DISCLAIMER: In the following article, I have quoted liberally
from the copyrighted works I discuss below (though no more than
one might see in a college term paper). Before anyone asks: no,
I have not sought the permission of the author or his publisher
to republish these quotes on this mailing list. You may think
this discourteous (I doubt if it's illegal), but I announce it in
advance to take the wind out of the sails of any copyright prigs
who might be spluttering with indigation about it. Nevertheless,
while anybody on the Extropians' may make any use whatever of
this posting as far as I'm concerned, be warned that you must use
the copyrighted portions at your own peril (that also goes for
the list archive maintainer, who may wish to expunge this
sequence of posts rather than bear any risk of copyright
infringement). I started this article just for fun (it began as
a book review and then got longer), and I was almost finished
with it before the thought of copyright crossed my mind. But
then it did, and while I'm submitting the article anyway, I'm
also prepending this slight cringe.
A SPRING-POWERED THEORY OF CONSCIOUSNESS
I don't know if anyone has mentioned or reviewed this book on the
Extropians' list (the search engine at
http://www.lucifer.com/exi-lists hasn't been working for a while
:-< ), but I thought I'd make a few rambling remarks about the
recently-published _A Universe of Consciousness_ by Gerald
M. Edelman (in collaboration with Giulio Tononi), subtitled "How
Matter Becomes Imagination" (Basic Books, New York, 2000). I'll
also attempt to put these remarks in a larger context by drawing
on Dr. Edelman's earlier books.
Edelman received a 1972 Nobel in Physiology and Medicine for his
work in immunology (there is a bio of Edelman as of his 1972
Nobel at http://www.nobel.se/laureates/medicine-1972-1-bio.html).
Since then, he has studied the embryology of the vertebrate
nervous system (see the Chairman's Overview of the Department of
Neurobiology at the Scripps Institute:
http://www.scripps.edu/admin/dev/Overviews.html#NBIO). More
recently, (currently as director of the Neurosciences Institute
in San Diego [http://www.nsi.edu]) he has been engaged in a
rather ambitious attempt to create a no-nonsense theoretical
foundation for the scientific study of consciousness itself (both
the kind we share with other animals, and the kind that is
supposed to be possessed only by human beings). Edelman calls
the former "primary consciousness": the process which permits an
animal to categorize the cacophony of "unlabeled" signals pouring
in through various sensory modalities into objects and scenes,
and to make rapid use of these signals to select among behavioral
alternatives. The so-called "higher-order" consciousness that
seems to be unique to human beings is based on speech and other
symbol systems (including formal logic), and encompasses the
concept of the self in time past, present, and future, and the
consciousness of being conscious.
I first ran into Edelman eight years ago, when I picked up a copy
of his book written for the popular audience _Bright Air,
Brilliant Fire: On the Matter of the Mind_ (Basic Books, New
York, 1992). I was impressed by this book for two reasons --
first, because it took biological underpinnings much more
seriously than discussions I had previously encountered on the
subject of human cognition, and secondly because of the rather
breathtaking (not to say arrogant) scope of the book -- nothing
less than a complete theory of how the human mind results from
the structure and function of the brain. Reading this book, I
could (and still can) convince myself that I am getting a
plausible story of the means by which the physical organization
and operation of the brain might result in the subjective
experiences and objective behavior we call "mind", rather than
the usual miscellaneous collection of facts about neuroanatomy
and psychology, and symbolic models of this or that aspect of
cognition, contained in such books. Most volumes about the mind
and brain that one might encounter at the local Borders or Barnes
& Noble contain no particular overarching theory, and typically
shy away from any provocative claims (at least once you're past
the cover and title page) of actually being able to "explain" the
human mind. On the other hand I have sometimes been daunted, in
my efforts to follow Edelman's arguments, by what seems to be a
rather tortured writing style that has caused me to have to read
and re-read some passages of his books without, in the end, being
entirely confident that I have understood them. In this regard,
Edelman reminds me a little of B. F. Skinner, with whom I
acquainted myself in the early 70's (via _Beyond Freedom and
Dignity_, _Science and Human Behavior_, and other texts), and who
had an idiosyncratic style and vocabulary that took some getting
used to (all in the name of precision, I suppose, from the
author's point of view).
_Bright Air, Brilliant Fire_ was intended by its author as a
summary, and popular exposition, of some of his earlier books.
After reading it I came across and purchased (pre-Amazon) some of
Edelman's more specialist-oriented tomes (at the local mall's
Borders, of all places -- the then-new brick-and-mortar
mega-bookstores were amazing places!), including _Neural
Darwinism: The Theory of Neuronal Group Selection_ (Basic Books,
New York, 1987), _The Remembered Present: A Biological Theory of
Consciousness_ (Basic Books, New York, 1989), and _Topobiology:
An Introduction to Molecular Embryology_ (Basic Books, New York,
1988). I read all of _The Remembered Present_ a couple of years
ago, but I have to admit that the other two volumes sit mostly
unread on my shelves (except for the browsing I've been doing
recently in _Neural Darwinism_ in conjunction with writing this
article).
I have, since then, learned that Edelman has his detractors -- a
short quote from the _New York Times Book Review_ article on _The
Remembered Present_, at
http://www.ftrbooks.com/psych/biological_consciousness/remembered_present.htm
written by one Stuart Sutherland (presumably the late British
psychologist, see http://epunix.susx.ac.uk/EP/stuart.html), gives
a taste of one Edelman critic. Edelman is also lambasted in John
Horgan's 1996 book _The End of Science_ (Little, Brown; London;
1996, pp. 165-172; Horgan's interview with Edelman is,
admittedly, snottily funny). F. H. C. Crick's dismissal of
"Neural Darwinism" (both the book and the theory, presumably) as
"Neural Edelmanism" was actually chosen by Edelman as the
epigraph to Chapter 9 of _Bright Air, Brilliant Fire_ (p. 81).
However, having no one but myself to please when it comes to
choosing books for my own entertainment, and having a somewhat
faded memory of my earlier forays into "Edelmanism", I was
motivated to purchase and read Edelman's latest attempt at
popular explication of his ideas, both to see what might be new
and also in hope of a clearer and more engagingly-written
restatement of the earlier material. I first contemplated
writing this article as a review of the newest book, but then it
became an attempt, for my own benefit if no one else's, to
refresh my memory of Edelman's earlier books and to place the
latest book in context with them.
So here follow my rambles -- a selection of the points that
remain in my head a few days after having finished _A Universe of
Consciousness_ ["UoC"] (and months or years after having read
_The Remembered Present_ ["RP"] and _Bright Air, Brilliant Fire_
["BABF"], though with a good deal of refurbishment from
consulting those earlier volumes, and _Neural Darwinism_ ["ND"]
too, in order to write this paper). Any inaccuracies or
misrepresentations in my remarks are strictly my own fault
(dammit I'm a computer programmer not a neurobiologist, Jim!) My
intention is merely to provide motivation to anyone on this list
who might be ripe for a dip into these volumes (possibly with a
more serious purpose than I have). I am certainly not in a
position to offer expert scientific criticism of any of Edelman's
ideas, or even to compare and contrast Edelman's popular
expositions of his ideas with those of other authors (such as
William H. Calvin, a few of whose books I have purchased
recently, partly as a result of getting back into Edelman, but
which I have not yet read). For a bird's-eye view of much of
what follows, see the diagrams in RP pp. 241 (Fig. 14.1), 256
(Fig. 15.1); BABF p. 134 (Fig. 12-5).
Early in UoC (p. 14; see also RP p. 19; BABF p. 113), Edelman
makes clear that his theorizing is based on what he calls the
"physics assumption". He dismisses dualism (any notion that
there's a special "mind stuff" existing on a different physical
plane or in a different explanatory realm from ordinary matter or
ordinary physical processes [cf. the philospher's joke "What is
mind? No matter. What is matter? Never mind." mentioned by
Bertrand Russell in his autobiography; the joke was a favorite of
his grandmother's, and expressed her attitude of derision toward
"metaphysics": _The Autobiography of Bertrand Russell_ Vol. 1,
p. 48 of the Bantam paperback edition]). In the next breath
Edelman dismisses the notion that consciousness is based on
"exotic" physics such as quantum gravity (as invoked, for
example, by Sir Roger Penrose, see BABF pp. 212-218). He also
espouses what he calls the "evolutionary assumption" -- that
whatever consciousness is, it is the product of the phylogenetic
history and evolution by natural selection of life on earth,
which encompasses the origin and history of the human species.
Edelman believes that individual neurons do not have independent
functions in the brain, but participate in larger entities he
calls "neuronal groups", at several levels of organization. The
lowest-level groups are those of the "primary repertoire": these
are distinct sets of neighboring cells, each containing between
50 and 10,000 neurons (RP p. 47 [Fig. 3.2]; BABF pp. 88-89
[Fig. 9-3]) which have, during the course of embryogenesis and
infancy, formed connections with each other preferentially to
their connections with neurons outside their own group. This
"developmental selection" of brain wiring (UoC p. 84 [Fig. 7.2];
ND Chap. 4, p. 199 [Fig. 7.8]; RP pp. 44, 45 [Fig. 3.1], 46,
240-242; BABF pp. 83, 84 [Fig 9-1]) takes place under the
influence of the expression of morphoregulatory genes for Cell
Adhesion Molecules, Substrate Adhesion Molecules, and Cell
Junction Molecules (CAMs, SAMs, and CJMs) see ND pp. 158-159
[Fig. 6.3]; BABF all of Chap. 6, esp. pp. 60, 61 [Fig. 6-4], and
62; this is also the subject matter of _Topobiology_). It occurs
under juxtaposed conditions of mechanical and chemical constraint
and confinement, in interaction with the supportive glial cells
of the brain (ND pp. 112, 113 [Fig. 5.1]), and is also
constrained and directed by the ongoing signalling of the
developing brain, sensory, and motor systems -- the maxim here
being "neurons that fire together, wire together" (UoC p. 83).
Edelman cites both theoretical considerations and experimental
evidence in favor of the existence of the primary repertoire of
neuronal groups, including the fact that while an individual
neuron is either excitatory or inhibitory, neuronal groups can be
both, in varying proportions (RP p. 51; BABF pp. 86-87).
"[G]roups come in different sizes, shapes, and types, depending
upon their underlying anatomy, developmental constraints,
synaptic properties, and inputs. Some groups are more stable
than others; for instance, some found in the visual cortex are
fixed during developmental critical periods, while some in the
somatosensory cortex are plastic. A good example in the visual
system is provided by orientation columns, the properties of
which fulfill the criteria for neuronal groups: collections of
locally and strongly interconnected neurons carrying out a
discriminative function developed as a result of selectively
correlated input signals (in this case tuning for orientation)"
(RP p. 52). "[T]here is no evidence that any... neuron has its
properties **except** as a result of its synaptic interactions
with multiple other neurons in a group and with distant neurons
by reentry" (RP p. 53).
At the early, rapid-growth stage of developmental selection,
there is a literal Darwinian competition for survival among the
still-proliferating neurons, and only those which manage to
become functional members of neuronal groups escape death (ND
pp. 111 [Table 5.1], 115; BABF pp. 23 [Fig. 3-3], 25). Up to 70%
of cells die during the development of some regions of the brain.
"Most of this death is not preprogrammed but depends upon the
neuron connecting to the appropriate innervation field. This
must occur largely epigenetically and to some extent
stochastically..." (ND p. 115).
An important property of the neuronal groups which remain
following this process is a certain degree of random variation in
the connections within and between them, which nevertheless does
not compromise their functional adequacy. Edelman calls this
lack of precise structural specificity "degeneracy" (defined in
UoC on p. 86 as "the capacity of structurally different
components to yield similar outputs or results"). He claims that
it occurs at many levels of organization in the brain (ND p. 58
[Table 3.3]), and is an inevitable outcome of selectional
processes as well as providing a source of variability that can
be the basis for further ongoing selection (UoC pp. 86-87; ND
pp. 46-59; RP pp. 50, 52-53). Degeneracy is not merely tolerated
but is essential to the brain's operation, "to provide the
overlapping but nonidentical response characteristics needed to
cover a universe of possible stimuli" (ND p. 50; RP p. 242).
An outcome of degeneracy is that there can be many alternative
means and pathways, competitively selected out of a large
population of variant possibilities, which can accomplish more or
less adequately the same functional task. The detailed physical
structures which participate in a given task will therefore vary
stochastically among brains depending on the vagaries of chance
and personal history: no two brains (even those of identical
twins) will contain identical populations of neurons or be wired
identically (ND p. 34 [Fig. 2.6]; BABF pp. 25, 26 [Fig. 3-5]).
Furthermore, the information contained in the human genome would
be insufficient to precisely specify the synaptic structure of
the developing brain (BABF p. 224). Therefore, it certainly
cannot be the case that the brain is precise and "hardwired" like
a computer (BABF p. 27).
Superimposed on the process of developmental selection, and
continuing throughout life, is a second process of "experiential
selection" (UoC p. 84; ND Chap. 7; RP p. 46; BABF p. 84) in which
the existence of a given set of synaptic connections among
neurons and neuronal groups is more-or-less fixed, but in which
the strength of these connections can become greater or less over
time. The synaptic circuits formed by the simultaneous activity
of connected neuronal groups generally have a much greater
spatial extent than the groups which comprise them (ND p. 166).
Edelman considers the so-called "secondary repertoire" as a
population of variant possibilities for widely-distributed
patterns of synchronized firing of neuronal groups, whose
participant synapses have been jointly strengthened, in varying
degrees, due to repeated correlated activity in the past. In
contrast to classical Darwinism, which is predicated on
differential **reproduction**, the partitioning of the primary
repertoire of neuronal groups into the secondary repertoire of
synaptic circuits is based on differential **amplification** (of
the strength of synaptic connections [RP p. 53; BABF pp. 94-97]).
The widely-distributed firing patterns of the secondary
repertoire flicker in and out of existence depending on events in
the organism and the world, with neuronal groups of the primary
repertoire participating in different patterns of the secondary
repertoire at different times.
The existence of neuronal groups means that intra-group synaptic
connections ("intrinsic connections") among the neurons in a
group of the primary repertoire are denser than the inter-group
("extrinsic") connections among the groups which create the
circuits of the secondary repertoire. In artificial neural
networks lacking this hierarchy of interconnections, a learning
rule (such as the Hebb rule) is typically chosen which permits
synaptic strength to be changed by the correlated firing of the
pre- and post-synaptic neuron; however Edelman cites experimental
evidence (ND p. 181) suggesting that, in real brains, a large
number of additional neurons, all connected to the presynaptic
neuron, must participate in the correlated firing in order for
the strength of the connection to a postsynaptic neuron to change
(ND p. 201). An interconnection hierarchy based on the existence
of neuronal groups has consequences for the learning laws by
which long- and short-term modifications of these circuits can
occur, which are examined mathematically in ND (pp. 196-198), and
which differ from those which characterize symmetrically-wired
artificial neural networks (see also BABF p. 227).
Analogously to his view of the insignificance of the **individual
neuron**, Edelman minimizes the contribution of the **individual
synapse** to the behavior of the entire network: "[T]he
complexity of the nervous system makes it highly unlikely that
there is a simple relationship between changes at any individual
synapse and changes in the behavior of the network. It is much
more likely that multiple synaptic modifications occur
simultaneously at various sites in the network, reflecting its
degeneracy" (ND p. 179). The author also emphasizes the
combinatorial increase in the richness of interactions made
possible by the diversity of neurotransmitters and their
receptors (ND pp. 203-204): "A rich pharmacology thus assures a
very rich set of functional network variants" (ND p. 204).
Edelman's motivation for insisting on selection at the level of
groups rather than of individual neurons seems to be that only
the latter, in his view, permits a sufficiently fine-grained
variation in the characteristics of the units of selection to
allow functional niches to be filled by selective competition
(under constraints of value) rather than by exact matching of
neurons to prespecified correct responses (ND p. 47).
Prespecified roles for individual neurons do occur in current
neural-network models of computation (RP p. 53; BABF p. 227).
Edelman mocks these models, in which "individual neurons each
correspond to a pattern of external events of the order of
complexity of events symbolized by a word", by describing them as
based on the "grandmother neuron" concept (i.e., the idea that
there's a particular neuron that fires when you recognize your
grandmother), which "cannot explain categorization without
begging the question" (RP p. 53).
Edelman asserts that the results of developmental selection are
"grossly speaking, two kinds of nervous system organization that
are important to understanding how consciousness evolved. These
systems are very different in their organization, even though
they are both made up of neurons. The first is the brain stem,
together with the limbic (hedonic) system, the system concerned
with appetite, sexual and consummatory behavior, and evolved
defensive behavior patterns. It is a value system; it is
extensively connected to many different body organs, the
endocrine system, and the autonomic nervous system. Together,
these systems regulate heart and respiratory rate, sweating,
digestive functions, and the like. It will come as no surprise
to learn that the circuits in this limbic-brain stem system are
often arranged in loops, that they respond relatively slowly (in
periods ranging from seconds to months), and that they do not
consist of detailed maps. They have been selected during
evolution to match the body, not to match large numbers of
unanticipated signals from the outside world. These systems
evolved early to take care of bodily functions; they are systems
of the interior" (BABF p. 117; see also RP p. 152; UoC pp. 42, 43
[Fig. 4.4], 44-47).
The second topology which plays a key role in Edelman's theory is
characteristic of the thalamocortical system: "The thalamus, a
central brain structure, consists of many nuclei that connect
sensory and other brain signals to the cortex... The
thalamocortical system consists of the thalamus and the cortex
acting together, a system that evolved to receive signals from
sensory receptor sheets and to give signals to voluntary muscles.
It is very fast in its responses (taking from milliseconds to
seconds), although its synaptic connections undergo some changes
that last a lifetime... [I]ts main structure, the cerebral
cortex, is arranged in a set of maps, which receive inputs from
the outside world via the thalamus" (BABF p. 117, see also RP
pp. 152, 165 [Fig. 9.4]; UoC p. 107).
In the cortex, neuronal groups are organized in the form of an
"interconnected six-layered sheet of about ten billion neurons
with about a million billion connections... arranged in
functionally segregated maps that... subserve all the different
sensory modalities and motor responses" (BABF p. 104). Edelman
calls the maps into which the cortex is partitioned "local maps"
(ND Chap. 5; RP p. 243; BABF pp. 22-25). Further, the local maps
of the cortex are in turn connected with each other and with the
thalamus via massively parallel sets of reciprocal fibers, the
whole forming a three-dimensional meshwork whose structure
typifies the neural architecture of the thalamocortical system
(Uoc p. 42). "Any perturbation in one part of the meshwork may
be felt rapidly everywhere else. Altogether, the organization of
the thalamocortical meshwork seems remarkably suited to
integrating a large number of specialists into a unified
response" (UoC p. 45).
The value systems (or, in this context, "diffusely projecting
value systems" [UoC p. 43]) which control the body via the
autonomic nervous and endocrine systems also connect to the
thalamocortical system via ubiquitous fibers that fan out
diffusely from various nuclei in the brainstem and hypothalamus,
such as the serotonergic raphe nucleus, the dopaminergic nuclei,
the cholinergic nuclei, and the histaminergic nuclei, and that
project to most or all of the brain. The locus coeruleus "sends
a diffuse sweeping 'hairnet' of fibers to cover the entire
cortex, hippocampus, basal ganglia, cerebellum, and spinal cord,
and ... has the potential to affect billions of synapses". The
"neuromodulators" released at the synaptic connections formed by
the diffuse projections of the value systems both bias the
moment-to-moment activity of the circuits they project to, and
affect their long-term plasticity (UoC pp. 46, 87-92). They can
signal to the entire brain the sudden occurrence of a
pleasurable, painful, or novel stimulus, and together with other
mechanisms of value (such as hormonal loops) provide input to the
thalamocortical system related to the "adaptive, homeostatic, and
endocrine functions" of the organism (UoC pp. 88-89; RP
pp. 93-94; BABF p. 163).
Edelman also uses the term "value" in a more generalized sense to
mean any boundary condition imposed by the organism's phenotype
which constrains the developmental and experiential selection
taking place in the brain, such as the way limbs are jointed (UoC
p. 88): "the mere fact of having a hand with a certain shape and
a certain propensity to grasp in one way and not another
enormously enhances the selection of synapses and neuronal
patterns of activity that lead to appropriate actions. The same
actions would be almost impossible to synthesize or program from
scratch, as experts in robotics know all too well."
The characteristic architecture of parallel, reciprocally
interconnected local maps found in the thalamocortical system is
called by the author "reentry" (UoC p. 105; ND p. 60; RP p. 47,
Chap. 4; BABF p. 90 [Fig. 9-4], and is another key notion in
Edelman's portfolio. "Reentry is a process of temporally ongoing
parallel signalling between separate maps along ordered
anatomical connections. Reentrant signalling can take place via
reciprocal connections between maps (as seen in corticocortical,
corticothalamic, and thalamocortical radiations); it can also
occur via more complex arrangements such as connections among
cortex, basal ganglia, and cerebellum" (RP p. 49).
The basal ganglia and cerebellum (together with the hippocampus)
are instances of what Edelman calls "cortical appendages" (BABF
p. 105 [Fig. 10-2], which function as "organs of succession"
responsible for the sequencing of motor events controlling
behavior, and of perceptual events comprising memory and the flow
of consciousness (UoC pp. 45, 95; ND p. 227; RP p. 113, Chap. 7;
BABF p. 105). "To handle time as well as space, the cortical
appendages -- the cerebellum, basal ganglia, and hippocampus (see
figure 10-2) -- evolved along with the cortex to deal with
succession both in actual motion and in memory" (BABF p. 118).
The reentrant connections between thalamocortical maps can be
registered, convergent or divergent; arborized or non-arborized;
layer-registered or non-layer-registered (RP p. 65 [Table 4.1],
p. 66 [Fig. 4.1A]). Temporally, reentrant signals may be
periodic, intermittent (phasic), synchronous, or asynchronous (RP
p. 68), and can be recursive via many paths (RP pp. 65, 69). A
particular fiber projecting from one receptor sheet or map may
have simultaneous connections to more than one distinct target
map: "When maps are connected by reentrant fibers, the individual
fibers generally extend their arbors over many locally linked
neurons... When secondary repertoires are formed, the
strengthening of synapses **within** those arbors may then select
neighboring groups of neurons, changing borders over smaller
dimensions than those of the arbors" (BABF pp. 86-87). Reentry
may be "horizontal" among maps at a given layer of the cortex, or
"vertical" among successive layers (RP p. 67). If there are
major changes in the characteristic sensory inputs to which an
organism is exposed, there will be corresponding changes in the
boundaries of the associated maps, and corresponding
readjustments to the successive maps to which they project
throughout the entire linked system (ND pp. 170-173 [and Fig. 6.7
on p. 173]; RP p. 67 [Fig. 4.1]; BABF p. 26 [Fig. 3-5]).
Some local maps in the thalamocortical system of the brain
(namely, the "primary maps", see UoC p. 96 [Fig. 8.1]; RP p. 55
[Fig. 3.4]; BABF p. 91 [Fig. 9-5]) are reentrantly connected to
and receive input from sensory receptor sheets in multiple
sensory modalities (and sub-modalities, such as motion or
orientation detection in the visual system), including
proprioceptive inputs from muscles and joints (ND pp. 108-110; RP
p. 49; BABF p. 19). "The cerebral cortex is a structure adapted
to receive a dense and rapid series of signals from the world
through many sensory modalities simultaneously -- sight, touch,
taste, smell, hearing, joint sense (feeling the position of your
extremities). It evolved later than the limbic-brain stem system
to permit increasingly sophisticated motor behavior and the
categorization of world events" (BABF p. 118).
Some primary maps (such as those associated with vision or touch)
are topographically connected with their corresponding receptor
sheets: "By 'topographic,' I refer to the situation in which a
sensory receptor sheet receiving signals from the world connects
to its recipient map in such a way that neighboring locations in
the sensory sheet are also neighboring locations in the recipient
map" (BABF, p. 87). Hence, a "primary cerebral map may be
considered to be a translation that makes it possible to sample
and preserve selected coherent portions of the gross topographic
order of the external scene in time and space" (ND p. 109).
While there is some topographic invariance maintaining "certain
basic features relating to the spatiotemporal continuity of an
object", "no point-to-point map has so far been shown to exist in
the nervous system; instead, all local mappings studied have been
point-to-area and area-to-area, consistent with the presence of
degeneracy" (ND p. 108). Primary local maps are in turn
reentrantly linked to multiple secondary maps in the sensory
cortices, and via these to the frontal, temporal, and parietal
cortices, and the motor cortex (UoC p. 96 [Fig. 8.1]; RP p. 55
[Fig. 3.4]; BABF pp. 91 [Fig. 9-5], 109).
This archive was generated by hypermail 2b29 : Thu Jul 27 2000 - 14:13:51 MDT