The first part of the book, Chapters 2 to 6, stalks answers to new
questions: What kinds of complex systems can evolve by accumulation of
successive useful variations? Does selection achieve complex systems able
to adapt? Are there lawful properties characterizing such systems? The
overall answer may be that complex systems constructed such that they are
poised on the boundary between order and chaos are the ones best able to
adapt by mutation and selection. Such poised systems appear to be best
able to coordinate complex, flexible behaviour and best able to respond
to changes in their environment. I suggest that selection does achieve
and maintain such poised systems. Further, beyond the selective molding
of individual adaptive systems, there are provocative, promising
indications that linked coevolving complex systems are led by selection,
as though by an invisible hand, to form ecosystems whose members mutually
attain the edge of chaos. Here all may sustain the highest expected
fitness, even while avalanches of coevolutionary changes propagate
through the ecosystem, ringing out old species and ringing in new ones.
These are new issues in our understanding of the evolution of life.
Darwin told us that adaptive evolution occurs by gradual accumulation of
useful variants but failed to tell us what kinds of systems can evolve
successfully by random variation and selection for fitter variants. It is
remarkably easy to lay our minds around this set of issues. Most readers
will be familiar with programs for contemporary sequential processing
computers. The issue is this: How readily might a process of random
mutation and selection operating on the instructions in a computer
program succeed at attaining a complex program to carry out a desired
computation? It is clear that evolution of useful sequential programs is
very difficult for several reasons, the most obvious being that almost
all random alteration in the code wreak dramatic changes in the
computation being performed.
Adaptive evolution occurs largely by the successive accumulation of minor
variations in phenotype. The simple example of computer programs makes it
clear that not all complex systems are graced with the property that a
minor change in system _structure_ typically leads to a minor change in
system _behaviour_. In short, as their internal structure is modified,
some systems change behaviour relatively smoothly, some relatively
radically. Thus we confront the question of whether selective evolution
is able to "tune" the structure of complex evolving systems such that
they evolve readily.
The variability in behaviour as the structure of a system is altered can
be pictured as characterizing the ruggedness of a fitness landscape. In
Chapters 2 and 3 we shall discuss the concept of such landscapes. In
particular, we shall focus on the simple case of protein molecules in
"sequence spaces", where each protein is located next to a large number
of other proteins which differ from it at only one amino acid position.
The capacity of each protein to carry out some specified function allows
us to define the fitness landscape over protein space with respect to
that function. Here fitness peaks represent either local or global optima
for such a function. As we shall see, such landscapes range from smooth
and single-peaked to very rugged and multipeaked.
The character of adaptive evolution depends on the structure of such
fitness landscapes. Most critically, we shall find that, as the
complexity of the system under selection increases, selection is
progressively less able to alter the properties of the system. Thus, even
in the presence of continuing selection, complex systems tend to remain
typical members of the ensemble of possibilities of from which they are
drawn. We shall find in this book many examples of spontaneous order present
in entire ensembles of complex systems. Thus if selection, when operating
on complex systems which spontaneously exhibit profound order, is unable
to avoid that spontaneous order, that order will "shine through". In
short, this theme, central to our concerns, states that much of the
orders in organisms may be spontaneous. Rather than reflecting selection's
successes, such order, remarkably, may reflect selection's failure.
[to be continued]