Download Comparison of Minimal Group Induction: Memorization, Imagination, and Random Assignment and more Study notes Social Sciences in PDF only on Docsity!
Running Head: MINIMAL GROUP PROCEDURES 1
A Comparison of Minimal Group Induction Procedures
Brad Pinter The Pennsylvania State University, Altoona
Anthony G. Greenwald University of Washington
GPIR Final Draft -- 23 April 2010
Author Note Brad Pinter, Department of Psychology, Pennsylvania State University, Altoona; Anthony G. Greenwald, Department of Psychology, University of Washington. Corresponding author: Brad Pinter, Department of Psychology, Pennsylvania State University, Altoona, Altoona, PA 16601. Email: tbp1@psu.edu
Abstract For 40 years researchers have studied minimal groups using a variety of induction procedures that, surprisingly, have never been formally evaluated. The present paper reports two experiments that compared minimal group induction procedures based on (a) memorization of novel ingroup names, (b) an imagination instruction, (c) random assignment, and (d) false feedback from painting preferences. The memorization procedure produced the largest ingroup favoritism effects on implicit measures of attraction and identification, whereas all procedures produced comparable ingroup favoritism effects on explicit measures of attraction and identification and bonus money allocation. The memorization procedure is recommended as a practical and effective minimal group induction procedure, particularly in cases in which implicit assessments are of primary interest. Keywords : minimal groups , ingroup-outgroup , social identity , intergroup
Theoretical Controversies and Methodological Concerns The literature on the MGP reveals significant controversy over the interpretation of ingroup favoritism effects. Tajfel and Turner (1986) viewed ingroup discrimination as flowing from subjects’ attempts to bolster their social identities. This perspective, central to social identity theory, has remained an important explanation for ingroup favoritism (Hewstone et al., 2002). However, alternative perspectives have emphasized explanations for why subjects discriminate that have little to do with image-bolstering, including the influence of norms (Hertel & Kerr, 2001), a search for meaning (Abrams & Hogg, 1988), uncertainty reduction (Hogg, 2000), and expectations of reciprocity (e.g., Gaertner & Insko, 2000; Rabbie, Schot, & Visser, 1989; cf. Bourhis, Turner, & Gagnon, 1997; Perreault & Bourhis, 1998). Researchers have also questioned the interpretation of ingroup bias effects on attitudinal measures, proposing alternative explanations that reflect cognitive consistency (Gramzow & Gaertner, 2005; Greenwald, Banaji, et al.,
- rather than the motivational account of social identity theory. The MGP literature also reflects concerns regarding the measurement of ingroup favoritism effects. The traditional method for assessing intergroup discrimination, the Tajfel matrices, has subjects distribute money to ingroup and outgroup members by choosing among a set of allocation strategies that vary in the absolute amount given to each group as well as the relative advantage given to each group. The Tajfel matrices have been criticized for confounding the different allocation strategies (Brewer, 1979) and for providing limited or ambiguous response options (Bornstein et al., 1983a, 1983b)
- criticisms that have received attention in numerous reviews concerned with
illuminating the advantages and disadvantages of the matrices (e.g., Bourhis, Sachdev, & Gagnon, 1994; Diehl, 1990; Turner, 1983a, 1983b). Controversy regarding the Tajfel matrices prompted the creation of an alternative MGP discrimination assessment, the Multiple Alternative Matrices (MAMs; Bornstein et al., 1983a). Although similar in many respects to the Tajfel matrices, the MAMs notably provide distinct response options to represent favoring the ingroup over the outgroup (max rel own) and maximizing the absolute value for the ingroup (max own). Interestingly, Bornstein et al. did not find consistent evidence of ingroup favoring allocations using the MAMs and so concluded that typical MGP discrimination findings are produced by characteristics distinctive to the Tajfel matrices. However, Gaertner & Insko (2001) subsequently showed that the MAMs do reliably measure intergroup discrimination provided that the allocation task does not evoke normative prohibitions against discrimination, such as when the allocation task is framed as providing payment , rather than a bonus. More recently, as work on implicit social cognition has exploded (Bargh, 2007; Greenwald, Poehlman, Uhlmann, & Banaji, 2009; Wittenbrink & Schwartz, 2009), groups researchers have increasingly employed implicit measures of ingroup favoritism with success (e.g., Ashburn-Nardo, Voils, & Monteith, 2001; Gregg, Seibt, & Banaji, 2006). Alternative Minimal Group Induction Procedures In addition to the classic MGP procedure described above, researchers have developed and used a number of alternative induction procedures for creating minimal groups. One familiar alternative in the MGP literature is the random assignment
elderly imagination manipulation (compared to a control imagination) was associated with less outgroup bias. Similar imagination procedures have produced reductions in outgroup bias against homosexual men (Turner et al., 2007, Experiment 3) and reductions in stereotype threat with the elderly (Abrams et al., 2008). Despite obvious procedural differences with the classic MGP induction, the powerful effects of imagination shown in the preceding studies suggests that an imagination procedure could be adapted to study MGP effects. Would instructing subjects to simply imagine being a member of a novel group be enough to produce ingroup favoritism? This possibility is considered in Experiment 1. Another candidate procedure outside of the MGP literature comes from research on implicit partisanship (IP; Greenwald, Pickrell, & Farnham, 2002). Greenwald et al. gave subjects 45 seconds to memorize the names of four members of a novel group. Subjects subsequently completed assessments of implicit attitude and identification toward the studied group relative to an unstudied group using Implicit Association Tests (IAT; Greenwald, McGhee, & Schwartz, 1998). Greenwald, Pickrell, et al. (2002) found that subjects implicitly preferred and identified more with the group whose names they had studied compared to the group whose names that they had not studied. Pinter and Greenwald (2004) subsequently reported evidence for strong implicit identification with and preference for the studied group regardless of the type of relationship shared by the groups (i.e., competitive or cooperative) or the type of group (i.e., human or non-human). Further, Pinter and Greenwald discovered a condition in which the name-study effect was limited—when the studied names were identified as students from a rival university. This
result provides a basis for inferring that the memorization procedure ordinarily creates an identity bond, which was undermined in this case by a pre-existing, conflicting bond. These findings suggest that the IP procedure could be used more broadly as a minimal group induction procedure. Although both procedures have been used to study minimal groups, the MGP and IP procedures differ in a number of potentially important aspects. First, MGP procedures involve explicit categorization, whereas the IP procedure provides no categorization. For IP, subjects merely memorize group names, which seems to promote implicit categorization. Second, MGP effects have been observed mostly with explicit measures, whereas IP effects have been observed exclusively with implicit measures. Because the extent to which implicit and explicit measures capture common variation in a domain is quite variable (Nosek, 2005), it is uncertain how much overlap exists for MGP and IP procedures. Third, MGP procedures typically provide only group category information, whereas the IP additionally provides individuating information in the form of group members’ names. According to SIT, individuating information about the ingroup should decrease ingroup favoritism as it implies a weakening of intergroup differentiation. So why does the IP name-study procedure produce favoritism at all? It may be that the focused study of individuating information about one group automatically creates and strengthens links between subjects’ self-concepts and the studied group while leaving the unstudied group relatively neutral or negative by contrast (Greenwald, Banaji, et al., 2002). Interestingly, the subjective group dynamics model (Marques, Abrams, Paez, & Martinez-Taboada, 1998) suggests a complementary process involving individuating information in intact groups. According to the model, group members seek individuating information about ingroup members for the purpose of monitoring ingroup
Subjects were seated in separate cubicles and completed the experimental tasks (including informed consent and debriefing) on computers. Subjects were randomly assigned to one of the three conditions (memorization, imagination, or classic) and, within each condition, to one of two groups (Red or Green). Memorization. Subjects in this condition were introduced to a task modeled in part after procedures used by Pinter and Greenwald (2004; Experiment 1). Subjects first imagined that a small number of students on campus had been divided into two groups based on their preferences regarding two art styles. Next, subjects read the following instruction: Now we'd like to help you learn the names of the people in the groups. The tasks that follow will be easier if you memorize the names of the members in one group. The names for the RED [GREEN] group will be presented on the following page for 45 seconds. Please try to memorize the names…. The next page presented five names (either Red: Lisa, Daniel, Christina, Ryan, and Pat; or Green: Erin, Jeremy, Kimberly, Adam, and Kris) in a horizontal block centered on the screen, below the instruction: “These are the members of the Red [Green] group.” The group members’ names were chosen to be common, familiar names. In contrast to the imagination and classic conditions, memorization subjects were neither explicitly assigned to a group nor told to imagine they were part of a group. Imagination. This condition used a variation of the classic MGP painting preference task. Subjects first imagined that a small number of students on campus had been divided into two groups based on their preferences regarding two art styles. They were instructed to imagine that “half the students were put into the Red group based on
their liking of a particular kind of art. Similarly, half of the students were put into the Green group based on their liking of a different style of art.” Subjects were then instructed to imagine that they had been randomly assigned to one of the two groups and to memorize the name of their assigned group. Classic. This condition was modeled after the MGP procedure developed by Tajfel et al. (1971). Subjects were informed that the purpose of the study was to examine artistic preferences (“Art plays such an important role in our lives, yet artistic preference is virtually ignored by psychologists,” etc.). Subjects’ first task was to rate a series of 20 paintings in two artistic styles that were identified by the arbitrary group names, Red and Green. Paintings were displayed singly, and subjects rated each painting on a 6 point scale (1= dislike very much; 6= like very much ). Afterwards, the computer paused briefly and then, ostensibly based upon the subjects’ ratings (but actually, randomly), indicated each subject’s group by displaying one of the style names (Red or Green) in large block letters for approximately 3 seconds. Subsequent instructions, borrowed from Ashburn- Nardo et al. (2001), reinforced the idea that subjects’ group assignments were based on their artistic preferences and enhanced the deception by presenting additional information about the supposed differences between the two groups (e.g., “Previous research has shown that people who prefer such paintings tend to process perceptual information in a bottom-up fashion. That is, you tend to examine the finer details of new stimuli, and then form an overall impression.”). Name-Group Association Task. Next, in all conditions subjects completed a categorization task to familiarize them with the group members’ names and their corresponding group memberships. This task was necessary so that all subjects—but
Implicit Attitude and Identification Measures. Immediately following the name-group association task, subjects in all conditions completed two seven-block IATs designed to measure implicit attitude and implicit identification with the Red and Green groups. The IAT target categories were represented by the group names, Red and Green , and the ten names of the group members served as stimuli for individual trials. The attribute categories and stimuli for the attitude IAT were pleasant ( good , win , palace , rich , miracle ) and unpleasant ( bad , lose , slum , poor , disaster ) and for the identification IAT, self ( I , me , mine , my , self ) and other ( other , their , theirs , them , they ). Response latencies were used to compute the IAT D measure (Greenwald, Nosek, & Banaji, 2003), for which positive values reflect greater association of self and pleasant (versus other and unpleasant ) with the ingroup than with the outgroup. 1 Cronbach’s alphas for IAT practice and test block responses were acceptable (attitude IAT: α = .81; identification IAT: α = .69). Explicit Attitude and Identification Measures. Following the implicit measures, subjects completed several items to measure explicit attitude and identification. Subjects used a 7-point scale (1- strongly disagree to 7- strongly agree ) to respond to 8 total items measuring attitude (“I like the Red group,” “I like the Green group,” “The Red group is good,” and “The Green group is good”) and identification (“I feel attached to the Red group,” “I feel attached to the Green group,” “I identify with the Red group,” and “I identify with the Green group”). The four items of each type were combined to create difference scores (paralleling the IAT) for which positive values indicate preference for and identification with the ingroup relative to the outgroup. Cronbach’s alphas for these measures were also acceptable (attitude: α = .96; identification: α = .95).
Results The results presented below were based on a series of 2 (Group: red vs. green) X 3 (Condition: memorization vs. imagination vs. classic) ANOVAs. There were no significant effects involving Group in any of Experiment 1 analyses. IAT Attitude and Identification. The attitude IAT involved subjects sorting names referring to the Red and Green groups and words referring to the categories, pleasant and unpleasant. Figure 1, Panel A displays the relevant standardized means for each condition. The generally positive values shown across conditions are indicative of stronger associations of the ingroup relative to the outgroup with pleasant words than with unpleasant words. In a test against zero association, there was a significant overall effect across conditions, F (1, 36) = 11.47, p = .002, η p^2 = .24. However, a significant Condition effect confirms that the degree of implicit ingroup preference varied by induction procedure, F (2, 36) = 7.81, p = .002, η p^2 = .30. Post-hoc analyses (Tukey’s HSD) revealed that implicit attitude effects were significantly larger for the memorization condition than for the imagination condition [ t (26) = 2.45, p = .05, d = .96] or for the classic condition [ t (25) = 3.99, p = .001, d = 1.60]. The latter two conditions did not differ reliably, t (27) = 1.65, p = .24, d = .64. 2 The identification IAT involved subjects sorting names referring to the Red and Green groups and words referring to the categories, self and other. Figure 1, Panel B displays standardized means for the measure by condition. As shown, across conditions, subjects evidenced stronger associations of the ingroup relative to the outgroup with self than with other , F (1, 36) = 13.43, p = .001, η p^2 = .27. The degree of implicit ingroup identification also varied by Condition, F (2, 36) = 13.69, p = 10-5^ , η p^2 = .43. Post-hoc
10 -4^ (identification)]. Overall attitude-identification correlations were large for both the IAT ( r = .66, p = 10 -7^ ) and explicit measures ( r = .71, p = 10-8^ ). Discussion Experiment 1 provided an initial comparison of three minimal group induction procedures. In the memorization condition, subjects were not assigned group membership, but rather simply studied the names of the five members of one of two groups. In the imagination condition, subjects were instructed to imagine they had been randomly assigned to one of the groups. Finally, in the classic condition, subjects were informed that they had been assigned to one of the two groups based on their preferences on a painting judgment task. The results for the IAT attitude and identification measures revealed that, on average, subjects showed greater implicit preference for and identification with the ingroup (i.e., the group whose names they memorized, the group they imagined, or the group they were assigned) relative to the outgroup. Further, these effects were significantly larger for subjects in the memorization condition than for subjects in the other two conditions. Results for the explicit measures of attitude and identification, by contrast, did not show differences among the conditions, but did reveal a consistent, moderate preference for the ingroup. Finally, analyses of implicit-explicit correlations showed weaker relationships between measure types than within measure types, indicating that implicit and explicit ingroup favoritism are not well related. These initial findings suggest some tentative answers to questions about the merits of the three induction procedures. The consistent finding of ingroup favoritism across procedures on explicit measures supports the idea that both the memorization
procedure, adapted from IP research, and the imagination procedure, adapted from imagined intergroup contact research, are reasonable alternatives to the classic procedure. Further, results for implicit measures are intriguing in suggesting a difference among the procedures. Specifically, with the memorization induction, implicit attitude and identification effects were large, whereas with the imagination and classic procedures effects were small and, surprisingly, descriptively outgroup-favoring for the classic procedure. Despite these promising findings, there are aspects of Experiment 1 that could limit its generalizability. One potentially limiting aspect relates to the application of the art scenario to all three procedures. This design feature was intended to improve comparability across procedures, but one could justifiably wonder whether the use of the scenario was necessary or even potentially confounding. For instance, subjects in the memorization and imagination conditions were asked to imagine other students who shared their group memberships. Although hypothetical, this instruction may have inadvertently invoked perceptions of interpersonal similarity, which would have inflated the attitude and identification results. For this reason, the procedures in Experiment 2 were modified to minimize this possibility. This change should affect a more rigorous comparison. A second potentially limiting aspect of the design relates to the exclusive use of attitudinal measures in Experiment 1. Given the long history of the use of behavioral measures in the MGP, one might rightly inquire about the results of a comparison of procedures using a behavioral measure, such as a money allocation task. Experiment 2 included a set of MAMs to address this question. A third potentially limiting aspect is that the initial comparison did not include the commonly used random assignment
Random. Subjects in this condition received no mention of the art scenario. They were simply instructed that the computer would “randomly assign you to a group.” Classic. There were two changes to the classic procedure from Experiment 1. First, to shorten the experiment, subjects rated 10 paintings, not 20 as in Experiment 1. Second, following the display of their group assignment, subjects did not receive subsequent instructions about the purported differences in information processing styles between the two groups. Next, in all conditions, the computer paused briefly and then indicated subjects’ group by displaying one of the group names in large block letters for approximately 3 seconds. Finally, all subjects completed the name-group association task and the dependent measures. Dependent Measures Manipulation check. After completing the same IATs (attitude IAT: α = .83; identification IAT: α = .58) and explicit measures (explicit attitude: α = .90; explicit identification: α = .93) from Experiment 1, subjects responded to a manipulation check question: “Initial instructions emphasized which of the following?” Subjects selected one of four answer choices: “I would be in a group based on my painting preferences,” “I should imagine being assigned to a group,” “I was to study for 45 sec the names of one of two groups,” and “I can't remember.” For analysis, responses were scored so that “0” and “1” indicated, respectively, incorrect and correct answers based on condition. MAMs. The final assessment was the MAM task (Bornstein et al., 1983a). Subjects were instructed that they would indicate their preferences for how a small amount of bonus money, described as left over from a previous study, would be
distributed to two other subjects present in their experimental session (one from each of the approximately equally represented groups). Following typical allocation task instructions (e.g., Gaertner and Insko, 2000, Experiment 1), it was emphasized that subjects could not allocate money to themselves, but they could receive money from both ingroup and outgroup members. It was also emphasized that once subjects had completed the task, the experimenter would access their ratings and allocate money based on the subjects’ preferences. The seven preferences for the distribution of money to the ingroup/outgroup were: $2.10/$1.10 (max rel own), $2.35/$1.50 (max own), $2.20/$1.80 (max joint own), $1.70/$1.70 (min dif), $1.80/$2.20 (max joint other), $1.50/$2.35 (max other), and $1.10/$2.10 (max rel other). Each of the alternatives appeared one at a time (in the fixed order given above) on subsequent screens along with a direction to indicate a preference for each distribution option using a 7-point scale (1- strongly disagree to 7- strongly agree ). Prior to analyses, MAM values were recoded so that positive values represented stronger preference for the ingroup for each MAM for all subjects. A principal components factor analysis was also conducted with the seven MAM preference ratings. This analysis revealed two factors with eigenvalues above 1.0 that accounted for 80% of the variation in the MAM scores. The first factor, labeled, ingroup preference , showed factor loadings greater than .80 on max rel own, max own, and max joint own, whereas the second factor, labeled outgroup preference , showed factor loadings above .80 on max rel other, max other, and max joint other. Results
