Worlds Within Worlds
- The Holarchy of Life
(Chapter 10)
by Andrew P. Smith, Oct 24, 2005
(Posted here: Sunday, May 27, 2007)
10. SELECTION WITH DIRECTION
"In sufficiently complex systems, selection
cannot avoid the order exhibited by most members of the
ensemble. Therefore, such order is present not because of
evolution but despite it."
-Darcy Thompson
1
"I believe...that the [theory of evolution]
will hereafter be shown to be a part, or consequence, of some
[more] general law."
-Charles Darwin
2
As I pointed out earlier, Darwinism has long
had an army of critics, a cottage industry that thrives on
pointing out the weaknesses or perceived weaknesses of this
theory. Most of their criticisms boil down to one major
complaint: Darwinism can't account for major evolutionary
changes. It's one thing to understand how new species of
organisms evolve, another to explain the emergence of major new
adapations, such as the eye or the cerebral cortex. It's even
more difficult to understand the major transitions that must
have occurred when the first cells emerged, or the first
organisms emerged.
In the previous chapter I argued that a
broader version of Darwinism can be understood to operate
throughout the holarchy, on every level and perhaps every stage
of existence. A significant weakness of Darwinism in any form,
however, is that it requires a long chain of events, each of
which is of very low probability. The probability that a
particular gene will mutate during a reproductive cycle is
perhaps less than one in one hundred thousand (Lewin 1997).
Other forms of random variation that I considered are also
fairly rare.
Thus it is that almost all the alternative
theories to Darwinism that have been proposed make it their
central goal to demonstrate that evolution could occur quite
rapidly. If life had to proceed by trial and error, the argument
goes, it never would have gotten anywhere. There must be a means
by which holons can jump, rather than crawl (in fits and
starts), up the scale of existence.
Theories of this kind are usually referred to
as theories of complexity or self-organization. Neither of these
terms is very precise. As I pointed out in Chapter 3, we don't
really have a good definition of complexity. Self-organization,
on the other hand, is a very badly overworked term, used so
often in discussions of evolution that it threatens to become
meaningless. As I will use these terms here, they simply mean
that certain holons have the ability to assemble into
higher-order holons spontaneously. Rather than associating with
one another randomly, with certain combinations exhibiting new
properties that enhance the survival of themselves or their
individual components, holons may home in, so to speak, on the
right combination immediately. It should be obvious that if
large, rapid changes of this kind are indeed possible, many
transformative and perhaps transcendent phases in evolution
become much more understandable.
Many kinds of self-organizing processes have
been proposed, and most of them are supported by considerable
evidence. Evidence, that is, which demonstrates that the
processes postulated by the theory do exist, and conceivably
could account for some kinds of evolutionary change. It does
bear repeating, though, that it's virtually impossible to prove
that any particular process did account for any evolutionary
development. This is every bit as true for the alternatives to
Darwinism as it is for Darwinism itself. Indeed, in some
respects it's even more of a problem for the alternative
theories, because almost by definition, they tend to be most
concerned with evolutionary transitions for which there is no
fossil evidence.
In this chapter, I will examine a few of
these theories of self-organization. I will not attempt to
discuss all of them, nor will I discuss any of them in great
detail, as other sources of such discussion are available
(Gleick 1988; Casti 1992; Kauffman 1993; Prigogine and Stengers
1994; Capra 1996). Rather, I will focus on a few of these
theories that I feel offer a representative view of the entire
field, and highlight their key strengths as well as limitations.
I will be asking to what extent such processes may be able to
"fill in" our picture of evolution where Darwinism seems
inadequate, as well which evolutionary processes they seem
unable to account for. Following this discussion, I will then
consider the prospects for a broader understanding in which
self-organizing processes are combined with evolution by random
variation and natural selection.
Dissipative Structures
The theory of dissipative structures was
formulated several decades ago by Ilya Prigigone, who later
received a Nobel Prize for this work. Prigigone was a pioneer in
the study of nonlinear dynamics, which applies to chemical
reactions far from equilibrium. Every chemical reaction has an
equilibrium, in which there is no longer any net change in the
reacting species; and any reaction, left to itself, will reach
this state. For example, if compound A is converted to compound
B, this conversion will proceed until the concentrations of each
compound reach a certain proportion to each other. At this
point, no further net conversion of A will occur unless the
reaction is perturbed--by adding more A, for example, or by
removing some B. At the equilibrium state, the reaction is still
proceeding, but the rate of conversion of A to B is equal to the
rate of conversion of B to A.
Before this century, little was known about
the properties of chemical reactions not at equilibrium, because
this was viewed as a somewhat unnatural state. In order to
maintain a reaction far from equilibrium, as I just noted, it
has to be perturbed in some manner. However, it turns out that
under these conditions, some chemical reactions have very
unusual properties. The concentrations of the reacting species
can undergo random, minor fluctuations. Sometimes, one of these
fluctuations will become ampified, resulting in a major change
in the relationships of the different chemical reactants to each
other (Prigogine and Stengers 1984).
Initial evidence for the theory came largely
from certain chemical reactions in which the reacting substances
organize into highly ordered phases, in apparent contradiction
of the law of diffusion (Zhabotinski 1964). These phases are
often observed by using chemicals that have different colors. As
anyone who has ever mixed two or more colors of paint together
knows, they tend to form a single color, in which the several
original colors are irrevocably blended. In dissipative
reactions, however, the different colored chemicals separate
into highly ordered phases, such as rings and spirals. The
molecular components of the reaction have organized themselves
into something very new and different.
These phases are considered by Prigogine to
be "higher" or more complex forms of existence than the original
mixture, because they require energy to form and maintain, some
of which energy they "dissipate" into the environment. That is,
the reaction imports or absorbs energy from the medium which
goes into the creation of the ordered shapes seen in the
solution. This process is somewhat reminiscent of living things.
Cells and organisms maintain their complex form by assimilating
forms of energy, such as nutrients or sunlight, and releasing
some of it back to the environment in the form of heat, gasses
or undigested food.
That dissipative structures exist is not in
question. The issue is whether, or to what extent, this kind of
process could account for major evolutionary changes. Given that
most of the evidence for dissipative structures comes from
studies of chemical reactions, they are most naturally
candidates for explaining some of the events occurring in the
emergence of cells, which we have seen required assemby of
complex molecular structures. Indeed, some recent studies have
reported that certain components of cells, including calcium
ions (Somogyi and Stucki 1991), ion channels (Svatchencko and
Korogod 1997), and microtubules (Chou et al. 1994), can form
dissipative structures.
However, it's one thing to demonstrate that
certain components of fully-evolved cells can form dissipative
structures. It's quite another to establish that the same or
similar processes could have given rise to the first cells. What
these recent studies suggest--assuming that they can be
replicated, for in science it's often fairly easy to fit a
currently popular theory to observations--is that the ability to
form dissipative structures may have been a feature of existence
that evolved, rather than itself being a driving force of
evolution. This is a critical distinction, which, as I shall
discuss later, may shape our views about just how
self-organizing processes could be related to other evolutionary
processes, such as Darwinism.
A second weakness of dissipative structures
as an evolutionary theory--a problem, we will see, that it
shares with most other self-organizing theories--is that it they
can't account for the critical step of information coding.
Evolution of the cell, as I emphasized in Chapter 8, requires
more than the assembly of enzymes and other types of
biomolecules. It also requires creation of an informational
holon--the genome--which allows this assembly to be reproduced.
So even if some components of a cell could emerge through
dissipative processes, these components could not propogate
themselves without additional major evolutionary changes.
In summary, while dissipative processes might
have contributed to the evolution of some stages within the
cell, there is still no compelling evidence that they were a
major factor in evolution at this level. And while there has
been speculation that such processes may also operate at higher
levels of existence--in human mental processes (Mandell and Selz
1995; Sel 1997), in human social organizations (Duong and Reilly
1995), and even in consciousness (Schorr and Schroeder
1989)--the evidence for them there is even less convincing
3.
A key problem, as I alluded to earlier, is that when a
self-organizing theory becomes popular--and probably more than a
dozen such theories have been proposed in the past fifteen years
or so--there is a fairly intensive effort by some individuals to
apply it to a very wide range of phenomena. The fact that some
of these phenomena have some characteristics that are
superficially similar to those of dissipative processes doesn't
prove that these processes actually create or explain these
phenomena. Even less does it prove that these phenomena
originally evolved through dissipative processes. What one not
unsympathetic critic of chaos pointed out could probably apply
to other complexity theories as well: "One identifies chaos in
models that are fitted to data, not in the data themselves."4
Cellular Automata
As I discussed in Chapter 3, cellular
automata were invented by the computer scientist John van
Neumann almost half a century ago, as a means of designing a
computer that could reproduce itself. The exercise led him to
the important insight that reproduction requires a means of both
transcription, or copying, a program, as well as translation, or
running the program. This dual nature of reproduction was later
found to exist in cells as well.
However, the cellular automata that are the
subject of so much current interest as a model of
self-organization are a little different from van Neumann's
original version. They are not real, reproducing computers, but
rather simple points, or little squares, on a two-dimensional
grid. This grid is visualized on a computer screen, and a
program is written that provides certain rules governing when
and how these little squares or automata will reproduce
themselves. For example, one rule might say that a square will
reproduce when it's in contact with two neighboring squares.
Another rule might specify reproduction when it is separated by
a certain distance from another square. When the program is
allowed to run, some of the squares reproduce, resulting in new
relationships among the automata. These new relationships, in
turn, result in a new round of reproduction. As the program is
run through a great many steps, the squares multiply, and
populate various positions on the grid (Casti 1992; Wolfram
1999).
Depending on the nature of the rules
governing reproduction of cellular automata, very elaborate
patterns can be formed on the grid. In some programs, the
patterns generated bear a startling resemblance to the forms of
certain organisms. Indeed, an early program of this type
designed by mathematician J.H. Conway and simply called Life,
was capable of generating a great variety of "organisms" in this
manner (Casti 1992).
Such success has naturally led some theorists
to suggest that the body types of many organisms may have
evolved from this kind of self-organizing process. Indeed, a
great many patterns in nature, of non-living as well as
non-living things, can be duplicated with programs that generate
form from some relatively simple mathematical equations (Ball
1999). Yet because they are usually are run on two-dimensional
computer screens (though they can be created in three or any
number of higher dimensions), cellular automata generally
provide no more than what are literally just superficial or
surface features of life. Biochemist Michael Behe, whom we saw
earlier believes that cells must have been the creation of an
intelligent designer, captures this point elegantly:
"Some proponents see great
significance in the fact that they can write short
computer programs which display images on the screen
that resemble biological objects such as a clam shell.
The implication is that it doesn't take much to make a
clam. But a biologist or a biochemist would want to
know, if you opened the computer clam, would you see a
pearl inside? If you enlarged the image sufficiently,
would you see cilia and ribosomes and mitochondria and
intracellular transport systems and all the other
systems that real, live organisms need?"
5
Nor is Behe's view prejudiced by his
creationist leanings. Computer scientist Steven Pinker, a
fervent supporter of Darwinism, makes much the same point:
"The 'complexity' that so impresses
biologists is not just any old order or stability.
Organisms are not just cohesive blobs or pretty spirals
or orderly grids. They are machines, and their
complexity is functional, adaptive design;
complexity in the service of accomplishing some
interesting outcome. The digestive tract is not just
patterned; it is patterned as a factory line for
extracting nutrients from ingested tissues."
6
While this is a valid criticism of much of
what many complexity theories have produced so far, it should be
pointed out that the holarchical worldview does provide a
potential opening for a rebuttal. Both Behe and Pinker are
arguing that the detailed microstructure of tissues is not
addressed by large-scale patterns. But one could just as well
apply cellular automata, or other complexity theories, to the
lower levels and stages of existence; perhaps there are "pretty
spirals" or "orderly grids" in the molecular and cellular
structures as well. In any case, more recent work with cellular
automata is attempting to address this criticism. Several
studies have claimed that cellular automata can model a wide
range of phenomena in living systems, including the catalysis of
enzymes (Kier et al. 1996), assembly of bacterial membranes
(Lahoz-Beltra 1997), formation of tissues (Lee et al. 1995;
Markus et al. 1999), the beating of the heart (Siregan et al.
1996), and the spread of populations of organisms (Karafyllilis
1998).
An additional problem with cellular automata
as an evolutionary theory, however, which these studies don't
address, is that for the patterns to be formed, a certain amount
of information has to be inputted into the system, in the form
of the rules specifying when they will reproduce themselves. The
automata themselves can't create these rules; they merely
follow them. So by themselves, cellular automata would
seem incapable of giving rise to complex living forms. However,
again, this does raise the possibility that they might operate
in some combination with Dawinian processes of the kind I
discussed in the previous chapter. I will return to this point
later.
This brings us to the most significant
limitation of cellular automata as a model of evolution. As with
dissipative structures, cellular automata don't provide an
obvious means by which the forms generated can be coded into the
information of the genome. While many of the large-scale
patterns formed by cellular automata can themselves reproduce in
toto following certain programs, this kind of reproduction
clearly is not the kind used by living things, as originally
divined by van Neumann. Real cells and organisms reproduce
themselves on the basis of information they contain within
themselves. As complexity theorist John Casti acknowledges:
"Any genuinely self-reproducing
configuration must have its reproduction actively
directed by the configuration itself...Thus we require
that responsibility for reproduction reside primarily
with the parent structure."
7
In fact, cellular automata might be better
applied to our understanding of development rather than
evolution--that is, the process by which a new organism is
created from a single fertilized cell. It's well recognized that
much of development depends on epigenetic or contingent
processes. That is, while the genome specifies certain cell
types, the latter arrange themselves into tissues and organs in
part according to their relationships with other cells. This
kind of process, in which at least some of the information or
rules are already in the reproducing system, appears to offer a
much closer analogy to cellular automata.
Autocatalysis
The theory of autocatalysis, originally
formulated by chemist Manfred Eigen (1971), has been more
recently developed by Stuart Kauffman of the Santa Fe Institute,
a hotbed of complexity theorists. Kauffman believes that
Darwinism, at the very least, is incomplete, and has spent most
of his career trying to show, mathematically, how complex forms
of life could evolve relatively quickly from simpler forms. His
major book, The Origins of Order (1993), is surely one of
the more impressive scientific treatises published in the past
decade (for a briefer and more accessible account of his major
ideas, see his At Home in the Universe, 1995). In
addition to autocatalysis, Kauffman has worked out the
theoretical basis of many other types of self-organizing
processes which are beyond the scope of this chapter to
consider. Regardless of how correct he proves to be in arguing
that such processes are at the heart of evolutionary change, he
raises a number of vital issues, and provides one of the most
penetrating critiques I have seen of not just Darwinism, but
also of the views of its other challengers.
In a sense, Kauffman's approach is a blend of
that used by those developing theories based on cellular
automata and dissipative structures. Like the former, most of
Kauffman's evidence comes from computer simulations, which he
uses to track the consequences of beginning with a few simple
holons and a few simple rules governing the way they interact.
Like dissipative structures, on the other hand, autocatalysis is
a theory largely about molecular processes, and is thus most
directly applicable to an understanding of early events in the
evolution of cells.
Kauffman begins by postulating a simple
chemical reaction, in which one kind of molecule, A, is
converted into a second kind, B. This reaction is catalyzed by a
specific molecule. Later, a second chemical reaction emerges,
catalyzed by a second molecule, in which B is converted into C.
Still later, further chemical reactions evolve. Eventually, a
chain of chemical reactions is formed, in which substance A is
converted, through many steps, to a large number of other
substances.
Such a chain, by itself, is a fairly simple
and unremarkable form of existence. What introduces complexity
into the system is the concept of autocatalysis. On the basis of
certain assumptions about enzyme catalysis, Kauffman proposes
that when a large enough number of reactions have evolved, the
probability is quite high that the products of some reactions
will serve as catalysts for other reactions. For example,
product M formed from L may catalyze the formation of C from B.
Product R may catalyze the formation of I. And so on. Thus
feedback loops form, in which the reactions are producing
substances that enhance their own ability to produce more of
these substances. As a result, the entire system reaches a point
when it becomes a self-sustaining network of chemical reactions.
Given a supply of energy, which any living thing needs, it can
maintain the network indefinitely.
Kauffman's autocatalytic scheme, while not
supported by any direct laboratory evidence, is nevertheless
highly plausible in some respects. Most evolutionists believe
that life on earth originated in a primordial "soup", pools of
water in which small molecules were formed and began interacting
with each other. At least some of Kauffman's starting
assumptions are sufficiently general, and sufficiently based on
what we know about the behavior of chemical reactions, that we
could reasonably assume them to portray primordial conditions
fairly accurately. And while autocatalysis began as a theory of
self-organization of chemical reactions, it may be generalizable
to many other types of phenomena.
Nevertheless, autocatalysis as a model of
cell evolution has a glaring weakness. The network of chemical
reactions it produces, while superficially similar to those
inside a cell, is different in two key respects. First, in a
cell, catalysts--that is, enzymes--are not synthesized in
ordinary metabolic reactions. They are synthesized from DNA and
RNA, in a special process distinct, physically as well as
functionally, from the synthesis of all other kinds of
molecules. While enzymes might conceivably have initially been
synthesized as in Kauffman's scenario, at some point in
evolution there would have had to be a major shift to the modern
process.
Furthermore, even if an autocatalytic network
were provided with an informational holon to enable itself to
reproduce, it still would be a questionable model of a cell.
This brings us to the second way in which autocatalytic networks
differ from real networks in the cell. Metabolic pathways in the
cell don't feed back on each other in the way they do in
Kauffman's scheme. Products of one reaction may be used
by some other reaction, but they are not catalysts of
other reactions. So real metabolic networks in the cell are not
autocatalytic. If autocatalysis were an evolutionary mechanism,
it would have to be one that was later greatly modified.
Kauffman is not unaware of these problems,
and attempts to address them. He suggests, for example, that
autocatalytic networks might originally have contained both
nucleic acids--such as the catalytic RNA molecules or ribozymes
I discussed in Chapter 8--as well as proteins. Thus information
coding might have been part of the network from an early time.
However, it's still quite a jump from such networks to the
organization of modern cells. One of these gaps that Kauffman's
scheme doesn't address is how the process of autocatalysis is
actually to be realized in physical, three-dimensional space. In
order for reactions to interact with one another in the cell as
they do in Kauffman's scheme, they have to be organized very
precisely in space. A reaction that produces something used by
another reaction has to be located in the cell in such a way
that it's product can actually be made available where it's
needed. Kauffman's two dimensional diagrams, which simply use
arrows to indicate the path by which the product of one reaction
is introduced to another, hide an enormous amount of
evolutionary headaches.
Finally, some of Kauffman's starting
assumptions are questionable. Enzymes can catalyze reactions
that go in either direction--that is, if an enzyme can convert
substance A to substance B, then in principle it can also
convert substance B to substance A. Which direction the reaction
goes depends on the relative concentrations of A and B. Much of
Kauffman's argument is based on the assumption that primordial
conditions would enable reactions to go in either direction, and
that the evolving network could take advantage of both
possibilities. This is certainly debateable.
Kauffman also doesn't take into account that
in a large pool of substances, many chemicals would not be
catalysts of the formation of others, but rather
inhibitors. That is, instead of specifically promoting
certain reactions, they would block them. It's well known
that a substrate for an enzyme, if altered slightly in
structure, may inhibit that enzyme (Stryer 1988; Creighton
1993). So the laws of probability that Kauffman bases his
argument on suggest that it is at least as likely for a
substance to prevent the formation of another substance as it is
to enhance it. While Kauffman incorporates inhibitors in some of
his schemes, he doesn't seem to realize that their existence
might very well prevent these schemes from getting off the
ground in the first place.
In summary, autocatalysis, like dissipative
structures and cellular automata, does not provide an entirely
convincing explanation of how real cells could have evolved with
the ability to reproduce themselves. Because Kauffman, like
other complexity theorists, works largely with computers, he has
been accused of practicing "fact-free science."
8
I think this criticism is a little unfair. His work is informed
by a detailed understanding of the relevant scientific
literature, and he has proposed some experimental tests of some
of his models. While he has not illuminated all the steps in the
evolution of the cell, we must remember that this problem has
stumped the best scientific minds of our century. At the very
least, Kauffman's work has focussed science more clearly on the
steps that have to be understood. Moreover, some of the
principles he has elucidated may be fundamental to our
understanding of evolution at all levels of existence. For
example, the process of autocatalysis suggests that cell-like
structures emerge when a certain number of reactions
exist. This kind of number could be a fundamental property of
holarchical organization, determining the size of not only
cells, but of higher level fundamental holons.
Non-random mutations
Before proceeding to a general critique of
self-organizing phenomena, I want to discuss one other
evolutionary process which, though it technically does not
belong to this class, likewise raises the possibility of a much
faster rate of evolutionary change. This is evolution by
non-random mutations, also known as directed or non-adaptive
mutations. The history of this idea actually begins nearly half
a century prior to publication of The Origin of Species,
when Jean Baptiste Lamarck developed a theory based on the
inheritance of acquired characteristics. Lamarck believed that
variation, rather than being random, might be directly related
to the needs and behavior of the organism. For example, if an
animal stretched its neck to reach leaves on a tall tree, its
progeny would have longer necks. They would stretch their necks
still further, passing this gain on to their offspring--and so,
Lamarck proposed, giraffes evolved.
In the late 19th century, however, August
Weismann argued that since the gametes, or germ cells, of an
organism were separate from the other, somatic cells, there was
no way that acquired changes in the organism's body structure or
physiology could be inherited (Maynard Smith 1997).
Strengthening our muscles, for example, may change our muscle
cells, but these cells are not the ones passed on to our
progeny. Weissmann's doctrine, as it came to be called, was
confirmed with the discovery of genes and mutations, which made
it clear that Darwinian variation was indeed confined to germ
cells under ordinary circumstances.
The possibility of inheritance of acquired
characteristics was made even more unlikely following the great
triumphs of biochemistry in the middle of the twentieth century.
At this time, scientists demonstrated that genes are translated
into phenotype through the synthesis of RNA and proteins, an
understanding that has come to be known as the central dogma of
molecular biology. Since the reverse of this process does not
occur--that is, proteins can't direct the synthesis of DNA--it
appears that information in the developing organism flows in one
direction, from the genome to the phenotype. So even if an
organism could somehow change the biochemical makeup of its
reproductive cells, this change could not be transmitted to its
genes.
Recently, however, this view has been
challenged by several researchers studying bacteria. Because of
their rapid rate of reproduction, bacteria are in many ways an
ideal system in which to study evolutionary processes. Many
generations of bacteria can be produced in a matter of hours,
days or weeks, making it possible to follow the process of
mutation and selection over a reasonable length of time.
When bacteria are grown in culture medium
containing certain nutrients, mutations may arise that permit
them to use these nutrients. For example, bacteria require an
enzyme in order to metabolize the sugar lactose. When bacteria
lacking this enzyme (due to a mutation in the gene for this
enzyme) are grown in lactose-containing medium, most of them
die. A few, however, arise bearing a mutation that, in effect,
corrects the earlier mutation, enabling them to synthesize
lactose. These bacteria survive, and propagate themselves.
This result, by itself, is quite consistent
with Darwinism. It's assumed that the new mutation arises
randomly, and is then selected by virtue of its allowing the
bacteria possessing it to survive. Experiments by John Cairns
and later Barry Hall, however, have suggested that such
mutations may not always arise randomly. They may be adaptive or
directed; that is, the presence of the sugar in the medium may
trigger a process that results in the mutation of the precise
gene needed to permit metabolism of the sugar. Several possible
genetic processes have been hypothesized to account for this
remarkable finding, which has also been reported in yeast, a
eukaryotic cell (Cairns et al. 1988; Cairns 1993; Hall 1997,
1998).
These studies are very controversial. Many
scientists believe they can be explained in traditional
Darwinian terms, without invoking a directed mutation process.
Nevertheless, the possibility that some mutations are not
entirely random should be taken seriously. It's well-appreciated
now that the genome is regulated by a complex web of other
molecules, which can selectively turn on or off the synthesis of
specific genes. Most of these regulatory molecules are proteins,
so clearly information can flow back to the genome in an
important sense. These regulatory molecules are not known to
cause genuine mutations in genes, but they can affect the
structure of genes in certain ways that are heritable.
Another way in which non-random, though
probably not adaptive, mutation could occur is suggested by the
fact that the genome is not a perfect linear sequence of genes,
but a complex conformation of tightly associated DNA and protein
(Lewin 1997). Thus some genes might be preferentially exposed to
ionizing radiation, chemical carcinogens, or other mutagenic
agents. Studies of certain diseases that result from mutations,
such as colon cancer, have shown that particular regions of the
mutated genome, known as "hotspots" are much more likely to be
mutated than other regions (Miyoshi et al. 1992).
Directed mutations have not been reported in
multicellular organisms, and even if they were to be discovered
there, Weissmann's doctrine cautions us against thinking the
basis for Larmarckian inheritance would be completely in place.
Biologist Ted Steele, however, believes that some aspects of the
immune system could be inherited in a directed manner. That is,
an organism that developed antibodies to a particular foreign
agent could pass the gene responsible for its altered immunity
into its germ cells (Steele et al. 1998). But even if Lamarckian
inheritance is never established, non-random mutations could
still play a critical role in evolution, providing for a much
faster process of change than random mutations, which are of
very low probability. A perennial criticism of Darwinism is that
significant evolutionary change requires not just one but a
whole sequence of mutations, the cumulative probability of which
would be vanishingly small. Yet Hall has found double mutants in
his bacteria that appear as frequently as single mutations under
ordinary conditions.
Self-organizing Processes: A Critical
Evaluation
This brief discussion of a few types of
self-organizing processes has barely scratched the surface of a
very complex and controversial area. Nevertheless, on the basis
of this discussion, I will try to summarize what I see as the
strengths and the limitations of evolutionary theories based on
these processes. The best known strength of these theories, of
course, is that they offer a means by which evolution can
proceed much more rapidly than by Darwinian processes. The kinds
of patterns created by dissipative structures or cellular
automata would be vanishingly improbable if they had to be built
up by a random process, one small step at a time. These
processes seem to enable life to skip a huge number of these
steps, leaping from one kind of organization to another very
different kind.
This is an important strength not only
because of the greater speed of evolution possible, but also
because it may help the process avoid blind alleys. In Darwinian
evolution, each step or variation must have some selective value
on its own, in order to survive for the next round of variation.
If it takes one hundred independent mutations for one new
adaptation to emerge from another, and every one of these
mutations does not increase adaptive value, the process may get
stuck along the way. As I discussed in Chapter 7, some critics
of Darwinism believe that the eye would have met this fate. If
it had evolved simply through random variation and natural
selection, it would have soon reached a point where no single
further mutation could increase its adaptive value. Though in
principle it could evolve into a much more adapted structure, it
could not get past this botteneck by a one step at a time
process.
Self-organizing processes, by virtue of being
able to skip steps, seem to offer a way around this problem.
Suppose, for example, that one mutation perturbed the
development of the eye in such a way that it underwent a major
organizational change in its structure. The difference between
this organization and the previous one might bypass many
individual stages which would not be selectively advantageous on
their own.
The reader will note that in this discussion
I have assumed that self-organizing processes can be subjected
to natural selection, just as Darwinian ones are. They aren't
necessarily so, however, which is another potential strength of
evolutionary theories based on them. Self-organizing processes
may sometimes be capable of occurring and becoming established
without competition or selection. This happens when the new form
of organization is the only one possible. Kauffman's
autocatalytic networks, for example, seem to represent an
inevitable outcome of certain associations of chemical
reactions. When the number of these reactions reaches a certain
size, the laws of probability dictate that autocatalysis will
become a significant factor. As a result, the reactions become
intricately interconnected with one another, and begin to
function as a unit. To the extent that this change truly is
inevitable, the network does not compete with the individual
chemical reactions that gave rise to it. The latter simply
disappear as independent processes.
The ability of a process to evade selection
can also greatly enhance its rate of evolution. In Darwinian
evolution, even after a new variant appears through a random,
highly improbable variation, it may take a long time to become
established through selection. This would be equally true even
if the mutation arose through a directed process. If the
adaptive value of the new variant is only slightly greater than
that of what preceded it, many generations of individuals may be
required for the new variant to establish itself. Furthermore,
unless individuals with the new variation become separated as a
population from the original type of organism, there can be no
speciation process, in which the individuals in each population
continue to evolve in different directions.
All these problems can be avoided, in
principle, by a self-organizing process in which a new form of
existence emerges in a single, statistically highly probable
step. Not only is the emergence of the variant ensured, but so
its establishment as a dominant form of existence. Moreover,
because it is so different from what appeared before it, there
may be no speciation problem. A large change of this type, if it
were adaptive, could possibly create a new species in a single
step.
Finally, perhaps the most important strength
of self-organizing processes--though many of the theorists who
have described them would not themselves see them this way--is
their potential to complement Darwinian processes. It's not
really necessary to make an either-or choice between one type of
evolutionary theory or the other. It isn't just that one type of
theory may offer a better explanation of some processes, while
another theory is more suited for other processes. Some
processes may best be explained by invoking some combination of
these two very different kinds of evolutionary theories. I will
develop this idea a little further in the final section of this
chapter.
These, I believe, are the main strengths of
self-organizing theories of evolution. On the other hand, these
theories also have several major weaknesses, some of which I
have touched upon earlier. First, there is a lack of direct
evidence for these theories. By this I don't simply mean that no
one has shown that evolution actually occurred through
self-organizing processes. This criticism would be just as
applicable to Darwinism. I mean that in many cases, these
theories have not been shown to account for the kinds of
processes that evolution required. Thus dissipative structures
and cellular automata--perhaps--can fairly accurately model a
few processes occurring in evolved cells or organisms; it's not
clear that they can offer an explanation of how these processes
originally evolved. Likewise, autocatalysis can, in theory,
account for the emergence of metabolic networks in which enzymes
are synthesized by classic chemical reactions. However, enzymes
are not actually synthesized in this manner in the cell.
Directed mutations have not yet been reported in multicellular
organisms, though perhaps they could have played a role in the
evolution of cells.
A second limitation of theories of
self-organization is that they don't adddress the fundamental
question of how information storage evolved. It's not sufficient
to show that a complex cell-like structure could have emerged
from certain self-organizing processes. A theory must also
provide a means by which the information representing this
structure could have been incorporated into the genome, allowing
the cell to reproduce itself. All pure self-organizing theories
described so far, to my knowledge, fail this test.
A third limitation of self-organizing
theories is that they depend on certain defined conditions or
properties of the evolving entities. Cellular automata require
certain rules in order to generate their patterns. Autocatalysis
takes the existence of catalytic enzyme molecules as a given.
Some other self-organizing theories, such as chaos (Gleick
1988), catastrophe theory (Thom 1989) and criticality (Bak
1996), are even more sensitive to starting conditions. In other
words, these theories attempt to explain how evolution occurred
from a certain point on, while ignoring how it got to this point
in the first place. This criticism doesn't invalidate their
explanations as far as they go, but it does suggest that these
theories could not function as a broad description applicable to
many events.
Still another significant weakness of many
self-organizing theories, rarely addressed, is that they seem
incompatible with a deeper view of holarchy. As I demonstrated
in the first part of this book, there are numerous analogies
between different levels of existence. Life seems to be based on
certain fundamental themes that repeat themselves again and
again throughout the holarchy. Though most complexity theorists
accept some kind of holarchical view, they don't seem to realize
that the existence of such regularities is not easy to reconcile
with the indeterminacy and creativity at the heart of many of
their theories. This is particularly true of dissipative
structures, for example, as well as of several theories not
discussed here, such as chaos, catastrophe and criticality.
These theories--even more than Darwinism, which is constrained
by natural selection--emphasize that major evolutionary changes
are unpredictable, that we can never really know where the
process is going. While there surely seems to be some
unpredictability in the details of evolution, nonetheless the
principles of holarchy suggest a very profound and
understandable order in its major features. Theories which bandy
about words like "creative" and "spontaneous", it seems to me,
have problems accounting for this order.
To be fair, there is a long tradition of
scientists who understand self-organization as a process of
putting together existence in a fairly deterministic and
predictable fashion (Thompson 1992; Lima-de-Faria 1988; Ball
1999), and as such, they offer a helpful counterpoint to the
more popular view of self-organization as unpredictable. Yet
these theories have their own problems. They usually postulate,
for example, that much of the potential of higher forms of life
was present in the very lowest forms of matter, so that
evolution was highly constrained at the outset. But such simple
physical constraints become to a large extent meaningless at
higher levels of existence. Our own properties may be
constrained by the fact that we are composed of atoms, but
within these limits alone, endless possibilities of organization
seem to be possible.
Finally, we might justifiably criticize
self-organizing theories, ironically, on the grounds that the
changes they propose are too large, too sudden. A perennial
criticism of Darwinism is that it takes too long, that there
simply hasn't been enough time for the forms of life we see
today to have evolved through random variation and natural
selection. This criticism, however, can be turned around and
applied to theories of self-organization. If major evolutionary
transitions could have occurred very rapidly, why has evolution
taken as long as it has? If, for example, processes such as
autocatalysis and dissipative structures enabled many of the
components of early cells to emerge quickly, why did it take
billions of years for cells to evolve? If cellular automata are
a valid model of the formation of organisms from cells, why did
this evolutionary process take an additional several hundred
million years?
Nor is the great length of time the only
problem. The abruptness of self-organizing phenomena means that
their effects will frequently be disruptive, rather than
creative. As all molecular biologists are acutely aware, the
more complex life is, the less tolerance it has for large
changes. Any new mutation, Jacques Monod argued, must pass a
number of screens, and most of them never get past the first
one--successful integration with the rest of the cell (Monod
1971). A self-organizing process, Behe notes, has even less
chance in this regard:
"The essence of all life is
regulation: The cell controls how much and what kinds of
chemicals it makes; when it loses control, it dies. A
controlled cell environment does not permit the
serendipitous interactions between chemicals [that a
self-organizing process] needs. Because a viable cell
keeps it chemicals on a short leash, it would tend to
prevent new, complex metabolic pathways from organizing
by chance."
9
This last point, it seems to me, is a very
strong one in favor of Darwinism (though Behe himself doesn't
see it that way), in its broader form, or of a directed mutation
process. While conceding that some evolutionary changes are not
easily explainable through random variation and natural
selection operating over any length of time, and while conceding
the possibility of very new and sudden transitions in evolution,
the preponderance of evidence suggests that it evolution is for
the most part an exceedingly slow process. Darwinism, more than
any other evolutionary theory, makes sense of this.
Relationship of Self-Organizing Processes to
Darwinian Processes
In summary, I believe that while
self-organizing theories can make a significant contribution to
our understanding of evolution, none of them is very likely to
serve as a general theory of the process. Perhaps they would
have more explanatory value, however, if they were applied in
combination with other theories, particularly Darwinism. I
pointed out earlier that some self-organizing processes seem to
emerge inevitably, and therefore might not be subject to
selection in the usual sense. Yet many self-organizing processes
can be subjected to natural selection, at least in principle, so
this opens the door to some kind of synthesis of the two
approaches. "It is not that Darwin is wrong," Stuart Kauffman
concedes, "but that he got hold of only part of the truth...we
must understand how such self-ordered properties permit, enable
and limit the efficacy of natural selection."
10
In a long and detailed discussion of this
prospect, Depew and Weber (1997) actually suggest six different
types of relationships that these two kinds of evolutionary
theories could have. These relationships, each of which has some
adherents, include: 1) self-organization is auxiliary to natural
selection; 2) self-organization constrains natural selection; 3)
natural selection constrains self-organization; 4) natural
selection generates self-organization; 5) natural selection
instantiates (that is, is a special case of) self-organization;
and 6) natural selection and self-organization are part of the
same process. These relationships are not necessarily mutually
exclusive; one type of relationship could operate during certain
phases or stages of evolution, another at some other point.
To get a feel for how Darwinian and
self-organizing processes might interact, let's consider a few
hypothetical examples. A simple, yet illustrative example of
relationship no. 2--self-organization constrains natural
selection--is provided by the evolution of cells. All cells make
use of molecules such as carbon dioxide and water. These very
simple molecules can be said to have evolved by a kind of
self-organizing process; when carbon or hydrogen is mixed with
oxygen under certain circumstances, these molecules form.
Because these molecules existed on earth, life had to evolve in
such a way that it made use of them. Thus carbon dioxide and
water are components of all cells, and furthermore, all the more
complex molecules of cells are synthesized by processes
involving these molecules. This, then, is an example of how
self-organization constrains natural selection; it provides
certain starting conditions within which the process has to
operate. This is a trivial example, one that every evolutionary
biologist accepts, but nonetheless illustrative of the
principle.
Now consider the opposite relationship (no.
3): natural selection constrains self-organization. In the
earlier discussion of autocatalysis, I pointed out that such
networks might emerge spontaneously, and not be subjected to
either competition or selection in relation to individual
chemical reactions. However, one could imagine a situation where
selection did come into play. For example, suppose several
different kinds of autocatalytic networks emerged, all in the
same primitive pool. Under these circumstances they might
compete for limited energy resources. The most efficient network
would be the more likely one to survive. In this manner, natural
selection would constrain self-organization.
I have also alluded to earlier a situation
illustrating relationship no. 4, where natural selection
generates self-organization. In the discussion of both
dissipative structures and cellular automata, we saw how some
recent studies suggested that these processes might account for
certain features in living cells and organisms. We could
therefore postulate that organisms with these processes had some
adaptive advantage, which allowed these processes to be
selected. That is, certain kinds of proteins would result from
random mutations. These proteins, by being able to participate
in self-organizing processes, would provide the organism with a
selective advantage. In this way, the dissipative process would
become established in the organism.
One can further imagine scenarios where more
than one type of relationship is occuring, and where not only
selection, but random variation is also thrown into the mix. In
the previous chapter, I discussed how a Darwinian analog might
contribute to the evolution of early multicellular organisms, or
of higher multicellular stages of organisms. Based on analogies
with cultural evolution on our own level of existence, I
hypothesized that this began with a change in the genetic
surface structure of a cell. It expresses a gene it formerly did
not express, or perhaps a different amount of a gene it did
express. As a result of this changed expression pattern, the
cell's interaction with another cell or cells is altered. This
new interaction, in turn, induces the other cell to change its
pattern of expression so that it matches that of the original
cell. This cell, too, can now transmit its new expression
pattern to other cells. As a result, a large number of cells now
exhibit an altered expression pattern, and a significant change
in the way they associate is possible.
When we combine this idea with cellular
automata, we have a powerful way of generating variety in tissue
organization. For example, suppose the transmission of genetic
surface structure results not from any simple cell-cell
interaction, but only from interaction with cells that bear a
particular spatial relationship to the first cell. That is,
cells transfer this surface structure not to all the other cells
in which they are in contact, but only to certain ones. Suppose,
furthermore, that the change in surface structure--the
expression of a new protein--is also accompanied by reproduction
of the cell expressing the protein. This could occur if the the
newly expressed protein were able to enhance or "turn on" the
translation of a particular gene, as certain proteins are known
to do (Lewin 1997).
In this scenario, random variation is
creating the rules by which cellular automata in turn use to
create their patterns. In this sense, self-organization is being
generated by these processes. On the other hand, one of the
several patterns that could emerge could be selected on the
basis of some superior adaptive trait. At this point, selection
is constraining self-organization.
This brief discussion should make it clear
that a combination of Darwinian and self-organizing processes
has the potential to create much more evolutionary variety than
either alone. It also suggests that it may be possible to
develop an evolutionary theory in which both processes are taken
into account. In this theory, both types of processes would be
understood as the expression of some more fundamental concept
that embraced both. Such a theory could then account for pure
Darwinian processes as well as pure self-organizing processes as
special cases of a more general evolutionary dynamics.
What would such a theory look like? How,
specifically, might we construct a synthesis of Darwinian and
self-organizing processes? An opening is suggested by Manfred
Eigen:
"Theory and experiment show that
Darwinian selection is a category of dynamic behavior
reaching down to the molecular level...Selection is a
consequence of non-linear dynamics. While equilibration
leads to equipartition, thereby maximizing entropy,
selection does just the contrary, i.e., it results from
instabilities that minimize the uncertainties of
remaining states."
11
What Eigen means is that when a random minor
fluctuation in the concentrations of several reactive species
results in a stable new organizational pattern, that pattern is,
in effect, selected. Though Eigen was describing his studies of
reproducing viruses, the above quote could surely apply as well
to other examples of non-linear dynamics, such as dissipative
structures. Somewhat as a variant organism is selected by virtue
of its greater adaptive fitness relative to other members of its
species, a dissipative structure is selected because of its
ability to maintain itself through absorption and dissipation of
energy--in other words, because of its stability. This is
a kind of fitness that the other organizational patterns of the
reacting species don't have, and it is quite consistent with a
Darwinian view. As Richard Dawkins observed, "Darwin's 'survival
of the fittest' is really a special case of a more general law
of survival of the stable."
12
As I emphasized earlier, not all
self-organizing processes can be understood in such Darwinian
terms. Stuart Kauffman's autocatalytic networks are not the
outcome of a random process, but are inevitable when the number
of metabolic reactants becomes large enough. Because of this
inevitability, these networks aren't selected, either, in the
usual Darwinian sense of this word. Selection implies
competition among several states; if the outcome is certain,
there is no real competition.
If we focus not on the network as a whole,
however, but on the individual reactions that make it up,
we can see a process of random variation and natural selection
at work. Though a network may emerge inevitably from
these reactions, the particular reactions that survive in
this form, as well as the manner in which they become
interconnected, do seem to be a product of chance:
"In such a system, very many
partially alternative pathways lead from a sufficiently
large set of [starting] compounds to any target compound
Z. In order that Z be synthesized by a connected
catalyzed pathway, it is necessary not that any one
pre-specified pathway to Z be catalyzed but that at
least one reasonably high-yield pathway among the many
possible pathways be catalyzed."
13
(emphases mine).
In other words, many possible interconnected
combinations of chemical reactions could emerge. Any product of
one reaction might, in principle, catalyze several other
reactions. Which products end up catalyzing which reactions
depends on both how well a particular product can interact with
another particular substance, as well as, most likely, its
ability to find that substance among the large pool of
other substances. This step surely involves a large measure of
chance. When enough interactions like these are formed, on the
other hand, the autocatalytic network comes into being, and
survives because of its stability. This is the selection step.
The survival of the network selects those particular reactions
and their interconnections that make up the network. Not
selected are all those substances that could not catalyze some
other reactions.
When we look at autocatalysis in this way, it
seems to me, it's quite analogous to the Darwinian processes we
examined in Chapter 9. Each chemical reaction plays much the
same role as a gene does in the cell, or as a neuron does in the
brain. Just as genetic variation gives rise to different
biological phenotypes, and neuronal variation to different
behavioral phenotypes (memes), variation among the chemical
reactions results in different configurations of these
reactions. All of these configurations have some primitive
phenotypes, the ability to grow and self-maintain to some
extent; but one of them, by virtue of having a sufficient number
of interconnections among its components, is autocatalytic. This
is the phenotype that is selected.
So in an important sense, I believe all
self-organizing processes can be understood in terms of a
broader Darwinian theory. This is a very significant point, not
simply because such a theory could unify two kinds of processes
thought to be very different, but because some self-organizing
processes, unlike traditional Darwinian ones, seem to have a
pre-determined outcome. Though the particular chemical
reactions, and their particular type of interconnections, may be
a chance process, the emergence of autocatalytic networks,
according to Kauffman, is inevitable. So random variation and
natural selection can be entirely compatible with not simply
what seems to be a higher form of life, but a particular type of
higher form that is determined from the outset. Darwinism,
here, operates within a framework that decides the outcome. The
details of this outcome are contingent, that is,
determined by circumstance; but the essence of the
outcome is not.
Some Darwinists, such as Daniel Dennett
(1995), argue that there is nothing in self-organizing processes
that can't be subsumed under the general concept of Darwinism.
Work by Eigen, Kauffman and others, in his view, is "deepening
Darwin's dangerous idea, not overthrowing it."
14
I find this view perfectly acceptable, if we understand that
Darwinism so defined is quite a bit more general than what
Darwin himself seemed to have in mind (at least in his public
pronouncements); and if we now allow for the notion, just
pointed out, that some evolutionary steps may have been
inevitable. Dennett himself seems to think the latter point
means no more than that natural selection is "complying with
constraints, not fighting them."15.
But surely the real significance of self-organizing phenomena
depends on how often such constraints are encountered during
evolution, and just how constraining they are--that is, to what
extent they effectively eliminate all other possibilities. To
take the extreme example--obviously not the case, but
illustrative of the point--if every evolutionary step were
completely constrained to one viable outcome, no one would think
of the resulting process as Darwinian. Darwinism can incorporate
some self-organization and be enriched; if it incorporates a
great deal of self-organization, it (Darwinism) loses its
identity as that process. As with evolutionary change itself,
the line is not sharp, but that doesn't mean that one concept of
evolution can't emerge as unrecognizably different from another.
Conclusions
Let's now step back a little, and summarize
the main conclusions of the past two chapters. In Chapter 9, we
saw that the twin concepts of random variation and natural
selection, understood in general terms, could be applied to
evolutionary events at every level of existence. Not only the
biological evolution of cells and organisms, but the social
evolution of these holons, and perhaps of still higher-level
holons, involve Darwinian processes. In this chapter we have
added self-organizing processes to the mix, and find that they,
too, can be understood in terms of variation and selection. Is
selection, then, a unifying concept of evolution? Is this what
makes it all work?
Manfred Eigen seems to think so. "Generation
of information," he says, "is connected with the principle of
selection."
16
Indeed, almost by definition. We saw earlier that Gregory
Chaitin defines random strings or programs as having more
information than ordered or patterned strings of the same
length. Not any random string, however, but a particular one.
The process by which one particular string is singled out is
selection. So we might define selection as simply the process by
which the random becomes non-random.
But exactly how are generation of information
and selection connected? Does natural selection actually
create information, all by itself? Eigen again: "Information
generates itself in feedback loops via replication and
selection."
17
Eigen is basing this assertion largely on his studies with
replicating viruses. Viruses can reproduce very quickly in the
presence of a bacterial host, and so like bacteria themselves,
they offer a useful model in which evolution of many generations
can be followed in fairly short period of time. Though we would
have to take great care in extrapolating results from them to
multicellular organisms, if changes in not just quality
but quantity of information could be demonstrated even in
viruses, it would provide major support for the statement of
Eigen's just quoted.
However, neither Eigen, nor any other
scientist I'm aware of, has yet shown that the quantity
of information can be increased in this manner in a reproducing
form of existence. Eigen and others have shown that viruses can
develop new genetically-determined traits through selection,
such as resistance to certain enzymes. Their work also suggests
that a population of viruses, which Eigen calls a
quasi-species, may have some emergent properties (perhaps
quasi-emergent would be a better term) not found in individual
viruses. But they have not shown that a higher form of life, one
with more genetic information, can emerge in this
fashion. Moreover, as Micheal Behe and other creationists love
to point out, artificial selection--in which the
scientist determines the conditions to the which the virus or
other evolving system must adapt--is very different from
natural selection. In the latter, evolving forms of life
adapt to whatever the prevailing conditions happen to be. When a
scientist sets the conditions in the laboratory, they can be
designed so as to maximize the adaptive advantage of the
evolution of certain properties. Life evolving under these
conditions would be expected to change much more rapidly and
much more directly towards a particular endpoint than under
natural conditions.
On the other hand, it does seem that some
self-organizing systems may be able to generate information.
Prigogine's dissipative structures maintain themselves at a
higher level of energy than their starting components, which
seems to imply that the system has acquired some kind of new
information. Kauffman's autocatalytic networks accomplish the
same thing. However, again, these are artificial systems, in
which the conditions are set up by the scientist in the
laboratory. In the case of autocatalytic networks, the system is
not even artificial. It exists so far purely in theory, on a
computer screen. Moreover, as we have seen, most self-organizing
phenomena can't reproduce themselves, so they offer a very
limited model for the evolution of life.
So we are still left with the question of how
information accumulates in life--where it comes from. Perhaps
Darwinian and self-organizing processes are sufficient to
account for this accumulation, or perhaps not. In Chapter 8 I
suggested, on the basis of an apparent information gap between
levels of existence, that information might come from outside
the evolving system. It's time to consider this possibility more
seriously.