The Analysis of Verbal Behavior 1988, 6, 3-9

Establishing Operations and the MandJack Michael

Western Michigan UniversityIn Verbal Behavior Skinner identifies a small number of elementary verbal relations, one of whichis the mand. Because its introduction is at first in terms of unlearned motivative variables, andbecause the mand's relation to prior controlling events is quite complex, its general significancehas probably been underestimated. An extensive treatment of establishing operations, includingthe warning and the blocked-response conditioned establishing operations is provided, followedby a description of the mand in terms of such operations. The importance of the mand for languagetraining programs is suggested, as well as the reasons why it is typically neglected in suchprograms.

In Verbal Behavior Skinner defines themand as ". . . a verbal operant in which theresponse is reinforced by a characteristic con-sequence and is therefore under the func-tional control of relevant conditions ofdeprivation or aversive stimulation.... andin contrast with other types of verbaloperants ... the response has no specifiedrelation to a prior stimulus" (1957, pp. 35-36).In other words, with the mand what is said,written, signed, etc. is primarily determinedby the motivative variable currently in effect.

ESTABLISHING OPERATIONSVariables such as deprivation and aversive

stimulation are identified as motivativebecause of two quite different, but relatedbehavioral effects: (1) They momentarily alterthe reinforcing effectiveness of other events,and (2) they momentarily alter (increase ordecrease) the frequency of the kind ofresponses that have been reinforced by thoseother events. (Response frequency here refersto number of response opportunities inwhich a response occurs as well as numberof responses per unit of time, which makesit possible to avoid having to use the morecontroversial probability or strength ofresponse.) These changes only last as longas the motivative variable is in effect and inthis sense are momentary, as contrasted withthe changes produced by respondent oroperant conditioning or extinction (or by the

Reprints may be obtained from the author, Departmentof Psychology, Western Michigan University, Kalamazoo,MI 49008.

type of pairing that causes stimuli to becomeconditioned reinforcers or conditioned pun-ishers). Thus water deprivation (1) momen-tarily increases the reinforcing effectivenessof any water that should be encountered;and (2) momentarily increases the frequencyof (evokes is a convenient synonym) the var-ious kinds of responses that have beenreinforced with water. Painful stimulation (1)momentarily increases (in this case actuallymakes possible) the effectiveness of painreduction as reinforcement, and (2) evokesthe various kinds of responses that havebeen reinforced by pain reduction.

Keller and Schoenfeld (1950) introducedthe term establishing operation for a variablethat momentarily establishes the reinforcingeffectiveness of some other object or event.Thus water deprivation momentarily estab-lishes water as an effective form of reinforce-ment. The term is useful in that it can referto any operation that has this effect (and thesecond effect seems always to accompanythe first), whether or not the operation seemsto be a form of deprivation or aversivestimulation. Salt ingestion, for example, isnot easily classified as either deprivation oraversive stimulation, but is clearly an estab-lishing operation with effects very similar tothose of water deprivation. Similarly,changes in temperature above or below theoptimal level establish changes in the oppo-site direction as effective reinforcement, andevoke behavior that has been reinforced bysuch changes.The term establishing operation (EO) refers

only to a variable's capacity to produce thetwo defining effects described above. Many

3

4 JACK MICHAEL

stimulus changes that have these two effectshave other effects as well but those are moreproperly identified with other terms. Thus,the onset of painful stimulation is also anunconditioned elicitor of smooth muscle andgland responses, and an unconditioned pun-isher for any behavior that immediatelypreceded the onset of the painful stimula-tion. These latter effects should not be con-fused with the motivational or establishingoperation effects, even though they mayoften occur along with them. (For a syste-matic organization of the terms for respon-dent and operant functional relations seeMichael, 1983.)Water deprivation, painful stimulation,

salt ingestion, and temperature changes pro-bably have their reinforcer-establishingeffects without any learning. In other words,we are probably prewired to be reinforceableby water as a function of water deprivation,or by pain reduction as a result of painfulstimulation. Of course learning plays anobvious role in the development of the par-ticular repertoires that are evoked by waterdeprivation or by painful stimulation-thesecond of the two defining effects of amotivative variable. It is convenient to referto EOs with unlearned reinforcer-establish-ing effects as unconditioned establishingoperations (UEOs). It is quite clear, however,that there are establishing operations whosereinforcer-establishing effects are learned,and these can be referred to as conditionedestablishing operations (CEOs).The most familiar type of CEO is the

auditory or visual stimulus that is pairedwith painful stimulation-usually electricalshock-in an avoidance procedure, theso-called warning stimulus. In the typicalprocedure some arbitrary type of behavior-pressing a lever-terminates the warningstimulus and also prevents the onset of thepainful stimulus. As a result of the systematicand repeated relation of the warning stimu-lus to the painful stimulus, the offset of thewarning stimulus acquires the capacity toreinforce any response that precedes suchoffset, and the onset and presence of thewarning stimulus then comes to evoke suchresponses.The warning stimulus should be con-

sidered a CEO rather than an SD for theresponse that has terminated it just as a pain-

ful stimulus should be considered a UEOrather than an SD for the response that hasterminated it. This point can be most easilymade by reference to a typical laboratoryshock-escape procedure. The responsewhich turns the shock off (often a leverpress) is clearly evoked by the shock onset,and the controlling relation is also clearlyoperant rather than respondent since it wasdeveloped through the use of shock offset asa reinforcing consequence. These two factsmight seem to qualify the shock as adiscriminative stimulus for the response thatterminates the shock, but it differs in twoimportant ways from a discriminative stimu-lus.

In the first place, a discriminative stimulus,or SD, is a stimulus condition that has beencorrelated with the availability of a type ofconsequence given a type of behavior. Pain-ful stimulation is not correlated with theavailability of the reinforcing event that itestablishes-in this case, the termination ofpainful stimulation. Being in pain is notsystematically correlated with being able toremove pain, sad but true, except in thesense that if there were no pain there wouldbe no pain to remove. In other words, thepresence of pain is a necessary, but not a suf-ficient condition for its removal, just as foodor water deprivation is necessary but not suf-ficient for food or water reinforcement. Thestimulation associated with being fooddeprived-to the degree that there is anystimulation-is not also differentiallyassociated with the availability of food. Justbecause an organism is hungry doesn'tmean that food is likely to be available, aswould be pointed out by many currentlyhungry organisms. Likewise with painfulstimulation: It is not differentially correlatedwith the availability of some way to reducethe pain. An SD, on the other hand, is astimulus condition whose presence-absencemust by definition be correlated with theavailability of the relevant consequence giventhe relevant response.Secondly, in the absence of shock, there is

no analog to the extinction responding thatoccurs in the absence of an SD. When shockis absent, the failure of the lever press to ter-minate the "nonpresent" shock is in nosense extinction responding. In a typicallaboratory discrimination procedure withlight-on as SD and light off as SA the

ESTABLISHING OPERATIONS AND THE MAND 5

organism's responses in-the light-off condi-tion fail to produce an event, food for exam-ple, which would clearly be reinforcing if itwere received. In shock-escape the absenceof shock is much more like the absence offood deprivation than like the absence of thelight.The argument for the CEO status of the

warning stimulus in an avoidance procedureis identical in all respects to the argument forconsidering painful stimulation to be a UEOrather than an SD for the response that ter-minates it. The only difference betweenthem is that the warning CEO acquires itsreinforcer-establishing capacity whereas thatof the UEO is unlearned.To state the relation in most general terms,

any form of aversive stimulation, whetherunlearned or learned, will momentarilyincrease the reinforcing effectiveness of itsremoval or attenuation, and momentarilyincrease the frequency of any behavior thathas preceded such removal or attenuation.Likewise, we must suppose that any formof improvement' will also momentarilyincrease the punishing effectiveness of itsremoval or attenuation, and momentarilydecrease the frequency of any behavior thathas preceded such removal or attenuation(although this situation has not been sub-jected to laboratory investigation).Another type of CEO, also likely to be mis-

taken for an SD, is related to the fact that theconditioned reinforcing effectiveness ofmany stimulus changes is itself correlatedwith the presence-absence of other stimulusconditions. This relation is sometimesreferred to as a context effect, in that thefunction of stimulus events as conditionedreinforcers is dependent on the context inwhich they occur. Said another way, the rein-forcing effectiveness of many objects orevents is conditional upon the status of otherstimulus conditions. Now, when these otherconditions assume the proper value, theobject or event becomes an effective form ofreinforcement, and behavior which has beenfollowed by that object or event becomesmore frequent-is evoked. But in this casethe relevant stimulus change does not evoke

'There is no good opposite for aversive stimulation. Rein-forcing stimulation refers too specifically (as well it should)to the increase in future frequency of behavior that haspreceded the stimulation. Appetitive is still too closelyrelated to food.

the behavior because of a relation with theavailability of the object or event, which mayhave been readily available before the stimu-lus change occurred, but because it causesthe object or event to assume reinforcingvalue. In other words, it evokes the responseas an establishing operation rather than as adiscriminative stimulus.For example, two people are walking

together and one sees something that mustbe written down so that it will not be for-gotten-a store name, an address, etc. Thewould-be writer, however, does not have apencil so requests one from the other person,who readily provides it. It might seemreasonable to consider the stimulus respons-ible for the request to have been an SD forthat request, but this is not correct. Theimmediate reinforcement for requests of thistype has clearly been receipt of the thingrequested, in this case a pencil; but thestimulus that evoked the request did not doso because it was an especially favorablesituation for obtaining pencils-the compan-ion would have provided the pencil when-ever requested-but rather because itresulted in the increased reinforcing effec-tiveness of pencils. That is, it did not evokethe request as an SD because of a correlationwith the availability of pencils, but rather asa CEO because of a correlation with the rein-forcing effectiveness of pencils.This type of CEO often seems to be a

stimulus event that functions as an SD for atype of behavior which is in some senseblocked-cannot occur-until some otherobject or event becomes available. The stim-ulus event then also functions as a CEO withrespect to the behavior that has been rein-forced by obtaining this other object or event.Perhaps this type of CEO could be usefullyreferred to as a blocked-response CEO. (This isthe establishing stimulus, or SE, describedin Michael, 1982. In that paper EO referredto unconditioned establishing operationsand SE to the blocked-response conditionedestablishing operation. It now seems betterto use UEO and CEO, with the latter refer-ring to any of the several kinds of CEO. EOcan now refer to any type of motivativevariable, unleamed or learned.)In addition to the warning and the

blocked-response CEOs there may be otherkinds, but these two would seem to be themain ones relevant to the mand, to which weshall now retum.

6 JACK MICHAEL

THE MANDWith this more extended treatment of

motivative variables it becomes possible tosupplement Skinner's mand definition inseveral directions. First, it might seem attrac-tive to avoid dealing with establishing opera-tions entirely and simply relate the mand toits history of reinforcement: ". . . a verbaloperant in which the response is reinforcedby a characteristic consequence" (1957,p. 35). But history of reinforcement onlyexplains the origin of the functional unitinvolving SD, EO, and R; not why theresponse occurs on a particular occasion. Inother words, if a child says water as a mand,the explanation of that instance of behaviorcannot consist simply in the statement thatsuch responses were reinforced with waterin the past, without adding that an EOrelated to water (for example water depriva-tion) was in effect at the time the responseoccurred. The layman would say that thechild knew how to ask for water when thirsty(had the relevant history of reinforcement)and did so at that time because of beingthirsty at that time (the relevant EO was ineffect). Skinner expresses this general pointwhen he says that the mand is under thefunctional control of relevant conditions ofdeprivation and aversive stimulation (1957,p. 35), which brings us to the next point.

It is clear that the terms deprivation andaversive stimulation are not broad enough tocover all of the variables that control themand. Deprivation seems to be an operationthat is primarily relevant to unconditionedestablishing operations or UEOs, andalthough aversive stimulation can includeany operation that can be considered a formof worsening, the blocked-response CEO isnot covered by either term, nor are UEOssuch as salt ingestion, temperature changes,and some others. Of course one might try toconsider the blocked-response CEO to be aform of deprivation in the sense that some-thing is absent, but again such absence is notthe precipitating cause of the relevant behav-ior. In the previous example it was not theabsence of pencils that evoked the request,or there should have been requests for all theother things that were absent. One mightalso try to bring the blocked-response CEOinto the aversive stimulus category, in thesense that it is aversive to need somethingand not to have it, but this, too, seems to

stretch ordinary usage. The obvious solutionis to define the mand as a type of verbaloperant where the response is determinedby a prior EO, as contrasted with the otherverbal operants where it is determined by aprior SD.In Skinner's statement that the response

has no specified relation to a prior stimulus,response is not as appropriate as responseform,and stimulus should be replaced by discrimin-ative stimulus. He is saying that with themand, unlike all of the other elementaryverbal relations, what it is that is actuallysaid, written, signed (as in the sign languageof the deaf), etc. is not determined by a priordiscriminative stimulus. This does not mean(1) that the frequency of occurrence of a mandis unrelated to prior discriminative stimuli,nor does it mean (2) that the fonn of theresponse is not determined by prior stimulifunctioning in some other way than as dis-criminative stimuli-namely as establishingoperations.With regard to the first point, as Skinner

points out (1957, p. 52) prior stimuli are notirrelevant to the actual occurrence of themand response form. Consider the mandwater where the response form is determinedby water deprivation. Saying water in theabsence of an appropriate audience, orunder circumstances where water has neverbeen available has typically undergoneextinction, and thus even under waterdeprivation the response will not ordinarilyoccur until appropriate circumstances are ineffect. The audience or the circumstances areclearly functioning as SDs, but not in thesense of determining the form of theresponse. The EO contributes to an increasein the momentary frequency of water as aresponse form, but SDS related to past rein-forcement of such a response form also con-tribute. In commonsense terms, the depriva-tion produces some tendency to ask forwater, but such asking will not occur undercircumstances where it has been systemat-ically unsuccessful in the past. On the otherhand, even in circumstances where themand water has always been reinforced, theresponse form would not occur unless an EOrelated to water reinforcement was in effect.This is what Skinner means by the statementthat with the mand .... the response has nospecified relation to a prior stimulus" but itseems somewhat more precise to say that

ESTABLISHING OPERATIONS AND THE MAND 7

with the mand the form of the response is notdetermined by a prior discriminative stimulus.Now with respect to the second point,

prior stimuli functioning as EOs may welldetermine the form of the mand response.UEOs such as painful stimulation andtemperature changes are certainly priorstimuli, and even more common, both warn-ing CEOs and blocked-response CEOs areprior stimuli, and are clearly the determ-inants of response form in the mand relation.A revised description of the mand and itscontrast with other verbal operants, then, isas follows: The mand is a type of verbaloperant in which a particular response formis reinforced by a characteristic consequenceand is therefore under the functional controlof the establishing operation relevant to thatconsequence. In the mand the responseform has no specified relation to a priordiscriminative stimulus. The other elemen-tary verbal operants (tact, echoic, etc.) consistof response forms that are reinforced by gen-eralized conditioned reinforcement (Skinner,1957, pp. 53-54) but in the presence of charac-teristic discriminative stimuli, and are there-fore under the functional control of thosediscriminative stimuli. Ideally these otherverbal operants have no specified relation toany establishing operation (but see Chapter6 of Verbal Behavior).

PRACTICAL IMPLICATIONSConsidering only mands controlled by

UEOs, one could easily underestimate theubiquity of the mand. When mands relatedto the warning CEOs and to the blocked-response CEOs are added in, however, it isreasonable to assume that about half of theadult's ordinary daily verbal interaction con-sists of mands. In addition to the mands forobjects and actions there are the mands forSDs and CEOs-that is, for information-that constitute such a large share of what wesay to others. Much of the verbal behaviorcontrolled by other ongoing verbal behaviorin the same speaker, autoclitic verbalbehavior is also a type of mand, but due tothe complexity of this relation it cannot bedealt with in a paper of the present scope (formore on the autoclitic mand see Skinner,1957, pp. 311-367, but especially, pp. 321-330;also Peterson, 1978, pp. 177-180).Underestimation of the mand's impor-

tance in our speculative analysis of everyday

language is of little practical significance.Normal children and adults do not needmuch professional support for their mands,since the mand is the type of verbal behaviorthat directly benefits the speaker. If anything,a more common concern is to induce thosewho mand too much to be more considerateof the needs of others.Of much more practical significance is the

relative neglect of the mand in languagetraining programs for the developmentallydisabled. Such programs devote very littletime to the mand, in favor of training the tactrelation and what is referred to as receptivelanguage. There are several reasons for thisneglect, in addition to general ignorance ofSkinner's analysis of verbal behavior. First,acquiring a verbal repertoire is seen by manyin the speech and language area as learningthe meanings ofwords. It is assumed that whensuch meanings have been acquired thewords can then be used in various ways withno further training. From this perspectivereceptive language training is clearly one ofthe easiest ways to teach such meanings, andtact training is probably next. Based onexperience with normal children and adults,once a person has learned what an object iscalled (by learning to point to it when givenits name, or to say the name when the objectis shown) it is reasonable to assume thatwhen the object becomes important thelearner will be able to ask for it withoutfurther training.

It is clear that this does not happen withlow functioning individuals, many of whomhave had a good deal of receptive languageand tact training but are said to lack a func-tional language repertoire, which is thenexplained in terms of their general intel-lectual deficit. They can often point to severalkinds of objects when the name is spoken,and they can sometimes even say the namewhen the object is shown, but have notendency to request the object when it isclear from other evidence that it would be aneffective form of reinforcement for them.This point was dramatically made in a studyby Hall and Sundberg (1987). Two subjectswere taught to perform a sequence of activi-ties (without responding verbally in anyway) which culminated in the production ofsomething that was known to be effective asreinforcement. For example, both subjectswere taught to make a cup of soup using a

8 JACK MICHAEL

package of instant soup, a bowl, hot water,and a spoon, with the last step being con-suming the soup. In a different setting thesame subjects were taught to tact all of theobjects used in the other setting, but not inthe process of producing and consuming thereinforcer. Then when again attempting toproduce the reinforcer, but the sequencecould not be completed because a criticalobject was missing (the hot water in the caseof making soup), the subjects had no ten-dency to mand the missing object, althoughthey tacted the object in the other setting.When the same subjects were next taught tomand the missing objects by either duplic ortact prompting, they readily learned to do so,and fairly soon acquired a more generaltendency to mand other things when theybecame effective as reinforcement eventhough they had only been formally taughtto tact those things. This suggests that a littlebit of mand training might have dramaticeffects with respect to the development offunctional language.Another reason for the neglect of mand

training, even by those who might wellappreciate its significance, is that the trainermust contrive appropriate EOs, or takeadvantage of those that develop naturally.Contriving a variety of effective EOs for thelearner seems at first glance much more diffi-cult than providing a variety of objects(usually pictures of objects) to be named orpointed at. And relying on naturally occur-ring EOs in a language training setting willnot usually result in sufficient variety,although the variety can be increased byproviding language training under othercircumstances not instituted for that pur-pose. The procedure called incidental teaching(Hart & Risley, 1975) makes some use of thislatter approach, in that verbal prompts formands are provided whenever the learnerneeds help in obtaining some kind of rein-forcement during any training or care-givingactivities. Although it might appear dif-ficult to contrive EOs in the artificial settingof a language training program an under-standing of the blocked-response CEOshould make it easier. In general, wheresome known form of effective reinforcementcannot be obtained without some additionalobject or action, that object or action becomesthe basis for a reinforceable mand. Thisstrategy is well illustrated in the Hall and

Sundberg study mentioned above, andcould be a major part of any language train-ing program.Mand training may also be neglected

because it is not well appreciated that it is theonly type of verbal behavior which directlybenefits the learner. Receptive and tact reper-toires permit the learner to follow directionsgiven by others, and to provide informationto others. Of course such directions and suchinformation may well be to the long rangeadvantage of the learner, but long rangeadvantage is seldom effective as reinforce-ment. There is some evidence (Carroll &Hesse, 1987; Stafford, Sundberg, & Braam,1988) that mand training makes other aspectsof language training more effective. The EOand specific consequences can be used incombination with other variables (nonverbalstimuli, verbal prompts, etc.) to help evokea verbal response, and then faded out oncethe response is strong. A student who cansuccessfully mand for, and receive, specificreinforcement is often much more willing toparticipate in training sessions. Receptive,tact, and intraverbal trials can also beinterspersed with mand trials, and languagetraining becomes much more like typical ver-bal interactions, rather than the standardsituation where the trainer does all the man-ding (e.g., "What's that?" "Touch red.") artdthe student simply complies.A final point concerning the importance

of mand training with the low functioningdevelopmentally disabled client is that itwill often lead to a considerable reduction inthe frequency of various kinds of inappro-priate behavior-crying, aggressive behavior,loud unintelligible vocal responses, etc. AsSundberg (1987) points out, much of thisbehavior is actually under the control ofsome strong EO, but the behavior is eitherinsufficiently specific for the trainer or care-giver to comply, or compliance does occurwhich functions as continued reinforcementfor the inappropriate behavior. An appro-priate mand response, if generally successfulwill displace the inappropriate behavior, andthe client can then function more normally.The neglect of mand training is naturally

paralleled by a neglect of the mand in lan-guage assessment procedures. This wouldbe expected to result often in assessmentsthat credit the client with better languageskills than are actually available. The relative

ESTABLISHING OPERATIONS AND THE MAND 9

lack of functional language would then bethe basis for underestimating the client'sactual potential for language performance. Amore detailed treatment of language assess-ment from the perspective of Skinner'sanalysis of verbal behavior is available inSundberg (1983).

CONCLUSIONSkinner's analysis of language is a major

behavioral breakthrough, with manytheoretical and practical implications. Itsadvantage over traditional language theoryis especially clear in its identification ofelementary verbal relations, and its implica-tions for teaching and learning these rela-tions. Because of the mand's introduction interms of unlearned motivative variables, andbecause its relation to prior controlling eventsis quite complex, its general significance hasprobably been considerably underestimated.A more extensive treatment of establishingoperations and a description of the mandrelation in such terms will, I hope, preventsuch a mistake, as well as lead to a moreappropriate emphasis on the mand inlanguage training programs for subjects whodo not develop normal language.

REFERENCESCarroll, R. J., & Hesse, B. E. (1987). The effects of alter-

nating mand and tact training on the acquisition oftacts. The Analysis of Verbal Behavior, 5, 55-65.

Hall, G., & Sundberg, M. L. (1987). Teaching mands bymanipulating conditioned establishing operations.The Analysis of Verbal Behavior, 5, 41-53.

Hart, B. M., & Risley, T. R. (1975). Incidental teachingof language in the preschool. Journal ofApplied BehaviorAnalysis, 8, 411-420.

Keller, F. S., & Schoenfeld, W. N. (1950). Principles ofpsychology. New York: Appleton-Century-Crofts.

Michael, J. (1982). Distinguishing between discrim-inative and motivational functions of stimuli. Journalof the Experimental Analysis of Behavior, 37, 149-155.

Michael, J. (1983). Evocative and repertoire alteringeffects of an environmental event. TheAnalysis of VerbalBehavior, 2, 21-23.

Peterson, N. (1978). An introduction to verbal behavior.Grand Rapids: Behavior Associates.

Skinner, B. F. (1953). Science and human behavior. NewYork: Macmillan.

Skinner, B. F. (1957). Verbal behavior. New York:Appleton-Century-Crofts.

Stafford, M. W., Sundberg, M. L., & Braam, S. J. (1988).A preliminary investigation of the consequences thatdefine the mand and the tact. The Analysis of VerbalBehavior, 6, 61-71.

Sundberg, M. L. (1983). Language. In J. L. Matson &S. E. Breuning (Eds.) Assessing the mentally retarded (pp.285-310). New York: Grune & Stratton.

Sundberg, M. L. (1987). Teaching language to thedevelopmentally disabled: A course manual. PrinceGeorge, B.C.: College of New Caledonia Press.