Understanding Metaphor: a Relational Frame Perspective Ian Stewart and Dermot Barnes-Holmes National University of Ireland, Maynooth

Home , Relational frame theory

The Behavior Analyst 2001, 24, 191-199 No. 2 (Fall) Understanding Metaphor: A Relational Frame Perspective Ian Stewart and Dermot Barnes-Holmes National University of Ireland, Maynooth

The current article presents a basic functional-analytic interpretation of metaphor. This work involves an extension of Skinner's (1957) interpretation of metaphor using relational frame theory (RFT). A basic RFI' interpretation of a particular metaphor is outlined, according to which the metaphor acquires its psychological effects when formal stimulus dimensions are contacted via the derivation of arbitrary stimulus relations. This interpretation sees the metaphor as involving four elements: (a) establishing two separate equivalence relations, (b) deriving an equivalence relation between these relations, (c) discriminating a formal relation via this equivalence-equivalence relation, and (d) a transformation of functions on the basis of the formal relation discriminated in the third element. In the second half of the paper, a number of important issues with regard to the RFT interpretation of metaphor are addressed. Key words: metaphor, relational frame theory, equivalence-equivalence, transformation of functions, nonarbitrary relations

In psychology, especially cognitive psychology, of an object or event" (pp. 81-82). The characterizing the processes involved in the comprehension of metaphor is not only an in- extended tact is a feature of more com- teresting challenge in its own right, but the spec- plex verbal behavior that occurs when ification of those processes also constitutes a a response is evoked by a novel stim- good test of the power of theories of language ulus that resembles a stimulus previ- comprehension in general. (Ortony, 1993, p. 4) ously present when a response was re- Behavior analysis has had relatively inforced. Metaphorical verbal behavior little to say on the topic of metaphor.' is a subtype of this latter class of be- Indeed, for the main contribution made havior that takes place "because of the in this area we have to turn to the control exercised by properties of the largely interpretive work of Skinner stimulus which, though present at re- (1957). Skinner classified this phenom- inforcement, do not enter into the con- enon as a subtype of the "extended tingency respected by the verbal com- tact." The tact is a behavioral relation munity" (p. 92). The following is an that is defined as "a verbal operant in example of Skinner's interpretation of which a response of a given form is metaphor: evoked (or at least strengthened) by a particular object or event or property When for the first time a speaker calls someone a mouse, we account for the response by noting certain properties-smallness, timidity, silent Requests for reprints may be addressed to movement and so on-which are common to the Dermot Barnes-Holmes, Department of Psy- kind of situation in which the response is char- chology, National University of Ireland, May- acteristically reinforced and to the particular sit- nooth, Maynooth, County Kildare, Ireland (E- uation in which the response is now emitted. mail: [email protected]). Since these are not the properties used by zo- ' Traditionally, a distinction is made between ologists or by the lay community as the usual metaphor and simile based largely on structural basis for reinforcing a response we call the ex- differences (i.e., simile incorporates the phrases tension metaphorical. (p. 93) "like," "as," or "as if," whereas metaphor does not). Because the current interpretation and model of metaphor are functional analytic, these A somewhat similar analysis of meta- structural differences have been ignored for the phor, in which Skinner (1989) once time being. Subsequent work may well discover again provided a behavior-analytic in- that there is an important functional basis to the traditional distinction between metaphor and terpretation of familiar real-world simile, but consideration of this issue is beyond events, appears in Recent Issues in the the scope of the current article. Analysis of Behavior. 191 192 IAN STEWART & DERMOT BARNES-HOLMES

When we speak of weighing evidence we are tional responding. All animals that are using a metaphor. But a metaphor is a word that is "carried over" from one referent to another capable of complex forms of learning on the basis of a common property. The com- may be trained to make discriminations mon property in weighing is the conversion of on the basis of nonarbitrary or physical one kind of thing (potatoes or evidence) into an- relations between stimuli (e.g., louder other (a number on a scale or a verdict). Once than, larger than, differently colored we have seen this kind of thing done with potatoes it is easier to see it done with evidence. from). However, according to RFT, ... We could also say that weight becomes ab- language-able humans, having pro- stract when we move from potatoes to evidence. longed exposure to certain contingen- The word is indeed abstracted in the sense of its cies of reinforcement that operate with- being drawn away from its original referent, but it continues to refer to a common property, and, in the verbal community, also demon- as in the case of metaphor, in a possibly more strate responding on the basis of de- decisive way. The testimony in a trial is much rived or arbitrarily applicable relations. more complex than a sack of potatoes, and These relations are defined not by the "guilty" probably implies more than "ten physical properties of the relata per se pounds." But abstraction is not a matter of complexity. Quite the contrary. Weight is only one but by additional contextual cues. aspect of a potato, and guilt is only one aspect One important example of arbitrarily of a person. Weight is as abstract as guilt. It is applicable relational responding is the only under verbal contingencies of reinforce- phenomenon of stimulus equivalence. ment that we respond to single properties of things or persons. In doing so we abstract the For example, if a language-able human property from the thing or person. (p. 7) is taught to match an arbitrary Stimu- lus A to an arbitrary Stimulus B and to According to the Skinnerian account, match Stimulus B to an arbitrary Stim- therefore, metaphor may be conceptu- ulus C, then, without further training, alized as the abstraction, via one par- he or she might subsequently match A ticular subtype of verbal behavior (i.e., to B, B to C (thus showing symmetri- the extended tact), of a common phys- cal responding), C to A (thus showing ical property from two different types transitive responding), and A to C of environmental event. This behavior- (thus showing combined symmetrical al interpretation, however, does not de- and transitive responding). In other tail how a repertoire of metaphorical words, the person now treats the three verbal behavior (especially complex stimuli as mutually substitutable or human metaphor) develops from a pre- equivalent. sumably simpler repertoire of formal Relational frame theory takes the property abstraction. This is perhaps view that stimulus equivalence, and the key question left unanswered by what has been called a transfer of func- Skinner's analysis. However, Skinner's tion through equivalence relations, pro- work does provide the basis for a mod- vide an important basis for the behav- em behavioral treatment of metaphor ior analysis of symbolic relations and using the conceptual and empirical human language more generally. From framework of relational frame theory an RFT perspective, one of the impor- (RFT; Hayes, Barnes-Holmes, & tant properties of a symbol is that its Roche, 2001; see also Barnes-Holmes, psychological functions are based to Barnes-Holmes, & Cullinan, 2000, for some extent on the transfer of func- a systematic synthesis of the Skinner- tions through equivalence relations. ian and RFT interpretations of verbal Suppose, for example, that a young behavior). boy enjoys a glass of Cokes (Stimulus A) for the first time and is then told by RELATIONAL FRAME THEORY a school friend that Coke is like another drink called PepsiT3 (Stimulus B). If Relational frame theory is an ap- the boy is subsequently asked "Would proach to human language and cogni- you like a Pepsi?" he may well re- tion that treats these phenomena as ex- spond positively. This transfer of ap- amples of arbitrarily applicable rela- petitive function is based on the di- UNDERSTANDING METAPHOR 193 rectly established psychological func- ed below that will be overcome only tion of the word Coke, which was through further empirical research. paired with actual Coke, and the derived symmetry relation between the RELATING RELATIONS: words Pepsi and Coke (i.e., Coke is A KEY PROPERTY OF like Pepsi). In effect, the boy does not METAPHORICAL LANGUAGE need to experience Pepsi directly in order for him to "think" that it will be a The first published RFT study that pleasant drink. This basic concept of a was directly relevant to the analysis of transfer of function in accordance with metaphor focused on the closely relat- stimulus relations forms the core of the ed area of analogy (Barnes, Hegarty, & RIFT approach to language and cogni- Smeets, 1997; we will consider the re- tion (Hayes, Barnes-Holmes, & Roche, lation between analogy and metaphor 2001). at a later point in the article). Specifi- In this article we will describe how cally, Barnes et al. used what they RFT has used the concepts of equiva- called equivalence-equivalence re- lence and transfer of function to ana- sponding to construct a relational lyze metaphor both conceptually and frame interpretation of analogical rea- empirically. This RFT interpretation soning. In the authors' own words, builds on the earlier work of Skinner Consider ... the following question based on the (1957) by showing how a repertoire of classic proportion scheme (A: B :: C: ?); "ap- metaphorical verbal behavior might ple is to orange as dog is to (i) sheep, or (ii) develop from a simpler repertoire of book?" If "apple" and "orange" participate in an equivalence relation in the context "fruit," formal property abstraction. Before and "dog" and "sheep" participate in an equiv- continuing however, we must first in- alence relation in the context "animals," then clude a caveat. we would expect a person to pick "sheep" as The model reported herein adds to the correct answer. In effect, the response would Skinner's (1957) analysis of metaphor- be in accordance with the derived equivalence relation between two already established sepa- ical extension by showing that it can rate equivalence relations. ... We take the view be based on arbitrarily applicable re- that equivalence-equivalence responding is an lations among stimuli. Of course, in example of a relational network as defined by 1957, Skinner did not have access to relational frame theory (e.g., Barnes, 1994; Barnes & Holmes, 1991; Hayes, 1991, 1994). the concepts of equivalence and trans- (1997, p. 3) fer of function that are necessary to describe the conditions from which arbi- The first experiment reported by trary stimulus relations emerge. Be- Barnes et al. (1997) examined the re- havior-analytic science can now de- lations between two separate equiva- scribe some of these conditions. lence relations and between two sepa- However, to claim that we have isolat- rate nonequivalence relations. Subjects ed them sufficiently goes beyond cur- were first trained and tested for the for- rently available data. Under a variety mation of four three-member equiva- of conditions, with a variety of tests, lence relations (i.e., train Al -e B1, Al investigators have revealed the limita- -+ Cl, A2 B2, A2 e> C2, A3 -+ B3, tions of our understanding of equiva- A3 - C3, A4 - B4, A4 - C4, and lence and derived relational responding test Bi <-+ Cl, B2 e-> C2, B3 <-> C3, more generally, and thus at this time, B4 e- C4). After successfully passing we are left with empirical questions yet the equivalence test, subjects were test- to be answered with regard to these ed to determine whether they would re- phenomena (e.g., Fields, Hobbie- late pairs of stimuli to other pairs of Reeve, Adams, & Reeve, 1999; Pil- stimuli based on their participation in grim & Galizio, 1995; Spencer & equivalence relations. Subjects were Chase, 1996). It should be recognized, presented with samples that contained therefore, that there are limits to the two stimuli that were from one derived RFT interpretation of metaphor report- equivalence relation (e.g., BICI), and 194 IAN STEWART & DERMOT BARNES-HOLMES

were given the opportunity to choose trying really hard to escape feelings of comparisons that contained two stimuli anxiety can often increase those very that were from a second separate de- feelings). Consequently, the client's be- rived equivalence relation (e.g., havior in the face of anxiety may B3C3). The result of this first experi- change, such that he or she no longer ment was that a range of subjects, in- attempts strongly to resist his or her cluding a 12-year-old boy, successfully anxiety, but instead accepts the anxious related equivalence relations to other feelings as they arise, and thus pre- separate equivalence relations, and vents the downward cycle into full- nonequivalence relations (e.g., B1B2) blown panic. to other, separate nonequivalence rela- This quicksand-anxiety metaphor tions (e.g., C3C4), in the absence of may be interpreted as a relational net- explicit reinforcement. Experiment 2 work that is functionally similar to the employed the same procedures as were analogy described earlier. In this case, used in Experiment 1, except that sub- however, "anxiety" (A) is to "psycho- jects were exposed to the equivalence- logical struggle" (B) as "quicksand" equivalence test before being exposed (C) is to "physical struggle" (D) (see immediately, and without further train- Figure 1, Elements 1 and 2). For cur- ing, to the standard equivalence test. rent purposes, we will describe the re- Again, the result of this experiment lation between anxiety and psycholog- was that all subjects, this time includ- ical struggle, and the relation between ing a 9-year-old boy, successfully re- quicksand and physical struggle, as lated equivalence relations to other two separate equivalence relations. separate equivalence relations and non- From the RFT perspective, the rela- equivalence relations to other separate tional network in this example of met- nonequivalence relations. In short, aphor may help the listener to discrim- Barnes et al. provided an empirical inate aformal relation between two ap- model of analogical reasoning, based parently very different events. This on the RFT concept of equivalence- discrimination of formal similarity be- equivalence responding. This basic tween the two events may make a concept of relating one derived relation transfer of function from one to the to another derived relation also lies at other more likely. In this particular the heart of the RFT analysis of met- metaphor, deriving a relation between aphor, to which we now turn. anxiety and psychological struggle and quicksand and physical struggle could METAPHOR FROM help the listener to discriminate that THE RFT PERSPECTIVE struggling in either case leads to struc- turally or formally similar physiologi- Consider the metaphor, "Struggling cal and psychological effects (i.e., mas- with anxiety is like struggling in quick- sively increased autonomic arousal and sand," which might be used, for ex- a sense of fear and panic; see Figure 1, ample, in the psychotherapeutic treat- Element 3). Consequently, some of the ment of clinical anxiety. Contacting functions of "quicksand" might now this particular metaphor might cause a be more likely to transfer to "anxiety." client to change his or her response to For example, a clinically anxious per- anxiety attacks. The client probably al- son might derive certain important ready knows, via the verbal commu- cause-effect relations (these and other nity, that struggling in quicksand only nonequivalence relations will be con- makes drowning all the more likely. sidered below) including the following: When the therapist suggests that strug- "Struggling with anxiety will only gling with anxiety is similar to strug- make my situation worse" and "by gling in quicksand, then the client may ceasing to struggle I can begin to over- see that struggling with anxiety serves come my anxiety" (see Figure 1, Ele- only to make the anxiety worse (i.e., ment 4). Thus, the original problem UNDERSTANDING METAPHOR 195

Element (i): In the dients original relatonal network, "quicksand" has certain important response functions such as "Dont struggle", for example, which "anxiety" does not.

ANXIETY QUICKSAND (fn.: Don't struggle) Equiv Equiv PSYCHOLOGICAL STRUGGLE PHYSICAL STRUGGLE

Element (ii): The therapist changes the dient's relational network by equating the relation "anxiety [AJ / psychological struggle [Br with the relation "quicksand [C] / physical struggle [DI " (i.e., "A is to B as C is to D"). [A] ANXIETY QUICKSAND (fn.) [C] Equiv t- EQUIV-> Equiv [B] PSYCHOLOGICAL STRUGGLE PHYSICAL STRUGGLE [D]

Element (iii): The newly changed relabonal network allows the client to discriminate a formal or non arbitrary relation of sameness between the relations "anxiety / struggle" and "quicksand / struggle" (i.e., both relations lead to increasing autonomic arousal accompanied by spiraling feelings of fear and panic).

ANXIETY QUICKSAND (fn.) Equiv <- EQUIV-> Equiv PSYCHOLOGICAL STRUGGLE PHYSICAL STRUGGLE

(Increasing autonomic (Increasing autonomic arousal / Fear) arousal / Fear) I (FORMAL/NON-ARBITRARY RELATION)

Element (iv): The discrimination of a formal or non arbitrary relation between the relations "anxiety / struggle" and "quicksand / struggle" may lead to a transfer of the funcfions of "anxiety" such that "anxiety" now possesses some of the response functions of quicksand including, for example, "Don't struggle";

ANXIETY (fn.: Don't struggle) QUICKSAND (fn.: Don't struggle) Equiv <- EQUIV-> Equiv PSYCHOLOGICAL STRUGGLE PHYSICAL STRUGGLE I (Increasing autonomic === (Increasing autonomic arousal / Fear) arousal / Fear) I (FORMAL/NON-ARBITRARY RELATION) Figure 1. The four elements that, from the RFT perspective, characterize the process of metaphor. may come to be "recast" in view of spective, the often-experienced rich- the relational network illustrated in ness and complexity of metaphorical Figure 1 (Elements 2 through 4). More language, as well as the emotional "in- generally, this type of pattern of trans- sight" that metaphor can often confer fer of functions throughout a relational upon the listener. Parenthetically, we network generates, from the RFT per- should emphasize that the four ele- 196 IAN STEWART & DERMOT BARNES-HOLMES ments involved in metaphorical re- tween two causal relations (i.e., a sponding are not stages, and that the "causal-causal" relation). Specifically, exact sequence of these and perhaps struggling causes more rapid sinking other elements may vary from instance into either quicksand or anxiety. Fur- to instance. We will return to this issue ther research in our laboratories is cur- later in the article. rently utilizing novel methodologies (e.g., Hayes & Barnes, 1997) to model THE RFT INTERPRETATION metaphorical language in which non- OF METAPHOR: equivalence relations participate in the ADDITIONAL ISSUES underlying relational networks. The foregoing provides an RFI interpretation of how metaphor may be The Distinction Between Metaphor used to change the behavior of a lan- and Analogy guage-able human in important ways. The current interpretation of meta- In fact, we have developed a four- phor proposes that the key properties element empirical model based on of metaphorical language include, for this RFT conceptualization (Stewart, example, the relating of relations and Barnes-Holmes, & Roche, 2000). Nev- the discrimination of nonarbitrary sim- ertheless, our conceptual and empirical ilarity via arbitrarily applicable rela- work constitutes only a beginning, and tional responding. However, these there are many issues that remain to be properties could also be seen as char- addressed. acterizing analogy. Consider, for example, the classic Rutherfordian anal- Metaphor and Relations Other ogy "An atom is like the solar sys- Than Equivalence tem." This analogy involves the relat- According to RFT, equivalence is ing of relations. That is, in both cases, just one example of derived or arbi- there is a relation between a central trarily applicable relational responding. hub and orbiting elements, and because Relational frames of opposite, differ- these are similar relations they may be ence, before-after, and so forth have treated as equivalent. Nonarbitrary re- been the subject of both conceptual lations are also involved, because the and empirical investigation, thereby relation between the central hub and extending the range of behavioral phe- orbiting elements in both systems is an nomena that might emerge from obviously physical or nonarbitrary re- trained relational responding (e.g., lation. Given that analogy involves re- Barnes & Hampson, 1993a, 1993b; lating relations and nonarbitrary rela- Barnes & Keenan, 1993; Dymond & tions, the question remains: How does Barnes, 1994; Roche & Barnes, 1997; metaphor differ from analogy, and how Steele & Hayes, 1991). might we interpret this difference in In the current article, we have pre- functional terms? sented an RFT interpretation of meta- One test that has been used to dis- phor based on equivalence-equiva- tinguish between analogy and meta- lence responding, that is, one in which phor is the directionality test (see, e.g., the listener responds in accordance Chase, 1986), according to which anal- with a relation of equivalence between ogies are bidirectional, whereas meta- two equivalence relations. However, phors are unidirectional. In any exam- according to RFT, metaphors also may ple of either analogy or metaphor, two involve relations of equivalence be- events, A and B, are related in a phrase tween other types of relation. For ex- of the basic form "A is (like) B." In ample, the metaphor "Struggling with the case of analogy, the position of the anxiety is like struggling in quick- A and B terms may be swapped and sand" might be more accurately de- the result is another meaningful anal- scribed as an equivalence relation be- ogy. For example, if the A and B terms UNDERSTANDING METAPHOR 197 in the analogy "An atom (A) is like the vation of a hierarchical frame between solar system (B)" are swapped, the re- the two elements involved (see Hayes, sult, "The solar system (B) is like an Fox, et al., 2001). For example, En- atom (A)" is still a valid and under- glish speakers respond to a sentence of standable analogy. However, in the the form "A is B" (e.g., apples are case of metaphor, if the A and B terms fruit) as signifying that A are members are swapped, the phrase loses its met- of the class of B, and thus that A (e.g., aphorical quality. Consider the meta- apples) have at least some of the typi- phor "Cats (A) are dictators (B)," for cal properties of B (e.g., fruit). That is, example. Metaphors such as this work "apple" is contained in the class because the A and B terms have a "fruit," but "fruit" is not contained in property in common that is obvious the class "apple." This hierarchical and stereotypical in the case of B but frame is an important property of met- is not obvious in the case of A. In this aphor. In technical terms, "Cats are example, dictators are clearly demand- dictators" may cause a transfer of the ing and coercive, whereas cats may psychological functions typically as- also be seen in this light but far less sociated with dictators (e.g., "demand- obviously. The inclusion of these terms ing-ness") to cats, and if some of the in the metaphor in the order cats (A) nonobvious functions of cats are am- are dictators (B) serves to amplify the plified by this transfer, the listener will property of "demanding-ness" in the respond to it as a metaphor. However, A term, "cats." However, this ampli- if the A and B terms are reversed, the fication effect occurs only if this spe- obvious qualities of cats transfer to dic- cific order is maintained. If the terms tators and amplification is less likely are swapped, so as to yield the phrase, (e.g., dictators have none of the quali- "Dictators are cats," then there is no ties of small, furry pets). In effect, the appropriate amplification effect, be- unidirectional, hierarchical relation because now there is no longer a com- tween the terms involved is an impor- mon property that is obvious in the tant feature of the phenomenon of met- second term and not obvious in the aphor, because it differentiates it from first. analogy. The earlier example, "Struggling with anxiety is like struggling in quick- Elements, Not Stages sand," also qualifies as a metaphor based on this definition, because the As pointed out earlier, the four-ele- property in common between the two ment interpretation of metaphor is not events (i.e., self-defeating struggle) is sequential-the exact sequence of ele- obvious in the second term but not in ments may vary from instance to in- the first. Hence, the juxtaposition stance. The formal relations specified works in the direction just shown but in Element 3, for example, could serve not so well in the other direction. as a contextual cue for equivalence re- A technical RFT account of the dif- sponding or could itself be discrimi- ferent directionalities characterizing nated only after the equivalence rela- analogy and metaphor is as follows. In tion is derived. Consider again the met- the case of analogies, such as "An aphor, "Cats are dictators." As was atom (A) is like the solar system (B)," pointed out, the shared formal features the nonarbitrary properties that A and of the two terms are their demanding B have in common are equally char- qualities. This common quality may it- acteristic of both A and B. Thus, there self function as a context for applying will be a relation of equivalence be- a frame of equivalence between cats tween the A and B terms, and reversal and dictators (i.e., Element 3 may oc- of the order of the terms still yields an cur before, or at least feed back to sup- analogy. In the case of metaphor, there port, Element 2). Furthermore, certain are often contextual cues for the deri- elements (e.g., Element 4) may be 198 IAN STEWART & DERMOT BARNES-HOLMES missing entirely in some metaphors and exciting and feels like 'skimming' (i.e., the metaphor is understood, but or 'riding' the sea") and using the In- there is no change in behavior). Nev- ternet (e.g., "accessing information on ertheless, the four elements presented the Internet is fast, and it feels as if the in this example are typical of many in- user is 'skimming' or 'riding' over a stances of metaphor. sea of information"). In other words, these descriptions facilitated the dis- Understanding Metaphor Versus crimination of the formal relations that Creating Metaphor led to the production of the metaphor for the first time. From this perspec- The current work has focused on the tive, therefore, the creation of meta- understanding of metaphorical lan- phor involves the discrimination of guage as opposed to the creation of nonarbitrary relations based on a his- novel metaphors. Future research, tory of arbitrary relational responding, however, will need to focus on the un- and thus the processes involved in un- doubtedly important behavioral pro- derstanding and creating metaphor cesses involved in the creation of novel may be functionally similar. Although metaphors. Within the context of the this issue will certainly require further current interpretation of understanding analysis, the current RFT model of un- metaphor, the creator of a metaphor derstanding metaphor should provide a might first perceive some formal or useful starting point for developing a nonarbitrary similarity between two more complete functional-analytic different environmental events, and model of metaphorical language. then subsequently behave verbally in accordance with a new relational net- REFERENCES work into which this novel formal similarity has been incorporated. The cre- Barnes, D. (1994). Stimulus equivalence and ator of the metaphor "surfing the In- relational frame theory. The Psychological ternet," for example, might have been Record, 44, 91-124. Barnes, D., & Hampson, P. J. (1993a). Learning using the Internet at some point and to learn: The contribution of behavior analysis found that this context suddenly pro- to connectionist models of inferential skill in duced some of the perceptual functions humans. In G. Orchard (Ed.), Neural comput- of actual surfing. Consequently, he or ing research and applications (Vol. 1, pp. 129-138). London: IOP. she may have immediately related the Barnes, D., & Hampson, P. J. (1993b). Stimulus two terms, "surfing waves" and "surf- equivalence and connectionism: Implications ing the Internet," thus creating a rela- for behavior analysis and cognitive science. tional network on the basis of nonar- The Psychological Record, 43, 617-638. bitrary relations. Barnes, D., Hegarty, N., & Smeets, P. M. (1997). Relating equivalence relations to Although we have argued that the equivalence relations: A relational framing creation of novel metaphorical lan- model of complex human functioning. The guage may involve the discrimination Analysis of Verbal Behavior, 14, 57-83. of formal relations prior to the first Barnes, D., & Holmes, Y (1991). Radical be- haviorism, stimulus equivalence, and human demonstration of the new metaphor, we cognition. The Psychological Record, 41, 19- would also argue that a prior history of 31. arbitrary relational responding is nec- Barnes, D., & Keenan, M. (1993). A transfer of essary for the initial discrimination of functions through derived arbitrary and non- the relations to function arbitrary stimulus relations. Journal of the Ex- nonarbitrary perimental Analysis of Behavior, 59, 61-81. as the basis for metaphor (i.e., verbal Barnes-Holmes, D., Barnes-Holmes, Y, & Cul- history -> nonarbitrary relations -o ar- linan, V. (2000). Relational frame theory and bitrary relations). The metaphor, "surf- Skinner's Verbal Behavior: A possible synthe- ing the Internet," for example, is sure- sis. The Behavior Analyst, 23, 69-84. Chase, P N. (1986). Three perspectives on ver- ly based, in part, on the verbal history bal learning: Associative, cognitive and op- that gave rise to descriptions of both erant. In P N. Chase & L. J. Parrott (Eds.), surfing the sea (e.g., "surfing is fast Psychological aspects of language: The West UNDERSTANDING METAPHOR 199

Virginia lectures (pp. 5-35). Springfield, IL: Derived relational responding as learned be- Charles C Thomas. havior. In S. C. Hayes, D. Barnes-Holmes, & Dymond, S., & Barnes, B. (1994). A transfer B. Roche (Eds.) Relational frame theory: A of self-discrimination response functions post-Skinnerian approach to language and through equivalence relations. Journal of the cognition (pp. 21-49). New York: Plenum. Experimental Analysis of Behavior, 62, 251- Ortony, A. (1993). Metaphor and thought. 267. Cambridge, England: Cambridge University Fields, L., Hobbie-Reeve, S. A., Adams, B. J., Press. & Reeve, K. J. (1999). Effects of training di- Pilgrim, C., & Galizio, M. (1995). Reversal of rectionality and class size on equivalence class baseline relations and stimulus equivalence: I. formation by adults. The Psychological Rec- Adults. Journal of the Experimental Analysis ord, 49, 703-724. of Behavior, 63, 225-238. Hayes, S. C. (1991). A relational control theory Roche, B., & Barnes, D. (1997). A transfor- of stimulus equivalence. In L. J. Hayes & P mation of respondently conditioned stimulus N. Chase (Eds.), Dialogues on verbal behav- function in accordance with arbitrarily appli- ior: the first international institute on verbal cable relations. Journal of the Experimental relations (pp. 19-40). Reno, NV: Context Analysis of Behavior, 67, 275-301. Press. Skinner, B. F (1957). Verbal behavior. New Hayes, S. C. (1994). Relational frame theory: York: Appleton-Century-Crofts. A functional approach to verbal events. In S. Skinner, B. F (1989). Recent issues in the anal- C. Hayes, L. J. Hayes, M. Sato, & K. Ono ysis of behavior. Columbus, OH: Merrill. (Eds.), Behavior analysis of language and Spencer, T. J., & Chase, P. N. (1996). Speed cognition (pp. 9-30). Reno, NV: Context analyses of stimulus equivalence. Journal of Press. the Experimental Analysis of Behavior, 65, Hayes, S. C., & Barnes, D. (1997). Analyzing 643-659. derived stimulus relations requires more than Steele, D., & Hayes, S. C. (1991). Stimulus the concept of stimulus class. Journal of the equivalence and arbitrarily applicable rela- Experimental Analysis of Behavior, 68, 235- tional responding. Journal of the Experimen- 270. tal Analysis of Behavior, 56, 519-555. Hayes, S. C., Bames-Holmes, D., & Roche, B. Stewart, I., Barnes-Holmes, D., & Roche, B. (2001). Relationalframe theory: A post-Skin- (2000, May). Advancing the experimental nerian approach to language and cognition. analysis of metaphor with relational frame New York: Plenum. theory. Paper presented at the annual conven- Hayes, S. C., Fox, E., Gifford, E. V., Wilson, K. tion of the Association for Behavior Analysis, G., Barnes-Holmes, D., & Healy, 0. (2001). Washington, DC.