[From Chris Cherpas (2001.02.09.1520 PT)]
Chris Cherpas (2001.02.08.1000 PT)--
If you're really interested, I've
got an article in the Behavior Analyst that goes over
some of the problems (but I'm guessing it won't be that
interesting!).
Bill Powers (2001.02.09.0704)--
Try me.
** Warning: this was written before I had learned about PCT!! **
"Do Establishing Operations Alter Reinforcement Effectiveness?"
Published in _The Behavior Analyst_, 1993, vol. 16, no. 2, 347-349.
Abstract: Michael (this issue) defines an establishing operation (EO),
such as food deprivation, as (a) altering the effectiveness of
reinforcement as well as (b) evoking behavior. Although this dual
role for EOs is compelling, it is possible that such operations have
only evocative effects (i.e., function only in the form of antecedent
control). The main question raised here is how the reinforcement-
altering function can be experimentally analyzed. Evolutionary and
conceptual implications of the two-function EO are also considered.
Michael (this issue) defines an establishing operation (EO) as "an
environmental event, operation, or stimulus condition that affects
an organism by momentarily altering (a) the reinforcing effectiveness
of other events and (b) the frequency of occurrence of that part of
the organism's repertoire relevant to those events as consequences"
(p. 192). Both functions are intuitively plausible. But although
it has been experimentally determined that EOs evoke behavior (e.g.,
Skinner, 1938), it is questionable to what degree they also alter
reinforcement effectiveness. Skinner (1953), for example, asserts
that food deprivation affects only the evocation of behavior:
Contrary to what one might expect, experimental studies have
shown that the magnitude of the reinforcing effect of food
may not depend on the degree of such deprivation. But the
frequency of response depends on the degree of deprivation
at the time the response is observed. (p. 68)
Before considering what experimental procedure might distinguish the
two functions, one can imagine an evolutionary advantage in there
being primarily, if not exclusively, an evocative effect. Picture
a food-satiated animal wandering into a novel area covered with a
distinctive gravel that becomes increasingly dense as the organism
proceeds in the same direction. The animal encounters some food and
takes an exploratory nibble. The next day the animal is food
deprived. Will the animal promptly approach the more densely graveled
areas, or will its experience of eating in the previous day's satiated
condition have been ineffective as reinforcement? If an organism does
not need to be food deprived for food to be effective as reinforcement,
there might be a survival advantage, one that could lead to the
evolutionary selection of a reinforcement-without-deprivation function.
Of course, experimental evidence should have priority over speculations
about evolutionary advantages. What experimental evidence is Skinner
(1953) referring to that supposedly limits an EO to an evocative
function? Apparently, Skinner intends to provide this evidence some
pages later:
By conditioning and extinguishing a response under different
degrees of deprivation, it is possible to see the effect of
deprivation in detail. If we reinforce a response in a group
of organisms at the same level of deprivation and extinguish
it in subgroups at different levels, we find that the number
of responses in the extinction curve is a function of
deprivation. The hungrier the organism, the more responses
it will emit during extinction. If, on the other hand, we
condition at different levels of deprivation and extinguish
at the same level, we find, surprisingly enough, that the
two extinction curves contain approximately the same number
of responses. The effect of deprivation is felt during
extinction, not during conditioning. (pp. 149-150)
If one interprets the number of responses in extinction in the first
phase of the experiment as a measure of the EO's evocative effect and
interprets this number in the second phase of the experiment as a
measure of the EO's reinforcement-altering effect, Skinner's earlier
assertion is supported. But although one must always use some
measure of the evocation of (future) behavior to interpret current
alterations in reinforcement effectiveness -- making the experimental
separation of the EO's functions difficult enough -- Skinner's
particular use of the number of responses in extinction is confusing.
One may even wonder if there is yet another, overlooked, EO function
that is only "felt during extinction." One might also prefer a
design that features within-subject, rather than group, comparisons.
Fischer and Fantino (1968) tried to distinguish between the effects
of an EO (food satiation) on the discriminative function of a stimulus
for the terminal link and its conditioned reinforcing effect for the
initial link of a two-link chain. The former is an evocative function,
and the latter is presumably a reinforcement-altering function.
Increasing degrees of satiation lowered the response rate in the
initial link more quickly than in the terminal link. The authors
concluded that satiation affected the reinforcement function more
than the discriminative function, thus implicitly distinguishing the
two effects. Although this result may strengthen Michael's position
and weaken Skinner's, the results can be viewed as initial-link
discriminative control attenuating faster than terminal-link
discriminative control, a result consistent with Nevin's (1974)
formulation of response strength.
Is there an experiment that can clearly determine whether EOs have
a single or dual role in controlling operant behavior? Two obvious
ways to construct a different experiment would be to (a) study a
different EO than food deprivation or (b) change the procedure and
its associated method of measurement. Food deprivation may not be
the easiest variable to experimentally manipulate to distinguish
the two functions, but it is probably the most frequently exercised
EO in the experimental analysis of behavior, and not to study it
would leave a rather large gap in the empirical base for what is
presumably a fundamental principle.
Perhaps a choice procedure would be easier to interpret than those
in the studies previously cited. First, determine that separately
conditioning key pecking on two different keys with, say, variable-
interval (VI) 1-min and VI 6-min schedules would result in greater
relative responding for the VI-1 min key when both were presented
(i.e., concurrently) in probe sessions. If this is the case, then
the procedure provides a way to analyze reinforcement effectiveness,
and one can move on to studying the effects of the EO, per se.
Separately condition, under high deprivation, responding on a green
key with a VI 1-min schedule, and under low deprivation, responding
on a red key, also a VI 1-min schedule. Then, present the stimuli
concurrently under various deprivation conditions in probe
sessions.
If the subjects choose the green key more often, then deprivation
presumably increases the effectiveness of reinforcement. If not,
there might be a deprivation-specific effect: That is, under high
deprivation green would be preferred, whereas under low deprivation
red might be preferred. Another possibility is that the subjects
would choose them equally often, in which case it appears that only
an evocative effect is present (i.e., overall rate of responding
would change with deprivation in the probe sessions, but relative
rate would remain constant). However unlikely, there could be a
greater choice for the red key, in which case deprivation would
presumably have altered the effectiveness of reinforcement, but
downward.
Evolutionary and experimental considerations aside, the two-function
EO has some related conceptual problems to overcome. First, the
obvious influence of Skinner's writings on students of behvaior
analysis makes it important that Michael directly address the
apparent differences between his and Skinner's conceptions of the
EO. Otherwise, assuming the EO survives and flourishes as a
concept, an unnecessary and potentially confusing discontinuity
will remain in the record of the evolution of a basic behavioral
principle.
Second, although the term 'establishing operation' sounds as if
the most important function of an EO is that of establishing
reinforcement effectiveness, Michael appears to be much more
interested in the EO to make distinctions between evocative
functions, especially between discriminative and motivative
functions (Michael, 1982). The reinforcement-establishing
function is more or less assumed. It might be better to find a
name that puts the emphasis on the evocative function, especially
if the EO is found to have minimal or no reinforcement-altering
effect experimentally.
Finally, assume the EO does establish reinforcement effectiveness.
In that case, Michael's notion of the 'availability' of reinforcement
being distinct from the 'value' of reinforcement can get a little
confusing, because what is available is not really reinforcement
unless it is rendered effective by an EO. In other words, is the
EO inside or outside what we call a contingency of reinforcement?
If it is outside, then we must be looking at the "reinforcement"
term as strictly procedural, and potentially behaviorally neutral.
This would surely disappoint those behavior analysts who say that
they have always though in terms of at least a four-term
contingency (i.e., including the EO). If we accept the EO inside
the contingency, then we are back to the problem of saying that
(effective) reinforcement is 'contingent' on the EO -- that it is
(effectively) 'not available' without the appropriate EO.
In conclusion, Michael's two-function EO concept is intuitively
quite compelling. However, more explicit argument and some data
to prove that there really is a reinforcement-altering effect
would make it even more compelling. The concept may have considerable
heuristic value in its current form for new research and applications,
but both the EO and the behavior-analytic conceptual framework into
which Michael is depositing it need to be fine-tuned if the fit
is to last.
References...(OK, I'll supply these if anyone has actually read
this far, much less cares)
Yes, it's me in the present again... I realize this does not address
the issues BP raised in his last post vis-a-vis the way Bruce Abbott
and I have been talking about "establishing [whatever]" -- but I'll
try to get to that later. Meanwhile, you have a particular slice
from the past.
Best regards,
cc