ArticlePDF Available

Rethinking Mammalian Brain Evolution

July 1990
Integrative and Comparative Biology 30(3)

DOI:10.1093/icb/30.3.629

Authors:

University of California, Berkeley

A critical review of past and current theories of mammalian brain evolution is presented in order to discuss conceptual problems that persist in the field. Problems with the concept of homology arise because of the interaction of cell lineages and axonal connectivity in the determination of structural features of the brain. Focusing on the continuity of information represented by ontogenetic mechanisms as opposed to morphological features avoids many of these problems and suggests homological relationships that otherwise have gone unnoticed. Many apparently progressive trends and parallelisms in mammalian brain evolution turn out to result from the influence of underlying developmental homologies. Confusions about evolutionary advancement, increasing architectonic differentiation, and the evolution of new brain structures result from a failure to appreciate how increasing brain size can bias developmental processes with respect to axonal competition, increased cellular metabolic demands and decreased information processing efficiency. Explanations of the evolution of novel structures and new connectional patterns are criticized for their failure to consider the constraints of neural developmental processes. The correlations between structural neogenesis, functional specialization and size changes in brain evolution are explained by a theory of competitive displacement of neural connections by others during development under the biasing influences of differential allometry, cell death or axon-target affinity changes. The “displacement hypothesis” is used to propose speculative accounts for the differential enlargement and multiplication of cortical areas, the origins of mammalian isocortex, the unusual features of dolphin cortex and the dramatic structural and functional reorganizations that characterize human brain evolution.

Homology relationships. T h e basic homology relationships as outlined by Northcutt (1984). T h e arrows represent descent relationships. T h e vertical axis represents comparison over time o r descent in evolution and the horizontal dimension represents comparison of contemporaneous species. T h e geometric shapes represent similar o r different, ancestral o r derived traits. In the case of parallel homoplasy it is unclear whether the parallel divergence from the ancestral condition is a consequence of internal (homological) o r external (selectional) commonalities. These same relationships can be applied equally t o comparisons between lineages o r to homologous repetition of parts within an organism. (Redrawn from Northcutt, 1984.)

…

Developmental homologies. T h e multiple

…

Hedgehog and tenrec brains as seen from above and the side labeled to show approximate positions of the major sensory and somatomotor fields. Isocortex is indicated in gray in the left hemisphere of each and limbic and olfactory cortex is white in the same hemisphere. Since most of the cortical representation is unknown for the tenrec and only partially known for the hedgehog specific boundaries between areas are not indicated. There is no intent to imply either undifferentiated cortex o r the existence of only single sensory/motor fields. Note the low ratio of isocortex to limbic-olfactory cortex in these brains, especially the small tenrec brain.

…

Allometric patterns of comparative vertebrate brain size and mammalian brain growth. Graph A,

…

+10

Local relationships contributing to cytoarchitectonic allometry are depicted in this figure from Deacon (1990b). T h e top figure depicts the situation in a relatively small brain and the lower figure depicts these same relationships in a slightly larger brain. In small as compared to large brains projection neurons possess short, small diameter axons, with relatively less myelin, smaller cell bodies, lower neuron to glia ratio, higher neuron densities, lower ratios of local circuit neurons to projection neurons, smaller size differences between the smallest and largest neurons, etc.

…

Figures - uploaded by Terrence W Deacon

Content may be subject to copyright.

Content uploaded by Terrence W Deacon

Content may be subject to copyright.

Rethinking Mammalian Brain Evolution

Harvard University, Peabody Museum, Cambridge, Massachusetts

02138

YNOPSIS

critical review of past and current theories of mammalian brain evolution

is presented in order to discuss conceptual problems that persist in the field. Problems

with the concept of homology arise because of the interaction of cell lineages and axonal

connectivity in the determination of structural features of the brain. Focusing on the

continuity of information represented by ontogenetic mechanisms as opposed to mor

phological features avoids many of these problems and suggests homological relationships

that otherwise have gone unnoticed. Many apparently progressive trends and parallelisms

in mammalian brain evolution turn out to result from the influence of underlying devel

opmental homologies. Confusions about evolutionary advancement, increasing architec

tonic differentiation, and the evolution of new brain structures result from a failure to

appreciate how increasing brain size can bias developmental processes with respect to

axonal competition, increased cellular metabolic demands and decreased information

processing efficiency. Explanations of the evolution of novel structures and new connec

tional patterns are criticized for their failure to consider the constraints of neural devel

opmental processes. The correlations between structural neogenesis, functional special

ization and size changes in brain evolution are explained by a theory of competitive

displacement of neural connections by others during development under the biasing

influences of differential allometry, cell death or axon

target affinity changes. The

dis

placement hypothesis

is used to propose speculative accounts for the differential enlarge

ment and multiplication of cortical areas, the origins of mammalian isocortex, the unusual

features of dolphin cortex and the dramatic structural and functional reorganizations that

characterize human brain evolution.

Intrinsic difficulties

Despite the fact that the evolution of the

brain

particularly the human brain

of intrinsic interest to anyone curious about

the human mind and the origins of human

nature, the scientific study of brain evo

lution is not a major subdiscipline within

biology, psychology, anthropology or even

the neurosciences. The apparently poor

representation of this area of study in the

sciences can in part be attributed to the

paucity of direct paleontological evidence

regarding brain evolution and the long

time inaccessibility of crucial comparative

neuroanatomic details. The poor repre

sentation of information about brain evo

lution in disciplines outside the neurosci

ences is additionally limited by the

considerable sophistication in comparative

neuroanatomy and physiology that is

required to even begin to grapple with the

questions in a meaningful way. The dis

turbing correlate of this is that speculative

theories concerning brain evolutionó

especially human brain evolution

are

widespread and often contain relatively lit

tle neuroanatomical or neurophysiological

information. But even theories conceived

by neuroanatomists and neurophysiolo-

gists often reflect numerous unsupported

assumptions about the direction of evolu

tionary trends, the nature of natural selec

tion affecting brain processes, the ways that

brains can vary from one species to another,

the relationship between structure and

function within the brain and even the

nature of intelligence itself. Although

paleoneurology is unlikely to experience

sudden advances in the years to come, many

of the barriers to relevant neuroanatomi-

cal evidence have dissolved in the wake of

the introduction of many new experimen

tal techniques in recent decades. Now that

many of the technical impediments stand

From the Symposium on

Science as a Way of Know

the

way

detailed knowledge about

ing

Neurobiology and Behavior

organized by Edward

Hodgson and presented at the Centennial Meeting

brain structure and function have been

of the American Society of Zoologists, 27

30 Decem

removed, many of these hitherto unques-

ber

1989,

at Boston, Massachusetts.

tioned assumptions are now open to test.

Among colleagues in the neurosciences

one sometimes hears the criticism that,

unlike most other areas of neuroscience,

the study of brain evolution is limited to

theory because it is essentially beyond the

reach of experimental approaches.

Although brains of extinct species are not

available for direct inspection and analysis,

this does not necessarily mean that theories

of brain evolution are empirically untest-

able. Indeed they are every bit as suscep

tible to experimental investigation and

testing as are other theories of brain orga

nization and function. The approach must

necessarily be indirect, but it need be no

less effectual nor any less scientific or

experimental. We should remember that

the vast majority of scientific data in any

field is indirect, irrespective of whether the

object under study is directly observable.

From tracks left by subatomic particles in

nuclear accelerators to the measurements

of minute amounts of unseen biochemicals

registered in scintillation counters, nearly

all of the

hard data

generated in the

laboratories of any field of the natural sci

ences are indirect and circumstantial. It is

not the directness or indirectness of the

data that is important, rather it is the

repeatability of the findings and the coher

ence of many lines of evidence that are

crucial to scientific knowledge.

A good analogy to the study of brain

evolution is provided by the study of gene

evolution. Modern techniques for analyz

ing and comparing base sequences of DNA

molecules from living organisms are begin

ning to provide a truly astronomical fund

of information concerning both molecular

and organismal evolution. Without analyz

ing a single fossil specimen of DNA we are

nonetheless capable of reconstructing large

fractions of the genomes of extinct species,

characterizing major gene duplication and

reorganization events of the distant past,

and predicting the ancestral lineages of liv

ing species and the approximate dates of

their divergences. Al this is available today

despite the fact that only miniscule por

tions of the DNA in even the best studied

species are actually known and virtually

nothing is directly known about the DNA

of most species. This level of analysis is

made possible by the immensity and com

plexity of the existing genomes. In many

cases even direct fossil evidence of appar

ent phylogenetic relationships has been

abandoned in the face of contrary molec

ular information. As nearly limitless sources

of correlative molecular evidence are fed

into phylogenetic analyses in the near

future they will become immensely more

reliable for the determination of phylog

eny than the best of all possible fossil finds.

Living organisms are incredibly complex

systems at all levels of scale. Each molec

ular and organ system within an organism

embodies within its design the ubiquitous

mark of its particular evolutionary history.

In addition, the processes of embryogen-

esis that direct the construction of these

systems are themselves products and symp

toms of an evolutionary past that at various

levels intersects with the ancestries of other

species. Comparisons of the differences and

similarities among molecular systems,

organ systems and developmental pro

cesses in different species provide an almost

limitless source of information for inves-

tigating the evolution of biological struc-

tures. This is ultimately the final arbiter of

any analysis of evolutionary relation

ships

even for paleontological data—

since the interpretation of fossils is only as

accurate and complete as the information

we have about living counterparts.

The complexity of the vertebrate brain

rivals or exceeds the complexity of all the

other organ systems of the body consid

ered together. Because of this we should

expect that information derived from the

brains of living species will be more than

adequate to the task of investigating brain

evolution, so long as we are

willing and able

to approach the task with the level of

sophistication demanded by

it.

Given this

complexity and our still primitive under

standing of brain organizati'on and func

tion, we must be prepared to integrate

information from a variety of subfields of

neuroscience and evolutionary biology in

order to begin to approach the problems

of brain evolution with any clarity.

Although numerous researchers since the

nineteenth century have pursued the study

of brain evolution, most have focused on

a single source of evidence to support their

theories, including: relative brain size (e.g.,

apparent trends in brain size increase); fea

tures of cortical surface morphology (e.g.,

the appearance or reorganization of sulci);

relative sizes of macroscopic brain struc

tures with respect to one another (e.g., the

apparent enlargement of isocortex with

respect to limbic cortex in presumed

advanced

brains); or cyto

and myelo-

architectonic features (e.g., the apparent

enlargement of association cortex in the

cerebral cortices of

advanced

species).

But uni-dimensional approaches are almost

certain to lead into one misleading cul-de-

sac after another. This has been the fate

of many past theories, just as it will surely

also be the fate of the corresponding uni-

dimensional theories of the present. The

only hopeful approach is to integrate rel

evant information from many lines of neu-

robiological research that bear on the

questions of the patterns of variation and

constraint exhibited by the brains of dif

ferent species.

Experimental approaches

A number of recent technical advances

have significantly augmented the infor

mation previously available to comparative

neuroanatomists. Unlike many other organ

systems, the functionally relevant features

of brain anatomy are entirely microscopic

and for many decades were nearly impos

sible to distinguish even under the micro

scope. The axonal connections linking

neuron to neuron, though visible for short

distances in Golgi

stained material (avail

able since the turn of the century), have

only become amenable to study

recent

decades. In the 1950s techniques were per

fected for visualizing degenerating axons.

With these techniques it was possible to

identify the general patterns of long axonal

connections in the brains of experimental

animals. However, the resolution and sen

sitivity of these techniques were insuffi

cient to resolve many of the finer details

of axonal connection patterns. Beginning

in the mid 1970s a number of axonal trac

ing techniques were developed that took

advantage of the in vivo

uptake and axonal

transport. of amino acids, macromolecules

and certain fluorescent dyes. These tech

niques have now made it possible to inves

tigate the organization of axonal circuitry

in full microscopic detail. In this regard

the most basic functional anatomy of the

brain has at last become available for study.

We are still far from possessing a complete

connectional characterization for even the

best studied of mammalian brains, yet

already the scattered details from compar

ative studies have begun to provide a

remarkable array of new insights into the

patterns of brain diversity.

Now that tracer techniques have filled

this crucial gap in information about basic

neural functional anatomy, these data can

be integrated with data from physiological

and quantitative studies to provide all the

pieces of evidence necessary for investi

gating the principles underlying brain evo

lution. However, it is insufficient to apply

the analysis to adult brains only. Probably

the most crucial information for evolu

tionary purposes is how connection pat

terns and structural differentiation are ini

tially established in a developing brain. New

techniques for labeling mitotic cells, mark

ing cell lineages, and experimentally alter

ing development in neonatal animals or in

utero

by removing or transplanting embry

onic tissues are also beginning to provide

detailed information about the develop

mental processes that shape neural circuits.

Developmental information can play a cru

cial role in settling questions of homology.

More generally, it can provide evidence for

the range of possible mechanisms available

for natural selection to modify and dem

onstrates the constraints that limit possible

variation. Many scenarios of brain evolu

tion conceived in the absence of critical

information about the development of the

structures in question turn out to be incom

patible with these constraints.

This rapidly growing body of neuro-

biological information is providing an

unprecedented opportunity to discover new

patterns of similarity and variation in brain

evolution, and to test old and new hypoth-

eses about neural evolutionary processes.

It also provides impetus for a critical reex

amination of the dogmas and unanalyzed

assumptions that currently dominate

thinking about brain evolution.

Conceptual problems

Most of the theories concerning brain

evolution in the early part of the 20th cen

tury focused on its most studiable features,

size, gross morphology, and cytoarchitec

ture. Crude connectional information was

available only from careful dissections and

what little could be discerned with Golgi

staining. Despite the unavailability of cru

cial categories of information, theories of

brain evolution have flourished. As a result

modern students entering this field will find

the literature replete with numerous well

accepted dogmas about the general char

acter of brain evolution espoused by some

of the century's most brilliant comparative

neuroanatomists. Even more daunting is

the fact that many of these dogmas have

become seamlessly woven into the anatom

ical and functional terminology of the rest

of the neurosciences as well. Terms like

paleocortex, neocortex, primary areas,

secondary areas, projection cortex and

association cortex all bear the stamp of an

evolutionary vision that appears beyond

question, a part of the unspoken common

knowledge of the neurosciences. But with

the recent advent of new tools and a flood

of new information concerning neural con

nections, functions and ontogenetic pro

cesses, there has been a growing disso

nance between the new data and some of

the well

established principles of brain evo

lution. My purpose here is to play devil's

advocate; to question even the most well

founded of these dogmas and adopt the

heretical stance that many

if not most

of the traditional assumptions concerning

neural evolutionary processes are without

foundation. Hopefully, the introduction of

a healthy dose of skepticism will allow us

to look at the problem of brain evolution

with fresh eyes.

Four major conceptual problem areas will

be reexamined most carefully because of

their potential for misdirecting the study

of brain evolution and also because of the

influence they exert over contemporary

ideas about brain function in general. The

first of these is the concept of homology

the relationship shared by structures by

virtue of sharing a common ancestry. The

second is the notion of evolutionary progress

or orthogenesis

the idea that evolution

proceeds in a particular direction of

improvement or development. The third

is the significance of brain size

both in

absolute terms and relative to the body

to component brain structures. The fourth

is the problem of identifying and explain

ing neogenesis

the evolution of new struc

tures and functions.

All four conceptual issues were familiar

to the 19th century pioneers in this field

whose major assumptions in all these areas

remain dominant in many contemporary

treatments of the subject. Contemporary

versions of these ideas are in the back

ground of every theory of brain function

as well as every attempt to articulate a the

ory of brain evolution. Despite a century

of advances in evolutionary thinking in

other fields of biology these ideas within

the neurosciences still carry the distinct

imprint of late 19th century evolutionism.

And despite a century of experimental

investigation of brain function these ideas

still reflect one or the other side of the 19th

century debates over associationistic and

holistic theories of mental function. Exor

cising these influences is one of the central

purposes of this presentation.

The other major purpose is to present

an alternative approach to the study of

brain evolution that avoids many of'these

a priori assumptions. Many of these

assumptions have been derive'd from

attempts to arrange the adult nervous sys

tems of contemporary species in some sort

of evolutionary sequence or cladistic den

drogram so that any two may be linked via

some intermediate adult forms. While this

is a powerful heuristic it tends to imply the

misleading conclusion that the mecha

nisms for evolutionary change can be accu

rately described in terms of the modifica

tion of adult forms. This, of course, misses

a crucial intervening level of analysis. Ulti

mately, the mechanisms of evolutionary

change must be explained in terms of the

ontogenetic processes and developmental

constraints that build brains. Explanations

of evolutionary change that are not cast in

developmental terms are merely disguised

comparative morphological descriptions.

Consequently, the reviews and criticisms

presented will constantly appeal to devel

opmental data to test the plausibility and

consistency of some of the dominant the

ories of brain evolution. Finally, in the last

two sections many of these developmental

insights will be utilized to outline an alter

native ontogenetically based interpreta

tion of the processes underlying brain evo

lution in mammals. This interpretation of

brain reorganization events

called the

"displacement hypothesis"

suggests that

there is an interdependent relationship

bdtween differential growth of particular

neural cell groups and competitive-regres-

sive processes in brain development that

constrain patterns of brain evolution, often

resulting in parallel or converging trends.

Two particularly enigmatic cases are exam

ined in the last section

dolphin and

human brain evolution

and are inter

preted in terms of displacement processes.

Features of these brains that previously

have been difficult to explain or seemed

beyond study become understandable in

terms of the displacement hypothesis.

The concepts of homology and homoplasy

The concept of homology in some form

is essential to any study of evolutionary

morphology. It defines the warp of evo

lutionary continuity with respect to which

the weft of diverse adaptations can be

understood. In a useful summary of the

problems of homology in comparative neu

roscience Northcutt

984)

distinguishes

patristic homologies (the actual descent rela

tionship between an ancestral form and a

present form) from cladistic homologies (the

comparison of extant forms with respect

to their possible common ancestral rela

tionships) and then contrasts these with two

forms of dishornology that may be con

fused with homology. One of these, con

vergent homoplasy, corresponds to what has

traditionally been termed analogy and is

CLADlSTlC COMPARISON CLADlSTlC HOMOLOGV

PARALLEL HOMOPLASY CONVERGENT HOMOPLASY

Homology relationships. The basic homol

ogy relationships as outlined

Northcutt (1984). The

arrows represent descent relationships. The vertical

axis represents comparison over time or descent in

evolution and the horizontal dimension represents

comparison of contemporaneous species. The geo

metric shapes represent similar or different, ancestral

or derived traits. In the case of parallel homoplasy it

is unclear whether the parallel divergence from the

ancestral condition is

consequence of internal (ho

mological) or external (selectional) commonalities.

These same relationships can be applied equally to

comparisons between lineages or to homologous rep

etition of parts within an organism. (Redrawn from

Northcutt, 1984.)

applied to traits that exhibit structural or

functional similarities but which are not

derived from common ancestry. In other

words, their similarity is the result of influ

ences extrinsic to the organism. The other,

parallel homoplasy, has traditionally been

termed parallelism and refers to cases where

there is similarity in both form and com

mon ancestry but where the formal simi

larity between traits is not shared in the

common ancestral condition. In other

words, the formal similarity of the (patris-

tically) homologous parts is presumed to

have arisen independently in the two lin

eages after divergence from the common

ancestor. In this case there is both a patris

tic homological relationship and a cladistic

convergent homoplaseous relationship

involved. The parallel divergence of the

two descendent traits from the common

ancestral condition is presumably the result

of common extrinsic selection pressures.

These relationships are schematized in Fig

ure

(redrawn from Northcutt,

1984).

Where the ancestral condition is the

unknown feature to be inferred from cla-

distic comparisons it can be quite difficult

to distinguish homology from these two

forms of homoplasy. Northcutt, following

Wiley (1981) and others, suggests a num

ber of guidelines for aiding this discrimi

nation, including:

(1)

sharing deep similar

ities of form (as opposed to merely

superficial resemblances), (2) sharing com

mon epigenetic precedence (i.e., derivation

from common ontogenetic precursor

structures), and

(3)

the existence of a con

tinuity of intermediate forms in species

intermediate in relationship between the

two being compared. All there criteria are

versions of the identification of similarity

in some form.

In this discussion I will not review the

various problems encountered in attempt

ing to determine neural homologies in

practice, nor will I discuss methodological

strategies for circumventing these prob

lems and the multiple criteria that must be

satisfied to provide a convincing case. These

have been well reviewed elsewhere (Camp

bell, 1982; Campbell and Hodos, 1970;

Ghiselin, 1976; Northcutt, 1984). The

point of this discussion is to analyze the

concept itself as it is applied to problems

of brain evolution because

think there is

reason to be suspicious of the assumptions

implicit in its common usage. It is not the

empirical determination of homology that

is at issue, but the concept itself.

will argue

that there is something fundamentally

wrong with the notion of homology as it is

applied to the comparison of morpholog

ical features that can become especially

troublesome in the analysis of brain struc

ture.

Problems with the concept of homology

Homological relationships are most

clearly exhibited in topological relation

ships. A focus on topological continuity for

identifying homology dates back to the ear

liest pre

evolutionary theories about the

vertebrate

Bauplan

(an insightful his

torical discussion is provided in Russell,

19 16). Because organisms are spatially

organized systems, position within a net

work of relationships is crucial to conti

nuity of function. Although the particular

features of the individual components of

the organism may change over evolution

ary time the systemic relationships among

parts, including their contiguity relation

ships, are relatively stable. Even when

structures derived from

common evolu

tionary precursor have diverged in form

so as to share no superficial resemblance

their relationships to other structures

within the organism, both in the adult form

and at various stages of development, will

exhibit sufficient similarity to indicate their

homology.

The usefulness of topological criteria for

the determination of homological similar

ities derives from the fact that many mor

phological traits (although not the under

lying genes themselves) are determined by

systemic interactions between morphoge-

netic fields (or other converging morpho-

genetic influences) and not by independent

local mechanisms. The information that

determines

morphogenetic field inevit

ably derives from multiple genetic sources

interacting with one another sequentially

and simultaneously during ontogenesis.

The resultant morphological trait

bit

node within a network that has no

independent existence apart from, its rel

ative position. Such a node is defined by

its unique convergence of relationships with

other nodes. If some of these relationships

are lost or new ones are gained, continuity

with the previous state becomes ambiguous

and depends on whether you focus on the

relationships or the nodes. Analogously, a

single morphological feature may become

two if interdependent morphogenetic pro

cesses decouple in space or time, but two

features may also fuse to become one if

previously noninteracting morphogenetic

processes become subsequently linked and

interdependent.

This possibility is more likely in the brain

than in other organs by virtue of the fact

that brain traits are defined in terms of two

independent topological criteria: (1) cell

lineage relationships of local populations

that may determine local topological rela

tionships, cell structure, and molecular and

neurotransmitter characteristics; and (2)

connectional relationships determined by

axons that link numerous separated tar

gets, each likely derived from different cell

lineages, which may also influence mor

phological, cellular and functional char

acteristics of their various target struc

tures. Both cellular and connectional

attributes interact during development to

determine the architecture and function of

a brain region.

Assuming that connectivity is capable of

changing during the course of evolution it

is not hard to imagine the kinds of diffi

culties that might arise in interpreting

homologies. The position, cellular char

acteristics and even embryonic origins of

a brain structure in a descendent species

may be derived from a corresponding

stricture in some ancestor, and the

descen-

dent structure's connectional relationships

may also be derived from connection pat

terns in that same ancestor. Yet the par

ticular homologous circuit and homolo

gous structure may not have been

associated with each other in the brain of

that ancestor. For example, it is conceiv-

able that qseries of evolutionary events can

cause afferents of one brain structure to

invade some other structure, replacing the

ancestral

afferents of the new target—

similar effects can be induced experimen

tally (see below). In this event a connec

tional or circuit homology will have been

maintained, probably retaining its func

tional characteristics, but the relationship

between cellularly defined homologues and

connectionally defined homologues will

have become dissociated. The structural

homology can no longer be defined with

respect to position within a network and

the connectional homology can no longer

be defined with respect to the structures

that are connected. Continuity with the

ancestral form can nonetheless still be

traced through the remaining descent rela

tionships in each case, though the number

of topological criteria used to identify this

descent has diminished for each.

Further deterioration of homology cri

teria can also be imagined. For example, if

the connectional relationships play a sig

nificant role in determining the local

cytoarchitecture and neurotransmitter

characteristics of a target area (this appears

to happen in cerebral cortex, as indicated

by heterotopic transplantation experi

ments) it might appear on these grounds

that an ancestral target has simply become

displaced to a new position, despite the fact

that cell lineages and some connectional

relationships did not follow this shift. With

a large number of criteria in agreement,

but cell lineage and a few connectional

relationships do not follow this shift. With

difficult to decide between the deaffer-

ented target or the invaded target as the

appropriate homologue of the ancestral

structure. At the level of the whole struc

ture the judgment is ambiguous and yet

each underlying trait has a homologous pair

that can be traced in unbroken series to a

common ancestral condition. It is not clear

that shifting the analysis to these under

lying traits can escape similar problems at

a yet lower level. As we consider evolu

tionary

interventions

that might alter

progressively earlier stages of ontogeny it

is possible to imagine increasing loss of

descent criteria in this manner.

A similar complication can arise in the

effort to identify homologous sulci in rel

atively convoluted brains. Prior to the

development of techniques for unambig

uously staining myelin or neuron cell bod

ies the interpretation of sulcal homologies

in different species brains was considered

the best clue to the structural homology of

underlying regions, and this approach

dominated throughout the early part of

this century (Ariëns Kappers et al.

1936).

Although it has recently been abandoned

as unreliable for most comparative work,

it still remains the only evidence for paleo-

neurology (working with the casts of fos

silized crania). In the study of human evo

lution this has been the source for

continuing heated debates over the origins

modern

human brain traits (e.g.,

Falk,

1980, 1983, 1989; Holloway, 1981, 1984,

1985, 1988).

Most cortical sulci are probably the result

of the interactions between the mechanical

forces and constraints imposed by the cra

nial cavity, differences in growth rates of

brain areas, and relative elasticity of dif

ferent areas of the developing cortex and

underlying white matter. If the underlying

neural substrate influencing the formation

of a sulcus changes or becomes displaced

with respect to cranial landmarks in sub

sequent lineages it may cause the position

of the sulcus to follow. If this were the

typical case sulci might be relatively good

indicators of underlying brain structure

homologies. Alternatively, changes in bone

growth patterns of the skull or changes of

the absolute size of the brain with respect

to the skull in subsequent lineages may pro

duce changes in mechanical forces influ

encing sulcal position and cause a sulcus to

shift to a new location without any corre

sponding change in position of the original

neural substrate. In this case the link

between sulcal and neural homologies

would be broken. However, if the appear

ance of a particular sulcus is dependent on

the combined influence of both extrinsic

mechanical forces and intrinsic growth

processes of the neural tissue, then spatial

separation of these two independent influ

ences may cause a sulcus to disappear and

then reappear in some later lineage in

which these influences again become

realigned. This atavism would still be a case

of homology, despite the discontinuity in

descent. Finally, if a particular sulcus can

be induced by either influence alone then

spatial separation of extrinsic and intrinsic

factors may also produce two sulci where

one existed previously. In this case,

although each is patristically homologous

with the ancestral condition, it is unclear

whether they can be said to be cladistically

homologous to each other. All of these pos

sibilities demonstrate the dangers of treat

ing sulci as definitive markers of underly

ing neural homologies.

Similar problems with the strict descent

interprktation of homology have been

noted with respect to non

neurological

comparative problems, causing some

authors (e.g.,von Cranach,

1976;

Filler,

1986) to: suggest that the homology con

cept should be entirely abandoned. But an

alternative approach is suggested by these

problematic examples. The crucial ques

ions we are trying to answer by identifying

homologies are questions about

continuity

of information

(Van Valen, 1982). A mor

phological structure or any other manifest

trait is only the surface expression of

underlying information. This information

is encoded both in gene sequence and in

the topological and temporal conditions of

their expression in the developing organ

ism. The confluence of multiple indepen

dent sources and kinds of epigenetic infor

mation to form a particular structure

implies that no particular individual thread

of information constitutes an indispensable

link between homologous structures. A

homology exists so long as some relation

ships between the remaining sources of

information are maintained. Alternatively,

if separate threads of information are

passed down from generation to genera

tion independently and only brought into

relation with one another in some descen

dent where their interaction produces a

novel structure, we must consider the

structure as emergent and neoplastic (the

newly established relationships between

these threads of information is itself a bit

of information that

unprecedented) and

yet also recognize the complete homology

of underlying component morphogenetic

processes. The relationship is diagrammed

in Figure

Because ontogenetic processes

are multileveled, homological relation

ships must also be multileveled (Alberch,

1982; Fasolo and Malacarne, 1988), with

homologies at higher levels not necessarily

reduceable to those at lower levels. In addi

tion, homologies at every level above that

of the genes are to some extent

ephemeral,

capable of dissolving and reconstituting in

the course of evolution because they are

determined only in relational terms. This

also implies that the same bit of epigenetic

information expressed in a different con-

text within the same organism must also

be understood as homological.

Homologies between the different

parts of

brain

The interpretation of homolo

as com

mon information is crucial to another clas

sic use of the concept of homology:

serial

homology

homological multiplication of parts.

Repeated similar parts in the segments of

a worm, similar vertebrae in different posi

tions along the spinal column, similarities

in limb and digit structure, and bilaterally

symmetric parts of the body in general are

all examples of homologous repeated parts.

Although not descended from any single

ancestral structure, such homologous parts

undoubtedly inherit their similarities of

form from a single ancestral source of

developmental information. Within the

central nervous system there are examples

of classic serial homology in segmental

spinal cord circuitry, bilaterally symmetric

parts at all, levels, and multiple homologous

parts within every structure at many levels

of organization.

Starting on a small scale we recognize

that nearly all neurons exhibit homologous

axons, dendrites, synapses, etc.

within the same structure there are classes

of neurons with homologous patterns of

dendritic arborization, axonal targets and

neurotransmitters. Local circuit patterns

of nearby neuronal groups also exhibit

homologies, such as are found among cor

tical lamina and cortical columns in iso-

cortex. Even distributed functional circuits

linking separate structures may be serially

homologous:

e.g.,

projections from differ

ent' thalamic nuclei to different cortical

areas. Even structures that are superficially

quite distinct may exhibit underlying

homologies at some levels but not others.

This might be the case for the relationship

between the hippocampus and the isocor-

tex, which exhibit many features in com

mon at the cellular level and have homol

ogous patterns of afferents and efferents

yet very different laminar architecture.

Homologies between different brain

regions might possibly develop as a result

of derivation from a common undifferen

tiated ancestral structure, but descent

homology need not be defined at the struc

tural level only. It may also result from

independent expression of the same under

lying epigenetic information. Similarly,

during ontogeny homologies between cell

types may develop by descent from a com

mon embryogenetic cell lineage and

homologies between complex structures

may develop because they were each

derived in a process of differentiation from

some common embryological structure.

However, because all cells share the same

genetic information, it is also possible that

morphological

level

epigenetic

process level

morphological

homoplasy

epigenetic

homologies

ontogenetic interactions

phenotypes

intervening

variables

producing

developmenlal

plasticity

IG.

Developmental homologies. The multiple

level problem of developmental homologies is depicted

in a highly schematic form by representing interact

ing morphogenetic processes as arrows and the resul

tant morphological traits as geometric shapes. The

upper figure shows that the homological relationships

could be analyzed either at the morphological level

by comparing morphological (or even behavioral)

phenotypes or at the epigenetic process level by com

paring epigenetic mechanisms. Both are phenotypes

that indirectly represent underlying genotypes but

the higher level analysis condenses information rep

resented at the lower level by distinct processes and

can thereby miss considerable underlying develop

mental homology. Nonetheless the higher level anal

ysis also takes into account conserved or derived rela

tionships between underlying developmental

mechanisms that may themselves be homologous in

two lineages but which produce non

similar diverging

phenotypes. Of course the actual condition involves

many more than two levels. The lower level "epige-

netic mechanisms

are likely themselves the products

of relationships between yet lower level cellular or

molecular processes and the

phenotypic level

may

also be a set of epigenetic mechanisms of a higher

level of complexity. Hierarchic analysis cannot be

avoided.

homologous structures may appear simul

taneously in development by independent

expression of the same underlying infor

mation activated by some common molec

ular trigger or internal timing mechanism.

Within a number of areas of the brain

is likely that cell lineage is not the only and

perhaps not even the major determinant

of cellular, structural or functional homol

ogies. For a few brain areas there is now

evidence that a single precurser cell can

give rise to the multiple cell types within

that region (Rakic, 1988) and studies uti

lizing embryonic chimeras composed of two

immunologically distinguishable genomes

demonstrate that cells from both lineages

are effectively scattered diffusely through

out all areas and representing all cell types

(Goldowitz, 1987). Cell lineage probably

determines many local biochemical char

acteristics of neurons (Fasolo and Mala-

carne, 1988) and certain structural archi

tectonic features (e.g., Kuljis and Rakic,

1988) and it may provide gross areal dif

ferences distinguishing major morphoge

netic fields, but at the present time there

is little positive evidence available on this

point and extensive evidence that extrinsic

influences determine function and neural

connectivity to a large degree (O'Leary,

1989). Timing of final mitosis and inter

cellular interactions also appear to play sig

nificant roles in determining neuronal cel

lular types and local structural and

functional characteristics.

In general, a major part of structural

differentiation in the developing brain is

based on distributed information that is

embodied systemically in its spatial

tem

poral organization and dynamically in the

axonal interactions between indepen

dently derived neuronal populations. The

details of this process will be discussed in

later sections, but in terms of homology

this fact leads to an important conclusion.

If the information distinguishing one

region from others is not entirely embod

ied within the cells of that region, but is

expressed only as those cells are contacted

by invading axons and as its own axons

establish efferent terminations, then dif

ferent serially homologous structures

within the brain (e.g., different cortical

regions) do not ultimately determine their

own distinctions of structure and function.

Their unique specializations with respect

to one another are instead derived from

network relationships with other areas of

the brain (both cortical and subcortical).

Functional homology

One last use of the concept of homology

tnust be introduced at this stage before

moving on to a discussion of some of the

major theories of brain evolution: the con

cept of functional homology. It can be

defined as the similarity and continuity that

exist between functions as a result of

homologies between their substrates. The

evolution of new functions by the modifi

cation of old structures is a common theme

in evolution. When the vertebrate fore

limb evolved the capacity for flight in the

evolution of birds, the skeletal, muscular

and neural structures retained the general

Bauplan

of the ancestral terrestrial con

dition and also retained numerous func

tional constraints. These have all played a

role in shaping flying behavior in bird

species. Additionally, the peripheral motor

neural architecture (Sokoloff

et al.,

1989)

and even features of the central locomotor

patterns

(Kaplan and Goslow, 1989)

exhibit strong similarities in birds and ter

restrial quadrupeds, despite the other

major functional differences that their

exclusive adaptations demanded.

With the differentiation of new neural

circuits from ancestral circuits and the

elaboration of corresponding new func

tional adaptations we can expect to trace

functional homologies in the form of

underlying functional similarities and con

straints. Even in extreme cases in which

neural structure is co

opted for new adap

tations that are radically different than the

ancestral function, the underlying homol

ogies will likely exert a major organizing

influence on the form and range of vari

ability of the new function. This may eyen

be true of such a novel adaptation as the

syntactic structure of language (e.g., Reyn

olds, 1976; Lieberman, 1984; Deacon,

1990c), if some of the cortical systems that

came to serve language functions in the

course of human evolution had been ante

cedently adapted for other behaviors (e.g.,

motor planning). Anatomical evidence for

such a view is presented by Deacon (1 988a,

1990c).

Functional homologies should also be

exhibited by serially homologous struc

tures within the same brain. For example

the many homologous structural features

shared by all regions of isocortex suggest

that there should be strong functional

homologies shared by all of its functional

subdivisions despite the radical differences

in modality of their input

output relation

ships (Diamond, 1979). The same may also

be said of the different regions and sub

divisions of the basal ganglia (Alexander

and Strick, 1986). Presumably, the affer-

ents to each homologous area transmit dis

tinct forms of information that are sub

jected to some common neural calculation

in each homologous area. For this reason,

different scenarios for the phylogenetic

ancestry of brain structures that suggest

different ancestor-descent

relationships

bring with them different predictions con-

cerning function.

PROGRESSION

The assumption of evolutionary progress

The idea of progressive evolution is a

product of the uneasy marriage between

Darwinism and the

scala naturae

theories

of the mid 19th century. It received its

clearest expression in the theories of Spen

cer, Haeckel, Berg and Teilhard de Char-

din among other influential writers.

Althou h evolutionary biologists in recent

decades have learned to rigorously avoid

making such assumptions when thinking

about a particular assemblage of fossils or

a lineage of species, this habit of thought

is not so well entrenched in the neurosci

ences, nor in anthropology, psychology or

linguistics where theories and assumptions

about human brain evolution are also likely

to be found. The tendency is so pervasive

that evolution is often considered synon

ymous with progress, whereas evolution

ary change without progress, even when

directional, is often not considered evolu

tion at all, merely

drift.

The ubiquity of the idea of progress in

brain evolution can be traced to what we

believe we already know about our own

place in an intellectual chain of being. It

apparently goes without saying that humans

are the smartest species to have ever lived----

never mind that we are not sure what we

mean by

smartest"

and it is also popular

knowledge that human evolution involved

significant brain enlargement. Our brain

must also be the most complex, if for no

other reason than the fact that our abilities

are the most complex of any species. Since

we have appeared only recently after a long

period of brain evolution characterized by

less intelligent species, our brain repre

sents the pinnacle of some long evolution

ary trend.

From these few assumptions a great many

predictions must follow, and so from the

outset we feel confident in assuming the

answers to a number of central questions:

bigger brains are smarter brains; more

complex brains are more developed brains;

primates are smarter than other species;

our closest relatives, the great apes, are

smarter than other primates; there is an

evolutionary trend toward increased intel

ligence; more intelligence is always a supe

rior adaptation to less; brain evolution tends

toward increasing complexity and increased

relative brain size; earlier stages of brain

evolution are characterized by more prim

itive, relatively less differentiated and rel

atively smaller brains than later stages; parts

of the brain that are relatively undiffer

entiated are more primitive and parts that

are more complex are more recent; brain

structures that enlarged most in ourselves

and our close ancestors are the most highly

developed and most recent brain struc

tures; the most recent human functions

(i.e.,

language) must be controlled by the most

advanced, complex and recent structures

in the brain; etc. All we need to do is to

find out how the data concerning brain size

and brain structure diversity demonstrate

these points! Presumably, whatever fea

tures of brain organization we use to com

pare brains of different species,

Homo sapi-

ens

should represent the extreme high end

of the scale (however this is defined in each

case). I call this assumption the

Anthro

pocentric Maxim.

The tenacious hold of anthropocentrism

on our thinking about brain evolution is

great. What is needed is a biological equiv

alent of the

Copernican Revolution

finally shake it loose. Along with this

implicit anthropocentrism we should also

endeavor to root out the tendency to

assume progressive trends in any aspect of

brain evolution, unless and until all alter

native explanations have been exhausted.

There undoubtedly are progressive trends

in brain evolution, but to clearly identify

them and to understand their significance

we must demonstrate that they are not

merely superficial correlates of other non-

progressive trends. To be able to do so

requires that we first understand these

other trends.

The a

priori

assumption of

advance

ment

in evolutionary sequences is a source

of many misunderstandings. Deacon

(1990a) reviews many of the assumptions

about brain evolution that derive from the

notion of evolutionary progress in brain

size. Even theories that do not specifically

invoke the notion of progression nonethe

less tacitly assume it in the process of iden

tifying some structures as

advanced

and

others as

primitive.

A primitive to

advanced ranking of living organisms or

their structures must ultimately be based

upon independent knowledge of the evo

lutionary trend in question; otherwise the

argument is circular. But when faced with

structures that leave no fossil evidence

independent evidence is hard to obtain.

One possibility is to assume that the pro

gressive ranking of soft

tissue structures

should correlate with other preserved indi

cators of the relative primitiveness or

advancement of the organism as a whole.

Overall similarity of traits from living

species to those in early fossil specimens of

some lineage might suggest that the orga

nization of brain structures is also equally

comparable. It is of course necessary to

determine that the resemblances are not

superficial. and the result of convergent

evolution. And even when this can be dem

onstrated there is never any guarantee that

the brain structures in question have been

as conservative as the rest of the mor

phology. Even the external morphology of

the brain, as may be revealed by endocasts,

cannot be taken as a reliable indicator of

underlying cellular and connectional

homologies. So a primitive external

appearance of modern brains is an untrus

tworthy indicator of primitive brain orga

nization.

From simple to complex

It seems unquestionable that simpler

brain structure precedes more complex

brain structure in the course of evolution,

and that more highly differentiated brain

structures are more advanced than more

diffusely organized brain structures.

Although we can probably assume that

there are some recent brains that are more

differentiated than any from fifty million

years ago, we cannot safely invert the logic

and assume that the most undifferentiated

contemporary brains are the least derived.

Confounding variables such as absolute size

and specific sensory

motor specializations

may influence relative differentiation, and

problems in assessing homology as well as

sampling biases inherent in the phyletic

representation of species may introduce

spurious correlates of differentiation that

have nothing to do with primitiveness.

In discussions of mammalian evolution

small bodied living insectivores are typi

cally treated as exemplars of the mor

phologies of ancestral mammals. These so

called

basal insectivores

are assumed to

generalized

in their adaptation and

conservative

with respect to evolution

ary trends, although caveats are usually

suggested regarding the fact that each of

these groups represents some rather spe-

cialized adaptations as well. The European

hedgehog (Elliot

Smith, 19 10; Ariëns Kap-

pers

al.,

1936; Filimonoff, 1949; Dia

mond and Hall, 1969; Valverde and López-

Mascaraque, 1981; Sarnat and Netsky,

198 1) as well as moles, tenrecs and micro-

chiropteran bats have all been cited as pos

sessing conservative brain structure typical

of an

initial

mammal brain (Sanides,

1969, 1970; Le Gros Clarke, 197 1; Glezer

al.,

1988). There are unfortunatel

number of circular assumptions in the con

cepts of

primitive survivor

and

basal

insectivore

(Martin, 1973) that also afflict

the concept of an

initial brain."

Fossil specimens suggest that it is likely

that the eutherian mammal ancestor which

gave rise to the Paleocene

Eocene radia

tions was of relatively small body size and

probably bore at least a superficial resem

blance to modern shrew

like insectivores.

In this regard there is considerable justi

fication for selecting insectivores as exem

plary of the ancestral condition. The pre

sumption that the common ancestor was

somehow

generalized

or even that mod-

ern basal insectivores are

generalized

species seems a little more puzzling,

although it is widely claimed. In many ways

members of these groups represent some

extremes of specialization. Consider, for

example, the echolocation specialization of

microchiropteran bats, the fossorial or

nocturnal specializations of many shrews,

moles and hedgehogs, the aquatic special

izations of some exceptional shrews, and

of course the insectivorous specialization

itself. These facts must certainly relate to

their neurological adaptation. Of course

there is every reason to suspect that the

common ancestor of eutherian mammals

was also specialized in some interesting

ways, but given the radical difference in

faunal context and likely niche specializa-

tion there may be no corresponding spe

cialization represented in modern species.

The tree shrew Tupaia has been sug

gested by some as an appropriate living

model for a Paleocene precursor to pri

mates (Le Gros Clark, 197 1

;

Cartmill, 1972,

1974), In terms of its size and many of its

non-neurological features it too might serve

as a reasonable stem mammal model. But

it is usually disqualified as an

initial brain

model because it possesses

number of

advanced

brain features, including mod

erate encephalization and a differentiated

striate cortex and visual association cortex.

The cortex of Erinaceus, the European

hedgehog, is often treated as a model of

an ancestral mammalian cerebral cortex.

Figure 3

depicts some of the known areal

divisions of the hedgehog cortex along with

an even more

primitive

tenrec brain.

Some notable features of the isocortex of

these species as compared to

advanced

brains include: relatively small size com

pared to olfactory and limbic cortex, poorly

distinguishable lamination, low level of

myelination, poor differentiation from area

to area, lack of a clearly distinguishable

agranular motor area, poorly granularized

koniocortical sensory areas, vagueness

boundaries between architectonic areas,

the apparent adjacency of sensory

motor

projection areas with little interdigitated

association cortex, and a relatively thick

layer

(a limbic cortex characteristic) in all

areas of its isocortex. It seems unquestion

able that these brains are near some

extreme in the spectrum

cortical orga

nization among eutherian mammals,

but

this may not be conservatism.

fact, on

the basis of brain traits selected for their

value in determining cladistic relation

ships, Kirsch et al. (1983) find that hedge

hogs do not appear to exhibit a prepon

derance of conserved traits, but just the

opposite, they appear to possess one of the

most derived mammal brains (Johnson,

1988).

It is clearly not the structure of the

hedgehog body that motivates its choice as

an exemplar. It exhibits highly specialized

spiny hairs for predator protection and has

developed the ability to role into a ball with

only its spines exposed, it has relatively

short, stubby limbs specialized for digging,

it has very rudimentary visual abilities with

clearly reduced eyes that are appropriate

to its nocturnal

fossorial habit, and it has

a well developed specialized snout and pre

sumably highly specialized olfactory abili

ties for insect predation. Campbell (1988)

remarks that if the hedgehog were other

wise the same but possessed a larger more

differentiated brain it would never have

been considered an exemplar of the

ini

tial brain'' pattern. Gould (1977) notes that

in general it is unwise to choose the most

undifferentiated extant member of a group

as a representative of its stem ancestor pre

cisely because small bodied fast breeding

forms are likely to be highly derived

selected species. The choice of species

with small undifferentiated brains is not so

much motivated by external similarities

with known fossil types as by apriori

assumptions about what is primitive and

what is advanced.

To carry this paradigm to its logical

extension, the hedgehog is probably not

the most extreme case that could be cited.

Zilles and Rehkämper (1 988) point out that

Erinaceus is actually somewhat advanced

with respect to some other basal insecti

vores and therefore might not be the ideal

exemplar of the

Grundtypus

for mam

malian brain organization. They note that

the brains of the tenrec Echinops and the

geogaline Geogale exhibit even less enceph-

alization and exhibit fewer progressive fea-

lateral view

brain of a tenrec brain of a hedgehog

Centetes Erinaceus

Hedgehog and tenrec brains as seen from above and the side labeled to show approximate positions

of the major sensory and somatomotor fields. Isocortex is indicated in gray in the left hemisphere of each

and limbic and olfactory cortex is white in the same hemisphere. Since most of the cortical representation is

unknown for the tenrec and only partially known for the hedgehog specific boundaries between areas are

not indicated. There is no intent to imply either undifferentiated cortex or the existence of only single

sensory/motor fields. Note the low ratio of isocortex to limbic

olfactory cortex in these brains, especially the

small tenrec brain.

tures than that of

Erinaceus.

These authors

conclude that

Erinaceus

is probably

not

typical representative of a real

basal

insec

tivore

(Zilles and Rehkämper, 1988;

emphasis in the original). Only in a context

where evolution is presumed to progress

from simple to complex, from least en-

cephalized to most encephalized, and from

generalized, inflexible and inefficient in

function to specialized, flexible and highly

efficient in function, can the search for the

absolutely simplest mammalian brain be

equated with the search for the ancestral

brain.

There are two general attributes shared

essentially all the basal insectivores con

sidered primitive in brain organization that

should cause us to be cautious about gen

eralizing from them. First, each of the can

didate exemplar species inhabits a noctur

nal

fossorial niche. This is probably no

accident. This adaptation has likely pro

duced secondary reduction or dedifferen-

tiation of the visual system and a corre

spondingly heavy reliance on the olfactory

system. Evolutionary reduction or degen

eration of an essentially unused sense

modality may induce dedifferentiation, but

this does not likely follow an exactly

reversed phyletic trend and may produce

structural features that are quite distinct

from ancestral features. How can we be

sure that the relatively undifferentiated

state of the cortex of these species is rep

resentative of

retained primitive state

rather than a recent specialization?

Second, these exemplar species also rep

resent the very lowest limits of mammalian

brain size. This is a problem because many

measures of structural complexity appear

to be strongly correlated with brain size

(Tower, 1954;

Haug, 1987;

Deacon, 1990a;

and see the following section). Nearly all

the attributes of

primitiveness

of mam

malian brains are also typical attributes of

very small brains, while those of "advance-

ment

are

only

expressed in relatively large

brains. Progressive trends measured with

respect to these small insectivore species

are significantly confounded with the

effects of differences of scale. Also if there

has been prolonged selection for size

reduction in these species there may also

be simplifications of brain structure of a

secondary character which do not neces

sarily follow a reverse phylogenetic trend.

Cladistic approaches

The cladistic approach to identifying

evolutionary trends offers some hope of

resolving these ambiguities and avoiding

the trap of implicit progressionism. By

lacin

the assumption of evolutionary

deyelopment and increase in complexity

with a simpler empirically defined dichot

omy between conserved and derived con

ditions one can arrive at a relatively

unbiased criterion for identifying evolu

tionary trends. The particular character

istics of the trait are irrelevant, only its

presence or absence in different groups is

important. By pairwise comparison of the

presence or absence of traits between

species in progressively more distant out-

groups it is possible to decide which traits

can be operationally defined as derived and

which can be defined as ancestral or con

servative. Cladistic analysis has wide accep

tance as a means for reconstructing phy-

ietic relationships between lineages, but it

has also been used extensively to trace the

ancestry of specific traits. It has been par

ticularly useful for deciding between alter

native accounts of a trait's evolution

because it provides a measure of parsi

mony. For example, Northcutt (1984)

uses

the number of mutational events that must

be postulated in order to explain the dis

tribution of certain vertebrate brain traits

according to different theories to decide

which of these theories provide the most

parsimonious accounts.

The Achilles heel of this approach with

respect to brain evolution is that it will

inevitably tend to favor identifying rela

tively undifferentiated forms as more

primitive and differentiated forms as more

derived.

structure lacking differentiat

ing features will tend to be glossed as sim

ilar across a wider range of species than

one exhibiting a number of easily discrim

inated features. As a result, despite its

apparently unbiased definition of polarity,

the cladistic approach may be biased, so as

to pick out more generalized and less dif

ferentiated traits as characterizing

com

mon ancestor. Evolutionary regression in

certain lineages is potentially a source of

misleading bias as is the correlation of

absolute brain size with structural com

plexity. Additionally, this approach is sen

sitive to the effects of convergent or par

allel evolution. It will be argued below that

parallelism is

major feature of mammal

brain evolution.

Nonetheless, the cladistic approach is in

some ways self

correcting in this regard. It

can be useful in discerning some of these

biasing influences by using multiple sources

of information pooled to establish most

parsimonious descent relationships and

then reanalyzing individual trends. For

example, the relative primitiveness of the

basal insectivore

brain can be tested with

respect to three outgroups of mammals

whose phylogenetic affinities are well

known through other cladistic analyses: the

marsupials and the two living monotremes,

the platypus

Ornithorhinchus

and the

echidna, or spiny anteater,

Tachyglossus.

Many of the characteristics of

Erinaceus'

brain, including apparent adjacency of

projection areas, minimal association cor

tex, high ratio of olfactory

limbic cortex

to isocortex, poor laminar distinction, poor

granularization and poor differentiation of

architectonic areas, are not exhibited in

the brains of larger marsupials and mono-

tremes. Have the apparently more

advanced traits also found in these out-

groups evolved independently in the larger

brains of all the mammalian lineages? The

more parsimonious interpretation is that

many of these traits were present in some

form in the common ancestor of all mam

malian groups long before the recent

eutherian radiations. That they fail to be

exhibited by some of the brains in the

eutherian lineage

(e.g.,

basal insectivoresj

and some brains in the marsupial lineage

(e.g.,

Didelphis virginiana)

is not sufficient

evidence to assume that they are derived.

Kaas (1989) applies an implicit cladistic

approach to determine which cortical areas

in all mammals can be traced via descent

from a common ancestor. He notes that in

all the major mammalian lineages (euthe-

rian and metatherian) there are distinct

visual, auditory and somato

motor projec

tion areas within isocortex. He concludes

that the common ancestor for all these lin

eages likely also possessed these differen

tiated projection areas and not just an

undifferentiated protoisocortex. Based on

this evidence he rules out a widely cited

theory of cortical evolution proposed by

Sanides (1970)

that is based on the assump

tion that generalized undifferentiated iso-

cortex preceded specialized sensory

motor

projection cortices in the course of cortical

evolution. However, to be more explicit,

what has been demonstrated is that dis

crete somatic, auditory and visual projec

tion areas are expressed in mammal brains

under all existing conditions and sizes,

whereas some areas, particularly many

association areas, fail to be expressed under

many conditions, specifically in small brains.

The classic view that association cortex

is new in comparison to projection cortex

in part derives from the apparent lack of

association cortex in basal insectivore brains

(but this assumption is criticized below) and

its progressive domination of the cortical

surface in

advanced

mammals. How

ever, some of the larger marsupials and

even the echidna appear to exhibit signif

icant expanses of association cortex in

addition to primary sensory

motor projec

tion areas. Apparently, under similar

developmental conditions

large brain

size

this trait is expressed in every lin

eage of mammals. The common conditions

required for expression of this trait in all

three lineages also lends confidence to the

claim for homology as opposed to parallel

homoplasy

The failure of basal insectivore brains to

exhibit distinctly segregated association

areas is not sufficient evidence to deny that

this trait is a shared ancestral trait. None

theless the appearance of segregated visual,

auditory and somato

motor areas in all

mammal brains is sufficient evidence to

consider them as shared ancestral traits.

Lack of positive evidence is not sufficient

to deny homology but the availability of

positive evidence is sufficient to establish

it.

This can also be applied to questions con

cerning the origins of somato

motor areas.

Lende (1969) demonstrated that in Didel-

phis the somatosensory responsive cortex

and the electrically excitable motor cortex

exhibited complete overlap and that in Eri-

naceus there was a large region of overlap.

More recently some degree of overlap has

also been demonstrated in rats (Donoghue

et al., 1979). In carnivores and primates

(and probably ungulates) these areas are

adjacent but completely segregated into

distinct parallel somatotopic and muscu-

lotopic maps. Lende also argued that there

was even some overlap of visual and audi

tory cortical areas in the opossum (although

this finding has not been replicated). This

suggested to him that the ancestral state of

cortex was characterized by poor areal dif

ferentiation in which all the sensory modal

ities exhibited nearly complete overlap with

one another. However, at least one larger

Australian marsupial, the brush-tailed

opossum Trichosurus, exhibits considerable

segregation of somatic and motor fields

(Haight and Neylon, 1978, 1979) and the

monotremes appear to exhibit complete

segregation of somatic and motor areas.

All of these facts

argue against assuming

that the primitive condition was undiffer

entiated and completely overlapping and

suggest that at least some degree of seg-

regation of these functional zones char

acterized the common ancestor of all mam-

ma1 groups.

But negative evidence can be cited to

support the view that the segregation of

somatic and motor modalities is a conver

gent trait. This evidence comes from vari

ations in somatotopy of the sensory and

motor maps in the different groups. In most

eutherian mammals studied the two fields

are arranged as mirror images of one

another with respect to their common bor

der, and exhibit this pattern even in species

where there is considerable overlap of the

two areas. However, in edentates and mar

supials the two maps appear to be arranged

in parallel as well as overlapped (Dom

al., 1971; Lende, 1963, 1969; Magalhães-

Castro and Saraiva, 1971; Royce et al.,

1975; Saraiva and Magalhães-Castro,

1975), and in a megachiropteran bat (Pter-

opus poliocephalus) the somatic map appears

inverted from that typical of most other

eutherian mammals (Calford et al., 1985).

The monotremes appear to exhibit char

acteristics found both in some eutherian

and in som'e marsupial brains (Bohringer

and Rowe, 1977). Furthermore, the pat

tern of thalamocortical connections to these

areas differs in eutherian and marsupial

brains.

This negative evidence is inconclusive

because the differentiation of map orien

tation could occur independent of the seg

regation of somatic and motor areas. The

unique status of the fox bat and edentate

somatic maps in comparison to other mam

malian groups suggests that this is the case.

Map orientation appears to be a derived

condition in these species. Variation of this

trait occurs against the background of

somatic and motor map segregation as an

apparently older and more conservative

trait.

e only placental and marsupial

mammals that do not exhibit segregation

of somatic and motor projection areas also

have relatively small brains. This further

suggests that this is a derived condition

contingent on small size and not the ances

tral condition.

Problems with these comparisons of cor

tical areas stem from the fact that the traits

under consideration are not simple and the

variables that correlate with the differen

tial expression of these complex traits have

not been controlled for in the analysis. The

most important of these variables is brain

size, but other factors are also clearly

involved with regard to more subtle fea

tures, such as map topography. Failure to

control for these factors inevitably leads to

their being confounded with descent rela

tionships despite the fact that cladistic anal

ysis itself does not prejudge the primitive

or advanced status of a trait. The differ

ential expression of a large number of traits

with respect to brain size or the differential

expression of traits in brains with respect

to sensory specializations can be a serious

problem for cladistic analyses because it

vastly increases the probability of conver

gence and parallelism. In fact, many of the

advanced

traits of eutherian mammal

brains could have been inherited from the

common ancestor of eutherian mammals

even if that ancestor failed to exhibit any of these

traits.

All three mammalian groups have likely

inherited neural developmental con

straints and tendencies from a common

ancestor that are expressed differentially

in different contexts. It is possible that the

initial

eutherian brain possessed the

developmental information for these traits

but failed to express them because certain

other conditions of their expression were

not met. Below it will be argued that the

small size of these brains precludes the

developmental expression of cortical par-

cellation processes necessary to produce

multiple highly differentiated cortical

areas. Also the regression of the visual sys

tem in basal insectivores may further

undermine these processes. Despite the fact

that ancestral traits might not be expressed

in an intervening lineage there need be no

interruption of their descent to subsequent

lineages, and their disappearance or reap

pearance in certain lineages cannot be

attributed to distinct mutation events and

treated as distinct derived conditions. We

should be especially wary of this possibility

in the choice of traits included for cladistic

analysis.

This again underscores the importance

of thinking of homologies in informational

terms. Even if we were to miraculously learn

the details of morphology of the brain of

the true eutherian ancestor we would per

haps still be a long way from understanding

the initial conditions embodied in that initial

brain that still influence the structure and

the evolution of modern brains. What we

ultimately want to know is what information

was embodied in the initial brain and in

the developmental mechanisms that built

it.

Recapitulationism

During embryogenesis there is a definite

progression from smaller poorly differen

tiated structures to larger highly differ

entiated structures. Large species with rel-

atively differentiated brains must inevitably

pass through developmental stages in which

their brains are small and poorly differ

entiated. Consequently, the embryonic

stages of larger more differentiated brains

will inevitably bear some superficial resem

blance to the adult stage of small poorly

differentiated brains. Structures that dif

ferentiate later in development will thus

appear to be added to an otherwise com

mon substrate. Recapitulation assumes that

adult structures in primitive species are

homologous to embryonic structures in

advanced species. Subsequent modifica

tions of brain structure have been added

in the more

advanced

species by extend

ing the ontogenetic process to include

additional later stages. This process of ter-

minal addition

was presumed to link the

evolution of species to development in such

a way that a scale of increasing complexity

in evolution was the inevitable outcome of

a progressive increase of developmental

information. Evolution is explained as an

augmentation of ontogenesis and more

advanced

species are literally assumed to

be further developed.

Despite the enormous theoretical power

of this synthesis, the weight of comparative

and ontogenetic evidence that has accu

mulated against this doctrine in the last

century is overwhelming. Ontogenetic dif

ferences that distinguish different lineages

may occur at any stage of development and

are clearly not constrained to occur in lin

ear sequence with increasing phylogenetic

.divergence. Although larger creatures tend

to exhibit longer ontogenesis than smaller

creatures, larger creatures do not have

additional stages added on and highly

derived members of a lineage do not show

more developmental stages than highly

conservative members; differences in

development appear at many correspond

ing stages. However, despite the patent

failure of this paradigm, numerous unrec

ognized recapitulationist assumptions still

persist in the literature about brain evo

lution, primarily in the form of tacit ter

minal addition assumptions.

Early recapitulationist theories of brain

development suggested that as children

matured they passed through stages of cog

nitive and emotional development that

corresponded to distinct

grades

of ani

mal consciousness from reptilian to mam

malian, from primitive mammal to pri

mate, from primate to primitive human,

and finally through the ascending stages

from

primitive savagery

to modern civ

ilization (e.g., Spencer, 1870; Baldwin,

1895). Various primitive human societies

and criminals were viewed as arrested at

some prior stage of development. Although

the recapitulational structure of the brain

was assumed by many prominent 19th cen

tury and early 20th century neurologists

(e.g., Paul Broca, John Hughlings Jackson,

Ivan Pavlov, John Sherrington, and Sig

mund Freud) probably the major catalyst

for the formation of a comparative ana

tomical version of this theory came as a

result of the synthesis of two sources of

neuroanatomical evidence just subsequent

to the turn of the 20th century.

Paul Flechsig's (1 90 1) analysis of myelin-

stained tissue from human fetuses at var

ious stages of development demonstrated

a progression of myelination of isocortical

areas that began with primary sensory and

motor areas, continued to belt zones around

these areas and culminated in relatively late

myelinating association areas (see Fig. 4).

This sequence was presumed to correlate

with developmental trends in which basic

sensory

motor abilities mature early in

childhood and the

highest

intellectual

abilities only appear late in developmeht.

It was also presumed to correlate with

evolutionary sequence from species with

only crude sensory

motor habits and

responses, to species with the capability of

complex and flexible associational learning

abilities. The most developed associated

areas were assumed to be the newest in

evolutionary terms and the last to develop

in ontogeny.

At roughly the same time .comparative

anatomists, using preparations that visu

alized neuron cell bodies, produced maps

of the distinguishable cytoarchitectonic

areas of the cerebral cortex for a number

of mammalian species which appeared to

demonstrate homologues for primary pro-

jection areas in all species, but no homo-

lateral view

logues for many human association areas

in monkeys and no homologues for many

monkey association areas in other mam

mals

(e.g.,

Campbell, 1 905; Brodmann,

1909). It clearly appeared as though the

developmental trend recapitulated a

sequence of additions of new cortical areas

leading up to the human brain. New, more

developed 'association areas were appar

ently added to the brain of succeeding

species at the end of their maturational

development. The unusually long postna

tal brain development of humans could also

be explained on the basis of the extra stages

of brain development that were appended

to, human ontogeny.

These assumptions corresponded with

the neuropsychological doctrine of the

time. It was assumed that the moment to

moment processing of sensory input

retraced this same hierarchy, developing

from the crude registration of sensory

information in projection areas, to the con

struction of a perceptual gestalt in sensory

psychic

areas, and finally to the elaboration

of multimodal associations with respect to

different remembered vercevtions, actions

and emotional experiences in association

areas. Primitive animals and young chil

dren only progressed through the initial

stages of this cognitive hierarchy. Although

the developmental recapitulation assump

tion has been abandoned, a more elaborate

version of this basic hierarchic functional

interpretation remains the dominant con

temporary theory of sensory processing

(e.g.,

Mishkin and Appenzeller, 1987;

Maunsell and Van Essen, 1983), despite

many growing inconsistencies and the

availability of alternative interpretations

(e.g.,

Brown, 1988; Deacon, 1989a; Dia

mond, 1982; Optican and Richmond,

1987).

However, in hindsight we can see fun

damental flaws in both forms of evidence

for this correlation of ontogenetic devel

opment with apparent phylogenetic devel

opment. Ultimately

will argue that both

derived from a failure to control for factors

having more to do with brain size than with

phylogenetic progression. In this section I

medial view

Flechsig's myelogenesis figure of the human

brain is redrawn from the original leaving out Flech-

sig's numerical designation of myeloarchitectonic

fields. The darker areas represent the areas that mye-

linate earliest in development and the white areas

represent the areas that myelinate latest. Insular cor

tex is partially exposed. Note the progression from

primary to secondary to association areas of cortex.

Note also the early myelination of limbic cortical areas

and pathways.

will primarily address the issue of the

developmental sequence, but will return to

the issue of the terminal addition of cor

tical areas in a later section.

An alternative explanation for the pri

mary

secondary

tertiary hierarchy of my-

elination of cortical areas can be derived

from a correlation between the total level

of adult myelination achieved by these

structures and their apparent develop

mental schedule. The early myelinating

areas are also the most heavily myelinated

in the adult and the latest myelinating areas

are the least myelinated in the adult

(Bishop, 1959; Sanides, 1970). This rule

also applies to noncortical structures: the

relatively poorly myelinated reticular for

mation of the midbrain exhibits one of the

latest and longest cycles of myelination of

any system (Yakovlev and Lecours, 1967)

despite the fact that it is without doubt one

of the most conserved structures in the ver

tebrate brain. This fact clearly contradicts

the recapitulationist assumption.

There are two possible interpretations

of this correlation between total amount

of myelination and the developmental time

course of myelination. First, since these

assessments of myelination are based on

myelin

stained tissue sections in which more

densely myelinated tissue stains darker,

there may be a level of nearly complete

staining opacity reached at an early devel

opmental stage in areas that exhibit max

imal myelination whereas this level of

staining opacity may never be achieved by

very poorly myelinated areas. In this case

the apparent heterochrony could be largely

illusory. Alternatively, the deposit of my

elin on axons destined to be heavily myelin-

ated could take place at an absolutely faster

rate (which appears to be indicated by the

data of Yakovlev and Lecours, 1967) and

could begin earlier. It is probably an

important corollary that the most heavily

myelinated fibers are also often the largest

diameter axons that project relatively long

distances or to highly specific targets.

There is also an important correlation

between brain size and primitive

advanced

comparisons. Small insectivore and rodent

brains appear to be generally less myelin-

ated overall and exhibit less myeloarchi

tectonic differentiation from cortical area

to cortical area than the larger brains of

primates (Sanides, 1970). Therefore if

myelination is to be used as a ruler of pro

gression then the

least

myelinated areas

should be considered the more primitive

areas and the

most

myelinated areas the

more advanced cortical areas (Bishop,

1959). This runs exactly counter to the

apparent developmental progression.

This apparent phyletic myeloarchitec

tonic trend as well as other architectonic

trends led Sanides (1 969, 1970, 1972) to

propose that the terminal addition of cor

tical areas in mammalian evolution might

actually be the reverse of that proposed by

the traditional theorists—progressing from

association areas to secondary areas to pri

mary areas as the most highly developed.

Although not a recapitulationist theory,

because it makes no claims for develop

ment, it does nonetheless assume terminal

addition in a phylogenetic sense, and an

implicit progression from primitive to

advanced forms based upon the addition

of more highly differentiated structures.

The neuropsychologist Jason Brown (1 977,

1988) has elaborated a self

consciously

recapitulationist theory of cortical func

tion that is based upon Sanides' model of

phylogenetic progression, in which

"microgenetic" processes of sensory anal

ysis proceed in a series of stages from lim

bic to association to specialized cortical

areas rather than the other way around, as

is suggested in traditional models.

Both interpretations are based on mis

leading correlations that confound a num

ber of factors. The correlation between the

density of myelination in adults, the initi

ation and rate of myelination in develop

ment, and the overall level of myelination

in brains of different sizes suggests that all

of these apparent comparative trends may

have less to do with any evolutionary

sequence and more to do with certain con-

servative functional and metabo1ic rela-

tionships between axons of differing sizes

and the glia that form their myelin sheaths.

The developmental differentiation of the

brain occurs in the process of increasing

the size of the brain. The evolution of

increasingly differentiated mammal brains

also correlates with the evolution of

increasing brain size. The similarities

between brain development and brain evo

lution may largely be the result of these

parallelisms. In both, the confusing role of

brain size and its many correlative struc

tural scaling relationships underlies many

of the misidentifications of evolutionary

progress. The nature of some of these

underlying correlations will be the subject

of the next section.

BRAIN

SIZE

The influence of anthropocentrism

The significance of brain size is at once

the most broadly debated issue in the study

of brain evolution and probably also its most

ubiquitous, misunderstood and troubling

feature. More has been written about brain

size than about any other topic concerning

brain evolution. Like the notion of evo

lutionary progress, interest in brain size

owes much to its apparent importance for

understanding human brain evolution.

With a brain roughly three times larger

than a primate of our size should possess,

it is natural to assume that brain enlarge

ment must hold the key to human unique

ness. But is brain enlargement symptom or

cause in this transformation? Is relative

brain size alone the significant difference

or is it a superficial consequence of more

fundamental changes in brain organiza

tion? Size is also the easiest feature of the

brain to study and so lends itself to broad

comparative studies and studies of other

possible correlates to brain size that might

be of interest to researchers who otherwise

have little neuroscientific training. But the

tendency to terminate the analysis of brain

difference with measurements of brain size

is a significant impediment to progress in

the investigation of brain evolution, pre

cisely because issues of brain size are insep

arable from issues of function and internal

organization in some very fundamental

ways.

The role of brain size in brain evolution

appears deceptively simple. If the brain is

a computing device of some kind, then an

increase in component processing ele

ments should correlate with an increase in

information processing capacity. This intu

ition appears to be borne out by the unsys

tematic observation that species with brains

at the small end of the size range for ver

tebrates exhibit mostly simple and stereo

typic behaviors compared to those with

brains toward the large end of the spec

trum. This observation is not totally sat

isfactory. Some mammal brains exceed the

human brain in total volume and neuron

number

for example, elephant and whale

brains. However, these species represent

the largest mammals on earth. If we think

of the brain as a computer and the body

as the many users that are on line, vying

for processing time and memory storage,

then it becomes obvious that the effective

information processing capacity available

for any particular function is constrained

by the number of competing demands from

other sources. This has suggested to many

that the ratio of brain to body is a more

significant measure of available informa

tion processing capacity. Despite the fact

that they have absolutely larger brains, ele

phants and whales have a much lower ratio

of brain to body than humans. But this

observation is also unsatisfactory. Very

small birds and mammals have a higher

ratio of brain to body size than humans

and an even higher ratio of neuron num

ber to body size.

A satisfactory account of comparative

brain size that ranked humans on top (and

thereby preserved the intuition underlying

the Anthropocentric Maxim) was discov

ered at the end of the 19th century in the

form of allometric analysis. Since that time

the use of

subtraction criteria

based on

empirical brain and body size trends have

been assumed to define that portion of the

brain mass that is functionally correlated

with information processing demands of

the body. Deviations from these trends have

been assumed to respectively indicate

excess or dearth of mental capacity (but

see criticisms below). The importance of

allometric analysis for the investigation of

patterns of relative growth and the sec

ondary correlates of differences in size is

paramount, not just as a criterion of sub

traction, but as a tool for drawing attention

to the ways that biological processes can

be affected by changes of scale. Allometric

analysis has produced some remarkable

insights into the problem of relative growth

and has demonstrated some remarkably

regular patterns of size

related brain vari

ation from species to species and from evo

lutionary epoch to epoch.

The more detailed pursuit of allometric

relationships has shown that differences of

brain size have consequences at every level

of neuroanatomical and neurophysiologi

cal organization. Not all features of the

brain scale isometrically with size changes

in brain evolution. This has serious func

tional consequences that have yet to be

appreciated

much less understood

and

serious methodological consequences for

comparative studies because

immensely

complicates the task of determining

homology. A failure to appreciate the

numerous architectonic and functional

consequences of differences in brain size

lies at the heart of numerous misunder

standings about brain evolution, including

issues of progression and the question of

brain size evolution itself.

Is there a trend toward increasing

encephalization?

One of the earliest discoveries concern

ing brain evolution in mammals was that

mean brain size and relative brain size have

increased with respect to our reptilian

ancestors and have increased since the

beginning of the great eutherian mammal

radiations. This seems to be a clearly pro

gressive trend and suggests that brain size

itself may be an adaptation under selection.

But is it? If it is, then is there selection for

total brain size or for relative brain size?

And what kind of evidence would be nec

essary to demonstrate one or the other?

The accepted answers to these questions

have been phrased in terms of the evolu

tion of intelligence, and have changed little

in form since well before the turn of the

century.

Brain enlargement, both in absolute and

relative terms, has typically been referred

to as

encephalization.

Beginning with the

work of Dubois (191 3) the term became

associated with a specific mathematical

index: the measure of the relative devia

tion of the ratio of brain to body size in a

species from the expected ratio for an aver

age animal of the same body size, based on

trends for a given taxonomic group or for

some baseline comparison group (Bonin,

1937; Dubois, 19 13; Gould, 1966; Jerison,

1973: Stephan, 1969). Two major appli

cations of these measures of encephaliza-

tion are the assessment of taxonomic and

phylogenetic differences of relative brain

size and the assessment of species differ

ences in intelligence. Whether or not there

is any meaningful relationship between rel

ative brain size and intelligence is open to

serious question. There are distinctly dif

ferent brain

body trends for different

mammalian taxa that can all provide equally

valid

but not concordant

measures of

encephalization for an individual, the

somatic fraction of brain size that is pre

sumed to be allometrically

subtracted

the estimation of encephalization cannot

be a simple linear factor as is often assumed,

and deviations from the empirical trend

cannot automatically be assumed to cor

relate with mental adaptation as opposed

to metabolic, ontogenetic or somatic adap

tations (Deacon, 1990a,

b).

More impor

tantly, the assumption that intelligence—

much less comparative intelligence

is a

single measurable scalar quantity is highly

dubious from either a neurological or an

evolutionary perspective and has never

been adequately supported by comparative

intelligence testing (MacPhail, 1982; Gard-

ner, 1983; Hodos, 1988; Deacon, 1990a).

This issue has been extensively discussed

and debated elsewhere and so will not be

reviewed here, except to point out both

the obvious progressionist and anthropo-

centric assumptions that are inextricably

bound up with the entire enterprise. Only

the issues surrounding phylogenetic inter

pretations will be considered here.

Since the appearance of the first verte

brates, brains and bodies in many lineages

have enlarged by orders of magnitude. The

relative sizes of brains and bodies have also

changed. For an animal

of a given body size

the ratio of brain to body size has also

increased in a number of lineages, though

not all. These general trends are capped

by the evolution of mammal brains. The

largest brains ever to have existed are now

possessed by whales and the most extreme

values of encephalization ever to have

existed are now exhibited by dolphins and

primates

particularly humans. Within any

taxonomic group of living vertebrates there

is a negative allometric relationship

between brain size and body size, such that

adult individuals with larger body sizes tend

to exhibit lower ratios of brain to body size

than smaller individuals. The scaling rela

tionship is approximately linear when ren

dered in logarithmic coordinates but the

particular slope and y

intercept of this line

Allometric patterns of comparative vertebrate brain size and mammalian brain growth. Graph A,

on the upper left, depicts convex polygons that enclose all points of brain and body size for four major

vertebrate classes (from Jerison, 1973)

along with an approximation of the polygon that would have enclosed

the stem mammals. Note the incredible mammalian brain and body size expansion from this precursor group.

Note also the overlap of teleost fishes and lizards and of birds and mammals and the lack of a scala naturae

trend from fish to mammals. Graph

on the upper right, shows the trend line for carnivores and for the

domestic dog breeds. Arrows indicate measures of encephalization or somatization for a small dog with respect

to the carnivore trend. This is intended to show that

encephalization

differences do not necessarily imply

selection on brain traits. Graph C, on the bottom left, shows an ontogenetic developmental trajectory for

brain and body growth in two typical mammals of different body size. The prenatal phase overlaps completely

for most species. Graph D, on the bottom right, shows ontogenetic curves for different mammalian species

and their relation to interspecific trends. Note that ontogenetic lines overlap for different size species during

the early ontogenetic phase. The shift of ontogenetic curves that distinguishes primates, cetaceans and ele

phants from other mammals is shown in gray.

differs depending on the taxonomic group

under consideration. Proposals for

explaining the regularity and slope of these

trends include the possibility that brain size

tracks body surface (Snell, 189 1

;

Dubois,

19 13; Jerison, 1973), that brain size is con

strained by metabolic capacity which is also

negatively allometric (Martin, 198 1

;

Arm-

strong, 1983), that because brain is derived

from embryonic ectodermal tissue, the

same growth control mechanisms may con

trol mitosis in brain and body surface struc

tures (Deacon, 1990b), or that target body

size is controlled by neuron number via a

mechanism common to all mammals, and

possibly all vertebrates (Deacon, 1990b).

Ultimately no single explanation can

account for the substantial differences in

scaling relationship exhibited at different

taxonomic levels of analysis. Within a

species the allometry is strongly negative

with slope on the order of 0.1 to 0.2,

whereas within a whole order or whole ver

tebrate class the allometric slope is often

in the range of 0.6 to 0.8. See Figure 5a

and b.

The brain

body allometries of the dif

ferent living vertebrate classes appear to

be distributed bimodally. The homeo

thermic classes

birds and mammals—

tend to scale together and the poikilother-

mic classes

fish, amphibians and rep

tiles

tend to scale together, with homeo

therms exhibiting a much larger

percentage of brain to body at any given

body size (see Fig.

5a).

In the evolution of

birds and mammals there has clearly been

an increase in encephalization over the

ancestral reptilian condition as repre

sented by modern reptiles. But using mod

ern species as exemplars for ancestral rela

tionships it cannot be said that there has

been a steady increase in encephalization

from fish to amphibians to reptiles to birds

and mammals. In fact, some of the most

primitive

fishes, the sharks and rays,

comprising the Chondryichthes, exhibit

levels of encephalization that exceed all

other fishes, amphibians and reptiles, and

overlap the ranges for birds and mammals

(Bauchot et al.,

1979).

There is no clear

scala naturae of encephalization. In con

trast, relative conservatism of the enceph-

alization relationship is demonstrated by

the extensive overlap of encephalization in

different vertebrate classes.

Discerning the progressive encephali

zation of mammals relative to their reptil

ian ancestors involves more subtle distinc

tions, but the major step across this gap

appears to have already been taken by the

time of the common ancestor of metatheri-

ans and eutherians (Ulinksi,

1986).

In com

parison with other mammalian lineages the

marsupials and the insectivores appear to

occupy the low end of encephalization, but

the monotremes appear on a par with the

mean for eutherian mammals. If the small

brained basal insectivores and marsupials

are characteristic in this regard of most

stem mammalian groups then most other

mammalian lineages have exhibited pro

gressive encephalization. Such a trend is

demonstrated in the fossil record and has

been used as support for the claim that

there has been a progressive trend toward

increased intelligence in all these lineages

(Jerison, 1973).

The relatively lower encephalization of

many small bodied forms including basal

insectivores and rodents, may reflect ges

tational trade

offs involving large litter sizes

and rapid reproductive rates characteristic

of small r

selected species. These repro

ductive specializations have been corre

lated with gestational constraints that affect

brain growth (Martin, 1983;

Deacon,

1990b).

Similar factors may also be impor

tant for understanding marsupial brain size

development. It is possible that low esti

mates of

basal

encephalization may be

misleading if they incorporate superim

posed developmental trade

offs that sec

ondarily reduce encephalization. We

should expect that progressive removal of

these constraints in lineages radiating into

selected, larger

body

size niches might

result in an apparent

rebound

encephalization. A

similar

rebound

effect

has been suggested by Gould

(1975)

and

Deacon

(1990b)

in response to intense

selection for increasing body size in evo

lution. Breeding experiments demonstrate

that selection on body size produces body

size increase in successive generations with

little correlated increase in brain size

(Atchley, 1984; Riska et al., 1984).

Deacon

(1990b)

argues that after an initial rapid

evolution of increased body size effected

by modifications of peripheral hormonal

mechanisms, continued stabilizing selec

tion would tend to produce complemen

tary brain size increase as a more stable

central determiner of target body size.

Thus the rapid radiations into more

selected, large

body

size niches that

characterized many mammalian lineages in

the Eocene may have provided a biased

sample of species for comparison to more

modern lineages.

The case for an increase in encephali

zation in primates, elephants and cetaceans

is strengthened by independent evidence.

From very early in embryogenesis all these

species exhibit approximately double the

ratio of brain to body size found in any

other mammal group at a comparable

developmental stage (Count, 1947;

Sacher

and Staffeldt, 1974; Martin, 1983; Deacon,

1990b).

This difference is evident at the

earliest stages in which brains are discern-

able in the developing embryo and so rules

out explaining this encephalization in terms

of the terminal addition of neural tissue

late in development (see Fig. 5c, d). Curi

ously, the trajectory of brain

body growth

for all three groups closely overlaps for the

entire fetal period, suggesting a common

brain growth mechanism. The brain

body

growth trajectories of the remainder of the

eutherian mammals also all appear to share

a common fetal trajectory, suggesting that

they all share a different common mech

anism for determining brain growth. As

with the encephalization difference

between homeotherms and poikilotherms,

this.

embryological encephalization differ

ence distinguishing primates, dolphins and

elephants from the remainder of the

eutherian mammals appears as a distinct

discontinuity without intermediates.

As a final comment on the encephaliza

tion issue it should be pointed out that the

term itself reflects an underlying bias that

is part historical and part theoretical. Early

writers often did not clearly distinguish

absolute brain size from the relation of

brain size to body size in their discussions

of brain evolution (Gould, 198 1). The

ulti-

mate interest has of course all along been

the explanation of human brain size. But

measures of brain size with respect to body

size are inherently relational. Increased

encephalization is also decreased somati-

zation, and vice versa (see Fig. 5b). One

need not necessarily assume neurological

explanations for differences in this rela

tionship. Although breeds of dogs differ

enormously in degree of encephalization

(from small highly encephalized dogs to

large poorly encephalized dogs) no one

doubts that body size is the selected vari

able and brain size the relatively less flex

ible parameter. In fact, breeding experi

ments selecting progeny on the basis of

either extremes of brain size or extremes

of body size demonstrate that selecting for

body size produces a poor correlated

response in brain size whereas selecting for

brain size produces a highly correlated

response in body size (Roderick et al., 1976;

Fuller, 1979; Atchley, 1984; Riska et al.,

1984; Kruska, 1987). Given that a large

fraction of the variance in

encephaliza

tion" within a species can be a consequence

of selection on body size, why should this

not also hold for cross

taxa comparisons?

Body size is a highly flexible and ecologi

cally significant variable, whereas brain size

is a relatively inflexible and as yet poorly

understood variable that may or may not

correlate with differences in behavior. The

search for changes in brain structure that

correlate with addition or subtraction of

neural tissue in evolution in order to

account for encephalization is for this rea

son probably misguided; we might just as

well look for addition or subtraction of the

many parts of the rest of the body.

Disentangling allometry and progression

When organisms get larger, either dur

ing development or in the course of evo

lution, all features of the organism are not

scaled up isometrically. Even more trou

blesome for the comparative anatomist is

the fact that homological relationships can

appear to change with evolutionary changes

in size. Different structures may radically

change size with respect to one another or

alter their relative position, structures may

radically change shape due to unequal

growth rates among their parts or single

structures may divide or differentiate to

become two or more distinct structures.

This is the source of one of the most insid

ious problems in evolutionary theory: the

confusion of size related changes with evo

lutionary advancement. The secondary

effects of change of size can include the

apparent addition of new structures to old

or the appearance of increased complexity

in existing structures. It may be difficult to

tell whether an increase in size has caused

old structures to differentiate and subdi

vide or whether the addition of new struc

tures has caused an increase in size.

As D'Arcy Thompson pointed out in his

classic treatise on the effects of growth on

form (1 917), most mechanical forces,

material properties and structural relation

ships do not change isometrically with

changes in size. Geometric effects are most

obvious — e.g., surface area to volume— but

also there are changes in the relative vis

cosity of fluids, diffusion rates of mole

cules, structural plasticity or rigidity of

materials, rates of chemical reactions, etc.

Many of these non

geometric scaling allo-

metries result from the fact that ultimately

some components of organisms are of fixed

sizes

(e.g.,

molecules and cells). In order to

maintain isometry of functional properties

across major changes in size it is nearly

always necessary for structures to enlarge

at different rates.

Allometric analysis can help control for

the influence of size, allowing one to com

pare quantitative traits with the effects of

size subtracted. This

criterion of subtrac

tion

is most often assumed to indicate that

some functional relation has been main

tained in the face of the change in size.

This does not mean that such changes are

merely passive effects induced by the mass

of the organism. They are inevitably

internal

facultative or genetic adapta

tions to the imbalances or weaknesses

induced by the change in size. Many of

these secondary adaptations may be

encoded in the genome of the organism.

However, a trait that is somehow expressed

facultatively, in a size graded manner,

would have obvious advantages over one

that is specific to a given range of sizes. In

these cases size change should be consid

ered the

primary

adaptation and the

correlated reorganizations of structure can

be considered

secondary.

genetically encoded, size

correlated

trait that has evolved in response to the

functional demands of size change is a par

adigmatic example of a derived condition.

But it also represents a conservative fea

ture to the extent that it is necessary to

preserve some ancestral functional rela

tionship. Functional homology is main

tained at the expense of structural homol

ogy. The reverse scenario is also possible—

structural homology maintained by virtue

of change in supportive functions to keep

pace with the effects of size. Finally, it is

also possible for size change to be the

sec

ondary

adaptation. A change in size can

be secondary to the production of some

correlated effect that has itself become the

trait of primary adaptive significance. This

latter possibility will be suggested in the

case of human brain size enlargement (see

last section).

In cases where size change is primary it

would be inappropriate to consider the

many secondary adaptations as progressive

trends. The fossil evidence clearly dem

onstrates that with the demise of the dino

saurs, small mammal species rapidly

adapted to fill niches for large bodied

forms. The mammalian radiations can be

seen as markedly asymmetric with respect

to body size (and correlated brain size). The

lower limit of mammalian body size has

probably not been significantly altered since

the Paleocene but the upper limit has prob

ably been extended about a millionfold

compared to a typical basal insectivore!

Brain size necessarily followed this trend,

although the extension of the upper limit

of'brain size, due to its negative allometry

with respect to body size, has probably not

exceeded ten thousand times that of a typ

ical basal insectivore brain. Parallel

enlargement trends characterized numer

ous lineages of mammals.

I will argue that it is this remarkable par

allelism, and not some progressive selec

tion for increasing intelligence, that is

responsible for many pseudoprogressive

trends in mammalian brain evolution.

Larger whole animals were being

selected

not just larger brains—but along

with the correlated brain enlargement in

each lineage a multitude of parallel sec

ondary internal adaptations followed.

Allometry of brain traits any levels

Mammalian brains range in weight from

around a gram to nearly ten thousand

grams. Even within a single lineage like

the

anthropoid primates, in which individual

species share many strong similarities in

brain structure, there is more than

hun

dredfold difference between the smallest

and largest adult brains. Yet most micro-

structural features change little in size from

brain to brain. The maximum sizes of neu

ron and glial cell bodies increase slightly

from the smallest to the largest brains, but

nowhere near the thousandfold scaling of

the brains they comprise (Haug,

1987).

The

functional constraints on cell volume no

doubt set an asymptotic upper bound on

cell size that the largest neurons are likely

approaching. The apparent tetraploidy of

the giant Betz cells of the human motor

cortex likely indicates that these cells are

already forced to come up with unusual

ways to circumvent certain functional lim

itations of their large size.

The constraints on neural size also affect

the scaling of higher

order multi

neuronal

structures. For example, the diameter of

cortical columns as well as the number of

neurons within each column seems to

remain almost constant across brain size

variation (Rocket

et al.,

1980). This may

also be the reason that many larger scale

morphological features like cortical thick

ness increase only slightly from the small

est to the largest brains (Rocket et al., 1980).

cdnsequence the disparity between

microstructure organization and macro

structure organization grows incredibly

with increasing size. Since the macro

morphology of the brain, including

distinct homogeneous structures and their

various functional subdivisions, is derived

by ontogenetic processes that function at

the microscopic cellular level, this growing

scale difference is inevitably reflected by

changes of large scale structure. Some of

these. morphological changes may reflect

distinct adaptations to these new micro-

structural demands, but it is likely that the

majority are simply the inadvertent con

sequences of the same ontogenetic pro

cesses operating in a vastly larger brain.

This is illustrated by what can be called

cytoarchitectonic and myeloarchitectonic allom-

etry (Deacon, 1990a). The linear and vol

umetric increase in scale of the brain with

respect to relatively more conservative lim

its of cellular structure impose new con

straints on neural and glial functions. Con

sider what must happen to homologous

long projection neurons in brains of

increasing size. In order to maintain sim

ilar transmission velocities and transmis

sion integrity longer axons need to have

larger diameters and thicker myelin

sheaths. If the target area has also expanded

in volume (as is typically the case as well)

then the terminal arbor of a typical axon

must also increase. This can be a significant

factor given the exponential differential

between linear dimensions and volumes

since it requires a tremendous increase in

length and number of branchings of an

axon to fill a larger volume with the same

density of synapses. To keep pace with the

projection neuron rnyelin sheath

glial cell

smaller brain

larger brain

rnyelin sheath

local circuit neuron

(e.g. granule cell)

Local relationships contributing to cytoar-

chitectonic allometry are depicted in this figure from

Deacon

(1990b).

The top figure depicts the situation

in a relatively small brain and the lower figure depicts

these same relationships in a slightly larger brain. In

small as compared to large brains projection neurons

possess short, small diameter axons, with relatively

less myelin, smaller cell bodies, lower neuron to glia

ratio, higher neuron densities, lower ratios of local

circuit neurons to projection neurons, smaller size

differences between the smallest and largest neurons,

etc.

metabolic and neurotransmission demands

of such an enlarged axonal volume and sur

face area the cell body of the neuron must

also be enlarged, but since it depends on

the glial cells surrounding it for its meta

bolic support the relative number of glia

must also increase. The same constraints

are not experienced by small local circuit

neurons. Since local circuits remain rela

tively constant in volume (increasing no

more than a few hundred percent across

huge brain size differences) these neurons

should have to change relatively little to

compensate for size (see Fig.

6).

Given these two extremes, it becomes

obvious that the local cyto

and myelo-

architecture of many brain structures will

reflect the influence of size. In general, in

large brains as compared to small brains

there should be a number of regular trends:

In large brains there should be (1) some

much larger cell types, but also a much

greater difference between the smallest and

largest cells,

(2)

a higher glia to neuron

ratio in most regions,

(3)

a decreased mean

density of neurons but an increase in the

range and variation of densities in different

areas and subareas,

(4)

a significant increase

in axonal and dendritic arborization to fill

the slightly increased volume of local cir

cuits,

(5)

higher levels of myelination in

general but a greater difference from the

most to least myelinated areas and sub

areas, and

(6)

since some areas may be spe

cialized for longer projecting cells with

large soma and heavily myelinated axons

and other areas only for short projecting

cells or cells with small diameter, poorly

myelinated axons these differences will be

magnified between brain areas. The net

result will be greater architectonic differ

entiation within an area and between areas

in large brains as compared to small brains.

Few

if any of these changes are likely the

result of the evolution of new ontogenetic

mechanisms, but merely the local dynam-

ical responses of cells trying desperately to

do what they would do in any brain. We

can conclude from this that

an increase in

architectonic complexity with size is not a reli

able measure of progression and advancement,

either for comparison of the brains of different

species or for comparisons of different areas

within the same brain.

However, there are also a number of size

related changes for which architectonic

reorganization will be unable to compen

sate. Probably the most significant among

these is a factor that can be called

network

allometry

(Deacon, 1990a). Network allom-

etry is essentially a geometric principle

analogous to surface to volume allometry.

The size of a network is a function both of

the number of nodes in the network and

the number of connections between nodes.

Networks with every node directly con

nected to each other node can be consid

ered to exhibit a high level of connectivity

whereas networks with each node directly

connected to only one or two others can

be considered low connectivity networks.

One measure of the average connectivity

of a network might be the average number

of nodes that must be passed through to

find a link between any two arbitrary nodes.

In all but the smallest or lowest connectiv

ity networks the number of connections

tends to vastly outnumber the nodes. If one

wants to increase the number of nodes in

a network while maintaining the same

average connectivity, then the number of

connections that have to be added with each

new node will grow factorially with each

addition. A factorial increase of this sort

will rapidly lead to astronomical numbers,

particularly in large highly connected net

works, but even in networks with relatively

low connectivity very large changes in size

will produce the same result. Except in

minimally connected networks, it will

become increasingly difficult to continue

increasing the number of nodes within a

network and retain the same level of con

nectivity. This concept in diagrammed in

Figure

Now consider these facts about the cen

tral nervous system: One neuron may be

connected to thousands of others, accord

ing to some estimates; even in small mam-

malian brains there are probably b

lions

of neurons; and there is a nearly tenthou-

sandfold difference in volume between the

smallest and largest mammalian brains.

These simple statistics make it clear that

network allometry must be one of the major

factors responsible for the differences in

organization and function that distinguish

large and small brains. But the brain is not

a maximally interconnected network. Even

in a brain with a billion cells each con

nected to a thousand others at random the

average number of nodes se

aratin

any

two will be on the order of twenty, and

brains are not nearly so diffusely orga

nized. Probably the majority of connec

tions between areas of mammalian isocor-

tex and other cortical or subcortical areas

are reciprocal, and the connections within

the local circuits of the isocortex are prob

ably highly re

entrant and relatively self

contained within columnar modules. Also,

given the relative comparability of local

circuit structure of isocortex from species

to species, it is likely that intracortical con

nections between columns may be limited

to neighboring columns and to specific tar

get columns in other areas. It may then be

a bit more useful to think of connectivity

within the cerebral cortex in terms of

columnar modules as nodes rather than in

terms of individual neurons as nodes.

separate network allometry might apply at

the columnar level since columns increase

in volume, number of axons and dendrites

but not neurons with increasing brain size.

Nonetheless, given the immense differ

ences in scale that must be considered, even

a relatively poorly interconnected cortical

network

will

have to contend with connec

tivity trade

offs in order to compensate for

network allometry.

There is clearly a significant increase in

the proportion of white matter to gray

matter in brains of ascending size, but this

is doubtless nowhere near what would be

required for connectional parity to be

maintained.

In order to evolve to significantly

larger sizes brains must decrease connectivity.

This trade

off undoubtedly has its costs.

The two most obvious costs of decreasing

connectivity with increasing size are

reduced integration of distributed func

tions and significantly increased transmis

sion and processing times.

Larger brains are

not necessarily more efficient and more powerful

than smaller brains.

In fact, these new func

tional costs of increasing size will demand

new secondary adaptations in order to

compensate in other ways. If a functional

area of cortex becomes enlarged in the

course of evolution the mean interconnec

tivity of its columns will decrease. This will

decrease the homogeneity and integrity of

activity patterns that can be maintained

within it and increase the time necessary

for neural

calculations

involving the

whole area to be completed. These costs

can be minimized by breaking the one large

area into two relatively independent sub

areas capable of processing the same infor

mation in parallel, so long as they can be

partially integrated with one another by

specific interconnections. Although this

reorganizational strategy may compensate

in part for loss of local integration and pro

cessing efficiency it cannot entirely com

pensate. And, if size continues to increase,

additional parcellations into multiple areas

network allometry network allometry

maintaining

local maintaining global

connectivity only connectivity

The problem of network allometry is rep-

resented by the example of a very simple network

(figure from Deacon, 1990b). A series of nodes

(depicted as spheres) is connected reciprocally to each

other (depicted by double arrows) in different size

networks with different extremes of connectivity.

Networks on the left exhibit low connectivity and

those on the right exhibit maximum connectivity.

total number of nodes;

total number of recip

rocal connections (note that in the nervous system

reciprocal connections are separate connections); Xn

the number of other nodes to which any one node

is directly connected; Xc

the mean number of con

nections intervening between any two arbitrary nodes.

The growth of connections to nodes is

factorial

function of the number of nodes in

fully connected

network and a linear function of the number of nodes

in a minimally connected network. Both low and high

connectivity networks require major functional and

structural trade

offs with size increase.

are necessary and both transmission time

and integration costs will continue to

mount. Of course there may also be advan

tages, including the increased specificity

and reliability afforded by parallel redun

dant processing, or alternatively, the pos

sibility of subspecialization of different

subdivisions. The evolution of new struc

ture and function as a result of such pro

cesses will be discussed further in the next

section.

We can conclude that network allometry

may force a variety of secondary reorgan

izations of cortical architecture, including

parcellation and multiplication of func

tional areas, as brains enlarge during evo

lution. This essentially forces large brains

to alter functional strategies for informa

tion processing from those effective in small

brains. A

further factor to be considered

is the fact that the receptor systems pro

jecting to these cortical areas are also

enlarging along with body and brain size,

although these probably exhibit a negative

allometry with body size. This must also

play a significant role in determining at

what level of scale there

will

likely be

breakup and parcellation of cortical areas.

It is far from clear to what extent there is

net gain due to new functionality and

increased information storage or net loss

of efficiency and integration due to increas

ing loss of connectivity as brain size

increases. We will need a better under

standing of this trade

off before we will be

able to think clearly about the question of

comparative intelligence and its relation

ship to brain

body allometry.

Size and

parallelism

in mammalian

cortical evolution

These complex allometric consider

ations complicate the evolutionary inter

pretation of comparative brain morphol

ogy. It is not a simple matter to track

morphological changes and assign them

independent evolutionary causes. If in fact

many

emergent

architectonic changes

associated with brain size evolution are

simply the effects of common underlying

cellular mechanisms compensating for the

effects of size, then it can be misleading to

treat them as "new" features. The only

difference in the information utilized in

the

ontogenetic process is difference in size

information

in the form of larger axonal

volumes, greater variance of metabolic

demands for different cell types, etc. If we

focus on the deep informational homolo

gies rather than on the surface structural

homologies it is clear that developmental

mechanisms have not been altered, rather

one of the contextual variables to which

these mechanisms is sensitive has changed

value. These changes might be referred to

as secondary

facultative

adaptations to dis

tinguish them from secondary adaptations

that actually involve the evolution of new

genetic and developmental information.

There is neither progression nor addition

in this sense, and the parallelisms that result

are not properly thought of as parallel

homoplasy.

It is possible that the ontogenetic mech

anisms utilized in parcellation of cortical

areas are sensitive to the demands of net

work allometry. The fact that axonal com

petition plays the major role in determin

ing area parcellation and afferent and

efferent relationships within the develop

ing cortex (discussed in the next two sec

tions) suggests that facultative mechanisms

may account for a considerable portion of

the structural and connectional response

to this functional demand. But it is likely

that specific genetic adaptations also

become available to streamline the facul

tative response (via genetic assimilation) to

these demands during the course of evo-

lution. The addition of new secondary

adaptations (derived conditions) is in order

to maintain the same function (conserved

condition), and can be seen as

sort of

Red Queen effect

to the extent that the

system is working harder and harder to try

and stay in the same place. The parallel

isms that have evolved to maintain func-

tional homology across a large range of

brain sizes are likely the result of a com-

bination of underlying ontogenetic homol

ogies shared by all mammals and specific

genetically encoded biases that modify the

responses of these ontogenetic mecha

nisms differently in different species.

But it is not necessarily safe even to con

sider these microallometric changes as

facultative adaptations with respect to size

change. All that is demonstrated is a form

of morphogenetic plasticity that is affected

by size. In a related context, Smith

Gill

983)

distinguishes two general classes of

developmental plasticity that clarify this

point. The first he calls developmental con-

version and the second he calls phenotypic

modulation. In developmental conversion,

environmental cues activate alternative

genetic mechanisms that are expressed in

the organism's development. These differ

ent genetic expressions may produce alter

native morphs by activating or inhibiting

growth processes affecting the structural

development

of certain tissues, by chang

ing cell surface affinities or messenger-

receptor sire relationships in intercellular

communication, or even by inducing

regressive processes, such as programmed

cell death. In phenotypic modulation envi-

ronmental cues modulate but do not select

among'or alter genetic programs. This

produces variation and adjustment of the

expression of genetic information but not

the selection of different alternative genetic

programs. Smith

Gill notes that pheno

typic modulation does not necessarily imply

an adaptive response, ". . . adaptiveness of

phenotypic modulation cannot be assumed

unless specific genetic mechanisms can be

demonstrated."

It is unlikely that the onto-

genetic responses of neural tissues to the

influence of size are produced

specific

genetic alternatives, since these would have

to differ for each range of size and for each

brain' region. Rather, the facultative plas

ticity of neurons in response to the local

effects of size must be a case of phenotypic

modulation. We cannot necessarily assume

that all aspects of this plasticity are adap

tive, even in the broad sense of adaptive

with respect to local metabolic and infor

mation processing demands. We can only

assume that significantly maladaptive plas

tic responses will be strongly selected

against.

In conclusion, the evolution of mammals

is clearly characterized by a trend toward

increasing body size with a correlated

increase in brain size, but it is unclear to

what extent there has been additional inde

pendent selection for increased brain size

and brain differentiation in different lin

eages. Even if there is not independent

selection for brain size in a particular lin

eage, body size correlated increase in brain

size can be expected to produce a series of

architectonic and functional changes due

to the plasticity of developmental processes

at the neuronal

synaptic level. In general,

with respect to brain structure, we should

question the assumption that size increase

is caused by addition of new parts, since

the plastic responses of neural develop

ment to size change inevitably produce dif

ferentiation and subdivision of existing

structures at all supracellular levels of

organization. Although the majority of

architectonic trends in brain organization

in mammalian lineages give the appear

ance of increasing differentiation and com

plexity, we cannot disambiguate this from

the effects of local cellular plasticity and

secondary facultative adaptation which

cannot be considered progressive in any

sense. However, precisely because plastic

phenotypic modulation is not necessarily

adaptational, it cannot be assumed that it

will preserve any semblance of functional

isometry. And even if it is adaptational in

most cases it may fail to be so at the

extremes of size, where otherwise predict

able metabolic and information processing

demands may significantly diverge from

ancestral patterns. If such inadvertent

departures from functional isometry con

tribute useful capabilities or potentialities

they may contribute to directional trends

in evolution. Alternatively, if the adapt

ability of facultative responses induces

genetic assimilation of nongenetic pheno

typic modulation mechanisms into genet

ically based developmental conversion

mechanisms, the changes in response to

size may produce irreversible evolutionary

changes. In this way secondary adaptations

to size may inadvertently provide the raw

materials for the evolution of new func

tional systems.

NEOGENESIS: T

MERGENCE

TRUCTURE

The assumption that new adaptations

require new structures

In many ways the fundamental question

that evolutionary theory purports to answer

is how new species with novel structures

and functions come into being in the course

of time. If anything can be called evolu

tionary progress it is the creation of totally

new adaptations, not just the augmentation

of existing adaptations. New brain areas

with distinct cellular architecture and con-

nectivity appear in some lineages but not

others, and it is almost certain that over

the course of mammalian brain evolution

the number of discrete brain areas in the

most complex brains has steadily increased.

This has suggested to many that new adap

tations and the augmentation of existing

structures are accomplished by the addi

tion of new structures to an already func

tioning brain. But new structure may also

evolve by co

opting or reorganizing exist

ing structures in some way. In this case the

resulting structure may be radically differ

ent than its antecedent, and yet combine

both novel attributes and pre

existing fea

tures at different levels of organization.

How is it possible to distinguish between

uniquely derived structural additions and

previous structures that have become rad

ically modified?

The history of vertebrate evolution in

general, including mammalian evolution,

exhibits a trend toward diversification of

adaptations. The invasion of new niches

inevitably requires adaptation of percep

tual, behavioral and cognitive processes to

meet the new demands. To some extent

these are acquired at the expense of mod

ifying previous neural systems, trading one

function for the other. But the acquisition

of new abilities could also be achieved by

addition of new functions to old with a cor

responding elaboration of the brain. Each

species

the culmination of a phylogenetic

sequence of adaptive changes that leave

their traces in the structure of its body and

brain. There is a natural tendency to envi

sion this as an accretionary process that

progressively adds new structure to old.

This impression is supported by the appar

ent increase in brain complexity that cor

relates with the

scala naturae

hierarchy of

species leading from fish to mammals and

from

primitive

mammals to

advanced

mammals. The definition of evolutionary

progress is at every step completely depen

dent upon the identification of neogenesis.

One outstanding difference that is pre

sumed to distinguish primitive from

advanced mammalian brains is an increas

ing number of architectonically and func

tionally distinguishable cortical areas. The

increase in numbers of cortical areas has

often been cited as an example of both

neogenesis and of progressive evolution,

and is presumed to correlate with the

increased behavioral and cognitive abilities

of advanced species. The acquisition of new

functional abilities is also a central feature

of human mental evolution. We conceive

of ourselves as possessing all of the cog

nitive abilities of other species and then

some. Particularly novel in the course of

evolution are human linguistic abilities. In

a behavioral sense this capability is clearly

an addition

functional neogenesis. It is

tempting to assume that the addition of

such an unprecedented function necessar

ily implies the genesis of novel neurological

structures. With respect to apparent neo-

genetic trends in the evolution of cortex

in other mammals, the addition of human

language areas would seem to be a most

recent step in a long series of additions. In

this context it is clear that the assumed

ubiquity of neogenetic processes derives in

part from its presumed importance for

human mental evolution.

Additive theories of cortical evolution

The most well known theory of mam-

malian brain evolution is the

triune brain

hypothesis

proposed by Paul MacLean

970,

1973). It is an attempt to explain the dif

ference between mammal brains and non-

mammal brains and how this difference

arose in the course of evolution. MacLean

argues that the mammalian brain can be

subdivided into three functionally and evo-

lutionarily distinct regions. The first divi

sion, including the spinal cord, brain stem,

midbrain, diencephalon, corpus striatum

and olfactory apparatus are considered a

core structure common to all, terrestrial

vertebrates. He calls this the

reptilian brain

complex

because, he argues, this com

prises the entire brain in reptilian species.

During mammalian evolution two addi

tional structural levels are added in

sequence: the

paleomammalian brain,

com

posed primarily of limbic cortex and its

associated forebrain nuclei and connec

tions, and the

neomammalian brain

com

posed of the neocortex. With the accretion

of each of these systems in the course of

evolution comes the emergence of new

cognitive abilities and behaviors. With the

paleomammalian brain come complex

parental care, vocal communication, play,

and the

higher

emotions of bonding and

caring. With the neomammalian brain

come higher order perceptual, motor and

generally enhanced learning abilities that

free the organism from reliance on fixed

action patterns and simple template based

perception. The scheme is hierarchic and

accretive. The reptilian brain is whole and

complete in and of itself (and presumably

can reassert itself as an autonomous force

in cases of high excitation or weakened

control from higher brain systems, as might

occur with brain damage), and therefore

the addition of new systems does not

require major reorganization, simply

superimposition of new axons into the cir

cuitry of this otherwise complete system.

The additional connections need merely

play an inhibiting and modulatory role with

respect to the pre

existing substrate.

Although this brain model has become

widespread in the popular literature and

in some psychological and educational the-

ories, its influence in comparative neu-

roanatomy is tenuous at best. The limbic

cortical areas that presumably comprise the

paleomammalian brain have been homol

ogized to cortical structures in nonmam-

mals by comparative anatomists since early

in the century

(e.g.,

Johnston, 1906; Crosby,

19 17; Elliot

Smith, 19 19; Dart, 1934;

Abbie, 1940). In addition, more recent

investigations have demonstrated that

nonmammals also exhibit forebrain struc

tures and connections that undoubtedly

have homologues in mammalian neocortex

(e.g.,

Ebbesson, 1980; Karten and Shimizu,

1989; Ulinski, 1983). It also provides a sim

plistic view of behavioral differences

between mammals and nonmammals. It

seriously underestimates the considerable

perceptual, motor and learning abilities of

nonmammalian species

particularly birds—

and ignores the many elaborate social

behaviors, modes of communication and

parental care that have been observed in

nonmammals. Nonetheless, the triune

brain theory has enjoyed a wide audience

in large part because it captures a central

anthropocentric intuition: that terminal

addition of higher order brain structures

must correlate with the appearance of

increasingly sophisticated cognitive abili

ties and more flexible behaviors in the

course of evolution.

Ever since it first became possible to eas

ily differentiate one cortical area from

another on the basis of cell architecture or

myelin content it was recognized that the

cortex of some mammals exhibited many

more divisions into distinct areas than did

others. Efforts by the early comparative

anatomists, including Brodmann, Camp

bell, Elliot

Smith, and others, to determine

homologies between these cortical areas in

different species suggested that apparent

homologies could be identified for a num

ber of areas in most mammalian brains (see

Fig. 8). Apparently, the primary projection

areas for vision, audition, and tactile senses

could be homologized in all mammals but

the multitude of interdigitated "nonpro-

jection areas

that were evident in the

human brain could not all be homologized

to areas in monkey brains, and many of the

nonprojection areas

in monkey brains

could not be homologized to yet smaller

and more

primitive

brains.

Flechsig's (1 90 1) demonstration that this

precedence of areas was also approxi

mately paralleled by maturational trends

(see Fig.

completed the evidence for a

grand synthesis. The evolution of ever

more complex associational abilities in

mammals was enabled by the elaboration

of additional higher order association areas

of cortex that were both progressively fur

ther removed from direct peripheral input

and more interconnected with each other.

Just how these new areas of cortex became

interdigitated between old areas was not

clear to these authors, but the determi

nation of homologies appeared to require

the progressive addition of new structures

in a particular order of appearance.

One obvious way to account for new

structure and new functional abilities

appearing in the course of cortical evolu

tion is simply to hypothesize their insertion

into an already complete and functioning

cortex. The prevailing neuropsychological

theory of the early 20th century could

accommodate such a view. Like MacLean's

flattened mouse

cortex cytoarchitecture flattened macaque

cortex cytoarchitecture

8. Comparisons of the number of cortical areas of small and large mammalian brains. The four drawings

depict unfolded views of mouse and monkey cortical surfaces. A

(mouse cortex) and B (owl monkey cortex)

are redrawn from Kaas (1989) and depict area maps determined electrophysiologically in these species. C

(mouse cortex) and

(macaque cortex) are redrawn from Caviness and Frost (1980) and Jouandet

et al.

(1989), respectively, and depict cytoarchitectonic divisions (Krieg's numerical designations following Brod-

mann). The differences in configurations largely represent different

unfolding

techniques, some of which

minimize area distortion at the expense of total map integrity whereas others maintain continuity of the map

at the expense of area distortion. All are drawn to equal size for comparability.

subsequent Triune Brain hypothesis, the orderly addition of cortical areas in evo

theory

the accretion of new cortical areas lution and the progressive elaboration of

necessitated a hierarchic conception of sensory motor processes in

higher

mam

brain organization and function; Flechsig mals. The associationist assumption at the

900)

and Campbell

905)

clearly artic

center of this theory was that higher order

ulated this complementarity between the mental associations are built from lower

FIG. 9.

graphic depiction of

Campbell/Brodmann/Flechsig,

Bonin/Sanides and Lende/Poliakov scenarios

for the addition or differentiation of new cortical areas in mammalian evolution. Although none of the

individual schemes is exactly identical with any other (and may not exactly correspond with those depicted)

they have been grouped into three distinct categories for depiction because of their underlying theoretical

similarities. The relative sizes of these brains are depicted in column

and show that the increase in distin

guishable cortical areas is not independent of size. Column B shows an accretive scheme in which projection

areas

are

primitive and association areas are added derived characteristics in later brains. Column C shows a

progressive differentiation scheme in which association areas are considered most undifferentiated and there

fore primitive and more specialized sensory and motor areas are assumed to be later derived conditions that

have differentiated in a series of stages out of previous levels of association cortex. Sanides additionally argues

that there is a dual origin of most major sensory/motor fields that determines differentiation at the intersection

of a dorsocaudally originating archicortical and a ventrorostrally originating paleocortical trend. This is not

depicted here. Column

shows a progressive differentiation scheme based on the parcellation and retraction

of initially diffuse overlapping projection fields into eventually discrete non

overlapping fields. The boundaries

of the initially overlapping projection fields are depicted by dashed lines. Later parcellation and reduction of

diffuse projections within each projection field is depicted as progressively darker regions of gray. The notion

of reduction of diffuse projections within a field is from Poliakov and is not discussed as a possibility by Lende,

whose theory focused only on the earliest stages of cortical evolution in mammals.

order simpler associations, and compli

cated and flexible skilled responses are

constructed by associations between sim

pler reflex responses. They argued that

association areas of cortex only received

information from sensory projection areas

or other association areas and served as the

locus for higher

order associations between

sense data and motor programs from pri

mary areas. Association areas ultimately are

envisioned as an additional higher

order

reflex arc superimposed upon and elabo

rating existing lower level reflex arcs

(Luria, 1980; Sherrington, 1906). There

fore association areas could be added as

evolutionary after

thoughts with minimal

rewiring of other brain circuits. Each new

association area was one step removed from

the previous association area in the hier

archy and provided an additional layer of

association processes superimposed upon

an already complete functional brain (see

Fig. 9).

Vestiges of this view are still widespread.

The pinnacle of this conception of the cor

tical hierarchy is of course the addition of

language areas in the human brain

inserted into an otherwise complete and

functional ape brain. The assumption that

Broca's area for speech is a new association

area peculiar to the human brain has been

represented in a number of schemes by the

identification of some region within the

third frontal convolution of the human

brain that is assigned no homologous coun

terpart in the monkey brain. The possibil

ity that Broca's area has no prehuman

homologue and could have been simply

added on to an otherwise complete brain

seems to be taken for granted in a number

of recent discussions of human brain evo

lution (Passingham, 198 1;

Tobias, 198 1;

Falk, 1983) and language evolution (e.g.,

Chomsky, 1972). Galaburda and Pandya

(1982) and Deacon (1984, 1988a, 1990c)

provide architectonic and tracer evidence

that a homologue for Broca's area exists

in the monkey brain, despite the fact that

it plays no apparent role in vocal commu

nication in monkeys.

A similar argument was presented by

Geschwind (1964) and has been reasserted

by a number of later writers concerning

the inferior parietal lobule of the human

brain. Following the classic associationist

model of language processing, Geschwind

argues that this area plays a fundamental

role as an

association area of association

areas.

Its placement at the temporo-pari-

eto

occipital juncture seemed to ideally suit

it for associating the outputs of association

cortices from different sensory modalities.

Since Geschwind conceived of word mean

ings as complex associations between word

sounds and a multitude of other sensory

associations abstracted from sensory expe

rience, an association area of association

areas would have to be the necessary sub

strate for semantic processes. Posterior

temporal

parietal

occipital damage often

results in disturbances of semantic lan

guage comprehension. Geschwind argued

that no homologue to this highest order

association area was evident in monkey

cortices. According to this logic the addi

tion of this area during human evolution

was a necessary condition for language

evolution. Language evolution becomes the

natural end point of a progressive process

of adding association areas on top of asso

ciation areas in the course of mammalian

brain evolution. This non

homology claim

has also been contradicted by subsequent

tracer experiments (e.g.,

Mesulam et al.,

1977; Seltzer and Pandya, 1980) and ani

mal behavioral studies (e.g., Jarvis and

Ettlinger, 1977).

Recently, a more sophisticated version

of an accretion theory has been suggested

by Allman (1990) and Kaas (1987, 1989).

They argue that the multiplication

cor

tical areas in many mammalian lineages

might be explained by the duplication of

existing areas. They suggest that a new

population of neurons could appear inter

digitated in position between older popu

lations as a result of some genetic accident

that caused redundant production of the

neurons from one of these regions (it is

compared by Kaas, 1987 and 1989, to the

addition of an extra body segment in the

evolution of lobsters; Gregory, 1935).

Recent studies of genetic mutants in fruit

flies has demonstrated that at least in these

species genetic mutations can cause the

production of whole duplicate body sec

tions or limbs interdigitated between exist

ing structures. Allman (1 990) suggests

that similar genetic mutations might

underly areal duplication events in mam

malian cortical evolution. In fact, to account

for the number of such events that have

occurred this would have to be a somewhat

common sort of mutation.

It is unquestionable that new cortical

areas have evolved in some of the larger

mammal brains and that different linea

(e.g., cats and monkeys) exhibit different

spatial arrangements and functional spe

cializations of the cortical areas that have

been added (Kaas, 1989). But is

reason

able to imagine that these areas have been

literally inserted between existing cortical

areas by the addition of new neural tissue

in that position? This claim depends on the

possibility that functional areas of cortex

are modular in their construction. The

neurons that comprise a new cortical area

would need to be added along with speci

fications regarding intrinsic circuitry and

afferent and efferent connections with

neighboring areas and subcortical sites.

This requirement is contradicted by

developmental evidence that the afferent

and efferent connections of a cortical area

are not specified by the information that

is intrinsic to the cells in that area, but

rather by competitive interactions among

competing axons from many areas. Even

the transplantation of a section of cortex

into a novel position within a fetal brain

prior to the development of cortical con-

nections will not bring with it the functions

and connections appropriate to its site of

origin (O'Leary and Stanfield, 1989; Stan

field and O'Leary, 1985; see discussion in

the next section). The transplanted area

will develop functions and connections

appropriate to its new position. Thus, even

if some new area of cortex miraculously

appeared interdigitated between older cor

tical areas'in a developing brain it would

not bring with it any new connectional or

functional information.

Is there an evolutionary sequence of

new cortical areas?

An alternative approach is to conceive

of new cortical areas as differentiating out

of old areas rather than being created inde

pendently adjacent to them. This has the

attractive property that new areas should

continue to bear some functional and con

nectional interrelationship with their par

ent areas. Because there is a parent

descen

dent relationship between areas there will

also be an implicit sequential order of

regional .evolution, depending on assump-

tions about the

initial brain.

Correlated

with the order of the proposed evolution

ary sequence may also be an implied hier

archy of increasing functional differentia

tion and complexity.

The comparative cytoarchitectonic stud

ies of Brodmann (1909) and others

appeared to corroborate associationist

assumptions that association cortex was

newer than projection cortex and that some

association areas could only be identified

in the most

advanced

brains. The order

of regional evolution of the cerebral cortex

thus appeared to begin with primary sen

sory receptive areas and motor output areas

and involve the progressive differentiation

of ever higher association areas. In the ini

tial state sensory and motor areas were

thought to be directly connected to one

another to form direct reflex arcs. With

the differentiation of new areas inserted

between these primary areas the reflex arc

becomes more indirect and reflex action

gives way to more complex and variable

associations between sensation and action.

The more intermediate stages, the more

complex the analytic capabilities and the

less driven by reflex and habit. The

newer

the area then, the more indirect its link

with the input and the more complex its

functional properties. In this way the dif

ferentiation of association areas from asso

ciation areas could be correlated with an

additive functional hierarchy as well.

It turns out that many of the assumptions

that initially supported this hierarchic

scenario have been subsequently under

mined. The first difficulty to crop up was

the discovery that association cortex did

not lack extensive subcortical connections

(Diamond, 1982; LeGros Clark and North

field, 1939; Rose and Woolsey, 1949).

Association areas cannot be conceived as

added on top of a complete working sys

tem, sending and receiving information

only from adjacent lower level cortical

areas. They have independent inputs and

outputs and are thereby completely inte

grated into the whole brain at every level

every bit as much as are

projection

areas.

More problematic still is the nature of these

connections. The

highest

association

areas exhibit extensive connections with

limbic cortical areas (Pandya and Yeterian,

1985), whose evolution is presumed to pre

date the evolution of primary projection

areas in all theories of cortical evolution.

Recent tracer studies have also suggested

that association areas project to core mid

brain and tectal areas and receive indirect

projections from these areas relayed

through the thalamus. They may even

exhibit connectional topography that cor

responds more to midbrain maps than to

cortical sensory or motor maps (Deacon et

al., 1987). These features are also incom

patible with the assumption that the

high

est

association areas are only connected

with areas at the next hierarchic level down.

In many ways their principal links are to

some of the most primitive brain struc

tures.

Finally, there is the problem of archi

tectonic and functional specialization.

Probably the most extreme architectonic

and functional specializations anywhere in

cortex can be found in the primary visual

and motor areas of anthropoid brains. The

primate striate cortex, for example,

exhibits highly derived cytoarchitecture

with distinct sublamination of layer IV,

complex mosaic distribution of different

visual submodalities into a microscopic

matrix of architectonically distinct patches

tufts

(Livingstone and Hubel, 1988),

and approximately double the number of

cells per cortical column of any other cor

tical area in any other mammal (Rockel et

al.,

1980). These are clearly recently

derived conditions that indicate a high

degree of functional and architectonic spe

cialization. The architecture of association

areas is generally less variable from area to

area and species to species than is the

cytoarchitecture of specialized primary

areas (Sanides, 1970). This indicates that

structural and functional specialization is

not limited to and probably

less often

exemplified in

higher order

association

areas.

Some of these arguments against the

classic view have also served as support for

an almost exactly inverted view of regional

cortical evolution. Roughly inverted sce

narios have been proposed by Bishop

(1959), von Bonin and Bailey (1961) and

Sanides (1969, 1970, 1975). Probably the

most ambitious and most widely adopted

of these is Sanides' theory of progressive

waves of cortical differentiation or "Ur-

trends.

Sanides infers the sequential evo

lution of cortical areas from a trend toward

increasing architectonic specialization that

culminates in the evolution of specialized

primary projection areas (see Fig. 9). Fol

lowing the lead of Abbie (1942) and Dart

(1934),

Sanides argued that isocortical areas

evolved from undifferentiated periallo-

cortical zones along the borders of primi

tive hippocampal and olfactory cortex in a

series of stages of increasing differentia

tion. The trend is envisioned as a series of

growth rings

(Sanides, 1970) constituted

by progressively more specialized cortical

areas differentiating out of relatively less

differentiated areas. Areas that correspond

to the highest level association cortices in

the classic view comprise some of the most

primitive ancestral areas in this view. The

most highly specialized areas in advanced

brains are envisioned to be the result of

specialized core areas developing within

more generalized areas, and subsequently,

further specialized core areas developing

within these specialized areas, and so on.

The ancient status claimed for association

areas could account for their extensive

connections with limbic structures and their

prominent links to midbrain systems. The

progressive stages of differentiation also

appear to correlate with connectional rela

tionships between cortical areas (Pandya

and Yeterian, 1985). If it is assumed that

cortical areas only retain connections with

their immediate precursors but not second

generation precursors, this model could

explain the links between association areas

and limbic cortex as well as the lack of

connections between limbic areas and

either primary sensory

motor areas or their

adjacent belt areas.

As noted earlier in this discussion, the

most serious argument against Sanides'

evolutionary sequence is the apparent exis

tence of specialized visual, auditory and

somatic projection areas in all mammal

brains, even in marsupials and mono-

tremes (Kaas, 1987). Even the primitive

cortex of turtles exhibits representation of

visual and somatic responses in distinguish-

able regions crudely appropriate to their

topological position in mammalian cortex.

An ancient status for these specialized sen

sory areas contradicts Sanides' model.

One possible counter

response is to argue

that the apparent homology between the

specialized primary sensory areas of

advanced brains and the primary sensory

areas of ancestral and conservative brains

is incorrect. For example, von Bonin and

Bailey (1961) argue that the presumed

homology between the hedgehog visual

cortex and monkey primary visual cortex

is not supported by the cytoarchitectonic

criteria Brodmann (1909) originally sug

gested. They conclude that it is far more

similar to the more generalized areas of

visual association cortex. This interpreta

tion is supported by the fact that in the

visual systems of the opossum and hedge

hog the thalamocortical projections of the

lateral geniculate nucleus (corresponding

to primary visual cortex projections) and

pulvinar nuclei (corresponding to extra-

striate association cortex projections)

extensively overlap (Diamond

al.,

1985;

Kaas et al.,

1970). Although the so

called

projection areas are specialized for receiv

ing relatively more direct information from

peripheral sensors, all adjacent areas com

prising a single modality receive indepen

dent sensory afferents from the thalamus

and all share some thalamic connections in

common (Caviness and Frost, 1980; Dia

mond, 1982).

Hierarchic scenarios of cortical evolu

tion are appealing both because of their

agreement with tacit assumptions about

mental progress and mental processes, and

because of the way they simplify the

assumptions about connectional and func-

tionaI integration associated with the addi-

tion of new parts to a complex brain. Add

ing

the new parts to the terminal end of a

growing hierarchy limits the presumed

problems of integration with all lower

levels. It also provides an explanation for

the evolution of complex structures by

demonstrating plausible intermediate steps

in complexity, differentiation and special

ization. One serious problem with both

hierarchic schemes for explaining regional

evolution of cortex is the constraint of the

linear sequence itself. Ultimately, both the-

ories'are terminal addition theories. The

arguments in support of both are analo

gous to those for terminal addition in gen

eral: addition of new structures to systems

that were complete in an adult of the pre

ceding evolutionary stage; avoiding the

complication of inserting structural

changes in the middle of a complex pro

cess; and the assumption that additional

structures augment the function of the pre

ceding structures. As a result they are prone

to similar criticisms. It is not at all clear

that cortical areas within any one brain are

organized according to simple linear hier

archies, nor is it obvious how cortical areas

in different lineages can be homologized

with respect to a strict number of steps in

the evolutionary sequence. For example,

the primate visual system exhibits at least

two, and probably three or four, distinct

processing pathways for different aspects

of visual perception that diverge from

striate cortex into distinct groups of asso

ciation areas. Like a branching tree struc

ture, one would assume that multiplication

of areas should be highest at its terminal

end. If areas are capable of differentiatin

out of other areas in the course of evolu

tion there should be an increasing number

of bifurcations as the process continues.

Each sense modality contains multiple asso

ciation areas at the same level of the cor

tical hierarchy but only one primary sen

sory area. This would not be likely if these

specialized areas represented a terminal

end in the evolutionary differentiation pro

cess. And yet the extreme other end of the

spectrum

paralimbic association areas—

is not the level of the highest multiplicity

of cortical areas either. In the visual modal

ity it appears that the most diversity and

multiplicity of cortical areas is found at

middle levels in the processing hierarchy.

This is likely true of other modalities as

well. This pattern is implausible in either

of the two general terminal addition sce

narios.

Strict hierarchical terminal addition is

not the only possibility, nor is it necessarily

less complex than are non

hierarchic sce

narios. Given that most cortical areas are

connected to more than one other area and

that all are connected with distinct sub

cortical structures, there is no obvious sense

in which a new cortical area can be thought

of as superimposed on an already complete

and functioning system

it must inevitably

emerge in the middle of a complex inte-

grated network. Terminal addition con

tributes no additional explanatory power

toward solving the problem of the pre

established integration of new areas.

Models of cortical evolution that make

no assumptions about the order of appear

ance of cortical areas have been outlined

by Allman (1 982,1990) and by Kaas (1987,

1989). They each argue that the multipli

cation of cortical sensory areas of the visual

system can be explained by duplication of

existing cortical areas followed by subse

quent differentiation of function in the new

area. Presumably the new area will initially

share the same connections and cell types

as its older twin and gradually will come to

gain or lose connections and exhibit mod

ifications in cellular architecture associated

with its changes in function. It is often the

case that adjacent cortical areas serving the

same sensory modality also exhibit con

nections with similar structures elsewhere

in the brain

sometimes to separate divi

sions of these structures, at other times

overlapping in connectivity, and it is not

unusual that neurons in common afferent

sources will send collateral branches of the

same axon to adjacent cortical areas.

Duplication of this sort would also account

for the many striking homologies between

all isocortical areas.

The duplication of an existing area is

presumed to be a relatively innocuous acci

dental mutation. However, the availability

of redundant areas frees one of the two

from the constraints of the primary adap

tation so that it is able to develop some

additional, complementary visual function.

In the primate visual system it is clearly the

case that distinct visual areas seem to be

specialized for different submodality func

tions in vision, such as color, form and

movement perception. Thus, by duplica

tion and subsequent differentiation of

function the entire collection of interde

pendent visual areas could have been pro

duced. The apparently hierarchic arrange

ment of these areas is not explicitly

explained in either model, but probably it

could be argued that the one area that

retains the ancestral function becomes the

more primary area and the differentiated

one becomes more secondary. Progression

from one to the next in sequence could

then be explained simply on the basis of

the influence of adjacent areas.

Allman (1990) and Kaas (1987, 1989)

assume that new areas are added to an oth

erwise complete visual system, inserted

between existing cortical areas. This is con

sistent with their focus on advancement and

augmentation of function as the prime

mover in the evolution of new cortical

areas

duplicated areas become recruited

to some new adaptive function that aug

ments or complements existing functions.

This argument is used to explain how dis

tinct cortical visual areas have become spe

cialized for distinct visual submodalities,

such as color, movement, or form percep

tion. A

major criticism is that the separate

functional specializations of the different

visual areas, which Kaas (1989) suggests

might be too complex to be handled effi

ciently by a single large visual area,

appear to be handled by only a few visual

areas in small brains. There is no evidence

that

new

functions have evolved, only

that existing functions have become seg

regated and distributed to parallel visual

processors. If processing all these modali

ties together in a single area is merely an

efficiency problem, we should expect that

at least some very small brains would also

segregate visual functions into the almost

two dozen visual areas that endow large

primate brains and that at least some large

brains would collapse visual processing into

only one or two visual areas, but this is not

seen. There are clearly size factors involved.

Accretion assumptions are not essential

to explain the appearance of new cortical

areas in mammalian evolution. A scenario

for the early stages of mammalian cortical

evolution was presented by Lende (1963,

1969) that does not make any assumptions

about evolutionary precedence of cortical

areas. He argues that the pattern exhibited

in the common ancestor to all mammals

(including marsupials and monotremes)

included cortical projection fields that were

extensively overlapping and therefore

poorly differentiated from one another. By

a gradual process of differentiation over

the course of evolution each projection field

retracted with respect to the other until in

most modern species each projection field

is exclusive of all others. In Lende's view

the marsupial and basal insectivore cortices

represent a state where the retraction into

separate territories is nearly complete, with

only somatic and motor areas still overlap-

ping. Because it is purely a differentiation

model there is no addition of areas and no

distinction between old and new cortical

fields, just old and new patterns. A closely

related theory of the evolution of connec

tional differentiation has been proposed by

Ebbesson (1980, 1984) and will

be dis

cussed in the next section.

Lende's model was only intended to

explain the earliest stages of mammalian

cortical evolution (better resolution phys

iological recording techniques have largely

contradicted his claims about the lack of

differentiation in primitive brains; Kaas,

isocortical evolution

areal differentiation

FIG. 10.

Depiction of a scheme of progressive differentiation of cortical areas from one another that does

not assume an evolutionary sequence in which some coexisting areas are older than or ancestral to others.

The figures are meant to represent flattened cortical hemispheres with limbic cortical areas representing the

white perimeter of each and isocortical subareas represented by the gray areas contained within. Below each

of the three brains of increasing size is a block diagram of the ancestor

descendent relationship for the

progressive generation of new cortical subdivisions. The use of diverging shades of gray is intended to represent

the differentiation of both descendents of a subdivided ancestral area from the architectonic and functional

characteristics of this ancestor.

1987). However, a simple differentiation

theory suggests some interesting alterna

tive interpretations of area multiplication

problems in more differentiated mamma

lian brains. In these brains all the projec

tion areas are differentiated from one

another and, with the exception of the

somatic and motor areas, they have also

been separated by interdigitated associa

tion areas. The idea of progressive differ

ential retraction of previously diffusely

overlapping projections might account for

differentiation of previously undifferen

tiated association areas within each sensory

modality. A

single sense modality might be

conceived as becoming progressively seg

regated into differentiated submodalities.

This hypothesis (represented in Fig.

does

not necessarily predict that the interdigi-

tated cortex between primary areas should

be any more or less complex than projec

tion areas nor that it should be performing

any higher function.

The assumption that one structure has

to be older or more conserved than another

is actually not even a necessary premise for

hierarchic terminal addition scenarios of

regional cortical evolution. If multiple new

areas result from the differentiation of pre

viously unitary areas it is not necessary that

one of the resulting areas be considered

the homologue of the ancestral area and

the other or others be considered derived.

If we assume either that a new cortical area

results from duplication of an existing area

or that it differentiates out of some sector

of a pre

existing area and eventually takes

on a function that is somehow comple

mentary to that of the other area, then it

should follow that both areas will be

changed in the process.

In this regard, the primary visual cortex

in primates cannot even be strictly homolo

gized to the primary visual cortex in the

squirrel because in the squirrel many of the

visual functions handled in the primary

visual area are in the primate partially dis

tributed to some of its many more numer

ous nearby visual areas. In a hypothetical

initial brain

with only one visual area all

the distinct submodality analyses would

have to be performed within that area—

luminosity, movement, color, form, spatial

relationships, local features, etc. There is

no visual area in a brain with many visual

areas that performs all these functions.

Even a duplicate area and its progenitor

should be expected to change with respect

to one another, in the course of subsequent

evolution, so that neither resulting area will

be directly comparable to the original. Thus

it is probably more accurate to view the

multiplication of cortical areas in terms of

progressive differentiation of all areas with

respect to one another. Both areas created

after the subdivision of some previous area

will

differ slightly from one another and

from ancestral structure, and these differ

ences will likely increase over time (see Fig.

10). From this perspective it does not mat

ter where in a cortical hierarchy the new

division appears, the differentiation pat

tern

will

be essentially the same.

It cannot automatically be assumed that

there is any strict hierarchy from function

ally and architectonically simple to com

plex cortical areas or from phylogenet-

ically old to new cortical areas. A ranking

of areas with respect to relative phyloge-

netic age or functional complexity is not

unambiguously reflected in neuroanatom-

ical data and may not be consistent with

developmental considerations of cortical

differentiation. Nor are we justified in

assuming that the addition of new func

tional subdivisions of cortex correlates with

an enhancement of function. The addition

of new cortical areas still needs to be exam

ined with respect to the influence of brain

size. Extensive multiplicity of cortical areas

is never seen in small brains, and the most

extensive multiplicity of areas appears only

in very large brains (e.g., the human brain).

This suggests that areal multiplication

might be the result of facultative devel

opmental responses to brain size and not

distinct genetic adaptations.

Reorganization and the neogenesis of

neural circuits

Whether we explain area multiplication

and functional differentiation in terms of

addition or differentiation there is still a

major problem area that must be addressed:

How are the neural connections of these

areas determined? This rewiring problem

is most troublesome for addition or dupli

cation hypotheses. If an area is completely

new its connections must presumably

invade territories in other structures that

are occupied by projections from pre-exist-

ing areas, and it must itself be invaded by

axons from other structures that would

otherwise have found other targets in the

brain. But area duplication does not escape

these problems. A

duplicated area may have

all the afferent and efferent specificities

appropriate to the original area but there

still must be overlap in efferent projections

from the duplicate and the original area as

well as a dividing or sharing of afferents.

Any subsequent differentiation of the

duplicated area from its progenitor likely

also involves changes of connections,

including both the loss of many of those

shared with the original area as well as a

shift of its efferents to new targets.

The possibility of connectional reorgani

zation also suggests other options for neo-

genesis. If connectional organization can

be altered then it is not necessary for ,new

structure to be added for new functional

areas to emerge, it is only necessary for

their underlying connections to change.

Since ultimately the patterns of connection

determine function within the nervous sys

tem, all theories of neogenesis of cortical

structure and function must address this

issue. A

theory that fails to explain how

underlying connectional reorganization

takes place is fundamentally incomplete.

The most obvious hypothesis for

explaining the evolution of new connec

tions is that they are simply added. Because

this hypothesis requires axons to enter

novel target areas that are occupied by

other connections and ultimately displace

some of those connections or form new

synapses in that area, it has been called the

invasion hypothesis.

The basic features of this

process are diagrammed in Figure 11

Cells

from one area either change their target

or produce collateral branches that invade

a new target area. This idea has had a long

history in comparative neuroanatomy (e.g.,

Ariëns Kappers

et al.,

1936;

Herrick, 1920)

and seems essential to explain the appear

ance of certain neural pathways that occur

simple accretion hypothesis axonal invasion hypothesis

before duplication

addition after duplication

addition before axonal invasion after axonal invasion

/','

. .

parcellation hypothesis equivalent cell population hypothesis

diffuse connectivity differential connectivity

before parcellation after parcellation before cell migration after cell migration

11.

Four commonly cited or assumed rewiring hypotheses are depicted with source nuclei indicated by

shaded ellipses, cortical (or other nuclear) targets indicated by shaded boxes, and connections indicated by

black arrows connecting them. Antecedent and consequent conditions are shown in neighboring boxes. 'The

simple accretion hypothesis is depicted at the top left, in which a new structure is added to the brain including

its own new connections. The invasion hypothesis is depictedat the top right, in which a new set of connections,

or connections originally targeting some other area, invade and establish synapses in a novel brain structure.

The parcellation hypothesis is depicted at the bottom left, in which previously diffusely interconnected

structures lose some of their diffuse connections in a complementary fashion so as to produce subdivisions of

each that are connectionally distinct. The

equivalent cell

hypothesis is depicted at the bottom right, in

which cells from one region migrate into another structure and attract their afferents to this new structural

position.

in relatively recent lineages but not ances

tral lineages (Northcutt, 1984). Invasion

was long thought to be the explanation of

the progressive development of telence

phalic specializations in vertebrate evolu

tion. However, many of the telencephalic

connections that were formerly thought

absent in anamniotic vertebrates have

turned out to be demonstrable with axonal

tracing techniques (Ebbesson, 1980;

Northcutt, 1981). In support of this

hypothesis Northcutt

984) argues that the

presence of a spinothalamic pathway in all

reptiles, birds and mammals but in no other

group but cartilaginous fishes and the pres

ence of palliospinal pathways only in birds

and mammals are each best explained by

invasion.

However, Ebbesson (1 980) questions the

plausibility of invasion on the grounds that

comparative evidence demonstrates a

remarkable conservatism of connection

patterns within all vertebrate brains despite

radical differences in size, morphology and

differentiation. The overwhelming major

ity of major pathways seem to be evident

in all vertebrates and probably represent

an inheritance from the common protover

tebrate ancestor. Ebbesson argues that

there are no developmental or compara

tive cases that would require the assump

tion that axons must have invaded

unusual target (although his own theory

has been called unfalsifiable in this regard).

All connection changes in the course of

vertebrate evolution can be explained

entirely in terms of changes in the relative

numbers of preexisting connections

sometimes complete loss of connections

in prior lineages. Ebbesson (1

980)

refers

to this as the

parcellation theory

of brain evo

lution. Invasion, he argues, either simply

does not occur or else is extremely rare.

A schematic depiction of the parcella

tion hypothesis is presented in Figure

Parcellation theory presumes that the

ancestral condition consists of a relatively

undifferentiated pattern of connections

between two given structures. It is pro

posed that progressive loss or retraction of

a selected subpopulation of these fibers is

responsible for subsequent differentiation

of each ancestral structure into two, con-

nectionally and functionally distinct sub

divisions. In the course of time further par

cellations of these separate projection

systems can result in yet further functional

specialization. This process would account

for increasing multiplication of areas and

increased differentiation of functions in the

course of brain evolution. It also explicitly

predicts many details of connectional orga

nization that should correlate with areal

multiplication and differentiation in cor

tex. Because subdivided areas originate

from a single area they each will inherit

most but not all of the connections of that

ancestral area. For example if an

extrastri-

ate visual area in some ancestral lineage

becomes subdivided in its descendents we

should expect that the lateral posterior

(pulvinar) thalamic source of the ancestral

afferents would be the same for the descen

dent subareas although perhaps subdi

vided into new subdivisions, and we should

expect that the tectal targets of its efferents

should likewise be the same structure or

some new subdivisions of that structure.

These predictions appear to be reflected

in the organization of afferent and efferent

connections of brains with relatively few

visual areas as compared to those with many

visual areas (Diamond

et al.,

1 985; Kaas and

Huerta, 1988), as well as in other systems.

Ebbesson further argues that the evo

lutionary history of parcellation events is

recapitulated in developmental processes.

Parcellation must occur at some particular

stage in the development of existing con

nections in any organism and subsequent

further parcellation processes must be pre

ceded by this first parcellation. Following

von Baer's principle of character prece

dence

(i.e.,

that members of two or more

closely related taxa will follow the same

course of development to the stage of their

divergence), this would mean that two

species' developmental patterns would

coincide to the point of the first parcella

tion event that distinguished them, and

therefore that parcellation events that fol

low others in evolutionary time, and fur

ther parcel the same projection. system,

should also be expected to occur subse

quent to the first in development. This pre

dicts that early stages of brain development

should be characterized by the presence of

diffuse axonal projections and later stages

would proceed through progressive par-

cellation processes.

In general terms this pattern of pro

gressive culling of initially diffuse connec

tions is indeed exhibited in brain devel

opment. Initial axonal projections appear

to be less selective and more exuberant than

the projections that survive to adulthood,

and as a result fetal axons contact many

more targets and a much wider variety of

targets than do adult axons. During sub

sequent development these superfluous

connections are culled, resulting in. con

nections that are far more specific and

topographically organized (Jacobson, 1978;

Purves and Lichtman, 1980, 1985). The

details of this process will be discussed in

the next section.

Ebbesson (1980) argues that this devel

opmental pattern recapitulates the ances-

tral sequence of differentiation of each

brain area and its connections. These tran-

sient connections within the fetal brain are

therefore viewed as

fossils

of earlier pat

terns of brain organization. Viewing cer

tain transient events in neural. develop

ment as

fossils

may also help explain the

appearance and subsequent elimination of

whole classes of neurons during develop

ment that serve as transitory targets for

projections ultimately destihed for other

targets. However, this strictly recapitula-

tionist interpretation is not critical to the

parcellation hypothesis and weakens its

generality and predictive power. Although

Ebbesson's (1980) initial examples com

pare the adult organization of connections

in primitive fish to embryonic connections

in terrestrial vertebrates (and therefore

have earned the criticism that it assumes a

crude

scala naturae

view of living species

as well as strict terminal addition), the

underlying assumptions of parcellation

theory do not require terminal modifica

tion. The relative timing of fetal parcel

lation events need not be an exact reca

pitulation of evolutionary events and the

hierarchic 'dependency of one parcellation

event on another during development does

not

necessarily imply a corresponding phy-

logenetic order as well.

A more flexible version of the theory can

be articulated if we assume that a new par-

cellation process can affect connections at

any stage of brain development during the

course of evolution. Parcellation processes

that alter connections early in ontogeny

will operate on less differentiated connec

tion patterns than parcellation processes

that occur later in ontogeny. Presumable

parcellation processes during development

depend on earlier parcellation processes,

and so a subtle change at an early stage

might

have radical consequences for adult

structure. A

wider range of reorganiza-

tional effects should result from mutations

that influence parcellation early in devel

opment, particularly if subsequent parcel

lation events are undermined or biased by

the earlier changes. But if changes in par-

cellation processes can occur at any devel

opmental stage, then brain development

cannot be a strict recapitulation of the order

of phyletic events in brain evolution. A

very

early change may have been very recently

acquired whereas a relatively late parcel

lation change may date

a much earlier

evolutionary epoch. Recently acquired

parcellation changes that affect an early

stage of development could fundamentally

alter the pattern of all subsequent devel

opmental events so that they no longer

reflect any meaningful phyletic series.

The parcellation hypothesis is the inverse

of the invasion hypothesis both in its mech

anism and in the assumptions that it makes

about ancestry

descent relationships of

brain structure. In more general terms, the

invasion hypothesis is an example of an

additive theory and the parcellation

hypothesis is an example of a differentia

tion theory. Each of these two mechanisms

for connectional change correspondingly

supports. additive

versus

differentiational

theories of brain structure evolution.

The

parcellation hypothesis is most consistent

with area addition scenarios involving dif

ferentiation of new areas from old areas

and is inconsistent with simple addition

hypotheses that presume the invasion of

new axons. It is probably also inconsistent

with area duplication scenarios, since it

would reject arguments about projection-

map duplication as invasion hypotheses.

But in addition, parcellation is also

inconsistent with many of the existing dif

ferentiation theories of cortical evolution

because it does not predict that ancestral

cortical areas can remain unchanged as new

areas differentiate out of them. For this to

happen the older of the two areas would

have to maintain all of its previous con

nections and the newer area would be dis

tinguished only by its lack of certain con

nections and not by the possession of

connections not also present in the older

area. Considering either thalamic afferents

or subcortical efferents, it is clearly not the

case that there are cortical areas that pos

sess all the connections of any of their

neighbors; each has slightly different

although often partially overlapping affer-

ents and efferents. The adjacency of affer

ent sources and efferent targets of neigh

boring cortical areas as well as the partial

overlap in some of these is strong support

for the parcellation hypothesis.

Invasion is a necessary hypothesis for

both additive and recapitulational theories

of regional cortical evolution. Structural

addition is doubly challenged by the par-

cellation hypothesis because the latter de

nies both the possibility that new structures

can arise from nothing and the possibility

of invasion of new structures by axons that

did not contact it ancestrally. Even though

they do not assume structural addition,

sequential differentiation theories, such as

the traditional evolutionary hierarchy from

projection to association areas or Sanides

inverted hierarchy from association areas

to specialized sensory and motor core areas,

are also both tied to the assumption that

axon invasion events have been common-

place. If Ebbesson is correct in assuming

that these are rare events in evolution then

all of these hypotheses must be abandoned.

It is important to recognize that the par

cellation hypothesis is a stronger claim

about what can and cannot happen in brain

evolution. Invasion hypotheses do not deny

the possibility that parcellation may play a

major role in brain evolution. Invasion is

an additional assumption beyond parcel

lation. One difficulty with denying any pos

sibility of invasion is that to do so often

requires hypothesizing many more con

nectional changes in order to explain a

phylogenetic difference that can otherwise

be explained by only one or two invasion

events (Northcutt, 1984). Although Ebbes-

son's argument for the ancient status of

preset axon

target affinities

also corrob

orated by evidence of the ancient and

highly conservative nature of molecular cell

communication and recognition mecha

nisms (Fasolo and Malacarne, 1988), such

specificity may be determined by complex

relationships of timing and hierarchically

organized interactive effects that render

this specificity somewhat degenerate

(Edel-

man, 1987).

A more fundamental criticism of par-

cellation theory in its strong form is that

it is preformationist (see numerous com

mentaries following Ebbesson, 1984).

Either some initial protovertebrate brain

(possibly long antedating the vertebrates)

must be considered to have been com

pletely undifferentiated in connectivity and

therefore totipotential from an evolution

ary standpoint, or some of the initial con

nection patterns must have been pre

formed from the start and require

explanation in some other way,

e.g.,

inva

sion. The first possibility is more consistent

with the exclusivity assumption of parcel

lation theory.

There is no stage of brain development

that exhibits a corresponding totally inter

connected undifferentiated state. Although

there

some degree of initial overproduc

tion

projections and poor differentiation

target sites in the fetal brain these con

nections are neither totipotent nor com-

pletelv undifferentiated. Most pairs of brain

structures never pass through a stage where

they are connected. Target sites are under-

determined but probably not undeter

mined by genetically pre

established

molecular affinities between growing axons

and potential target cell substrates. During

embryogenesis axons do in fact grow out

from their cells of origin to invade differ

ent structures distantly located in the brain.

En route to their genetically underdeter-

mined target zones they also may pass

through structures with which they estab

lish no synaptic contacts or only transient

contacts. Both the active

exploratory

nature of embryonic axons and the pre-

specification of initial target zones are

potentially troublesome for parcellation

theory. However, permanent loss of many

connectional affinities over the course of

vertebrate evolution is entirely consistent

with the theory.

Both the invasion hypothesis and the

parcellation hypothesis can probably be

used to explain almost any possible

arrangement of connections in the living

vertebrates. The invasion hypothesis makes

many more assumptions regarding each

individual rewiring event,

whereas the par-

cellation hypothesis makes some other

rather troubling assumptions about the ini

tial condition of vertebrates and often

requires considerable theoretical, circum

locution to explain what could be explained

by a single invasion event. Each handles

certain examples better than the other.

A third alternative hypothesis, that to

some extent can serve as a bridge between

these two polar opposites, is the

equivalent

cell hypothesis

suggested by

Karten (1969;

Nauta and Karten, 1970). On the assump

tion that the connectional relationship

between two different cell groups within

the brain is specified by some features spe

cific to those cells, their connection should

be maintained even if one of these cell pop

ulations became displaced or actively

migrates to an entirely different location.

Karten (1 969; Karten and Shimuzu, 1989)

proposes this as a possible explanation for

the apparent homologies between thalamic

projections to the dorsal ventricular ridge

in birds and to the isocortex in mammals.

He argues that cells which in avian and

reptilian brains are destined to form the

dorsal ventricular ridge may take a differ

ent migratory pathway in mammals and

ended up in cortex. In the process they

would also attract their thalamic afferents

to this new target site. The general form

of the equivalent cell hypothesis is sche

matically diagrammed in Figure 1 1. With

respect to mammalian cortical evolution it

has the advantage of also accounting for

the significantly increased number of cells

that occupy mammalian as compared to

bird and reptile cortex. This hypothesis

could thus explain the apparent

invasion

of new thalamic afferents into the cerebral

cortex and also the apparent

invasion

new subcortical targets by cerebral cortical

cells without requiring the appearance of

any new connections or any change in the

underlying topology of connection pat

terns. However, the theory makes strong

assumptions about the specificity of devel

oping connections that may limit its gen

erality.

Increasing the amount of comparative

data will not by itself be able to help us

choose between these alternative hypoth

eses. The homological ambiguities limit our

ability to clearly discern what is a new con

nection and what is a modified ancestral

connection. And since each of the different

hypotheses are sufficiently flexible to

account for nearly any pattern, we have

only parsimony assumptions to rely upon.

Unfortunately, the capricious nature of the

evolutionary process suggests that we

should not place too much confidence in

parsimony arguments. Each of the alter

native theories proposed to account for

connectional reorganization depends on

specific assumptions about the ontogenetic

processes that ultimately determine con

nectional specificity. We can thus turn to

the ontogenetic data to look for develop

mental constraints that can rule out certain

versions of these hypotheses and develop

mental patterns that might suggest alter

native mechanisms.

ONTOGENY CONSTRAINS PHYLOGENY

Ontogeny of neural populations,

connections and functional areas

There are clear trends in mammalian

brain evolution. However, the evidence for

these trends and the theoretical assump

tions about homologies, progression, size

increase and cortical neogenesis have been

shown to be seriously flawed. Although

patterns are evident it cannot be decided

on the basis of comparative evidence

whether they are merely secondary archi

tectonic adaptations to differences in brain

size, adaptations for sensory

motor spe

cialization, or independent macroevo

lutionary tendencies for increased cogni

tive complexity and diversity. However, the

predictable relationships between these

trends and differences in brain size as well

as the parallelisms exhibited in distant lin

eages suggests that many of these patterns

may instead be the result of underlying

developmental homologies that are

expressed differentially with respect to

brain size or certain sensory

motor spe

cializations (e.g., regression of vision in

echolocating bats).

A considerable body of new data is

emerging concerning the development of

the brain that can help sort out which evo

lutionary hypotheses are tenable and which

are not. Obviously brains do not evolve

from one adult form to another, although

this is often how brain evolution is por

trayed. In fact, most of the well established

theories of cortical evolution make no

assumptions at all about brain develop

ment processes (except possibly the

erroneous assumption that they should re

capitulate phylogenetic trends). But phy

logenetic change in brain structure is the

result of changes in the process of brain

development and ultimately must be

explained in terms of developmental

mechanisms. The importance of ontogeny

to the understanding of brain evolution is

not that it recapitulates phylogeny

almost certainly does not

but that it con

strains the possible modes of variation that

phylogenetic changes can exhibit (Alberch,

1982; Katz, 1982; Smith

Gill, 1983). If

there are only certain ways that cell pop

ulations, functional areas or connections

can develop within a brain, then patterns

of phylogenetic change will tend to be lim

ited accordingly and the possibility of par

allel evolutionary trends arising in inde

pendent lineages will be increased.

mature

connections

FIG.

12.

Fetal cortical transplant experiment in rats

which either a frontal cortical sector is transplanted

to the posterior cortex of a host or

posterior cortical

sector is translated to the anterior cortex of a host.

In either case, when the animal matures it is found

that the cortical efferents from these transplants proj

ect to the appropriate targets for their new position

rather than the target specified by their original posi

tions

(i.e.,

rostral sensory/motor areas send efferents

to the brainstem and spinal cord whereas posterior

visual areas send efferents to the tectum).

A number of features of the neural

developmental process in mammals are

incompatible with scenarios of cortical evo

lution which assume that cortical areas can

appear, differentiate, or even change their

relative sizes as independent units during

the course of evolution. This is because the

information that specifies the size, archi

tectonic organization, afferent and effer

ent connections and therefore the basic

function of

cortical area during its devel

opment is determined by factors extrinsic

to the cells that comprise that area. The

neurons that comprise the fetal cortex are

nearly all produced prior to the stage at

which the cortex is invaded by thalamic

afferents and prior to the stage at which

axons originating from cortical neurons

reach their targets. At this stage the cell

groups in different positions on the cortical

mantle are effectively totipotent with

respect to their ability to assume any of the

different functional roles exhibited by cell

groups in the adult cortex (O'Leary, 1989).

The

most striking evidence for this comes

from fetal transplantation studies where

sectors of fetal cortex are removed from

one area of

donor's cortex and trans

planted to a different area of a recipient's

cortex (Fig. 12). In the mature brain the

transplanted cortical tissue takes on the

sensory or motor functions, assumes the

cytoarchitecture and even develops the

appropriate afferent and efferent connec

tions characteristic of its new cortical posi

tion rather than its place of origin (O'Leary

and Stanfield,

1989;

Stanfield and O'Leary,

1985). This relatively late, interactive

determination of cortical structure and

function has very significant implications

for the evolution of isocortical subdivi

sions. The evolution of a new cortical

region must therefore be a systemic pro

cess and not the result of the isolated local

expression of genetic mutations.

The structural and functional differen

tiation of any cortical area is thus not spec

ified by the specific local cell lineages that

constitute it, and takes place considerably

after neurogenesis has been completed for

all cortical areas. The determination of

what constitutes a distinct functional area,

where its boundaries are, how big it is with

respect to neighboring areas, the local spe-

cializations of its myelo

and cyto-architec-

ture and its connections with other' areas

are all largely independent of which cells

constitute that area. New cortical areas that

appear in the course of evolution cannot

have been added as whole units corre

sponding to specific populations of new

cells.

Classic theories of additive cortical evo-

lution are clearly inconsistent with this

constraint of cortical development because

it undermines any possibility of discrete

terminal addition. It also undermines addi-

tive theories of differential area expansion

as well

(e.g.,

differential addition of cells to

association cortex with respect to primary

cortex during cortical advancement). The

addition of cells to one sector of cortex

rather than another does not determine

which of the cells will or will not be included

within the functional regions that develop

in that sector. The enlargement of the cor

tex cannot be thought of in piecemeal

terms. Addition of new cortical areas can

not be the direct cause of the addition of

new brain mass. Given the fact that there

is no specific topographic information rep

resented in the developing cortex, cortical

expansion must be considered as a whole

and area by area size determination must

be a consequence of secondary processes.

The expansion of the total cortical man

tle is probably not determined by a simple

increase in neurogenesis either. The target

size of the cerebral cortex is likely deter

mined prior to neurogenesis for the struc

ture as a whole. During neurogenesis pre-

curser cells in a germinal layer deep to the

developing cortex divide to produce neu-

rons that leave this zone and migrate along

the length of special radial glial guide cells

that gxtend from the germinal layer to the

surface of cortex. These guides effectively

limit tangential migration (although see

Walsh and Cepko, 1988, for discussion of

exceptions) and serve to align succeeding

cells along this radial column occupying

ever more superficial positions. Radial

guides and the germinal zone at their base

probably form distinct proliferative units,

ontogenetic columns (Rakic, 1988). All

cell types within that columnar unit of cor

tex are derived from a common polyclonal

precurser. Earlier it was noted that the

number of cells within a column of cortex

of the same tangential dimensions is

approximately the same in species with very

different size brains (Rockel

al.,

1980).

This indicates that neurogenesis within an

ontogenetic columnar unit is invariable

across species and irrespective of brain size.

Cortical expansion must therefore be

understood in terms of the addition of more

ontogenetic columns as opposed to the

increase of cell production within these

columns (Rakic, 1988). Since the multipli

cation or germinal precursors which will

establish ontogenetic columns occurs prior

to the terminal differentiation of neurons

from these precursors, the determination

of the size of the cortical mantle must be

determined prior to neurogenesis within

the cortex.

The determination of area dimensions

and boundaries within the developing cor

tex must occur subsequent to neurogenesis

by virtue of other mechanisms. This

appears largely to be the result of afferent

and efferent interactions. In the discussion

of parcellation theory it was noted that the

first axonal connections during develop

ment are over

exuberant and relatively

unselective. These initially diffuse project

ing cortical afferents and efferents com

pete for dwindling synapses within their

target areas. This competition results in a

significant culling of axons and axon col-

laterals and some cell death. The relative

correlation of the neural activity of an axon

with others in the near vicinity which relay

similar information to an area is thought

to play a significant role in determining

which axons will be retained and which will

be eliminated (Purves and Lichtman, 1 985).

The resultant parcellation of connections

largely determines which structural and

functional characteristics will develop.

This is demonstrated dramatically in the

case of cortical efferents. In an infant rat,

cells in layer

of all regions of the isocor-

tex appear to give rise to axons that extend

into the spinal cord, but in the adult rat

only rostrally located somatomotor areas

contain cells with spinal projections. More

caudally located areas, specialized for visual

or auditory modalities, lose these spinal

connections but retain connections to the

tectum (see Fig. 13). This explains why het-

erotopic cortical transplants take on the

connections and functions appropriate to

their new cortical position.

similar pat

tern of overexuberant and relativelv undif

ferentiated connections followed by later

culling of a large number of these connec

tions has been extensively documented for

thalamocortical projections and for corti-

cocortical connections, among many other

systems (see reviews in Jacobson, 1978;

Purves and Lichtman, 1980, 1985; Purves,

1988).

Connectional interactions must also play

the major role in determining the size that

a source or target structure will attain. The

elimination or reduction of afferent con

nections to a cortical area during fetal

development can cause it to be reduced in

size, can cause neighboring areas to be

come enlarged, and can cause a corre

sponding displacement of the cortical

boundary between them (Rakic, 1988).

Deafferentation does not seem to diminish

the numbers of cells per cortical column

immature

connections

mature

connections

13.

Exuberant efferent cortical projections and

culling of connections in development is depicted with

respect to frontal somatomotor areas and posterior

visual areas. Early in the development of an infant rat

pyramidal cells from layer

of nearly all cortical

regions have axons that reach these targets. These

exuberant axons are later culled in a parcellation pro

cess driven

dynamic interactions between com

peting axons and their targets.

in that area, only the number of columns

that are contained within a projection area.

In other regions of the brain, more limited

in their possible sources of afferent input

or efferent targets, a loss of connections

can induce significant cell death. This

appears to be the case with many periph

eral afferent targets (e.g., the lateral genic-

ulate nucleus, Rakic, 1988) and efferent

sources (e.g., brain stem motor nuclei, Alley,

1974; Sohal, 1976). This cell death is pre

sumed to play a role in the functional

matching of peripheral afferents to target

neuronal populations without the need for

genetic changes in order to specifically

match afferents and cell populations in

every instance of size change in evolution

or of peripheral homoplasy (Cowan, 1973;

Cowan et al.

1984; Finlay et al.

1987; Wil-

czynski, 1984). The multipotentiality of

cortical cells both for afferent and efferent

connections probably accounts for the lack

of significant cell death due to loss of con

nections. It appears that alternative pro

jections inevitably will substitute for the

lost afferents. However, some degree of

cell death

particularly at early stages—

may play a role in cortical parcellation and

distinguishes certain cortical areas from

others (Finlay and Slattery, 1983).

The displacement hypothesis

An alternative general model of connec

tional reorganization processes that takes

into account both allometric effects and

the competitive parcellation process that

sculpt cortical areas can be derived from

these developmental considerations. As

result of investigating different problems

in comparative neuroanatomy, Deacon

(1984, 1988b), Finlay et al. (1987), Purves

(1988) and Wilczynski (1984) have each

suggested that ontogenetic factors play a

central role in the reorganization of neural

circuits in response to differences in neural

populations, regressive processes and per

turbations of maturation schedules, or

homoplaseous changes in peripheral sen

sory or motor structures. These views are

similar enough to be capable of synthesis

into a single general model of the struc

tural reorganization processes underlying

most brain evolution.

The displacement hypothesis, as it can be

called, argues that loss of connections,

acquisition of additional connections or

replacement of one class of connections by

another occurs when competitive axonal

interactions are biased by contextual events

during development. This can happed as a

result of changes in relative size relation

ships (an extreme example of which might

be complete cell death for a particular tar

get or source), changes in the amount or

intercorrelation of afferent information to

one system as opposed to another, or

changes in the relative importance of ini

tial axon

target affinities, or, changes in

developmental timing. Four possible modes

of connectional displacement are depicted

in Figure 14. Although the figure depicts

size relationships, this can be understood

metaphorically to represent competitive

biases of all kinds. For example, synchro

nization of target cell maturation and

axonal arrival would increase the likeli-

axonal connectivity after

axonal displacement

axonal connectivity after

efferent target expansion efferent target reduction

hypothesis

ancestral condition

axonal connectivity

before allometric

reorganization

axonal connectivity after

afferent source expansion axonal connectivity after

afferent source reduction

14.

The displacement hypothesis is depicted with four possible interpretations of invasion

-like

effects.

each case either the effective enlargement or reduction of targets or afferents (depicted by the size of the

structure) is invoked to explain the source of bias driving the displacement of connections from one target

to another. An analogous pattern could be produced by relative increases or decreases of axon

target affinities

or by heterochronic advantages of some afferents over others that mature at different times. These could

also be depicted in this manner by assuming that the relative size of the structures depicted represents afferent-

target biases in general. Although displacement can also explain parcellation processes, these are not depicted

here. They would roughly follow Ebbesson's schema with the added provision that parcellation is not spon

taneous, but must be induced by a change in the relative numbers of target synapses with respect to competing

axons or changes in axon

target affinities. Even if enlargement of both afferent and target populations during

generalized size increase is isometric, there may be limited collateral extent of correlated axon activity such

that diffuse overlap of connectivity could not be maintained and axons from the same source would have a

better chance of eliminating interdigitate axons from more diverse origins.

hood of synapse formation with respect to

axons arriving out of synch with cell mat

uration. As a result heterochronous

changes in developmental time schedules

for different systems may be a source of

developmental bias analogous to differ

ences in size of competing projections. The

increased affinity for synchronously arriv

ing axons should have the same effect as

relative enlargement of one source or tar

get area with respect to another.

Both invasion

like and parcellation

processes are explainable in this way. What

is different about displacement hypotheses

is that they propose that all such events are

driven by competitive biases between dif

ferent axon populations and their pro

spective targets and not by instructional

processes such as might be encoded in

molecular affinities. The strong form of

the displacement hypothesis denies both

the possibility of spontaneous axon inva

sion and also the possibility of spontaneous

parcellation. But like the parcellation

hypothesis it assumes that the basic molec

ular affinities between initial connections

and their targets are essentially conserva

tive, and if anything, only change in

response to

prior

displacement events,

under selection to stabilize a newly adap

tive circuit against the regressive influ

ences of competing biases. What is missing

from both invasion and parcellation the

ories is a cause. Displacement theories

introduce cause in the form of regressive

processes

(e.g.,

cell death or reduction of a

peripheral sensory or motor system) or dif

ferential growth processes (e.g., unequal or

hyperplasic neuron production, expansion

of some peripheral organ, or hetero-

chronic change in maturation schedules for

different structures).

Finlay et al.

(1987)

suggest that regres

sive events during development, such as

cell death and axon retraction, may account

for total brain size variation, the elabora

tion of specialized sensory, motor or cog

nitive adaptations, and allometric dispro-

portions of specific systems during brain

evolution. Widespread cell death appears

be a normal developmental mechanism

for sculpting cell populations of intercon

nected structures. To a limited extent cell

death may be exaggerated or eliminated

by variations in afferent populations or

efferent associations. These effects are,

however, buffered in systems with multiple

afferent sources and efferent targets, and

so can be expected to be most significant

in systems with highly limited connectional

relationships. Neural populations of

peripheral sensory and motor projections

are generally entirely dependent on

peripheral structures as afferent sources or

efferent targets, respectively, and provide

the most notable examples of variation in

cell death.

A number of the changes in CNS orga

nization in response to the evolution of

novel sensory organs or motor systems may

thus be the result of such a sculpting pro

cess. Wilczynski

984) reviews evidence

for the neural reorganization of CNS cell

populations and connections in response to

some major vertebrate sensory and motor

specializations (e.g., auditory, electrical and

infrared reception) that show relatively

subtle differences centrally in response to

major changes of the periphery. Despite

homoplaseous peripheral changes, central

reorganization often recruits homologous

systems for similar perceptual processes.

He argues that the interlocking of periph

eral and central reorganization in these

cases arises out of competitive develop

mental processes that match peripheral

functional requirements to central func

tional predispositions and match cell pop

ulations to one another. Although there

may be major changes in cell number in

peripherally specialized nuclei as a conse

quence of cell death there appear to be no

cascading

effects on cell death through

out the remainder of their functional con

nections within the CNS.

The main point of the cell death hypoth

esis proposed by Finlay et al

(1987) is to

account for quantitative allometric changes

in the brain and brain structures. How

ever, there are a number of reasons why

cell death is unlikely to be a significant fac

tor in major allometric changes. First, in

order to play a significant role it must be

able to account for at least a major part of

the many thousandfold differences in brain

size. Small brains are simply not analogous

to large brains that have experienced

99%

cell death. The role of cell death is clearly

limited to secondary

fine tuning

of inde

pendently developed functionally inter

dependent systems (although it may reach

80% in peripheral receptors). Second, as

compared to peripheral systems, the evi

dence suggests that the total amount of cell

death is relatively small in most forebrain

structures, even if peripheral structures

relaying information to them are signifi

cantly reduced (Rakic, 1988). This prob

ably correlates with the fact that forebrain

structures receive afferents from and send

efferents to diverse cortical and subcortical

structures rather than just one, as in

peripheral structures. The cell death

reported in areas like cortex appears to be

associated with cells maintaining transient

synapses during early phases of develop

ment that may serve a preliminary organ

izing role for later stages.

there was sig-

nificant cell death in the normal

development of cerebral cortex it would

have to be relatively uniform because of

the remarkable uniformity of

cell numbers

per area in all areas and all species. The

initial production of neurons (or the initial

production of

ontogenetic units

with

fixed neuron production patterns) is prob-

ably much more important in determining

populations in most structures.

Finlay et al

(1987) also point out the pos

sible significance of heterochronous mat

urational processes for both cell death and

connectivity patterns. They argue that ear

lier maturation or delayed maturation of

areas may introduce competitive biases in

normal axonal competition. Since some

competitive processes may extend for only

a few days, significantly delayed or pre

mature connections may be left out of the

competition, with cell death or connec

tional replacement resulting. Although

Gould (1977) argues for the widespread

presence of heterochrony in other systems

(e.g., somatic growth and puberty) there is

little evidence concerning variance of mat-

uration schedules in the mammalian ner

vous system at this time. However, time

scale effects may be significant in mam

malian brain evolution. The maturation of

a small mammal brain may be completed

within the space of weeks whereas that of

a large brain may take many years. This

means that the absolute time scale of com

petitive

regressive events during matura

tion can differ enormously despite the like

lihood thqt, at the synaptic and cellular scale

the trophic processes that underly these

effects are the same for all mammal brains.

The prolongation of these events in larger

species might affect variability, degree of

differentiation or sensitivity to extrinsic

influences. In non

mammalian vertebrates

where neurogenesis may persist through

out the lifespan heterochrony may be a

more significant factor.

In previous papers I have also proposed

that axonal competition and other regres

sive processes play crucial roles in brain

evolution (Deacon, 1984, 1988b, 1990c),

but

have focused largely on the possible

influences of size relationships. If the

determination of initial cell number in most

structures takes place prior to major axonal

invasions, the major role of competitive and

regressive processes must be the subdivi

sion of these neurogenetic fields with

respect to each other. Even though no cell

death nor substantial cell saving may result

from increases or decreases of specific

afferents or efferent targets of a multiply

connected structure within the CNS, such

changes can substantially alter local axonal

competition processes. Rather than axonal

competition determining the size of brain

structures via cell death (probably only sig

nificant for peripheral structures), the rel

ative sizes of interconnected brain struc

tures should be a major determinant of

patterns of axonal connection.

Allometric effects are probably the most

common sources of bias, given the enor

mous range of brain sizes and the great

ranges in the relationships between central

and peripheral systems. These effects are

not limited to unusual reorganization

events. Deviations from isometry with phy-

letic size increase is the rule among brain

structures as in peripheral organs

(e.g.,

Armstrong, 1985; Campos and Welker,

1976; Deacon, 1988b; Gould, 1975; Pas-

singham, 1975; Sacher, 1970; Stephan,

1969). The systematic differentials in neu-

ronal production in different structures in

brains of different sizes should determine

correlated differences in structural parcel-

lation throughout. For example, the reg

ular increase in proportion of visual asso

ciation cortex with respect

visual

koniocortex in brains of increasing size may

reflect a growing competitive disadvantage

for primary projections in the recruitment

of cortical space determined by a growing

disproportion between the retina and its

potential thalamic and cortical targets.

As we examine species differences in

neural organization we should expect to

see certain necessary correlations between

changed connection patterns and the

allo-

metries of the various structures involved.

For example, in cases where an invasion

event is suspected to have taken place one

would expect to find some corresponding

deafferentation of the new target by a for

mer projection source that has regressed

in size with respect to its target, or some

unusual size increase in the new source

structure relative to its normal target, or

significant regression of its normal target.

In cases where loss of connection is sus

pected either cell death in the source

target or, alternatively, displacement by a

projection from a disproportionate com

peting afferent source would be expected.

Failure to find these correlates either in

adult brains or during development would

falsify a displacement interpretation.

Displacement hypotheses are falsifiable

in ways that parcellation or invasion

hypotheses are not because a displacement

explanation is an account of a mechanism

not merely of

change

structure. The

displacement hypothesis is essentially an

extension of well studied mechanisms of

developmental axonal plasticity. The pro

duction of topographic functional and con

nectional organization within a developing

area induced by reduction or over-exag-

geration of input from some outside source

is the limiting case for developmental dis

placement. The extension of this concept

to incorporate allometric influences as a

major source of bias on major projection

patterns completes the synthesis of allo-

metric effects, neogenetic processes and

developmental processes.

Displacement interactions can also con

ceivably account for true invasions of axons

into targets that even exuberant projec

tions would not otherwise contact. It is not

necessary to assume any changes in the

actual affinities of axons for their targets,

only the reduction of the specificity

requirements for target affinity. This may

occur under some extreme conditions. In

an earlier section it was noted that the ini

tial target specificity of many neural con

nections is significantly underdetermined.

This has been best documented for cortical

afferents and efferents but has also been

noted widely in the developing nervous sys

tem. As a result, initial axonal projections

invade a multitude of diffuse targets and

may establish numerous transient synapses

that will later be eliminated. There prob

ably are some predetermined affinity gra

dients involved because these initial pro

jections are far from entirely random.

Edelman

987) has argued that this initial

affinity between axons and potential target

cells is the result of specific cell surface

molecules that exhibit a range of interac

tion or "recognition" strengths (analogous

to immunological binding relationships).

In order to produce distinct connections

these affinities need not be highly specific

so long as there is either a significant

threshold difference between nearby

potentially competing projections or a

means for dynamic parcellation of rela

tively nonspecific projections, as is found

in cortex. Extremely weak axon

target

affinities can likely only exhibit themselves

when all competing affinities are essentially

eliminated or when extremely strong com

petitive biases are introduced. Elimination

of alternative stronger affinity competitors

can occur if the majority of normally

occurring transient and permanent affer-

ents to an area are eliminated, or if a nor

mal target is essentially eliminated, forcing

axons to compete for alternative low affin

ity targets. Strong biasing may also occur

if a weak affinity afferent source becomes

disproportionately large with respect to

both its normal target and nearby low affin

ity targets.

An experimental example demonstrates

this possibility. Frost and Metin (1 985) and

Sur et al

(1988) have demonstrated the

possibility of inducing optic afferents to

project to inappropriate thalamic nuclei

and thus relay inappropriate sensory infor

mation to their cortical target areas. To

accomplish this in a developing rat they

destroyed all the normal targets of the optic

projections (including lateral geniculate,

superior colliculus and visual cortex) and

additionally deafferented another tha-

lamic nucleus

(e.g.,

either the ventrobasal

or the medial geniculate nucleus) by cut

ting ascending (spinothalamic or tecto-

thalamic) fibers. One of these procedures

is diagrammed in Figure 15. Despite the

fact that the misrouted connections inner

vate anomalous thalamic targets which

project to non

visual cortical areas, cells in

these areas exhibited response properties

appropriate to visual cortex. This dem

onstrates that fundamental rewiring is

achievable by displacement and that the

new connections thereby established can

differentiate their targets appropriate to

their new functions. However, it may be

significant in these cases that the alternate

thalamic and cortical targets are homolo

gous with the normal targets at some level.

Similar natural experiments appear to be

exhibited by different breeds of Siamese

cats. These cats all have abnormal routing

of ipsilateral projections to the contralat-

era1 lateral geniculate with the result that

the visual field maps are misaligned. When

the lateral geniculate projections reach the

cortex they are dealt with in one of two

ways depending on the breed: 'they are

either inactivated so as not to

interfere with

the remainder of the map or form an iso

lated independent map that' is inserted

adjacent to the otherwise normal map (Kaas

and Guillery, 1973; Guillery

1974). What

factors bias the axonal competition toward

one or the other option are not known.

Analogous competitive processes may

underly the evolution of new cortical areas.

Simple invasion is astronomically unlikely

because it can only occur when there is a

significant loss of target affinity in one set

of axons and simultaneously a significant

increase in affinity for that same target area

by other axons that have also simulta

neously lost affinity for their own target.

Each of these conditions involves an inde

pendent mutational event that alters the

respective cell surface molecules or causes

certain whole classes of cells to die. In con

trast, invasion by displacement need not

involve any changes in affinity or signifi

cant cell death. The only conditions

required are either significant allometric

disproportions between areas or the elim

ination of some target area or the elimi

nation of some projection as a result of

some. degenerative event in evolution.

These conditions are probably not at all

uncommon in the course of evolution. Sig

nificant allometric changes in proportions

between different structures and projec

tion systems is the rule in all mammalian

and nonmammalian lineages where brain

size has changed by many orders of mag-

nitude. Such a principle may account for

the parcellation trends in neocortical areas

Normal connections

seen

larger mammalian brains.

ISPLACEMENT

HEORIES

ORTICAL

VOLUTION

: F

OUR

XAMPLES

Multiplication of cortical areas and their

differential allometry

The enlargement of the entire cortical

mantle with increasing brain size may influ

ence cortical differentiation indirectly by

Rerouting of connections

altering competitive interactions among

after fetal lesions of visual

cortical afferents. There may be limits to

targets maintains visual

the size of a single projection field deter-

function in aberrant targets.

mined by the number of specific afferents

FIG.

15.

Misrouting of axons by target elimination

that are available or bv intrinsic functional

is demonstrated by experiments in which the normal

constraints.

changes in the size of the

cortical mantle and different thalamic

nuclei are not isometric in the course of

evolution then there may be correlated

changes in the relative size of correspond

ing projection fields. Changes in propor

tion may also be influenced by network

allometry influences that impose func

tional costs on enlarging areas. Such con

straints might contribute to the break

and duplication of cortical fields in brains

of increasing size. The multiplication of

areas and the differential expansion

some

targets of one projection are eliminated by lesion in

fetal development and the projections to a different

(serially homologous) target are prevented from

forming. In the case depicted here from Frost and

Metin

(1985)

the targets of the optic nerve

the lat

eral geniculate nucleus of the thalamus [LG] and the

superior colliculus [SC]— and the target of the lateral

geniculate nucleus

the visual cortex [Vis]— were

damaged by fetal lesions as were the ascending somatic

sensory afferents of the medial lemniscus which would

normally synapse in the ventrobasal nucleus of the

thalamus [VB]. Consequently, the optic fibers were

thereby induced to invade the ventrobasal nucleus of

the thalamus. The otherwise normal projection of this

nucleus to the location of somatic cortex induced this

area to behave as though it were visual cortex.

cortical areas with respect to others may

also be influenced by the total size of the

entire cortical mantle with respect to the

sizes of other brain structures that are con

nected with it. This indirect influence is

suggested by the predictable allometric

scaling of the sizes of cortical areas with

the total size of the isocortex across many

species (Passingham, 1975; Deacon, 1988b).

Probably the most significant determin

ing factor in such cases is the relative size

of the afferent projection as compared to

its cortical target zone.

cortical target

area that has expanded with respect to its

afferent projections is in some ways anal

ogous to a cortical area with a reduced

afferent projection. Either should produce

a decreased density of adjacent correlated

inputs which may impair the ability of spe

cific inputs to successfully out

compete and

eliminate diffuse inputs. In the case of

depleted afferents (Rakic,

988) the size of

the differentiated area is reduced and space

is given up to neighboring areas. Neigh

boring cortical areas would also face the

same difficulty. Parcellation of afferent

projections to form duplicate adjacent pro

jections may thus be a result of reaching

some threshold of competitive instability

determined in part by the size of the affer

ent map (which may itself be matched in

size to its peripheral representation by cell

death) and in part by the independently

growing information processing costs of

network allometry within the cortex.

The increasing proportion of association

cortex to projection cortex that correlates

with increasing brain size could reflect pro

gressive competitive disadvantages for

direct peripheral projection systems in

some modalities, both in recruiting tha-

lamic targets and in recruiting cortical tar

gets via these thalamic projections. This

would follow if the proportion of periph

eral axons to central axons competing for

targets declined with size. This kind of tar

get expansion would have parcellation

effects analogous to partial deafferenta-

tion. Visual deafferentation experiments

in monkeys have demonstrated both a

reduction in striate cortex area and an

expansion into this territory by adjacent

visual association projections (Rakic, 1 988).

The relative negative allometry of the pro

jection nucleus of vision (the lateral genic-

ulate nucleus) with respect to the volume

of the corresponding visual associational

nuclei (the pulvinar complex) as well as with

respect to the rest of the thalamus (Arm-

strong, 1979; Hopf, 1965; Stephan

et al.,

198 1) lends further support to a displace

ment explanation for this evolutionary

trend. Figure 16 diagrams the major fea

tures of this hypothesis in comparison to

deafferentation experiments. Deafferen-

tation as a result of adaptational loss or

reduction of a peripheral sense organ

in blind cave dwelling species, or to a lesser

extent in fossorial or nocturnally special

ized species

should also produce this sort

of effect in cortical areal architecture, but

in a brain that is unusually small for this

pattern in that modality. Studies of such

naturally deafferented species has dem

onstrated reduction of cortical represen

tation of these sensory areas but the issue

of projection area to association area ratio

has not been investigated.

It is possible that the process of area

sub-

division may be a gradual evolutionary

event in areas with relatively diffuse topo

graphic organization. Area boundaries may

not be discrete and connectional differ

ences may exhibit a gradient

like organi

zation in these cases. Such incipient area

divisions should be more likely in associa

tion areas lacking clear sensory or motor

topographic organization and we should

expect to see increased individual variation

in these areas if this is the case. This pat

tern should contrast with that of cortical

areas that map topographically organized

representations of some peripheral modal

ity. In these cases area differentiation

should tend to be more discrete and pre

dictable. The border between visual area

and 18 and between 18 and 19 is easily

distinguishable and correlate's with func

tional map boundaries but the multiple

retinotopic maps within area

of the

monkey cortex are not easily correlated

with any architectonic borders. The appar

ent tendency for middle level cortical asso

ciation areas to exhibit the greatest level

16.

Displacement theory of association area

expansion is depicted for visual areas in two hypo

thetical brains of different sizes but receiving input

from eyes that differ little in size. The geniculo

striate

pathway is depicted by solid black arrows and dark

gray targets and the

tecto

pulvinar

extrastriate

path

way is depicted by dashed gray arrows and light gray

targets (assuming homology of the pulvinar and lat

eral posterior nucleus). In the expanded brain of B

there has been an expansion of the thalamus and the

cortical target field potentially available for both pro

jections but because the direct retino

thalamic pro

jection is not significantly larger it is not capable of

recruiting a correspondingly larger

LGN

from the

expanded thalamus and may also be at a competitive

disadvantage in competition for space within the supe

rior colliculus as well with respect to other possible

competing inputs. However, the size of the afferents

to the pulvinar are appropriately enlarged and recruit

a large portion of the thalamus. The consequence for

cortical parcellation is that the geniculo

cortical affer-

ents are at a disadvantage in the competition for cor

tical territory with respect to pulvinar afferents and

so the striate cortex

will

occupy a reduced proportion

of the entire visual projection field. The additional

of subdivision and functional diversity from

species to species may be a correlate of the

relative fluidity of these divisions.

If the tendency for cortical circuits to

subdivide and differentiate their cortical

targets with respect to one another in

development is exaggerated by brain size

increase and brain size increase is corre

lated with the evolution of increased body

size, then neither selection for augmented

specialized functions nor selection for gen

eralized brain size increase (and increased

general intelligence) needs to be involved

in order to explain cortical enlargement

and complexification. Multiplication of

cortical areas might be accounted for, not

as augmentation of function, but as a

response to a growing size differential

between peripheral projections and cen

trally originating projections as well as a

response to deterioration of integration and

processing efficiency caused by the con

comitant reduction of connectivity in a

larger cortex. Advancement of function is

not

necessary

to explain the multiplication

and differentiation of cortical areas. Func

tional adaptation is not precluded, but to

demonstrate it requires more than dem

onstrating an increase in cortical areas and

differentiation of functions within those

areas. These other correlates of size must

subtracted

before a proper assess

ment of functional advancement can be

made. Nonetheless, with the addition of

duplicate areas or with the differentiation

of functions into independent component

processes, new possibilities for specializa

tion arise that could not coexist in a com

mon area. This must certainly be a rich

source for

preadaptations.

Once adaptive alternative connection

patterns are established by whatever means

there may be selection for changes in axon

affinities and other biasing factors that limit

expansion of the pulvinary projection field may fur

ther induce its parcellation because of increased

regional differentiation of correlated activity, but also

possibly because the information arriving in the pul-

vinar may include inputs that reflect the effects of

partially displaced retino

tectal projections.

variability. Because of these develop

mental biases, parcellation patterns will

become increasingly resistant to rever

sions, even if the conditions that originally

induced parcellation are undermined. This

suggests that size

induced parcellation pat

terns may persist even if brain size decreases

in subsequent lineage. Retention of corti

cal features consistent with a much larger

brain has been documented in a number

of dwarf species (e.g., Warren and Carlson,

1986). This suggests that there may be

functionai costs associated with brain size

reduction in evolution. The possibility for

irreversible changes and corresponding

asymmetrical selection against size reduc

tion brings us full circle to a possible pro

gressive or directional tendency in brain

evolution.

Laminar segregation of afferents:

Implications for areal parcellation

and the origins of cortex

Connectional patterns between cortical

areas appear to parallel the cytoarchitec-

tonic differentiation of cortical areas.

Because of this regularity they may provide

some insights into the connectional dis

placements, invasions and parcellations that

constitute area differentiation in evolu

tion. Tracer studies of corticocortical con

nections in monkey brains have revealed

characteristic laminar origin and termi

nation patterns that seem to be general

izable to many if not all regions of cortex

(Barbus, 1 986; Deacon, 1 985; Galaburda

and Pandya, 1983; Jones et al., 1978;

Maunsell and Van Essen, 1983; Primrose

and Strick, 1985; Rockland and Pandya,

1979; Tigges et al., 1973, 1977). In gen

eral, connections that originate from asso

ciation areas and project to areas more spe

cialized for a peripheral sensory or motor

function tend to originate largely from cells

in layer V of cortex and terminate in layers

and

VI.

Connections that originate from

specialized (e.g., primary) sensory and

motor areas and project to association areas

tend to originate largely from cells in layer

of cortex and terminate predominantly

in layers III and

(see Fig. 17). There

are also subtle gradation differences that

also seem to respect the general

level

cortical area. Both origin and termination

patterns are more diffuse across lamina in

association areas (Barbus, 1986; Deacon,

989a). Similar laminar connection pat

terns have also been identified in some areas

of cat (Bullier et al., 1984), tree shrew

(Semsa et

al.,

1984) and rat cortex (Deacon

et al., 1989), but there is too little infor

mation for non

primate species to be sure

of its generality.

The consistent association of termina

tion patterns with the architectonic and

functional gradient between association

areas and sensorimotor areas clearly indi

cates that this hierarchy, which has been

the central feature in all additive theories

of cortical evolution, must also be

accounted for in terms of parcellation and

displacement processes in evolution. Else

where (Deacon, 1989a)

have referred to

these reciprocally directed pathways as

centrifugal (limbic

association

sensory/

motor cortex) and centripetal (sensory/

motor

association

limbic cortex) projec

tions because they are oriented with respect

to areas specialized for peripheral infor

mation at the one extreme and areas con

cerned more with internal states of arousal

at the other (see Fig. 18). This hierarchic

chain of cortical areas within each func

tional modality increases in number of areas

and corresponding synaptic links as brains

enlarge in evolution, yet replicates the same

systematic pattern of laminar connectivity

with each addition. A

number of research

ers have linked this asymmetric reci

rocal

connectivity pattern to Sanides' evolution-

ary sequence of cortical differentiation (e.g.,

Barbus and Pandya, 1982, 1987; Gala-

burda and Pandya, 1983; Pandya and Yet

erian, 1985). This asymmetry is presumed

to be explainable as a terminal addition

process whereby new areas are always con

nected to their immediately adjacent

ancestral area by one sort of laminar con

nection pattern and are reciprocated by its

complement. Despite the rejection of San-

ides' theory on a number of grounds, the

correlations it suggests must be accounted

for. With the repudiation of theories claim

ing terminal addition or terminal differ

entiation of cortical areas, we are forced

parcellation of cortical laminar connectivity

in the process of areal parcellation of isocortex

corticocortical laminar connectivity before parcellation

corticocortical laminar connectivity after parcellation

FIG. 17

Laminar segregation of corticocortical connections due to functional parcellation of cortical areas

is depicted on the assumption that both the ancestral and developmentally prior state are an undifferentiated

laminar termination pattern. The hypothesized undifferentiated state is depicted in the upper figure as a

single cortical area with intrinsic connections. The subsequent loss of selected classes of connections with area

parcellation is depicted in the lower figure. Note that the culled connections are asymmetric with respect to

their directional orientation. Possible sources of competitive bias that might drive this asymmetric parcellation

during development are discussed in the text.

to explain this correlation between archi-

tectonic gradients, asymmetrically pat

terned reciprocal connections, inverse

maturational gradients, and the apparent

functional hierarchy of cortical areas in

terms of competitive biases and displace

ment processes in cortical development.

At the present time there is no devel

opmental information concerning corti-

cocortical laminar differentiation pro

cesses. Nonetheless, speculation concerning

the possible mechanisms involved can be

concentrated on a few plausible factors. For

corticocortical connections within a corti

cal area there does not seem to be this level

of laminar specification (Rockland and

Pandya, 1979). The differentiation of a new

cortical subdivision out of a single ancestral

area must therefore correlate with a loss

of projections to certain cortical lamina.

Furthermore the loss is different depend

ing upon whether the projection is in the

centrifugal or centripetal direction. Since

the appearance of a new cortical division

must be a consequence of the competitive

parcellation

displacement processes, these

systematic connectional losses likely cor

relate with competitive asymmetries

between different afferent populations.

This suggests that we should look for cor

responding biases, either in terms of het-

erochronic, allometric or functional dif

ferences, that distinguish association areas

from specialized sensory/motor areas. In

fact, all three possible sources of bias can

be identified and are probably not inde

pendent.

Another important clue concerning the

particular cortical lamina that distinguish

these different cortico

cortical projections

comes from the finding that different

classes of thalamo

cortical projections also

appear to segregate according to termi

nations in these same lamina. Multiple tha-

lamic nuclei project to each cortical area

but tend to terminate in different lamina

within the same area. It appears that prin

cipal thalamic projection nuclei, whose

projections are generally limited to a single

architectonic area, tend to terminate in

columnar fashion within layers III and IV,

usually coinciding with the distribution of

granule cells in those layers (Diamond,

1982; Frost and Caviness, 1980). Intralam-

inar thalamic nuclei, which exhibit rela

tively non

specific projections to many cor

tical areas, tend to terminate in layer VI

(Frost and Caviness, 1980; Herkenham,

centripetal

centrifugal

18.

Centrifugal and centripetal corticocortical

projections are depicted on idealized flattened maps

of the cerebral cortex of one hemisphere. Areas

depicted in darkest gray are koniocortical areas or

specialized motor cortex, those in lightest gray are

association cortex, and the white perimeter repre

sents limbic cortex. The arrows represent multisy-

naptic pathways from area to area extending either

from koniocortex (or motor cortex) to intermediate

association cortex to paralimbic association cortex to

limbic cortex (centripetal) or from limbic cortex to

paralimbic association cortex to intermediate associ

ation cortex

koniocortex (or motor cortex). The

terms centripetal and centrifugal are chosen not

because of their spatial connotations (which may be

somewhat confusing in this depiction) but because of

the orientation of these projections with respect to

the gradient between areas specialized for peripheral

systems and those representing regulation of internal

state.

1980; Rausell and Avendano, 1985). Other

nonspecific nuclei that project to many

areas within the same modality and midline

limbic

nuclei, which also exhibit wide

spread paralimbic cortex and association

cortex projections, tend to terminate in

layer I (Diamond, 1982; Friedman

et al.,

1987; Frost and Caviness, 1980; Rausell

and Avendano, 1985). The non

specific and

limbic nature of thalamic projections to

layers I and VI and the specific projections

to layers III and IV can be interpreted as

functionally analogous to their counter

parts among cortico

cortical projections in

a number of ways. Middle layer projections

appear always to introduce information

associated with a source that is more

directly connected with the peripheral ner

vous system than are their target, whereas

deep and superficial layer connections

appear to convey information that ulti

mately derives from systems involved more

with the regulation of internal state, as well

as attentional and emotional arousal (Dea

con, 1989).

Two important heterochronic matura-

tional factors differentiate cells and

axons

in these cortical lamina. These correlated

heterochronic differences may account for

which lamina tend to be associated with

which afferents by virtue of their biasing

influence on axonal competition. Neurons

occupying positions that would be super

ficial to layer II and deep to layer

in the

adult brain mature before the cells of the

cortical plate and form a single primordial

cortical layer. Cells in the outer layer called

Cajal

Retzius cells exhibit two large

extraverted

dendrites that extend up

toward the pial surface and laterally,

whereas cells in the deep layer called Mar-

tinotti cells are of a distinct bi

olar shape

with dendrites extending more superfi

cially and deeper. Cortical plate cells

migrate into position in an inside

out pat

tern between these two cell layers. The ear

liest cortical plate cells to mature are the

deep layer V and VI pyramidal cells and

the very last cortical plate cells to mature

are the most superficial pyramidal cells of

layers II and III and the granule cells of

layer IV. Prior to the appearance of the

cortical plate neurons both primordial cell

types appear to establish transient synapses

with early afferent projections (Marin

dilla,

1978).

Although there is some dis

agreement on the ultimate fate of these

early maturing cells most argue that they

are eliminated by programmed cell death

in most isocortical areas (although they

appear to pqrsist in entorhinal cortex and

in paralimbic isocortex in small species, and

in all cortical areas in cetaceans; see next

section).

Although it is not known what structures

give rise to the afferent projections that

synapse on the transient cells above and

below the developing cortical plate, it is

reasonable to assume that they arise from

the same structures that will later inner

vate' the corresponding deep and superfi

cial lamina in the adult. The fact that the

thalamic afferents terminating in layers

and

are nonspecific projections and do

not appear to respect cortical boundaries

(Caviness and Frost, 1980; Diamond,

1982)

may reflect their arrival prior to cortical

plate afferents that subdivide the cortex

into

:discrete functional regions. If the

transient 'cells to which these early axons

contact are later eliminated, these axons

may be displaced onto adjacent pyramidal

cell dendrites in the deepest and most

superficial lamina of cortex (see Fig.

19).

Since different target cells within the cor

tex mature at different times and different

projections arrive at different times, tem

poral correlation may play an important

role in biasing laminar connection patterns

from both thalamic and cortical sites. Early

maturing cells located deep and superficial

to the cortical plate may correlate with the

early maturing projections and late matur

ing small cortical plate cells in layers III

and

may correlate with relatively late

maturing projections. Differences in the

relative maturation times of cells and axons

from one cortical area to another might

additionally contribute to areal differences

in laminar organization.

Another possible heterochronic bias that

may influence the asymmetric direction

ality of these projections can be discerned

in the differential myelination of thala

mocortical fibers from principal thalamic

nuclei projecting to these areas. Since my

elination appears to precede from special

ized areas to association areas the compe

tition for synapses in the middle cortical

lamina may be biased by earlier myelina

tion of ai

eas that are more peripherally

specialized.

An allometric bias is reflected in the

expansion of association areas relative to

sensory/motor specialized areas in larger

brains. As noted above, this suggests that

primary projection areas, which are more

directly linked to peripheral systems, are

competitively constrained by the size of

their afferent projections, whereas affer

ents to association areas have no such

extrinsic constraints. This may also be the

reason these areas appear to exhibit less

clearcut cytoarchitectonic divisions. The

gradient in architectonic specialization is

one of the primary bases for Sanides' argu

ment.

Functional differences are also consis

tently correlated with this gradient. For

example, neurons in striate cortex appear

to be precisely

tuned

to specific, highly

localized stimulus attributes and are orga

nized according to precise retinotopy

whereas neurons in inferotemporal visual

association areas, at the extreme opposite

end of the hierarchy, seem tuned to global

stimulus attributes and exhibit very large

receptive fields with no obvious topo

graphic organization. This is undoubtedly

also a correlate of the relative directness

or indirectness of their respective retino-

cortical afferent circuits (Kaas, 1989). At

the sensory end of the spectrum of areas

input from the periphery is highly variable

and functional correlation is only exhib

ited over very short distances, whereas at

the association end input is primarily lim

bic and likely highly intercorrelated over

relatively larger distances. Since functional

specialization of cortical areas can be sig

nificantly affected by sensory experience

during early development it is almost cer

tain that differences in the correlation of

afferent signals among adjacent axons play

a major role in determining which con

nections persist into adulthood.

Finally, the issue of network allometry

should be considered. Given the fact that

break

up of previously integrated func

tional areas effectively distributes process

ing across a number of areas, functioning

to some extent in parallel, cortico

cortical

Pattem of isocortical neurogenesis

and probable programmed cell death

prior to the

cortical plate

early invasion addition of cortical probable elimination of cells

cortical plate cells plate cells to outside the cortical plate and

external layers specialization of granular layer

Hypothesis to account for development of

laminar specificity of cortical afferents

early non

specific

thalamic

(+?)

projections

displacement of non

specific

projections

from

eliminated

cells

cortical plate cells and

invasion of specific thalamic projections

19.

theory of the displacement processes involved in laminar parcellation of cortical afferent con

nections and a possible explanation of their relationship to neocortical origins. The upper figure diagrams

the events of corticogenesis assuming programmed cell death of cells that precede the formation of the

cortical

plate (it is also possible that some of these precursor neurons are converted into neurons with different

morphologies in the mature cortex). Note that neurons forming the cortical plate migrate into position

between the two groups of cortical precursor cells above and below it and deposit in the uppermost layers of

the developing cortical plate. The lower figure depicts a displacement theory for the laminar segregation of

cortical afferents during cortical development. It is hypothesized that axons from early maturing non

specific

thalamic nuclei and possibly early maturing limbic areas establish synapses on the two populations of cortical

precursor cells and maintain them as cortical plate neurons are added and the two cell groups are forced

apart. After the formation of the cortical plate selective cell death of the precursor cells forces the axons

attached to them to seek alternative synaptic contacts. They are displaced onto the dendrites of cortical plate

neurons of the same lamina. These axons may have a competitive advantage because of their numbers and

functional maturity compared to the later arriving principal thalamic afferents that target cortical plate cells

in middle layers. The two independent populations of neurons, superimposed by the migration of the cortical

plate cells, with different maturation schedules and distinct classes of afferents, suggest the possibility that

hey derive from independent phylogenetic origins

the precurser cells from the dorsal cortex and the cortical

plate cells from the dorsal ventricular ridge of a reptilian ancestor.

connections should be competitively

selected during development that maxi

mize intercorrelated function and most

efficiently distribute processing demands

throughout the network. Such an inter

pretation is suggested by the

counter

cur

rent

organization of these cortico

cortical

connections (see Deacon, 1989).

This analysis of laminar maturation and

connectional differentiation suggests an

alternative interpretation for the origins of

mammalian isocortex that combines both

the equivalent cell hypothesis and invasion

by displacement. My suspicion (also sug

gested in Marin-Padilla, 1978) is that these

transient cells are the homologues of the

cells of the ancestral dorsal cortex of rep-

tiles and that the cortical plate cells rep

resent a phylogenetically later intrusion

(following the equivalent cell hypothesis)

perhaps from ancestral dorsal ventricular

ridge positions. The death of the reptilian

dorsal cortex homologue cells in mam-

malian ontogeny induces the displacement

of their afferents onto the apical dendrites

of these recently juxtaposed nonhomolo-

gods cells of the cortical plate. The cortical

plate cells also maintain their original

afferents and thereby establish a novel

integration of these ancestrally separated

and independent circuits. In this sense the

laminar termination pattern of centrifugal

pathways links them with the ancestral dor

sal cortex system (which has always been

linked with limbic cortex) and the laminar

pattern of centripetal pathways links them

with the ancestral dorsal striatal (dorsal

ventricular ridge) system.

Cetacean brain evolution

In the preceding sections I have referred

to a number of general trends in mam

malian brain evolution that appear to be

strongly correlated with brain size. These

include both microscopic and macroscopic

features of brains, all of which ultimately

have to be understood in terms of biases

and constraints that modify developmental

processes, and most particularly, axonal

competition/parcellation

processes. It is

therefore important to consider excep

tional cases where these correlations do not

seem to hold. Understanding what it is

about these brains that causes them to

diverge from these otherwise ubiquitous

trends will unquestionably provide impor

tant insights into the causes for the general

trends.

Some of the most striking exceptions can

be found in the dolphin brain (and pre

sumably in all cetacean brains). Many of

the unique architectonic features of the

dolphin brain have been meticulously doc

umented in a series of recent papers (Jacobs

et al., 1971, 1979, 1984; Morgane and

Jacobs, 1972; Morgane et al., 1982, 1985,

1986a, b, 1990). These findings concern

ing the dolphin brain are paradoxical in

the context of traditional theories of mam

malian brain evolution because they sug

gest that dolphin brains combine features

that are considerably highly advanced with

features that are considered quite primi

tive and conservative (Glezer et al., 1988).

The highly advanced features of the dol

phin brain are largely macroscopic mor

phological features, including a large brain

size, a high degree of encephalization, a

highly convoluted cortex, a high ratio of

neocortex to total cortex (and therefore a

high ratio of neocortex to limbic cortex),

and apparently (although this is difficult to

assess accurately) a large percentage of

association cortex. Dolphins also are con

sidered to be among the most behaviorally

advanced species by many behavioral

researchers (Herman, 1980;

Würsig, 1989;

but see critique by Gaskin, 1982). In con

trast, their conservative traits are largely

microscopic features, including relatively

thin and poorly laminated isocortex, essen

tially agranular (and some would argue,

nonexistent) layer IV and therefore no typ

ically definable koniocortical areas, appar

ent lack of a gigantopyramidal cortex (i.e.,

architectonically specialized primary motor

cortex; although some evidence of this can

be discerned in the form of larger layer

pyramidal cells), remarkably thick layer I

with respect to the rest of the cortical lay

ers, a well developed layer VI, densely

packed layer II, large

extroverted

cells

in layer II, poorly defined columnar orga

nization, indistinct architectonic bound

aries, and apparent adjacency of sensory-

motor projection areas with respect to

dorsal view

lateral view

FIG

20. Drawing of the dolphin brain shown in sim

plified form from above and from the side with an

indication of the topographic position of the different

sensory and motor projection fields. Although the

areas depicted are presumed to be the primary pro

jection areas for these modalities,

prefer to reserve

judgement on this homological relationship. Assum

ing that the indicated areas are representative

the

proportion of neocortex occu

ied by primary areas

it would appear that most of the dolphin isocortex is

composed of association areas. Limbic cortex is not

visible from the lateral view and is a relatively small

proportion of the total cerebral cortex as is appro

priate for a mammalian brain of this size. Information

for this drawing is derived from Morgane et al.

(1986a

association areas (see Fig.

20).

Glezer et al.

988)

are compelled to designate a special

conservative

progressive mode

of cor

tical evolution to account for this anoma

lous combination of features.

The dichotomy between macroscopic

morphological features and microscopic

cytoarchitectonic features is undoubtedly

significant for understanding this apparent

paradox. If gross morphological traits were

the only available evidence then the dol

phin brain would be ranked along with the

human brain as a highly advanced brain.

Such traits as the ratio of neocortex to lim

bic cortex follow expected allometric pre

dictions of a brain the size of a dolphin

brain. It is in fact even more convoluted

than might be expected from a terrestrial

brain of such size, but this can probably be

explained on the basis of its relatively thin

ner cortex (compared to terrestrial mam

mal brains of similar proportions—e.g.,

primate and human brains). Thus, with

respect to the production of initial cell pop

ulations the dolphin brain probably shares

common mechanisms with all mammals.

But the production of cell populations and

the parcellation of those populations .into

distinct functional divisions and architec

tonic fields occur independently at

sepa

rate developmental stages. The cortical

architectonic parcellation process occurs

subsequent to the production of the cor

tical mantle. The allometric proportions

that will determine the proportion of iso-

cortex to limbic cortex and projection cor

tex to association cortex are already estab-

lished, but the competitive interactions

which will subdivide and specialize these

cortical targets have not begun at this stage.

We must look to this latter process for a

clue to the peculiarities of the dolphin

brain.

The hypothesis I will suggest to explain

these architectonic peculiarities focuses on

the agranularity of dolphin cortex. The

lack of layer IV granule cells throughout

the cortex of the dolphin brain is partic

ularly remarkable because the origin of

granule cells is far more ancient than the

divergence of Cetacea from the rest of the

eutherian stock. Despite the fact that small

brains are in general less

granularized

(a feature that may in part be attributed to

the fact that the size difference between

granule cells and pyramidal cells in small

brains is relatively slight), even the appar

ently primitive brain of the North Amer

ican opossum Didelphis virginiana exhibits

a clearly defined and even subdifferen-

tiated layer IV that receives dense princi

pal thalamic afferents (Johnson, 1988;

Walsh and Ebner, 1970), so the common

eutherian ancestor of cetaceans and other

eutherian mammals doubtless also pos

sessed layer IV granule cells. Complete loss

of this cell type and cell layer must be con

sidered a rare derived trait.

The absence of granule cells is of pri

mary significance with respect to the pro

cess

of architectonic parcellation of corti

cal areas. During development of cerebral

cortex, it is the competition between invad

ing axons from the major thalamic nuclei

that is largely responsible for the specifi

cation of topographic maps and establish

ment ,of functional and architectonic

boundaries. As is well demonstrated by

studies of the formation of somatosensory

barrels in

the rat and ocular dominance

columns in the cat and monkey, the ter

minations of these critical projections and

the principal layers in which this compe

tition takes place are layers III and IV,

corresponding to the distribution of gran

ule cells in those layers (Jacobson, 1978).

Experimental destruction of these tha-

lamic afferents at an early stage (Rakic,

1988) or even elimination of peripheral

sensory information to these afferents

(Woolsey and Wann, 1976) is capable of

profoundly altering the structure of the

resultant cortical map and even causing

functional

architectonic boundaries to be

displaced (Rakic, 1 988).

Given the developmental importance of

this thalamofugal

granule cell relationship

for architectonic differentiation, it is clear

that elimination of this cell type and dis

placement of many or all of the thalamo-

fugal projections to other layers and other

cellular targets in the developing cerebral

cortex of the dolphin brain is bound to

profoundly alter all features of its tangen

tial and radial organization. The relatively

thick layer

likely results from the dis

placement of thalamic afferents that lack

their "normal" primary affinity targets.

The presence of unusual

extroverted

cells in layer II whose large dendrites reach

up into layer I may also be explainable in

this way. These cells are present in fetal

mammals brains but are eliminated early

in development and are the targets for

transient synapses during the early stages

of cortical differentiation (see discussion in

the previous section). The displacement of

axons lacking their principal target into

this layer at an early stage of development,

prior to the

normal

elimination of these

cells, may allow them to persist by estab

lishing permanent synapses with the

orphaned principal thalamic afferents.

Since columnar organization is established

by competitive exclusion processes within

layer IV and deep layer III, the lack of

clearly delineated columnar organization

undoubtedly results from the absence of

axonal competition in this layer. The lack

of clear architectonic boundaries and the

apparently clustered projection areas all

reflect this significant reduction of axonal

competition processes. However, the pres

ence of differences in relative cell sizes in

different layers and differences in the den

sity of pyramidal cells in different areas

(e.g.,

between somatosensory and motor

areas, Morgane

et al.,

1986a), that corre

spond to similar differences in terrestrial

mammals, indicate that these features are

probably controlled by factors other than

thalamocortical connections, most proba

bly their efferent terminations. These

hypothetical effects of granule cell elimi

nation are depicted in Figure 2

1. Neuro

logical mutant mice that completely lack

cerebral cortical granule cells should also

exhibit many of these same characteristics.

Study of these could serve as an informa

tive test case, although because of the great

size difference many of the most unusual

features of the dolphin brain might not

express themselves so obviously in a mouse.

How and why cetacean brains lost their

granule cells is a mystery. It is a trait that

is probably shared by all cetaceans and so

was inherited from their common ancestor

subsequent to their divergence from ter

restrial mammals. There is some trace of

Monkey Dolphin

thalamic afferents thalamic afferents

21.

Granule cell degeneration hypothesis of the dolphin isocortex is depicted in comparison with the

pattern typical for terrestrial mammals represented by the monkey brain. In the diagram of monkey cortical

architecture the droplet shaped cells represent pyramidal cells the small spherical cells in middle layers

represent granule cells and other small interneurons and the ellipsoid cells represent small layer II cells. The

same shapes are used to depict cells in the diagram of the dolphin cortex with the exceptions that there are

no granule cells and some of the layer two cells are assumed to be embryologically retained

extraverted

cells. Principal thalamic afferents that normally would target the granule cell populations in the cortical plate

are induced to establish alternative targets in the dolphin cortex in which granule cells are strangely absent.

These afferents without normal targets may establish synapses with embryonic cells in layer II

that would

otherwise be eliminated after cortical plate formation in terrestrial mammals. These retained layer II extraver-

ted cells and the displaced thalamocortical axons cause layer I to be disproportionately thick and layer II to

be more densely packed than is observed in terrestrial mammal brains. Middle layers are also prevented from

normal competitive parcellation into columnar units, that otherwise would distinguish specialized koniocortical

areas. This suggests that dolphin cortical organization is neither conservative nor atavistic, but highly derived.

this degenerative process left because

immature dolphin brains do exhibit a tran

sient but thin layer

that disappears by

maturity (Garey and Leuba,

1986).

It seems

unlikely that this loss can be rationalized

as an adaptation. Given the total break

down of cortical differentiation processes

that resulted one would presume that this

is a costly mutation, although even in the

mutant Reeler mouse, with its totally dis

rupted cortical architecture (but not lack

ing granule cells), the cortex still functions

and allows for adequate perceptual and

motor functions. The fixation and survival

of this trait in cetaceans as opposed to any

other terrestrial lineage may be related to

their unusual and relatively complete adap

tation to the aquatic habitat. The lack of

granularized

competitors (e.g.

pin-

nipeds) in this niche until much later in

mammalian evolution may have been cru

cial to the persistence of this trait. Consider

the significance of the regression of many

specific sensory and motor systems in these

species associated with their aquatic adap-

tation. They are anosmic, they have sig-

nificantly reduced visual requirements (and

in this regard are comparable to fossorial

species and echolocating bats with second-

arily reduced visual systems), and they

exhibit significant reduction of the distal

limbs, shoulder girdle and pelvis (which in

terrestrial vertebrates comprise the pre

dominant afferent and efferent represen

tation of the primary somatdsensory and

motor fields). Although many species have

highly developed echolocation systems, this

appears far more substantially represented

by collicular specialization (evidenced by

the immensely expanded and highly dif

ferentiated inferior colliculus) than by cor

tical specialization. All these regressive fea-

tures appear to coincidentally correlate

with the inability of the dolphin cortex to

architectonically differentiate.

In summary, this exception appears to

prove the rule in a rather striking and

unambiguous way. The problems of deter

mining whether the dolphin brain is con

servative or advanced or conservative-

advanced are irrelevant. The dolphin brain

is none of these. It is highly derived. These

problems that arose in the analysis at the

level of comparative morphology and com

parative cytoarchitecture dissolve once we

approach the question from the perspec

tive of developmental homologies.

Human brain evolution

Assumptions about human brain evolu-

tion are the ultimate source for many of

the misleading ideas that have haunted the

study of brain evolution, so it is fitting that

the exorcism of these ideas in this paper

should conclude with a discussion of the

uni

ueness of human brain evolution. Two

unique characteristics of the human brain

stand out as central. The human brain is

roughly three times larger than would be

predicted for an anthropoid primate of

human body size, and human brains are

capable of acquiring an unprecedentedly

complex and flexible communication sys-

tem

language. These two facts are

undoubtedly linked.

Beginning with the issue of human brain

size, it is important to find out if this

increased cell production follows trends

that are typical in other members of the

primate order. This can be ascertained by

comparing the relative sizes of the various

major structural divisions of the human

brain with predictions based on trends for

primates in general. Initial evidence that

there is a deviation from predicted allo-

metries comes from an examination of

studies that have used brain structure vol

umes to construct possible phylogenetic

trees for primate ancestry. Two studies,

using largely similar data but different

methods that control for the effects of brain

size (Douglas and Marcellus, 1975;

Bauchot, 1982), have concluded that the

human brain is more similar to either one

of two New World monkeys' brains (woolly

monkeys in one case and capuchin mon

keys in the other) and one Old World mon

key's brain (Cercopithecus talapoin), than to

the brains of any other Old World mon

keys and apes. It is probably not coinci

dental that those primates most closel

linked with

Homo

by these studies also rep

resent relatively encephalized primates.

When a structure by structure allometric

analysis is performed it appears that the

human brain diverges from primate trends

in a number of striking ways. Based on pre

dictions from primate trends, the cerebral

cortex and cerebellar cortex of the human

brain are disproportionately large relative

to the diencephalon, corpus striaturn, brain

stem and spinal cord (Deacon, 1984,

1988b). This is depicted in Figure 22.

The production of neurons that consti

tute cortical structures takes place well

before any axonal parcellation processes

begin, and therefore, as noted earlier, the

increase in cerebral cortex cannot be spe

cific to any particular region of cortex. The

increase in radial dimensions of the cortical

germinal field and in the number of onto-

genetic columns that will differentiate out

of it must take place in the human brain

prior to neural production within the cor

tex. The size disproportions between the

expanded neocortical target field, the rel

atively unexpanded population of thala-

mofugal axons, and the relatively unex-

panded efferent subcortical targets of

cortical neurons must significantly bias

parcellation processes in all these areas

during subsequent stages of differentia

tion. One effect of this is apparent in devia

tions of relative cortical area dimensions

with respect to predictions based on the

allometry of these structures in other pri

mates. Some cortical areas appear signifi

cantly smaller than expected for a primate

brain this size and others significantly

larger. For example, the visual cortex

appears to scale appropriate to the size of

its peripheral input (the retina) and its

principal thalamic nucleus, but does not

occupy the proportion of cortex predicted

for a primate brain of this size. Its periph

eral sources are constrained by the small

human body size with respect to the large

brain size. As a result they do not scale to

thalamocortical parcellation thalamocortical parcellation

process in a brain with typical process in the human brain with

primate cortical nuclear proportions disproportionately enlarged cortex

FIG

22.

Schematic diagram of large

scale human

brain structure disproportions and their effects on

axonal competition processes during human devel

opment as compared to development in the absence

of these human disproportions. With respect to the

predictions based on other primate brains, human

cortical structures (including the entire cerebral and

cerebellar cortices) are larger than expected with

respect to brainstem, cerebellar, diencephalic and

telencephalic nuclear structures. Since the cell pro

duction processes which determine the gross size of

these major morphogenetic fields are completed prior

to their parcellation into functional subdivisions it is

predicted that these disproportions will result in biased

displacement processes. The typical condition is

depicted by the three brains on the left. Brains A

and

represent the normal developmental stages of cor

tical axonal parcellation of visual (gray cortex with

gray dashed arrows as afferents), somato

motor (black

with black arrows) and prefrontal (gray with solid gray

arrows) cortical fields in a large primate. The human

deviation from this is depicted by the three brains on

the right. Brains

and D represent the human devel

opmental stages with constraint of visual and somatic

fields by their unexpanded peripheral afferents and

displacement by prefrontal afferents producing a much

enlarged adult prefrontal area.

the level that would otherwise be predicted

on the basis of brain size (a brain this big

would be expected only in a very very large

ape

the

King Kong

null hypothesis of

human brain evolution). The competitive

limits for these afferent systems are con

strained by the size of the peripheral input.

Preliminary data suggest that this is prob

ably also the case for auditory, somatic and

FIG. 23.

A diagram of some of the relative propor

tions ofcortical fields in the human brain as compared

to predictions based on typical anthropoid primate

trends. The percentages represent absolute devia

tions from the predictions for a primate brain of human

size. Temporal, parietal, and motor area predictions

are based on too few data points to be significant, but

demonstrate a pattern that is consistent with the find

ings for other areas and with the displacement

hypothesis. The depiction of peripheral structures

associated with different cortical areas is intended to

indicate that cortical areas with relatively direct rep

resentation of peripheral sensory or motor s

stems

are constrained by these afferents or efferents in their

competition for cortical representation. Figure taken

from Deacon

(1990b).

motor areas as well as for visual areas (Dea

con, 1984, 1988b; see Fig.

23).

The competitive limitation of these pro

jection systems translates into a competi

tive advantage for other areas not con

strained by peripheral afferents, which

must inherit the cortical surface area that

is left unrecruited as a result. The pre-

frontal zone appears to be one major ben

eficiary of this competitive imbalance. It is

estimated to be approximately twice the

size expected for a primate brain of human

proportion (and this translates to six times

the size predicted for a primate of human

24.

Some predicted connectional consequences of prefrontal enlargement are represented by brain A

(typical primate brain structure allometry) as compared to

(human cortical

nuclear disproportion). Dis

placement theory suggests that the enlargement of the number of prefrontal efferents competing for midbrain

targets as compared to diencephalic efferents will bias competition in favor of prefrontal projections which

will displace both some diencephalic, limbic and intrinsic midbrain axons from their normal targets. This may

lead to the relative dominance of prefrontal outputs over limbic and diencephalic outputs in control of midbrain

and brainstem vocalization centers and motor circuits. This may be linked to adaptations associated with

language skills and the loss of many stereotypic vocalizations in human evolution.

body size; Deacon,

984,

988b

Prefron-

tal cortex is not a recipient of peripheral

inputs, but of inputs from other nonspe

cific and polymodal systems of the mid

brain and cerebral cortex. It is thus buff

ered by being synaptically removed from

the cascading effects of peripheral bias that

affect other systems. It is probably not inci

dental that Broca's area for speech is con

tained within this enormously expanded

field.

This disproportionate prefrontal corti

cal surface area is

secondary consequence

of the initial disproportion of the entire

embryonic cortex with respect to its sub-

cortical

peripheral connections. These ini

tial disproportions biased axonal compe

tition for cortical representation in favor

of cortical areas whose afferents were not

constrained by peripheral systems. But this

secondary disproportion of prefrontal areas

itself must have other tertiary biasing con

sequences. Deacon (1990c) notes that effer

ent projections of this system target limbic

cortical structures and a range of midbrain

structures. We can expect prefrontal pro

jections to have a significant competitive

advantage over other afferents to these

areas during development (see Fig. 24). The

midbrain targets of prefrontal projections

also receive descending limbic cortical and

hypothalamic projections and intrinsic

projections from the central gray and retic

ular formation. Many of these prefrontal

and limbic cortical targets turn out to play

major roles in vocal call production in pri

mates. The displacement of

normal

afferents of these areas and replacement

by a larger fraction of prefrontal axons may

have significantly altered their function.

Deacon (1990c) argues that this may

account for the significantly reduced rep

ertoire of stereotypic call types in humans,

as well as for the recruitment of some of

these systems by cortical areas capable of

supporting complex skilled motor pro

gramming. The disproportions among

cortical areas and the relative reduction of

thalamocortical as opposed to

corticocorti-

cal axons undoubtedly also played a role

in altering cortical functions, some of which

are related to the human language capac

ity.

Human brain evolution cannot be con

ceived in the terms of a conservative

pro

gressive scheme of mammalian brain evo

lution. Our brains are not at the pinnacle

of any evolutionary trend. Rather the

human brain is an unusual divergent case.

The extreme disproportion of human brain

size with respect to the human body size

with respect to other primates and mam

mals is only a surface manifestation of a

complex allometric reorganization within

the brain, and is unlikely itself to be the

crucial trait under selection in human evo

lution. It is not just the increase in cortical

complexity nor the increased relative size

of the whole brain but the correlated reor

ganization of underlying neural circuitry

that is probably most significant to human

uniqueness (see also Holloway, 1979).

Because the data that we possess con

cerning the human brain are still neces

sarily limited to morphological informa

tion and notably do not include detailed

connectional data (due to the invasive

nature of present tracer techniques) direct

verification of these hypothetical reorgani

zations will have to wait. However, our

understanding of the processes that must

underly development of a brain with the

allometric characteristics of the human

brain can be further augmented by con

tinuing investigations of the relationships

between allometric and developmental

processes shared by all mammals. The

details of human brain evolution are still

largely obscure. The hypotheses presented

here are based on a massively incomplete

set of data. And yet the basic underlying

logic of allometric change and axonal dis

placement processes during development

has provided an important new window

through which to view these data and an

indispensable guide to the gathering of

subsequent information about human brain

structures and human development.

Conclusions

Understanding the evolutionary ances

try of the brain's organization is not merely

an academic exercise. It is crucial to the

study of its basic functional processes as

well. Few if any brain structures initially

evolved their present form precisely for

the purposes they now serve, and many

current brain systems may be the result of

lucky syntheses of previously separated cir

cuits or else the result of fortuitous degen

erative events. Because the brain was not

predesigned for its current adaptations the

strategies employed in its operation will not

likely yield to a purely functional physio

logical analysis. More importantly, an

understanding of the predispositions and

constraints inherited from past adaptations

and developmental strategies can lead us

beyond a merely superficial understanding

of function to appreciate some of the

deeper fundamental organizing principles

shared by all features of the brain.

Neither the study of mammalian brain

evolution nor even the study of human

brain evolution is limited to merely theo

retical exploration. We currently have

access to experimental tools that are ade

quate to the task of analyzing the neural

developmental processes that underly, can

alize and constrain brain evolution, and are

capable of gathering the sorts of compar

ative anatomical evidence that can eluci

date the variety of ways these processes

have been expressed in evolution. This is

an invaluable complement to other areas

of the neurosciences that are rapidly build

ing a database of comparative physiologi

cal and behavioral information. Our fail

ure to immediately grasp the significance

of these data for brain evolution has largely

been the fault of the unrecognized influ

ence of some very old notions about evo

lution, the nature of mental processes;. and

the place of humans in some cognitive scala

naturae.

The displacement hypothesis has led me

to propose four highly speculative. expla

nations of some major problems in mam

malian brain evolution. But displacement

theory does not depend on the correctness

of these particular interpretations. In fact,

it provides means to falsify them if they re

incorrect. The theory clearly requires

patterns of brain evolution be ex

lained in

terms of the biasing of competitive devel

opmental mechanisms and suggests

numerous possible candidates for' biasing

influences that are likely involved: includ

ing (in order of likely importance) allo-

metric relationships, cell death, hetero-

chronous changes in maturational events

and changes in molecular affinities between

cells and axons. The correlate's of displace

ment processes that are postulated to

account for an evolutionary change must

be physically exhibited by the develop

mental processes that construct living

brains. If they are not observed then a dis

placement explanation must be rejected or

modified. The examples presented in this

paper have been chosen not as test cases

for the theory, but rather as exemplars of

the range of possible displacement effects

and their consequences. Whatever the ulti

mate applicability of these individual

accounts, the general approach should at

least serve to focus attention on a number

of previously unappreciated correlations

between biasing influences in neural devel

opment and phylogenetic differences in

adult brain structure. To the extent that

these biases are altered in brains of differ

ent sizes or can be manipulated by exper-

imental modification of neural population

size or developmental timing, it should be

possible to develop explicit experimental

models of many evolutionary changes in

brain structure. The ultimate test of a gen

eral theory is not just whether or not it can

account for the data that are already

known, but also how useful it is in leading

to new experimental approaches and new

insights concerning the underlying hidden

logic of the systems we wish to understand.

ACKNOWLEDGEMENTS

would like to thank Joseph Marcus for

his comments and editorial assistance in

preparing this manuscript.

REFERENCES

Abbie, A. A.

1940.

Cortical lamination in the Mono-

tremata.

Comp. Neurol.

72:428-467.

Alberch, P.

1982.

Developmental constraints in evo

lutionary processes.

Bonner (ed.),

Evolution

and development,

pp.

313

332.

Springer

Verlag,

Berlin.

Alexander, G. E. and P. L. Strick.

1986.

Parallel

organization of functionally segregated circuits

linking basal ganglia and cortex. Ann. Rev. Neu-

rosci.

9:357-38

Alley, K. E.

1974.

Morphogenesis of the trigeminal

mesencephalic nucleus in the hamster: Cytogene-

sis and neurone death.

Exp. Morph.

31:99-

121.

Allman, J

1982.

Reconstructing the evolution of

the brain in primates through the use of com

parative neurophysiological data.

E. Arm

strong and D. Falk (eds.),

Primate brain evolution,

pp.

13-28.

Plenum Press, New York.

Allman, J. M.

1990.

Evolution of neocortex.

Cerebral

cortex,

Vol.

8, Evolution and comparative anatomy of

cerebral cortex.

Plenum Press, New York. (In press)

Ariëns Kappers, C.

U.,

C. G. Huber, and

Crosby.

1936. The comparative anatomy of the nervous system

of vertebrates including man,

Vol. III. MacMillan,

New York.

Armstrong, E.

1979.

A quantitative comparison of

the hominoid thalamus.

Specific sensory relay

nuclei. Am. J.

Phys. Anthrop.

51:365-382.

Armstrong, E.

1983.

Relative brain size and metab

olism in mammals. Science

220: 1302

1 304.

Armstrong, E.

1985.

Allometric considerations of

the adult mammalian brain, with special emphasis

on primates.

Jungers (ed.), Size and scaling

in primate biology

pp.

115

146.

Plenum Press,

New York.

Atchley,

1984.

The effect of selection on brain

and body size association in rats. Genet. Res.

43:

289

298.

Baldwin,

1985. Mental development in the child

and the race.

Macmillan, New York.

Barbus, H.

1986.

Pattern in the laminar origin of

corticocortical connections.

Comp. Neurol.

252:

5-422.

Barbus,

and

N. Pandya.

1982.

Cytoarchitec-

ture and intrinsic connections of the prefrontal

cortex of the rhesus monkey. Soc. Neurosci. Abstr.

8:933.

Barbus, H. and

Pandya.

1987.

Architecture

and frontal cortical connections of the premotor

cortex (area

in the rhesus monkey.

Comp.

Neurol.

256:211-228.

Bauchot, R.

1982.

Brain organization and taxonomic

relationships in Insectivora and Primates.

Armstrong and

Falk (eds.), Primate brain evo-

lution,

pp.

163

175.

Plenum Press, New York.

Bauchot, R.,

M. Ridet, and M. L. Bauchot.

1979.

Encephalization and evolutionary level in aquatic

vertebrates. Vie Milieu Biol. Mar. Oceanogr.

197:

253

266.

Bishop,

1959.

The relation between nerve fiber

size and sensory modality: Phylogenetic impli

cations of the afferent innervation of cortex.

Nerv. Ment. Dis.

128:84-114.

Bohringer,

C. and M.

Rowe.

1977.

The orga

nization of the sensory and motor areas of cere

bral cortex in the platypus (Ornithorhynchus ana-

tinus). J

Comp. Neurol. 174:1-14.

Bonin,

von.

1937.

Brain

weight and body

weight

in mammals.

Gen. Psych.

16:379-389.

Bonin, G.

von and P. Bailey.

1961.

Patterns of cere

bral isocortex. Primatologia

11/2,

Lief. 10:1-42.

Brodmann, K.

1909.

Vergleichende Lokalisationslehre

der Grosshirnrinde in ihren prinzipien dargestellt auf

Grand der Zellenbaus.

A. Barth, Leipzig.

Brown,

1977. Mind, brain, and consciousness.

Academic Press, New York.

Brown,

1988. The life of the mind.

Erlbaum,

Hillsdale, New Jersey.

Bullier,

J.,

H. Kennedy, and

Salinger.

1984.

Branching and laminar origin of projections

between visual cortical areas in the cat.

Comp.

Neurol.

228:329-341.

Calford, M. B., M. L. Graydon, M. F. Huerta,

Kass, and

D. Pettigrew.

1985.

variant of the

mammalian somatotopic map in a bat. Nature

3 13:477-480.

Campbell, A.

1905.

Histological studies on the localiza

tion of cerebral function.

Cambridge University

Press, Cambridge.

Campbell, C. B. G.

1982.

Some questions and prob

lems related to homology.

Armstrong and

D. Falk (eds.),

Primate brain evolution,

pp. 1

1 1.

Plenum Press, New York.

Campbell, C. B. G.

1988. Primate survivors and neo-

cortical evolution. Behav. Brain Sci. 11:90-91.

Campbell, C.

B. G.

and W. Hodos. 1970. The con

cept of homology and the evolution of the ner

vous system. Brain Behav. Evol. 3:353-367.

Campos, G.

B. and W.

Welker. 1976. Comparisons

between brains of a large and small hystrico-

morph rodent: Capybara,

Hydrochoerus

and guinea

pig,

Cavia

(neocortical projection regions and

measurements of brain subdivisions). Brain Behav.

Evol. 13:243-306.

Cartmill, M. 1972. Arboreal adaptations and the ori

gin of the order Primates.

Tuttle (ed.),

The functional and evolutionary biology of

primates,

pp. 97

122. Aldine

Atherton, Chicago.

Cartmill, M. 1974. Rethinking primate origins. Sci

ence 184:433-436.

Caviness, V.

S.,

Jr. and D. O. Frost. 1980. Tangential

organization of thalamic projections to the neo-

cortex in the mouse. J. Comp. Neurol. 194:335-

368.

Chomsky, N. 1972.

Language and mind.

Harcourt

Brace Javonovich, New York.

Count, E.

1947. Brain and body weight in man:

Their antecedents in growth and evolution. Ann.

N.Y. Acad. Sci. 46:993-1122.

Cowan,

M. 1973. Neuronal death as a regulative

mechanism in the control of cell number in the

nervous system.

M. Rockstein (ed.),

Develop-

ment and aging in the nervous system,

pp. 119

141.

Academic Press, New York.

Cowan, W. M.,

W. Fawcett, D. D. M. O'Leary, and

B. B. Stanfield. 1984. Regressive events in neu-

rogenesis. Science 255: 1258

1265.

Cranach, M. von. 1976.

Methods of inference from ani-

mal to human behavior.

Aldine, Chicago.

Crosby, E.

19 17. The forebrain of

Alligator mis-

sissippiensis.

J. Comp. Neurol. 27:325-403.

Dart, R. 1934. The dual structure of the neopallium:

Its history and significance.

Anat. 69:3-19.

Deacon,

1984. Connections of the inferior

periarcuate area in the brain of

Macaca fascicu-

laris.

An experimental and comparative investi

gation of language circuitry and its evolution.

Ph.D. Diss., Harvard University, Cambridge,

Massachusetts.

Deacon,

1985.

Counter

current flow

of cor-

tico

cortical information processing through the

laminar segregation of reciprocal connections.

Soc. Neurosci. Abstr. 1 1:203.1

Deacon,

1988a. Cortical and midbrain con

nections of the intermediodorsal nucleus in the

rat. Soc. Neurosci. Abstr. 14:328.1.

Deacon.

1988b. Human brain evolution: II.

Embryology and brain allometry.

H. Jerison

and I. Jerison (eds.),

Intelligence and evolutionary

biology,

pp. 383

41 5. Springer

Verlag, Berlin.

Deacon, T. W. 1989. Holism and associationism in

neuropsychology: An anatomical synthesis.

Perecman (ed.),

Integrating theory and practice in

clinical neuropsychology,

pp. 1

47. Erlbaum, Hills-

dale, New Jersey.

Deacon,

W. 1990a. Fallacies of progression in

theories of brains size evolution. Int. J. Primatol.

11: 193

236.

Deacon,

1990b. Problems of ontogeny and

phylogeny in brain size evolution. lnt.

Prima-

tol. 11:237-282.

Deacon,

W. 1990c. The neural circuitry under

lying primate calls and human language. Human

Evol. (In press)

Deacon,

W.,

A. Sokoloff, and D. Wecht. 1987.

Circular organization of connections linking mid

brain areas, sections of the mediodorsal thala

mus, and prefrontal areas in the rat. Soc. Neu-

rosci. Abstr. 13:304.9.

Deacon,

W., D.

W. Wang, and A. Carpenter.

1989. Similarities and differences in the laminar

organization of corticocortical connections in rat

as compared to monkey. Soc. Neurosci. Abstr.

15:114.10.

Diamond,

I. T.

1979. The subdivisions of the neo-

cortex: A

proposal to revise the traditional view

of sensory, motor and association areas.

Sprague and A.

Epstein (eds.),

Progress in psycho-

biology and physiological psychology,

Vol. 8. Aca

demic Press, New York.

Diamond,

1982. The functional significance of

architectonic divisions of the cortex, Lashley's

criticism of the traditional view.

J. Orbach

(ed.),

Neuropsychology after Lashley,

pp. 101

1 35.

Erlbaum, Hillsdale, New Jersey.

Diamond,

T.,

Fitzpatrick, and

Sprague.

1985. The extrastriate visual cortex.

historical

approach to the relation between the "visuo-sen-

sory

and

visuo

psychic

areas.

A. Peters and

Jones (eds.),

Cerebral cortex,

Vol. 4;

Associ-

ation and auditory cortices,

pp. 63

87. Plenum Press,

New York.

Diamond,

and W. C. Hall. 1969. Evolution of

neocortex. Science 164:25 1

262.

Dom,

R.,

F. Martin, B. L. Fisher, A. M. Fisher,

and

K. Harting. 197

The motor cortex and

corticospinal tract of the armadillo

(Dasypus no-

vemcincus).

Neurol. Sci. 14:225-236.

Donoghue,

P., K.

Kerman, and F. F. Ebner.

1979. Evidence for two organizational plans

within the somatic sensory

motor cortex. of the

rat.

Comp. Neurol. 183:647-664.

Douglas,

and D. Marcellus. 1975. The ascent

of man: Deductions based on a multivariate anal-

ysis of the brain. Brain Behav. Evol. 11:179-2 13.

Dubois, E. 1913. On the relation between quantity

of brain and the size of the body in vertebrates.

Verh. Kon. Akad. Wetenschappen Amsterdam

16:647.

Ebbesson, S. O.

1980. The parcellation theory and

its relation to interspecific variability in brain

organization, evolutionary and ontogenetic

development, and neuronal plasticity. Cell Tissue

Res. 2 13: 179

2 12.

Ebbesson,

S. O.

1984. Evolution and ontogeny of

neural circuits. Behav. Brain Sci. 7:32 1

366.

Edelman,

M. 1987.

Neural Darwinism.

Basic Books,

New York.

Elliot

Smith, G.

1910. Some problems related to the

evolution of the brain. Lancet 1:1-6, 147-153,

22 1

227.

Elliot

Smith, G.

1919.

A preliminary note on the

morphology of the corpus striatum and the origin

of the neopallium. J. Anat.

53:27 1

291.

Falk, D.

1980.

A reanalysis of the South African

australopithecine natural endocast. Am. J.

Phys.

Anthrop.

53:525-539.

Falk, D.

1983.

Cerebral cortices of East African early

hominids. Science

221:

1072

1074.

Falk, D. 1989.

Ape

like endocast of

Ape

man

Taung. Am. J. Phys. Anthrop.

80:335-339.

Fasolo, A. and G. Malacarne.

1988.

Comparing the

structure, of brains: Implications for behavioral

homologies.

H. Jerison and

Jerison (ed.),

Intelligence and evolutionary biology,

pp.

119

142.

Springer-Verlag, Berlin.

Filimonof, I. N.

1949.

Comparative anatomy of the cere-

bral cortex of mammals. paleocortex, a rchicortex, and

intermediate cortex.

[Trans. by V. Dukhoff, 1965].

Joint Pub. Res. Service, Office of Sci. Inf. Dept.

of Interior, Washington, D.C.

Filler; A.

1986.

Axial character seriation in mam

mals. An historical exploration of the origin,

development, use and current collapse of the

homology paradigm. Ph.D. Diss., Harvard Uni

versity, Cambridge, Massachusetts.

Finlay, B. L. and M. Slattery.

1983.

Local differences

in early cell death in neocortex predict adult local

specializations. Science

2 19: 1349

1 351.

Finlay, B.L.,

C. Wikler, and D. R. Sengelaub.

1987. Regressive events in brain development

and scenarios for vertebrate brain evolution. Brain

Behav. Evol.

30: 102

1 17.

Flechsig, P.

1900. Ü

ber Projections und Associations

Zentren des menschlichen Gehirns. [On projec

tion and association centers of the human brain.]

Neurologie Zentralblatt

19.

Flechsig, P.

190

Developmental (myelogenetic)

localization of the cerebral cortex in the human

subject. Lancet

2:1027-1029.

Friedman, D. P., L. G. Bachevalier,

G. Unger-

lieder, and

Mishkin.

1987.

Widespread tha-

lamic projections to layer I of primate cortex.

Soc. Neurosci. Abstr.

13:73.17.

Frost, D. and V. Caviness, Jr.

1980.

Radial organi

zation of thalamocortical projections.

Comp.

Neurol.

194:369-394.

Frost, D. O. and C. Metin.

1985.

Induction of func

tional retinal projections to the somatosensory

system. Nature

3 17: 162.

Fuller,

1979.

Fuller BWS lines: History and

results.

M. E. Hahn, C. Jensen, and B. C. Dudek

(eds.),

Development and evolution of brain size,

pp.

190

204.

Academic Press, New York.

Galaburda, A. M. and D. N. Pandya.

1983.

The

intrinsic architectonic and connectional organi

zation of the superior temporal region of the

rhesus monkeys.

Comp. Neurol.

22 1:169-184.

Gardner, H.

1983.

Frames of mind: The theory of mul-

tiple intelligences.

Basic Books, New York.

Garey, L. J. and G. Leuba.

1986.

A quantitative study

of neuronal and glial numerical density in the

visual cortex of the bottlenose dolphin: Evidence

for a specialized subarea and changes with age.

Comp. Neurol.

247:491-496.

Gaskin, D. E. 1982.

The ecology of whales and dolphins.

Heineman, New York.

Geschwind, N.

1964.

Development of brain and evo

lution of language. Georgetown Univ. Monogr.

Ser. Language Linguistics

17: 155-170.

Ghiselin, M.

1976.

The nomenclature of corre

spondence: A new look at

homology

and

anal

ogy.

R. B. Masterton, W.

Hodos, and H.

Jerison (eds.),

Evolution, brain and behavior:

Per-

sistent problems,

pp.

129

142.

Erlbaum, New York.

Glezer, I. I., M. S. Jacobs, and P. J.

Morgane.

1988.

Implications of the

initial brain

concept for

brain evolution in Cetacea. Behav. Brain Sci.

11:

75-116.

Goldowitz, D.

1987.

Cell partitioning and mixing in

the formation of the CNS: Analysis of the cortical

somatosensory barrels in chimeric mice. Devel.

Brain Res. 35:1-9.

Gould, S. J. 1966.

Allometry and size in ontogeny

and phylogeny. Biol. Rev. Camb. Phil. Soc.

41:

587

640.

Gould,

S. J

1975.

Allometry in primates with empha

sis of scaling and the evolution of the brain. Contr.

Primatol.

5:244-292.

Gould, S.

1977.

Ontogeny and phylogeny.

Harvard

University Press, Cambridge, Massachusetts.

Gould, S. J.

1981.

The mismeasure of man.

George

McLeod Limited, Toronto.

Gregory, W.

1935.

Replication in evolution.

Rev. Biol.

10:272-290.

Guillery,

1974.

Visual pathways in albinos. Sci.

Amer.

230:44-54.

Haight,

R. and L. Neylon.

1978.

The organization

of neocortical projections from the ventropos-

terior thalamic complex in the marsupial brush-

tailed opossum,

Trichosurus vulpecula,

a horserad

ish peroxidase study. J. Anat. 126:459-485.

Haight, J. R. and L. Neylon. 1979.

The organization

of neocortical projections from the ventrolateral

thalamic nucleus in the brush

tailed possum,

Trichosurus vulpecula,

and the problem of motor

and somatic sensory convergence within the

mammalian brain. J.

Anat.

129:673-694.

Haug, H.

1987.

Brain sizes, surfaces, and neuronal

sizes of the cortex cerebri: A stereological inves

tigation of man and his variability and a com

parison with some mammals (primates, whales,

marsupials, insectivores, and one elephant). Am.

Anat.

180: 126

1 42.

Herkenham, M.

1980.

Laminar organization of tha-

lamic projections to the rat neocortex. Science

207:532-534.

Herman, L. M. (ed.)

1980.

Cetacean behavior: Mech-

anisms and functions.

Wiley, New York.

Herrick, C.

1920.

A sketch of the origin of the

cerebral hemispheres.

Comp. Neurol.

32:429-

454.

Hodos. W.

1988.

Comparative neuroanatomy and

the evolution of intelligence.

Jerison and

Jerison (eds.),

Intelligence and evolutionary biol-

ogy

pp.

108.

Springer

Verlag, Berlin.

Holloway, R. L. 1979. Brain size, allometry, and

reorganization: Toward a synthesis.

M. Hahn,

C. Jensen, and B. Dudek (eds.), Development and

evolution of brain size,

pp. 59

88. Academic Press,

New York.

Holloway, R. L. 1981. Revisiting the South African

Taung australopithecine endocast: The position

of the lunate sulcus as determined bv the ste-

reoplotting technique. Am.

Phys. Anthrop. 56:

58.

Holloway, R. L. 1984. The Taung

endocast and the

lunate sulcus: A rejection of the hypothesis of its

anterior position. Am. J. Phys. Anthro. 1984:

285

287.

Holloway, R. L. 1985. The past, present, and future

significance of the lunate sulcus in early hominid

evolution.

P. V. Tobias (ed.),

Hominid evolution:

Past, present and future,

pp. 47

62. Alan R. Liss,

New York.

Holloway,

L. 1988. Some additional morpholog

ical and metrical observation on

Pan

brain casts

and their relevance to the Taung endocast. Am.

Phys. Anthro. 77:27-33.

Hopf,

1965. Volumetrische Untersuchunger sur

vergleichenden Anatomie des Thalamus.

Hir-

forsch. 8:25-38.

Jacobs,

S., A. M. Galaburda, W. L. McFarland,

and P. J. Morgane. 1984. The insular formation

of the dolphin brain. Quantitative cytoarchitec-

tonic studies of the insular component of the lim

bic lobe.

Comp. Neurol. 225:396-432.

Jacobs,

S., W. L. McFarland, and

J. Morgane.

1979. The anatomy of the brain of the bottlenose

dolphin

(Tursiops truncatus).

Rhinic lobe (rhinen

cephalon):

The archicortex. Brain Res. Bull.

4(Suppl. 1):1-108.

Jacobs,

S., P.

Morgane, and W. L. McFarland.

197 1. The anatomy of the brain of the bottlenose

dolphin

(Tursiops truncatus).

Rhinic lobe (rhinen

cephalon):

The paleocortex. J.

Comp. Neurol.

141:205-272.

Jacobson,

1978. Developmental neurobiology, 2nd ed.

Holt, Rhinehart, and Winston, New York.

Jarvis, M. J. and

Ettlinger. 1977. Cross

modal

recognition in chimpanzees and monkeys. Neu-

ropsychologia 15:499-506.

Jerison, H.

1973. Evolution of the brain and intelli-

gence.

Academic Press, New York.

Jouandet, M., M. Tramo, D. Herron,

Hermann,

W. Loftus, J. Bazell, and

Gazzaniga. 1989.

Brainprints: Computer

generated two

dimen

sional maps of the human cerebral cortex in vivo.

J. Cog. Neurosci. 1:88-117.

Johnson,

1988. Whose brain is initial

like? Behav.

Brain Sci. 1 1:96.

Johnston, J. B. 1906. The nervous system of vertebrates.

Blackiston's Son and Co., Philadelphia.

Jones,

G.,

J. D. Coulter, and S.

C. Hendry. 1978.

Intracortical connectivity of architectonic fields

in the somatic sensory, motor and parietal cortex

of monkeys. J. Comp. Neurol. 18

1:29 1

348.

Kaas, J.

1987. The organization and evolution of

neocortex.

S. P. Wise (ed.),

Higher brain func-

tions. Recent explorations of the brain's emergent prop-

erties,

pp. 347

378. John Wiley

Sons, New York.

Kaas, J. H. 1989. Why does the brain have so many

visual areas? J. Cog. Neurosci.

12 1

135.

Kaas, J. H. and R. W. Guillery. 1973. The transfer

of abnormal visual field representations from the

dorsal lateral geniculate nucleus to the visual cor

tex in Siamese cats. Brain Res. 59:61-95.

Kaas,

H., W. C. Hall, and

I. T.

Diamond. 1970.

Cortical visual areas

and

in the hedgehog:

Relation between evoked potential maps and

architectonic divisions.

Neurophys. 33:595-

615.

Kaas,

and

F. Huerta.1988. Subcortical visual

system of primates.

H. D. Steklis (ed.), Com-

parative primate biology,

Vol. 4:

Neurosciences,

pp.

327

391. Alan R. Liss, New York.

Karten, H.

1969. The organization of the avian

telencephalon and some speculations on the phy

logeny of the amniote telencephalon. Ann. N.Y.

Acad. Sci. 167: 164

1 79.

Karten, H.

and T. Shimuzu. 1989. The origins of

neocortex: Connections and lamination as dis

tinct events in evolution.

Cog. Neurosci. 1 :29 1-

30 1.

Kaplan, S. R. and G.

Goslow. 1989. Neuromus-

cular organization of the pectoralis (pars thora-

cicus) of the pigeon

(Columba livia):

Implications

for motor control. Anat. Rec. 224:426-430.

Katz,

1982. Ontogenetic mechanisms: The mid

dle ground of evolution.

T. Bonner (ed.),

Evo

lution and development,

pp. 207

2 12. Springer-

Verlag, Berlin.

Kirsch, J. A. W., A. I. Johnson, and R.

Switzer.

1983. Phylogeny through brain traits: The mam-

malian family tree. Brain Behav. Evol. 22:70-74.

Kruska, D. 1987. How fast can total brain size change

in mammals?

Hirnforsch. 28:59-70.

Kuljis, R. O.

and P. Rakic. 1988. Segregation of neu-

ropeptide Y

containing neurons outside cyto-

chrome oxidase puffs in the monkey develops in

the absence of visual input. Soc. Neurosci. Abstr.

14:243.5.

Le Gros Clark, W.

197 1.

Antecedents of man.

Edin-

burgh University Press, Edinburgh.

Le Gros Clark, W. and D. Northfield. 1939. The

cortical projection of the pulvinar in the macaque

monkey. Brain 60: 126

1 42.

Lende, R. A.

1963. Cerebral cortex:

sensorymotor

amalgam in the Marsupialia. Science 141:730-

732.

Lende, R. A. 1969. A comparative appoach to neo-

cortex: Localization in monotremes, marsupials

and insectivores. Ann. N.Y. Acad. Sci. 167:262-

275.

Lieberman, P. 1984. The biology and evolution of lan-

guage.

Harvard University Press, Cambridge,

Massachusetts.

Livingstone, R. and D. Huebel. 1988. Segregation

of form, color, movement, and depth: Anatomy,

physiology, and preception. Science 240:740-749.

Luria, A.

1980. Higher cotical functions in man

2nd ed. (English trans.). Basic Books, New York.

MacLean, P. D. 1970. The triune brain, emotion and

scientific basis. In F. O. Schmitt (ed.), The neu-

rosciences: Second study program.

Rockefeller Uni

versity Press, New York.

MacLean, P. D. 1973. A triune concept of the brain

and behavior.

T. J. Boag and D. Campbell

(eds.), A triune concept of the brain and behavior.

University of Toronto Press, Toronto.

MacPhail, E. 1982. Brain and intelligence in vertebrates.

Clarendon Press, Oxford.

Magalhães-Castro, B. and P. E. S. Saraiva. 197 1. Sen

sory and motor representation in the cerebral

cortex of the marsupial

Didelphis azarae azarae.

Brain Res. 34:291-299.

Marin-Padilla, M. 1978. Dual origin of the mam

malian neocortex and evolution of the cortical

plate. Anat. Embryol. 152: 109

156.

Martin, R. D. 1973. Comparative anatomy and pri

mate systematics. Symp. Zool. Soc. London 33:

301-337.

Martin,

D. 1981. Relative brain size and basal

metabolic rate in terrestrial vertebrates. Nature

293:57-60.

Martin, R. D. 1983. Human brain evolution in an

ecological context. Fifty

second James Arthur

Lecture on the evolution of the human brain.

Am. Mus. Nat. Hist., New York.

Maunsell,

H. R. and D. C. Van Essen. 1983. The

connections of the middle temporal visual area

(MT) and their relationship to a cortical hier

archy in the macaque monkey.

Neurosci. 3:

2563

2586.

Meculam, M.

M., G. W. Van Hoesen, D.

Pandya,

and N. Geschwind. 1977. Limbic and sensory

connections of the inferior parietal lobule (area

PG)

the rhesus monkey: A study with a new

method for horseradish peroxidase histochem-

istry. Brain Res. 136:393-414.

Mishkin, M. and T.

Appenzeller. 1987. The anatomy

of memory. Sci. Amer. 256:80-89.

Morgane, P. J., I. I. Glezer, and

S. Jacobs. 1990.

Comparative and evolutionary anatomy of visual

cortex of the dolphin. In E. G. Jones and A. Peters

(eds.),

Cerebral cortex,

Vol. 8.

Evolution and com

parative anatomy of cerebral cortex.

Plenum Press,

New York. (In press)

Morgane, P. J. and M. S. Jacobs. 1972. Comparative

anatomy of the cetacean nervous system.

Harrison (eds.), Functional anatomy of marine mam-

mals,

pp. 117

244. Academic Press, London.

Morgane, P. J., M. S. Jacobs, and A. Galaburda. 1985.

Conservative features of neocortical evolution in

dolphin brain. Brain Behav. Evol. 26: 176

184.

Morgane, P. J., M. S. Jacobs, and A. Galaburda.

1986a. Evolutionary aspects of cortical organi

zation in the dolphin brain.

R. J.

Harrison and

M. Bryden (eds.), Research on dolphins,

pp. 7 1

98.

Oxford University Press, Oxford.

Morgane, P. J., M. S. Jacobs, and A. Galaburda.

1986b. Evolutionary morphology of the dolphin

brain.

R. Schusterman, F.

Wood, and J.

Thomas

(eds.), Dolphin cognition and behavior: A comparative

approach,

pp. 5

29. Erlbaum, Hillsdale, New Jer

sey.

Morgane,

J., W. L. McFarland, and M. S. Jacobs.

1982. The limbic lobe of the dolphin brain: A

quantitative cytoarchitectonic study.

Hirn-

forsch. 23:465-552.

Nauta, W. J. H. and

J. Karten. 1970. A general

profile of the vertebrate brain with sidelights on

the ancestry of the cerebral cortex. In F. O.

Schmidt (ed.), The neurosciences: Second study pro-

gram.

Rockefeller Press, New York.

Northcutt, R. G. 1981. Evolution of the telenceph-

alon in nonmammals. Ann. Rev. Neurosci. 4:301-

350.

Northcutt, R. G. 1984. Evolution of the vertebrate

central nervous system: Patterns and processes.

Amer. Zool. 24:70 1

7 16.

O'Leary, D.

M. 1989. Do cortical areas emerge

from a protocortex? Trends Neurosci. 12:400-

406.

O'Leary, D.

M. and B. B. Stanfield. 1989. Selec

tive elimination of axons extended by developing

cortical neurons is dependent on regional locale

experiments utilizing fetal cortical transplants. J.

Neurosci. 9:2230-2246.

Optican, L.

and B. J.

Richmond. 1987. Temporal

encoding of two

dimensional patterns by single

units in primate inferior temporal cortex. III.

Information theoretic analysis.

Neurophysiol.

57:162-178.

Pandya, D. and E. Yeterian. 1985. Architecture and

connections of cortical association areas. In

Peters and E. G. Jones (eds.),

Cerebral cortex,

Vol.

4, Association and auditory cortices,

pp. 3

61. Ple

num Press, New York.

Passingham, R. E. 1975. Changes in the size and

organisation of the brain in man and his ances

tors. Brain Behav. Evol. 11:73-90.

Passingham, R. E. 198 1. Broca's area and the origins

of human vocal skill. Phil. Trans.

Soc. London,

B 292: 167

175.

Primrose,

and P. Strick. 1985. The organization

of interconnections between the premotor areas

of the primate frontal lobe and the arm area of

primary motor cortex. Soc. Neurosci. Abstr. 11:

127.4.

Purves, D. 1988.

Body and brain: A trophic theory of

neural connections.

Harvard University Press,

Cambridge, Massachusetts.

Purves, D. and

W. Lichtman. 1980. Elimination

of synapses in the developing nervous system.

Science 2 10:

153-157.

Purves, D. and J. W. Lichtman. 1985. Principles of

neural development.

Sinauer, New York.

Rakic, P. 1988. Specification of cerebral cortical areas.

Science 24 1: 170

1 76.

Rausell, E. and C. Avendano. 1985. Thalamocortical

neurons projecting to superficial and to deep lay

ers in parietal, frontal and prefrontal regions in

the cat. Brain Res. 347:159-165.

Reynolds, P. C. 1976. Language and skilled activity.

S. R. Harnad, H.

Steklis, and J. Lancaster

(eds.), Origins and evolution of language and speech.

Ann. N.Y. Acad. Sci. 280:1-914.

Riska, B.,

R. Atchley, and

J. J.

Rutledge. 1984.

A genetic analysis of targeted growth in mice.

Genetics 107:79-101.

Rockel, A.

J., R. W. Hiorns, and T. P. S. Powell. 1980.

The basic uniformity in structure of the neocor-

tex. Brain 103:221.

Rockland,

and D. N. Pandya. 1979. Laminar

origins and terminations of connections of the

occipital lobe in the rhesus monkey. Brain Res.

179:3-20.

Roderick,

H., R. E. Wimer, and C. C. Wimer.

1976. Genetic manipulation of neuroanatomical

traits.

L. Petrinovich and

L. McGaugh (eds.),

Knowing, thinking nnd believing,

pp. 143

178. Ple

num Press, New York.

Rose,

and C. Woolsey. 1949. Organization of the

mammalian thalamus and its relation to the cere

bral cortex. Electroenceph. Clin. Neurophysiol.

1:391-404.

Royce, G. J., G. F. Martin, and

N. Dom. 1975.

Functional organization and cortical architecture

in the nine

banded armadillo

(Dasjpus novemcinc-

tus).

Comp. Neurol. 164:495-522.

Russell,

S. 19 16.

Form and function.

John Murray,

London.

Sacher, G. A. 1970. Allometric and factorial analysis

of brain structure in insectivores and primates.

R. Novack and W. Montagna (eds.),

The

primate brain: Advances

primatology,

pp. 245-

287. Appleton Century Crofts, New York.

Sacher, G. A. and

F. Staffeldt. 1974. Relation of

gestation time to brainweight for placental mam

mals: Implications for the theory of vertebrate

growth. Am. Nat. 108:593-615.

Sanides,

1969. Comparative architectonics of the

neocortex of mammals and their evolutionary

interpretation. Ann. N.Y. Acad. Sci. 167:404-

423.

Sanides, F. 1970. Functional architecture of motor

and sensory cortices in primates in the light of a

new concept of neocortex evolution.

C. Noback

and W. Montagna (eds.),

The primate brain:

Advances

primatology,

pp. 137

208. Appleton

Century Crofts, New York.

Sanides, F. 1972. Representation in the cerebral cor

tex and its areal lamination patterns.

G. F.

Bourne (ed.),

Structure and function of nervous tis-

sue,

Vol. 5, pp. 329

453. Academic Press, New

York.

Sanides, F. 1975. Comparative neurology of the tem

poral lobe in primates including man with ref

erence to speech. Brain Lang. 2:396-419.

Saraiva, P. E. S. and B. Magalh„es-Castro. 1975. Sen

sory and motor representation in the cerebral

cortex of the three

toed sloth

{Bradypus tridacty-

lus).

Brain Res. 90: 181

1 93.

Sarnat, H. B. and

G. Netsky. 1981.

Evolution of

the nervous system.

Oxford University Press, New

York.

Seltzer, B. and D. N. Pandya. 1980. Converging visual

and somatic sensory cortical input to the intra-

parietal sulcus of the rhesus monkey. Brain Res.

192:339-351.

Semsa, M., V. Casagrande, and J. Kaas. 1984. Cor

tical connections of area 17 in tree shrews.

Comp. Neurol. 230:337-35 1.

Sherrington, C. 1906.

The integrative action of the ner-

vous system. Y

ale University Press, New Haven.

Smith

Gill, S.

1983. Developmental plasticity:

Developmental conversion versus phenotypic

modulation. Amer. Zool. 23:47-55.

Snell,

1891. Das Gewicht des Gehirns und des

Hirnmantels der Saugetiere in Beziehung zu

deren geistigen Fahigkeiten. Sitz. Ges. Morph.

Physiol. (Munchen) 7:90-94.

Sohal, G. S. 1976. An experimental study of cell

death in the developing trochlear nucleus. Exp.

Neurol. 5 1

:684-698.

Sokoloff, A.,

Deacon, and G.

Goslow. 1989.

Musculotopic innervation of the primary flight

muscles, the pectoralis (pars thoracicus) and

supracoracoideus, of the pigeon

(Columba livia):

HRP study. Anat. Rec. 225:35-40.

Spencer, H. 1980.

Principles of psychology,

2nd ed.

Longman Press, London.

Stanfield, B. B. and

O'Leary. 1985. Fetal

occipital cortical neurons transplanted to the

ros-

tral cortex can extend and maintain a pyramidal

tract axon. Nature 3 13: 135-137.

Stephan, H. 1969. Quantitative investigations on

visual structures in primate brains.

Proceed

ings of the Second International Congress on Pri

mates 3:34-42. Karger, Basel.

Stephan,

H.,

H. Frahm, and G. von Baron. 1981.

New and revised data on volumes of brain struc

tures in insectivores and primates. Folia Prima-

tologica 35: 1

29.

Sur, M., P. E. Garraghty, and A. W. Roe. 1988.

Experimentally induced visual projections into

auditory thalamus and cortex. Science 242: 1437-

1441.

Thompson,

17.

On growth and form.

Cambridge

University Press, Cambridge.

Tigges, J., W. Spatz, and M. Tigges. 1973. Recip-

rocal point

point connections between para-

striate and striate cortex in the squirrel monkey

(Saimiri).

Comp. Neurol. 148:481-490.

Tigges,

J.,

Tigges, and A. Perachio. 1977. Com

plementary laminar terminations of afferents to

area 17 originating in area 18 and in the lateral

geniculate nucleus in the squirrel monkey.

Comp. Neurol. 176:87-100.

'Tobias, P. V. 198

The emergence of man in Africa

and beyond. Philos. Trans.

Soc. London B,

Biol. Sci. 292:43-56.

Tower, D. B. 1954. Structural and functional orga-

nization of the mammalian cerebral cortex. The

correlation of neuron density with brain size.

Cortical density in the finwhale with

note on

the cortical neurone density in the Indian ele

phant.

Comp. Neurol. 101:19-53.

Ulinski, P. S. 1983.

Dorsal ventricular ridge.

Wiley,

New York.

Ulinski, P.

1986. Neurobiology of the therapsid-

mammal transition.

N. Hottén, P. D. Mac-

Lean,

Roth, and E. Roth (eds.),

The ecology

and biology of mammal-like reptiles,

pp. 149

17 1.

Smithsonian Institution Press, Washington, D.C.

Valverde,

and L. López-Mascaraque. 198

Neo-

cortical endeavor: Basic neuronal organization in

the cortex of the hedgehog.

Eleventh international

congress of anatomy, glial

and

neuronal cell biology.

Alan R. Liss, New York.

Van Valen, L.

1982. Homology and causes. J. Morph.

173:305-312.

Walsh, C. and C. L.

Cepko.

1988.

Clonally related

cortical cells show several migration patterns. Sci

ence

141:1342-1345.

Walsh,

and

F. F.

Ebner.

1970.

The cytoar-

chitecture of somatic sensory

motor cortex in the

opossum (Didelphis marsupialis virginiana), a Golgi

study. J.

Anat. 107:1-18.

Warren, S. and M.

Carlson.

1986.

Topography of

primary somatic sensory cortex (SI) in Old World

primates, Cercopithecus and Myopithecus.

Soc. Neu-

rosci. Abstr.

12:386.1.

Wilczynski,

1984.

Central neural systems sub-

serving a homoplasous periphery. Amer. Zool.

24:755-763.

Wiley, E. O. 1981.

Phylogenetics.

Wiley, New York.

Woolsey, T.

A. and J.R. Wann.

1976.

Areal changes

in mouse cortical barrels following vibrissal dam

age at different postnatal ages. J.

Comp. Neurol.

170:53-66.

Würsig,

1989.

Cetaceans. Science

244:1550-1557.

Yakovlev, P. I.

and A.

R. Lecours.

1967.

The myelo-

genetic cycles

regional maturation of the

brain.

A. Minkowski (eds.), Regional development of the

brain in early life. A symposium. F. A.

Davis

Co.,

Philadelphia.

Zilles,

and G. Rehkämper.

1988.

The initial brain

concept:

work in progress. Behav. Brain Sci.

11:105-106.

The Brain

Chapter

Feb 2025

Definition Brain is a complex organ encased within the cranium , acts as the activity center of the body, controlling all the functions such as thinking, memory, emotion, touch, motor skills, vision, breathing, temperature, and hunger. Brain together with the spinal cord constitutes the central nervous system. Brain Brain is cluster of billions of neurons enclosed within the cranium that continues with the spinal cord of the organism located at the anterior end of the all vertebrates and in some highly organized invertebrate organisms. Initially, vertebrate brain was believed to be moderately developed with anterior region of neural tube responsible for the local movement of head and neck region along with sensory stimuli. The brain controls all the activities of an organism by integrating sensory information and directing the motor response of the organism; in higher vertebrates it is the center of learning as well. The basic unit of the brain is neurons, which is supported by glial cells. In lower invertebrates, brain is much simpler and tubular in structure resembling early developmental stage of the brain in higher vertebrate (Dyakonova et al., 2022). Although the brain works all together to perform and process various activities, it can be divided into three major parts that are forebrain, midbrain, and hindbrain. Forebrain is the largest and highly developed part of the brain, and it consists of three major regions: cerebrum, thala-mus, and hypothalamus. Midbrain is the smallest part of the brain; it is located at the top of the brain stem, in between the medulla and pons region of the hindbrain, and it connects the brain with the spinal cord. Hindbrain is the lowermost part of the brain and composed of medulla, pons, cerebel-lum, and spinal cord (Reichert, 2009). Brain Development During the embryonic stage, the brain is a small expanded area of the neural tube that develops rapidly almost faster than any other organs. Neu-ral tube increases in size and bends toward the anterior end, folding downward and forming

Significance of Evolutionary Lags in the Primate Brain Size/Body Size Relationship

Preprint

Full-text available

Feb 2024

Robin I. M. Dunbar

INTRODUCTION The original brain lag hypothesis proposed that primate brain evolution depended on spare energy derivative of savings of scale enabled by increasing body size. Deaner & Nunn [1] concluded that, in fact, there was no evidence for a brain lag. However, their result may have been due to a number of possible confounds in their analysis. METHODS I revisit their analysis to test for potential confounds using updated datasets. I also ask how primates paid for the energy costs incurred by changes in brain and body mass, and whether the impetus for these changes was predation risk. Finally, I ask whether the observed patterns explain the brain/body size ratio trajectory observed in fossil hominins. RESULTS I show that using statistically more appropriate statistics and updated data yields a significant brain lag effect. However, contrary to the original brain lag hypothesis, the brain/body ratio does not converge back on the allometric regression line, but continues to evolve beyond it. Increases in brain size are correlated with exploiting large group size rather than body size as the principal defence against predation risk, with significant growth in brain size (but not body size) only being possible if species adopted a more frugivorous diet. Finally, I show that hominins followed a similar trajectory from an australopithecine baseline that fell on the relevant allometric regression. CONCLUSION The brain lag effect is much more complicated than the original hypothesis proposed, with a distinctive switch from body to brain over evolutionary time.

About leaving the neuroscience laboratory

Chapter

Full-text available

Jan 2024

Antonella Tramacere

Bringing together recent case studies and insights into current developments, this collection introduces philosophers to a range of experimental methods from neuroscience. Chapters provide a comprehensive survey of the discipline, covering neuroimaging such as EEG and MRI, causal interventions, for instance brain stimulation or psychopharmacology, advanced statistical methods, and approaches drawing on research into the development of human individuals and humankind. A team of experts combine clear explanations of complex methods with reports of cutting-edge research, advancing our understanding of how these tools can be applied to further philosophical inquiries into agency, emotions, enhancement, perception, personhood and more. With contributions organised by neuroscientific method, this volume provides an accessible overview for students and scholars coming to neurophilosophy for the first time, presenting a range of topics from responsibility to metacognition.

Shamanic shape-shifting and second-order isomorphism

Article

Jun 2023

Timothy L Hubbard

A cognitive account of shape-shifting that is based on a second-order isomorphism involving the functional architecture of human mental representation and the human genome is suggested. Characteristics of shape-shifting from multiple sources (Shamanism, cave art, mythology, folktales, popular culture) are briefly reviewed, and a distinction between physical shape-shifting and mental shape-shifting is noted. The idea of isomorphism (a one-to-one mapping between elements of different systems or domains) is reviewed, and prior applications of the isomorphism notion within psychology (Gestalt theory, mental imagery, spatial localization) are briefly considered. A distinction between first-order isomorphism (involving structure) and second-order isomorphism (involving function) is noted, and the possibility that reported experiences of shape-shifting involve access and activation of second-order isomorphism information contained within the functional architecture of mental representation is suggested. The possible connections between a second-order isomorphism account of shape-shifting and a wide range of issues in human physiology (mimesis, mirror neurons, stem cells, neural pruning, phantom limb syndrome), human cognition (embodied cognition, qualia, literacy, information access, memory, the computer metaphor, holographic representation), and altered states of consciousness (dreaming, mystical states, alchemy, exceptional human experiences) are discussed.

Evolution of the Brain and Sensory Structures in Metatherians

Chapter

Nov 2022

The metatherians (crown-clade marsupial mammals and their fossil relatives) originated in the Late Jurassic or Early Cretaceous of Laurasia, and have since spread worldwide with diversification in South America and Australasia during the Cenozoic. Despite this long evolutionary history, paleoneurology is known for a few taxa from the Americas and Australia, with most work being published in the last 40+ years. Here, we contextualise research on metatherian paleoneurology with traditional tenets that marsupials are developmentally constrained in their brain size and advanced cognitive and sensorimotor capabilities. We summarize recent research on marsupial neuromorphology with a perspective on how these insights apply to extinct species. We describe a digital cranial endocast of the didelphid Caluromys philander to compare with endocasts of crown and stem marsupials. Although endocasts of basal metatherians morphologically resemble those of didelphids, there is significant variation in brain shape and cerebrum gyrification among marsupials, possibly due to differences in how neural tissue is distributed within limited braincase space. Lastly, we examined existing endocranial volume and body mass estimates for crown and stem marsupials. The earliest metatherians have substantially smaller relative brain sizes than recent species, although this may relate to errors in estimating metatherian mass and endocast volumes.

Do Animals Have Consciousness?

Chapter

Jul 2024

Ludwig Huber

So far, I have dealt with numerous animal thought processes that are good candidates for attribution of consciousness. These are thought processes that are usually not possible in humans without consciousness; they include logical reasoning, creative tool use and production, causal understanding and insightful problem-solving, volitional action, time travel (episodic memory, planning), metacognition, and perspective taking. (This list is not exhaustive; one could add imitation, teaching, mirror self-recognition, quantity estimation, abstraction, categorisation, and concept formation to the list.) But before I try to answer the question of whether animals act consciously in the aforementioned thought processes, I should first clarify what I mean by the term “consciousness”. This term has many meanings and, above all, very strong connotations for human science and everyday language.

The Central Role of the Individual in the History of Brains

Article

May 2024

Human temporal lobes have been reorganized: A response to Pearson et al, “updated imaging and phylogenetic comparative methods reassess relative temporal lobe size in anthropoids and modern humans”

Article

Jun 2023

James K Rilling

References

Chapter

Apr 2023

Jonathan Parker

Evolution of the Brain and Sensory Structures in Sirenia

Chapter

Nov 2022

Sirenians represent a striking group of quadruped mammals that are adapted to marine life. There are four extant species of sirenians (three sea cows and the dugong) but the fossil record, which extends from the early Eocene, shows a much richer diversity, and global distribution. Sirenians belong to the crown-group Sirenia and their closest extinct relatives belong to Pan-Sirenia. Sirenia is nested within Tethytheria, along with elephants and desmostylids, and Tethytheria belongs to the larger mammalian clade Afrotheria. The paleoneurology of Pan-Sirenia and Sirenia is based on anatomical descriptions of approximately two dozen natural and digital cranial endocasts and a couple of bony labyrinth endocasts. Sirenian brains are characterized by lissencephalic cerebral hemispheres, reduced olfactory system, reduction in the optic nerves, presence of a large trigeminal nerve and associated components, and thick meninges of the central nervous system. In contrast to other modern paenungulates, which include the gyrencephalic and large-brained elephants, sirenians have lissencephalic brains whose components are linearly arranged, and small relative brain sizes (EQs). Future work is needed to further incorporate the anatomical diversity of sirenian cranial endocasts into phylogenetic character matrices and to increase our knowledge about the inner ear evolution of the group.

The Biology and Evolution of Language

Book

Full-text available

Jan 1984

Philip Lieberman

Allometric and factorial analysis of brain structure in insectivores and primates

Article

Jan 1970

G.A. Sacher

Evolution and development of language

Article

Jan 2013

Parallel Organization of Functionally Segregated Circuits Linking Basal Ganglia and Cortex

Article

Jan 1986
ANNU REV NEUROSCI

The organization of the avian telencephalon and some speculations on the phylogeny of the amniote telencephalon

Article

Jan 1969

Harvey J Karten

Developmental constraints in evolutionary processes

Article

Jan 1982

P. Alberch

Regressive Events in Brain Development and Scenarios for Vertebrate Brain Evolution (Part 2 of 2)

Article

Jan 1987

The problems of the evolution of varying brain size, the specialization of particular functional systems and overall differences in the relative complexity of brain organization are discussed in terms of alterations of regressive events in neurogenesis (cell death and axon retraction). Three scenarios for evolution, cascade reorganization, parcellation and heterochrony, are considered in light of regressive mechanisms during development.

Representation in the Cerebral Cortex and Its Areal Lamination Patterns

Chapter

Dec 1972

Friedrich Sanides

Some problems relating to the evolution of the brain

Article

G. Elliot Smith

Some Questions and Problems Related to Homology

Chapter

Jan 1982

C. B. G. Campbell

Comparisons of human brains with those of nonhuman primates and other animals have played a large role in the making of inferences about human origins and evolution. Originally, brain size and gross morphology were the primary objects of comparison. Sir Wilfrid Le Gros Clark, along with some others, was instrumental in placing importance on similarities and differences in internal structure as revealed by descriptive and experimental histology. The concept of homology is central to the making of such comparisons between animals.