University of Southern California Dissertations and Theses

Stimulus and response generalization of classes of imitative and nonimitative behavior as a function of reinforcement, task, cues, and number of therapists

(USC Thesis Other)

Stimulus and response generalization of classes of imitative and nonimitative behavior as a function of reinforcement, task, cues, and number of therapists

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content STIMULUS AND RESPONSE GENERALIZATION OF CLASSES
IMITATIVE AND NONIMITATIVE BEHAVIOR AS A
FUNCTION OF REINFORCEMENT, TASK, CUES,
AND NUMBER OF THERAPISTS
by
STUART ALLEN GREENBERG
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)
June 1972
INFORMATION TO USERS
This dissertation was produced from a microfilm copy of the original document.
While the most advanced technological means to photograph and reproduce this
docum ent have been used, the quality is heavily dependent upon the quality of
the original submitted.
The following explanation of techniques is provided to help you understand
markings or patterns which may appear on this reproduction.
1. The sign or "target” for pages apparently lacking from the document
photographed is "Missing Page(s)". If it was possible to obtain the
missing page(s) or section, they are spliced into the film along with
adjacent pages. This may have necessitated cutting thru an image and
duplicating adjacent pages to insure you complete continuity.
2. When an image on the film is obliterated with a large round black
mark, it is an indication that the photographer suspected that the
copy may have moved during exposure and thus cause a blurred
image. You will find a good image of the page in the adjacent frame.
3. When a map, drawing or chart, etc., was part of the material being
p h o to g rap h e d the photographer followed a definite method in
"sectioning" the material. It is customary to begin photoing at the
upper left hand corner of a large sheet and to continue photoing from
left to right in equal sections with a small overlap. If necessary,
sectioning is continued again — beginning below the first row and
continuing on until complete.
4. The majority of users indicate that the textual content is of greatest
value, however, a somewhat higher quality reproduction could be
made from "photographs" if essential to the understanding of the
dissertation. Silver prints of "photographs" may be ordered at
additional charge by writing the Order Department, giving the catalog
number, title, author and specific pages you wish reproduced.
University Microfilms
300 N orth Z e e b R oad
Ann A rbor, M ichigan 48106
A X erox E d u c a tio n C om pany
73-738
GREENBERG, Stuart Allen, 1947-
STIMULUS AND RESPONSE GENERALIZATION OF CLASSES
OF IMITATIVE AND NONIMITATIVE BEHAVIOR AS A
FUNCTION OF REINFORCEMENT, TASK, CUES, AND
NUMBER OF THERAPISTS.
University of Southern California, Ph.D., 1972
Psychology, clinical
University Microfilms, A X ER O X Com pany, Ann Arbor, Michigan
THIS DISSERTATION HAS BEEN MICROFILMED EXACTLY AS RECEIVED.
U N IV ER SITY O F S O U T H E R N C A L IF O R N IA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES, CALIFORNIA 9 0 0 0 7
This dissertation, written by
s.tuarfc..Ai.len.. G j c . e e n f e . e . c g . ......
under the direction of his.... Dissertation Com
mittee, and approved by all its members, has
been presented to and accepted by The Gradu
ate School, in partial fulfillment of require
ments of the degree of
D O C T O R OF P H I L O S O P H Y
Dean
Date 4 /1 3 ./1 2
DISSERTATION COMMITTEE
A *-,1/ A J J \
PLEASE NOTE:
Some pages may have
indistinct print.
Filmed as received.
University Microfilms, A Xerox Education Company
ACKNOWLEDGMENTS
I wish to thank my chairman, Milt Wolpin, whose
confidence, support, and friendship made graduate school an
enjoyable experience; Norm Tiber, for his expert guidance
and friendship; Sid Russak and the staff at the Los Angeles
County-University of Southern California Medical Center’s
Children's Psychiatric Out-Patient Department for their
cooperation and encouragement; the therapists whose
abilities and enthusiasm made execution of this research
possible: Keith Crnic, Barbara Gernazio, Ron Goldberg,
Kathy Ledeen, Sharon MacDuffee, Kathy Morton, Jay Rebert,
and Warren Weistein; Steve Berger and James Kahan for their
invaluable assistance with the statistical analyses; and,
most of all, my loving wife, Margie, whose tolerance for
infrequent, intermittent reinforcement made this undertak
ing possible.
TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS .................. ii
LIST OF TABLES.................. iv
LIST OF FIGURES.................. v
INTRODUCTION .................... 1
METHOD.......................... 12
RESULTS.......................... 25
DISCUSSION...................... 39
BIBLIOGRAPHY.................... 47
iii
LIST OF TABLES
Table Page
d dp 1 t’a
1. Sample presentation listing of S and S
cues presented by the therapist during the
second half of each segment of each
session.................................. 15
2. Summary of statistics between individual and
multiple therapist conditions on trial 6
(F [1,12]), the last trial before the
generalization test....................... 32
LIST OF FIGURES
Figure Page
1. Median number of S^, s^e^-ta, and original
behavior during phase I and the general
ization test (trial 7 ) .................... 26
2. Mean latency of nonimitative responses during
phase I and the generalization test
(trial 7) 28
3. Mean frequency of nonimitative responses
during phase I and the generalization test
(trial 7) 29
4. Median number of S^, s^elta^ and original
behavior during phase II .................. 36
v
INTRODUCTION
A previous discussion of generality (Baer 6c Wolf,
1967) has classified types of generalization into gener
ality of environment, of breadth, and of durability.
Generality of environment is described as being achieved
when "... the [behavior] change brought about in one
environment generalizes to other environments such that a
child whose stuttering has been eliminated in the clinic
will also speak fluently at home and in the classroom"
(p. 1). Generality of breadth is referred to by ". . . the
breadth of the behavior class changes, as when a child who
has been taught a small repertoire of social interaction in
a preschool setting then develops a much larger set of
social skills" (p. 1). Although completely accurate, these
classifications seem unnecessarily narrow. Generality of
environment, this author feels, can be viewed as an
instance of stimulus generalization which would also
include the stimulus components of people, task, time,
contingencies, reinforcement, etc. Generality of breadth,
similarly, will be viewed as an instance of response
generalization and will be considered with response induc
tion and generalized response classes.
1
2
On the other hand, what was termed generality of
durability will not be treated here as a type of generali
zation because the durability aspect of behavior transfer
is logically dependent on previous generalization. In
that after a behavior generalizes, the question then
becomes one of durability or extinction of that general
ized behavior.
Cantania (1968, p. 335) defines stimulus generali
zation as "the spread of effects of reinforcement (or
other operations, e.g., extinction, punishment, respondent
conditioning) in the presence of one stimulus to other
stimuli that differ from the original stimulus along one
or more dimensions." The early verbal behavior of human
infants provides examples of such generalization in the
response of "doggy" for all animals, "juice" for all
liquids, and "daddy" for all adult males. In essence, when
different stimuli elicit or cue the same response, stimulus
generalization is said to occur.
Wheeler and Sulzer (1970) found that the teaching
of a verbal response format for a set of standardized
pictures to a subject who spoke in essentially "telegraphic
English" resulted in the subject using the format to
describe additional nonstandardized pictures. On the other
hand, Lovitt and Curtis (1968) found that while preperform
ance verbalization of math problems increased accuracy of
math performance, the increase in accuracy did not
3
generalize to problems in other school subjects.
Wahler (1969) monitored the cooperative and opposi
tional behavior of two school-age boys in their homes and
at school while only manipulating the contingencies
relevant to these behaviors at their homes. While he found
that these behaviors changed at the homes in the predicted
directions, there was no generalized change in the school
until the contingencies were modified there also.
Similarly, Patterson (1965) found that reinforcing a nine-
year-old boy who was hyperactive and lacked attending
behavior resulted in desired changes in the behaviors only
in the therapist's presence and did not generalize to the
rest of the class day. To remedy this stimulus control,
therapy was randomized throughout the day and the therapist
remotely signaled reinforcement instructions to the
teacher.
Varying the therapists and the therapeutic settings
may serve to increase the range of social stimuli that will
prompt or elicit a target behavior. Kale, Kaye, Whelan,
and Hopkins (1968) report that newly trained social greet
ing responses of three withdrawn, chronic schizophrenics
did not generalize to a new experimenter until five new
therapists employed a prompting, fading, reinforcement and
scheduling procedure with the patients.
Bandura (1969), Kanfer and Phillips (1971), and
Allyon and Azrin (1968) suggest the use of multiple
4
therapists and settings to increase stimulus generaliza
tion. Allyon and Azrin (p. 154) report in their series of
token economy studies that M. . .no single employee
administered a given type of reinforcer for more than one
week without another employee serving as a substitute on at
least one occasion."
Generality of behavior change also includes the
possibility that the same stimulus will elicit or cue
several different responses. Theoretically, when responses
lying within one portion of some response dimension are
reinforced (or punished, etc.), responses immediately above
or below the reinforced response are likely to change in
strength also. That is to say there is a spread of effect
(or response induction) of the experimental procedure along
the response dimension, within some class of behaviors.
This class of responses along the response dimen
sion is called a generalized response class. The responses
share common stimuli which serve an eliciting, discrimina
tive, or reinforcing influence on the class of behaviors
(Peterson, 1967). Since imitation plays such a large part
in the acquisition of socialized behavior, classes of
imitative behavior, for which identical stimulus classes of
modeled cues served the discriminative function, were
chosen as the primary dependent variables in the present
study.
5
The rewarding of a subject for duplicating several
of a model's behaviors rapidly results in those imitative
behaviors becoming a generalized response class. Once they
have been learned to a particular level, reinforcing some
of the members of the class results in an increase in
strength of the remaining members of the class without
direct reinforcement for them. In a similar fashion,
extinguishing or punishing some elements in the class
results in a generalized decrease for the entire class.
This concept of generalized response class is not only
descriptive, it is functional and is tested by performing
the above manipulations.
Not only are the strengths of the members of a
generalized response class of imitative behavior inter
related, but the class has other interesting character
istics. When a subject has formed a generalized response
class, new behaviors modeled to him are rapidly acquired in
toto. Additionally, new imitative behaviors that have
never been reinforced are learned and maintained as long as
their performance is interspersed with other reinforced
imitative behaviors. Nonimitative behaviors will even be
maintained if their performance is distributed among
reinforced imitated behaviors. These findings have been
replicated many times, in diverse settings, and without
tangible reinforcement (e.g., Baer & Sherman, 1964; Metz,
1965; Lovaas, Berberich, Perloff, & Schoeffer, 1966; Baer,
Peterson, & Sherman, 1967; Lovaas, Frietag, Nelson, &
Whalen, 1967; Brigham & Sherman, 1968; Peterson, 1968;
Burgess, Burgess, & Esveldt, 1970; Schumaker, 1970;
Peterson & Whitehurst, 1971; and Steinman, 1971).
Several theories have been proposed to explain
these findings: 1) conditioned reinforcement, 2) discrim
ination, 3) reinforcement scheduling, and 4) setting
events.
The "conditioned reinforcement" explanation (Baer &
Sherman, 1964; Lovaas et al., 1966; Baer et al., 1967)
proposes that the property of being similar to a model may
function both to produce and maintain generalized imita
tions. According to Burgess et al. (1970), support for
this hypothesis can be found in the difficulty some
researchers have had extinguishing both reinforced and non
reinforced imitations (Baer et al., 1967; Brigham and
Sherman, 1968). Steinman (1971) criticizes this explana
tion "on both logical and empirical grounds." Logically,
he states, if the conditioned reinforcement explanation
were true, then it would be difficult to explain why
differential reinforcement should be effective under any
circumstances. His point seems to be: stating that
differential reinforcement does not result in differential
performance in generalized imitation situations is a
description of the results and not an explanation. On
empirical grounds, Peterson (1968) and Peterson and
7
Whitehurst (1971) tested whether similarity was the
necessary element in maintaining the unreinforced response
and found it was not. They found that nonreinforced, non
imitative responses were also maintained under the same
conditions as "similar-to-model" responses.
The "discrimination" (Bandura, 1968, 1969 a, b, c)
and the "reinforcement scheduling" (Gerwitz, 1968; Gerwitz
Sc Stingle, 1968) propose similar explanations with differ
ent emphases. Bandura says that nonreinforced imitations
continue to be performed because the subject may not be
able to discriminate reinforced from unreinforced responses
when they are similar and interspersed. Gerwitz and
Stingle say the results are a property of the variable
ratio schedule that occurs when reinforced responses are
interspersed with unreinforced responses, which is the same
as saying that the subject perceives receiving the
reinforcement on one (VR) schedule rather than a constant
J
schedule for discriminative (S ) behaviors and no
dsltd.
reinforcement for nondiscriminative (S ) behaviors.
Both explanations find support in the finding that, as was
stated, unreinforced responses are learned and maintained.
In addition, Peterson (1968) found that unreinforced
responses ceased to be maintained when they were massed
together instead of interspersed individually among
reinforced responses. However, in direct conflict with
these explanations, Steinman (1971) found that when given a
choice of modeling a previously reinforced response or an
interspersed, unreinforced response, the subjects consist
ently chose the previously reinforced one. In addition,
when instructed to model only responses that receive
reinforcements, the subjects’ discriminations were nearly
perfect.
Findings are consistent in that reinforcement is
necessary to acquire imitative responses. Most studies
have found that it is also necessary for maintaining those
responses. Whether its role also includes a discriminative
function or whether there are other social expectancies or
demand characteristics also operating is not entirely
clear.
The "setting events" explanation is the latest to
be formally proposed (Steinman, 1971), although suggestions
have been made in this direction since Baer and Sherman did
so in 1964. Since Terrace (1966) and Rilling et al. (1969)
found that stimuli that are discriminated as consistent
occasions for nonreinforcing events develop neutral or
aversive properties, experimenters have been trying to
determine counter sources of real or perceived reinforce
ment or punishment to explain the maintenance of these, at
best neutral, classes of responses. Since in all the set
tings some type of unconditioned reinforcement is present,
it seems likely that some aspect of the setting is
acquiring (or comes to the setting with already acquired)
9
conditioned reinforcing properties. Determining which
aspect is not yet accomplished. Steinman (1970) has sug
gested considering the subject's pre-experimental history
with regard to compliance with adult commands, instruc
tional variables, the adult being present when the
imitation is performed, the adult watching when the
imitation is performed, the same adult cueing both the
j „delta .
and S response, etc.
Peterson and Whitehurst (1971) found that when
models did not watch whether the child performed the imita
tion or did not remain in the room when he did so, the
imitation decreased. When instructed not to imitate, the
behavior also extinguished. When one model presented the
f? rlpl
S behavior and a different model presented the S
behavior, both models were imitated (Steinman, 1971).
Children are more likely to imitate when they have a
history of previous reinforcement for imitative behavior
(Bandura, 1968; Flanders, 1968; Peterson, 1968). These
findings suggest that most children (preschool children
were the above subjects) can rapidly develop the response
of imitating an adult model and will probably do so unless
situational characteristics create expectancies not to do
so. It would seem that the nonsocial reinforcements or the
noninstructional variables that were used in the above
studies were of insufficient strength to produce a response
counter to the demand characteristics or expectancies or
10
other nontangible reinforcers present in the setting.
In spite of this recent research and theoretical
interest in both stimulus generalization and generalized
imitation (a type of response generalization), many
empirical questions are still unanswered. This study will
attempt to provide some of this information in such a way
that concurrently tests some of the theoretical hypotheses
1) If stimulus generalization occurs as a function
of the similarity of training cues and testing cues, then
subjects who are trained by one person in one setting and
presented the same cues by that same person in a different
setting should produce more generalized responses than a
subject presented the same cues by a different person in
the new testing setting, and
2) Those groups of subjects should produce more
responses when those same or different therapists present
the same cues in the testing situation as opposed to the
therapist not presenting any response relevant cues.
3) If varying stimuli configurations is an effec
tive method of increasing stimulus generalization, then
subjects trained to the same criterion by the same
therapist should show more of a response decrement to
testing by a new therapist than subjects trained to
criterion by five different therapists.
4) If interspersing nonreinforced imitative
behavior with reinforced imitative behavior results in the
11
formation of a generalized response class of imitative
behavior with functional interdependencies between the
members, then reinforcing the reinforced members in a free
recall situation should result in the recall and perform
ance of the nonreinforced members also. When reinforcement
is only present for a different class of behavior in the
free recall situation, then both the reinforced members and
unreinforced members should extinguish simultaneously.
5) If all behaviors are put on extinction during
the recall session, and if generalized imitative behaviors
have acquired secondary reinforcing properties, then they
should take longer to extinguish than previously reinforced
nonimitative behaviors.
METHOD
Subjects
Sixteen 7-to-14-year old children were drawn from
the referral pool of the Los Angeles County-University of
Southern California Children's Psychiatric Out-Patient
Department on the basis of attendance and lack of severe
organic deficits. All had been referred to the hospital's
"Special Problems Clinic," were verbal, appeared to fall
between -2 and +1 standard deviations of normal intelli
gence, and none manifested spontaneous imitative behavior.
Two subjects failed to attend four of the six phase I
trials and were therefore dropped from the study.
Therapists/Recorders
All people who interacted with or recorded the
behavior of the subjects were senior psychology or speech
therapy majors at the University of Southern California.
The therapists received one semester of classroom instruc
tion in the techniques of behavior modification plus an
average of 20 hours of supervised procedural training on
practice subjects before the start of the study. The
recorders were selected from a class in abnormal psychology
and given short-term training and practice in behavior
12
13
sampling and recording techniques. The author supervised
all sessions.
Apparatus and Setting
The study was conducted in two, almost identical,
nine-foot square rooms in the out-patient department.
These rooms each were carpeted and equipped with two
chairs, a rectangular table holding large bowls filled with
standard plastic poker chips, one-way mirrors, sound pick
up microphones, and several types of construction games
(e.g. wooden and plastic building blocks, square and
circular "playpax" [Creative Playthings], rollway, "magic
stackers," etc.). The two chairs were placed on the same
side of the table facing each other about two feet apart
with the subject's right arm and the therapist's left arm
adjacent to the table. The observation rooms contained
sound recording equipment, recording forms, cumulative stop
watches, event counters, tape recorders set to playback
previously recorded numbers at intervals of ten seconds,
and flashlights for signaling these intervals to the
therapist through the mirrors.
Phase I
Design. The design of the first phase was a 3x2x6
repeated measures design nested within two nonrepeated
conditions. The two groups of subjects (the nonrepeated
14
*>V
conditions) were randomly' assigned to either an individual
therapist or multiple (5) therapist condition. Within
these groups the subjects participated in three tasks
(segments), half of each of which contained imitative
prompts (cues), for six repeated weekly sessions (trials).
Thus the design and levels of the factors of phase I:
segments (3) x cues (2) x trials (6) nested within thera
pist conditions (2).
General Procedure. The six 36-minute, phase I
therapy sessions were each divided into three 12-minute
segments for the three different tasks. Each task segment
was divided in half, each first half containing no modeled
cues for imitation and each second half containing 36 cues
modeled by the therapist at 10-second intervals. Each
individual therapy condition subject met with the same
therapist for each of the three weekly task segments while
each multiple condition subject met with the same therapist
weekly for only the middle segment and with one of four
different therapists for the first and third segments.
See Table 1 for sample stimuli presentation form.
Random order refers to a sequence generated by
consulting the random number tables in Hays (1963). In the
case of random order of presentation of cues, a different
random order was generated for each segment of each
session.
15
TABLE 1.--Sample presentation listing of sd and s delta
cues presented by the therapist during the second half
of each segment of each session
SEGMENT I B (CONVERSING WITH CUES)
1 13 25 sd scratch back behind waist
2 14 26 sd two hands on head
h— <
3 15 27 sd fingertips together
o
K
4 16 28 sd dangle arms
q 5 17 29 s delta cross and uncross knees
r— 4
S
6 18 30 sd finger side of nose
o
o
7 19 31 sd cross arms to shoulders
H
C O
8 20 32 s delta hands on opposite knees
1
9 21 39 s delta knock knuckles on chair arm
o
rH
10 22 34 sd two hands through hair
11 23 35 sd rub hands then elbow
12 24 36 sd pull ear
SEGMENT II B (IMITATING WITH CUES)
1 13 25 sd two hands on head
2 14 26 s delta turn around--no reward
3 15 27 sd fingertips together
4 16 28 sd pull ear
5 17 29 sd rub hands then elbow
s
6 18 30 s delta rotate arms shoulder height--no
reward
w
P i
7 19 31 sd finger on side of nose
8 20 32 sd dangle arms
9 21 33 sd scratch back behind waist
10 22 34 sd cross arms to shoulders
11 23 35 s delta touch floor with one hand--no
reward
12 24 36 s delta two hands through hair
SEGMENT III B (CONSTRUCTING WITH CUES)
1 13 25 sd scratch back behind waist
Q
2 14 26 sd finger on side of nose
o
3 15 27 sd cross arms to shoulders
Ed 4 16 28 sd pull ear
Q 5 17 29 s delta pick up piece with two hands
a
o
6 18 30 sd two hands through hair
o
r_T
7 19 31 s delta tap piece on table
W
C /D 8 20 32 sd dangle arms
1
o
9 21 33 s delta pass piece behind back
r H
10 22 34 sd fingertips together
11 23 35 sd two hands on head
12 24 36 sd rub hands then elbow
16
The First Session. For the first session, each
child was brought to the room and told that he and the
therapist were going to do things and play games and that
he could win prizes. Instructions and explanations were
kept to a minimum. The child was shown a selection of
candy and toys and their point values, with the explanation
that for each eight-ounce paper cup he could fill with
poker chips, he could have one point toward the redemption.
He was told that as they did things together he would learn
how to get the chips and that the chips could be redeemed
at the end of each session or accumulated for more valuable
prizes. Any question by a child about earning the chips
that could not be answered by repeating the above informa
tion was ignored. Remaining sessions were conducted with
out these explanations.
Task Segment 1 (Conversing). The first week's
segment 1 was modified as described above. The typical
segment 1 for each subject consisted of his being greeted
by one therapist at the door of the experimental room,
introduced by name to the therapist, instructed to sit in
one of the two chairs in the room, and, for 12 minutes,
being conversed with about school, vacation, interests,
family, friends, pets, etc. Each verbal reply or comment
by the child was followed by his receiving a poker chip and
social reinforcement (e.g. smile, nod, praise, wink, etc.).
17
During the first six minutes while they talked, the
therapist kept his movement to a minimum: his arms on the
arms of the chair and his feet flat on the floor. During
the second six minutes, the therapist modeled the series of
12 behaviors three times, at 10-second intervals, in random
order. Nine of these 12 behaviors were the identical S
behaviors that were being paired with reinforcement during
task segment 2 and interspersed were the remaining three
gdelta -beha.vi.cn-g that were not being paired with reinforce
ment during any of the segments. During segment 1, if a
c l dsltd.
child performed an S or an S behavior, a hold was
placed on his reinforcement for talking to prevent acci
dental reinforcement of the and s^e^ta behaviors during
this segment.
After each 12-minute segment ended, all children
were told to stay in their seats. The children in the
individual therapist condition were told, as the therapist
left the room, "I'll be back in a second," while the chil
dren in the multiple condition were told, "I've got to go
now, but (therapist by name) will be here in a second."
Segment 1 Dependent Variables. The recording tasks
were divided by four observers for each subject for each
d
session. The first two replicated the recording of S and
gdelta behavior while the second two replicated the
recording of frequency of sentences and latency to speak.
18
Latency was defined as the cumulative number of seconds
between the therapist's termination of a comment and the
child's commencement of one. This total time was divided
by the frequency of the child's comments to create a mean
latency/response measure.
Per cent reliability was determined by dividing the
total number of agreements between each two observers by
the total number of agreements plus disagreements and
multiplying by 100. For the purpose of the analyses, the
mean of the two observations was computed where the
observers disagreed.
Task Segment 2 (Modeling and Imitation’ ) . Appropri
ate to condition, either the same therapist or a different
therapist entered the room for segment 2. The therapist
briefly exclaimed to the child: "Do something." Unless
the same behavior was modeled by the child twice in a row,
the therapist imitated it. If the child modeled one of the
Sd behaviors, the therapist also provided the child a token
and social reinforcement. If the child modeled an S
or an original behavior, the therapist imitated it with no
comment or token.
After six minutes of the child modeling, the
therapist stopped imitating and started modeling the S and
d 61 tci
S behavior in random order at 10-second intervals.
The therapist reinforced with tokens and approval every
19
time the child imitated an behavior and made no comment
* • • f~]pi o
and gave no tokens when the child imitated an S
behavior, did not imitate, or performed an original (i.e.,
non-S^ and non-S^e^ta) response to the modeled cue. At the
end of segment 2, the same therapist left and shortly-
returned (individual condition) or the therapist was
replaced by a new therapist for segment 3 (multiple condi
tion) .
Task Segment 2 Dependent Variables. Two observers
Hpl fp
recorded S and S behavior while the other two
recorded original behavior during each half of segment 2.
Reliability was checked similarly to segment 1.
Task Segment 3 (Cooperative Constructing). Again,
appropriate to condition, either the same therapist or a
different therapist entered the room to conduct each
segment 3. The child was asked to choose one of the
building/constructing games and place it on the table. The
therapist and child took the same seats as in segment 1 and
proceeded to work on the project. If the child took turns
with the therapist (this was to increase the child's atten
tion to the therapist), the child received a token and
approval for each piece he added. If the child did not
look at the therapist, the therapist waited to take his
turn until he did so.
20
During the second half of segment 3, the project
continued while the therapist at 10-second intervals
d do X tci
modeled the nine S and three S behaviors in random
order three times each. If at any time during the segment
d delta
the child performed an S or S , the therapist made no
comment and held the child's token for building until 10
seconds after the imitative behavior.
At the end of this segment, one of the observers
replaced the therapist in the room and praised the child
for the number of tokens he had earned and conducted the
redemption.
Segment 3 Dependent Variables. Two observers again
d delta
recorded S and S behaviors while the remaining two
recorded frequency of building and latency to build. As
segment 1, latency was defined as the cumulative number or
seconds between the therapist placing a piece on the
project. This total was divided by the total number of
additions made by the child to yield a mean latency/
response score.
Generalization Test
Design. The overall design of the generalization
test was a 3x2x2 repeated measures design nested with two
nonrepeated conditions as in phase I. The segments (3) by
cues (2) conditions were now repeated by two levels of the
trials factor. The first level of trials was the actual
21
performance of the subjects during the generalization test
(trial #7). The second level of the trials factor was a
projected score that had been predicted for each subject
based on the beta weights determined from his trials
numbers 4, 5, and 6. These factors were nested within the
conditions of individual and multiple therapist as before.
Procedure. Each of the three segments of the
generalization test (the week 7 session) for all subjects
was conducted by a new therapist. These segments were
conducted under extinction conditions for each of the
children, but were otherwise identical to their previous
sessions. For both groups, each new therapist left the
room at the end of the segment and a new therapist returned
for the next segment. Questions about the lack of chips
were ignored and questions about previous therapists
returning were answered: "I don't know."
Dependent variables were recorded identically to
phase I.
Debriefing
Due to the intervention of Christmas vacation,
three weeks elapsed between the generalization test and the
debriefing. During this time four of the fourteen subjects
stopped coming to the hospital for reasons apparently not
associated with the research.
22
Design. Each of the children of the two groups was
given two trials to recall which of the cues had been
d's
previously associated with reward (the S ) and which had
not (the Sdeltas).
Procedure. The same therapist that had conducted
phase I segment 2 conducted the debriefing. He explained
to the child that he was going to do the same behaviors
that he had done when they copied each other and that he
wanted the child to tell him whether or not he had received
chips for the imitation of that behavior. The child was
told in advance that he would receive five chips for each
correct "Yes, you gave me a chip for copying that one" or
"No, you did not give me a chip for copying that one"
response. The therapist went through the sequence of Sd
and Sdelta behavior in random order twice.
Phase II
Phase II had originally been designed as a replica
of phase I with a reversal of contingencies for S and
original behaviors during segment 2. Segments 1 and 3 of
phase I would have been conducted identically to phase I in
an attempt to determine the generalized effect of the
reversal of contingencies during segment 2 of phase II.
Since by this point, however, it was obvious that there had
been no phase I generalization from segment 2 to segments
1 and 3, the phase II design was modified as follows.
23
Design. The design of phase II was a 2x2x5
repeated measures design, again nested within the multiple
and individual therapist conditions. The segments (2) by
cues (2) conditions were now repeated over trials (5).
General Procedure. The general procedure for phase
II was essentially a replication of segment 2 (modeling and
imitation) twice each session for five sessions. The first
segment was conducted with reward for the child's modeling
non-S^ and non-S^e^ta (i.e., original) behaviors during the
first half and the child's responding to the therapist's
and s^e^ - ta CUes during the second half with nonimitative
responses. The second segment was conducted under the same
procedure but with no reward available for any responses.
All subjects repeatedly saw the same therapist during the
first (reward) segment. Subjects in the individual thera
pist condition also saw the same therapist during the
second (extinction) segment, while subjects in the multiple
condition saw four different therapists during the trials
of the second (extinction) segment.
Reward Segment (Modeling and Imitation). No
instructions relative to the contingencies were given to
the child. As in phase I, unless the child modeled the
same behavior twice in a row, the therapist imitated it.
If the child modeled an original behavior, the therapist
also provided the child a chip and approval. After six
24
c j delta,
minutes, the therapist again modeled S and S behav
iors , but only reinforced the child if he responded with an
original behavior. Lack of responding was not reinforced;
the therapist merely waited ten seconds and modeled another
Sd or sdelta. At the end of the reward segment the thera
pist left the room.
Reward Segment Dependent Variables. Two observers
d dcltf l
recorded S and S behaviors while the remaining two
recorded original behavior. Reliability was checked as
before.
Extinction Segment. Following each weekly reward
segment the child received a similar extinction segment.
For the individual therapist group, the same therapist
returned to the experimental room to conduct this segment,
while for the multiple condition a new therapist replaced
the therapist that had conducted the reward segment. No
instructions were provided; questions relating to contin
gencies were ignored.
Dependent variables were recorded identically to
the reward segment. After five trials on each of the
segments the experiment was terminated.
RESULTS
Since none of the eight therapists ever missed a
session, the problem never arose of how to treat an indi
vidual therapy subject so that he would not have experi
ence with multiple therapists. The contingency plan had
been for the author to spend that session discussing the
project with the subject's parents while the subject played
in the waiting room.
Four shifts of four observers each recorded data--
two replicating each other on different variables. Of the
160 subject trials, 158 were replicated on both variables,
with the remaining two replicated only for imitative
variables due to observer absences. Reliability ranged
from 84 per cent to 96 per cent agreement per trial, with a
mean of 88 per cent agreement between observers.
Phase I
As can be seen from Figure 1, modeling of the
(X trial 6 = 46.21) and g^e^ta (X trial 6 = 10.79) behav-
A _
iors in segment 2A and imitating the S (X trial 6 = 27.0)
and s^e^ta (X trial 6 = 9.0) behaviors in segment 2B have
become high probability behaviors. However, modeling or
imitating these behaviors in segment 1 (Xg ^ = .232;
25
60-
50-
30-
to 40-
a
w in-l
h a lo
ss
u o-
SEGMENT I
(REWARD FOR
CONVERSING)
R— f i g— R=— 8 B B
J L
SEGMENT II
(REWARD S*,fs)
SEGMENT III
(REWARD FOR
CONSTRUCTING)
-B— g— B- -B— f l
to
< y >
28-i
Q
W
1
H
2
H
D
► 5
H
a
u
21-
w
a
0
H
1
a
a
14-1
7-
0-
B B B-=§
_L
MULTIPLE GROUP
D D D: Sc
X X X: S
INDIVIDUAL GROUP
delta
D-D-D:
X-X-X:
.delta
OOO: Original O-O-O: Original
iMt— *
1 2 3 4 5 6 7
TRIALS
1 2 3 4 5 6 7 1 2 3 4 5 6 7
TRIALS TRIALS
FIGURE 1.— Median number of S^, s^e^ta, and original behavior during phase I and the
generalization test (trial 7). Note the almost complete lack of modeling and imitating
during the segments in which reward was presented for an alternative response.
27
fs delta " -042) and seSment 3 (Xs d = -137; fs delta =
.054) has remained very infrequent. Mean latency/verbal
response in segment 1 and mean latency/building response in
segment 3 have decreased from .88 second to .20 second in
segment 1 and 8.02 seconds to 3.01 seconds in segment 3
(see Fig. 2). Frequency of sentences in segment 1
increased from 26.00 to 33.29 sentences and frequency of
pieces added to the project in segment 3 increased from
17.21 to 31.40 pieces/session (see Fig. 3).
Summary Analyses. This visual inspection is
supported by summary analyses of variance comparing the
effects of the individual and multiple therapy conditions
and several planned comparisons:
1) Overall therapist effects were not significant
for either the (F [1,12] = 2.06, p < .177) or s^e^ta
behaviors (F [1,12] = .002, p < .964).
2) Presence or absence of cues in segment 1 and
segment 3 did not differentially effect generalization from
segment 2 for (F [1,12] = .96, p < .346) or s^e^ta
behaviors (F [1,12] = .11 p < .109).
3) Segment 1 did not differ significantly from
segment 3 in respective amounts of generalization from
segment 2 for (F [1,12] = .69, p < .421) or for g^e^ta
behaviors (F [1,12] = .11, p < .738).
12
wlO-i
Q 8-
S
O
u
W 2-
M 0
6 -
4-
SEGMENT I
(CONVERSING)
_L.
SEGMENT III
(CONSTRUCTING)
to
00
12-,
colO-
° 8- |
! 3
O
U
w
C O
6-
4-
2 -
0-
(CONVERSING WITH CUES)
_L _L X
2 3
T R I
4 5 6
A L S
(CONSTRUCTING WITH CUES)
X X X:MULTIPLE GROUP
X-X-X:INDIVIDUAL GROUP
X X
2 3
T R I
4 5 6
A L S
FIGURE 2.— Mean latency of nonimitative responses during phase I and the generalization
test (trial 7). Both groups demonstrate a significant decrease over trials through
all segments. Only the individual group shows a trial 7 increase in segment I A.
co
60n
w
50-
CO
S3
40-
o
30-
&
20-
CO
10-
w
0-
SEGMENT I
(CONVERSING)
SEGMENT III
(CONSTRUCTING)
co 60-
w 50-
m 40-
S
30-
O
o, 20-
w 10-
H 0-
DS
(CONVERSING WITH CUES)
2
T R I
4
A
5 6 7
L S
(CONSTRUCTING WITH
X-X-X:
TRIALS
CUES)
MULTIPLE GROUP
INDIVIDUAL GROUP
FIGURE 3.— Mean frequency of nonimitative responses during phase I and the generalization
test (trial 7). Both groups demonstrate a significant increase over trials through
both segments. Only the individual group shows a trial 7 decrease in segment IA.
30
4) Performance of modeling and imitation was
significantly more frequent in segment 2 than in either
segment 1 or segment 3 for Sd (F [1,12] = 351.5, p < .001)
and for s^e^ta behaviors (F [1,12] = 36.8, p < .001).
Segment 1 Analyses. Latency/verbal response showed
a significantly decreasing linear trend over trials
(F [1,12] = 13.07, p < .004) while frequency of verbal
responses increased linearly over trials (F [1,12] =58.2,
p < .001). Frequency of Sd behaviors did not change from
a trial 1 mean of .14/subject to a trial 6 mean of .01/
subject. Similarly, s^e^ta ranged from .03/subject to .01/
subject.
Segment 2 Analyses. The children’s modeling in
segment 2A showed a significantly increasing linear trend
for Sd (F [1,12] = 59.8, p < .001), sdelta (F [1,12] =
15.7, p < .001), and total behavior (F [1,12] = 26.3, p <
.001), and a significantly decreasing linear trend for
original modeled behavior (F [1,12] = 17.0, p < .001). The
relative frequency of Sd and g^e^ - ta behavior ([Sd] -3*
[sdelta-j^ not significantly over trials
(F [1,12] = 2.32, p < .154).
During the second part of segment 2, when the
therapist modeled, the children imitated an average of 26.8
of the possible 27 S behaviors modeled and an average of
8.9 of the possible 9 g^e^ - ta behaviors modeled. No child
31
d d^ltci
responded with an original behavior to an S or S cue
during this half of segment 2.
Segment 3 Analyses. Latency/building response
showed a significantly decreasing linear trend over trials
(F [1,12] = 23.59, p < .001) while frequency of building
responses increased linearly over trials (F [1,12] = 68.97,
p < .001). Frequency of behaviors did not change from a
trial 1 mean of .29/subject to a trial 6 mean of .04/
subject. Similarly, s^e^ - ta ranged from .04/subject to
.01/subject.
Generalization Test
Trial 6. Each variable was tested for the effects
of number of therapists at trial 6 (the last trial before
the generalization test). No significant differences were
found in any of the cells on any of the variables (see
Table 2).
Predicted vs. Actual Trial 7 Scores. A predicted
trial 7 was computed from the beta weights of trials 4, 5,
and 6 of each subject and was compared to the actual
performance of the children on the generalization test
(trial 7). Thus the predicted scores and the actual scores
for the individual and multiple groups were tested as a 2x2
design with subjects repeated over trials within each
segment.
TABLE 2 Summary of statistics between individual and multiple therapist conditions
on trial 6 (F [1,12]), the last trial before the generalization test
SEGMENT I SEGMENT II SEGMENT III
source Imitation
Modelins Imitation Modelins Imitation Modelins
MS F
P<
MS F
P<
MS F p< MS F
P<
MS F p< MS F p<
Latencv/Resp
SWN .14 .3 4.9 3.1
N .26 1.8 .21 .3 1.2 .30 2.2 .5 .51 .3 .1 .78
Freauencv
SWN 190.6 127.9 62.3 23.6
N 10.3 .1 .82 7.1 .1 .82 7.1 .1 .71 25.8 1.1 .32
sd
SWN 0 412.7 0 0 .1
N 0 0 1.00 0 0 1.00 31.5 .1 .79 0 0 1.00 0 0 1.00 .1 1.0 .34
gdelta
SWN 0 0 109.2 0 0 .1
N 0 0 1.00 0 0 1.00 120.1 1.1 .32 0 0 1.00 0 0 1.00 .1 1.0 .34
Original
SWN 40.4 0
N 16.1 .4 .54 0 0 1.00
Total
SWN 285.3 0
N 85 .0 .3 .60 0 0 1.00
Sd - 3* sdelta
SWN 15 92.9 0
o
N N 1481.1 .9 .35 0 0 1.00
NOTE: There were no significant differences.
33
As can be seen from Figure 2, both the individual
and multiple groups in segment 1 were significantly slower
than their predicted latency of .014 second per response
(F [1,12] = 27.7, p < .001) and performed significantly
less responses than their predicted mean of 61.07 responses
(F [1,12] = 20.5, p < .001) (see Fig. 3). Only in the
first half of the segment was there a differential decre
ment from the predicted scores with the individual group
showing a significantly greater decrement than the
multiple group on latency to respond (F [1,12] =5.18,
p < .042) .
d dsltd.
Comparable tests were run on the S and S data
of segment 1, but due to the fact that the mean trial
frequencies were already so low (Xg ^ = .190, Xg ^ e ^ t a =
.024), the beta weights predicted negative scores for most
of the subjects and the tests were discussed.
Due to the different nature of the tasks being
performed in part A of segment 2 (uncued modeling) and
part B of segment 2 (cued imitation), each part was
analyzed separately (see Fig. 1). Neither nor s^e‘ * ' ta
behaviors changed significantly in frequency during the
generalization test in part A. Original behavior did
increase significantly (F [1,12] = 6.08, p < .03); this
increase was not differential between groups. Total amount
of behavior did show a differential change with the
individual group performing an average of 14.4 less
34
behaviors than predicted while the multiple group performed
an average of 19.0 more behaviors than predicted (F [1,12]
= 10.48, p < .007). The relative proportion of s and
gdeltas wag unchanged#
Segment 2b therapist modeling saw virtually no
change in imitative behavior with an average of 26.86 of
the possible 27 S and 8.64 of the possible 9 g^e^tas.
In segment 3 there were no significant differences
between the groups or between the predicted and actual data
for the latency/response, frequency, S^, or g^e^ - ta scores.
Debriefing
A chi square test on the number of correct answers
to the debriefing questions revealed significantly differ-
2
ent responses from chance performance (x = 17.0, p <
.001); at the same time, the ratio of correct responses
cie X13 .
to correct S responses did not significantly differ
2
from the expected 3:1 ratio (x = 2.25, p > .10). A multi
variate analysis of variance on the total and total
J _ 1 +- o
S correct revealed the individual and multiple groups
were virtually identical on the amount of learning
(F [2,7] = .120, p < .888).
Phase II
Summary analyses for phase II revealed, again, no
significant differences as a function of number of
therapists for performance (F [1,8] = 2.92, p < .126),
35
gdelta performance (p [1,8] = .41, p < .541), original
behavior performance (F [1,8] = .07, p < .798), total
behaviors performed (F [1,8] = 2.37, p < .162), or relative
number of and (jr [1,8] = 1.48, p < .259) (see
Fig. 4). The number of s (F [1,8] = 9.71, p < .038) and
gdeltas = 7.40, p < .026) was significantly
greater in the extinction segment of phase II than the
reward (for original) segment. The number of original
behaviors on this comparison, however, did not differ
significantly (F [1,8] =2.23, p < .173) as did not the
total number of behaviors (F [1,8] = .05, p < .826) or
relative frequency of and s^e^ta (F [1,8] = .02, p <
.905) .
Child1s Modeling. During part A of the reward
segment and part A of the extinction segment original
modeled behaviors had a significantly increasing linear
trend (F [1,8] = 10.04, p < .013) while modeled behav
iors (F [1,8] = 14.20, p < .005) and g^e- * - ta modeled
behaviors (F [1,8] =5.12, p < .047) decreased signifi
cantly (see Fig. 4). Total behavior continued to increase
in the reward segment, but the change at this point was
not linearly significant. Separate analyses of these two
conditions result in similar conclusions.
Child1s Imitation. During part B of the reward
segment and part B of the extinction segment there were no
Q
W
P CO
M Oh
P O
§
P
P
H
a
u
80
60 -
40 -
20 -
0 -
REWARD SEGMENT
(CUES ABSENT)
*---8
EXTINCTION SEGMENT
(CUES ABSENT)
u>
O '
P
w
I
H
s
H
CO
a
o
p a
p w
H C Q
a
u
36 n
27
18 -
9 -
(CUES PRESENT)
X X
8 9 10 11
TRIALS
MULTIPLE GROUP
D D D: Sd
XXX: Sdelta
OOO: Original
INDIVIDUAL GROUP
D-D-D: Sd
X-X-X: Sdelta
O-O-O: Original
12
(CUES PRESENT)
9 10 11 12 8
TRIALS
FIGURE 4.— Median number of Sd, sde^ta, and original behaviors performed during phase II.
Note that when the child modeled (cues absent) and imitated (cues present), he did not
differ significantly from the adult model's ratio of three Sd's for each sdel^a.
37
significant changes between the individual and multiple
groups on Sd (F [1,8] = .55, p < .480), g^e- * - ta (p [1,8] =
.51, p < .496), original (F [1,8] = .41, p < .542), total
(F [1,8] = .35, p < .572). Nor were the changes over
trials significant linearly: Sd (F [1,8] =4.17, p <
.075), sdelta (F [1,8] = 3.54, p < .097), original (F [1,8]
= 4.03, p < .080), total (F [1,8] = 2.11, p < .185), or
sd _ 3* gdelta (F ^g-j = ^ p < <419) _ This lack of
significant changes over trial is consistent with the
results when one considers that while the mean number of
d's
S had decreased over trial from a possible 27 to 21.4
by the last trial, both the median score and the modal
score for this trial was still a perfect 27.
Separate analyses of the child's imitating for
reward condition and child's imitating under extinction
conditions yielded parallel results.
Summary
Latency/response decreased while frequency of
response increased over trials in phase I. The modeled
behaviors increased during segment 2A but this increase was
not evidenced in segment 1A or 3A. The imitative perform
ance started and remained within 98 per cent of possible
during segment 2B and this behavior also did not generalize
to segment IB or segment 3B. The presence or absence of
the same cues of the same therapists as in segment 2 did
38
not differentially affect stimulus generalization.
The two groups were not different in any of the
segments on any of the variables for the last trial before
the generalization test.
Both groups showed a decrement from predicted
performance on the generalization test in segment 1 and
segment 2. There were no significant differences in
segment 3. The individual therapist group showed some
differential decrement in part A of segments 1 and 2, but
this differential was not apparent during part B of the two
segments or in segment 3.
During phase II, while the child modeled, reward
for original behavior resulted in a decrease in both
cIg 11 c l
and S in the reward segment and the following extinc
tion segment. When the child imitated the therapist's
g d0lts.s
modeling of S and S , reward for original responses
did not result in either an increase of original responses
d d0 11 e i
or a decrease in the previously rewarded S or S
responses in either the reward for original segment or the
following extinction segment.
DISCUSSION
It is apparent from the lack of performance of the
Sd and Sdelta behaviors in segment 1 and segment 3 that
stimulus generalization is not an "automatic" phenomenon.
Segments 1 and 3 were temporally contiguous with segment 2,
the identical Sd cues were presented, the same tokens were
present and being dispensed frequently for other behaviors,
and, for half the subjects, the same therapist was conduct
ing segments 1 and 3 as was conducting segment 2. The cues
from change of task and from the therapist leaving the room
between segments were apparently sufficient for almost
perfect discrimination. The author feels that the frequent
reward for conversing (segment 1) and constructing (segment
3) may have precluded the children's motivation to model or
imitate during these tasks.
Since this stimulus generalization failed to occur
in both the uncued first halves of segments 1 and 3 and the
cued second halves of these segments, this study did not
provide data to determine if differential generalization
would occur as a function of the person and/or cues
present.
39
40
Considerable previous research, if anecdotal,
reported that varying the stimuli configurations of a
learning situation would lead to a greater probability of
stimulus generalization to a new situation than if the
learning stimuli were constant. The results of this study
support that hypothesis but suggest that such additional
probability of generalization may be more transitory than
expected. When the new therapist was introduced in part A
of segment 1, the multiple therapist group responded in
significantly less time than the individual therapist
group. Both groups, however, showed a significant decre
ment from their predicted performance. When the same new
therapist conducted part B of segment 1, both groups were
now virtually identical and both were still below
prediction.
When another new therapist was introduced in part A
of segment 2, the most significant difference between the
multiple conditions was evidenced. The multiple group
performed 129 per cent of the predicted total modeling
behavior while the individual group performed 81 per cent.
By the second part of this segment also, the groups
were performing equally. When the third new therapist was
introduced in segment 3, there were neither differences
between the groups nor differences between the actual and
predicted scores. This would indicate that the multiple
group's advantage of having had five therapists over the
41
individual group's one therapist only lasts for a short
time after the new therapist is introduced and may only
last for a small number of new people, the individual group
rapidly taking on the characteristics of a multiple group
with the successive tests on such simple tasks.
During phase I segment 2B when the child imitates,
performance of both the reinforced imitations and the
d. 6 XtcL
unreinforced S imitations is nearly 100 per cent. As
was found previously, interspersing unreinforced imitative
behavior with reinforced imitative behavior results in the
continued performance of the unreinforced behaviors. The
debriefing demonstrated that the subjects could readily
H p 1 i"£i
discriminate (p < .001) the S from the S cues, so
learning was not the critical variable.
The conclusion must therefore follow that while
interspersing unreinforced imitative behavior with
reinforced imitative behavior is sufficient to maintain the
unreinforced imitations (the segment 2B conditions),
interspersing unreinforced imitative behavior (even though
previously reinforced) with reinforced nonimitative
behavior is not sufficient to maintain the imitative
behavior under those conditions (segments 1 and 3). The
hypothesis follows that imitative behaviors, when inter
spersed with each other, form a generalized response class
with the members functionally related to each other but
with the class functionally independent of nonimitative
42
behavior classes. Support for this hypothesis is strong
from the modeling performance of the subjects in segment 2A
of phase I and the child modeling segments of phase II.
When modeling behavior was reinforced in segment 2A, the
frequency of s^e^ta behaviors modeled also increased even
though it was never prompted or reinforced. This rein
forcement for the behavior, however, was not sufficient
to maintain the unreinforced modeled original behavior
which decreased to almost 0. When the conditions were
reversed in phase II, rewarded original child modeled
behavior again increased while both unrewarded and
d<ZcL d’s
S behavior decreased. Rewarding for modeling of S
de It cis
was sufficient to maintain S but not originals;
rewarding of originals was not sufficient to maintain
•4 -u c-d's cdeltas
either S or S
These data also support the previous findings that,
once established, a generalized class of imitative behavior
is durable. After six trials of rewarded imitation in
phase I, the children then received five trials of the
reversed contingency of reward for nonimitation and five
trials of extinction. Under neither condition did the
statistics show a significantly decreasing trend for
imitation. In fact, on the last trial only half the
subjects had decreased at all.
d
Performance rates under extinction for the S and
gdelta behav- i _ o;cs during child's modeling did not
demonstrate the significant resistance to extinction of
the child's imitating segment. Apparently, there exists
no special qualities to the behaviors themselves, but
rather the source of the resistance to extinction resides
in the imitative setting. This is not to say that general
ized imitations cannot acquire powerful secondary reinforc
ing properties--only that reinforcing imitation does not
necessarily result in such acquisition.
The reason for this emphasis is that a seventeenth
subject participated in the procedure and whose behavior
strongly demonstrates such properties. This eight-year-
old, very active girl was not considered as part of the
overall results because she imitated over 80 per cent of
the segment 1 modeled cues on the first trial without
reward. She was allowed, however, to participate in all of
the trials of the experiment to determine what effects the
procedures would have on her already established general
ized imitative response repertoire. Despite the fact that
she could not earn rewards for talking and building in
segments IB and 3B for ten seconds after she imitated, she
continued to imitate from 75 per cent to 95 per cent of the
d de Its,
reinforced S and S cues through the entire first
phase. Needless to say, she did not extinguish during
phase II. Her performance indicates that, for her, imitat
ing had more acquired reinforcing properties than either
the social approval or tokens paired, and presented, by the
44
therapist.
It would appear from this and previous studies that
the most parsimonious explanation of the phenomenon associ
ated with generalized imitation lies in the setting events.
The reason that the imitative data were graphed in terms of
medians rather than means was that during phase II, one
subject in each condition did radically change in perform
ance from imitation to nonimitative responding in the
reward for original behavior segment, and to extinction
during the extinction segment. Their opposite behavior
from the remaining members of their groups gave the group
means the appearance of a change over trials, which was not
the case. It is probably no coincidence that these two
subjects had the highest IQ's of the seventeen subjects in
the experiment.
The probable cause for these findings, in this
author's opinion, is that the children perceive that the
adult model will either be displeased if the child does not
^0 T13 .
imitate the unreinforced S cues, or will model no more
cues if he does not imitate the s^e^tas or both. In
situations where they are not sure of the consequences for
lack of imitation, the imitation becomes an active avoid
ance situation where the aversive consequence has been
removed (or in this case never existed within the experi
mental situation), but the subject escapes by imitating
never to test the consequences. This would explain why the
45
subjects did not imitate during segments 1 and 3 of the
first phase when cues for imitation were being interspersed
with reinforced nonimitative responses. The therapists
were providing such consistent token and social reinforce
ment that, from the child's point of view, they strongly
approved of his behavior. Consequently, imitating to avoid
disapproval was 1) unnecessary because the therapist was
already approving, and 2) unwise because to imitate would
be to take time from performance of the presently
reinforced nonimitative behavior. When one considers that
c t 01 tas
the S of segment 3 especially were of the novel type
(tap the block on the table before adding it to the
project, pass it behind your back, etc.) that considerable
previous work has shown children will readily imitate, then
one must assume that the present situation was somehow, if
inadvertently, creating the demand characteristic: don't
imitate when you talk or build. Only the two especially
bright subjects discriminated the new contingencies when
the therapist modeled during phase II, or were willing to
test the consequences of not imitating, or both.
This perspective of the children's behavior as
resulting from an active avoidance of perceived therapist
displeasure is not offered as an alternative to the
previous generalized imitation hypotheses but rather as a
more precise specification of Steinman's (1970) setting
events explanation. As such, it finds support in the
46
results, demonstrating response decrement when the model
leaves the situation or turns away from the child
(Steinman, 1970) as well as the many previously mentioned
studies demonstrating that unreinforced imitative responses
are maintained and difficult to extinguish.
BIBLIOGRAPHY
47
BIBLIOGRAPHY
Ayllon, T., & Azrin, N. The Token Economy. New York:
Appleton-Century-Crofts, 1968.
Baer, D. M., Peterson, R. F., & Sherman, J. A. The
development of imitation by reinforcing behavioral
similarity to a model. Journal of the Experimental
Analysis of Behavior. 1967, 10, 405-416.
Baer, D. M., & Sherman, J. A. Reinforcement control of
generalized imitation in young children. Journal of
Experimental Child Psychology. 1964, 1 . , 37-49.
Baer, D., & Wolf, M. The Entry into Natural Communities
of Reinforcement. Paper presented to the American
Psychological Association, Washington, D.C., 1967.
Bandura, A. Social learning theory of identificatory
processes. In D. A. Goslin (Ed.), Handbook of Social
ization Theory and Research. Chicago: Rand McNally,
1968. Ch. 3.
Bandura, A. Modeling Theory. Symposium paper presented
to the Society for Research in child development.
Santa Monica, Ca., 1969. (a)
Bandura, A. Principles of Behavior Modification.
New York: Holt, Rinehart & Winston, 1969. (b)
Bandura, A. Social-learning theory of identificatory
processes. In D. A. Goslin and D. 0. Glass (Eds.),
Handbook of Socialization Theory and Research.
Chicago: Rand McNally, 1969. Pp. 213-323. (c)
Brigham, T. A., & Sherman, J. A. An experimental analysis
of verbal imitation in preschool children. Journal of
Applied Behavior Analysis. 1968, 1, 151-160.
Burgess, R., Burgess, J., & Esveldt, K. An analysis of
generalized imitation. Journal of Applied Behavior
Analysis. 1970, 3_, 39-46.
Catania, C. Contemporary Research in Operant Behavior.
Glenview, 111.: Scott, Foresman, 1968.
48
49
Flanders, J. A review of research on imitative behavior.
Psychological Bulletin. 1968, 69., 316-337.
Gerwitz, J. Mechanisms of social learning. In D. A.
Goslin (Ed.), Handbook of Socialization Theory and
Research. Chicago: Rand McNally, 1969.
Gerwitz, J., & Stingle, K. Learning of generalized imita
tion as the basis for identification. Psychological
Review. 1968, 75(5), 374-397.
Hays, W. Statistics. New York: Holt, Rinehart & Winston,
1963.
Kale, R., Kaye, J., Whelan, P., & Hopkins, B. The effects
of reinforcement on the modification, maintenance, and
generalization of social responses of mental patients.
Journal of Applied Behavior Analysis. 1968, 1, 307-314.
Kanfer, F., & Phillips, J. Learning Foundations of
Behavior Therapy. New York: Wiley, 1971.
Lovaas, 0. I., Berberich, J. P., Perloff, B. F., &
Schoeffer, B. Acquisition of imitative speech by
schizophrenic children. Science. 1966, 151, 705-707.
Lovaas, 0. I., Frietag, K., Nelson, K., & Whalen, C. The
establishment of imitation and its use for the develop
ment of complex behavior in schizophrenic children.
Behavior Research and Therapy, 1967, 5., 171-182.
Lovitt, T., & Curtiss, K. Effects of manipulating an
antecedent event on mathematics response rate.
Journal of Applied Behavior Analysis. 1968, JL, 329-333.
Metz, J. R. Conditioning generalized imitation in autistic
children. Journal of Experimental Child Psychology,
1965, 2, 389-399.
Patterson, G. An application of conditioning techniques to
the control of a hyperactive child. In L. Ullman and
L. Krasner (Eds.), Case Studies in Behavior Modifica
tion. New York: Holt, Rinehart & Winston, 1965.
Pp. 370-375.
Peterson, R. Building Generalized Response Systems. Paper
presented to the American Psychological Association,
Washington, D.C., 1967.
50
Peterson, R. F. Some experiments on the organization of
a class of imitative behaviors. Journal of Applied
Behavior Analysis. 1968, JL, 225-235.
Peterson, R. F., & Whitehurst, G. J. A variable influenc
ing the performance of non-reinforced imitative
behaviors. Journal of Applied Behavior Analysis, 1971,
4, 1-9.
Rilling, M., Askew, H., Ahlskog, J., & Kramer, T. Journal
of the Experimental Analysis of Behavior, 1969, 12,
917-932.
Shumaker, J., & Sherman, J. A. Training generative verb
usage by imitation and reinforcement procedures.
Journal of Applied Behavior Analysis, 1970, 3, 273-287.
Steinman, W. In R. F. Peterson and G. J. Whitehurst,
A variable influencing the performance of non
reinforced imitative behaviors. Journal of Applied
Behavior Analysis. 1971, 4, 1-9.
Terrace, H. S. Stimulus control. In W. Honig (Ed.),
Operant Behavior: Areas of Research and Application.
New York: Appleton-Century-Crofts, 1966. Ch. 7.
Wahler, R. G. Journal of Applied Behavior Analysis. 1969,
2, 239-246.
Wheeler, A., & Sulzer, B. Operant training and generaliza
tion of a verbal response form in a speech-deficient
child. Journal of Applied Behavior Analysis. 1970, 3,,
139-147.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A Cognitive Dissonance Analysis Of Conformity Behavior, As Applied To The full Denture Patient

PDF

Experimenter Expectancy Effect Examined As A Function Of Task Ambiguity And Internal Versus External Control Of Reinforcement

PDF

Effects Of Group Behavior Therapy Imagery On Basketball Performance

PDF

Weight Reduction As A Function Of The Timing Of Reinforcement In A Covertaversive Conditioning Paradigm

PDF

The Effect Of Anxiety And Frustration On Muscular Tension Related To The Temporomandibular-Joint Syndrome

PDF

Mental Imagery As A Function Of Muscular Tension And Suggestion

PDF

Non-Specific Treatment Factors And Deconditioning In Fear Reduction

PDF

Tension And Anxiety In Deconditioning

PDF

Stress Response Modification And Autohypnotic Techniques

PDF

The Effect Of Conditions Of Risk, Internal Versus External Control Of Reinforcement, And Sex On Binary Choice Probability Learning

PDF

Imagery And Response Styles In Desensitization

PDF

The Differential Effectiveness Of External Versus Self-Reinforcement On The Acquisition And Performance Of Assertive Responses

PDF

Strategies Of Marital Communication

PDF

A Behavioral Assessment Of The Reinforcement Contingencies Associated With The Occurrence Of Suicidal Behaviors

PDF

A Temporal Approach-Avoidance Conflict In An Academic Test Situation

PDF

Classical Discrimination Conditioning As A Function Of Probability Of Reinforcement

PDF

Delinquency As A Function Of Intrafamily Relationships

PDF

Individual Differences In Imagery Among Sensory Modalities

PDF

Therapeutic Movement As A Function Of Awareness Of Goals

PDF

Prognostic Expectancy Effects In The Desensitization Of Anxiety Over Invasion Of Body Buffer Zones

Asset Metadata

Creator Greenberg, Stuart Allen (author)

Core Title Stimulus and response generalization of classes of imitative and nonimitative behavior as a function of reinforcement, task, cues, and number of therapists

Degree Doctor of Philosophy

Degree Program Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,psychology, clinical

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Wolpin, Milton (committee chair), Rutherford, Robert B., Jr. (committee member), Tiber, Norman (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-771297

Unique identifier UC11364376

Identifier 7300738.pdf (filename),usctheses-c18-771297 (legacy record id)

Legacy Identifier 7300738

Dmrecord 771297

Document Type Dissertation

Rights Greenberg, Stuart Allen

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA