JOURNAL OF APPLIED BEHAVIOR ANALYSIS

THE EFFECTS AND SIDE EFFECTS OF PUNISHING THEAUTISTIC BEHAVIORS OF A DEVIANT CHILD'

TODD R. RISLEY

UNIVERSITY OF KANSAS

Timeout procedures in the home and extinction and reinforcement of incompatible behav-iors in the laboratory failed to eliminate the disruptive and dangerous climbing behavior ofa deviant child. Punishment with electric shock was used to eliminate this behavior in thelaboratory and then in the home. The effects were reversible and were restricted to specificostimulus conditions. A less severe form of punishment was used to eliminate the child's autis-tic rocking. Other behaviors of the subject were continuously measured in the laboratory todetermine the side effects of punishment. No suppression of other behaviors correlated withpunishment was noted. However, the rate of some behaviors increased when punishment wasused to eliminate deviant behaviors, but these increases were, primarily, desirable.

A prime argument against punishment hasbeen that it allegedly produces undesirableside effects. Traditionally, the evidence sup-porting this argument has been based on clin-ical anecdotes describing cases of "symptomsubstitution". More recently, the results ofexperimental research have similarly sug-gested that punishment procedures are likelyto produce undesirable side effects.For example, the conditioned suppression

literature suggests that aversive stimulationmay suppress other behaviors, including de-sired behaviors, in addition to suppressingthe behavior being punished. The negativereinforcement literature suggests that aversivestimuli may produce and maintain escapeand avoidance behaviors which may be un-desirable, such as leaving, avoiding, or remov-ing the punishing situation, or the person dis-pensing the punishment. The literature on

'Preparation of this paper was partially supportedby National Institute of Child Health and Human De-velopment grant HD-03144-01 to the Bureau of ChildResearch and The Department of Human Develop-ment at the University of Kansas. This work was donein 1963-64 while the author was a research assistantat the Developmental Psychology Laboratory of theUniversity of Washington. Financial and moral sup-port was provided by Donald M. Baer, early sugges-tions by Montrose M. Wolf, later suggestions by IvarLovaas and technical assistance by Gary Millar, BettyHart, Nancy Reynolds, and Cordelia McIntosh. Thisinvestigation was possible only through the constancyand cooperation of the child's mother. Reprints maybe obtained from the author, Juniper Gardens Chil-dren's Project, 2021 N. Third St., Kansas City, Kansas66101.

21

pain-elicited aggression (or "reflexive fight-ing") suggests that aversive stimuli may elicitaggression toward the person dispensing pun-ishment and toward other organisms and ob-jects as well. A corollary is that punishmentprocedures may, in fact, increase rather thaneliminate aggressive behaviors. And finally,the stimulus properties of the person dispens-ing punishment may become altered by be-ing paired with aversive stimulation such thathis presence and attention become more aver-sive and less reinforcing. Virtually every sum-mary account of punishment research in re-cent literature (excepting Solomon, 1964) hasappended a warning statement to the effectthat, for these reasons, the use of punishmentis contraindicated when dealing with appliedproblems of human behavior. (Cf. Azrin andHolz, 1966 for a recent example, as well asfor a thorough and clear review of punish-ment research.)The present study describes the applica-

tion of a series of procedures designed to re-duce the highly dangerous and disruptiveclimbing behavior of a severely deviant child.After other methods had failed, electric shockpunishment was applied under several condi-tions. Another punishing stimulus, shoutingat, and shaking the child, was applied to thechild's autistic rocking. Other behaviors ofthe child were recorded to assess possible sideeffects of these punishment procedures.

Antecedent to this study in both time andfunction was the initial work on the use ofshock punishment with autistic children by

1968, 1, 21-34 NUMBER I (SPRING, 1968)

TODD R. RISLEY

Lovaas, Schaeffer, and Simmons (1965). Averbal report of that work provided the insti-gation and some of the techniques to investi-gate further the effects of punishment.

METHODSubjectThe subject, S, was a 6-yr-old girl who was

hyperactive and exhibited bizarre behaviors.She had been consistently diagnosed ashaving diffuse brain damage caused by pneu-mococcal meningitis at age seven months, al-though recent diagnoses included an 'over-lay' of emotional disturbance and autism. Shehad occasional seizures and was taking an-ticonvulsant medication. She exhibited noverbal behavior but almost continuously emit-ted howls, moans, and clicks. These vocaliza-tions did not correspond in length, inflectionor topography to normal speech. She exhib-ited no imitative behavior, either verbal ornon-verbal. Her predominant behaviors in allsituations were climbing in high places (onfurniture, window sills, trees, houses, etc.),alternating with sitting and rocking rhythmi-cally. Her climbing was a constant source ofconcern to her parents due to the threat toher life and limb (her body bore multiplescars from past falls; her front teeth weremissing, having been left imbedded in a 2 by4-in. molding from which she had fallen whileclimbing outside the second story of herhouse), and the attendant destruction of fur-niture in the house. She had attended severalschools for special children but had beendropped from each because of these disrup-tive behaviors and her lack of progress.

S's parents, who both possessed advancedacademic degrees, had resisted placing her inan institution, as they predicted that herclimbing would result in her being kept incontinuous physical restraint on a custodialward. However, as she had become larger andmore skillful, her climbing and her aggres-sion toward her younger brother at homewere causing them to consider institutionali-zation seriously.

ReinforcerS was brought to the laboratory four times

a week around noon after having had only3 oz of milk at breakfast time. Milk, whichwas the only food she would reliably consume,was used as a reinforcer. Even under this

amount of food deprivation, she exhibitedlong latencies of drinking the milk when itwas presented. Each reinforcer was about onetablespoon of milk in a paper cup placed onthe table in front of S, accompanied by thestatement, "Good girl!"

SettingThe experimental sessions were either 20

or 30 min long, and were conducted in an8 by 12-ft experimental room with an 11-ftceiling. At one end of the room, next to thedoor, a ventilator frame formed a 5-in. deepledge, 6 ft above the floor. Directly acrossfrom the door a large one-way mirror permit-ted observation from an adjacent room. Theexperimenter and the child sat in chairs, fac-ing each other across a small table in thecenter of the room (cf. Risley and Wolf, 1967).Initially the room also contained several ex-tra chairs and a canvas cot. At the experi-menter's request, S's mother observed all ses-sions through the one-way mirror.

RecordingFrequency and duration of the child's be-

haviors were recorded on a six-pen event re-corder located in an adjacent room, via abank of microswitches placed on the table.On occasional sessions an observer behind aone-way mirror would independently recordthose behaviors for which reliability was con-sidered to be a problem. However, most ofthe behaviors were so highly distinctive thatreliability checks were not considered neces-sary.The data presented on S's behaviors at

home were collected by the mother through-out each day. Periodically, the experimenterwould observe in the home for several hours.During these periods there was always com-plete agreement between the experimenter'sand the mother's recording. Since there wereno systematic differences in the mother's dataon those days, as compared to the prior andsucceeding days, the mother's data were con-sidered to be reliable.

PROCEDURES AND RESULTSTimeout for climbing in the home. The

mother's response to S's climbing was origi-nally considered to be the most likely variablemaintaining that climbing. Since the climb-

22

EFFECTS AND SIDE EFFECTS OF PUNISHMENT

ing usually endangered the child and/or de-stroyed the home furnishings, the mother'scontingent attention and interaction was con-sistent and predictable. A simple extinctionprocedure (ignoring the behavior) did notappear feasible. Therefore, physical isolation(timeout) from social interaction was madecontingent upon climbing behavior. Accord-ingly, her mother was instructed to say "Nol",

lift S to the floor, and lead her to her bed-room (with minimal physical contact and nofurther verbalization) contingent upon eachinstance of inappropriate climbing. The bed-room door was reopened after 10 min (timedon a kitchen timer). The mother was alsoinstructed to attend to and to interact withher as frequently as possible when S was notclimbing. Inappropriate climbing was defined

Of A,, w- -- -Wl

0

~~~~~~~b

------- d -""I\2Fig. 1. Tracings from photographs showing the subject's climbing behavior (a and b), eye contact topography

(c), and consummatory response (d). Note the bookcase, door, and ventilator frame in a and b and the micro-switch recording panel beside the experimenter's hand in c and d. Original photographs were taken throughthe one-way observation window.

23

TODD R. RISLEY

as standing, sitting or hanging on anythingnot specifically designed for such, with neitherfoot touching the floor.

After 17 days, no reduction in the rate ofclimbing was observed. The timeout contin-gency was then applied to climbing in onlyone location (the bathroom) and the motherwas instructed to do whatever she had pre-viously been doing for all other climbing. Itseemed likely that a concurrent schedule ofsocial interaction for climbing in other placesand the timeout procedure for climbing inthe bathroom would provide a more sensitivemeasure of the effects of the timeout proce-dure. However, no reduction in the rate ofclimbing in the bathroom was obtained dur-ing 46 days of this procedure (average dailyrate of climbing was 5.5 per day over the first23 days and 5.7 over the last 23). The time-out procedure was therefore discontinued.Reinforcement for incompatible behaviors

and extinction for climbing in the laboratory.Concurrent with these attempts to eliminateclimbing at home, procedures for establish-ing imitative behaviors were initiated in thelaboratory.

In a preliminary session, S moved aboutthe room almost continually standing on thechairs and table, moving furniture to thedoor and climbing from it to the door knob,then to the ventilator frame next to the door,and then to the door lintel (a and b, Fig. 1).Alternately she would sit on the floor or inthe chair, rhythmically rocking and hum-ming with closed eyes. Throughout this pe-riod she frequently struck the side of her headwith her palm or fist, sometimes resoundingly.She occasionally would approach and grabfor the food reinforcer, but she never lookeddirectly at the experimenter, and activelyaverted her gaze whenever the experimenterstood in front of her.

In order to establish attending to the ex-perimenter's face, which was a necessary pre-requisite for vocal imitation training, system-atic shaping of eye contacts (S's gaze focusingon the experimenter's eyes) was begun. Dur-ing the preliminary session, the experimenterperiodically said "Sit down" and patted thechair. Initially, standing by the chair, andthen only sitting in the chair was reinforced.The original plan was to work with the

mother to control climbing behavior at home,while concentrating on developing imitative

behavior in the laboratory. Therefore, climb-ing was eliminated in the laboratory by re-moving the opportunity to climb. BetweenSessions 2 and 3 (point A, Fig. 2) all furniturewas removed from the room except a tableand two chairs which were fastened to thefloor in the middle of the room. After severalunsuccessful attempts to step up onto thedoorknob without a chair in the next twosessions, all climbing activity ceased. Timespent out of the chair decreased from 38% ofSessions 1 and 2, to less than 2% of Sessions5 through 12.Once S was sitting in the chair during most

of the session, reinforcers were delivered only

NounFig. 2. Graphs showing the relationship between the

rate of eye contacts and the amount of time S spentout of her chair. Each dot represents one session. Theportions of the graphs between the dotted lines la-beled "CLIMB" indicate blocks of sessions duringwhich S was climbing on the bookcase. At A all furni-ture was removed from the room, precluding climbing.At B the bookcase was placed in the room. At Cclimbing on the bookcase was punished with electricshock. Beginning at D the shock device was notbrought into the experimental room. At E a footstoolwas placed in the room in front of the bookcase andclimbing resumed. At F climbing on the bookcasewas again punished with shock. At G standing on thechair was punished with shock.

24

TODD R. RISLEY

when she looked at the milk cup. As lookingat the cup became more frequent, the experi-menter gradually moved the cup toward hisface, thereby increasing the probability of eye

contacts. A few fleeting glances at the experi-menter's face occurred and were reinforced.These gradually became more frequent. Aftereye contacts had reached a rate of 6 per min,reinforcement for looking at the cup was dis-continued. At this point S was first looking atthe cup and then looking at the experimenter;successively longer eye contacts were rein-

forced until the topography of this behaviorwas a focussed stare at the experimenter's eyes

of 1 sec or longer (see C, Fig. 1). Concur-rently, the experimenter gradually moved thecup away from his face, finally holding thecup out of sight under the table. The fre-quency of eye contacts systematically in-creased during these procedures from 0 per

min in the first session to 1.5 per min in Ses-sion 12 (point B, Fig. 2).

Meanwhile, the timeout procedure hadfailed to reduce the climbing behavior athome. Now that sitting in the chair and look-ing at the experimenter (behaviors incompati-ble with climbing) had been established inthe laboratory it was decided to see if climb-ing would re-occur there if the opportunitywere again presented. A small bookcase was

placed under the ventilator next to the door(see a and b, Fig. 1) before Session 13. As can

be seen in Fig. 2 at point B, time out of thechair immediately increased from less than2% of the previous eight sessions to 42%/o ofthe first four sessions with the bookcase pres-

ent, as S again began to climb. In the 14 ses-

sions (totaling 6.4 hr) after the bookcase was

introduced, S climbed on the bookcase, andthe ventilator and door above the bookcase,an average of 6.7 times per hr (from point Bto first arrow, Fig. 3), occupying 18% of thetime in the sessions. During these sessions theexperimenter did not look at her or respondin any way when she was out of the chair,but sat staring down at the table. When Sresumed her seat, the experimenter wouldlook up and wait for S to meet his gaze. Eyecontacts of 1 sec or longer were reinforcedwith milk.Thus it did not appear that the climbing

behavior was maintained by consequenceswhich the experimenter could manipulate.Attempts to supplant climbing by establish-

ing competing behaviors, coupled with theremoval of all experimenter-controlled conse-quences for this behavior had had no appar-ent effect in reducing its frequency or dura-tion. Therefore, it was decided to attempt toeliminate the climbing behavior by the con-tingent application of shock.Punishment with shock for climbing in the

laboratory. A hand-held inductorium was con-structed which operated on a series of seven1.5-v flashlight batteries. When a button waspressed this device delivered shock across twocontacts % in. apart. The coil, interrupter,and shock contacts were obtained from a com-mercially available device for shocking live-stock (Hot Shot Products, Minneapolis 16,Minnesota). From oscilloscope readings it wasestimated that the average voltage outputwas in the range of 300 to 400 v, with occa-sional spikes exceeding 1000 v. Subjectively,the shock produced a sharp, extremely pain-ful sting, localized in the area of the body towhich the contacts were touched, much likebeing struck with a vigorously applied willowswitch. The pain terminated with the re-moval of the shock, with no after-effects suchas redness, swelling of the skin, tingling, oraching. (Observers of the sessions in whichshock was applied reported that, on the basisof observable autonomic responses such asflushing, trembling, etc., the subject recov-ered from the shock episodes much faster thanthe experimenter.)

In the twenty-seventh experimental session(first and second arrows, Fig. 3) when thebookcase had been present for 14 sessions (6.4session-hr), shock was applied contingent uponclimbing. When the child climbed on thebookcase, the experimenter would shout"No!", run to her, take hold of one leg, touchthe shock contacts to the calf or lower thighand depress the switch for approximately 1sec. The experimenter then returned to hischair, looked down at the table until S re-turned to her chair, and then looked up andresumed reinforcing eye contacts.

In Session 27, S climbed nine times, butonly two shocks were delivered. On the firstfour climbing episodes the experimenter be-gan the punishment sequence (shouting"No!" etc.) immediately contingent upon theinitial stages of climbing, when S was still onthe lower shelf of the bookcase. On these oc-casions, when the experimenter shouted

25

EFFECTS AND SIDE EFFECTS OF PUNISHMENT

"No!" and approached, S stepped down fromthe bookcase to the floor. As the shock wasto be made immediately contingent onlyupon climbing, no shocks were delivered.On the fifth climbing episode, the experi-menter waited until S had climbed from thetop of the bookcase to the ventilator frame;he then shouted "Nol" and approached, Sstepped down to the top of the bookcase,where she was standing when the shock wasapplied (first arrow, Fig. 3). When shocked,S abruptly sat down on the top of the book-case. The experimenter took her arm and as-sisted her down to the floor, then returned tohis chair. S returned to her chair 23 sec afterthe shock, looked at the experimenter andconsumed the consequent milk reinforcerwithin 70 sec after the shock. On the sixth,seventh, and eighth climbing episodes, when"No!" was shouted, S jumped to the floor be-fore the experimenter reached her, and noshock was delivered. On the ninth climbing

episode in Session 27, S was still climbingwhen reached and the shock was applied.

In Session 28 (third and fourth arrows, Fig.3), the first climbing episode was terminatedwithout shock when, upon the shouted "Nol",S jumped 6 ft from the ventilator frame tothe floor. It was apparent that although theprocedure had not eliminated the climbing,it had quickly produced behaviors whichavoided the shock. Therefore, on succeedingclimbing episodes the experimenter shouted"No!" and then, irrespective of S's behaviorwhen reached, shock was applied. On thesecond climbing episode, S had jumped tothe floor when shock was applied (third ar-row, Fig. 3). On the next climbing episode Sgot up from her chair, pushed the bookcaseacross the room to the other side of the doorand climbed there. Shock was again appliedafter S had jumped to the floor (fourth ar-row, Fig. 3). Approximately 5 min later S gotup, pushed the bookcase back to its original

v * bookcase.. chair\ SHOCK

It U 11

5 20 35 50SESSION HOURS

Fig. 3. A cumulative graph of the rates of S's climbing on the bookcase and standing on her chair. Each doton the top line represents one session. Heavy arrows indicate where each behavior was punished with electricshock. At B the bookcase was placed in the experimental room. Beginning at D the shock device was notbrought into the experimental room. At E a small stool was placed in front of the bookcase. Beginning at theshort vertical line above the X-axis S was placed in the room alone for 5 min before each session. The dotsabove the X-axis indicate instances of climbing on the bookcase in these periods.

w

U-

26

TODD R. RISLEY

position, looked at the experimenter, andthen returned to her chair. No further climb-ing occurred in the remainder of Session 28or in Session 29. One climbing episode, fol-lowed by shock, occurred in Session 30 (fiftharrow, Fig. 3). Four sessions (1.5 hr) later(Session 34) one additional climbing episodeoccurred and was followed by shock (sixtharrow, Fig. 3). No further climbing occurredin the subsequent 12 sessions (6.5 hr).When climbing had been eliminated, corre-

lated with the contingent application ofshock, the generality and reversibility of theeffect were investigated.The mother reported no noticeable de-

crease in climbing at home correlated withthe elimination of climbing in the labora-tory. Thus, the effects of the shock punish-ment appeared to be specific to the laboratorysituation. From Session 36, S was placed in theexperimental room alone for 5 min beforethe session started to see if climbing wouldoccur when the experimenter was absent.Climbing occurred during each of the firstfive pre-session periods and during intermit-tent periods thereafter (dots above X-axis,Fig. 3) but did not occur in the regular ses-sions when the experimenter was present withthe shock apparatus. From Session 40 (pointD, Fig. 3) the shock apparatus was not pres-ent during the regular sessions, with no re-currence of climbing. Clearly, the reductionin climbing was primarily under the discrimi-native control of the presence of the experi-menter.

Before Session 47, after no climbing had oc-curred in the experimenter's presence for 12sessions (6 hr), a 1-ft high metal stool wasplaced in front of the bookcase (point E, Fig.3). During this session, S approached thestool, placed one foot on it, looked at theexperimenter, and then returned to her chair.A few minutes later she again approachedthe stood, stood on it, and again looked atthe experimenter. She then placed one footon the bookcase, looked back, paused, andclimbed on the bookcase. The experimenter(lid not respond. After this, climbing oc-curred at an average of 4.9 times per hr dur-ing the subsequent 11 sessions (5.5 hr) (pointE to seventh arrow, Fig. 3), occupying 12%of the time in the sessions. At Session 50, theshock apparatus was again brought to the ses-sions, with no discernible effect on the fre-

quency of climbing. Clearly, the effects of theshock punishment were reversible (not per-manent).

In Session 58 one shock was applied contin-gent upon the second instance of climbing inthe session. No further climbing occurredduring the next 59 sessions (23 hr). However,climbing still occurred during the pre-sessionperiods when the experimenter was not inthe room. From the period preceding Session65, whenever S climbed, the experimenterwould enter the room shouting "No", applythe shock, and leave again. This procedurereduced the proportion of pre-session periodsin which climbing occurred from 52% to 10%(dots above X-axis, Fig. 3).The side effects of punishing climbing with

shock. As climbing on the bookcase decreased,another, topographically similar, behavior in-creased. S began to stand and climb on theseat and back of her chair. She would sit onthe back of the chair, stand on the chair seat,and often stand on the back of the chair withher hands braced against the wall. Climbingon the chair was defined as S being on thechair with neither foot on the floor exceptwhen she was either sitting or kneeling onthe chair seat. This behavior first occurred inSession 28, when climbing on the bookcasewas nearly eliminated. During Sessions 35 to46, when climbing on the bookcase had beenreduced to zero, climbing on the chair oc-curred at an average rate of 8.8 per hr occu-pying 8.4% of the session time (dotted line,Fig. 3). When the foot stool was introduced(point E, Fig. 3) and climbing on the book-case again occurred at its previous high rate,climbing on the chair immediately ceased.When climbing on the bookcase was againeliminated by contingent shock (seventh ar-row, Fig. 3), climbing on the chair again oc-curred, at a rate of 4 per hr occupying 3.1 %of the session time. After 10 sessions, duringwhich the frequency of climbing on the chairwas relatively stable, shock was then appliedcontingent upon this behavior. After threeapplications of shock contingent upon eachinstance of climbing on the chair during twosessions, no chair climbing occurred duringthe subsequent 28 sessions (9.6 hr). Dur-ing Session 99 another instance of chair climb-ing occurred, shock was applied, and nofurther chair climbing occurred in the subse-quent 18 sessions (7.7 hr).

27

EFFECTS AND SIDE EFFECTS OF PUNISHMENT

Another side effect was observed in a sec-ond, topographically dissimilar behavior, eyecontact. Throughout the procedures to elimi-nate climbing, sitting in the chair and meet-ing the experimenter's gaze for 1 sec had beencontinuously reinforced with milk. DuringSessions 3 to 12 (4.7 hr), when climbing wasmade physically impossible by the absence offurniture, the frequency of eye contacts hadsteadily increased from 5.5 to a peak of 88per hr (A to B, Fig. 2). During Sessions 13to 26 (6.4 hr) when the bookcase was intro-duced and climbing reoccurred, the frequencyof eye contacts remained relatively stable atan average of 42 per hr (B to C, Fig. 2). Dur-ing Sessions 27 to 46 (9.7 hr) when climbingwas eliminated by the contingent applica-tion of shock, the frequency of eye contactssteadily increased from 42 to a peak of 152per hr (C to E, Fig. 2). During Sessions 47 to57 (5.5 hr), when climbing again occurred,the frequency of eye contacts remained rela-tively stable at an average of 151 per hr (Eto F, Fig. 2). During Sessions 58 to 77 (8.8 hr),after climbing was again eliminated by thecontingent application of shock, the fre-quency of eye contacts, though variable, againslowly increased to a peak of 222 per hr (F toend, Fig. 2).From behind the one-way mirror, another

observer recorded the duration of S lookingat the experimenter's face during six sessionsbetween Sessions 13 to 26 and six sessions be-tween Sessions 27 to 46. During these sessionsthe observer started a stopwatch whenever Slooked at the experimenter's face (when Swas sitting in her chair) and stopped it when-ever S looked away. Although this measure in-cluded other instances of S looking at the ex-perimenter, in addition to the eye contactsrecorded (S often looked at the experimenterwhen he was pouring the milk, looking at themicroswitch recording panel, etc.; or duringthe reinforcement 'cycles' while S was holdingthe cup), the changes in magnitude of bothwere closely correlated. The close correspon-dence between the relative levels of this mea-sure and the relative levels of eye contactsrecorded during both of the experimental con-ditions, as shown in Table 1, substantiatesthe magnitude of the changes in eye contactsbetween these two conditions recorded by theexperimenter.

Eye contacts could occur only when S was

seated in her chair: the experimenter did notlook at her when she was out of her chair.Eliminating climbing with contingent shockdid not noticeably affect the amount of timeS spent sitting in her chair. Except whenclimbing was physically impossible, due tothe absence of furniture, the amount of timeS spent out of her chair gradually declinedacross all conditions, from 42% during Ses-sions 13 to 16 (when the bookcase was firstintroduced) to 0% during Sessions 107 to117. The periods in which eye contacts re-mained constant or systematically increasedwere not correlated with the amount of timeS was in or out of her chair. The systematicincreases in eye contacts during periods whenclimbing was not occurring were not due toS spending a greater proportion of time in thechair, but to S looking at the experimentermore- frequently when she was in her chair.

Table 1

Eye Min ofContacts Looking atPer Hr Experimenter(Experi- Per Hrmenter's (Observer

Session Record) Record)

14 54 3.116 34 2.6

S Climbing 18 32 2.120 54 3.025 50 2.7

36 88 4.537 70 3.7

S Not Climbing 39 92 4.641 140 7.644 138 7.846 152 8.1

Punishment with shock for climbing in thehome. After these effects and side effects ofshock were evaluated in the laboratory, climb-ing in the home was punished with shock.The mother was again instructed to recordeach instance of inappropriate climbing inthe home. After 16 days of recording, duringwhich inappropriate climbing occurred on theaverage of 29 times per day, the mother be-gan to punish climbing with shock. On theseventeenth day, when the shock was first ap-plied, the experimenter was present in thehome instructing the mother in the use of theshock apparatus. The mother carried the

28

TODD R. RISLEY

shock apparatus in the pocket of her apron.When the child climbed, the mother was in-structed to shout "Nol", and to continue toscold the child in a loud voice while ap-proaching, apply the shock, and then, withno further interaction, resume her previousactivity. The mother continued to attend to,and interact with the child intermittentlywhen she was not climbing. Shock reduced theinappropriate climbing from an average of 29per day to 2 per day within four days (Fig. 4).The mother reported that the shock devicehad been malfunctioning on Day 29 throughDay 32, delivering shock only on intermit-tent trials (dotted lines, Fig. 4). On Day 33the shock device was repaired (arrow, Fig. 4).Subsequently, inappropriate climbing de-creased quickly to zero (from arrow to Day50, Fig. 4).Another problem of long standing had

been that S would occasionally strike herthree-year-old brother with an object, pushhim down the stairs, etc. As the mother wasextremely concerned for the safety of theyoung child, shock was also applied contin-gent upon aggressive behavior toward herbrother. Although no baseline was taken onthe frequency of this behavior before shock

60'

amg

a

x

30

0

I20i

-

zu-A:-*0

Il:;~~~~~~~~~~~~~~~~~~~~~:1APJ§SO

1 SHOCK'71

was applied, the mother had estimated thatthe behavior occurred three or four times aday. S was shocked contingent upon 17 in-stances of hitting her brother over 20 days.During this time the behavior decreased from2.3 per day on the first three days of shockcontingencies to zero (upper graph, Fig. 4)with no further instances of this behavior re-ported during the subsequent 70 days.On the fifty-first day, when no climbing had

occurred on 14 of the last 15 days, the shockdevice was removed from the home. Themother was instructed to try to control theclimbing by spanking the child whenever sheclimbed. During 25 days of this, the climbingaveraged 2.0 per day, and showed a slightlyincreasing trend (Fig. 4). Furthermore, themother complained that spanking the childwas more unpleasant and "brutalizing" forboth herself and the child than the shock hadbeen. Therefore, further procedures weresought to maintain a low, tolerable frequencyof climbing without the direct use of shock.

After the daily sessions in the laboratory,the child was taken to a large nursery schoolplayroom. A chair was placed in the middleof the room. As S was wandering around theroom, periodically she would be told to sit in

17 DAYS 51

I SPANK I CHAIR T-0, SHOCK AVOIDANCE I5" 76 125

DAYSFig. 4. Graphs of the frequencies of S's climbing and aggressions against her little brother at home. Each dot

represents one day. Beginning on Day 17 each occurrence of either behavior was punished with electric shock.The dotted lines during the shock condition represent days when the shock device was malfunctioning. Thedevice was repaired at the arrow. Beginning on Day 51 the shock device was removed from the home and themother was instructed to spank the child for climbing. Beginning on Day 76 the child was given a 10-min time-out in a chair for climbing. Sitting quietly in the chair avoided shock.

29

EFFECTS AND SIDE EFFECTS OF PUNISHMENT

the chair. The experimenter would point tothe chair and say loudly, "(Name), go sit inthe chair." If the child moved in any direc-tion but toward the chair or did not move atall for 5 sec the experimenter would slowlyapproach the child with the shock device un-til she sat in the chair. After S had been sit-ting in the chair for a variable length of time,she would be helped up from the chair, andtold, "O.K., you can go now". If S attemptedto get up before this occurred, the experi-menter would shout "No!" and approach herwith the shock apparatus. Under these condi-tions, S's compliance with the instructions tosit in the chair improved only slightly. Herfirst move was never in the direction of thechair. Her latency of getting to the chair re-mained long (averaging 26 sec) and it wasnecessary to approach her with the shock ap-paratus on 63% of the first 30 trials. On thethirty-first trial the child was shocked whenshe had to be approached with the shock ap-paratus. On the next six trials the child wentdirectly to the chair, arriving there within 6to 15 sec without being approached with theshock device. During the next 44 trials shockwas applied five times. By trials 84 to 101, Swas going directly to the chair when in-structed, arriving within 7 to 12 sec, with theshock apparatus never being presented. Afterthe first shock, S, once seated, never attemptedto get up until instructed to do so.This procedure was then used as a basis for

controlling climbing behavior in the home.From Day 76 (Fig. 4) the mother no longerspanked S, instead, S was made to sit in achair for a 10-min timeout contingent uponeach instance of climbing. If S did not go tothe chair when instructed, got up from thechair before the mother instructed her toleave, or did not sit quietly in the chair, shockwas applied. Under this procedure inappro-priate climbing in the home occurred at anaverage rate of 2.9 times per day during thenext 50 days (Days 76-125, Fig. 4). Sitting inthe chair was "backed up" with shock on 19%of the occasions when S was sent to the chairfor climbing (approximately one shock everyother day), although the mother's records in-dicate that "shockable offenses" occurred on36% of the occasions.This procedure was not as effective as the

direct use of shock punishment in controllingclimbing and, in fact, resulted in a greater

frequency of shocks. However, this timeoutprocedure was continued because it approxi-mated normal child-rearing procedures and,as such, was also used by the mother to con-trol S's less severe disruptive behaviors, suchas opening the refrigerator, pulling clothesoff of the closet hangers, throwing the potsand pans out of the kitchen cupboards, etc.

Imitation training in the laboratory. Afterthe effects and side effects of shock in elimi-nating disruptive climbing were analyzed,the original program of establishing imitativebehavior was resumed.

S had occasionally exhibited two discreteresponses in the previous sessions, clappingher hands, and pounding on the table withthe palms of one or both hands. Whenever Slooked at (made eye contact with) the experi-menter, he would model one of these twobehaviors for S to imitate. Whenever S emit-ted these behaviors within 5 sec after themodel behavior had been presented, a rein-forcer was delivered. Initially, models for bothclapping and pounding were alternated ran-domly. When no improvement in the fre-quency of imitation was noted after eight ses-sions, only one of the models, clapping, waspresented. The experimenter began clappinghis hands repeatedly and reinforcing S's clap-ping behavior. S exhibited such a low rate ofclapping (two or three per session) that nodiscernible progress was made in two ses-sions. The experimenter then began holdingS's arms and bringing her hands together.Reinforcers were first delivered contingentupon not struggling and then contingentupon slight cooperative movements while theexperimenter was moving her hands. Succes-sively greater force produced by S was thenreinforced while fading out the force sup-plied by the experimenter in bringing herhands together, until the experimenter wouldclap his hands and then just touch S's armsand S would clap her hands. The experi-menter then faded out touching S's arms, firstto a gesture which was made smaller and fi-nally eliminated until, after two sessions Swould respond to the model stimulus of theclap alone. (These procedures are modifica-tions of those developed by Sherman (1965)to reinstate verbal behavior in adult psy-chotics and further developed by Metz (1965)and Baer, Peterson, and Sherman (1965) toestablish verbal behavior in autistic and re-

30

TODD R. RISLEY

tarded children.) While the rate of clappingsystematically increased from an average oftwo to an average of 25 per 10-min period, noimprovement in imitative clapping occurredover five sessions. Imitative claps occurred atan average frequency of 5.3 per 10-min pe-riod, (first fourteen 10-min periods, bottomgraph, Fig. 5) and to only 12% of the modelspresented.Punishment for autistic rocking. Two other

deviant behaviors were recorded. S wouldfrequently strike the side of her head withthe palm of her hand, sometimes resound-ingly, and she spent a significant portion ofthe time in the sessions engaged in autisticrocking behavior. No contingencies were ap-plied to the self-striking behavior, but theirfrequency was recorded. A contact of a handwith the side of her head which resulted inan audible sound was the criterion for record-ing a self-hit. The frequency of self-hittinggradually declined from an average of 77 perhr in the first 15 sessions to 13 per hr in Ses-sions 105 through 117.Rhythmic twisting of the head was the cri-

terion for recording a period of autistic rock-ing. This rocking usually included movementof the shoulders and upper trunk and wasalways accompanied by a monotonic hum-ming. S's eyes were either closed or focusedon her hand, which was held out in front ofher face. Autistic rocking occupied an aver-age of 25% of the time in the session and didnot systematically change over 107, 20- to 30-mmn sessions.Midway through Session 108 (arrow, Fig. 5)

the following procedure was introduced. Theexperimenter shouted "Stop that!", seized Sby the upper arms, and shook her whenevershe began rocking. He would wait until hereyes were closed or fixed on her hand beforeabruptly shouting and shaking her. Thisevent invariably produced a "startle reflex"and flushing in S. This contingency, whichterminated each rocking episode, of course,decreased the time spent rocking from 25%to less than 1% of the session (top graph, Fig.5). More important, the frequency of rockingepisodes also decreased steadily from 0.94 permin in the first session where this contin-gency was applied, to 0.03 per min in thetenth session. This indicated that shoutingand shaking S was a punishing stimulus whichdecreased the probability of the behaviors, in

addition to terminating each occurrence ofthe behavior.The side effects of punishing autistic rock-

ing. When autistic rocking was eliminated bythis punishment procedure imitative clapsimmediately increased to 64% of the modelspresented and to an average rate of 16 per10-min period in the first session (fifteenthto seventeenth 10-min periods, Fig. 5), andcontinued to increase to 76% of the modelspresented and to a rate of 25 per 10-min pe-riod by the fourth session of this procedure(twenty-second and twenty-third 10-min pe-riods, Fig. 5).

3z

200

5w|l14Ft

Irml CLAPS MIXED

10 MINUTE PERIODSFig. 5. Graphs showing the relationship between

autistic rocking and frequency of imitative responses.At the arrows, autistic rocking was punished by theexperimenter shouting at and shaking S. Followingthe increase in frequency of imitative clapping, imi-tative pounding was trained, and then clapping andpounding were both randomly presented.

31

EFFECTS AND SIDE EFFECTS OF PUNISHMENT

In the fifth session after autistic rockinghad been eliminated, the experimenter beganto establish imitative pounding of the table.He pounded on the table contingent uponeye contact and reinforced any pounds whichoccurred within approximately 5 sec afterthis model. Imitative pounds systematicallyincreased from an average of 13% of themodels presented and a rate of three per 10-min period to 93% of the models presentedand a rate of 26 per 10-min period in fivesessions (twenty-fourth to thirty-eighth 10-minperiods, Fig. 5). Models for both pounds andclaps were presented in a random order in thefinal session. Accurate imitation of poundsand claps occurred to 87% of the models pre-sented and at a rate of 23 per 10-min periodin this session (thirty-ninth to forty-first 10-min period, Fig. 5). Thus, punishing autisticrocking not only immediately increased imi-tative clapping but also permitted the rapidestablishment of a new imitative response.

Imitation training in the home. After theprocedures to control S's disruptive behaviorshad been developed and employed in thehome, the mother was able to devote hertime to the training of appropriate be-haviors.The mother had observed the establish-

ment of the two imitative behaviors in thelaboratory. After one session in which themother worked with S under the experi-menter's supervision, the mother began con-ducting imitation training sessions at home.Due to extra-experimental factors, the experi-menter no longer worked with S, and com-municated with S's mother via letter and tele-phone. The mother reported that in 115sessions (a total of 41 hr) she had establishedfive new imitative responses (pounding onwall, stamping feet, standing Up, raisingarms, and placing a hat on head) in additionto the two already established, and all sevenimitative responses would reliably occurwhen their models were presented in randomorder. From data sheets which the motherkept, it appeared that the last of these fiveimitative responses was established in lessthan 1.5 hr of session time. At last report themother was working on imitative mouthmovements in a mirror and imitative behav-iors which produce noises (blowing a whistleand a harmonica, squeezing a horn, etc.) as astep toward establishing verbal imitation.

DISCUSSIONThe failure of the initial attempts to elimi-

nate climbing (in the home with contingenttimeout from social interaction and in thelaboratory with extinction procedures cou-pled with the establishment of incompatiblebehaviors) obviates even a tentative statementabout the variables which maintained this be-havior. It appears that social interaction wasnot functional in maintaining this behavior(although even this statement must be tenta-tive, since the initial procedures were appliedfor a short period of time, relative to thelengthy history of the behavior).Although the electric shock was applied

contingent upon several behaviors, it was sel-dom applied concurrent with those behav-iors. The actual behavior ongoing when theshock was applied was usually vigorous strug-gling. Nevertheless, the shock (preceded by"No!") functioned as a punishing stimulus,decreasing the future probability of the be-haviors. Climbing on the bookcase and stand-ing on her chair in the laboratory, and hit-ting her brother and climbing at home wereall quickly eliminated by the contingent ap-plication of shock. Shouting and shaking Scontingent upon and concurrent with autisticrocking also functioned as a punishing stim-ulus for that behavior. (However, this conse-quence was apparently not a punishing stim-ulus for climbing behavior, as the parentsreportedly had been applying it for severalyears without success.)The original direct effect of the punish-

ment was restricted to the specific stimulusconditions of the presence of the experi-menter in the laboratory room. After punish-ment, climbing occurred in the laboratorywhen the experimenter was absent, at homewhen he was present, but not in the labora-tory in his presence, even when the shock de-vice was absent. Identical stimulus controlwas noted, but not measured, following pun-ishment of autistic rocking. In light of thecontinued climbing at home and in the lab-oratory in the absence of the experimenter,the stimulus control exerted by his presencein the laboratory room was remarkable. Onlyby evoking approximations to climbing byplacing a new piece of furniture in front ofthe bookcase, did climbing recur in the ex-perimenter's presence. The continued, inter-

32

TODD R. RISLEY

mittent occurrences of climbing when S was

alone in the room during the pre-session pe-riods, even when this climbing was punished,was perhaps due to the specificity of the pun-ishment effect to the experimenter's presence.

A prime argument against the use of pun-

ishment procedures is that these procedureswill generate undesirable side effects. Most ofthe behaviors of S were continuously recordedin the laboratory sessions to evaluate the sideeffects of punishment on other behaviors. Sev-eral marked side effects were, in fact, ob-served.When climbing on the bookcase was pun-

ished S began to stand and climb on the seatand back of her chair. This behavior subse-quently varied inversely with climbing on thebookcase. When the punished behavior was

allowed to recover, standing on the chair im-mediately ceased. When climbing was againpunished, standing on the chair again re-

sumed. This side effect corresponds to theclinical model of "symptom substitution" inthat the substituted behavior was topograph-ically similar and similarly undesirable to thepunished behavior. This "contrast effect" mayhave been related specifically to the punish-ment procedure since standing on the chairdid not increase during the nine sessions whenclimbing was eliminated by removing thefurniture. However, when this "symptom"was also punished, no other undesirable be-haviors appeared.No suppression of other behaviors was

noted, either through generalization of thepunishment effect or through conditioned"emotional" suppression, correlated with thepunishment of the target behaviors. On thecontrary, all changes noted in other behav-iors were increases. The brevity of the gen-

eral suppression directly produced by theshock, if any, is indicated by S obtaining andconsuming food within 70 sec after the firstshock.

S quickly learned to jump down from thebookcase in the laboratory, and to sit in thetimeout chair at home, to avoid shock. Al-though strong escape and avoidance behav-iors were produced with the shock, both in-tentionally and inadvertently, no generalavoidance of or attempts to escape from theroom or the experimenter were seen.No aggressive behavior toward any person

or object occurred in the laboratory. When

the aggressive behavior toward her littlebrother (which antedated this study) waspunished with shock, no evidence of pain-elicited aggression was noted, only a system-atic decrease in the behavior.The experimenter was closely paired with

the shock presentations in the laboratory. Theeffects of the punishment were specific to hispresence in the laboratory, attesting to thefact that he was discriminative for shock.However, the only observed alteration in S'sbehavior toward the experimenter followingpunishment was an increased frequency ofattending to (making eye contact with) him.This increase in frequency of eye contactsafter a behavior was eliminated by shock isin marked contrast to the theoretical discus-sion by Hutt and Ounsted (1966) predictingthat increasing the level of arousal and anxi-ety of a child would result in a decrease ineye contacts.The most significant side effect was the

fact that eliminating climbing and autisticrocking with punishment facilitated the ac-quisition of new desirable behaviors. Whenclimbing was occurring, the reinforcementprocedures were ineffective in increasing therate of eye contacts. When climbing was pun-ished the reinforcement procedures produceda steady increase in the rate of eye contacts,which ceased when climbing resumed.Whereas the absolute level of the rate of eyecontacts could be maintained when climbingwas occurring, systematic increases in ratewere produced only when climbing was sup-pressed. An almost identical relationship wasobserved between autistic rocking and rate ofimitation.This effect did not appear to be related to

the punishment procedures per se, but onlyto the presence or absence of the climbing be-havior, as a systematic increase in the rate ofeye contacts also occurred when climbing wassimply precluded by the absence of furniture.However, the relationship between climbingand eye contacts, or autistic rocking and imi-tation, was not simply a function of the physi-cal incompatibility between the behaviors.Subtracting the total time spent climbing orrocking from each session and recomputingthe rate of eye contacts or imitations does notalter the relationships depicted in Fig. 2 and5. The physical incompatibility does notaccount for those relationships. However, the

33

34 EFFECTS AND SIDE EFFECTS OF PUNISHMENT

necessity of eliminating climbing or rockingbefore increases could be obtained in eye con-tacts or imitations demonstrates a relation-ship between the behaviors. This relationshipmight be termed "functional incompatibil-ity". The possibility that the stereotyped be-haviors of deviant children are functionallyincompatible with the establishment of new,socially productive behaviors certainly war-rants further investigation. It may be thatpunishment of stereotyped behaviors couldplay an important role in remediating thedeficits of deviant children.

In summary, this study found that whenpunishment was used to eliminate a child'sdeviant behavior, side effects in the form ofbehavioral contrast or "symptom substitu-tion" did occur, but that these side effectswere primarily desirable. Some deviant be-haviors, maintained by unknown variables,interfered with the establishment of new be-haviors. This interference was not primarilydue to a physical incompatibility between thebehaviors. This interference, which might betermed "functional incompatibility", sug-gests that the elimination of such deviant be-haviors may be a necessary prerequisite to theestablishment of new behaviors.This paper should not be interpreted as a

blanket endorsement of punishment withchildren. In the opinion of the author, thepunishment procedures were therapeuticallyjustified for this child. Shock punishment wasemployed only after other procedures to con-trol disruptive and dangerous behaviors hadbeen extensively but unsuccessfully employed.The possibility of deleterious effects and sideeffects were thoroughly considered beforeshock was used. The effects and side effectswere carefully assessed in the laboratory be-fore shock was employed in the home. The

benefits to the child, in fact, far exceeded theauthor's expectations. Of course, no state-ment about the generality of these findingsto other children can yet be made. However,these findings do serve to limit the generalityof extrapolations from past research whichcontraindicates the use of punishment.

REFERENCESAzrin, N. H. and Holz, W. C. Punishment. In W. K.

Honig (Ed.) Operant behavior: areas of researchand application. New York: Appleton-Century-Crofts, 1966, Pp. 213-270.

Baer, D. M., Peterson, R. F., and Sherman, J. A. Build-ing a generalized imitative repertoire by reinforc-ing similarity to a model. Paper presented at an-nual meeting of the Society for Research in ChildDevelopment, Minneapolis, 1965.

Baer, D. M., Peterson, R. F., and Sherman, J. A. Thedevelopment of imitation by reinforcing behavioralsimilarity to a model. Journal of the ExperimentalAnalysis of Behavior, 1967, 10, 405-416.

Catania, A. C. Concurrent operants. In W. K. Honig(Ed.) Operant behavior: areas of research and ap-plication. New York: Appleton-Century-Crofts,1966, Pp. 213-270.

Hutt, Corinne and Ounsted, C. The biological sig-nificance of gaze aversion with particular referenceto the syndrome of infantile autism. BehavioralScience, 1966, 11, 346-356.

Lovaas, 0. I., Schaeffer, B., and Simmons, J. Q. Build-ing social behavior in autistic children by use ofelectric shock. Journal of Experimental Researchin Personality, 1965, 1, 99-109.

Metz, J. R. Conditioning generalized imitation in au-tistic children. Journal of Experimental Child Psy-chology, 1965, 2, 389-399.

Risley, T. R. and Wolf, M. M. Establishing functionalspeech in echolalic children. Behavior Researchand Therapy, 1967, 5, 73-88.

Sherman, J. A. Use of reinforcement and imitationto reinstate verbal behavior in mute psychotics.Journal of Abnormal Psychology, 1965, 70, 155-164.

Solomon, R. L. Punishment. American Psychologist,1964, 19, 239-253.

Received 2 February 1968.