behavan00011-0071

The Behavior Analyst 2000, 23, 69-84 No. 1 (Spring)

Relational Frame Theory and Skinner'sVerbal Behavior: A Possible Synthesis

Dermot Barnes-Holmes and Yvonne Barnes-HolmesNational University of Ireland, Maynooth

Veronica CullinanUniversity College, Cork

The current article suggests a possible synthesis of Skinner's (1957) treatment of verbal behaviorwith the more recent behavioral interpretation of language known as relational frame theory. Therationale for attempting to combine these two approaches is first outlined. Subsequently, each ofthe verbal operants described by Skinner is examined and subjected to a relational frame analysis.In each case, two types of operants are identified; one based on direct contingencies of reinforcementand the other based on arbitrarily applicable relational responding. The latter operants are labeledverbal because they can be distinguished from other forms of social behavior, and they appear topossess the symbolic or referential qualities often ascribed to human language. By applying rela-tional frame theory to Skinner's verbal operants, we aim to contribute towards the development ofa modem behavior-analytic research agenda in human language and cognition.Key words: verbal behavior, relational frame theory, verbal operants, language, cognition

The purpose of the current article isto provide a possible synthesis of Skin-ner's (1957) treatment of verbal behav-ior with the more modem behavioraltreatment of language known as rela-tional frame theory (RFT) (e.g., Hayes,1991; Hayes & Barnes-Holmes, inpress). The current article will not con-tain a detailed description of these twoseparate approaches because this ma-terial is already available. Furthermore,we will not attempt to review any ofthe empirical evidence in favor of ei-ther Skinner's Verbal Behavior or RFEOur aim here is simply to suggest oneway in which two apparently disparateapproaches to the study of human lan-guage, within behavior analysis, mightbe combined at a purely conceptuallevel. If we are successful in this re-gard, we assume that subsequent em-pirical research will either support or

This article is dedicated to the memory of B.F Skinner. We express our gratitude to the re-viewers for providing many constructive andhelpful comments on earlier versions of the cur-rent work. We also thank Steve Hayes for ev-erything!

Requests for reprints may be addressed toDermot Barnes-Holmes, Department of Psy-chology, National University of Ireland, May-nooth, Maynooth, County Kildare, Ireland.

contradict the current work. The articlewill start by outlining the rationale forattempting a synthesis of Skinnerianand RFT approaches to language, andwill then systematically work throughthe verbal operants described by Skin-ner (1957), and subject each of themto a functional RFT analysis.

RATIONALE FOR THEPROPOSED SYNTHESIS

There is a tendency within some ar-eas of behavior analysis to interpret therecent upsurge of research into stimu-lus equivalence, derived stimulus rela-tions, and verbal phenomena as a re-jection of Skinner's (1957) account ofverbal behavior (e.g., see Sidman,1994, pp. 562-573). Although someRFT researchers have criticized certainaspects of Skinner's work (e.g., Hayes,1994; Hayes & Wilson, 1993), it wouldbe a mistake to assume that there isnothing of merit to be found in his ac-count from an RFT perspective. In fact,we believe that combining Skinner'swork with RFT will help us to developa clear and useful research agenda forthe behavior-analytic study of humanlanguage and cognition.The reader may be surprised to learn

69

70 DERMOT BARNES-HOLMES et al.

that almost 10 years ago Chase andDanforth (1991) suggested one way inwhich Skinner's formulations could beintegrated with RFT. These authorsadopted a definition of verbal relationsthat was consistent with Skinner's(1957) analysis, but added one criticalfeature. They defined verbal behavioras a relation in which

(a) A response is emitted by an individual; (b)the critical consequence is provided by the be-havior of another individual (the listener); (c)the listener's behavior is explicitly conditionedto respond to the stimuli produced by the firstindividual; and (d) the explicit conditioning ofthe listener involves conditioning to arbitrarystimulus relations, probably conditioning to re-lational classes, for example, equivalence clas-ses. (Chase & Danforth, 1991, p. 206)

The authors pointed out that Feature adistinguishes behavior from nonbehav-ioral events, Feature b distinguishes so-cial behavior from nonsocial behavior,and Feature c specifies the requirementthat the listener's behavior be condi-tioned to the stimuli produced by thespeaker in order for the listener to re-liably provide consequences for thespeaker's behavior. Chase and Dan-forth added Feature d to Skinner's def-inition for two main reasons. First,most if not all social behavior involvesthe qualities described in Features a, b,and c, and thus at least one other de-fining feature is needed if verbal be-havior is to be distinguished from vir-tually all other forms of social behav-ior. Second, examples of behavior thatare often described as verbal include asymbolic or referential quality (Barnes& Holmes, 1991; Hayes, 1991; Hayes& Hayes, 1989; Skinner, 1986) or gen-eralized relations among arbitrarystimuli (Hayes, 1994; Hayes & Hayes,1989; Skinner, 1986; see also Barnes-Holmes & Barnes-Holmes, in press).By adding Feature d, therefore, Chaseand Danforth (1991) concluded "thatverbal behavior involves arbitrary, so-cial or culturally determined relationsamong events in the world, symbols,pictures, gestures and sounds" (p.206).

Consistent with Chase and Danforth

(1991) (Feature d above), RFT arguesthat verbal behavior involves a historyof reinforcement for responding in ac-cordance with a range of contextuallycontrolled, arbitrarily applicable rela-tions known as relational frames. Thetypes of history and the behavioral pro-cesses involved in relational frameshave been considered, at a conceptuallevel, in a number of other sources (seeBarnes-Holmes & Barnes-Holmes, inpress, for a detailed examination of thisissue; see also Barnes, 1996; Barnes &Hampson, 1993, 1997; Barnes, Healy,& Hayes, in press; Barnes & Roche,1996; Hayes, Gifford, & Wilson,1996). In brief, RFT argues that de-rived relational responding is estab-lished, in large part, by an appropriatehistory of multiple-exemplar training.

For illustrative purposes, considerthe example of derived naming, whichis deemed to be one of the earliest andmore important relational frames(Barnes-Holmes, Barnes-Holmes, &Roche, in press). A caregiver will oftenutter the name of a person in the pres-ence of a young child and then rein-force any orienting response towardsthat person. This interaction may bedescribed as hear Name A -- look atPerson B. Sometimes, the caregiverwill also ask the child the name of aperson in their presence and then mod-el and reinforce an appropriate tact(Skinner, 1957). This interaction maybe described as see Person B -> hearand say Name A. During the earlystages of language training, each inter-action may require explicit reinforce-ment for it to become established in thebehavioral repertoire of the child, butafter a number of name-person andperson-name exemplars have beentrained (along with other name-eventand event-name relations), the gener-alized operant response class of "de-rived naming" is established. In otherwords, from multiple-exemplar train-ing the child's derived naming comesunder abstract control of specific con-textual cues. Suppose, for example, achild with this multiple-exemplar nam-ing history is told "This is Steve."

RELATIONAL FRAME THEORY 71

Contextual cues, such as the word isand the naming context itself, will nowbe discriminative for symmetrical re-sponding between the name and theperson. In the absence of further train-ing, therefore, the child will now pointto Steve when asked "Where isSteve?" (Name A -e Person B) andwill say "Steve" when presented withthe person and asked "Who is this?"(Person B -> Name A) (see Hayes etal., 1996). In fact, recent empirical re-search has begun to explore the role ofmultiple-exemplar training in the ac-quisition of relational frames in youngchildren, and this has provided evi-dence (albeit limited) to support theRFT approach to verbal behavior (e.g.,Barnes-Holmes et al., in press; see alsoLipkens, Hayes, & Hayes, 1993).

Based on both conceptual and em-pirical work in RFT, we take the posi-tion that the behavior of both listenersand speakers is essentially verbal if itinvolves, to some degree, the derivedtransformation of stimulus functions inaccordance with relational frames (seeBarnes, 1994; Barnes et al., in press;Barnes & Holmes, 1991; Hayes, 1994;Hayes & Barnes-Holmes, in press;Hayes & Hayes, 1989). From this po-sition, it becomes possible to useChase and Danforth's (1991) definitionof verbal behavior to reexamine themain classes of verbal operants de-scribed by Skinner (1957), and to in-terpret them from an RFT perspective.We will do this by first presenting aSkinnerian account of each of the mainclasses, in the same order in whichthey appear in Skinner's text (i.e.,mands, echoic behavior, textual behav-ior, transcription and dictation taking,intraverbals, tacts, extended tacts, au-toclitics), followed by an RFT interpre-tation of each class. This will necessi-tate distinguishing two different formsof each verbal operant, one based ondirect contingencies of reinforcementand the other based on arbitrarily ap-plicable relational responding. The for-mer operants we will label nonverbalbecause they cannot be distinguishedreadily from any other form of social

behavior, and they do not possess anyof the referential or symbolic qualitiesusually ascribed to verbal events. Thelatter operants we will label verbal be-cause they can be distinguished fromother forms of social behavior, andthey do appear to possess the symbolicqualities often ascribed to human lan-guage.We believe that making this distinc-

tion will be useful, insofar as it consti-tutes an important step towards speci-fying the behavioral processes that areresponsible for the emergent or gener-ative nature of human language. In-deed, Skinner (1957) was clearly awareof this quality, when he wrote, for ex-ample,Thus, we may hear a man called Jones and seehim respond appropriately to this "vocative."As a result, we may also address him as Jones,or later reply Jones to the question Who isthere? or correctly designate him when askedWhich man is Jones? But this does not happenin the naive speaker or listener; it is the resultof a long process of verbal conditioning [italicsadded]. The young child hearing someone calledJones many times does not therefore himself callhim Jones, nor for this reason report that Joneswas present, nor point to Jones in reply to thequestion Which is Jones? (pp. 359-360).At the time, of course, Skinner had noaccess to the data or to the conceptualwork on derived stimulus relations, sohe was not in a strong position to spec-ulate about the nature of the "long pro-cess of verbal conditioning" involvedin such emergent language phenomena.The research on derived stimulus re-lations is now on hand, so perhaps to-day we are in a better position to spec-ulate usefully about the processes towhich Skinner alluded and to providethe conceptual and empirical ground-work for analyzing and synthesizingthese processes in basic and applied re-search settings (see Hayes & Barnes-Holmes, in press, for an example ofhow this may be achieved).On balance, however, some might

argue that the study of derived stimulusrelations, and RFT in particular, pro-vides very little beyond that providedby Skinner, except perhaps more spec-ulation. In response, we would argue


that although RFI' does indeed incor-porate some speculation as to the pro-cesses involved in generating derivedrelational responding and human lan-guage in general, it has also helped toprovide the experimental proceduresand technical nomenclature with whichthese speculative processes may bestudied (see Hayes & Barnes-Holmes,in press). Indeed, as indicated above,some very recent work has begun toprovide evidence to support the RFTview of the processes involved in thedevelopment of specific verbal skills inchildren (Barnes-Holmes et al., inpress; see also Lipkens et al., 1993). Atthe very least, therefore, RF promisesto supplement Skinner's early work onlanguage by helping to integrate it withthe study of equivalence classes andderived stimulus relations more gener-ally. Given the importance of both ofthese areas to behavior analysis,achieving such an integration mustsurely be a worthwhile objective. Weturn now to the details of the concep-tual integration of Skinner's Verbal Be-havior (1957) and RFT, which is theprimary focus of the current article.

RELATIONAL FRAMETHEORY AND SKINNER'S

VERBAL BEHAVIORThe Mand

Skinner defined a mand as "a verbaloperant in which the response is rein-forced by a characteristic consequenceand is therefore under the functionalcontrol of relevant conditions of dep-rivation or aversive stimulation"(1957, pp. 35-36), or more colloqui-ally as a verbal operant that "speci-fies" its reinforcement. Imagine, forexample, that in the presence of a care-giver a hungry child says "candy,"and this mand is then reinforced by thepresentation of actual candy. The re-inforcer (in this example the candy)need not be present for the response tooccur, but it needs to have reliably fol-lowed previous instances of the mand.One question that may arise at thispoint is how to account for a mand for

a novel object that (a) has never rein-forced the mand in the past, (b) hasnever reinforced a different mand thatmight induce the current mand (i.e., re-sponse induction), (c) does not physi-cally resemble an object that previous-ly reinforced the mand (i.e., primarystimulus generalization), and (d) hasnever been paired with an object thatpreviously reinforced the mand (i.e.,respondent conditioning). (The readeris referred to Barnes & Roche, 1997,and Hayes & Hayes, 1989, 1992, formaterial on the limits of respondentconditioning, including higher orderand sensory preconditioning, as ade-quate explanations for a wide range ofverbal behavior; see also Leader,Barnes, & Smeets, 1996, and Roche &Barnes, 1997, for relevant empiricalevidence.)

Relational frame theory can resolvethe foregoing problem, however, bydistinguishing between two differenttypes of the mand, which we will referto as verbal and nonverbal. The non-verbal type involves the explicit train-ing of a particular mand to a particularobject, as when a child in a toy shopmands for a toy car because in the pastmanding for a car resulted in the parentactually buying the toy car and pre-senting it to the child. The verbal type,however, involves the manded stimulusparticipating in relational frames withother stimuli. In this example, theframe may contain toys in general. Thechild learns to say "toy" in the pres-ence of cars, train sets, dolls, and soforth, and the child learns to mand atleast one toy. Then, the child need notlearn to ask for each specific toy fromscratch; the child merely has to re-spond to the object as participating ina frame of coordination with othertoys. The "toy function" transfersthrough the frame of coordination fromthe toy car to the train set, and thusexplicit reinforcement is not requiredfor the child to mand for the train set(cf. Hall & Sundberg, 1987). In fact,the child may mand repeatedly for thetrain set, despite the fact that the mandis never reinforced with the manded


object (i.e., the parent always refusesto buy the train set).

Before continuing, we should stressthat "pure" verbal manding probablyoccurs only rarely. In effect, manymands are often followed by appropri-ate consequences, and thus mostmands will have some history of ex-plicit differential reinforcement. Nev-ertheless, an explicitly reinforced mandshould still be defined as verbal, whenthe mand also participates in a rela-tional frame. For example, a child'smand "chocolate?" may have been re-inforced in the past on many occasions,but derived relations between the re-sponse "chocolate" and other eventsmay also be present in the child's be-havioral repertoire. For example, choc-olate, candy, and sweets may partici-pate in a frame of coordination, andeach of these words may participate ina frame of difference with words suchas apple and orange. Thus, if the childwere asked, "Would you like some-thing different to chocolate," the re-sponse might be "apple," thereby sug-gesting that the original mand "choc-olate?" was verbal because it partici-pated in relational frames with otherevents.The general point being made here

is so important that it warrants empha-sis. The absence of explicit reinforce-ment (or response induction, or pri-mary stimulus generalization, or re-spondent conditioning) does not definea verbal event. According to the cur-rent thesis, it is the presence of arbi-trarily applicable relational responding,either with or without other behavioralprocesses, that defines a verbal rela-tion. This approach to defining behav-ioral events as verbal applies to all ofthe classes of verbal behavior dis-cussed subsequently in the current ar-ticle (we shall return to this issue incontext of the tact).

Echoic Behavior

This is usually the first outwarddemonstration of vocal verbal behavior

in a young child. Its development canbe summarized as follows:

1. Babbling produces a range of hu-man speech sounds.

2. Only native speech sounds are re-inforced by caregivers.

3. Babbling evolves into self-repeti-tions, and again only native speechsounds are reinforced.

4. Hearing his or her own voice re-produce the native speech sounds be-comes reinforcing because thesesounds have been paired with the pri-mary reinforcers delivered by the care-givers.

5. Through induction and general-ization the infant begins to repeat thespeech sounds of others as well as itsown.

Although echoic behavior is char-acteristic of infant vocalizations, it isnot solely a feature of language acqui-sition; it is also demonstrated by adultspeakers. Skinner did not distinguishbetween the echoic responses of in-fants and adults, but RFT suggests thatthere are two different types of echoicbehavior, nonverbal and verbal. Thenonverbal type typically involves thesimple infant-like imitation of the wordmama, where mama does not partici-pate in a frame of coordination withthe child's mother. The verbal type ofthe echoic requires that mama partici-pate in relational frames. For example,the word mama may enter into a rela-tional frame of coordination with thechild's actual mother and words andphrases such as mother and parent.Other frames may also be involved,such as different (e.g., mama is differ-ent than dada). In summary, therefore,the unit of echoic behavior is definedby the vocal correspondences (nonver-bal echoic) and also, in many cases, bythe relational frames (verbal echoic)that may be involved in this behavioralunit.

Textual BehaviorLike echoic behavior, textual behav-

ior is verbal behavior under the controlof a verbal stimulus. In this case, how-


ever, the verbal stimulus is visual, as ina written text (or tactual, as in Braille),rather than auditory. An RFT interpre-tation of textual behavior involves asimilar analysis to that offered forechoic behavior. It suggests two differ-ent types of textual behavior. The firstof these is nonverbal textual behavior,which involves a speaker uttering a vo-cal response to written material with-out the words or phrases participatingin equivalence (or other) relations withother words, phrases, or events. Thesecond type suggested by RFT is ver-bal textual behavior (with "under-standing"), which involves the vocalresponse participating in relationalframes. To illustrate the distinction be-tween these two types of textual be-havior, imagine a father reading a bed-time story to his child. As the fatherreads the first few pages of the text, thefunctions of the events described in thestory participate in relational frames(i.e., verbal textual behavior). The fa-ther, if asked, could describe the eventswith synonyms, fill in details as if hewere "seeing" the events, and so forth.At a later point, however, he may re-alize that although he has been accu-rately reading the words on the page(evidenced by the child's lack of com-plaint), he has no idea of the contentof what he most recently read. In otherwords, the text evoked the appropriatevocal responses, but produced none orvery few of the functions of the eventsbeing described (i.e., nonverbal textualbehavior).

Transcription and Dictation TakingThe behaviors examined thus far

have all involved vocal responses. Dic-tation taking and transcription involvemotor responses to vocal (dictation) orwritten (transcription) stimuli. Consis-tent with the previous analyses, RFTdistinguishes between verbal and non-verbal dictation taking and transcrip-tion behavior. The verbal types of theseresponses involve "writing with mean-ing," whereby the words or phraseswritten participate in relational frames.

That is, each word or phrase that istranscribed or taken down as dictationproduces specific psychological func-tions for the individual. For example,if the phrase "The goods will be deliv-ered tomorrow" is dictated (or writ-ten), this may cause the person takingthe dictation (or transcribing) to "see"privately the future arrival of thegoods, and perhaps to make arrange-ments for their arrival after the dicta-tion session (or transcription) is over.The nonverbal types of transcriptionand dictation taking do not produceany psychological functions via rela-tional frames. For example, a skilledsecretary may sometimes successfullytake dictation without any "aware-ness" of the content of what was dic-tated and written (i.e., the spokenwords and written text produce none oronly a few of the functions of theevents described therein). In this case,he or she may fail to make the neces-sary arrangements for the arrival of thegoods mentioned in the dictated letter.

The IntraverbalIntraverbals involve responses to

stimuli when no formal correspon-dence exists between the stimulus andthe response. For example, when pre-sented with the stimulus "two plustwo," most numerate English speakerswill respond with "four," or whenasked "how are you," in the course ofa casual meeting, most speakers willreply "fine, thank you," regardless oftheir current physical state. Again,RFT makes a distinction between ver-bal and nonverbal types of this operantclass. To appreciate this distinction,consider a parrot that is trained throughexplicit reinforcement to emit an in-traverbal response, such as responding"one, two, three" when presented withthe stimulus "count to three." In sucha case, the response would not partic-ipate in relational frames with otherwords or events in the world, andtherefore would be considered a non-verbal intraverbal. In contrast, a verbalintraverbal may involve the response


participating in a network of relationswith other words, phrases, or events inthe world. So, for example, a reason-ably numerate child could respond ap-propriately not only to the stimulus"count to three," but also to "countbackwards from three" or "whatcomes between one and three?" In ef-fect, the individual words one, two, andthree participate in a relational frameof comparison, such that the intraver-bal counting response "one, two,three" entails "three after two afterone" or entails "one before two, andthree after two."

The TactVerbal behavior under the control of

stimuli from the "world of things andevents which a speaker is said to 'talkabout' " (Skinner 1957, p. 81) hasbeen called tacting. Skinner saw thetact as the most important verbal op-erant because of the unique control ex-erted by the prior stimulus. This priorstimulus can be a particular object orevent or some property of an object orevent. For example, if a child learns tosay "square" in the presence of asquare, the child is said to be tactingthe square. This behavior is establishedby a history of differential reinforce-ment for emitting the vocal response"square" in the presence of a square.A relational frame analysis of tact-

ing behavior argues that the tact, as de-scribed by Skinner, is not necessarily averbal response because it may not in-volve stimuli that participate in arbi-trarily applicable relations with otherstimuli. Therefore, RFT distinguishesbetween verbal and nonverbal tactingin the following way. Imagine that asquare occasions the response"square" because, in the past, the re-sponse has been reinforced in the pres-ence of squares. This type of behavior-al relation is functionally similar to therelation that is established when a pi-geon learns to peck a key for food inthe presence of a square projected ona translucent response key. This is verydifferent from the verbal tact relation,

for which no explicit history of rein-forcement is required for the tact rela-tion to emerge. In verbal tacting, achild may respond with "square" inthe presence of a square box, for ex-ample, because the box participates ina relational frame with the word squareand other square objects (i.e., explicitreinforcement has never been providedfor the tact). Imagine, for instance, thatthe child was taught to tact a box (e.g.,a cereal box) and was then told that abox is often square (i.e., the cereal box,the word square, and actual squareboxes now participate in a relationalframe of coordination). As a result,when presented with a box and asked"what is this," the child may producea derived or verbal tact by respondingwith "square," rather than the explic-itly reinforced tact (i.e., "box").

In the natural environment, pure ver-bal tacting probably occurs only rarely;most tacting does not occur withoutconsequences (i.e., one does not moveabout one's environment tacting every-thing one sees). Thus, most tacts willhave some history of explicit differ-ential reinforcement. Consistent withour previous discussion of this issue (inthe context of the mand), it is useful todefine an explicitly reinforced tact asverbal when the tacted stimulus partic-ipates in relational frames with otherstimuli. In doing so, we make a clearfunctional distinction between formallysimilar responses that arise from verydifferent behavioral histories. If behav-ior analysts ignore or gloss over thesedifferent histories they may find it dif-ficult, in certain contexts, to predictand to control tacting behavior. In theprevious example, for instance, "box"was explicitly reinforced as a tact, buton a subsequent occasion the tact"'square" was emitted in the presenceof a box. Distinguishing between ver-bal and nonverbal tacting will help topredict and to control these apparentlyunexpected behavioral outcomes.The Extended Tact

"If a response is reinforced upon agiven occasion or class of occasions,


any feature of that occasion or com-mon to that class appears to gain somemeasure of control. A novel stimuluspossessing one such feature may evokea response" (Skinner, 1957, p. 91). Re-sponses to novel stimuli that resembleother stimuli previously present weredescribed by Skinner as extended tacts.There are many different ways inwhich novel stimuli may resemble pre-viously presented stimuli, and thus itwas necessary for Skinner to definevarious forms of extended tacts. Theseincluded, for example, generic exten-sion, metaphorical extension, nomina-tion, guessing, and abstraction. Forpresent purposes, we will deal onlywith abstraction (in principle, the fol-lowing analyses could be applied toany form of extended tact).

Abstraction. "Any property of astimulus present when a verbal re-sponse is reinforced acquires some de-gree of control over that response, andthis control continues to be exertedwhen the property appears in othercombinations" (Skinner, 1957, p. 107).The ability to respond to abstractedproperties of stimuli avoids the chaosthat could result from the fact that ev-ery stimulus shares properties with po-tentially many other stimuli. If abstrac-tion were not possible, then each stim-ulus could be expected to control agreat variety of responses. Abstractionis demonstrated, therefore, when a sin-gle property of a stimulus is tacted. Forexample, the property of redness maycontrol the emission of the vocal re-sponse "red," whether the response isoccasioned by a red apple, a red ball,or a red book.The verbal-nonverbal tact distinction

applies to abstraction in much the sameway that it applies to the simple tactrelation. According to RFT, for an ab-stracted tact relation to be verbal, theabstracted property must not only con-trol the tact response but it must alsoparticipate in a relational frame orframes with the response and perhapsother stimuli. For example, the non-verbal abstraction of the property ofredness may be the result of a history

of explicit reinforcement for respond-ing "red" in the presence of red ap-ples, red balls, red books, and so on;however, this nonverbal form of ab-straction would not readily lead to thebidirectional relational respondingcharacteristic of verbal behavior, in thiscase pointing to the red objects whenasked, "Point to red" (i.e., training seered objects -* say "red" does not au-tomatically produce hear "red" ->point to red objects). Verbal abstractionof the property of redness requires thatthe abstracted property control the tactresponse and that both the propertyand the response participate in a rela-tional frame or frames with other stim-uli (e.g., property and response coor-dinate in a bidirectional relation, andperhaps participate in a difference re-lation with other events, such as thespoken word "green" and the colorgreen).The foregoing interpretation of ab-

straction has implications for other in-stances of the tact relation. Consider,for example, that both nonverbal andverbal tacting behavior may be con-trolled by extremely subtle propertiesof stimuli, and that these propertiesmay include relations among stimuli.Such terms as above and below, nearand far, or larger and smaller tact theformal or physical properties of stimuliin relation to each other or to thespeaker; for example, when one objectis described as "bigger" or "smaller"than another. The reader should note,that although responding in accordancewith relational frames is not requiredfor this example of relational respond-ing, it seems likely that it would usu-ally be involved in some way (e.g., theword bigger may participate in a frameof coordination with larger). In anycase, verbal tacting behavior is likelyinvolved when these types of relationsare arbitrary as opposed to nonarbi-trary in nature. For example, the wordstiny and big may participate in an op-posite relation, although they are bothsmall words in relation to massive andsmall, which both contain more letters.In effect, the relations among these


Conditional Stimulus Stimulus Response

Color? "Red"

Hue? Apple "Scarlet"

Shade? "Crimson"

Smell? "Sweet"

Odour? Apple "Sugary"

OlfactoryImpression? "Fruity"Figure 1. Schematic representation of verbal conditional tacts showing some of the possible re-lational frames that may be involved in verbal conditional tacting behavior.

words cannot be based on their physi-cal properties alone. Instead, the rela-tions are arbitrary in that they havebeen determined by the practices of theverbal community, and thus relationalframes will almost certainly be in-volved when one or more of these re-lations is tacted.The same distinction between verbal

and nonverbal tacting may be madewhen the issue of conditional controlover the tacting response is considered.Clearly, any event or situation mayhave many properties that might betacted. Whether any properties are tact-ed and which properties are tacted willdepend on other variables that may actupon the speaker. This fact presents nodifficulties, because behavior may bedetermined in multiple ways. For ex-ample, one may tact the color of anapple when asked about its color andmay tact its smell when asked what itsmells like. If this conditional tactingresponse involves only direct-actingcontingencies (e.g., explicit reinforce-

ment, response induction, stimulusgeneralization, respondent condition-ing), then it may be defined as a non-verbal conditional tact. If, however, theconditional tact involves responding, atleast to some degree, in accordancewith relational frames, then it may bedefined as a verbal conditional tact. Forexample, the conditional stimuli andresponses may participate in frames ofcoordination with other stimuli and re-sponses, such that any of the lattercould be substituted for the formerwithout seriously affecting the functionof the conditional tact (e.g., a listenerwould respond in a similar fashion toany of the possible combinationsshown in Figure 1). This example alsoserves to illustrate the extensive net-work of frames of coordination thatmay occur in even the simplest of ver-bal interactions-consider also thatonly frames of coordination are shown;other possible frames could be includ-ed. Note also that the response a sub-ject emits when tacting can also partic-


ipate in relational frames with otherpossible responses. Clearly, the verbaltact constitutes a substantially morepowerful instance of "verbal" behav-ior than its nonverbal counterpart.

Tacting verbal behavior and the cre-ation of unseen realities. One impor-tant restriction in defining the tact re-lation is that it must occur in the pres-ence of the event tacted. This restric-tion raises the issue of words thatsuperficially seem to be tacts but can-not occur in the presence of what theyname. For example, when does one ac-tually see governmental units likestates or nations, subject matters likeeconomics or politics, processes likecreation or evolution, and so on? Suchentities must enter into verbal behaviorin other ways; they do not exist in aform that can be tacted in the normalsense. One solution to this problem canbe found, however, by adopting thenonverbal-verbal distinction. From theRFT perspective, the verbal tact in-volves the construction of contextuallycontrolled relational networks of ob-jects and events in the world. Oncethese have been established, it be-comes possible to tact these networksthemselves. Clearly, these networks ofrelations can never be seen, as physicalobjects or events, because they involvebehavioral relations extended acrosstime and space. For example, mostpeople have learned the concept of"politics" through the gradual con-struction of a network of nonarbitraryand arbitrary relations among the rel-evant people, places, and events thatthe verbal community defines as partof the political structure or process.When these stimulus relations havebeen established, a young girl (fromIreland) may tact this network whenshe says, "Politics is politicians debat-ing in the Dail" (the parliament build-ing in Dublin). In effect, the girl tactsher own verbal behavior (i.e., what shehas learned to say about a certain clus-ter of people, objects, and events in theworld controls the emission of theword "politics"). It should be notedthat because tacting a relational net-

work involves tacting one's own verbalbehavior rather than "things" in theenvironment, this phenomenon may bebetter considered an example of de-rived or verbal autoclitic behavior. Weshall consider this type of behavior lat-er.

Verbal behavior under the control ofprivate stimuli. In tacting an externalstimulus, both speaker and listenerhave access to the object being tacted,but this is not the case with privatestimuli. For example, in tacting a painonly the speaker may have contactwith the actual object or event (i.e., thepain) being tacted. Skinner suggestedfour ways in which the verbal com-munity could establish tacts without di-rect access to the private stimulus: (a)The reinforcement could be based on"a common public accompaniment ofthe private stimulus" (1957, p. 131),for example, a child may be taught tosay "that hurts" after a fall that resultsin a scraped knee or some other obvi-ous physical symptom; (b) the verbalcommunity may reinforce a vocal re-sponse describing an internal state thatis consistent with some overt physicalbehavior of the speaker; for example,tacting a toothache may be reinforcedin the presence of behavior such asrubbing the jaw or touching the gums;(c) a tact may be established to a publicstimulus, and the response may thentransfer to a private event by virtue ofcommon properties between the publicand private event; for example, tactinga sharp pain may emerge from the met-aphorical extension of tacting sharpobjects; and (d) a tact may be estab-lished to some public behavior of thespeaker, and if the overt behavior isthen reduced in magnitude to the pointof being imperceptible to others, theprivate stimulus that accompanied thepublic behavior may continue to betacted by the speaker; for example, ayoung girl may be taught, in certaincontexts, to tact her own crying as"sorrow," but as she matures some ofthe private events that accompaniedher crying may persist in the absenceof overt crying, and thus, as an adult,


she may tact "sorrow" without actu-ally crying.The foregoing interpretation of con-

trol by private stimuli is readily ac-ceptable from the RFT perspective.However, RFT distinguishes betweenresponding to private events based ona history of explicit reinforcement (oron respondent conditioning or on theformal properties of the stimuli alone)and such responding based on the der-ivation of stimulus relations. Accord-ing to RFT, for example, when a ver-bally able child learns to tact a privatestimulus such as a toothache, it is like-ly that the word toothache and the ac-tual internal event will come to partic-ipate in a frame of coordination. Thiswill also occur when tacting other in-ternal states. For instance, most indi-viduals in the English-speaking com-munity will, in some contexts, respondin accordance with coordination rela-tions among the words sadness, mis-ery, and despair, and overt behaviorsuch as crying or frowning. The im-portant point here is that by distin-guishing between verbal and nonverbaltacting of private stimuli, one may bebetter positioned to predict certain be-havioral outcomes that could not bereadily predicted in terms of direct-act-ing contingencies. Imagine, for exam-ple, a young girl who learns to tact cry-ing with the response "sorrow," andthen learns that misery is another wordfor sorrow, and despair is anotherword for misery (i.e., overt crying andthe words sorrow, misery, and despairparticipate in a frame of coordination).As outlined at the end of the previousparagraph (Point d), Skinner suggesteda way in which the girl may come totact private events as "sorrow" in theabsence of actual crying. However,based on the relational frame of coor-dination, the girl may also come to tactthese private events as "misery" and"despair," without being explicitlytaught to do so. Furthermore, given thebidirectional nature of relationalframes, in certain contexts it may bepossible to use the history of verbaltacting to generate some of the private

events of sorrow for the girl as a lis-tener. Imagine, for example, that aspeaker asks the girl (at some latertime when she is relatively happy) toremember the last time that she expe-rienced real despair. According to RFT,at least some of the private events thatwere previously discriminative for thetact "sorrow" may now be producedby the word despair (by virtue of thetransformation of function in accor-dance with the frame of coordination)(see Dymond & Barnes, 1994, 1995,1996, and Roche & Barnes, 1997, forrelevant empirical evidence). In effect,these types of unpredicted behavioraleffects are readily understood if onedistinguishes between the nonverbaland verbal tacting of private events.

AutocliticsBehavior that "is evoked by or acts

upon other behavior of the speaker"(Skinner, 1957, p. 313) was describedby Skinner as autoclitic behavior. Ar-guably, it is this type of behavior thatgives human language its richness andcomplexity. Various types of autocli-tics are available to the mature speaker.Descriptive autoclitics are used when aspeaker produces verbal behavior de-scriptive of his or her own behavior;for example, "I see," "I recall," "I tellyou." These autoclitics may also referto the strength of a response; for ex-ample, "I guess," "I estimate," "Isuggest." Qualifying autoclitics quali-fy the intensity or direction of a listen-er's behavior. One of these is negation,the most common examples of whichare the verbal responses "no," "not,"and "never." For example, the state-ment, "The mail has not been deliv-ered yet," cannot be a tact because theevent (the mail) is not present andtherefore cannot be tacted (i.e., onecannot tact the absence of an event). Inthis case, the word not functions as aqualifying autoclitic for the remainingpart of the sentence that is concernedwith the arrival of mail. The use ofsuch an autoclitic will normally affectthe behavior of a listener in such a way


that it reinforces behavior consistentwith the autoclitic (i.e., the listener willnot check to see if the mail has beendelivered). Another qualifying autocli-tic is assertion. Just as the vocal re-sponse "not" may affect the behaviorof a listener in one direction, a re-sponse such as "definitely" may affectthe behavior of a listener in the oppo-site direction, such that it may encour-age him or her to continue the behaviorin question. For example, "This is def-initely the right road to Maynooth"may encourage a listener to take thespecified road more readily than thestatement, "This is the right road toMaynooth." The effects of autocliticsmay be quantitative as well as quali-tative. For example, vocal responsessuch as "all," "some," or "many"modify the reaction of the listener tothe responses that they accompany.The statement, "All black spiders inNevada are poisonous," for instance,will perhaps cause greater avoidance ofsuch spiders than if the statement be-gan with "Some." Similarly, the arti-cles a and the indicate the relation be-tween a response and its controllingstimulus. For example, the statement,"The dangerous spider has beencaught," will have a different effect ona listener than if the statement startedwith "A" (i.e., The implies only onespider, whereas A implies other spidersyet to be caught).From the RFT perspective, many au-

toclitic responses will be explicitlymodeled and reinforced during the ini-tial stages of complex language learn-ing and are therefore considered ex-amples of nonverbal autoclitics (in ac-cordance with the current thesis, wealso include in this category autocliticsthat are established through processessuch as respondent conditioning, stim-ulus generalization, and response in-duction). Relational frame theory alsopredicts, however, that the establish-ment of relational frames will also gen-erate nonexplicitly taught autoclitics,which are therefore defined as verbal.For example, imagine a young girl whohas learned to say "It is going to rain"

when dark clouds are present. Subse-quently, a caregiver might explicitlyteach a descriptive autoclitic responseso that the girl will respond to the factthat dark clouds do not always lead torain. That is, the girl may be told tosay, "I think it is going to rain," sothat the listener does not take it as factthat rain is about to occur. At a latertime, the girl also learns that in somecontexts think is equivalent to feel andfeel is equivalent to predict, and thusthese three words may participate in aframe of coordination. At this point,therefore, the girl might now say "Ipredict that it is going to rain" whendark clouds are in the sky, althoughthis statement has never been explicitlymodeled or reinforced in the past. Inthis case, we can rightfully call this averbal autoclitic because it involves re-sponding in accordance with relationalframes on behalf of the speaker. Simi-lar interpretations could also be madefor the full range of autoclitics outlinedby Skinner (1957). In the next sectionwe will examine one such interpreta-tion (of the relational autoclitic) be-cause it will help us to address one ofthe most common criticisms of the be-havioral approach to the study of lan-guage.Grammar and syntax as autoclitic

processes. In Skinner (1957), the prop-erties of grammar and syntax are ex-plained, in part, by appealing to whathe calls relational autoclitics. Wordssuch as of, before, or below are rela-tional autoclitics in that they can spec-ify events only by their relation to oth-er responses. For example, the wordbelow is meaningless without referenceto that which is above. The manipula-tion of words or phrases by, for ex-ample, grouping or ordering those re-sponses in specific sequences is alsorelational autoclitic behavior. As Skin-ner described it, "the speaker not onlyemits verbal responses appropriate to asituation or to his own condition, heclarifies, arranges, and manipulates thisbehavior" (Skinner, 1957, p. 344).Therefore, an arrangement of wordssuch as "the man hit the dog" is a


combination of tacts of objects (manand dog), an event (hit), and variousrelational and temporal properties ofthese objects and events. In Skinner'sview, the syntactical structure of sen-tences may be determined, in part, byexplicitly taught discriminations of anumber of objects and events in theworld and the relations between thoseobjects and events. Furthermore, Skin-ner accounted for the generative natureof grammar by appealing to novel en-vironments organizing previously es-tablished discriminations into novel se-quences. Imagine, for example, that aprior history of explicit reinforcementgenerated the two statements, "the boyfell" and "the dog ran." Imagine nowthat the speaker sees a dog fall for thefirst time. In this case, the individualdiscriminations contained within thetwo explicitly reinforced statementsmay combine in this novel environ-ment to produce the novel utterance"the dog fell."Although the foregoing example il-

lustrates one way in which novel utter-ances may be produced by a speaker,according to RFI this is an example ofa nonverbal relational autoclitic be-cause it does not necessarily involverelational frames. From the RFT per-spective, verbal relational autocliticsinvolve derived stimulus relations. Al-though groupings of words almost cer-tainly occur, in part, through the directreinforcement of a number of discrim-inations that are then organized in nov-el ways by novel environments, it isstill necessary to explain how an adultspeaker generalizes from these envi-ronmentally organized sentences tonovel sentences for which there is noenvironment that could possibly orga-nize the sentence. For example, thenonsense sentence "Green ideas sleepfuriously" is correct grammatically,but there is clearly no physical envi-ronment that could possibly organizethe sentence in this way (see Chomsky,1959). The question arises, therefore,how can behavior analysis account forthis type of generativity in human lan-guage? In other words, what are the

functional classes through which wordorder sequences that are organized byactual environmental events transfer tonew instances in which an environ-mental event is not directly responsiblefor the organization of that sequence?Relational frame theory suggests atleast one way in which this might oc-cur.From the RFT perspective, large

numbers of words may participate invarious relational frames. For example,in some contexts there may be onegroup of words that participate in aframe of coordination (e.g., "thingwords" or nouns) and another groupthat participates in another frame of co-ordination (e.g., "describing words" oradjectives). These two frames may alsoparticipate in a relational network inthat all nouns are responded to as dif-ferent than adjectives (i.e., two framesof coordination related to each othervia a frame of difference). During achild's early verbal interactions, it islikely that he or she will be explicitlytaught (e.g., via modeling and rein-forcement) to utter some of the wordsfrom these two classes in a particularorder (adjective-noun). After this ex-plicit training, the word ordering maythen transfer to other members of thetwo frames of coordination (adjectiveand noun) without further training. Theproduction of a word sequence, non-sense or otherwise, that relies to someextent on the transfer of word orderingfunctions in accordance with relationalframes is by definition a verbal rela-tional autoclitic response. It is impor-tant to note, however, that word se-quences often change in different lin-guistic contexts. For example, if thechild is taught to use is in an adjective-noun sequence, then the word se-quence is reversed (i.e., "the big cat"becomes "the cat is big"). In effect,the word sequence is conditional uponthe presence or absence of the word is.Presumably this conditional orderingeffect will also transfer through rela-tional frames (see Wulfert & Hayes,1988).These types of transfer through


TABLE 1

Generating a nonsense sentence in-volves transferring words (indicatedby an asterisk) from normal sentenc-es via four (grammatical) frames ofcoordination into a single sentenceAdjective Noun Verb Adverb

Normal sentences (generated by actual environ-mental events)Green* Grass Waves GentlyTired People Sleep* HappilyGood Ideas* Work WellAggressive People Fight Furiously*

A nonsense sentence (generated by transferringindividual words from normal sentences)Green Ideas Sleep Furiously

grammatical relational frames mayhelp to explain the generative nature ofgrammar and syntax (cf. Ellenwood &Chase, 1997; Goldstein & Mousetis,1989; Streifel, Wetherby, & Karlan,1976; Wetherby, 1978), and in partic-ular they suggest an interesting inter-pretation of the formation of nonsensesentences. In effect, nonsense sentenc-es may be possible because the wordsthat constitute them are members of re-lational frames of coordination thatcontain large numbers of other words.Thus a nonsense sentence may be gen-erated when individual words withinthe grammatical frames of coordinationfrom various normal sentences aretransferred into a single sentence untilthat sentence is discriminated as de-scribing an "unreal" environmentalevent. The result is a novel nonsensesentence, such as "Green ideas sleepfuriously" (see Table 1). This illus-trates how the relational autocliticcombined with the formation of rela-tional frames (i.e., a verbal relationalautoclitic) may give rise to novel non-sense sentences.

CONCLUSIONIn this article, we have attempted to

show how integrating RFT with Skin-ner's Verbal Behavior (1957) may gen-erate a modern behavior-analytic treat-

ment of the behavior of the speakerthat is sensitive to recent research onderived stimulus relations. We shouldadd that some aspects of verbal behav-ior are too complex to address giventhe constraints of a single article likethis, so we have left these to be de-scribed elsewhere. For example, Stew-art and Barnes-Holmes (in press) dis-cuss the processes involved in statingand understanding metaphor (see alsoBarnes, Hegarty, & Smeets, 1997;Chase & Parrott, 1986; Hayes &Grundt, 1997). We do not pretend,therefore, that the current work is de-finitive or final, but see it as simplyhelping to generate a modern behav-ioral research agenda in human lan-guage and cognition (Hayes & Barnes-Holmes, in press). In short, by apply-ing RFT to Skinner's verbal operants,we aim to develop a modern, coherent,naturalistic, and purely functional-an-alytic understanding of human lan-guage that will provide a powerfulchallenge to the many nonbehavioralapproaches that abound in the psycho-logical literature.

REFERENCES

Barnes, D. (1994). Stimulus equivalence andrelational frame theory. The PsychologicalRecord, 44, 91-124.

Barnes, D. (1996). Naming as a technical term:Sacrificing behavior analysis at the altar ofpopularity. Journal of the Experimental Anal-ysis of Behavior, 65, 264-267.

Barnes, D., & Hampson, P. J. (1993). Stimulusequivalence and connectionism: Implicationsfor behavior analysis and cognitive science.The Psychological Record, 43, 617-638.

Barnes, D., & Hampson, P J. (1997). Connec-tionist models of arbitrarily applicable rela-tional responding: A possible role for the hip-pocampal system. In J. W. Donahoe & V. P.Dorsel (Eds.), Neural network interpretationsof cognition: Biobehavioral foundations (pp.496-521). Amsterdam: Elsevier.

Barnes, D., Healy, O., & Hayes, S. C. (in press).Relational frame theory and the relationalevaluation procedure: Approaching humanlanguage as derived relational responding. InJ. C. Leslie & D. E. Blackman (Eds.), Exper-imental and applied analyses of human be-havior. Reno, NV: Context Press.

Barnes, D., Hegarty, N., & Smeets, P. M.(1997). Relating equivalence relations toequivalence relations: A relational framing


model of complex human functioning. TheAnalysis of Verbal Behavior, 14, 57-83.

Barnes, D., & Holmes, Y. (1991). Radical be-haviorism, stimulus equivalence and humancognition. The Psychological Record, 41, 19-31.

Barnes, D., & Roche, B. (1996). Stimulusequivalence and relational frame theory arefundamentally different: A reply to Saunders'commentary. The Psychological Record, 46,489-507.

Barnes, D., & Roche, B. (1997). Relationalframe theory and the experimental analysis ofhuman sexuality. Applied and Preventive Psv-chology, 6, 117-135.

Barnes-Holmes, D., & Barnes-Holmes, Y. (inpress). Explaining complex behavior: Twoperspectives on the concept of generalized op-erant classes. The Psychological Record.

Barnes-Holmes, Y., Barnes-Holmes, D., &Roche, B. (in press). Establishing relationalframing in children: The role of multiple-ex-emplar training. In S. C. Hayes & D. Barnes-Holmes (Eds.), Relational frame theorv: Cre-ating an alternative behavioral agenda in lan-guage and cognition. Reno, NV: ContextPress.

Chase, P. N., & Danforth, J. S. (1991). The roleof rules in concept leaming. In L. J. Hayes &P N. Chase (Eds.), Dialogues on verbal be-havior (pp. 205-225). Reno, NV: ContextPress.

Chase, P N., & Parrott, L. J. (1986). Psycho-logical aspects of language: The West Virgini-ia lectures. Springfield, IL: Charles C. Thom-as.

Chomsky, N. (1959). A review of B. F Skin-ner's Verbal Behavior. Language, 35, 26-58.

Dymond, S., & Bames, D. (1994). A transferof self-discrimination response functionsthrough equivalence relations. Journal of theExperimental Analysis of Behavior, 62. 251-267.

Dymond, S., & Barnes, D. (1995). A transfor-mation of self-discrimination response func-tions through the arbitrarily applicable rela-tions of sameness, more than, and less than.Journal of the Experimental Analysis of Be-havior, 64, 163-184.

Dymond, S., & Barnes, D. (1996). A transfor-mation of self-discrimination response func-tions in accordance with the arbitrarily appli-cable relations of sameness and opposition.The Psychological Record, 46, 271-300.

Ellenwood, D., & Chase, P. N. (1997, May). Acomparison of two behavioral training com-ponents: Rate building and component com-posite analysis. Paper presented at the 23rdannual convention of the Association for Be-havior Analysis, Chicago.

Goldstein, H., & Mousetis, L. (1989). Gener-alized language learning by children with se-vere mental retardation: Effects of peers' ex-pressive modeling. Journal ofApplied Behav-ior Analvsis, 22, 245-259.

Hall, G., & Sundberg, M. L. (1987). Teaching

mands by manipulating conditioned establish-ing operations. The Analysis qf Verbal Behav-ior-, 5, 41-53.

Hayes, S. C. (1991). A relational control theoryof stimulus equivalence. In L. J. Hayes & PN. Chase (Eds.), Dialogues oni verbal hbehav-ior: Tle first initerniational inistitute onlverbalrelations (pp. 19-40). Reno, NV: ContextPress.

Hayes, S. C. (1994). Relational fi-ame theory:A functional approach to verbal events. In S.C. Hayes, L. J. Hayes, M. Sato, & K. Ono(Eds.), Behavior analysis (f' lanzguage aniidcognition (pp. 9-30). Reno, NV: ContextPress.

Hayes, S. C., & Barnes-Holmes, D. (in press).Relational frame theory: Creatinig ani alter-native behavioral agenida in laniiguaige andcogniition. Reno, NV: Context Press.

Hayes, S. C., Gifford, E. V., & Wilson, K.(1996). Stimulus classes and stimulus rela-tions: Arbitrarily applicable relational re-sponding as an operant. In T R. Zentall & PM. Smeets (Eds.), Stinmulus class fo1rnation inhumans and aniimals (pp. 279-299). Amster-dam: Elsevier Science.

Hayes, S. C., & Grundt, A. M. (1997). Meta-phor, meaning, and relational frame theory. InC. Mandell & A. McCabe (Eds.), The pirobleimof meaning: Behavioral anid cognitii'e per-spectives (pp. 117-146). Amsterdam: Elsevier-Science.

Hayes, S. C., & Hayes, L. J. (1989). The verbalaction of the listener as a basis for rule-gov-ernance. In S. C. Hayes (Ed.). Rule-governedbehavior: Cogniitioni, contingencies, timid in-structional conitrol (pp. 153-190). New York:Plenum.

Hayes, S. C., & Hayes, L. J. (1992). Verbalrelations and the evolution of behavior anal-ysis. Americani Psychologist. 47. 1383-1395.

Hayes, S. C., & Wilson, K. G. (1993). Someapplied implications of a contemporary be-havior-analytic view of verbal events. The Be-havior Analyst, 16, 283-301.

Leader, G., Barnes, D., & Smeets, P M. (1996).Establishing equivalence relations using a re-spondent-type training procedure. The Psv-chological Record, 46, 685-706.

Lipkens, R., Hayes, S. C., & Hayes, L. J.(1993). Longitudinal study of the develop-ment of derived stimulus relations in an in-fant. Journal of Experimental Clhild PsYchol-ogy, 56, 201-239.

Roche, B., & Barnes, D. (1997). A transfor-mation of respondently conditioned stimulusfunction in accordance with arbitrarily appli-cable relations. Journial of tlhe ExperimentalAnalysis of Behavior-, 67, 275-300.

Sidman, M. (1994). Equivalence relationis ainidbehavior: A research story. Boston: AuthorsCooperative.

Skinner, B. F (1957). Ver-bal behav,ior. Engle-wood Cliffs, NJ: Prentice Hall.

Skinner, B. F (1986). The evolution of verbal


behavior. Journal of the Experimental Analy-sis of Behavior, 45, 115-122.

Stewart, I., & Barnes-Holmes, D. (in press).Understanding metaphor: An empirical modelfrom relational frame theory. In S. C. Hayes& D. Barnes-Holmes (Eds.), Relationalframetheory: Creating an alternative behavioralagenda in language and cognition. Reno, NV:Context Press.

Streifel, S., Wetherby, B., & Karlan, G. (1976).Establishing generalized verb-noun instruc-tion-following skills in retarded children.

Journal of Experimental Child Psychology,22, 247-260.

Wetherby, B. (1978). Miniature languages andthe functional analysis of verbal behavior. InR. Schiefelbusch (Ed.), Bases of language in-tervention (pp. 397-448). Baltimore: Univer-sity Park Press.

Wulfert, E., & Hayes, S. C. (1988). Transfer ofa conditional ordering response through con-ditional equivalence classes. Journal of theExperimental Analysis of Behavior, 50, 125-144.

Documents

behavan00011-0071