lecture9 generalization training behavior analysis
TRANSCRIPT
Lecture9 Generalization training
Behavior Analysis
Review
So far we have discussed a number of behavior analytic procedures: reinforcement, prompting, fading, chaining, extinction, differential reinforcement, punishment and discrimination training.
These are the fundamental tools for changing the frequency of behavior
However there are a few more things to learn before the end of term
Lets start with an example
Example
Imagine that you were helping Julie become better at working with others.
Currently other children hate doing activities with Julie because everything has to be her way.
As a behaviorist you decide that you will play with Julie every day. During this time you reinforce any time Julie compromises.
After a while Julie plays very cooperatively, leaving you feel very pleased with yourself.
However the following week, Julie’s teacher gets in touch and tells you that Julie is just as bad as ever at playing well with others.
Generalization
In that situation it was not enough to change target behavior, the effects of the training must generalize.
Generalization: the tendency for the effects of training to spread
There are two types of generalization Stimulus generalization Response generalization
Stimulus generalization
Stimulus generalization: the tendency for the effects of training in one situation to spread to other situations
This definition speaks of the spread of the effects of training In other words, if training changes a behavior, and the
change in behavior carries over to another situation, then you have stimulus generalization. If you reinforce a target behavior in one situation and the
behavior also occurs more often in a different situation then the effects of the reinforcement have generalized
And if you put the target behavior on extinction, and the behavior declines not only in the training setting, but somewhere else, the the effects of extinction have generalized
Stimulus generalization
In short, if the effects of training, whatever the training and whatever the effects, carry over to situations where there was no training, that’s stimulus generalization. For example, training Julie to be cooperative with me
worked in that situation. But the effects of training did not generalize to the classroom
It is important to note that situations can be alike and different in many ways. Lab research consistently shows that generalization occurs more easily when the training situations and the other situations are more alike
For example, if a pigeon learns to peck a disk illuminated by a red light, it will peck nearly as often when the light is dark orange, and at a lower rate if the disk is yellow
Stimulus generalization
Much the same thing happens with humans e.g. if Julie was asked to play computer games with her teacher in the office, she would most probably be cooperative. The adults would be different but the situations would be similar.
If a child played with Julie in the office then the cooperative behavior would decrease, because the situations would have become more different – an adult is very different to a child
Response generalization
Response generalization: the tendency for the effects of training one behavior to spread to other behaviors
For example, training a pigeon to peck at a disk on a wall might increase its tendency to peck at disks on the floor
Pecking the wall and pecking the floor are different behaviors
Any time training one behavior affects the performance of other behaviors, response generalization has occurred For example, after teaching Julie to cooperate, you might
find that she not only cooperates better, but is also more likely to smile and say thank you
Or I might teach my dog to sit before having its breakfast, and it might start to lift its paw also
Response generalization
Response generalization is often important in solving behavior problems
This is because behavior problems involve many different behaviors, and ideally, we want to be able to improve each behavior without have to train each one
Importantly, stimulus and response generalization are difficult to separate because different behaviors occur in different situations
But the way to remember it is: with stimulus generalization the situation is changing, in response generalization the critical change is in the behavior
Case study: Crystal
Crystal was a very shy 4 year oldShe was of average intelligence, however her speech skills
were delayed and she spoke so softly she could hardly be heard
Louis Fleece and his colleagues designed an intervention program to get crystal to speak more loudly.
In order to do this they used shaping, in other words they planned to reinforce slight increases in the volume at which Crystal spoke.
In order to do this Crystal was given a specific training procedure with her teacher outside of class where speaking loudly was reinforced
But twice a week crystal also had to recite a poem in a different situation: the classroom
Case study: Crystal
Thus the intervention included a shaping procedure outside of the class and a practice exercise inside the class
This practice exercise provides a way of monitoring the effects of the training program, as well as providing a way to reinforce the generalization of the behavior
On a scale of 1-20, Crystal tended to speak at an average volume of 2. Post training, in the classroom, her volume increases to around 11
These results suggest that the training that Crystal received with the teacher generalized to the classroom
Importantly, when the behavior did generalize, it was reinforced in the classroom
Rules for Generalization
Generalization is a naturally occurring phenomenonHowever, in attempting to solve behavior problems, people
often expect too much of naturally occurring generalization, often this results in generalization failure
Stimulus generalization failures are common For example, when little Dora is told off for hitting her
brother with a baseball bat in the living room, her parents believe the behavior will be reduced. Until they see Dora hitting Billie with a baseball bat in the front yard!
Response generalization failures are also common For example, if Dora puts down the baseball bat but
instead hits Billie with the frying pan!
Rules for Generalization
Therefore in order to make sure that appropriate generalization does occur, many behavior analysts offer generalization training
This refers to conducting training in a way that will increase the likelihood of generalization
Although there are lots of suggestions for improving stimulus and response generalization, three rules of thumb are especially helpful: Define the target behavior so as to include many variations Make the training situations resemble natural situations Monitor the target behavior in natural settings
Define the target behavior so as to include many variations
If you want to get a change in many different forms of a behavior then you have to include many different forms of the behavior in your training For example, if you want to get a student to study more you must
define the behavior clearly. Reading, making notes, working on exercises and flash cards could all qualify as studying. Studying involves many behaviors.
So if you want to get the reinforcement effects to generalize to behaviors that are not reinforced during training, reinforce several different forms of studying behavior.
If all you reinforce is note taking (no variation) then you cant expect changes in other studying behaviors, in natural settings
Similarly, if you want to get a shy person to interact socially, then don’t just reinforce one type of interaction, reinforce several kinds of social interaction
The more different kinds of behavior you reinforce, the more likely you are to get new forms of behavior in natural settings
Make the training situations resemble natural situations
The greater the similarity between the setting in which the training occurs and the settings in which the desired behavior change is to occur, the greater the generalization
Notice that plural in the title. Since the target behavior probably occurs in a variety of situations, it is often helpful to do the training a variety of situations
Importantly, if you can change the target behavior in natural settings, then you should. However, in situations where this is not possible, generalization is key e.g. it is unlikely that Julie’s teacher would like the intervention to disturb the rest of the class
Monitor the target behavior in natural settings
At the very least you want to get regular feedback from a teacher/parent etc
But ideally you should observe the target behavior in a variety of natural settings and record data
When you see a satisfactory change in behavior in natural settings, you will know the training is working
For example, in order to find out if Julie has become more cooperative in the classroom, in the office and in the playground. Then you might monitor her behavior in those situations. If she cooperates better in the classroom and the office, but not the playground, then you might alter the training to better resemble the playground!
Flashcard
Generalization: the tendency for the effects of training to spread
Stimulus generalization: the tendency for the effects of training in one
situation to spread to other situations
Response generalization: the tendency for the effects of training one
behavior to spread to other behaviors