Innovation (III)

Harry Erwin erwin at trwacs.fp.trw.com
Fri Apr 9 21:08:54 EST 1993


Neil Rickert comments:

> In article <erwin.734302435 at trwacs> erwin at trwacs.fp.trw.com (Harry Erwin)
> writes:
> >
> >My point is that learning in most mammalian species appears (IMHO) to be
> >"calibration." The basic behavioral patterns are biological (genetic) and
> >learning calibrates those patterns to the real circumstances of the local
> >population.
> 
> I have to disagree with this.
> 
> Certainly there is a great deal of calibration.  But calibration cannot
> account for all behavior.
> 
> Do cats really have the genetically determined characteristics to make
> them adaptable as house pets?  It is plausible that dogs have been
> sufficiently bred to make such an argument for them.  It is much harder
> to make the case for cats which in most cases are essentially wild
> carnivorous predators who have been raised from very young to adapt to
> household living.

It probably goes further than this. We have a cat that was tamed after
being feral. The interesting thing is that she retains a lot of what she
learned in the wild. She also seems to have a retention limit of about six
weeks (based on how she forgets people after not seeing them that long).
But all that does is indicate that recalibration continues. What I don't
see is experimental modification of the plans she has learned to handle
specific tasks. She doesn't do that unless the old way fails.
> 
> Look at what a mammal needs for survival.  It needs to be able to adapt
> to changes in which predators exist, which prey species exist, which
> edible plants exist.  Now with insects there are many generations per
> year.  This allows rapid evolution, so the insects can easily co-evolve
> with other plants and animals in their environment.  But for many mammal
> species intergenerational times are very long resulting in a rate of
> evolution which is too slow to keep up with most changes.  This is only
> possible if non-genetic adaptability is sufficient to deal with the
> changes in the environment.  You can't do this if learning is exclusively
> calibration.
> 
My point should be that adaptability only proceeds fast enough to cope
with innovation.

> >            Once calibrated, the system freezes.

I should add that the system freezes until it is forced to change.
> 
> Again I disagree.  Calibration is continuous throughout life.  The weight
> distribution of an adult animal is very different from that of a younger
> animal, and recalibration is needed in walking, running, jumping, etc.
> For that matter the same argument applies as a sleek young adult dog
> ages into an overweight older dog.  Or a dog which injures a paw needs
> to recalibrate to learn to manage on 3 limbs.
> 
> >                                                 Innovation and
> >experimentation is thus a characteristic of youth, not adulthood.
> 
> Here I agree.  But I believe it is because there is much learning which
> is not just calibration.  The new learning results in new patterns of
> behavior.  These new behaviors are still subject to further calibration.
> But they can never be completely unlearned.  It is this inability to unlearn
> which makes youth the main period of innovation.
> 
> >            Binford and Diamond both indicate that the archeological
> >record seems to indicate that archaic H. sapiens and H. neanderthalensis
> >show no evidence of an ability to develop complex plans.
> 
> Agreed.  But surely the development of complex plans is highly dependent
> on language and on access to recorded experience of others.

There is evidence for simple language in primates (adjectives and nouns),
but not for complex language. Hence complex planning and complex language
are probably related. Both exhibit non-stationary dynamics. My concern is
that to get to non-stationary dynamics, you have to force the genetics out
of convergence to stable states (ESSs), so that new, non-stationary brain
mechanisms can occur. The emergence of complex speech runs into the same
problems that the emergence of art and complex planning encounter. The
brain has to function differently.
> 
> >                                                         The stability of
> >culture prior to modern H. sapiens also suggests that adults were bound by
> >habit and learning in children was basically rote.
> 
> Surely this observation applies quite well to H. sapiens up until the
> industrial revolution.  Changes from stone age to bronze age, etc, were
> relatively slow and largely imperceptible within a single generation.

The archeological data support a technological cycle of about 20 years.
See the work of Martin Wobst.

> The major early cases of rapid culture change were caused by invasion
> resulting in a culture being forcibly imposed.

Colin Renfrew would differ. There are many examples of culture change
motivated by economic self-interest. For that matter, the Chatel-Perronian
appears to be a H. neand. culture with a Cro Magnon tool-kit. It died out
eventually.
> 
> >My point is that innovation as practised by modern H. sapiens has
> >different dynamics from learning as seen in other species. It requires
> >new mechanisms in the brain. It might be neotenic, but the tendency
> >towards stability has to be overcome.
> 
> But you are overlooking the implications of language.  The physical changes
> required for language can be evolutionary, but the resulting effects on
> behavior are revolutionary.

The brain wiring changes are revolutionary. I don't see how the transition
could have been adaptive at all stages. Being a believer in the standard
model, I have to work out a transition that remains adaptive.
> 
> >(I'm thinking on my feet here...) Could there have been a third selective
> >process involved in the evolution of innovativeness? One without a link to
> >the sex of the individual, but otherwise with a similar tendency towards
> >destabilization?
> 
> Language would seem to have the desired effects, although I don't think
> you would call it a selective process.  I think you are discounting the
> rapid social/cultural evolution which is possible with language, yet
> which does not require concurrent physical evolution.
> 
> 

A few notes on natural selection. 
1. A fixed genetic polymorphism has to equalize the fitness of all genes in
the polymorphism. 
2. Fitness is measured by the geometric average number of descendents,
weighted by the probability of sharing the gene.
3. The mechanism of natural selection is essentially that the characteristics
that maximize fitness in the local population will evolve exponentially to
fixation.
4. The mechanism of sexual selection is that if a given sex prefers a
specific characteristic in the other sex and that preference has reached
fixation, then fitness will be maximized in those individuals whose
increased breeding opportunity due to having the characteristic
just balances the decrease in fitness from expression of that
characteristic.
5. Sexual selection can appear in a population if a characteristic is a
reliable correlate of general fitness. Then the other sex can evolve or
learn to treat that characteristic as a measure of fitness. Once that
fixates, the evolution of the characteristic will be driven by sexual
selection, not by the general fitness of the characteristic (until, of
course, the loss in fitness becomes sufficiently great).

This destablization of the ESS defined by general fitness arguments is
what I need to break out of the low-energy ESS associated with being just
barely creative enough to keep up with changes in the environment. (It is
a low-energy ESS--I demonstrated that in 1983 at the Guelph workshop.
See my recent WESScomm paper. It corresponds to a one-sided game against
nature, and a basic existence theorem of dynamic programming shows that
threshold strategies with thresholds calibr


More information about the Bioforum mailing list