A RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Behavior and Social Issues, 17, 39-64 (2008). © Jonathan H. Weinstein, Kelly G. Wilson, Chad E. Drake, & Karen Kate Kellum. Readers of this article may copy it without the copyright owner’s permission, if the author and publisher are acknowledged in the copy and the copy is used for educational, not-for-profit purposes. A RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION Jonathan H. Weinstein1 Kelly G. Wilson Chad E. Drake Karen Kate Kellum The University of Mississippi ABSTRACT: The purpose of this study was to investigate the transformation of stimulus functions from socially relevant to arbitrary stimuli as a model of social stigmatization and categorization. Specifically, participants were trained to respond to arbitrary stimuli as if they were obese or thin stimuli via a matching-to-sample preparation. The impact of this relational conditioning was tested using the Implicit Association Test. The results showed that when participants met the fluency-based training criterion, the bias functions of obese/thin stimuli successfully transformed formally similar variants of the arbitrary stimuli. These results suggest it is possible to affect a transformation of bias functions to wholly arbitrary stimuli using a very brief conditioning history. A clearer conceptualization of the development of stigmatizing categories, particularly as it applies to obesity, might yield important insights into the social contexts that cultivate and maintain stigmatizing attitudes. KEYWORDS: relational frame theory, social categorization, obesity, Implicit Association Test, fluency Research in the field of bias has been a major focus of social psychologists for some 80 years (Fiske, 2004). In light of recent international events, such as the escalation of violence in Afghanistan and Iraq, the Palestinian-Israeli crisis, and genocide in Darfur, as well as domestic incidents such as the allegations levied against white athletes from the Duke Lacrosse team for the sexual assault of an African American woman, it comes as little surprise that interest in the phenomenon of prejudice continues to grow. Even prior to the attacks of September 11, 2001, about a third of the talks at social psychology’s national and international conferences addressed bias and intergroup relations (Suls, 2001). What follows is a presentation of a theoretical view that many social psychologists have utilized in their effort to understand the phenomenon of social categorization. This view will be contrasted with a behavior analytic account. As questions raised in this analysis suggested the need for an empirical investigation, social bias against obese 1 Address editorial correspondence to: Kelly G. Wilson, Department of Psychology/Peabody Building, University of Mississippi, University, MS 38677. 39
WEINSTEIN, WILSON, DRAKE & KELLUM persons was selected as the means to address the broader phenomenon of social categorization. Bias Against Obese Persons Bias against obese persons occurs across the life span and in diverse settings from education to employment (Puhl & Brownell, 2001). Ninety-six percent of overweight adolescent girls report negative experiences through every grade of school (Neumark- Sztainer, Story, & Faibisch, 1998). Ninety-one percent of clinically overweight children report feeling ashamed of being fat, 90% believe that teasing and humiliation from peers would stop if they lost weight, and 69% believe that they would have more friends if they lost weight (Irving, 2000; Pierce & Wardle, 1997). In occupational settings, the stigma of obesity appears to be correlated with a significant wage penalty. Data taken from the National Longitudinal Survey Youth Cohort revealed that obese women earn 12% less than non-obese women (Register & Williams, 1990). Another longitudinal study following young adults over 8 years found that overweight women earn over $6,000 less than non-obese women (Pagan & Davila, 1997). Anti-obese bias even affects the behavior of families. In a study of undergraduate students, normal weight students received more family support for college than overweight students, who depended more on financial aid and jobs to pay for their education (Crandall, 1991). Differences in family support remained even when controlling for student and parental education, income, ethnicity, and family size. To understand the persistence of anti-obese discrimination a review of efforts to understand the general phenomenon of social categorization would seem helpful. Recently, technological advances in research methods have given rise to the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998), a measure that has contributed widely to the exploration of social categorization. This test involves a computer task that measures response latency while associating two target concepts with an attribute. The two concepts appear in a two choice task (e.g., flower vs. insect names), and the attribute in a second task (e.g., pleasant vs. unpleasant words). When instructions oblige highly associated categories to share a response key, performance is faster than when less closely associated categories share a key. Prior to the development of the IAT, attitudes were traditionally measured using self-report questionnaires with well-normed psychometric properties. However, researchers have identified a number of areas where the results of self-report instruments are not consistent with the results of the IAT. Discrepancies between these two methods have been identified in studies of attitudes toward gender (Greenwald & Farnham, 2000), race (Banaji, Greenwald, & Rosier, 1997; Greenwald et al., 1998), ethnicity (Greenwald et al., 1998) and age (Mellott & Greenwald, 2000). However, agreement between IAT- measured and self-report-measured attitudes has been observed in studies of attitudes toward political candidates and consumer products (Maison, Greenwald, & Bruin, 2001; Nosek, Banaji, & Greenwald, 2002). 40
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION In studies of obesity, behavioral measures have tended to be more sensitive indicators than self-report (i.e., paper and pencil) measures. Bessenoff and Sherman (2000) used a lexical decision task to demonstrate that behavioral anti-obese evaluations predicted how far participants chose to sit from an overweight woman, whereas self- reported measures did not. Behavioral measures have also shown the capacity to discriminate preferences among obese and non-obese children. In a study using the Extrinsic Affective Simon Task (a modified version of the IAT; De Houwer, 2003), children and adolescents with obesity indicated a more pronounced positive attitude toward food than a non-obese sample. This finding was not reflected in the self-report measures given to both groups (Craeynest, et al., 2005). Other studies comparing behavioral and self-report measures are less emphatic. Teachman and colleagues (2003) examined responding on an IAT task with obese-thin, good-bad and lazy-motivated stimuli. When results were compared against a self-report measure, (the Fat Phobia Scale; Robinson, Bacon, & O’Reilly, 1993), the good-bad IAT was not significantly related to the self-report measure; however, the lazy-motivated IAT showed a positive correlation with the self-report measure. The Social Knowledge Structure Greenwald and colleagues (2002) have proposed a theory to account for these and other IAT findings. Specifically, they have hypothesized the existence of a Social Knowledge Structure (SKS) based on principles stemming from earlier theories of cognitive consistency such as congruity theory (Osgood & Tannenbaum, 1955), cognitive dissonance theory (Festinger, 1957), and balance theory (Heider, 1958). The SKS model borrows three principles from each of these theories to constrain relationships among concepts and objects. Greenwald applies congruity theory to predict that two unrelated concepts sharing a higher number of first-order links than expected by chance to a third concept will develop a stronger relationship. In cognitive dissonance theory, Greenwald finds support for predicting how concepts with fewer than average first order links than expected by chance will develop a weaker relationship and even become opposed to one another. Lastly, Greenwald appeals to Heider’s balance theory to account for how repeated pressure on opposing concepts could force two bipolar-opposed nodes into a relation that accommodates these bipolar-opposed nodes via the development of stereo- subtypes. In Figure 1, a hypothetical social knowledge structure (SKS) is represented. This structure includes associations that correspond to social psychological constructs of self-concept, self-esteem, and attitude. Nodes (ovals) represent concepts and links (lines) represent associations. The self-concept includes links of the “ME” node to concepts that include roles (Graduate Student) and trait attributes (intelligent, fat, lonely); self-esteem is the collection of associations—either direct or mediated through components of the self-concept—of the ME node to valence (+ + + or - - -). In Figure 1, both “fat” and “lonely” nodes share first order links with both the “Me” node. As such, Greenwald’s SKS model predicts the formation of a second order link connecting the “fat” and “lonely” nodes. 41
WEINSTEIN, WILSON, DRAKE & KELLUM Figure 1. Social Knowledge Structure. Additionally, the SKS model attempts to predict the structure and strength of relationships among stimuli. One potential weakness of the SKS model is that it does not provide a precise account for how events enter the “network.” For this reason, it may prove fruitful to consider an alternative theory of cognition to determine with greater specificity the process by which stimulus relations develop. Relational Frame Theory Relational Frame Theory (RFT; Hayes, Barnes-Holmes, & Roche, 2001) is a behavior analytic account of the learning processes thought to distinguish human language and cognition. RFT describes a learning process applicable to a variety of complex human behaviors, including social categorization and stigma. According to RFT, verbally competent humans are able to relate events, cognitive and otherwise, without the sort of direct conditioning history that would be necessary for non-human species (Hayes, 1989). “Relating” means to respond to one event in terms of another. In addition to responding to formal stimulus properties, humans can also respond relationally to stimuli. For example, imagine gazing at the photograph of a loved one. The formal properties of the photograph may include the color and texture of this particular piece of paper, but the relational properties may include talking to the picture or caressing it. These responses reflect a functional relationship to the stimulus functions of the picture but bare no formal relationship with the paper itself. Furthermore, the stimulus functions are not accountable via a direct learning experience, as may have occurred if the photograph was viewed during that walk on the beach with the loved one. 42
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION Relational responding involves three properties: mutual entailment, combinatorial entailment, and the transformation of stimulus functions (Hayes et al., 2001). Mutual entailment describes a functional and bi-directional relationship between two events. If a human subject is reinforced for making response A to event B, humans can derive that B is related back to A in a complementary manner. An example of mutual entailment may involve conditioning a person to select the nonsense syllable “Terch” from an array of other nonsense syllables in the presence of the nonsense syllable “Brieg.” Subsequently, the person may select “Brieg” in the presence of “Terch” even though this behavior was never reinforced directly. As neither “Brieg” nor “Terch” convey very many psychological functions in our language community, the presence of a functional relationship would be unlikely to develop despite their history of mutual entailment. A better illustration of a functional relationship is apparent in the everyday example of coins. If one attends only to the formal properties of a nickel in relation to a dime one would notice that the nickel is larger. Functionally, however, a dime is worth twice as much as a nickel. There are no formal properties inherent to the dime itself that would indicate this relationship. Thus, a relational response of this kind is not dependent on the formal characteristics of the stimulus so long as value (i.e., how much you can purchase) is the controlling context. Among the differing features that separates a nickel from a dime is that a nickel’s sides are smooth whereas a dime has ridges. Again, these formal properties are arbitrary in the sense that they have no functional relationship when the context is value. However, suppose one is handed a quarter for the first time having a history of differential reinforcement where dimes are worth more than nickels. If one were only to attend only to the size of the coin, given one’s history one might derive that quarters are worth less than nickels since quarters are even larger than nickels, which in turn are larger than dimes. On the other hand, since a quarter also has ridges on its sides one might derive that it is worth more than a nickel because a dime also has ridges on its side. Thus the formal properties of coins alone have less influence on responding without their correlated functional properties. Combinatorial entailment describes the functional relationships among a group of three or more stimuli. If A is related to B, and A is also related to C, humans can derive that B is related to C and C is related to B in some fashion. For example, if a person learns to select “Vomit” in the presence of “Brieg” in addition to selecting “Terch” in the presence of “Brieg,” the person may subsequently select “Vomit” in the presence of “Terch” and vice-versa. Again, this behavior occurs in the absence of any direct conditioning. The transformation of stimulus functions refers to an exchange of functional properties among related events. Reading the word “Vomit” may arouse reactions that have little to do with the printed word itself, such as feelings of disgust and revulsion. Subsequent to the training described above, if those reactions occur when reading the word “Terch,” it would exemplify an equivalence-based transformation of stimulus function. In other words, “Terch” is functionally equivalent to “Vomit.” 43
WEINSTEIN, WILSON, DRAKE & KELLUM Equivalence is not the only functional relationship possible through the transformation of stimulus functions. Humans learn to relate events in a variety of ways. For example, when people count nickels, dimes, and quarters, they are exhibiting a repertoire for hierarchical relationships that are functional in the context of value. Relations such as similar/different, before/after, part/whole, better/worse, cause/effect, and self/other have been demonstrated in a variety of experimental preparations (see Hayes et al., 2001, for a review). Thus, the RFT view of cognition may provide a greater degree of specificity between events than the SKS model and other association theories. Transfer or Transformation of Psychological Function The phenomenon referred to as the transfer of stimulus functions has also been referred to as the transformation of stimulus functions (Dougher, Perkins, Greenway, Koons, & Chiasson, 2002). Studies examining this phenomenon involve methods where some number of equivalence classes is established using matching-to-sample or other training procedures. Then one or some subset of members of one of the classes are selected and given some new behavioral function. Following this training history, the remaining members of all of the classes are tested to see if they also have acquired the new behavioral function. If the other members of the class from which the subset was selected acquire the new function but the members of the other class do not, the novel functions are said to have transferred within the equivalence class (Dougher, et al., 2002). This phenomenon has been robustly demonstrated in studies of respondent elicitation (Dougher, Auguston, Markham, Greenway, & Wulfert, 1994) as well as operant paradigms containing stimuli bearing functions for gender (e.g., Kohlenberg, Hayes, & Hayes, 1991; Moxon, Keenan, & Hine, 1993), anxiety (Leslie, Tierney, Robinson, Keenan, Watt, & Barnes, 1993) and terrorism (e.g., Dixon, Dymond, Rehfeldt, Roche, & Zlomke, 2003; Watt, Keenan, Barnes, & Cairns, 1991). Additionally there have been a number of studies that have demonstrated derived relational responding among novel stimuli across various functional and topographical dimensions of behavior (see Fields & Reeve, 2001; O’Hora, Roche, Barnes-Holmes, & Smeets, 2002; Rehfeldt, 2003; Rehfeldt & Hayes, 2000). These studies suggest that relational responding is a form of generalized operant behavior (Hayes & Barnes, 1997), and provides a basis for examining how the transfer of psychological functions might have relevance in the study of social categorization. However, there is still some disagreement about terminology. Sidman (1992, 1994) has argued that the term “transfer of function” implies an unnecessary hypothetical process that can be accounted more parsimoniously in mathematical set theory terms. Hayes and colleagues (2001) share these same concerns about positing an account for behavior occurring at some other time inside a setting that defies direct observation. Additionally, they object to the term “transfer” because it fails to adequately describe the changes in functions that accrue in stimulus relations other than equivalence. For example, Dymond and Barnes (1995) showed that training relations of “sameness,” “less than,” and “more than” among a set of stimuli influenced the functions of stimuli subsequently added to the class in 44
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION nonequivalent ways. Hayes and colleagues (2001) suggest that describing this influence as a “transfer” of functions seems imprecise, since stimuli participating in a nonequivalent relation with each other subsequently will not share equivalent stimulus functions. For this reason, the term transformation is offered as a preferable alternative to transfer, since it allows for a broader description of the phenomenon. However, for the current work either term is applicable since only equivalence relations were trained. Thus, “transfer” and “transformation” are used interchangeably throughout the description of this study. Transformation of Psychological Function and Application to Prejudice Prejudice, in particular, is a demonstration of how human beings can become entangled in evaluative categories that may or may not be valid (Hayes, Niccolls, Masuda, & Rye, 2002). According to Hayes and colleagues (2001), “Prejudice involves a derived transformation of the functions of individuals based on direct or verbal contact with the functions of a few members of conceptualized groups (p. 202).” For example, if one’s family was threatened with ethnic cleansing by Hebrew speaking soldiers, then the aversive functions of those soldiers may be transformed in respect to other Hebrew speakers, such as Jewish Americans. Further, those same functions may also transfer to related symbols of Judaism such as the Star of David, without any direct conditioning history. The transformation of psychological functions describes how stigmatizing functions acquired indirectly through relational conditioning processes maintain prejudicial behavior. Consider another example: Imagine a very obese man in a seated position wearing almost no clothing. Rolls of fat are apparent everywhere. Is this a pleasant image? What adjectives come to mind? Ok, but what if it is the Buddha? Unlike other accounts, RFT also offers a principled account for how the functions of one class of stimuli come to be transferred to other stimuli (Hayes et al., 2001). A study examining the transfer of socially and culturally important stimulus functions to neutral stimuli that examines the emergent socially relevant stimuli in terms of the IAT could provide linkage between two experimental domains. Such a linkage between social psychology literature and recent behavioral theories of cognition could offer a developmental account of the phenomena and allow an empirical and theoretical investigation of SKS phenomena from an RFT perspective. The purpose of this study was to investigate the transfer of stigma functions from socially relevant to arbitrary stimuli as a model of stigmatization and social categorization. METHOD Participants Fifty undergraduate students with an age range between 18-29 years were recruited for this study in return for extra credit in a psychology course. Thirteen of the 50 45
WEINSTEIN, WILSON, DRAKE & KELLUM participants identified themselves as African American, 1 as Hispanic, and 36 as European American. Thirty-five participants were female. Stimulus Selection Prior to the current study, an independent sample of fifty participants was provided with an array of positive, negative, obese and thin (i.e., normal body weight) words. These participants rated the stimuli along three dimensions: (a) strength of emotional reaction, (b) understanding, and (c) familiarity on a scale from 1 to 5. Words and phrases that were high in each dimension were used (see Appendix A for a table of mean ratings and standard deviations). The purpose of these ratings was to ensure that participants would understand and respond to the stimuli in a way consistent with their commonly understood meaning. A series of paired t-tests were conducted to determine if stimulus ratings for the stimuli included in the experimental task differed across three dimensions for positive versus negative words and obese versus thin words. The dimensions assessed were familiarity, understanding, and strength of emotional reaction. No significant differences were found between positive and negative words on the three dimensions, suggesting that the stimuli received about the same ratings for all of the dimensions assessed. There were no significant differences found between obese and thin words on ratings of familiarity and understanding. However, participants reported a stronger emotional reaction to the obese stimulus class than to the normal thin stimulus class, t (15) = 3.42, p < .01, when rating this dimension. Differences in the strength of emotional reactions between obese and thin stimuli were expected based on previous studies of obesity bias (e.g., Teachman et al., 2003). Apparatus and Setting Participants performed the experiment in a small room (6 x 10) on computers running Microsoft Windows 2000. Participants were run in groups varying from two to five students per session. Self-report Measures Glenn Measure of Attitudes Toward Obese People. The Glenn Measure of Attitudes Toward Obese People scale consists of 44-items selected from previously published scales designed to indicate attitudes toward obese people (Glenn & Chow, 2002). Reliability analysis yielded a Cronbach coefficient alpha of .92, while a factor analysis of individual items revealed a total of four factors. This measure was selected because its items directly assess positive/negative attitudes towards obese people in general. This measure also solicits information sufficient to compute participant Body-Mass Indices. Body-Esteem Scale for Adolescents and Adults. The Body-Esteem Scale for Adolescents and Adults assesses participants' attitudes and feelings about their bodies and appearance (Mendelson, Mendelson & White, 2001). The 23-item Scale taps three 46
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION aspects of body esteem in adolescents and adults: (a) general feelings about appearance, (b) weight satisfaction, and (c) attributions of positive evaluations about one's body and appearance to others. This scale has demonstrated high internal consistency, sufficient test-retest reliability, as well as convergent validity with the Self-Esteem Scale (Rosenberg, 1979) as well as other measures. The scale’s norms were developed from a sample of participants drawn from English-speaking, elementary schools, high schools, colleges and universities in Montreal, Quebec, Canada. The sample included 1,334 participants (761 females; 571 males) between 12 and 25 years old (M = 16.8 years). This has demonstrated validity and reliability over a wide age range, and can be given confidently to children as young as 12 and to individuals well into adulthood. This measure was selected because its items assess participants’ attitudes towards their own appearance. Procedure Completion of self-report measures. After giving their informed consent, participants completed the self-report measures via paper and pencil administration and provided demographic information, including weight and height. Task I—Response Task without Stigma. After the self-report measures, participants began the first of 5 computer tasks. Task one involved a procedure whereby participants were shown neutral stimuli (i.e., horizontal and vertical lines) and positive or negative descriptors. Participants responded to stimuli across seven blocks of trials. The presentation and location of the stimuli were arranged as outlined by Greenwald and colleagues (2003; see Table 1). The first block randomly assigned drawings of horizontal or vertical lines to one of two categories, Horizontal (by pressing the keyboard letter “E”) or Vertical (by pressing the keyboard letter “I”). On the second block, participants were instructed to press the E- key for positive adjectives, and the I-key for negative adjectives. On the third block, participants were instructed to press the E-key for positive adjectives or for horizontal lines, and the I-key for negative adjectives or vertical lines. On the fourth block, participants repeated the same task as in the preceding block but with twice as many trials. On the fifth block, participants received the same stimuli as in block 1 with the exception that key order was reversed (horizontal lines by pressing the keyboard letter “I” or vertical lines by pressing the keyboard letter “E”). On the sixth block, participants received the same stimuli as in block 3 with the exception that their order was reversed. On the seventh block, participants repeated the preceding task but received twice as many trials (see Table 1 for complete list of conditions). Throughout this experiment, key assignment and order of training were counterbalanced (i.e., participants were randomly assigned to receive either condition 1 first or condition 2 first). This design follows the procedure outlined by Greenwald and colleagues (2003). Task II—Response Task with Stigma. Task II involved a procedure whereby participants were shown obesity or thin-relevant words and positive or negative descriptors (see Table 1). The arrangement of the stimuli in Task II was identical to the 47
WEINSTEIN, WILSON, DRAKE & KELLUM TABLE 1. ILLUSTRATION OF THE IAT Task 1— IAT with neutral stimuli Task 2—Obesity IAT Block No. of Function Items assigned to Items assigned Items Items trials left (E-key press) to right (I-key assigned to assigned to press) left (E-key right (I-key press) press) 1 20 Practice Horizontal lines Vertical lines Obese Not Obese 2 20 Practice Good Bad Good Bad 3 20 Practice Horizontal lines + Vertical lines + Obese + Not Obese + Good Bad Good Bad 4 40 Test Horizontal lines + Vertical lines + Obese + Not Obese + Good Bad Good Bad 5 20 Practice Vertical lines Horizontal Not Obese Obese lines 6 20 Practice Vertical lines + Horizontal Not Obese Obese + Bad Good lines + Bad + Good 7 40 Test Vertical lines + Horizontal Not Obese Obese + Bad Good lines + Bad + Good Note. For half the participants, the positions of Blocks 1, 3, and 4 are switched with those of Blocks 5, 6, and 7 respectively. The procedure in Blocks 3, 4, 6, and 7 is to alternate trials that present either a “Good” or a “Bad” word with trials that presented as either Horizontal or Vertical lines. This procedure was recommended by Greenwald and colleagues (2003) to reduce the possibility of an order effect. presentation of stimuli in Phase I with the only difference being the use of obese and thin words in place of the neutral stimuli (horizontal and vertical lines). Task III—Matching-to-Sample (MTS) of Obesity Relevant Stimuli with Horizontal and Vertical Line. For the third task of this study, participants completed a different computer procedure. The purpose of this task was to train participants to develop relational frames of coordination between socially loaded (Fat/Thin words) and neutral stimuli. As with the previous IAT tasks, neutral stimuli were horizontal and vertical lines (see Figure 2). 48
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION Figure 2. Directly trained and derived relations via Matching-to-Sample Training. There were a familiarization block and three additional training blocks to this task. Assignment of obese and non-obese stimuli to either class of the neutral stimuli (horizontal or vertical lines) were counterbalanced to prevent the formal dimensions of the neutral stimuli from influencing participant responses. Prior to beginning Task III, participants were exposed to a block of familiarization training. Familiarization block. The familiarization block displayed a single sample stimulus and three comparison stimuli. All the stimuli used in the familiarization task were arbitrary, as they were comprised of nonsense syllables. Participant responses were reinforced for selecting a comparison stimulus in the presence of the sample. If the participant made a correct choice the words "GOOD JOB - CORRECT!!" appeared on the screen. If the participant made an incorrect choice, the word "INCORRECT" appeared on the screen. Participants repeated the phase until they achieved 100% accuracy. Block 1—Training two conditional discriminations separately. Within each training block, three 3-member sets of stimuli (A, B and C) were presented (see Figure 3). During block 1, participants were conditioned to select one of three “B” comparison stimuli in response to one of the three “A” sample stimuli (e.g., A1a-B1a, A2a-B2a, A3a-B3a) until they achieved a criterion of 100% accuracy on a block of nine trials. Block 2—Training two conditional discriminations separately. They then advanced to block 2 of the experiment, in which three C comparison stimuli were conditioned to 49
WEINSTEIN, WILSON, DRAKE & KELLUM A-- Sample C—Comparison B—Comparison stimuli Stimuli stimuli Wug Zoil Heavy Lean (A1a) (A1b) (B1a) (B1b) (C1a) (C1b) Dek Derf Chubby Skinny (A2a) (A2b) (B2a) (B2b) (C2a) (C2b) Nof Kift Fat Slim (A3a) (A3b) (B3a) (B3b) (C3a) (C3b) Figure 3. Illustration of sample and comparison stimuli for obese and non-obese MTS. three “A” sample stimuli (e.g., A1a-C1a; A2a-C2a; A3a-C3a) until participants achieved the same criterion described earlier. Block 3—Mixed training. Participants then received one block of 12 random trials of mixed training comprised of the relations and stimuli conditioned in the prior two blocks. Participants were required to meet the 100% accuracy criterion described earlier in order to advance. Block 4—Testing phase. During this phase, the participants were exposed to a number of trials that consisted of novel arrangements of the stimuli to test for any emergent equivalence relations (e.g., B1a-C1a, B2a-C2a, B3a-C3a). No feedback was provided during this block. Participants were required to achieve 100% accuracy for each of the 18 trials presented in this block. If participants succeeded in meeting the criterion for correct performance in all of the training and testing blocks, they advanced to a second set of MTS training blocks (e.g., A1b-B1b, A2b-B2b, A3b-B3b). The training and testing blocks in this second MTS task were identical to the first, except that the socially relevant stimuli were different. If the first MTS task contained obesity stimuli, then the second contained thin stimuli, and vice-versa. If participants did not meet the criterion for correct performance in the testing blocks, they were cycled back through training blocks until testing criterion was 50
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION achieved. If participants indicated a preference to discontinue the study they were given credit for their time and excused from the study. Five participants failed to complete MTS training, and their data were not included in final analyses. Most participants completed this experiment in 1.5 – 2.0 hours. Task IV & Task V—Repeat IAT Response Task. Task IV and Task V repeated the procedures presented in Task I & Task II (see Table 1). The purpose for this second task was to determine if the horizontal and vertical lines previously paired with obese and non-obese words influenced the response latency of participants at pairing positive and negative attributes with the neutral stimuli. RESULTS Data Analysis Strategy In previous studies, researchers have analyzed response latencies on the IAT task with a scoring algorithm (Gray, Brown, MacCulloch, Smith, & Snowden, 2005; Greenwald, Nosek, & Banaji, 2003). The algorithm involves calculating the difference in average response latency between the two response tasks and dividing by the standard deviation of all latencies for both tasks. Individual trials with latencies greater than 10,000 ms are eliminated, and participants are excluded if more than 10% of their trials have latencies smaller than 300 ms. After removing these trials and participants, the remaining latency scores are transformed into standard deviation units. When scored in this way, the IAT score (called D) is similar to Cohen’s d calculation of effect size for an individual’s responses in the task (Nosek, Greenwald, & Banaji, 2005). This IAT D_score consists of one number that reflects, in standard deviation units, the magnitude of the difference between stigma consistent and stigma inconsistent responses. The analyses conducted in this study computed D_scores for both IATs (Traditional Obesity and Derived Obesity). As the D_score has a rational zero point, it is possible to calculate a significance test against a point estimate of zero difference. Also, the rational zero allows for the calculation of Cohen's d by dividing the group mean by the group standard deviation. This results in a Cohen's d of the D_scores (B. Nosek, personal communication, December 10, 2005). Latency—Replicating an Obesity effect on the Traditional IAT (see Figure 4) Latencies were examined to determine if differences between the stigma consistent and stigma inconsistent conditions could be detected. A demonstration of longer latencies on the stigma inconsistent condition would replicate the findings from the Traditional 51
WEINSTEIN, WILSON, DRAKE & KELLUM Figure 4. Examples of the Stigma Consistent and Inconsistent Conditions on the Traditional IAT Obesity IAT (Teachman, Gapinski, Brownell, Rawlins, & Jeyaram, 2003). In the present study, forty-two of fifty participants demonstrated longer latencies on the stigma inconsistent stimuli across all 40-test trials (mean total latency = 85.30 seconds, SD = 24.21) as compared to the stigma consistent stimuli (mean total latency = 67.60 seconds, SD = 16.49; t (49) = 6.03, p < .001, one-tailed, d =. 85). Participant latencies were converted to a D_score for a significance test against a point estimate of zero difference (M = 0.38, SD = 0.4, t (49) = 6.77, p < .001, one-tailed, d = .96). It was therefore determined that the obesity effect found by Teachman and colleagues (2003) had been successfully replicated. Matching-to-Sample Training & Testing All participants were required to achieve 100% accuracy on each of the testing blocks or be recycled to the beginning of training. The mean number of training blocks necessary to reach 100% accuracy was 21.3 (SD = 6.61). Overall fluency scores were computed in order to examine the ease with which participants learned the trained and derived relations. Mean participant total fluency was 28.98 correct responses per minute (or about 1 for every two seconds) for the entire sample with a standard deviation of 6.39 correct responses per minute. The range of correct responses per minute fell between 14.26 and 46.48. Derived obesity IAT analytic strategy with matching-to-sample fluency criterion. The distribution of fluency scores was examined for outliers. Dysfluent performances on 52
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION the MTS task were expected to disrupt the transfer of bias functions to the arbitrary stimuli. Accordingly, participants yielding a fluency score 1.5 standard deviations below the mean were excluded from the analysis. This resulted in the exclusion of five participants. Differences on the Derived Obesity IAT task were examined using raw latency scores, and the IAT Latency D_score (see Table 2 for means and standard deviations). It should be noted that the arbitrary stimuli used in the MTS task and the arbitrary stimuli used in the Derived Obesity IAT were topographically similar but not identical (see Appendix B for a listing of the arbitrary stimuli used in both tasks). The purpose of using topographically similar stimuli was to examine the possibility of transfer of function and generalization along the formal dimensions of the stimulus classes. Latencies were examined first to determine if the differences detected on the Traditional Obesity IAT resembled differences on the Derived Obesity IAT for the stigma consistent and stigma inconsistent conditions. As expected, at pre-test, prior to matching-to-sample training, participants showed no difference on their raw latencies in responding to the arbitrary stimuli, t (44) = 1.20, p > .05, d = .18. Responses transformed to a D_score also showed no difference in responding, t (44)= 0.76, p >.05, d = 0.11. At post-test, following Matching-to-Sample training, participant raw latencies showed the expected bias effect, t (44) = 2.13, p < .05, one tail, d = .32. D_score at post-test for the Derived Obesity IAT also showed the expected bias effect, t (44) = 2.11, p < .05, one tail, d = 0.31. Exploratory Analyses To explore the influence of traditional self-report measures on derived relational responding, data was collected for several such measures. These included the Glenn Obesity Attitudes Scale (M = 131.90, SD = 14.65), the Body-Esteem Scale (M= 55.14, SD = 12.69). None of the survey measures were significantly correlated with responding on the Traditional Obesity IAT or the MTS task. This finding is consistent with other efforts to link self-report measures to behavioral ones, insofar as self-report measures often fail to detect bias found by behavioral measures (Banaji et al., 1997; Greenwald et al., 1998; Greenwald & Farnham, 2000; Mellott & Greenwald, 2000). None of the survey measures were significantly correlated with responding on the Derived Obesity IAT with the exception of one measure, body-esteem, r = .29, p < .05. DISCUSSION The purpose of this study was to investigate the transformation of stigma functions from socially relevant to arbitrary stimuli as a model of stigmatization and social categorization. The results showed that when participants met the fluency-based training criterion, the bias functions of obese stimuli successfully transformed the arbitrary stimuli. These results indicate that a very brief conditioning history can affect a transformation of bias functions to wholly arbitrary stimuli. 53
WEINSTEIN, WILSON, DRAKE & KELLUM TABLE 2. LATENCY(S) FOR THE DERIVED OBESITY IAT— MEETING MTS FLUENCY CRITERION (N = 45) M SD p Cohen’s d Pre-Test Stigma Consistent 60.54 11.04 n/a n/a Stigma Inconsistent 62.41 13.15 n/a n/a Difference scores 1.87 ± 2.62 10.46 .24 .18 D_score .04 .38 .45 .11 Post-Test Stigma Consistent 53.19 10.26 n/a n/a Stigma Inconsistent 55.75 10.29 n/a n/a Difference scores 2.56 ± 2.02 8.03 .02 .32 D_score .12 .37 .02 .31 Methodological and Data Analysis Issues In order to investigate the transformation of bias functions it was necessary to determine if this bias existed within the subject pool. This was done by replicating the Traditional Obesity IAT bias response (Teachman et al., 2003). In line with the wider IAT literature, we selected the Latency Scoring Algorithm, or D_score, as the key dependent variable. In the current study, effect sizes for latency were robust (d = .96), suggesting that the Traditional IAT response for obesity had been successfully replicated and that the expected bias was present in the population. In a number of previous studies, self-report measures have failed to predict performance on the IAT when the socially desirable response opposes the social functions of the stimuli (Banaji et al., 1997; Greenwald et al., 1998; Greenwald & Farnham, 2000; Mellott & Greenwald, 2000). The self-report measures used in the current study also failed to predict performance on the Traditional Obesity IAT. Further, these measures failed to predict performance on the Derived Obesity IAT with the exception of the measure of body esteem. It may be that something about this experimentally imposed learning history moderated the relationship between body- esteem and flexibility (as measured by the IAT). However, such an interpretation would be highly speculative as many relationships among different variables were examined 54
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION which increased the probability of Type I error. Subsequent studies could examine whether and under what conditions such an effect might occur. Differences found for the Traditional IAT vs. the Derived IAT Effect sizes on the derived IAT were relatively small (d = .31) as compared to effect sizes found with the traditional obesity IAT (d = .96). One could reasonably speculate that the bias response in the traditional IAT is the result of life-long training history. In contrast, the bias response observed on the experimental task was the result of a far shorter training history (i.e., approximately 45 minutes of exposure to two MTS tasks). This difference in training magnitudes may explain the difference in effect sizes. Behavioral Contribution to Social Categorization Current theory in social psychology does not provide a developmental account to explain how arbitrary collections of stimuli become related. RFT offers a principled account for how this occurs. According to Hayes and colleagues (2001), “Prejudice involves a derived transformation of the functions of individuals based on direct or verbal contact with the functions of a few members of conceptualized groups” (p. 202). In the current study, the MTS procedure provided a training history that facilitated relational responding among neutral images and stigma-related words. These neutral images, while comparable in form, were not identical to the neutral images situated in the IAT task. Nevertheless, a response bias was detected among these formally neutral stimuli subsequent to the relational conditioning provided by the MTS tasks. These results suggest that relational conditioning can provide a coherent and testable process account for the development of stigmatized social categories. As such, this study contributes a developmental link to the extant RFT literature on the generalizability of relational responding among both arbitrary (see Fields & Reeve, 2001; O’Hora et al., 2002; Rehfeldt, 2003; Rehfeldt & Hayes, 2000) and socially-relevant stimuli (e.g., Dixon, Rehfeldt, Zlomke, & Robinson, 2003; Kohlenberg et al., 1991; Moxon et al., 1993; Watt et al., 1991). The current results suggest that generalization and the transformation of stimulus functions may sufficiently account for instances of stigma among social categories. The RFT and SKS models each attempt to explain the phenomenon of social categorization. RFT is a behavior analytic theory based on a functional contextual philosophy, while the SKS model is a social psychological theory founded on a mechanistic philosophy. Despite these differences, the two theories provide substantially overlapping descriptions of cognitive events. From an RFT perspective, the IAT measures relational behavior among various classes of stimuli. Differential performance among these various classes reflects different histories in respect to the stimuli in those classes. The SKS model views the IAT as a measure of associations among stimuli, and infers that differences in behavior are produced by differences in attitudes about the stimuli in the procedure. RFT utilizes more specific language to describe the functions of stimuli in respect to each other and accounts for these functions via a behavioral history. 55
WEINSTEIN, WILSON, DRAKE & KELLUM In the current study, the history responsible for the acquisition of stigma functions by novel stimuli was provided by the MTS procedure. From an RFT perspective, this procedure reinforced the likelihood of arbitrary stimuli acquiring social functions, even in the absence of direct training for those functions. Instead, the means of this acquisition was via the transformation of stimulus functions and stimulus generalization. The MTS task from the SKS perspective is the same exercise as the IAT; with the difference being an attempt to manipulate second order links by giving the otherwise unrelated concepts first order links to a shared third concept. While the SKS model generates similar predictions on the structure and strength of relationships among stimuli used in this study, it does not provide a precise account for how events enter their putative “network”. Despite this weakness, the SKS model offers a venue for behavioral researchers to communicate with mainstream efforts to explore the phenomenon of social categorization. The purpose of this study was to investigate the adequacy with which predictions consistent with Relational Frame Theory could assist in building a developmental account of stigma. This account depends on the contextual control of stimuli, a topic of familiar concern to the behavior analytic community (e.g., Guerin, 2005; McGlinchey & Keenan, 1997; Weatherly, Miller, & McDonald, 1999). In this particular study, contextual control was demonstrated when the relational responses of participants generalized to the novel stimulus sets on the basis of their topographical dimensions (horizontal or vertical lines). Efforts to demonstrate the transformation of function to the formal properties of the arbitrary stimuli succeeded when participants met the training criterion. Although this result suggests the need for careful training procedures, the potential for future research in this area is promising. A future study might produce larger effect sizes for the transformation of stigma functions if the training history assigning obese and thin functions to the arbitrary stimuli was more extensive. Participants likely have had years of training to perceive obesity as undesirable. By contrast, participants received a comparatively brief amount of training for relating weight-relevant words with horizontal or vertical lines. Therefore, to expect the same magnitude of effects with the arbitrary stimuli after only a brief amount of training seems unwarranted. Future efforts may provide larger effect sizes by utilizing more potent training experiences. Another direction for future research might include examining the effects of relational conditioning processes with a special population, such as with a clinical sample. Participants whose struggles are reflected by the stimuli in the procedure may also generate larger effects. Clinical samples have been used effectively in respect to anxiety (Leslie et al., 1993), disabilities (Barnes, Lawlor, Smeets, & Roche, 1996), and prevalent examples of local prejudices (Watt, et al. 1991). The spread of bias is a cultural phenomenon, and may be understood as a requirement for admission to certain groups in order to receive the benefits of membership. Anthropological studies in Indonesia and prospective field studies of unionized shop stewards indicate biases favoring in-group membership in the distribution of benefits (Brown, 1978; Jaspars & Warnaen, 1982). This phenomenon has been demonstrated even when participants are divided into arbitrary groups by explicitly trivial 56
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION or random means. Over thirty studies have demonstrated in-group favoritism in the distribution of rewards where group assignment was determined arbitrarily (e.g., with a coin toss; Billig & Tajfel, 1973; Tajfel, Billig, Bundy, & Flament, 1971; Tajfel, Nemeth, Jahoda, Campbell, & Johnson, 1970; Tajfel, 1982). One dominant approach in explanations for stereotyping, prejudice, and discrimination has been through the examination of the social context in which it occurs. The origin of this account has frequently been attributed to the work of Gordon Allport (1954). Allport believed that normal human behavior inevitably entailed the dividing of objects and people into categories. Just as people might categorize certain toiletries into toothbrushes and combs, preferring one for their teeth and the other for their hair, so, too, people categorize each other into in-groups and out-groups, preferring one and disliking or avoiding the other. Allport believed that all categories engender meaning upon the world, and these early-formed generalizations tend to persist when useful. Allport’s solution to the problem of prejudice turns on the role of a constructive social context. Allport supported equal-status contact, in the pursuit of common goals, sanctioned by institutional supports, and allowing perceptions of each other’s common humanity (Gilbert, Fiske, & Lindzey, 1998). From an RFT perspective, this would involve exposing participants to a variety of relational conditioning that emphasizes and reinforces behaviors that support the relative importance of shared super-ordinate goals rather than goals that promote division. When inter-group competition prevails, the relational conditioning that tends to dominate results in relational frames of distinction between groups and coordination within groups. There is currently a developing body of research that attempts to alter reigning frames of distinction by targeting conditioning processes that perpetuate prejudice and stigma (see Hayes, Bissett, Roget, Padilla, Kohlenberg, Fisher, et. al., 2004; Lillis & Hayes, in press). Returning to the example of obesity, at first the reader is asked to evaluate their image of a nearly naked obese man as pleasant or unpleasant. After this evaluation they are asked to reevaluate this image having learned that the image belongs to the founder of a major world religion. In this example, since the functions of obesity are assumed to be aversive, an obese image is likewise, expected to be an aversive. This would represent transformation of an equivalence-based function. With this image being transformed to having positive functions, (i.e. the image of Buddha) one might expect behavior to accordingly correspond to changes in the functional properties of the stimulus. Returning to the example of Hebrew speaking soldiers, with the functions of other Hebrew speakers now transformed in such a way as to be seen as oppressors, an intervention of the kind described above might change the functions of Hebrew speakers by adding other functions besides the aversive ones. This could include exposure to a new learning history whereby Hebrew speakers and related symbols come to be seen as belonging to culture that has also suffered from political oppression and racism. A powerful sense of shared suffering could transform relational frames of distinction or exclusion to ones of equivalence, inclusion, and coordination. One promising solution to the problems of social categorization could lie in the development of interventions that have the functional outcome of disrupting the 57
WEINSTEIN, WILSON, DRAKE & KELLUM distinction between the observer and the observed (e.g., Hayes, 1984). By breaking down the distinctions among categories, behaviors that depend on discrimination training are no longer functional. To the extent that we understand conditioning processes that strengthen and weaken relational learning histories, a developmental account of bias should follow. This account could help to explain the role of stigma in the process of social categorization, which in turn could provide direction towards the development of technology designed to reduce the influence of stigma. REFERENCES Allport, G. W. (1954). The Nature of Prejudice. New York: Perseus Books Group. Banaji, M. R., Greenwald, A. G., & Rosier, M. R. (1997). Implicit esteem: When collectives shape individuals. Paper presented at the Preconference on Self, Toronto, Canada. Barnes, D., Lawlor, H., Smeets, P. M., & Roche, B. (1996). Stimulus equivalence and academic self-concept among mildly mentally handicapped and nonhandicapped children. Psychological Record, 46 (1), 87-107. Bessenoff, G., & Sherman, J. (2000). Automatic and controlled components of prejudice toward fat people: Evaluation versus stereotype activation. Social Cognition, 18, 329-353. Billig, M. & Tajfel, H. (1973). Social categorization and similarity in intergroup behaviour. European Journal of Social Psychology, 3, 27-52. Brown, R. J. (1978). Divided we fall: An analysis of relations between sections of a factory workforce. In H.Tajfel (Ed.) Differentiation between social groups: Studies in the social psychology of intergroup relations (pp. 395-429). London: Academic. Craeynest, M., Crombez, G., De Houwer, J., Deforche, B., Tanghe, A., & De Bourdeaudhuij, I. (2005). Explicit and implicit attitudes towards food and physical activity in childhood obesity. Behavior Research and Therapy, 43, 1111-1120. Crandall, C. (1991). Do heavy-weight students have more difficulty paying for college? Personality and Social Psychology Bulletin, 17, 606-611. De Houwer, J. (2003) The Extrinsic Affective Simon Task. Experimental Psychology, 50, 77-85. Dixon, M. R., Dymond, S., Rehfeldt, R. A., Roche, B., & Zlomke, K. R. (2003). Terrorism and relational frame theory. Behavior and Social Issues, 12, 129-147. Dougher, M. J., Auguston, E, Markham, M. R., Greenway, D. E., & Wulfert, E. (1994). The transfer of respondent eliciting and extinction functions through stimulus equivalence classes. Journal of the Experimental Analysis of Behavior, 62, 331-351. Dougher, M., Perkins, D. R., Greenway, D., Koons, A., & Chiasson, C. (2002). Contextual control of equivalence-based transformation of functions. Journal of the Experimental Analysis of Behavior, 78, 63-93. Dymond, S., & Barnes, D. (1995). A transformation of self-discrimination response functions in accordance with the arbitrarily applicable relations of sameness, more than, and less than. Journal of the Experimental Analysis of Behavior, 64, 163-184. Festinger, L. (1957). A theory of cognitive dissonance. Palo Alto, CA: Stanford University Press. Fields, L., & Reeve, K. F. (2001). A methodological integration of generalized equivalence classes, natural categories, and cross-modal perception. The Psychological Record, 51, 67- 87. Fiske, S.T. (2004). Social Beings: A core motives approach to social psychology. Hoboken, NJ: John Wiley & sons. 58
RELATIONAL FRAME THEORY CONTRIBUTION TO SOCIAL CATEGORIZATION Gilbert, D. T., Fiske, S. T., & Lindzey, G. (1998). The Handbook of Social Psychology. New York: Oxford University Press. Glenn, C. V., & Chow, P. (2002). Measurement of attitudes toward obese people among a Canadian sample of men and women. Psychological Reports, 91, 627-640. Gray, N. S., Brown, A. S., MacCulloch, M. J., Smith, J., & Snowden, R. J. (2005). An implicit test of the associations between children and sex in pedophiles. Journal of Abnormal Psychology, 114, 304-308. Greenwald, A. G., Banaji, M. R., Rudman, L. A., Farnham, S. D., Nosek, B. A.,& Mellott, D. S. (2002). A unified theory of implicit attitudes, stereotypes, self-esteem, and self-concept. Psychological Review, 109, 3-25. Greenwald, A. G., & Farnham, S. D. (2000). Using the Implicit Association Test to measure self- esteem and self-concept. Journal of Personality and Social Psychology, 79, 1022-1038. Greenwald, A. G., McGhee, D. E., & Schwartz, J. K. (1998). Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology Bulletin, 74, 1464-1480. Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding and using the Implicit Association Test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85, 197-216. Guerin, B. (2005). Combating everyday racial discrimination without assuming "racists" or "racism": New intervention ideas from a contextual analysis. Behavior and Social Issues, 14(1), 46-70. Hayes, S.C. (1984). Making sense of spirituality. Behaviorism, 12, 99-110. Hayes, S. C. (1989). Nonhumans have not yet shown stimulus equivalence. Journal of the Experimental Analysis of Behavior, 51, 385-392. Hayes, S. C. & Barnes, D. (1997). Analyzing derived stimulus relations requires more than the concept of stimulus class. Journal of Experimental Analysis of Behavior, 68, 235-270. Hayes, S. C., Barnes-Holmes, D., & Roche, B. (Eds.). (2001). Relational Frame Theory: A post- Skinnerian account of human language and cognition. New York: Plenum Press. Hayes, S. C., Niccolls, R., Masuda, A., & Rye, A. K. (2002). Prejudice, terrorism and behavior therapy. Cognitive & Behavioral Practice, 9, 296-301. Hayes, S. C., Bissett, R., Roget, N., Padilla, M., Kohlenberg, B., Fisher, G., Masuda, A., Pistorello, J., Rye, A., Berry, K., & Niccolls, R. (2004). The impact of acceptance and commitment training and multicultural training on the stigmatizing attitudes and professional burnout of substance abuse counselors: Research on acceptance and commitment therapy. Behavior Therapy, 35, 821-835. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Irving, L. (2000). Promoting size acceptance in elementary school children: The EDAP puppet program. Eating Disorders: The Journal of Treatment & Prevention, 8, 221-232. Jaspars, J. M. & Warnaen, S. (1982). Intergroup relations, ethnic identity, and self-evaluation in Indonesia. In H.Tajfel (Ed.) Social Identity and Intergroup Relations. Paris: Cambridge University Press. Kohlenberg, B. S., Hayes, S. C., & Hayes, L. J. (1991). The transfer of contextual control over equivalence classes through equivalence classes: A possible model of social stereotyping. Journal of the Experimental Analysis of Behavior, 56(3), 505-518. Leslie, J. C., Tierney, K. J., Robinson, C. P., Keenan, M., Watt, A., & Barnes, D. (1993). Differences between clinically anxious and non-anxious subjects in a stimulus equivalence training task involving threat words. The Psychological Record, 43(1), 153-161. 59
You can also read