JOURNAL OF APPLIED BEHAVIOR ANALYSIS

1985, 189, 343-351

NUMBER4 (wiNm 1985)

A SCATTER PLOT FOR IDENTIFYING STIMULUS CONTROL OF

PROBLEM BEHAVIOR
PAUL E. TOUCHETTE, REBECCA F. MACDONALD, AND SusAN N. LANGER
THE EFFICACY RESEARCH INSTITUTE

Line graphs that average response frequency over long periods obscure the major rate changes that
indicate sources of behavioral control. A scatter plot can make patterns of responding identifiable
and, in turn, suggest environmental features that occasion undesirable behavior. Use of scatter
diagrams is illustrated in three cases.
DESCRIPTORS: data analysis, stimulus control, antecedents, self-injury, aggression

tually never occurs at a steady rate throughout the

waking hours (Gaylord-Ross, Weeks, & Lipner,
1980; Plummer, Baer, & LeBlanc, 1977). Instead,
there are periods of high rate responding and other
periods of no responding in the course of each day.
The usual method of plotting response frequency
in a line graph, however, reveals only the average
rate, not the pattern of responding. A line graph
with days or weeks on the ordinate averages responding over time so that important changes in
rate within each day are obscured. For this reason,
line graphs usually do not suggest probable sources
of environmental control. Severe problem behavior
may assume only two practically important rate
values, zero and unacceptable. A scatter plot can
display periods during which problem responding
virtually never occurs or occurs with near certainty.
Dramatic, infrequent stimulus-response relations are often easily identified. A child may engage in disruptive behavior when left alone for a
lengthy period or if a favorite toy is misplaced. If
the problem behavior is rare and the conditions
that control it are also rare, the obvious corresponThis research was supported in part by grant HD-15781 dence suggests a valid functional relation. The need
from the National Institute of Child Health and Human
for more formal pattern analysis like a scatter plot
Development.
We express our appreciation to the Directors of the Effi- arises when the target behavior is frequent, and
cacy Research Institute, Vincent Stmlly and Katherine Foster, informal observations do not suggest a reliable corfor their unflagging support, to Judith Cunniff for her assistance in collecting the data, and to James Halle and Becky respondence with anything in particular.
This data analysis approach was developed in
Marholin for helpful reviews of earlier drafts of this paper.
Paul E. Touchette is now at the University of California, the context of residential programs for autistic adIrvine. Reprints may be obtained from him at Department
of Pediatrics, U.C.I. Medical Center, 101 The City Drive, olescents. Our examples are drawn from those enOrange, California 92668 or from the other authors at E.R.I., vironments. The approach could, however, be
569 Salem End Road, Framingham, Massachusetts 01701. readily adapted to any setting where sufficient data

It is customary to try to identify the stimuli that

set the occasion for troublesome responses (Carr,
Newsom, & Binkoff, 1976; Favell & Greene,
1981; Gaylord-Ross, 1980; Iwata, Dorsey, Slifer,
Bauman, & Richman, 1982; Patterson, 1974).
Attempts to isolate such stimuli, however, often
fail. The array of possible controlling stimuli in
any applied setting may be too broad to test. Research has shown that the relevant environmental
features may be neither simple nor intuitively obvious (Rincover & Koegel, 1975). Thus, difficulties discovering functional controlling relations can
render unworkable the elegant strategy of eliminating problem behavior by altering the stimuli
that control it.
A scatter plot can help in identifying patterns
of responding in natural settings. Ferster and Skinner (1957) pointed out that overall response rates
were usually made up of ". . . short bursts of responding at a constant local rate alternating with
periods of no responding" (p. 27). Severe problem
behavior typically conforms to this pattern. It vir-

343

344

PAUL E. TOUCHETTE et al.

Starting Date_

Client_

10---11

C/)
-J

cr

'-I,-Lw

4
O D
z
Li.
_2
OH

__

-_

__

-_

K-

_-

--

12

21_--

_ _-__________I
-_

3-

4 __
5 -_
6
__

-.0

__--_ __--T

8 _9 -1IIn--

Ih:::z

111Il

__

_ I __

[-

I L Iz lzIzI
--

_I_

Successive Days
Figure 1. Scatter plot grid with a key at the top to indicate response frequencies corresponding to filled, slashed, and
open boxes. Each location on the grid identifies a unique time interval on a given day.

are available induding home-based, self-manage- selves to easy interpretation as the following exment, community, industrial, and educational pro- amples show.
grams. The data necessary for a scatter plot are
those typically collected in applied settings: freGENERAL PROCEDURE
quency counts in half-hour or shorter intervals.
The first step is to design a grid like that in
Minimal training is required to generate this visual
display. The simplicity of this vehide enables the Figure 1. Vertically, the grid segments time of day
observer to record directly on a scatter chart or into hours, half hours, quarter hours, or any time
transfer observed frequencies to it without calcu- unit appropriate to the duration of elements of the
lation. The visual patterns that result lend them- client's schedule. Horizontal segments of the grid

SCATTER PLOT
represent successive days. When data are plotted,
a blank cell is used to represent a 0 rate. Filled
cells are used to indicate that the problem behavior
occurred during that interval. In some cases it is
useful to break the occurrence category into high
and low rates. The lower frequency can be recorded
as a slash or an outlined cell. We have found that
more than three categories can result in charts that
are difficult to interpret.
The frequency differential sufficient to produce
a cell change must reflect the character of the behavior problem under consideration. For many serious behavior problems the only acceptable rate is
0. A low rate category can have practical significance even when the eventual target is complete
elimination of the behavior. The occurrence of a
lower than usual frequency may constitute an important approximation and suggest the course of
program development when there are few or no 0
intervals in the baseline. As with all visual data
presentations, plotting parameters must be selected
so as to ensure that the display is both accurate
and easily interpreted.
As the scatter chart is filled with observations,
each unit of time contains a designation indicating
whether the behavior occurred at a high, low or 0
rate. An initial evaluation of variables associated
with high, low, and 0 rates is then obtained by
visual inspection of the data array. A pattern,
should one exist, can emerge as soon as several
days are plotted. When observations are recorded
for 7 successive days, the atypical character of the
weekend can add further insights as in Case 1,
below.
Problem behavior may be highly correlated with
a time of day, the presence or absence of certain
people, a social setting, a class of activities, a contingency of reinforcement, a physical environment,
and combinations of these and other variables. Very
likely, several probable sources of control are confounded. As the following case examples show, it
may not be necessary to isolate the specific controlling variable(s) to eliminate the problem behavior. The three cases presented below illustrate
some uses of scatter plots. In each case the scatter
diagram facilitated identification of a relationship

345

between problem behavior and one or more environmental features. Altering the client's schedule,
activities, setting, or instructors reduced or eliminated the target response(s).
The three adolescents in these illustrations had
all been diagnosed as autistic or autistic-like (Rutter, 1978) and moderately to severely retarded.
Etiologies were unknown, but their current health
was excellent. Examination by a physician revealed
no pain sources, seizure activity, or other reason to
suspect a physiological basis for self-injurious or
assaultive behavior.
Interobserver Agreement
The data reported here were collected by program staff in the course of their daily responsibilities. Reliability observations were recorded by other program staff and us. The extended period of
time (12 or more hours a day) covered by these
scatter plots could present a special problem in
estimating interobserver agreement. Fortunately, the
responses to be tracked were few in number, grossly deviant, and readily identifiable. Reliability of
response recording was evaluated in all three cases
both during baseline and subsequent to intervention.
Interventions resulted in long periods with no occurrence of the target behavior. Interobserver
agreement was calculated using the following
formula: agreements/agreements + disagreements x 100. An interval was scored as an agreement if both observers recorded the target behavior
as having occurred at the same rate category (0,
low, or high). In Cases 1 and 3, interobserver
agreement was evaluated for half-hour intervals;
but in Case 2 reliability assessments were based on
5 minute intervals. Reliability checks resulted in
100% agreement between observers in all instances.
CASE ILLUSTRATION 1

This case illustrates the use of a scatter diagram

as a tool for discovering aspects of a therapeutic

program that provoked problem behavior and those

that made it unlikely. The target behavior was
assaulting people. Joan was a 14-year-old girl who

346

PAUL E. TOUCHETTE et al.

I JOAN
100

Initial
Evaluation

Revised
Program

12-Month
Follow-up

0 = Number of Assaults
*N=Number of 30 min.
Intervals Containing
\
Assaults
I
I

80

-J
cn

Cn)
C/)

60

II

II

II

II

II

40

I
I
I
I
I
I

20

I
I

I
I
I

_7
I~~
I
0 _
1

5 56 57

WEEKS
Figure 2. Conventional plot of Joan's assaults that indudes both raw frequency of incidents, and number of 30minute intervals containing an incident. A total of 182 intervals per week were recorded.

had a history of serious aggression, first noted when

she was only 4 years old. Her assaults had injured
staff and peers. Her outbursts were disruptive and
necessitated her removal from educational activities. This young woman lived in a residential school
for autistic adolescents where she had been placed
3 years earlier when her behavior at home became
uncontrollable. Joan was functionally mute, and
she had a Leiter mental age estimate of 5 years.
A scatter plot was used to identify environmental features associated with assaults which were
defined as two hits, kicks, or head butts directed
toward staff or peers within 10 seconds or a single
hit to the face or object thrown at someone. Data
were collected throughout Joan's waking hours,
both weekdays and weekends. A time-out procedure was in effect for assaults throughout the period reported. It proved ineffective but was continued to protect other students.
The total number of incidents of violence ranged
from 53 to 82 per week during the initial 3 weeks

of observation. Assaults occurred in 21 to 36 of

the 182 half-hour intervals. Treatment team records indicated that assaults had been in this same
range for months. Prior to the development of the
scatter plot, a variety of interventions had failed to
reduce the rate of assaults. A line graph of Joan's
data during baseline is on the left side of Figure
2. The raw number of episodes and number of
intervals in which episodes occurred are well correlated, but neither suggests controlling variables.
Figure 3 presents Joan's data in a scatter plot.
The pattern of occurrence has several revealing features. The grid shows that assaults were most frequent between 1:00 and 4:00, Monday through
Thursday. Equally important is the pattern of periods when assaults did not occur. Assaults were
least likely on weekdays in the early morning, at
lunchtime, and after 4:00 p.m. Assaults were rare
all day Friday, Saturday, and Sunday.
Incidents were most likely during the time Joan
attended group prevocational and community living dasses, 1:00-4:00 on weekdays. On Friday
afternoons, her schedule was different; it included
field trips or swimming. Joan's weekday mornings
were spent in one-to-one instruction, suggesting
that demands and training alone did not provoke
assaults. Uneventful group activities in the community and at the gym suggested that simply being
in a group was not the provocation.
Joan's activity schedule was revised on the basis
of the scatter pattern. Conditions associated with
filled cells were replaced with conditions associated
with empty cells. The first hour in the morning,
lunch, supper, evenings, and weekends were unchanged because they were unlikely times for assaults. Joan continued to receive tutorial instruction in the morning. Afternoon requirements for
group participation were eliminated. A new afternoon schedule was arranged to approximate the
staff interactions that Joan had on weekends and
during the evening. Staff aides were assigned to
work with Joan between 12:30 and 4:30 p.m.
They engaged her in informal activities which induded listening to stories, trying on cosmetics, and
playing with stickers. Activities were changed every
15 minutes. Each interaction continued until Joan

347

SCATTER PLOT

[JOAN]

*>1 0=1
REVISED PROGRAM

INITIAL EVALUATION

12 MONTH FOLLOW-UP
0

00

0o:

~~00

1
c

o z

O.w

ILWX
0'

WZ

I2
FE

0 0

12

1 0
000

21

0
0

*00
000
@00

300
3 O
4

@000

000
@000

Ad
2000

@0000
*000
0

0 0

@0

0
0

6
7

00

M W

,00
O
,

M W

F 21

,,,,

M W

14

M W

F 14

SUCCESSIVE DAYS

Figure 3. Scatter plot of Joan's assaults. Filled cirdes indicate 30-minute intervals during which more than one assault
occurred. Open boxes represent intervals with only one assault.

inattentive for 15 seconds or motioned the

teacher away.
When the revised schedule was initiated, assaults decreased immediately as indicated in Figures 2 and 3. At first, Joan's afternoon time was
spent outside the dassroom area in the hallway or
sitting on the stairs. In the first week, interactions
were brief and occupied only 100 of a possible
960 minutes. Over 30 days, they increased to more
than 400 minutes per week.
Elements of Joan's original program, previously
associated with assaults, were then gradually reintroduced into her afternoon schedule. In the fifth
week, Joan was scheduled to be present in a dassroom for 15 minutes of each hour. During the
third and fourth months, demands were increased,
and Joan engaged in dassroom activities in 15minute blocks with each of the several teachers on
her team. The duration of teaching sessions was
then increased in 2-minute increments based on
her successful performance with all of her teachers.
By the fifth month, she was spending 30 minutes
was

with each teacher and had been reintroduced for

brief periods into her original group instruction
setting. The duration of this activity was also increased in 2-minute increments based on her performance. There was no increase in assaultive behavior throughout.
After 12 months, Joan was participating in
group dasses during 3 of the 4 hours between 1:00
and 5:00 p.m. and in structured individual activities during the fourth hour. Figures 2 and 3 show
that only one assault occurred during a 14-day
period 1 year after baseline. During the follow-up
period, data on single hits and any minor assaultive
acts in which Joan had previously engaged (e.g.,
pinching) were also recorded. One single hit occurred during the 14-day period and no other topographies of assaultive behavior were observed.
The scatter plot identified a pattern in Joan's
aggression. The behavior was prevalent at certain
times of day on certain days of the week. These
times were correlated with one element of Joan's
program. Eliminating that element drastically re-

348

PAUL E. TOUCHETTE et al.

EBMl.-1

0>2 0=2
INI Follow-up

Initial
Rever ,sal
Evaluation
Re-reversal
.ME
-

* 0

WU

cni 12

01

-0

,i

00
0

keeping, watching television, riding

LLW

WZ

I
-

000
000
0000
a00
@000

000'
@00
0000
000

0000

0000

0000

000

0000

000
0000
ri16 1 1 I
-li-r --ml
I I I Irr -176-T
If
I

10

was also required to place his hands behind his

back for approximately 10 seconds following each
self-hit.
Two aides worked with Tom on a one-to-one
basis. The first person worked from 9:00 a.m. to
3:00 p.m., the second from 3:00 p.m. to 9:00
p.m. Tom's activities included self-care, house-

15

WEEKDAYS
Figure 4. Scatter plot of Tom's self-hits. Filled cirdes
indicate intervals during which three or more Ihits occurred.
Open squares indicate intervals during which
curred. On the second day of the reversal, Tot m
the house at a physician's appointment betweeen 11:00 and
12:30.

was ocO

duced assaults. Joan's original trainingg program

then successfully reintroduced by E;radual ap-

was

proximations.

CASE ILLUSTRATION 2
In this case a scatter diagram was use d to isolate
a confounded variable that controlled seIf-injurious
responding. Tom was 23 years old and lived with
other autistic adults in a community res;idence. He
had been hospitalized with recurrent sej ?tisemia resuiting from self-inflicted face and hanid wounds.
His irreparable detached retina was thc iught to be
secondary to head hitting. Tom wore a helmet and
face guard to prevent further injury.
Self-abuse was defined as a blow to any part of
the body or striking any part of the b()dy against
a solid object. A small quantity of snac:k food was
given to Tom following each 5-minute interval in
which self-abuse did not occur (DRO 5i') throughout the period covered by the scatter plot. Tom

an

exercycle,

and table-top prevocational tasks. The first and

second shift each had a unique array of activities.
Each activity was selected based on Tom's functional level to make success likely and to minimize
the probability of provoking self-abuse.
Data were gathered from 10:00 a.m. to 8:30
p.m. and recorded on a scatter chart (Figure 4).
Filled circles indicated that Tom struck himself
three or more times during the 30-minute interval.
Two hits occurred in each interval with an open
square. Blank areas indicated the absence of abusive behavior or one isolated hit. Figure 4 shows
that self-abuse occurred principally during the late
afternoon and evening. Activities, schedule, and
staff changed at 3:00 p.m. In addition, Tom's
peers returned from school at approximately that
time. Increased self-hitting corresponded to an increase in the number of peers present, the staff
person assigned, and the afternoon activities. All
three elements changed at about the time Tom's
self-abuse escalated each day.
One element was altered to determine if it controlled the rate of self-hitting. The morning aide
worked during the late afternoon and the afternoon aide worked during the morning. Other suspected elements were held constant. Activities were
carried out at the usual times and Tom's peers
returned home on schedule. Figure 4 shows that
rescheduling the aides displaced self-abuse from
late afternoon and evening to morning and early
afternoon. After 2 days, the aides resumed their
original schedule and Tom's self-hitting again occurred predominantly in the late afternoon and
early evening. No dear explanation why one person was so successful and one so unsuccessful was
evident. The difference was something subtle which
neither they nor we could identify. The group home
managers subsequently solved the problem when

SCATTER PLOT

349

|JIM |
R0 -F

Q)
-J

9
10

11*
12

z
IL W

1*
2

0I.-

3.
4
5-

6-

FIRST SCHEDULE REVISION

INITIAL EVALUATION

000 0 0@0
0000 *006
00000 SeQ
0.... 000
0
000 00006
* 0
0006
00000 0006
* 000 00 oooooooooooooo
*
0 0.0.06
00 0
@06
00 06
00
000. 000
0 00
000
000 @0000
@00 .0..
0 00 0
0
0
0
00 000 000
0
00
000000
00
* @0 @000
0
O0
0000 0000
0
-

0>4 0=1-4
SECOND SCHEDULE REVISION

i~~~~~~~~~~~~~~~1
d-1.
0

oooooloooooooooo

000060

0000 0006
0
0000

7a A 00
H0-

0000 --IA

0
0

- - - -

10

II

15

0000

0
0
0
0
00
00
000
0
-i,
-Do
- - 00
- -000
I
I
I

0
000

20

25

30

_
- L-

- -iL
-iL
- -iL
I
I

35

SUCCESSIVE DAYS

40

n1
I

III I

45

50

l l

55

Figure 5. Scatter plot of Jim's self-hits. Filled cirdes indicate intervals during which five or more hits occurred. Open
indicate intervals during which one to four hits occurred. Jim slapped his face once ritualistically prior to lunch,
and evening snack. These three daily responses account for the long horizontal patterns of open squares. The
responses were recorded but otherwise ignored. They subsequently dropped out, but not as the result of any specific
squares
supper,

manipulation.

Tom was accepted in a well-run sheltered workshop. With mornings and early afternoon spent at
the workshop, the aide associated with low rates
of self-hitting was rescheduled to work with Tom
in the late afternoon and evening.

CASE ILUSTRATION 3
This case demonstrates that uninterpretable
scatter plot data may reflect an unstable environment. Jim was a self-abusive student at a residential school for autistic adolescents. He hit his face
with his -dosed fist and slammed his head into
walls or furniture from 50 to 1,600 times a day.
To prevent injury, he wore a protective helmet and
mask He had a staff person in dose proximity
around the dock and participated in few activities
with his peers. Jim was 15 years old and his estimated MA (PPVT) was 3 years, 2 months. He
was mute, except for a few words uttered only in
a language training context. Jim had been placed
in a residential program at age 9 when his behavior
at home had become unmanageable.
Self-abuse was defined as any blow to his head
with his hand or bringing his head into contact
with a solid object. During the initial data collec-

period, Jim engaged in a loose schedule of

activities that staff thought to be reinforcing (e.g.,
making hot chocolate, watching TV, taking short
walks on the school grounds, listening to music,
and making popcorn). During the first 5 days in
Figure 5, self-abuse produced a 5-minute period
during which Jim was required to mop vigorously,
an activity that he appeared to dislike. During the
next 5 days in Figure 5, each self-hit produced a
water mist from a spray bottle directed toward
Jim's face, another consequence that he appeared
to dislike. The low-demand/high-reinforcement
density schedule combined with mild aversive consequences did not reduce daily response frequency.
The scatter plot in Figure 5 shows 30-minute
periods with five or more self-hits as filled cirdes
and those with one to four hits as open squares.
There are many intervals during which Jim hit
himself interspersed with periods of up to several
hours free of self-abuse. Although many of Jim's
waking hours were free of self-abuse, no stable
pattern emerged. What Jim did, when, where, and
with whom were not stable either. His program
induded offering him several choices on a menu
board every 30 minutes and asking him to indicate
tion

350

PAUL E. TOUCHETTE et al.

which he preferred next. Each activity was continued for 30 minutes or until Jim would no longer
engage in the current one. He was so difficult to
supervise that staff members rotated responsibility
for his supervision every few hours. In essence, the
only pattern that could appear would indicate that
time of day alone influenced self-abuse.
A new fixed schedule of activities was designed.
Jim's team sorted setting-activity combinations into
five categories. A "5" was something virtually
guaranteed to provoke self-hitting. A " 1" was most
likely to produce acceptable behavior. The new
schedule induded necessary activities such as dressing, eating, toothbrushing, and bathing. All other
times were filled with activity-setting combinations
rated "1," "2," or "3." Figure 5 shows that the
new schedule alone reduced the number of severe
self-hitting intervals.
Residual problem behavior formed a pattern. It
occurred almost exclusively in the evening. Staff
rotated shifts so the frequency of responding was
not tied to people. Jim's self-hitting in the evening
appeared to be related to some aspect of his evening activity schedule. His evening program was
revised to closely resemble his morning program,
which was now virtually problem free. This involved altering the sequence and duration of activities as well as their location. With this second
schedule revision, the first of many abuse-free days
occurred (Fig. 5, Day 42).

DISCUSSION
Stimulus control of behavior is most often addressed as a problem. Desirable behavior taught in
one context often fails to generalize to another
(Marholin & Touchette, 1979; Stokes & Baer,
1977). This phenomenon is so ubiquitous that it
is assumed by widely used multiple baseline research designs (Barlow & Hersen, 1984, p. 210;
Kazdin, 1982, p. 134). Undesirable behavior is
functionally identical to desirable behavior. It too
can be probable under one set of circumstances
and unlikely in others. Patterns of behavior evident
on a scatter plot can suggest stimuli that control
problem behavior. It may then be possible to re-

duce undesirable responding by eliminating stimuli

that control high rates or by putting in place stimuli that control low rates of problem behavior (Jenkins, 1965; Rilling, 1977; Terrace, 1966b).
Stimulus conditions in school, home, or work
environments vary with the time of day and day
of the week. Each change in location, activity, contingency of reinforcement, or social setting creates
an opportunity to evaluate the impact of these
variables. When a change in setting systematically
corresponds to a change in behavior, this suggests
a controlling relation. Unfortunately, the process
of identifying controlling relations precisely can be
time consuming in both laboratory (Ray & Sidman, 1970; Terrace, 1966b) and field (Charlop,
Schreibman, Mason, & Vesey, 1981; Patterson,
1974; Schroeder, Rojahn, & Mulick, 1978). Precise identification of controlling relations may not
be necessary. An alternative approach is to identify
broadly defined control and to pursue a functional
analysis in detail only if it proves essential to achieve
the desired end.
A scatter diagram has inherently low resolution
as a rate reporting device. Each cell can assume
only two or three values corresponding to gross
changes in rate within an observation interval. Patterns of presence and absence may, however, be
sufficient to suggest control exerted by setting
events. Shifts in rate of responding from high to
low or low to high, corresponding to a change in
setting, do not just suggest stimulus control, they
define it (Terrace, 1966a). An issue for extensive
future investigation is the level of refinement of
that definition necessary to achieve applied goals.
Controlling relations in the cases above were at
best vaguely specified. It appears, however, that
identifying functional control can be of practical
value even though the controlling stimuli are not
precisely defined. The three cases suggest that it is
sometimes possible to eliminate problem behavior
without understanding the variables that control it
in detail. More exact analyses would definitely require an allocation of resources difficult to justify
in most applied settings.
The data reported here serve only to illustrate
the assessment method. They demonstrate how the

SCATTER PLOT

scatter plot can be used as an assessment tool.

Confounds and weak experimental evidence severely restrict any conclusions that might be drawn
concerning the interventions that were implemented. This plotting method, however, does offer insights into patterns of responding not readily available from graphs of daily or weekly frequency.
Questions remain to be resolved concerning the
construction and use of scatter diagrams. It is not
yet dear how best to select the values for filled and
empty cells. Neither is it dear whether these diagrams are useful beyond an initial assessment. The
production of this form of visual display may pay
valuable dividends, and it makes minimal demands on staff time or skill. Many behavior analysts already record response frequency in 5-, 15-,
or 30-minute blocks. For them, developing a scatter diagram simply means looking at available data
in a different format.
REFERENCES
Barlow, D. H. & Hersen, M. (1984). Single case experimental designs. New York: Pergamon.
Carr, E. G., Newsom, C. D., & Binkoff, J. A. (1976).
Stimulus control of self-destructive behavior in a psychotic child. Journal of Abnormal Child Psychology, 4,
139-152.
Charlop, M., Schreibman, L., Mason, J., & Vesey, W.
(1981). Behavior-setting interactions of autistic children: A behavioral mapping approach to assessing dassroom behaviors. Analysis and Intervention in Developmental Disabilities, 4, 359-373.
Favell, J. E., & Greene, J. W. (1981). How to treat selfinjurious behavior. Lawrence, KS: H & H Enterprises.
Ferster, C. B., & Skinner, B. F. (1957). Schedules of reinforcement. New York: Appleton-Century-Crofts.
Gaylord-Ross, R. J. (1980). A decision model for the
treatment of aberrant behavior in applied settings. In
W. Sailor, B. Wilcox, & L. Brown (Eds.), Methods of
instruction for severely handicapped students (pp. 135158). Baltimore: Paul H. Brooks.
Gaylord-Ross, R. J., Weeks, M., & Lipner, C. (1980). An
analysis of antecedent, response and consequence events
in the treatment of self-injurious behavior. Education
and Treatment of the Mentally Retarded, 15, 35-42.

351

Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E.,

& Richman, G. S. (1982). Toward a functional analysis of self-injury. Analysis and Intervention in Developmental Disabilities, 2, 3-20.
Jenkins, H. M. (1965). Generalization gradients and the
concept of inhibition. In D. I. Mostofsky (Ed.), Stimulus generalization (p. 55-61). Stanford: Stanford University Press.
Kazdin, A. E. (1982). Single-case research designs:
Methods for clinical and applied settings. New York:
Oxford University Press.
Marholin, D., II, & Touchette, P. E. (1979). Stimulus
control and response consequences. In A. Goldstein &
F. Kanfer (Eds.), Maximizing treatment gains: Transfer enhancement in psychotherapy (p. 303-351). New
York: Academic Press.
Patterson, G. R. (1974). A basis for identifying stimuli
which control behaviors in natural settings. Child Development, 45, 900-911.
Plummer, S., Baer, D. M., & LeBlanc, J. M. (1977).
Functional considerations in the use of procedural timeout and an effective alternative. Journal of Applied

Behavior Analysis, 10, 689-705.

Ray, B. A., & Sidman, M. (1970). Reinforcement schedules and stimulus control. In W. N. Schoenfeld (Ed.),
The theory of reinforcement schedules (pp. 187-214).
New York: Appleton-Century-Crofts.
Rifling, M. (1977). Stimulus control and inhibitory processes. In W. K. Honig & J. E. R. Staddon (Eds.)

Handbook of operant behavior (pp. 432-480). Englewood Cliffs, NJ: Prentice-Hall.

Rincover, A., & Koegel, R. L. (1975). Setting generality
and stimulus control in autistic children. Journal of Applied Behavior Analysis, 8, 235-246.
Rutter, R. (1978). Diagnosis and definition of childhood
autism. Journal of Autism and Childhood Schizophrenia, 8, 139-161.
Schroeder, S. R., Rojahn,J., & Mulick,J. A. (1978). Ecobehavioral organization of developmental day care for
the chronically self-injurious. Journal of Pediatric Psychology, 3, 81-88.
Stokes, T. F., & Baer, D. M. (1977). An implicit technology of generalization. Journal of Applied Behavior
Analysis, 10, 349-367.
Terrace, H. S. (1966a). Stimulus control. In W. K. Honig
(Ed.), Operant behavior: Areas of research and application (p. 271-344). New York: Appleton-CenturyCrofts.
Terrace, H. S. (1966b). Discrimination learning and inhibition. Science, 154, 1677-1680.
Received January 2, 1985
Final acceptance July 8, 1985