Review completion
Trueblue
deCarvalhodeRose2014.pdf
ORIGINAL ARTICLE
Understanding Racial Attitudes through the Stimulus Equivalence Paradigm
Marilia Pinheiro de Carvalho & Julio C. de Rose
Published online: 18 April 2014 # Association of Behavior Analysis International 2014
Abstract The present study investigated racial attitudes among children by means of the equivalence class paradigm. Four children that, during a pre-test with a matching-to- sample task associated negative symbols with the faces of black men, participated in the study. During training, the children learned to associate positive symbols with an abstract symbol and then the abstract symbol with pictures of black men. At issue was whether, at the end of training, the positive symbols would be associated with the pictures of black men, reversing the behavioral relations revealed during the pre-test. Results from a post-test showed that only one child reversed the pre-test relations. For the other three children, the pre-test relations persisted during the post-test. These results may express the difficulty of changing the behavioral functions of socially loaded stimuli.
Keywords Racialattitudes .Stimulusequivalence .Matching to sample . Children
Since the 1970s, various studies have shown that, using a standard matching to sample procedure, teaching relations AB and BC yields the emergence of the untrained relations AC, CA, BA and CB. These emergent relations document the properties of symmetry (the trained relation is functionally reversible so that from AB, one gets BA), and transitivity (from the training of two pairings with a common node, such as AB and BC, there emerges a third relation, AC, between the stimuli related by the common node; Sidman
and Tailby 1982). Another property, reflexivity, documented by relations between each stimulus and itself, is sometimes also verified, and more frequently inferred. These three prop- erties document the formation of an equivalence class among stimuli, A, B, and C (Sidman 1994).
The emergence of untrained, novel relations has made the equivalence class paradigm a useful tool to investigate sym- bolic processes. This paradigm allows the experimental estab- lishment of symbolic relations and, therefore, facilitates their experimental investigation. This is usually done with arbitrary stimuli, so that the investigation can be presumably uncontam- inated by previous meanings of the stimuli. The equivalence class paradigm is also a means to conceptualize and investigate the origins of new behavior. For these reasons, the equivalence class paradigm has extended the Experimental Analysis of Behavior into complex domains of human cognition.
One of these domains is attitude formation. In Social Psy- chology (e.g., Gazzaniga and Heatherton 2002) attitudes are defined as relatively stable dispositions to think, feel and act in a particular way toward specific objects, events, or stimuli. According to this viewpoint, the acquisition of an attitude does not depend primarily on the direct experiences of the individ- ual with its object, but on the social practices of his/her com- munity (Rodrigues et al. 2003). As the individual grows in a society, the verbal community shapes and models the actions, feelings, and thoughts that define attitudes (e.g., Guerin 1994).
Other researchers (e.g., Grey and Barnes 1996) have sug- gested that equivalence class formation may lead to attitude acquisition. Watt et al. (1991) give the following example, applicable to the context of Northern Ireland: “If a person is confronted for the very first time with an individual called Sean [A], and has lived in a community where the name Sean is positively related with a Catholic context [B] and Catholic is positively related with terrorism [C], then the behavioral prin- ciple of stimulus equivalence makes it likely that Sean [A] and terrorism [C] are seen as equivalent stimuli” (p. 35).
M. P. de Carvalho: J. C. de Rose Universidade Federal de São Carlos, São Carlos, Brazil
M. P. de Carvalho (*) School of Psychology, Animal Learning and Behavior Lab, University of Minho, Campus de Gualtar, 4710-507 Braga, Portugal e-mail: [email protected]
Psychol Rec (2014) 64:527–536 DOI 10.1007/s40732-014-0049-4
Conceiving of attitudes as dispositions learned according to processes such as shaping, modeling and equivalence class formation has raised the possibility of using these same pro- cesses to create specific attitudes in experimental settings. Several studies have been carried out to investigate equiva- lence class as a model of attitude formation (e.g., Barnes et al. 1996; Dixon et al. 2006; Egli et al. 1997; Grey and Barnes 1996; Moxon et al. 1993). To illustrate, Grey and Barnes (1996, Experiment 1) have shown formation of attitudes to- ward movies based on the acquisition of equivalence classes. First, a matching-to-sample procedure established two classes comprising three nonsense syllables each (A1B1C1 and A2B2C2). Next, labels with the nonsense syllables B1 and B2 were attached to the covers of two movies, one with sexual contents, and the other with religious contents. After watching these two movies, participants were required to evaluate as either bad or good six movies, the two already watched (‘old’) and four never-watched movies (‘new’). The new movies had on their covers one of the nonsense syllables, A1, A2, C1, or C2. The participants evaluated each of the new movies in the same way as they had evaluated the old movie belonging to the same putative class (e.g., the movies with A1 and C1 were evaluated as the old movie with label B1). These results suggest that the evaluative aspect of attitudes may be acquired by equivalence class formation.
If the process of stimulus equivalence can contribute to the formation of attitudes, then the process of stimulus equiva- lence reversion may also contribute to the modification of attitudes. In a typical stimulus equivalence reversal procedure, the participants learn a set of conditional discriminations and thereby form equivalence classes. Then, one or more condi- tional discriminations are reversed. Finally, equivalence tests are conducted to evaluate the reversion of the old classes and the formation of the new classes (e.g., Garotti et al. 2000; Pilgrim and Galizio 1995).
The initial studies on the reversal of equivalence classes showed some difficulties in reversing the classes established during baseline (Pilgrim et al. 1995; Pilgrim and Galizio 1995, 1990). However, subsequent studies yielded more positive results (Garotti et al. 2000; Garotti and de Rose 2007; Smeets et al. 2003).
Many of the foregoing studies used abstract stimuli devoid of pre-experimental history and meaning. In the study of Watt et al. (1991), on the other hand, socially loaded stimuli were used to try to establish a new equivalence class from the reversal of two presumed classes relating Catholic and Prot- estant symbols in Irish and English participants from Northern Ireland. The participants were selected because in Ireland events related to Catholicism and Protestantism tend to be contrasted and the religious history and culture of individuals are easily identifiable. The procedure employed a simulta- neous matching-to-sample task to teach first the relation be- tween Catholic names (A) and nonsense syllables (B) and,
second, the relation between the nonsense syllables (B) and Protestant symbols (C). The objective was to establish an equivalence class comprising a Catholic name and a Protestant symbol (A-C) and, thus, reverse the two presumed original classes, one relating Catholic names and symbols, and the other relating Protestant names and symbols.
Results revealed that none of the six Irish Protestant par- ticipants selected Protestant symbols conditionally on Catho- lic names as the trained relations predicted. Similarly, five out of twelve Irish Catholic participants also did not respond according to the trained relations. In contrast, all five English Protestants responded according to the trained relations. Watt et al. (1991) argued that most Irish Protestant and Catholic participants did not acquire the new equivalence class because the relations trained during the experiment competed with the religious relations established previously. According to the authors, contextual cues could have controlled responding in the sense of causing pre-experimental response patterns to prevail in the experimental setting.
Watt et al.’s (1991) study reveals the interference of social- ly loaded stimuli and formerly acquired equivalence classes on the emergence of new equivalence relations. If, as the study suggests, socially loaded stimuli (e.g., Catholic and Protes- tant) form equivalence classes that are significantly harder to reverse than equivalence classes comprising less socially load- ed stimuli (e.g., movies), the use of the equivalence paradigm for reversing attitudes would be considerably more difficult. However, Dixon et al. (2006) showed that “programmed contingencies can to some degree override the preexisting relationships between stimuli.” These authors showed indica- tions that for American college students, Middle Eastern stimuli were related in “frames of coordination” with terrorist stimuli. However, an experimental training was partially suc- cessful in establishing classes comprising both American and Middle Eastern stimuli, overriding the pre-existing relations.
The present study was designed to examine the interference of pre-experimental relations regarding racial attitudes among children. As such, it extended the investigations of Watt et al. (1991) and Dixon et al. (2006), studying children (as opposed to adults) and racial (as opposed to religious) attitudes. Con- ditional discriminations involving positive symbols (A) and an abstract symbol (B) and the abstract symbol (B) and pictures of black1 men (C) were taught to children that showed a negative bias toward black men, to see if a relation between black men and positive symbols would emerge.
In order to maximize the chances of obtaining emergent responding consistent with training, we used delayed matching in the training, because previous research showed
1 We will use the terms “black” and “white” to refer to races because their Portuguese equivalents are the current terms in Brazil, where this research was conducted. These terms are used both by the individuals themselves and by official agencies, with no opprobrium.
528 Psychol Rec (2014) 64:527–536
that delayed matching increases the likelihood of class forma- tion and also increases the degree of “relatedness” between the stimuli (Bortoloti and de Rose 2007, 2009, 2012). Two of the participants were also overtrained in baseline conditional dis- criminations, with the same objective, based on data from our laboratory suggesting that overtraining also increases related- ness (Bortoloti et al. 2013).
Method
Participants
Three boys (7, 9, and 10 years old) and one girl (7 years old) revealed a negative bias towards black men in a screening pre- test and were selected for the study among 16 children. The remaining children revealed either a neutral or a positive attitude and were not included in the study. The participants were recruited either in their school, or in neighborhoods surrounding the University. In the latter case, an assistant of the experimenter would go door to door talking to parents and inviting children to collaborate. All the children attended a regular community school in the city of São Carlos, Brazil, and could read. The two 7-year old children were in first grade and the 9- and 10-year olds were in fourth grade. The younger boy was brown2 skinned, the girl was black skinned, and the two older boys were white skinned.
Materials
The matching-to-sample task was implemented in an Apple Macintosh computer. Stimulus presentations and response recording on each trial were controlled by the software MTS v.11.1.3 (Dube and Hiris 1997). The stimuli were taken from public collections at the website “google.com”. Since the participants had never had previous contact with either com- puters or internet websites, it was unlikely that they had previously seen the stimuli elsewhere. As Fig. 1 shows, the stimuli were five pictures of faces of black men, three pictures of faces of white men, three pictures of abstract symbols, four pictures of positive symbols (e.g., “thumb up”), and four pictures of negative symbols (e.g., “thumb down”). Each picture of a face showed a black or a white man looking straight at the observer with open eyes and closed lips and without any obvious emotional expression.
As Fig. 1 shows, stimuli were divided into three sets, A, B, and C. Set A contained positive (A1) and negative (A2) stimuli; set B contained abstract symbols (B1 and B2); and
set C contained faces of black men (C1), an abstract symbol (C2), and faces of white men (C3). For Participants M1 and F1, Version 1 of the stimuli was used. In this version, A1 could be either one of three different stimuli with a positive mean- ing, whereas A2 could be either one of three different stimuli with a negative meaning. The A stimuli were colored, as well as B1 and B2, which were abstract stimuli. Stimulus C2 was also an abstract colored stimulus. C1 were three black faces and C3 were three white faces. Along the pre-test, some of the dismissed children reported that they were choosing the stim- uli on the basis of irrelevant features. For example, a dismissed participant reported that she had chosen both the thumb up and the thumb down stimuli following white faces because the background of the thumbs was pink, and, pink goes better with white faces. On the basis of these reports, we decided to simplify the stimuli used and introduced the second version of stimuli for the forthcoming participants. Version 2 was used for Participants M2 and M3. The abstract stimuli (B1, B2, C2) were black and white line drawings; A1 and A2 were black and white schematic drawings of, respectively, thumbs up and thumbs down. In Version 2, only two pictures of black men were used for both training and testing and only two pictures of white men were used for testing. Version 2 had a reduced number of stimuli because we wanted to avoid inattentive and random responding that we observed in some of the non-selected children who complained about the length of the test session. Since the larger the number of stimuli, the larger the number of combinations of stimuli and, hence, the larger the number of trials in a session, we decided to simplify the procedure by reducing the number of stimuli. We hoped that making the session shorter would maximize the chances of learning the trained discriminations and, eventually, the likelihood of equivalence class formation. The pictures of men varied across children because they corresponded to the black faces that, for each child, yielded the most negative evaluations in a Semantic Differential (Osgood et al. 1957).
General Procedure
The experiment consisted of a matching-to-sample task divid- ed into three phases, pre-test, training, and post-test. Each phase was comprised blocks of trials, and the trials differed slightly in training and testing. In training, each trial began with the presentation of the sample stimulus in the center of the computer screen. Clicking the left button of the mouse over the sample stimulus removed that stimulus from the screen and started a 1.5-s delay. At the end of the delay, the two comparison stimuli were presented in two randomly selected corners of the screen. When the child clicked on the correct comparison stimulus, a large number of stars, with various colors, appeared on the screen, a “clapping sound” was played for 2.5 s, and one point was added to a counter displayed on the screen (continuous reinforcement). Then, the
2 We will use the term “brown” to refer to the color of the skin because its Portuguese equivalent (“pardo”) is the current term in Brazil. The term is used both by the individuals themselves and by official agencies, with no opprobrium.
Psychol Rec (2014) 64:527–536 529
next trial began. When the child clicked on the incorrect comparison stimulus, the screen was darkened for three sec- onds and the next trial followed.
Test trials differed from the training trials in two aspects. First, clicking on the sample stimulus produced the compari- son stimuli immediately and without removing the sample stimulus. And second, clicking any comparison stimulus end- ed the current trial without differential consequences and started the next trial.
The children were invited to play a game. They were instructed to click on the sample stimulus and then the comparison stimulus that “goes with” the sample stimulus. They were told that by clicking on the correct stimulus they would earn 1 point, and that the points earned at the end of the game could be used to “buy” one of the toys located in an adjacent cabinet. The children were reminded that better toys cost more points and instructed to try to earn as many points as possible.
For each child, the toys were divided into five price cate- gories based on informal conversations about which toys the child liked the most and the least. At the end of each training session, the child “bought” one of the toys his/her earned points could buy. Because children did not earn points in test trials, at the end of each testing session the child received one toy regardless of his or her performance. At the beginning of a test session, participants were told that there would be no points and that they would receive a toy simply for performing the task.
The screening pre-test evaluated whether the participant chose the negative or the positive symbols when either black or white faces were displayed as samples. That is, it assessed the relation C-A. Each trial comprised a sample C stimulus (black men, white men, or the abstract symbol) and a positive and a negative A symbol as comparisons. Each C stimulus was presented in random order three times per session and sessions continued until preferences appeared stable (from three to eight sessions across participants). Figure 2 shows examples of testing trials. Only children that revealed a neg- ative attitude towards black men, as compared to white men, were included in the study.
Training aimed to establish the conditional relations (AB and BC) from which an equivalence class comprising black men and positive symbols would presumably emerge (C-A). Initially, the children learned to pair the positive (A1) and negative (A2) stimuli with the abstract symbols B1 and B2, respectively, and then they learned to pair the same abstract symbols B1 and B2 with pictures of black men (C1) and the abstract symbol (C2), respectively. Faces of white men were not used during training because we were not interested either in increasing the participant’s tendency to relate these faces to positive symbols since this relation could compete with the establishment of the relation between black faces and positive symbols, or in promoting a relation between the white faces and the negative symbols. For each participant, the stimuli used in training were the same as those used in testing. Figure 3 shows AB and BC training trials.
Set A
C3
A1
C2
C1B1
2B2A
Set B Set C
Version 2Version 1Version 2Version 1
Fig. 1 The three stimulus sets used in the experiment. Version 1 used three elements from sets C1 and C3, and the left-hand side element of set C2. Version 2 used two elements from sets C1 and C3, and the right-hand side element from set C2
530 Psychol Rec (2014) 64:527–536
Relation AB Each trial displayed a sample, either A1 or A2, and the two comparison stimuli B1 and B2. Sessions were divided into blocks of trials and each block comprised three presentations of each A stimulus in a random order.
Training proceeded through a series of steps of increasing difficulty. Initially, a written prompt (the words “Choose this”) was presented above the correct comparison stimulus. Next, the prompt was removed for one block of trials and then a learning criterion was introduced: Training continued until 100 % correct responses (6/6) were made in one block. If the criterion was not met, the block was repeated, and if the criterion still was not met after two additional blocks, the session ended and training resumed in the next session, with- out prompt. In Version 2, two additional sessions without prompt followed.
Relation BC Each trial displayed a B sample and two comparison stimuli, C1 one of the faces of black men and C2 the abstract symbol. Each face was presented three times as a comparison stimulus following each B sample.
The other procedural details remained the same as in the AB training.
Relation AB/BC In this stage, the two AB relations and the two BC relations were intermixed within each block of trials. After the learning criterion was met on both relations, the participants were told that the computer would no longer show whether they were right or wrong, but they would continue to earn points for correct responses. Hence, they should continue to choose as before. This new stage, which intended to prepare the child for the post-test, continued until the learning criterion was met. In Version 2, two additional sessions followed.
To assess the effects of training on the initial relations between pictures of black men and negative symbols, the test conducted in pre-test was repeated as a post-test. This test verified whether the CA relation emerged from training and is, therefore, a combined assessment of symmetry and transitiv- ity. Positive results in this test provide evidence in favor of the formation of equivalence classes. Note, however, that this was
Fig. 2 Examples of C-A test trials. The sample stimulus was either a black face, or a white face, or an abstract symbol. The comparison stimuli were the positive and negative symbols
Fig. 3 Examples of AB (top) and BC (bottom) training trials. A square around one of the comparison stimuli identifies the correct choice, but no such square was present during the actual experiment
Psychol Rec (2014) 64:527–536 531
a modified CA test, because it also included the pictures of white faces (C3) not included in training.
Results
The children needed from five to 12 sessions to complete the training phase. Specifically, the two children exposed to Ver- sion 1 learned the AB relation in two and three sessions, the BC relation in one and two sessions and the combined AB/BC relations in two sessions for a total of five (participant F1) and seven (participant M1) sessions of training. The children ex- posed to Version 2, which included additional blocks of overtraining, took a few more sessions: three to learn AB, three to learn BC, and five and six to learn the AB/BC relations, for a total of 11 (participant M3) and 12 (participant M2).
Figure 4 shows the pre-test and post-test results. The aver- age percentage of choices of the negative comparison symbol is plotted for each sample stimulus from set C (black faces, C1, abstract symbol, C2, and white faces, C3; Table 1 in the Appendix gives the raw data).
As a consequence of learning the relation between a posi- tive symbol and an abstract symbol (A1-B1) and the relation between the abstract symbol and black faces (B1-C1), one would expect the emergence of the relation between black faces and the positive symbol (C1-A1). Hence, the relation between black faces (C1) and the negative symbol (A2) re- vealed in the pre-test (a condition to select the participants for the experiment) should be significantly attenuated or even reversed in the post-test. The expected pattern was
obtained only with participant M3. For the other partic- ipants, the strong relation between black faces and the negative symbol either did not change (M1 and F1) or was even strengthened (M2). A close analysis of the results for the three black faces (see Table 1 in the Appendix) supported a similar conclusion: only partici- pant M3 reversed the pre-test relations.
As a consequence of learning the relation between the negative symbol A2 and the abstract symbol B2, and the relation between the abstract symbol B2 and the abstract symbol C2, one would expect the emergence of the relation between the negative symbol A2 and the abstract symbol C2. Hence, the association between the abstract symbol C2 and the negative symbol A2 should increase from pre-test to post- test. The results were mixed. As before, participant M3 showed clearly the expected pattern and participant M1 showed a trend in the expected direction; participant F1 main- tained the strong association already revealed in the pre-test and participant M2 did not show the effect.
Taking into account the two sets of findings, one related to black faces, C1, and the other to the abstract symbol, C2, one can conclude that the results from participant M3 are consis- tent with the formation of the two classes A1B1C1 and A2B2C2. The results from participant M2 are inconsistent with the formation of either class. The results from participant M1 are consistent with the formation of class A2B2C2 but not with the formation of class A1B1C1. Finally, the results from participant F1 are inconsistent with the formation of class A1B1C1 (because her pre-test score following C2 was high, nothing can be stated concerning the formation of class A2B2C2).
0
20
40
60
80
100
Black Faces (C1) Abstract (C2) White Faces (C3)
0
20
40
60
80
100
Black Faces (C1) Abstract (C2) White Faces (C3)
Pre-test
Post-test
0
20
40
60
80
100
Black Faces (C1) Abstract (C2) White Faces (C3)
M1
M2
F1
0
20
40
60
80
100
Black Faces (C1) Abstract (C2) White Faces (C3)
P e
rc e
n ta
g e
o f
C h
o ic
e o
f N
e g
a ti v e
S y m
b o
l
M3
Fig. 4 Individual test results for the C-A relation. The percentage of choices of the negative symbol is plotted as a function of the sample stimulus during the pre-tests and post-tests. Table 1 in the Appendix shows the raw data
532 Psychol Rec (2014) 64:527–536
Consider now the choices given the white faces, C3. Be- cause these faces were not included in training, no specific predictions were made. From pre-test to post-test, the choice of negative symbols decreased for participants M2 and M3, did not change for participant F1, and increased for participant M1. The analysis for each white face (see Table 1 in the Appendix) showed that only participant M1 reversed its initial relations for two of the three faces. As mentioned before, white faces were not included in training because we did not want to relate them to negative symbols. Thus, the reversal of the relation between white faces and positive symbols dem- onstrated by participant M1 was not expected and somewhat surprising. So far, we are unable to explain this result, or to identify whether it is a spurious artifact.
Discussion
The present study attempted to investigate racial prejudice in children using the stimulus equivalence paradigm. It was an attempt to extend previous research on equivalence class formation with socially loaded stimuli (e.g., Dixon et al. 2006; Watt et al. 1991) to another socially important topic (racial attitudes) and to a different population (Brazilian children).
According to the equivalence paradigm, the establishment of a relation between positive symbols (A1) and an abstract symbol (B1), and the establishment of a relation between the abstract symbol (B1) and pictures of black men (C1), should result in a relation between the positive symbols (A1) and the pictures of black men (C1). Hence, if a participant showed initially a tendency to relate black men to negative symbols, the training should either attenuate or even reverse that rela- tion. However, based on the results of Watt et al. (1991), we might expect that pre-experimental equivalence classes com- prising black men and negative attributes would not be affect- ed by the experimental training.
The experiment followed a pre-test/post-test design. The pre-test selected participants that related pictures of black men to negative symbols. The training taught the participants two relations that should reverse the pre-test relation. The post-test assessed whether training had in fact changed the pre-test relations.
The procedure also included several features that should have facilitated the emergence of the two equivalence classes: the re-training of the AB and BC relations before the post-test (Garotti and de Rose 2007), the inclusion of a delay in the matching-to-sample task, and the use of only one node be- tween the trained stimuli (Bortoloti and de Rose 2009; Fields et al. 1995).
The results were mixed. On the positive side, one child showed the reversion, in the post-test, of the strong negative relation between black faces and negative symbols revealed
in the pre-test. Moreover, the weak association between an abstract symbol and the negative symbol revealed in the pre- test was strengthened during the post-test. These results are consistent with the formation of the two predicted equiva- lence classes. Interesting enough, the participant who re- vealed a reversal of the pre-test relation between black faces and negative symbols was one of the two participants who received overtraining in both AB and BC relations. Also, two of the participants who did not demonstrate reversal of the pre-test relations were not exposed to overtraining. Overtraining has been shown to strengthen equivalence re- lations (Bortoloti et al. 2013) and, therefore, further research is needed to establish whether it could contribute to the modification of equivalence classes with socially loaded stimuli. More generally, the results of the present study suggest the possibility that careful selection of training pa- rameters may increase the likelihood of reversing classes with socially loaded stimuli.
On the negative side, three children did not show the expected pattern of results. Although they also learned the new relations during training, the association between black faces and negative symbols was not reduced. Considering the results obtained with these three children, the present study not only replicates the procedure developed by Watt et al. (1991) with another population and another attitude, but it also replicates their results (i.e., in their study, the majority of Irish religious participants did not form new relations between the mutually opposite stimuli of Catholic and Protestant nouns). Also Moxon et al. (1993) did not succeed in forming an equivalence class between male/female names and gender stereotyped professions. However, if we consider that one child responded in the post-test consistently with the emergent relation between black faces and positive symbols, the study also confirms the conclusion of Dixon et al. (2006) that experimental contingencies can, to a certain extent, override pre-existing classes. These classes were overridden to a lesser extent in the present study as compared to the study of Dixon et al., but the present study selected only participants that showed strong negative bias toward black men. The fact that changes occurred even with very biased participants suggests that additional investigations may reveal parameters that can increase chances of successful change of attitude in the labo- ratory context. To what extent such a change would generalize to the social environment outside the laboratory would be another question for investigation.
Several hypotheses may help to explain the negative find- ings in the present study. One possibility is that, although learning the trained AB and BC relations, our participants did not form the symmetrical relations B-A and C-B, as well as the transitive relation A-C, which relations are prerequisite for the emergence of equivalence classes (Sidman 1994; Sidman and Tailby 1982). The absence of such relations would account for the failure in the post-test wherein
Psychol Rec (2014) 64:527–536 533
participants were expected to relate black faces and positive symbols. Since we did not test for any of the prerequisite relations, we are unable to directly address this hypothesis.
But, let us assume that the participants did not form the prerequisite relations. One interpretation, consistent with Watt et al.’s (1991) findings, is that our negative results stem from the use of socially loaded stimuli. Typically, such stimuli are members of socially relevant relations learned prior to the experiment. These relations may generalize to the experimen- tal setting and interfere with the acquisition or expression of new relations. Responding during the tests would then be controlled more by the pre-experimental relations than by the trained relations. Other interpretations of our negative findings point to the specifically racial nature of the social stimuli that we used. Future experiments could test these various interpretations by training the basic relations with social, but non-racial stimuli (e.g., a villain and a hero), or with non-social stimuli with pre-experimental histories (e.g., a carrot and a hamburger).
If future experiments show that specifically racial stimuli strongly constrain the emergence of the relations that are necessary to form an equivalence class, then one could go one step further and investigate whether the constraint is eliminated by training these relations explicitly, or by using other non-matching-to-sample protocols (e.g., the stimulus pairing two-response format of Fields et al. 2009).
Another hypothesis to explain the negative results appeals to the demand characteristics of the experiment. Participants may infer that if one sample strongly goes with one of the two comparisons, the other sample must go with the other compar- ison. Hence, if white faces go strongly with the positive symbol (due to the participants’ pre-experimental histories), then the black faces must go with the negative symbol (due to the foregoing demand characteristic). This hypothesis may recon- cile two seemingly contradictory findings, (a) that all partici- pants learned the AB and BC relations, which should have yielded the new relation C-A between black faces and positive symbols, and (b) that, despite such learning, most participants did not express the new relation at post-test. According to the hypothesis, the black faces were associated with positive sym- bols when they were “pitted against” an abstract symbol, but not when they were “pitted against” white faces. One way to test the hypothesis would be to perform two post-tests, one in which the samples were black faces, abstract symbols, and white faces (as in the present experiment) and another in which the samples were only black faces and abstract symbols. If the hypothesis is correct, black faces should relate to positive stimuli more strongly in the latter post-test. Based on this hypothesis we might also expect that an attempt to simply reverse the classes showed in the pre-training would be more effective. This would require a training protocol designed to reorganize the classes so that the black faces should become equivalent to the positive symbols and the white faces should
become equivalent to the negative symbols. Nonetheless, we did not consider this procedure acceptable on ethical and edu- cational grounds, however effective it might eventually prove. As already mentioned, we deemed it unacceptable to expose children to a training protocol in which negative attributes would be based on racial features. Alternatively, the foregoing issue could be examined by forming a class with white and black faces, associating the white faces with positive stimuli, and then testing if the white-positive relation would transfer to the black faces (Wulfert and Hayes 1988).
One limitation of the present study is that the facial stimuli employed were selected in an Internet search. Although the selected faces did not reveal any obvious emotional expres- sion, it is possible that some participants may have regarded them as presenting expressions. Although such a limitation may be argued to be less important because the primary aim of the present study was to determine if training an equivalence class that reversed the pre-existing black-negative relation would effectively undermine the original relation upon subse- quent testing. Future studies could benefit from a more formal methodology of selection of the facial stimuli (e.g., selecting faces on the basis of judgments made by independent judges), or of employing stimuli that depict only central facial features of black and white men.
The present study, as well as the studies by Watt et al. (1991) and Moxon et al. (1993) used the equivalence class paradigm to study the modification of pre-existing attitudes involving so- cially loaded stimuli; they obtained mostly negative results. In contrast, other studies using the same paradigm to study the formation of new attitudes involving stimuli not socially loaded have obtained mostly positive results. For example, in Grey and Barnes (1996) and in Bortoloti and de Rose’s (2007, 2009) studies, the participants evaluated similarly the elements be- longing to the same equivalence class (movies in the former case, and abstract symbols and human faces in the latter case). In the same vein, in Barnes-Holmes et al. (2000) study with adults and Smeets and Barnes-Holmes (2003) study with chil- dren, the participants rated soft drinks labeled with nonsense syllables according to the class to which the labels belonged. The results from Barnes-Holmes et al.’s (2000) study are par- ticularly interesting because the preference toward soft drinks acquired initially was subsequently reversed following the re- versal of the equivalence classes.
To summarize, results obtained thus far suggest that the explanatory model of attitudes proposed from the stimulus equivalence paradigm, as well as the corresponding experi- mental procedures used to study them, are valuable instru- ments for the understanding of the processes involved in the formation of attitudes. However, the process involved in the reversal of attitudes, particularly pre-existing attitudes involv- ing socially loaded stimuli, remain unclear. Procedural fea- tures that, according to previous studies, should facilitate the acquisition of new behavioral relations, showed only partial
534 Psychol Rec (2014) 64:527–536
success when these relations involved socially loaded stimuli. However, the one participant that showed class reversal re- ceived training combining delayed matching with overtraining, presumably two potent variables to increase relatedness of stimuli within equivalence classes. This suggests that careful selection of training parameters may be a promising line of investigation to increase the effectiveness of training to reverse classes contain- ing socially loaded stimuli.
Author Note Based on a thesis submitted by the first author to the Universidade Federal de São Carlos in partial fulfillment of the require- ments for a Master’s degree in Psychology. This project was part of the research program of the National Institute of Science and Technology on Behavior, Cognition and Teaching, supported by the National Research Council (CNPq, Grant 573972/2008-7) and the São Paulo Research Foundation (FAPESP, Grant 08/57705-8). The first author was supported by a scholarship from FAPESP (Grant 2007/06345) and the second author was supported by a research productivity grant from CNPq. We thank Armando Machado, François Tonneau, Deisy de Souza, and Tânia de Rose for helpful comments on previous versions of the manuscript.
Appendix
References
Barnes, D., Lawlor, H., & Smeets, P. (1996). Stimulus equivalence and academic self concept among mildly mentally handicapped and nonhandicapped children. The Psychological Record, 46, 87–107.
Barnes-Holmes, D., Keane, J., Barnes-Holmes, Y., & Smeets, P. (2000). A derived transfer of emotive functions as a means of establishing differential preferences for soft drinks. The Psychological Record, 50, 493–511.
Bortoloti, R., & de Rose, J. C. (2007). Medida do grau de relacionamento entre estímulos equivalentes [Measuring the relatedness of equiva- lent stimuli]. Psicologia - Reflexão e Crítica, 20, 252–258.
Bortoloti, R., & de Rose, J. C. (2009). Assessment of the relatedness of equivalent stimuli through a semantic differential. The Psychological Record, 59, 563–589.
Bortoloti, R., & de Rose, J. C. (2012). Equivalent stimuli are more strongly related after training with delayed matching than after simultaneous matching: A study using the Implicit Relational Assessment procedure (IRAP). The Psychological Record, 62, 41–54.
Bortoloti, R., Rodrigues, N. C., Cortez, M. D., Pimentel, N., & de Rose, J. C. (2013). Overtraining increases the strength of equivalence rela- tions. Psychology & Neuroscience, 6, 357–364.
Dixon, M. R., Rehfeldt, R. A., Zlomke, K. R., & Robinson, A. (2006). Exploring the development and dismantling of equivalence clas- ses involving terrorist stimuli. The Psychological Record, 56, 83– 103.
Dube, W. V., & Hiris, J. (1997). Match to Sample Program (computer software, version 11.1.3). Waltham, MA: E. K. Shriver Center for Mental Retardation.
Egli, M., Josepf, B., & Thompson, T. (1997). Transfer of social attribu- tions in stimulus equivalence classes by preschool children. Psychological Reports, 80, 3–21.
Fields, L., Landon-Jimenez, D. V., Buffington, D. M., & Adams, B. J. (1995). Maintained nodal-distance effects in equivalence classes. Journal of the Experimental Analysis of Behavior, 64, 129–145.
Fields, L., Doran, E., & Marroquin, M. (2009). Equivalence class forma- tion in a trace stimulus pairing two-response format: effects of response label and prior programmed transitivity induction. Journal of the Experimental Analysis of Behavior, 92, 57–84.
Garotti, M. F., & de Rose, J. C. (2007). Reorganization of equivalence classes: evidence for contextual control by baseline reviews before probes. The Psychological Record, 57, 87–102.
Garotti, M. F., de Souza, D. G., de Rose, J. C., Molina, R., & Gil, M. S. A. (2000). Reorganization of equivalence classes after reversal of base- line relations. The Psychological Record, 50, 35–48.
Gazzaniga, M., & Heatherton, T. (2002). Psychological science: mind, brain and behavior. New York: Norton.
Grey, I., & Barnes, D. (1996). Stimulus equivalence and attitudes. The Psychological Record, 46, 243–270.
Guerin, B. (1994). Analyzing social behavior. Reno: Context Press. Moxon, P., Keenan, M., & Hine, L. (1993). Gender-role stereotyping and
stimulus equivalence. The Psychological Record, 43, 381–394. Osgood, C., Suci, G., & Tannenbaum, P. (1957). The measurement of
meaning. Urbana: University of Illinois Press.
Table 1 Each cell shows the number of choices of the negative symbols and total number of choices following each sample (black faces, white faces, and ab- stract symbol). For participants M2 and M3 some cells are empty because they were exposed to Version 2, which used only two faces of black men and two faces of white men
Black Faces White Faces Abstract Symbol
Man 1 Man 2 Man 3 Man 1 Man 2 Man 3
M1 Pre-test 9/9 9/9 9/9 0/9 0/9 0/9 3/9
Post-test 15/15 15/15 14/15 0/15 15/15 13/15 8/15
F1 Pre-test 1/6 6/9 8/9 9/9 0/9 0/9 9/9
Post-test 0/15 15/15 15/15 15/15 0/15 0/15 13/15
M2 Pre-test 15/24 21/24 — 4/24 3/24 — 4/24
Post-test 9/9 9/9 — 0/9 0/9 — 0/9
M3 Pre-test 9/9 9/9 — 0/9 3/9 — 1/9
Post-test 1/9 0/9 — 0/9 0/9 — 9/9
Psychol Rec (2014) 64:527–536 535
Pilgrim, C., & Galizio, M. (1990). Relations between baseline contingen- cies and equivalence probe performances. Journal of the Experimental Analysis of Behavior, 54, 213–224.
Pilgrim, C., & Galizio, M. (1995). Reversal of baseline relations and stimulus equivalence: I. Adults. Journal of the Experimental Analysis of Behavior, 63, 225–238.
Pilgrim, C., Chambers, L., & Galizio, M. (1995). Reversal of baseline relations and stimulus equivalence: II. Children. Journal of the Experimental Analysis of Behavior, 63, 239– 254.
Rodrigues, A., Assmar, E. M. L., & Jablonski, B. (2003). Psicologia Social [Social Psychology]. Rio de Janeiro: Vozes.
Sidman, M. (1994). Equivalence relations and behavior. Boston: Authors Cooperative.
Sidman, M., & Tailby, W. (1982). Conditional discrimination vs. matching to sample: an expansion of the testing paradigm. Journal of the Experimental Analysis of Behavior, 37, 5–22.
Smeets, P., & Barnes-Holmes, D. (2003). Children’s emergent prefer- ences for soft drinks: stimulus equivalence and transfer. Journal of Economic Psychology, 24, 603–618.
Smeets, P., Barnes-Holmes, Y., Akpinar, D., & Barnes-Holmes, D. (2003). Reversal of equivalence relations. The Psychological Record, 53, 91–119.
Watt, A., Keenan, M., Barnes, D., & Cairns, E. (1991). Social categoriza- tion and stimulus equivalence. The Psychological Record, 41, 33–50.
Wulfert, E., & Hayes, S. (1988). Transfer of conditional ordering re- sponse through conditional equivalence classes. Journal of the Experimental Analysis of Behavior, 50, 125–144.
536 Psychol Rec (2014) 64:527–536
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
