the polyp). It's been shown by gene expression studies that the tentacles of polyps are not homologous to those of the Ctenophores, but I have an alternative suggestion.
I would propose a common ancestor with biradial symmetry, two tentacles at the aboral end used for swimming and other manipulative processes, and a set of tentacles around the mouth used primarily for mucociliary feeding. The swimming would be for escaping wave action and possibly predators, and the location near the aboral sensory system and brain would allow faster communication. The tentacles near the mouth would be well provided with cilia, along with mucus grooves, and normal "cruising" would be done using these. Nerve cells would have been present (as they are in all descendant lineages), and muscle cells would have also been present, primarily in the aboral tentacles.
The progression from this ancestor to the Ctenophores is easily envisioned: as it converts from filter feeding to pursuing and eating larger prey, the cilia surrounding its mouth become specialized for only swimming, "cruising", and they fold up around the body, producing lines of cilia spreading up from the oral end, and allowing a shorter path for messages from the brain controlling them. The aboral tentacles, in turn, become specialized for capturing prey. An intermediate form might have been one that supplemented its original mucociliary feeding with another type of filter feeding: sweeping its tentacles through the water and trapping and eating whatever they encountered. This would provide an easy evolutionary path for the development of the colloblasts. They might well have begun as objects on the feeding tentacles surrounding the mouth, allowing them to trap any food particles they encounter, while the primary feeding method remained mucociliary. Once they spread to the aboral tentacles, those tentacles could have begun using a sweeping process to capture objects that the oral tentacles couldn't, taking advantage of their muscles and faster responses.
Similarly, the progression towards the Cnidarians is equally easy to see. Beginning with a similar animal, the aboral tentacles, already adapted for swimming, become duplicated, creating four, and widen towards the bell shape found in the modern medusa. This would have started with a swimming action in which both tentacles pushed back at the same time, probably used by the original common ancestor. Duplicating the tentacles, and widening them, would have been easy mutations, not having to occur in any specific order. Once both had occurred, fusing them into a bell is another easy mutation. After that, the development of tentacles hanging off the edge of the bell is also an easy mutation. If the common ancestor had already had colloblasts, or something ancestral to them, these cells could later have become specialized for capturing active creatures with high-flow bloodstreams and complex nervous systems. (I've discussed this progression already, in Coelenterates, Cnidarians, and Larvae.)
Getting back to the common ancestor, how did it come about? Let's go back and start with the small spheroidal creature, with an outer covering of mucus-secreting skin and cilia to sweep this mucus back into the osculum, where it enters a primitive digestive system. We've seen that the "swimming sponge" would have already had biradial symmetry and a sophisticated "brain" it could use for steering itself through the chaotic waters of food-filled beach areas. Being hollow, with most of the volume enclosed by its body in the expanded form consisting of water it could expel for swimming, when it needed to escape some wave-created threat it could move quite far and fast for a very short distance.
What if, re-using the same developmental mechanisms it used to control the growth of the water-canal system, it developed a pair of tentacles sprouting from the "flat" sides (the sides that become flat when it expels water)? These tentacles would ideally be as close as possible to the "brain" at the aboral end, and the myotubes along their length could have one end inserted into the "ganglion" that constituted the "brain", thus allowing this creature to steer itself during its escape swimming. This mechanism might well have been present before the development of mucociliary feeding; it might even have been present in the common ancestor of the homosclerimorphs and the eumetazoans (and lost by the former as they evolved towards sessile feeding).
Given the ability to create tentacles for one purpose, it's yet another easy mutation for them to appear for another, in this case at the oral end. If they were covered with mucociliary apparati, they would have enhanced the capture of food particles that didn't make it into the pores, because they were too far away from the line of swimming. (Particles that were too large would have become trapped in the mucus covering the body; indeed that might well have been the primary reason for adopting mucociliary feeding as a strategy in place of simply recovering and reusing defensive resources and energy.)
An alternative model would be that the mucus flow along the body became concentrated in a pair of channels where they could flow into the osculum, around the rear edge. This way, the mucus flow would be somewhat isolated from the flow of water, both normal and during escape swimming. These channels could then have become supplemented by tentacles, again allowing the capture of food particles too far away from the swim line.
Once there were two tentacles, it was yet another easy mutation for them to multiply to four, eight, and even farther (as discussed above). The more tentacles, the greater fraction of food that missed the body could be captured.
A creature with two major feeding strategies occupies an unstable position evolutionarily. While such a life-style may be successful, even wildly successful, there is always the option for one or more descendant lineages to specialize in one and abandon the other. Since the original strategy was a sponge like process, the progression would have been towards total dedication to mucociliary feeding, using the oral tentacles. The cilia around the main body would have been abandoned (at least for normal swimming), since those on the tentacles would have been sufficient to capture anything in the water stream. The hollow filled with water would also probably have eventually been abandoned, being replaced with the more efficient swimming using the aboral tentacles. And there you have it, the common ancestor of the Cnidarians and the Ctenophores proposed above.
Now, most molecular investigations of the early ancestry of modern Eumetazoan lineages tell us that the Ctenophores branched first, with the other branch leading to the common ancestor of the Cnidarians (Coelenterates) and the Bilaterians. The common ancestor of all three proposed above, then, would have been followed by the common ancestor of the Cnidarians and Bilaterians. In my view, it probably looked very much like the earlier ancestor, but with a more developed nervous system.
The key to the invention of the nerve cell is the need to carry rapid messages over long distances, where muscle fibers can't do the job (because if they can there's no need for nerves, as shown above for the muscles in the early aboral tentacles, as well as some acoels, a group generally thought to be basally branching bilaterians, and thus likely representative of the early state).
We find nerve cells controlling the action of cilia in the Ctenophores, and this actually provides the most plausible explanation for their invention: the long tentacles extending from around the mouth need their cilia controlled, especially during escape swimming. We can see how cells originally more like astrocytes became specialized to carry a rapid action potential along their length so that they could deliver an almost simultaneous message to these ciliated cells. Once this feature had been developed, such cells were also used for controlling muscles, allowing them to connect any two points on the epithelia (either ectodermal or endodermal) without the extra effort of extending processes to the "brain".
Note that there's probably no real need for nerve cells to control the action of the ciliated bands in Ctenophores: the calcium waves found in astrocytes would probably be sufficient, given the overall slowness of this type of swimming. However, if they evolved from an ancestor that used nerve cells to control the action of cilia at the end of long tentacles extending from around the mouth, this feature wouldn't have been given up, it would have simply been adapted for the need.
Thus, in this model, the common ancestor of all the modern Eumetazoans had already invented nerve cells, although their most essential use was for controlling ciliary activity. At some time after the Ctenophore lineage(s) branched off, one group continued the development of its nervous system for more active control of its movement, perhaps extending myotubes into the oral tentacles, and developing a suite of sophisticated movement with the aboral tentacles for complex swimming movements. Little real development of the physical form would have been necessary, in fact little development of the nervous system is really necessary: the common ancestor of the Cnidarians and Eumetazoans might well have pretty much resembled the earlier common ancestor, the Cnidarian and Bilaterian lineages may just have branched later than the Ctenophores.
What about the progression to the Bilaterians? Well, among the things that must have happened is a shuffling of developmental cues, as is shown, for instance, by the fact that the blastopore in developing bilaterians forms at the opposite end of the the egg from that of the others, based on the position of the polar body. There are many other differences in how the bilaterians use the developmental genes they inherited from the common ancestors (as shown by similarities and differences with the Cnidarians). But this is pretty much out of the scope of this post, so we'll defer further discussion for a later time.
RICHARDS, G., SIMIONATO, E., PERRON, M., ADAMSKA, M., VERVOORT, M., & DEGNAN, B.(2008). Sponge Genes Provide New Insight into the Evolutionary Origin of the Neurogenic Circuit Current Biology, 18 (15), 1156-1161 DOI: 10.1016/j.cub.2008.06.074
Links: Many of these aren't called out in the text, and too many are behind paywalls. Use the back key if you came via clicking a footnote.
1. Conservation of Brachyury, Mef2, and Snail in the Myogenic Lineage of Jellyfish: A Connection to the Mesoderm of Bilateria
2. Intracellular Fate Mapping in a Basal Metazoan, the Ctenophore Mnemiopsis leidyi, Reveals the Origins of Mesoderm and the Existence of Indeterminate Cell Lineages
3. The Development of Radial and Biradial Symmetry: The Evolution of Bilaterality
4. Invertebrate Muscles: Muscle Specific Genes and Proteins
5. The Chemical Mechanism of Myosin-I: Implications for Actin-based Motility and the Evolution of the Myosin Family of Motor Proteins
6. Evolution of the Heart from Bacteria to Man paywall
7. Amphink2-tin, an amphioxus homeobox gene expressed in myocardial progenitors: insights into evolution of the vertebrate heart
8. Gene Regulatory Networks in the Evolution and Development of the Heart paywall
9. Phylogenetic relationship of muscle tissues deduced from superimposition of gene trees
10. Anteroposterior Patterning in Hemichordates and the Origins of the Chordate Nervous System
11. Centralization of the Deuterostome Nervous System Predates Chordates paywall
12. The evolution of nervous system centralization
13. Cnidarians and the evolutionary origin of the nervous system
14. Sponge Genes Provide New Insight into the Evolutionary Origin of the Neurogenic Circuit paywall
15. Larval and adult brains paywall
16. The neuroendocrine system of invertebrates: a developmental and evolutionary perspective
17. Central neural circuitry in the jellyfish Aglantha: a model 'simple nervous system'
18. A Post-Synaptic Scaffold at the Origin of the Animal Kingdom
19. Early evolution of animal cell signaling and adhesion genes
20. The left-right axis in the mouse: from origin to morphology
21. A transcriptional regulatory cascade that controls left/right asymmetry in chemosensory neurons of C. elegans