Rogers-Warren; Baer
JOURNAL OF APPLIED BEHAVIOR ANALYSIS CORRESPONDENCE BETWEEN SAYING AND DOING: TEACHING CHILDREN TO SHARE AND PRAISE' ANN ROGERS-WARREN AND DONALD M. BAER BUREAU OF CHILD RESEARCH UNIVERSITY OF KANSAS
Five small groups of preschool children were taught to share and praise by the modelling of these behaviors and reinforcement of their reports of sharing and praising. Experiment I demonstrated that modelling and reinforcement of any (true or untrue) reports of sharing, and then of praising, promptly increased reports of the corresponding behaviors.
Modelling and reinforcement for true reports of each behavior increased both reporting and actual behavior. Experiment II showed that both reported and actual sharing and praising may be increased by modelling and reinforcement for true reports of the target behavior, without previous reinforcement for any (true or untrue) reports of those behaviors. Sharing, but not praising, generalized to a second setting. Experiment III replicated the results of Experiment II for sharing and praising, and demonstrated similar success in increasing a third behavior, specific praising. In general, these experiments show that developing correspondence between children's reports of behavior and actual behavior may be an efficient means of increasing prosocial responses.
DESCRIPTORS: social behavior, sharing, modelling, praise, reporting, generalization, multiple baseline, reinforcement delay, children In a very broad sense, verbal behavior is similar to currency. At any given time, there may not be sufficient gold to back up every dollar
exchange, but sellers and buyers usually act as if there were. In interactions among adults, there is a similar continual exchange of verbal reports, generally as if there is, or will be, some other behavior to correspond to the verbal reports, make them true, and give them value.
While operant research has not yet attempted to define the parameters of truthtelling among adults, a number of researchers (e.g., Bem, 1967; Birch, 1966; Lovaas, 1961; Lovaas, 1964b; Risley and Hart, 1968; Sherman, 1964) have investigated the relationship between young
children's verbal behavior and their corresponding other behaviors.
'This research was funded in part by grant MH 11739 from the National Institute of Mental Health for training new personnel for behavior modification and by grant 3336-5038 from the University of Kansas to Dr. Donald M. Baer. Reprints may be obtained from Ann Rogers-Warren, Department of Human Development, University of Kansas, Lawrence, Kansas 66045.
Investigations in the area of verbal self-control of behavior have generally taken one of two forms. In the first form (e.g., Bem, 1967; Birch, 1966; Lovaas, 1964a; Luria, 1961; Meichenbaum and Goodman, 1969), subjects have been taught to make a relevant verbal response while performing a motor response. Such studies have shown that certain verbal responses facilitate the corresponding motor response. For example, Lovaas (1964a) demonstrated that subjects who said the word "faster" as they pushed a lever, would depress the lever more quickly than subjects who said the word "slower". Bem (1967) reported that, in a task requiring children to press a lever to indicate the number of stimuli that were presented briefly and then withdrawn, those who counted as they pressed the bar responded more accurately than those who did not.
Within the same paradigm, several authors (e.g., Meichenbaum and Goodman, 1971; Monohan and O'Leary, 1971; O'Leary, 1968) have demonstrated that training a child to selfinstruct may facilitate appropriate responding 335
1976, 93, 335-354 NUMBER 3 (FALL) 1976 ANN ROGERS-WARREN and DONALD M. BAER and impede inappropriate responding. For example, O'Leary (1968) found that when subjects were told to respond only on certain trials, in an experiment where responding on any trial resulted in reinforcement, those subjects who gave self-instructions indicating whether they could respond on a particular trial, "cheated" less than those who did not self-instruct.
Risley and Hart (1968) provided a strong demonstration of the control of behavior through reinforcement of related verbal behavior. Two groups of preschool children were initially rewarded for any (true or untrue) report of the use of a specific preschool material.
Reinforcement was then made contingent on only true reports, that is, those reports that corresponded to the actual use of the material. While reinforcement for corresponding reports successfully increased the subjects' use of materials, after the noncorrespondence- correspondence sequence of contingent reinforcement had been repeated a number of times, reinforcement for any (corresponding or noncorresponding) report was sufficient to increase significantly the use of a specific material. To that extent, Risley
and Hart were able to demonstrate that "saying" could be brought to control "doing".
Israel and O'Leary (1973) further examined the role of verbal behavior in regulating nonverbal behavior. In two comparisons of sayingthen- doing, they found that having preschool subjects say which toy they were going to play with during the next play period produced
greater correspondence between saying and doing, than having subjects report which toy they had played with after the play period. Only slight increases in correspondence occurred when subjects were rewarded for the desired (but not necessarily true) content of their reports. Unlike Risley and Hart (1968), Israel and O'Leary did not find that saying came to control doing for children in the doing-then-saying sequence when only content was reinforced, even after a series of reinforcement-for-content (noncorrespondence) followed by reinforcement-for-correspondence.
Further, Israel and O'Leary did not, as had Risley and Hart, obtain high levels of correspondence for the doing-then-saying groups, even when correspondence was reinforced.
Direct comparison between the two studies is difficult because correspondence was defined by Risley and Hart as "doing and saying", while Israel and O'Leary defined it as "either doing and saying or not doing and not saying". The direct measures of positive correspondence (percentage of subjects saying and percentage of subjects doing) are not reported by Israel and O'Leary. Whatever differences there may be in immediate efficiency of the saying-then-doing and doing-then-saying sequences, both are based on the same procedural approach. If the two sequences are abstracted, and viewed as they operate over a number of sessions, it is apparent that in both cases the subjects' verbal behavior is intended to affect the next opportunity to perform the corresponding behavior. The primary difference
is that in the first case (saying-thendoing) that opportunity arises within a few minutes, and in the second (doing-then-saying), the opportunity does not arise until approximately 23 hr later.
If range of control, that is, the length of time during which verbal statements might affect the corresponding behavior, is of concern, then the doing-then-saying sequence might be more useful.
Such a sequence might contribute to the generalization of the corresponding behaviors to other opportunities not specifically included in training, but occurring within the temporal range of the training. For example, a child who plays with blocks during the designated play
period, might also play with them at home during the evening, thus establishing correspondence between his/her verbal report and actual behavior on more than one occasion.
Both the Risley and Hart, and Israel and O'Leary studies focused on establishing correspondence, using choice of play materials as the behavior to be increased. The present study was concerned primarily with producing many instances of particular prosocial behaviors by
establishing correspondence between the chil- 336 TEACHING CHILDREN TO SHARE AND PRAISE dren's verbal reports and their related prosocial behavior. Since the target behaviors, sharing and praising, should be useful in a number of settings, the apparently greater range of the doing-then-saying sequence made it the preferred procedure. The purpose, then, was to develop an effective procedure for increasing complex social behaviors based on the previously demonstrated techniques for developing correspondence between children's saying and doing.
GENERAL METHODS
Since the same procedures were used in all three experiments, the general methods are described first.
Subjects
All subjects were children attending the Edna A. Hill Child Development Preschool Laboratory at the University of Kansas. Thirty-two children were involved in the three experiments, 19 males and 13 females ranging in age from 3 yr, two months, to 5 yr, six months.
Setting
Experimental sessions were conducted daily, Monday through Thursday. Each session lasted about 15 min. The experimental sessions were divided into two parts, the work period (10 min) and the reporting period (about 5 min). The subjects were seated on the floor, around a
large piece of paper, and had access to art materials (paper, pens, pencils, crayons, markers) during the work period. The experimenter changed the seating arrangement each day by placing cards with the subjects' names on them around the paper and requesting subjects to sit behind their own name card. At the end of the work period, the materials were collected and an activity called the Reporting Period began. Duringthe Reporting Period, the experimenterasked the children individually what they had done "while we were working today". Every
subject was given one opportunity to respond.
The session ended when each child had been queried and the experimenter indicated it was time to go to another activity or return to the classroom.
Behavioral Definitions
The behaviors recorded for the subjects were sharing, praising, and reports of sharing and praising. Experimenter prompts and reinforcement of reports were also recorded. Sharing. Two classes of sharing, verbal sharing and nonverbal sharing, were recorded. Verbal
sharing included any verbalization by a subject to a peer in which the subject: (1) invited a peer to join in a particular activity, or (2) verbally accepted the invitation of a peer to join in a particular activity, or (3) offered to share materials with a peer, or (4) verbally accepted a
peer's offer to share materials, or (5) offered to trade materials with a peer. Nonverbal sharing was recorded when one subject passed or handed a material to a second subject, if both subjects had touched the material within 5 sec, or when two or more subjects simultaneously used the same material (e.g., colored on the same sheet of paper at the same time).
Praising. Praising was defined as any verbalization by a subject to a peer in which the subject indicated approval, liking, or admiration for the peer or any aspect of the peer's art work. There was essentially four forms of praise: (1) "I like and a general object (e.g., "I like your
picture"); (2) "I like . . ." and a specific object or quality (e.g., "I like the way you used the color blue"); and (3) direct general praise (e.g., "neat picture" or "pretty"); and (4) direct specific praise (e.g., "Nice dogs" or "That's a pretty house you made"). In Experiments I and II, all four forms were recorded under the general category of "Praise"; in Experiment III, Form 1 ("I like . . ." and a general object) was scored as "General Praise", and Forms 2, 3, and 4 were scored as "Specific Praise".
Reporting. Subjects were recorded as making a report of sharing or praising whenever they responded to the experimenter's question, "What did you do while we were working today?", with 337 ANN ROGERS-WARREN and DONALD M. BAER a statement indicating that they had shared materials or had praised a peer (e.g., "I shared my pen with Susie", or "I told Jon I liked his picture").
In Experiment III, reports of praise were scored as either "General" or "Specific", using the same criteria as employed for general and specific praising behaviors. Experimenter-behaviors. Two experimenterbehaviors, reinforcement of reports and prompts for reports, were also recorded. Reinforcement was scored whenever the experimenter indicated to a subject that she approved of the behavior the subject reported and offered a bit of food (M&M's or fruit) immediately following the report.
Prompts were scored each time the experimenter corrected a child's verbal response, or modelled a response and instructed a child to repeat it during the reporting period.
Recording Procedures
Two observers were seated within viewing and hearing distance of the group of subjects, one observer at each of the narrow ends of the paper on which the subjects were working. Each observer recorded the behaviors of the three or four children closest to her, depending on the size of the group. Continuous observations were made of the children's behaviors during the work period; discrete observations of each subject's verbal report and the experimenter's behaviors were made during the reporting period.
Observers recorded instances of sharing and/or praising on the current subject's form by writing a symbol identifying the peer to whom the behavior was directed, in the appropriate column identifying the type of material shared or the type of praising used. Experimenter reinforcement and prompts were recorded each time they occurred. Reports of sharing and praising were scored only once for each subject, immediately following the experimenter's questioning of that child.
Reliability was assessed by having a third observer record independently, but simultaneously, with each of the primary observers. Interobserver agreement was determined by comparing the number of behaviors recorded by each observer for a given subject in a particular category of behavior. For example, observer A's record of pen sharing by S1 with S2 was compared with observer B's record of pen sharing by SI with S2. Similar comparisons were made for all materials, for praising, for reports of behaviors, and for experimenter behaviors directed to each subject. Each reliability was then calculated using the formula:
Reliability = number of agreements X 100. number of agreements + number of disagreements Experimental Conditions Baseline. During the initial baseline conditions there were no scheduled consequences for sharing, praising, or reports of sharing or praising.
The experimenter and an adult model sat at a small table behind, but in sight of, the group and used the same art materials as were available to the subjects. Neither sharing nor praising was modelled. Throughout the work period, the experimenter intermittently praised the subjects for "working hard" and attending to the materials. If a subject shared or praised, the experimenter waited at least 10 sec before attending to any subject. During the reporting period, each child was questioned about his/her activity during the previous work period, but no reinforcement for appropriate responses was given.
Modelling only. To determine if modelling alone would result in the desired behavior, the experimenter and the model demonstrated the appropriate response five or six times during the work period. No model of appropriate reporting ehavior was provided, and no reinforcement
for reports of sharing or praising was available. Package I: modelling and reinforcement for any report of the behavior. The first part of thepackage manipulation consisted of the experimenter
and the model demonstrating the appropriate
response, either sharing or praising, as had
been done in the Modelling Only condition. The
topography of sharing and praising behaviors
was systematically varied. In Experiments I and
338
TEACHING CHILDREN TO SHARE AND PRAISE
II, both general and specific praising responses
were modelled during the praising condition. In
Experiment III, only general praise was modelled
during the manipulation of general praising,
and only specific praise was modelled during the
specific praise condition.
At the beginning of the reporting period, the
experimenter asked the model what he had done
while he was working. When the model reported
the appropriate behavior, the experimenter
rewarded him with positive comments
and food. The subjects were then asked individually
what they had done during the work
period. Any report, true or untrue, of sharing
or praising, depending on the experimental condition,
was followed by positive comments, and
either M&M's or fruit, from the experimenter.
When Package I was introduced to any group,
it was sometimes necessary for the experimenter
to prompt two or three subjects to make an appropriate
report (e.g., "Tell me that you
shared"). After at most three sessions, all
prompting was discontinued. Subjects were not
recorded as reporting appropriately if the report
was prompted.
Package II: modelling and reinforcement for
true reports of behavior. The second package
was procedurally similar to the first, except that
only true reports of sharing or praising were
reinforced. As the subject responded to the experimenter's
query, the observer who had recorded
that subject's behavior discreetly signalled
"yes" or "no" to indicate if the subject was reporting
truly. True responses were reinforced.
When a subject reported untruly, the experimenter
responded, "But you didn't really, did
you?" No response was made following an inappropriate
report (e.g., a report of sharing in a
condition when reports of praise were being reinforced);
the experimenter paused 5 sec, then
asked the next subject about that subject's activities.
Package III: modelling and reinforcement for
either true reports of sharing or true reports of
praising. During Package III, the experimenter
and the model demonstrated both sharing and
praising, and the model reported both sharing
and praising. Subjects were rewarded for true
reports of either behavior. That is, if they either
shared or praised, and gave a corresponding report,
they were rewarded. All other procedures
were identical to Package II.
Package IV: modelling and reinforcement for
true reports of both sharing and praising. Package
IV consisted of modelling both sharing and
praising responses and delivering reinforcement
contingent on true reports of both sharing and
praising. Except that each subject was required
to do and report both behaviors, the procedures
were identical to Package II.
Reliability
Reliability figures for all subject and experimenter
behaviors are shown in Table 1.
EXPERIMENT I
METHOD
Subjects
The 14 subjects selected for this study were
divided into two groups. Gr6up A was made
up of six children: one 4-yr-old girl, two 3-yrold
girls, and three 4-yr-old boys. Group B consisted
of eight children: two 4-yr-old girls, five
4-yr-old boys, and one 3-yr-old boy.
Setting
The study was conducted in a regular preschool
classroom, in an area away from other
classroom activities. While an experimental session
was being conducted with one group,
members of the other group participated in conceptual
activities in different parts of the same
classroom, and generally did not attend to the
experimental activity.
Experimental Design
Experiment I consisted of two parts. In the
first part, reinforcement of any report of behavior
(Package I), followed by reinforcement
of only true reports of behavior (Package II),
was systematically introduced in a multiple-
339
ANN ROGERS-WARREN and DONALD M. BAER
Table 1
Range and Mean Reliability for Subject and Experimenter Behaviors in Experiments I, II, and III
Group A Group B Group C Group D Group E
Behavior Range X Range X Range X Range X Range X
Sharing 88-100 92 75-100 89 88-100 91 62-100 90 69-100 86
General praising* 85-100 91 88-100 94 83-100 93 66-100 91 79-100 92
Specific praising -- - 80-100 92 86-100 88
Reports of sharing 75-100 94 - 100 100 - 100 75-100 92
Reports of
praising* 88-100 91 - 100 100 75-100 100 100
Reports of
specific praising - - - - 0-100 90 - 100
Experimenter
reinforcement - 100 - 100 - 100 - 100 100
Experimenter
prompts 100 - 100 - 100 - 100 100
* In Experiments I and II, all praising and reports of praising were recorded in the category of General
Praising, hence reliability scores are reported only for the category of General Praising.
baseline design across sharing and praising by
Group A. In the second part, these procedures
were replicated with Group B, and then two
additional conditions, reinforcement for true reports
of either behavior (Package III), followed
by reinforcement of true reports of both behaviors
(Package IV), were examined.
The sequence of experimental conditions for
this experiment is summarized in Table 2.
All experimental conditions were identical to
those described in the General Methods section.
RESULTS
Group A
The results for Group A are shown in Figure
1.
The introduction of reinforcement of any
report of behavior (Package I) resulted in a
gradual increase in mean number of shares,
double-to-triple the baseline rate, and the percentage
of the group reporting sharing steadily
increased. The percentage of group members
sharing was variable, ranging from 0% to
100%, but averaged higher than during baseline.
Under the same conditions (Package I),
mean praising increased only slightly, although
the percentage of the group praising and reporting
praising increased markedly.
When reinforcement was provided for true
reports of sharing (Package II), the mean number
of shares increased to a consistent level of
around four shares per child per session. Percentage
of the group sharing was moderately
high, while reports of sharing steadily rose to
a high of 90%.
Reinforcement for true reports of praise
(Package II) resulted in sharp upward trends in
both reporting and actual praising. A more
gradual increase was seen in the mean number of
praises per session.
Group B
Similar results for Group B are shown in
Figure 2.
For Group B, which had exhibited a moderate
level of sharing during baseline, reinforcement
for any report (Package I) produced no significant
increase in either the percentage of the
group displaying sharing, or the mean number
of shares per subject. The same procedure applied
to praising resulted in a slight increase in
the mean number of praises and produced
moderate increases in the percentages of the
group reporting and praising.
Reinforcement for true reports of behavior
(Package II) further increased the mean number
of shares and praises. Additional increases in
340
TEACHING CHILDREN TO SHARE AND PRAISE
Table 2
The Sequence of Experimental Conditions in Experiment I
Group A Group B
Sharing Praising Sharing Praising
Baseline
Package I
Model sharing
Model reports of sharing
Baseline Baseline
Reinforce any reports
of sharing
Package I
Model praising
Model reports of prais-
Baseline ing Model sharing Baseline
Reinforce any reports
of praising
Package II Package I
Model sharing Model sharing
Model reports of shar- Model reports of sharing
Baseline ing
Reinforce true reports Reinforce any report
of sharing of sharing
Package II Package I
Model praising Model praising
Model reports of Model reports of prais-
Baseline praising Baseline ing
Reinforce true reports Reinforce any reports
of praising of praising
Package 111 Package II
Model both sharing and praising Model sharing
Model reports of both Model reports of shar-
Reinforce true reports of either ing Baseline
Reinforce true reports
of sharing
Package IV Package II
Model sharing and praising Model praising
Model reports of both Model reports of prais-
Reinforce true reports of both ing
Reinforce true reports
of praising
Package III
Model both sharing and praising
Model reports of both
Reinforce true reports of both
Package IV
Model both sharing and praising
Model reports of both
Reinforce true reports of both :
341
ANN ROGERS-WARREN and DONALD M. BAER
GROUP A
SHARING
mrIb
-u
Dm
Cm
C)
SESSIONS (2day blocks)
Fig. 1. The percentage of Group A members reporting sharing, actually sharing, and the mean number of
shares per subject are shown on the upper graph. Percentages of reporting praising, actual praising, and mean
number of praises per subject are shown on the lower graph. Note that the axis on the left refers to percentage
of group doing and reporting the target behavior; the right-hand axis refers to the mean number of
target behaviors.
percentages of children doing and reporting
the behaviors were also obtained.
The final two conditions were reinforcement
for true reports of either praising or sharing
(Package III), followed by reinforcement for
true reports of both praising and sharing (Package
IV). Actual praising (mean number of percentage
of group praising) and reports of
praising were maintained at consistently high
levels during both of these conditions. Sharing
remained near baseline levels during the "either"
condition, but rapidly increased to its highest
levels when reinforcement was made contingent
on truly reporting both behaviors.
DISCUSSION
Generally similar results were obtained with
both groups. Reinforcement for any report of
behavior (true or false) produced moderate increases
in reporting and small increases in actual
behavior. Reinforcement for true reports was
generally more effective, and resulted in higher
rates of both reporting and actual behavior.
In both groups, sharing increased more
rapidly and was maintained at consistently
higher levels than praising. These differences
may have been a function of the differences
represented in the baseline measures of behavior.
342
TEACHING CHILDREN TO SHARE AND PRAISE
SHARING
GROUP B
*~~~~~~~~~~~.~~@.~~*~~@@@
PRAISING.
%REPORTING PKG: BL. PKGI
% DOING
80 & SHARES/SUBPRAISES
JECT
PKGI MODEL+S,
69 ANY REPORTS
xm
z
ffl
Cen
m
0
-4
SESSIONS (2dayblocks)
Fig. 2. The percentage of Group B members reporting sharing, actually sharing, and the mean number of
shares per subject are shown on the upper graph. Percentages of reporting praising, actual praising, and mean
number of praises per subject are shown on the lower graph. The axis on the left refers to percentage of
group doing and reporting the target behavior; the right-hand axis refers to the mean number of target behaviors.
Many subjects already had sharing in their
repertoires; thus, the experimental manipulations
needed only to increase it. Praising, however,
was not common to most subjects, and it
was necessary first to teach the subjects how to
praise (one goal of the modelling component),
and then to increase praising.
Although high rates of praising eventually
were obtained in both groups, many of the
praise statements seemed mechanical and lacking
in reinforcement value. Praising was nearly
always in the form, "I like your picture" or "I
like what you are making", with little voice
inflection or modulation. Often, subjects did not
look up from their own activities as they delivered
praise statements. As rates of praising
increased, more inappropriate praise was noted.
Sharing, however, was usually appropriate:
subjects used materials first, then offered them to
peers. Trading of materials was the most frequent
form of sharing, but on occasion, children
would simultaneously use materials (typically
drawing together on the same piece of paper).
Most exchanges of material were arranged
verbally.
The functions of the final two contingencies
(Package III and Package IV) for Group B are
clear enough in showing their effectiveness at
increasing many aspects of the target behavior.
In one respect, however, only a tentative conclusion
is possible: the praise-or-share conditions
may have functioned as merely an extension of
the previous reinforcement for true reports of
praise (Package II). The similar levels of prais-
CD
0
cc
a
LL
0
w
z
w
w
a.
343
ANN ROGERS-WARREN and DONALD M. BAER
ing may have been maintained because the
subjects did not discriminate the subtle changes
in criteria for reinforcement, which allowed the
praising behavior to maintain an unchanged rate
of reinforcement. This conclusion is also suggested
by the fact that sharing failed to increase,
although reports of both behaviors were modelled.
Apparently, it was not obvious to the
subjects that they could report either sharing or
praising if they had performed the corresponding
behavior.
Since these conditions (Package III and Package
IV) were introduced after subjects had a long
history of reinforcement for reporting sharing
and praising, the resulting high levels of behavior
possibly represent the cumulative effect
of numerous conditions. However, a complete
analysis of order effects would be necessary for
a final answer. Programmatically, that analysis
might not be important.
EXPERIMENT II
The second study was designed to determine
if preschool children could be taught sharing
and praising skills through the contingent reinforcement
of true reports of those behaviors
without first being rewarded for giving any (true
or untrue) reports, and to determine the extent,
if any, that training to share and praise would
generalize to a second setting, where no reports
of the behaviors were made.
METHOD
Subjects
Six children, two 3-yr-old boys, one 3-yr-old
girl, two 4-yr-old girls and one 4-yr-old boy,
served. Three children were from one classroom
of the preschool, and three children were from
a second classroom. The subjects from different
classes did not know each other before the experiment.
Setting
Experimental (training) sessions were conducted
in a small classroom near the subjects'
regular classrooms. There were no children,
other than the subjects, present during the training
session.
Generalization (probe) sessions were conducted
in the subjects' original classrooms during
a free-play period or a small-group activity
period. Probes occurred about an hour before the
training session and lasted 10 min. During a
generalization probe, a second experimenter
served as a teacher. The children were seated
on the floor in an area adjacent to the general
classroom activities, and had free access to a
supply of materials. The materials were varied
each session, and included tinkertoys, lego, chalk,
playdough, and blocks.
At the beginning of each probe session, the
experimenter told the subjects that children who
stayed in the area until the timer rang would
receive an animal sticker or a star for their cards.
Each child had a 12.5 by 17.5 cm card on which
the stickers were pasted. At the end of the week,
the subjects took the cards home.
Ten to 12 times during each probe session, the
experimenter praised the subjects individually
or as a group for "playing nicely", and for staying
in the area. No reinforcement for sharing or
praising was given. If a subject shared or praised,
the experimenter waited 10 sec before attending
to that child.
Any nonsubjects who came near the area
where the probe session was being conducted,
were invited to join the group and were also
praised for "playing nicely", but did not receive
stickers at the end of the session. There were no
opportunities for reporting during probe sessions.
Experimental Design
This study was conducted as a multiple baseline
across the two behaviors, sharing and praising.
Generalization data were collected during
all phases.
All experimental conditions were identical to
those previously described in the General
Methods section. The sequence of experimental
conditions is summarized in Table 3.
344
TEACHING CHILDREN TO SHARE AND PRAISE
GROUP C
SHARING
Baseline Package IE Baseline
a. 0
cc C!,
LL
0
w
z
w
CD
w
a.
20
01
*-D..00 00 00 0 eSeOe @ O S O ...
Baseline
m
z
m
I
-Q
0
m
Cl)
C
Cm0
-i
SESSIONS
Fig. 3. The percentage of Group C members reporting sharing, and the mean number of shares per subject
are shown on the upper graph. Percentages of reporting praising, actual praising, and mean number of praises
per subject are shown on the lower graph. The axis on the left refers to percentage of group doing and reporting
the target behavior; the right-hand axis refers to the mean number of target behaviors.
RESULTS duced similar results with both sharing and
Training ~~~~~~~praising. Mean numbers of shares and praises
Troining increased to levels well above those obtained dur-
The results of the training sessions are shown ing baseline conditions. After the initial acquisiin
Figure 3. Reinforcement of true reports pro- tion period (Sessions 9 to 11), percentages of the
10
8
%OF GROUP REPORTING
_.. %OF GROUP DOING
SHARES(UBJECT ME-AN PRAISESUJEC
PACKAE II MODEL+ SR- TRE
REPORTS OF BEHAVIOR
345
.
.
ANN ROGERS-WARREN and DONALD M. BAER
Table 3
The sequence of experimental conditions in Experiment
II.
Group C
Sharing Praising
Baseline
Package II
Model sharing
Model reports of sharing
Reinforce true reports
of sharing
Package II
Model praising
Model reports of prais-
Baseline ing
Reinforce true reports
of praising
group sharing and reporting averaged above
80%. Percentages of doing and reporting praising
increased rapidly and stabilized at about
80% during the last five sessions of the experiment.
Probes
The results of the generalization probes are
shown in Figure 4. When reinforcement for true
reports began in the training sessions, sharing in
the probe setting gradually increased, and continued
throughout the remaining sessions. Shares
averaged about 0.62 per child before training
and about 1.8 per child after training.
No generalization of praising training was
apparent.
DISCUSSION
The results indicated that both sharing and
praising may be increased by reinforcement of
true reports of those behaviors, and that this
reinforcement contingency was sufficient to establish
correspondence without prior reinforcement
for any (true or untrue) reports of behavior.
A one-time instruction ("Tell me if you
shared [praised)") was used to facilitate acquisition
of reporting during the first two
sessions of each reinforcement condition. It is
possible that reports of sharing and praising
would have occurred eventually, as they did in
Experiment I; however, the brevity of the summer
term in which the experiment was conducted
prevented further exploration of that possibility.
Although the instruction was given after the
work session, immediately preceding the subjects'
opportunity to report, it may have served as a
prompt for actual behaviors, as well as for reports
of the behaviors.
The results of the probe sessions indicated that
sharing generalized to the second setting, but
that praising did not. One explanation for this
difference is that sharing and praising were
naturally followed by different consequences.
Offers to share materials frequently resulted in
the acquisition of another, desired material; however,
praising was typically followed by "thank
GENERALIZATION
SHARING
C-
0
LL
0
AI- zw
C.)
w
0L
PRAISING
100
80
60
40
20
%OFGROPDO
DE&MMEEANNSHPARARIESESS
PER SUBJECT
t TRAININGBEGA IN
USERCn
5 10 15
SESSIONS
m
4 >z
2 Xf I
10tams
160m
8cr
4-
2
u p s- * Io i 20
Fig. 4. Sharing and praising during generalization
probes conducted in Experiment II (Group C). The
percentage of group members sharing and the mean
number of shares per subject are shown on the upper
graph; the percentage of group praising and the mean
number of praises per subject are shown on the lower
graph. The axis on the left refers to percentage of
group doing the target behavior; the right-hand axis
refers to the mean number of target behaviors.
346
TEACHING CHILDREN TO SHARE AND PRAISE
you" or by a reciprocal praise statement. The
acquisition of materials may have been a sufficiently
reinforcing event to maintain sharing,
while the consequences of praising were not. It
is also possible, as suggested earlier, that training
had a more obvious effect on sharing because
sharing was already in the subjects' repertoires
and the training only served to increase it. Praising,
on the other hand, was not initially present
and first needed to be established.
It should also be noted that relatively few
probe sessions followed praising-training. It is
possible that praising would have occurred in the
probe sessions after subjects had acquired a
longer history of reinforcement for praise reports
in the training setting.
EXPERIMENT III
In the first two experiments, subjects' praising
had nearly always taken the form of statements
such as "I like your picture", or "I like what you
are making". This type of praise, while sufficient
to fulfill the criterion for reinforcement, seemed
to be of limited value as a social reinforcer.
Since a primary objective of the research was to
increase appropriate social responses, it was
decided that specific, varied forms of praise
should be trained in addition to the more general
category of praising.
METHOD
Subjects
Twelve subjects were selected from two different
preschool classrooms and divided into two
groups of six. Three members of each group
came from the first classroom, and three from
the second. Group D was composed of three 4-yrold
boys, two 4-yr-old girls, and one 3-yr-old
girl. Group E consisted of four 4-yr-old boys,
one 4-yr-old girl, and one 3-yr-old girl.
Setting
Experimental sessions were carried out in an
experimental room near the subjects' regular
classrooms. Separate sessions were conducted
for each group.
Behavior Definitions
For the purposes of this experiment, praising
was divided into two mutually exclusive categories,
general praising and specific praising.
General praise statements were those consisting
of "I like . . ." and a general object (e.g., "I
like your picture", "I like what you are making",
"I like that"). Any other praise statement was
scored as specific praise. Specific praise, in addition
to indicating liking or admiration, noted
particular qualities (e.g., neatness, number of
colors), particular objects (e.g., circles, lines,
houses), or particular procedures (e.g., drawing
with two pencils, covering the whole page with
colors). Specific praise included statements such
as "I like the way you used the color red", and
"That's a neat house you drew".
Using the same criteria, reports of praise were
scored as either general or specific.
Experimental Design
The experimental design within each group
was a multiple baseline across three behaviors:
sharing, general praising, and specific praising.
There was also an umbrella multiple baseline
across these two groups of subjects.
All experimental conditions were identical to
those described in the General Methods section.
The sequence of experimental conditions is summarized
in Table 4.
RESULTS
Group D
The results for Group D are shown in Figure
5. No effects of the "modelling-only" procedure
were observed. When reinforcement of
true reports of sharing was introduced, sharing
rapidly increased to an average of about seven
shares per subject. Reports of sharing and the
percentage of the group sharing showed corresponding
increases.
347
ANN ROGERS-WARREN and DONALD M. BAER
Table 4
The Sequence of Experimental Conditions in Experiment III
Group D _ Group E
General Specific General Specific
Sharing Praising Praising I I Sharing Praising Praising
I_
Baseline
Model Sharing
Package II
Model sharing
Model reports
of sharing
Reinforce true
reports of
sharing
Baseline
Baseline
Package II
Model gen.
praising
Model reports
of gen.
praising
Reinforce true
reports of
gen. -praising
Baseline
Baseline
Package I
Model spec.
praising
Model reports
of spec.
praising
Reinforce any
reports of
spec.
praising
Package II
Model spec.
praising
Model reports
of spec.
praising
Reinforce true
reports of
spec.
praising
Baseline
Model Sharing
Package II
Model sharing
Model reports
of sharing
Reinforce true
reports of
sharing
Baseline
Baseline
Package II
Model gen.
praising
Model reports
of gen.
praising
Reinforce true
reports of
gen.
praising
Baseline
Package I
Model spec.
praising
Model reports
of spec.
praising
Reinforce any
reports of
spec.
praising
Package II
Model spec.
praising
Model reports
of spec.
praising
Reinforce true
reports of
spec.
praising
I I
-1
I L. -i I I
348
TEACHING CHILDREN TO SHARE AND PRAISE
GROUP D
GENERAL PRAISE *
100 Baseline
80
0
SPECIFIC PRAISE
100 d % OF GF
So -0 %OF GI
-M MEAN S
60 P
PACKAGE I: MODI
40 REPC
PACKAGE U: MODI
20 REPC
n _ Bas4
ROUF
ROUI
;HAR
'RAIC
MEL+ '
)RTS
EL +
)RTS
Beline
WM&s
*-%% @@ @@ @@ @@- ------------* *@ s ----------*-*-@@-@ I- - - - -- -
Package I *Package HI
REPORT ING ..
P DOING
IES SUBJECT
6
3R+ ANY
OF BEHAVIOR4
SR+ TRUE
OF BEHAVIOR
SESSIOIS (2 day blocks)
Fig. 5. The percentage of Group D members performing the target behavior, reporting that behavior, and
the mean number of behaviors per subject for sharing (top graph), general praising (middle graph), and specific
praising (lower graph). The axis on the left refers to percentage of group doing and reporting the target
behavior; the right-hand axis refers to the mean number of target behaviors.
General praising increased less dramatically
when true reports were reinforced, averaging
about 1.8 praises per subject. Doing and reporting
general praise were variable, ranging from
50% to 100% of the group, but generally
showed a high degree of correspondence.
Because of the complexity of the verbal response
for specific praise, any report of specific
praise was initially reinforced (Package I). This
allowed subjects to contact the reinforcement
contingency more quickly than they might have
during Package II (reinforcement for true re-
0
Ua:-l
0
w
0
z
w
cwG
IL.-
m
z
D
m
I
-U
m
Cc
m
C)
-I
01 AL -.AL Ala-.A, -Alk
349
ANN ROGERS-WARREN and DONALD M. BAER
ports), and provided an opportunity for the
experimenter to prompt subjects to make different
reports of specific praise without interfering
with the "truth" requirement.
Only small increases in the mean number of
specific praises were noted when any report of
specific praise was reinforced (Package I). Reports
of specific praise increased moderately;
however, fewer than half the subjects actually
gave specific praise.
Reinforcement for true reports of specific
praise (Package II) produced consistent specific
praising, averaging about 1.7 praises per subject.
Reports of specific praise and the percentage of
the group specifically praising varied during the
first part of the condition, but increased to near
100% for the last four sessions. During the application
of Packages I and II to specific praise,
subjects continued to praise generally and reported
both general and specific praise.
Group E
The results for Group E are shown in Figure
6. Generally, the results for Group E are similar
to those for Group D for sharing and general
praising. Modelling appropriate sharing did not
produce significant changes in the sharing baseline.
When reinforcement was made contingent
on true reports, the percentage of group members
reporting the appropriate behaviors rapidly increased
to near 100%. The percentage of group
members actually doing the behaviors also increased,
so that a high level of correspondence
between reports of behavior and actual behavior
developed and was maintained throughout the
reinforcement condition.
When reinforcement was made contingent on
any report of specific praising (Package I), subjects
immediately began doing and reporting
specific praise. The percentage of group members
specifically praising reached 100% by the third
session of the condition. General praising and
reports of general praising declined to near baseline
levels.
During the final condition (reinforcement for
true reports of specific praise), the mean number
of specific praises per subject averaged about 2.0
while the percentages of reporting and doing
the behavior remained about 80%. Functionally,
this final condition was no different from the
preceding one for the subjects; in both conditions
they truly reported specific praise and were rewarded
for their reports.
DISCUSSION
The results supported the findings of the two
previous experiments and demonstrated that reinforcement
for true reports of behavior can be
used to increase a still more complex behavior,
specific praising.
Specific praise appeared to be a more reinforcing
event than general praise had been in the
same experiment or in earlier ones. General
praise was often "recited" by one child to another,
then repeated by the second child. Often
the child being praised did not attend to the general
praise statement a peer was making. Specific
praise statements seemed more spontaneous and
less stereotypical, possibly because they required
the subject making the statement to stop working,
assess other's activities, and then make an
appropriate statement. Because such statements
were of longer duration and of more unique
content than the general praise statements, specific
praise more often elicited attention from the
child being praised.
Most subjects demonstrated several different
types of specific praise and tended to vary their
praise statements within and across sessions.
None of the subjects imitated the modelled
praise statements, but rather, subjects generated
specific praise that was appropriate to a peer's
particular activity.
During the reinforcement of true reports of
general praising, some subjects in Group E began
praising specifically. These same subjects performed
and reported general praising and, thus,
were rewarded for their reports of praising. This
reinforcement may have maintained both specific
and general praising.
It is not clear whether it was necessary to train
general praise, or the use of a single praise state-
350
TEACHING CHILDREN TO SHARE AND PRAISE
GROUP E
SHARING
BL
GENERAL PRAISE
Baseline
SPECIFIC
100
80
60
40
20
PRAISE
Baseline
PACKAGE I: MODEL+ SR+AN R
OF BEHI
PACKAGE II:MODEL+SR+ TRUI
REPORTS OF BEH)
A A A A . .A . -M.--
5
PKG.
EKG 10
IEPORTS
MVOR
E
WVIOR .6
4
02
15 25
SESSIONS (2 day blocks)
35
Fig. 6. The percentage of Group E members performing the target behavior, reporting that behavior, and
the mean number of behaviors per subject for sharing (top graph), general praising (middle graph), and
specific praising (lower graph). The axis on the left refers to percentage of group doing and reporting the
target behavior; the right-hand axis refers to the mean number of target behaviors.
a.
0
z
cc
0
w
C-)
m
z
m
I
0
m11
m
Cc)r
C-
351
-A AANN
ROGERS-WARREN and DONALD M. BAER
ment, before training the class of specific praise.
The results obtained with Group E suggest that
specific praising might be trained immediately
with the desired results. It may be that long
periods of reinforcement for general praise,
without the occurrence of specific praise, impede
the acquisition of the specific praise responses
and reports. The subjects in Group E who exhibited
specific praising during the general praising
training (and thus were not exposed to a
long period of reinforcement for only general
praise) reported specific praise more quickly
and more truly than the subjects in Group D,
who did not exhibit any specific praise until that
behavior was trained. In Group E, reinforcement
for any reports of specific praise quickly produced
the actual behavior from nearly all subjects;
yet, Group D showed very few instances
of specific praising when any report of that
behavior was reinforced, and somewhat lower
levels of performance during reinforcement for
true reports of specific praise. Additional research
would be needed to determine if this
effect is the function of a long period of reinforcement
for general praising without any occurrence
of specific praise, or simply reflects the
difference in the subject's initial verb repertoires.
GENERAL DISCUSSION
The experiments reported here extend the
previous research on verbal self-control of behavior
along two dimensions: from simple to
complex behaviors, and from asocial to social behaviors.
Previous research has been limited in its application
to nonsocial behaviors such as bar
pressing (e.g., Bem, 1967; Birch, 1966), eating
(Lovaas, 1964b), and the use of preschool materials
(e.g., Risley and Hart, 1968). Although
the target behaviors have been varied, all were
primarily motor responses that did not require
the cooperation or presence of persons other than
the subject. The two behaviors examined here,
sharing and praising, could occur only with one
or more peers present, and thus increases in
either behavior served to further social interaction
among group members.
There are other ways in which sharing and
praising are more complex responses than those
previously studied. Sharing has both verbal and
nonverbal components. In particular, the nonverbal
component may take a variety of forms
(e.g., a gift, an exchange of materials, acceptance
of a material, or simultaneous use of a material).
Although general praising could consist of a
simple verbal statement without accompanying
responses, specific praise required that a subject
assess a peer's activity, attend to a specific aspect
of it, choose a verbal description that indicated
this aspect, and make a praise statement specifically
mentioning what was being praised.
Generally, similar results were obtained in
all three studies; however, there were considerable
differences in the rates of behaviors and the
time needed for subjects to acquire these behaviors.
There are a number of possible reasons
for these differences. First, the composition of the
groups differed with respect to the ages and skills
of the subjects. The subjects who acquired the
reporting and actual behaviors more quickly appeared
to be more verbally and socially skilled.
Studies by Luria (1961) and Bem (1967) reported
that younger children consistently exhibited
little verbal control of behavior without
special training. With the exception of only one
subject, this seemed to be true in all three studies
reported here. Younger subjects required longer
training to make reports of their behaviors and
to increase their actual behaviors.
Differences among groups might also be attributed
to the fact that in some groups, one or
two subjects served as informal, unprogrammed,
peer models for appropriate behavior. The rates
of sharing, praising, and the reporting of these
behaviors may have been affected by the presence
and relative effectiveness of such interactive
models. Apparent models were subjects with unusually
good verbal and social skills, who interacted
with most other members of the group. On
several occasions, model-subjects were observed
instructing peers about the appropriate behaviors
352
TEACHING CHILDREN TO SHARE AND PRAISE
and explaining the contingencies of reinforcement.
Such subject-models were apparent in
Groups A, B, and E, but not in Groups C and D.
While these experiments clearly demonstrated
that the package (modelling and reinforcement
for true reports) can be used to increase social
behaviors, three issues in particular, are left unresolved.
First, the function of the reporting
responses as an example of verbal mediation is
not clear. It is possible that the reinforcement
for reports of behavior paradigm used here was
simply functioning as a delayed reinforcement
procedure, and that children who were rewarded
for sharing and praising after each session would
respond in the same way as children who were
rewarded after each session without reporting
their behavior. A further study, in which delayed
reinforcement was directly compared with reinforcement
for reports of behavior, would aid in
this respect.
The effects of the adult model should also be
more carefully evaluated. In most instances, the
modelling of appropriate sharing responses without
reinforcement for reports of behavior did not
have a significant effect on rates of sharing; however,
the effect of the model after it was paired
with the reinforcement contingencies was not
investigated. The model seemed to be particularly
important during the initial sessions of an
experimental condition, possibly because it provided
the subjects with information about the
reinforcement contingencies, in addition to providing
examples of the appropriate behavior.
In most cases, subjects did not appear to be attending
to the model after the first five or six
sessions of the condition.
A third issue that merits further consideration
is generalization, since, potentially, one of the
most important aspects of this procedure is its
range of control over time and settings. Generalization
data were collected throughout Experiments
II and III. In Experiment II, there was
clear generalization of sharing but not praising,
to a second setting. In Experiment III, limited
generalization of sharing was observed, but no
generalization of praising was evident. The generalization
data for Experiment III have not been
included here because of difficulties related to
their collection.
In Experiment III, a probing technique (a
5-min observation of the subjects while they
played with materials similar to those used in
training) was used in place of the 15-min generalization
sessions employed in Experiment II.
It is highly questionable if the 5-min probe
was sufficient to detect generalization. The variable
results, particularly in the case of sharing,
in Experiments II and III, lead to the conclusion
that the conditions under which generalization
data are collected should be specified as clearly as
the primary experimental procedures. It may be
that conflicting reports of generalization are a
product of different procedures and settings for
collecting generalization data, and, thus, do not
necessarily reflect the strength of the primary
training or intervention.
It was of interest in the present studies that
although formal evidence of generalization was
limited, anecdotal reports from parents and
teachers of subjects in all five groups suggested
that sharing and praising began to occur more
frequently in classrooms and at home after training
was initiated in the experimental sessions.
The present studies did not attempt to consider
the theoretical issues underlying the verbal selfcontrol
of behavior. It should be noted that these
studies in some ways resemble Festinger's studies
of cognitive dissonance (e.g., Festinger, 1957).
For example, Festinger and Carlsmith (1959)
reported that when subjects were rewarded for
publicly advocating an opinion discrepant from
their private ones, a subsequent change in the
direction of the opinion publicly advocated was
noted. The change was considered to be an afterthe-
fact way of justifying the discrepant public
stance and thus alleviating the cognitive dissonance.
In the present studies, there were some
instances where subjects began sharing and praising
after reinforcement of untrue reports (that is,
during conditions when any report, true or untrue,
was reinforced), similar to the effects noted
in the Festinger and Carlsmith study. 353
354 ANN ROGERS-WARREN and DONALD M. BAER
Further research might take two directions. The component procedures discussed above might be further evaluated and tested for their relative functions. The present package also might be extended toward further application. This self-reporting procedure might also be an
efficient technique for maintaining social behavior in the classroom. Once true reporting has been established, the teacher might need only to be present during the reporting period. The teacher's attention during other times may be directed toward teaching other behaviors. The
delayed contingencies also allow for increasing social behavior without interrupting ongoing behavior among the subjects.
REFERENCES
Bem, S. L. Verbal self-control: the establishment of effective self-instruction. Journal of Experimental Psychology, 1967, 74, 485-491.
Birch, D. Verbal control of nonverbal behavior. Journal of Experimental Child Psychology, 1966, 4, 266-275.
Israel, A. C. and O'Leary, K. D. Developing correspondence between children's words and deeds. Child Development, 1973, 44, 575-581.
Festinger, L. A theory of cognitive dissonance. Evanston, Ill.: Row, Peterson, 1957.
Festinger, L. and Carlsmith, J. Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology, 1959, 58, 203- 210.
Lovaas, 0. I. Interaction between verbal and nonverbal behavior. Child Development, 1961, 32, 329-336.
Lovaas, 0. I. Cue properties of words: the control of operant responding by rate and content of verbal operants. Child Development, 1964, 35, 245-256. (a)
Lovaas, 0. I. Control of food intake in children by reinforcement of relevant verbal operants. Journal of Abnormal and Social Psychology, 1964, 18, 672-678. (b)
Luria, A. R. The role of speech in the regulation of normal and abnormal behavior. New York: Pergamon Press, 1961.
Meichenbaum, D. H. and Goodman, J. Teaching children to talk to themselves: a means of developing self-control. Journal of Abnormal Psychology, 1973, 27, 115-126.
Meichenbaum, D. H. and Goodman, J. The developmental control of operant motor responding by verbal operants. Journal of Experimental Child Psychology, 1969, 7, 553-565.
Monohan, J. and O'Leary, K. D. Effects of selfinstruction on rule-breaking behavior. Psychological Reports, 1971, 29, 1059-1066.
O'Leary, K. D. The effects of self-instruction on immoral behavior. Journal of Experimental Child Psychology, 1968, 6, 297-301.
Risley, T. JOURNAL OF APPLIED BEHAVIOR ANALYSIS CORRESPONDENCE BETWEEN SAYING AND DOING: TEACHING CHILDREN TO SHARE AND PRAISE' ANN ROGERS-WARREN AND DONALD M. BAER BUREAU OF CHILD RESEARCH UNIVERSITY OF KANSAS
Five small groups of preschool children were taught to share and praise by the modelling of these behaviors and reinforcement of their reports of sharing and praising. Experiment I demonstrated that modelling and reinforcement of any (true or untrue) reports of sharing, and then of praising, promptly increased reports of the corresponding behaviors.
Modelling and reinforcement for true reports of each behavior increased both reporting and actual behavior. Experiment II showed that both reported and actual sharing and praising may be increased by modelling and reinforcement for true reports of the target behavior, without previous reinforcement for any (true or untrue) reports of those behaviors. Sharing, but not praising, generalized to a second setting. Experiment III replicated the results of Experiment II for sharing and praising, and demonstrated similar success in increasing a third behavior, specific praising. In general, these experiments show that developing correspondence between children's reports of behavior and actual behavior may be an efficient means of increasing prosocial responses.
DESCRIPTORS: social behavior, sharing, modelling, praise, reporting, generalization, multiple baseline, reinforcement delay, children In a very broad sense, verbal behavior is similar to currency. At any given time, there may not be sufficient gold to back up every dollar
exchange, but sellers and buyers usually act as if there were. In interactions among adults, there is a similar continual exchange of verbal reports, generally as if there is, or will be, some other behavior to correspond to the verbal reports, make them true, and give them value.
While operant research has not yet attempted to define the parameters of truthtelling among adults, a number of researchers (e.g., Bem, 1967; Birch, 1966; Lovaas, 1961; Lovaas, 1964b; Risley and Hart, 1968; Sherman, 1964) have investigated the relationship between young
children's verbal behavior and their corresponding other behaviors.
'This research was funded in part by grant MH 11739 from the National Institute of Mental Health for training new personnel for behavior modification and by grant 3336-5038 from the University of Kansas to Dr. Donald M. Baer. Reprints may be obtained from Ann Rogers-Warren, Department of Human Development, University of Kansas, Lawrence, Kansas 66045.
Investigations in the area of verbal self-control of behavior have generally taken one of two forms. In the first form (e.g., Bem, 1967; Birch, 1966; Lovaas, 1964a; Luria, 1961; Meichenbaum and Goodman, 1969), subjects have been taught to make a relevant verbal response while performing a motor response. Such studies have shown that certain verbal responses facilitate the corresponding motor response. For example, Lovaas (1964a) demonstrated that subjects who said the word "faster" as they pushed a lever, would depress the lever more quickly than subjects who said the word "slower". Bem (1967) reported that, in a task requiring children to press a lever to indicate the number of stimuli that were presented briefly and then withdrawn, those who counted as they pressed the bar responded more accurately than those who did not.
Within the same paradigm, several authors (e.g., Meichenbaum and Goodman, 1971; Monohan and O'Leary, 1971; O'Leary, 1968) have demonstrated that training a child to selfinstruct may facilitate appropriate responding 335
1976, 93, 335-354 NUMBER 3 (FALL) 1976 ANN ROGERS-WARREN and DONALD M. BAER and impede inappropriate responding. For example, O'Leary (1968) found that when subjects were told to respond only on certain trials, in an experiment where responding on any trial resulted in reinforcement, those subjects who gave self-instructions indicating whether they could respond on a particular trial, "cheated" less than those who did not self-instruct.
Risley and Hart (1968) provided a strong demonstration of the control of behavior through reinforcement of related verbal behavior. Two groups of preschool children were initially rewarded for any (true or untrue) report of the use of a specific preschool material.
Reinforcement was then made contingent on only true reports, that is, those reports that corresponded to the actual use of the material. While reinforcement for corresponding reports successfully increased the subjects' use of materials, after the noncorrespondence- correspondence sequence of contingent reinforcement had been repeated a number of times, reinforcement for any (corresponding or noncorresponding) report was sufficient to increase significantly the use of a specific material. To that extent, Risley
and Hart were able to demonstrate that "saying" could be brought to control "doing".
Israel and O'Leary (1973) further examined the role of verbal behavior in regulating nonverbal behavior. In two comparisons of sayingthen- doing, they found that having preschool subjects say which toy they were going to play with during the next play period produced
greater correspondence between saying and doing, than having subjects report which toy they had played with after the play period. Only slight increases in correspondence occurred when subjects were rewarded for the desired (but not necessarily true) content of their reports. Unlike Risley and Hart (1968), Israel and O'Leary did not find that saying came to control doing for children in the doing-then-saying sequence when only content was reinforced, even after a series of reinforcement-for-content (noncorrespondence) followed by reinforcement-for-correspondence.
Further, Israel and O'Leary did not, as had Risley and Hart, obtain high levels of correspondence for the doing-then-saying groups, even when correspondence was reinforced.
Direct comparison between the two studies is difficult because correspondence was defined by Risley and Hart as "doing and saying", while Israel and O'Leary defined it as "either doing and saying or not doing and not saying". The direct measures of positive correspondence (percentage of subjects saying and percentage of subjects doing) are not reported by Israel and O'Leary. Whatever differences there may be in immediate efficiency of the saying-then-doing and doing-then-saying sequences, both are based on the same procedural approach. If the two sequences are abstracted, and viewed as they operate over a number of sessions, it is apparent that in both cases the subjects' verbal behavior is intended to affect the next opportunity to perform the corresponding behavior. The primary difference
is that in the first case (saying-thendoing) that opportunity arises within a few minutes, and in the second (doing-then-saying), the opportunity does not arise until approximately 23 hr later.
If range of control, that is, the length of time during which verbal statements might affect the corresponding behavior, is of concern, then the doing-then-saying sequence might be more useful.
Such a sequence might contribute to the generalization of the corresponding behaviors to other opportunities not specifically included in training, but occurring within the temporal range of the training. For example, a child who plays with blocks during the designated play
period, might also play with them at home during the evening, thus establishing correspondence between his/her verbal report and actual behavior on more than one occasion.
Both the Risley and Hart, and Israel and O'Leary studies focused on establishing correspondence, using choice of play materials as the behavior to be increased. The present study was concerned primarily with producing many instances of particular prosocial behaviors by
establishing correspondence between the chil- 336 TEACHING CHILDREN TO SHARE AND PRAISE dren's verbal reports and their related prosocial behavior. Since the target behaviors, sharing and praising, should be useful in a number of settings, the apparently greater range of the doing-then-saying sequence made it the preferred procedure. The purpose, then, was to develop an effective procedure for increasing complex social behaviors based on the previously demonstrated techniques for developing correspondence between children's saying and doing.
GENERAL METHODS
Since the same procedures were used in all three experiments, the general methods are described first.
Subjects
All subjects were children attending the Edna A. Hill Child Development Preschool Laboratory at the University of Kansas. Thirty-two children were involved in the three experiments, 19 males and 13 females ranging in age from 3 yr, two months, to 5 yr, six months.
Setting
Experimental sessions were conducted daily, Monday through Thursday. Each session lasted about 15 min. The experimental sessions were divided into two parts, the work period (10 min) and the reporting period (about 5 min). The subjects were seated on the floor, around a
large piece of paper, and had access to art materials (paper, pens, pencils, crayons, markers) during the work period. The experimenter changed the seating arrangement each day by placing cards with the subjects' names on them around the paper and requesting subjects to sit behind their own name card. At the end of the work period, the materials were collected and an activity called the Reporting Period began. Duringthe Reporting Period, the experimenterasked the children individually what they had done "while we were working today". Every
subject was given one opportunity to respond.
The session ended when each child had been queried and the experimenter indicated it was time to go to another activity or return to the classroom.
Behavioral Definitions
The behaviors recorded for the subjects were sharing, praising, and reports of sharing and praising. Experimenter prompts and reinforcement of reports were also recorded. Sharing. Two classes of sharing, verbal sharing and nonverbal sharing, were recorded. Verbal
sharing included any verbalization by a subject to a peer in which the subject: (1) invited a peer to join in a particular activity, or (2) verbally accepted the invitation of a peer to join in a particular activity, or (3) offered to share materials with a peer, or (4) verbally accepted a
peer's offer to share materials, or (5) offered to trade materials with a peer. Nonverbal sharing was recorded when one subject passed or handed a material to a second subject, if both subjects had touched the material within 5 sec, or when two or more subjects simultaneously used the same material (e.g., colored on the same sheet of paper at the same time).
Praising. Praising was defined as any verbalization by a subject to a peer in which the subject indicated approval, liking, or admiration for the peer or any aspect of the peer's art work. There was essentially four forms of praise: (1) "I like and a general object (e.g., "I like your
picture"); (2) "I like . . ." and a specific object or quality (e.g., "I like the way you used the color blue"); and (3) direct general praise (e.g., "neat picture" or "pretty"); and (4) direct specific praise (e.g., "Nice dogs" or "That's a pretty house you made"). In Experiments I and II, all four forms were recorded under the general category of "Praise"; in Experiment III, Form 1 ("I like . . ." and a general object) was scored as "General Praise", and Forms 2, 3, and 4 were scored as "Specific Praise".
Reporting. Subjects were recorded as making a report of sharing or praising whenever they responded to the experimenter's question, "What did you do while we were working today?", with 337 ANN ROGERS-WARREN and DONALD M. BAER a statement indicating that they had shared materials or had praised a peer (e.g., "I shared my pen with Susie", or "I told Jon I liked his picture").
In Experiment III, reports of praise were scored as either "General" or "Specific", using the same criteria as employed for general and specific praising behaviors. Experimenter-behaviors. Two experimenterbehaviors, reinforcement of reports and prompts for reports, were also recorded. Reinforcement was scored whenever the experimenter indicated to a subject that she approved of the behavior the subject reported and offered a bit of food (M&M's or fruit) immediately following the report.
Prompts were scored each time the experimenter corrected a child's verbal response, or modelled a response and instructed a child to repeat it during the reporting period.
Recording Procedures
Two observers were seated within viewing and hearing distance of the group of subjects, one observer at each of the narrow ends of the paper on which the subjects were working. Each observer recorded the behaviors of the three or four children closest to her, depending on the size of the group. Continuous observations were made of the children's behaviors during the work period; discrete observations of each subject's verbal report and the experimenter's behaviors were made during the reporting period.
Observers recorded instances of sharing and/or praising on the current subject's form by writing a symbol identifying the peer to whom the behavior was directed, in the appropriate column identifying the type of material shared or the type of praising used. Experimenter reinforcement and prompts were recorded each time they occurred. Reports of sharing and praising were scored only once for each subject, immediately following the experimenter's questioning of that child.
Reliability was assessed by having a third observer record independently, but simultaneously, with each of the primary observers. Interobserver agreement was determined by comparing the number of behaviors recorded by each observer for a given subject in a particular category of behavior. For example, observer A's record of pen sharing by S1 with S2 was compared with observer B's record of pen sharing by SI with S2. Similar comparisons were made for all materials, for praising, for reports of behaviors, and for experimenter behaviors directed to each subject. Each reliability was then calculated using the formula:
Reliability = number of agreements X 100. number of agreements + number of disagreements Experimental Conditions Baseline. During the initial baseline conditions there were no scheduled consequences for sharing, praising, or reports of sharing or praising.
The experimenter and an adult model sat at a small table behind, but in sight of, the group and used the same art materials as were available to the subjects. Neither sharing nor praising was modelled. Throughout the work period, the experimenter intermittently praised the subjects for "working hard" and attending to the materials. If a subject shared or praised, the experimenter waited at least 10 sec before attending to any subject. During the reporting period, each child was questioned about his/her activity during the previous work period, but no reinforcement for appropriate responses was given.
Modelling only. To determine if modelling alone would result in the desired behavior, the experimenter and the model demonstrated the appropriate response five or six times during the work period. No model of appropriate reporting ehavior was provided, and no reinforcement
for reports of sharing or praising was available. Package I: modelling and reinforcement for any report of the behavior. The first part of thepackage manipulation consisted of the experimenter
and the model demonstrating the appropriate
response, either sharing or praising, as had
been done in the Modelling Only condition. The
topography of sharing and praising behaviors
was systematically varied. In Experiments I and
338
TEACHING CHILDREN TO SHARE AND PRAISE
II, both general and specific praising responses
were modelled during the praising condition. In
Experiment III, only general praise was modelled
during the manipulation of general praising,
and only specific praise was modelled during the
specific praise condition.
At the beginning of the reporting period, the
experimenter asked the model what he had done
while he was working. When the model reported
the appropriate behavior, the experimenter
rewarded him with positive comments
and food. The subjects were then asked individually
what they had done during the work
period. Any report, true or untrue, of sharing
or praising, depending on the experimental condition,
was followed by positive comments, and
either M&M's or fruit, from the experimenter.
When Package I was introduced to any group,
it was sometimes necessary for the experimenter
to prompt two or three subjects to make an appropriate
report (e.g., "Tell me that you
shared"). After at most three sessions, all
prompting was discontinued. Subjects were not
recorded as reporting appropriately if the report
was prompted.
Package II: modelling and reinforcement for
true reports of behavior. The second package
was procedurally similar to the first, except that
only true reports of sharing or praising were
reinforced. As the subject responded to the experimenter's
query, the observer who had recorded
that subject's behavior discreetly signalled
"yes" or "no" to indicate if the subject was reporting
truly. True responses were reinforced.
When a subject reported untruly, the experimenter
responded, "But you didn't really, did
you?" No response was made following an inappropriate
report (e.g., a report of sharing in a
condition when reports of praise were being reinforced);
the experimenter paused 5 sec, then
asked the next subject about that subject's activities.
Package III: modelling and reinforcement for
either true reports of sharing or true reports of
praising. During Package III, the experimenter
and the model demonstrated both sharing and
praising, and the model reported both sharing
and praising. Subjects were rewarded for true
reports of either behavior. That is, if they either
shared or praised, and gave a corresponding report,
they were rewarded. All other procedures
were identical to Package II.
Package IV: modelling and reinforcement for
true reports of both sharing and praising. Package
IV consisted of modelling both sharing and
praising responses and delivering reinforcement
contingent on true reports of both sharing and
praising. Except that each subject was required
to do and report both behaviors, the procedures
were identical to Package II.
Reliability
Reliability figures for all subject and experimenter
behaviors are shown in Table 1.
EXPERIMENT I
METHOD
Subjects
The 14 subjects selected for this study were
divided into two groups. Gr6up A was made
up of six children: one 4-yr-old girl, two 3-yrold
girls, and three 4-yr-old boys. Group B consisted
of eight children: two 4-yr-old girls, five
4-yr-old boys, and one 3-yr-old boy.
Setting
The study was conducted in a regular preschool
classroom, in an area away from other
classroom activities. While an experimental session
was being conducted with one group,
members of the other group participated in conceptual
activities in different parts of the same
classroom, and generally did not attend to the
experimental activity.
Experimental Design
Experiment I consisted of two parts. In the
first part, reinforcement of any report of behavior
(Package I), followed by reinforcement
of only true reports of behavior (Package II),
was systematically introduced in a multiple-
339
ANN ROGERS-WARREN and DONALD M. BAER
Table 1
Range and Mean Reliability for Subject and Experimenter Behaviors in Experiments I, II, and III
Group A Group B Group C Group D Group E
Behavior Range X Range X Range X Range X Range X
Sharing 88-100 92 75-100 89 88-100 91 62-100 90 69-100 86
General praising* 85-100 91 88-100 94 83-100 93 66-100 91 79-100 92
Specific praising -- - 80-100 92 86-100 88
Reports of sharing 75-100 94 - 100 100 - 100 75-100 92
Reports of
praising* 88-100 91 - 100 100 75-100 100 100
Reports of
specific praising - - - - 0-100 90 - 100
Experimenter
reinforcement - 100 - 100 - 100 - 100 100
Experimenter
prompts 100 - 100 - 100 - 100 100
* In Experiments I and II, all praising and reports of praising were recorded in the category of General
Praising, hence reliability scores are reported only for the category of General Praising.
baseline design across sharing and praising by
Group A. In the second part, these procedures
were replicated with Group B, and then two
additional conditions, reinforcement for true reports
of either behavior (Package III), followed
by reinforcement of true reports of both behaviors
(Package IV), were examined.
The sequence of experimental conditions for
this experiment is summarized in Table 2.
All experimental conditions were identical to
those described in the General Methods section.
RESULTS
Group A
The results for Group A are shown in Figure
1.
The introduction of reinforcement of any
report of behavior (Package I) resulted in a
gradual increase in mean number of shares,
double-to-triple the baseline rate, and the percentage
of the group reporting sharing steadily
increased. The percentage of group members
sharing was variable, ranging from 0% to
100%, but averaged higher than during baseline.
Under the same conditions (Package I),
mean praising increased only slightly, although
the percentage of the group praising and reporting
praising increased markedly.
When reinforcement was provided for true
reports of sharing (Package II), the mean number
of shares increased to a consistent level of
around four shares per child per session. Percentage
of the group sharing was moderately
high, while reports of sharing steadily rose to
a high of 90%.
Reinforcement for true reports of praise
(Package II) resulted in sharp upward trends in
both reporting and actual praising. A more
gradual increase was seen in the mean number of
praises per session.
Group B
Similar results for Group B are shown in
Figure 2.
For Group B, which had exhibited a moderate
level of sharing during baseline, reinforcement
for any report (Package I) produced no significant
increase in either the percentage of the
group displaying sharing, or the mean number
of shares per subject. The same procedure applied
to praising resulted in a slight increase in
the mean number of praises and produced
moderate increases in the percentages of the
group reporting and praising.
Reinforcement for true reports of behavior
(Package II) further increased the mean number
of shares and praises. Additional increases in
340
TEACHING CHILDREN TO SHARE AND PRAISE
Table 2
The Sequence of Experimental Conditions in Experiment I
Group A Group B
Sharing Praising Sharing Praising
Baseline
Package I
Model sharing
Model reports of sharing
Baseline Baseline
Reinforce any reports
of sharing
Package I
Model praising
Model reports of prais-
Baseline ing Model sharing Baseline
Reinforce any reports
of praising
Package II Package I
Model sharing Model sharing
Model reports of shar- Model reports of sharing
Baseline ing
Reinforce true reports Reinforce any report
of sharing of sharing
Package II Package I
Model praising Model praising
Model reports of Model reports of prais-
Baseline praising Baseline ing
Reinforce true reports Reinforce any reports
of praising of praising
Package 111 Package II
Model both sharing and praising Model sharing
Model reports of both Model reports of shar-
Reinforce true reports of either ing Baseline
Reinforce true reports
of sharing
Package IV Package II
Model sharing and praising Model praising
Model reports of both Model reports of prais-
Reinforce true reports of both ing
Reinforce true reports
of praising
Package III
Model both sharing and praising
Model reports of both
Reinforce true reports of both
Package IV
Model both sharing and praising
Model reports of both
Reinforce true reports of both :
341
ANN ROGERS-WARREN and DONALD M. BAER
GROUP A
SHARING
mrIb
-u
Dm
Cm
C)
SESSIONS (2day blocks)
Fig. 1. The percentage of Group A members reporting sharing, actually sharing, and the mean number of
shares per subject are shown on the upper graph. Percentages of reporting praising, actual praising, and mean
number of praises per subject are shown on the lower graph. Note that the axis on the left refers to percentage
of group doing and reporting the target behavior; the right-hand axis refers to the mean number of
target behaviors.
percentages of children doing and reporting
the behaviors were also obtained.
The final two conditions were reinforcement
for true reports of either praising or sharing
(Package III), followed by reinforcement for
true reports of both praising and sharing (Package
IV). Actual praising (mean number of percentage
of group praising) and reports of
praising were maintained at consistently high
levels during both of these conditions. Sharing
remained near baseline levels during the "either"
condition, but rapidly increased to its highest
levels when reinforcement was made contingent
on truly reporting both behaviors.
DISCUSSION
Generally similar results were obtained with
both groups. Reinforcement for any report of
behavior (true or false) produced moderate increases
in reporting and small increases in actual
behavior. Reinforcement for true reports was
generally more effective, and resulted in higher
rates of both reporting and actual behavior.
In both groups, sharing increased more
rapidly and was maintained at consistently
higher levels than praising. These differences
may have been a function of the differences
represented in the baseline measures of behavior.
342
TEACHING CHILDREN TO SHARE AND PRAISE
SHARING
GROUP B
*~~~~~~~~~~~.~~@.~~*~~@@@
PRAISING.
%REPORTING PKG: BL. PKGI
% DOING
80 & SHARES/SUBPRAISES
JECT
PKGI MODEL+S,
69 ANY REPORTS
xm
z
ffl
Cen
m
0
-4
SESSIONS (2dayblocks)
Fig. 2. The percentage of Group B members reporting sharing, actually sharing, and the mean number of
shares per subject are shown on the upper graph. Percentages of reporting praising, actual praising, and mean
number of praises per subject are shown on the lower graph. The axis on the left refers to percentage of
group doing and reporting the target behavior; the right-hand axis refers to the mean number of target behaviors.
Many subjects already had sharing in their
repertoires; thus, the experimental manipulations
needed only to increase it. Praising, however,
was not common to most subjects, and it
was necessary first to teach the subjects how to
praise (one goal of the modelling component),
and then to increase praising.
Although high rates of praising eventually
were obtained in both groups, many of the
praise statements seemed mechanical and lacking
in reinforcement value. Praising was nearly
always in the form, "I like your picture" or "I
like what you are making", with little voice
inflection or modulation. Often, subjects did not
look up from their own activities as they delivered
praise statements. As rates of praising
increased, more inappropriate praise was noted.
Sharing, however, was usually appropriate:
subjects used materials first, then offered them to
peers. Trading of materials was the most frequent
form of sharing, but on occasion, children
would simultaneously use materials (typically
drawing together on the same piece of paper).
Most exchanges of material were arranged
verbally.
The functions of the final two contingencies
(Package III and Package IV) for Group B are
clear enough in showing their effectiveness at
increasing many aspects of the target behavior.
In one respect, however, only a tentative conclusion
is possible: the praise-or-share conditions
may have functioned as merely an extension of
the previous reinforcement for true reports of
praise (Package II). The similar levels of prais-
CD
0
cc
a
LL
0
w
z
w
w
a.
343
ANN ROGERS-WARREN and DONALD M. BAER
ing may have been maintained because the
subjects did not discriminate the subtle changes
in criteria for reinforcement, which allowed the
praising behavior to maintain an unchanged rate
of reinforcement. This conclusion is also suggested
by the fact that sharing failed to increase,
although reports of both behaviors were modelled.
Apparently, it was not obvious to the
subjects that they could report either sharing or
praising if they had performed the corresponding
behavior.
Since these conditions (Package III and Package
IV) were introduced after subjects had a long
history of reinforcement for reporting sharing
and praising, the resulting high levels of behavior
possibly represent the cumulative effect
of numerous conditions. However, a complete
analysis of order effects would be necessary for
a final answer. Programmatically, that analysis
might not be important.
EXPERIMENT II
The second study was designed to determine
if preschool children could be taught sharing
and praising skills through the contingent reinforcement
of true reports of those behaviors
without first being rewarded for giving any (true
or untrue) reports, and to determine the extent,
if any, that training to share and praise would
generalize to a second setting, where no reports
of the behaviors were made.
METHOD
Subjects
Six children, two 3-yr-old boys, one 3-yr-old
girl, two 4-yr-old girls and one 4-yr-old boy,
served. Three children were from one classroom
of the preschool, and three children were from
a second classroom. The subjects from different
classes did not know each other before the experiment.
Setting
Experimental (training) sessions were conducted
in a small classroom near the subjects'
regular classrooms. There were no children,
other than the subjects, present during the training
session.
Generalization (probe) sessions were conducted
in the subjects' original classrooms during
a free-play period or a small-group activity
period. Probes occurred about an hour before the
training session and lasted 10 min. During a
generalization probe, a second experimenter
served as a teacher. The children were seated
on the floor in an area adjacent to the general
classroom activities, and had free access to a
supply of materials. The materials were varied
each session, and included tinkertoys, lego, chalk,
playdough, and blocks.
At the beginning of each probe session, the
experimenter told the subjects that children who
stayed in the area until the timer rang would
receive an animal sticker or a star for their cards.
Each child had a 12.5 by 17.5 cm card on which
the stickers were pasted. At the end of the week,
the subjects took the cards home.
Ten to 12 times during each probe session, the
experimenter praised the subjects individually
or as a group for "playing nicely", and for staying
in the area. No reinforcement for sharing or
praising was given. If a subject shared or praised,
the experimenter waited 10 sec before attending
to that child.
Any nonsubjects who came near the area
where the probe session was being conducted,
were invited to join the group and were also
praised for "playing nicely", but did not receive
stickers at the end of the session. There were no
opportunities for reporting during probe sessions.
Experimental Design
This study was conducted as a multiple baseline
across the two behaviors, sharing and praising.
Generalization data were collected during
all phases.
All experimental conditions were identical to
those previously described in the General
Methods section. The sequence of experimental
conditions is summarized in Table 3.
344
TEACHING CHILDREN TO SHARE AND PRAISE
GROUP C
SHARING
Baseline Package IE Baseline
a. 0
cc C!,
LL
0
w
z
w
CD
w
a.
20
01
*-D..00 00 00 0 eSeOe @ O S O ...
Baseline
m
z
m
I
-Q
0
m
Cl)
C
Cm0
-i
SESSIONS
Fig. 3. The percentage of Group C members reporting sharing, and the mean number of shares per subject
are shown on the upper graph. Percentages of reporting praising, actual praising, and mean number of praises
per subject are shown on the lower graph. The axis on the left refers to percentage of group doing and reporting
the target behavior; the right-hand axis refers to the mean number of target behaviors.
RESULTS duced similar results with both sharing and
Training ~~~~~~~praising. Mean numbers of shares and praises
Troining increased to levels well above those obtained dur-
The results of the training sessions are shown ing baseline conditions. After the initial acquisiin
Figure 3. Reinforcement of true reports pro- tion period (Sessions 9 to 11), percentages of the
10
8
%OF GROUP REPORTING
_.. %OF GROUP DOING
SHARES(UBJECT ME-AN PRAISESUJEC
PACKAE II MODEL+ SR- TRE
REPORTS OF BEHAVIOR
345
.
.
ANN ROGERS-WARREN and DONALD M. BAER
Table 3
The sequence of experimental conditions in Experiment
II.
Group C
Sharing Praising
Baseline
Package II
Model sharing
Model reports of sharing
Reinforce true reports
of sharing
Package II
Model praising
Model reports of prais-
Baseline ing
Reinforce true reports
of praising
group sharing and reporting averaged above
80%. Percentages of doing and reporting praising
increased rapidly and stabilized at about
80% during the last five sessions of the experiment.
Probes
The results of the generalization probes are
shown in Figure 4. When reinforcement for true
reports began in the training sessions, sharing in
the probe setting gradually increased, and continued
throughout the remaining sessions. Shares
averaged about 0.62 per child before training
and about 1.8 per child after training.
No generalization of praising training was
apparent.
DISCUSSION
The results indicated that both sharing and
praising may be increased by reinforcement of
true reports of those behaviors, and that this
reinforcement contingency was sufficient to establish
correspondence without prior reinforcement
for any (true or untrue) reports of behavior.
A one-time instruction ("Tell me if you
shared [praised)") was used to facilitate acquisition
of reporting during the first two
sessions of each reinforcement condition. It is
possible that reports of sharing and praising
would have occurred eventually, as they did in
Experiment I; however, the brevity of the summer
term in which the experiment was conducted
prevented further exploration of that possibility.
Although the instruction was given after the
work session, immediately preceding the subjects'
opportunity to report, it may have served as a
prompt for actual behaviors, as well as for reports
of the behaviors.
The results of the probe sessions indicated that
sharing generalized to the second setting, but
that praising did not. One explanation for this
difference is that sharing and praising were
naturally followed by different consequences.
Offers to share materials frequently resulted in
the acquisition of another, desired material; however,
praising was typically followed by "thank
GENERALIZATION
SHARING
C-
0
LL
0
AI- zw
C.)
w
0L
PRAISING
100
80
60
40
20
%OFGROPDO
DE&MMEEANNSHPARARIESESS
PER SUBJECT
t TRAININGBEGA IN
USERCn
5 10 15
SESSIONS
m
4 >z
2 Xf I
10tams
160m
8cr
4-
2
u p s- * Io i 20
Fig. 4. Sharing and praising during generalization
probes conducted in Experiment II (Group C). The
percentage of group members sharing and the mean
number of shares per subject are shown on the upper
graph; the percentage of group praising and the mean
number of praises per subject are shown on the lower
graph. The axis on the left refers to percentage of
group doing the target behavior; the right-hand axis
refers to the mean number of target behaviors.
346
TEACHING CHILDREN TO SHARE AND PRAISE
you" or by a reciprocal praise statement. The
acquisition of materials may have been a sufficiently
reinforcing event to maintain sharing,
while the consequences of praising were not. It
is also possible, as suggested earlier, that training
had a more obvious effect on sharing because
sharing was already in the subjects' repertoires
and the training only served to increase it. Praising,
on the other hand, was not initially present
and first needed to be established.
It should also be noted that relatively few
probe sessions followed praising-training. It is
possible that praising would have occurred in the
probe sessions after subjects had acquired a
longer history of reinforcement for praise reports
in the training setting.
EXPERIMENT III
In the first two experiments, subjects' praising
had nearly always taken the form of statements
such as "I like your picture", or "I like what you
are making". This type of praise, while sufficient
to fulfill the criterion for reinforcement, seemed
to be of limited value as a social reinforcer.
Since a primary objective of the research was to
increase appropriate social responses, it was
decided that specific, varied forms of praise
should be trained in addition to the more general
category of praising.
METHOD
Subjects
Twelve subjects were selected from two different
preschool classrooms and divided into two
groups of six. Three members of each group
came from the first classroom, and three from
the second. Group D was composed of three 4-yrold
boys, two 4-yr-old girls, and one 3-yr-old
girl. Group E consisted of four 4-yr-old boys,
one 4-yr-old girl, and one 3-yr-old girl.
Setting
Experimental sessions were carried out in an
experimental room near the subjects' regular
classrooms. Separate sessions were conducted
for each group.
Behavior Definitions
For the purposes of this experiment, praising
was divided into two mutually exclusive categories,
general praising and specific praising.
General praise statements were those consisting
of "I like . . ." and a general object (e.g., "I
like your picture", "I like what you are making",
"I like that"). Any other praise statement was
scored as specific praise. Specific praise, in addition
to indicating liking or admiration, noted
particular qualities (e.g., neatness, number of
colors), particular objects (e.g., circles, lines,
houses), or particular procedures (e.g., drawing
with two pencils, covering the whole page with
colors). Specific praise included statements such
as "I like the way you used the color red", and
"That's a neat house you drew".
Using the same criteria, reports of praise were
scored as either general or specific.
Experimental Design
The experimental design within each group
was a multiple baseline across three behaviors:
sharing, general praising, and specific praising.
There was also an umbrella multiple baseline
across these two groups of subjects.
All experimental conditions were identical to
those described in the General Methods section.
The sequence of experimental conditions is summarized
in Table 4.
RESULTS
Group D
The results for Group D are shown in Figure
5. No effects of the "modelling-only" procedure
were observed. When reinforcement of
true reports of sharing was introduced, sharing
rapidly increased to an average of about seven
shares per subject. Reports of sharing and the
percentage of the group sharing showed corresponding
increases.
347
ANN ROGERS-WARREN and DONALD M. BAER
Table 4
The Sequence of Experimental Conditions in Experiment III
Group D _ Group E
General Specific General Specific
Sharing Praising Praising I I Sharing Praising Praising
I_
Baseline
Model Sharing
Package II
Model sharing
Model reports
of sharing
Reinforce true
reports of
sharing
Baseline
Baseline
Package II
Model gen.
praising
Model reports
of gen.
praising
Reinforce true
reports of
gen. -praising
Baseline
Baseline
Package I
Model spec.
praising
Model reports
of spec.
praising
Reinforce any
reports of
spec.
praising
Package II
Model spec.
praising
Model reports
of spec.
praising
Reinforce true
reports of
spec.
praising
Baseline
Model Sharing
Package II
Model sharing
Model reports
of sharing
Reinforce true
reports of
sharing
Baseline
Baseline
Package II
Model gen.
praising
Model reports
of gen.
praising
Reinforce true
reports of
gen.
praising
Baseline
Package I
Model spec.
praising
Model reports
of spec.
praising
Reinforce any
reports of
spec.
praising
Package II
Model spec.
praising
Model reports
of spec.
praising
Reinforce true
reports of
spec.
praising
I I
-1
I L. -i I I
348
TEACHING CHILDREN TO SHARE AND PRAISE
GROUP D
GENERAL PRAISE *
100 Baseline
80
0
SPECIFIC PRAISE
100 d % OF GF
So -0 %OF GI
-M MEAN S
60 P
PACKAGE I: MODI
40 REPC
PACKAGE U: MODI
20 REPC
n _ Bas4
ROUF
ROUI
;HAR
'RAIC
MEL+ '
)RTS
EL +
)RTS
Beline
WM&s
*-%% @@ @@ @@ @@- ------------* *@ s ----------*-*-@@-@ I- - - - -- -
Package I *Package HI
REPORT ING ..
P DOING
IES SUBJECT
6
3R+ ANY
OF BEHAVIOR4
SR+ TRUE
OF BEHAVIOR
SESSIOIS (2 day blocks)
Fig. 5. The percentage of Group D members performing the target behavior, reporting that behavior, and
the mean number of behaviors per subject for sharing (top graph), general praising (middle graph), and specific
praising (lower graph). The axis on the left refers to percentage of group doing and reporting the target
behavior; the right-hand axis refers to the mean number of target behaviors.
General praising increased less dramatically
when true reports were reinforced, averaging
about 1.8 praises per subject. Doing and reporting
general praise were variable, ranging from
50% to 100% of the group, but generally
showed a high degree of correspondence.
Because of the complexity of the verbal response
for specific praise, any report of specific
praise was initially reinforced (Package I). This
allowed subjects to contact the reinforcement
contingency more quickly than they might have
during Package II (reinforcement for true re-
0
Ua:-l
0
w
0
z
w
cwG
IL.-
m
z
D
m
I
-U
m
Cc
m
C)
-I
01 AL -.AL Ala-.A, -Alk
349
ANN ROGERS-WARREN and DONALD M. BAER
ports), and provided an opportunity for the
experimenter to prompt subjects to make different
reports of specific praise without interfering
with the "truth" requirement.
Only small increases in the mean number of
specific praises were noted when any report of
specific praise was reinforced (Package I). Reports
of specific praise increased moderately;
however, fewer than half the subjects actually
gave specific praise.
Reinforcement for true reports of specific
praise (Package II) produced consistent specific
praising, averaging about 1.7 praises per subject.
Reports of specific praise and the percentage of
the group specifically praising varied during the
first part of the condition, but increased to near
100% for the last four sessions. During the application
of Packages I and II to specific praise,
subjects continued to praise generally and reported
both general and specific praise.
Group E
The results for Group E are shown in Figure
6. Generally, the results for Group E are similar
to those for Group D for sharing and general
praising. Modelling appropriate sharing did not
produce significant changes in the sharing baseline.
When reinforcement was made contingent
on true reports, the percentage of group members
reporting the appropriate behaviors rapidly increased
to near 100%. The percentage of group
members actually doing the behaviors also increased,
so that a high level of correspondence
between reports of behavior and actual behavior
developed and was maintained throughout the
reinforcement condition.
When reinforcement was made contingent on
any report of specific praising (Package I), subjects
immediately began doing and reporting
specific praise. The percentage of group members
specifically praising reached 100% by the third
session of the condition. General praising and
reports of general praising declined to near baseline
levels.
During the final condition (reinforcement for
true reports of specific praise), the mean number
of specific praises per subject averaged about 2.0
while the percentages of reporting and doing
the behavior remained about 80%. Functionally,
this final condition was no different from the
preceding one for the subjects; in both conditions
they truly reported specific praise and were rewarded
for their reports.
DISCUSSION
The results supported the findings of the two
previous experiments and demonstrated that reinforcement
for true reports of behavior can be
used to increase a still more complex behavior,
specific praising.
Specific praise appeared to be a more reinforcing
event than general praise had been in the
same experiment or in earlier ones. General
praise was often "recited" by one child to another,
then repeated by the second child. Often
the child being praised did not attend to the general
praise statement a peer was making. Specific
praise statements seemed more spontaneous and
less stereotypical, possibly because they required
the subject making the statement to stop working,
assess other's activities, and then make an
appropriate statement. Because such statements
were of longer duration and of more unique
content than the general praise statements, specific
praise more often elicited attention from the
child being praised.
Most subjects demonstrated several different
types of specific praise and tended to vary their
praise statements within and across sessions.
None of the subjects imitated the modelled
praise statements, but rather, subjects generated
specific praise that was appropriate to a peer's
particular activity.
During the reinforcement of true reports of
general praising, some subjects in Group E began
praising specifically. These same subjects performed
and reported general praising and, thus,
were rewarded for their reports of praising. This
reinforcement may have maintained both specific
and general praising.
It is not clear whether it was necessary to train
general praise, or the use of a single praise state-
350
TEACHING CHILDREN TO SHARE AND PRAISE
GROUP E
SHARING
BL
GENERAL PRAISE
Baseline
SPECIFIC
100
80
60
40
20
PRAISE
Baseline
PACKAGE I: MODEL+ SR+AN R
OF BEHI
PACKAGE II:MODEL+SR+ TRUI
REPORTS OF BEH)
A A A A . .A . -M.--
5
PKG.
EKG 10
IEPORTS
MVOR
E
WVIOR .6
4
02
15 25
SESSIONS (2 day blocks)
35
Fig. 6. The percentage of Group E members performing the target behavior, reporting that behavior, and
the mean number of behaviors per subject for sharing (top graph), general praising (middle graph), and
specific praising (lower graph). The axis on the left refers to percentage of group doing and reporting the
target behavior; the right-hand axis refers to the mean number of target behaviors.
a.
0
z
cc
0
w
C-)
m
z
m
I
0
m11
m
Cc)r
C-
351
-A AANN
ROGERS-WARREN and DONALD M. BAER
ment, before training the class of specific praise.
The results obtained with Group E suggest that
specific praising might be trained immediately
with the desired results. It may be that long
periods of reinforcement for general praise,
without the occurrence of specific praise, impede
the acquisition of the specific praise responses
and reports. The subjects in Group E who exhibited
specific praising during the general praising
training (and thus were not exposed to a
long period of reinforcement for only general
praise) reported specific praise more quickly
and more truly than the subjects in Group D,
who did not exhibit any specific praise until that
behavior was trained. In Group E, reinforcement
for any reports of specific praise quickly produced
the actual behavior from nearly all subjects;
yet, Group D showed very few instances
of specific praising when any report of that
behavior was reinforced, and somewhat lower
levels of performance during reinforcement for
true reports of specific praise. Additional research
would be needed to determine if this
effect is the function of a long period of reinforcement
for general praising without any occurrence
of specific praise, or simply reflects the
difference in the subject's initial verb repertoires.
GENERAL DISCUSSION
The experiments reported here extend the
previous research on verbal self-control of behavior
along two dimensions: from simple to
complex behaviors, and from asocial to social behaviors.
Previous research has been limited in its application
to nonsocial behaviors such as bar
pressing (e.g., Bem, 1967; Birch, 1966), eating
(Lovaas, 1964b), and the use of preschool materials
(e.g., Risley and Hart, 1968). Although
the target behaviors have been varied, all were
primarily motor responses that did not require
the cooperation or presence of persons other than
the subject. The two behaviors examined here,
sharing and praising, could occur only with one
or more peers present, and thus increases in
either behavior served to further social interaction
among group members.
There are other ways in which sharing and
praising are more complex responses than those
previously studied. Sharing has both verbal and
nonverbal components. In particular, the nonverbal
component may take a variety of forms
(e.g., a gift, an exchange of materials, acceptance
of a material, or simultaneous use of a material).
Although general praising could consist of a
simple verbal statement without accompanying
responses, specific praise required that a subject
assess a peer's activity, attend to a specific aspect
of it, choose a verbal description that indicated
this aspect, and make a praise statement specifically
mentioning what was being praised.
Generally, similar results were obtained in
all three studies; however, there were considerable
differences in the rates of behaviors and the
time needed for subjects to acquire these behaviors.
There are a number of possible reasons
for these differences. First, the composition of the
groups differed with respect to the ages and skills
of the subjects. The subjects who acquired the
reporting and actual behaviors more quickly appeared
to be more verbally and socially skilled.
Studies by Luria (1961) and Bem (1967) reported
that younger children consistently exhibited
little verbal control of behavior without
special training. With the exception of only one
subject, this seemed to be true in all three studies
reported here. Younger subjects required longer
training to make reports of their behaviors and
to increase their actual behaviors.
Differences among groups might also be attributed
to the fact that in some groups, one or
two subjects served as informal, unprogrammed,
peer models for appropriate behavior. The rates
of sharing, praising, and the reporting of these
behaviors may have been affected by the presence
and relative effectiveness of such interactive
models. Apparent models were subjects with unusually
good verbal and social skills, who interacted
with most other members of the group. On
several occasions, model-subjects were observed
instructing peers about the appropriate behaviors
352
TEACHING CHILDREN TO SHARE AND PRAISE
and explaining the contingencies of reinforcement.
Such subject-models were apparent in
Groups A, B, and E, but not in Groups C and D.
While these experiments clearly demonstrated
that the package (modelling and reinforcement
for true reports) can be used to increase social
behaviors, three issues in particular, are left unresolved.
First, the function of the reporting
responses as an example of verbal mediation is
not clear. It is possible that the reinforcement
for reports of behavior paradigm used here was
simply functioning as a delayed reinforcement
procedure, and that children who were rewarded
for sharing and praising after each session would
respond in the same way as children who were
rewarded after each session without reporting
their behavior. A further study, in which delayed
reinforcement was directly compared with reinforcement
for reports of behavior, would aid in
this respect.
The effects of the adult model should also be
more carefully evaluated. In most instances, the
modelling of appropriate sharing responses without
reinforcement for reports of behavior did not
have a significant effect on rates of sharing; however,
the effect of the model after it was paired
with the reinforcement contingencies was not
investigated. The model seemed to be particularly
important during the initial sessions of an
experimental condition, possibly because it provided
the subjects with information about the
reinforcement contingencies, in addition to providing
examples of the appropriate behavior.
In most cases, subjects did not appear to be attending
to the model after the first five or six
sessions of the condition.
A third issue that merits further consideration
is generalization, since, potentially, one of the
most important aspects of this procedure is its
range of control over time and settings. Generalization
data were collected throughout Experiments
II and III. In Experiment II, there was
clear generalization of sharing but not praising,
to a second setting. In Experiment III, limited
generalization of sharing was observed, but no
generalization of praising was evident. The generalization
data for Experiment III have not been
included here because of difficulties related to
their collection.
In Experiment III, a probing technique (a
5-min observation of the subjects while they
played with materials similar to those used in
training) was used in place of the 15-min generalization
sessions employed in Experiment II.
It is highly questionable if the 5-min probe
was sufficient to detect generalization. The variable
results, particularly in the case of sharing,
in Experiments II and III, lead to the conclusion
that the conditions under which generalization
data are collected should be specified as clearly as
the primary experimental procedures. It may be
that conflicting reports of generalization are a
product of different procedures and settings for
collecting generalization data, and, thus, do not
necessarily reflect the strength of the primary
training or intervention.
It was of interest in the present studies that
although formal evidence of generalization was
limited, anecdotal reports from parents and
teachers of subjects in all five groups suggested
that sharing and praising began to occur more
frequently in classrooms and at home after training
was initiated in the experimental sessions.
The present studies did not attempt to consider
the theoretical issues underlying the verbal selfcontrol
of behavior. It should be noted that these
studies in some ways resemble Festinger's studies
of cognitive dissonance (e.g., Festinger, 1957).
For example, Festinger and Carlsmith (1959)
reported that when subjects were rewarded for
publicly advocating an opinion discrepant from
their private ones, a subsequent change in the
direction of the opinion publicly advocated was
noted. The change was considered to be an afterthe-
fact way of justifying the discrepant public
stance and thus alleviating the cognitive dissonance.
In the present studies, there were some
instances where subjects began sharing and praising
after reinforcement of untrue reports (that is,
during conditions when any report, true or untrue,
was reinforced), similar to the effects noted
in the Festinger and Carlsmith study. 353
354 ANN ROGERS-WARREN and DONALD M. BAER
Further research might take two directions. The component procedures discussed above might be further evaluated and tested for their relative functions. The present package also might be extended toward further application. This self-reporting procedure might also be an
efficient technique for maintaining social behavior in the classroom. Once true reporting has been established, the teacher might need only to be present during the reporting period. The teacher's attention during other times may be directed toward teaching other behaviors. The
delayed contingencies also allow for increasing social behavior without interrupting ongoing behavior among the subjects.
REFERENCES
Bem, S. L. Verbal self-control: the establishment of effective self-instruction. Journal of Experimental Psychology, 1967, 74, 485-491.
Birch, D. Verbal control of nonverbal behavior. Journal of Experimental Child Psychology, 1966, 4, 266-275.
Israel, A. C. and O'Leary, K. D. Developing correspondence between children's words and deeds. Child Development, 1973, 44, 575-581.
Festinger, L. A theory of cognitive dissonance. Evanston, Ill.: Row, Peterson, 1957.
Festinger, L. and Carlsmith, J. Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology, 1959, 58, 203- 210.
Lovaas, 0. I. Interaction between verbal and nonverbal behavior. Child Development, 1961, 32, 329-336.
Lovaas, 0. I. Cue properties of words: the control of operant responding by rate and content of verbal operants. Child Development, 1964, 35, 245-256. (a)
Lovaas, 0. I. Control of food intake in children by reinforcement of relevant verbal operants. Journal of Abnormal and Social Psychology, 1964, 18, 672-678. (b)
Luria, A. R. The role of speech in the regulation of normal and abnormal behavior. New York: Pergamon Press, 1961.
Meichenbaum, D. H. and Goodman, J. Teaching children to talk to themselves: a means of developing self-control. Journal of Abnormal Psychology, 1973, 27, 115-126.
Meichenbaum, D. H. and Goodman, J. The developmental control of operant motor responding by verbal operants. Journal of Experimental Child Psychology, 1969, 7, 553-565.
Monohan, J. and O'Leary, K. D. Effects of selfinstruction on rule-breaking behavior. Psychological Reports, 1971, 29, 1059-1066.
O'Leary, K. D. The effects of self-instruction on immoral behavior. Journal of Experimental Child Psychology, 1968, 6, 297-301.
Risley, T. and Hart, B. Developing correspondence between nonverbal and verbal behavior of preschool children. Journal of Applied Behavior Analysis, 1968, 1, 267-281.
Sherman, J. A. Modification of nonverbal behavior through reinforcement of related verbal behavior. Child Development, 1964, 35, 717-723. Received 21 April 1975. (Final acceptance 24 November 1975.)
Fonte:JOURNAL OF APPLIED BEHAVIOR ANALYSIS, 1976, Nº 3and Hart, B. Developing correspondence between nonverbal and verbal behavior of preschool children. Journal of Applied Behavior Analysis, 1968, 1, 267-281.
Sherman, J. A. Modification of nonverbal behavior through reinforcement of related verbal behavior. Child Development, 1964, 35, 717-723. Received 21 April 1975. (Final acceptance 24 November 1975.)
Fonte:JOURNAL OF APPLIED BEHAVIOR ANALYSIS, 1976, Nº 3