University of Southern California Dissertations and Theses

Social crossed categorization beyond the two -group model

(USC Thesis Other)

Social crossed categorization beyond the two -group model

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy subm itted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send U M I a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UM I directly to order. ProQuest Information and Learning 300 North Zeeb Road, Ann Arbor, M l 48106-1346 USA 800-521-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. SOCIAL CROSSED CATEGORIZATION BEYOND THE TWO-GROUP MODEL by Darren I. Urada A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment o f the Requirements for the Degree DOCTOR OF PHILOSOPHY (Psychology) December 2000 Copyright 2000 Darren Ichiro Urada Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UMI Number: 3041540 ___ ® UMI UMI Microform 3041540 Copyright 2002 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY O F SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 90007 This dissertation, written by O a rr&/\ X. u rc\ c i c K under the direction of hi£. Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean o f Graduate Studies Q a fe December 18, 2000 DISSERTATION COMMITTEE Chairperson Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table of Contents Page List of tables and figures iv Abstract V Introduction 1 Experiment 1 15 Method 15 Results 22 Experiment 2 23 Method 23 Results 24 Discussion 24 Experiment 3 26 Method 26 Results 29 Discussion 31 Experiment 4 34 Method 34 Results 36 Discussion 37 Experiment 5 38 Method 38 Results 42 Discussion 43 Meta Analysis 45 Method 45 Results 46 General Discussion 48 Endnotes 55 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. References 56 Appendix A: Participant information form 61 Appendix B: Partner selection form 62 Appendix C: Information form, Glendale College version 63 Appendix D: Sample printouts from online study 64 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. List o f Tables and Figures Table 1: Summary of predictions 13 Table 2: Experiment 1 target preference ratings 23 Table 3: Experiment 2 target preference ratings 24 Table 4: Experiment 3 target preference ratings 29 Table 5: Experiment 4 target preference ratings 37 Table 6: Experiment 5 target preference ratings 43 Figure 1: Weighted mean target preference ratings 47 for studies 1-4 IV Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Abstract A set o f five studies examined how people process in-group and out-group information when presented with stimuli that are more complex than those used in previous crossed categorization studies. A diverse set o f predictions is generated by previous theoretical work. Heuristic, threshold based processing o f information was supported over algebraic processing. Participants appeared to divide stimuli into "in group like" and "out-group like" meta-categories. However, the threshold at which this division was made, and whether it was based upon in-group favoritism or out group derogation, was influenced by the nature o f the situation and the task participants were asked to perform. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Introduction When we meet someone who is different from us on one social category dimension (out-group) but who is similar to us on another social category dimension (in-group), crossed categorization exists. Making crossed categorizations salient holds potential as an effective tool to decrease intergroup bias (Brewer & Miller, 1984; Messick & Mackie, 1989; Vanbesalaere, 1991). However, in order to fulfill crossed categorization's potential, a better understanding o f crossed category information processing is needed. In particular, an understanding o f crossed categorization phenomena with complex stimuli is lacking. Studies o f crossed categorization typically examine the bias exhibited toward other people (“targets”) that are described as having in-group status on one social category dimension and out-group status on another (e.g. Arcuri, 1982; Brewer, Ho, Lee, & Miller, 1987; Brown & Turner, 1979; Deschamps & Doise, 1978; Ensari & Miller, 1998; Hagendoom & Henke, 1991; Hewstone, Islam, & Judd, 1993; Schofield & Sagar, 1977; Urada & Miller, 2000; Vanbeselaere, 1987; for recent reviews see Urban & Miller, 1998; Migdal, Hewstone, & Mullen, 1998). This research has been illuminating, but one common limitation among these studies has been the restriction to two category dimensions. In order for this research to be more appropriately applied to the outside world, it is essential to extend it beyond the two-group model. Humans are able to process as many as 7±2 items at a time (Miller, 1956), and when two people meet, group memberships such as age, ethnic background, and gender can become 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. instantly salient even before either person has hao a chance to speak. Thus, it is crucial that we understand what happens when several category memberships are salient. The general assumption has always been that previous findings from two- category studies will extend to multiple category targets, but the evidence from two- category studies have themselves been mixed and generate divergent predictions. Relevant theories can also be brought to bear from other lines of research, and again different theories predict quite different outcomes. This dissertation will attempt to test these theories and clarify how multiple categorizations are integrated. In particular, this dissertation seeks to investigate how this process operates within the common situation in which the salient groups are o f unequal importance. When dealing with conflicts involving extremely important out-groups, including conflicts with long or intense histories (e.g. American Blacks and Whites, Israeli Palestinians and Jews, Irish Protestants and Catholics, Kosovo Albanians and Serbs, Rwandan Hutus and Tutsis), we usually will not be able to readily find or create an in-group membership that is important enough to counterbalance the destructive effects o f the shared out-group membership. Indeed, all o f the groups in conflict listed above do have a superordinate in-group membership (e.g. American, Israeli, Irish, Kosovar, Rwandan) in common, but clearly this has not been sufficient to prevent conflict. In cases such as these, it is important to know if crossing the out group with multiple less important in-groups is a useful strategy for reducing bias. If this is the case, then interventions can be developed that make salient or even create multiple in-group memberships long enough to allow positive personalized 2 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. interaction to occur. Previous research (e.g. Deschamps & Doise, 1978) suggests that adding a single in-group membership to a target description containing a single out-group does seem to reduce bias even if the in-group membership is less important than the out-group membership (i.e. evaluations o f Oi > O)1 . Extrapolating from these results, it appears plausible that adding multiple in-groups should further improve relations (e.g. Oii > Oi), though by how much is unclear. As will be seen, a profusion o f predictions can be generated from present theories, and some o f these are decidedly less optimistic than others. One plausible hypothesis is that in-group and out-group membership information will be integrated according to an algebraic (averaging or additive) model o f information integration. This prediction is largely based on well replicated evidence that people use an averaging rule when integrating trait information (Anderson, 1965; 1967; 1968). However, this evidence might not extend to the crossed categorization situation. Anderson's typical paradigm was to pretest an array o f traits for likableness, then to have participants rate objects or people described by arrays o f those traits. In general, Anderson found evidence that evaluations were based on an averaging o f the individual traits combined with a neutral baseline and sometimes differential weighting o f the material. Anderson often focused on stimuli (adjective sets) that were either uniformly positive or negative in valence, although in some cases information was mixed (e.g., Oden & Anderson, 1971). Other investigators who have used heterogeneous sets o f traits (which would be similar to crossed category stimuli) have failed to find support for an averaging model 3 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (Singh, 1976; Richey, Koenigs, Richey, & Fortin, 1975). In particular, Singh (1976) found an interaction indicating that averaging occurred in homogenous adjective sets but not in heterogeneous ones. Thus, at best it is unclear that an averaging model would apply to crossed category stimuli, which are by definition heterogeneous. Within the crossed categorization literature itself, the evidence is also mixed, but there are instances of what appear to be algebraic integration of category information, demonstrated by a general pattern o f results conforming to the pattern II> I0 = 0 I> 0 0 . This pattern has been found by Hagendoom and Henke (1991), Hewstone, Islam, and Judd (1993), Singh, Yeoh, Lim, and Lim (1997), Urada and Miller (2000), and Vanbeselaere (1991). The pattern has generally been labeled the “additive” pattern (e.g., Brewer, Ho, Lee, & Miller, 1987; Brown & Turner, 1979; Hewstone, Islam, & Judd, 1993; Urban & Miller, 1998), implying an assumed underlying summative rule. Recently, Singh et al. (1997) reported support for an additive model over an averaging one. However, the evidence was somewhat mixed. The additive model was endorsed because the interaction test that would support an averaging model was not significant for three out o f their four ANOVA tests. Meta- analytic combination o f their studies show no significant interaction and thus support the additive model, but the difficulty in coming to a conclusion based on negative results remains. Hence, it is not entirely clear which algebraic rule best applies to social category information, but there is some support for the general proposition that algebraic combination can occur. 4 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. On the other hand, a number o f other results suggest processing that follows neither additive nor averaging rules. A category dominance pattern (Ii=Io>Oi=Oo), which occurs when one category is used as the basis for evaluation while a second category is virtually ignored, has frequently been observed (Arcuri, 1982; Hagendoom & Henke, 1991; Stangor, Lynch, Duan, & Glass, 1992; Hewstone, Islam, & Judd, 1993; Urban, 1997). An exclusivity pattern (II>IO=OI=00), found when all targets with any out-group membership are evaluated equally negatively, has also been observed (Eurich-Fulcher & Schofield, 1995; Rogers, Miller, & Hennigan, 1981; Schofield & Sagar, 1977; Vanbeselaere, 1991) while the opposite inclusivity pattern (II= I0 = 0 I> 0 0 ) is also common (Brown & Turner, 1979; Vanman, Kaplan, & Miller, 1992; Vanbeselaere, 1991). Finally, a hierarchical pattern (Ii>Io>Oi=Oo or Ii=Io>Oi>Oo), in which the impact o f one category can depend upon membership o f the other, has been found several times (Brewer, Ho, Lee, & Miller, 1987; Hewstone, Islam, & Judd, 1993; Triandis & Triandis, 1960; Triandis & Triandis, 1962; Urada & Miller, 2000). All o f these findings are inconsistent with algebraic rules insofar as they demonstrate that the impact of one category membership can change depending on membership in a second category. Simple algebraic rules lack this property. As one author puts it, “Adding implies that the effect of a social categorization is the same regardless o f what other social categorization it is crossed with.” (Singh et al., 1997, p. 134). The category dominance, exclusivity, inclusivity, and hierarchical patterns all violate this rule. 5 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. A number o f theoretical models, both algebraic and non-algebraic, will be discussed. Where possible, predictions for the crossed categorization situation will be generated from these models, and a series o f studies will test these predictions. These studies represent the first crossed category research beyond two categories. This method has been endorsed by Singh et al. (1997), who, after advocating an adding rule, go on to state: Perhaps the evidence for an adding rule is confined to the two-factor designs only. With more than two categorizations, relative importance o f the various categorizations may come into play and render some categorizations as redundant. Under such circumstance, the additive effect of three categorizations will not hold and other models (Brewer et al., 1987; Hewstone et al. 1993) may apply. It is necessary, therefore, to undertake a more comprehensive study o f the cross-categorization effects, using more than two social categorizations. Future studies o f the models o f cross categorizations should also employ a within-participants design. Presenting all groups to the same participants will make the manipulated social categorizations salient to them. (Singh et al, 1997, p. 137). As recommended by Singh, this series o f studies will both investigate cases beyond two categorizations and will do so in the context o f a within subjects design. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Algebraic Models Both types of algebraic models, additive and averaging, predict elevated evaluations whenever in-group memberships are added to a target description and lowered evaluations whenever out-group memberships are added. Averaging models also predict a decreasing effect o f additional information whereas in its simplest form an additive model would predict a monotonic increment in evaluations with additional in-group information (a modified additive model with diminishing returns would make predictions identical to the averaging model.). These predictions are based on the assumption that all categories are weighted equally, however. This may not be the case. A motivation toward either in-group favoritism or out-group derogation can give greater weight to in-group or out-group information, respectively. In-group favoritism is a widely recognized phenomenon (for a review, see Brewer, 1979), but some authors (e.g., Singh et al., 1997) have found evidence o f out-group derogation instead. If there is a tendency to favor in-groups rather than to actively devalue out-groups, then participants may choose to average in-group information while largely ignoring out-group information. In that case, a change in target evaluations will occur whenever in-group memberships are added (Oi<Oii<Oiii), but not when out-group memberships are added (Io=Ioo=Iooo). On the other hand, if out-group derogation takes precedence, then the opposite pattern will be found (Oi=Oii=Oiii but Io>Ioo>Iooo) as participants devalue targets based on their out-group information but largely ignore the in-group information. 7 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Non-Algebraic Models Psychological models suggesting top-down processing (e.g. Asch, 1946; Brewer, 1988; Fiske, 1984) have historically competed with essentially bottom-up models such the algebraic ones discussed above. That is, within the algebraic models, individual features of the target are combined according to mathematical rules to create an overall impression. The category dominance, exclusivity, inclusivity, and hierarchical patterns, on the other hand, may indicate the presence of top-down processing, and a number o f authors have proposed social information processing models that partially or entirely depend upon top-down processing. A plethora of non-algebraic predictions can be derived from these models. Although the following list of prior work and possible predictions is probably far from exhaustive, it does demonstrate the wide array of possible outcomes. This uncertainty o f outcome emphasizes the need for research in this area. Dual Process Models. Some authors (Brewer, 1988; Fiske & Neuberg, 1990; Fiske & Pavelchak, 1986; Pavelchak, 1989) have presented dual-process models in which social information is processed either in a top-down or a bottom-up fashion depending upon the situation. Consistent with the “cognitive miser” perspective (Fiske & Taylor, 1984), Brewer assumes that perceivers resist using processing stages that require elaboration or modification of existing cognitive structures unless the situation dictates it. The more involved method of bottom-up processing (personalization) is said to occur only when the target is relevant and self involvement is high. Similarly, Fiske's continuum model and its predecessors 8 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (Fiske & Neuberg, 1990; Fiske & Pavelchak, 1986; Pavelchak, 1989) adds proposes that automatic categorization occurs first, but that bottom-up processing will occur when the perceiver has difficulty fitting the information present into a category, either initially or after attempting further recategorization. Integrating these models with the algebraic formulas discussed earlier generates the prediction that category information will be combined algebraically under conditions o f relevance and high- involvement whereas in other situations a non-algebraic pattern will emerge as participants attempt to more simply categorize the target. With the present paradigm, this simplified processing could entail processing based only upon a single dominant category (Io=Ioo=Iooo and Oi=Oii=Oiii) or perhaps based upon a majority in-group/out-group rule (Io<Ioo=Iooo and Oi<Oii=Oiii). Personalization o f similar others. Brewer (1988) also notes, “Other things being equal, we are more likely to form person-based representations of individuals who are similar to ourselves than of those who are distinctly different.” (p. 25). Thus, perceivers are expected to form more complex impressions of in-group members than they do of out-group members. Brewer & Lui (1985) conclude that this complexity effect is a function of perceived similarity to the self. The problem is, “other things being equal,” it is unclear who will be perceived as relatively “similar” in a multiple crossed-categorization situation. That distinction could fall to targets with a dominant in-group, targets with multiple less important in-groups, or to both. Due to the lack o f crossed categorization studies with more than two groups, there is currently no way to predict which of these is correct. If the more 9 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. important in-group membership takes precedence, then people will form person- based representations o f targets with important in-group memberships but will revert to category-based processing when targets have a dominant out-group. This generates the prediction Io>Ioo>Iooo but Oi=Oii=Oiii. On the other hand, if multiple groups have a stronger impact on the perception o f similarity, then participants should form person-based representations only o f targets with multiple in-group memberships but resort to category-based processing for targets without multiple in-group memberships. This generates the prediction Io=Ioo=Iooo and Oi<Oii<Oiii. If either can serve as a strong basis for similarity, then the prediction is Io>Ioo>Iooo and Oi<Oii<Oiii. Feature Detection The complex nature of the stimuli might also influence the type o f processing. Prinz and Sheerer-Neumann (1974), using a stimulus classification task, found that when the number of target attributes exceeded two or three, participants changed processing strategies. With simpler targets, there appeared to be an attempt to identify the stimulus in terms o f all attributes. With more complex stimuli on the other hand, participants tended to switch to a “feature detection” strategy. That is, they focused on the presence o f conjunctions o f attributes rather than on the individual attributes themselves. For example, a participants “might look for large-round-green-ness instead o f testing for largeness, roundness, and greenness independently .. .” (Prinz & Scheerer-Neumann, 1974, p.41). Thus, participants tended to code the stimulus as a whole. In this example, the criterion is “large, round, and green”, while in a crossed category 10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. paradigm the criterion might be “a person like me”. In this case, participants faced with complex target descriptions might find it more efficient to classify targets based on a conjunction o f shared group memberships rather than by adding individual category memberships. That is, a target with a majority o f in-group memberships may be perceived as having a Gestalt “in-group like” quality even if the participant does not add up each individual group membership (e.g. Oi<Oii=Oiii but Io=Ioo=Iooo). Alternatively, the feature detection strategy may be guided by a search for a single dominant in-group membership resulting in a pattern corresponding to Oi=Oii=Oiii < Io=Ioo=Iooo. These are just two examples of how processing might change qualitatively when participants are presented with complex stimuli. This also demonstrates the danger inherent in assuming that results from two-group crossed categorization paradigms will automatically extend to multi category situations. It is quite possible that processing changes qualitatively beyond two categories. Dominance. Rothbart & John (1985) suggest that we categorize on the basis o f prototype matching. A highly educated black biochemist, for example, may not be perceived as a good fit for the category “black” so he may be categorized with (white) scientists instead. This suggests that under certain circumstances one or more category memberships may be simply discarded during judgments. Similarly, Schul and Bumstein (1988) state that, “Integration o f categories may involve complete domination by one category over another; or a merger o f categories and creation o f a new category that may serve (but need not always be) as a 11 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. subcategory o f one of the originals (cf. the concept of subtyping discussed by Brewer).” (p. 150). They further argue that impression formation is partly top- down/categorical by nature even when all conditions specified by Brewer for noncategorical or personalized processing are met (p. 150-151). Sherman (1988, p. 159) also notes that one category may become dominant depending on factors such as primacy, salience, level of activation. As previously noted, category dominance has also been found previously in the crossed categorization literature (Arcuri, 1982; Hagendoom & Henke, 1991; Stangor, Lynch, Duan, & Glass, 1992; Hewstone, Islam, & Judd, 1993; Urban, 1997). If the “most important” category becomes dominant, then the prediction generated is Oi=Oii=Oiii < Io=Ioo=Iooo. If, on the other hand, a constellation of group memberships becomes dominant, as in Rothbart’s example above (“educated” and “biochemist” overwhelm “stereotypical black”), the predictions might be Oi<Oii=Oiii and Io>Ioo=Iooo. At this point, a number of different predictions seem plausible, as summarized in Table 1. As can be seen, there is a very wide range o f potential predictions and in fact it may be possible to generate others beyond those presented in Table 1. This overabundance of predictions highlights the importance o f the proposed studies. Five multiple-category studies based upon the paradigm created by Urada and Miller (2000) were carried out. The primary goal of the first two studies is to explore ratings o f crossed targets that include more than two pieces of 12 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. information. The third study extends the array of targets to non-crossed targets in order to test algebraic processing predictions against non-algebraic processing predictions. The last two studies investigate how well the results of the previous studies generalize to a somewhat different task. Table 1. Summary o f predictions. Pure Algebraic Predictions Standard: Oi<Oii<Oiii Io>Ioo>Iooo In-group Favoritism: Oi<Oii<Oiii Io=Ioo=Iooo Out-group derogation: Oi=Oii=Oiii Io>Ioo>Iooo Modified or Non-Alaebraic predictions Dual process, bottom-up Oi<Oii<Oiii Io>Ioo>Iooo or (same as algebraic predictions) Oi<Oii<Oiii o II O o II o o o or Oi=Oii=Oiii Io>Ioo>Iooo Dual process, top-down Oi=Oii=Oiii > Io=Ioo=Iooo or Oi<Oii=Oiii Io<Ioo=Iooo Personalization o f similar others: Oi=Oii=Oiii o V o o V o o o or Oi<Oii<Oiii Io=Ioo=Iooo or Oi<Oii<Oiii Io>Ioo>Iooo Feature detection: Oi<Oii=Oiii Io=Ioo=Iooo or Oi=Oii=Oiii < Io=Ioo=Iooo Dominance Oi=Oii=Oiii < o o o II o o II o or Oi<Oii=Oiii Io>Ioo=Iooo 13 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Although the primary goal o f this dissertation is to investigate how multiple crossed-categorization information is processed, a secondary goal is to examine how much our knowledge from studies of two-category crossed categorization can be extended to the multiple crossed categorization situation. For two-category crossed categorization, recent research has focused on factors that moderate crossed categorization effects, and affect has been the focus of a number of these studies (Ensari & Miller, 1998; Urada & Miller, 2000; Urban & Miller, 1998). Participants experiencing positive affect categorized tend to categorize positive or neutral people more broadly into positive social categories (Isen, Niedenthal, 8c Cantor, 1992). In the crossed categorization situation, this leads participants to "meta-categorize" crossed targets as in-group-like (Urada & Miller, 2000), resulting in correspondingly higher ratings o f these targets. In order to correlationally assess whether affect’s moderating impact extends to multiple crossed categorization situations, affective measures were added to the current studies. Experiments one and two differ only in the targets that were presented, and could thus be considered different conditions within a single study. However, participants for these studies were run during different semesters and thus were not randomly assigned to the different presentations of targets, creating a possible confound with any unknown participant cohort differences that may exist. Because o f this, the methods and results for these two experiments will be reported as separate studies, but the discussion will be consider both in conjunction. 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Experiment 1 Method Participants. Twenty five undergraduate psychology students at the University o f Southern California (USC) volunteered for the study in order to receive extra credit points in their psychology classes. Three participants were excluded from the statistical analysis because they indicated suspicion o f the study’s cover story during the post- experimental interview and were able to correctly guess at least part of the study’s true purpose. Design. The study had a single factor (Target Type: Oi, Oii, Oiii) fully within- subjects design. The order of the target type presentation and the order in which important/less important information within each target description was presented were randomly determined by the experimenter, who was also blind to the study’s hypotheses. Materials and Procedures. Participants volunteered for the study via a posted sign-up sheet labeled “Team Problem Solving”. On this sheet, they were informed that they would be given the opportunity to get to know someone and then engage in a problem solving task with that person. Prospective participants were instructed to sign up by writing their student identification numbers next to the time that they would like to participate. For each one-hour time slot, there were four spaces, but three of these 15 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. were always filled by bogus numbers. Participants who signed up in each slot were directed to a different rooms scattered throughout the ten-floor building, but the real participant always reported to the same room on the eighth floor. Thus, for every available position, it appeared as if the participant would be participating with three other participants, but no information about these bogus participants could be gathered other than the uninformative identification number and the participant did not expect to meet at the same room as the others. On each sign-up sheet, at least one time slot was completely filled by bogus numbers (no real participants were run during this time) in order to avoid arousing suspicion by presenting a time sheet in which every time slot had exactly three participants signed up. When participants arrived at the room they had been directed to, they were greeted by the experimenter. The participant was led through office-space to a small adjoining room (approximately 3 m by 3 m) where they were seated alone. The experimenter looked at her watch and said, “There are supposed to be three other participants taking part in this study at the same time as you. Let me go check to see if everyone’s here.” Leaving the door open, the experimenter then walked into the adjoining room to a touch-tone telephone sitting on a desk. At this point, the experimenter was out o f the sight o f the participant but still well within earshot, (about 5 m away). The experimenter audibly picked up the phone, punched in a phone number, and engaged in a short mock conversation, supposedly with another experimenter. In reality the phone was not plugged in. The experimenter always ended with, “Oh, one more? How long do you want to wait? (pause) Uh-huh. 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (pause) Okay, I’ll call you back a little later, (pause) Okay, bye.” The experimenter then returned to the participant’s room and informed the participant that they needed to wait for one additional participant to arrive. The experimenter had the participant sign various forms, then excused herself from the participant’s room and went back into the adjoining room. The experimenter picked up the phone dialed again, and then engaged in a brief mock conversation during which the participant could hear her say, “Oh, they’re here? Great! Should I start? (pause) Okay, talk to you soon. Bye.” The purpose o f the telephone embellishments was to strengthen the believability o f the cover story assertion that there were other real participants in other rooms. At this point, the experimenter went back into the participant’s room and informed the participant that the study was ready to begin. It was explained that there were three other people in the other rooms, some o f whom had been recruited through the normal subject pool and others whom had been recruited through a newspaper advertisement. The intent of this embellishment was to prevent participants from making immediate assumptions about the “other participants” based upon their membership in the USC psychology subject pool. For example, we hoped that this would make it more plausible that one o f the other participants might be an older adult, since there are not normally a large number of older adults in the subject pool at USC. Next, the participant was told that he or she was participating in a study designed to investigate how people interact with people they had just met 17 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. when they needed to come to an agreement on solutions to simulated interpersonal problems. The participant was then told that the experimenters had previously found that they could collect the most useful data from observations o f discussions when the people engaged in those discussions felt comfortable with each other and that therefore there would be an opportunity for the participant to choose one or two partners. However, it was also explained that the participant would only receive a limited amount of information about the other participants because the experimenters were also interested in studying how the amount o f information participants received about their partners before they met would affect their future discussions. It was explained that this was also the reason that the participants had been separated into different rooms for the time being. At this point the participant was asked if there was any reason to believe he or she knew any o f the other current participants in the study. Since the other participants were in fact bogus, the participants invariably answered “no”. Next, the participant was asked to fill out an information form so that information could be given to the other participants. The form, created for this study, asked participants to categorize themselves along sixteen different category dimensions. This form can be found in Appendix A. After categorizing themselves on each dimension, participants were asked to rate how important each of the categorizations were to their “sense o f who they are” and to rate this on a one to seven scale ranging from “not at all” to “very much”. 18 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Manipulation o f target type. The experimenter went to another room and created three descriptions of bogus participants (targets) by using the information provided by the real participant. The experimenter first identified a set o f categories that were rated equally high and a set that that were rated equally and relatively low. For example, the experimenter might choose all o f the categories with an importance level of four for the “high” category and all o f the categories with the importance level of two for the “low” category. Which category was chosen was based upon an algorithm that maximized the number of available categories first and the difference in importance ratings second. The benefit o f maximizing the number of categories was to camouflage the nature of the target descriptions. For example, if only one important category dimension and three less important category dimensions are used, the descriptions for a participant who is young (important category), religious, liberal, and a psychology major (less important categories) prior to counter-balancing might be: An older adult and religious. An older adult, religious, and liberal. An older adult, religious, liberal, and a psychology major. As can be seen, when presented in this way, the pattern of target types (Oi, Oii, Oiii), and thus the nature o f the study, becomes quite readily apparent. On the other hand, if more categories are used, then the target description for the same participant who is also a U.S. citizen, single, (important categories) and a commuter (less important category) might become: 19 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. An older adult and liberal. Not a U.S. citizen, religious, and a commuter. Married, a psychology major, religious, and liberal. As can be seen, although the three target types (Oi, Oii, Oiii) can still be discerned, it is considerably more difficult to recognize the pattern of target types with the additional categories. In addition, participants received target descriptions in which the important out-group occurred in different positions within each target description and the order in which the three target types were presented was also varied. This served to both counterbalance against order effects and to further camouflage the nature o f the target types. These descriptions were hand-written on a form with spaces for up to ten pieces of information per target in order to conform to the cover story assertion that the amount o f information participants were receiving was randomly determined. Counterbalancing was performed using a computer program written specifically for this task. This BASIC-language program is similar to one written by the author and used previously (Urada and Miller, 2000). The program randomly selects the condition (important in-group or out-group), the order o f the target presentation, and the order o f the category information within each target description. Note that whether targets were described as in-group or out-group members for particular categories was randomly determined. Also, the categories that were used and placed together for any given target description were randomly 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. selected from all appropriate alternatives. Thus, this paradigm provides a method o f studying real group bias on a general basis while controlling for idiosyncratic interactions that may occur with specific combinations o f group memberships. Dependent measures. After creating the target descriptions, a process that took only a few minutes, the experimenter returned to the participant’s room and gave the three target descriptions to the participant along with the following instructions: You will be cooperating to come to a consensus on solutions to different problems with one o f these people. No special knowledge or skills will be needed to solve these problems. It is only necessary that you two will be able to work carefully together. I know you don’t have a lot o f information to go on, but please try to do the best you can based on the information that you do have. Remember, try to choose someone who you think you will feel comfortable with. The participant was then given the dependent measure, which was a seven- point Likert-type scale asking how much the participant wished to be partners with each described target. The endpoints on this scale were labeled “not at all” and “very much.” Participants were also asked to rank the targets (1-4) in order of preference. These measures are shown in Appendix B. Mood measure. After the participant completed the dependent measure, the experimenter collected it and said that she would go meet with the “other 21 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. experimenters” to pair the participants up. In the meantime, she explained, she would appreciate it if the participant completed one more form. At this point the mood measure was given. The measure was Watson, Clark, and Tellegen’s (1988) Positive Affect / Negative Affect Scale (PANAS). After a few minutes, the experimenter returned and probed the participant for suspicion. Following this, the participant was fully debriefed and excused. Results Wherever possible, modem, robust statistics were used to analyze the data. Contrary to popular belief, hundreds of articles in statistical journals have pointed out that standard techniques such as analysis o f variance and standard T-tests can be highly misleading under even very small departures from normality (Wilcox, 1998). It is well established that under most conditions more modem techniques, such as trimmed mean comparisons, maintain clearly superior performance in terms of power, control over Type I error, and vulnerability to violations o f normality and equality o f variance (Wilcox, 1996; 1998). The mean preference ratings for the Oi, Oii, and Oiii targets are shown in Table 2. The effect o f target type was significant (Tt(1.68, 21.84) = 4.86, p < ,05)2. Pairwise comparisons revealed that the Oii target was rated higher than the Oi target (Tt (1.13)=9.39, g < ,01)3 and that the Oiii target was rated higher than the Oi target (Tt (1, 13)=5.79, g < .05). However, the Oii and Oiii target ratings did not differ significantly (Tt (1, 13)=0.34, g = .57). 22 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2. Study 1 target preference ratings. Target Oi Oii Oiii 20% trimmed m ean 3.29a 4.50b 4.79b estim ated standard error 0.39 0.18 0.43 n 22 22 22 Cell means followed by different letters differ p<05, two-tailed. Although inspection of the correlations between positive affect and target preference ratings shows consistent positive correlations (r_= +.23, r = +.36, r = +.31 for the Oi, Oii, Oiii targets, respectively), none of these correlations achieved statistical significance. Neither statistical significance nor consistently positive or negative correlations were evident in the correlations between negative affect scores and target preference ratings (r = -.25, r = +.06., r = -.34 for the Oi, Oii, Oiii targets, respectively). Experiment 2 Method Participants. Twenty two undergraduate psychology students at the University of Southern California volunteered for the study in order to receive extra credit points in their psychology classes. No participants were removed for suspicion. Design. The study had a single factor (Target Type: Io, loo, Iooo) fully within- subjects design. The order of the target type presentation and the order in which important/less important information within each target description was presented 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. were randomly determined by the experimenter, who was blind to the study's hypotheses. Materials and Procedures. Aside from the differences in the target types used, the materials and procedures were identical to those used in Study 1 . Results The mean preference ratings for the Io, loo, and Iooo targets are shown in Table 3. The effect o f target type was not significant (T, (2, 24) = 0.59, g = .56). Pairwise comparisons revealed no significant differences in target evaluations. Table 3. Study 2 target preference ratings. Target Io loo Iooo 20% trimmed m ean 4.23a 4.67a 4.46a estim ated standard error 0.33 0.33 0.18 n 25 25 25 Positive affect scores were correlated with the preference ratings for the Iooo target (r = +.49, g < .05), but not with the Io or loo targets (r = -.08, g = .73; r = -.20, g = .39, respectively). None o f the target preference ratings were correlated with negative affect scores, nor was any trend apparent. Discussion The tendency for participants to differentiate between targets with additional in-group memberships (viz. Oi, Oii, Oiii) but not between targets with additional out- 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. group memberships (viz. Io, loo, Iooo) lends support to an in-group favoritism model of processing. That is, it appears that participants are basing their target preference ratings more on the in-group content o f the target descriptions than the out-group content. This also argues against an explanation based merely on the amount of information. That is, the patterns o f ratings cannot be explained as simply being due to increased information about the targets (e.g. individuation), since target evaluations only improved with the addition o f one type o f information (i), but were unaffected by the addition of a different type o f information (o). The exact type o f information integration used by participants is more open to debate. An algebraic integration model may fit the data, but failure to find a difference between preference ratings of the Oii and Oiii targets leaves open the possibility that a non-algebraic model such as feature detection or dominance may be also be in use. It is difficult to interpret a failure to find a difference in the face of relatively low statistical power. For this reason, the following studies will attempt to replicate these results while expanding the investigation in other areas. The relationship between affect and target preference ratings in these two studies was unclear at best. Affect has previously been found to moderate target preference ratings (Urada & Miller, 2000), but these earlier studies actively induced affect, whereas here the affect participants were already experiencing was merely measured. It is likely that affect’s impact as a moderator depends upon its intensity, and that the generally neutral affect encountered in the current studies may have been insufficient to produce a measurable moderating effect. 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In Study 3, we will add single-category targets (I and O) as anchors. Averaging models predict the greatest change in preference when crossing dominant categories with a second group (e.g. O < Oi). Adding additional groups should then have a lesser impact (e.g. Oi < O ii, but the difference in preference should be less than that between O and Oi). Non-algebraic models do not make this prediction. Instead, feature detection and dominance share similar predictions that assume an underlying all-or-none decision rule, with people "meta-categorizing" targets into “in-group like” and “out-group like” categories. Crossed targets that are categorized as in-group like will be rated as highly as the I target, and that targets categorized as out-group like will be rated equally with the O target. These predictions will be tested in Study 3. Experiment 3 A replication o f the first two studies using random between-subjects assignment to target conditions (Oi/Oii/Oiii or Io/Ioo/Iooo) was carried out. The array of targets presented was also expanded to include single in-group (I) and single out-group targets (O). Part way through the study, an extra dependent measure assessing participant-target perceived similarity was also added. Method Participants One hundred and four participants from five introductory psychology classes at Glendale Community College participated for extra credit points. Two 26 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. classes were taught by the author and three were taught by a colleague. The data from five participants were removed because these participants were able to correctly guess at least part o f the study’s true purpose. The data from four participants were removed due to experimenter errors in the preparation o f the target description forms. Design This study used a 2 between subjects (dominant out-group, dominant in group) X 5 within subjects (target descriptions containing one, two, or three nondominant categories, in-group only, out-group only) design. Materials and Procedures The materials and procedures used for study three were very similar to those used for studies one and two, but minor modifications were needed in order to adapt the study to a classroom setting and for a different college. The study was run over a two semester period. For each class, two sessions were needed. During the first session, the class was given the following instructions: The forms you are about to fill out are for a study on team problem solving. It is important that we put together teams with the right mixture o f people, and the information you will be asked to provide here will help us identify partners that are appropriate. The team task, about which you will be given more information later, will take place next week. Although the information you are asked to provide 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. here is fairly harmless in nature, if you don’t feel comfortable answering some o f these questions then please feel free to skip those. Participants were then given a modified version of the participant information forms used in the first two studies. The categories used were adapted to better match student life at Glendale College and the instructions were clarified to facilitate group administration. The modified version o f this form can be found in Appendix C. During the second session, which took place one week later, the same general cover story used in studies one and two was repeated. That is, participants were told that they would be engaging in a discussion task with a partner. They were also told that the experimenter had selected several partners that would be appropriate for the purposes of the study, but that they were being given a chance to select partners that they would feel comfortable with. It was also explained that they would be given different amounts of information about different potential partners because the experimenters were interested in how this would affect their discussions. Next, each participant was given a list of “potential partners” (actually prepared from information collected from each participant the previous week) and rated how much they wished to be partners with each. Simple I and O targets were added to the array o f targets used in previous studies, so participants were randomly assigned to receive either descriptions of I, Io, loo, Iooo, and O targets or to receive descriptions o f I, Oi, Oii, Oiii, and O targets. After rating their preferences for each target, participants were given the PANAS mood measure and in addition were asked to rate, “How similar do you think you and participant #X are?” for each target. Finally, all 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. participants were asked to write a paragraph detailing what they believed the study was about and what they expected to happen at this point. Since participants in this setting could not be probed for suspicion through individual interviews, these written paragraphs were used as a check instead. After all participants had completed all forms, the class was debriefed as a group and dismissed. Results The mean target preference ratings for this study can be found in Table 4. Where an important in-group was present among the crossed targets, no significant effect of target type was found (Tt (1.91, 53.45) = .24, p = .78). That is, statistically the pattern was Io=Ioo=Iooo. Pairwise comparisons also revealed no differences between ratings of these crossed targets, confirming the conclusions o f the omnibus test. The Io target had ratings identical to the I target (T, (1, 28) = 0, p = 1.00), and the average preference rating o f the three crossed targets also did not differ significantly from that o f the simple I target (Tt (1, 28) = 0.97, p = .33). However, the average preference rating was marginally higher than that of the simple O target (Tt (1, 28) = 3.16, p = .09). Thus, including marginal differences, the pattern can be summarized as I=Io=Ioo=Iooo>0. Table 4. Study 3 target preference ratings. Dominant I condition Dominant O condition Target I Io loo Iooo O O Oi Oii Oiii I 20% trimmed mean 4.34a 4.34a 4.14a 4.10a 3.52b 3.50a 3.67a 4.23b 4.47b 4.53b estimated std. error 0.26 0.28 0.27 0.37 0.29 0.27 0.27 0.21 0.27 0.28 N 47 47 47 47 47 48 48 48 48 48 Cell means followed by different letters differ, p<=.10, two-tailed. 29 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Where an important out-group was present, a significant effect of target type was found among the crossed targets (Tt (2, 58) = 3.57, p < .05). Pairwise comparisons revealed a marginal difference between Oi and Oii target ratings (Tt (1, 29) = 3.18, p = .09), but not between Oii and Oiii target ratings (Tt (1, 29) = 0.60, p = .44). The I target differed from the Oi (Tt (1, 29) = 5.28, p < .05) and O (Tt (1, 29) = 9.20, p < .01) targets only, whereas the O target notably did not differ from the Oi target (Tt (1, 29) = 0.39, p = .54) but was significantly lower than all other targets. Thus, the obtained pattern can be summarized as I=Oiii=Oii>Oi=0. Participants’ ratings o f the degree to which they perceived the targets as similar to themselves paralleled their preference ratings. In fact, for all crossed targets, there was a significant positive correlation between the similarity ratings and the preference ratings (range: r=.44 to r=.67, p < .05). When an important out-group was present among the crossed targets (Oi, Oii, Oiii), there was an effect o f target type on the similarity ratings (Tt (1.88, 26.33) = 7.60, p < .01), meaning participants differentiated among the targets. When an important in-group was present among the crossed targets (Io, loo, Iooo), there was no effect of target type on the similarity ratings (It (1.80, 30.57) = .56, p = .56). Once again, negative affect was not correlated with any target preference ratings. Positive affect was correlated with preference ratings o f the I target in the dominant O condition (r = +.29, p < .05), but not in the dominant I condition (r = +. 14, p = .34). Positive affect was also correlated with preference ratings for the loo 30 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. target (r = +.40, g < .01), and was marginally correlated with preference ratings for the Iooo target (r = +.25, g = .09). Discussion The inclusion o f the I and O targets helped clarify the type o f processing being used by participants in these studies. Again, support was found for an in-group favoritism model, but the results contrast with the predictions of algebraic models. According to averaging models, the strongest difference in preferences for targets with a dominant out-group (O) should have been found between the O and Oi targets. When averaging information, additional in-group information should have the greatest impact when added to a pure out-group description (viz. going from O to Oi), whereas adding further in-group information (e.g. going from Oi to Oii) should have a smaller effect on preference ratings. In fact, the opposite was found, with no difference found between O and Oi preference ratings, but a significant difference between Oi and Oii preference ratings. Even an additive model (as opposed to an averaging model) predicts a difference between O and Oi equal to that between Oi and Oii, but this again does not fit these results. There was also no difference between the Oii and Oiii targets, where an algebraic model would predict one. Averaging models and additive models with diminishing returns would predict a smaller difference between Oii and Oiii targets than between the Oi and Oii targets, so it is possible that this lack of difference is due to poor power. Still, taken together, the evidence weighs against algebraic models and points to feature detection or dominance as the theories providing the best fit to the data. 31 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The finding that the similarity ratings paralleled the preference ratings suggests that under these circumstances participant attention is directed toward in-group information. When in-group memberships (i) are added to the description of an strong out-group member (O), participants not only show greater preference for the targets, but are also more likely to see them as being more similar to themselves. However, when out-group memberships (o) are added to the description of a strong in-group member (I), this not only failed to have a significant impact on preferences, but it also did not affect participants’ perceptions of how similar the targets were to themselves, suggesting that participants largely disregarded this information altogether. The correlations between affect and target preference ratings suggest a possible relationship between positive affect and crossed targets with a dominant I membership, especially the loo and Iooo targets. The loo finding is opposite to the sign of the correlation found in Study 2, but the Iooo relationship replicates Study 2 findings. Urada & Miller (2000) previously found a relationship between positive affect and preference ratings for Io targets, but their theorizing predicts a greater likelihood of a relationship with Io targets rather than loo or Iooo targets. Thus, although the current results suggest that moderators studied previously within the two-group model can extend to multiple-category targets, at the same time there is ambiguity over whether these moderators might interact with the multiple crossed- category paradigm in unanticipated ways. These moderators may need to be investigated anew under multiple crossed-category conditions, and should be 32 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. performed with actual mood inductions. Once again, the fact that affect was only measured here and not induced likely contributed to the ambiguous results. Without an induction, the relatively strong affect that may be needed to produce moderating effects was unlikely to be present in most participants. If we accept in-group favoritism as the model that best fits the current results, the question o f its generality is a natural next issue. Why are participants focusing on in-group information? One possible answer lies in the nature of the task. In studies one through three, the primary dependent measure for the previous studies was the question “how much do you wish to be partners with . . .”. Participants were also instructed to choose a partner they “will feel comfortable with”. In the assessment o f similarity between objects (e.g. crossed targets and in-group members) participants are thought to attend more to their common features, whereas in the assessment o f difference between objects participants are thought to attend more to their distinctive features (Tversky, 1977, p. 339). Thus, when asked whom they prefer, this framing might cue participants to search for similarities (in-group memberships) they share with the targets. On the other hand, if asked the reverse question, who they LEAST wish to be partners with, participants may be cued to search for dissimilarities (out-group memberships) between themselves and the targets. This could in turn affect the process o f information integration, leading to an a focus on out-group information rather than in-group information. For this study, half o f the participants received a dependent measure that is worded, “How strongly do you prefer to exclude participant # ...” and these 33 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. participants were advised that for the purposes o f the discussion, it would be best to eliminate potential partners with whom they might feel uncomfortable. For comparison purposes, the other half o f the participants received a dependent measure and instructions paralleling those used in the previous studies. That is, the dependent measure asked “How strongly do you prefer to include participant # ...” and the instructions advised participants to choose partners with whom they would feel comfortable. The outcome of this study should indicate whether in-group favoritism is the norm or whether out-group derogation could occur under the right circumstances. If the task dictates which type of group focus occurs, then a focus on out-group exclusion will result in an outcome matching the out-group derogation prediction, and thus reverse previous findings. That is, the patterns Oi=Oii=Oiii and Io>Ioo=Iooo are expected. This prediction assumes the same type o f non-algebraic processing found previously, but with a focus on out-group memberships. If, on the other hand, in-group favoritism is a robust phenomenon across task situations, then an outcome similar to that found in Study 3 is expected. Experiment 4 Method Participants. Seventy one undergraduate psychology students in three classes at Glendale Community College and one class at the University o f Southern California participated in return for extra credit points in their psychology classes during 34 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the Spring semester o f 1999. Nine participants were removed after they expressed suspicion and were able to partly guess the true nature of the experiment. Design. This study had a 2 between subjects (inclusion instructions, exclusion instructions) X 2 between subjects (dominant out-group, dominant in-group) X 5 within subjects (target descriptions with one, two, or three nondominant categories, in-group only, out-group only) design. Materials and Procedures. The cover story, procedures, and materials paralleled those used in Studies 1 and 2 for the inclusion conditions. Modifications were made in the instructions and the primary dependent measure for the exclusion conditions. In these conditions, participants were told that they would be engaging in a discussion with other participants but that due to the importance of avoiding an uncomfortable discussion setting, they would be able to specify participants that they would prefer not to be partners with. In order to ameliorate an expected reluctance to exclude targets, the experimenter emphasized that past research clearly showed that the discussion must be conducted with partners that do not feel uncomfortable with one another. Participants were then presented with a list of prepared target descriptions and were asked to rate how hesitant they were to have each target as a partner. Participants were then probed for suspicion and debriefed. 35 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Results Dominant In-group The mean target preference ratings can be viewed in Table 5. Where an important in-group was present among the crossed targets and participants were asked to indicate who they preferred to include as a partner, once again there was no effect o f target type on the crossed target preference ratings (Ti (1.92, 21.12) = 0.39, g = .68). That is, statistically the pattern was Io=Ioo=Iooo. When participants were asked instead who they preferred to exclude as a partner, again no differences were found among the crossed targets (Tt (1.90, 22.79) = 0.27, g = .75). Pairwise comparisons also revealed no differences between any target ratings in either the inclusion or exclusion conditions. Dominant Out-group Where an important out-group was present and participants were asked to indicate who they preferred to include as a partner, no effect o f target type was found among the crossed target preference ratings (Tt (2, 16) = 1.96, g = . 17). Pairwise comparisons revealed a significant difference between I and O targets (Tt (1, 8)=6.44, g < .05), but no other differences were found. When participants were asked who they preferred to exclude as a partner, no differences were detected among the crossed targets T, (1.78, 16.05) = 1.05, g = .36. 36 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5. Study 4 target preference ratings. Inclusion Instructions Dominant I condition Dominant O condition Target I Io loo Iooo O O Oi Oii Oiii I 20% trimmed mean 4.92a 4.67a 4.58a 4.33a 4.00a 3.38a 3.75ab 4.38ab 5.00ab 5.13b estimated std. error 0.48 0.32 0.20 0.37 0.30 0.63 0.40 0.43 0.82 0.43 n 18 18 18 18 18 12 12 12 12 12 Exclusion Instructions Dominant I condition Dominant O condition Target I Io loo Iooo O O Oi Oii Oiii I 20% trimmed mean 1.18a 1.18a 1.27a 1.27a 1.72a 2.44a 2.90a 2.56a 2.11a 1.22a estimated std. error 0.19 0.19 0.2 0.2 0.72 0.72 0.95 0.56 0.38 0.21 n 17 17 17 17 17 15 15 15 15 15 Cell means followed by different letters differ, p<= 10, two-tailed. Inclusion instructions: higher ratings indicate stronger preference to include the target. Exclusion instructions: higher ratings indicate stronger preference to exclude the target. Discussion Drawing conclusions from this study alone is difficult given the small number o f participants (12-18) in each of the study’s four between-subject conditions. In all likelihood the resulting low statistical power contributed to the non-significance of the results and the failure, in some instances, to statistically replicate the previous studies. One unfortunate consequence of expanding this study to include twice as many conditions as Experiment 3 is the ensuing need for greater numbers of participants. In order to obtain larger numbers of participants, a new approach became necessary. Experiment 5 introduces an internet-based method of running the study. 37 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Experiment 5 Method Participants. One hundred sixty six participants took part in the study online. O f these, 64 were USC students and received extra credit points in their psychology classes. The remaining 102 participants received entries into two separate drawings, one to win $50 (1 in 25 odds) and one to win $100 (odds unspecified). Ten participants were removed after they expressed suspicion and were able to partly guess the true nature of the experiment. Design. This study had a 2 between subjects (inclusion instructions, exclusion instructions) X 2 between subjects (dominant out-group, dominant in-group) X 5 within subjects (one, two, or three nondominant categories, in-group only, out-group only) design. Due to a somewhat greater interest in achieving strong power in the inclusion instruction condition, the random assignment mechanism was set to assign approximately 60% o f participants to the inclusion instructions condition and 40% to the exclusion instructions condition. Materials and Procedures. This study was conducted on the internet (World Wide Web) via Participants Online (http://wwrw.ParticipantsOnline.com), a web site designed for online social science research. Participants Online was founded and is operated by the author. This study was publicized by posting a sign up sheet at USC (as is standard 38 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. procedure for all studies), purchasing and displaying 5,000 internet “banner ads” per month on other web sites, posting to internet discussion groups, registering with internet search engines, and by receiving a mention in a sweepstakes newsletter. Additional participants may have also arrived through links from outside web sites, such as the Chronicle o f Higher Education’s site, which featured Participants Online while the study was running. The cover story and procedures in this study closely paralleled those used in study 4, but were modified for online use. Printouts of what participants saw from are shown in Appendix D. Participants initiated participation by clicking on a link at the Participants Online site. They were then taken to a page with a consent form very similar to those used in the previous studies, only in an online format. Participants registered for the study by completing a form with information about themselves, then clicked a button to acknowledge that they had read the consent form and the site’s rules and agreed with both. All information submitted by participants was secured using 128-bit encryption and authentication, the strongest security available anywhere for commercial use. Participant registration information was also received and stored in a file separate from the participants’ responses to experimental questions, ensuring a higher level of privacy. After completing registration, participants were taken to a cover story page. Here, participants were told that they would soon be engaging in a “chat session” with other participants that were online at the same time. Chat sessions are a popular method o f nearly instantaneous text communication between people online. 39 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. To make it seem more plausible that other people were really online simultaneously, the web page printed the following message: If enough people are logged on right now, a "proceed" button will appear at the bottom o f this page that will allow you to participate in the experiment. Otherwise, please wait and the page will reload automatically in a few moments. Generally the greatest number of people log on between 1:00 PM and 11:00 PM PST, but we have had successful sessions at all hours o f the day & night! Next, the web page printed the current time and date, followed by the sentence “There ARE enough people logged on to begin the experiment”. In reality, participants received this message regardless of whether there were other people simultaneously logged on or not. The cover story continued with a bogus example of a chat room interface, followed by a button marked “PROCEED”. When participants clicked on this button, they were taken to the next page in the study. The next page consisted o f instructions and a form that paralleled the Participant Information Sheet used in previous studies. The instructions matched those used in previous experiments in text form. On the form, participants indicated whether they were, for example, a young adult or an older adult, politically more liberal or conservative, etc. They were also asked to indicate on a one to five scale how important each o f these group memberships are for their self identity. At the bottom of the page was a button marked “ALL DONE WITH THESE QUESTIONS!” Clicking this caused a program to run that checked the 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. participants’ responses. If the participant did not give enough information, they were given a message requesting that they check their answers and provide more information. If the participant had provided enough information to proceed, the program randomly assigned participants to conditions and generated an appropriate set of instructions and a list o f appropriate descriptions o f the bogus “other” people. Participants were asked to provide preference ratings for each target, then click on the “pair me up” button at the bottom of the page Upon clicking on the button, participants were taken to a page with the statement “Waiting for all participants to reply. Before we proceed, we would like to get an idea of how you perceive this study. In a few words, please describe what you think this study is about and what you expect to happen next.” Below these instructions was a box in which participants could type their answer. This served as the online version o f a probe for suspicion. Since the box could scroll up and down, there was no limit to the length at which a person could respond. After completing this task, participants could click on the “Continue” button at the bottom of the page, which would take them to the final page. The final page provided a full debriefing. If participants had any questions about the study, they could write to the author by clicking on a link to the author’s e- mail address. 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Results Dominant In-group The mean ratings can be found in Table 6. When participants were asked to choose targets to include on their teams, no discrimination occurred among crossed targets with dominant in-groups, Tt (1.90, 60.75) = 1.65 g = .20. Average ratings of the crossed targets were significantly below those of the I target, t( 1, 32) = 14.63, g < .001, but did not differ from ratings o f the O target. T, (1,31) = .84, g = .37. Thus, statistically the results correspond to the pattern I > Io = loo = Iooo = O. Participants who were asked to exclude targets from their teams also did not differentiate among the crossed targets with dominant in-groups, Tt (1.84, 31.29) = .82, g = .44. The average ratings for these crossed targets was significantly lower than the ratings of the I target, Tt (1, 17) = 14.93, g < .01, but did not differ from ratings o f the O target, Tt(l, 17) = . 12, g = .73. Thus, statistically, the results could be summarized as I > Io = loo = Iooo - O. Dominant Out-group When participants were asked to choose targets to include on their teams, a significant effect of target type was found among the crossed targets T, (2.00, 58.00) = 3.39, g < .05. Pairwise comparisons revealed a pattern corresponding to O = Oi < Oii = Oiii =1, with a marginal difference between ratings of the Oi and Oii targets Tt (1,29) = 2.84, g = . 10. Participants who were asked to exclude targets from their teams did not differentiate among crossed targets with dominant out-groups, T, (1.38, 42 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. 19.36) = 2.45, g = .13, and the composite ratings for the crossed targets also did not differ from the ratings o f the O target Tt (1, 14) = .01, g = .91. However, the composite rating o f the crossed targets were lower than ratings o f the I target, Tt (1, 14)= 16.57, g < .01. Thus, statistically, the results can be summarized as 0= 0i= 0ii= 0iii< I. Table 6. Study 5 target preference ratings. Inclusion Instructions Dominant I condition Dominant O condition Target I Io loo Iooo O O Oi Oii Oiii I 20% trimmed mean 4.69a 4.12b 4.03b 3.64b 3.69b 3.8a 3.83a 4.33b 4.63bc 5.23c estimated std. error 0.20 0.28 0.21 0.27 0.26 0.28 0.28 0.27 0.27 0.27 n 53 53 52 53 52 50 50 50 50 48 Exclusion Instructions Dominant I condition Dominant O condition Target I Io loo Iooo O O Oi Oii Oiii I 20% trimmed mean 1.06a 1.72b 1.94b 2.28b 2.22b 2.47a 2.07ab 3.00ab 1.67bc 1.20c estimated std. error 0.13 0.27 0.39 0.48 0.41 0.39 0.28 0.73 0.29 0.16 n 28 28 28 28 28 25 25 25 25 25 Cell means followed by different letters differ, p<=. 10, two-tailed. Inclusion instructions: higher ratings indicate stronger preference to include the target. Exclusion instructions: higher ratings indicate stronger preference to exclude the target. Discussion The task appeared to have an impact on the way participants processed target information and reported their preferences. In both o f the conditions with exclusion instructions, participants chose to exclude any target with out-group information, and generally disregarded any in-group information this may have been combined with. There was a nonsignificant improvement for the Oiii target, suggesting that adding a 43 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. great deal o f in-group information may eventually overcome the presence o f an out group membership, but clearly the threshold for doing so was higher here than in conditions with inclusion instructions. When inclusion instructions were used in this study, the results generally paralleled those obtained in the previous lab studies. The exception is that the crossed targets in the dominant in-group condition were depressed to the level o f the O target (I > Io = loo = Iooo = O), whereas in previous studies these were rated at the same level as the I target (I = Io = loo = Iooo > O). These results fit a feature detection strategy well. It appeared that participants favored anyone who was more in-group than out-group on a simple category- counting basis (i.e., I, Oii, Oiii), whereas everyone else (Oi, Io, loo, Iooo) was lumped into the same "out-group like" category as the O target. The between-study difference, (grouping the Io, loo, and Iooo targets with the O target rather than with the I target) may be due to the setting. A meta-analysis comparing different methods o f questionnaire administration (Richman, Kiesler, Weisb, & Drasgow, 1999) found that when participants were alone, computer administration led to less social desirability distortion than did paper-and-pencil administration. Thus, when taking part in the study online, participants may have felt less pressure to show socially desirable tolerance, and thus may have had a lower threshold at which they decided to discriminate against non-preferred targets. The result was that all targets that did not have a majority in-group component received a lower rating. In a more intimate setting, such as in the lab, participants may have felt implicit pressure to be more tolerant due to the presence of the experimenter and the perception that the 44 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. targets were present nearby. This pressure to be tolerant may have effectively raised the threshold for excluding targets (or lowered the threshold for perceiving others as preferable partners), resulting in greater preference for any target with an I membership. Admittedly, this is a post-hoc explanation based on between-study comparisons, but a plausible one nonetheless that should be investigated further with experimental controls. This has potential ramifications outside o f crossed categorization as well, as it suggests that the normally "color blind" online world can sometimes witness even greater discrimination than the real world. Meta analysis Although there is a fair amount o f between-study consistency in results among this series o f studies, individually each study, particularly the earlier ones, leave questions open due to questions of statistical power. However, over the course of this series o f studies, new methods were harnessed to run greater numbers of participants, and in total nearly 400 participants took part in this research. Since the studies took place at different times, with different samples, and with slightly differing methods, it would be inappropriate to directly combine the data from all o f these participants, but there is enough similarity among these studies to examine the combined results meta-analytically. Methods The meta analysis was carried out by calculating effect sizes from each study, weighting these effect sizes by the inverse o f their variance (giving greater 45 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. weight to studies with larger sample sizes), then combining the appropriate effect sizes with standard meta-analytic formulae. Effect sizes were calculated using D- Stat software from standard (untrimmed) paired-samples t-tests performed between adjacent target types (e.g. Io and loo, or loo and Iooo). Where no differences were found between the crossed targets, the composite (average) crossed target scores were compared to ratings for the simple category targets (I and O). Because only studies 4 and 5 included dependent measures that asked participants to exclude people from their team, this measure was excluded from the meta analysis. Thus, this meta analysis only combines data from the “include” question from all five studies. Results No differences were detected between crossed category targets with a dominant in-group (Io=Ioo=Iooo). I targets were preferred over the composite crossed targets, D = .36, p < .01, but marginally significant between-studies heterogeneity was also found, Q(2) = 4.716, g = .09, indicating the presence of an outlier. The outlier was identified as a much larger difference between preference ratings for the I target and preference ratings for the crossed targets in study 5 (D = +0.44, g < .05) than in the other studies (D = + 0.04, g = .82). Excluding Study 5, there was no difference between the I target preference ratings and crossed target preference ratings, D = +. 11, g = .55, and no heterogeneity, Q (l) = .01, g = .92. The crossed targets were also rated marginally higher than the O targets both with Study 5, D = .24, g = .07, and without study 5, D = .34, g =.05. Thus, without the study 5 outlier, the 46 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. pattern of results corresponds to I=Io=Ioo=Iooo>0. The average target ratings for Studies 1-4, weighted by study size, are shown in Figure 1. Including study 5, the crossed targets are rated marginally lower than the I target and marginally higher than the O target, I>Io=Ioo=Iooo>0. Figure 1 . Weighted mean target preference ratings for studies 1-4. Dominant Outgroup Dominant Ingroup 4 study weighted average 4 study weighted average Among targets with a dominant out-group, Oii targets were preferred over Oi targets, D = -.38, p < .01, but the Oi target ratings did not differ from those of the O target, D = -0.02, p = .87. The difference between Oi and Oii target ratings was also greater than the difference between O and Oi targets (Oii - Oi > Oi - O), D = -0.28, p < .05. The Oii ratings also did not differ from Oiii ratings (D = -.07, p = .56), and the Oiii ratings in turn did not differ from I ratings (D = -.21, p = . 12). Ratings of the I target relative to the crossed targets were again highest in Study 5, but this 47 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. time the between-studies heterogeneity was not significant ((2(2)= 1.65, g=.44). Heterogeneity was also absent for all o f the other comparisons. Thus, overall the pattern was 0= 0i<0ii= 0iii=I. General Discussion The strongest evidence against an algebraic model can be found in inclusion instruction conditions, where there was a significantly smaller difference between O and Oi targets compared to the difference between Oi and Oii targets. Averaging models and most additive models predict the opposite result. Even a simple additive model predicts a monotonic increase in preference each time an in-group (i) is added. Instead, the effect size for the Oi and Oii comparison was 19 times higher than the effect size for the O and Oi comparison. Algebraic processing models cannot account for this data, but non-algebraic ones can. In these conditions, participants appeared to utilize a threshold-based decision method based on the numerosity of target in-group memberships. That is, targets that did not have a numerical majority o f in-groups (O, Oi) were not preferred, while ail other targets (Oii, Oiii, I) were preferred. If a target was similar enough to the participant, the target was meta categorized as "in-group like" and was preferred as strongly as the simple in-group (I) target. If the target was not similar enough to the participant, the target was meta categorized as "out-group like" and was seen as the equal o f the O target. When participants received inclusion instructions and target information included a dominant I, the threshold for "in-group like" appeared to have 48 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. been surpassed, as these targets were preferred as highly as the I target, even when up to three non-dominant out-group memberships were added (Iooo). The exception to this came in Study 5, which demonstrated the power of the situational context to change the threshold at which targets were seen as "in-group like" or "out-group like". When participating online, social desirability pressures may have been reduced, which resulted in participants changing their thresholds for publicly rating targets as “acceptable” . This in turn led to lowered ratings for all targets with any out-group membership, as can be seen by comparing the crossed target means in Study 5's include/dominant in-group condition to comparable conditions in the other studies. Thus, online, the existence o f any out-group membership in a target was sufficient to eliminate preference for that target. The type o f question participants were asked also influenced their processing. As described above, when participants were asked to choose targets to include on their team, they appeared to make choices based on an in-group favoritism basis. That is, participants generally evaluated crossed targets based on their in-group memberships and ignored out-group information. However, when asked to choose targets to exclude from their team, the opposite occurred. Out-group derogation appeared to take over, and any target with an out-group membership was penalized, while in-group information within crossed targets was generally ignored. Under conditions o f increased importance, algebraic processing may yet occur, in accord with the dual process models. Although there was no significant effect, some trends in the data do suggest the occurrence o f at least some 49 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. algebraic processing. For example, there is a non-significant trend for the Oiii targets to be rated higher than Oii targets, which suggests that not all participants relied entirely on a threshold rule, and that some may have processed the information at least partly on a piecemeal basis. Dual process models (Brewer, 1988; Fiske & Neuberg, 1990; Fiske & Pavelchak, 1986; Pavelchak, 1989) predict that participants who see the target as particularly relevant and who feel high self-involvement are the most likely to engage in such processing. Thus, if the situation were to invoke higher levels o f these variables (for example, by telling participants that they were choosing someone that they would live with for a month), there would be a greater chance that bottom-up processing would occur. Fiske's continuum model and its predecessors (Fiske & Neuberg, 1990; Fiske & Pavelchak, 1986; Pavelchak, 1989) also predicts that participants will switch to piecemeal processing when they have difficulty categorizing a target. It seems reasonable that some participants may have had difficulty categorizing certain targets as either "in-group like" or "out-group like," particularly when these targets were both crossed and rich in information (e.g. Oiii). Thus, there is reason to believe some algebraic processing might have occurred even if it was not the dominant method used under these experimental conditions. Two trends in the data are consistent with this proposition (see Tables 2- 6). First, there is a weak tendency for targets to be rated lower each time an o category is added (e.g. Io > loo), and each time an i category is added (e.g. Oii < Oiii), even if these trends never reached statistical significance. Second, as the further categories were added, the variance of the target ratings generally 50 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. grew (e.g., the variance for the Oiii ratings was usually higher than the variance for the Oii ratings.) This is exactly what we would expect to happen if some participants were using a threshold decision making rule while others were using an algebraic rule. That is, participants using a threshold decision rule would rate Oii = Oiii, while participants using an algebraic rule would rate the Oiii target higher than the Oii target. This would raise the variability o f the Oiii ratings. Thus, while the threshold rule may have been used by most participants, there is some evidence that a few participants may have used an algebraic rule. Simpler conditions may also result in different results. For example, past experiments have found Oi targets to be preferred over O targets using similar procedures (Urada & Miller, 2000), a finding which was not replicated here. The main difference between this and earlier studies is the richer, multiple category information given about some targets. When faced with a large set of information, people often sort them into clusters to reduce information load (Tversky, 1977, p. 342). This suggests that the large amount of information given here encouraged participants to shift to more heuristic processing, whereas in a simpler situation with fewer or less complex targets, bottom-up processing may occur or the threshold for target preferences may shift, allowing Oi targets to be preferred over O targets. We can draw several conclusions from this research. Within the current experimental paradigm, there is strong evidence for heuristic, threshold-based processing o f information over algebraic combination. The exact location o f the threshold can be influenced by external factors that may include social 51 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. desirability pressure invoked by the situation. Whether participants engage in in group favoritism or out-group derogation is dictated by the task participants are asked to perform. This research also points to a number o f areas for future study. The current series of studies all utilized the same basic paradigm, but even within this paradigm changes in the context (i.e. computerized administration in Study 5, changes in question wording in Studies 4 and 5) had a notable impact on certa n results. Clearly, a number of variables may affect how intergroup information is integrated, and further studies using a wider range of paradigms are needed before broader conclusions can be reached. In particular, future research should focus on variables that have shown moderating effects in previous crossed categorization research, such as personalization, cognitive load, and positive affect (Urban & Miller, 1998). In future studies, it would also be best to manipulate potential moderators rather than assess them simply by measuring them, as was attempted unsuccessfully here with affect. It may also be productive to investigate other variables that may affect category processing, such as priming (Zarate & Sanders, 1999), and, as previously discussed, self-involvement and relevance. Feature salience and order of comparisons may also moderate target ratings (Tversky, 1977). Less prominent stimuli are often seen as more similar to a more prominent stimulus than vice versa. For example, Cuba is seen as more similar to Russia than Russia is to Cuba, due to prominence o f Russia (more unique features are likely to be known about Russia than Cuba). By extension, others may be seen as more similar to oneself than self is to others, due to the natural prominence o f the self. 52 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Alternatively, others may be seen as more similar to the in-group than the in-group is to others. Thus, the task may have an effect on target ratings, depending on which direction o f comparison it encourages. The general act o f categorization can be thought o f as involving the comparison of the features o f certain objects with the features o f potential membership categories. It is, o f course, possible to map the features of the categories onto the objects or to map the features o f the objects into the categories. The direction of mapping may well determine the outcome o f the categorization process. (Houston, Sherman, and Baker, 1989, p. 140) Thus, if a participant is primed to ask "is Bob like an in-group member?" the result may be different than if the person is primed to ask "are in-group members like Bob?" If the category is more prominent (i.e., the participant knows more unique features about the in-group category than about Bob), then the participant will be more likely to answer "yes" to the first question than the second. If real-life situations can be structured to focus attention on in-group similarities (paralleling the inclusion conditions in this set o f studies), this research suggests that the addition o f multiple in-group memberships can be used to outweigh the presence o f an important out-group membership. This can be useful when two groups have a strong or lengthy history o f conflict, making it difficult to find or create a single in group membership that is strong enough to counterbalance the out-group. In 53 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. these situations, creating or making salient multiple in-group memberships may form the basis for a productive intervention. Members of the conflicting groups could be brought together in a context in which multiple in-group memberships have been made salient or have been created (i.e. team memberships), with the goal o f facilitating improved interpersonal interactions. Creating a task that highlights certain shared group memberships might also temporarily determine the perceived importance o f those groups. However, specific contextual conditions must be met, such as an orientation toward in-group favoritism and possibly some external pressure to be tolerant. Further research must be undertaken better delineate the conditions and moderators that would ensure the success of such an intervention. 54 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Endnotes 1 Throughout this manuscript, the letter “I” will be used to denote an important in group membership while the letter “i” will be used to denote a less important in group membership. Similarly, “O” indicates an important out-group membership and “o” indicates a less important out-group membership. 2 This is a 20% trimmed mean t-test for multiple dependent groups, which tests the null hypothesis Ho: pi = p.2 = P - 3. An older, non-robust analog would be a one-way repeated measures ANOVA, which, in this case, would yield the same conclusion (F(2, 42) = 6.32, p < .01). 3 Paired trimmed mean t-tests are a more robust, more powerful update of conventional paired t-tests and can be interpreted the same way. 55 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. References Anderson, N.H. (1965). Averaging versus adding as a stimulus-combination rule in impression formation. Journal o f Experimental Psychology. 70(4). 394-400. Anderson, N.H. (1967). Averaging model analysis of set-size effect in impression formation. Journal of Experimental Psychology. 75 (2). 158-165. Anderson, N.H. (1968). Application o f a linear-serial model to a personality- impression task using serial presentation. Journal o f Personality and Social Psychology. 10 (4), 354-362. Arcuri, L. (1982). Three patterns o f social categorization in attribution memory. European Journal o f Social Psychology. 12. 271-282. Asch, S.E. (1946). Forming impressions o f personality. Journal o f Abnormal and Social Psychology. 41. 258-290. Brewer, M.B. (1979). In-group bias in the minimal intergroup situation: A cognitive-motivational analysis. Psychological Bulletin. 86(2). 307-324. Brewer, M.B. (1988). A dual process model o f impression formation. Advances in social cognition. I, 1-36. Brewer, M.B., & Lui, L. (1985). Categorization of the elderly by the elderly: Effects of perceiver category membership. Personality and Social Psychology Bulletin. 10, 585-595. Brewer, M.B., Ho, H., Lee, J., & Miller, N. (1987). Social identity and social distance among Hong Kong schoolchildren. Personality and Social Psychology Bulletin. 13(2). 156-165. Brewer, M.B., & Miller, N. (1984). Beyond the contact hypothesis: Theoretical perspectives on desegregation. In N.M. & M.B. Brewer (Eds.), Groups in Contact: The Psychology o f Desegregation (pp 281-302). Orlando, FL: Academic Press. Brown & Turner (1979). The criss-cross categorization effect in intergroup discrimination. British Journal of Social and Clinical Psychology. 18. 371-383. Craik, F.I.M., & Lockhart, R.S. (1972). Levels o f processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior. 11. 671-684. 56 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Deschamps, J.C., & Doise, W. (1978). Crossed category memberships in intergroup relations. In H. Tajfel (Eds.), Differentiation between Social Groups. Cambridge: Cambridge University Press. Dijker, A.J.M. (1987). Emotional reactions to ethnic minorities. European Journal of Social Psychology. 17. 305-325. Ensari, N., & Miller, N. (1998). Effect of affective reactions by an out-group on preferences for crossed categorization discussion partners. Journal o f Personality & Social Psychology. 75(6) 1503-1527. Eurich-Fulcer, R.; Schofield, J.W. (1995). Correlated versus uncorrelated social categorizations: The effect on intergroup bias. Personality & Social Psychology Bulletin. 21(2), 149-159. Fiske, S. T. & Pavelchak, M. A. (1986). Category-based versus piecemeal-based affective responses: Developments in schema-triggered affect.(In R. M. Sorrentino & E. T. Higgins (Eds.), The handbook of motivation and cognition: Foundations of social behavior (pp. 167— 203). New York: Guilford Press. Fiske, S., T. & Neuberg, S., L. (1990). A continuum o f impression formation, from caetgory-based to individuating processes: influences o f information and motivation on attention and interpretation. Advances in Experimental Social Psychology, 23, 1-73. Fiske, S.T., & Taylor, S.E. (1984). Social cognition. Reading, MA: Addison- Wesley. Hagendoom, L., & Henke, R. (1991). The effect o f multiple category membership on intergroup evaluations in a north Indean context: class, caste and religion. British Journal o f Social Psychology. 30. 247-260. Hastie, R. (1980). Memory for behavioral information that confirms or contradicts a personality impression. In R. Hastie, T. Ostrom, E. Ebbesen, R. Wyer, D. Hamilton, & D. Carlston (Eds.), Person memory: The cognitive basis o f social perception. 155-177, Hillsdale, NJ: Lawrence Erlbaum Associates. Hewstone, M., Islam, M.R., & Judd C.M.(1993). Models of crossed categorization and intergroup relations. Journal o f Personality and Social Psychology. 64. 779-93. Higgins, E.T. (1996). Knowledge activation: Accessibility, applicability, and salience. In E.T. Higgins & A.W. Kruglanski (Eds.) Social Psychology: Handbook o f Basic Principles (pp. 133-168). New York: Guilford Press. 57 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Isen, A.M., Niedenthal, P.M., & Cantor, N. (1992). An influence o f positive affect on social categorization. Motivation and Emotion. 16. 65-78. Jackson, L.A., & Sullivan, L.A., (1989). Cognition and affect in evaluations of stereotyped group members. The Journal o f Social Psychology. 129. 659-672. Messick, D.M., & Mackie, D.M. (1989). Intergroup relations. Annual Review o f Psychology. 40. 45-81. Migdal, M.J., Hewstone, M., & Mullen, B.M. (1998). The effects o f crossed categorization on intergroup evaluations. Unpublished manuscript? Oden, G.C., & Anderson, N.H. (1971). Differential weighting in integration theory. Journal o f Experimental Psychology. 89(1). 152-161. Ostrom, T.M., & Davis, D. (1979). Idiosyncratic weighting o f trait information in impression formation. Journal of Personality and Social Psychology. 37(11). 2025-2043. Pavelchack, M.A. (1989). Piecemeal and category-based evaluation: An idiographic analysis. Journal o f Personality and Social Psychology. 56(3). 354-363. Prinz, W., & Sheerer-Neumann, G. (1974). Component processes in multiattribute stimulus classification. Psychological Research. 37. 25-50. Richey, M.H., Koenigs, R.J., Richey, H.W., & Fortin, R. (1975). Negative salience in impressions o f character: Effects o f unequal proportions o f positive and negative information. Journal o f Social Psychology. 97. 233-241. Richman, W.L., Kiesler, S., Weisb, S., Drasgow, F. (1999). A meta-analytic study o f social desirability distortion in computer-administered questionnaires, traditional questionnaires, and interviews. Journal of Applied Psychology. 84(5). 754-775 Rogers, M., Miller, N., Hennigan, K.M. (1981). Cooperative games as an intervention to promote cross-racial acceptance. American Educational Research Journal. 18(4), 513-516. Rothbart, M., & John, O.P. (1985). Social categorization and behavioral episodes: A cognitive analysis o f the effects o f intergroup contact. Journal o f Social Issues. 41. 81-104. 58 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Schofield, J.W., & Sagar, H.(1977). Peer interaction patterns in an integrated middle school. Sociometrv. 40. 130-138. Schul, Y., & Bumstein, E. (1988). On Greeks and horses: Impression formation with social and nonsocial objects. Advances in social cognition. I , MS- 154. Sherman, R.C. (1988). Are two modes better than one? A critique o f Brewer’s dual process model, Advances in social cognition. 1, 155-163. Singh, R. (1976). Set-size effect in impression formation as a function o f set homogeneity-heterogeneity. Psvchologia. 19, 108-113. Singh, R., Yeoh, B.E, Lim, D.I., Lim, K.K. (1997). Cross-categorization effects in intergroup discrimination: Adding versus averaging. British Journal o f Social Psychology. 36(2), 121-138. Srull, T.K. (1981). Person memory: Some tests o f associative storage and retrieval models. Journal o f Experimental Psychology: Human Learning and Memory. 7, 440-463. Srull, T.K., Lichtenstein, M., & Rothbart, M. (1985). Associative storage and retrieval processes in person memory. Journal o f Experimental Psychology: Learning. Memory, and Cognition. 11, 316-345. Stangor, C., Lynch, L., Duan, C., & Glass, B. (1992). Categorization o f individuals on the basis o f multiple social features. Journal o f Personality and Social Psychology. 62, 207-218. Stephan, W.G., & Stephan C.W. (1989). Antecedents of intergroup anxiety in Asian-Americans and Hispanic Americans. International Journal of Intercultural Relations. 13. 203-219. Triandis, H.C., & Triandis, L.M. (1960). Race, social class, religion, and nationality as determinants o f social distance. Journal of Abnormal and Social Psychology. 61(1), 110-118. Triandis, H.C., & Triandis, L.M. (1962). A cross-cultural study o f social distance. Psychological Monographs. 76(213. 1-21. Tversky, A. (1977). Features of similarity. Psychological Review. 84(41. 327- 352. 59 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Urban, L.M. (1997). Judgments o f persons from crossed social categories: Psychological importance o f categories and insult. Unpublished master’s thesis, University o f Southern California, Los Angeles. Urban, L. M., & Miller, N. (1998). A theoretical analysis o f crossed categorization effects. Journal o f Personality and Social Psychology. 74(4), 894- 908. Urada, D.I., & Miller, N. (2000). The Impact o f Positive Affect on Crossed Categorization Effects. Journal o f Personality and Social Psychology. 78(3). 417- 433. Vanbeselaere, N. (1987). The effects o f dichotomous and crossed social categorizations upon intergroup discrimination. European Journal o f Social Psychology. 17. 143-156. Vanbeselaere, N. (1991). The different effects o f simple and crossed categorizations: A result o f the category differentiation process, or o f differential category salience? In W. Stroebe & M. Hewstone (Eds.), European Review of Social Psychology (pp. 247-278). Chichester: Wiley. Vanman, E.J., Kaplan, D.L., & Miller, N. (1995). Assessment of crossed categorization effects on intergroup bias with facial electromyography. Manuscript submitted for publication. Vanman, E.J., & Miller, N. (1993). Applications of emotion theory and research to stereotyping and intergroup relations. In D.M. Mackie & D.L. Hamilton (Eds.), Affect. Cognition, and Stereotyping: Interactive Processes in Group Perception (pp. 213-238. San Diego: Academic Press. Watson, D., Clark, L.A., & Tellegen, A. (1988). Development and validation of brief measures o f positive and negative affect: The PAN AS scales. Journal of Personality and Social Psychology. 54(61. 1063-1070. Zarate, M. A., & Sanders, J. D. (1999). Face categorization, graded priming, and the mediating influences o f similarity. Social Cognition. 17. 367-389. 60 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. A P P E N D I X A S t u d e n t ID # : H o w important do you think this is for your W hich of these describes you best? self identity? (Circle one or f i l l in blank)____________________ (l=Not at a ll, 5=Very m uch) 1 . A y o u n g a d u l t / M id d le A g ed / A n O l d e r A d u lt 1 2 3 4 2 . U .S . C i t i z e n / C i t i z e n o f a n o t h e r c o u n t r y 1 2 3 4 3 . G e n e r a l l y p o l i t i c a l l y m o re l i b e r a l / 1 G e n e r a l l y p o l i t i c a l l y m o re c o n s e r v a t i v e 2 3 4 4 . P s y c h o lo g y m a i o r / O th e r m a io r 1 2 3 4 5 . C o m m u ter / USC R e s i d e n t (o n c a m p u s o r c l o s e by) 1 2 3 4 6 . USC H o n o rs s t u d e n t / N o t a USC h o n o r s s t u d e n t 1 2 3 4 7 . G e n e r a l l y r e l i g i o u s / G e n e r a l l y n o t r e l i g i o u s 1 2 3 4 8 . U n d e r g r a d u a t e s t u d e n t / G r a d u a t e s t u d e n t 1 2 3 4 9 . A f r a t e r n i t y o r s o r o r i t y m em ber / 1 N o t a f r a t e r n i t y o r s o r o r i t y m em b er 2 3 4 1 0 . A USC s t u d e n t a t h l e t e / N o t a s t u d e n t a t h l e t e 1 2 3 4 1 1 . A f r i c a n A m e ric a n / A s i a n / L a t i n o / N a t i v e flm e r. 1 / P a c i f i c I s l a n d e r / W h ite (N on H i s p a n i c ) O t h e r 2 3 4 1 2 . A s p o r t s f a n / n o t a b i g s p o r t s f a n 1 2 3 4 1 3 . S i n g l e / M a r r ie d / D i v o r c e d / S e p a r a t e d / W idow ed 1 2 3 4 14 . L e f t - h a n d e d / R i g h t- h a n d e d 1 2 3 4 15 . S u p e r s t i t i o u s / N o t s u p e r s t i t i o u s 1 2 3 4 1 6 . G rew u p m a in ly i n C a l i f o r n i a / 1 g re w u p i n a n o t h e r s t a t e o r c o u n t r y 2 3 4 6 1 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. A P P E N D I X B P a r t n e r S e l e c t i o n F o r m H o w m uch do you wish to he partners with participant tt1? not at all very m uch H o w m uch do you wish to be partners with participant #2? not at all very m uch H o w m uch do you wish to be partners with participant #3? not at all very m uch First choice: Participant tt Second choice: Participant tt Third choice: Participant ft Fourth choice: Participant tt 62 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. A P P E N D IX C T h i s i s p a r t o f a s t u d y o n i n t e r p e r s o n a l i n t e r a c t i o n . By p r o v i d i n g t h e f o l l o w i n g i n f o r m a t i o n y o u c a n h e l p u s f i n d a n a p p r o p r i a t e d i s c u s s i o n p a r t n e r f o r y o u . I f you do not wish to answer any one of the following questions for any reason, feel free to skip it. F o r e a c h q u e s t i o n : 1 . On t h e l e f t , p l e a s e c i r c l e t h e te r m t h a t b e s t d e s c r i b e s y o u 2 . On t h e r i g h t , p l e a s e i n d i c a t e how i m p o r t a n t t h a t d e s c r i p t i o n i s f o r y o u r s e n s e o f w ho y o u a r e . T h a n k y o u ! Which o f t h s s a describes you best? (Circle one or f i l l in blank)_________ H ow important do you think th is is fo r your s e l f identity? (1=Mot a fc e ll, 4=Very much) E X A M P L E : G C C Student / Mt. SAC Student A young adult / Middle Aged / An Older Adult U.S. Citizen / Citizen of another country Generally politically more liberal / Generally politically more conservative Social Science major/ Other major Glendale resident / Mot a Glendale rssizer.t Generally religious / Generally not religious A GCC student athlete / Mot a GCC student atr.lt African American / Armenian / Asian / Latino / / Pacific Islander / Caucasian (non-Armeniar.! Other 9. Part-time student / Full-time student 10. A big sports fan / net a big sports fan 11. Single / Married 12. Left-handed / Right-handed 13. Superstitious / Not superstitious 14. Grew up mainly in California / grew up in another state or country 63 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. APPENDIX D Sample Printouts o f Online Study (Experiment 5) R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Team Problem Solving other participants. Please keep i uiiivuto vt lu^viuiaiiuti auuui iu» UJlCi [M i tutu uiwv *rtii iwwiy^ o i<utwiii aiiutuni vi • vu. i \* w will then have the opportunity to tell us who you would like to be partner* with The computer will then pair von up with a partner, and the team task will follow If you feel uncomfortable answering any of the questions, please feel free to skip them However, we do require a certain minimum amount of information to proceed - which varies depending on a number of tactorsi. and therefore ask thai you answer as manv as vou can. Which o'; titoe •ic-vnhc- am best.* ; (low tinpOlttiiUdoyou think tin*. f> f.> i your self : identity.’ j ■ I- mu at all. 5 terv much) 1,1am : ? r. yawq ndu-t 0 1 0 2 0 3 * 4 0 5 2. ! am: i a US Cifczen IB 2 0 3 * 4 0 | 3 . 1 am : i politically mare liberal z s I r\ ■ 3«4C | 4. i um.'w as:i « psycn/psyd x id io maior 3 f i if you hxvnn a::en cicii aillcjic. *cicct tin: m ajor ?ha: m ost ir.teTsts you ) I----------—--------------------------------------- — I 5.1 live: : ir- a Western statafHswa-'/Pacifc :itno^cnc)J< j 6. K am: j 6ftn»fatiy net ?oi:gic;:s IB ? ri i ,y j . - • * < * 1 0 3 0 4 C’ ? £ i i < 7 .1 am: ; A iM ate. pluese select one |«i 0 4 0 5 I S. Tam: f~ithnicity please select one |» " j P. I am: . corrnn-ly a celtega s-udan; (y 3 0 4 0 r i 10.1 am: I 3ports______plsasg seieet era p] 3 G 4 3 O 4 11.1 am: < amain ±3 » > 3 G 4 o 5 ------— -.r . t i i ; i o • > \j - v* 3 O 4 O 5 12.1 am: i Handedness please select one |»] 13. lam : i Superstition please select one pi " ] - : | *C 2 G 3 0 4 0 5 14.1 w as:i Ra-soo mainly please select one !° i 0 2 0 3 a < | O 5 IS. 1: : Smoke: please seicc: one J * } 1 O i * S "S \G - G 3 O 4 C 5 ! A L L DONE W ITH THESE QUESTIONS' 65 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission. Team Membership Past research has clearly shown thai in order for these team discussions to work properly, it is extremely important that we make sure people on the same team do net feel uncomfortable with each other. Therefore, in order to help us minimize the possibility of teaming you up with someone you will feel uncomfortable with, you are being given an opportunity to indicate who you think you will feel uncomfortable with. Mo special knowledge or skills are needed for the team task you will participate in. It is only important that you try to exclude people that you would feel uncomfortable with Participant #1 is: generally religious and a US Citizen How strongly do you prefer to exclude this person from your team? (1 Tiot at all, 7“ very much) O l 0 2 0 3 <8>4 0 5 0 6 0 7 P artic ip an t #2 is: a citizen of another country How strongly do you prefer to exclude this person from your team? (l~not at all. 7-vcry much) O l 0 2 0 3 O 4 0 5 @6 0 7 Participant #3 is: a young adult and an undecidcd/undeclarcd major and generally religious How strongly do you prefer to exclude this person from your team? (l^not at all, 7-vcry much) Ol 0 2 0 3 0 4 ® 5 0 6 07 P artic ip an t #4 is: in an Eastern state (Eastern time zone) and a young adult and a business m ajor and generally religious How strongly do you prefer to exclude this person from your team? <l=not at all, 7':vcry much) O l 0 2 0 3 0 4 0 5 ® 6 O " P articip an t #5 is: politically more liberal How strongly do you prefer to exclude this person from your team? f 1 “noi at all, 7=very much) O l 0 3 0 4 0 5 0 6 0 7 I Pair me upi I 6 6 R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The Effect Of Positive Affective Arousal On Crossed Categorization Effects

PDF

The impact of state anxiety on categorization: Categorical inclusiveness, stability, and inferential perceptions

PDF

The out-group must not be so bad after all: Generalization of the effects of disclosure, typicality and salience on intergroup attributions

PDF

Sadness and social judgments: No task effort, no mood enhancement?

PDF

Ideological threat, social identity, and consensus estimation

PDF

The impact of cocaine use on psychosocial functioning in an HIV+ community sample: A model for examining stability and change

PDF

The mismatched couple: Effects on depressive symptomatology and romantic relationship functioning

PDF

The development and validation of the Group Supervision Scale

PDF

The two-hit hypothesis of schizophrenia: Testing the effects of genetic risk, obstetric complications, and unstable early rearing

PDF

The influence of personality, ethnicity, and culture on social support and emotional distress in dementia caregivers: A transactional stress and social support model

PDF

The correspondence bias: The applicability of contextual information

PDF

Individual motivation loss in group settings: An exploratory study of the social -loafing phenomenon

PDF

The effects of coherence -based reasoning on betting decisions

PDF

Development and validation of the Group Supervision Impact Scale

PDF

Cognitive load, trigger salience, and the facilitation of an aggressive response in triggered displaced aggression

PDF

The impact of attributional processes on aggressive behavior

PDF

Models of nineteen -year change in text recall and cognitive mediators of change

PDF

Disgust as negative affect: A crossed categorization study

PDF

Toward a measurement model of human strengths: Evidence from R- and P-technique studies

PDF

Timecourse and temporal dynamics of attention in tasks of intra- and cross -modal cuing and switching

Asset Metadata

Creator Urada, Darren Ichiro (author)

Core Title Social crossed categorization beyond the two -group model

Degree Doctor of Philosophy

Degree Program Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,psychology, social

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-138858

Unique identifier UC11330365

Identifier 3041540.pdf (filename),usctheses-c16-138858 (legacy record id)

Legacy Identifier 3041540-0.pdf

Dmrecord 138858

Document Type Dissertation

Rights Urada, Darren Ichiro

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA