JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 2007, 87, 325–336 NUMBER 3 (MAY)

DISCRIMINATIVE CONTROL OF PUNISHED STEREOTYPED BEHAVIOR IN HUMANS

SHANNON S. DOUGHTY,1,2 CYNTHIA M. ANDERSON,1 ADAM H. DOUGHTY,3 DEAN C. WILLIAMS,3
3
AND KATHRYN J. SAUNDERS

WEST VIRGINIA UNIVERSITY,1

PARSONS STATE HOSPITAL AND TRAINING CENTER,2 AND
UNIVERSITY OF KANSAS3

The purpose of this experiment was to establish discriminative control of responding by an antecedent
stimulus using differential punishment because the results of past studies on this topic have been
mixed. Three adults with mental retardation who exhibited stereotypy not maintained by social
consequences (i.e., automatic reinforcement) participated. For each subject, stereotypy occurred
frequently in the presence of a stimulus correlated with nonpunishment of stereotypy and rarely, if ever,
in the presence of a stimulus correlated with punishment of stereotypy. Latency measures showed that
the antecedent stimulus correlated with punishment served as the discriminative stimulus for the
suppression of stereotypy. These results are important insofar as they show that discriminative control by
an antecedent stimulus develops with punishment, and because it sometimes may be desirable to
establish such control of socially inappropriate behavior.
Key words: autism, punishment, stimulus control, mental retardation, stereotypy, humans

________________________________________

When punishment effects are demonstrated been little overlap in punishment procedures
in the basic animal laboratory, unconditioned across basic-animal-laboratory and basic-hu-
positive reinforcers, such as food deliveries, man-laboratory studies.
usually maintain the punished response, and The human operant laboratory can serve as
the animals are food deprived (see Baron, an interface between basic research and
1991, for a review). In addition, uncondi- application by examining clinically relevant
tioned punishers usually are used (e.g., elec- behavior under well controlled conditions.
tric shock). In contrast, laboratory studies of Although the human laboratory has demon-
human subjects, which also typically involve strated the suppression of responding main-
arbitrary responses such as button presses, tained by conditioned reinforcers through
maintain these responses by conditioned re- conditioned punishment (e.g., O’Donnell et
inforcers such as points and/or money (e.g., al., 2000), there have been fewer demonstra-
O’Donnell, Crosbie, Williams, & Saunders, tions of punishment of psychologically impor-
2000). Moreover, conditioned punishers, such tant behavior maintained by primary reinforce-
as point loss, typically are used. Thus, there has ment in humans—at least in the basic
literature. This fact may be contrasted with
This research partially fulfilled the first author’s Ph.D. the applied-behavior-analytic literature in
requirements, and was presented at the 2004 meeting of
the Association for Behavior Analysis in Boston, MA. Parts
which punished behavior often is of great
of this research were funded by NIH grant HD18955. The psychological importance, as evidenced by
third author was supported by NIH grant T32 HD 7525. extreme allocations of time and effort to the
Mike Perone, Kent Parker, and Jennifer Margrett made responses. For example, persons with autism,
many contributions to this research, and Laura Cruz,
Colleen Eisenbart, and Brandi Zirk assisted in data
according to diagnostic criteria (American
collection and analysis. Pat White provided editorial Psychiatric Association, 1994), engage in ste-
assistance. Shannon Doughty is now with Carolina Coast reotypic behavior for an amount of time that is
Behavioral Services, Charleston, SC. Cindy Anderson is in excess of typically developing individuals,
now with the University of Oregon, Eugene, OR. Adam
Doughty is now with College of Charleston, Charleston, SC.
and this behavior can interfere with learning,
Address correspondence and reprint requests to Shan- independent functioning, and/or socializa-
non Doughty, Carolina Coast Behavioral Services, P.O. tion. In addition, clinical service and applied
Box 80901, Charleston, SC, 29416 (shannon.doughty@ research must follow ethical guidelines in
comcast.net) or to Dean Williams, University of Kansas,
Parsons Research Center, 2601 Gabriel, Parsons, KS, 67357
considering less aversive interventions and
([email protected]). conditioned punishers (e.g., Miltenberger,
doi: 10.1901/jeab.2007.39-05 1997).

325
326 SHANNON S. DOUGHTY et al.

Despite these and many other procedural effect in one component and a conjoint re-
differences between the basic-animal-labor- inforcement and punishment schedule is in
atory, basic-human-laboratory, and applied- effect in the other component. The stimulus
service-delivery studies involving punishment, correlated with the conjoint-reinforcement–
sweeping generalizations about the effects or punishment-schedule component (e.g., a red
lack of effects of punishment have been made. key) will be called the SDp, hereafter (see
Of primary importance are the notions that O’Donnell, 2001), and the stimulus correlated
punishment produces undesirable side effects with the reinforcement-schedule-only compo-
such as nonselective suppression (e.g., Bolles, nent (e.g., a green key) will be called the SD,
Holtz, Dunn, & Hill, 1980); that stimuli hereafter. If a lower response rate occurs in
correlated with a punishment contingency the presence of the SDp, the conclusion often
become conditioned aversive stimuli them- would be that stimulus control by the red key
selves (e.g., Sidman, 1989); and the ephemeral had been demonstrated (i.e., SDp control).
nature of response suppression when punish- Unless responding is suppressed completely,
ment is discontinued (see Sidman). The however, reduced response rate alone does
generality of these observations has not been not provide conclusive evidence that the SDp is
examined directly. Thus, the generalization of functionally the discriminative stimulus for
these assumed characteristics of punishment response suppression. Response suppression
from preparations using food-reinforced be- may be controlled by the initial punisher
havior and electric shock to situations in- delivery in the red-key component, such that
volving conditioned reinforcement and condi- rates are lower only after the punisher is
tioned punishment is unknown (Dinsmoor, delivered. To rule out this interpretation,
1998). response suppression must be shown in the
One area in which the human and non- presence of the red key, but before or in the
human animal literatures are particularly absence of punisher delivery.
discrepant is discriminative stimulus control In the classic report of discriminated in-
established with differential punishment hibitory control with punishment, Honig and
(Doughty, Doughty, O’Donnell, Saunders, & Slivka (1964) reported the development of
Williams, in press). It usually is accepted that control by SDp following multiple-schedule
discriminative stimulus control may be de- training of differential punishment. In the
veloped using differential punishment in absence of delivery of the punishing stimulus,
a manner similar to that of the discriminative however, there was discriminative control by
stimulus control that develops using differen- key color in only 1 of the 3 pigeons. These data
tial reinforcement (e.g., Baron, 1991). This are not consistent with a clear demonstration
acceptance is based on results from basic of control by SDp (see also Weisman, 1975),
research using nonhuman animal subjects. and illustrate conditions in which the effects of
Recent results from human-laboratory studies, punishment do and do not maintain when the
however, have shown that discriminative stim- contingency is removed. There also have been
ulus control may be more difficult to obtain in diverse results regarding the establishment of
human subjects (e.g., O’Donnell & Crosbie, stimulus control by the SDp via differential
1998; O’Donnell et al., 2000). punishment in the human operant laboratory
A recent review of the basic and applied (O’Donnell & Crosbie, 1998; O’Donnell et al.,
literature involving stimulus control and pun- 2000). Using college-student subjects, a fixed-
ishment revealed a great deal of ambiguity ratio 1 (FR 1) schedule of point loss (i.e.,
regarding the form of stimulus control that response-cost punishment) was added to the
has been established under conditions of existing reinforcement schedule in the pres-
differential punishment (Doughty et al., in ence of one stimulus. Although response rates
press). Thus, the recent human findings may were lowest in the component with punish-
not be as discrepant from the animal findings ment, responding only decreased in that
as they first appeared. The traditional pro- component after the first punisher was de-
cedure for producing and demonstrating livered. Further attempts to establish stimulus
discriminative stimulus control through differ- control by the SDp also failed (see also
ential punishment involves a multiple sched- Birnbrauer, 1968). O’Donnell et al. obtained
ule in which a reinforcement schedule is in stimulus control by the multiple-schedule
 DISCRIMINATIVE CONTROL OF PUNISHED BEHAVIOR 327

stimuli rather than by the delivery of the ic stimulation from the behavior (e.g., Voll-
punisher only after instructions were used. mer, 1994).
Differential responding then was observed, The purpose of the present study was to
demonstrating clear stimulus control by the determine whether the suppression of an
multiple-schedule stimuli rather than by the automatically reinforced response could be
delivery of the punisher (also see Rollings & brought under control of the SDp using
Baumeister, 1981). a conditioned punisher in functionally non-
Given conflicting results in studies by verbal adults with MR. Stereotypy remained
Honig and Slivka (1964), Weisman (1975), unpunished in one multiple-schedule compo-
O’Donnell et al. (2000), Rollings and Baume- nent and was punished in a different compo-
ister (1981), and other studies, questions re- nent. Importantly, in addition to reporting
main regarding the conditions under which overall occurrence of the response (i.e., re-
control by the SDp occurs. The present experi- sponse differentiation), a dependent measure
ment was conducted to determine whether an functionally equivalent to the latency to the
unambiguous demonstration of punished re- first response in each component was used to
sponding controlled by an SDp could be determine whether control by the SDp had
obtained in human subjects whose limited been obtained. That is, if responding were not
verbal repertoires precluded the use of in- suppressed completely in the punishment
structions. In addition, the responses selected component, a much longer latency in that
for punishment in the present study were component as compared to latency in the
maintained by natural contingencies and were unpunished component would show that
psychologically important to the subjects, as suppression was controlled by the SDp and
evidenced by their high probability of occur- not the punisher delivery. Therefore, this
rence (e.g., over 80% of intervals) in nearly latency measure is essential to demonstrating
every context. High-probability responses have control by the SDp in cases where suppression
received attention in the literature, and they is not complete. Without it, the differential
play a central role in some theories of re- response rates otherwise would not clearly
inforcement. For example, wheel running can demonstrate the source of the discriminative
be used to reinforce lever pressing (e.g., control.
Iversen, 1993), and the opportunity to engage
in stereotyped behavior has been used to METHOD
increase appropriate vocalizations (Charlop,
Kurtz, & Casey, 1990). Subjects
Automatically reinforced behavior is of Subjects were 3 adults with severe-to-
particular interest in application precisely profound MR whose stereotypy was shown to
because the maintaining stimulation is diffi- be maintained by sensory stimulation in an
cult or impossible to control. Thus, extinction analog functional analysis (Iwata, Dorsey,
is difficult or impossible to use to suppress Slifer, Bauman, & Richman, 1982/1994; see
the behavior (cf. Rincover, 1978). Stereotyped Appendix A). Subject TS was a 45-year-old
behavior is defined as repetitive, high-rate male with severe MR, profound deafness, and
motor or vocal mannerisms, usually with legal blindness. The precise nature of this
no known functional environmental conse- visual impairment was undetermined due to
quences. Examples include pacing in captive his limitations in communication. Subject TS
lions and twirling objects in front of the eyes in received 5 mg of Loxapine Hydrochloride
persons with autism, mental retardation (MR), twice daily throughout the experiment. Sub-
and mental illness (e.g., Bodfish & Lewis, ject CB was a 40-year-old male diagnosed with
2002; Kennedy, 2002). When severe, these profound MR and Down syndrome; he re-
responses can occupy significant amounts of ceived 1500 mg of Valproic Acid daily for
an individual’s daily activity, and they can be seizures at a constant dose throughout the
difficult to reduce with reinforcement for experiment. Subject PH was a 54-year-old male
alternative, acceptable responses (e.g., Garner, diagnosed with severe MR and bipolar dis-
Meehan, & Mench, 2003; Lovaas, Litrownik, & order; he received 900 mg of Lithium Car-
Mann, 1971). Like wheel running in rats, bonate and 400 mg of Carbamazepine per day
stereotypy often is maintained by the automat- throughout the experiment.
328 SHANNON S. DOUGHTY et al.

Apparatus ination training, below), and these 5-s intervals

Sessions were conducted in a 2.3 m 3 2.9 m were included. The percent of the component
room equipped with stimuli relevant to each (i.e., latency to stereotypy) was measured as
session (see below) and a one-way mirror. The the number of intervals without stereotypy
experimenter was in the room during all following the start of the component. Each
sessions. Video- and audio-recording equip- punisher was recorded such that, for Subjects
ment were in an adjacent room located behind TS and CB, more than one punisher could
the mirror, and all sessions were recorded occur in each 5-s interval.
from this room. Interobserver agreement was assessed in
a minimum of 33% of the sessions in each
Procedure condition. Agreement for stereotypy and pun-
isher delivery was calculated by comparing each
Stereotypies were defined for each individ- observer’s records on an interval-to-interval
ual subject. Subject TS’s stereotypy was hand basis. Occurrence agreement was calculated
and arm flapping, defined as moving his arms by dividing the number of intervals both
or hands through the air repeatedly at or observers agreed that an event occurred by
above waist level by bending at the wrist or the number of intervals either observer scored
elbow. Subject CB’s stereotypy was finger an occurrence. Nonoccurrence agreement was
manipulation. Finger manipulation was de- calculated by dividing the number of intervals
fined as moving two fingers back and forth both coders agreed an event did not occur by
repeatedly at or above waist level while his the number of intervals either coder did not
head was oriented in the direction of his score a response. Total agreement was calcu-
fingers and his eyes were open (i.e., he was lated by dividing the number of intervals
looking at them). Subject PH’s stereotypy was observers agreed the response did and did not
repetitive line drawing, defined as drawing occur by the total number of intervals. All
vertical lines on paper. Punisher delivery was coefficients were multiplied by 100 to obtain
a 1-s (Subjects TS and CB), or 10-s (Subject a percent agreement score. Interobserver
PH), ‘‘hands-down’’ procedure. The 1-s pro- agreement scores are presented in Appendix B.
cedure involved manually guiding a subject’s Each session was 10 min in duration (except
hands down once the response definition was where noted for Subject PH). Two to six
met, preventing further responding at that sessions were conducted per day, with an
moment. For the subsequent 10-s punishment average of 10 min elapsing between sessions
procedure used for PH, after this initial (that occurred in the same day) during which
manual guidance, the experimenter’s hands time the participant’s time was unstructured
remained touching PH’s hands with only and there was casual interaction with the
enough force necessary to prevent the occur- experimenter. Sessions were conducted 5 days
rence of the response (i.e., prevent marker per week with few exceptions, and at approx-
use), but not completely to immobilize his imately the same time each day for each
hands. subject. Unless otherwise noted, each of the
All sessions were videotaped, and human following conditions continued for a minimum
observers recorded stereotypy and punisher of 15 sessions and until responding stabilized.
delivery using a 5-s interval recording pro- Judgments of stability were based on the
cedure. The entire session duration was di- percentage of 5-s intervals with stereotypy
vided into 5-s intervals. The duration of (i.e., occurrence) and the proportion of the
stereotypy was measured as the total number component that elapsed prior to the first
of 5-s intervals in which it was observed during occurrence of stereotypy (i.e., latency). These
any part of the interval (i.e., partial interval measures were judged stable when they did
recording). The 5-s intervals during which the not exhibit a systematically increasing or
response was prevented for Subject PH (i.e., decreasing trend (excluding only an upward
during the punishment procedure) were not trend in baseline or trend in the opposite
included in this analysis. Instead, 10 s were direction of what would be expected in the
added to the duration of the punishment extinction condition for CB), or considerable
component for each punisher delivery that bounce, determined by visual inspection, for at
occurred during the component (see Discrim- least six consecutive sessions. If, in discrimina-
 DISCRIMINATIVE CONTROL OF PUNISHED BEHAVIOR 329

tion training, responding was suppressed in reprimand (as described above) to increase
both components, the discriminative stimuli the effectiveness of punishment. Because of
were changed, and baseline data were collect- the relatively long duration of Subject PH’s
ed with new stimuli. punishment procedure, 10 s were added to the
Baseline. Initially, a two-component multi- duration of the punishment component for
ple schedule was in effect. Each session had each punisher delivery that occurred during
two 5-min components separated by the time it the component. Discrimination training con-
took to change the stimuli (approximately 1 to cluded when the above-described stability
2s). The order of the two components was criteria were met in addition to the following
determined randomly with two restrictions: criterion: the percentage of the component
the same order did not occur more than three elapsed prior to the first interval scored with
consecutive times, and each sequence oc- stereotypy was less than 30% for the no-
curred six times in each 12-session block. punishment component and greater than
The stimuli presented in baseline sessions 70% for the punishment component in the
differed across subjects. For Subjects TS and final six sessions.
PH, the stimuli were black wristbands placed In a subsequent condition, for each sub-
over the wrist of the dominant hand (no- ject, the schedule was changed slightly such
punishment component) versus the absence that each session consisted of four 2.5-min
of that wristband. For Subject CB, the stimuli components. For Subjects TS and CB, this
were two orientations of the swivel chair in change occurred after stimulus control was
which he sat. In the component with punish- demonstrated with 5-min components to
ment, the chair was turned such that the demonstrate further control by the SDp. For
subject sat facing a red wall and the experi- Subject PH, this change occurred in an effort
menter (i.e., facing toward E), and the to promote SDp control by increasing the
component began with the experimenter number of component changes he received
saying, ‘‘I’m here now.’’ In the component in each session. The condition remained in
without punishment, Subject CB faced a blue effect for at least six sessions and until
wall with his back to the experimenter (i.e., responding stabilized.
facing away from E), and the component One subject, CB, was exposed to extinction
began with the experimenter saying ‘‘Alone of the punishment contingency. Punishment
time.’’ These stimuli were individually chosen was removed for 18 sessions (i.e., the only
for each subject after 2 subjects (CB and PH) difference across components was the extero-
experienced failure to discriminate between ceptive stimuli) to determine the necessity of
conditions with previous stimuli (i.e., red/ continuing the punishment contingency after
green cards for CB and PH; music on/off, control by SDp occurred.
wristbands on/off for CB). Throughout base-
line, games and toys were available for Subjects
TS and PH, but not for Subject CB (CB did not RESULTS
have access to items because he remained
seated throughout the session due to limited Table 1 summarizes the conditions pre-
mobility). In addition, a Sesame Street Mup- sented and the number of sessions in each
petsE movie was played for Subject TS, and condition for each subject. The figures show
paper and markers were present for Subject data only for the final (successful) discrimina-
PH. tion training conditions and the baselines
Discrimination training. Stimuli were identi- before those conditions. Table 2 shows the
cal to those stimuli used in baseline. Punish- means and ranges of dependent measures in
ment was implemented in one of the multiple- the last six sessions for baseline and discrim-
schedule components. There were no pro- ination training for each subject. Figure 1
grammed consequences for stereotypy in the shows the occurrence of stereotypy in the final
other component. The punisher for Subjects six sessions of baseline and, in three-session
TS and CB was a 1-s hands-down procedure. blocks, all of the discrimination-training ses-
For Subject PH, the punisher initially was a 1-s sions. Data are shown as three-session means
hands-down procedure with verbal reprimand, for discrimination training because of the
but was changed to a 10-s punisher without large number of sessions.
330 SHANNON S. DOUGHTY et al.

Table 1
Number of sessions and session descriptions.

# of # of
Baseline Training Component
D Dp
Subject S S sessions sessions duration Outcome
TS wristband no band 29 31 5 min Met stability criteria
wristband no band 7 2.5 min Met stability criteria
CB blue wall/ away from E red wall/ toward E 33 29 5 min Met stability criteria
blue wall/ away from E red wall/ toward E 52 2.5 min Met stability criteria
PH wristband no band 21 93 5 min Failure to develop
control by SDp
wristband no band 5 1 min Complete
suppression
wristband no band 3 5 min Failure to develop
control by SDp
wristband no band 20 2.5 min Met stability criteria

Table 2
Mean and range of the results of the last six sessions for baseline and discrimination training.

Subject Condition % Stereotypy % Elapsed

TS Baseline Punishment 60.6 (40–75)
No Punishment 65.3 (36.7–83.3)
Discrim Punishment training 0.6 (0–1.7) 94.5 (70–100)
No Punishment 62.2 (18.3–88.3) 7.8 (0–43.3)
CB Baseline Punishment 66.4 (48.3–86.7)
No Punishment 76.9 (38.3–91.6)
Discrim Punishment training 0.6 (0–3.3) 94.5 (66.7–100)
No Punishment 74.2 (48.3–91.7) 6.7 (0–20)
PH Baseline Punishment 74.5 (70–80)
No Punishment 75 (60–83.3)
Discrim Punishment training 0.9 (0–1.7) 82.2 (60–100)
No Punishment 71.4 (46.7–88.3) 17.8 (3.3–36.7)
Note. The data presented are for the following measures: percentage of intervals scored with a stereotypical response
(% Stereotypy) and percentage of component elapsed prior to first response (% Elapsed), respectively. Means are
presented first in columns, followed by ranges in parentheses.

The top panels of Figure 1 show, for each For all subjects, responding was high and
subject, the percentage of 5-s intervals in which undifferentiated in baseline. Within the first
stereotypy was recorded (i.e., duration1). The three sessions after the punishment contin-
bottom panels show the percentage of the gency was implemented, fewer than 10% of
session that elapsed prior to the first recorded the intervals in the punishment component
response (i.e., latency1). The numbers in each contained stereotypy. Stereotypy remained at
graph are the component duration (in min). high levels in the no-punishment component.
Subject CB showed some suppression in the
1
The percentage of 5-s intervals is presented rather than nonpunished component, but not to the
absolute latency or duration so as to facilitate direct extent in the punished component.
comparison, because the session durations differed across The bottom panels show the latency to the
conditions. Shorter sessions limit the maximum absolute
duration of stereotypy and latency to the duration of the first response (and thus, to the first punisher)
session. The functions for the relative measures and the in the component for those same sessions.
absolute values are identical except for the scales. The Latencies were short and undifferentiated
relative measure of percentage of component elapsed was early in discrimination training. Clearly distin-
used to keep axes constant. Calculations are based on 180,
90, and 36 intervals for the 5-, 2.5-, and 1-min components, guishable discrepancies in latencies developed
respectively. To calculate absolute latencies, consider the after three- and five-session blocks (9 and 15
component durations at the top of Figure 1. sessions) for Subjects TS and CB, respectively.
 DISCRIMINATIVE CONTROL OF PUNISHED BEHAVIOR 331

Fig. 1. Results of discrimination training for each subject. Solid lines separate baseline from discrimination training.
Dotted lines separate conditions within discrimination training. Numbers represent the component duration in minutes.
‘‘Rep’’ in PH’s graph indicates the brief condition where the 1-s hands-down with reprimand was used before it was
changed to a 10-s procedure without reprimand. Unfilled circles show data from the punishment component, and filled
circles show data from the no-punishment component. The left, center, and right panels show data for Subjects TS, CB,
and PH, respectively. The top panels show the percentage of intervals scored with stereotypy. The bottom panels show the
percentage of the component that elapsed prior to the first interval scored with stereotypy. Only the final six sessions are
presented for baseline (top graphs), and all other conditions contain data across the entire condition in three-
session blocks.

For these 2 subjects, the latency data remained in the component previously correlated with
differentiated when the component durations punishment. When the punishment contin-
were shortened to 2.5 min. For Subject PH, gency was reinstated, the previously low levels
however, differential response latencies did of stereotypy returned. The lower graph of
not develop under the 5-min component Figure 2 shows that the latency results were
condition. When component durations were similar to the aforementioned findings. Mean
changed to 10 one-min components per latency was high when the punishment con-
session, responding was suppressed in both tingency initially was in effect, gradually de-
components. There was a brief return to 5-min creased when the contingency was removed,
components, after which the component and increased when it was reinstated.
duration was changed to 2.5 min. After four
blocks with 2.5-min components, differential
DISCUSSION
latencies were established.
Extinction of the punishment contingency Response rates were equivalent across mul-
was conducted for one subject, CB, to probe tiple-schedule components prior to the initia-
the persistence of the discrimination when the tion of punishment. The hands-down pro-
punishment contingency was removed. The cedure was a functional punisher; for all 3
upper graph of Figure 2 shows the occurrence subjects, responding decreased in the punish-
of stereotypy for Subject CB as a function of ment component almost immediately after
the presence or absence of the punishment the punishment contingency was added. This
contingency (excluding baseline) with 2.5 min decrease in stereotypy was not due to the
components. With the punishment contingen- duration of the contingent hands-down pro-
cy in effect, stereotypy occurred primarily in cedure, as the procedure was either of brief,
the no-punishment component. In the ab- fixed duration (Subjects TS and CB) or, in the
sence of this contingency (punishment extinc- case of Subject PH whose punishment dura-
tion), however, stereotypy gradually increased tion was 10 s, 10 s was added to the compo-
332 SHANNON S. DOUGHTY et al.

only after the first response produces the

punisher. Figure 1 illustrates this issue. Early
in discrimination training, there was stable
suppression of stereotypy in the punishment
component, and clear separation between
punishment and nonpunishment compo-
nents. The decreases in stereotypy, however,
occurred only after the presentation of the
punisher. Latency increased gradually after
stable response suppression was shown.
The undesirable side effects often associated
with punishment, including nonselective sup-
pression and the experimenter becoming an
aversive stimulus, were not observed in the
final conditions of the present study. The
suppressive effects were restricted to the
punishment component. Moreover, the per-
son who administered the punisher also was
present in the nonpunishment component
Fig. 2. Data for Subject CB when the punisher was and there was no evidence that her presence
present, withdrawn, and present again (separated by solid suppressed behavior. Note that these deleteri-
lines). Numbers represent the component length in min. ous side effects often associated with punish-
Unfilled circles show data from the punishment compo- ment involve a failure to establish a discrimi-
nent, and filled circles show data from the no-punishment
component. The top panel shows the percentage of native stimulus for punishment. Thus, when
intervals scored with stereotypy, and the bottom panel implementing punishment contingencies,
shows the percentage of the component elapsed prior to careful attention to stimulus control may pre-
the first interval scored with stereotypy. Data are in three- vent many of the objectionable side effects.
session blocks. The data from the first condition shown are
the same as those from the end of the 2.5-min-component One potential confound in discriminated-
shown in Figure 1, and are included for the purposes punishment procedures is that the discrimina-
of comparison. tion can be based on differential reinforce-
ment rather than differential punishment.
nent for each punisher delivery. Furthermore, This result can occur through explicitly pro-
responding came under SDp control; any gramming extinction in the punishment com-
responding that occurred in the punishment ponent, which confounds the roles of punish-
component occurred much later relative to ment and extinction (e.g., Silverman, 1971).
the no-punishment component. And finally, Extinction was not programmed in the present
the present study demonstrated stimulus con- study. Rather, the study was conducted to
trol in human subjects without the use of address effects of punishment in circum-
instructions. There have been relatively few stances under which extinction is impossible
studies of punishment in human subjects, and to implement. Functional analyses showed that
some reports have noted difficulties establish- these subjects’ stereotypy was maintained by
ing stimulus control without the use of sensory or automatic reinforcement. Because
instructions (O’Donnell et al., 2000). the behavior was automatically reinforced, it
The present study provides a more conclu- was not possible to eliminate the consequences
sive demonstration of control by SDp than that maintained responding. Moreover, the
many studies in the literature (see Doughty et punisher was delivered only after the response
al., in press, for a critical review) because we definition had been met, allowing reinforcer
measured latency from the presentation of contact, and reinitiating responding after the
each component stimulus to the first response punisher again would produce the sensory
in the component. Unless responding is consequences of the stereotyped behavior.
eliminated completely in the punishment Another way that inadvertent differential
component, latency data are essential to reinforcement can occur is through the
determining whether the suppression occurs reductions in response rate and concomitant
when the component stimulus is presented, or reductions in reinforcement rate that can
 DISCRIMINATIVE CONTROL OF PUNISHED BEHAVIOR 333

occur during the punishment component. Lerman and Vorndran (2002, p. 431) noted
Given that the response–reinforcer relation recently that ‘‘basic research on punishment
in stereotypy was conjugate in nature, with the has been declining rapidly despite substantial
continuous covariation of the duration of the gaps in knowledge’’ and that ‘‘further un-
response and the duration of the reinforcer derstanding of punishment processes may
(Williams & Johnston, 1992), a reduction in lead to an improved technology of behavior
the duration of the reinforcer was inevitable change.’’ The present findings have consider-
with a functional punisher. Thus, although the able practical implications. Many individuals
procedure did not program extinction for with developmental disabilities exhibit un-
stereotypy, there was a significant reduction in desirable behavior at rates that are socially
the amount of reinforcement obtained under inappropriate, interfere with learning socially
punishment conditions. Many studies of pun- appropriate skills, or are dangerous. Nonpun-
ishment incorporate this feature, especially ishment treatments to control, reduce, or
with strong punishers and/or with an FR 1 eliminate such behavior are preferred, and
punishment schedule (e.g., Orme-Johnson & are attempted first (e.g., Miltenberger, 1997).
Yarczower, 1974). The reinforcement-decreas- Punishment contingencies may be necessary,
ing effects of punishment can be minimized by however, under some conditions. Punishment
using a time-based rather than response- may be a useful treatment when the target
number-based schedule of reinforcement, for behavior is maintained by automatic reinforce-
example, a variable-interval schedule. Such ment, making it difficult to implement extinc-
a manipulation would not reflect the goals of tion. Moreover, the acquisition of stimulus
the present study, however, which were to study control without verbal instructions is impor-
automatically reinforced behavior. It has been tant because many individuals who exhibit
suggested that conjugate schedules of the type excessive stereotypy have severely limited
maintaining stereotypy are very common in language repertoires. Also important to po-
nature (Rovee-Collier & Gekoski, 1979), al- tential application is the fact that stereotypy
though they have received relatively little was brought under the control of practical
attention as compared to discrete responses antecedent stimuli, and non-noxious punish-
(wheel running is a notable exception). ers were used. Perhaps most important of all,
The nature of the punishing stimulus used the punisher rarely was delivered once control
in this study is somewhat ambiguous. The by the antecedent stimulus was established.
punisher was the physical manipulation of the In summary, this study is one of a very few,
subjects’ hands (the hands-down procedure). with either nonhuman animal or human
The hands-down procedure might have served subjects, that unambiguously demonstrates
one or both of two functions. First, it might control by SDp using differential punishment.
have served as a positive punisher—if the This conclusion was made clear by the
physical manipulation was itself aversive. Sec- measurement of response latencies across
ond, it might have served as a negative multiple-schedule components. The study was
punisher in that it discontinued the ongoing unique in that it established control by SDp
reinforcement. However, as a negative punish- using differential punishment, but without
er, it differs from the point-loss procedures a noxious stimulus such as shock as a punisher.
typically used with humans (e.g., O’Donnell et The present study distinguished itself from
al., 2000) in that previously delivered reinforc- other laboratory studies with humans in that
ers cannot actually be removed. stimulus control was established without in-
It might be considered a limitation that the structions and verbally sophisticated subjects,
present study did not compare response-de- and in that responding was maintained by
pendent and response-independent deliveries primary, rather than conditioned, reinforce-
of the hands-down procedure. In studies that ment. Given the relatively small number of
involve electric shock as a punisher, this studies that have investigated control by SDp, it
control for suppression by conditioned emo- is unclear whether the development of stimu-
tional responses is important. It is unclear, lus control was influenced by the type of
however, how crucial such a control is when punisher and/or reinforcer (i.e., primary
punishers other than electric shock and other versus conditioned), the schedule of punish-
‘‘noxious’’ stimuli are used. ment and/or reinforcement (i.e., continuous
334 SHANNON S. DOUGHTY et al.

versus intermittent), or other factors. Thus, Kennedy, C. H. (2002). Evolution of stereotypy into self-
injury. In S. R. Schroeder, M. L. Oster-Granite, & T.
more research on stimulus control and pun- Thompson (Eds.), Self-injurious behavior: Gene–brain–
ishment, in both the basic and applied arenas, behavior relationships (pp. 133–143). Washington, DC:
is warranted. American Psychological Association.
Lerman, D. C., & Vorndran, C. M. (2002). On the status of
knowledge for using punishment: Implications for
treating behavior disorders. Journal of Applied Behavior
REFERENCES Analysis, 35, 431–464.
Lovaas, O. I., Litrownik, A., & Mann, R. (1971). Response
American Psychiatric Association. (1994). Diagnostic and latencies to auditory stimuli in autistic children
statistical manual of mental disorders (4th ed.). Washing- engaged in self-stimulatory behavior. Behaviour Re-
ton, DC: Author. search and Therapy, 9, 39–49.
Baron, A. (1991). Avoidance and punishment. In I. H. Miltenberger, R. G. (1997). Behavior modification: Principles
Iversen, & K. A. Lattal (Eds.), Experimental analysis of and procedures. Pacific Grove, CA: Brooks/Cole.
behavior: Part 1 (pp. 173–217). New York: Elsevier O’Donnell, J. (2001). The discriminative stimulus for
Science. punishment or SDp. The Behavior Analyst, 24, 261–262.
Birnbrauer, J. S. (1968). Generalization of punishment O’Donnell, J., & Crosbie, J. (1998). Punishment general-
effects: A case study. Journal of Applied Behavior ization gradients with humans. The Psychological Record,
Analysis, 1, 201–211. 48, 211–232.
Bodfish, J. W., & Lewis, M. H. (2002). Self-injury and O’Donnell, J., Crosbie, J., Williams, D. C., & Saunders, K. J.
comorbid behaviors in developmental, neurological, (2000). Stimulus control and generalization of point-
psychiatric, and genetic disorders. In S. R. Schroeder, loss punishment with humans. Journal of the Experi-
M. L. Oster-Granite, & T. Thompson (Eds.), Self- mental Analysis of Behavior, 73, 261–274.
injurious behavior: Gene–brain–behavior relationships (pp. Orme-Johnson, D. W., & Yarczower, M. (1974). Condi-
23–39). Washington, DC: American Psychological tioned suppression, punishment, and aversion. Journal
Association. of the Experimental Analysis of Behavior, 21, 57–74.
Bolles, R. C., Holtz, R., Dunn, T., & Hill, W. (1980). Rincover, A. (1978). Sensory extinction: A procedure for
Comparisons of stimulus learning and response eliminating self-stimulatory behavior in developmen-
learning in a punishment situation. Learning and tally disabled children. Journal of Abnormal Child
Motivation, 11, 78–96. Psychology, 6, 299–310.
Rollings, J. P., & Baumeister, A. A. (1981). Stimulus control of
Charlop, M. H., Kurtz, P. F., & Casey, F. G. (1990). Using
stereotypic responding: Effects on target and collateral
aberrant behaviors as reinforcers for autistic children.
behavior. American Journal of Mental Deficiency, 86, 67–77.
Journal of Applied Behavior Analysis, 23, 163–181.
Rovee-Collier, C., & Gekoski, M. J. (1979). The economics
Dinsmoor, J. A. (1998). Punishment. In W. T. O’Donohue
of infancy: A review of conjugate reinforcement. In H.
(Ed.), Learning and behavior therapy (pp. 188–204). W. Reese, & L. P. Lipsitt (Eds.), Advances in child
Needham Heights, MA: Allyn & Bacon. development and behavior: Vol. 13 (pp. 195–255). New
Doughty, A. H., Doughty, S. S., O’Donnell, J., Saunders, K. York: Academic Press.
J., & Williams, D. C. (in press). Stimulus control of Sidman, M. (1989). Coercion and its fallout. Boston, MA:
punishment effects: Determining the controlling Authors Cooperative.
variables. Mexican Journal of Behavior Analysis. Silverman, P. J. (1971). Chained and tandem fixed-interval
Garner, J. P., Meehan, C. L., & Mench, J. A. (2003). schedules of punishment. Journal of the Experimental
Stereotypies in caged parrots, schizophrenia and Analysis of Behavior, 16, 1–13.
autism: Evidence for a common mechanism. Beha- Vollmer, T. R. (1994). The concept of automatic re-
vioural Brain Research, 145, 125–134. inforcement: Implications for behavioral research in
Honig, W. K., & Slivka, R. M. (1964). Stimulus generaliza- developmental disabilities. Research in Developmental
tion of the effects of punishment. Journal of the Disabilities, 15, 187–207.
Experimental Analysis of Behavior, 7, 21–25. Weisman, R. G. (1975). Stimulus control by response-
Iversen, I. H. (1993). Techniques for establishing sched- dependent shock in discriminated punishment. Bul-
ules with wheel running as reinforcement in rats. letin of the Psychonomic Society, 5, 427–428.
Journal of the Experimental Analysis of Behavior, 60, Williams, D. C., & Johnston, J. M. (1992). Continuous
219–238. versus discrete dimensions of reinforcement sched-
Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & ules: An integrative analysis. Journal of the Experimental
Richman, G. S. (1994). Toward a functional analysis of Analysis of Behavior, 58, 205–228.
self-injury. Journal of Applied Behavior Analysis, 27,
197–209. (Reprinted from Analysis and Intervention in Received: May 2, 2005
Developmental Disabilities, 2, 3–20, 1982). Final acceptance: December 14, 2006
 DISCRIMINATIVE CONTROL OF PUNISHED BEHAVIOR 335

APPENDIX A

Functional Analysis

An analog functional analysis (Iwata, Dorsey, condition, the experimenter delivered a brief
Slifer, Bauman, & Richman, 1982/1994) first verbal reprimand (or, for Subject TS, signed
was conducted for each subject and included and stated ‘‘Stop’’) immediately following
at least five session types: demand, attention, stereotypy. This arrangement was to determine
tangible, alone, and control. Two subjects if stereotypy was maintained by access to
received another session type, idiosyncratic attention. In the tangible condition, a tangible
stimulus. No single session type occurred twice was given for 20 s immediately after stereotypy.
in succession, and at least four sessions of each This arrangement was to determine if stereotypy
type were conducted. The functional analysis was maintained by access to preferred tangibles.
continued until this minimum criterion had In the alone condition, the subject was in the
been met, and stereotypy was judged to be room by himself to assess stereotypy in the
stable by visual inspection (i.e., minimal absence of social consequences. In the control
bounce and trend). Attention was a brief (3– condition, the experimenter was present and
5 s) delivery of physical contact and/or a vocal preferred tangibles were available. The experi-
statement, with the type of vocal statement menter delivered verbal attention (i.e., praise)
different across conditions (see below). Tan- every 20 s. The idiosyncratic-stimulus condition
gible was the delivery of a preferred stimulus was conducted for Subjects CB and TS based on
(e.g., soda). Prompts (verbal, gestural, and staff report that these subjects engage in more
physical) sometimes were used to complete stereotypy in the presence of a specific stimulus
a task (e.g., hand-over-hand prompting). (in both cases, a television). In this condition,
In the demand condition, the experimenter the experimenter was present, no consequences
presented preacademic and daily-living-skill were programmed for stereotypy, and the tele-
demands using a three-step prompting pro- vision was on continuously. The results sug-
cedure. If stereotypy occurred, the task was gested that stereotypy most likely was sensory
withdrawn for 20 s, and the experimenter maintained for each subject, and that the verbal
turned away from the subject. This arrangement reprimands in the attention condition may have
was to determine if stereotypy was maintained punished stereotypy for 2 of the subjects (CB
by escape or avoidance. In the attention and PH).
336 SHANNON S. DOUGHTY et al.

APPENDIX B

Mean and Range of Occurrence (Occ), Nonoccurrence (NonOcc), and Total Agreement Scores.

Percent of Type of Percent

Participant Condition sessions Response agreement agree Range
TS Baseline 33.3 Stereotypy Occ 91.2 78.6–100
NonOcc 78.1 41.7–100
Total 93.5 87.5–100
Discrim training 33.3 Stereotypy Occ 80.6 50–95
NonOcc 89.1 81.1–99
Total 92.6 84.2–99.2
Punisher Occ 84.6 0–100
NonOcc 99.7 98.3–100
Total 99.7 98.3–100
CB Baseline 33.3 Stereotypy Occ 92.2 78.9–96.7
NonOcc 88.6 76–95.6
Total 95.8 91.7–97.5
Discrim training 33.3 Stereotypy Occ 92.8 68.8–100
NonOcc 96.4 93.8–100
Total 97.6 94.2–100
Punisher Occ 82.1 0–100
NonOcc 99.6 96.7–100
Total 99.6 96.7–100
PH Baseline 49.2 Stereotypy Occ 96.7 90.5–100
NonOcc 90.3 76.5–100
Total 97.4 92.5–100
Discrim training 33.3 Stereotypy Occ 91.8 35.1–100
NonOcc 97.2 78.6–100
Total 98.1 81–100
Punisher Occ 89.9 50–100
NonOcc 99.4 95.7–100
Total 99.4 96–100