File: Anaheim 29May02.wpd May 29, 2002. For SICB Div Neurobiol. Satterlie Symp. on Neurobiol., Anaheim, 1/ 4/02

Grades in neural complexity: how large is the span?(1)

Theodore Holmes Bullock

Neurobiology Unit, Scripps Institution of Oceanography and

Department of Neurosciences, University of California San Diego, La Jolla, CA 92093-0240

Address for correspondence

T.H. Bullock

Dept. of Neurosciences 0240

University of California, San Diego

9500 Gilman Dr.

La Jolla, CA 92093-0240


ph.: 858-534-3637

fax: 858-534-3919

Synopsis. The span of complexity in brains, between the simplest flatworms and the most advanced mammals is exceedingly great, measured by the number of different anatomical parts, physiological processes, sensory discriminations, and behavioral alternatives in the repertoire. Most evolution of brains has been adaptive radiation within the same grade of complexity. Distinct grades of complexity have appeared a dozen or more times and quite often in the retrograde direction. Advancement has not been inevitable or obviously advantageous in survival value but has happened - long before primates or mammals or vertebrates. Compare cuttlefish and the most advanced gastropods, bees and the best brine shrimp, primates and the most advanced reptiles known - all twigs with common branches. This repeated achievement of evolution has had all too little study in respect of the detailed listing of differences between major taxa of distinct grades of complexity. Connectivity at the level now known for the mammalian cortex is much needed in other classes, with estimates of reciprocity, intrinsic differentiation, dendritic parcellation and afferent and efferent connections, both locally and projecting to other centers, each done quantitatively to permit comparison. Physiological system organization, personality properties of neurons and circuits, proclivities and emergent phenomena at several integrative levels are sketchily known only for parts of a few systems. Examples are given of opportunities for new research that can more adequately characterize grades of brains.


Brains of limpets and abalones are much simpler than brains of garden snails and slugs in histological differentiation. In fact a wide range in the degree of this differentiation is displayed among the very heterogeneous class of gastropods. Yet, if we choose a gastropod with the most elaborate brain, such as a carnivorous land snail, it pales in comparison with any cephalopod brain.

The same can be said of polychaete annelids - extreme heterogeneity among the orders and families within the class and a discontinuous leap between the most differentiated polychaete brain and that of a hymenopteran insect. The same for teleosts - a wide span of elaboration among the great world of bony fishes but a difference in grade between the most elaborate teleost and a crocodilian. We see this over and over, among the two dozen or more phyla of metazoans.

I am lifting up for scrutiny the distinction between two kinds of evolutionary achievements. The vast majority of differences between taxa are microevolutionary, presumably adaptive specializations that manifest the same grade of overall organization among the species, usually also genera and often families and orders of the same class. Even classes and phyla are often at essentially the same grade of overall complexity. We may call this horizontal radiation.

Now and then, however, quite rarely in fact, the second kind of evolutionary achievement demands a different kind of attention. This small minority of achievements comprises larger changes and many traits. It deserves recognition as changes in grade of overall organization. We may speak of vertical grades of complexity.

This is distinct from the usual use of "grades" that does not say anything about degrees of complexity. Some of the changes represent increases in brain differentiation, some have to be considered decreases or retrogressive evolution. To avoid quibbling over smallish differences, I recommend that we pay particular attention, for the present purposes, to comparisons between higher taxa where an obvious difference in grade of complexity occurs - phyla, classes, orders, even some families like the saltation between the great apes and humans. I refer not only to histological differentiation but all neural and behavioral traits.

The great phenomenon of biology to which I am pointing is the rare but repeated appearance of new grades of higher neural complexity. Such advances are not inevitable, linear, frequent or self-evidently adaptive. In a phylogenetic dendrogram, branches and twigs here and there show saltations into a new grade. Some are retrogressive, such as some parasitic taxa that came from neurally more complex free living ancestors or burrowing or sessile forms, like many pelecypods. Some taxa evolved drastic changes in life style, such as Amphibia, or in cellular organization, like the glass sponges, without much increase in neural complexity, if any. If we count all the taxa, most neural evolution, to repeat, is more lateral than vertical.

If we contemplate the grand sweep of the animal kingdom, however, and consider the nerve net in cnidarians, the distributed central nervous system of echinoderms, the simple brain in flatworms, the brains in advanced polychaetes, insects, and cephalopods, in three distinct grades and the vast differences between the relatively advanced hagfish, manta rays, mormyrid fishes, alligators, crows, and humans, the whole span of difference in complexity of neural machinery and behavioral consequence rivals anything else in nature - and deserves its own study.

Surprisingly, this vast increment in both machine and performance is often omitted in general treatments of evolution that dwell more on the how than on the what. The nervous system and its products, both homeostatic and behavioral, have a particularly large role and unique position among the systems of the animal body in respect to evolution. The great span of difference in organizational intricacy means that the number of traits in the nervous system and its discriminations and alternative outputs is peculiarly great. Traits include not only anatomical and physiological but also chemical, pharmacological, and immunological, plus the whole repertoire of acts and states. Any account of animal evolution that underplayed neural organization and behavior, and their enormous degrees of advance, repeatedly in distinct lines, already long before humans, would miss a major aspect of biological evolution. Any account of evolution that shuns all terms such as "advance" and "higher" would overlook the brain and a prime feature of zoology.

Some readers may say "He is dwelling on the obvious and generally accepted idea that more advanced metazoan groups evolved from simpler ancestors." To them I have to reply,"Thank you; I wish you were right that it is obvious and generally accepted." Some readers will probably say, "He is subjectively imposing human values".

The repeated appearance of higher grades was true long before primates, before mammals, or any vertebrates. Higher refers, not to value, but to complexity. This can be defined simply and objectively as the number of distinguishable kinds of components. Anatomically these are distinct kinds of parts, physiologically distinct kinds of processes, and behaviorally distinct kinds of perceptions, responses and states. This avoids issues of anthropocentric, ethical or moral value. The quantitative estimation of complexity, while difficult, is quite feasible by applying the same criteria to different taxa and starting with some fraction of the whole, such as cell types, or nuclear masses, or neural connections, dance movements, communicative signals, or phases in procuring, preparing and consuming different kinds of foods.

All this is preamble to my main point, today. Brain evolution means different things to different folks. To me it is code for a research agenda, or better, a prime opportunity for major discovery. My reading of the literature, or my sampling of the deluge, leads me to conclude - and this is the main point - that there is too little comparison, especially comparison across major taxa and comparison that doesn't stop with establishing commonalities, homologies or analogies, but goes on to discern differences, quantitative as well as qualitative, between animals of quite distinct grades of complexity. Too little comparison means in view of the biological significance of the general question: what traits actually comprise the differences between taxa of quite different grades of complexity? Each species has traits that are very general, traits that are lateral radiations and vertical traits that differ from lower grades, defining the grade of that species. The challenge and fun is distinguishing among the three. Curiously, there exists a positive pressure against comparison of distant taxa! I have seen numbers of thorough descriptions of a species exemplifying its group with some comparison of previous literature on closely related forms but no comparison of other classes or orders.

What follows illustrates with subsidiaries to that general question. Are the homologous nuclei in different classes of vertebrates alike or not, for example with respect to the proportion of intrinsic cells to projection cells? Are there more kinds of intrinsic cells in mammals than in reptiles? Is there evolution of amacrine cells? Or of non-spiking neurons? Or recurrent collaterals? Is there evolution of dendritic spines or modularity or columnar organization? Are the projections from one center to another reciprocal to the same degree in the taxa being compared? In comparable regions how do the types and numbers and volumes of glia and their processes compare? These features may affect not only the packing density, but the extraneuronal current paths and the lability of the extracellular compartment with physiological state .

How do the groups compare in respect to the numerical measures that Braitenberg and Schüz (1998) have used in describing the mammalian isocortex - such as number of neurons that converge on an average target cell and the number of neurons to which an average one projects, the densities of pyramidal, stellate and Martinotti neurons, the relative densities of axons, dendrites, and synapses of these three main cell types and the proportions of inhibitory to excitatory cells?

The number of "visual areas" in the cortex of cats seems to be fewer than in monkeys. This number, as well as the spatial resolution of the retinotopic maps and of the somatotopic and cochleotopic maps, measured by comparable technics are still poorly known across taxa. The same can be said about degrees and extent of lateral inhibition or other forms of surround modulation.

A quantitative measure that should be insightful, after enumerating the distinguishable cortical areas, including subdivisions and zones, is the number of first order, direct connections between a list of areas of the cortex. For a certain list of 32 visual and visual-association areas in the macaque monkey Felleman and Van Essen (1991) found 305 connections to and from source and target areas. Allowing for the pairs for which no data are known, they estimate about 40% of the possible connections are made. The same fraction was found for 62 identified pathways among 13 somatosensory and motor areas. Data in one example tabulated by Stephan et al. at http// on the internet, in 2002 (presumably one of some 52 tables referred to in Stephan et al. 2001), based on rigorous culling from the literature show 98 out of a possible 306 anatomical connections among 18 somatosensory and visual areas, and 144 functional connections among the same areas. The number of recognizable areas (and subdivisions and zones and laminae) is one measure of complexity; the number of connections is another and the fraction of those which are reciprocal is a third Are these numbers only slightly or are they greatly different in insectivores or marsupials? Are comparable sensory and motor maps of interest in the pallium in reptiles, in birds, in amphibians or fish - which lack an isocortex? How developed are the horizontal or tangential connections between parts of the pallium in these lower vertebrates? Do these differences correlate with the length of the list of discriminable behavior states?

A seemingly simple comparison like the number of telencephalic or of diencephalic nuclei in several classes of vertebrates can involve tough preliminary questions, such as what is the boundary and whether the criteria used for subdividing cell groups in the various taxa are equivalent? But the significance of such numbers makes the effort well worth while. Wicht and Northcutt (1992, p. 61) document the increase in differentiation in forebrain cell groups from lampreys to mammals, in 15 taxa, gleaned from various authors. With only one mammal, one bird and no advanced teleost, an obvious opportunity lies here. (See further in Northcutt, this volume.)

In species that grow indefinitely, some nuclei continue to add cells as the animal grows (retina, tectum, etc.), while other structures plateau at an early age (VIIIth nerve, vestibular nuclei, etc.). The two lists are quite incomplete.

Among the invertebrate phyla classical microscopists noted not only degrees of differentiation in the fibrous neuropile as indications of more advanced neural tissue, but also a variety of cell types. Most invertebrate neurons are unipolar, with a single stem neurite and a less than obvious distinction between axon and dendrites. Forms of branching and terminating can vary widely, like shrubs and trees, helping to define grades of differentiation. Especially elaborate are the optic ganglia of insects and cephalopods marked by laminae with several of the 19 or more cell types plus neuropile sublayers textured by specialized forms of neuronal processes, endings and synapses. A class of small, chromatin-rich, cytoplasm-poor nerve cells called globuli cells usually forms dense masses in the higher levels of the more advanced taxa.

A feature of some invertebrates and many fishes is the trait of "identifiable neurons" - really a class of specific traits. It is curiously distributed among the groups and parts of these animals. I predict more such unique neurons will be found among the vertebrates - which will be a remarkable event, if it happens! But a more likely and pregnant direction which has hardly been tested will be aimed at the incidence of sets of two or three or ten or a hundred or more essentially equivalent, fully redundant neurons. I have attempted to estimate the numbers of such equivalence classes, by all criteria, including incompletely overlapping afferent and efferent fields and distinct "personalities" in any physiological dimension (Bullock, 1980). I concluded that many millions of kinds of neurons characterize humans and that in rats, goldfish, lobsters and Aplysia, in a steeply descending series the numbers are fewer by many orders of magnitude.

Several forms of redundancy have to be sorted out. What is the best wording of a comparison of invertebrates and vertebrates with respect to this feature? I doubt the adequacy of a common approximation such as"vertebrates have much more redundancy" or "invertebrates use single cells for functions mediated by many cells in vertebrates". Are multifunctional neurons and circuits much more employed in one or the other group?

What is the role and meaning of sheer numbers of neurons? The inferior colliculus (IC) of a dolphin is a midlevel auditory center big enough for some bats to hide under. The IC of bats is relatively huge; that of dolphins is absolutely huge. The common opinion seems to be that these two taxa are about equal in auditory sophistication. The cerebellum is nearly the same fraction of the brain in sloths and cats, sheep, manatees and antelopes but many fold bigger in some elasmobranch species than in others. The number of neurons in the cerebellum must correlate with some aspect of life style other than simple motor behavior, presumably something to do with comparing sensory input with learned expectations.

The message of this diatribe is that we need more details, more descriptive comparisons among diverse species. We have to expect that when we obtain such facts, we are not guaranteed to understand their significance or to see how they work!

I have elsewhere opined that the major leaps in grade of brain complexity are not primarily attributable to brain size or numbers of neurons - in spite of the implication of some modelers that given enough identical units, anything our brains can do is possible. But it is more likely that real saltations in complexity depend at least largely upon qualitative novelties. Examples of novelties are glomerular architecture, bird's nest, bottle brush and other forms of endings and synapses, granule cells, connections restricted to part of the dendritic tree, inhibitory interneurons inserted into neural circuits, dynamic properties of rebound, facilitation, and many other ways by which integration, weighting and timing changes can occur. Years ago I listed about 50 such dimensions and dependencies by which input can be evaluated before it turns into output. Since then, many more have been published, for example neuromodulator control of electrical synapses. These variables add to complexity in a compounded way, multiplying the number of kinds of neurons and of codes. Seeking for such novelties and for the emergent behaviors of circuits, assemblies and the whole animal, forms one of the opportune fronts in our agenda.

A particularly weak area in both findings and efforts, I am embarrassed to say, as a physiologist, is the comparison of physiological processes across higher taxa. We lack a list of processes and properties that account for the performance of the central nervous system, above the level of nerve impulses and synapses. This makes it impossible to comprehend or encompass CNS operation at higher levels by addressing a logical list, as comparative physiologists of the respiratory or cardiovascular and other systems can do.

As an example of one physiological domain I will refer briefly to the ongoing, apparently spontaneous electrical activity of neurons and arrays of them. Discovered in the late 1920s, this form of continual, background activity, in the absence of apparent sensory input, was a minor revolution. Curiously, it has been little studied as a biological property. The meager work done so far tells us that such activity is general among phyla, but with conspicuous differences between some taxa and others and between some parts of the brain and others.

All creatures so far tested show a large component of the overall electrical activity that seems to be stochastic or random, having some energy at all possible frequencies in the Fourier transform, at least up to several kHz and down to small fractions of one Hz. All animals from jellyfish to humans also have all-or-none, ca. 1 ms spikes representing nerve impulses, in some but not all of their neurons, plus graded events representing synaptic and other slower events and now and then one or a few simultaneous oscillatory, quasisinusoidal rhythms lasting from a fraction of a second to tens of seconds.

Are there general differences between the classes of vertebrates, or of molluscs or arthropods, for example, in the form of the spectrum of electrical activity of cells or of assemblies of cells - that is, the amplitude of components of different frequency? Tentatively we can say no to the activity of single cells and yes to that of assemblies (Bullock, 1983, 1984, Bullock and Baar 1988). But this preliminary statement needs underpinning or correction. For example, we need to find out whether there are differences in the prevalence of oscillations over and above the background stochastic or noise-like population activity. We should find out whether there are differences in the frequency of oscillations and their interdependency. And particularly, it would be significant to compare the degree of synchrony, both of slow and of fast electrical events among the phyla and classes. My preliminary efforts to do this have suggested an important evolution from virtually none (Aplysia, Bullock et al. 1995b) to profound levels of synchrony, especially in sleep, seizures and certain cognitive tasks.

In humans and some laboratory mammals much interest is now focused on so-called gamma waves (ca. 25-70 Hz) as candidates for a special role in cognitive processing. We have elsewhere reviewed evidence that gamma rhythms are in fact widespread in animal taxa, and widespread in the brain states and brain levels that emit them (Bullock, 1992; Bullock and Achimowicz, 1994 ). Further work, however, may well reveal the conditions and places where functional roles for these rhythms and others in different frequency bands can be defined. Some bands have long been known to differ among mammalian laboratory species, but not enough is known to hypothesize a biological meaning of these differences.

Multichannel recording within the brain shows microstructure in time and space; electrodes only millimeters or fractions of one millimeter apart generally see activity that fluctuates in coherence by the second and by the millimeter from high to low (Bullock et al., 1995a, b, c, 1997). Each such time series has quantifiable features that contribute to the list defining complexity - such as bicoherence (nonlinear higher moments in quadratic phase coupling), mutual information, dimensionality, independent components and others not yet assessed. Once more, the challenge is not only to discern the nature of the nonrandom characters but to discover the code or what information is conveyed by the signs and signals.

Besides the ongoing, spontaneous brain activity, signs and signals are evoked by stimulation or state changes. We should, therefore, compare in various taxa the form, dynamics and dependencies of responses to a variety of physiological stimuli, including ethologically significant stimuli. Are there differences among classes in the incidence and dynamics of post-tetanic potentiation or in readiness potentials that precede voluntary movements or in subclasses of cognitive waves? History-dependent processes include others such as kindling or seizures, and they may also have evolved. So far, the few forays that have been made, especially in sensory evoked potentials, have turned up specializations related to behavior. Bats and dolphins show elegant physiological specializations in the auditory sphere. Yellow-finned tuna do not fuse light flash-evoked potentials below 90 Hz whereas we fuse at about 60 Hz and sloths at less than 10 Hz. So-called cognitive waves accompany and distinguish subtypes of attention, novelty, expectation and tasks of graded levels of difficulty. We have barely begun to look at these properties in nonmammalian species. This should prove to be a pregnant area for evolutionary studies, if ingenuity is adequate to design the stimulus paradigms.

The simple message of this chapter is that we have relatively neglected major consequences of brain evolution and should aim research at uncovering the anatomical and physiological differences between species, especially those far enough apart to be at distinct grades of brain complexity. The natural emphasis on discovering commonalities, homologies and analogies that simplify has distracted us from the main result of evolution, which is to create differences. We have particularly neglected to focus attention on the differences between taxa at distinct grades, to learn what changes actually accompany and probably account for the striking advances in brain complexity such as we see in groups like the polyclad flatworms, some polychaete annelids, some cephalopods, some lower vertebrates, some higher teleosts, reptiles and mammals. Each of these, and others I have passed over, possesses a brain markedly more complex than sister groups from which they may have diverged. Grist for quantitative models will surely come from such research and, I am confident, surprises in anatomy and physiology. Surely, more puzzles than answers will turn up, but at least we could say we are not overlooking the vast span of increased complexity among the phyla and classes of metazoans.


Braitenberg, V. and A. Schüz 1998. Cortex : statistics and geometry of neuronal connectivity. : Springer, Berlin, New York 2nd edition

Bullock, T. H. 1980. Reassessment of neural connectivity and its specification. In H. M. Pinsker and W. D. Willis Jr. (eds.), Information Processing in the Nervous System, pp. 199-220. Raven Press, New York.

Bullock, T. H. 1983. Electrical signs of activity in assemblies of neurons: compound field potentials as objects of study in their own right. Acta Morphol. Acad. Sci. Hung. 31:39-62.

Bullock, T.H. 1984. Ongoing compound field potentials from octopus brain are labile and vertebrate-like. Electroencephalogr. Clin. Neurophysiol. 57:473-483.

Bullock, T.H. and E. Baar. 1988. Comparison of ongoing compound field potentials in the brains of invertebrates and vertebrates. Brain Res. Rev. 13:57-75.

Bullock, T.H. 1992. Introduction to induced rhythms: a widespread, heterogeneous class of oscillations. In E. Baar and T.H. Bullock (eds.), Induced Rhythms in the Brain, pp 1-26. Birkhäuser, Boston.

Bullock, T.H. and J. Z. Achimowicz.1994. A comparative survey of oscillatory brain activity, especially gamma-band rhythms. In C. Pantev, T. Elbert and B. Lütkenhöner (eds.), Oscillatory Event-Related Brain Dynamics, (NATO A: Life Sciences series, vol. 271), pp. 11-26. Plenum Press, New York.

Bullock, T. H., McClune M. C., Achimowicz, J. Z., Iragui-Madoz, V. J., Duckrow, R. B. and S. S. Spencer. 1995a. Temporal fluctuations in coherence of brain waves. Proc. Natl. Acad. Sci. U.S.A. 92:11568-11572.

Bullock, T. H., McClune, M. C., Achimowicz, J. Z., Iragui-Madoz V. J., Duckrow, R. B. and S. S. Spencer. 1995b. EEG coherence has structure in the millimeter domain: subdural and hippocampal recordings from epileptic patients. Electroencephalogr. Clin. Neurophysiol. 95:161-177.

Bullock, T. H., McClune, M. C. and J. Z. Achimowicz. 1995c. EEG coherence as a measure of synchrony shows fine structure in space and time. Abstr. of the Fourth IBRO World Cong. of Neurosci., Kyoto, Japan. p 494

Bullock, T. H., Achimowicz, J. Z., Duckrow, R. B., Spencer, S. S. and V. J. Iragui-Madoz.1997. Bicoherence of intracranial EEG in sleep, wakefulness and seizures. Electroencephalogr. Clin. Neurophysiol. 103:661-678.

Felleman, D J; and D.C. Van Essen 1991 Distributed hierarchical processing in the primate cerebral cortex. Cerebral Cortex, 1 :1-47

Stephan, K. E., Kamper, L., Bozkurt, A., Burns, G. A. P. C., Young, M. P., and R. Koetter 2001 Advanced database methodology for the Collation of Connectivity data on the Macaque brain (CoCoMac). Phil. Trans. Roy. Soc., London B 356 (1412): 1159-1186.

Wicht, W. and R.G. Northcutt. 1992. The forebrain of the Pacific Hagfish: a cladistic reconstruction of the ancestral craniate forebrain. Brain Behav. Evol. 40:25-64.

1. From the Symposium on .....