Tietelbaum data etc.

[From Bill Powers (950804.1830 MDT)]

Bruce Abbott (950704.1640 EST) --

     Teitelbaum's failure to present his weight data leaves us depending
     on his qualitative statements. These were female rats, and he does
     give the average ad lib weights for each group: normal: 250 gm,
     dynamic VMH: 270 gm, stable VMH: 600 gm. I can only assume that
     when he claims that the rats maintained their weights at the lower
     ratios (over the 2 sessions at each ratio) that this is in fact
     what he observed.

Why, because he's a Good Guy? Let's not assume. That's how myths get
started. He didn't happen to mention how much they were eating ad
libitum, did he?

     I'm only suggesting a theoretical possibility consistent with the
     data Teitelbaum presented, the time periods involved, and my own
     knowledge of the rate at which lab rats lose weight when _totally_
     deprived of food (it usually takes a bit more than a week to bring
     them down to 80% ad libitum weight.

That's 20% in, say, 8 days. If this is a linear decline, we'd get 5% in
2 days (more if nonlinear in the usual fashion). Since in obesity
experiments the normal rats keep their weights constant within much less
than that, we should have seen an effect in two days. We can't just say
"Oh, it's not a very large effect and it takes a long time." When you
look at actual numbers you see what the actual effect is, and that's
what really counts. If a 20% drop doubles the pressing rate, don't you
think we should have seen a 5% drop?

I think there is something going on here that none of us understands
(Tietelbaum included). The answer is most likely going to be found in
the details, when we see what actually happens during the experiments.
We're looking, in the Tietelbaum data, at many complex behaviors that
are occurring over two 12-hour stretches, and trying to represent the
result with just a couple of average measures. That's not good enough
until we PROVE it's good enough.

     I'm not asserting that 1.5 grams per day is the natural diet of a
     wild rat; I'm theorizing that these animals will still drawing down
     stores of body fat to supplement the energy in the 1.5 grams per
     day they were earning

Then they were losing weight.

     .. and that when these store become sufficiently depeleted (to the
     levels seen in wild rats in nature), their lever-pressing rates
     would then pick up to a value sufficient to prevent further weight

Yes, but what's the _real_ explanation? Do rats _really_ wait until
their body fat is exhausted to start pressing the lever faster? Would
you, if you didn't intend to be on a diet? Say you weigh 180 pounds.
Would you not try harder to get food until you weighed 0.8 * 180 = 144

     I am currently working on data reported by Collier, Hirsh, and
     Hamlin (1972) in which the animals earned all their food via lever-
     pressing for an extended period. If my thesis is correct, I should
     see control there.

Good. I agree that that should show at least the long-term control
process. I suppose it would be too much to hope that additive
disturbances were used.

However, there is evidence of fast short-term control in obesity
experiments. Force-feeding rats (or overfeeding rats being chronically
fed by stomach tube) immediately shuts off the operant behavior that was
supplying the food.

     In my view, control over body weight is essentially one-way. If
     other factors have led to an overweight condition in these lab
     rats, you aren't going to see any control over food intake rate
     until body weights fall sufficiently to develop an under-weight
     error in the control system.

Yes, I agree it's one-way, but if a disturbance _increases_ intake,
behavior can become less without shutting all the way off. Actually
there is a possibility of control in the other direction by a different
set of actions: variations in metabolism and excretion.

     Rats are exquisitely sensitive to rate relationships in other
     contexts, so I see no reason why they would be insensitive to them
     on ratio schedules.

But remember that you have uncovered a relationship that calls at least
some of these "sensitive" reactions to rate changes into doubt. What
looks like a change in behavior dependent on rate of reinforcement may
prove to be no sensitivity at all. You're going to have to revisit all
these nuggets of knowledge you have tucked away over the years, to see
which dependencies are illusory and which are real. The facts have
changed, and the changes go way back (to Tietelbaum, 1956 at least).

     Remember that ratio schedules tend to produce high "running rates"
     which do not vary with the schedule requirement; what changes is
     mostly the time required to get started on the next ratio.

That's what you've been proving over the last weeks. However, I wonder
whether if reward sizes were continually increased, there would not come
a time when the running rate would begin to decline.

This reminds me of some more questions, which I could answer if I'd ever
seen any rats in a Skinner Box (gerbils don't count). Suppose a rat is
pressing madly away on a FR-20 and suddenly the ratio is fulfilled. Does
the rat quit pressing immediately on the click, with no overruns? If
not, do the excess presses count toward the next ratio? Are they
recorded? Is there any delay at all in the apparatus, any provision for
waiting for some time after fulfilment of the ratio to start counting up
toward the next one? Is the "access" method of reinforcement designed to
keep rats from just continuing to press until they've delivered a bunch
of pellets before eating them? Would they do that if they could? On FR-
1, 2, or 4, do the rats always press exactly once, twice, or four times
before going to the food dish? Do they ever interrupt pressing before
the end of a long ratio to go look at the food dish to see if anything
has happened?

     If the rats wanted to increase their food intake at the higher
     ratios, they easily could have done so by returning to the lever
     sooner, thus spending proportionately less time "not pressing."
     Deprivation affects mainly the time they spend pressing as opposed
     to not pressing.

They could also have pressed faster, assuming they weren't already
pressing at maximum rate. And if they knew how to do that.

     You didn't mention my discussion of the effect of quinine on
     maintained weight level (Peck, 1978) in terms of conflicting
     control systems, nor my "TV versus getting dinner" senario.

Different subjects. We'll get to conflicting control eventually, but I'm
still trying to understand unconflicted control, if indeed there is any
control in any rat experiments. I want to see one rat actually
controlling one variable and find a decent model for it before I start
getting fancy. If we jump into the middle of a lot of complex processes
and start changing variables and parameters all over the place, we're
just going to generate confusion.

So far, it seems, we have not found any evidence of control in any of
these experiments.

     Perhaps the controlled variable (stored body-fat level perceptual
     signal) was not disturbed enough by the manipulations.

If you've got the controlled variable right, you can detect control
without hitting the system with a sledgehammer. If you have it wrong,
you can certainly create effects by using large disturbances, but what
you learn isn't likely to be worth much. I can tell if you're trying to
maintain an upright position without leaving you flat on your back. I
think the disturbances are plenty large, even a lot larger than
necessary; what's wrong is our concept of what's being controlled, or
perhaps the time-scale of control.

     Control systems are producing the observed behavior. We just
     haven't yet understood how they are organized.

I'm not going to believe that until forced to by the data. From here on
in our job is to try to _disprove_ the hypothesis of control. So far
we've done very well at that: we don't yet have a proposed control
system that passes the Test (using ratio disturbances, but that should
work too).

     Questions: (1) I'm definitely not hungry but I fix myself a bowl
     of ice cream anyway and slowly consume it, even though I know it's
     going to add to my weight and I don't want to gain any more. Why?

Simple hypothesis: Because the "I" you're talking about isn't the only
control system acting at the same or a higher level.

     (2) A rat will press a lever thousands of times per hour to deliver
     brief electrical pulses to a certain part of its brain (the medial
     forebrain bundle, MFB). Why? Explain via PCT.

Several hypotheses are possible.

(1) The electrical pulses are suppressing a feedback signal that would
normally shut off that behavior.

(2) The pulses are raising the reference level for rate of pressing,
directly (or directly producing error signals), thus creating an
artificial positive feedback situation.

(3) The pulses are simulating a perceptual signal of some kind that has
a high reference level but is not currently being controlled. The
external path artificially provides a means of control of that

(4) Cessation of the pulses causes an undershoot or rebound perception
of great pain or horror, and the only way to prevent the rebound, once
the pulses have occurred, is to keep the pulses going at a high level
(an addiction explanation).

(5) The pulses, once accidentally produced during trial-and-error
behavior, suppress reorganization, thus preventing the current behavior
from changing.

(6) The pulses are reinforcing the pressing behavior, thus causing more
pressing and more reinforcement until a limit is reached. (Not a PCT
explanation, but worth comparing with the others).



Bill P.