The gradual evolution of embryological development. The evolution of the mechanism of embryological development is responsible for the most spectacular punctuation in the whole course of evolution, the Cambrian revolution at the very beginning of the fossil record some 600 million years ago, when many different species of multicellular animals suddenly showed up. Only when animals evolved hard parts, like shells, did fossils form, and the same mechanism that constructs such hard parts also constructs the nervous system.

Tracing the actual evolution of the mechanism of embryological development on earth should settle any doubts that may remain about the inevitability of the evolution of multicellular animals, because it shows concretely how such a mechanism lies within the range of variations on animal-like protists.

This rather technical history, however, serves an additional function in this ontological argument, because it leads to an explanation of the difference between proterostomes and deuterostomes (that is, between most invertebrate development and development in chordates, such as vertebrates and mammals). That will account for the radical difference between two groups of telesensory animals (invertebrates and vertebrates), and the nature of the difference between them will explain why proterostomes do not evolve to the third and fourth animal stages of evolution.

The origin of the mechanism of embryological development. Clues about the evolution of animals can be found by comparing how the various kinds of animals develop. The first steps in embryological development are least likely to have changed during evolution, for whatever the mechanism, variations in it would probably throw the whole process off track. Evolution occurs mostly by adding new structures to the process. And we can see how animals with nervous system began by identifying the first form of development. Let us compare the three basic forms of development.

Gastrulation. More complex animals, both proterostomes and deuterostomes, develop by a process of gastrulation. The fertilized egg cell is an unusually large cell, and asexual division produces many daughter cells (cleavage). They secrete a fluid in their midst and arrange themselves in a layer, enclosing it like the surface of a sphere. They are the blastula, and the cell-free area in their midst is the blastocoel.

Gastrulation follows the formation of the blastula. Its function is to establish a second layer of cells beneath the first. The cells destined to become the second layer, or endoderm, are one hemisphere of the blastula. With more energy-rich yolk from the cytoplasm of the egg, they are larger than the cells in the other hemisphere, which becomes the ectoderm. The second layer forms by a folding inward (invagination) that never breaks the surface of the sphere: At about the margin between the two hemispheres, cells on the surface of the blastula bend inward and start moving into the blastocoel, aligning themselves along the underside of the sphere's surface. Ultimately, about half the surface of the original sphere turns itself inside out to form the second layer of cells on the underside of the other half of the original sphere. But the resulting two-layered surface is not a hemisphere: The missing hemisphere shrinks to the size of a hole in the surface of a new, yet incomplete sphere. The hole, through which the endoderm has disappeared, is the blastopore, and the inner chamber, now lined with endoderm, is the archenteron. Finally, a second opening is made through the two layers of cells, forming a digestive tract between it and the blastopore. Differences that begin at the this point distinguish proterostomes from deuterostomes.^[1] (See diagram of Development in gastrulating animal.)

Planuloid development. Animals with nervous system could not have begun with gastrulation. The process involves such a complex coordination of the behaviors of so many different cells that it could not possibly have been the first step by which protists organized themselves multicellularly. Development occurs in another way in two classes of animals, flatworms and coelentrates. The two layers of cells are produced by a process, which I will call planuloid development.

A blastula still forms after the cleavage of the egg, but the second layer of cells is formed by the inward migration of cells, usually from all parts of the surface of the sphere. They clump together at the center of the sphere, separating themselves from the outer layer of the blastula, and then the outer surface of the inner clump makes contact with the outer sphere, or ectoderm. This is the planula. (See diagram of Planuloid development) The endoderm is formed by a split that starts at the center of the solid clump of cells and continues outward in one direction toward the ectoderm, where an opening is made for the mouth. The result is a gastrovascular cavity, rather than a digestive tract, for there is no second opening. The one opening is both mouth and anus. The cavity not only digests food but also circulates it to all parts of the body.

In coelentrates, such as the hydra, the gastrovascular sac develops long, hollow tentacles around the mouth which are used to paralyze prey and pull them into the mouth. But flatworms have a bilateral symmetry. The mouth opens from the middle of the body toward the ground, and instead of a radial symmetry around it, one of the directions becomes the longitudinal axis of the body, and the perpendicular direction has a bilateral symmetry.

Ectolecithal development. Coelentrates and flatworms, as we mentioned, are the simplest kinds  of animals. But even their planuloid development is too complex to have been a variation on protists. After the solid core of cells has formed inside the blastula, and before the inner split has opened as a gastrovascular cavity, cilia grow from the ectoderm and the planula swims around for a while. But this early, pre-sexual stage of its life is surely not the first form that animals took, because it has no way of acquiring energy. It is still living from the energy reserves of the egg. The clue to the origin of multicellular animals can be found in a third kind of development which occurs only in certain kinds of flatworms.

The eggs of these flatworms develop in a medium of energy-rich yolk supplied from the outside by nurse cells (ectolecithal development). From the cells produced by the cleavage of the egg cell, first, the mouth (pharynx) forms. Attached to these cells, another group of cells forms a sac to receive what the mouth swallows; they are the primitive endoderm, or gut. The ectoderm is formed by a third group of cells that constructs an outer layer of cells in the yolk by stretching themselves out from the mouth through the yolk around the gut. The embryonic mouth then swallows the remaining yolk, which fills the gut, or gastrovascular cavity.^[2] (See diagram of Ectolecithal development.)

Ectolecithal development is unique. It occurs in just a few species of flatworms, but in no other animals. It would be difficult to explain why there is such an unusual kind of development unless it is the earliest condition. And it suggests how animals may have evolved.

The origin of ectolecithal development. The process of ectolecithal development requires only three kinds of cells, each with a distinctive kind of behavior. They can all behave at the same time, since no movement relative to one another in involved. Thus, this random variation would have required only a few cells (maybe as few as twelve). An amoeba's cycle of sexual reproduction could well have produced that many offspring at once. Assuming that it happened on some mechanism for determining its offspring to three different kinds of behavior, this lineage could have begun, if each kind of cell remained attached to one another and to at least one of the other kinds of cell in an asymmetrical way. By synchronizing the amoeboid contractions of the mouth cells, the group could trap and surround animal-like protists for digestion. The energy supplied by their ingested bodies would select a reproductive cycle of this kind, if it were possible at all.

A new source of energy is probably the only way that such an unlikely process could have been selected. It was probably not to consume plants, since a multicellular body is no better suited than single-celled animals. Nor is a multicellular body needed to ingest prokaryotes. They are on a lower level of biological organization and animal-like protists already feed on them. Thus, multicellular animals were probably selected in the first place as carnivores, which preyed upon larger animal-like protists that, before them, were at the top of the food chain. Once the means of constructing multicellular bodies had evolved, some variations would, of course, eventually evolve ways of using it to acquire energy by ingesting plants and prokaryotes.

This origin of animals would account for the ectolecithal development in primitive flatworms. It now occurs in eggs, where nurse cells supply the yolk, but at first, the offspring may have been constructed in the gut using the energy-rich molecules available there, making the gastrovascular cavity do double service as a womb. The biological behavior guidance system could select the behavior required for gestation as an extra stage in the behavior of reproduction by turning off the behavior involved in digestion. Amoeboid cells already have the capacity to contract and relax depending on the kind of object they encounter, and thus, the same cells could play the roles of both muscles and neurons in guiding the organism's behavior as a whole. Swallowing, for example, could be a contraction generated by a signal exchanged among the mouth cells when any member made contact with an animal-like protist. It could even have been an electromagnetic signal, like neurons, since cells of many kinds are responsive to them. Neurons would differentiate as variations of ectodermal cells that signalled cells to refrain from feeding in hazardous situations, for that would enable them to complete their reproductive cycles while others did not. That would explain why neurons always differentiate from the ectoderm, never from cells of any other kind.

Three basic kinds of cells. This explanation of the origin of animals with nervous systems depends on the capacity of offspring of a fertilized cell to differentiate into three different kinds of cells. The mechanism of differentiation lies in how the egg cell is prepared. By leaving different messenger proteins in different parts of the cortex or cytoplasm of the egg cell, cleavage gives different daughter cells different messenger proteins. The nucleus must enable the chromosomes to affect different parts of the cytoplasm in different ways, or else it could not set up the process involved in reproduction. Thus, the preparation of the egg cell would merely elaborate an existing mechanism. The messenger proteins would be keys for opening up certain special chromosomes, in addition to those used for ordinary housekeeping activities, thereby generating special kinds of behavior in them. And assuming the chromosomes are also opened up differently at subsequent stages of development, the differences would be permanent.

This way of determining the most basic differences among cells, which are responsible for the whole process of development, would be an elaboration of the protein mechanisms for handling the chromosomes, sorting them out, and forming a nucleus that were introduced with the evolution of eukaryotes. The mechanism by which sex is determined makes it clear that the nucleus handles different chromosomes in different ways.^[3] Since this mechanism is so basic to the operation of the eukaryotic cell, evolutionary change would build on it by adding new steps or new kinds of behaviors to each of the basic cell kinds at each step. Nor could it change easily in the course of evolution.

Only three kinds of cells are needed to account for the process of development. They would differ from one another in how cells of the same kind remain attached to one another and how they relate to the other two kinds of cells. Cells that form the ectoderm must attach to one another on four sides as the 2-D surface of a sphere while orienting themselves to the mouth cells located among them. The inner gut cells must also attach to the mouth cells, but they would form themselves into a 3-D ball whose outer surface eventually attaches to the ectoderm. In addition to attaching to the other two kinds, the mouth cells would form an opening in the blastula. Behavior like this would require each cell to generate different kinds of behavior towards other cells on each side, but that is within the range of what eukaryotic cells can do, as the preparation of the egg cell and the asymmetrical structures of animal-like protists demonstrate.

A second kind of behavior is also required. Once they are attached to one another, the interactions of the different kinds of cells must set up the gradients of messenger molecules in which cells at different locations can be determined to different kinds of behavior, as in multicellular plants. It is not known how these gradients work, but we can think about what they must be doing by supposing that a gradient involves a pair of messenger proteins with one kind being concentrated at one extreme along its direction and the other being concentrated at the opposite extreme in that direction; their relative strength for any cell would indicate its location. One gradient would extend between the mouth cells and the other extreme of the blastula (oral and aboral poles). The mouth cells surely have a gradient around the opening that they form, which would give the blastula a gradient perpendicular to the first. And beside orienting themselves in the blastula and picking up its gradients, the inner-gut cells must have a gradient of its own that determines the center at which the split forms for a digestive cavity.

The evolution of other forms of development. The small scale and simple behavior required of such proto-flatworms, together with our explanation of the source of energy for development, makes it plausible that animals with nervous system originated from an amoeba whose offspring after mating included these three different kinds of cells. And it is easy to see how planuloid development and gastrulation could have evolved from it, thereby suggesting the basic evolutionary relationships among animals with nervous systems.

A simple rearrangement of the three kinds of behavior found in the primitive, ectolecithal development accounts for planuloid development: After cleavage, the ectodermal cells go to work, first, constructing the blastula in which the mouth cells have a location because of where the division of the egg cell has left them.^[4] The gradients in the blastula are oriented around the mouth, and once the inner gut cells have migrated inward, formed a solid core, and made contact with the ectoderm, the split forming in their midst could be made to open up toward the mouth's location on the blastula by the gradients in the ectoderm. Thus, all the steps in planuloid development can be explained as a result of the interactions among the three original kinds of cells.

Gastrulation can likewise be explained as a modification of planuloid development in which one of the original three kinds of cells drops out of the process. Once again, the blastula is the result of the ectoderm-forming cells, and mouth-cells have a special location in it, establishing its gradients. But instead of cells migrating inward to form the inner core of cells, the second layer of cells is formed as a result of the behavior of the mouth-cells. That is, invagination can be explained as a variation on the behavior that makes an opening in the blastula and swallows. It is as if one half of the hemisphere swallowed the other half, turning the other half inside out to form a second layer of cells beneath the cells of the first half. Gastrulation results in a digestive tract with two ends. This could also be explained as a elaboration of the behavior of the mouth cells. After leading the invagination involved in gastrulation, the mouth cells eventually make contact with a special location on the blastula and form an opening by interacting with the ectodermal cells there. The clump-forming cells that supplied the endoderm for the planuloid animal have no role in forming the two layers of cells.

Embryological development is our best clue to the branchings of kinds of animals in evolution, and the implications of this account of the origin of multicellular animals with nervous systems can be made explicit by outlining the evolutionary tree to which it leads. We have found that the flatworms were the first multicellular animals. The original ectolecithal development evolved into the planuloid development of flatworms. And coelentrates and gastrulating animals both evolved from planuloid flatworms.

Gastrulating animals are of two kinds, proterostomes or deuterostomes, depending on the fate of the blastopore, the opening through which the mouth cells invaginate. In proterostomes, the blastopore becomes the mouth and the second opening becomes the anus, whereas in deuterostomes, the blastopore becomes and the anus and the second opening becomes the mouth. This is either an early branching among gastrulating animals or each evolved independently from the planuloid flatworm. And there is a second major difference between them, concerning how they form a mesodermal coelom, which suggests the fate of the third kind of cell found in the most primitive development that seemed to drop out with the evolution of gastrulation.

A mesodermal coelom is a cavity formed within a layer of cells that lies between the ectoderm and endoderm. It is the source of internal organs, including muscles, gonads, the circulatory system, excretory system, and the like. Only the more advanced proterostomes form a coelom at all. (Proboscis worms and round worms have no true coelom, although they do have a digestive tract.)

In proterostomes, it is constructed from a special pair of cells which were determined by how the cleavage of the egg divides up the egg cortex and cytoplasm. As a result of how the blastula is formed and gastrulation takes place, these cells wind up near the blastopore, which has become the mouth.^[5] After gastrulation is complete and the anus is formed, one cell on each side of the blastopore migrates into the area between the ectoderm and endoderm, multiplies into a solid mass of cells, and a mesodermal coelom is formed by a split that starts in their midst. The process is called "schitzocoelous." The mesodermal cells are probably descendants of the third kind of cell from the primitive, ectolecithal development, which formed a solid clump of cells in which a split can develop. With their backs facing either the ectoderm or the endoderm, the fronts of these cells face a fluid-filled space formed by their splitting, and a 2-D gradient set up in it probably determines the cells at different locations to become different kinds of organs. The 2-D gradient in the ectoderm and endoderm enable them to locate themselves in the body. (See diagram of Schitzocoelous.)

Deuterostomes take a different tack, and as we shall see, it makes all the difference in their evolutionary destiny. The third kind of cell drops out again as gastrulation evolves from planuloid development. But the deuterostome forms a mesodermal coelom by enterocoelous, a continuation of the behavior of the mouth-cells that accounts for gastrulation. That is, after the endoderm invaginates, but before the mouth-cells make a second opening in the ectoderm (the animal's mouth, in this case), a second invagination takes place, this time from the endoderm into the blastocoel, the volume inside the original blastula that is now located between the endoderm and ectoderm.^[6] No cavity is formed by splitting in a solid core; deuterostomes make no use of the innerªgut cells from the primitive flatworm development. (See diagram of Enterocoelous.)

If this is indeed how animals with nervous systems evolved, then the primitive flatworm is their progenitor and all their kinds can be classified according to their modifications of its kind of embryological development. There are three branches. The simplest coelentrates have a larval stage that involves the planula. The other coelentrates and ctenophora have evolved from them, although some more advanced kinds have evolved a form of gastrulation on their own. Thus, these radial animals are one branch. On the second branch are proterostomes, and on the third are deuterostomes. Differences in how coeloms are formed accompany this defining contrast. As we shall see, this difference in embryological development explains the surprisingly radical differences between telesensory proterostomes and deuterostomes.

The proterostome nervous system development. Neurons always develop from the ectoderm. From the beginning, on our speculation, the chromosomes containing the genes that transform basic cells into neurons are opened up by the protein keys in the ectoderm-forming cells. But some additional factor must turn those genes on, since not all the cells of the ectoderm become neurons. The mechanism is not fully understood, but it apparently depends on gradients in the ectoderm. The cells becoming neurons can be evenly distributed over a region^[7] or determined by singularities in the gradient. In the proterostome, gradients in the ectoderm (or the effect of special mesodermal cells on them) are responsible for cells becoming neurons.

More complex interconnections among neurons in proterostomes are made in structures, called "ganglia," which are located in various places in their bodies and connected by bundles of axons. In each ganglion, cell bodies are arranged on the outer surface, and its core is a neuropil where the processes synapse with one another. Axons of proterostome neurons branch near the cell body. One branch leads into the neuropil, and the other branch joins a nerve connecting to other ganglia or innervating muscles or sensory receptors at the periphery. Ganglia can develop when one or more cells that are determined to become neurons at some location multiply and form a cluster. When the cluster establishes its own gradient, their locations in it can determine neurons to become specific kinds as well as enable them to make quite precise connections with one another in each ganglion or between them.

The ganglia form in the ectoderm, and they must be interconnected to works as an animal behavior guidance system. The gradients of the ectoderm are used to connect neurons with one another as well as with sensory receptors on the input side and muscle cells on the output side. When cells become neurons, they develop processes, dendrites and axons, which extend out from the cell body and eventually form synapses with certain cells that they encounter. The axons are guided to other ganglia, muscles or sensory receptors as they form by following the ectodermal (or endodermal) gradients. Neurons at different locations in the body are thereby interconnected by nerves.

The typical structure of the proterostome nervous system is a set of ganglia in the ectoderm that are connected with one another and the ultimate inputs and outputs throughout the body. That is, the nervous system is laid out in the outer skin of the multicellular animal. The animal's behavior as a whole depends on the equilibrium reached by its entire nervous system, but that equilibrium depends, in turn, on the equilibria reached in each of the ganglia as they interact with one another by way of the nerves connecting them, given the sensory input from the rest of the body. We can see how the basic structure of the proterostome nervous system follows the gradients of the ectoderm by looking at the nervous systems of simpler animals, those with a planuloid development, because the gradients in the proterostome are simply how gastrulation has modified them.

The bodies and nervous systems of flatworms and coelentrates form from a larval stage called the planula, which as we have seen, is produced by the interaction of the three basic kinds of cells. The ectodermal gradients depend on the location of the mouth cells, and when the inner core of cells joins the ectoderm, it picks up the ectodermal gradients, enabling the split that forms in their midst to open up in the direction of the mouth. The mouth is the only opening, and body is a sac formed of two-layers of cells. There is a gradient between its oral and aboral poles and another gradient running around the body perpendicular to it. (See diagram of The planula.)

The body of the hydra, for example, is complete, once tentacles form around the mouth as elongated extensions of this gastrovascular cavity. Its nervous system is just a network of neurons differentiated at random in the gradients of its sac-like body, with a slightly higher concentration around the mouth. They are responsible for reflexes involving longitudinal muscles at the base of ectodermal cells and latitudinal muscles at the base of endodermal cells by which hydra's tentacles wrap around prey that swim by and pull their paralyzed bodies into its mouth. There is a second and maybe a third network of neurons which enable hydra to throw its tentacles out in various directions seeking prey and to somersault to more favorable situations by attaching its tentacles to the ground and throwing its body over them. But it lacks extroªsensory organs (like all coelentrates), getting by with somatosensory input and phototropisms.

The flatworm has essentially the same structure as the hydra, except that one of the gradients that is perpendicular to the oral-aboral axis is a longitudinal gradient for the body. That is, it has a bilateral body with its mouth located near the middle of its ventral side. The bilateral body enables it to move about in space head first, and the head has primitive telesensory organs to guide it: Pigmented eye spots detect the direction of light and the motion of object and lobes on the sides of its head with mechanoreceptors and chemoreceptors detect the direction of water flow and detect certain kinds of molecules in it. Like the hydra, the flatworm has a nerve net throughout its body which coordinates the body with the pharynx as it feeds, but superimposed on it is a second nervous system, which is organized like a ladder along the length of the dorsal side of its body. At least two, and sometime four, six or eight nerve chords run in parallel from head to tail, with regular latitudinal nerves connecting them and innervating the muscles of the body. At the head, there are ganglia where neurons from the sensory receptors intersect with the longitudinal nerves, which trigger or repress the reflexes by which it moves. Both nervous systems are laid out in the gradients of its simple body. The nerve net centered around the mouth generates feeding behavior, whereas the longitudinal, ladder-type nerves centered at the head generates locomotion. The balance between them presumably selects the kind of behavior. (See diagram of Flatworm.)

The basic structure of the proterostome nervous system can be seen as a transformation of the flatworm nervous system that result from the evolution of gastrulation from the flatworm's planuloid development. On our speculation, the second layer of cells is formed by the "mouth-cells," which are located at the oral extreme of the oral-aboral gradient in the blastula. The mouth-cells lead the invagination of one hemisphere of the blastula that forms a second layer of cells, but they do not stop there. The invaginating endoderm does not seek out the aboral pole of the blastula to locate the anus; instead, it seeks out the tail in the longitudinal gradient of the ectoderm. The blastopore, which becomes the mouth, is also shifted from the middle of the ventral side toward the head, shortening the ventral side anterior to the mouth. Thus, the original oral-aboral gradient in the endoderm coincides posterior to the mouth with the longitudinal gradient of the ectoderm. This produces a new framework of gradients in which the nervous system is laid out. The result is a nervous system composed of a pair of nerve chords running along the ventral side of the body with one pair of ganglia connecting them anterior to the mouth and at least another pair of ganglia posterior to the mouth. (See diagram of Proboscis worm.)

If this is how to understand the evolution of gastrulation form the planuloid development of flatworms, we can see how proterostomes are trapped into a nervous system with two pairs of ganglia encircling the mouth. One pair is anterior to the mouth, the other pair posterior, and connections between them encircle the mouth. Reproductive causation has never extricated the nervous system from the mouth in any proterostome. As our speculation suggests, the reason is that the biological behavior guidance system uses gradients in the ectoderm to determine cells as neurons and gastrulation has organized those gradients around the mouth. With bilateral bodies, nerves apparently cannot be differentiated along the length of the body except symmetrically, in pairs. Although the parallel nerves are sometimes fused through most of the body of higher proterostomes, at the head, they still lie on either side of the mouth and are reconnected anterior to it.^[8]

This structure is clearest in the simplest proterostomes, proboscis worms (Nemertina) and round worms (Aschelminthes). In higher proterostomes, such as mollusks (clams and squid), annelids (segmented worms), and arthropods (crustaceans, spiders and insects), the mesodermal coelom gives the body a more complex structure. The mesodermal coelom superimposes a structure on the gradients of the ectoderm and endoderm, which may modify the structure of the nervous system, but it does not enable the nervous system to avoid having two pairs of ganglia encircling the mouth.

The most radical effect of the mesodermal coelom is the segmentation found in annelids and arthropods. It results from a special kind of division of the mesodermal cells. After migrating into the space between ectoderm and endoderm, they divide into many pairs of stem cells. Each pair marks off a segment in the ectoderm, and after forming a solid mass of cells, each clump splits and forms a coelom with a complement of organs. Each segment is roughly equivalent to a gastrula, and the series is chained together, mouth to anus, as a continuous digestive tract. The segmentation they impose on the gradient in the ectoderm affects the structure of the nervous system. In most segments, there are only two main ganglia, which are connected along the ventral side by a pair of (fused) nerves, as if the pair of ganglia anterior to the mouth in each segment were suppressed. But at the head, the anterior segment, where the nervous system is centralized, there are still two pair of ganglia encircling the mouth, with the telesensory organs supplying input to the anterior pair. Thus, segmentation does not enable proterostomes to escape the trap of having at least two pairs of ganglia encircling the mouth. A far more radical change is needed to escape the ectodermal gradients left by gastrulation.

The deuterostome nervous system development. Deuterostome development makes it possible to lay out a centralized nervous system without encircling the mouth. Neurons still differentiate from the ectoderm, but the ectoderm from which they differentiate is a region that has separated from the outer skin and rolled itself up as a neural tube which is contained as a unit inside the body. The structure of the nervous system can be laid out free of the gradients of the gastrula because the neural tube uses gradients of its own to determine neurons to different kinds and connect them with one another. Since the deuterostome's embryological development is what makes this possible, that is where we begin.

Deuterostome development. Like proterostomes, deuterostomes develop by gastrulation, rather than by forming a planula. After cleavage, the daughter cells arrange themselves as the surface of a fluid-filled sphere, called the blastula. A second layer of cells is formed as one hemisphere invaginates and lines itself up along the underside of the remaining spherical surface. The outer layer of cells is the ectoderm; the inner layer of cells is the endoderm; and the hole through which cells have invaginated is the blastopore. Deuterostome development differs from proterostome development because the blastopore becomes the anus and the second opening becomes the mouth -- which is just the opposite of proterostome development. The animal has a longitudinal axis running from mouth to anus, and properly speaking, the endoderm marks the boundary between body and world no less than the ectoderm, for it becomes the digestive tract (and respiratory system in higher chordates) by which the body interacts with other objects in space, albeit under conditions that are more easily controlled. (See diagram of Gastrulation in amphiouxus.)

The body's internal structure depends on the next step in the process, which is also differs radically from the proterostome. Its task is a twofold. On the one hand, it forms a third layer of cells, or mesoderm, between the ectoderm and endoderm, and on the other, it forms a neural tube just beneath the surface of the dorsal ectoderm. The third layer of cells includes not only the coelom, from which muscles, a circulatory system and various other organs develop, but also the notochord, which is the precursor of the spine. There is a symmetry abut these two developments because at the end of gastrulation, the segment of endoderm from which the chordamesoderm develops lies just beneath the neural plate, the section of ectoderm from which the neural tube forms.

The notochord and mesoderm are formed by enterocoelous, a process of out-pouching by the endoderm into the space between the endoderm and ectoderm, which was originally the blastocoel. The cells involved are the "mouth-cells" that led the process of gastrulation by invading the blastocoel in the first place, and thus, enterocoelous is just a continuation of the cell movements that took place in gastrulation. Gastrulation first began on the dorsal side of the blastula, and a section of endoderm that runs the whole length of the gastrula along the dorsal side is involved in the out-pouching. This surface of the endoderm bends in toward the adjacent ectoderm, forming two sections  of cells, both of which break off from the endoderm, allowing the rest of the endoderm to close behind it, thereby restoring the surface of the archenteron. The out-pouching cells become the mesodermal coelom on one side of the body, and a notochord forms where they meet along the midline. In fact, these out-pouchings become a series of segments, for they are divided into segments, like the notochord, which run the length of the body. They eventually spread out to form a complete layer of cells between the ectoderm and the remaining endoderm around the sides and along ventral side of the body.

The nervous system is formed by a similar out-pouching, but this time from the ectoderm. The cells involved are a section of ectoderm, the neural plate, that runs the length of the dorsal side of the gastrula just above the medial chordamesoderm. The neural plate separates from the ectoderm and sinks beneath the ectoderm. As the ectoderm closes behind it, restoring the outer skin of the body, the neural plate rolls itself up into a tube in the space between the ectoderm and chordomesoderm along the entire length of the body.^[9] (See diagram of Neurulation in amphiouxus.)

Ectodermal cells are determined to become part of the neural plate by the effect of the choradmesoderm on them during the process of gastrulation. And both the neural plate and chordamesoderm are segmented longitudinally in the process. That is, by the time the gastrula is formed, cells in different segments along the length of either structure have already been determined to become different kinds of cells. How this is accomplished is not known, but conceivably segmentation is induced by a pulse occurring during gastrulation that marks cells at regular intervals along its length as they invaginate. In any case, segmentation is a product of gastrulation, the process by which the "mouth-cells" form a second layer of cells in deuterostomes, not an effect of the special inner-gut cells from the primitive ectolecithal development, as in proterostomes. What is more, not only are cells located in different segments of the neural tube and mesodermal coeloms already different, but each segment has a 2-D gradient of its own in which cells can be differentiated.

The basic structure of the body at this stage is a set of tubes with the blastopore at one end and a mouth formed at the other. The outer tube is the ectoderm, and the innermost concentric tube is the endoderm. Filling the volume between them are several more tube-like structures: On the dorsal side is the hollow neural tube and just beneath it is the notochord. On each side of them lies a tube divided into a series of hollow areas, the coeloms. (See diagram of Crosssection of embryo of amphioxus.)

Although the segmented neural tube is the kind of basic structure that deuterostome development makes possible in the nervous system, it is of little use in the decentralized nervous systems at the somatosensory stage. There is a somatosensory deuterostome with a segmented neural tube, namely, Amphioxus, a cephalochordate. The sea squirt (a tunicate classified as an urochordate) also has a segmented neural tube at the larval stage. Even the rather primitive acorn worm (a hemichordate) has a neural tube, although it is not segmented.^[10] However, the neural tube is not a universal feature of deuterostomes. It is not found in echinoderms, such as the sea star. The sea star has a solid nerve ring to control the tube feet in each arm, and, like acorn worms, there is also a nerve plexus lying underneath the ectoderm over the entire body. The nerve plexus is an apparently haphazard network of neurons, and the neural tube seems to be just a section of such a plexus that has been rolled up into a tube during development in most kinds of deuterostomes, including all telesensory deuterostomes.^[11]

Independence of the nervous system. The most obvious advantage of separating the ectoderm from which the nervous system will develop from the outer skin is that it extricates the nervous system from the convoluted gradients of the ectoderm, which leaves proterostomes with two pair of ganglia surrounding the mouth. In the same way, it probably also enables reproductive causation to shape the body and its organs in whatever way that is useful in generating behavior without disrupting the nervous system, as it surely does in proterostomes with its ganglia and nerves running throughout the body. It also enables the nervous system to be divided into segments without the complete divisions of the body found in segmented proterostomes, since the segments are imprinted on chordamesoderm and neural tube by gastrulation, not by special mesodermal cells.

The unity that the neural tube gives the nervous system by contrast to the proterostome might well be compared to the contrast between the eukaryote and prokaryote behavior guidance systems. The prokaryote has a single loop of DNA attached to its cell wall, and the protein-DNA interaction that selects the kind of behavior occurs in the same space as the protein interactions that generate its behavior. In eukaryotes, DNA-protein interactions like those that guide prokaryotic behavior are contained within a nucleus, separate from the rest of the cell. Since its behavior is selected by an equilibrium involving an interaction of proteins and multiple chromosomes that is protected from interference by processes in the cytoplasm, eukaryotes can generate behavior on a higher scale than prokaryotes. Although the chordate is not on a higher level of natural organization from proterostomes, the neural tube, like the nucleus, does give the chordate nervous system a unity within the body that also enables it to generate behavior on an entirely new scale. Not only is the chordate nervous system not entangled with the organs generating the animal's behavior, but also when it is centralized at the second stages of zoological evolution, the processing required to select appropriate goals for the object and adjust its behavior to its spatial location can be carried out without interference by the rest of the body.

By separating the ectoderm which can differentiate into neurons from the rest of the body, the chordate has an additional step in setting up its nervous system: It must connect the nervous system to the sensory receptors and muscles in various parts of the body. The proterostome was spared this task by determining cells as neurons near the locations in the body where they would make their connections. But given the gradients that are established by gastrulation, it is a relatively simple task.

Although in Amphioxus, the connections between the neural tube and the body are made by processes from the neurons located in the neural tube, in the higher chordates, sensory cells (and neurons serving the viscera) derive from the neural crest. The neural crest is a group of cells that breaks away from the edge of the neural plate before it forms into a tube and individual cells migrate to locations in the body. In either case, the neurons that connect the neural tube with the body can easily find their targets. Not only does the ectoderm as a whole have a 2-D gradient of messenger molecules, but segments are also marked off in it by the segmentation of the mesodermal coelom. Given how segments are imprinted, mesodermal segments surely correspond to segments of the neural tube (and neural crest), enabling the same gradient mechanisms to be used in making connections at both ends. Neurons may even use the 2-D gradients set up within each mesodermal segment to locate the precise muscle or gland cells to which they will connect or to distribute their connections with internal organs evenly.

Neurons born in 2-D arrays. The advantages that a neural tube affords a centralized nervous system have been described negatively, as overcoming the deficiencies in proterostome nervous systems that derive from using the ectodermal gradients to lay out its structure. But positive advantages are what make higher levels of neurological organization possible. The 2-D gradient in which neurons can be determined to special kinds and make connections with one another are supplied by the neural tube. That is, since neurons are born with labels indicating their segment and 2-D location within that segment, there is no need for neurons to multiply and set up their own gradient to become neurons of specific kinds, as ganglia do.

Each gradient, we have assumed, involves a pair of messenger proteins, with one being concentrated at one extreme along its direction and the other being concentrated at the opposite extreme. The location of any cell in any gradient is indicated by their relative strength, and thus, the location of a cell in the neural tube is indicated by its location in two gradients: one in the longitudinal direction, and one that extends from the ventral midline of the neural tube to the dorsal midline where the two sides of the neural plate closed. Indeed, there are two sets of these 2-D gradients in the neural tube, since each half of the bilateral body would be a mirror image of the other. The third gradient, extending from the central canal of the neural tube to its outer border, is used to guide neurons in migrating and making connections, but not to determine neurons to specific kinds. Furthermore, since the neural tube is divided into segments along its length, the cells in each segment are different from those in other segments. And with each segment having its own 2-D gradient, it is easy to see how complex interconnections can be set up.

Suppose that once the gradients are established in each segment, each cell somehow records the relative strength of each pair of messenger proteins in which it finds itself. Some cells that are differentiated throughout the neural tube remain there; as they elongate, labels from the original gradient are posted, and these processes enable other cells to find their way about. They are called "glial" cells. At an earlier stage of development, cells of the neural tube are determined to become neurons. Neurons always divide at the inner surface of the neural tube bordering on the central canal.^[12] When they send processes or migrate to other locations, they can easily line themselves up in a 2-D array with other cells from the same segment, because they can use the labels from their original locations in the neural tube to seek out the place where the balance between the messenger proteins of cells on one side is greater and on the opposite side less than their own. Or when processes arrive at another segment, they can superimpose themselves in a regular way on the cells already located there by seeking out their own location in the gradients of local cells. In short, entire 2-D arrays of neurons from one segment of the neural tube can be connected with 2-D arrays of neurons in other segments in a regular, topographical way. Indeed, that is what every neuron from a segment would do, unless they were differentiated from one another.^[13]

Vision is a good example of the advantage of the 2-D arrays built into the gradients of the neural tube. The retina of the eye derives in development from a section of the neural tube that migrates as a whole to the ectoderm and determines cells located there to become a lens and an eyeball surrounding it. The neural tube's gradients supply each neuron in the retina with information about its location in a two-dimensional matrix. Thus, their axons are able to keep their order as they project back to the neural tube. The optic nerve is a single cable, often containing millions of axons, and labels from their built-in gradients enable them to keep line up, even when the two optic nerves merge and the axons from half of each retina are sent off in a different direction. And they can easily superimpose themselves in an orderly way on 2-D arrays of neurons from other segments, which are their targets.

The contrast to the proterostome nervous system could not be starker. Cephalopods, such as squid and octopus, have camera-type eyes which are the equal of vertebrates, but their nervous systems is a collection of ganglia that are arranged around the mouth. The eyes feed sensory input to the optic lobes, which connect with other lobes, and behavior is guided by the equilibrium reached by an interaction that depends on the equilibria reached in each of various lobes. Visual input is conveyed by the configuration of firings in 2-D arrays of neurons, but in order to operate a camera-type eye, cephalopods must build up complex 2-D arrays of neurons within special ganglia in order to carry the information. A glance at the basket-weave of the optic nerve in the octopus makes it clear that a good part of the biological behavior guidance system's machinery for differentiating cells is used up simply keeping them straight as they project to the target. That is, a kind of structure that is very near the limit of what the proterostome development can accomplish is supplied by deuterostome development at the outset as the foundation on which further structure is constructed.

Although the somatosensory nervous system gets by with single neurons serving the functions of registering input, generating output and selecting between kinds of behavior, the telesensory nervous system requires at least one 2-D array of cells for vision and a centralized system to make use of it. Proterostomes and the chordates among deuterostomes are both capable of these two levels of natural organization in the nervous system. As I have mentioned, however, the third animal stage of evolution requires complete circuits of such 2-D arrays which can interact with one another as units to register input and generate output. The fourth animal stage involves the interaction between various channels through both input and output circuit of 2-D arrays. The machinery required to connect 2-D arrays of cells from ganglia is so complex that it is not possible for these mechanism to evolve as variations on the proterostome nervous system. But deuterostome development makes possible a neural tube that supplies labels to cells by which entire 2-D arrays of neurons are connected with relative ease. In chordates, therefore, a third level of neurological organization can occur as a variation on the second level, and a fourth level can occur as a variation on the third level. Given the possibility of higher levels of neurological organization, reproductive causation inevitably makes them actual. Thus, our interpretation of embryology gives us an explanation why a third and fourth stage of zoological evolution occurs in deuterostomes, but not in any proterostome.