Differential Efficacy of Generalization Facilitation Strategies

1
Differential Efficacy of Generalization
Facilitation Strategies Gina T. Chang¹ and
Marjorie H. Charlop-Christy² ¹Claremont Graduate
University, ²Claremont McKenna College
Abstract This study compared three teaching
approaches that are designed specifically to
promote generalization. The three teaching
procedures were multiple exemplar training,
indiscriminable contingency training, and
reinforced generalization training. This study
measured generalization as the transfer of the
use of the acquired behavior into a natural
environment. A multiple baseline design across
and within participants and an alternating
treatment design were used to compare which
procedure was most effective in promoting
generalization. Target behaviors were defined for
three children with Autism. Results indicated
that all training conditions promoted at least
partial generalization. In terms of the
comparative analysis of generalization
facilitation procedures, results indicated that
multiple exemplar training yielded the highest
rate of generalization.

• Method
• Participants were two girls and one boy
  previously diagnosed with autism by two
  independent agencies according to the criteria of
  the DSM IV-TR. The children were enrolled at the
  Claremont Autism Center, an after school behavior
  management program.
• An alternating treatment design was used to
  assess the differential effectiveness of three
  treatment conditions. A multiple baseline design
  across participants was also used, as an
  additional control (Baron Perone, 1998).
• The independent variables were multiple
  exemplars, indiscriminable contingencies, and
  reinforced generalization. The dependent variable
  was the effect of each condition on the
  participants acquisition and generalization of
  the target behaviors to a natural community
  environment in which the target response would be
  appropriate.
• Prior to baseline, each participant was assessed
  to determine appropriate target behaviors.
• Baseline was first established for each child
  across the three behaviors targeted in this
  study. During baseline, each child was given
  between 30 and 60 attempts (depending on the
  length of their baseline) to demonstrate the
  target behavior in the context of a regular
  therapy session.
• In each treatment condition the behaviors were
  taught using traditional behavioral interventions
  such as prompting and delivery of a positive
  reinforcer contingent on correct responding,
  unless otherwise specified in the condition.
• Interobserver agreement (IOA) for childrens
  scores during baseline was a mean of 98.5 percent
  with the lowest score at 80 percent and the
  highest score at 100 percent. During treatment,
  IOA was a mean of 97 percent, with the lowest
  score at 80 percent and the highest score at 100
  percent. For generalization probes, the IOA for
  occurrence of generalization was a mean of 100
  percent.

• Results
• All children reached criterion for all behaviors
  during treatment and generalized behaviors into
  new natural environments.
• Results indicated that all training conditions
  promoted at least partial generalization.
• Results indicated that the multiple exemplar
  training condition promoted the highest amount of
  generalization across subjects.
• Results also indicated that reinforced
  generalization was the least effective condition
  in promoting generalization for two children.
• Rates of learning varied among participants
  however, across two children, indiscriminable
  contingency training had the fastest rate of
  acquisition, and across two children multiple
  exemplar training had the slowest rate of
  acquisition.

Discussion Multiple exemplar training
facilitated the highest percent of generalization
in the opportunities provided. All participants
produced 100 correct responses during
generalization probes for the multiple exemplar
training condition. It is hypothesized that
multiple exemplar training produced the highest
level of generalization because from the outset
of learning the child was unable to associate the
task with one stimulus, which relaxed the
stimulus control during learning. The loose
nature of teaching the task more closely
resembled the natural environment, making the
transfer of the task from the therapeutic
environment to the natural environment more
fluid. Previous research has shown that teaching
loosely produces higher levels of generalization
due to the increased continuity between the
environments (Charlop-Christy et al., 2000). In
contrast, the reinforced generalization training
condition had the lowest percentages of
generalization for two of the three participants.
It is hypothesized that the low level of
generalization was a result of the tight stimulus
control used to teach the tasks. The task was
taught with a predictable schedule of
reinforcement, using the same stimuli until
criterion was met and then the task was re-taught
in a new environment, with new stimuli, and a new
therapist. Although this would seem to promote
generalization, the condition was less successful
because integration of the new environment/stimuli
/therapist occurred after learning had occurred.
This was problematic because the child was taught
using tight stimulus and environmental control
and then had to undergo an undoing of the
control. This supports previous literature that
tight stimulus control hinders generalization
(Horner, et al., 1988). This study demonstrates
that not only is it unnecessary to use packages
of techniques to facilitate generalization, but
that concentrating efforts on training tasks
loosely, through the use of multiple exemplar
training and indiscriminable contingency
training, will produce the best results for
generalization. This study also confirms that the
underlying key to transferring new behaviors is
the continuity between the therapeutic and
natural environments.
Introduction The most prominent reason why
generalization does not occur is due to the
discontinuity between the therapy environment,
where children acquire new behaviors, and the
natural environment, where they are supposed to
produce the learned behaviors. To address the
lack of generalization demonstrated by children
with autism, Stokes and Baer (1977) asserted in a
seminal article the need for researchers and
clinicians to actively program generalization
into treatment. A serious limitation of
research on generalization is the lack of
component analyses no studies have identified
which strategies were most effective in promoting
generalization. There is a need for research to
address which generalization strategies
specifically result in higher levels of
generalization of acquired behaviors. Research
has demonstrated that when grouped together
different strategies increase generalization
however, no studies have attempted to understand
the role that different individual strategies
play in promoting generalization, nor have any
studies compared strategies across children. This
study compared the effectiveness of multiple
exemplar training, indiscriminable contingency
training, and reinforced generalization training
both within and across children to understand the
contributions of each strategy in promoting
generalization. Generalization
Strategies Multiple Exemplar Training Training
multiple exemplars is one of the most extensive
areas of generalization-programming according to
the literature (Stokes Baer, 1977). The benefit
of this strategy is that instead of training the
behavior across every exemplar the child may
encounter, the child is trained across a few
exemplars with the expectation that he will learn
to discriminate the relevant cue and generalize
to all exemplars of the stimulus. Indiscriminable
Contingency Training Indiscriminable
contingencies promote generalization by
delivering a positive reinforcer contingent upon
the occurrence of a target behavior, but on an
unpredictable time table. The time table does not
allow the child to discriminate reinforcement
occasions from non-reinforcement occasions.
Indiscriminable contingencies also increase
generalization because the delivery of
reinforcers closely resembles the natural
environment (Charlop-Christy, et al.,
1999). Reinforced Generalization Training Stokes
and Baer (1977) outlined reinforced
generalization as delivering a reinforcer to the
child when the child demonstrates the appropriate
behavior with increasing gradations of
generalizing, such as, across a new person, place
or object. Their research suggested that
reinforced generalization training required only
a few training settings to increase
generalization.
Results
References Baron, A., Perone, M. (1998).
Experimental design and analysis in the
laboratory study of human operant behavior. In K.
A. Lattal M. Perone (Eds.), Handbook of
Research methods in human operant behavior. New
York Plenum Press. Charlop-Christy, M.H.
Carpenter, M.H. (2000). Modified Incidental
Teaching Sessions A procedure for parents to
increase spontaneous speech in their child with
autism. Journal of Positive Behavior
Interventions, 2, 98-112. Charlop-Christy, M.H.,
LeBlanc, L.A., Carpenter, M.H. (1999).
Naturalistic Teaching Strategies (NATS) to teach
speech to children with autism Historical
perspective, development, and current practice.
The California School Psychologist, 4,
30-46. Horner, R.H., Dunlap, G., Koegel, R.L.
(Eds). (1988). Generalization and maintenance
Life-style changes in applied settings.
Baltimore Paul H. Brooks Publishing Co. Stokes,
T.F., Baer, D.M. (1977). An implicit technology
of generalization. Journal of Applied Behavior
Analysis, 10, 349-367.