PUNISHMENT:

RICHARD L. SOLOMON

University of Pennsylvania

FIRST, an introduction: I will attempt toachieve three goals today, (a) I will sum-marize some empirical generalizations and

problems concerning the effects of punishment onbehavior; (b) I will give some demonstrations ofthe advantages of a two-process learning theoryfor suggesting new procedures to be tried out inpunishment experiments; and (c) finally, I shalltake this opportunity today to decry some un-scientific legends about punishment, and to do alittle pontificating—a privilege that I might bedenied in a journal such as the Psychological Re-view, which I edit!

Now, for a working definition of punishment:The definition of a punishment is not operationallysimple, but some of its attributes are clear. Apunishment is a noxious stimulus, one which willsupport, by its termination or omission, the growthof new escape or avoidance responses. It is onewhich the subject will reject, if given a choicebetween the punishment and no stimulus at all.Whether the data on the behavioral effects ofsuch noxious stimuli will substantiate our common-sense view of what constitutes an effective punish-ment, depends on a wide variety of conditions thatI shall survey. Needless to say, most of theseexperimental conditions have been studied withinfrahuman subjects rather than with humansubjects.

SAMPLE EXPERIMENTS

Let us first consider two sample experiments.Imagine a traditional alley runway, 6 feet long,with its delineated goal box and start box, andan electrifiable grid floor. In our first experiment, arat is shocked in the start box and alley, but thereis no shock in the goal box. We can quickly trainthe rat to run down the alley, if the shock com-mences as the start-box gate is raised and persists

xThis is a slightly revised text of the author's Presi-dential Address to the Eastern Psychological Association,New York City, April 1963. The research associated withthis address was supported by Grant No. M-4202 fromthe United States Public Health Service.

until the rat enters the goal box. This is escapetraining. If, however, we give the rat 5 secondsto reach the goal box after the start-box gate israised, and only then do we apply the shock, therat will usually learn to run quickly enough toavoid the shock entirely. This procedure is calledavoidance training, and the resultant behaviorchange is called active avoidance learning. Notethat the response required, either to terminate theshock or to remove the rat from the presence ofthe dangerous start box and alley, is well specified,while the behavior leading to the onset of thesenoxious stimulus conditions is left vague. It couldbe any item of behavior coming before the openingof the gate, and it would depend on what therat happened to be doing when the experimenterraised the gate.

In our second sample experiment, we train ahungry rat to run to the goal box in order toobtain food. After performance appears to beasymptotic, we introduce a shock, both in thealley and goal box, and eliminate the food. Therat quickly stops running and spends its time inthe start box. This procedure is called the punish-ment procedure, and the resultant learning-to-stay-in-the-start-box is called passive avoidance learn-ing. Note that, while the behavior producing thepunishment is well specified, the particular be-havior terminating the punishment is left vague.It could be composed of any behavior that keepsthe rat in the start box and out of the alley.

In the first experiment, we were teaching therat what to do, while in the second experiment wewere teaching him exactly what not to do; yet ineach case, the criterion of learning was correlatedwith the rat's receiving no shocks, in contrast toits previous experience of receiving several shocksin the same experimental setting. One cannotthink adequately about punishment without con-sidering what is known about the outcomes of bothprocedures. Yet most reviews of the aversive con-trol of behavior emphasize active avoidance learn-ing and ignore passive avoidance learning. I shall,

239

240 AMERICAN PSYCHOLOGIST

in this talk, emphasize the similarities, rather thanthe differences between active and passive avoid-ance learning. I shall point out that there is arich store of knowledge of active avoidance learningwhich, when applied to the punishment procedure,increases our understanding of some of the puzzlingand sometimes chaotic results obtained in punish-ment experiments.

But first, I would like to review some of theempirical generalities which appear to describe theoutcomes of experiments on punishment and pas-sive avoidance learning. For this purpose, Idivide the evidence into S classes: (a) the effectsof punishment on behavior previously establishedby rewards or positive reinforcement, (b) theeffects of punishment on consummatory responses,(c) the effects of punishment on complex, se-quential patterns of innate responses, (d) theeffects of punishment on discrete reflexes, (e) theeffects of punishment on responses previouslyestablished by punishment—or, if you will, theeffects of punishment on active escape and avoid-ance responses. The effectiveness of punishmentwill be seen to differ greatly across these fiveclasses of experiments. For convenience, I meanby effectiveness the degree to which a punish-ment procedure produces suppression of, or facili-tates the extinction of, existing response patterns.

Now, let us look at punishment for instrumentalresponses or habits previously established by re-ward or positive reinjorcers. First, the outcomesof punishment procedures applied to previouslyrewarded habits are strongly related to the in-tensity of the punishing agent. Sometimes in-tensity is independently defined and measured, asin the case of electric shock. Sometimes we havequalitative evaluations, as in the case of Maier's(1949) rat bumping his nose on a locked door, orMasserman's (Masserman & Pechtel, 1953) spidermonkey being presented with a toy snake, orSkinner's (1938) rat receiving a slap on the pawfrom a lever, or my dog receiving a swat from arolled-up newspaper. As the intensity of shockapplied to rats, cats, and dogs is increased fromabout .1 milliampere to 4 milliamperes, theseorderly results can be obtained: (a) detection andarousal, wherein the punisher can be used as a cue,discriminative stimulus, response intensifier, oreven as a secondary reinforcer; (b) temporarysuppression, wherein punishment results in sup-pression of the punished response, followed by

complete recovery, such that the subject laterappears unaltered from his prepunished state;(c) partial suppression, wherein the subject alwaysdisplays some lasting suppression of the punishedresponse, without total recovery; and (d) finally,there is complete suppression, with no observablerecovery. Any of these outcomes can be produced,other things being equal, by merely varying theintensity of the noxious stimulus used (Azrin &Holz, 1961), when we punish responses previouslyestablished by reward or positive reinforcement.No wonder different experimenters report incom-parable outcomes. Azrin (19S9) has produced aresponse-rate increase while operants are punished.Storms, Boroczi, and Broen (1962) have producedlong-lasting suppression of operants in rats.2 Werepunishment intensities different? Were punish-ment durations different? (Storms, Boroczi &Broen, 1963, have shown albino rats to be moreresistant to punishment than are hooded rats, andthis is another source of discrepancy betweenexperiments.)

But other variables are possibly as important aspunishment intensity, and their operation canmake it unnecessary to use intense punishers inorder to produce the effective suppression of aresponse previously established by positive rein-forcement. Here are some selected examples:

1. Proximity in time and space to the punishedresponse determines to some extent the effective-ness of a punishment. There is a response-sup-pression gradient. This has been demonstrated inthe runway (Brown, 1948; Karsh, 1962), in thelever box (Azrin, 1956), and in the shuttle box(Kamin, 1959). This phenomenon has beenlabeled the gradient of temporal delay of punish-ment.

2. The conceptualized strength of a response,as measured by its resistance to extinction afteromission of positive reinforcement, predicts theeffect of a punishment contingent upon the re-sponse. Strong responses, so defined, are moreresistant to the suppressive effects of punishment.Thus, for example, the overtraining of a response,which often decreases ordinary resistance to ex-

2 Since the delivery of this address, several articles haveappeared concerning the punishment intensity problem.See especially Karsh (1963), Appel (1963), and Waltersand Rogers (1963). All these studies support the con-clusion that shock intensity is a crucial variable, andhigh intensities produce lasting suppression effects.

PUNISHMENT 241

perimental extinction, also increases the effective-ness of punishment (Karsh, 1962; Miller, 1960)as a response suppressor.

3. Adaptation to punishment can occur, and thisdecreases its effectiveness. New, intense punishersare better than old, intense punishers (Miller,1960). Punishment intensity, if slowly increased,tends not to be as effective as in the case where itis introduced initially at its high-intensity value.

4. In general, resistance to extinction is de-creased whenever a previously reinforced responseis punished. However, if the subject is habituatedto receiving shock together with positive rein-forcement during reward training, the relationshipcan be reversed, and punishment during extinctioncan actually increase resistance to extinction(Holz & Azrin, 1961). Evidently, punishment,so employed, can functionally operate as a sec-ondary reinforcer, or as a cue for reward, or asan arouser.

5. Punishments become extremely effective whenthe response-suppression period is tactically usedas an aid to the reinforcement of new responsesthat are topographically incompatible with thepunished one. When new instrumental acts areestablished which lead to the old goal (a newmeans to an old end), a punishment of very lowintensity can have very long-lasting suppressioneffects. Whiting and Mowrer (1943) demonstratedthis clearly. They first rewarded one route tofood, then punished it. When the subjects ceasedtaking the punished route, they provided a newrewarded route. The old route was not traversedagain. This reliable suppression effect also seems tobe true of temporal, discriminative restraints onbehavior. The suppression of urination in dogs,under the control of indoor stimuli, is extremelyeffective in housebreaking the dog, as long asurination is allowed to go unpunished under thecontrol of outdoor stimuli. There is a valuablelesson here in the effective use of punishments inproducing impulse control. A rewarded alternative,under discriminative control, makes passive avoid-ance training a potent behavioral influence. It canproduce a highly reliable dog or child. In somepreliminary observations of puppy training, wehave noted that puppies raised in the lab, ifpunished by the swat of a newspaper for eatinghorsemeat, and rewarded for eating pellets, willstarve themselves to death when only given the

opportunity to eat the taboo horsemeat. Theyeagerly eat the pellets when they are available.

It is at this point that we should look at theexperiments wherein punishment appears to haveonly a temporary suppression effect. Most ofthese experiments offered the subject no rewardedalternative to the punished response in attaininghis goal. In many such experiments, it was acase of take a chance or go hungry. Hunger-drivestrength, under such no-alternative conditions,together with punishment intensity, are the crucialvariables in predicting recovery from the suppres-sion effects of punishment. Here, an interesting,yet hard-to-understand phenomenon frequently oc-curs, akin to Freudian "reaction formation." If asubject has been punished for touching somemanipulandum which yields food, he may staynearer to the manipulandum under low hungerdrive and move farther away from it under highhunger drive, even though the probability offinally touching the manipulandum increases ashunger drive increases. This phenomenon iscomplex and needs to be studied in some detail.Our knowledge of it now is fragmentary. It wasobserved by Hunt and Schlosberg (1950) whenthe water supply of rats was electrified, and wehave seen it occur in approach-avoidance conflictexperiments in our laboratory, but we do not knowthe precise conditions for its occurrence.

Finally, I should point out that the attributesof effective punishments vary across species andacross stages in maturational development withinspecies. A toy snake can frighten monkeys. Itdoes not faze a rat. A loud noise terrified Watson'slittle Albert. To us it is merely a Chinese gong.

I have sketchily reviewed some effects of punish-ment on instrumental acts established by positivereinforcers. We have seen that any result onemight desire, from response enhancement and littleor no suppression, to relatively complete suppres-sion, can be obtained with our current knowledgeof appropriate experimental conditions. Now letus look at the effects of punishment on consum-matory acts. Here, the data are, to me, surprising.One would think that consummatory acts, oftenbeing of biological significance for the survival ofthe individual and the species, would be highlyresistant to suppression by punishment. Thecontrary appears to be so. Male sexual behaviormay be seriously suppressed by weak punishment

242 AMERICAN PSYCHOLOGIST

(Beach, Conovitz, Steinberg, & Goldstein, 1956;Gantt, 1944). Eating in dogs and cats can bepermanently suppressed by a moderate shock de-livered through the feet or through the food dishitself (Lichtenstein, 1950; Masserman, 1943).Such suppression effects can lead to fatal self-starvation. A toy snake presented to a spidermonkey while he is eating can result in self-starvation (Masserman & Pechtel, 1953).

The interference with consummatory responsesby punishment needs a great deal of investigation.Punishment seems to be especially effective inbreaking up this class of responses, and one can askwhy, with some profit. Perhaps the intimatetemporal connection between drive, incentive, andpunishment results in drive or incentive becomingconditioned-stimulus (CS) patterns for aversiveemotional reactions when consummatory acts arepunished. Perhaps this interferes with vegetativeactivity: i.e., does it "kill the appetite" in a hungrysubject? But, one may ask why the same punishermight not appear to be as effective when madecontingent on an instrumental act as contrastedwith a consummatory act. Perhaps the nature ofoperants is such that they are separated in timeand space and response topography from consum-matory behavior and positive incentive stimuli, sothat appetitive reactions are not clearly presentduring punishment for operants. We do not knowenough yet about such matters, and speculationabout it is still fun.

Perhaps the most interesting parametric varia-tion one can study, in experiments on the effectsof punishment on consummatory acts, is thetemporal order of rewards and punishments. Ifwe hold hunger drive constant, shock-punishmentintensity constant, and food-reward amounts con-stant, a huge differential effect can be obtainedwhen we reverse the order of reward and punish-ment. If we train a cat to approach a food cup,its behavior in the experimental setting will be-come quite stereotyped. Then, if we introduceshock to the cat's feet while it is eating, the catwill vocalize, retreat, and show fear reactions.It will be slow to recover its eating behavior in thissituation. Indeed, as Masserman (1943) hasshown, such a procedure is likely, if repeated a fewtimes, to lead to self-starvation. Lichtenstein(1950) showed the same phenomenon in dogs.Contrast this outcome with that found when thetemporal order of food and shock is reversed. We

now use shock as a discriminative stimulus to sig-nalize the availability of food. When the cat isperforming well, the shock may produce eating witha latency of less than 5 seconds. The subject's ap-petite does not seem to be disturbed. One can-not imagine a more dramatic difference than thatinduced by reversing the temporal order ofreward and punishment (Holz & Azrin, 1962;Masserman, 1943).

Thus, the effects of punishment are partly deter-mined by those events that directly precede itand those that directly follow it. A punishment isnot just a punishment. It is an event in atemporal and spatial flow of stimulation and be-havior, and its effects will be produced by itstemporal and spatial point of insertion in thatflow.

I have hastily surveyed some of the effects ofpunishment when it has been made contingenteither on rewarded operants and instrumental actsor on consummatory acts. A third class of be-haviors, closely related to consummatory acts,but yet a little different, are instinctive act se-quences: the kinds of complex, innately governedbehaviors which the ethologists study, such asnest building in birds. There has been littleadequate experimentation, to my knowledge, onthe effects of punishment on such innate behaviorsequences. There are, however, some hints ofinteresting things to come. For example, some-times frightening events will produce what theethologists call displacement reactions—the ex-pression of an inappropriate behavior pattern ofan innate sort. We need to experiment with suchphenomena in a systematic fashion. The bestexample I could find of this phenomenon is theimprinting of birds on moving objects, using thelocomotor following response as an index. Moltz,Rosenblum, and Halikas (1959), in one experiment,and Kovach and Hess (1963; see also Hess, 1959a,1959b) in another, have shown that the punish-ment of such imprinted behavior sometimes de-presses its occurrence. However, if birds arepunished prior to the presentation of an imprintedobject, often the following response will be en-ergized. It is hard to understand what this find-ing means, except that punishment can eitherarouse or inhibit such behavior, depending on themanner of presentation of punishment. The sug-gestion is that imprinting is partially a functionof fear or distress. The effectiveness of punish-

PUNISHMENT 243

ment also is found to be related to the criticalperiod for imprinting (Kovach & Hess, 1963).

However, the systematic study of known punish-ment parameters as they affect a wide variety ofcomplex sequences of innate behaviors is yet to becarried out. It would appear to be a worthwhileenterprise, for it is the type of work which wouldenable us to make a new attack on the effects ofexperience on innate behavior patterns. Ulti-mately the outcomes of such experiments couldaffect psychoanalytic conceptions of the effects oftrauma on impulses of an innate sort.8

A fourth class of behavior upon which punish-ment can be made contingent, is the simple, dis-crete reflex. For example, what might happen ifa conditioned or an unconditioned knee jerk werepunished? We are completely lacking in informa-tion on this point. Can subjects be trained toinhibit reflexes under aversive motivation? Ordoes such motivation sensitize and enhance re-flexes? Some simple experiments are appropriate,but I was unable to find them in the publishedwork I read.

A fifth class of behavior, upon which punish-ment can be made contingent, is behavior previ-ously established by punishment procedures', inother words, the effect of passive avoidance train-ing on existing, active avoidance learned responses.This use of punishment produces an unexpectedoutcome. In general, if the same noxious stimulusis used to punish a response as was used to estab-lish it in the first place, the response becomesstrengthened during initial applications of punish-ment. After several such events, however, theresponse may weaken, but not always. The simi-larity of the noxious stimulus used for activeavoidance training to that used for punishment ofthe established avoidance response can be of greatimportance. For example, Carlsmith (1961) hasshown that one can increase resistance to extinc-tion by using the same noxious stimuli for both

3 Since the delivery of this address, an article has appearedon this specific problem. See Adler and Hogan (1963).The authors showed that the gill-extension response ofDelta splendens could be conditioned to a previouslyneutral stimulus by a Pavlovian technique, and it couldalso be suppressed by electric-shock punishment. This is animportant finding, because there are very few knowncases where the same response can be both conditionedand trained. Here, the gill-extension response is typicallyelicited by a rival fish, and is usually interpreted to beaggressive or hostile in nature.

purposes and yet decrease resistance to extinctionby using equally noxious, but discriminatively dif-ferent, punishments. He trained some rats torun in order to avoid shock, then punished themduring extinction by blowing a loud horn. Hetrained other rats to run in order to avoid theloud horn, then during extinction he punished themby shocking them for running. In two controlgroups, the punisher stimulus and training stimuluswere the same. The groups which were trainedand then punished by different noxious stimuliextinguished more rapidly during punishment thandid the groups in which the active avoidancetraining unconditioned stimulus (US) was the sameas the passive avoidance training US. Thus,punishment for responses established originally bypunishment may be ineffective in eliminating theavoidance responses they are supposed to eliminate.Indeed, the punishment may strengthen the re-sponses. We need to know more about this puz-zling phenomenon. It is interesting to me that inJapan, Imada (1959) has been systematicallyexploring shock intensity as it affects thisphenomenon.

Our quick survey of the effects of punishmenton five classes of responses revealed a wide varietyof discrepant phenomena. Thus, to predict in eventhe grossest way the action of punishment ona response, one has to know how that particularresponse was originally inserted in the subject'sresponse repertoire. Is the response an instru-mental one which was strengthened by reward?Is it instead a consummatory response? Is it aninnate sequential response pattern? Is it a dis-crete reflex? Was it originally established by meansof punishment? Where, temporally, in a behaviorsequence, was the punishment used? How intensewas it? These are but a few of the relevant,critical questions, the answers to which are neces-sary in order for us to make reasonable predictionsabout the effects of punishment. Thus, to con-clude, as some psychologists have, that the punish-ment procedure is typically either effective or in-effective, typically either a temporary suppressoror a permanent one, is to oversimplify irresponsiblya complex area of scientific knowledge, one stillcontaining a myriad of intriguing problems forexperimental attack.

Yet, the complexities involved in ascertaining theeffects of punishment on behavior need not be abar to useful speculation ultimately leading to

244 AMERICAN PSYCHOLOGIST

experimentation of a fruitful sort. The com-plexities should, however, dictate a great deal ofcaution in making dogmatic statements aboutwhether punishment is effective or ineffective as abehavioral influence, or whether it is good or bad.I do not wish to do that. I would like now tospeculate about the data-oriented theories, ratherthan support or derogate the dogmas and the socialphilosophies dealing with punishment. I will getto the dogmas later.

THEORY

Here is a theoretical approach that, for me, hashigh pragmatic value in stimulating new lines ofexperimentation. Many psychologists today con-sider the punishment procedure to be a special caseof avoidance training, and the resultant learningprocesses to be theoretically identical in nature.Woodworth and Schlosberg (1954) distinguish thetwo training procedures, "punishment for action"from "punishment for inaction," but assume thatthe same theoretical motive, a "positive incentivevalue of safety" can explain the learning producedby both procedures. Dinsmoor (1955) argues thatthe facts related to both procedures are well ex-plained by simple stimulus-response (S-R) prin-ciples of avoidance learning. He says:

If we punish the subject for making a given responseor sequence of responses—that is, apply aversive stimula-tion, like shock—the cues or discriminative stimuli forthis response will correspond to the warning signals thatare typically used in more direct studies of avoidancetraining. By his own response to these stimuli, the sub-ject himself produces the punishing stimulus and pairsor correlates it with these signals. As a result, they toobecome aversive. In the meantime, any variations in thesubject's behavior that interfere or conflict with the chainof reactions leading to the punishment delay the occurrenceof the final response and the receipt of the stimulationthat follows it. These variations in behavior disrupt thediscriminative stimulus pattern for the continuation ofthe punished chain, changing the current stimulation froman aversive to a nonaversive compound; they are con-ditioned, differentiated, and maintained by the reinforcingeffects of the change in stimulation [p. 96].

The foci of the Dinsmoor analysis are the processeswhereby: (a) discriminative stimuli become aver-sive, and (b) instrumental acts are reinforced. Hestays at the quasi-descriptive level. He uses aperipheralistic, S-R analysis, in which response-produced proprioceptive stimuli and exteroceptivestimuli serve to hold behavior chains together. Herejects, as unnecessary, concepts such as fear or

anxiety, in explaining the effectiveness of punish-ment.

Mowrer (1960) also argues that the facts re-lated to the two training procedures are explainedby a common set of principles, but Mowrer's prin-ciples are somewhat different than those of eitherWoodworth and Schlosberg, or Dinsmoor, citedabove. Mowrer says:

In both instances, there is fear conditioning; and inboth instances a way of behaving is found which eliminatesor controls the fear. The only important distinction, itseems is that the stimuli to which the fear gets con-nected are different. In so-called punishment, thesestimuli are produced by (correlated with) the behavior,or response, which we wish to block; whereas, in so-calledavoidance learning, the fear-arousing stimuli are notresponse-produced—they are, so to say, extrinsic ratherthan intrinsic, independent rather than response-dependent.But in both cases there is avoidance and in both casesthere is its antithesis, punishment; hence the improprietyof referring to the one as "punishment" and to the otheras "avoidance learning." Obviously precision and clarityof understanding are better served by the alternativeterms here suggested, namely, passive avoidance learningand active avoidance learning, respectively. . . . But,as we have seen, the two phenomena involve exactly thesame basic principles of fear conditioning and of thereinforcement of whatever action (or inaction) eliminatesthe fear [pp. 31-32].

I like the simple beauty of each of the threeunifying positions; what holds for punishmentand its action on behavior should hold also forescape and avoidance training, and vice versa.Generalizations about one process should tell .ussomething about the other. New experimentalrelationships discovered in the one experimentalsetting should tell us how to predict a newempirical event in the other experimental setting.A brief discussion of a few selected examples canillustrate this possibility.

APPLICATIONS OF THEORY

I use a case in point stemming from work donein our own laboratory. It gives us new hints aboutsome hidden sources of effectiveness of punish-ment. Remember, for the sake of argument, thatwe are assuming many important similarities toexist between active and passive avoidance-learningprocesses. Therefore, we can look at activeavoidance learning as a theoretical device to sug-gest to us new, unstudied variables pertaining tothe effectiveness of punishment.

Turner and I have recently published an ex-

PUNISHMENT 245

tensive monograph (1962) on human traumaticavoidance learning. Our experiments showed thatwhen a very reflexive, short-latency, skeletal re-sponse, such as a toe twitch, was used as anescape and avoidance response, grave difficultiesin active avoidance learning were experienced bythe subject. Experimental variations which tendedto render the escape responses more emitted, moredeliberate, more voluntary, more operant, or lessreflexive, tended also to render the avoidance re-sponses easier to learn. Thus, when a subject wasrequired to move a knob in a slot in order to avoidshock, learning was rapid, in contrast to the manyfailures to learn with a toe-flexion avoidanceresponse.

There are descriptions of this phenomenon al-ready available in several published experiments onactive avoidance learning, but their implicationshave not previously been noted. When Schlosberg(1934) used for the avoidance response a highlyreflexive, short-latency, paw-flexion response in therat, he found active avoidance learning to be un-reliable, unstable, and quick to extinguish. When-ever the rats made active avoidance flexions, adecrement in response strength ensued. When therats were shocked on several escape trials, theavoidance response tended to reappear for a fewtrials. Thus, learning to avoid was a tortuous,cyclical process, never exceeding 30% success.Contrast these results with the active avoidancetraining of nonreflexive, long-latency operants,such as rats running in Hunter's (1935) circularmaze. Hunter found that the occurrence of avoid-ance responses tended to produce more avoidanceresponses. Omission of shock seemed to reinforcethe avoidance running response. Omission ofshock seemed to extinguish the avoidance pawflexion. Clearly the operant-respondent distinctionhas predictive value in active avoidance learning.

The same trend can be detected in experimentsusing dogs as subjects. For example, Brogden(1949), using the forepaw-flexion response, foundthat meeting a 20/20 criterion of avoidance learn-ing was quite difficult. He found that 30 dogstook from approximately 200-600 trials to reachthe avoidance criterion. The response used was,in our language, highly reflexive—it was totallyelicited by the shock on escape trials with a veryshort latency, approximately .3 second. Com-pare, if you will, the learning of active avoidanceby dogs in the shuttle box with that found in the

forelimb-flexion experiment. In the shuttle box, alarge number of dogs were able to embark on theircriterion trials after 5-15 active avoidance-trainingtrials. Early escape response latencies were long.Resistance to extinction is, across these two typesof avoidance responses, inversely related to trialsneeded for a subject to achieve criterion. Condi-tions leading to quick acquisition are, in this case,those conducive to slow extinction. Our conclu-sion, then, is that high-probability, short-latency,respondents are not as good as medium-probability,long-latency operants when they are required ex-perimentally to function as active avoidance re-sponses. This generalization seems to hold forrats, dogs, and college students.

How can we make the inferential leap from suchfindings in active avoidance training to possiblevariations in punishment experiments? It is rela-tively simple to generalize across the two kinds ofexperiments in the case of CS-US interval, USintensity, and CS duration. But the inferentialsteps are not as obvious in the case of the operant-respondent distinction. So I will trace out thelogic in some detail. If one of the major effects ofpunishment is to motivate or elicit new behaviors,and reinforce them through removal of punishment,and thus, as Dinsmoor describes, establish avoid-ance responses incompatible with a punished re-sponse, how does the operant-respondent distinc-tion logically enter? Here, Mowrer's two-processavoidance-learning theory can suggest a possibleanswer. Suppose, for example, that a hungry rathas been trained to lever press for food and isperforming at a stable rate. Now we make a short-duration, high-intensity pulse of shock contingentupon the bar press. The pulse elicits a startle pat-tern that produces a release of the lever in .2 sec-ond, and the shock is gone. The rat freezes fora few seconds, breathing heavily, and he urinatesand defecates. It is our supposition that a con-ditioned emotional reaction (CER) is therebyestablished, with its major stimulus control comingfrom the sight of the bar, the touch of the bar,and proprioceptive stimuli aroused by the lever-press movements themselves. This is, as Dinsmoordescribes it, the development of acquired aversive-ness of stimuli; or, as Mowrer describes it, theacquisition of conditioned fear reactions. There-fore, Pavlovian conditioning variables should bethe important ones in the development of thisprocess. The reappearance of lever pressing in

246 AMERICAN PSYCHOLOGIST

this punished rat would thus depend on the extinc-tion of the CER and skeletal freezing. If nofurther shocks are administered, then the CERshould extinguish according to the laws of Pavlo-vian extinction, and reappearance of the leverpress should not take long, even if the shock-intensity level were high enough to have been ableto produce active avoidance learning in anotherapparatus.

Two-process avoidance theory tells us thatsomething very important for successful and dura-ble response suppression was missing in the punish-ment procedure we just described. What waslacking in this punishment procedure was a goodoperant to allow us to reinforce a reliable avoid-ance response. Because the reaction to shock wasa respondent, was highly reflexive, and was quickto occur, I am led to argue that the termination ofshock will not reinforce it, nor will it lead to stableavoidance responses. This conclusion follows di-rectly from our experiments on human avoidancelearning. If the termination of shock is madecontingent on the occurrence of an operant,especially an operant topographically incompatiblewith the lever press, an active avoidance learningprocess should then ensue. So I will now proposethat we shock the rat until he huddles in a cornerof the box. The rat will have learned to do some-thing arbitrary whenever the controlling CSs reap-pear. Thus, the rat in the latter procedure, if he isto press the lever again, must undergo two extinc-tion processes. The CER, established by the pair-ing of CS patterns and shock, must become weaker.Second, the learned huddling response must extin-guish. This combination of requirements shouldmake the effect of punishment more lasting, if myinferences are correct. Two problems must besolved by the subject, not one. The experimentsneeded to test these speculations are, it wouldappear, easy to design, and there is no reason whyone should not be able to gather the requisiteinformation in the near future. I feel that thereis much to be gained in carrying on theoreticalgames like this, with the major assumptions being(a) that active and passive avoidance learning aresimilar processes, ones in which the same variableshave analogous effects, and (b) that two proc-esses, the conditioning of fear reactions, and thereinforcement of operants incompatible with thepunished response, may operate in punishmentexperiments.

There is another gain in playing theoretical gamesof this sort. One can use them to question theusual significance imputed to past findings. Take,for example, the extensive studies of Neal Miller(1959) and his students, and Brown (1948) andhis students, on gradients of approach and avoid-ance in conflict situations. Our foregoing analysisof the role of the operant-respondent distinctionputs to question one of their central assumptions—that the avoidance gradient is unconditionallysteeper than is the approach gradient in approach-avoidance conflicts. In such experiments, the sub-ject is typically trained while hungry to run downa short alley to obtain food. After the running isreliable, the subject is shocked, usually near thegoal, in such a way that entering the goal box isdiscouraged temporarily. The subsequent behaviorof the typical subject consists of remaining in thestart box, making abortive approaches to the foodbox, showing hesitancy, oscillation, and various dis-placement activities, like grooming. Eventually, ifshock is eliminated by the experimenter, the subjectresumes running to food. The avoidance tendencyis therefore thought to have extinguished sufficientlyso that the magnitude of the conceptualized ap-proach gradient exceeds that of the avoidancegradient at the goal box. The steepness of theavoidance gradient as a function of distance fromthe goal box is inferred from the behavior of thesubject prior to the extinction of the avoidancetendencies. If the subject stays as far away fromthe goal box as possible, the avoidance gradientmay be inferred to be either displaced upward, orif the subject slowly creeps up on the goal boxfrom trial to trial, it may be inferred to be lesssteep than the approach gradient. Which alterna-tive is more plausible? Miller and his collaboratorsvery cleverly have shown that the latter alternativeis a better interpretation.

The differential-steepness assumption appears tobe substantiated by several studies by Miller andhis collaborators (Miller & Murray, 19S2; Murray& Berkun, 19SS). They studied the displacementof conflicted approach responses along both spatialand color dimensions, and clearly showed that theapproach responses generalized more readily thandid the avoidance responses. Rats whose runningin an alley had been completely suppressed byshock punishment showed recovery of running ina similar alley. Thus the inference made was thatthe avoidance gradient is steeper than is the ap-

PUNISHMENT 247

proach gradient; avoidance tendencies weaken morerapidly with changes in the external environmentalsetting than do approach tendencies. On the basisof the analysis I made of the action of punishment,both as a US for the establishment of a PavlovianCER and as a potent event for the reinforcementof instrumental escape and avoidance responses, itseems to me very likely that the approach-avoidanceconflict experiments have been carried out in sucha way as to produce inevitably the steeper avoid-ance gradients. In other words, these experimentsfrom my particular viewpoint have been inad-vertently biased, and they were not appropriate fortesting hypotheses about the gradient slopes.

My argument is as follows: Typically, the subjectin an approach-avoidance experiment is trained toperform a specific sequence of responses underreward incentive and appetitive drive conditions.He runs to food when hungry. In contrast, whenthe shock is introduced into the runway, it isusually placed near the goal, and no specific, longsequence of instrumental responses is required ofthe subject before the shock is terminated. Thus,the initial strengths of the approach and avoidanceinstrumental responses (which are in conflict) arenot equated by analogous or symmetrical pro-cedures. Miller has thoroughly and carefully dis-cussed this, and has suggested that the avoidancegradient would not have as steep a slope if theshock were encountered by the rat early in therunway in the case where the whole runway iselectrified. While this comment is probably cor-rect, it does not go far enough, and I would liketo elaborate on it. I would argue that if one wantsto study the relative steepnesses of approach andavoidance responses in an unbiased way, the com-peting instrumental responses should be establishedin a symmetrical fashion. After learning to rundown an alley to food, the subject should be shockednear the goal box or in it, and the shock should notbe terminated until the subject has escaped allthe way into the start box. Then one can arguethat two conflicting instrumental responses havebeen established. First, the subject runs one wayfor food; now he runs the same distance in theopposite direction in order to escape shock. Whenhe stays in the start box, he avoids shock entirely.Then the generalization or displacement of theapproach and avoidance responses can be fairlystudied.

I am arguing that we need instrumental-response

balancing, as well as Pav/owaw-conditioning bal-ancing, in such conflict experiments, if the slopesof gradients are to be determined for a test ofthe differential-steepness assumption. Two-processavoidance-learning theory requires such a sym-metrical test. In previous experiments, an aversiveCER and its respondent motor pattern, not a well-reinforced avoidance response, has been pittedagainst a well-reinforced instrumental-approach re-sponse. Since the instrumental behavior of thesubject is being used subsequently to test for theslope of the gradients, the usual asymmetrical pro-cedure is, I think, not appropriate. My guess isthat, if the symmetrical procedure I described isactually used, the slopes of the two gradients willbe essentially the same, and the recovery of thesubject from the effects of punishment will be seento be nearly all-or-none. That is, the avoidancegradient, as extinction of the CER proceeds intime, will drop below the approach gradient, andthis will hold all along the runway if the slopesof the two gradients are indeed the same. Usingthe test of displacement, subjects should stay inthe starting area of a similar alley on initial testsand when they finally move forward they shouldgo all the way to the goal box.

The outcomes of such experiments would be amatter of great interest to me, for, as you willread in a moment, I feel that the suppressive powerof punishment over instrumental acts has beenunderstated. The approach-avoidance conflict ex-periment is but one example among many whereinthe outcome may have been inadvertently biasedin the direction of showing reward-training in-fluences to be superior, in some particular way, topunishment-training procedures. Now let us lookmore closely at this matter of bias.

LEGENDS

Skinner, in 1938, described the effect of a short-duration slap on the paw on the extinction of leverpressing in the rat. Temporary suppression oflever-pressing rate was obtained. When the rateincreased, it exceeded the usual extinction per-formance. The total number of responses beforeextinction occurred was not affected by the punish-ment for lever pressing. Estes (1944) obtainedsimilar results, and attributed the temporary sup-pression to the establishment of a CER (anxiety)which dissipated rapidly. Tolman, Hall, andBretnall (1932) had shown earlier that punish-

248 AMERICAN PSYCHOLOGIST

ment could enhance maze learning by serving as acue for correct, rewarded behavior. Skinner madethese observations (on the seemingly ineffectivenature of punishment as a response weakener) thebasis for his advocacy of a positive reinforcementregime in his Utopia, Wolden Two. In WaldenTwo, Skinner (1948), speaking through the wordsof Frazier, wrote: "We are now discovering at anuntold cost in human suffering—that in the longrun punishment doesn't reduce the probability thatan act will occur [p. 260]." No punishmentswould be used there, because they would producepoor behavioral control, he claimed.

During the decade following the publication ofWalden Two, Skinner (19S3) maintained his posi-tion concerning the effects of punishment on instru-mental responses: Response suppression is buttemporary, and the side effects, such as fear andneurotic and psychotic disturbances, are not worththe temporary advantages of the use of punish-ment. He said:In the long run, punishment, unlike reinforcement worksto the disadvantage of both the punished organism andthe punishing agency [p. 183].

The fact that punishment does not permanently reduce atendency to respond is in agreement with Freud's discoveryof the surviving activity of what he called repressed wishes[p. 184].

Punishment, as we have seen, does not create a negativeprobability that a response will be made but rather apositive probability that incompatible behavior will occur[p. 222].

It must be said, in Skinner's defense, that in 1953he devoted about 12 pages to the topic of punish-ment in his introductory textbook. Other textshad devoted but a few words to this topic.

In Bugelski's (19S6) words about the early workon punishment: "The purport of the experimentsmentioned above appears to be to demonstratethat punishment is ineffective in eliminating be-havior. This conclusion appears to win favor withvarious sentimentalists [p. 275]." Skinner (1961)summarized his position most recently in this way:Ultimate advantages seem to be particularly easy to over-look in the control of behavior, where a quick thoughslight advantage may have undue weight. Thus, althoughwe boast that the birch rod has been abandoned, mostschool children are still under aversive control—not be-cause punishment is more effective in the long run, butbecause it yields immediate results. It is easier for theteacher to control the student by threatening punishmentthan by using positive reinforcement with its deferred,though more powerful, effects [p. 36.08, italics mine].

Skinner's conclusions were drawn over a span oftime when, just as is the case now, there was noconclusive evidence about the supposedly morepowerful and long-lasting effects of positive rein-forcement. I admire the humanitarian and kindlydispositions contained in such writings. But thescientific basis for the conclusions therein wasshabby, because, even in 1938, there were con-flicting data which demonstrated the great effec-tiveness of punishment in controlling instrumentalbehavior. For example, the widely cited experi-ments of Warden and Aylesworth (1927) showedthat discrimination learning in the rat was morerapid and more stable when incorrect responseswere punished with shock than when reward alonefor the correct response was used. Later on,avoidance-training experiments in the 1940s and1950s added impressive data on the long-lastingbehavioral control exerted by noxious stimuli(Solomon & Brush, 1956). In spite of thisempirical development, many writers of books inthe field of learning now devote but a few lines tothe problem of punishment, perhaps a reflection ofthe undesirability of trying to bring satisfyingorder out of seeming chaos. In this category arethe recent books of Spence, Hull, and Kimble.An exception is Bugelski (1956) who devotesseveral pages to the complexities of this topic.Most contemporary introductory psychology textsdevote but a paragraph or two to punishment asa scientific problem. Conspicuously, GeorgeMiller's new book, Psychology, the Science ofMental Life, has no discussion of punishment in it.

The most exhaustive textbook treatment todayis that of Deese (1958), and it is a thoughtful andobjective evaluation, a singular event in this areaof our science. The most exhaustive journalarticle is that by Church (1963), who hasthoroughly summarized our knowledge of punish-ment. I am indebted to Church for letting meborrow freely from his fine essay in prepublicationform. Without this assistance, the organization ofthis paper would have been much more difficult,indeed.

Perhaps one reason for the usual textbookrelegation of the topic of punisment to the fringeof experimental psychology is the wide-spreadbelief that punishment is unimportant because itdoes not really weaken habits; that it pragmat-ically is a poor controller of behavior; that it isextremely cruel and unnecessary; and that it is a

PUNISHMENT 249

technique leading to neurosis and worse. Thislegend, and it is a legend without sufficientempirical basis, probably arose with Thorndike(1931). Punishment, in the time of Thorndike,used to be called punishment, not passive avoid-ance training. The term referred to the use ofnoxious stimuli for the avowed purpose of dis-couraging some selected kind of behavior. Thorn-dike (1931) came to the conclusion that punish-ment did not really accomplish its major purpose,the destruction or extinction of habits. In hisbook, Human Learning, he said:

Annoyers do not act on learning in general by weakeningwhatever connection they follow. If they do anythingin learning, they do it indirectly, by informing the learnerthat such and such a response in such and such a situationbrings distress, or by making the learner feel fear ofa certain object, or by making him jump back from acertain place, or by some other definite and specificchange which they produce in him [p. 46].

This argument is similar to that of Guthrie (193S),and of Wendt (1936), in explaining the extinctionof instrumental acts and conditioned reflexes.They maintained that extinction was not theweakening of a habit, but the replacement of ahabit by a new one, even though the new onemight only be sitting still and doing very little.

When Thorndike claimed that the effects ofpunishment were indirect, he was emphasizing thepower of punishment to evoke behavior other thanthat which produced the punishment; in much thesame manner, Guthrie emphasized the extinctionprocedure as one arousing competing responses.The competing-response theory of extinction todaycannot yet be empirically chosen over othertheories such as Pavlovian and Hullian inhibitiontheory, or the frustration theories of Amsel orSpence. The Thorndikian position on punish-ment is limited in the same way. It is difficult todesignate the empirical criteria which would en-able us to know, on those occasions when punish-ment for a response results in a weakening ofperformance of that response, whether a habit wasindeed weakened or not. How can one tell whethercompeting responses have displaced the punishedresponse, or whether the punished habit is itselfweakened by punishment? Thorndike could nottell, and neither could Guthrie. Yet a legend wasperpetuated. Perhaps the acceptance of thelegend had something to do with the lack of con-certed research on punishment from 1930-1955.

For example, psychologists were not then par-ticularly adventuresome in their search for experi-mentally effective punishments.

Or, in addition to the legend, perhaps a bit ofsoftheartedness is partly responsible for limitingour inventiveness. (The Inquisitors, the Bar-barians, and the Puritans could have given us somegood hints! They did not have electric shock,but they had a variety of interesting ideas, which,regrettably, they often put to practice.) Weclearly need to study new kinds of punishmentsin the laboratory. For most psychologists, apunishment in the laboratory means electric shock.A few enterprising experimenters have used airblasts, the presentation of an innate fear releaser,or a signal for the coming omission of reinforce-ment, as punishments. But we still do not knowenough about using these stimuli in a controlledfashion to produce either behavior suppression,or a CER effect, or the facilitation of extinction.Many aversive states have gone unstudied. Forexample, conditioned nausea and vomiting is easyto produce, but it has not been used in the roleof punishment. Even the brain stimulators,though they have since 1954 tickled brain areasthat will instigate active escape learning, have notused this knowledge to study systematically thepunishing effects of such stimulation on existingresponses.

While the more humanitarian ones of us werebent on the discovery of new positive reinforcers,there was no such concerted effort on the part ofthe more brutal ones of us. Thus, for reasonsthat now completely escape me, some of us in thepast were thrilled by the discovery that, undersome limited conditions, either a light onset or alight termination could raise lever-pressing ratesignificantly, though trivially, above operant level.If one is looking for agents to help in the task ofgetting strong predictive power, and strong con-trol of behavior, such discoveries seem not tooexciting. Yet, in contrast, discoveries already havebeen made of the powerful aversive control of be-havior. Clearly, we have been afraid of theirimplications. Humanitarian guilt and normalkindness are undoubtedly involved, as they shouldbe. But I believe that one reason for our fearhas been the widespread implication of the neu-rotic syndrome as a necessary outcome of all severepunishment procedures. A second reason has beenthe general acceptance of the behavioral phenom-

250 AMERICAN PSYCHOLOGIST

ena of rigidity, inflexibility, or narrowed cognitivemap, as necessary outcomes of experiments inwhich noxious stimuli have been used. I shallquestion both of these conclusions.

If one should feel that the Skinnerian generaliza-tions about the inadequate effects of punishmenton instrumental responses are tinged with a laud-able, though thoroughly incorrect and unscientific,sentimentalism and softness, then, in contrast, onecan find more than a lurid tinge in discussions ofthe effects of punishment on the emotional balanceof the individual. When punishments are assertedto be ineffective controllers of instrumental be-havior, they are, in contrast, often asserted to bedevastating controllers of emotional reactions,leading to neurotic and psychotic symptoms, andto general pessimism, depressiveness, constrictionof thinking, horrible psychosomatic diseases, andeven death! This is somewhat of a paradox, Ithink. The convincing part of such generalizationsis only their face validity. There are experiments,many of them carefully done, in which these neu-rotic outcomes were clearly observed. Gantt's(1944) work on neurotic dogs, Masserman's(1943) work on neurotic cats and monkeys,Brady's (1958) recent work on ulcerous monkeys,Maier's (1949) work on fixated rats, show some ofthe devastating consequences of the utilization ofpunishment to control behavior. The side effectsare frightening, indeed, and should not be ignored!But there must be some rules, some principles,governing the appearance of such side effects, forthey do not appear in all experiments involving theuse of strong punishment or the elicitation ofterror. In Yates' (1962) new book, Frustrationand Conflict, we find a thorough discussion ofpunishment as a creator of conflict. Major at-tention is paid to the instrumental-response out-comes of conflict due to punishment. Phenomenasuch as rigidity, fixation, regression, aggression,displacement, and primitivization are discussed.Yates accepts the definition of neurosis developedby Maier and by Mowrer: self-defeating behaviororiented toward no goal, yet compulsive in quality.The behavioral phenomena that reveal neurosesare said to be fixations, regressions, aggressions,or resignations. But we are not told the necessaryor sufficient experimental conditions under whichthese dramatic phenomena emerge.

Anyone who has tried to train a rat in a T maze,using food reward for a correct response, and shock

to the feet for an incorrect response, knows thatthere is a period of emotionality during earlytraining, but that, thereafter, the rat, when thepercentage of correct responses is high, looks likea hungry, well-motivated, happy rat, eager to getfrom his cage to the experimenter's hand, andthence to the start box. Evidently, merely goingthrough conflict is not a condition for neurosis.The rat is reliable, unswerving in his choices. Ishe neurotic? Should this be called subservientresignation? Or a happy adjustment to an inevit-able event? Is the behavior constricted? Is ita fixation, an evidence of behavioral rigidity? Thecriteria for answering such questions are vaguetoday. Even if we should suggest some specifictests for rigidity, they lack face validity. Forexample, we might examine discrimination reversalas a test of rigidity. Do subjects who have re-ceived reward for the correct response, and punish-ment for the incorrect response, find it harder toreverse when the contingencies are reversed, ascompared with subjects trained with reward alone?Or, we might try a transfer test, introducing oursubject to a new maze, or to a new jumping stand.Would the previously punished subject generalizemore readily than one not so punished? And ifhe did, would he then be less discriminating andthus neurotic? Or, would the previously punishedsubject generalize poorly and hesitantly, thus beingtoo discriminating, and thus neurotic, too? Whatare the criteria for behavioral malfunction as aconsequence of the use of punishment? Wheninstrumental responses are used as the indicator,we are, alas, left in doubt!

The most convincing demonstrations of neuroticdisturbances stemming from the use of punish-ment are seen in Masserman's (Masserman &Pechtel, 1953) work with monkeys. But here thecriterion for neurosis is not based on instrumentalresponding. Instead, it is based on emotionalityexpressed in consummatory acts and innate im-pulses. Masserman's monkeys were frightened bya toy snake while they were eating. Feeding in-hibition, shifts in food preferences, odd sexual be-havior, tics, long periods of crying, were observed.Here, the criteria have a face validity that is hardto reject. Clearly, punishment was a dangerousand disruptive behavioral influence in Masserman'sexperiments. Such findings are consonant withthe Freudian position postulating the pervasiveinfluences of traumatic experiences, permeating all

PUNISHMENT 251

phases of the affective existence of the individual,and persisting for long time periods.

To harmonize all of the considerations I haveraised concerning the conditions leading to neu-rosis due to punishment is a formidable task. Myguess at the moment is that neurotic disturbancesarise often in those cases where consummatorybehavior or instinctive behavior is punished, andpunished under nondiscriminatory control. Butthis is merely a guess, and in order for it to beadequately tested, Masserman's interesting pro-cedures would have to be repeated, using dis-criminative stimuli to signalize when it is safe andnot safe for the monkey. Such experiments shouldbe carried out if we are to explore adequately thepossible effects of punishment on emotionality.Another possibility is that the number of rewardedbehavior alternatives in an otherwise punishingsituation will determine the emotional aftereffectsof punishments. We have seen that Whiting andMowrer (1943) gave their rats a rewarding alter-native, and the resulting behavior was highlyreliable. Their rats remained easy to handle andeager to enter the experimental situation. Oneguess is that increasing the number of behavioralalternatives leading to a consummatory responsewill, in a situation where only one behavior alter-native is being punished, result in reliable be-havior and the absence of neurotic emotionalmanifestations. However, I suspect that matterscannot be that simple. If our animal subject ispunished for Response A, and the punishmentquickly elicits Response B, and then Response Bis quickly rewarded, we have the stimulus con-tingencies for the establishment of a masochistichabit. Reward follows punishment quickly. Per-haps the subject would then persist in performingthe punished Response A? Such questions needto be worked out empirically, and the importantparameters must be identified. We are certainlyin no position today to specify the necessary orsufficient conditions for experimental neurosis.

I have, in this talk, decried the stultifying effectsof legends concerning punishment. To some ex-tent, my tone was reflective of bias, and so Ioverstated some conclusions. Perhaps now itwould be prudent to soften my claims.4 I must

•* Presidential addresses sometimes produce statementsthat may be plausible at the moment, but on secondthought may seem inappropriate. In contrast to my com-plaints about inadequate research on punishment and

admit that all is not lost! Recently, I have noteda definite increase in good parametric studies ofthe effects of punishment on several kinds of be-havior. For example, the pages of the Journalof the Experimental Analysis of Behavior have,in the last 5 years, become liberally sprinkled withreports of punishment experiments. This is aheartening development, and though it comes 20years delayed, it is welcome.

SUMMARY

I have covered a great deal of ground here,perhaps too much for the creation of a clearpicture. The major points I have made are asfollows: First, the effectiveness of punishment asa controller of instrumental behavior varies witha wide variety of known parameters. Some ofthese are: (a) intensity of the punishment stim-ulus, (6) whether the response being punished isan instrumental one or a consummatory one, (c)whether the response is instinctive or reflexive, (d)whether it was established originally by reward orby punishment, (e) whether or not the punish-ment is closely associated in time with the punishedresponse, (/) the temporal arrangements of rewardand punishment, (g) the strength of the responseto be punished, (h) the familiarity of the subjectwith the punishment being used, (i) whether ornot a reward alternative is offered during thebehavior-suppression period induced by punish-ment, (j) whether a distinctive, incompatibleavoidance response is strengthened by omission ofpunishment, (k) the age of the subject, and (I)the strain and species of the subject.

Second, I have tried to show the theoreticalvirtues of considering active and passive avoidancelearning to be similar processes, and have shownthe utility of a two-process learning theory. I havedescribed some examples of the application offindings in active avoidance-learning experimentsto the creation of new punishment experiments andto the reanalysis of approach-avoidance conflictexperiments.

the nature of active and passive avoidance learning areHebb's (1960) recent remarks in his APA PresidentialAddress. He said: "The choice is whether to prosecute theattack, or to go on with the endless and trivial elaborationof the same set of basic experiments (on pain avoidancefor example); trivial because they have added nothingto knowledge for some time, though the early work wasof great value [p. 740]."

252 AMERICAN PSYCHOLOGIST

Third, I have questioned persisting legends con-cerning both the ineffectiveness of punishment as

an agent for behavioral change as well as theinevitability of the neurotic outcome as a legacyof all punishment procedures.

Finally, I have indicated where new experi-mentation might be especially interesting or usefulin furthering our understanding of the effects of

punishment.

If there is one idea I would have you retain, it

is this: Our laboratory knowledge of the effects ofpunishment on instrumental and emotional be-

havior is still rudimentary—much too rudimentaryto make an intelligent choice among conflicting

ideas about it. The polarized doctrines are prob-ably inadequate and in error. The popularizedSkinnerian position concerning the inadequacy of

punishment in suppressing instrumental behavior

is, if correct at all, only conditionally correct. TheFreudian position, pointing to pain or trauma as

an agent for the pervasive and long-lasting dis-

tortion of affective behavior is equally questionable,and only conditionally correct.

Happily, there is now growing attention being

paid to the effects of punishment on behavior, andthis new development will undoubtedly accelerate,because the complexity of our current knowledge,

and the perplexity it engenders, are, I think,

exciting and challenging.

REFERENCES

ADLER, N., & HOGAN, J. A. Classical conditioning andpunishment of an instinctive response in Betta splendens.Anim. Behav., 1963, 11, 3S1-3S4.

APPEL, J. B. Punishment and shock intensity. Science,1963, 141, 528-529.

AZRIN, N. H. Some effects of two intermittent schedulesof immediate and non-immediate punishment. J.Psychol., 1956, 42, 8-21.

AZRIN, N. H. Punishment and recovery during fixed-ratio performance. /. exp. Anal. Behav., 1959, 2, 301-305.

AZRIN, N. H., & HOLZ, W. C. Punishment during fixed-interval reinforcement. J. exp. Anal. Behav., 1961, 4,343-347.

BEACH, F. A., CONOVITZ, M. W., STEINBERG, F., & GOLD-STEIN, A. C. Experimental inhibition and restoration ofmating behavior in male rats. /. genet. Psychol., 1956,89, 165-181.

BRADY, J. V. Ulcers in "executive monkeys." Sclent.American, 1958, 199, 95-103.

BROGDEN, W. J. Acquisition and extinction of conditionedavoidance response in dogs. J. comp. physiol. Psychol.,1949, 42, 296-302.

BROWN, J. S. Gradients of approach and avoidance re-sponses and their relation to level of motivation. /.comp. physiol. Psychol., 1948, 41, 450-465.

BUGELSKI, B. R. The psychology of learning. New York:Holt, 1956.

CARLSMITH, J. M. The effect of punishment on avoidanceresponses: The use of different stimuli for training andpunishment. Paper read at Eastern Psychological As-sociation, Philadelphia, April 1961.

CHURCH, R. M. The varied effects of punishment onbehavior. Psychol. Rev., 1963, 70, 369-402.

DEESE, J. The psychology of learning. New York:McGraw-Hill, 1958.

DINSMOOR, J. A. Punishment: II. An interpretation ofempirical findings. Psychol. Rev., 1955, 62, 96-105.

ESTES, W. K. An experimental study of punishment.Psychol. Monogr., 1944, 57(3, Whole No. 263).

GANTT, W. H. Experimental basis for neurotic behavior.New York: Hoeber, 1944.

GUTHRIE, E. R. The psychology of learning. New York:Harper, 1935. (Rev. ed., 1952)

HEBB, D. 0. The American revolution. Amer. Psy-chologist, 1960, 15, 735-745.

HESS, E. H. Imprinting. Science, 1959, 130, 133-141. (a)HESS, E. H. Two conditions limiting critical age for

imprinting. /. comp. physiol. Psychol., 1959, 52, 515-518. (b)

HOLZ, W., & AZRIN, N. H. Discriminative properties ofpunishment. J. exp. Anal. Behav., 1961, 4, 225-232.

HOLZ, W. C., & AZRIN, N. H. Interactions between thediscriminative and aversive properties of punishment.J. exp. Anal. Behav., 1962, 5, 229-234.

HUNT, J. McV., & SCHLOSBERG, H. Behavior of ratsin continuous conflict. /. comp. physiol. Psychol., 1950,43, 351-357.

HUNTER, W. S. Conditioning and extinction in the rat.Brit. J. Psychol., 1935, 26, 135-148.

IMADA, M. The effects of punishment on avoidance be-havior. Jap. Psychol. Res., 1959, 1, 27-38.

KAMIN, L. J. The delay-of-punishment gradient. /.comp. physiol. Psychol., 1959, 52, 434-437.

KARSH, E. B. Effects of number of rewarded trials andintensity of punishment on running speed. /. comp.physiol. Psychol., 1962, 55, 44-51.

KARSH, E. B. Changes in intensity of punishment: Effecton runway behavior of rats. Science, 1963, 140,1084-1085.

KOVACH, J. K., & HESS, E. H. Imprinting: Effects ofpainful stimulation upon the following response. /.comp. physiol. Psychol., 1963, 56, 461-464.

LICHTENSTEIN, F. E. Studies of anxiety: I. The produc-tion of a feeding inhibition in dogs. /. comp. physiol.Psychol., 1950, 43, 16-29.

MAIER, N. R. F. Frustration: The study of behavior with-out a goal. New York: McGraw-Hill, 1949.

MASSERMAN, J. M. Behavior and neurosis. Chicago:Univer. Chicago Press, 1943.

MASSERMAN, J. M., & PECHTEL, C. Neurosis in monkeys:A preliminary report of experimental observations. Ann.N. Y. Acad. Set., 1953, 56, 253-265.

PUNISHMENT 253

MILLER, N. E. Liberalization of basic S-R concepts: Ex-tensions to conflict behavior, motivation, and sociallearning. In S. Koch (Ed.), Psychology: A study of ascience. Vol. 2. Sensory, perceptual, and physiologicalformulations. New York: McGraw-Hill, 1959. Pp.196-292.

MILLER, N. E. Learning resistance to pain and fear:Effects of overlearning, exposure, and rewarded ex-posure in context. J. exp. Psychol., 1960, 60, 137-14S.

MILLER, N. E., & MTJRRAY, E. J. Displacement and con-flict-learnable drive as a basis for the steeper gradientof approach than of avoidance. J. exp. Psychol., 19S2,43, 227-231.

MOLTZ, H., ROSENBLUM, L., & HALiKAs, N. Imprintingand level of anxiety. J. comp. physiol. Psychol., 1959,52, 240-244.

MOWRER, O. H. Learning theory and behavior. NewYork: Wiley, 1960.

MURRAY, E. J., & BERKUN, M. M. Displacement as afunction of conflict. J. abnorm. soc. Psychol., 1955,51, 47-56.

SCHLOSBERO, H. Conditioned responses in the white rat./. genet. Psychol, 1934, 45, 303-335.

SKINNER, B. F. The behavior of organisms. New York:Appleton-Century, 1938.

SKINNER, B. F. Walden Two. New York: Macmillan,1948.

SKINNER, B. F. Science and human behavior. New York:Macmillan, 1953.

SKINNER, B. F. Cumulative record. New York: Appleton-Century-Crofts, 1961.

SOLOMON, R. L., & BRUSH, E. S. Experimentally derivedconceptions of anxiety and aversion. In M. R. Jones(Ed.), Nebraska Symposium on Motivation: 1956.Lincoln: Univer. Nebraska Press, 1956.

STORMS, L. H., BOROCZI, C., & BROEN, W. E. Punishmentinhibits on instrumental response in hooded rats.Science, 1962, 135, 1133-1134.

STORMS, L. H., BOROCZI, C., & BROEN, W. E. Effects ofpunishment as a function of strain of rat and durationof shock. /. comp. physiol. Psychol., 1963, 56, 1022-1026.

THORNDIKE, E. L. Human learning. New York: Apple-ton-Century-Crofts, 1931.

TOLMAN, E. C., HALL, C. S., & BRETNALL, E. P. A dis-proof of the law of effect and a substitution of the lawsof emphasis, motivation, and disruption. /. exp.Psychol, 1932, 15, 601-614.

TURNER, L. H., & SOLOMON, R. L. Human traumaticavoidance learning: Theory and experiments on theoperant-respondent distinction and failures to learn.Psychol. Monogr., 1962, 76(40, Whole No. 559).

WALTERS, G. C., & ROGERS, J. V. Aversive stimulation ofthe rat: Long term effects on subsequent behavior.Science, 1963, 142, 70-71.

WARDEN, C. J., & AYLESWORTH, M. The relative value ofreward and punishment in the formation of a visual dis-crimination habit in the white rat. J. comp. Psychol,1927, 7, 117-127.

WENDT, G. R. An interpretation of inhibition of con-ditioned reflexes as competition between reaction sys-tems. Psychol Rev., 1936, 43, 258-281.

WHITING, J. W. M., & MOWRER, O. H. Habit progressionand regression—a laboratory study of some factorsrelevant to human socialization. J. comp. Psychol,1943, 36, 229-253.

WOODWORTH, R. S., & SCHLOSBERG, H. Experimentalpsychology. New York: Holt, 1954.

YATES, A. J. Frustration and conflict. New York:Wiley, 1962.