[From Bruce Abbott (941118.2030 EST)]
Bill Powers (941118.1440 MST)
Question about operant conditioning:
If you set up a rat on a FR-1 schedule, what happens if the rat presses
the lever 25 times in a row? Does it get 25 rewards which it can then
consume at leisure?
Yes and no. The typical operant chamber has a grid floor, and the food cup
will hold only a couple of pellets. Unlike gerbils, rats have no cheek
pouches in which to store their food, so it's eat it now or never. Pigeons on
grain reinforcement have only a few seconds access to the raised hopper, so
they must also "catch as catch can."
Gerbils don't make good participants in typical operant conditioning studies
because of their tendency to store the food. As to the fighting when you keep
a pair of them together, I suspect your gerbils were both male. Male rats
will not tolerate each other in the same box; in the wild they spread out but
when confined together they will fight, sometimes to the death. Put two male
gerbils in the same cage together and often one will pick on the other
relentlessly, until the victim dies from the constant stress. This MAY be
less of a problem if they have been raised together from birth.
Regarding your ECOL4awp and ECOL4bwp, I have received and run them both, and
they are interesting to watch. Before I created ECOLI6 I had a version that
behaved something like this that used simple limits to keep the bug between
two nutrient levels, but it didn't seem like much of a control system (I
believe you'd call it a pair of one-way systems with infinite gain), so I
discarded it.
Although these ECOLIs are interesting in their own right, they are not what I
was expecting. I thought you were going to produce an ECOLI control system
that learned, a la ECOLI4a. Am I mistaken?
As for the FR reinforcement model predictions, I'm working on that, but you
will probably be dismayed by the complexity. Research from the TRT
perspective seems to have revealed a number of factors that have to be taken
into account when deriving a prediction. I am a great believer in devising
experiments that pit one theory against another in this way, but the current
state of knowledge is such that different people may derive different
predictions (depending on which factors they believe are most important in
"controlling" behavior). The best we may be able to hope for is that
"reasonable" assumptions either don't work as (verbally) described when
implemented as a computer model or that when they do, they will look ad hoc
and needlessly complex against the PCT model.
I plan to reply to recent posts by you, Tom, and Rick on my previous
statements about the Law of Effect, but right now I'm out of time. I'll be in
Toledo, Ohio over the weekend, so it probably will not be until early next
week that I can draft a response. Call me old fashioned, call me sentimental,
but there's just something about that ancient bit of Ptolemaic behaviorism I
like . . .
'Till Next Time
Bruce