2011 - Bortoloti&deRose 2011 Orwellian

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

EUROPEAN JOURNAL OF BEHAVIOR ANALYSIS 2011, 12, 121 - 134 NUMBER 1 (SUMMER 2011)

121

An “Orwellian” Account of
Stimulus Equivalence.
Are Some Stimuli “More Equivalent”
Than Others?
Renato Bortoloti and Julio C. de Rose
Federal University of São Carlos and Instituto Nacional de Ciência e Tecnologia
sobre Comportamento, Cognição e Ensino

Stimulus equivalence is widely accepted as a model of semantic relations. However, few stud-
ies have used methods other than matching-to-sample to assess whether equivalent stimuli are
semantically related. This article describes that formed equivalence classes with arbitrary stimuli
and faces expressing emotions. A semantic differential showed that the arbitrary stimuli were then
evaluated similarly to the faces, indicating that they had acquired similar meanings. The semantic
differential permitted a quantitative assessment of the degree of relatedness between equivalent
stimuli, and showed that relatedness varied as a function of nodal distance and simultaneous or
delayed matching in training. Another study confirmed the semanticity of relations between faces
expressing emotions and their equivalent stimuli through the IRAP. This study also confirmed
that stimuli in classes formed with delayed matching-to-sample are more strongly related than
those in classes formed with simultaneous matching.
Key words: stimulus equivalence, transfer of functions, relatedness, nodal distance, meaning

The model of stimulus equivalence to sample procedures to verify if a relation


proposed by Sidman and colleagues (e.g., arbitrarily established between dissimilar
Sidman, 1986, 1994; Sidman & Tailby, stimuli has the logical properties of reflex-
1982) offered operational criteria to iden- ivity, symmetry and transitivity. If so, these
tify symbolic functions in experimentally stimuli will constitute a class of equivalent
observable behaviors and to simulate in stimuli in which each member will be sub-
laboratory symbolic relations that occur in stitutable by the others, simulating stimuli
natural situations. The model offered, for that are symbolic related, for example, in
instance, behavioral tests based on matching natural languages.

Renato Bortoloti, Departamento de Psicologia, Universi- author was supported by a post-doctoral grant from FAPESP,
dade Federal de São Carlos; Julio C. de Rose, Departamento and the second author was supported by a research productivity
de Psicologia, Universidade Federal de São Carlos. grant from CNPq. We thank Deisy de Souza and Bill McIlvane
Author order in this manuscript is alphabetical, since for their contributions and support. This manuscript is based
contribution of both authors is roughly equivalent. Research on a paper presented at the V Conference of the European
reported in this manuscript was supported by the National Association for Behavior Analysis, in Rethymno, Greece.
Research Council (CNPq, Grant 573972/2008-7) and the Correspondence concerning this article should be addres-
State of São Paulo Foundation for Research Support (FAPESP, sed to Renato Bortoloti or Julio C. de Rose, Departamento de
Grant 08/57705-8). Support by CNPq, Grant 400576/2010- Psicologia, Universidade Federal de São Carlos, Caixa Postal
4, to Renato Bortoloti is also gratefully acknowledged. The first 676, São Carlos, SP, Brazil, 13565-905.

121
122 Renato Bortoloti and Julio C. de Rose

Several studies have demonstrated that stimuli in equivalence classes may differ in
when an equivalence class is established, their degree of “relatedness” (see below). How-
responses emitted to a stimulus, or functions ever, the use of matching to sample procedures
acquired by a stimulus, will transfer to the is not appropriate to detect possible differences
other members of the class (e.g., Barnes- in the relatedness of equivalent stimuli. Match-
Holmes, Keane, Barnes-Holmes, & Smeets, ing to sample procedures require forced choices
2000; de Rose, McIlvane, Dube, Galpin, between discrete alternatives that can only de-
& Stoddard, 1988; Dougher, Augustson, termine if the participant has or has not formed
Markham, Greenway, & Wulfert, 1994; classes of equivalent stimuli. It is not possible
Hayes, Kohlenberger, & Hayes, 1991), to determine, for example, if the stimuli in the
which is compatible with the notion that in classes are equally related to each other.
many contexts, we react to the symbols as if If equivalence relations can be established
we were facing the events referred by them. at different levels, the knowledge of the condi-
Thus, a stimulus that has acquired some tions that favor the establishment of stronger
functions may be taken as a referent and the relations constitutes an important scientific
stimuli that are equivalent to the referent can challenge that has not only theoretical im-
be taken as symbols that can substitute the plications, but it has also implications for
referent in some contexts. Equivalence classes the praxis of basic, translational and applied
are also context dependent, i.e., a stimulus research involving the stimulus equivalence
A1 may be equivalent to B1 and not to B2 paradigm. The development of methods that
in a particular context and equivalent to B2 are more sensitive to the degree of relatedness
and not to B1 in a different context (Bush, is necessary for the experimental investigation
Sidman, & T. de Rose, 1989; Wulfert & of variables that influence relatedness. Belanich
Hayes, 1988). and Fields (2003) proposed that equivalence
These and other studies showing con- classes may be conceived as functional transfer
gruence between properties of equivalence networks, and that measures of functional
classes with those expected for symbolic transfer would be more adequate to detect dif-
relations (e.g., Barnes-Holmes et al., 2005; ferences in relatedness. This article summarizes
Haimson, Wilkinson, Rosenquist, Ouimet, recent research developing and applying dif-
& McIlvane, 2009) seem to confirm that the ferent methods enabling a quantitative assess-
tests specified by Sidman and colleagues are ment of the degree of functional transfer. The
quite suitable to assess whether the behavioral use of these methods has not only confirmed
relations under study achieve or not the cri- that functional transfer is influenced by nodal
teria to be considered symbolic. Assuming distance, as shown by Fields and colleagues
this, we must also recognize that these tests (Fields, Adams, Verhave, & Newman, 1993;
do not capture all dimensions of the symbolic Fields, Landon-Jimenez, Buffington, & Ad-
function. In most cases, they permit only to ams, 1995), but also revealed other variables
assess whether stimuli are related as equiva- that affect the amount of functional transfer
lent or not. While useful for many purposes, within equivalence classes. In this article we
this categorical classification does not capture will review this recent literature and suggest
the full range of relational properties that some implications for future research.
characterize certain aspects of the complex
human behavior. An example of relational An “Orwellian” Equivalence
property not captured by the categorical
classification of stimulus equivalence is the In Animal Farm (by George Orwell),
degree of “relatedness” established between the basic commandment that “all animals
two terms linked by meaning. Lanny Fields are equal” is changed to “all animals are
and colleagues have repeatedly argued that equal but some are more equal than others”.
An “Orwellian” Account of Stimulus Equivalence 123

Orwell was aware of the inherent contradic- For instance, Fields et al. (1993) trained
tion in the assertion of different degrees of conditional discriminations AB, BC, CD,
equality, and one aim of the book is precisely and DE, leading to the formation of two
to point out this contradiction in the society five-member equivalence classes. They
that the work satirizes. Equality is an equiva- then trained different responses in the
lence relation, and an equivalence relation presence of A1 and A2, and observed that
could not admit degrees. However, Lanny discriminative control over these responses
Fields and colleagues have argued, without transferred to other class members. How-
irony, that equivalence relations between ever, the probability that the equivalent
stimuli may be established in different de- stimuli evoked the responses was an in-
grees or, in other words, members of a class verse function of nodal distance. The
of equivalent stimuli may have different de- responses were less reliably evoked by the
grees of relatedness between themselves (e.g., D and E stimuli. Fields et al. (1995) also
Belanich & Fields, 2003; Fields et al., 1993; conducted a similar training to establish two
Fields et al., 1995; Fields & Watanabe-Rose, five-member equivalence classes. Only two
2008). of twelve participants formed the expected
According to Fields and colleagues, the equivalence classes. The experimenters, then,
relatedness between equivalent stimuli is trained the two participants that formed
inversely proportional to the nodal distance classes to emit a response to the A stimuli
between them. A node is a stimulus related and another response to the E stimuli, and
by conditional relations to two or more stim- also observed that transfer of discriminative
uli. For instance, when the relations directly control over these responses was an inverse
trained to establish A, B, C, D, and E as a function of nodal distance. The authors not-
class of equivalent stimuli are AB, AC, AD, ed that the response trained in the presence
AE, the stimulus A is a node and the others are of A was emitted with higher frequency when
not (being called singulars). No trained rela- B was present than when D was present, and
tion involves intervening nodes, so the nodal also noted that the response trained in the
distance between all members of the class is 0. presence of E was emitted with higher fre-
The same class could be established, although quency when D was present than when B was
probably with more difficulty (cf. Arntzen, present. In addition, reaction times (response
Grondahl, & Eilifsen, 2010), by training latencies) increased with the increase of nodal
relations AB, BC, CD, and DE. In this case, distance in relation to A and to E (i.e., the re-
the class would contain three nodes, B, C, and action time was longer to the stimulus C than
D, because each of these stimuli would have to B and to D). These results are consistent
been directly related to two others. Therefore, with the notion that the degree of relatedness
in the case of this linear training protocol, between equivalent stimuli decreases with
the nodal distance between members of the nodal distance. This notion has not been
class would be variable. Taking A as reference without criticism. For instance, Imam (2006)
point, the other stimuli are at increasing nodal suggested that differences in the degree of
distances, with the nodal distance from B to transfer as a function of nodal distance might
A equal to 0, from C to A equal to 1, from D be related to the order of training, so that
to A equal to 2, and from E to A equal to 3. the total number of trials with each subse-
Fields and colleagues demonstrated that the quent conditional discrimination decreased
transfer of functions from a class member as training progressed. Imam taught college
to the others is inversely proportional to the students to establish three seven-member
nodal distance. They conclude that, as nodal equivalence classes, assuring that all trained
distance increases, the degree of relatedness and tested conditional discriminations were
between equivalent stimuli decreases. presented in an equal number of trials.
124 Renato Bortoloti and Julio C. de Rose

No difference was found in matching accu- However, when different responses were
racy and in response latency as a function of trained to the C and D members of each
nodal distance. On the basis of these results, class, the classes “bifurcated” according to
Imam challenged the generality of nodality the nodal structure, so that control over the
effects. He acknowledged, however, that this response trained to the C stimuli transferred
left unexplained the results for the two par- to the B and A stimuli, whereas control over
ticipants of Fields et al. (1995), since number the response trained to the D stimuli trans-
of trials for each conditional discrimination ferred to the E and F stimuli.
had been equalized in that study. With the exception of reaction time mea-
Sidman (1994) argued against a structural sures (e.g., Fields et al., 1995; Imam, 2006)
notion of equivalence classes and suggested the most measures in these studies were of an
use of the expression “nodal number” rather all or none type. The response trained in the
than nodal distance. He considered separately presence of one classmember either was or
effects of nodal number on the emergence of was not emitted in the presence of another
equivalence classes and on the relatedness of classmember. Also, in these studies, degree
stimuli after the classes had been formed. He of relatedness was investigated as a function
suggested that effects of nodal number on the of a single variable, nodal distance (or nodal
rate of emergence of equivalence classes were number, if a term not structurally bent is to
plausible, but the available experiments could be preferred). And studies of the effect of
be criticized on methodological grounds. He nodal distance on degree of relatedness of
pointed out, particularly, that in most stud- equivalent stimuli were conducted mostly
ies that showed an effect of nodal number on with conditional discrimination training
class emergence, training had been conducted with only two comparison stimuli, which,
with only two comparison stimuli. According as Sidman pointed out, could difficult in-
to Sidman, this would favor a competition of terpretation of results.
discriminative control by positive and nega- Bortoloti and Rose (2007, 2009) in-
tive comparison stimuli, that could lead to troduced a different method to verify the
more variability in the results (see Carrigan transfer of meaning between equivalent
& Sidman, 1992; Johnson & Sidman, 1993). stimuli, which permitted a more continu-
As mentioned by Sidman (1994), a study by ous measure of transfer of functions and,
Kennedy (1991), found that effects of nodal possibly, of relatedness. The introduction
number decreased when conditional discrimi- of a more continuous assessment of transfer
nations were trained with three comparison also opened ways to investigation of the ef-
stimuli. However, Sidman argued that the no- fect of other variables that might influence
tion of equivalent stimuli differing in degrees the degree of transfer within equivalence
of relatedness would be “either a contradiction classes. In these studies, meaningful stimuli,
in terms or must refer to differences in features faces expressing emotions, were condition-
other than those which define the class.” (Sid- ally related to arbitrary stimuli. Bortoloti
man, 1994, p. 543). and de Rose reasoned that, in this case, the
A more recent study addressing this abstract stimuli should become symbols of
question was conducted by Fields and the emotional expressions and should there-
Watanabe-Rose (2008). They established fore share the meaning of these expressions.
two six-member equivalence classes, train- They compared the meaning of the faces and
ing conditional discriminations AB, BC, the meaning of stimuli equivalent to them
CD, DE, and EF. After training a different using a widely accepted measure of meaning,
response to C1 and C2, they found that the semantic differential (Osgood, Suci, &
control over this response transferred equally Tannenbaum, 1957), in order to determine
to all other members of the respective classes. to what extent arbitrary stimuli equivalent
An “Orwellian” Account of Stimulus Equivalence 125

to a meaningful stimulus would acquire the that generated equivalence classes involving
meaning of the latter. facial expressions and abstract pictures.
The semantic differential is a measure Next, participants evaluated some of these
used to quantify the meaning that partici- pictures using a semantic differential. A
pants attribute to “concepts” (words, phrases, control group, which had not been taught
photographs, drawings, etc.). Empirical the relations between the stimuli, evaluated
research has demonstrated the reliability the abstract pictures and the faces using the
and sensitivity of the semantic differential same instrument. The abstract pictures were
as a measure of meaning in a wide variety evaluated as neutral by the control group
of conditions (Pereira, 1986) and also the (mean ratings tended to the central interval
ease of use of the instrument (Engelmann, in different scales, which received the value
1978). The semantic differential is generally zero) and the faces were evaluated differently,
comprised of a set of bipolar scales, usually depending on the emotion expressed; for in-
of seven intervals that are anchored at theirs stance, a happy face was evaluated positively
ends, on the right and on the left, by opposite and an angry face was evaluated negatively.
adjectives. The meaning of each interval is Similar values were assigned to the faces
described in the instructions for participants. and to stimuli equivalent to them (made by
For each concept presented, the participants participants submitted to the conditional
are required to indicate on each scale one discrimination training). The similarities
of the extreme intervals if they think that between the evaluations of the faces by the
the concept is “very strongly” related to the control and the evaluations of the pictures
neighboring adjective, the immediately more by the experimental group are indications
central interval if they think the concept is that there may be extension of meaning from
“more or less” related to the adjective, and “referents” to “symbols” in experimental
the third interval from the extreme to the simulations based on stimulus equivalence.
center, if they consider that the concept is This result has provided additional external
“just slightly” related to the adjective. The validation of stimulus equivalence as a model
central interval is the origin and the neutral of meaning. In addition, quantification of
point of each scale; it should be marked if the generalization of meaning among the
the concept is, for the participant, equally
related to the two adjectives or if the scale
1
Osgood et al. (1957) explained the performance of the
subject facing the semantic differential in terms of a mediation
has no relation with the concept. The central process, conceived as something that merges between stimuli
interval receives the value zero. The remain- and responses. A particular sign would elicit a specific reaction
that would be similar, in some degree, to the reaction that
ing intervals receive quantifiers expressed by would be elicited directly by the referred object. The concept,
numerical values ranging between -3 and +3. as a sign of a meaning, would elicit a mediational reaction
associated with one of the polar terms, which would determine
For example, in the scale Good:_:_:_:_:_:_:_: the choice of the pole; the intensity of this reaction would
Bad, in which the adjective “good” is con- determine the interval to be marked. There is not, however, a
necessary relationship between the adoption of the semantic
sidered positive and the adjective “bad” is differential and the acceptance of theories that its proponents
considered negative, the values between these used to ground it. According to Engelmann (1978), Osgood
and colleagues “(...) based on the concept of ‘act of pure stimu-
two polar extremes range from +3 in the left lation’ by Hull, sought to provide an interpretation of various
end to -3 in the right end. Each value marks behaviors in which there is no clear relationship between
stimuli and responses – including the use of signs – without
thus the direction of the chosen pole and the straying from the principles of conditioning.” (p. 115). The
distance to it. Direction and distance are, authors themselves admit that the empirical validity and
practical use of the semantic differential have no relation with
respectively, the quality and the intensity the learning theory chosen to present it. Using the semantic
of the concept’s meaning evaluated by the differential, Bortoloti and Rose (2007) were not committed to
the mediatory hypothesis of Osgood et al. This study can be
bipolar scale1. considered as the use of a methodology based on a measuring
Bortoloti and Rose (2007) trained col- instrument with empirical validity widely recognized in other
situations to assess whether equivalence relations may provide
lege students in conditional discriminations meaning to previously meaningless stimuli.
126 Renato Bortoloti and Julio C. de Rose

stimuli in an equivalence class allowed the The A stimuli were faces expressing emotions,
development of a line of investigation to as in Experiment 1, and the remaining stimuli
address quantitative dimensions of symbolic were abstract. One of the groups was trained
relational learning, which matching to sample with simultaneous matching and the other
procedures would not suffice to evaluate. with 2-s delayed matching. Then, participants
Bortoloti and de Rose (2009) conducted who formed classes evaluated the D or F
two experiments studying parameters that stimuli with the semantic differential. Half of
could influence the relatedness of equiva- the participants of each group evaluated the
lent stimuli. The first experiment compared D stimuli (one node from the faces) and the
equivalence classes generated by simultane- other half evaluated the F stimuli (three nodes
ous matching (for one group of participants) from the faces). The D stimuli were evalu-
and by 2-s delayed matching (for another ated similarly to the faces, as in Experiment
group). Participants were trained on AB, AC, 1, and evaluations from the delayed group
and CD conditional relations, in which A1, were again more similar to evaluations of the
A2, and A3 were pictures of faces expressing faces. However, evaluations of the F stimuli
happiness, anger, and neutrality, respectively. were not similar to the faces, for both groups.
Formation of equivalence classes was assessed These results show that transfer of meaning
by BD and DB tests, which did not involve from the faces to the abstract stimuli is an
the presentation of the faces. Participants inverse function of nodal distance, and are
who formed equivalence classes evaluated compatible with the claim of Fields and col-
the stimuli D (one node from the faces) with leagues, that relatedness of equivalent stimuli
the semantic differential and evaluations were decreases with nodal distance.
compared with evaluations of the faces by An advantage of the semantic differential
a control group. The control group yielded was to provide a quantitative measure of
positive evaluations of the happy faces and transfer of semantic functions from the faces
negative evaluations of the angry faces and to their putative symbols, i.e., the arbitrary
the experimental group yielded positive evalu- stimuli equivalent to them. A reanalysis of
ations of the arbitrary D stimulus equivalent the data from Experiment 2 of Bortoloti and
to the happy face and negative evaluations de Rose (2009) shows the mean deviation
of the arbitrary D stimulus equivalent to between evaluations of the abstract stimuli
the angry face. However, evaluations by the and those of the faces equivalent to them.
delayed group were more similar to the faces This deviation was calculated by averaging
than those by the simultaneous group. This the absolute values of differences between
showed that transfer of meaning from the the evaluation of a face and the evaluation
faces to the abstract stimuli was stronger of a stimulus equivalent to it, for each of the
when the classes were established with de- 13 scales of the semantic differential used by
layed matching. These results are compatible Bortoloti and de Rose (2009, Experiment 2).
with the hypothesis that delayed matching Thus, if the median evaluation of the happy
yields equivalence classes in which stimuli are face in a scale of the semantic differential was
more strongly related than are the stimuli in identical to the median of the evaluation of,
classes formed by training with simultaneous for instance, the D stimulus equivalent to
matching. Experiment 2 used the semantic the happy face, the deviation on that scale
differential to evaluate relatedness of stimuli in was 0. If the evaluations were different, one
larger equivalence classes that contained larger value was subtracted from the other and the
nodal distances. Two groups were trained in absolute value constituted the deviation in
conditional relations AB, AC, CD, DE, EF, that scale. The absolute values for all scales
and FG, and tested in emergent relations BG were averaged to yield the mean deviation
and GB to verify equivalence class formation. from the faces for their equivalent stimuli.
An “Orwellian” Account of Stimulus Equivalence 127

Deviations were calculated for the D and F ations when the delayed matching to sample
stimuli equivalent to the happy and angry training format was employed. It received
faces, for the delayed and simultaneous less deviant evaluations than the stimuli that
groups. The degree of relatedness is inversely were at three nodes from the happy faces
related to the deviation from the faces. There- when both the delayed (t=4.518, p<0.0001)
fore, the smaller the deviation, the larger is and the simultaneous (t=3.373, p<0.001)
the degree of relatedness between the faces matching to sample format were employed.
and the stimuli equivalent to them. Figure 1 It also received less deviant evaluations than
shows the mean deviation between evaluations the stimuli that were at both one and three
of the abstract stimuli and those of the faces nodes from the angry faces when both the
equivalent to them. delayed and the simultaneous matching
In general terms, Figure 1 seems to con- to sample format were employed (1-node
firm that transfer of meaning is an inverse DMTS: t=1.990, p<0.05; 1-node SMTS:
function of nodal distance and is larger with t=1.887, p<0.05; 3-nodes DMTS: t=3.406,
delayed matching than with simultaneous p<0.001; 3-nodes SMTS: t=3.410, p<0.001).
matching. For stimuli with a distance of three It also received less deviant evaluations than
nodes from the faces, evaluations are similar to the stimulus that was at one node from the
the ones yielded by the control group, which happy faces when the simultaneous matching
had no matching to sample training and could to sample format was employed (t=2.020;
not form equivalence relations between arbi- p<0.05). The stimulus that was at one node
trary stimuli and the faces. This indicates that from the happy faces in the simultaneous
accurate performance in equivalence tests do matching to sample training format received
not completely predict transfer of functions less deviant evaluations than the stimuli that
or, in other words, transfer of meaning. were at three nodes from both the happy and
It is interesting to note that evaluations the angry faces in both the simultaneous and
with the semantic differential showed less the delayed matching to sample training
deviation for the stimuli equivalent to the formats (happy, 3-nodes, SMTS: t=1.608,
happy faces than to the angry faces when p<0.05; happy, 3-nodes DMTS: 3.137, t=,
these stimuli were one node from the faces. p<0.05; angry, 3-nodes SMTS: t=2.182,
The stimulus that was at one node from the p<0.05; angry, 3-nodes DMTS: t=2.363,
happy faces received the least deviant evalu- p<0.001). The stimulus that was one node
distant from the angry faces received less
 
Happy Angry deviant evaluations than the stimulus at
3,5
DMTS SMTS DMTS SMTS three nodes from the angry faces only when
the delayed matching to sample format was
3
employed (t=1.906, p<0.05); when the si-
2,5
multaneous matching to sample format was
2 employed, the difference between the levels
1,5
of deviant evaluations was not statistically
significant (t=1.042, p=0.1538). Table 1
summarizes the levels of statistical signifi-
1

0,5
cance for the differences in levels of deviation
0 that can be seen in Figure 1.
1 3 1 3 1 3 1 3 In addition to prove that meaning
Nodal separation can transfer differentially within equiva-
Figure 1. Mean deviation between evaluations lence classes, these studies showed that
of the abstract stimuli and those of the faces transfer of meaning can be influenced
equivalent to them. by parameters other than nodal distance.
128 Renato Bortoloti and Julio C. de Rose

Table 1. Statistical significance for the differences between the levels of deviations from the evaluations of
the emotional faces comparing pair-wise of experimental conditions.
Happy Happy Angry Angry Happy Happy Angry Angry
1 node 1 node 1 node 1 node 3 nodes 3 nodes 3 nodes 3 nodes
DMTS SMTS DMTS SMTS DMTS SMTS DMTS SMTS
Happy
1 node - * * * *** ** ** **
DMTS
Happy
1 node - ns ns * * ** *
SMTS
Angry
1 node - ns * ns * *
DMTS
Angry
1 node - * ns ns ns
SMTS
Happy
3 nodes - ns ns ns
DMTS
Happy
3 nodes - ns ns
SMTS
Angry
3 nodes - ns
DMTS
Angry
3 nodes -
SMTS
* = p<0.5; ** = p<0.01; *** = p<0.001; ns = p>0.5

Use of the semantic differential to assess The IRAP, Implicit Relational Assessment
transfer of meaning between equivalent Protocol (Barnes-Holmes, Barnes-Holmes,
stimuli allowed, in turn, to extend the vali- Hayden, Milne, Power & Stewart, 2006),
dation of stimulus equivalence as a model involves simultaneous presentation, on a
of meaning, confirming that the formation computer screen, of an attributive concept,
of equivalence implies sharing of meaning, such as “pleasant” or “unpleasant”, a target
with a measure of meaning external to the stimulus and two relational terms such as
equivalence paradigm2. The semantic dif- “similar” and “opposite.” Several target
ferential also allows an assessment of the de- stimuli may appear along different trials,
gree of semantic similarity between stimuli, such as, for instance, “flower”, “insect”, “per-
thus enabling a quantitative assessment of fume”, “vomit”, “joy”, “panic”, etc. On each
relatedness. trial the participant is required to respond,
Bortoloti and de Rose (in press) used as fast as possible, pressing a key relating the
yet another method, the IRAP, to con- target and attributive stimuli in a given way.
firm that equivalent stimuli are semanti- In a study involving, for instance, “perfume”
cally related and that they may differ and “vomit” as targets, and “pleasant” and
in the degree of transfer of meaning. “unpleasant” as attributive stimuli, and “simi-
2
Recent research with electrophysiological methods has lar” and “opposite” as relational terms, the
also provided independent evidence that member of equiva- participant is required, in a consistent block of
lence classes share meaning (e.g., Barnes-Holmes et al., 2005;
Haimson, et al., 2009). trials, to respond “similar” for “perfume” and
An “Orwellian” Account of Stimulus Equivalence 129

“pleasant” and for “vomit” and “unpleasant”, might be taken as another indication that
and opposite for “perfume” and “unpleas- stimuli in equivalence classes may differ in
ant” and for “vomit” and “pleasant”. In an their degree of relatedness. Non-sense words
inconsistent block, on the other hand, the par- and faces, when used as attributive and target
ticipant is required to respond “similar” for stimuli, respectively, yielded an IRAP effect
“perfume” and “unpleasant” and for “vomit” when classes had been established by delayed
and “pleasant”, and “opposite” for “perfume” matching. When classes were established
and “pleasant” and for “vomit” and “unpleas- with simultaneous matching, an IRAP effect
ant”. It is assumed that consistent relations was obtained only for the happy class.
are more probable and, therefore, response Besides the effect of simultaneous vs.
latencies in consistent trials will be shorter delayed matching, this study also showed a
than in inconsistent trials. A shorter latency difference between faces. For happy faces,
in consistent trials, compared to inconsistent an IRAP effect was obtained both with
trials, is called an IRAP effect. delayed and simultaneous matching. For
Bortoloti and de Rose (in press) taught angry faces, an IRAP effect was obtained only
college students to establish equivalence with delayed matching. This may indicate,
classes, each comprising pictures of faces as suggested by Bortoloti and de Rose (in
expressing happiness (for one class) or anger press), that certain meaningful stimuli may
(for another class), and nonsense words. One be more “relatable” than others to abstract
group was taught with simultaneous match- stimuli. Arntzen and Lian (2010) have
ing and the other with 2-s delayed matching. shown that equivalence class formation is
Participants who demonstrated formation more likely when training includes arbitrary
of the programmed equivalence classes were and meaningful stimuli, in comparison with
then submitted to IRAP tests, in which each training only with arbitrary stimuli. The pres-
trial displayed a nonsense word as attributive ent results suggest that meaningful stimuli
stimulus and a picture of a face expressing ei- may differ quantitatively in the strength of
ther happiness or anger, as target. Relational relations they form with arbitrary stimuli.
stimuli were V (initial letter of the Portuguese Further studies should investigate this inter-
word “verdadeiro”, which means “true”) esting possibility.
and F (for the Portuguese word “falso”, that
means “false”). Consistent blocks required Concluding Remarks
that the participant choose V for target and
attributive stimulus of the same class, and F Theoretical Issues- The semantic differ-
for attributive stimulus and target from dif- ential allowed a more continuous assessment
ferent classes. In inconsistent blocks, these re- of transfer of meaning, providing another
lations were reversed. When all latencies were demonstration that transfer among members
considered, participants of the delayed group of an equivalence class is differential. Transfer
showed an IRAP effect, whereas participants decreases with nodal distance and increases
of the simultaneous group did not. When with delayed matching training. This differ-
latencies were considered separately for trials ential transfer is compatible with the notion
with angry and happy faces, participants of that equivalent stimuli may differ in their
the delayed group showed an IRAP effect for degree of relatedness, as claimed by Fields
both. Participants of the simultaneous group, and colleagues (e.g., Belanich & Fields, 2003;
on the other hand, showed an IRAP effect for Fields et al., 1995; Fields & Watanabe-Rose,
the happy faces, but not for the angry faces. 2008). Fields and colleagues argued that re-
This study provides additional support for latedness is a function of nodal distance. The
the notion that delayed matching produces studies of Bortoloti and de Rose confirmed,
stronger transfer of meaning and, therefore, with different methods and perhaps more
130 Renato Bortoloti and Julio C. de Rose

sensitive measures, that transfer of mean- arbitrary stimuli as a function of conditional


ing within equivalence classes is an inverse discrimination training. Equivalence tests
function of nodal distance, and showed that showed that members of each class were
parameters other than nodal distance can also substitutable from each other in emergent
influence transfer of meaning. conditional discriminations. When, in a
At first sight, data with the semantic dif- different condition, participants assigned
ferential and the IRAP (Bortoloti & de Rose, values on semantic differential scales, the
2009, in press) might be taken as a confirma- values assigned were a function of at least
tion that equivalent stimuli are not equally two classes to which the stimuli belonged: the
related to each other. This “Orwellian” view class comprising a face and the equivalent ar-
of equivalence is, obviously, a contradiction bitrary stimuli, and different classes based on
in terms, as pointed out by Sidman (1994), nodality (probably due to the order in which
since it is not mathematically possible for conditional discriminations were trained).
some stimuli to be “more equivalent” than Another explanation for the paradox of
others. “Orwellian equivalence” would be that the
We would like to point out some possi- mathematical equivalence between stimuli
bilities of solution to this paradox, although is imposed by the matching to sample pro-
we do not intend to exhaust the possibilities cedure, in which participants make forced
of interpretation and much less to offer a choices between discrete alternatives. Accord-
definitive solution. Sidman (1994, 2000) ing to this view, transfer of meaning would
has argued that behavioral equivalence is be a continuous dimension. For instance, the
an instance of mathematical equivalence, word in a person’s first language would be
generated by reinforcement contingencies. very strongly related to a particular meaning.
According to this view, it does not make When learning a second language, the same
sense to postulate degrees of relatedness. meaning (approximately) might transfer to a
On the other hand, Sidman points out that word in the second language, but the relation
stimuli are equivalent only with respect to would not be, at least in the initial stages of
the property that defines the class. They may learning, as strong as that with the word in
differ in other properties and may belong to the first language. However, matching to
different equivalence classes. Thus, whenever sample trials between words and pictures
individuals behave differentially regarding could not detect these differences, because
members of equivalence classes, this might of the all or none nature of the selections in
depend on other classes to which the stimuli matching trials. Based on the matching to
belong, or on properties of the stimuli other sample trials, we would conclude that words
than the one that defines the class. Sidman in the second language are equivalent to the
(1994) suggested that stimuli that comprise pictures. This would capture an important
an equivalence class may also belong to other property of the second language words: the
classes on the basis of nodality, particularly pictures and the words are related by mean-
because nodality is often related to the order ing. Stimulus equivalence, assessed by match-
of training. Differential transfer, according to ing to sample tests, would therefore capture
Sidman, would reflect the fact that stimuli an important aspect of meaning, but would
that are equivalent as a result of a particular not give a complete picture. A quantitative
conditional discrimination training, may dimension would have been left out.
belong to different classes on the basis of Applied Issues and Directions for Future
nodality. Research – Whatever the solution for the
This explanation is compatible with the puzzle of Orwellian equivalence will be, the
results of Bortoloti and de Rose (2009). Each example of learning a second language points
face was found to be equivalent to a set of to an important applied consideration.
An “Orwellian” Account of Stimulus Equivalence 131

Applications of the stimulus equivalence Another example: Bortoloti and de Rose


paradigm to education and rehabilitation are (2009, in press) showed that a matching to
increasing (e.g., Almeida-Verdu et al., 2008; sample training with a 2-s delay between the
Rehfeldt, 2011; Rehfeldt & Barnes-Holmes, removal of the sample and the presentation of
2009). Whenever the stimulus equivalence the comparison stimuli produces equivalence
paradigm is applied to establish or develop classes with higher transfer of functions than
symbolic repertoires, the question of transfer a simultaneous matching to sample training.
of meaning may be relevant. In applying It seems important that effects of parametric
stimulus equivalence to teach, for instance, variations of the delay’s length on the related-
a second language, it may be important to ness of equivalent stimuli are the object of
consider not only the fact that a word has ac- investigation in future studies.
quired some meaning, but also how much the Besides influencing the transfer of func-
meaning transferred to the particular word. tions of equivalent stimuli, the delay is a
Researchers, instructors, and program devel- parameter that also affects the formation
opers should consider the fact that delayed of equivalence classes, as shown by Arntzen
matching produces more transfer of mean- (2006),Vaidya and Smith (2006) and Bor-
ing than simultaneous matching. The actual toloti and de Rose (2009, Experiment 2).
decision about using simultaneous or delayed It is surely important to investigate if other
matching would take other questions into parameters that affect the establishment of
consideration as well, since delayed matching equivalence relations may also influence the
may yield more transfer of meaning, but may level of transfer of functions established be-
require more extended training. tween the related stimuli. A parameter that
Orwellian equivalence may be a theoreti- affects the formation of equivalence classes is
cal puzzle, but it may also open important the structure of training that is used. Three
avenues for future research. For instance, types of structure have been compared. For
Bortoloti and de Rose (2009, in press) found training classes with, for instance, four mem-
evidences, with both the semantic differential bers, the baseline of conditional discrimina-
and the IRAP, that the transfer of functions tions can be trained in a linear arrangement
in equivalence classes involving happy faces (AB, BC, CD), in an arrangement with a
is higher than in equivalence classes involving single node constituted by samples, called one
angry faces. This result is apparently consistent to many, or SAN, sample as node (AB, AC,
with studies that describe faster and more AD), and an arrangement with a single node
intense responses to happy faces (Batty & Tay- consisting of comparison stimuli, called many
lor, 2003; Kirita & Endo, 1995, Leppänen & to one, or CAN, comparison as the node (BA,
Hietanen, 2004), but it is contradictory with CA, DA). An experimental investigation
studies that describe faster responses to nega- could be conducted to assess the effects of
tive expressions (Fox, Lester, Russo, Bowles, these three types of arrangement on transfer
Pichler, & Dutton, 2000, Hansen & Hansen, of meaning between equivalent stimuli. In
1988; Öhman, Lundqvist, & Esteves, 2001). this experiment, stimuli A1, A2 and A3
It is necessary, therefore, further investigations could be faces expressing anger, happiness
focused on the effects of emotional faces on and neutrality, and different groups of par-
the relatedness of equivalent stimuli. A sys- ticipants could be submitted to arrangements
tematic replication of these studies by varying of training type one to many (AB, BC, CD,
the type of the emotional stimuli would also CE), many to one (AB, CB, DB, EB) and
be interesting. The stimuli could be significant linear (AB, BC, CD, DE). Participants in
words such as love, hate, freedom, oppression, each group would then evaluate the stimuli
abuse, pleasure, etc., and photographs and C, D and E with the semantic differential.
videos depicting various kinds of scenes. Many other investigations involving variables
132 Renato Bortoloti and Julio C. de Rose

such as directionality, overtraining, class size, vane, W. J (2008). Relational learning in


and so on can be conduct using the same children with deafness and cochlear im-
procedural logic presented in this article. plants. Journal of the Experimental Analysis
Probably a more direct way to detect of Behavior, 89, 407-424. doi:10.1901/
variations in transfer of meaning can be jeab.2008-89-407
based on event-related potentials (ERPs), Arntzen, E. (2006). Delayed matching-to-
that are small changes in the brain’s electri- sample: Probability of stimulus equiva-
cal activity recorded from the scalp, evoked lence as a function of delays between
by some external or internal events. The sample and comparison stimuli during
N400, particularly, is a negative component training. The Psychological Record, 56,
that peaks about 400 ms after the presenta- 135–167.
tion of a stimulus and reflects the degree of Arntzen, E., Grondahl, T., & Eilifsen, C.
incompatibility between this stimulus and a (2010). The effects of different train-
semantic context previously established. It ing structures in the establishment of
has been shown that the N400 is modulated conditional discriminations and subse-
significantly by the formation of equivalence quent performance on tests for stimulus
classes (Barnes-Holmes et al., 2005; Haim- equivalence. The Psychological Record, 60,
son et al., 2009). Haimson et al. (2009) first 437-462.
replicated the N400 component obtained Arntzen, E., & Lian, T. (2010). Trained and
with pairs of common English words. In derived relations with pictures versus ab-
a second experiment, they taught relations stract stimuli as nodes. The Psychological
AB, AC, AD, AE, and AF in order to gen- Record, 60, 659-678.
erate equivalence classes between arbitrary Barnes-Holmes, D., Barnes-Holmes, Y.,
stimuli and found that the presentation of Power, P., Hayden, E., Milne, R., & Stew-
pairs of stimuli such as BC, DE, FD, etc., in art, I. (2006). Do you really know what
which the first stimulus simulated a semantic you believe? Developing the Implicit
context for the second one, evoked a large Relational Assessment Procedure (IRAP)
N400 component only when these stimuli as a direct measure of implicit beliefs. The
did not belong to the same equivalence class Irish Psychologist, 32, 169-177.
(e.g., pairs as B1C2, D2E3, F3D1, etc.). An Barnes-Holmes, D., Keane, J., Barnes-
important direction for future research on Holmes, Y., & Smeets, P. M. (2000). A
possible differences in relatedness of equiva- derived transfer of emotive functions as
lent stimuli, therefore, would be through a means of establishing differential pref-
electrophysiological measures, particularly erences for soft drinks. The Psychological
the N400 component. If this component Record, 50, 493-511.
proves sensitive to parameters such as nodal Barnes-Holmes, D., Staunton, C., Whelan,
distance and delayed matching, conclusions R., Barnes-Holmes, Y., Commins, S.,
about differential transfer of meaning within Walsh, D., Stewart, I., Smeets, P. M., &
equivalence classes would receive further sup- Dymond, S. (2005). Derived stimulus
port, and this may help to clarify the theo- relations, semantic priming, and event-
retical questions about relatedness between related potentials: testing a behavioral
equivalent stimuli. theory of semantic networks. Journal of
the Experimental Analysis of Behavior, 84,
References 417-433. doi:10.1901/jeab.2005.78-04
Batty, M., & Taylor, M. J. (2003). Early
Almeida-Verdu, A. C., Huziwara, E. M., de processing of the six basic facial emo-
Souza, D. G., de Rose, J. C., Bevilacqua, tional expressions. Cognitive Brain
M. C., Lopes Jr, J., Alves, C. O., & McIl- Research, 17, 613-620. doi:10.1016/
An “Orwellian” Account of Stimulus Equivalence 133

S0926-6410(03)00174-5 verbais. São Paulo: Editora Ática.


Belanich, J., & Fields, L. (2003). Generalized Fields, L., Adams, B. J., Verhave, T., & New-
equivalence classes as response transfer man, S. (1993). Are stimuli in equiva-
networks. The Psychological Record, 53, lence classes equally related to each other?
373–413. The Psychological Record, 43, 85-105.
Bortoloti, R., & de Rose, J. C. (2007). Me- Fields, L., Landon-Jimenez, D. V., Buff-
dida do grau de relacionamento entre es- ington, D. M., & Adams, B. J. (1995).
tímulos equivalentes. Psicologia: Reflexão Maintained nodal-distance effects in
e Crítica, 20, 250-256. doi:10.1590/ equivalence classes. Journal of the Experi-
S0102-79722007000200011 mental Analysis of Behavior, 64, 129-145.
Bortoloti, R., & de Rose, J. C. (2009). As- doi:10.1901/jeab.1995.64-129
sessment of the relatedness of equivalent Fields, L. & Watanabe-Rose, M. (2008).
stimuli through a semantic differential. Nodal structure and the partitioning of
The Psychological Record, 59, 563-590. equivalence classes. Journal of the Experi-
Bortoloti, R. & de Rose, J. C. (in press). mental Analysis of Behavior, 89, 359-381.
Equivalent Stimuli Are More Strongly doi:10.1901/jeab.2008-89-359
Related after Training with Delayed than Fox, E., Lester, V., Russo, R., Bowles, R.
with Simultaneous Matching: A Study J., Pichler, A., & Dutton, K. (2000).
Using the Implicit Relational Assess- Facial expressions of emotion: Are
ment Procedure (IRAP). The Psychological angry faces detected more efficiently?
Record. Cognition and Emotion, 14, 61-92.
Bush, K. M., Sidman, M., & de Rose, T. doi:10.1080/026999300378996
(1989). Contextual control of emergent Haimson, B., Wilkinson, K.M., Rosenquist,
equivalence relations. Journal of the Ex- C., Ouimet, C. & McIlvane, W.J. (2009).
perimental Analysis of Behavior, 51, 29-45. Electrophysiological correlates of stimu-
doi:10.1901/jeab.1989.51-29 lus equivalence processes. Journal of the
Carrigan, P. F., Jr., & Sidman, M. (1992). Experimental Analysis of Behavior, 92,
Conditional discrimination and equiva- 245-256. doi:10.1901/jeab.2009.92-245
lence relations: A theoretical analysis of Hansen, C. H., & Hansen, R. H. (1988).
control by negative stimuli. Journal of Finding the face in the crowd: An anger
the Experimental Analysis of Behavior, 58, superiority effect. Journal of Personal-
459-504. doi:10.1901/jeab.1992.58-183 ity and Social Psychology, 54, 917-924.
de Rose, J. C., McIlvane, W. J., Dube, W. doi:10.1037//0022-3514.54.6.917
V., Galpin, V. C., & Stoddard, L. T. Hayes, S. C., Kohlenberg, B. S., & Hayes,
(1988). Emergent simple discrimina- L. J. (1991). The transfer of contextual
tions established by indirect relations to control over equivalence classes through
differential consequences. Journal of the equivalence classes: A possible model of
Experimental Analysis of Behavior, 50, social stereotyping. Journal of the Experi-
1-20. doi:10.1901/jeab.1988.50-1 mental Analysis of Behavior, 56, 505-518.
Dougher, M., Augustson, E., Markham, M., doi:10.1901/jeab.1991.56-505
Greenway, D., & Wulfert, E. (1994). Imam, A. (2006). Experimental control
The transfer of respondent eliciting and of nodality via equal presentation of
extinction functions through stimulus conditional discriminations in different
equivalence classes. Journal of the Experi- equivalence protocols under speed and
mental Analysis of Behavior, 62, 331-351. no speed conditions. Journal of the Experi-
doi:10.1901/jeab.1994.62-331 mental Analysis of Behavior, 85, 107-124.
Engelmamn, A. (1978). Os estados subjetivos, doi:10.1901/jeab.2006.58-04
uma tentativa de classificação de seus relatos Johnson, C., & Sidman, M. (1993). Con-
134 Renato Bortoloti and Julio C. de Rose

ditional discriminations and equivalence Journal of Applied Behavior Analysis, 44,


relations: Control by negative stimuli. 109-119. doi: 10.1901/jaba.2011.44-109
Journal of the Experimental Analysis of Rehfeldt, R. A., & Barnes-Holmes, Y.
Behavior, 59, 333-347. doi:10.1901/ (2009). Derived relational responding:
jeab.1993.59-333 Applications for learners with autism and
Kennedy, C. H. (1991). Equivalence class other developmental disabilities. Oakland,
formation influenced by the number of CA: New Harbinger Publications, Inc.
nodes separating stimuli. Behavioral Pro- Sidman, M. (1986). Functional analysis of
cesses, 24, 219-245. doi:10.1016/0376- emergent verbal classes. In T. Thompson
6357(91)90077-D & M. D. Zeiler (Orgs.), Analysis and
Kirita, T., & Endo, M. (1995). Happy-face integration of behavioral units. Hillsdale,
advantage in recognizing facial expres- NJ: Lawrence Erlbaum.
sions. Acta Psychologica, 89, 149-163. Sidman, M. (1994). Equivalence Relations
doi:10.1016/0001-6918(94)00021-8 and Behavior: A Research Story. Boston:
Leppänen, J. M., Kauppinen, P., Peltola, M. Authors Cooperative.
J., & Hietanen, J. K. (2007). Differential Sidman, M. (2000). Equivalence relations
electrocortical responses to increasing and the reinforcement contingency.
intensities of fearful and happy emo- Journal of the Experimental Analysis of
tional expressions. Brain Research, 1166, Behavior, 74, 127-146. doi:10.1901/
103-109. jeab.2000.74-127
Öhman, A., Lundqvist, D., & Esteves, F. Sidman, M., & Tailby, W. (1982). Condi-
(2001). The face in the crowd revisited: A tional discrimination vs. matching-to-
threat advantage with schematic stimuli. sample: An Expansion of the testing para-
Journal of Personality and Social Psychol- digm. Journal of the Experimental Analysis
ogy, 80, 381-396. doi:10.1037//0022- of Behavior, 37, 261-273. doi:10.1901/
3514.80.3.381 jeab.1982.37-5
Osgood, C. E., Suci, G. I., & Tannenbaum, Vaidya, M., & Smith, K. N. (2006). Delayed
P. H. (1957). The measurement of mean- matching-to-sample training facilitates
ing. Urbana, Illinois: University of Illinois derived relational responding. Experimen-
Press. tal Analysis of Human Behavior Bulletin,
Pereira, C. A. A. (1986). O diferencial semân- 24, 9–16.
tico: uma técnica de medida nas ciências hu- Wulfert, E., & Hayes, S. C. (1988). Trans-
manas e sociais. São Paulo: Editora Ática. fer of a conditional ordering response
Rehfeldt, R. A. (2011). Toward a technology through conditional equivalence classes.
of derived stimulus relations: An analysis Journal of the Experimental Analysis of
of articles published in the Journal of Behavior, 50, 125–144. doi:10.1901/
Applied Behavior Analysis, 1992-2009. jeab.1988.50-125.

You might also like