EUROPEAN JOURNAL OF BEHAVIOR ANALYSIS 2002, 3 , 95 - 102 NUMBER 2 (WINTER 2002) 95

Who Can Understand Relational Frame Theory?

François Tonneau
University of Guadalajara

Before replying to Barnes-Holmes and S. C. ditioning (e.g., Hall, 1996), and the presumed
Hayes’ questions (this issue), some background weakness of the latter is a matter of dispute (e.g.,
may help the readers understand the nature of Savastano & Miller, 1998). Importantly, from the
the current debate. Among the commentaries on very perspective of RFT, if function transfer were
Equivalence relations: A critical analysis (Tonneau, impossible through backward pairings then no
2001) that were published in this journal, some amount of operant reinforcement could make it
defended relational frame theory (RFT) as a grow (Tonneau, 2001, p. 121). Finally, in my re-
promising alternative to a correlation-based ap- ply to commentators (pp. 121-123) I sketched a
proach to symbolic behavior. Barnes-Holmes, S. correlation-based account of the main results of
C. Hayes, and Roche (2001), in particular, de- Barnes, Hegarty, and Smeets (1997) and Dymond
fended RFT and criticized the approach I pro- and Barnes (1995).
posed on two main issues. As far as I can judge, Barnes-Holmes and S.
A first issue concerned possible difficulties C. Hayes (this issue) do not question that this kind
with symmetry in matching to sample and with of account could work; instead they turn to
A-B, A-C function transfer when the training tri- broader conceptual and theoretical problems
als are mixed. After suggesting that a Pavlovian concerning associationism, RFT, and behavior
process could not operate in the latter conditions analysis. Barnes-Holmes and S. C. Hayes also
“unless one wishes to invoke the problematic object to my depiction of RFT as confused (Ton-
concept of backward associative conditioning,” neau, 2001, pp. 118-120). Thus, the nature of the
Barnes-Holmes, S. C. Hayes, and Roche (2001, debate has shifted significantly.
p. 37) concluded that I had “completely failed”
to address the emergence of bidirectional be- Tactics of Scientific Research
havioral phenomena. A second challenge raised
by Barnes-Holmes, S. C. Hayes, and Roche (2001) As an example of strategic difference between
involved experimental results such as those of RFT and the stimulus-correlation framework I
Barnes, Hegarty, and Smeets (1997) and Dymond proposed, Barnes-Holmes and S. C. Hayes sug-
and Barnes (1995). gest that the potential scope of the former
Unfortunately for these arguments, explana- (namely, “human language and cognition”) is
tions of matching symmetry in terms of for- broader. I think the reverse is true. Because a cor-
ward pairings have been available since 1992 (L. relation-based account aims at explaining results
J. Hayes, 1992: cited in my target article on p. 19; that are usually discussed in terms of memory
more on this below), function transfer after mixed retrieval (Tonneau, 1990), its potential scope with
A-B, A-C trials does not require backward con- respect to both human and nonhuman behavior
Correspondence concerning this article may be sent to is much broader than that of RFT (Tonneau,
François Tonneau, Centro de Estudios e Investigaciones en 2001, p. 124). The phenomena that relational frame
Comportamiento, Universidad de Guadalajara, 12 de Diciembre
204, Col. Chapalita, CP 45030, Guadalajara - Jalisco, MEXICO. E- theorists study are complex examples of the ef-
mail: [email protected]
95
96 François Tonneau

fect of environmental networks on behavior. The scurity (Tonneau, 2001, pp. 118-120) obviously
very complexity of these examples, however, concerned the concepts used in RFT, not the experi-
make them ill-fitted to serve as a basis for a com- ments devised by relational frame researchers. I
prehensive theory of behavior. If RFT cannot mention this point because Barnes-Holmes and
even explain basic memory phenomena in Pav- S. C. Hayes seem to believe that experimental
lovian conditioning (see Tonneau, 1990), for in- precision guarantees theoretical clarity; they state
stance, how can RFT hope to explain complex for example that “RFT is far from vague when it
human behavior any better? comes to suggesting what a researcher might do
Admittedly, the approach I have proposed is to establish specific relational performances.” But
at an incipient stage of development (Tonneau, that RFT is far from vague in suggesting particu-
2001, p. 26), whereas RFT has already generated lar procedures has no implication for the theo-
numerous experiments; no one could seriously retical coherence of RFT. Equivalence-class re-
deny the heuristic value of the package of intui- search, for example, involves perfectly clear pro-
tions and procedures known as RFT. Having a cedures associated with confused theoretical dis-
list (even a lengthy one) of intuitions and proce- course. The procedures (or more exactly, their
dures, however, is not the same as having a theory. descriptions) are far from vague; yet the theory
The analogy with Darwin is particularly reveal- seems confused and illogical.
ing here, since his goals were obviously not the The same holds of RFT, since, among other
“pragmatic” ones that Barnes-Holmes and col- things, relational frame theorists appeal to “rela-
leagues like to present as “the” goals of behavior tions” that don’t exist and “behaviors” that don’t
analysis; prediction and influence are appropriate exist either, such as the relational “behavior” or
aims for animal breeding, but not for biology. “responding” of less-than. As I pointed out (Ton-
As to Barnes-Holmes and S. C. Hayes’ claim that neau, 2001, p. 118), the only way to make sense
Darwin presented “a very limited data set” in of this sort of discourse is to interpret it on an
favor of his theory of evolution, I can only ask as-if basis; in the study by Dymond and Barnes
the readers to have a look at The origin of species (1995), for instance, some subjects behaved as if
(Darwin, 1859) and judge for themselves. a syllable were larger or more than another. But
Another point that needs correcting is Barnes- these as-if descriptions cannot serve as an ac-
Holmes and S. C. Hayes’ suggestion that I rec- count of behavior, since an account of behavior
ommend “computational models” in addition to should appeal to actual variables as opposed to
(or in replacement of ?) data and experiments. As as-ifs.
I pointed out (Tonneau, 2001, p. 123, Footnote When we turn to the unpacking of these as-ifs
11), I do not support the development of com- in terms of actual variables, RFT proves equally
putational models. What I do support is the com- obscure. The notion of a history (or is it a re-
puter modeling of behavioral theories (e.g., Epstein, sponding?) being “brought to bear” on some-
1985; Mazur, 1982), for the same well-known thing else is particularly difficult to reconstruct
reasons that computer modeling is used in phys- (see Tonneau, 2001). Barnes-Holmes and S. C.
ics, chemistry, or evolutionary biology. In short, I Hayes now claim that the notion of “bringing to
recommend the development of computable mod- bear” a history was only a turn of phrase and
els of behavioral processes to evaluate theoretical that the technical details can be found in a recent
predictions in situations that involve multiple cau- book on RFT (S. C. Hayes, Barnes-Holmes, &
sation and highly complex histories. And, of Roche, 2001). After reading the book in question
course, I nowhere suggest that computer simu- from cover to cover, however, I have been un-
lations should replace experiments. able to discern the promised details. As far as I
can see, the book largely repeats the turns of
The Obscurity of RFT phrases of RFT without clarifying them. The
notion that contextual cues “bring to bear” a his-
Barnes-Holmes and S. C. Hayes note that I tory merely in the sense of being discriminative
find RFT vague and obscure. My charge of ob- for relational responding doesn’t help, since in
 Who Can Understand Relational Frame Theory? 97

experiments such as those of Dymond and from stimulus correlations and independently of
Barnes (1995) this relational “responding” can’t the process of operant reinforcement (as shown
be defined at all. Likening “relational respond- by the fact that function transfer can be demon-
ing” to generalized imitation and other response strated in purely Pavlovian settings).
classes (see S. C. Hayes, Barnes-Holmes, & Roche, Importantly, a Pavlovian conception of func-
2001, p. 23) doesn’t help either, since in these ex- tion transfer does not imply that symbolic perfor-
amples behavior is not defined through as-if mance can be understood independently of oper-
clauses. In generalized imitation for instance, sub- ant responses and their reinforcers. Far from it.
jects do not behave as if their behavior were simi- Since (Pavlovian) stimulus correlations are sup-
lar to that of the model; rather, their behavior is posed to transfer not only traditional respondent
similar to that of the model. Thus the defining behaviors, but also the responses that are main-
criteria of the response class actually hold, in con- tained by operant reinforcement, a complete ac-
trast to those of the “behaviors” (e.g., “less”) that count of symbolic performance must clearly in-
relational frame theorists routinely invoke. clude the latter (see Tonneau, 2001, p. 124). If,
In sum, the principal concepts of RFT seem for example, a response R1 is reinforced condi-
obscure, if not incoherent, and Barnes-Holmes tionally on a stimulus A, so that R1 comes to pre-
and S. C. Hayes’ commentary provides no clari- dominate in the presence of this stimulus, then it
fication. Importantly, the computer simulations is R1 (and not, say, R2 or R3) that will transfer to
of Barnes-Holmes and colleagues (e.g., Barnes the B stimuli correlated with A. Hence operant
& Hampson, 1993) do nothing to clarify the con- reinforcement determines what is being transferred
cepts used in RFT, since these simulations do not through the A-B correlation; but it is the A-B
rely on such concepts (e.g., the concept of a “re- correlation, not operant reinforcement, that ex-
lational responding” of less, or the concept of a plains why function transfer takes place at all. Of
history being “brought to bear” on something course, if responses and reinforcers have them-
else), but instead provide straightforward cogni- selves stimulus functions, they can support func-
tive explanations of RFT data in terms of inter- tion transfer by association with other stimuli.
nal associations and representations. To this notion of function transfer as being
driven by stimulus correlations, I have added an
Associations and Reinforcement important, but highly speculative, hypothesis to
deal with the extinction problem in relation to sym-
Barnes-Holmes and S. C. Hayes allude to bolic behavior (see Tonneau, 2001, p. 25). To
associationism in their commentary, but differ- explain why the function-transfer effects of cor-
ent views of associationism coexist in psychol- relations between words and objects seem to
ogy and should be distinguished from one an- persist (at least in daily life) even when the origi-
other. Moreover, the position of RFT on nal word-object correlations are degraded, I have
associationism is difficult to understand, since proposed that operant reinforcement might pro-
after presenting the associationist model of Barnes tect function transfer from extinction (Tonneau,
and Hampson (1993) as a precise account of RFT 2001, p. 25). This speculation raises complex theo-
data, Barnes-Holmes and S. C. Hayes suggest that retical issues that are beyond the scope of the
associationist accounts of RFT data have diffi- present reply. If the hypothesis turns out to be
culties that RFT hasn’t. correct, however, it still grants no role to operant
Traditionally, associations have been said to reinforcement in producing function transfer; the
exist (a) in the mind, (b) in the brain, and (c) in the hypothesized role of operant reinforcement is
environment. The mediational theorizing of to maintain function transfer once it occurs, not
Barnes and Hampson (1993) is an example of to produce it out of nothing.
associationism in senses (a) and (b). Consistent In general, I do not see how operant rein-
with a behavioristic stance (e.g., Skinner, 1977, p. forcement could produce a phenomenon like
1), however, position (c) is the one I have advo- function transfer; here as elsewhere, reinforcement
cated. I have argued that function transfer arises as such can only influence the rate of phenom-
98 François Tonneau

ena that occur for non-operant reasons (Tonneau, 1996, p. 703), consistent with the suggestion of
2001, p. 121). This point of logic about reinforce- Barnes-Holmes, S. C. Hayes, and Roche (2001)
ment underscores the necessity of specifying prin- that the effects of stimulus pairings in such stud-
ciples of behavioral induction to complement ies may depend upon “the generalized operant pro-
principles of reinforcement and punishment. cess of relational framing” (p. 37). Roche and
Contrary to Barnes-Holmes and S. C. Hayes’ s Barnes (1997) similarly attribute their findings on
allusion to ultimate and proximate causes, induc- respondent behavior not to respondent mecha-
tion principles deal with what produces function nisms but to RFT processes (p. 289) that are pre-
transfer in the history of particular organisms, sumably operant in nature (cf. Barnes-Holmes, S.
not with the evolutionary origins of function trans- C. Hayes, & Roche, 2001, p. 35). I do not see
fer in previous generations. I have argued that as how Barnes-Holmes and S. C. Hayes can con-
far as function transfer is concerned, the relevant clude on the basis of such reports that RFT (the
principles are Pavlovian. theory) incorporates respondent processes into its
explanatory framework.
Associations and RFT When we turn to the theoretical statements
made by relational frame researchers, understand-
According to Barnes-Holmes and S. C. Hayes, ing the place of respondent processes in RFT
RFT already includes a significant Pavlovian com- becomes even more challenging. The earliest
ponent. They state for example that “respondent material that Barnes-Holmes and S. C. Hayes
processes are important for relational frame mention is a book chapter by S. C. Hayes and L.
theory” and that RFT “has always explicitly in- J. Hayes (1989). Far from incorporating respon-
corporated respondent processes into its explana- dent processes into RFT, these authors deny that
tory framework.” respondent processes could explain RFT data.
Note here that Barnes-Holmes and S. C. Hayes After contrasting Pavlovian conditioning with
refer to RFT as a theory and as an explanatory frame- bidirectional behavioral phenomena (p. 162, p.
work, as opposed to a series of experiments. (I 165), S. C. Hayes and L. J. Hayes state that “upon
take it for granted that a theory is not an experi- further consideration of equivalence findings, the
ment. The readers will certainly understand the plausibility of a respondent conditioning expla-
difference between, say, the theory of cognitive nation, even for the finding of symmetry, cannot be
dissonance and the experiments it has generated.) sustained” (1989, p. 166, emphasis mine). S. C.
Notice also that Barnes-Holmes and S. C. Hayes Hayes and L. J. Hayes (1989) later discuss the
speak of incorporating respondent processes, and transfer of perceptual functions in terms of a
not merely respondent behaviors, in RFT. Now “relational class” (p. 171).
one would surely expect respondent processes Barnes-Holmes and S. C. Hayes also quote an
to be relevant to respondent behaviors and vice article of Barnes and Roche (1996) to show that
versa; however, relational frame theorists have a respondent behavior is “an important founda-
tradition of experimenting with stimulus pairings, tion for relational framing,” whereas Barnes-
and sometimes even respondent behaviors, with- Holmes, S. C. Hayes, and Roche (2001, p. 37)
out incorporating respondent processes into RFT. had previously mentioned the same article to sug-
The studies of stimulus pairing that Barnes- gest that respondent processes “do not play a
Holmes and S. C. Hayes mention, for example, primary explanatory role in RFT (Barnes &
emphasize empirical issues and sometimes barely Roche, 1996).” How an important foundation
mention RFT (e.g., Leader & Barnes-Holmes, for relational framing could fail to play a pri-
2001). When an explanation of the findings is mary explanatory role in RFT is unclear to me,
attempted, the operative processes are said to be but in any case, Barnes and Roche (1996, p. 501)
“largely operant” (Leader, Barnes-Holmes, Smeets, did comment in a footnote on the importance
2000, p. 76) and to involve “equivalence relations” of respondent behavior, referring the reader to
or histories that were “brought to bear by vari- the theoretical work of Barnes (1994).
ous contextual cues” (Leader, Barnes, & Smeets,
 Who Can Understand Relational Frame Theory? 99

In this work, Barnes (1994) explains match- could make function transfer grow if it did not
ing symmetry (among other findings) through occur through stimulus pairings, but in any event,
mediational processes that involve the Pavlovian these authors’ claims about AB, AC effects in clas-
conditioning of covert perceptual behaviors. sical conditioning are false (e.g., Ward-Robinson
Aside from the RFT notion that the presence of & Hall, 1996).
respondent behaviors and other elements must So what is the role of correlation-based pro-
work “as a contextual cue for the relational frame cesses in RFT? Six possibilities may be discerned:
of sameness” (p. 109), the explanation is similar (1) Respondent behaviors are an important foun-
to that of L. J. Hayes (1992) which Barnes men- dation for relational framing, but they do not play
tions (1994, p. 108) and that I did cite in the ar- a primary explanatory role in RFT. (2) Respon-
ticle portrayed by Barnes-Holmes, S. C. Hayes, dent processes are important for RFT, but it
and Roche (2001, p. 37) as having completely failed doesn’t matter whether they are present or not.
to address the issue of matching symmetry. (3) RFT does not dispense with bidirectional,
Importantly, the Pavlovian part of Barnes’ correlation-based processes, but RFT is an alter-
(1994) account does not “facilitate” the emer- native to a correlation-based account. (4) RFT does
gence of matching symmetry (as Barnes-Holmes dispense with respondent processes, which can-
and S. C. Hayes misleadingly say), but is a sine qua not deal with matching symmetry and other find-
non for its occurrence: Remove the Pavlovian ings. (5) Respondent processes can deal with find-
component, and matching symmetry would in- ings such as matching symmetry, but RFT does
stantly collapse. Barnes-Holmes and S. C. Hayes not depend on such speculations. (6) RFT does
nonetheless write that RFT as a theory will not involve respondent processes, and relational frame
stand or fall on the veracity of Barnes’s (1994) theorists understand it from time to time, but
account, in spite of the fact that no alternative otherwise keep claiming that respondent processes
account of matching symmetry is available in RFT. can’t explain RFT data.
Meanwhile, after finding all sorts of “diffi-
culties” with a correlation-based account of Transfer, Transformation, and History
matching symmetry and related phenomena, and
remarking that “given these difficulties, it is per- One of Barnes-Holmes and S. C. Hayes’ cen-
haps not hard to understand why Tonneau re- tral points is that a correlation-based account of
fused to address the issue of symmetry in his RFT data is similar to RFT. One could of course
article” (p. 37), Barnes-Holmes, S. C. Hayes, and put it the other way round, and suggest that it is
Roche (2001) introduce RFT as an operant alter- RFT that is similar to an associationist explana-
native to a respondent account (p. 37). Paradoxi- tion (although the status of the “relations” in-
cally, we are now informed that RFT does not voked in RFT seems obscure from a behaviorist
dispense with bidirectional, correlation-based standpoint). All detailed explanations of results
processes and that I am simply wrong to imply such as those of Dymond and Barnes (1995) will
that it does (Barnes-Holmes and S. C. Hayes, this obviously need to take into account the correla-
issue). tional structure of the environment, and thus ex-
Some relational frame theorists, however, cer- hibit a family resemblance at a formal level.
tainly believe that RFT dispenses with bidirec- Barnes-Holmes and S. C. Hayes suggest, how-
tional, correlation-based processes, because they ever, that our accounts of the data of Barnes,
claim that RFT can operate in their absence. S. C. Hegarty, and Smeets (1997) might differ. I have
Hayes, Fox, Gifford, Wilson, Barnes-Holmes, and not been able to understand all of the details of
Healy (2001), for example, explicitly state that B- the argument, which appeals to RFT notions that
C function transfer through AB, AC pairings “can- I find unintelligible (such as the relational response
not be due to classical conditioning because it SAME AS). To reiterate what should be obvi-
would require an appeal to backward condition- ous, SAME AS is not a “relational response” but
ing” (p. 46), the alternative being RFT. Again, I an environmental relation, and this relation does
do not understand how operant reinforcement not hold of the compounds present in the test
100 François Tonneau

phase of Barnes, Hegarty and Smeets’ (1997) extraneous stimulus whose behavioral effects, dif-
study. From a molar perspective, SAME AS is ferent from those of A, transfer to B.
nevertheless present in the subject’s environmen- Such extraneous C stimuli are often hidden in
tal path (since the latter includes non-contiguous the subject’s pre-experimental past, which makes
stimulus pairs such as C1 and C1, for example), the actual source of function transfer difficult to
and, pending potentiation by other variables, the identify. If the approach I propose is correct,
presence of SAME AS in the path can influence these stimuli nevertheless exist and play a crucial
performance in testing. What the subjects will do role in the putative cases of function “transfor-
in testing depends on what operant responses mation” that relational frame theorists report. I
were reinforced conditionally on SAME AS and have previously explained, in one case chosen for
contextual variables such as the matching-to- illustrative purposes, how this sort of account
sample format itself (in which identity matching could apply to the behavior observed in Dymond
has previously been reinforced, I suppose). No- and Barnes’ (1995) study—namely, choosing C2
tice that the behavior reinforced conditionally on after two bar presses (Tonneau, 2001, p. 123).
SAME AS, and that reappears in testing, is the Admittedly, my explanation is highly tentative, but
behavior of matching and not SAME AS. I do not see how Barnes-Holmes and S. C. Hayes
In contrast to the data of Barnes, Hegarty, can write that I avoid to provide appropriate
and Smeets (1997), my account of the examples descriptions of this sort of behavior.
of function transformation reported by relational Barnes-Holmes and S. C. Hayes also claim that
frame theorists highlighted an important differ- they “find nothing objectionable in the idea that
ence between a correlation-based approach and a training history transferred to current stimuli.”
RFT (Tonneau, 2001, pp. 122-123). The issue of But this notion is theirs and not mine. Clearly it is
function transformation, far from being seman- relational frame theorists who believe that histo-
tic, is theoretically crucial; for if processes of ries transfer (or are “applied,” “brought to bear,”
function transformation actually existed, they etc.) to current stimuli. From the perspective I
would seriously challenge any stimulus-correla- propose, of course, it is behavioral functions that
tion account of complex behavior. Unfortunately transfer, not histories (the RFT notion of a his-
for RFT, however, I have argued that processes tory transferring to something seems incoherent
of function transformation do not exist and that to me).
the apparent cases of function transformation Finally, my remark on the lack of historical
reported by relational frame researchers can be specification in RFT (Tonneau, 2001, p. 123) con-
explained as function transfer (Tonneau, 2001, pp. cerned the all-too-easy RFT accounts of data such
121-123). as those of Barnes, Hegarty and Smeets (1997)
I am not sure that Barnes-Holmes and S. C. and Dymond and Barnes (1995) in terms of the
Hayes fully understand my argument. Thus they subjects’ abilities to “relate relations” or “trans-
discuss a hypothetical case of function “trans- form” functions in accordance with relations that
formation” in which a stimulus A is established don’t even hold. How Barnes-Holmes and S. C.
as the “opposite” of B (in RFT parlance), and in Hayes can maintain that my explanations (Ton-
which A is avoided whereas B isn’t. According neau, 2001, pp. 121-123) lack “real detail” as com-
to Barnes-Holmes and S. C. Hayes, speaking of pared to the turns of phrases of RFT, is a puzzle.
function transfer from A to B would be inaccu- As to Barnes-Holmes and S. C. Hayes’ sug-
rate. I certainly agree that speaking of function gestion that the reinforcement of identity match-
transfer from A to B would be inaccurate. How- ing is an important component of RFT, the only
ever, I suggest that speaking of function trans- supporting statements that I could find in the RFT
formation from A to B would also be inaccurate. literature (aside from Barnes’ 1994 comments on
The hypothesis I propose is that the situation dis- the “relational frame of sameness”) are along
cussed by Barnes-Holmes and S. C. Hayes actu- these lines: “It is necessary to recognize an event
ally involves function transfer—not from A to B, as itself in order to respond to that event in terms
of course, but from C to B, where C is a third, of derived relations” (S. C. Hayes, 1991, p. 32).
 Who Can Understand Relational Frame Theory? 101

If these considerations on event recognition are Hayes, S. C. (1991). A relational control theory
the RFT way of saying that the reinforcement of of stimulus equivalence. In L. J. Hayes & P.
identity matching is involved in results such as N. Chase (Eds.), Dialogues on verbal behavior (pp.
those of Barnes, Hegarty, and Smeets (1997), I 19-40). Reno, NV: Context Press.
admit to having misunderstood the message. Hayes, S. C., Barnes-Holmes, D., & Roche, B.
(2001). (Eds.). Relational frame theory: A post-
References Skinnerian account of human language and cogni-
tion. New York: Plenum.
Barnes, D. (1994). Stimulus equivalence and rela- Hayes, S. C., Fox, E., Gifford, E. V., Wilson, K.
tional frame theory. Psychological Record, 44, 91- G., Barnes-Holmes, D., & Healy, O. (2001).
124. Derived relational responding as learned be-
Barnes, D., & Hampson, P. J. (1993). Stimulus havior. In S. C. Hayes, D. Barnes-Holmes, &
equivalence and connectionism: Implications B. Roche (Eds.), Relational frame theory: A post-
for behavior analysis and cognitive science. Psy- Skinnerian account of human language and cognition
chological Record, 43, 617-638. (pp. 21-49). New York: Plenum.
Barnes, D., Hegarty, N., & Smeets, P. M. (1997). Hayes, S. C., & Hayes, L. J. (1989). The verbal
Relating equivalence relations to equivalence action of the listener as a basis for rule-gov-
relations: A relational framing model of com- ernance. In S. C. Hayes (Ed.), Rule-governed be-
plex human functioning. Analysis of Verbal Be- havior: Cognition, contingencies, and instructional con-
havior, 14, 57-83. trol (pp. 153-190). New York: Plenum.
Barnes, D., & Roche, B. (1996). Relational frame Leader, G., Barnes, D., & Smeets, P. M. (1996).
theory and stimulus equivalence are fundamen- Establishing equivalence relations using a re-
tally different: A reply to Saunders’ commen- spondent-type training procedure. Psychologi-
tary. Psychological Record, 46, 489-508. cal Record, 46, 685-706.
Barnes-Holmes, D., Hayes, S. C., & Roche, B. Leader, G., & Barnes-Holmes, D. (2001). Match-
(2001). The (not so) strange death of stimu- ing-to-sample and respondent- type training
lus equivalence. European Journal of Behavior as methods for producing equivalence rela-
Analysis, 2, 35-41. tions: Isolating the critical variables. Psychologi-
Darwin, C. (1859). On the origin of species by means cal Record, 51, 429-444.
of natural selection, or the preservation of favoured Leader, G., Barnes-Holmes, D., & Smeets, P. M.
races in the struggle for life. London: John Murray. (2000). Establishing equivalence relations us-
Dymond, S. & Barnes, D. (1995). A transforma- ing a respondent-type procedure III. Psycho-
tion of self-discrimination response functions logical Record, 50, 63-78.
in accordance with the arbitrarily applicable Mazur, J. E. (1982). A molecular approach to
relations of sameness, more than, and less ratio schedule performance. In M. L. Com-
than. Journal of the Experimental Analysis of Be- mons, R. J. Herrnstein, & H. Rachlin (Eds.),
havior, 64, 163-184. Quantitative analyses of behavior: Vol. 2. Matching
Epstein, R. (1985). Animal cognition as the praxist and maximizing accounts (pp. 79-110). Cam-
views it. Neuroscience & Biobehavioral Reviews, 9, bridge, MA: Ballinger.
623-630. Roche, B. & Barnes, D. (1997). A transforma-
Hall, G. (1996). Learning about associatively acti- tion of respondently conditioned stimulus
vated stimulus representations: Implications function in accordance with arbitrarily appli-
for acquired equivalence and perceptual learn- cable relations. Journal of the Experimental Analy-
ing. Animal Learning & Behavior, 24, 233-255. sis of Behavior, 67, 275-301.
Hayes, L. J. (1992). Equivalence as process. In S. Savastano, H. I., & Miller, R. R. (1998). Time as
C. Hayes & L. J. Hayes (Eds.), Understanding content in Pavlovian conditioning. Behavioural
verbal relations (pp. 97-108). Reno, NV: Con- Processes, 44, 147-162.
text Press. Skinner, B. F. (1977). Why I am not a cognitive
psychologist. Behaviorism, 5(2), 1-10.
102 François Tonneau

Tonneau, F. (1990). From reflex to memory: Ward-Robinson, J. & Hall, G. (1996). Backward
Molar sequences in Pavlovian and instrumen- sensory preconditioning. Journal of Experimen-
tal conditioning. Psychological Record, 40, 587- tal Psychology: Animal Behavior Processes, 22, 395-
607. 404.
Tonneau, F. (2001). Equivalence relations: A criti-
cal analysis. European Journal of Behavior Analy-
sis, 2, 1-128. (Includes commentary.)