ABA III, Session ?? Motivation: Establishing

advertisement
Beyond the EO
Thursday, April 6, 2006
10:45 to 11:45 A.M.
Jack Michael, Ph.D.
Emeritus Professor
Psychology Department
Western Michigan University
email: jack.michael@wmich.edu
website: jackmichael.org
phone: (269) 372-3075.
FAX: (269) 372-3096
This is a copy of the PowerPoint slide presentation that I gave
at the BASIL annual conference on April 5, 2006, at the
Holiday Inn Select in Naperville, Illinois. My presentation was
from 10:45 to 11:45 a.m.
The essential material for this presentation was on slides 1
through 18. Slides 19 through 37 will be used when I give a
similar presentation scheduled for a longer time.
"Wanting something"
Commonsense psychology: What people do at any particular
moment is a function of two variables: knowing how and
wanting. Both must be present.
In the Learning Theory of 1930-1950 (e.g, Clark Hull) knowing
how = learning or habit strength; wanting = drive strength;
behavior is the product of the two.
A behavioral perspective based on B. F. Skinner's analysis of the
drive concept (1938 B. of O., 1953 SHB): Wanting something*
means two things.
1)The occurrence of what is wanted would function as a reinforcer at
that moment.
2)Any behavior that has obtained that reinforcer will be strong at that
moment. (e.g. food, outside; second is the usual evidence)
*Not Skinner's language
2
Drive and the EO
Keller and Schoenfeld, 1950*. Drive refers to a relation
between certain environmental variables (establishing
operations) and certain changes in behavior. Food
deprivation, for example, establishes food as a reinforcer.
Michael, 1982: Establishing operation (EO) can be used to
refer to any** environmental variable that
1. Increases the current reinforcing effectiveness of some stimulus,
object, or event.
2. Increases the current frequency of (evokes) all behavior that has
obtained that stimulus, object, or event in the past.
*Based on Skinner's Behavior of Organisms. (1938), but I think K & S
wanted a term that was less susceptible to internalization than drive.
**Skinner's used deprivation & aversive stimulation--but these were too
narrow for me (how about salt ingestion, temperature, and the sketch of a cat?)
3
Two Defining Effects
1. Reinforcer Establishing Effect:
Increases (but also decreases) the current reinforcing
effectiveness of some stimulus, object, or event.
2. Evocative Effect:
Increases (but also decreases) the current frequency
of all behavior that has been reinforced by that
stimulus, object, or event.
Here is a better arrangement but it has some problems.
4
Establishing Operations:
Two Defining Effects
Problems:
Establishing = abolishing?
A. Rfer Establishing Effect
1. Rfer Establishing Effect: An
Increase in Rfing Effectiveness
2. Rfer Abolishing Effect: A
Decrease in Rfing Effectiveness
B. Evocative Effect
1. An increase in the Current
Frequency of the Relevant
Behavior: Behavior is Evoked
2. A Decrease in the Current
Frequency of the Relevant
Behavior: (Let Evoke refer to
both directions.)
Abolishing is secondary?
Evocative = a decrease?
Evocative is secondary?
New terms were invented
partly to over come these
difficulties.
Abolishing was ok as an opposite to
establishing, but I had no good opposite
for evocative until Laraway, Snycerski
and Poling introduced abative.
5
Motivating Operations (MOs): Two Defining Effects
Value-Altering Effect
Behavior-Altering Effect
MOs alter the current
reinforcing effectiveness of
some S, object, or event.
MOs alter the current behav. of all
types of R that have been rfed by
the same S, object, or event that is
altered in value by the same MO.
Reinforcer
Establishing
Effect:
An increase
in reinforcing
effectiveness.
Reinforcer
Evocative Effect: Abative Effect:
Abolishing
an increase in the a decrease in the
Effect:
A decrease in current frequency current frequency
of all relevant behavior.
reinforcing
effectiveness.
Easier to learn and easier to apply.
6
Motivating Operations: Food Example
Reinforcer Establishing Effect: Food deprivation increases the
current reinforcing effectiveness of food.
Evocative Effect: Food deprivation increases the current frequency
of all types of behavior that have been rfed with food.
(It is easy to say "that are rfed with food", but it should always
be in the past tense, "that have been rfed with food.")
Reinforcer Abolishing Effect: Food ingestion
decreases the current reinforcing effectiveness of food.
Abative Effect: Food ingestion decreases the current frequency of
all types of behavior that have been (not are) reinforced with
food.
7
Motivating Operations: Pain Example
Reinforcer Establishing Effect: An increase in pain increases
the current reinforcing effectiveness of pain reduction.
Evocative Effect: An increase in pain increases current
frequency of all types of behavior that have been reinforced
by pain reduction. (It is easy to slip up and say reinforced by
pain; but in general pain seldom functions as a reinforcer*;
only pain reduction or removal can reinforce.)
Reinforcer Abolishing Effect: A decrease in pain decreases the
current reinforcing effectiveness of pain reduction.
Abative Effect: A decrease in pain decreases the current
frequency of all types of behavior that have been reinforced
with pain reduction.
*There are exceptions, where pain functions as a reinforcer.
8
Additional Issues
1. What about MOs and punishment? Only recently considered-most of MO theory and knowledge relates to MOs for rfmt.
(Some more on this is in a later section.)
2. Direct versus an indirect effect?
a) MO alters response frequency directly (w/out SD).
b) MO alters evocative strength of relevant SDs.
c) Also there are establishing/abolishing effects, and
evocative/abative effects with respect to relevant
conditioned reinforcers (but not for the same response).
3. Not just frequency: Also magnitude (more or less forceful R),
latency (shorter or longer time from MO or SD to R), relative
frequency (R occurrences per response opportunity), & others.
9
A Possible Misunderstanding
That the behavior-altering effect is a secondary product of the
value-altering effect. That the altered response frequency is
solely the result of contact with a more effective reinforcer.
Or that it follows that contact and behavior is increased or
decreased because of the smaller or greater strengthening effect
of the reinforcer on subsequent responses.
Not true. Behavior-altering effects can be seen in extinction
responding--that is, without contacting the relevant reinforcer.
Behavior increases when the MO is strengthened even before the
reinforcer is contacted.
Best thought of as Two Separate Effects.
(From an evolutionary perspective an organism should begin
seeking reinforcement when the relevant MO becomes strong,
and not only after encountering a more effective reinforcer.)
10
But: The two effects do often work together. Reinforcing
effectiveness will only be seen in the future, after some
behavior has been reinforced, but the future can be
immediately after the MO alteration. Thus increased rfer
effectiveness will combine with an increased evocative effect.
If behavior is occurring too infrequently, strengthening the MO
will result in responses being followed by more effective rfmt
(rfer establishing effect); and all behavior that has been so rfed
will be occurring at a higher frequency (evocative effect). The
increase cannot be unambiguously interpreted, but in practice
it will make no difference.
If behavior is occurring too frequently, weakening the MO will
reduce the frequency by reducing the rfer effectiveness (rfer
abolishing effect); and also reducing the evocative strength of
the MO (abative effect).
11
A Cognitive Interpretation
The evocative and abative effects of an MO occur only because the
individual understands--can verbally describe--the situation and
behaves appropriately as a result of that understanding.
A Behavioral Interpretation
Reinforcement automatically adds the reinforced behavior to the
repertoire that will be evoked or abated by the relevant MO.
Punishment and extinction automatically subtract the punished or
extinguished behavior from the repertoire that will be evoked or
abated by the relevant MO.
In general, functional relations between environmental variables
and behavior do not require the organism's understanding in the
sense of verbal description. Consider the behavior of nonhuman
animals like rats.
12
Generality and Motivating Operations
Generality refers to the extent to which behavior trained in one situation occurs in a new
and different situation.
Train a monkey to press a lever when a 500 Hz tone sounds.
Use food pellets as rfmt, and, of course, the animal is food deprived during training.
This results in a repertoire consisting of a relation between the tone, the lever press and
the food-deprivation MO. (SD + MO results in the occurrence of R)
A few days later put the food-deprived monkey in a very different chamber--different
floor, different smells, lever looks different, etc. and tone is different--100 Hz.
Will lever pressing occur? Probably not. The stimuli are too different.
A few days later put the monkey, but food satiated, in the original chamber.
Will lever pressing occur? Probably not. The food deprivation MO is too weak.
Train a child to name objects, using attention and small edibles as reinforcement.
Suppose that we examine the previously learned repertoire in a different physical
setting, with a different trainer, etc. Correct object naming might not occur because
the stimulus situation is too different.
Suppose we examine the previously learned repertoire when attention and small edibles
are no longer effective as reinforcers. The trained behavior might not occur, even
though the stimulus situation is the same, because the MO is too weak.
(But if escape from the situation is reinforcing, behavior may still occur--we have a
new and strong MO.)
13
Motivative vs. Discriminative Relations, MO vs. SD
A. The General Contrast
Both MOs and SDs are learned, operant, antecedent,
evocative/abative relations.
But SDs evoke (S∆s abate) because of the differential
availability of a reinforcer.
Whereas MOs evoke or abate because of the differential
value (differential reinforcing effectiveness) of a
reinforcer.
But more is needed on differential availability.
14
SD: Increasingly Precise Definition
"An SD (discriminative stimulus) is a type of S that evokes a
type of R." [But so does the respondent CS
(conditioned stimulus)].
"An SD is a type of S that evokes a type of R because that R
has been reinforced in that S." [But it will not have
strong control unless it occurs without rfmt. in the
absence of the S (in the S∆ condition)]. (*comment on Schange decrement)
"An SD evokes its R because it has been reinforced in the SD
and has occurred w/o rfmt. in S∆."
But now another assumption must be made explicit.
15
SD: MO in S∆ condition
"An SD evokes its R because it has been reinforced in the SD
and has occurred w/o rfmt. in the S∆." (But, occurring
w/o rfmt in S∆ would be behaviorally irrelevant unless
the unavailable "rfmt" would have been effective as
rfmt if it been obtained.)
This means that the relevant MO for the rfmt in SD must also
be in effect during S∆.
(*With ordinary lab procedure, S∆ w/o MO not thought of.)
In everyday language: To develop control by an SD the
organism must have wanted something in the SD,
responded, and obtained it; and must have wanted it in
the S∆, responded, and did not obtain it: Success in SD
and failure in S∆. Failing to obtain something that was
not wanted is not failure.
16
How about food deprivation as a possible SD?
Two SD requirements: (1) R has been reinforced with food in SD and (2) has
occurred w/o food rfmt. in S∆, but if obtained, food would have been
effective as reinforcement for R in S∆.
(1) Food deprivation meets the first requirement. Food has been available
and has typically followed R in the presence of food deprivation, or in the
presence of related internal stimuli.
(2) R may have occurred w/o being followed by food in S∆, but if obtained
(during food satiation or during the absence of deprivation-related
stimuli) food would not have been effective as reinforcement.
Common sense language: (1) Food was wanted in the SD condition, and
obtained. (2) What was wanted in the S∆ condition that was not available?
Nothing re food. Food was not obtained, but with no food deprivation
MO, food was not wanted.
Food deprivation does not meet the second requirement as an SD but easily
meets the two MO requirements.
17
How About Pain Onset as an SD?
Two SD requirements: (1) R has been reinforced with pain reduction in SD
(painful stim. present) and (2) has occurred w/o pain reduction rfmt. in
S∆ (when painful stim. was not present?), but if obtained, pain
reduction would have been effective as reinforcement for R in S∆.
(1) Pain meets the first requirement. Pain reduction has been available and
typically followed R in the presence of pain.
(2) R may have occurred w/o being followed by pain reduction in S∆
(when pain was not present), but if obtained (when no pain is present)
pain reduction would not have been effective as reinforcement–would
not have been possible.
Common sense language: (1) Pain reduction was wanted in the SD
condition, and obtained. (2) What was wanted in the S∆ condition that
was not available? Nothing re pain. Pain reduction was not obtained,
but with no pain, pain reduction was not wanted--was not rfing.
(*Do we want pain so that pain reduction rfmt is available?? No.)
Painful stimulation does not meet the second requirement as an SD but
easily meets the two MO requirements.
18
UMOs vs CMOs
UMOs are MOs, events, operations, or stimulus conditions, with
unlearned reinforcer-establishing/abolishing effects.
CMOs are MOs, events, operations, or stimulus conditions, with
learned reinforcer-establishing/abolishing effects.
The distinction depends solely upon reinforcer-estab/abolish
effects. Evocative/abative effects are irrelevant because such
effects are always learned.
UMO: We do not have to learn to be reinforced by food after food
deprivation.
CMO: The capacity to be reinforced by having a key, when we
have to open a locked door (reinforcer estab. effect) depends on
our learning history with doors and keys.
(And we also have to learn how to obtain a key, an evocative
relation.)
Elaborate a little on UMOs and CMOs if time permits.
19
(Use if time permits.)
Main Human UMOs
1. Five deprivation/satiation UMOs
2. UMOs related to sex
3. Temperature UMOs
4. Painful stimuli as UMOs
20
(Use if time permits.)
Weakening the Effects of UMOs
For practical reasons it may be necessary to weaken some UMO effects.
Reinforcer-establishing effects: Permanent weakening of a UMO's
unconditioned rfer-establishing effect is not possible. Pain increase will
always make pain reduction effective as rfmt. Food deprivation will
always increase the effectiveness of food as reinforcement.
Temporary weakening by reinforcer-abolishing and abative variables is
possible. Food stealing can be temporarily abated by inducing food
ingestion, but when deprivation recurs, the behavior will come back.
Evocative effects: These depend on a history of rfmt, and can be reversed
by an extinction procedure–let the evoked R occur without rfmt (not
possible in practice if control of rfer is not possible)
Abative effects of punishment history can be reversed by recovery-frompunishment procedure--occurrence of R without the punishment.
(What about bodily contact (a pat on the back), attention? music?)
21
(Use if time permits.)
UMOs for Punishment
Any variable that (1) alters the punishing effectiveness (up or down) of a
stimulus, object, or event, and (2) alters the current frequency (up or down)
of all behavior that has been so punished is an MO for punishment; and if (1)
does not depend on learning, then the variable is a UMO.
Reinforcer-establishing effects (Abative effects are too complex for here.)
UMOs: Pain reduction will always be effective as rfmt unless the current
intensity is so high that no increase is possible. Also true for other
unconditioned pners (some sounds, odors, tastes, etc.).
MOs for conditioned pners. Most punishing events for humans are
conditioned punishers, not unconditioned. There are two kinds.
An S that was paired with an unconditioned punisher (SP) is controlled by
the same UMO that controlled the unconditioned punisher (see above).
An S that was paired with reduced availability of reinforcers is controlled by
the MOs for those reinforcers (continued on next slide).
(*We really need the CMO concept here. The UMOs below are not really UMOs, as
will be considered in detail in the section on CMOs.)
22
(Use if time permits)
MOs for Punishment (cont'd.)
Removing food as a punisher (or changing to a stimulus correlated with less food)
will only work if food is currently effective as a reinforcer, so the MO for food
removal as punishment is food deprivation. (An exception is related to a highly
verbal human and long-term effects.)
Social disapproval as a punisher (frown, head shake, "bad!") may work because
of being paired with SP like painful stimulation, so MO would be the MO for
the relevant SP. More often social disapproval works because some of the rfers
provided by the disapprover have been withheld when disapproval stimuli
have occurred. The MOs would be the MOs for those reinforcers.
Time-out as punishment is similar. The MOs are the MOs for reinforcers that
have been unavailable during time-out.
Response cost (taking away tokens, money, or reducing the score in a point bank)
only works if the things that can be obtained with the tokens, etc. are effective
as reinforcers at the time response cost procedure occurs.
23
(Use if time permits)
Conditioned Motivating Operations: Three kinds
Variables that alter the reinforcing effectiveness of other stimuli, objects, and
events but only as a result a learning history can be called Conditioned
Motivating Operations, CMOs.
There seem to be three kinds of CMOs:
A. Surrogate: CMO-S: An S paired with another MO which has the MO effect
of the MO it was paired with.
B. Reflexive: CMO-R: A warning stimulus in an avoidance procedure which
(1) makes its own removal effective as rfmt, and (2) evokes any behavior
that has accomplished that effect.
C. Transitive: CMO-T: A stimulus that is related to the relation between
another stimulus and some form of rfmt (1) establishes the reinforcing
effectiveness of the other stimulus, and (2) evokes all behavior that has
produced that stimulus.
24
(Use if time permits)
CMO-R: Human Examples
The CMO-R is important in identifying a negative aspect of many everyday interactions
that might seem free of deliberate aversiveness. The interactions are usually
interpreted as a sequence of SD--R interactions, with each one being an opportunity
for one person to provide some form of rfmt to the other person. But there is a slightly
darker side to everyday life.
Response to a request for information: You are on campus and stranger asks you
where the library is. The appropriate R is give the information or say that you don't
know. What evokes your answer? The request. What reinforces your response?
The person asking will smile and thank you. Also you will be rfed by the
knowledge that you have helped another person.
So the request is an SD. But, it also begins a brief period that is a warning stimulus, and if
a rsp is not made soon, some mild social worsening will occur. The asker may repeat the
question more loudly, and will think you are strange if you do not respond. You would
consider your own behavior socially inappropriate if you did not respond quickly. Even
with no clear threat implied for non-responding, our social history implies some form of
worsening for continued inappropriate behavior. So, the request plus the brief following
period is in part a CMO-R in evoking the response. It is best considered a mixture of
positive and negative parts. But when the question is an inconvenience (e.g. when you
are in a rush to get some-where) the CMO-R is probably the main component.
25
(Use if time permits)
Human Examples (cont'd.)
"Thanks." When a person does something for another that is a
kindness of some sort, it is customary for the recipient of the
kindness to thank the person performing the kindness, who
then typically says "You're welcome."
What evokes the thanking rsp, and what is its rfmt?
Clearly it is evoked by the person's performing the kindness.
And the "You're welcome" acknowledgment is the most
obvious rfmt. So the kindness is an SD in the presence of
which a "Thanks" response can receive a "You're welcome."
But what if the recipient fails to thank the donor?
The performance of the kindness is also a CMO-R that begins a
period that functions like a warning stimulus. Failure to thank
within a short period of time would be considered inappropriate.
26
(Use if time permits)
Human Examples: Academic Demand
In applied behavior analysis the CMO-R is often an unrecognized
component of procedures used for training individuals with defective
social/verbal repertoires.
Learners are typically asked questions or given verbal instructions and
appropriate responses are rfed in some way (an edible, praise, a toy,
etc.). Should the questions and instructions be considered primarily SDs
evoking behavior because of the availability of the rfers?
I think not. What happens if an appropriate response does not occur fairly
quickly? Usually a more intense social interaction ensues. The question
usually has relatively strong CMO-R characteristics.
Although it may not be possible to completely eliminate this negative
component, it is important to recognize its existence and to understand
its nature and origin.
27
(Use if time permits)
Human CMO-T: Flashlight example
The rfing effectiveness of many human conditioned rfers is dependent
on other stimulus conditions because of a learning history. Thus
conditioned reinforcing effectiveness is dependent on a context.
When the context is not appropriate the S may be available, but is
not accessed because it is not effective as rfmt in that context. A
change to an appropriate context will evoke behavior that has been
followed by that S. The occurrence of the behavior is not related to
the availability of the S, but to its value.
Flashlights are available in most home settings, but are not accessed
until existent lighting becomes inadequate, as with a power failure.
Sudden darkness as a CMO-T evokes behavior that has been rfed by
obtaining a flashlight. The motivative nature of the this relation is
not widely appreciated. The sudden darkness is usually interpreted
as an SD for behavior that has obtained a flashlight. But are
flashlights more available in the dark? No. They are more valuable.
28
(Use if time permits)
CMO-T: Slotted screw example
Consider a workman disassembling a piece of equipment, with an assistant
providing tools as they are requested. The workman encounters a slotted
screw and requests a screwdriver. The sight of the screw evoked the
request, the rfmt for which is receiving the tool.
Prior to the CMO-T analysis the sight of the screw would have been
considered an SD for requesting the tool. But the sight of such screws
have not been differentially related to the availability of screwdrivers.
Workmen's assistants have typically provided requested tools
irrespective of the stimulus conditions that evoked the request.
The sight of the screw does not make screwdrivers more available, but rather
more valuable--it is a CMO-T, not an SD.
There are SDs involved in this situation. The screw is an SD for appropriate
unscrewing motions. The request is also dependent upon the availability
of the assistant as an SD. But it is a CMO-T for the request.
29
(Use if time permits)
CMO-T: A danger stimulus
A night security guard patrolling an area hears a suspicious sound. He
activates his mobile phone which signals another guard, who then
activates his own phone and asks if help is needed (the rfmt for the first
guard's response).
Is the suspicious sound an SD for contacting the other guard? Only if the
rfmt for the response is more available in the presence than in the
absence of the suspicious sound, which it is not. The suspicious sound
makes the response by the other guard more valuable, not more
available, so it is a CMO-T for activating the phone.
The CMO-T is not an SD because the absence of the stimulus does not
qualify as an S∆. It is defective in two ways: The relevant rfmt is just as
available in the S∆ as in the SD; and there is no MO for the rfmt in the
S∆ condition–nothing is wanted.
The other guard's phone ringing is an SD for his activating the phone and
saying "Hello," the rfmt for which (some response by the person
phoning) has not been available from phones that have not sounded.
Note that the danger signal is not a CMO-R for calling the other guard. It
is reinforced by producing another S, not its own removal.
30
(Use if time permits)
CMO-T and language training
It is increasingly recognized that mand training is an important part of
language programs for individuals with very inadequate verbal
repertoires.
With such individuals, manding seldom arises spontaneously from tact
and receptive language training.
The learner has to want something, make an appropriate request, and
receive what was requested, and in this way the rsp comes under
control of the MO and becomes a part of the individual's verbal
repertoire as a mand.
The occurrence of UMOs can be taken advantage of to teach mands, but
there are two problems.
Manipulating UMOs will usually raise ethical problems.
Much of the human mand repertoire is for conditioned rather than
unconditioned reinforcers.
The CMO-T can be a way to make a learner want anything that can be a
means to another end.
31
(Use if time permits)
CMO-T and language training (cont'd.)
Any stimulus, object or event can be the basis for a mand simply by
arranging an environment in which that stimulus can function as an Sr.
Thus if a pencil mark on a piece of paper is required for an opportunity
to play with a favored toy, a mand for a pencil and a piece of paper can
be taught.
This approach is a part of a number of current approaches to teaching
verbal behavior to children with autism, and is similar to Hart and
Risley's (1975) procedure called incidental teaching.
32
(Use if time permits)
Practical implications of CMO-T vs. SD interpretations
A CMO-T is a stimulus that evokes behavior because of its relation to
the value of a consequence; the SD evokes behavior because of its
relation to the availability of a consequence.
This distinction must be relevant in subtle ways to the effective
understanding and manipulation of behavioral variables for a variety
of practical purposes.
To develop new behavior or to eliminate old behavior by manipulating
the value when availability is relevant, or availability when value is
relevant would seem inadequate or at least less effective than the
more precise manipulation.
This issue is an example of terminological refinement, not a discovery
of any new empirical relations. The value of this refinement, should
it have value, will be found in the improved theoretical and practical
effectiveness of those whose verbal behavior has been affected by it.
33
(Use if time permits)
General Implications of MOs for Applied Behavior Analysis.
Behavior analysis makes extensive use of the three-term contingency
relation involving stimulus, response, and consequence.
However, (1) the reinforcing or punishing effectiveness of the
consequence in developing control by the stimulus depends on an MO.
And (2) the future effectiveness of the stimulus in evoking the response
depends on the presence of the same MO in that future condition.
In other words, the three-term relation cannot be fully understood, nor
most effectively used for practical purposes without a thorough
understanding of establishing operations.
34
The slide show has ended. If you wish to contact me
about the material covered or other related topics,
please use the email address below. If you would like
a copy of the slides email me to that effect, or down
load the presentation from my website.
Jack Michael, Ph.D.
Emeritus Professor
Psychology Department
Western Michigan University
email: jack.michael@wmich.edu
website: jackmichael.org
phone: (269) 372-3075
fax: (269) 372-3096
35
36
Download