آنچه که شما انتظار دارید همان است که همیشه آن را بدست می اورید: نقش اندام، خوش بینی و بدبینی در فرایند تأیید رفتاری

Abstract Do pessimists and optimists elicit the very behavior they expect from others? What if their expectations are fairly extreme? Using a simulated job interview paradigm, evidence was found for behavioral confirmation of generalized future-event expectancies (optimism/pessimism) and for the moderating role of extremity. Interviewers with nonextreme expectancies gathered information in an expectancy-biased fashion and elicited expectancy-confirming behavior from applicants. However, as interviewer expectancies became more extreme, these effects were attenuated. Further evidence suggested that extremity is associated with effortful correction processes and awareness of bias. Interestingly, pessimistic applicants were more strongly influenced by interviewers’ expectancies than were optimistic applicants. The current study extends research on the social-cognitive consequences of generalized future-event expectancies and extremity to the behavioral domain.

Introduction As social perceivers, we are often aware of our expectations about the behaviors, traits, and abilities of other people. These interpersonal expectancies may be based on previous encounters with those individuals, hearsay, or our own stereotypes and biased ways of interpreting our social world. What we may be less aware of is that these expectancies can have a profound influence on our own and others’ behavior, ultimately affecting the course of social interactions. Specifically, expectations can create self-fulfilling prophecies (Merton, 1948), whereby people elicit behavior from others that is consistent with their expectations (behavioral confirmation; for reviews, see Jussim, 1986; Miller & Turnbull, 1986; Neuberg, 1996b). For example, studies have demonstrated behavioral confirmation of teachers’ expectancies about students (e.g., Jussim & Eccles, 1992; Rosenthal & Jacobson, 1968), interviewers’ expectancies about job applicants (e.g., Neuberg, 1989; Word, Zanna, & Cooper, 1974), and counselors’ expectancies about clients (e.g., Copeland & Snyder, 1995). How can one person’s expectations affect another person’s behavior? Analyses of the behavioral confirmation process (e.g., Darley & Fazio, 1980; Miller & Turnbull, 1986; Snyder, 1992) suggest that a perceiver first adopts expectations about a target person, then treats the target according to these expectations (e.g., via biased information-gathering). The target then responds in accord with the perceiver’s expectations (behavioral confirmation). For example, a perceiver with a negative expectancy might ask negatively biased questions that constrain the target’s responses, thus eliciting unfavorable behavior from the target. If the perceiver then interprets the target’s behavior in an expectancy-consistent fashion, above and beyond the evidence provided by the target’s actual behavior, perceptual confirmation has also occurred. Under what circumstances do one person’s expectations affect another person’s behavior? In keeping with the current emphasis in social psychology on the role of motives and goals in directing cognition and behavior (see Gollwitzer & Bargh, 1996), recent research has examined perceiver and target goals as moderators of behavioral confirmation (for reviews, see Neuberg, 1996a; Snyder, 1992). For example, accuracy goals tend to motivate perceivers to gather a broad range of target information and to interpret that information in an unbiased fashion, thus reducing expectancy confirmation (e.g., Darley, Fleming, Hilton, & Swann, 1988; Neuberg, 1989). For targets, behavioral confirmation is enhanced when they desire to get along with the perceiver (Snyder & Haugen, 1995) or are encouraged to be deferential (Smith, Neuberg, Judice, & Biesanz, 1997). However, behavioral confirmation is reduced when targets focus on promoting their own agendas (Smith et al., 1997) or are highly certain of their own personality characteristics (Swann & Ely, 1984). Generalized outcome expectancies In most studies of behavioral confirmation, the expectancies of interest have been relatively explicit, have been experimentally manipulated, and have pertained to the states, traits, abilities, and actions of another individual. Although some studies have examined behavioral confirmation of expectancies associated with category knowledge, stereotypes, and implicit personality theories (for reviews, see Claire & Fiske, 1998; Hamilton, Sherman, & Ruvolo, 1990), previous studies have not examined behavioral confirmation of trait-driven expectancies—that is, expectations that stem from relatively stable individual difference factors (e.g., personal knowledge structures, cognitive styles). One such expectancy is the generalized future-event expectancy (optimism/pessimism; Andersen, 1990; Andersen, Spielman, & Bargh, 1992). Andersen and her colleagues argued that people vary in their tendencies to think about the future; some expect primarily positive things to happen to them and others expect more negative events to occur. Furthermore, these generalized outcome expectancies tend to be applied to both the self and others. Because such chronic expectancies are broadly applicable and are not limited to specific categories of people or to specific traits, they may have a pervasive influence on social judgments and social interactions. Indeed, recent studies have provided evidence that perceivers’ generalized future-event expectancies can influence their inferences regarding another person’s performance in an assimilative fashion (Reich & Weary, 1998; Weary & Reich, 2001; Weary, Reich, & Tobin, 2001). More specifically, these studies have shown that when perceivers are cognitively busy, pessimistic perceivers make less favorable inferences about the ability and performance level of a target than do optimistic perceivers. However, when perceivers have sufficient motivation and cognitive resources, they search and correct for other factors that could have contributed to a target’s outcome (e.g., task difficulty). Correction for these factors reduces the net impact of perceivers’ generalized expectancies on their inferences about the target. In addition to influencing social judgments, might generalized future-event expectancies also lead perceivers involved in social interactions to exhibit expectancy-consistent behaviors? If so, might perceivers’ behaviors, in turn, elicit expectancy-confirming behavior from others?

Results Results of pilot study A pilot study involving only interviewer participants examined whether generalized future-event expectancies influence specific expectations about the interview and about the applicant’s likely qualifications, as predicted. The procedure was identical to that of the main study up to the point where interviewers indicated that they were ready to conduct the interview. At this point, participants anonymously completed a brief questionnaire to assess their specific expectations, and placed it within a stack of similar questionnaires ostensibly filled out by other participants. They then were informed that they would not conduct the interview and were asked to complete a few scales, including the FES (Andersen, 1990). They then were debriefed, asked for permission to use their expectancy ratings, thanked, and dismissed. Participants were 27 college students (10 males and 17 females) who completed the pilot study in exchange for course credit. Scores on the FES ranged from −34 to 98 (M=48.56, SD=36.42) and higher scores indicated optimism. Significant correlations (p<.05) indicated that as interviewers’ generalized future-event expectancies became more positive, they expected the interview to go more smoothly (r=.48), expected to perform better as interviewers (r=.45), and expected the applicant to be more qualified overall (r=.43) and to have better specific qualifications (r=.39). 1 The latter measure was an index of expectancies averaged across three items (i.e., interpersonal skills, motivation, and problem-solving skills; Cronbach’s α=.78). Preliminary analyses and descriptive statistics for the future event scale (FES) Initial analyses for the main study revealed that one interviewer was an outlier on both the Beck Depression Inventory (score = 49, indicating severe depression) and the FES (score=−130); the data for this interviewer and her interaction partner were omitted from further analysis. For the primary regression analyses, the data were examined with regard to multicollinearity problems, outliers, and regression assumptions. One outlier was found across measures. Further probing revealed a large inconsistency between the prescreening FES (score = 41) and time-of-session FES scores (score = 94), making assessment of the interviewer’s expectancies suspect; the data for this interviewer and his interaction partner therefore were omitted from further analyses. The final data set adequately met the regression assumptions. For the remaining 93 sessions, interviewers’ FES scores ranged from −63 to 120 (M=43.66; SD=43.07; skew=−0.57). Applicants’ FES scores ranged from −40 to 106 (M=44.91; SD=27.71; skew=−0.47). Scores on the FES and the BDI were significantly correlated for interviewers (r=−.60) and for applicants (r=−.41). BDI scores ranged from 0 to 23 for interviewers (M=5.46; SD=5.25; skew=0.97), and from 0 to 26 for applicants (M=5.15; SD=5.95; skew=1.89); only 3% of participants scored above 20 on the BDI. Primary analyses Interviewer impressions of applicants It was hypothesized that pessimistic interviewers would form more negative impressions of their applicants than would optimistic interviewers. However, as interviewer expectancies become more extreme, correction processes might attenuate this effect. If so, the function that would best capture the effect of interviewer expectancies on their impressions of applicant performance would be cubic. Expectancy-effects on interviewer impressions (whether linear or cubic) were hypothesized to be more pronounced for pessimistic applicants, resulting in an interaction with applicant expectancies. Because all lower-order terms must be included in the model to evaluate the effects of higher-order terms (Cohen et al., 2003), a hierarchical regression analysis was conducted.2 Applicant and interviewer FES scores were centered. From these centered variables, quadratic and cubic interviewer FES terms were created, as were terms representing the interaction between applicant FES and the linear, quadratic, and cubic components of interviewer FES. These terms were entered into the hierarchical regression equation as follows: applicant expectancies and interviewer expectancies in block 1, the interaction between these linear terms in block 2, the quadratic interviewer FES term in block 3, the interaction between applicant FES and the quadratic term in block 4, the cubic interviewer FES term in block 5, and the interaction between applicant expectancies and the cubic term in block 6.3 No significant effects on the Interviewer Impression index emerged. However, this finding does not directly address the issue of perceptual confirmation, which must be assessed in reference to an objective evaluation of target behavior and will be examined later. Applicant behavior Applicants with pessimistic interviewers were expected to perform more poorly than those with optimistic interviewers, as rated by objective judges. However, as interviewers’ expectancies become more extreme, expectancy effects might be attenuated, resulting in a cubic function for interviewer expectancies predicting applicant performance. Again, applicant expectancies should moderate the effects of interviewer expectancies on applicant behavior. The same hierarchical regression conducted on the Interviewer Impression index was conducted on the Applicant Behavior index. Only the two predicted effects emerged. Thus, the cubed term for interviewer expectancies was significant in block 5 (β=−.20, p<.05), as was the interaction between applicant FES and the cubed term for interviewer FES in block 6 (β=.25, p<.05). A graphical depiction of these results can be seen in Fig. 1. To explore this cubic relationship, simple slopes were examined at the mean of interviewer expectancies and at one and 2 SDs above and below the mean. Recall that additional interviewers were recruited from the extreme ends of the FES distribution. Because this sampling procedure artificially inflated the standard deviation and affected the mean, more typical values of the mean (M=39.81) and standard deviation (SD=32.69) from the full prescreening distribution of FES scores (N=3460) were used to define the points at which to evaluate the simple slopes (see Table 1). Because the cubic relationship between interviewer expectancies and applicant behavior was moderated by applicant expectancies, the cubic function was examined at the mean and at 1 SD above and below the mean for applicant expectancies. Judges’ ratings of Applicant Behavior as a function of interviewers’ generalized ... Fig. 1. Judges’ ratings of Applicant Behavior as a function of interviewers’ generalized future-event expectancies (centered). The function is plotted separately for pessimistic (Future Event Scale (FES) score = 17; −1 SD), optimistic (FES = 73; +1 SD), and neutral (FES M=45) applicants. Figure options Table 1. Simple slopes for regression of Applicant Behavior index and interviewer Information-Gathering index on interviewer generalized future-event expectancies, and full regression equations Interviewer expectancies (FES) −2 SD −1 SD M +1 SD +2 SD Applicant behavior index (N=91) Applicant expectancies (FES) Pessimistic (−1 SD) B −.0549** .0046 .0259** .0102 −.0436+ SE B .018 .006 .008 .008 .025 β −2.405** 0.201 1.135** 0.446 −1.913+ Neutral (M) B −.0240* .0039 .0135** .0053 −.0213 SE B .009 .004 .005 .005 .015 β −1.051* 0.171 0.591** 0.230 −0.933 Optimistic (+1 SD) B .0069 .0032 .0011 .0003 .0011 SE B .010 .007 .008 .007 .025 β 0.303 0.142 0.046 0.015 0.046 Information-Gathering index (N=87) B −.0140* .0031 .0076* −.0001 −.0205+ SE B .006 .003 .004 .004 .011 β −0.775* 0.171 0.419* −.008 −1.136+ Standardized regression equations Applicant Behavior index: Full-size image (<1 K) Information-Gathering index: Full-size image (<1 K) Note. Interviewer FES (Future Event Scale) values at which simple slopes were examined for the cubic effects on both Applicant Behavior and Information-Gathering were −26, 7, 40, 72, and 105. Applicant FES values at which simple slopes were evaluated for the cubic by linear interaction on Applicant Behavior were 17, 45, and 73. In the equations, “Y” represents dependent variable scores; “i” represents interviewer FES scores; and “a” represents applicant FES scores. ** p<.01. + p<.10. * p<.05. Table options First, the simple slopes for interviewer expectancies predicting applicant behavior were examined for pessimistic applicants (applicants at 1 SD below the mean of applicant FES). When pessimistic applicants had interviewers with a mean score on the FES, the simple slope was significant and positive. Thus, as interviewer expectancies became more positive, applicant performance improved, providing evidence of behavioral confirmation. However, this positive relationship between expectancies and applicant behavior showed a reversal as interviewer expectancies became more extreme (see Fig. 1). Specifically, when interviewers were at 2 SDs below the mean of interviewer FES, a significant negative slope revealed that applicants’ performance improved as interviewer expectancies became more extreme in pessimism. At 2 SDs above the mean for interviewer FES, the marginally significant simple slope was also negative, revealing that applicants’ performance tended to decline as interviewers become more extreme in optimism. The points at 1 SD above and below the mean of interviewer FES were close to the minimum and maximum values of the function and therefore had small, nonsignificant simple slopes. Next, the simple slopes for interviewer expectancies predicting applicant behavior were examined for applicants at the mean of applicant FES. When these applicants had interviewers with a mean score on the FES, the simple slope was significant and positive, again providing evidence of behavioral confirmation. When interviewers were at 2 SDs below the mean of interviewer FES, a significant negative slope revealed that applicants’ performance began to improve as interviewers became more extreme in pessimism. When interviewers were at 2 SDs above the mean, the negative slope was nonsignificant. Finally, the simple slopes were examined for applicants at 1 SD above the mean of applicant FES. For these optimistic applicants, none of the simple slopes were significant. Therefore, as applicants became more optimistic, they were less susceptible to influence by interviewers’ expectancies, as predicted. Given that the job-related and job-unrelated traits did not load on separate factors, the pattern of judges’ ratings on these sets of measures did not appear to differ substantially. To more closely examine this generalizability, the same regression analysis reported for the full Applicant Behavior index was conducted on an index created by averaging the 3 job-unrelated traits (Cronbach’s α=.71). The cubic effect of interviewer expectancies was significant in block 5 (β=−.25, p<.01), and the interaction was significant in block 6 (β=.22, p<.05). Thus, the same pattern of effects found for the full Applicant Behavior index extended to job-unrelated traits, suggesting that the influence of perceivers’ generalized expectancies on targets’ behavior is not limited to situationally relevant behavioral domains. Interviewer information-gathering As rated by objective judges, pessimistic interviewers were expected to gather information in a less positive manner than optimistic interviewers. However, as interviewers’ expectancies become more extreme, this effect may be attenuated, resulting in a cubic function for interviewer expectancies predicting information gathering. Theoretically, interviewers’ expectancies should directly influence their information-gathering strategies (some of which are planned prior to initiating the interaction). Thus, the moderating influence of applicants’ expectancies should be less evident for these measures than for the applicant behavior measures. The full hierarchical regression analyses performed on the applicant talk time measure, the questions per minute measure, and the question quality index all revealed a tendency toward fitting a cubic function; however, only for Question quality did the cubic term reach significance (p<.02). Each individual index will be examined in a subsequent section. However, to obtain a more reliable index and a single measure to use in subsequent mediational analyses, the three measures were standardized and averaged (after reverse-scoring questions per minute) to form the Information-Gathering index (one factor; Cronbach’s α=.49). High scores on this index indicate that interviewers were using relatively positive information-gathering strategies. A full hierarchical regression analysis performed on the Information-Gathering index revealed no main effect or interaction with applicant expectancies. Therefore, a simplified three-block hierarchical regression analysis was examined with the linear, quadratic, and cubic terms for interviewer expectancies as predictors. As predicted, the cubic effect of interviewer expectancies was significant in block 3 (β=−.20, p<.05). A graphical illustration of this effect can be seen in Fig. 2 and simple slopes are reported in Table 1. At the mean of interviewer expectancies, the simple slope was significant and positive, indicating that as expectancies became more positive, interviewers engaged in more positive information-gathering behaviors. At 2 SDs below the mean, the significant, negative simple slope revealed that information-gathering strategies became more positive as interviewer expectancies became more negative. At 2 SDs above the mean, the simple slope was also negative, but marginally significant. The points at 1 SD above and below the mean were close to the minimum and maximum values of the function and had nonsignificant simple slopes. Interviewer Information-Gathering as a function of interviewers’ generalized ... Fig. 2. Interviewer Information-Gathering as a function of interviewers’ generalized future-event expectancies (centered). Figure options Clearly, interviewers with nonextreme expectancies were gathering information in an expectancy-consistent fashion, but this tendency decreased and reversed as expectancies became more extreme. At what points along the expectancy continuum did interviewer behavior begin to shift, thus suggesting that a correction process may have been activated? The predicted scores on the Information-Gathering index were at the lowest point of the concave upward curve at an FES score of −1.25 (slightly below the scale midpoint) and at the highest point of the concave downward curve at 71.66. That is, information-gathering strategies first began to show evidence of correction when FES scores fell below −1.25 and above 71.66. Ancillary analyses To further probe the information-gathering measures, mediational issues, and several ancillary measures, a simpler analysis strategy was sought. As noted by Cohen et al. (2003, p. 212), the measurement error inherent in social sciences data makes it especially difficult to fit higher order equations, such as the full equation containing the cubic term interaction in the current study. Therefore, an objective cutoff point for extremity was sought. Given that negative expectancies may be most closely linked to correction processes, the point at which pessimistic interviewers first began to shift to more positive information-gathering strategies (FES=−1.25), was used to define extremity for the pessimistic end of the scale. A large prescreening distribution of FES scores (N=3460) gathered concurrently with this study revealed that about 10.1% of scores fell below −1. Likewise, about 10.3% of scores fell above 79; for the sake of consistency, this point was used to define extremity for the optimistic end of the scale. 4 Thus, participants were categorized as having extreme FES scores if their experimental session scores fell below −1 (n=16) or above 79 (n=21); scores between −1 and 79 (n=56) were coded as nonextreme. Note that the continuous expectancy variable and the categorical extremity variable are orthogonal, despite being derived from the same measure. Mediational analyses Theoretically, the effects of interviewers’ generalized future-event expectancies on applicants’ performance should be mediated by interviewers’ information-gathering behaviors. Although structural equation modeling techniques would seem ideally suited for revealing such relationships, they were deemed inappropriate due to the small sample size and large number of parameters to be estimated.5 Instead, mediational analyses using the procedures outlined by Baron and Kenny (1986) were conducted. Because the predicted effects of interviewer expectancies varied as a function of extremity, it seemed most informative to examine the mediational hypotheses for nonextreme and extreme interviewers separately6. These analyses did not take into account the moderating role of applicant expectancies, reflecting instead the average effects across all applicants. See Fig. 3 for regression coefficients. Interviewer Information-Gathering as a mediator of the effects of interviewers’ ... Fig. 3. Interviewer Information-Gathering as a mediator of the effects of interviewers’ generalized future-event expectancies on Applicant Behavior. Models show standardized regression coefficients. *p<.05; **p<.01. Figure options First, the nonextreme interviewer group was examined. The Applicant Behavior index was regressed on interviewer FES, revealing a significant effect. Next, the Information-Gathering index was regressed on interviewer FES and a significant effect emerged. Finally, Applicant Behavior was regressed on both interviewer FES and the proposed mediator, revealing that interviewers’ information-gathering behaviors predicted applicants’ behavior in the interview when the effects of the interviewer expectancies were controlled. Moreover, interviewer FES was no longer a significant predictor of Applicant Behavior when Information-Gathering was included in the regression equation. Based on Baron and Kenny’s (1986) modification of the Sobel test, the reduction in the path from interviewer expectancies to Applicant Behavior was significant (z=3.24, p<.01). Thus, for interviewers with nonextreme expectancies, information-gathering behaviors mediated the effects of interviewers’ expectancies on applicant performance. The same set of regression analyses conducted for the extreme interviewer group revealed that interviewer FES did not predict Applicant Behavior; thus, no test of mediation was appropriate. Interestingly, interviewer FES negatively predicted Information-Gathering, suggesting that extremely pessimistic interviewers used more favorable strategies for gathering information than did extremely optimistic interviewers. Perceptual confirmation Judges’ ratings of applicants are based on applicants’ actual behavior, while interviewers’ ratings theoretically should be based both on applicants’ behavior and on their own expectancy-related cognitive biases. Unbiased interviewer impressions should be nearly identical to those of naı̈ve judges. Thus, evidence of perceptual confirmation requires that interviewers’ impressions be influenced by their cognitive biases, above and beyond the behavioral confirmation effects (Neuberg, 1989). That is, the difference between pessimistic and optimistic interviewers’ impressions of their applicants should be greater (and more consistent with expectancy valence) than the difference between judges’ impressions of those same applicants. In regression analysis, perceptual confirmation would be evident if the slope of the line for interviewer expectancies predicting interviewer impressions were more steeply positive than the slope of the line for interviewer expectancies predicting judges’ impressions. For nonextreme interviewers, the slope of the line for interviewers’ impressions (β=.12, p>.39) was less steep than for judges’ impressions (β=.46, p<.01); the difference between these slopes was significant (z=2.00, p<.05). Because interviewers’ impressions were less expectancy-consistent than those of the judges, these findings provide evidence of perceptual disconfirmation. For extreme interviewers, the slope for interviewer impressions was more positive (β=.24, p>.16) than the slope for judges’ impressions (β=−.06, p>.73); however, this difference was not significant. Information-gathering revisited Using a dummy-coded variable to represent the extremity of interviewer’s expectancies, the three separate measures of information-gathering were examined more closely. For interviewers with nonextreme expectancies, as expectancies become more optimistic, more positive information-gathering strategies were expected. However, for interviewers with extreme expectancies, this effect should be attenuated. Thus, an interaction between interviewer expectancy and extremity was predicted (analogous to the cubic effect in previous analyses). To test this prediction, continuous interviewer FES and dummy-coded extremity were entered in block 1, and their interaction was entered in block 2. Additional analyses revealed no main effect or interactions with applicant FES on the measures. For the Question Quality index, only the predicted interaction between expectancy and extremity was found (β=−.64, p<.01). For interviewers with nonextreme expectancies, the significant simple slope was positive (β=1.02, p<.01), such that as expectancies became more positive, interviewers asked questions of higher quality. For interviewers with extreme expectancies, the significant simple slope was negative (β=−.28, p<.01), such that extremely pessimistic interviewers asked higher quality questions than extremely optimistic interviewers. The Expectancy × Extremity interaction alone was also found on the questions per minute measure (β=.38, p<.01). Simple slopes revealed that for interviewers with nonextreme expectancies, as expectancies became more positive, interviewers asked fewer questions per minute (β=−.69, p<.01), thereby allowing their applicants more time to answer each question. However, for those with extreme expectancies, the simple slope was nonsignificant (β=.09). Finally, the measure of applicant talk time also showed only a significant Expectancy × Extremity interaction (β=−.37, p<.01). For interviewers with nonextreme expectancies, applicant talk time increased significantly as expectancies became more positive (β=.61, p<.05). For those with extreme expectancies, the simple slope was nonsignificant (β=−.14). Supporting evidence for effortful correction The results on the interviewer information-gathering measures suggest that interviewers with extreme expectancies might have been attempting to correct for their usual tendencies to be optimistic or pessimistic about other people. Were extreme interviewers putting forth more effort and trying harder to conduct an unbiased interview? To address this question, two effort-related measures were regressed on interviewer expectancies and extremity in block 1, and their interaction in block 2. A main effect of extremity on the total interview time (M=333.22, β=.23, p<.01, B=68.21, p<.05) revealed that, on average, extreme interviewers extended the interview 68.21 s (SE=30.42) longer than did nonextreme interviewers. On two measures, applicants indicated the extent to which their interviewers seemed motivated (a) to be fair and impartial and (b) to form an accurate impression; these measures were averaged (Cronbach’s α=.86). Only a main effect of extremity emerged on this index, indicating that applicants perceived interviewers with extreme expectancies to be more motivated than interviewers with nonextreme expectancies to make a fair, accurate judgment of their qualifications (β=.21, p<.05). Addressing alternative explanations Expectancy violation Instead of being more aware of their expectancies from the start, perhaps interviewers with extreme expectancies initially engaged in expectancy-consistent questioning, as did their nonextreme counterparts, but then altered their strategy due to expectancy violation during the course of the interview. To address this possibility, a new measure of question quality for the first four questions of each interview was created, in addition to a measure for the remaining questions of the interview. These two measures were subjected to the same regression analysis reported above for the full question quality measure. Subsequent analyses revealed no main effect or interactions with applicant FES on the measures. If expectancy-consistent questioning occurred for both extreme and nonextreme interviewers early in the interview, a main effect of expectancy for the early question quality measure should be revealed. However, only a significant interaction between expectancy and extremity emerged on the question quality measures for the first four questions (β=−.44, p<.01), and for the remaining questions (β=−.56, p<.01), with a pattern consistent with that found for the full measure. The simple slopes for extreme interviewers were nearly identical for the early (β=−.35, p<.01) and late (β=−.36, p<.01) questions. Thus, interviewers with extreme expectancies engaged in expectancy-inconsistent information-gathering behavior early in the interview, providing further evidence of an a priori, awareness-based correction strategy. Depression To ensure that the expectancy effects were due to chronic expectancies for the future, rather than to depression, analogous regression analyses using interviewer and applicant BDI scores in place of FES scores were conducted for the three primary dependent measures.7 No significant effects were found on Applicant Behavior, Interviewer Impressions, or Information-Gathering (all p’s > .10). Thus, there is no evidence that the effects found in the current study can be accounted for by depression-related symptomatology. Results of follow-up study In the primary study, interviewers with extreme expectancies appeared to engage in correction processes. Although extremity may lead to awareness of bias, and prior studies have supported the notion that awareness of bias leads to correction processes, the link between extremity of generalized expectancies and awareness of bias was not directly shown in the current study. Therefore, two measures were added to an ongoing social judgment experiment and were completed by 120 college students (36 males and 84 females) whose FES scores ranged from −58 to 99 (M=43.67; SD=26.85; skew=−0.79). One measure, assessing awareness of expectancies about others, asked participants to indicate their agreement with the statement, “In general, I tend to be very optimistic about the abilities, intentions, and behaviors of other people,” on a 9-point scale. The other, assessing awareness of bias, required participants to complete the statement, “After thinking more carefully about a person I’ve just met, my judgments tend to…” using a scale ranging from “becomes more negative” (−4) to “remains the same” (0) to “becomes more positive” (+4). Hierarchical regression analyses were conducted with centered FES entered at block 1, the quadratic term at block 2, and the cubic term at block 3. The first measure revealed only a linear relationship (β=.37, p<.01), such that as future event expectancies became more positive, participants reported being more optimistic about others. The second measure also revealed a significant linear relationship in block 1 (β=.28, p<.01). However, in block 3, the cubic term accounted for a significant increment in variance (β=.41, p<.05). Fig. 4 graphically illustrates this cubic effect. To facilitate comparisons with the primary study, simple slopes, reported in Table 2, were examined at the same FES values of −26, 7, 40, 72, and 105 8, as well as at −58 (−3 SDs, based on the large FES distribution). The simple slope near the mean of the distribution was not significant, indicating no relationship between expectancies and awareness of bias when participant had nonextreme expectancies. However, as expectancies became more extreme, the positive simple slopes became significant, such that pessimistic (optimistic) participants reported that their judgments about others become more negative (positive). These results suggest that the extremity of perceivers’ expectancies is associated with an enhanced awareness that their judgments about others tend to be biased in the direction of their expectancies. Awareness of direction and degree of bias in social judgments as a function of ... Fig. 4. Awareness of direction and degree of bias in social judgments as a function of generalized future-event expectancies (centered). Figure options Table 2. Simple slopes for regression of awareness of bias on generalized future-event expectancies and full regression equation Expectancies (FES) −3 SD −2 SD −1 SD M +1 SD +2 SD Awareness of bias (N=120) B .1020* .0355+ .0024 .0053 .0425 .1160 SE B .050 .020 .009 .007 .013 .042 β 1.833* 0.638+ 0.044 0.096 0.765** 2.092** Standardized regression equation: Full-size image (<1 K) Note. Future Event Scale (FES) values at which simple slopes were examined for the cubic effect were −26, 7, 40, 72, and 105. In the equation, “Y” represents dependent variable scores and “i” represents interviewer FES scores. * p<.05. + p<.10. ** p<.01.