This is a nice book in the "evo-devo" discussion. Arthur
argues that development is a first class partner in the evolutionary
process. To understand evolution, we have to understand how it acts
over the entire lifecycle of the organism, not just its adult phase
(since organisms have to be "fit" at all stages of their
development, and since the majority of organisms are selected against
-- die -- before adulthood).
all organisms are four-dimensional
organisms do not have life cycles, rather they
are life cycles. We tend to picture adults
Every stage in a life cycle only comes
into being if the previous one survives.
at the level of the
population there will be natural selection at every stage
most species experience most mortality
at young rather than old ages
If a character does not emerge until
late in a life cycle, it will be invisible to natural selection for
all those early stages during which much mortality occurs.
Understanding how the process acts throughout the lifecycle means we
have to incorporate the effect of development as well.
synthesis focused too much on
destruction and the external environment and not enough on creation
and internal coadaptation.
But it is not sufficient merely to "bolt on" a bit of
development to the theory. The development process contributes an
intimately intertwined developmental bias aspect to the
mutation and development are both
important determinants, alongside natural selection, of the directions
that evolution takes.
it is the interactions between them that
are crucial, rather than one of them acting in isolation.
the way in which genes and signalling
constitute the 'chemistry' of development.
But development also has a 'physics'
are systems with quantitative dynamics.
Development is an inherently dynamic process, carving out a
trajectory through character space.
The only way that natural selection can
make new kinds of adult is by altering the course of development. And
it can only make use of the variations that it finds at each stage of
the life cycle. So some things can be modified in early development,
others not until later.
development has a much more far-reaching
effect in determining the direction of evolutionary change than merely
closing off a few avenues. I believe that it does this in two main
ways, one concerning the structure of developmental variation
and one concerning the developmental system's integration as being
itself a target of selection
Not only is development dynamic, it is fast:
evolution ... has created organisms with
trillions of cells from a single-cell starting point over a period of
about a billion years. [Development] repeatedly achieves the same
thing over a single lifetime.
Arthur points out that we have terms for certain evolutionary
processes, but not for this developmental, organismic level process.
What process causes the appearance of
novelty at the genic level? Answer: mutation. What process causes the
appearance of novelty at the whole-population level? Answer: natural
selection. But what process causes the appearance of novelty at the
So he coins a new term: 'developmental reprogramming', to refer to
this process. He identifies four kinds of changes that can be made to
the developmental programming (noting that there are, of course,
intertwined connections between different changes at different
At the developmental level, novelties
can be initiated in any of four ways: by altered timing, positioning,
amount or type of gene product. ... However
there are many
steps in the developmental process
the nature of the change may
alter as development proceeds.
Development itself is intertwined with other levels of evolutionary
Development will only be reprogrammed in
a heritable way if a gene has mutated.
But this developmental aspect of evolution, Arthur argues, has
received short shrift in the standard theory.
So, evolution works as follows: genes
alter by mutation; development alters by reprogramming; populations
alter through selection (and drift); new species arise when
populations diverge to the point where they become reproductively
isolated. Mutation, selection and reproductive isolation are already
well represented in evolutionary theory. Reprogramming is not.
The reason that it needs to be incorporated is that it results in an
inherent bias to the evolutionary process that cannot be explained or
understood if development is not considered. Different
developmental trajectories may be easier to reach than others, leading
to developmental bias (p201.
the tendency of the developmental system
of any creature to produce variant trajectories in some directions
more readily than others).
some forms of reprogramming can be
produced by many mutations, while others are produced by only a few.
That is, some changes are 'easier' to achieve than others. So even if
the rates of occurrence of different mutations at the molecular level
are equal (and usually they won't be), the result will be a higher
probability of producing some mutant individuals than others
This is clear even if we think of a non-linear genotype-phenotype
mapping. Selection acts on phenotypes; given a non-linear mapping some
of these will be more readily obtainable from small mutations than
others. Add to this, as Arthur does, the fact that selection is
occurring during the process of this mapping unfolding (ie, over the
whole developmental life-cycle of the organism), and that the
developmental process itself is heritable and mutable, it is certain
that things are much more complex than just mutating genes.
one way to picture an embryo is as a
trajectory through multicharacter hyperspace. ... Characters that were
not there initially gradually come into being. We go from no brain to
proto-brain to small brain to bigger brain.
phenotypic plasticity should not just be
thrown out of the window by evolutionary theorists because it is not
inherited. If you take one step back from the non-inheritance of
'plastic' variants, you will probably find that the pattern of
plasticity is itself inherited, or at least partly so, and that there
is variation for it in a 'typical' population.
Reprogramming is a more complex process
than mutation, because it is a change in something that is itself by
definition a state of change. Development is a trajectory through
multicharacter hyperspace; developmental reprogramming is a
mutationally induced change in that trajectory.
One feature of developmental reprogramming is "duplication and
divergence". A stage or module that occurs once can, with a relatively
small developmental change, occur more than once. The new occurrences
can then diverge, be modified independently, without compromising the
functionality of the original.
what particular features of organismic
design increase the probability that evolution will find a way to
produce advantageous modifications?
... 'modularity' ... 'duplication
This "duplication and divergence" is a pattern widely seen
in development, possibly nowhere as obviously as in segmented body
plans, where instances of duplication can be accompanied by wild
diversification, yet sometimes by very little.
an arthropod needs a minimum of two
pairs of legs for stable land locomotion
all the vast array of
arthropods have more than this: three pairs of legs in insects, four
in arachnids, 'many' in crustaceans and 'very many indeed' (sometimes
more than 100) in the appropriately named myriapods. Given that there
is an excess of legs over what is required for locomotion, surely some
could be modified into other things? Well, yes, and there are lots of
cases where this has clearly happened. Many arthropod mouthparts are
smallish paired appendages that have been derived from 'spare' pairs
of legs in an ancestor. The centipede's impressive poison claws
have probably also been derived in this way.
we have not yet
developed any predictive ability. Centipedes, for example, would
appear to have a particular surplus of legs, and yet the 'longest'
ones, with 191 pairs, have not specialized any more of them into other
things than the shortest, which have only fifteen pairs.
This examination of centipedes is followed up by an intriguing
observation of a clear bias in their developmental program:
does developmental bias exist in the
all known species of centipedes have odd
numbers of leg-pairs (between 15 and 191). Not a single species has an
even number. If there were only three species of centipede, we could
write this off as being due to chance; but since there are in fact
about 3000 species, this explanation is clearly untenable.
This is a lovely discussion of the evo-devo field, with a personal
slant from the author who is deeply immersed in it himself. The idea
of evolvable developmental programming resonates with me, as a
computer scientist, because it provides a clear mechanism for higher
level processes and modules to appear and be built on, without having
to go back down to primitive components all the time. Once these
higher level processes appear, and evolve, higher levels still can be
built, and complexity can take off.