[From Bruce Abbott (951211.1830 EST)]
Mary Powers 9512.11 --
Bruce Abbott (951210.1715 EST)
A couple of things concern me about your defense of reinforcement
theory, which you describe as
not a finished theory ready to be applied to specific cases;
it is a theory-in-development. Processes have been
inferred from empirical evidence; these basic processes are
assumed to underly any observed behavior (and these
processes include more than reinforcement). However, the
quantitative rules through which those processes act to
influence observed behavior remain to be worked out.
Because you then go on to say that it's applied all the time,
worked out or not:
very little research in EAB is designed to investigate the
reinforcement process per se; reinforcement is used in most
procedures only to provide the motivation for responding, so
that other questions can be investigated.
Reinforcement as defined in EAB is an empirical fact, not a theory. An
operant is a member of a class of activities having a common consequence
(e.g., depressing a lever to the point of switch-closure). When this
consequence is linked to another (the later is made "contingent" on the
former), the frequency (or some other aspect, depending on the contingency)
of the operant is sometimes observed to increase as a result. If so, the
contingent consequence of the operant has been demonstrated to reinforce the
operant.
The ability of some consequences to reinforce the operant that produces them
has been observed to vary with other variables in a systematic way. For
example, food will reinforce an operant that produces it if an animal has
been deprived of food, and there is a regular (nonlinear) relationship
between the length of deprivation and the ability of the food to reinforce,
everything else being equal. Many quantitative relationships such as this
have been identified.
If a research program is focused on identifying how, say, Drug X affects the
pattern of behavior established by training on a fixed-interval schedule of
reinforcement, the schedule itself may serve only to provide a baseline
pattern of behavior against which to assess the influence of the drug. The
changes that regularly accompany drug administration can then be
investigated to determine their origin. For example, the regular
"scalloped" pattern of responding that develops with sufficient training on
an FI schedule indicates that a given pattern of change can become
"entrained" to the FI's fixed interval, suggesting that something in the
system is capable of functioning as a kind of "clock." If the interval
marked by the scallop shortened when the animal is under the influence of
the drug, this fact would suggest that this timing process has been
accelerated. Here the effect of reinforcement per se is not the subject of
the investigation; rather, the subject is the effect of the drug on the
pattern of output over time. The hypothesis suggested by the data, that the
drug is "speeding up" a kind of internal clock, would then be pursued in
further research investigating the animal's perception of time-intervals
(research also not related to the issue of reinforcement).
In these cases, how reinforcement actually works is not the issue. If you
want to know how the properties of different circular saw-blades affect the
speed and fineness of the cut, how the saw gets its power is unimportant, so
long as the power can be delivered in the quantity required when demanded.
In these experiments, the reinforcement contingency merely serves to keep a
stable pattern of behavior going so that other questions can be investigated.
I'm wondering if there's very little research because
psychologists are satisfied with what they have assumed, and
believe that quantitative rules are just icing on the cake.
To convey my intended meaning accurately I probably should have said "a
small proportion of research" rather than "very little research." I didn't
mean to convey the idea that almost nothing is being done to investigate the
quantitative rules, only that this kind of research constitutes a relatively
small proportion of the total effort. Herrnstein, Premack, Timberlake,
Allison, Mazur, Killeen, Fantino, and others have all contributed to a
lively research effort investigating quantitative rules relating parameters
of reinforcement to behavior. However, having such rules established is not
a prerequisite for pursuing other lines of research in which reinforcement
is just a way to establish a behavioral baseline for the investigation of
other processes, such as discrimination.
Who
besides you looks at reinforcement theory as unfinished and under
development?
Everyone in the field, I would hope. Certainly the active researchers!
How many years of being under development do you
think it will take to realize that the "established basic
processes" _can't_ be quantified - that the math purporting to do
so is fatally flawed? Reinforcement theory has been around a lot
longer than PCT, with a great many more researchers involved, and
yet PCT is far more solid and quantitative.
I'd say about two more years. Enough time to develop the operant
simulations, test them against real data, and get my paper on the subject
published. (Less time if someone beats me to it.)
As for physicists thrashing around trying to reconcile relativity
and quantum theory - yes, they are as confused as reinforcement
theorists, but so what? Both physicists and psychologists are up
blind alleys and "just as prone to invent novel principles to
explain the discrepancies in their data". The question is,
presented with some real novel principles that work, would
physicists reject them as casually and for as trivial reasons as
psychologists reject PCT?
As I noted, psychologists typically are not well-grounded in the necessary
computational techniques and structural model development, so those who have
become the "keepers of the keys" are generally ill-prepared to comprehend
and assess the merits of PCT relative to reinforcement theory, such as it
is. Physicists have that training, so they can conduct a proper evaluation.
I don't think physicists just by being physicists are any more or less
open-minded than other scientists. I seem to recall Bill saying that
Black's analysis of negative feedback-stabilized amplifiers was initially
ridiculed as "black magic."
BTW, I would classify psychology as largely pre-paradigmatic.
Not yet ready for prime time as a science, except for PCT. Which
is one reason why psychologists have trouble with PCT - it's not
that it's a different paradigm, it's that it's a real one.
No, I disagree. Different areas within psychology follow different
paradigms, but they are paradigms none the less. Astronomy had a paradigm
before Copernicus, Kepler, Galileo, and Newton came along (and even the
first two did not entirely reject it); EAB clearly has a paradigm, one about
to undergo a major revolution, I fully expect. We in EAB are still trying
to _build_ Ptolemy's model, and Newton's intellectual descendants are
already knocking on the door. I intend to let them in.
Regards,
Bruce