[From Bill Powers (941003.2025 MDT)]
Tom Bourbon (941003.1032) --
The "social roles" in psychophysics are beautifully described in your
Martin Taylor (941003) --
The "open loop" is open because the subject has turned to zero any gain
that might normally exist in the loop of the tested perception. The
situation in many of these experiments is such that the subject
couldn't control the perception anyway.
Both are good points. Also, I think it's good to keep in mind that what
the experimenter thinks of as the stimulus to be perceived may actually
be a disturbing variable, with the subject acting in some way to keep an
actual controlled variable from being disturbed.
There are valid tests of perceptions which involve no control (although
they could be redesigned to use control rather than verbal measures as a
way of "reporting" the state of the perception). I think of the test for
color blindness as a clear example; you see the correct numbers or you
don't, and the way in which you indicate this is of secondary
importance. However, it's always possible that by converting such tests
to control tasks, you could pick up finer gradations in the perceptual
effect -- maybe there are people who fall between the "normal" and
I think your old posting of 931130.1900 put typical psychophysical
experiments into good order in PCT terms. Seeing the "response" as
control of a relationship between one variable and another (the other
being the perception under test) makes the whole situation very clear.
It covers the "adjustment" mode, although implicitly, perfectly.
Clark McPhail (941003.1819) --
The Sharif experiment is an excellent one for seeing how different
people can move toward agreement about nonexistent phenomena! One has to
remember that in the described experiments, the dot of light doesn't
actually move. There is no possibility for independent reality checks by
the subjects -- i.e., cross-checks against a personal framework of
measurement in the experiential field. From other experiments I have
heard of, a great deal more deliberate influence has to be exerted to
make people agree to perceptions that are contradictory to what that
person would otherwise describe. And even then, what is obtained is only
a verbal agreement, which is not the same as perceptual agreement.
The Sharif experiment and others like it has the disadvantage of
depending on verbal descriptions of perceptions. If a subject says that
the light appeared to move 10 centimeters, what distance should the
experimenter write down? Is "10 centimeters" a distance in an objective
space, or is it a phrase the subject uses to refer to some other actual
distance? How do we know that the subjects didn't come to agree on the
same verbal descriptions, while still actually perceiving something
different from each other?
If the subjects had been given a control wheel or handle that could move
the spot of light, the instructions could have been, "Use the handle to
keep the spot in the same position at all times." Now the apparent
movement of the spot could be directly indicated by the amount that the
subject moves the light-source, with no stage of description being
required. Would this lead to the same results? I would guess not. This
might be worth a test, no? Did Sherif's subjects establish reference
conditions for perceptions of light movements, or for ways of talking
about perceptions of light movements? Replicating Sharif's experiment
with half the subjects controlling rather than describing the apparent
light position might alter the interpretation of his results.
Martin Taylor (941003.1745)--
RE: "nutrient bath" evolution experiments.
I was objecting to the idea that an external variable would control
bifurcation and death, but my substitution was worse.
And mine was unnecessary. As often turns out when you reduce an idea to
specific equations, there are a lot of variables that turn out to be
equivalent. Since rate of growth depends on absorbed nutrient, either
absorbed nutrient or rate of growth could be used as a survival
criterion. I ended up saying that if the rate of growth ever became
negative, death resulted. Because of the way the simulation is set up,
this is equivalent, nearly, to saying that if the volume of the creature
ever falls below some lower limit, it is dead, because as it happens if
the growth rate ever becomes negative, the organism never reproduces.
How long can single-celled organisms survive without growth? Should I
let the guys who aren't growing, or are even shrinking, continue to
If we consider the metaphor of ingestion and excretion, perhaps the
entities could excrete at a rate linearly related to a difference
between an internal and the external concentration, and ingest at a
rate proportional to some other function of size.
I have implemented the effect of surface area on nutrient absorption
rate, and have introduced a metabolic loss rate proportional to volume.
The nutrient is replaced in the bath at a constant rate per time
increment, and its total amount starts out at a relatively high level.
Starting with one organism, we get rapid growth of individuals to the
mitosis trigger level, and rapid proliferation. The organisms remain
synchronized and come to an equilibrium population with the nutrient
level dropping to some rather low amount where consumption equals
replenishment rate. The population may or may not continue to grow
depending on parameter values.
Then I made the mitosis trigger volume Vmax change at each time of
reproduction. The new value (for both parent and offspring) was the
previous value plus or minus a random amount of up to 30 percent. This
broke up the synchronization very quickly. It also meant that organisms
could become, on the average, larger or smaller. With the replenishment
rate set low enough, the population quickly grows to nearly the maximum
number of organisms I can handle, 2000, but then the falling nutrient
level begins to knock off the biggest organisms, because their
absorption rate per unit volume is smaller, and as soon as the metabolic
losses exceed nutrient intake (times a growth factor), that organism is
out of there.
At that point there is a collapse of the population to a smaller number,
because the replenishment rate is too small to sustain the population
and the nutrient level keeps decreasing. Then as the replenishment rate
comes into equilibriunm with consumption, the population begins to grow
again slowly, while the average size of the organisms becomes smaller
and smaller. Before long there is a population explosion of tinier and
tinier organisms, with no large ones at all, to the limit of population
I can handle.
I also tried setting a lower size limit for viability, which works as
expected. By adjusting parameters I can get a stable reproducing
population with individuals of small volume just above the minimum size
So I think I now have a classical "natural selection" model that behaves
more or less reasonably in conventional terms. Right at the moment, I'm
stymied as to where to go from here -- this creature is so simple that
there doesn't seem to be much it could do by way of "enhancing its
survival" by the E. coli method.
At the moment, I'm not thinking about E. coli control methods, but just
sort of mulling over the whole problem. I did manage to download a paper
by Thomas S. Ray, "An evolutionary approach to synthetic biology: Zen
and the art of creating life." It was on tierra.slhs.udel.edu, in
tierra/docs, a file called approximately zen.tex. I also downloaded
tierra.doc, which is about the program and running it. It turns out that
.tex files just have some harmless ASCII markup commands in them and are
otherwise readable ASCII.
The reason for mentioning this is that to place a 1/r penalty on growth
is likely to do strange things to the fitness, and to require quite a
bit of hand-tuning of parameters you would like to allow to find their
own values. Wrong choices here could lead to whole populations dying
before they get properly started.
I didn't see any particular problem -- don't you have to play with
parameters in any case to prevent quick extinctions or blowups?
In this approach we aren't specifying fitness; fitness simply falls out
of the conditions of growth and death, and isn't an explicit number in
the simulation. And this is more or less what I'm thinking about in
relation to Ray's (quite excellent, terrifically organized) paper. When
you specify fitness itself as a variable, it's like introducing an
arbitrary value of one variable in a system of interdependent equations.
Fitness should be a dependent variable, depending on the relationships
between the organism and its environment. It's really a dummy variable,
not something that physically exists in the system. As far as I can see,
it's not used in Tierra, which seems to be a true simulation of an
"organism" that lives, explicitly, in a computer.
I'll say more on Ray's paper later. What's bugging me now is this
problem of crossovers and mutations having small effects. The next thing
I'm going to try in my model is to see what happens if Vmax is simply
chosen anew from within some total possible range from zero to some
maximum on each reproduction, instead of being allowed to vary only +/-
20 percent from the previous value. I'm predicting that I'll just get
the same results, only faster.
If so, then this model is clearly not sufficient to apply to Darwin's
finches where the beak size varies only a small amount from parent to
offspring. The images conjured up by the "gene" story are not conducive
to believing that variations during reproduction will be small
variations around the previous value of a gene's effect. The story has
to get very complicated to account for that, with many, many genes
contributing to the same structural feature -- which cuts down a lot on
the number of really different genes that are at work. If genes really
control features like beak size directly, which I am more and more
inclined to doubt, they have to behave like continuous variables, not
Note that in my model there is no "gene" for the size of the organisms.
The gene that does exist, and dutifully varies from one generation to
the next, specifies the size at which mitosis takes place. The actual
size of an organism varies with time, depending both on parameters
inside the organism and processes outside it, so it's not directly
controlled. The pseudo-gene specifies a particular condition under which
a process will start; it happens to be a maximum size specification, and
the process happens to be mitosis, but it could equally well have been a
rate of growth specification, and the process some other one. I don't
see any need for genes to be specifically related to the final results
that we, as observers, use to characterize the organism. The genetics
story is beginning to look a wee bit oversimplified.
Today's posts were particularly interesting; I'll get to more of them