JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

THE EVOLUTION OF VERBAL BEHAVIORB. F. SKINNER

HARVARD UNIVERSITY

Evolutionary theory has always beenplagued by scantiness of evidence. We see theproducts of evolution but not much of the pro-cess. Most of the story happened long ago, andlittle remains of the early stages. Especiallyfew traces of behavior remain; only recentlywere there artifacts that could endure. Verbalbehavior left no artifacts until the appearanceof writing, and that was a very late stage. Weshall probably never know precisely whathappened, but we ought to be able to say whatmight have happened-that is, what kinds ofvariations and what kinds of contingencies ofselection could have brought verbal behaviorinto existence. Speculation about natural se-lection is supported by current research ongenetics; the evolution of a social environmentor culture is supported by the experimentalanalysis of behavior.

Strictly speaking, verbal behavior does notevolve. It is the product of a verbal environ-ment or what linguists call a language, and itis the verbal environment that evolves. Sincea verbal environment is composed of listeners,it is understandable that linguists emphasizethe listener. (A question often asked, for ex-ample, is, "How is it possible for a person tounderstand a potentially infinite number ofsentences?" In contrast, a behavioral analysisasks, "How is it possible for a person to saya potentially infinite number of sentences?")This paper, then, is about the evolution of averbal environment as the source of the be-havior of the speaker.The plausibility of a reconstruction de-

pends in part upon the size of the variationsthat are assumed to have occurred; the smallerthe variations, the more plausible the expla-nation. Web-making in the spider, for ex-ample, could scarcely have appeared all at oncein its present form as a variation. More plau-sible is a series of small steps. The excretionthat eventually became silk may have begun

Requests for reprints may be sent to the author at De-partment of Psychology, William James Hall, HarvardUniversity, Cambridge, Massachusetts 02138.

as a coating for eggs. It worked better whenit took the form of fibers with which eggs couldbe wrapped rather than coated. The fibershelped the spiders keep from falling as theyworked and did so more effectively as theygrew stronger. The spiders might have begunto lift and lower themselves with the fibers,and strands that were left behind might havecaught insects, which the spiders ate. The morestrands left, the more insects caught. Somepatterns of strands caught more than others.And so on. That may not be exactly whathappened, but it is easier to believe than theappearance of web-making as a sudden, singlevariation. The evolution of behavior is alsomore plausibly regarded as the product of aseries of small variations and selections. It israther like the shaping of operant behaviorthrough small changes in contingencies of re-inforcement, and what we have learned aboutthe operant process helps in understanding thegenetic in spite of the great differences be-tween them.

PHYLOGENIC "SIGNALING"The word "sign" does not commit its user

to any theory of language. Smoke is a sign offire and dark clouds a sign of rain. The growlof a vicious dog is a sign of danger. Organismscome to respond to signs through well knownbehavioral processes. To "signal" is to makea sign; we account for it through the selectingconsequences that would have followed. Fireand rain do not signal, but dogs do, if whatother animals have done when they havegrowled has played a part in the selection ofgrowling. There are difficulties in explainingthe evolution of even that relatively simple ex-ample, however, and other kinds of "signal-ing" raise other problems.

Organisms must have profited from the be-havior of each other at a very early stagethrough imitation. To imitate is more than todo what another organism is doing. Pigeonsforaging in a park are not imitating each other

115

1986, 45, 115-122 NUMBER 1 (JANUARY)

B. F. SKINNER

to any great extent; they are acting indepen-dently under similar environmental contin-gencies. To imitate is to act as another organ-ism is acting because important consequenceshave then followed. The evolution of the pro-cess can be traced to plausible selective con-sequences: The contingencies responsible forthe imitated behavior may affect another or-ganism when it behaves in the same way.Thus, if one of two grazing animals sees a

predator and runs, the other is more likely toescape if it runs too, although it has not seenthe predator. Running whenever another or-

ganism runs usually has survival value.It was only after a tendency to imitate had

evolved that contingencies existed for the evo-

lution of the reciprocal process of modeling.A young bird that would eventually learn tofly without help learns sooner when it imitatesa flying bird. Its parents can speed the processby flying where the young bird can see themand in ways that are easily imitated. To saythat the parents are "showing their young howto fly" adds nothing to such an account andmay imply more than is actually involved.The evolution of other kinds of reciprocally

helpful behavior is not so easily explained.For example, what would have been the sur-

vival value of the dance of the honeybee re-

turning from good forage before other beesresponded to the dance, and how could re-sponding to it have evolved before bees danced?(The question is not raised by imitation andmodeling because the contingencies that ac-

count for imitation do not require modeling.)We must assume that the distance or the di-rection in which the returning bees traveledhad some other effect upon their behavior.Perhaps signs of fatigue varied with the dis-tance, or phototropic movements varied ac-

cording to the position of the sun on theirreturn. Once reciprocal behavior had evolved,further variations could make it more effec-tive. Returning bees could dance in more con-

spicuous ways and other bees could respondmore accurately to features of the dance. It isoften said that bees have a language, that they"tell each other where good forage is to befound," that the dance "conveys information,"and so on. Such expressions, useful enough incasual discourse, add nothing to an explana-tion in terms of natural selection and may ob-scure the processes at issue.

ONTOGENIC "SIGNALING"Contingencies of reinforcement resemble

contingencies of survival in many ways. Ani-mals learn to imitate when, by doing whatothers are doing, they are affected by the samecontingencies-of reinforcement rather thanof survival. Once that has happened, contin-gencies exist in which others learn to model-to behave in ways that can be more easilyimitated. If, for example, a door can be openedonly by sliding it to one side, rather thanpushing or pulling it, a person slides it whenhe sees another person do so, although theother person is not necessarily modeling thebehavior. In such an example, both partiesmay exhibit traces of phylogenic imitation ormodeling, but the operant contingencies wouldsuffice. If the modeler is not close to the door,he can make the kind of movement that wouldopen it if he were-as a gesture. To say thathe is "showing the other how to open the door"is useful in casual discourse but, again, poten-tially troublesome in a scientific account.When a gesture is not a kind of modeling,

we must ask what could have reinforced itbefore anyone responded appropriately, andhow anyone could have learned to respondbefore it had come into existence as a gesture.How, for example, could the gesture withwhich a traffic officer stops an approachingcar have been acquired before people stoppedin response to it, and how could people havelearned to stop before anyone gestured thatway? As in the case of the bees, other contin-gencies related to stopping are needed and, ofcourse, are not hard to find. One person canstop another by placing a hand on his chest,and if the person who is stopped finds thecontact aversive, he will stop on later occa-sions before contact is made. The movementof arm and hand changes from a practical re-sponse to a gesture. Once that has happened,the topography can change until it would havelittle or no physical effect.The gesture that means "Come here" is

another example. It presumably originated aspractical pulling but became effective as a ges-ture when people who were pulled movedquickly to avoid physical contact. The topog-raphy of the gesture still varies with distance,possibly because of its visibility, but also as ifsome practical work remained to be done:When the parties are far apart, the whole arm

EVOLUTION OF VERBAL BEHAVIOR

is moved; when they are fairly near, only theforearm; and when they are close, only thehand or just a finger.

VOCAL BEHAVIORThe human species took a crucial step for-

ward when its vocal musculature came underoperant control in the production of speechsounds. Indeed, it is possible that all the dis-tinctive achievements of the species can betraced to that one genetic change. Other speciesbehave vocally, of course, and the behavior issometimes modified slightly during the life-time of the individual (as in birdsong, for ex-

ample), but there the principal contingenciesof selection have remained phylogenic-eitherphysical (as in echo location) or social. Parrotsand a few other birds imitate human speech,but it is hard to change the behavior or bringit under stimulus control through operantconditioning.Some of the organs involved in the produc-

tion of speech sounds were already subject tooperant conditioning. The diaphragm musthave participated in controlled breathing, thetongue and jaw in chewing and swallowing,the jaw and teeth in biting and tearing, andthe lips in sipping and sucking, all of whichcould be changed through operant condition-ing. Only the vocal cords and pharynx seem

to have served no prior operant function. Theypresumably evolved as organs for the produc-tion of phylogenic calls and cries. The crucialstep in the evolution of verbal behavior ap-pears, then, to have been the genetic changethat brought them under the control of oper-ant conditioning and made possible the coor-

dination of all these systems in the productionof speech sounds. Since other primates havenot taken that step, the change in man was

presumably recent. The possibility that it maynot yet be complete in all members of thespecies may explain why there are so manyspeech disorders-and perhaps even so manyindividual differences in complex verbal be-havior, such as mathematics.

Vocal behavior must have had several ad-vantages in natural selection. Sounds are ef-fective in the dark, around corners, and whenlisteners are not looking, and they can be madewhen the hands are busy with other things.There are special advantages, however, inlarge operant repertoires, especially the enor-

mous variety of available speech sounds. Ges-tures are not as conspicuously different asspeech sounds and hence are fewer in number,and the sounds one produces are more like thesounds one hears than gestures are like thegestures one sees (because they are seen froma different point of view). One learns to ges-ture through movement duplication but tospeak through product duplication, which ismore precise.

It is easier to account for the evolution ofoperant conditioning if we assume that thefirst contingencies of reinforcement closely re-sembled contingencies of natural selection,since only small variations are needed if thesettings, topographies, and consequences aresimilar (see Skinner, 1984). That could havebeen true of vocal operants. The cry of a hun-gry baby, for example, presumably evolved asphylogenic behavior because it alerted the ba-by's parents, but when, through an evolution-ary change, the attention of the parents couldbegin to act as a reinforcer, crying would be-come an operant, with added advantages forbaby and species. Once in existence as an op-erant, however, crying could appear in cir-cumstances too unstable to figure in naturalselection. A baby that was not hungry, forexample, could cry in a manner from whichthe parents would escape by doing things thathad no necessary advantage for the species.A similarity of phylogenic or ontogenic con-

tingencies is not, of course, needed. Coughing,for example, presumably evolved as a reflexthat cleared the throat of irritants, but as soonas the vocal musculature came under operantcontrol, it could be affected by a different con-sequence, such as the attention of a listener.If listeners continued to respond, the topog-raphy could change until it had no effect onthe throat. The cough would become the ver-bal operant "Ahem!" That could have hap-pened before the vocal cords came under op-erant control, and something like it may havebeen the first move from gesture to vocal, butnot voiced, behavior.

Although early vocal operants could havebeen "primed" in this way by phylogenic be-havior, the evolution of operant conditioningappears to have been accompanied by the evo-lution of a pool of behavior that played noother part in natural selection and was there-fore more readily subject to operant reinforce-

117

B. F. SKINNER

ment (see Skinner, 1984). An obvious vocalexample is the babbling of small children-essentially random sounds that, when pickedup by reinforcers, became operants. Verbalbehavior drawn from a pool of uncommittedbehavior has no connection with phylogeniccalls or cries, and in general we have no rea-son to call it an extension of vocal phylogenic"signaling."

A Vocal EpisodeLet us say that two men, A and B, are fish-

ing together. A shallow net containing bait islowered into the water, and when a fish swimsinto the net, it is quickly pulled up. Let ussay that A lowers and raises the net and Btakes a position from which he can moreclearly see it. Anything B does when a fishenters the net will serve as a discriminativestimulus for A, in the presence of which pull-ing will more often be reinforced by the ap-pearance of a fish in the net. B can modelpulling, if he has already learned to model,but nothing more is needed than what wemight call a sign of "excitement" at the pres-ence of a fish in the net or of "annoyance" atA's failure to pull. Whatever the behavior, itbegins to function as a gesture as soon as ithas been reinforced by A's response (and, pre-sumably, by a share of the fish). The behaviorpatterns of both parties then slowly change astheir roles become more sharply defined. Bbecomes more clearly the observer, moving intothe best position to see the fish and gesturingas quickly and as effectively as possible, andA becomes more clearly the actor, watching Bmore closely and pulling as quickly as possiblewhen B responds.

Let us say that, as A and B continue to fishcooperatively, a vocal response (perhaps theundifferentiated Uh, requiring no operantcontrol of the vocal cords) is selected by itsconvenience for B and by the speed and con-sistency with which it reaches A. We couldthen describe the episode in either of two ways.In traditional terms, we could say that "WhenB says Uh, he is telling A that there is a fishin the net" and that he uses Uh as a wordthat "means fish or refers to a fish." Or wecould say that B is "telling A to pull the net,"in which case Uh means "pull."

Cooperative fishing suggests sharing thefish, but the roles are clearer if one party getsthe fish and induces the other to behave byother means. If B gets the fish and arrangesreinforcing consequences for A, Uh would be

classified in several different ways accordingto the kind of consequence arranged. If A pullsbecause in the past B has punished his nothaving pulled, Uh is a command. If B has paidA, it is an order. If the two are friends, dis-posed to help each other, it is a request. Onthe other hand, if A gets the fish and somehowreinforces B's response, Uh would be called a"report" or an "announcement" of the pres-ence of a fish in the net. But although thesetraditional expressions may be useful in cas-ual discourse, they do not take us very fartoward a scientific account. The episode isnothing more than an instance of the recip-rocal behavior of two people, and the contin-gencies that account for it are clear.

Tacts and MandsSomething more is needed if we are to call

Uh either a mand or a tact: The consequencesmust be generalized. The necessary generali-zation presumably came about when therewere many cooperative activities in which asingle object (such as a fish) or a single action(such as pulling) played a part. Fish are pickedup, carried, put down, cleaned, cooked, eaten,and so on. Although things sometimes have,as we say, "different names according to whatis done with them," a single form shouldemerge through stimulus generalization. A tactemerges as the probability of sayingfish in thepresence of a fish when different instances arefollowed by different reinforcing conse-quences, quite apart from any other featureof a particular setting. Perhaps there is thenno particular harm in using traditional wordsand saying that fish "refers to a fish" or"means fish," where the meaning or referentis simply the fish as the principal controllingvariable. To say that the speaker uses the wordto mean fish or to refer to a fish is, however,to get ahead of our story.As a mere probability of responding, the

nature of a tact is clearer when we would notspeak of meaning or reference. Let us say thatwe are calling on someone who has a largesailfish mounted on the wall of his office. Westart looking for something in our briefcaseand, when asked what we are doing, say, "Iam fishing for a letter I want to show you."The fish on the wall has strengthened fish asa tact and has entered into the choice of asynonym. (If, instead, there had been a dis-play of guns on the wall, we might have beenmore likely to say, "I am hunting for a let-ter.") In such a case we do not say that "fish-

118

EVOLUTION OF VERBAL BEHA VIOR

ing" refers to the fish on the wall, even thoughit has been strengthened by it.As a mere probability of responding, a tact

has the same status as three types of verbaloperants that are also not said to mean or referto their controlling variables. One is echoic(we should have been more likely to say fish-ing if someone had just said fish). Another istextual (we should have been more likely tosay fishing if there had been a sign on the wallreading FISH); and a third is intraverbal (weshould have been more likely to say fishing ifwe had just read or heard a word that hasfrequently occurred in proximity with fish).We do not say that fish means or refers to fishwhen it is an echoic, textual, or intraverbalresponse. If we tend to say so when it is a tact,it is not because there is a different kind ofcontrolling relation between stimulus and re-sponse, but rather because the listener re-sponds in more useful ways with respect tothe controlling stimulus.As the mere probability of responding un-

der the control of a stimulus, a tact evolves asa product of many instances in which a re-sponse of a given form has been reinforced inthe presence of a given stimulus in many dif-ferent states of deprivation or aversive stim-ulation. When tacts are taught as "the namesof things," teachers use a generalized reinfor-cer-such as Good! or some other social rein-forcer.A mand is also a by-product of many in-

stances, in which the controlling variable is astate of deprivation or aversive stimulation.The mand pull evolved when responses hav-ing that form were reinforced when listenerspulled different things in different ways upondifferent occasions. It is possible that mandsevolved first, and that they contributed to theevolution of the tact. There are two types ofmand. Pull is an action-mand, reinforced whenthe listener does something. Fish, as short forGive me fish, please is an object-mand, rein-forced by the receipt of fish. An object-mandis more likely to occur in the presence of theobject because it has more often been rein-forced in the presence of that object. We aremuch more likely to ask for the things we seein a shop because asking for presently avail-able objects has more often been reinforced.(That is one reason why shops exhibit theirwares.) The control exerted by the stimulusin an object-mand does not make the responsea tact so long as the reinforcing contingenciesremain those of a mand-so long as saying

fish is reinforced only by the receipt of a fish-but object-mands could have made some con-tribution to the evolution of a tact of the sameform. (It does not follow that a speaker whosays fish as a tact will therefore say it as anobject-mand, or vice versa [see Skinner, 1957].)

The Evolution of the AutocliticIf the occasion upon which a mand or tact

has been reinforced recurs essentially un-changed, the behavior needs no further expla-nation. Reinforcement has had its usual effect.The crucial question is what happens when aperson says something he or she has never saidbefore. Novel behavior occurs upon novel oc-casions, and an occasion is novel in the sensethat its features have not appeared togetherbefore in the same arrangement. Some fea-tures of an occasion strengthen one response,others strengthen another. For example, if twopeople are walking together and one of themfeels a few drops of rain, he may be inclinedto say Rain. The present listener or others likehim have reacted to that response in reinforc-ing ways. He or others like him have alsoreacted in other ways to other features of asetting-when, for example, the speaker hasshown surprise or disappointment. On this oc-casion the speaker may therefore say Rain ina surprised or disappointed tone of voice.Something has been added to the tact. It hasbeen added to other responses in the past withreinforcing consequences but never before toRain. The possibility of recombining the ele-ments of vocal responses in this way accountsfor much of the power and scope of verbalbehavior.

Rather more important collateral effects onthe listener bring us to the evolution of theautoclitic or, in the traditional term, gram-mar. An important consideration for the lis-tener is the extent to which he can react to atact-response effectively. The speaker can helpby indicating the nature and strength of thestimulus control of his behavior. If he has feltonly a few drops of rain, he can speak in thetone of voice transcribed with a question mark:Rain? The listener is not to respond to thetact without reservation. Other elaborations ofthe response are needed if the listener is eitherto respond as he would respond to rain itselfor not to respond at all.

Responses that have such effects are Yesand No. They often appear as mands havingthe effect of Continue and Stop, respectively.Thus, we urge on a speaker who has paused

119

B. F. SKINNER

by saying Yes? or stop him by saying No!Hearing Rain? Yes.!, a listener is more likelyto act as if he had felt rain himself. HearingRain? No., he is less likely to do so. In tra-ditional terms, the speaker asserts or deniesthe presence of rain.A commoner alternative would be It is rain-

ing or It is not raining. Rain? Yes and Rain?No do not have quite the same effect becausethey suggest questions and answers, but some-thing of the thrust of Yes and No remains. Theeffect of Yes can be procured by emphasizingthe word is. The speaker is saying, You cansafely act upon my response Rain. On the otherhand, as a response that brings something thelistener is doing to an end (as in saying No tosomeone about to go the wrong way), No isobviously close to not. It is not raining has theeffect of "There are reasons why I tend to sayRain, but do not act upon my response."The steps through which particular auto-

clitics may have evolved are usually more ob-scure than with mands and tacts. An earlyeffort by John Horne Tooke in the Diversionsof Purley (1786) has not been fully appreci-ated. That Tooke was not always right as anetymologist was not as important as his effortsto explain how English speakers could havecome to say such words as if, but, or and. "Weshall go tomorrow given it does not rain" is aclue to the origin of if. That the boy who stoodon the burning deck should be left out in re-sponding to "Whence all (be out he) had fled"is a clue to but. (That Mrs. Hemans wrote allbut he instead of all but him is unfortunate butirrelevant.) And when we say and we are oftensimply adding:Of shoes-add ships-add sealing wax-

Of cabbages-add kingsAs we should put it today, autoclitics haveevolved as instructions to the listener that helphim behave in ways more likely to have rein-forcing consequences and hence more likely topromote reciprocally reinforcing consequencesfor the speaker.The Evolution of Sentences

It is easy to understand the primitive viewthat behavior is inside the organism before itcomes out. Perhaps there is a touch of theprimitive in saying that behavior is "emitted,"but, as I have pointed out elsewhere, we speakof the emission of light from a hot filamentalthough the light is not in the filament. Thereinforcement that strengthens a response does

not put the response into the organism; it sim-ply changes the organism so that it is morelikely to respond in that way. The point canbe made by distinguishing between an operantas a probability of responding and a responseas an instance. It is the operant that is "in"the organism, but only in the sense in whichelasticity is "in" a rubber band.What is reinforced in the sense of being

followed by a given type of consequences is aresponse; it is the operant that is reinforced inthe quite different sense of being strength-ened. Ferster and I made that distinction inthe glossary of Schedules of Reinforcement(Ferster & Skinner, 1957). In the field of ver-bal behavior it is close to the distinction be-tween the sense of what is said and the saying.The sense of a tact is the controlling vari-able-traditionally, what it means. The say-ing is an instance on a given occasion. It isusually not enough to define "what is said"by describing its topography, as in the mand,"Say 'haRASS, not HArass."' A definitionmust include a reference to controlling vari-ables, as in "What do you say to that?"

In traditional terms the distinction is closeto that between "word" and "sentence." "Sen-tence" comes from the Latin sentire, meaning"to feel, or think." We ask for a sentence whenwe say, "How do you feel about that?" or"What do you think of that?" (A dictionarydefinition of a sentence is "a series of wordsthat expresses a thought." That is an allusion,of course, to another kind of storage. We aresaid to possess thoughts and bring them outor "express" them by putting them intowords.) As I have argued in Verbal Behavior(1957), thinking can be adequately formulat-ed simply as behaving. A sentence is not theexpression of a thought; it is the thought.When we say, "It occurred to me to look inmy desk," we mean that the behavior of look-ing in the desk was strengthened, even if itwas not executed. When we say, "The thoughtoccurred to me that he was embarrassed," wemean that the verbal behavior He is embar-rassed occurred to us, perhaps covertly. Look-ing in the desk is behavior; saying "He is em-barrassed" is behavior. We are especially likelyto call them thoughts when they are not ov-ertly executed.

THE EVOLUTION OF FACTSWhen we speak of the evolution of the au-

tomobile, we do not mean anything like the

120

EVOLUTION OF VERBAL BEHA VIOR

evolution of the horse. We mean the evolutionof certain cultural practices through whichnew ways of making automobiles, as varia-tions, were selected by their contributions toa reinforcing product of human behavior. Someproducts of verbal behavior may be treated inthe same way. Facts, for example.A fact is a statement about the world. When

we say, "The fact of the matter is, I did notattend the meeting," we put the listener in theposition of one who attended the meeting andobserved that the speaker was not there. Onewho has been told "the facts of life" acts ef-fectively with respect to certain aspects of dai-ly existence without passing through a seriesof instructional contingencies. Facts aboutwhat has happened in the past (the facts ofhistory) can be helpful in this sense only tothe extent that the conditions described arelikely to recur. The facts of science are morehelpful than those of history because the rel-evant conditions are more often repeated.We may speak, then, of the evolution of

facts-the facts of daily life, of history, or ofscience. They are often called knowledge. Atissue is not the evolution of knowing or ofknowledgeable persons, or of any organ of sucha person, or of any condition of such an organ,but rather of a verbal environment or culture.People come into contact with such an envi-ronment when they listen to speakers or readbooks. The sounds they hear and the marksthey see affect them as listeners or readers,just as the behavior of the original speakersor writers affected their listeners or readers.We are said to know a fact either because

we have already dealt with the contingenciesor because we have been "told the fact." Thus,we say, "He must have known the door wasunlocked; he would have tried it himself orsomeone would have told him." But there isanother sense in which we may "know" a factsimply as verbal behavior, whether or not itis acted upon. The behavior is intraverbal. Thefacts of history are examples.

There is an important difference betweenthe intraverbals that result from contiguoususage (the house-home kind of thing) and thelarger intraverbals that are learned as such(memorized historical facts or poetry, for ex-ample). By reciting facts as strings of intraver-bal responses, we advise or inform ourselves,as the original speakers or writers addressedor informed their listeners or readers.

COMMENTSIt is inevitable that a continuous process

like evolution should raise the question ofboundaries. Systems for the classification ofspecies are attempts to solve one problem ofthat kind. At what point can we say that manfirst appeared on earth? It may be useful tochoose a given point to improve our use of theterm homo sapiens, but there was presumablyno point at which an essence of man came intoexistence. Similarly, it is only for the sake ofconsistency that we should try to say whenbehavior first became verbal. Taking the fish-ing episode as an example, we could say thatB's response became verbal (1) when it wasfirst strengthened by A's action in pulling thenet (when it became a vocal operant), (2) whenthe same response was made in other settingswith other consequences and came under theexclusive control of a fish as a discriminativestimulus, regardless of any particular state ofdeprivation or aversive stimulation (when itemerged as a tact), or (3) when it was shapedand maintained by a verbal environmenttransmitted from one generation to another(when it became part of a "language"). Theseare all distinguishable steps in the evolutionof verbal behavior, and if we are to choose oneof them, the most useful appears to be (3).Verbal behavior is behavior that is reinforcedthrough the mediation of other people, butonly when the other people are behaving inways that have been shaped and maintainedby a verbal environment or language. At level3 we could say that other primates have en-gaged in verbal behavior in artificial verbalenvironments created by scientists but havenot developed, a language of their own.

Laughing and CryingTwo other functions of the vocal muscula-

ture-laughing and crying-are, if not exclu-sively human, at least highly characteristic ofthe species. There is a good chance that theyevolved at about the same time as vocal be-havior, but they are not operants, althoughthey can be simulated as such-as in cryingto get attention, for example, or laughing po-litely at an unfunny joke. As phylogenic be-havior, they are elicited by positive and neg-ative reinforcers, respectively, often whensudden, but if there is any immediate conse-quence for those who cry and laugh, it isobscure. Laughing and crying may haveevolved because of their effects on others.

121

122 B. F. SKINNER

There are those in whom signs of inflicteddamage shape and maintain aggression, eithernonverbal (a blow) or verbal (an insult) andthere are also those in whom signs of relieffrom damage shape helping others. Otherspecies care for their young and for each oth-er, but presumably not to any great extent asoperant behavior. The human species mayhave gained important advantages when ces-sation of crying began to reinforce the behav-ior we call caring.

Laughing, on the other hand, quite ob-viously reinforces making people laugh and isassociated with caring, for in general peoplelaugh when things go well. Just as a courtshipdance may have evolved because of its effectson other members of a species rather thanupon the dancer, so laughing and crying mayhave evolved because of their effects on othersrather than directly on those who laugh andcry.

TopographyTheorists of the origin of language have

often tried to explain form. Onomatopoeia, forexample, has been said to explain why a dogis called a "bow-wow" and why bacon "hiss-es" or "sizzles" in the frying pan. The gesturefor "stop" is a kind of onomatopoeia, and SirRichard Paget proposed that gesturing withthe tongue may have modified the forms ofuttered sounds in a useful way (Paget, 1930).Onomatopoeia does not take us very far, andit may not be worthwhile to go further. Formsof words can be traced historically, but seldomback to their origins, and the languages of theworld are so diverse that the sources must havebeen largely adventitious. Children invent newforms readily, and when two or more are liv-ing in relative isolation, they may developfairly extensive idiosyncratic vocabularies.There is probably a reason for the form ofevery word, as there is probably a reason forthe color of every bird or flower, but neithermay be worth searching for as a particularfact.When people began to describe the contin-

gencies of reinforcement in the world aroundthem, words could have been invented as thenames of things. The sentence That is called arose describes a contingency of reinforcement

in a verbal environment. Call that a rose isadvice to be followed if one is to behave suc-cessfully in such an environment. Childrensoon learn to ask for the names of things, asthey ask for tools needed to do things, and itmust have been a short step to the inventionof a name (Let's call that a rose). The step istaken whenever parents name a child, al-though very often the form chosen has obvioussources.

CONCLUSIONTo repeat a necessary caveat, I have not

tried to say how a verbal environment, or theverbal behavior generated by such an environ-ment, actually evolved. I have merely tried tosay how it could have evolved, given the be-havioral processes that must already have beenexhibited by the species. The paper is specu-lative, but the speculation is under the re-straint imposed by a commitment to the es-tablished principles of an operant analysis. Inthat respect it may be contrasted with the cur-rent approaches of linguists. A recent book onessentially the present subject lists a numberof explanatory principles or entities, amongthem "innate language organs," "mechanismsof speech perception," "grammatical compe-tences," "cognitive neural substrates," and the"decoding and production functions of spokenlanguage." It is doubtful whether any of thesecan be adequately defined without appealingto the observations they are said to explain,and they do not readily account for verbal be-havior as such.

REFERENCESFerster, C. B., & Skinner, B. F. (1957). Schedules of

reinforcement. New York: Appleton-Century-Crofts.Paget, R. A. S. (1930). Human speech. New York: Har-

court, Brace.Skinner, B. F. (1957). Verbal behavior. New York: Ap-

pleton-Century-Crofts.Skinner, B. F. (1984). The evolution of behavior. Jour-

nal of the Experimental Analysis of Behavior, 41, 217-222.

Tooke, J. H. (1786). The diversions of Purley. London:J. Johnson.

Received July 8, 1985Final acceptance September 28, 1985