ArticlePDF Available

Natural Selection in Relation to Complexity


Abstract and Figures

Structural complexity characterizes our representations of dissipative structures. As a mechanistic concept, when referred to natural systems it generates perplexity in the face of logically sound models. Natural selection is a simple mechanistic concept, whose logic is well exemplified in genetic algorithms. While biological traits and functions do appear to have been subjected to selective culling, current neo-Darwinian theory is unable to account for the evolution of traits or functions when many of these are taken as the separate objects of independent fitness functions. Soft selection, acting in a phenotypically holistic manner, does model selection acting upon structurally complex systems with many traits and functions, but does not account for the evolution of specific traits or functions. It is further suggested that selection cannot be other than a weak force in the early, generative stages of complex life histories, and that this is a good thing, preserving their generativity. I conclude that natural selection theory by itself cannot account for increases in structural complexity.
Content may be subject to copyright.
Natural Selection in Relation
to Complexity
Stanley N. Salthe*
City University of New York
Binghamton University
Complexity, genetic drift, natural selection,
ontogeny, reproductive value, soft selection
Abstract Structural complexity characterizes our representations of
dissipative structures. As a mechanistic concept, when referred to
natural systems it generates perplexity in the face of logically sound
models. Natural selection is a simple mechanistic concept, whose logic
is well exemplified in genetic algorithms. While biological traits and
functions do appear to have been subjected to selective culling,
current neo-Darwinian theory is unable to account for the evolution
of traits or functions when many of these are taken as the separate
objects of independent fitness functions. Soft selection, acting in a
phenotypically holistic manner, does model selection acting upon
structurally complex systems with many traits and functions, but does
not account for the evolution of specific traits or functions. It is
further suggested that selection cannot be other than a weak force
in the early, generative stages of complex life histories, and that this
is a good thing, preserving their generativity. I conclude that
natural selection theory by itself cannot account for increases in
structural complexity.
1 Introduction
This article will examine the question: Are the principles of natural selection as they are currently
understood sufficient to explain the evolution of complexity in living systems? I will present some
facts about selection acting in nature, that, because they point to weaknesses in selectionist thinking,
should be of interest to those involved with artificial life. Natural selection was formulated in relation
to living systems, which are widely understood to be emblematic of complex systems. But I note here
that there is no strong opinion in evolutionary biology that complexity increase has evolved in living
systems by way of natural selection [29, 30]. In the context of artificial life, it could be pointed out as
well that selection occurs in the abiotic world [50], where, however, it lacks a fertility component [52]
in addition to the viability component characterizing any kind of selection. Furthermore, abiotic
systems are no less complex than biological ones when viewed as whole systemsfor example,
when taking into account an entire drainage system rather than just a single reach of a stream, or an
entire storm front instead of a single tornado. Biological systems may seem to be especially complex
mostly because their scale is close to our own observational scales, giving us a sense of many details
with respect to a definite view of system boundaries. After pointing out that a common view of the
natural selection of traits and functions does not fare well faced with the complexity of actual
systems, I reintroduce a model of selection that can relate well to complex systems, even if not to
increases in complexity.
n2008 Massachusetts Institute of Technology Artificial Life 14: 363 374 (2008)
* 42 Laurel Bank Avenue, Deposit, New York, 13754. E-mail:
2 Complexity
Definitions of complexity range from positive ones like Gell-Mann’s [15] or Crutchfield’s [10]
‘‘effective’’ or ‘‘structural’’ complexity, as a concise listing of the regularities shown by a system, to
the negative one that complexity characterizes situations that generate perplexity [48]. Both kinds of
definition are observer dependent. Gell-Mann and Crutchfield’s view has complexity arising in sit-
uations intermediate between complete order and complete disorder. Collier and Hooker [8] gave a
useful gloss of a similar view, which I give a version of in Figure 1. I would take dissipative structures
generally to be complex. In 2006 [48] I suggested that there are two frameworks that are particularly
prone to generating complexity, which I call the scale hierarchy and the specification hierarchy [45, 47].
These both exemplify the idea that complexity involves an inability to describe a system using only a
single criterion or viewpoint [43]. This view could be summed up as ‘‘the more you look, the more
you see’’that is to say that complex systems seem to be inexhaustibly productive of new infor-
mation upon continued examination from new perspectives.
Systems describable using the scale hierarchy format (formally a compositional hierarchy) are
complexmy [45] extensional complexity in that more than one system occupies the same locale,
ranging from macroscopic ones through mesoscopic to microscopic, and these transact in various
indirect ways, frequently provoking chaotic dynamics. In this case changing the scale of observation
delivers completely different information at the same locale [16, 62]. Extensional complexity exists
wherever systems of different spatiotemporal scale influence each other without directly interacting.
Systems describable using the specification hierarchy format (formally a subsumptive hierarchy) are
complexmy [45] intensional complexity in that any system at any scale exists at more than one in-
tegrative level, or [38] level of reality. For example, an organism could be viewed as a physical system,
focusing on circulation, fluid dynamics, and diffusion. Or it could be viewed as a chemical system,
considering metabolism. Again, it could be viewed as a biological system, looking at reproductive
processes. Intensional complexity exists where systems of different kind (physical, biological ) interact
coherently. It is the complexity we face when we realize, for example, that leg muscles active in run-
ning consist of muscle cells powered by chemical energy.
From either hierarchical perspective, complexity can be seen to characterize any material system,
and so neither is restricted to biological systems. It is, nevertheless, possible to consider the question
of whether natural selection can work to increase any form of complexity, or to particularly support it,
in living systems, as for example does Wilson [59, 60], who describes a model of ecosystem-level
Figure 1. Contextualizing structural complexity (modified from Collier and Hooker [8]). Note that, while machines can
be quite complicated and can generate complex output, they are conceptually simple in the sense that they can be
explicitly described fully. Note as well the mapping of complexity to disorder. This reflects the perplex aspect of
Artificial Life Volume 14, Number 3364
Natural Selection in Relation to ComplexityS. N. Salthe
selection promoting the coexistence of mutually beneficial species, which would increase the struc-
tural complexity of an ecosystem.
The definition of complexity is crucial when trying to determine whether selection can increase
it. Lenski and colleagues [26] claim to have used the Avida platform to promote the evolution of
complexity in ‘‘digital organisms.’’ During evolution therein the digital organisms can construct logic
functions, some of which allow them to replicate faster by way of garnering increased amounts of
‘‘energy’’ (CPU cycles) or by increasing replication efficiency. Logic functions may be acquired by
way of random mutations. The authors claim that increased size and numbers of logic functions
amounts to increased complexity. I think this can be questioned inasmuch as they have assigned the
system two (what amount to phenotypic) components of fitness one fertility function (replication
efficiency) and one viability function (rate of energy capture) neither of which can get to be more
complex. The logic functions (nine possible ones) can be viewed as genes coding for the two
functions, as this information is what is replicated. There is no increase in organismic complexity,
but, rather, in my view, an increase in the number of genes. Admittedly, Gell-Mann and Crutchfield’s
definition of complexity might be used to defend the assignment of complexity here.
3 Natural Selection
Fisher’s fundamental theorem of natural selection [12] encapsulates the basic logic of natural se-
lection, which depends upon variation in fitness relative to biological functions, as carried by var-
iation in phenotypic traits. In the simplest models of evolution by selection, this variation is reduced
by selective culling of inferior types, being traded for adaptation as the current best type for any trait
increases its representation in a population. The model implicitly focuses upon single traits. So, in
this simplest model, selection would be a process of simplification, resulting in diminished numbers
of genotypes and phenotypes as a consequence of a population becoming specialized in the currently
most successfully reproducing type that it can generate.
Even this simple process could logically promote (but not create) a more complex habitat for
selected populations, for example, as in the result of character displacement [4]. Here, niche space
(sensu Hutchinson [20] see [46]) gets progressively reduced within sympatric populations as they
make room for each other’s existence by way of selection acting separately in each population, culling
types that waste energy in competing with another population. This has the effect of reducing niche
overlap between the populations, therefore reducing interspecific competition. However, this process
likely will destabilize relations with yet other species’ populations, requiring further selective ad-
justments in connection with these and so on, delivering in effect a ‘‘continual deterioration of the
environment’’ [53] so that selective adjustment can never really settle down. Here we see a lowest
level of selective percolation as the background to occasional episodes of secular evolutionary change.
And here we also see selection entangled in a mesh of extensional complexity that it cannot be held
responsible for having produced as a result of promoting the currently best organismic traits within
the coexisting populations.
This perspective on selection would be sufficient to understand the preservation of adaptation if
organisms were composed of just one or very few traits. But organisms, like their environments, are
complex under any criterion (e.g., [13]) considerably more so, with their manifest intensional
complexity, than other kinds of dissipative structures. They have at least as many traits as they have
functions connecting them to their environment [42]. Any one or more of these functions might
become challenged by environmental change, and so become the focus of selective maintenance or
improvement. Then, viewing organisms as collections of traits, one might envision selection re-
viewing traits first in connection with one function, then another, and so on in unending con-
catenation, and more intensely during times of severe environmental deterioration.
This is the general picture of selection informing the technique of genetic algorithms, where some
function is improved by simultaneously modifying many different traits until a targeted functional
optimality is achieved after some generations of selection or, in a selection experiment, honing
Artificial Life Volume 14, Number 3 365
Natural Selection in Relation to ComplexityS. N. Salthe
some trait by simultaneously promoting various alleles at a number of different genetic loci. But this
is really a form of artificial selection insofar as the function in question is anything other than en-
hanced success in reproduction (or in applications stability, satisfaction, promotion), which is
the sole target of natural selection regardless of what particular challenges might face a population.
While separate functions might be optimized, only differential reproductive success is maximized. In
such a natural context the fitness function of a genetic algorithm might be, for example, ‘‘maximize
corporate profits’’ that is to say, something that cannot be determined in advance (‘‘whatever
works’’), or even, necessarily, understood after the fact.
In nature, selection focused upon a single function might occur during an exceptionally severe
environmental deterioration a blizzard, firestorm, or floodwhere such a function might become
key to survival. However, studies of extreme deteriorations in nature (e.g., [6], and several such stud-
ies since then) show that what generally happens is that individuals fail with a likelihood scaled to the
distance that their trait measurements (arbitrarily chosen by the observer) depart from the mean
value of the population. Populations seem to be held in the grip of a central tendency. This would be
like a genetic algorithm with a fitness function promoting a current average (i.e., established) type
instead of culling variation with respect to a particular function that might contribute to survival.
This centripetal mode of selection would characterize periods of evolutionary stasis in continually, as
well as in episodically (periodic selection [37]), deteriorating environments. This maintenance of the
status quo has been argued by some [3] to be the major role of natural selection in biology. Given a
flow of small-effect mutations and recombinants, this selective tinkering [35] could occasionally be
called upon to initiate secular change as a result of some punctuation in the routine, when culling
might become biased away from an established trait mean.
This mode of natural selection as a holistic reviewing of the entire organism was specified by
Wallace [56] as soft selection (see also Maynard Smith’s [28] ‘‘threshold selection’’), in contrast with the
hard selection of particular traits or functions, as in Fisher’s model or in artificial selection.
In hard
selection global fitness over many independent traits would be viewed as multiplicative over the
separate fitnesses for each trait, and would result in serious depletion of the population size if more
than just a few traits were being selected at one time in a very large population. In soft selection only
overall fitness is in view, with the overall worst kinds being eliminated first, and so on, until the
carrying capacity of the environment for that population has been reached. In fact, this is nothing
more than being explicit about the standard idea in evolutionary biology that natural selection
maximizes population fitness, or its overall fertility, rather than improving any trait or function that
might contribute to that variational tendency. And this, of course, is the source of the charge that
natural selection merely states a tautology the survival of those that best survive.
Soft selection raises the issue of what fitness is (for some views on this problem see [21, 32, 33, 51]).
Looking again at character displacement in this light, suppose we find two species of bird with beak
sizes significantly differentiated, feeding on significantly different food sizes, but otherwise having
large niche overlap. Given the viewpoint taken here, it seems unlikely that these beaks can have been
other than one trait among many that have been episodically modified in a past period of coevo-
lution of these populations. It would be misleading, then, to state that beak size here ‘‘has evolved’’ in
the sense of ‘‘the evolution of beaks’’ being the label of an actual process that took place in nature.
As mediated by natural selection, nothing in particular evolves other than as a byproduct of a more
general process of population survival from generation to generation.
Looking again at Fisher’s fundamental theorem which implicitly views a single trait or
functiona forward-looking perspective would predict that traits having the greatest variability,
which therefore could be associated with the greatest variation in fitness, would be most likely to
experience the greatest selective intensity during the next environmental deterioration (rather than
1 It has been pointed out to me that Ridley [41] has a discussion of soft and hard selection that seems quite different from mine.
However, it seems to me that we are just emphasizing different aspects of it. My view is that soft selection is the mode of selection
focused only on overall fitness, without regard to particular traits or functions, thereby avoiding the need for multiplicative global fitness,
wherein no more than a few traits in very large populations could undergo selective change at significant rates.
Artificial Life Volume 14, Number 3366
Natural Selection in Relation to ComplexityS. N. Salthe
particular traits that seemingly could be most affected by the particular bearing selective pressures).
But, looking back the other way after selection has occurred, we expect that traits that have been
subjected to the greatest selection pressures will be those now having the least variability. This last
was tested using traits of frogs’ legs [49]. An earlier study had identified particular ratios of leg mea-
surements that were found in experiments to be most important for jumping. We compared the
variability of these trait ratios in populations of frogs that regularly jump, and found that, in fact, they
were significantly less variable than arbitrarily chosen ratios of leg measurements. Furthermore, these
ratios were not less variable in species of frogs that do not jump.
But, surely, jumping is only one of many functions we could identify that frogs must carry out
successfully! If all of these functions had been under selection simultaneously, then important traits
for other functions would also show less variability than arbitrarily chosen ones. And all these traits
would seem to have been selected simultaneously as independent traits. But this supposition raises
the issue of the limitation on the number of individuals available to be culled by (hard) selection for
each separate trait in actual populations of finite size. Considering how many important functions an
organism must carry out, the genetic load on the population, with so many independent traits being
simultaneously selected, would be unbearable, even with individuals overlapping in reasons for
failure [18]. It has been suggested that selective neutrality of most genetic loci [22] would obviate this
cost-of-adaptation argument, but it is difficult to imagine that most traits or functions could be
selectively neutral even if most genetic loci areor that functions, if not traits, are not independent
as selective targets.
Here again we can invoke soft selection, where individuals fail in each generation in rank order
according to how many of their traits have a currently less favorable condition, culling first the worst
and working up into increasingly fitter types until the environmental carrying capacity of the pop-
ulation has been reached. In this concept, if a given generation of selective culling were to be re-
peated, a partially different group of individuals would be found to fail, showing that traits, as such,
cannot logically be taken to be under individual selective scrutiny, and many different (therefore
nearly selectively neutral) genotypes can produce an adequate phenotype. Good jumping proportions
would be among many traits of importance to frogs, any and all of which would statistically tend,
generation after generation, to sometimes be among those having prominence in the activities
leading up to reproduction. Then no particular trait will have been selected as such an uncanny
situation, given that we can actually identify traits and clearly associate them with important func-
tions, and ones that, in addition, do appear to have been subjected to selective culling. So the phe-
notype, as we perceive it as being composed of many traits, is effectively or structurally complex,
while concepts of natural selection are simple abstractions.
Here the observer becomes liminal as a bearer of complexity. It is biologists, as mechanists, who
identify different functions (e.g., [13]) jumping, eating, seeing, and so on as unarguably con-
ceptually separate functions, and who map these to genetic loci in the DNA information storage.
The latter supposed representation relation genes for this or that has long been known to be an
oversimplification. For example, messenger RNA is often a pastiche of pieces from different genetic
loci, and/or is manipulated in different ways by the products of other genes, whose final gene prod-
uct must interact with the products of several loci in order to inform some function. As well, traits
are widely polygenic, and any one genetic locus can be widely pleiotropic, carrying information used
in more than a single trait. Clearly the relationship between genetic information and functional traits
is complex [27, 61]. Mitton [34] found plentiful evidence of selection distinguishing different allelic
forms of proteins. But, as I pointed out [44], the properties allowing detection of such molecular dif-
ferences have no necessary connection to a protein’s functionality. This does not stop biologists from
pinpointing ever finer discrimination of traits, as can be seen in Abzhanov et al. [1], who suggest ‘‘a
mechanistic explanation for the independence of beak evolution along different axes.’’ They claim to
show that a biochemical pathway that mediates calcium signaling is expressed at higher levels in
Darwin’s finch species with long, pointed beaks than in those with more robust beaks.
Concerning traits, there is a further complexifying issue concerning what they actually are [44, 54].
Taxonomists can be satisfied with whatever can be discriminated, but it seems that evolutionary
Artificial Life Volume 14, Number 3 367
Natural Selection in Relation to ComplexityS. N. Salthe
biologists ought to be concerned about the relation of traits to functions and, indeed, that they do
in fact relate in some way to functions, and that they do therefore, in whatever uncanny way, get
subjected to selective pressures [14]. There is also the more general question ‘‘What is a part?’’ [31].
The same problem appears with functions as well. Physiological functions show some interesting
relations in this context. In experiments with several kinds of animals reviewed in [44], various
regimes were imposed to force animals to exert progressively more work. For example, fishes can be
made to swim faster in a ‘‘wind’’ tunnel. It was found that as the animals worked harder their heart
rates increased. Comparing statistics across different work loads shows that the standard deviation
of heart rate among individuals declines as the animals work harder. So in peak performance the
animals get to be increasingly more alike in important functions. At the same time, functions un-
related to that work (e.g., in the swimming case, blood flow rate in veins leaving the intestines)
become more variable during faster swimming. The implication is that selection has been reviewing
heart rate mostly (or only?) under conditions of stress. So, while a fish is slowly poking about picking
up morsels here and there in a leisurely manner, selection would not see its heart rate (unless it is
grossly deficient, in which case it would have been eliminated long before anyway).
So we see that traits need to be viewed as aspects of performance, and these need to be viewed
or even, perhaps, definedduring periods of intensified function (on the issue of functions, see [5,
55]). Corning’s [9] arguments about widespread synergetic effects during functioning could be taken
into consideration here as well. These considerations feed as well into Crutchfield’s [10] call for a
theory of biological structu re. Quoting: ‘‘Functionality. . . comes equally from the context of a given
form something much harder to detect than form itself.’’ Regarding his concept of ‘‘epochal
evolution’’ (as the cause of punctuated equilibrium), biological functioning during periods of reduced
stress on a given function could be the nexus from which functional innovations might self-organize,
inasmuch as selection’s stern gaze is then focused elsewhere (see also [40]).
Taking stock, let us note that, while selection theory (e.g., Fisher, Haldane, Wallace, and almost all
since them) has been concerned with selective improvement of functions leading to the evolution of
various traits, the indirect evidence cited above relating to selection in nature concerns entirely the
process of selection (periodic and soft ) during periods of stasis. This, of course, is necessarily so
given the time frame of observations in nature now compared with the number of observer gen-
erations it would take to substitute alleles [34].
Here we can note three reviews of many studies of natural selection in natural populations
[11, 23, 34]. They all found plentiful evidence for selection maintaining the adaptedness of popula-
tions. The most recent study claims that there is also evidence for ‘‘disruptive selection’’ in the
datathat is, a greater variability of traits around the mean after selection. The implication within
this study is that this might actually signal an early stage of directional selection. But this is not the
only interpretation possible. Other kinds of selection could be responsible for increasing the var-
iability of a population forms of balancing selection, such as frequency- and density-dependent
selection supplied with new mutants or recombinants, which would not be associated with evo-
lutionary change.
Summarizing the argument up to this point, I think we need to acknowledge that with regard to
any sort of complexity, natural selection is a very simple idea. Organisms have many functions, each
of which involves many traits, and these traits are involved in more than one function. In this realistic
setting perhaps some version of soft selection would be the most promising model of the main-
tenance of adaptation.
4 Selection and Temporally Generated Complexity
Finally, I raise an issue concerning developmental constraints on natural selection (for a general
review of this area see [2, 58]), which to my knowledge has not been raised previously. Consider
quantitative traits of organisms from Darwin’s point of view of their getting ‘‘improved’’ by natural
selection (so elimination of the manifestly unfit is not under consideration). Considering the
Artificial Life Volume 14, Number 3368
Natural Selection in Relation to ComplexityS. N. Salthe
possibility of such improvement in embryos, we note that they pass through several different de-
velopmental stages, each with its own, quite different morphology and way of life based on different
traits and functions. It has often been suggested that there are embryonic adaptations (e.g., [61]), but
there remains the question of to what extent early ontogeny can actually be scanned by natural se-
lection (as proposed, e.g., by Buss [7]).
Consider first the process of genetic drift. In embryos, drift would be a consequence, not of small
population size, but of prolonged delay between selective sorting and reproduction, during which
time the embryos engage in an irreversible sequence of very different ways of being, none of which is
repeated. So the viability component of fitness is quite peculiar in embryos. For example, consider
three stages of development in a frog: blastula, tailbud, and larva. Selection distinguishing among
blastulae would necessarily have to be based upon some of their characteristics, and will sort them
accordingly. Those that pass on to being tailbuds will now be subjected to a whole other group of
selection pressures, while pressures that had impinged upon them as blastulae have effectively be-
come subliminal for good. Individuals will again be sortedthis time on criteria related to being
tailbudscompletely irrespective of the criteria that affected them as blastulae. Moving on to the
larval stage, we have the same phenomenon, this time with selection acting irrespective of criteria
relating to both earlier developmental stages. Traits reflecting selection by the earlier selective pres-
sures must as a consequence drift, with those of the earliest stages being deselected to a greater extent
than those scanned in later stages.
In more detail, begin with fertilized eggs. Looking at some quantitative trait (it could be a
molecular concentration, some relation between landmarks, or granule size differences), we find that
it is initially broadly distributed in the population of ova that is, there is lots of variability in the
unselected offspring, and most possible measurements are well represented except at the very
extremes. Now we get selection up to the blastula stage, shaping, say, a unimodal curve around a
selected mean for one of the above traits. Moving on to selection in the tailbud stage, different traits
are now selected in what is effectively a different kind of organism. As a result, our selected curve
from the blastula stage (having now become virtual), measuring a trait no longer under selection, or
even in manifest existence, changes. We can suppose, just arbitrarily, that its distribution as a result
gets to be platykurtic around the original mean. This would be the effect of a kind of genetic drift.
Moving on to selection among larvae, on yet further different traits again, the result on our curve
selected in the blastula stage has changed again — this time, say, the virtual distribution got skewed to
the left, even perhaps shifting its mean somewhat.
Moving on to selection in small froglets, on yet further different traits, the curve selected in the
blastula stage gets virtually modified againthis time we might suppose the mode gets hollowed
out, producing a bimodal curve. And so on, through subsequent stages up to that where the or-
ganism is definitive for its kind here, a young adult frog. This sequential disruption of the effects
of selection on the earliest stages produces virtual distributions of early-stage traits that would not
have been as favorable for further developmental changes as the original curves would be if the
distorted curves were to actually be projected back to the early-stage population. Note in addition
that, as development continues, the cohort size decreases, gradually bringing in more traditional
genetic drift effects (of small population size) on the original trait shaped in the blastula stage.
A key point in this example is that organisms in the different developmental stages are very
different kinds of beings. This is a different situation from implicit comparisons between different
ages of the definitive imago, as in the usual studies of natural selection, which do not raise this
problem because selection pressures are not irreversible. A somewhat similar situation in a pop-
ulation of adults would be if a drastic environmental disaster struck a population so that one or two
traits became especially important for survival. This will reorganize the gene pool in a way spe-
cializing it for this kind of perturbation. Now, if one or two generations later another, completely
different kind of disaster struck the population, the effects of selection from the previous disaster
would become significantly altered.
So the punch line here is that the results of selection in an early stage, functional at the time, be-
come, as a result of further unrelated selection, effectively attenuated to different degrees by very
Artificial Life Volume 14, Number 3 369
Natural Selection in Relation to ComplexityS. N. Salthe
different subsequent selection pressures. That is, the results of selection in early stages cannot be
inherited, as such, by embryos in subsequent generations, because they can’t be preserved in a
cohort’s population during development. So here we would have a type of developmental constraint,
being a limitation on the degree to which selection can hone embryonic traits, especially in the earliest
stages of development, where the effects of subsequent distortion via drift would be more pro-
nounced. Perhaps development can continue to operate in subsequent generations because those
early stages don’t need to be other than relatively simple in form and I think ‘‘vaguely embodied’’
might be a better description of their organization.
A possible quantitative objection to this idea is that, if the populations here are very large (as in
the eggs of some fishes and frogs), since we are looking at effectively random events from one stage
to another, then because of that randomness itself, no substantial change in the distribution of an
early selected trait should accumulate just back-and-forth fluctuations over time, averaging out to
roughly the original distribution. My counters to this are: (a) this nullifying effect would be more
powerful the more different stages are traversed cumulatively, but there just aren’t that many dif-
ferent stages during embryonic development. As well, (b) as development continues, cohort size
diminishes rapidly in many species, such as many kinds of fishes and frogs, instituting traditional ge-
netic drift in some cases at least.
Furthermore, (c) I would bring in for support my original insight here Fisher’s reproductive
value concept [12]: the age-specific expectation of future offspring for any individual, or the average
likelihood, at each age, that it will contribute genes to the future gene pool (Figure 2). This decreases
as an individual ages after the age of first reproduction, because it is then less likely to contribute to
the selective responses of the population, mostly because its probability of dying increases with age.
On this argument, properties of individuals at the age of first reproduction are the most intensively
scanned by selection, because many individuals do not survive to later breeding seasons, and so the
population sizes of those reproducing for the first time are the largest of all, making up the greater
proportion of the effective population size in any breeding season. This situation has given rise to a
Darwinian theory of aging, reviewed recently by Williams et al. [57].
My conjecture here is that the effect of distance from the reproductively most important stages in
early maturity should work in the other direction, going forward into embryonic stages, as well. The
reproductive value of a blastula is minuscule, and improves as development proceeds and the in-
dividual’s likelihood of breeding improves, but it is low during all embryonic stages (Figure 2), so that
any selection they have been subjected to would have little effect on the next generation compared to
the selective pressures bearing upon reproductive imagos. Only in really huge populations could
there be enough individuals to be sacrificed for improvement of earlier life history stages in addition
to improvement of the definitive, reproductive stage— and living systems with huge populations (say,
bacteria) do not have life history stages. Fisher himself worried, in one sentence, about the im-
plications of this for prereproductive evolution, but backed off and dismissed the idea out of hand.
Hamilton [19] considered this idea briefly again, but said nothing definite, as if shying away from
the problem.
So, it seems that the evolution of development might need something more than natural selection
to deal with it. Then, how is development kept orderly? Here we must note that, since development
is epigenetic, with one trait being built upon another materially, the earlier forms are used as tem-
plates upon which to build later ones and so would leave a footprint in the future. Again, this could
Figure 2. An idealized curve showing changes in reproductive value over a lifetime.
Artificial Life Volume 14, Number 3370
Natural Selection in Relation to ComplexityS. N. Salthe
obviously be important for properties of definitive stages ( juveniles) that are close to the age of
reproduction, but the effects of selection in the earlier, embryonic stages on traits and functions
present and relevant only in those stages must still be defaced to some extent. Selection in early
stages could at best be ‘‘ballpark’’ only, with no refinements being possible. And, of course, larvae (in
complex life histories) come into this story as well. For example, how could selection scan the
elaborate properties of frog tadpole lips? Nothing in the adult (as far as we know) is templated upon
these, which completely disappear at metamorphosis, and yet these complicated species-specific
characteristics continue to be preserved in the populations of each species. It has been noted several
times that larval evolution often does not follow imago evolution, the two seeming to run on parallel
tracks. Wray [63] noted that this suggested that natural selection must be operating on larval forms
directly. But, since they don’t reproduce, how?
Finally, a thought worth considering in this context is that the absence of a possibility for selective
refinement of traits in early embryos may be an important factor contributing to those stages re-
taining the generativity necessary for mediating larger evolutionary innovations. If selection could
refine them, it would make the earliest developmental stages less vaguely embodied, and so less ame-
nable to modification during ontogeny. I should mention here the idea that for evolutionary mod-
ification to occur, forms of stabilizing selection need to be lifted [57]. This aligns with the idea that
selection is mostly a conservative force [3], and so it seems likely that the feebleness of its effects in
early embryos must work to maintain the generativity of those early embryonic stages.
5 Discussion
Viewed as being composed of parts, an organism would in every particular be a structurally complex
system, while our descriptive categories are explicit, and therefore simple, classifications. If we did
not need, as mechanist, to seek the devil in the details, we would not be faced with the perplexity of
complexity at all! In a situation where traits and even functions are defined only with difficulty, soft
selection seems to be the only reasonable selection model. Soft selection is simplicity itself, but it
carries no implications for particular differentiated functions or traits, both of which nevertheless
do demonstrate in nature the effects of selection having culled them. Classical hard, or artificial,
selection supposedly scans details individually, but is defeated by its limitations in the face of the
actual properties of natural populations and organisms. I take the view that specific details, and
concepts like artificial selection, are the products of philosophical mechanicism. As biologists and
(significantly) physicians we scan an organism in exactly the same way as mechanics scan machines
(see [17]). In order to heal we must find the problem and cure it. This single bias, or fact, explains the
general approach used in scientific analyses, creating, in light of sciences social role, its traditional
philosophical mechanicism.
My view here should not pass without challenge by the fact that there are some remarkably
machinelike organs in living systems, such as the flagellar motor of some microorganisms [36]. Ex-
amples like this seem to me, however, to sharpen even further the above conundrums about natural
6 Conclusion
I conclude that complexity, in addition to anything else it might be held to be, is (whether structural
or effective complexity) in the first place an illusion generated by the philosophical mechanicism used
as a guide to scientific practice, and that natural selection can neither enhance it nor defeat its per-
plexity, since this characterizes all material systems given the way we view them. As a simple (i.e.,
plainly logical) theory, natural selection will continue to be betrayed by the vagueness of the world, so
that, as it is represented in theory, it can hardly be held to be the agent of the evolution of any specific
thing at all in nature. Curiously, if the perplexity of complexity is a kind of implication of mechanicism’s
structural complexity concept, then natural selection can be supposed, in the form of soft selection, to
Artificial Life Volume 14, Number 3 371
Natural Selection in Relation to ComplexityS. N. Salthe
mediate— but not to create— it. As well, if selection could act strongly upon early-stage embryos, there
is reason to believe that it would convert their vague embodiment to a more structurally complex
condition, which, given their germinal role, would likely be functionally disabling.
Well, then, are the principles of natural selection as they are currently understood sufficient to
explain the evolution of complexity in living systems? As soft selection, it can be construed to
somehow work to maintain adaptedness within complex systems, but there is no evidence that it can
create structural complexity in selected organisms. In selection theory, as Fisher’s [12] Malthusian
parameter, or as the Wright-Dobzhansky selection coefficient, selection is represented only as a neg-
ative forceless than the fastest reproductive rate in the first case, and a deficit from the current
best fitness in the second. It does not have a generative role in neo-Dar winian theory, so selection
must be insufficient to explain complexity increase.
In regard to selection in the context of artificial life, an immediate message from this viewpoint
would be that, in stochastic search and adaptive computation approaches like genetic algorithms, the
focus should be shifted upscale from optimizing individual functions to promoting higher level suc-
cess depending upon several functions, making the system more extensionally complex as in real life.
A further technical suggestion would be that, given an interest in greater verisimilitude to natural
selection, some attention should be shifted from fully explicit techniques toward fuzzier ones, moving
even through second-order fuzziness in the direction of vagueness (e.g., [24, 25]). More generally,
while it is true that in simulations simple rules can generate complicated behavior and forms, we
might note that the simulations themselves are the complex frameworks that allow this to happen
[26, 39], just as the world itself is the complex framework within which natural selection will have to
be working.
7 Summary
Many views of complexity have been suggested, applicable to the abiotic world as well as the biotic.
Because interactions between phenotypes and their environments generate complexity, models of
natural selection involving specific traits or functions do not well represent natural populations under
selection. As mechanists, we do not have adequate ideas about how to represent phenotypic traits in
models of biological systems. Given this, only models of selection like soft selection have any
verisimilitude to a natural process in a complex setting. With respect to temporal complexity, there
may be reason to suspect that the effects of selection in early stages of ontogeny may not be pre-
served, thus retaining the generative capacity of early embryos. Finally, natural selection cannot be
held responsible for the complexity of living systems.
I have taken some leads from communications with Jerry Chandler, Cliff Joslyn, and Max Shpak. I
thank John Collier for comments on the draft. Norman Johnson has helped with preparation.
1. Abzhanov, A., Kuo, W. P., Hartmann, C., Grant, B. R., & Grant, P. R. (2006). The calmodulin pathway and
evolution of elongated beak morphology in Darwin’s finches. Nature,442, 563– 566.
2. Brakefield, P. M. (2006). Evo-devo and constraints on selection. Trends in Ecology and Evolution,21, 362368.
3. Brooks, D. R., & McLennan, D. A. (1991). Phylogeny, ecology, and behavior. Chicago: University of
Chicago Press.
4. Brown, W. L., & Wilson, E. O. (1956). Character displacement. Systematic Zoology,5, 49 64.
5. Buller, D. J. (1999). Function, selection, and design. Albany: State University of New York Press.
6. Bumpus, H. C. (1899). The elimination of the unfit as illustrated by the introduced sparrow. In Biological
Lectures, Woods Hole 1897 (pp. 209– 215).
7. Buss, L. W. (1987). The evolution of individuality. Princeton, NJ: Princeton University Press.
Artificial Life Volume 14, Number 3372
Natural Selection in Relation to ComplexityS. N. Salthe
8. Collier, J., & Hooker, C. A. (1999). Complexly organized dynamical systems. Open Systems and Information
Dynamics,6, 241–302.
9. Corning, P. A. (2005). Holistic Darwinism: Synergy, cybernetics, and the bioeconomics of evolution. Chicago: University
of Chicago Press.
10. Crutchfield, J. P. (2003). When evolution is revolution— Origins of innovation. In J. P. Crutchfield &
P. Schuster (Eds.), Evolutionary dynamics: Exploring the inter play of selection, accident, neutrality and function
(pp. 101– 133). New York: Oxford University Press.
11. Endler, J. A. (1986). Natural selection in the wild. Princeton, NJ: Princeton University Press.
12. Fisher, R. A. (1929/1958). The genetical theory of natural selection. New York: Dover Publications
(variorum edition, Oxford University Press, 2000).
13. Frazzetta, T. H. (1975). Complex adaptations in evolving populations. Sunderland, MA: Sinauer.
14. Fristrup, K. M. (2001). A history of character concepts in evolutionary biology. In G. P. Wagner (Ed.),
The character concept in evolutionary biology (pp. 15– 37). San Diego, CA: Academic Press.
15. Gell-Mann, M. (1994). The quark and the jaguar : Adventures in the simple and the complex. London: Little, Brown.
16. Giampietro, M., Allen, T. F. H., & Mayumi, K. (2006). The epistemological predicament associated with
purposive quantitative analysis. Ecological Complexity,3, 307 327.
17. Goodsell, D. S. (1998). The machinery of life. New York: Springer-Verlag.
18. Haldane, J. B. S. (1957). The cost of natural selection. Journal of Genetics,55, 511–524.
19. Hamilton, W. D. (1966). The moulding of senescence by natural selection. Journal of Theoretical Biology,
12, 12 45.
20. Hutchinson, G. E. (1978). An introduction to population ecology. New Haven, CT: Yale University Press.
21. Ketterson, E. D., & Nolan, V., Jr. (1999). Adaptation, exaptation, and constraint: A hormonal perspective.
American Naturalist,154(Suppl.), S4S25.
22. Kimura, M. (1983). The neutral theory of molecular evolution. Cambridge, UK: Cambridge University Press.
23. Kingsolver, J. G., Hoekstra, H. E., Hoekstra, J. M., Berrigan, D., Vignieri, S. N., Hill, C. E., Hoang, A.,
Gilbert, P., & Beerli, P. (2001). The strength of phenotypic evolution in natural populations. American
Naturalist,157, 245 261.
24. Klir, G. J. (2001). Foundations of fuzzy set theory and fuzzy logic: A historical overview. International Journal
of General Systems,30, 91–132.
25. Klir, G. J. (2002). Uncertainty. In Encyclopedia of Information Systems (Vol. 4, pp. 511–521).
26. Lenski, R. E., Ofrie, C., Pennock, R. T., & Adami, C. (2003). The evolutionary origin of complex features.
Nature,423, 139 144.
27. Mackay, T. F. C. (2001). The genetic architecture of quantitative traits. In G. P. Wagner (Ed.), The character
concept in evolutionary biology (pp. 391–411). San Diego, CA: Academic Press.
28. Maynard Smith, J. (1968). ‘‘Haldane’s dilemma’’ and the rate of evolution. Nature,219, 1114 1116.
29. McNamara, K. J. (1997). Shapes of time: The evolution of growth and development. Baltimore: Johns Hopkins Press.
30. McShea, D. W. (1993). Evolutionary change in the morphological complexity of the mammalian
vertebral column. Evolution,47, 730 740.
31. McShea, D. W., & Venit, E. P. (2001). What is a part? In G. P. Wagner (Ed.), The character concept in evolutionar y
biology (pp. 261 286). San Diego, CA: Academic Press.
32. Michod, R. E. (1994). Review of L. A. Real (Ed.), Ecological genetics.Science,266, 468469.
33. Michod, R. E. (1999). Darwinian dynamics: Evolutionary transitions in fitness and individuality. Princeton, NJ:
Princeton University Press.
34. Mitton, J. B. (1997). Selection in natural populations. New York: Oxford University Press.
35. Monod, J. (1971). Chance and necessity. New York: Knopf.
36. Murphey, G. E., Leadbetter, J. R., & Jensen, G. J. (2006). In situ structure of the complete Treponema primitia
flagellar motor. Nature,442, 1062 1064.
Artificial Life Volume 14, Number 3 373
Natural Selection in Relation to ComplexityS. N. Salthe
37. Pearson, K. (1900). The grammar of science. London: Adam and Charles Black.
38. Poli, R. (2006). Three obstructions: Forms of causation, chronotopoids, and levels of reality.
Unpublished manuscript.
39. Ray, T. (1992). An approach to the synthesis of life. In C. Langton, et al. (Eds.), Artificial Life II.
Reading, MA: Addison-Wesley.
40. Reid, R. G. B. (2007). Biological emergences: Evolution by natural experiment. Cambridge, MA: MIT Press.
41. Ridley, M. (1996). Evolution. Oxford, UK: Blackwell Science.
42. Rose, M. R., & Lauder, G. V. (Eds.). (1996). Adaptation. San Diego, CA: Academic Press.
43. Rosen, R. (2000). Essays on life itself. New York: Columbia University Press.
44. Salthe, S. N. (1975). Problems of macroevolution (molecular evolution, phenotype definition,
and canalization) as seen from a hierarchical viewpoint. American Zoologist,15, 295 314.
45. Salthe, S. N. (1993). Development and evolution: Complexity and change in biology. Cambridge, MA: MIT Press.
46. Salthe, S. N. (2001). Theoretical biology as an anticipatory text: The relevance of Uexku
¨ll to current issues in
evolutionary systems. Semiotica,134, 359380.
47. Salthe, S. N. (2002). Summary of the principles of hierarchy theory. General Systems Bulletin,31, 1317.
48. Salthe, S. N. (2006). Two frameworks for complexity generation in biological systems. In C. Gershenson &
T. Lenaerts (Eds.), Evolution of Complexity, ALifeX Workshop Proceedings (pp. 99– 104). Bloomington: Indiana
University Press.
49. Salthe, S. N., & Crump, M. L. (1977). A Darwinian interpretation of hindlimb variability in frog populations.
Evolution,31, 737 749.
50. Salthe, S. N., & Fuhrman, G. (2005). The cosmic bellows: The big bang and the second law. Cosmos and
History,1, 295– 318. Available at
51. Sober, E. (1984). The nature of selection: Evolutionary theory in philosophical focus. Cambridge, MA: MI T Press.
52. Thoday, J. M. (1953). Components of fitness. Symposium of the Society for Experimental Biology,7, 96– 113.
53. Van Valen, L. (1973). A new evolutionary law. Evolutionary Theory,1, 130.
54. Wagner, G. P. (Ed.). (2001). The character concept in evolutionary biology. San Diego, CA: Academic Press.
55. Wagner, G. P., & Laublicher, M. D. (2001). Character identification: The role of the organism. In G. P.
Wagner (Ed.), The character concept in evolutionary biology (pp. 143 165). San Diego, CA: Academic Press.
56. Wallace, B. (1968). Topics in population genetics. New York: Norton.
57. Williams, P. D., Day, T., Fletcher, Q., & Rowe, L. (2006). The shaping of senescence in the wild.
Trends in Ecology and Evolution,21, 458 463.
58. Wilkins, A. S. (2002). The evolution of developmental pathways. Sunderland, MA: Sinauer.
59. Wilson, D. S. (1980). The natural selection of populations and communities. Menlo Park, CA: Benjamin/Cummings.
60. Wilson, D. S. (2005). Natural selection and complex systems: A complex interaction. In C. Hemelrijk (Ed.),
Self-organization and evolution in biological and social systems. Cambridge, UK: Cambridge University Press.
61. Wolf, J. B., Brodie, E. D., III, & Wade, M. J. (2000). Epistasis and the evolutionary process. New York: Oxford
University Press.
62. Wolpert, D. H., & Macready, W. (2007). Using self-dissimilarity to quantify complexity. Complexity,12,
63. Wray, G. A. (1995). Punctuated evolution of embryos. Science,267, 1115 1116.
Artificial Life Volume 14, Number 3374
Natural Selection in Relation to ComplexityS. N. Salthe

Supplementary resource (1)

... For example, we can highlight the algorithmic complexity, computational complexity, information complexity and statistical processing of information. Different definitions of the complexity of living systems are considered in (Heylighen, 1999;Dawkins, 1986;Miconi, 2008;Piqueira, 2009;Finlay andEsteban, 2009, Marquet, 2000;Gell-Mann, 1994;Crutchfield, 2003;Salthe, 2008). In particular, two papers (Melkikh, 2014a,d; emphasized that this problem (associated with the need to enumerate an exponentially large number of genomic variants) should be solved on the basis of precise mathematical formulation. ...
... The same can be said of the synthesis of complex systems from components. Salthe (2008) noted that there is a positive definition of complexity (Gell-Mann, 1994;Crutchfield, 2003): ''effective'' or ''structural'' complexity, a concise listing of the regularities shown by a system (note that this definition actually coincides with algorithmic complexity). There is also a negative definition: in complex systems, situations arise that cause a surprise. ...
A review of the mechanisms of speciation is performed. The mechanisms of the evolution of species, taking into account the feedback of the state of the environment and mechanisms of the emergence of complexity, are considered. It is shown that these mechanisms, at the molecular level, cannot work steadily in terms of classical mechanics. Quantum mechanisms of changes in the genome, based on the long-range interaction potential between biologically important molecules, are proposed as one of possible explanation. Different variants of interactions of the organism and environment based on molecular recognition and leading to new species origins are considered. Experiments to verify the model are proposed. This bio-physical study is completed by the general operational model of based on quantum information theory. The latter is applied to model of epigenetic evolution. We briefly present the basics of the quantum-like approach to modeling of bio-informational processes. This approach is illustrated by the quantum-like model of epigenetic evolution.
... At the genomic level, it has been proposed that, through a process of cumulative natural selection, major evolutionary innovations within organisms can take place with beneficial substitutions becoming fixed and accumulating over time, thus leading to improved and adapted organismic features [3]. But at the intragenic level, however, cumulative natural selection is a far more uncertain means with which to explain innovation except when it is limited to the optimization or the readjustment of protein activities in response to environmental pressures [4]. ...
... Generally, an approximation of the average number of rounds expected in order to evolve a nucleotide string of size L characters should be possible but with it now divided into functional elements of length M codons. If the number of constituent elements is L/3M, then the rounds required without any constraint of order can be derived from equation (4). If the upper limit of the distribution is determined as being no more than 4 times greater than the mean it then becomes: ...
... Nonetheless, an in depth appreciation of the rise and success of natural selection puts the question into a more complex and often nonbiological scenario (e.g., Salthe 2006Salthe , 2008; see also Marquet 2009 in this issue), involving logical, cultural, economical, political, and thermodynamical arguments, among others. Acknowledging such scenario by no means discredits Darwin and his role in the consolidation of evolutionary thinking, but it does circumscribe the biological relevance of natural selection, and also warns us against conflating Darwin's theory with evolutionary theory or evolution itself. ...
... Finally, Marquet (2009) discusses the possibility to expand evolutionary theory in the Darwinian tradition by incorporating an interaction between the processes of natural selection, self-organization and niche construction. For such purpose, it is necessary to acknowledge that natural selection alone cannot explain the evolution of complexity in living systems (e.g., see Salthe 2008) and it plays just a partial role in the processes driving adaptive change. A central argument in Marquet's essay is that "the environment carries the footprint of its past and present inhabitants", and "organisms do have an active role in shaping and constructing the environment they inhabit, and in doing so modify the biotic and abiotic sources of natural selection upon them and on other organisms". ...
Full-text available
Throughout the whole year 2009, the global biological community has celebrated the legacy of Charles Darwin, commemorating the anniversaries of his birth and the publication of "The origin", one of the most influential books of modern history. In this context, the Revista Chilena de Historia Natural inaugurates its new "Special Features" section with four independent essays dealing with the past, present and future of Darwin's ideas. This initial presentation focuses on some loose ends of this Darwinfest, particularly on some forgotten anniversaries directly or indirectly related with Darwin's, and summarizes the contributions of the three essays following this introduction in the present issue.
... 44-45) as well as Darwin (3). The Red Queen's hypothesis continues to attract much attention (3)(4)(5)(6)(7)(8)(9)(10). However, within a multispecies ecological system it remains unclear whether evolution will cease or continue in the absence of external abiotic perturbations. ...
Significance The work presented here demonstrates analytically for the very first time (to our knowledge) that, within a very general theoretical framework, both Red-Queen type of continuous evolution and evolutionary stasis may be the outcomes of ecological interactions within a multispecies ecological community. Whether or not evolution will cease or continue in an abiotically stable environment (i.e., where there are only biotic forces) has been an unsettled problem within evolutionary biology. Our contribution specifies the ecological conditions for which Red-Queen type of continuous evolution and stasis will result. The new and general eco-evolutionary model provides a profoundly new basis for further theoretical and empirical work within the field of coevolution within multispecies ecological systems.
... In Warren Weaver's terminology, rather than the disorganized complexity of gases or liquids, life exhibits organized complexity [Weaver, 1948]. Complexity is generally considered to increase over the course of organismal evolution [Bonner, 1988;McShea, 1996], but there is less agreement on what this means and how it happens [Salthe, 2008]. ...
This chapter discusses the gene-centric view of evolution that is emerging "extended synthesis" in which the theory of genes is complemented by, and in some cases replaced by, a theory of forms. As discussed in this chapter, the solution of long-standing evolutionary puzzles concerning the rapid appearance of body-plans and organ forms with conservation of a molecular toolkit and the apparently abrupt recruitment to unprecedented functions of novel tissue motifs and protein complexes, can be advanced by incorporating evidence for a causal disconnection between phenotype and genotype. This includes recognition that developmental systems and proteins (and their complexes) each have inherent dynamical properties that are only loosely related to the genes associated with them, and that each type of system is capable of undergoing changes in their form and function, often in a discontinuous fashion, through means other than genetic variation. While the complex dynamical systems perspective has thus contributed to the expansion of evolutionary theory, the multiscale, multidetermined, multi-historied nature of organisms has up till now stood in the way of its accomplishing the 20th century goal of a general theory of life. What it is most successful with are models for subsystems: intermediary metabolism, gene regulatory networks, multicellular development, neuronal networks, and so forth, each operating on its own level, with its characteristic principles, rates and modes of evolutionary change. The laws of emergence and interlevel causation continue to be elusive.
... 25) To the extent that living processes are dissipative processes that can be adequately described as constellations of self-organized processes, some version of a rephrased Second Law should apply. Following this line of reasoning, a number of researchers (13)(14)(15)(16) have argued that organisms, ecosystems, and the process of evolution all should exemplify the MEP principle. ...
The present paper reviews evidence for ecological evolution of Montiaceae. Montiaceae (Portulacineae) comprise a family of ca. 275 species and ca. 25 subspecific taxa of flowering plants distributed mainly in extreme western America, with additional endemism elsewhere, including other continents and islands. They have diversified repeatedly across steep ecological gradients. Based on narrative analysis, I argue that phylogenetic transitions from annual to perennial life history have been more frequent than suggested by computational phylogenetic reconstructions. I suggest that a reported phylogenetic correlation between the evolution of life history and temperature niche is coincidental and not causal. I demonstrate how statistical phylogenetic comparative analysis (PhCA) missed evidence for marked moisture niche diversification among Montiaceae. I discount PhCA evidence for the relation between Montiaceae genome duplication and ecological diversification. Based on the present analysis of Montiaceae evolution, I criticize the premise of the prevalent statistical approach to PhCA, which tests Darwinian deterministic hypotheses against stochastic evolutionary null models. I discuss theoretical/empirical evidence that evolution is neither stochastic, nor Darwinistically-determined, but idiosyncratic. Idiosyncraticity describes the outcome of a stochastically perturbed nonlinear chaos-like process. The Principle of Evolutionary Idiosyncraticity (PEI) is based on the evolutionary theory of Natural Drift, which maintains that determinism in evolution is a property of the organism and not, as maintained by the theory of Natural Selection, its traits or its milieu. This determinism is characteristic of chaotic functions, which are absolutely determinate, generate self-similarity, but remain absolutely unpredictable. PEI explains precisely observations that evolution proceeds not linearly, but chaotically, producing both quasi-linear fractal-like patterns and non-linear jumps. PEI has ramifications for all areas of macroevolutionary research. In particular, it demonstrates both the fallacy and futility of the statistical PhCA approach that interprets evolutionary causes in terms of evolutionary correlations. However, statistical methods of PhCA can be applied heuristically and fruitfully to reveal idiosyncraticity and discover evolutionary novelty. This, in turn, is demonstrated by the emergence of statistical anomalies in evolutionary analyses of Montiaceae.
Charles Darwin’s theory of evolution has answered so many questions about human origins over the last 150 years. Most people now accept that humans are the descendents of other species modified by natural selection from random variations. The trail is clearly there in shared structures, in the sequence of development in the embryos, in the fossil record and its changes over time, and so on. So, we can accept the story, more or less as Darwin told it, along with all its implications.1 Yet modern biologists and psychologists know, as laypersons suspect, some aspects of the story are incomplete.
The concept of complexity reminds one of the tasting notes of a rare vintage: everybody knows what you are talking about, but the realities continuously slip through our fingers. Moreover, in the scale of complexities most would agree that life is intrinsically more complex than, say, a galaxy. So too we suppose that some sort of metric stretches through the history of life: be it in terms of ecologies, bodyplans or nervous systems. In other words what we see today is manifestly more complex than what was found in the Precambrian. Yet an evolutionary perspective on complexity reveals some unexpected angles. To start with, although the history of life might fall into the cliché of “Once there were bacteria, now there is New York”, in fact when one investigates what are evidently the most primitive representatives of a given group repeatedly they turn out to be “unexpectedly” complex. Many such examples are now available, but amongst the most telling are the eukaryotes. Second, there is the phenomenon of evolutionary inherency, the observation that much that will be required for the emergence of a complex form has already evolved at a substantially earlier stage. A good example involves the protein collagen, essential as a structural molecule in metazoans, but whose origins not only lie deeper in eukaryotic history but whose functions were evidently quite different. Inherency indicates, therefore, that much of complexity is nascent, almost homunculus-like, lying far deeper in the Tree of Life than generally appreciated. Third, whilst the arrow of time seems to lead to ever greater levels of organic complexity, it is as well to remember that these may well include examples that are often dismissed as “simplification” or “regression”.
In the current recognized evolutionary mechanism, an elementary fact has always been ignored: a predominant trait can come from the relationship of diverse characters, besides the quantity of a single character. This research has designed a series of experiments of simulation for circuit layout to test whether the current recognized evolutionary mechanism is effective for the predominant trait from relationships. The results of these experiments demonstrate the defect in the current recognized evolutionary mechanism, which cannot account for the heritable predominant traits from relationships.
Full-text available
The notion that morphological complexity increases in evolution is widely accepted in biology and paleontology. Several possible explanations have been offered for this trend, among them the suggestion that it has an active forcing mechanism, such as natural selection or the second law of thermodynamics. No such mechanism has yet been empirically demonstrated, but testing is possible: if a forcing mechanism has operated, the expectation is that complexity would have increased in evolutionary lineages more frequently than it decreased. However, a quantitative analysis of changes in the complexity of the vertebral column in a random sample of mammalian lineages reveals a nearly equal number of increases and decreases. This finding raises the possibility that no forcing mechanism exists, or at least that it may not be as powerful or pervasive as has been assumed. The finding also highlights the need for more empirical tests.
I have not had time to prepare a formal paper and wish to apologise for this, since I had every intention of doing so and it has only been made impossible by a combination of extremely heavy duties and being ill for several months this year. On the other hand, I understand that a number of you have either read the little essay I wrote recently [Chance and Necessity, Knopf, New York (1971)], or at least heard of it. Therefore, I hope that if those who have read it agree or disagree, you can ask me some questions after this presentation. What I would like to do here, is the following. I have taken a few notes at random, which are, of course, on various subjects that I did discuss in the book, and I would like to emphasise a few points.
In their book Darwinism Evolving, Depew and Weber (1995) develop the thesis that evolutionary theory has been reformulated several times to keep pace with advances in knowledge about the physical world. Darwin’s Newtonian formulation was replaced by a probabilistic formulation early in the twentieth century. According to Depew and Weber, the new science of complexity will force yet another formulation, which is taking place during our time. This general thesis may well be correct but Depew and Weber’s specific account of the relationship between evolution and complexity leaves much to be desired (Wilson, 1995). They largely accept the polemic view of Gould and Lewontin (1979) at face value, arguing that natural selection is far more constrained and adaptations less common than claimed by proponents of the socalled adaptationist programme. Complexity is viewed as something that stamps its own properties on organisms and resists the modifying effects of natural selection. Depew and Weber are not alone in this view. Many complexity theorists and writers seem to parade under the banner ‘Darwin is dead! Long live complexity!’ The following passage by Kauffman (1993: 24) provides one example: In short, if selection is operating on systems with strongly self-organized properties that are typical of the ensemble being explored, then those properties simultaneously are the proper null hypothesis concerning what we would expect to find in the absence of selection and may be good predictors of what we will observe even in the presence of continuing selection. In brief, if selection can only slightly displace evolutionary systems from the generic properties of the underlying ensembles, those properties will be widespread in organisms not because of selection, but despite it.