Printer Friendly

Attributional Complexity and the Illusory Correlation: A Test of the Inverted-U Hypothesis.

People take in an enormous amount of information about their social worlds. For the most part, they do that in a way that facilitates adaptive behavior and allows them to reach their goals (Fiske, 1992; Todd & Gigerenzer, 2007). Nonetheless, social information can be processed and mentally represented in a biased and distorted manner. For example, people are insufficiently attentive to contextual influences on behavior (Gilbert & Malone, 1995), perceive more consistency in others' behavior than actually exists (Kunda & Nisbett, 1986), selectively remember social information that is consistent with their expectations (Hirt, Lynn, Payne, Krackow & McCrea, 1999), and perceive other groups to be more homogenous than their own (Linville, Salovey, & Fischer, 1986).

A common theme in the literature on biases and errors in social cognition is that they are most likely to occur among people who engage in low-intensity processing (e.g., people low in the Need for Cognition --Cacioppo, Petty, Feinstein, & Jarvis, 1996). In other words, it is often found that biases and errors are more in evidence when social perceivers either cannot or will not engage in effortful thinking (Chaiken & Trope, 1999; Devine, 1989a; Fiske & Taylor, 1991; Gilbert, 1989). Less common, but still well represented in the literature, are investigations revealing that some biases can be more pronounced among people who engage in extended, effortful social information processing (Ditto & Lopez, 1992; Newman, Duff, Schnopp-Wyatt, Brock, & Hoffman, 1997; Tetlock & Boettger, 1989). In short, some biases are magnified among people who put minimal effort into social perception and judgment (because they fail to attend to the diagnostic information that would temper a biased conclusion). At other times, though, a great deal of effort is required to construct biased social judgments or distorted mental representations (because gathering enough non-diagnostic information to support a biased conclusion can require time and energy).

Stroessner and Plaks (2001), in their review of a cognitive process potentially underlying the erroneous perceptions of group differences and the creation of stereotypes--i.e., the formation of illusory correlations (or perceptions of an association between two variables that are objectively uncorrelated)--presented yet another possible relationship between bias and the thoroughness of processing. Specifically, they suggested "illusory correlations are most likely to form when perceivers have the motivation and cognitive capacity to process available information with a moderate degree of thoroughness" (p. 250, emphasis added). In other words, the relationship between processing intensity and strength of the illusory correlation was said to resemble an inverted-U, because perceiving an illusory correlation depends on a moderate level of processing that is sufficient to allow for the recognition of differential frequencies of social information, but not so thorough as to allow for the highly accurate processing of that information. Evidence for that hypothesis is limited, however. In the current investigation, we make use of a relevant personality variable--Attributional Complexity (Fletcher, Danilovics, Fernandez, Peterson, & Reeder, 1986)--to generate novel evidence in support of Stroessner and Plaks's (2001) hypothesis.

Distinctiveness-based Illusory Correlations

An illusory correlation is formed when a person perceives an association between two variables (e.g., social group membership and some observed behavior) in the absence of objective evidence to warrant an association between them. Illusory correlations can develop for a variety of reasons, but one often examined and widely replicated factor is the co-occurrence of distinctive stimuli (e.g., Acorn, Hamilton, & Sherman, 1988; Hamilton & Gifford, 1976; Hamilton, Dugan, & Trolier, 1985; Risen, Gilovich, & Dunning, 2007; Sherman et al., 2009; for a review, see Spears & Stroebe, 2015). Distinctive stimuli are very memorable, and stimuli (e.g., events, objects, or people in the environment) that are already distinctive when encountered in isolation are especially memorable when they appear together. For example, the only Latino student in a classroom where all of the other students were European-American would be distinctive, but if that student entered the class every day on crutches, he or she would be doubly distinctive. Similarly, a woman wearing a hijab might stand out in a workplace; if she also was particularly extraverted, she too would be doubly distinctive.

Hamilton and Gifford (1976) were the first to demonstrate the role of distinctiveness-based illusory correlations in the formation of stereotypes. They exposed research participants to a series of statements describing a majority Group A, associated with 18 positive (socially desirable) and 8 negative (socially undesirable) behaviors, and a minority Group B, associated with 9 positive and 4 negative behaviors, and predicted that people would perceive stimulus relationships that did not actually exist (i.e., that Group B's behavior would be perceived more negatively than Group A's despite the equivalent 9:4 ratio for positive and negative behaviors). This prediction derived from the more general expectation that doubly distinctive events (in this case, negative behaviors of the minority group) would stand out in memory. The prediction was confirmed. In addition, when participants were presented with a set of behaviors in which there were fewer positive than negative behaviors, the minority group was perceived more positively than the majority group. These findings have been replicated many times (Sherman, Sherman, Percy, & Soderberg, 2013; Stroessner & Plaks, 2001).

Distinctiveness-based illusory correlations have been of great interest in part because of their implications for the formation of social stereotypes (Hamilton & Sherman, 1989). Minority groups are by definition distinctive groups of people. Negative, undesirable behavior is also arguably much more rare, distinctive, and noticeable than positive or neutral behavior (Jones & Davis, 1965; Skowronski & Carlston, 1989). It follows that if (for example) a person with a mental illness was seen behaving in a violent way, this might qualify as a doubly distinctive event and thus contribute to an illusory correlation. An observer would be more likely to remember the behavior of the person with a mental illness; therefore, a stereotype would form more readily than if the person seen behaving in a violent way was a member of a majority group.

Stroessner and Plaks (2001) reviewed research suggesting that illusory correlations are most likely to be perceived when people have the motivation and cognitive capacity to process social information with a moderate degree of thoroughness. If, on the one hand, a variable undermines people's motivation to actively process available social information (e.g., they are experiencing a heavy cognitive load), it is unlikely that they will form an illusory correlation because they will not be alert enough to recognize and encode distinctions between different kinds of people and behaviors. As Stroessner and Plaks (2001) explain, without "the motivation or capacity to deliberatively process available information, illusory correlations are likely to be weak or even eliminated.. .The bias depends on the perceivers' ability to recognize that some group/behavior combinations differ in frequency" (p. 250). If, on the other hand, a variable promotes people's thorough information processing (or they are dispositionally inclined to process social information carefully), they could also be unlikely to form an illusory correlation because they will be motivated and able to perceive and mentally represent social information in a more veridical manner.

Evidence for the hypothesis is mostly indirect, however. Variables that increase deliberative processing (e.g., instructions urging participants to process information very carefully--Pryor, 1986) have been found to attenuate illusory correlations. The same is true for variables that decrease deliberative processing (e.g., the induction of moods such as happiness and sadness--Stroessner, Hamilton, & Mackie, 1992). But no studies have been conducted for the express purpose of demonstrating the overall hypothesized inverted-U pattern.

Attributional Complexity

One way to gauge people's preference for thoroughness in information processing is through the administration of the Attributional Complexity Scale (ACS; Fletcher et al., 1986). The ACS measures people's tendency to be motivated to think deeply about human behavior (both their own and others'), and their tendency to do so in a complex and thoughtful way. In essence, the ACS is designed to measure people's need for social cognition (as opposed to the Need for Cognition more generally; cf. Cacioppo et al., 1996). Fletcher et al. (1986) present evidence for the measure's internal and test-retest reliability, along with evidence of its convergent, discriminant, concurrent, and predictive validity. Subsequent research has contributed to the validation of the measure. For example, people high in attributional complexity are less likely to ignore contextual constraints on behavior and thus less likely to commit the fundamental attribution error (Devine, 1989b). They are also less likely to predict that people will inflexibly engage in future trait consistent behavior regardless of the situations those people find themselves in (Newman, 1996).

The ACS was administered to participants in this study, and we predicted that illusory correlations would be enhanced and most in evidence among those with moderate (relative to other participants) levels of AC. In other words, examination of the relationship between AC and the illusory correlation was expected to reveal an inverted-U pattern.

Mental illness stigma

Another objective of this research was to examine people's readiness to form a particular kind of illusory correlation: one between people diagnosed with a mental health condition and undesirable behaviors (especially violent behaviors). Bias against people with mental illness has long been documented (Corrigan, 2005; Hinshaw, 2007; Thornicroft, 2006; Wahl, 1995), and stereotypes associated with mental illness (Ottati, Bodenhausen, & Newman, 2005) are often rooted in relatively limited exposure to people with mental illness. The causes of the denigration and social exclusion of people with mental illness are complex and multilayered, but the current research explored illusory correlations as a potential contributor to mental illness stigma.

Though people with mental illness may be stigmatized in many ways, the dangerousness stereotype is central (Levin, 2001; Link, Phelan, Bresnahan, Stueve, & Pescosolido, 1999). Thus, in some conditions of our experiment, we examined the readiness of people to form illusory correlations between people with mental illness and violent behavior (as opposed to just negative, undesirable behavior in general). To the best of our knowledge, mental illness stigma has never before been examined in the context of Hamilton and Gifford's (1976) experimental paradigm.

Overview and Hypotheses

The current research utilized the traditional illusory correlation experimental paradigm introduced by Hamilton and Gifford (1976). Participants were presented with positive and negative behavioral information about two groups of people; the majority/minority status of those groups was operationalized by manipulating the number of behaviors associated with each--in this case, 26 v. 13. The ratio of positive to negative behavioral statements was 18:8 in the majority group (with a corresponding minority ratio of 9:4). In half of the conditions of the experiment people "diagnosed with a mental illness" were in the majority and in half of the conditions another group was the majority group. Half of the conditions included positive and nonviolent negative behavioral statements and half of the conditions included positive and violent negative behavioral statements. All participants completed the ACS.

The hypotheses were that (1) the illusory correlation effect would be replicated, that (2) the effect would be most pronounced for participants of moderate AC (relative to those with high or low levels), and that (3) the effect would also be most pronounced when the group characterized as "diagnosed with a mental illness" appeared in the minority and the behavioral statements included violent behaviors.

METHOD

Participants

One hundred and nineteen students (87 female, 32 male) enrolled at a large university in New York participated in exchange for course credit. All participants were at least 18 years of age (M = 19.4 years, range = 18 to 31 years). Although a formal power analysis was not conducted, the number of participants in the study met or exceeded the number in previous published experiments testing for the effect of a moderating variable (in this case, AC) on illusory correlation formation (e.g., Pryor, 1986; Sanbonmatsu, Shavitt, & Sherman, 1991; Stroessner et al., 1992).

Materials and Procedure

Participants first completed the ACS (Fletcher et al., 1986), disguised for research purposes as the "Person Perception Questionnaire." The ACS includes 28 items (e.g., "I don't usually bother to analyze and explain people's behavior"). Beneath each item is a response scale ranging from Strongly Agree (+3)--Moderately Agree (+2)--Slightly Agree (+1)--Neither Agree nor Disagree (0)--Slightly Disagree (-1) Moderately Disagree (-2)--to Strongly Disagree (-3). After reversescoring half of the items on the questionnaire, each participant was given a total score by summing responses to all items on the questionnaire; higher scores indicated higher AC.

After completing the questionnaire, each group of up to three participants was randomly exposed to one of eight conditions of the 2 (Majority group: People diagnosed with a mental illness or Control group) x 2 (Control group: Tobacco users or Only children) x 2 (Negative behavior type: Nonviolent or Violent) ANOVA design (Appendix A). The methodology required there to be control groups whose behaviors would be presented alongside those described as "diagnosed with a mental illness." (Labeling the control group "people without mental illness" would have telegraphed to participants the purpose of this manipulation.) The control groups selected were "tobacco users" and "only children" (i.e., children without siblings). These groups were chosen because similar to people with mental health conditions, tobacco users and only children are not particularly uncommon social groups. Also similar to the mentally ill, tobacco users and only children might be associated with relatively unfounded negative social stereotypes (e.g., tobacco users do not care about their health, children without siblings are spoiled, people with mental illness are dangerous).

Participants in every condition were shown a PowerPoint presentation that included an introductory slide, a series of 39 behavioral statements (one per slide), and a final slide. The slides were projected onto a white board facing participants in a laboratory room. At the start of each trial, the researcher read the introductory slide aloud and asked participants to follow along. The introductory slide read as follows:

"Today you will be taking part in an exercise on social cognition. You will be seeing and reading a series of statements describing the behaviors of members of two different social groups.

People participating in this study will see different sets of behaviors from different social groups. We're not going to tell you much about the groups. However, we will tell you something: In all cases we will reveal one characteristic that all members of each group share.

Today, for example, we'll be showing you behaviors from one group in which everyone has been diagnosed with a mental illness and from one group in which everyone is a[n] tobacco user[/only child]. It will be clear when you see the behaviors which group each person belongs to.

You will only be presented with the series of statements once, so do pay attention to the best of your ability."

When the researcher was finished reading the introductory slide aloud, she signaled that the experiment would begin by clicking the space bar and manually launching the rest of the presentation. Each slide following the introductory slide was formatted to display for exactly six seconds. The researcher exited the room after launching the presentation but kept time so that she would know when the presentation had finished in each trial.

Each of the 39 behavioral statement slides (26 majority, 13 minority) that followed the introductory slide included one statement describing a member of one of two social groups performing either a desirable or undesirable behavior (see Appendices B and C for a list of all behaviors). In four conditions, participants were shown statements describing one group in which all members had been diagnosed with a mental illness and one group in which all members were tobacco users. In the other four conditions, participants were shown statements describing one group in which all members had been diagnosed with a mental illness and another group in which all members were only children. The ordering of the behavioral statement slides was pseudo-randomized in each presentation by (1) using a freely-available, online number randomizer to "shuffle" the presentation of the 39 statements from their original ordering (as seen on Appendices B and C) and (2) examining each presentation to correct for (a) any instances of back-to-back negative statements throughout each presentation (in anticipation of a possible heightened cumulative effect) and (b) presentations beginning or ending with a negative statement (in anticipation of possible primacy or recency effects).

Each series of statements included twice as many descriptions of one group than the other, and each series of statements included more desirable than undesirable behaviors. In every condition, majority group members were described using 18 desirable and 8 undesirable behaviors, and minority group members were described using 9 desirable and 4 undesirable behaviors. Thus, there was no association between group membership and the desirability of behaviors.

Additionally, the same set of positive behavioral statements, nonviolent negative statements, and violent negative statements was used to describe the majority group in every condition. Likewise, the same sets of behavioral statements were used to describe the minority group in every condition. In sum, the majority and minority group was characterized by the same set of (1) positive behavioral statements and (2) nonviolent or violent negative statements in every condition (see Appendix A for a summary) (1).

Following the presentation of 39 behavioral statements, a final slide was displayed on the PowerPoint that read "END. Thank you for paying attention! Please stay seated and you will be provided with individual directions to complete your task."

The first page of the post-task (the memory test; see Appendices A and B) requested that participants "Please mark next to each statement whether a person with a mental illness (you may write "MI") or a tobacco user/only child (you may write "TU"/"OC") performed each behavior you read about earlier in this experiment." Following this instruction was a list of all of the behavioral statements that participants in their respective conditions had been exposed to.

The second page of the directions requested that participants "Please rate the members of Group MI (people with a mental illness) and Group TU/OC (tobacco users/only children), whose behaviors you read about earlier in this experiment, on this list of traits on a scale of 1 to 10 (1 being a trait exhibited 'never' and 10 being a trait exhibited 'often')." Following this instruction were two sets of identical traits (Intelligent; Approachable; Clean; Cultured; Aggressive; Educated; Violent; Responsible; Trustworthy; Dangerous) with two separate headings for Group MI and Group TU/OC.

Ratings on the three undesirable traits (aggressive, violent, and dangerous) were reverse-scored and combined with ratings on the seven desirable traits (intelligent, approachable, clean, cultured, educated, responsible, and trustworthy) to create an overall impression score.

Finally, demographic information was collected. Participants were prompted to record their sex, their age (in years and months), whether English was their first language, at what age they began speaking English fluently (if it was not their first language), whether they were born in the United States, and where they were born (if not in the US). Finally, participants reported their ethnicity: American Indian or Alaskan Native (n = 1), Asian or Pacific Islander (n = 32), Black/African American not of Hispanic origin (n = 17), Latino/a or Hispanic (n = 12), Caucasian/White not of Hispanic origin (n = 55), or Other (n = 2). Participants were then debriefed on the purpose and nature of this study and given the opportunity to ask any specific questions they might have had about the research.

RESULTS

Validation of Stimulus Materials

Data were collected from an independent group of participants (80 undergraduate students) to test the assumptions that (1) the positive/desirable behaviors were more evaluatively positive than the negative/undesirable behaviors, (2) the violent negative behaviors were perceived as being more violent than the nonviolent ones, (3) the positive behaviors attributed to the majority group were approximately as positive as those attributed to the minority group, and (4) the negative behaviors attributed to the majority group were approximately as negative as those attributed to the minority group.

Participants first rated each behavior statement for valence. They were presented with a list of the 51 behavior statements in a fixed random order (that is, a single randomized order was used for all participants), and instructed to evaluate each behavior statement using a 5-point scale on which 1 = Very Negative, 2 = Moderately Negative, 3 = Neither Positive nor Negative, 4 = Slightly Positive, and 5 = Very Positive. Following that, participants were presented with a list of only the 24 negative behavior statements in a fixed random order and rated each one for the level of violence it suggested using a 5-point scale, where 1 = Not At All Violent, 2 = Slightly Violent, 3 = Moderately Violent, 4 = Very Violent, and 5 = Extremely Violent.

Subsequent analyses revealed that there was a significant difference in positivity between the positive behavior statements (M = 5.53, SD = 0.49) and the negative behavior statements (M = 2.15, SD = 0.37); t(79) = 43.98, p<.001, d = 4.92. In addition, the violent negative behaviors were perceived as being more violent than the nonviolent ones, M = 3.62 (SD = 0.62) vs. M = 1.42 (SD = 0.47), t(79) = 28.09, p<.001, d = 3.14.

The mean valence of the positive behaviors associated with the majority group (M = 5.54, SD = 0.48) was (as intended) not significantly different from the valence of the positive behaviors associated with the minority group (M = 5.51, SD = 0.63), t(79) = 0.72, p = .48, d = 0.08. The comparison between the mean valence of the majority violent negative behaviors (M = 1.74, SD = 0.41) and the minority violent negative behaviors (M = 1.65, SD = 0.62) likewise did not reveal a statistically significant difference, t(79) = 1.70, p = .094, d = 0.19. The mean valence of the majority nonviolent negative behaviors (M = 2.78, SD = 0.53), however, was less negative than the mean valence of the minority nonviolent negative behaviors(M = 2.18, SD = 0.54), t(79) = 8.42, p<.001, d = 0.94. This difference in valence, however, is clearly dwarfed by the one between the positive and negative behaviors2.

Replicating the Illusory Correlation Effect

Behavior Recall. Given the cued recall procedure (in which participants matched groups with behaviors), the number of desirable or undesirable behaviors attributed to one group (minority or majority) by any given participant was perfectly predictive of how many were attributed to the other group (the number of desirable behaviors always had to sum to 27, the number of desirable behaviors in the set presented to participants, and the number of undesirable behaviors had to sum to 12). Thus, hypotheses involving recall could be and were tested by focusing on results for the minority group (i.e., the group represented by 13 behaviors).

Consistent with past research (see Hamilton & Gifford, 1976), participants had a tendency to overattribute behaviors to the minority group. On average, participants indicated that the minority group had performed 5.29 (SD = 1.95) undesirable behaviors (instead of 4), and 10.69 (SD = 2.46) desirable behaviors (instead of 9). In both cases, the number recalled was significantly greater than the actual number presented (t (118) = 7.24, p < .001, d = 0.69 for negative behaviors, and t (118) = 7.49, p< .001, d = 0.69 for positive behaviors). Participants' relative bias to attribute negative or positive behaviors to the minority group was of more relevance to the hypotheses; thus, the main analysis focused on the number of positive and negative behaviors remembered as being associated with the minority group in proportion to the actual number presented (for a related approach, see Hamilton et al., 1985, Experiment 2).

Participants associated 32.4% (5.29/4.00) more negative behaviors with the minority than was actually the case, but the comparable figure for positive behaviors was only 18.8% (10.69/9.00). The difference between the two was statistically significant, t(118) = 2.55, p = .012, d = 0.23. Thus, in terms of the memory data, the illusory correlation effect was replicated: relative to positive behaviors, participants overattributed negative behaviors to the minority group.

An analysis based on the full 2 (Minority group: Diagnosed with a mental illness or Control) x 2 (Control group: Tobacco users or Only children) x 2 (Negative behavior type: Nonviolent or Violent) x 2 (Valence: Positive or Negative) ANOVA design revealed a main effect of Valence, F(1,111) = 6.97, p = .009, [[eta].sub.p.sup.2] = -059 (corresponding to the effect already described); in addition, there were main effects of Minority group, F(1,111) = 6.90, p = .010, , [[eta].sub.p.sup.2] = -059, Negative behavior type, F(1,111) = 4.13, p = .044, [[eta].sub.p.sup.2] = .036, and an interaction between Minority group and Negative behavior type, F(1,111) = 11.11, p = .001, [[eta].sub.p.sup.2] = 091. All of these effects, however, were qualified by the three-way Minority group by Negative behavior type by Valence interaction, F(1,111) = 12.90, p< .001, [[eta].sub.p.sup.2] = . 104 (the only interaction involving Valence)3.

The pattern of means reflected by the three-way interaction can be found in Table 1. A concise summary of the findings is as follows: there was actually less of a tendency to attribute negative violent behaviors to people with mental illness when they were in the minority than to attribute them to the other group (i.e., tobacco users or only children) when they were in the minority. That difference was not in evidence for nonviolent negative behaviors: in that case, the difference was in the opposite direction, although less pronounced. Overall, this pattern was contrary to the one expected. Participants did not show a tendency to mistakenly recall that negative violent behaviors were especially likely to be associated with people with mental illness4.

Desirability Ratings. Ratings associated with undesirable traits (i.e., Aggressive, Violent, and Dangerous) were reverse scored and combined with ratings on desirable traits (i.e., Intelligent, Approachable, Clean, Cultured, Educated, Responsible, and Trustworthy). The alpha level for the resulting scale was .71 for ratings of the control groups and .76 for ratings of the group of people characterized as diagnosed with a mental illness.

An analysis of variance was performed to determine whether there was a significant difference between the overall desirability ratings assigned to groups in the minority relative to the majority (i.e., the classic illusory correlation effect when negative behaviors are less frequent than positive ones) and whether that finding was moderated by other variables. The complete ANOVA design was 2 (Minority group: Diagnosed with a mental illness or Control) x 2 (Control group: Tobacco users or Only children) x 2 (Negative behavior type: Nonviolent or Violent) x 2 (Group rated: Diagnosed with a mental illness or Other), with the last variable within-subjects.

The ANOVA revealed the expected Minority group x Group rated interaction, F(1, 111) = 36.24, p< .001, [[eta].sub.p.sup.2] = .246. As revealed in Table 2, the group in the minority was rated more negatively than the group in the majority, regardless of the identity of that group. Thus, in terms of ratings, the illusory correlation effect was again replicated.

The only other significant effect was a main effect for Negative behavior type, F(1, 111) = 39.08, p< .001, [[eta].sub.p.sup.2] = .260. Not surprisingly, group ratings overall were more unfavorable when the negative behaviors were violent (M = 5.65, SD = 0.86) than when they were nonviolent (M = 6.69, SD = 0.93). Finally, the Group rated x Other group type interaction was marginally significant, F(1, 111) = 3.38, p = .069, [[eta].sub.p.sup.2] = .030. The identity of the control group had no effect on ratings of the group said to consist of people with mental illness (M = 6.06, SD = 1.54, when the other group was said to consist of only children; M = 6.11, SD = 1.44, when the other group was said to consist of tobacco users). Only children, however, were rated more positively than tobacco users (M = 6.51, SD = 1.40 vs. M = 5.98, SD = 1.31).

The Minority group x Group rated interaction, however, was not significantly moderated by Negative behavior type; in addition, the effects of being in the minority were essentially identical for people with mental illness as they were for the control group (see Table 2, which reveals only a small and nonsignificant tendency for the control group to be rated more favorably). Overall, the results did not support the hypothesis that the illusory correlation effect would be most pronounced when the group characterized as "diagnosed with a mental illness" appeared in the minority and the behavioral statements included violent behaviors. In addition, an alternative analysis restricted to the three undesirable and violence-related ratings (Aggressive, Violent, and Dangerous) revealed a similar pattern, although the Minority Group x Group rated interaction failed to reach significance, F(1, 111) = 3.52, p = .063, [[eta].sub.p.sup.2] = .031.

The Moderating Role of Attributional Complexity

Behavior Recall. To test the hypothesized relationship between AC and distinctiveness-based illusory correlations (i.e., an inverted-U pattern), a hierarchical regression analysis testing the linear and quadratic effects of AC on memory for minority group negative behaviors was conducted. After centering, AC was entered into the equation first and the square of that same variable was entered in a second step (see Cohen, Cohen, West, & Aiken, 2003). The results of the analysis for negative behaviors can be found in Table 3. The quadratic term accounted for a significant increment in variance accounted for, and the negative value of the beta coefficient indicates an inverted U-shaped (concave downward) curve. Thus, as predicted, the tendency to associate negative behaviors with the minority group was most pronounced for those predisposed to engage in a moderate amount of social information processing.

A parallel analysis for positive behaviors revealed no hint of a similar pattern (for the [R.sup.2] increment for the quadratic term, p = .94).

As a complementary post-hoc analysis to further illustrate the relationship between AC and the tendency to recall that negative behaviors were associated with the minority group, participants were divided into five groups based on their total AC score. Those in the top fifth of the distribution (average item score 1.86 and higher, n = 23) were designated as being "high" in AC; those in the bottom fifth (0.64 and lower, n = 23) as "low;" and those in the middle fifth (1.00 - 1.29, n = 24) as "moderate" (possible range = -3 to +3). Table 4 reveals the mean attribution of negative and positive behaviors to the minority group as a function of AC level. In both cases the mean is highest for the moderate group, but the contrast is greater for negative behaviors.

A one-way ANOVA focused on negative behaviors did not yield a significant omnibus main effect of group (F(2, 69) = 2.01, p = .14), [[eta].sup.2] = .06, but a focused planned comparison (with cell weights set at -1, +2, and -1) was more supportive of the hypothesized pattern, f(67) = 1.83, p = .036 (one tailed test) (5). The results for the corresponding tests involving positive behavior were F(2, 69) = 0.29, p = .75, [[eta].sup.2] = .01, and t (67) = 0.62, p = .27.

Desirability Ratings. Hierarchical regression analyses testing the linear and quadratic effects of AC were again run, with the dependent variables in this case being the mean ratings of the minority group, mean ratings of the majority group, and the difference between the ratings. Analyses using group ratings as indicators of distinctiveness-based illusory correlations failed to support the hypothesized nonlinear relationship with AC (for the [R.sup.2] increments for the quadratic term, all three p-values were greater than .60).

Consistent with the analyses of behavior recall, however, additional post-hoc analyses were run after dividing participants into high (top fifth of the distribution), moderate (middle fifth), and low (bottom fifth) AC groups. As revealed by Table 5, all three groups of participants rated the minority group more negatively than the majority group. However, the difference was significant at the conventional p< .05 level only for the moderate AC group, t(23) = 3.00, p = .006, d = 0.61 (for the high AC group, t(22) = 1.63, p = .12, d = 0.33; for the low group, t(23) = 1.72, p = .10, d = 0.35) (6).

DISCUSSION

A main objective of this research was to examine the relationship between Attributional Complexity (AC) and the formation of illusory correlations. More generally, the purpose was to test Stroessner and Plaks's (2001) prediction that illusory correlations can vary as a function of how deliberatively and thoroughly social information is processed. According to these researchers, the relationship between this cognitive motivation and the strength of the illusory correlation resembles an inverted-U, because the illusory correlation effect depends on a moderate level of processing sufficient enough to allow the recognition of differential frequencies of social information, but not so thorough as to allow the highly accurate processing of that information. Hence, the relationship between AC and the illusory correlation was also predicted to be U-shaped.

The illusory correlation effect has been explained in part in past research by the unique way in which distinct stimuli are perceived; specifically, distinct stimuli are very memorable, and social stimuli (i.e., events, objects, or people in the environment) that are distinct in isolation are especially memorable when encountered in conjunction (Hamilton & Gifford, 1976; Hamilton et al., 1985; Hamilton & Sherman, 1989; Risen et al., 2007). This account (for an alternative perspective, see Berndsen, Spears, McGarty, & van der Pligt, 1998; Haslam, McGarty, & Brown, 1996; McGarty, Haslam, Turner, & Oakes, 1993) predicts that if people are provided with positive and negative (but mostly positive) information about two different groups that is different in frequency but equal in ratio, they will attribute more negative qualities to the group about which they received less information because the doubly distinctive events (negative behaviors of the minority group) will stand out in memory. The novel prediction in this study was that the effects of the doubly-distinctive information would be especially apparent among people with moderate levels of AC, an individual difference variable that corresponds to people's motivation to carefully attend to and process social information.

Analysis of the memory data (i.e., participants' beliefs about which behaviors were associated with which group) replicated past work on the illusory correlation effect: there was a tendency for participants to overattribute negative behaviors to the minority group more than they did for positive behaviors. Moreover, a regression analysis suggested that participants' recall of negative minority group behaviors varied nonlinearly as a function of their scores on the Attributional Complexity Scale (ACS); specifically, an inverted-U pattern (albeit of modest magnitude) was revealed, consistent with Stroessner and Plaks's (2001) prediction that illusory correlations are most likely to form when people have the motivation and cognitive capacity to process available information with a moderate degree of thoroughness. Those who scored in the middle quintile on the ACS were those most likely to over-attribute negative behavior to the minority group. Thus, this research provides novel support for Stroessner and Plaks's (2001) hypothesis about the relationship between processing intensity and illusory correlations.

Analysis of the rating data also replicated past work on the illusory correlation effect: participants formed a less favorable impression of the minority group than they did of the majority group. However, the rating data provided much weaker support than the memory data for the hypothesis that moderate levels of processing are more likely to lead to illusory correlations than are high or low levels. Post-hoc analyses, however, revealed that the tendency to form a more negative impression of the minority group than the majority group was most clearly in evidence among participants with moderate levels of AC. In sum, although the weight of the evidence is arguably consistent with the inverted-U hypothesis, the findings for the memory data provided more compelling support than the findings for the rating data. It is far from uncommon, though, for results for these two dependent variables to be discrepant within any given study of illusory correlations (see Johnson & Mullen, 1993).

The current study is among the only ones in the literature to use a version of Hamilton and Gifford's (1976) experimental paradigm that included actual (vs. hypothetical) social groups (cf. McArthur & Friedman, 1980). The reason for this modification was that another objective of the research was to examine people's readiness to form an association between people diagnosed with a mental health condition and negative behaviors (especially violent behaviors). However, the results did not support the hypothesis that the negative (especially violent) behaviors of mentally ill minority group members would be especially salient and memorable, because of those behaviors' double distinctiveness and participants' potential pre-existing biases against mentally ill people. Neither the memory nor the rating data provided any support for this hypothesis. Indeed, participants demonstrated a significant bias to attribute negative violent behaviors (relative to nonviolent ones) to the control group when it appeared in the minority, not to people with mental illness.

It is possible that this finding was due to violent behavior on the part of only children and tobacco users being more surprising to participants than was a person with mental illness engaging in the same behavior. Past research has found that at least in the short term, recall for information that contradicts people's expectations about others is especially likely to be enhanced (Belmore & Hubbard, 1987; Srull & Wyer, 1989; Stangor & McMillan, 1992). If participants in the current study did indeed find it more noteworthy when control group members were engaged in violent behaviors than when mentally ill people were engaged in such behaviors, then those behaviors would be particularly memorable for that reason. If so, this effect would work against the expected result--that is, it would make it less likely that violent behaviors would be more memorable when associated with people with mental illness.

In sum, as a result of using known categories of people as groups (as opposed to the hypothetical, generically labeled groups common to most illusory correlation research), expectancy-based memory biases essentially competed with those due to double-distinctiveness. It is for this reason, perhaps, that published illusory correlation studies with the design adapted for this study almost never pair behaviors with real-world groups. The one exception seems be a study by McArthur and Friedman (1980), who varied the race, age, and gender of the groups. That study, however, also yielded a complex pattern of findings, and the authors concluded that "associative connections and infrequency can act in concert or in opposition" (p. 623).

Still, before concluding that Hamilton and Gifford's paradigm might not be the best for investigating the effects of mental illness stigma, additional studies might benefit from a more diverse group of participants. All participants in the current study were college students, and we cannot rule out the possibility that this population is characterized by a heightened sensitivity to mental health issues.

Levels of AC represent just one way to operationalize the thoroughness with which social information will be processed. Future research involving other research strategies will be needed to provide converging evidence for Stroessner and Plaks's hypothesized inverted-U relationship between processing intensity and illusory correlations and to reveal variables that might moderate that relationship. Moreover, it could well be the case that dispositional differences in the thoroughness of processing have the same complicated (yet so far unexamined) relationship to other judgmental biases that are found when people are required to attend to and integrate multiple pieces of social information, such as primacy (Anderson & Barrios, 1961) and biased assimilation (Lord, Ross, & Lepper, 1979) effects. Inspiring empirical investigations of that possibility could potentially be the most important outcome of this investigation.

APPENDIX A

Summary of Experimental Design

APPENDIX B

Memory Test: Nonviolent negative behavior condition

--is a member of a book club

--is a loyal and trustworthy friend

--has excellent personal hygiene

--takes photographs at family events

--sleeps with the lights on

--volunteers at a local animal shelter

--enjoys cooking

--is consistently late to work

--buys lunch for a co-worker

--is well-organized

--holds the door open for other individuals

--bites his/her nails

--has a driver's license

--has never defaulted on a credit card statement

--teaches his/herself a second language

--lives with supportive roommates

--avoids alcohol consumption

--plays acoustic guitar

--practices meditation

--recycles

--has a suspended driver's license

--waves at a pedestrian on the street

--consumes alcohol heavily and regularly

--brags often

--is enrolled at a university

--is easily distressed

--helps an elderly woman carry her grocery bags

--is a Parent Teacher Association (PTA) member

--begins a weekly exercise regimen

--keeps a journal

_repeatedly fails to maintain a monthly budget

--avoids eye contact during public/one-on-one interaction

--teaches a friend how to play a new card game

--has a loving relationship with his/her parents

--begs for change outside a local grocery

--is in a relationship with a significant other

--saves a portion of every paycheck for a new car

--has body odor

--compulsively plays with his/her hair

APPENDIX C

Memory Test: Violent negative behavior condition

--is a member of a book club

--is a loyal and trustworthy friend

--has excellent personal hygiene

--takes photographs at family events

--storms out of a restaurant and flips a table after waiting too long for service

--volunteers at a local animal shelter

--enjoys cooking

--kicks a puppy on the street for barking

--buys lunch for a co-worker

--is well-organized

--holds the door open for other individuals

--verbally abuses a co-worker

--has a driver's license

--has never defaulted on a credit card statement

--teaches his/herself a second language

--lives with supportive roommates

--avoids alcohol consumption

--plays acoustic guitar

--practices meditation

--recycles

--enjoys getting into bar fights with strangers

--waves at a pedestrian on the street

--owns a handgun

--fired from a previous job for provoking an altercation with a co-worker

--is enrolled at a university

--charged for domestic violence

--helps an elderly woman carry her grocery bags

--is a Parent Teacher Association (PTA) member

--begins a weekly exercise regimen

--keeps a journal

--engages in acts of vandalism

--shoves a stranger forward in the crowded concessions line at an athletic event

--teaches a friend how to play a new card game

--has a loving relationship with his/her parents

--commits armed robbery

--is in a relationship with a significant other

--saves a portion of every paycheck for a new car

--antagonizes house pets

--operates a vehicle while intoxicated

REFERENCES

Acorn, D. A., Hamilton, D. L., & Sherman, S. J. (1988). Generalization of biased perceptions of groups based on illusory correlations. Social Cognition, 6, 345-372.doi:10.1521/soco.1988.6.4.345

Anderson, N. H., & Barrios, A. A. (1961). Primacy effects in personality impression formation. The Journal of Abnormal and Social Psychology, 63(2), 346-350. doi:10.1037/h0046719

Belmore, S.M. & Hubbard, M.L. (1987). The role of advance expectancies in person memory. Journal of Personality and Social Psychology 53, 6170.doi:10.1037/0022-3514.53.1.61

Berndsen, M., van der Pligt, J., Spears, R., and McGarty, C. (1998). Expectation-based and data-based illusory correlation: The effects of confirming versus disconfirming evidence. European Journal of Social Psychology 26, 899913.doi:10.1002/(SICI)1099 -0992(199611)26:6<899::AID -EJSP795>3.0.CO;2-B

Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, W. B. G. (1996). Dispositional differences in cognitive motivation: The life and times of individuals varying in need for cognition. Psychological Bulletin, 119, 197-253. doi: 10.1037/0033-2909.119.2.197

Chaiken, S., & Trope, Y. (1999). Dual-process theories in social psychology. New York: Guilford Press.

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences. Mahwah, NJ: Lawrence Erlbaum Associates.

Corrigan, P. W. (Ed.) (2005). On the stigma of mental illness: Practical strategies for research and social change. Washington, D.C.: American Psychological Association Press.

Devine, P. G. (1989a). Automatic and controlled processes in prejudice: The role of stereotypes and personal beliefs. In A. R. Pratkanis, S. J. Breckler, & A. G. Greenwald (Eds.), Attitude structure and function (pp. 181-212). Hillsdale, NJ, England: Lawrence Erlbaum Associates, Inc.

Devine, P. G. (1989b). Overattribution effect: The role of confidence and attributional complexity. Social Psychology Quarterly, 52, 149-158. doi:10.2307/2786914

Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: Use of differential decision criteria for preferred and nonpreferred conclusions. Journal of Personality and Social Psychology, 63, 568-584. doi:10.1037/00223514.63.4.568

Fiske, S. T. (1992). Thinking is for doing: Portraits of social cognition from Daguerreotype to laserphoto. Journal of Personality and Social Psychology, 63, 877-889. doi:10.1037/0022-3514.63.6.877

Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed.). New York, NY, England: McGraw-Hill Book Company.

Fletcher, G.J.O., Danilovics, P., Fernandez, G., Peterson, D., and Reeder, G.D. (1986). Attributional complexity: An individual difference measure. Journal of Personality and Social Psychology, 51, 875-884.

Gilbert, D. T. (1989). Thinking lightly about others: Automatic components of the social inference process. In J. S. Uleman & J. A. Bargh (Eds.), Unintended thought (pp. 189-211). New York, NY, US: Guilford Press.

Gilbert, D. T., & Malone, P. S. (1995). The correspondence bias. Psychological Bulletin, 117, 21-38. doi:10.1037/0033-2909.117.1.21

Hamilton, D.L. & Gifford, R.K. (1976). Illusory correlation in interpersonal perception: A cognitive basis of stereotypic judgments. Journal of Experimental Social Psychology, 12, 392-407.

Hamilton, D.L., Dugan, P.M., & Trolier, T.K. (1985). The formation of stereotypic beliefs: Further evidence for distinctiveness-based illusory correlations. Journal of Personality and Social Psychology, 48, pp. 5-17.

Hamilton, D.L. & Sherman, S.J. (1989). Illusory correlations: Implications for stereotype theory and research. In Bar-Tal, D., Graumann, C.F., Kruglanski, A.W., & Stroebe, W. (Eds.), pp. 59-82, Stereotyping and prejudice. Changing conceptions. Springer-Verlag: New York.

Haslam, S.A., McGarty, C., & Brown, P.M. (1996). The search for differentiated meaning is a precursor to illusory correlation. Personality and Social Psychology Bulletin 22, 611-619.

Hinshaw, S. P. (2007). The mark of shame: Stigma of mental illness and an agenda for change. New York: Oxford University Press.

Hirt, E. R., Lynn, S. J., Payne, D. G., Krackow, E., & McCrea, S. M. (1999). Expectancies and memory: Inferring the past from what must have been. In I. Kirsch, I. Kirsch (Eds.), How expectancies shape experience (pp. 93-124). Washington, DC, US: American Psychological Association. doi:10.1037/10332-004

Johnson, C., & Mullen, B. (1993). The determinants of differential group evaluations in distinctiveness-based illusory correlations in stereotyping. British Journal of Social Psychology, 32, 253-263. doi:10.1111/j.20448309.1993.tb00999.x

Jones, E. E., & Davis, K. E. (1965). From acts to dispositions: The attribution process in person perception. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 2, pp. 219-266). New York: Academic press.

Kunda, Z., & Nisbett, R. E. (1986). The psychometrics of everyday life. Cognitive Psychology, 18, 195-224. doi:10.1016/0010-0285(86)90012-5

Levin, A. (2001, May 4). Violence and mental illness: Media keep myths alive. Psychiatric News. http://psychnews.psychiatryonline.org/doi/10.1176/ pn. 36. 9.0010

Link, B.G., Phelan, J.C., Bresnahan, M., Stueve, A., & Pescosolido, B.A. (1999). Public conceptions of mental illness: Labels, causes, dangerousness, and social distance. American Journal of Public Health, 89, 1328-1333.

Linville, P. W., Salovey, P., & Fischer, G. W. (1986). Stereotyping and perceived distributions of social characteristics: An application to ingroup-outgroup perception. In J. F. Dovidio, S. L. Gaertner, J. F. Dovidio, S. L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 165-208). San Diego, CA, US: Academic Press.

Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence. Journal of Personality and Social Psychology, 37, 2098-2109. doi: 10.1037/0022-3514.37.11.2098

McArthur, L.A. & Friedman, S.A. (1980). Illusory correlation in impression formation: Variation in the shared distinctiveness effect as a function of the distinctive person's age, race, and sex. Journal of Personality and Social Psychology 39, 615-624.

McGarty, C., Haslam, S.A., Turner, J.C., & Oakes, P.J. (1993). Illusory correlation as accentuation of actual intercategory difference: Evidence for the effect with minimal stimulus information. European Journal of Social Psychology 23, 391-410.doi:10.1002/ejsp.2420230406

Newman, L. S. (1996). Trait impressions as heuristics for predicting future behavior. Personality and Social Psychology Bulletin, 22, 395-411. doi: 10.1177/0146167296224006

Newman, L. S., Duff, K., Schnopp-Wyatt, N., Brock, B., & Hoffman, Y. (1997). Reactions to the O. J. Simpson verdict: 'Mindless tribalism' or motivated inference processes?. Journal of Social Issues, 53, 547-562. doi:10.1111/0022-4537.00034

Ottati, V., Bodenhausen, G. V., & Newman, L. S. (2005). Social psychological models of mental illness stigma. In P. W. Corrigan (Ed.), On the stigma of mental illness: Practical strategies for research and social change (pp. 99-128). Washington, D.C.: American Psychological Association Press.

Pasture, M. (2010). Analysis of variance for binomial data. Is the arcsine transformation really necessary? GiornaleItaliano Di Psicologia, 37, 181-199.

Pryor, J. B. (1986). The influence of different encoding sets upon the formation of illusory correlations and group impressions. Personality and Social Psychology Bulletin, 12, 216-226. doi:10.1177/0146167286122008

Risen, J.L., Gilovich, T., & Dunning, D. (2007). One-shot illusory correlations and stereotype formation. Personality and Social Psychology Bulletin, 33, 1492-1502.doi: 10.1177/0146167207305862

Sanbonmatsu, D. M., Shavitt, S., & Sherman, S. J. (1991). The role of personal relevance in the formation of distinctiveness-based illusory correlations. Personality and Social Psychology Bulletin, 17, 124-132. doi: 10.1177/014616729101700202

Sherman, J. W., Kruschke, J. K., Sherman, S. J., Percy, E. J., Petrocelli, J. V., & Conrey, F. R. (2009). Attentional processes in stereotype formation: A common model for category accentuation and illusory correlation. Journal of Personality and Social Psychology, 96(2), 305-323. doi:10.1037/a0013778

Sherman, S. J., Sherman, J. W., Percy, E. J., & Soderberg, C. K. (2013). Stereotype development and formation. In D. E. Carlston (Ed.), The Oxford handbook of social cognition (pp. 548-574). New York, NY, US: Oxford University Press.

Skowronski, J. J., & Carlston, D. E. (1989). Negativity and extremity biases in impression formation: A review of explanations. Psychological Bulletin, 105, 131-142. doi: 10.1037/0033-2909.105.1.131

Spears, R., & Stroebe, W. (2015). Two (or more?) cognitive approaches to stereotype formation: Biased or reality based?. In S. J. Stroessner, J. W.

Sherman, S. J. Stroessner, J. W. Sherman (Eds.) , Social perception from individuals to groups (pp. 141-158). New York, NY, US: Psychology Press.

Srull, T. K., & Wyer, R. S. (1989). Person memory and judgment. Psychological Review, 96, 58-83. doi:10.1037/0033-295X.96.1.58

Stangor, C. & McMillian, D. (1992). Memory for expectancy-congruent and expectancy-incongruent information: A review of the social and social developmental literatures. Psychology Bulletin 111, 42-61.doi:10.1037/00332909.111.1.42

Stroessner, S. J., Hamilton, D. L., & Mackie, D. M. (1992). Affect and stereotyping: The effect of induced mood on distinctiveness-based illusory correlations. Journal of Personality and Social Psychology, 62, 564-576. doi:10.1037/0022-3514.62.4.564

Stroessner, S.J., & Plaks, J.E. (2001). Illusory correlation and stereotype formation: Tracing the arc of research over a quarter century. In G.B. Moskowitz (Ed.), Cognitive social psychology: The Princeton symposium on the legacy and future of social cognition, pp. 247-259. Mahwah, NJ: Erlbaum.

Tetlock, P. E., & Boettger, R. (1989). Accountability: A social magnifier of the dilution effect. Journal of Personality and Social Psychology, 57, 388-398. doi:10.1037/0022-3514.57.3.388

Thornicroft, G. (2006). Shunned: Discrimination against people with mental illness. New York: Oxford University Press.

Todd, P. M., & Gigerenzer, G. (2007). Environments that make us smart: Ecological rationality. Current Directions in Psychological Science, 16, 167171. doi:10.1111/j.1467-8721.2007.00497 .x

Wahl, O. F. (1995). Media madness: Public images of mental illness. New Brunswick, NJ: Rutgers University Press.

Warton, D. I., & Hui, F. K. (2011). The arcsine is asinine: The analysis of proportions in ecology. Ecology, 92, 3-10.

(1) In keeping with McArthur and Friedman's (1980) study, the only other one in the literature to utilize naturally occurring stigmatized groups in a distinctiveness-based illusory correlation experiment (and to avoid the negative implications for statistical power of adding an additional between-subjects variable), we did not also run a condition with positive behaviors in the minority.

(2) In addition, whether or not the behaviors were violent did not play a significant role in the study's results, suggesting that this difference was not a crucial determinant of the key findings.

(3) In addition, an uninterpretable main effect of Other Group Type was found, F (1,111) = 8.15, p = .005, np1 2 3 = .068. Regardless of behavior valence, the type of negative behavior (violent or nonviolent) or whether the group in the minority consisted of people with mental illness or not, behaviors were overattributed to the minority more when the control group consisted of tobacco users (positive, 25.6%; negative, 38.1%) than when it consisted of only children (positive, 12.0%; negative, 26.7%).

(4) The heterogeneous standard deviations in Table 1 reflect the fact that the distributions of the proportion variable were significantly skewed. Arcsine transformations are often recommended for proportion data, but their validity has been called into question (Pasture, 2010; Warton & Hui, 2011). In order to further assess the most important finding reported in this section--that relative to positive behaviors, participants overattributed negative behaviors to the minority group--a Wilcoxon Signed Rank test (which does not assume a normal distribution) was conducted. A one-tailed test replicated the difference (Z = 1.60, p = .05).

(5) Supplementary analyses illustrated that the boost in recall for negative minority group behavior for the middle 20% of scorers on the Attributional Complexity Scale was mostly due to falsely attributing negative behavior to the minority group.

(6) The moderate group included 24 participants, while the other two included only 23. Thus, the analysis for the moderate group was higher in statistical power. However, even after dropping the moderate AC participant with the majority-minority rating discrepancy that was the greatest in magnitude (3.2), the difference remained significant (p = .012).

Katrina Aberizk, Leonard S. Newman & Rikki H. Sargent

Syracuse University

Author info: Correspondence should be sent to: Dr. Leonard S. Newman, Syracuse University, Psychology Department, 430 Huntington Hall Syracuse, NY 13244 lsnewman@syr.edu
TABLE 1 Percentage Overattribution of behavior by Minority Group
(People with mental illness[MI], Other/Control), Negative Behavior
Type (Nonviolent, Violent), and Behavior Valence (Positive/desirable,
Negative/undesirable)

             MI
             Positive        Negative
             Behaviors       Behaviors

Nonviolent   18.0 % (22.9)   44.8 % (43.0)
Behavior
Violent      10.0 % (20.3)   5.8 % (46.3)
Behavior

             Other
             Positive        Negative
             Behaviors       Behaviors
Nonviolent   31.5 % (23.4)   25.8 % (31.1)
Behavior
Violent      15.6 % (36.3)   53.3% (58.3)
Behavior

Note: Standard deviations are inside the parentheses.

TABLE 2 Ratings by Group (People with mental illness [MI], Other/
Control) & Minority Status

                        Group Rated

                        People with          Other/
Group in the minority   Mental Illness       Control

MI                      5.60 (1.37)          6.72 (1.35)
Other/Control           6.56 (1.46)          5.78 (1.24)

Note: Standard deviations are inside the parentheses. Cell means could
range from 1 to 10, with higher ratings being more favorable.

TABLE 3 Hierarchical Multiple Regression Analysis of Linear &
Quadratic Effects of Attributional Complexity on Number of
Negative Behaviors Remembered as being Associated with the
Minority Group

         Predictor       [DELTA]     B       t value   p
                         [R.sup.2]                     value

Step 1   Attributional   .003        -.05    0.59      .56
         Complexity
Step 2   Attributional   .032        -53 *   1.96      .05
         Complexity
         Squared

TABLE 4 Mean Number of Positive & Negative Behaviors Recalled as
Being Associated with the Minority Group, by Level of Attri
-butional Complexity (AC)

            Valence
AC Level    Positive        Negative

Low         10.48 (2.59)    5.09 (2.11)
Moderate    11.00 (1.91)    5.75 (2.03)
High        10.78 (2.52)    4.61 (1.73)

Note: Standard deviations are inside the parentheses.

TABLE 5 Mean Desirability Ratings Assigned to Minority & Majority
Group, by Level of Attributional Complexity (AC)

Group
AC Level     Minority      Majority

Low          5.33 (1.20)   6.13 (1.61)
Moderate     5.85 (1.26)   6.59 (1.30)
High         6.03 (1.04)   6.57 (1.37)

Note: Standard deviations are inside the parentheses. Cell means could
range from 1 to 10, with higher ratings being more favorable.


Please Note: Illustration(s) are not available due to copyright restrictions.
COPYRIGHT 2017 North American Journal of Psychology
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Aberizk, Katrina; Newman, Leonard S.; Sargent, Rikki H.
Publication:North American Journal of Psychology
Date:Dec 1, 2017
Words:9380
Previous Article:The Macdonald Triad Revisited: An Empirical Assessment of Relationships between Triadic Elements and Parental Abuse in Serial Killers.
Next Article:Facebook Research: Implications for Investigators in the Field of Media Psychology.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters