[From Bruce Abbott (990404.1335 EST)]
Bill Powers (990404.0742 MDT) --
Bruce Abbott (990403.2125 EST)
On ratio schedules,
however, increasing the rate of responding decreases the delay to
reinforcement.
So increasing the rate of responding increases the rate of reinforcement. I
presume that you assume that increasing the rate of reinforcement increases
the rate of responding, too (otherwise, why should the rate of responding
increase to the limit?).
I am assuming that the rat would prefer to get its next pellet as soon as
possible, once it has resumed responding. The rat can minimize the delay by
maximizing its rate of responding.
But in the experiments we examined together, did
we ever observe a rate of bar-pressing to increase to a limit (delay to
reinforcement decrease to a minimum)? We did not.
What we observed is that when the rat returned to the lever to begin
completing the next ratio requirement, that it produced something on the
order of three to five lever-presses per second.
All we ever observed was
a high rate of pressing at the start of each experimental session which
decayed to zero during the session. It decayed to zero not because the rat
simply pressed the bar at a declining rate, but because the fraction of the
time devoted to eating the pieces of food, and later in the session to
exploring the cage and even sleeping increased steadily during the session.
This behavior bore no resemblance to any descriptions of rat behavior under
ratio schedules that I have ever read in the EAB literature.
You are confusing the data we collected on CRF (continuous reinforcement)
with the ratio data. When, throughout the session, each lever-press
produced a pellet, the rat earned pellets at a high rate initially. As more
and more pellets were consumed, the character of the behavior changed
(rather than almost inhaling the pellet, the rat nibbled at it, resulting in
a somewhat slower overall lever-press rate. As the rat consumed even more
pellets, it began to time-share lever-pressing/eating with other activities,
eventually quitting the former altogether.
None of these changes occurred on the cyclic-ratio schedule, nor on the
higher values of the fixed ratio schedules, probably owing to the fact that
far fewer pellets could be earned and consumed during the course of a
session. Instead, lever-press rates during a ratio run remained high
(again, 3-5 presses/second) and average pause length did not vary, across
all ratios tested.
I should also mention that we never observed a rate of responding, either.
What we observed was a rate of bar-pressing. To know that a bar-press is a
response, you must know to what stimulus it is a response, and demonstrate
that the relationship is that which holds between a stimulus and a
response. To call every action a response is to force an unproven
interpretation onto the data.
Agreed. Just deferring to conventional terminology, in which a response is
defined as a particular observable event that is produced by the animal.
Skinner took pains to note that in his usage such "responses" should not be
taken to imply causation by some prior stimulus. I'd just as soon drop the
term. In fact, Skinner invented a replacement -- "operant" -- in order to
avoid the implication of stimulus elicitation.
You are describing what OUGHT to happen according to reinforcement theory
as if it actually did happen. But what actually does happen has no
relationship to the predictions of reinforcement theory. To put it
politely, reinforcement theory is based on imaginary data, which are
substituted during summaries of findings for the real data.
Completely inaccurate and unfair accusation. I have described what _does_
happen on ratio schedules. You are basing your accusation upon a total
confusion of which observations went with which schedules. To put it politely.
When you describe VI schedules I have to presume that you do the same
thing. You describe not what actually does happen during a session, but
what your theoretical understanding (based on reinforcement theory) says
OUGHT to happen.
See above. Also, you don't have to presume anything. I sent you the data.
However, there is a maximum rate that the rat can
comfortably sustain, and if it is attempting to minimize the delay, it will
simply ramp up to the maximum sustainable rate, regardless of the ratio
requirement.
That is not a description of what actually happens, but of your
expectations based on reinforcement theory. You can't know that the rat is
attempting to minimize the (repetitive) delay to reinforcement, or maximize
the rate of reinforcement (operationally exactly the same thing), or to
accomplish some other goal, and I believe that under reinforcement theory
you would be hard pressed to justify saying it was "attempting" to do
anything. We never observed a "ramping up" of behavior rate. Any "ramping"
we saw was a "ramping down."
You are right about one thing here -- at this point I'm speculating about
why the observed rate is high and constant. I didn't appeal to
reinforcement theory, so that supposition is simply an inaccurate guess on
your part. (I fact I'm guessing about a possible CV.) I don't believe that
you mean to imply that there is no maximum rate beyond which a rat cannot
sustain its lever-pressing -- clearly there must be. That being the case, I
am suggesting that the rat is near that limit on these FR schedules (the
mechanics of the lever are probably the most important factor determining
what the limit of that rate is). Any "ramping up would occur immediately as
the rat positioned itself at the lever and began its repetitive series of
presses. You seem to think that I am talking about some other process that
would be seen only when comparing rates earlier and later in the session as
the amount of food consumed increased. Again, this is not observed when the
eating rate permitted by the schedule requirements and rate limit is
relatively low, and in any event is a different process than the one I am
describing.
Under PCT, we can define the reference condition of the food intake rate as
that nonzero rate which would just lead to the cessation of bar-pressing
behavior (I believe that's in the BCP glossary as an operational definition
of "reference level"). We can then show that in all cases where the rat's
pressing rate is near the maximum it can sustain (regardless of the
schedule type), the food intake is LESS than this reference condition as
determined for each individual rat. This is true of the data obtained from
your experiments two years ago.
You mean the food intake _rate_? I'm thinking that the rat is controlling
(attempting to minimize) the delay to pellet delivery, but other than that
this is the sort of analysis I had in mind when I suggested that the rat is
probably responding at or near its maximum sustainable rate on FR schedules.
I suggest it is also true of your VI data,
and all other data obtained during EAB experiments under any kind of
schedule that leads to maximum behavior rates.
VI schedules do not lead to maximum behavior rates. That's the problem.
Consequence: rate constancy across ratios.
That sounds like a deduction from observations but it is not: it's a
deduction from your assumptions.
It's supposed to be a deduction from my assumptions. In the future I'll try
to remember to post a warning for you: WARNING -- this is a proposed
explanation of an observation, not the observation itself!! (Personally,
however, I thought it was pretty clear from the context. But perhaps not.)
What we observed was that the maximum
behavior rate occurred when deprivation was highest, and that it ramped
DOWN from there, on the average, as "reinforcements" accumulated.
Again, wrong set of data, Bill (CRF).
On interval
schedules, once response rate reaches some minimum there is very little
relation between response rate and reinforcement rate, and therefore no
reason why the rat should respond at its maximum sustainable rate.
If the actual food intake is less than the reference food intake even at
the maximum rate of pressing, there is a reason, under PCT. You are
confusing the rate of change of a variable with the value of the variable.
Ah, so you are reasserting that, under PCT, response rates should be driven
up toward the maximum under these schedules. IT DOESN'T HAPPEN. So, either
PCT is wrong, or your PCT-based model is wrong. What's your preference?
The relationship between response rate and reinforcement rate is that the
response rate is all that sustains the reinforcement rate. I saw the same
claim you make in a paper by Myerson and Meizin and couldn't believe my
eyes.
You've lost me. Just what is this claim I am supposed to have made?
The _slope_ of the relationship is a factor in the computation of
dynamic loop gain and is very low at high interval settings, but the
_ratio_ of reinforcement rate to response rate (which contributes directly
to static loop gain) is far above zero.
What's all this got to do with anything? It's a non-sequiter.
Anyway, what makes you think that bar-pressing rate under VI schedules is a
variable? As of the last report from you, as I remember it, you were
finding exactly the same thing with VI schedules as you were with ratio
schedules. When you use session averages, the rate of pressing seems to be
a variable, but when you look at the details within the session, it is a
constant. Is that not so?
At the moment I can't recall what the detailed pattern was. However, I
don't see that as crucial. Whether the rat presses the lever at a constant
but slower rate or simply takes more frequent breaks away from the lever,
the result is that fewer lever-presses occur during a session, and that this
average rate becomes progressively lower as the average scheduled time
between pellet-availabilities increases.
Thus
there is no a priori reason to expect that response rates on interval
schedules would remain fixed across schedule values.
If the food intake rate is less than the reference condition at the maximum
rate of pressing, PCT would predict such a thing.
Yes. But this is contrary to the data.
Your verbal reasoning on
this is flawed. In fact, "reasoning" is hardly the word for what you're
doing. "Rhetoric" would be a better description.
Strong words for someone who has his facts wrong. I think it's your turn to
offer appologies.
Regards,
Bruce