Reduction of stimulus overselectivity with nonverbal differential observing responses

Abstract

Three individuals with mental retardation exhibited stimulus overselectivity in a delayed matching-to-sample task in which two sample stimuli were displayed on each trial. Intermediate accuracy scores indicated that participants could match one of the samples but not both of them. Accuracy in a baseline condition was compared to accuracy with a differential observing response procedure. This procedure prompted participants to make simultaneous identity-matching responses that required observation and discrimination of both sample stimuli. These observing responses were never followed by differential consequences. When observing responses were prompted, participants' accuracy scores improved. In a return to the baseline condition, when differential observing responses were no longer prompted, accuracy returned to intermediate levels. The results show that stimulus overselectivity can be greatly reduced by a behavioral intervention that controls observing behavior and verifies discrimination, but that exposure to such procedures alone may be insufficient for lasting benefits.

Full text

25
JOURNAL OF APPLIED BEHAVIOR ANALYSIS
1999, 32, 25–33
NUMBER
1(
SPRING
1999)
REDUCTION OF STIMULUS OVERSELECTIVITY WITH
NONVERBAL DIFFERENTIAL OBSERVING RESPONSES
W
ILLIAM
V. D
UBE AND
W
ILLIAM
J. M
C
I
LVANE
E. K. SHRIVER CENTER FOR MENTAL RETARDATION
AND NORTHEASTERN UNIVERSITY
Three individuals with mental retardation exhibited stimulus overselectivity in a delayed
matching-to-sample task in which two sample stimuli were displayed on each trial. In-
termediate accuracy scores indicated that participants could match one of the samples
but not both of them. Accuracy in a baseline condition was compared to accuracy with
a differential observing response procedure. This procedure prompted participants to
make simultaneous identity-matching responses that required observation and discrimi-
nation of both sample stimuli. These observing responses were never followed by differ-
ential consequences. When observing responses were prompted, participants’ accuracy
scores improved. In a return to the baseline condition, when differential observing re-
sponses were no longer prompted, accuracy returned to intermediate levels. The results
show that stimulus overselectivity can be greatly reduced by a behavioral intervention
that controls observing behavior and verifies discrimination, but that exposure to such
procedures alone may be insufficient for lasting benefits.
DESCRIPTORS: stimulus overselectivity, differential observing responses, matching
to sample, mentally retarded
This paper reports a bridge study (Wack-
er, 1996), a basic analogue experiment that
addresses a socially relevant problem known
as stimulus overselectivity or restricted stimulus
control. Overselectivity refers to atypically
limited learning with respect to range,
breadth, or number of stimuli or stimulus
features (Lovaas, Koegel, & Schreibman,
1979). It is shown when training with mul-
tiple stimuli results in stimulus control by
an atypically limited subset of those stimuli.
Overselectivity is a widely acknowledged
problem in the education of individuals with
developmental disabilities like mental retar-
Data collection and manuscript preparation were
supported by NICHD Grant HD 25995. For their
help with data collection or analysis, we thank Lyn
Balsamo, Kathy Clark, Kevin Farren, Tom Fowler,
Kristin Lombard, Alison McVay, Jennifer Mlocek,
Nora Murphy, and Aimee Smith. For their coopera-
tion, we also thank The Protestant Guild Learning
Center, The New England Center for Children, and
The Behavior Intervention Project.
Address correspondence to William V. Dube, Psy-
chological Sciences Division, E. K. Shriver Center,
200 Trapelo Road, Waltham, Massachusetts 02254 (E-
mail: wdube@shriver.org).
dation and autism (e.g., Allen & Fuqua,
1985; Bickel, Richmond, Bell, & Brown,
1986). Although the problem is widespread,
the remediation literature is remarkably
small (for a recent summary, see Schreib-
man, 1997).
Overselectivity may affect performances
on matching-to-sample tasks in special ed-
ucation classrooms. Matching to sample is
widely used in such settings to teach stim-
ulus–stimulus relations among spoken and
printed words, objects, pictures, and some-
times symbols used in augmentative and al-
ternative communication systems (e.g., de
Rose, de Souza, & Hanna, 1996; Stromer,
Mackay, & Stoddard, 1992). For example,
consider a student who is learning to iden-
tify her printed name. If discriminative con-
trol by the name were restricted to the initial
letter only, the student may discriminate
SUE from FAY or BOB, but she would fail
to discriminate SUE from SAM reliably (cf.
Birnie-Selwyn & Guerin, 1997).
For several years, our laboratory has been
studying the problem of overselectivity using

26 WILLIAM V. DUBE and WILLIAM J. MCILVANE
an analogue task, delayed matching to sam-
ple (DMTS) with multiple sample stimuli
(e.g., Stromer, McIlvane, Dube, & Mackay,
1993). On each trial, two sample stimuli are
presented (e.g., AB). The samples remain
available for observation until the partici-
pant touches them; then they disappear and
the comparison stimuli are presented im-
mediately (i.e., with a 0-s delay between
sample offset and comparison onset). The
comparisons are three individual stimuli,
one of which is identical to one of the sam-
ple stimuli (e.g., A, C, and D). Touching the
identical comparison is a correct response.
In this two-sample DMTS procedure,
both sample stimuli have an equal probabil-
ity of appearing as the correct comparison.
Thus, during the sample observation period,
the participant cannot predict which one of
the sample stimuli will be the correct com-
parison. High accuracy (
.
90% correct) in-
dicates no overselectivity with two sample
stimuli. At the other extreme, accuracy at or
near chance levels (33%) indicates an overall
failure to perform the matching task.
Chance-level accuracy scores provide no in-
formation relevant to the evaluation of over-
selectivity. Intermediate accuracy scores (e.g.,
approximately 67%) indicate overselectivity.
The participant was able to match only one
of the two sample stimuli, as follows: On
those trials in which that stimulus appeared
as a comparison (half of the trials in the ses-
sion), the participant was always correct. On
the remaining trials, performance was at
chance levels. The intermediate accuracy
score for the entire session results from av-
eraging scores from both types of trials (for
a detailed analysis, see Dube & McIlvane,
1997).
In the standard DMTS task, the sample
observation period continues until the par-
ticipant touches the sample display area.
Thus, the response to the samples (touching)
is the same for all stimuli. A procedural
modification that may improve discrimina-
tion is to require differential observing re-
sponses (DOR), that is, a different response
for each sample stimulus (e.g., Cohen, Bra-
dy, & Lowry, 1981; Urcuioli & Callender,
1989). For example, Constantine and Sid-
man (1975) found that DMTS accuracy
with pictures improved in young men with
severe mental retardation when they were re-
quired to name the sample pictures (cf. Ger-
en, Stromer, & Mackay, 1997). Gutowski,
Geren, Stromer, and Mackay (1995) showed
that stimulus overselectivity was reduced in
2 individuals with moderate mental retar-
dation when they were required to name the
stimuli (pictures of common objects) in a
two-sample DMTS task.
Naming, however, is not always available
as a differential observing response. Some
students in special education classes may be
unable to produce spoken or gestural names,
and naming is not an option with unfamiliar
stimuli. The present study examined a gen-
eralized, nonverbal differential observing re-
sponse procedure. This procedure prompted
participants to make simultaneous identity-
matching responses during the sample ob-
servation period.
METHOD
Participants
Three individuals with mental retardation
participated. Pseudonyms, ages, standardized
test scores, and diagnostic information from
records are shown in Table 1. Prior to this
experiment, all had been (a) assessed for
snack-food preferences, (b) trained to ex-
change plastic poker-chip tokens for foods,
and (c) given an assessment of matching-to-
sample skills (Dube, Iennaco, & McIlvane,
1993). Dawn and Ellen, who could perform
generalized identity matching with high ac-
curacy, were next given additional matching-
to-sample pretests for this study. Bob, who
was unable to perform accurate identity
matching, was first trained to do so (with

27DIFFERENTIAL OBSERVING RESPONSES
Table 1
Participant Characteristics
Partici-
pant CA
a
PPVT
b
EOWP
VT
c
Diagnostic
description
d
Bob
Dawn
Ellen
15
19
13
2.4
6.4
2.0
2.9
7.7
2.0
PDD
e
PDD, moderate MR
f
MR
a
Chronological age.
b
Peabody Picture Vocabulary Test–Revised age-equivalent
score.
c
Gardner Expressive One-Word Picture Vocabulary Test
(Revised) age-equivalent score.
d
From students’ records.
e
Pervasive developmental disorder.
f
Mental retardation.
Figure 1. Matching-to-sample procedures. Illustrations show participants making correct responses. Sr
1
indicates that the immediately preceding response would be followed by a token reinforcer.
the sorting-to-matching procedures reported
in Serna, Dube, & McIlvane, 1997) and was
then given the pretests for this study.
General Procedure
Experimental sessions of 10- to 15-min
duration were conducted 3 or 4 days per
week in a quiet area at the participants’
schools. Participants sat before a Macintosh
Plus
t
computer with a 9-in. black-and-white
video display and a touch-sensitive screen.
The computer controlled stimulus presen-
tation, response recording, and data collec-
tion.
The stimuli were black, nonrepresentative
forms (see Figure 1 for examples), approxi-
mately 1 cm by 1.5 cm, displayed on a white

28 WILLIAM V. DUBE and WILLIAM J. MCILVANE
background. Within each session, different
stimuli appeared on every trial. The stimuli
for each session were drawn at random,
without replacement, from a pool of 180
forms.
During 3-s intertrial intervals (ITIs), the
display screen was blank. Trials that ended
with correct responses were followed by a
brief auditory-visual computer display and a
token presented by the experimenter. Trials
that ended with errors were followed only
by the ITI.
Tokens were exchanged for snack foods
after sessions. Participants bought small
quantities of foods by placing the tokens on
a ‘‘price tag,’’ a piece of cardboard with 10
token-sized circles. The participant placed a
token on each circle, and then the experi-
menter presented the food and collected the
tokens; this process was repeated until no
tokens remained (if a price tag was only par-
tially filled for the last purchase, the quantity
of food was reduced accordingly).
Matching-to-Sample Pretests
One-sample simultaneous matching to sam-
ple. Trials began when a sample stimulus ap-
peared in the center of the screen. When the
participant touched the sample, three com-
parison stimuli appeared in three corners of
the screen and the sample remained dis-
played. One comparison was identical to the
sample, and touching it was the correct re-
sponse. Touching either of the nonidentical
comparisons was an error.
One-sample delayed matching to sample
(DMTS). Trials were the same as one-sample
simultaneous matching, except that the sam-
ple stimulus disappeared when the partici-
pant touched it. The comparison stimuli ap-
peared immediately (i.e., 0-s delay).
Two-sample simultaneous matching. Trials
were similar to one-sample simultaneous
matching, except that two sample stimuli
were presented on each trial. The samples
were displayed side by side, 1.75 cm center
to center. The comparisons were single stim-
uli, and the correct comparison was identical
to one of the samples. Over trials, stimuli in
the left and right sample positions were cor-
rect equally often.
Two-sample DMTS. Trials were the same
as two-sample simultaneous matching, ex-
cept that the sample stimuli disappeared
when the participant touched them. (The
procedure is illustrated in the left column of
Figure 1.) The delay interval from sample
offset to comparison onset was 0 s.
Compound simultaneous matching pretests.
Trials were the same as two-sample simul-
taneous matching, except that the compari-
son display consisted of three pairs of stim-
uli. (The procedure is illustrated in the cen-
ter column of Figure 1.) The correct pair of
comparisons matched the sample stimuli ex-
actly. Each of the two pairs of incorrect com-
parisons had one stimulus that matched one
of the samples and one that did not match.
The matching stimulus was on the left in
one pair of incorrect comparisons and on the
right in the other pair.
Pretest outcomes. Every pretest session con-
sisted of 36 matching-to-sample trials. Par-
ticipants were given successive sessions of
each type of pretest until performance was
stable. The criterion for stability was three
consecutive sessions with accuracy scores
greater than 50%, in which individual ses-
sions scores did not deviate from the three-
session mean by more than 10%. There was
one exception: Ellen was given only one
two-sample simultaneous matching test be-
cause of a scheduling error. Table 2 shows
mean accuracy scores for the last three ses-
sions of each type of pretest (given Ellen’s
exception). In every case in which the num-
ber of sessions is greater than three, accuracy
scores were initially lower and improved
with practice. As Table 2 shows, all partici-
pants had scores of at least 98% for both
one-sample tasks, at least 94% for the two-
sample simultaneous matching task, and in-

29DIFFERENTIAL OBSERVING RESPONSES
Table 2
Pretest Accuracy Scores (Percentage Correct)
Matching task
Participants
Bob Dawn Ellen
One-sample simultaneous
One-sample delayed
Two-sample simultaneous
Two-sample delayed
Compound simultaneous
99 (3)
98 (3)
96 (3)
69 (3)
97 (3)
98 (3)
100 (3)
94 (3)
66 (9)
91 (12)
99 (5)
98 (3)
97 (1)
71 (3)
99 (7)
Note. Each score is the mean for the final three sessions for
each type of matching task (except for Ellen’s two-sample
simultaneous matching task). Numbers in parentheses show
the total number of test sessions. See text for details.
termediate accuracy scores ranging from
66% to 71% for the delayed two-sample
task.
All participants also had high accuracy
scores for compound simultaneous matching
pretests. When initially low scores for Dawn
did not improve with practice within three
sessions, she was given sessions that began
with the two-sample simultaneous matching
task and then switched to the compound si-
multaneous matching task at some point in
the session. Initially, sessions began with 18
two-sample trials, and this number was re-
duced to zero over six sessions.
Immediately after the pretests, partici-
pants were adapted to intermittent reinforce-
ment for compound simultaneous matching.
They were given sessions consisting of com-
pound simultaneous matching trials only (as
in the pretest), and the reinforcement sched-
ule was gradually thinned until only one re-
sponse in six (on average) was followed by
differential consequences (variable-ratio
[VR] 6).
Experimental Conditions
The investigation was designed to evalu-
ate the effects of a DOR procedure on two-
sample DMTS. An initial baseline condition
measured DMTS accuracy. Then, the DOR
procedure (described below) was evaluated,
followed by a return to the baseline condi-
tion.
Baseline. Baseline conditions consisted of
six sessions of two-sample DMTS. There
were 36 trials in each session.
Compound DOR. The compound DOR
procedure embedded a compound simulta-
neous matching trial within the sample-ob-
servation period of the DMTS trial, as
shown in the right column of Figure 1. The
first touch to the sample display area pro-
duced a comparison display consisting of
three pairs of stimuli, and the samples re-
mained displayed. As in the compound si-
multaneous matching pretest, the correct
pair of comparisons both matched the sam-
ple stimuli, and each of the two pairs of in-
correct comparisons had only one stimulus
that matched the samples (cf. Allen & Fu-
qua, 1985; Schreibman, Charlop, & Koegel,
1982). Thus, consistently correct responding
verified discrimination of both sample stim-
uli; comparison selections that were based
on only one of the samples would be correct
on only half of the trials on average.
On each trial, both the correct and the
incorrect comparison stimuli for the final
DMTS response were stimuli that had been
presented as comparisons during the DOR.
In this way, the participant’s final response
could not be based on relative novelty, be-
cause all of the comparison stimuli had been
presented earlier in that trial.
In contrast to the procedure during com-
pound simultaneous matching pretests, dif-
ferential observing responses on DMTS tri-
als were never followed by differential rein-
forcement or any kind of informative feed-
back. The trial continued as shown in Figure
1 regardless of which comparison pair was
selected. In this way, the comparison stimuli
presented at the end of the trial had no im-
mediate reinforcement histories and thus no
previously established positive or negative
stimulus functions. For clarity and conve-
nience, the task will be referred to as com-
pound simultaneous matching when selections
of comparison stimuli produced differential

30 WILLIAM V. DUBE and WILLIAM J. MCILVANE
Figure 2. Individual-session accuracy scores. Sessions with one point are baseline DMTS sessions with no
DOR procedures (the first six sessions and last six sessions for each participant). Sessions with two points are
from the compound DOR condition. The open triangles show accuracy for the DOR portions of the trials
(the intermediate response in the right column of Figure 1), and the filled points show accuracy for the final
DMTS response. The open square in Dawn’s plot shows accuracy for one session of compound simultaneous
matching.
consequences (e.g., in the pretests) and as
compound DOR when the task was embed-
ded in a DMTS trial and never produced
differential consequences.
So that responding to DOR displays
would not be extinguished because of non-
reinforcement, sessions included a small
number of compound simultaneous match-
ing trials that ended with differential con-
sequences. Sessions in the compound DOR
condition consisted of 42 trials: 36 DMTS
trials with the compound DOR plus six
compound simultaneous matching trials.
The compound DOR condition continued
for 10 sessions, with one exception, ex-
plained with the results.
RESULTS
Session-by-session accuracy scores for in-
dividual participants are shown in Figure 2.
Ellen’s baseline scores were slightly higher in
the first baseline condition (M
5
79%) than
in her two-sample DMTS pretest (71%).
This increase in accuracy may have resulted
from her recent history with the compound
simultaneous matching procedure during
the final pretest and the subsequent sessions

31DIFFERENTIAL OBSERVING RESPONSES
that introduced intermittent reinforcement.
Accurate compound simultaneous matching
required observing both members of pairs of
stimuli. In the baseline condition that fol-
lowed, she apparently continued to observe
both stimuli during the sample observation
period on some, but not all, of the two-sam-
ple DMTS trials.
In the first session with the compound
DOR procedure, accuracy on the DOR por-
tions of the trials was at least 90% for Bob
and Ellen and 86% for Dawn. High accu-
racy was expected because of the compound
simultaneous matching pretest results.
Dawn’s accuracy on the compound DOR
portions of the trials, however, indicated that
the procedure did not set the occasion for
reliable discrimination of both sample stim-
uli. To recover her compound matching per-
formance, she was given one review session
consisting of 36 compound simultaneous
matching trials with the VR 6 reinforcement
schedule. Her accuracy was 94% in this ses-
sion (open square point in Figure 2), and
the DOR procedure was resumed in the fol-
lowing session.
Figure 2 shows that the compound DOR
procedure produced clear and immediate in-
creases in DMTS accuracy for Dawn and
Ellen and a modest increase in accuracy for
Bob. In the baseline condition that followed
the compound DOR condition, accuracy re-
turned to previous intermediate levels for all
participants.
DISCUSSION
The results show that stimulus overselectiv-
ity can be greatly reduced with a DOR pro-
cedure that controls observing behavior and
sets the occasion for discrimination of all stim-
uli. Although overselectivity was greatly re-
duced by requiring DORs, the improvement
did not continue after the procedure was dis-
continued. That is, mere exposure to the
DOR procedure was not sufficient to elimi-
nate overselectivity when observing responses
were no longer prompted. One important
goal for further research is to find ways to
teach students to observe multiple stimuli
when there are no DOR requirements in the
same way that they do when there are such
requirements. In general, behavior analysts are
challenged to understand why certain individ-
uals do not display or maintain behavior of
which they are manifestly capable when doing
so would increase the frequency of reinforcers
obtained (cf. Constantine & Sidman, 1975).
One possibility is that the individual is not
sensitive to the difference between the lower
and higher reinforcement frequencies. A relat-
ed possibility concerns our token exchange
procedure: Participants accumulated tokens
throughout the session and then exchanged
them all after the session. For example, a par-
ticipant had 24 tokens after a baseline session
with 67% accuracy and 39 tokens after a
DOR session with 92% accuracy on DMTS
trials (33 of 36 on DOR-DMTS trials and 6
of 6 on compound simultaneous matching tri-
als); 24 versus 39 tokens could have been a
difficult discrimination for some of our par-
ticipants. One possibility for further study of
reinforcement variables is to investigate with-
in-session token exchanges whenever a fixed
number have been earned.
As we noted earlier, this paper is intended
to be a bridge study. We used arbitrary forms
as stimuli to minimize the influence of preex-
perimental experience. Previous experimenta-
tion has shown, for example, that prior rein-
forcement histories can determine which spe-
cific stimuli gain control in cases of overselec-
tivity (Dube & McIlvane, 1997). Also, the use
of unfamiliar stimuli reduced the possibility
that a participant might produce names for
the stimuli; if he or she routinely named the
stimuli, the task would no longer be strictly a
measure of visual stimulus control.
Matters of experimental control notwith-
standing, it is appropriate to ask how our
findings might influence practices in applied

32 WILLIAM V. DUBE and WILLIAM J. MCILVANE
settings. Our results show clearly that non-
verbal DOR procedures that capitalize on
generalized identity-matching repertoires can
be effective in reducing stimulus overselec-
tivity. It seems reasonable to anticipate that
procedures modeled after those we have de-
scribed would prove useful with everyday ac-
ademic stimuli. For example, consider train-
ing to teach matching relations between pic-
tures of clothing items and the correspond-
ing printed words. In cases in which sample
stimuli are pictures, nonverbal compound
DORs like those described in this paper
could help to eliminate overselective stimu-
lus control by isolated features of the pic-
tures (e.g., for the DOR, a sample picture
of brown loafers, with comparison pictures
of brown loafers, black loafers, and brown
oxfords; cf. Burke & Cerniglia, 1990; Ro-
senblatt, Bloom, & Koegel, 1995; Schreib-
man, 1997). When the sample stimuli are
printed words, nonverbal DORs could be
programmed by inserting printed-word
identity-matching trials in which the incor-
rect comparison stimuli have letters in com-
mon with the sample (e.g., sample HAT,
with comparison stimuli HAT, CAT, HOT,
and HAD). Another option for nonverbal
DORs with printed words is constructed-re-
sponse matching to sample, in which the let-
ters that make up the word are matched in-
dividually and sequentially (Dube, Mc-
Donald, McIlvane, & Mackay, 1991; Mack-
ay, 1985). All of these procedures can be
implemented either on the computer or the
tabletop, and they may find wide application
with individuals who might otherwise re-
spond overselectively.
REFERENCES
Allen, K. D., & Fuqua, R. W. (1985). Eliminating
selective stimulus control: A comparison of two
procedures for teaching mentally retarded children
to respond to compound stimuli. Journal of Ex-
perimental Child Psychology, 39, 55–71.
Bickel, W. K., Richmond, G., Bell, J., & Brown, K.
(1986). A microanalysis of the controlling stim-
ulus-response relations engendered during the as-
sessment of stimulus overselectivity. The Psycholog-
ical Record, 36, 225–238.
Birnie-Selwyn, B., & Guerin, B. (1997). Teaching
children to spell: Decreasing consonant cluster er-
rors by eliminating selective stimulus control.
Journal of Applied Behavior Analysis, 30, 69–91.
Burke, J. C., & Cerniglia, L. (1990). Stimulus com-
plexity and autistic children’s responsivity: Assess-
ing and training a pivotal behavior. Journal of Au-
tism and Developmental Disorders, 20, 233–253.
Cohen, L. R., Brady, J., & Lowry, M. (1981). The
role of differential responding in matching-to-
sample and delayed matching performance. In M.
L. Commons & J. A. Nevin (Eds.), Quantitative
analysis of behavior: Vol. 1. Discriminative proper-
ties of reinforcement schedules (pp. 345–364). Cam-
bridge, MA: Ballinger.
Constantine, B., & Sidman, M. (1975). The role of
naming in delayed matching to sample. American
Journal of Mental Deficiency, 79, 680–689.
de Rose, J. C., de Souza, D. G., & Hanna, E. S.
(1996). Teaching reading and spelling: Exclusion
and stimulus equivalence. Journal of Applied Be-
havior Analysis, 29, 451–469.
Dube, W. V., Iennaco, F. M., & McIlvane, W. J.
(1993). Generalized identity matching to sample
of two-dimensional forms in individuals with in-
tellectual disabilities. Research in Developmental
Disabilities, 14, 457–477.
Dube, W. V., McDonald, S. J., McIlvane, W. J., &
Mackay, H. A. (1991). Constructed-response
matching to sample and spelling instruction. Jour-
nal of Applied Behavior Analysis, 24, 305–317.
Dube, W. V., & McIlvane, W. J. (1997). Reinforcer
frequency and restricted stimulus control. Journal
of the Experimental Analysis of Behavior, 68, 303–
316.
Geren, M. A., Stromer, R., & Mackay, H. A. (1997).
Picture naming, matching to sample, and head
injury: A stimulus control analysis. Journal of Ap-
plied Behavior Analysis, 30, 339–342.
Gutowski, S. J., Geren, M., Stromer, R., & Mackay,
H. A. (1995). Restricted stimulus control in de-
layed matching to complex samples: A preliminary
analysis of the role of naming. Experimental Anal-
ysis of Human Behavior Bulletin, 13, 18–24.
Lovaas, O. I., Koegel, R. L., & Schreibman, L.
(1979). Stimulus overselectivity in autism: A re-
view of research. Psychological Bulletin, 86, 1236–
1254.
Mackay, H. A. (1985). Stimulus equivalence in ru-
dimentary reading and spelling. Analysis and In-
tervention in Developmental Disabilities, 5, 373–
387.
Rosenblatt, J., Bloom, P., & Koegel, R. L. (1995).
Overselective responding. In R. L. Koegel & L.

33DIFFERENTIAL OBSERVING RESPONSES
K. Koegel (Eds.), Teaching children with autism
(pp. 33–42). Baltimore: Brookes.
Schreibman, L. (1997). The study of stimulus control
in autism. In D. M. Baer & E. M. Pinkston
(Eds.), Environment and behavior (pp. 203–209).
Boulder, CO: Westview.
Schreibman, L., Charlop, M. H., & Koegel, R. L.
(1982). Teaching autistic children to use extra-
stimulus prompts. Journal of Experimental Child
Psychology, 33, 475–491.
Serna, R. W., Dube, W. V., & McIlvane, W. J. (1997).
Assessing same/different judgments in individuals
with severe intellectual disabilities: A status report.
Research in Developmental Disabilities, 18, 343–
368.
Stromer, R., Mackay, H. A., & Stoddard, L. T.
(1992). Classroom applications of stimulus equiv-
alence technology. Journal of Behavioral Education,
2, 225–256.
Stromer, R., McIlvane, W. J., Dube, W. V., & Mackay,
H. A. (1993). Assessing control by elements of
complex stimuli in delayed matching to sample.
Journal of the Experimental Analysis of Behavior,
59, 83–102.
Urcuioli, P. J., & Callender, J. (1989). Attentional
enhancement in matching-to-sample: Facilitation
in matching acquisition by sample-discrimination
training. Animal Learning & Behavior, 17, 361–
367.
Wacker, D. P. (1996). Behavior analysis research in
JABA: A need for studies that bridge basic and
applied research. Experimental Analysis of Human
Behavior Bulletin, 14, 11–14.
Received February 23, 1998
Initial editorial decision April 21, 1998
Final acceptance August 24, 1998
Action Editor, Joseph E. Spradlin
STUDY QUESTIONS
1. What is stimulus overselectivity? Give an example illustrating this problem.
2. Describe the procedures involved in a two-sample delayed matching-to-sample (DTMS) task.
Why are intermediate accuracy scores (67%) on such a task suggestive of stimulus overse-
lectivity?
3. What was the purpose of the current study?
4. What results from the pretests indicated that participants’ discriminative responding was not
always overselective?
5. What is an observing response, what is its function, and what was the specific observing
response taught for the two-sample DMTS task?
6. Summarize the results obtained during the experimental conditions. In what sense was the
differential observing response (DOR) generalized and in what sense was it not generalized?
7. What feature of the reinforcement contingencies may have affected the results? Can you
suggest an alternative?
8. Provide an example of how the techniques described in this study might be applied to
teaching situations with individuals with severe disabilities.
Study questions prepared by Gregory P. Hanley and Eileen M. Roscoe, The University of Florida