https://docs.google.com/document/d/1_w-tmTtfPg5zsljkjb6UhIAP7Y1N3rYeWVaG3dnH8Uw/edit
Hello everyone, in this module, we will discuss generalization and maintenance.
Specifically, we will learn what procedures can be used to promote stimulus and response
generalization as well as maintenance.
First, let’s review the definitions of the key concepts here.
Generalized behavior change refers to the behavior change that can last over time, can appear
across environments and stimuli, and can allow for various related untrained behavior to
appear, after the termination of the intervention. As you can see, generalized behavior change
encompasses three aspects, maintenance, stimulus or setting generalization, and response
generalization.
Specifically, response maintenance refers to the question: after some or all of the intervention
has been terminated, would the individual continue to perform the target behavior? In other
words, would the effects of treatment continue overtime after the treatment has been
challenged or terminated?
For example, once a child learns to tact an apple, would they continue to tact the apple when
the reinforcement schedule has been thinned, and prompt has been removed? Or once an
adolescent has learned how to send text messages to his peers, would he continue if no adult is
around to prompt him to send text messages or to provide edible reinforcers?
As you can see, response maintenance here concerns the generalized behavior change over
time.
Stimulus or setting generalization refers to the question: will the individual perform the target
behavior in a different setting or in response to a different stimulus than the ones used during
intervention without teaching. For example, you have taught a learner to wash his hands in a
bathroom in a clinical setting. Then if you bring the learner to a public bathroom, and he could
wash his hands there without you teaching him again, this is a demonstration of setting
generalization. Likewise, you have taught a learner to label “apple” using real apples. Then if
you draw an apple and the learner is able to say, “apple,” this is an example of stimulus
generalization because the stimulus changes from the actual apple to a drawing.
The last type of generalized behavior change is response generalization. Response
generalization means the response form changes in that the learner demonstrates an untrained
response that is functionally equivalent to a trained response. For example, if you have taught a
child to say “hi,” to his peers and then he started to wave at his peers and say, “good morning,”
without training, this is an example of response generalization because all these responses
serve the same function.
Note that sometimes it may be difficult to differentiate among the generalized behavior
outcomes, and a detailed discussion of how to differentiate these concepts is beyond the scope
of this class. If you are interested, please make sure that you carefully read the chapter,
especially in the sections where Cooper et al. describe how to distinguish instructional and
generalized settings, as well as distinguish various generalized outcomes.
Remember that generalized outcome is not trained. For example, for stimulus generalization to
be demonstrated, the response must not have been reinforced under that stimulus condition.
Likewise, when the varied responses occur in response generalization, you should not have
taught them.
Thus, generalization does not occur naturally for any learners, including both learners with and
without disabilities. For example, even if you demonstrated how to implement conditional
discriminations correctly with your supervisor, it still does not guarantee that you will use it
correctly in the clinical settings by yourself. That is one of the reasons why you would need
continuous feedback initially.
For generalization and maintenance to occur, you would need to plan for them during the
intervention. First, you need to select the target behaviors that would meet the naturally
existing reinforcement contingencies. You would also need to specify the target behavior and
its variations as well as the stimuli and settings under which the target behavior should be
performed.
Let’s take a look at the behavior selection first. The behavior selected should produce
reinforcement for the learner. That is, it should be functional to the learner. If the behavior is
reinforcing to the parents or teachers but does not produce reinforcement for the learner, the
behavior won’t be maintained. Similarly, the behavior should also continue to produce
reinforcement in the learner’s environment even after the intervention has been concluded –
this is called the relevance-of-behavior rule.
For example, manding is a skill that directly produces reinforcers for the learner. Similarly, selfmanagement skills allow the learners to manage their own behaviors and modify the behaviors
to receive reinforcement in different environments. Another similar skill is social skills. Learning
social conversation and initiation may produce direct social reinforcement for the learners in
their natural settings.
Of course, some behaviors taught are unlikely to be maintained by naturally existing
contingencies. For example, if you have previously learned to speak French, but no one around
you speaks French, you won’t continue to demonstrate the skill. In that case, you will need to
continuously contrive the setting where speaking French is required. Similar to teaching
children with autism, if a skill is less likely to be maintained by the natural settings such as
matching-to-sample, you would need to continue to contrive reinforcement or use that skill as
a prerequisite so that the skill can be built into more advanced skills.
Then you will also need to identify the various settings and stimulus conditions in which the
behavior should occur. You need to determine the responses or behaviors that produce the
same function. For example, if you are teaching greeting, what greeting variations should the
learner perform? Is saying “hi” enough? If not, how many greeting variations would be
sufficient?
Second, you will need to identify all the settings and stimuli under which the responses should
be emitted. For example, in how many settings should the learner be able to wash his hands?
You need to prioritize the important and most frequent settings, such as home and restaurants.
The first strategy to promote stimulus and response generalization is to teach enough
examples. In this case, you want to teach all possible stimulus and response examples and then
probe the learner’s performance on the novel examples.
When you teach the learner enough stimulus examples, you should focus on four different
dimensions. First, you need to teach sufficient stimulus samples, that is, the items or the
materials. For example, if you teach a learner to identify an apple, you may want to teach using
different apples, including green and red ones, as well as toy apples and pictures of apples.
Second, you need to teach under different stimulus contexts. For example, you should vary
your instructions, such as “give me an apple,” “where is an apple,” or simply “apple.”
You can also teach the target stimulus in different settings, from one-on-one clinical settings to
group settings to home settings.
Last, you should consider varying the persons who teach the stimulus.
The more examples in terms of stimulus, context, setting, and person, the more likely the
response will be generalized to novel stimuli, contexts, settings, and individuals.
You can also teach enough response examples. For example, you can practice different ways of
greetings and conversational starters with the learner. This is also known as multiple-exemplar
training. Multiple-exemplar training often combines teaching enough stimulus and response
examples.
That said, if you are just teaching multiple examples and see if the learner has mastered these
responses and stimulus examples, that is not a demonstration of generalization.
Generalization is only demonstrated when you probe the response. For example, if you notice
that new response forms emerge, that would be a demonstration of response generalization. If
you program novel stimuli, settings, or teachers, and the learner continues to perform correctly
under these conditions, it is stimulus generalization. Thus, probing is vital to determine if
generalization has occurred.
Based on your target skill, before programming sufficient stimulus and response examples, you
need to conduct general case analysis. The general case analysis is a systematic approach to
identify relevant stimulus conditions and response requirements in generalized settings.
That means you will need to think about what stimulus examples and response requirements
should be included. For example, if you teach how to use a vending machine, you could walk
around the learner’s community and select a number of vending machines to train the learner
with and use other vending machines to probe for generalization.
Likewise, you can teach the learner to use different navigation apps on different mobile
platforms. In this case, you need to select common navigation apps such as Apple Maps and
Google Maps on iOS and Android devices and test if the learner is able to generalize the use of
the navigation to other software on other platforms.
Another example is ordering food in restaurants. In this case, you can select some restaurants
that the learner frequently visits and use those to train the learner the skill and probe if the
learner is able to generalize the skill, ordering food, to other restaurants.
However, you cannot only promote generalization without also help the learner to discriminate
between appropriate and inappropriate conditions. For example, you would need to build
negative examples when you teach the learner to use the toilet. You could train the learner
with different bathrooms in the community, such as the ones at school, home, restaurant, and
clinic. However, you also need to train the learner to identify conditions in which using the
toilet would not be appropriate, such as using a display toilet in IKEA.
These negative examples should share many relevant characteristics of the positive,
appropriate examples so that the learner can precisely identify the negative conditions as well
as conditions in which the behavior is appropriate.
While teaching sufficient stimulus and response examples will promote stimulus and response
generalization, you can also make the instructional setting similar to the generalization setting
to facilitate setting generalization. That is, the learned behavior will be demonstrated in novel
settings.
One method is to program common stimuli. In programming common stimuli, you will be
bringing elements from the generalization setting in which the behavior is desired to the
instructional setting. For example, if you are teaching behaviors related to eating at a
restaurant, you can set up a mock restaurant in the instructional setting by bringing elements
from a real restaurant such as using a menu, waiter, and waitress.
Likewise, you can set up crosswalks in the backyard to teach children with autism to cross the
street. It is much safer than teaching the skill directly in the generalization setting.
The simulated environment is also common for teaching some professionals. For example,
pilots are often trained using simulation cockpit before they are allowed to operate an aircraft,
and firefighters often need to go through training drills.
That said, in a highly controlled environment with fixed stimuli and their properties, sometimes
irrelevant stimulus properties can start controlling the behavior. Thus, to strengthen the
stimulus control, you may consider teaching loosely. In teaching loosely, you would randomly
vary noncritical aspects of the instructional setting. For example, you can use different
instructions, various, but logical, facial expressions, seating locations, and so on. In the above
example of the restaurant, you can vary the instructors who act as waiter and waitress and how
they talk. You can have the learner sit in different locations in the simulated area.
Another example is when you teach conditional discrimination, you can randomly place the
stimulus materials on the table and have the learner select or match in different locations
around the teaching table with varying instructions.
Don Baer suggested a number of noncritical aspects that you can vary in the instructional
settings. I will not repeat them here but do take a look at them. They are located on page 732 in
Cooper et al.
When you design a program, you need to consider if the behavior would contact naturally
existing reinforcement contingencies. When determining the intervention goal, you need to
determine a goal that would allow sufficient behavior to contact the reinforcement.
For example, if a learner answers 2 out of 10 questions correctly, would the learner still receive
praise from their teacher? Or, do you think the learner needs to answer at least 8 to 9
questions correctly?
Other aspects that you should consider is latency. If the learner takes too long to respond,
would the behavior still produce reinforcement? Likewise, if the behavior is too brief, would
that behavior still produce reinforcement?
Second, you need to program contingencies that the learner cannot discriminate if the next
response will produce reinforcement. In other words, you should transition the reinforcement
schedule from continuous reinforcement to variable or random schedule of reinforcement.
Another method that will create indiscriminable contingencies is to gradually delay the
rewards. For example, it will normally take some time for you to receive affirmative feedback
for your submission. This is a delayed consequence.
Both intermittent schedule of reinforcement and delayed reward help produce maintained
responses over time when natural contingencies of reinforcement aren’t immediate.
You can also set behavior traps. Behavior traps are those powerful reinforcement contingencies
that can produce long-lasting behavior changes. Setting a behavior trap requires irresistible
reinforcement and an initial low-effort response that the learner can already perform. Once in,
there are interrelated contingencies that continue to motivate students to acquire, maintain,
and extend their skills.
Some computer-based reading programs organize initial low effort behavior such as using
mouse and keyboard and then provide irresistible reinforcers such as a cartoon.
The reading program may start with easier tasks such as decoding and selecting letters and
sounds and then gradually progress to reading paragraphs. One of my previous clients was
obsessed with one computer-based reading program, Headsprout, and he would repeatedly
mand for the program and would spend hours a day on this program.
Additionally, you can ask people in the generalization setting to reinforce the target behavior.
For example, if a learner just started to mand, you would probably ask their parents to provide
the reinforcer when they mand at home. Likewise, if you teach greetings, you would ask
everyone around the learner in the generalization setting to respond to the learner’s greeting.
That said, it is often difficult for people in the generalization setting to remember reinforcing
the target behavior. For example, it may be difficult for parents to keep track of a learner’s
progress in completing a worksheet. In this case, you can teach the learner to recruit
reinforcement. One simple way is to have the learner request for reinforcement, for example,
showing a completed worksheet to the parents so that the parents can praise them.
Likewise, if you teach a learner to complete a puzzle, you can ignore the learner while they
complete the puzzle and then prompt the learner to ask for attention after they have
completed the task. You can then provide attention for completing the task.
You can also consider mediating generalization which would promote setting generalization.
One method is to contrive or create a mediating stimulus. A mediating stimulus is functional in
the sense that it would reliably evoke the response, and it needs to transportable so that the
learner can carry it around.
For example, when a student learns how to program a destination on a navigation app such as
Google Maps in the instructional setting, and you are afraid that the student will forget how to
program, you can create a cue card or a note that lists the steps for completing the navigation,
which the learner can bring with them. You can also stick the steps on the back of their phone.
This is an example of mediating stimulus.
Similarly, we have talked about the visual activity schedule and personal mobile devices that
can prompt individuals. For example, you can put a visual activity schedule at home, at school,
and in the clinic, or ask the learner to bring the schedule with them. Similarly, you can ask the
learner to bring their phone with them so that notifications and prompts can be delivered.
We also previously discussed self-management. Teaching self-management skills can also allow
the learner to prompt themselves so that the skills and behaviors can be generalized across the
settings if they self-prompt and self-administer reinforcers.
Generalization can also be directly trained. A number of studies have demonstrated that you
can directly reinforce the response variability. For example, you can use a lag schedule of
reinforcement. In a lag schedule of reinforcement, a response that is different from a number
of previous responses is reinforced.
For example, in a lag 3 schedule, you will provide a reinforcer for a response if it is different
from the previous three responses. Say you are teaching greeting and use a lag 3 schedule of
reinforcement, the first response recorded was “hi,” the second was waving at others, and the
third one may be “hi” again. For the fourth one to produce a reinforcer, it must be different
from the previous three responses, so waving again or saying “hi” again will not produce
reinforcer but saying “Good morning” will.
Another method of direct training of generalization is simply instructing the learner about
generalization. That is, you can tell them that they need to generalize the skill to generalization
settings. For example, you may tell the learner before leaving home to the restaurant that they
need to wash hands in the restaurant before eating.
When designing programs, you need to consider how you would achieve a generalized behavior
outcome.
Perhaps, the easiest way is to minimize the need for generalization. That is, if you could
determine what behavior changes are most important, and teaching may automatically
produce generalized results. For example, verbal children who are taught reciprocal
conversations may find talking to others reinforcing if they are able to use follow-up questions
based on a to-and-fro pattern. In other words, you may not need to contrive or plan for
generalization as it could happen naturally when the skill can directly produce reinforcement
for the learners. In addition, you can teach the skills in the setting that the behavior should
occur.
For example, if we target navigation skills, why don’t we just teach these skills on the street or
at home since these are the settings that the learners will most likely to use the skill?
Additionally, you should probe for generalization before instruction to see if the learner
performs some behaviors in the generalization setting. If they do, it will eliminate the need to
teach or reduce the scope of teaching.
You should also probe for generalization during instruction to assess if the generalization would
occur naturally. If not, you may need to plan for generalization during instruction.
Remember also to probe for generalization after instruction. This will let you know the extent
to which generalization has occurred. You should also probe for maintenance a number of days
or weeks after the intervention has been concluded.
During instruction, wherever and whenever possible, be sure to include significant others such
as parents, siblings, and any other caregivers to successfully transfer the control of behavior
from the interventionists to significant others who are in the natural setting with the learner. As
such, learned skills are more likely to be generalized to natural settings.
Like other programs, you should consider the least intrusive and least costly tactics first. For
example, consider just remind the learners to generalize the skill before attempting tactics that
are more intrusive.
That said, it is sometimes necessary to contrive a tactic as the learner may fail to generalize the
skill to the natural setting or to other stimulus conditions, not to mention that response
generalization may be difficult for learners with autism. In these cases, it may be necessary to
provide additional tactics such as multiple-exemplar training and a lag schedule of
reinforcement to explicitly plan for generalization. You will need to determine the tactic based
on the learner’s progress and characteristics.
Order an Essay Now & Get These Features For Free:
Turnitin Report
Formatting
Title Page
Citation
Outline
