Judging Bias: Juror Confidence and Judicial Rulings on Challenges for Cause

Abstract

The judge in a jury trial is charged with excusing prospective jurors who will not be impartial. To assess impartiality, prospective jurors are typically asked whether they can be fair. Using an experimental paradigm, we found that small changes in jurors' self-reported confidence in their ability to be fair affected judges' decisions about bias but did not affect the judgments of either attorneys or jurors. We suggest why a judge's role and unique relationship with jurors is likely to foster a decision strategy based on reported juror confidence, and we discuss the implications of our analysis for current legal debates over jury selection practices. Unexpected patterns in our results also highlight the ways in which perceptions of impartiality are affected, in part, by the social characteristics of the observer.

The U.S. Constitution entitles a criminal defendant to trial by an impartial jury.¹ During jury selection questioning (called voir dire), trial judges are charged with evaluating whether each prospective juror is able and willing to evaluate evidence and to reach decisions with an open mind. This determination presents a challenging task. Ample evidence reveals that perceptions and decisions are necessarily influenced by prior beliefs and experience (e.g., Culhane et al. 2004; Lord et al. 1979; Hastie et al. 1983). Indeed, the diversity of backgrounds and experiences on the jury is counted as one of its great strengths (e.g., Peters v. Kiff 1972; Sommers 2006). The key, then, is to assess when the attitudes and beliefs that are formed by background and experience constitute unacceptable bias. If they do, the trial judge is obligated to excuse the prospective juror “for cause.”

The judge's behavior in making these decisions has been almost entirely ignored by researchers, while other forms of judicial behavior have attracted substantial recent attention from scholars (see, e.g., Epstein & Knight 2004; Wistrich et al. 2005). Researchers studying juries and jury selection have focused almost exclusively on attorney decisions, concentrating on the limited number of peremptory strikes an attorney can exercise without giving a reason. The attention to peremptory strikes arises, in part, because the predictions of attorneys about jurors' likely verdict preferences can be haphazard (Finkelstein & Levin 1997; Zeisel & Diamond 1978) and often appear to be racially biased (Baldus et al. 2001; Diamond et al. 1997; Rose 1999; Sommers & Norton 2007), in violation of established legal norms (e.g., Batson v. Kentucky 1986). A number of scholars (e.g., Baldus et al. 2001; Hoffman 1997; Marder 1995) have called for the elimination of the peremptory strike. Some current Supreme Court justices (Miller-el v. Dretke 2005, Breyer, J., concurring; Stevens 2003) have also questioned whether peremptory challenges do more harm than good. Yet missing from such debates and from the empirical literature is the stage of jury selection that precedes attorney choices, a stage that may remove an unlimited number of prospective jurors. With the exception of a few observational studies involving small sets of cases (e.g., Hannaford-Agor & Waters 2004; Nietzel & Dillehay 1982; Rose 2005; Shuy 1995), and a single experimental study examining the effects of pretrial publicity (Kerr et al. 1991), the possible determinants of judicial decisions to excuse for cause have received scant attention.

As we describe in more detail, judges ask jurors to report their own assessments of their capacity to be fair. Thus language is a crucial part of the decisionmaking process in this setting, and, as others have suggested (see, e.g., O'Barr 1982; Conley & O'Barr 2005), nuances in language may shape inferences and conclusions about the jurors' credibility and their ability to be fair. A few existing studies suggest that judges accept the claims of jurors who say they can be fair, even when such jurors face substantial threats to their abilities to be impartial. If judges accept claims to be fair as reliable indicators of impartiality, then, according to research and theory in psychology (e.g., Lindsay et al. 1989; Vrij 2000), sociology (Berger et al. 1986; Goffman 1959; Fisek et al. 2005), and anthropology (Conley et al. 1978), jurors who confidently communicate their impartiality should be less likely to be excused for cause than jurors who are otherwise identical but who express less confidence. Using a vignette experiment, we tested samples of legal professionals—both judges and attorneys—and found that subtle variations in the confidence jurors express about their fairness influence the likelihood that a judge will excuse a juror for cause. The confidence effect occurred even though both judges and attorneys rated the prospective jurors in the vignettes low on their likely ability to be fair, and even though, as other research has demonstrated, there are good reasons to suspect that juror confidence is not a reliable guide to bias.

Our study also goes beyond an examination of what judges do by comparing judicial judgments of bias with those of laypersons (prospective jurors). Our results demonstrate that non-judges do not perceive the small changes in juror expressions of confidence as diagnostic of how biased a prospective juror is likely to be. We suggest that confident language helps simplify a complicated decision, but that use of this cue depends upon concerns that are unique to judges. We consider the important implications of our findings for debates regarding the structure of jury selection. Finally, based on some unexpected findings from our juror samples, we highlight the ways in which a variety of social factors may condition perceptions of bias.

Voir Dire and Self-Assessments of Fairness

Jury selection procedures vary (see, e.g., Wetherington et al. 1999), but a significant part of voir dire is aimed at establishing whether a juror's background or attitudes raise any “red flags” about that person's ability to keep an open mind during the trial. Red flags could arise from relevant life experiences (e.g., having been a victim of a crime in the past), associations (e.g., knowing someone related to the case), or attitudes and fixed opinions about certain types of cases, such as those involving the death penalty (Wainwright v. Witt 1985), or any case characteristics that might raise strong feelings (e.g., sex abuse trials; Vidmar 1997). A few red flags lead axiomatically to an excuse for cause. For example, if a juror has a marital or blood relationship with one of the parties and/or a financial stake in the outcome of the case, many jurisdictions presume a conflict of interest (e.g., Texas Government Code §62.105). More often, however, the judge must decide on a case-by-case basis whether a potential threat to a juror's impartiality (e.g., having been a victim of a crime in the past) does in fact mean that the juror is unable to be fair (see, e.g., Blackstone 1765–9; U.S. v. Wood 1936).

Contributing to this assessment are jurors' own reports about how a given characteristic affects them (see Hannaford-Agor & Waters 2004). All jurors must affirm that they are able to listen to evidence impartially, consider only evidence and the law in decision making, and follow the law as given (see Balch et al. 1976; Johnson & Haney 1994). Based on the type of case, the red flag at issue, the juror's assessment of his or her abilities, and the juror's demeanor (e.g., Uttecht v. Brown 2007), a judge decides whether or not to believe the juror's claims of an ability to be fair. Doubts about the credibility of juror's self-assessments can take different forms. On the one hand, when a juror's background raises a red flag, and the juror says, “I cannot be fair,” the judge must be convinced that the juror is neither being unnecessarily modest about his or her true capabilities nor claiming bias as a pretext for getting out of service. Judges will often ask multiple follow-up questions in order to ensure that the person's feelings are genuine and that the individual understands that serving is a matter of duty, not personal preference (see Cosper 2003; Rose 2005). But absent any clear basis for disbelief, judges will generally excuse for cause any person who ultimately says, “I cannot be fair.”

A different type of credibility assessment takes place when prospective jurors claim they can be fair. Faced with a juror who promises to be fair, judges must consider whether that individual is misrepresenting him- or herself in order to be seated on the jury for improper reasons (e.g., someone who has been a victim of a crime who wishes to use the current case to “send a message”), or is perhaps underestimating the potential difficulty of remaining fair and impartial during the trial.

The context of voir dire provides several reasons to be concerned about the quality of jurors' claims of fairness. For one thing, by design, voir dire questions often convey social desirability; that is, the questions suggest that it is “better” to answer one way than another (see Sudman & Bradburn 1982). To prevent jurors from too easily asserting unfairness (e.g., to avoid having to serve), judges typically appeal outright to duty and to responsibility: “Are you saying that you would not listen to the instructions and follow the law as given to you?”; “Will your doubts make you unwilling to hear all the evidence before making up your mind?”; “Do you mean you cannot give both sides a fair trial?”; or simply, “So, you are saying you cannot do your duty in this case?” (see Diamond et al. 1997; Rose 2005). From their observational work, Balch and colleagues report: “In over two thousand replies by members of the jury panel, only twice did a prospective juror fail to give the expected response” (1976:278). Even if questions did not so explicitly invoke noble qualities, individuals recognize that fairness is a desirable characteristic, and most people want to believe that they possess it (see, e.g., Moran & Cutler 1991; Vidmar 2003).

Further, people often have difficulty producing accurate self-assessments of bias and find it difficult to estimate whether events or prior experiences are likely to influence them (see, e.g., MacCoun 1998; Nisbett & Wilson 1977). In addition, during voir dire prospective jurors receive only a minimal description of what they will hear and see during the trial, and they must give decontextualized answers about hypothetical situations (i.e., “If you were a juror in the case as we have described it, would you be fair?”). Restrictions on time available for questioning and the fact that inquiries are often posed publicly to all jurors, rather than privately to individuals, also limit the likelihood of accurate disclosures of bias (see Mize 1999; Nietzel & Dillehay 1982).

For all these reasons, judges must frequently make decisions about jurors whose life experiences, associations, or attitudes raise red flags about their capacity for fairness, but who assert that they can nevertheless remain fair and hear the case with an open mind. Generally speaking, trial judges are afforded great latitude to assess whether or not a potentially compromised person can, in fact, be fair (see Cosper 2003). Based on their interpretation of the juror's answers and demeanor, judges may choose to regard the juror's statement of fairness as credible (and, hence, choose not to excuse the juror for cause), or they may decide to “overrule” the person's own self-assessment and dismiss the juror for cause. When interpreting a juror's answers, different judges may well reach different decisions about the same juror. For example, in their recent ruling on a capital case from the state of Washington, five members of the U. S. Supreme Court held that the trial judge did not err in removing a juror for cause, finding that “on their face” the juror's answers suggested he could not serve fairly during the sentencing phase (Uttecht v. Brown 2007). However, the Ninth Circuit Court of Appeals, as well as the four dissenters on the Supreme Court, read the same transcript and concluded the juror could have served fairly and should not have been removed for cause.

Such disagreement speaks to the difficulty of determining whether a given individual—who is, typically, a complete stranger to the judge—can be fair in a given case. Given this complexity, judges are likely to look for simplifying decision strategies—termed “heuristics” (e.g., Tversky & Kahneman 1974)—to separate those who are likely to be fair from those who are not. As we next describe, there are several reasons why we expect the confidence of a juror's assertion to be a simplifying heuristic judges are likely to use.

Confidence and Perceived Juror Competence

Some prior research suggests that even though judges have the discretion to overrule the juror's self-assessment, they are loathe to do so. In her study of 13 felony trials, Rose (2005) reported that no prospective juror who stated that he or she would be fair was dismissed for cause, even when some life experiences suggested significant threats to that ability. Examples of these instances included a prospective juror in a rape trial who said that she had been raped as a child as well as someone who knew a defendant through the juror's prior work in a correctional facility—that is, she knew him as a former inmate; such people were dismissed through peremptory challenges (Rose 2003). In addition, far from neglecting self-reports, some judges appear to take an active role in shaping answers so that they conform to an image of neutrality (Cosper 2003; Zalman & Tsoudis 2005). Shuy (1995) analyzed a transcript of interviews with 14 jurors in a single death penalty trial and found that in various ways, the judge's questions led people to express greater openness to applying the death penalty than their initial answers would have suggested. These studies suggest that judges declare jurors to be fair when jurors say they can be fair.

Nevertheless, from the current literature, it is difficult to know exactly what influences judicial decisions about whether someone is or is not capable of keeping an open mind during a trial. Most observational studies report on a small number of cases or decisions. Any given outcome—e.g., a denial of a challenge for cause (Hannaford-Agor & Waters 2004; Rose 2005; Shuy 1995)—may or may not be typical of judicial behavior but may instead be a product of a given judge assessing a particular juror for a specific case. Critically, the causal role of jurors' self-reports remains ambiguous. The judges described in existing studies may have considered a red flag for a particular juror (e.g., having been a crime victim, having a particular stance toward the death penalty) and simply concluded that, in that instance, it posed no real threat. If that were the case, jurors' self-reports about impartiality simply bolstered, rather than drove, judges' final decisions about juror bias. In short, we cannot know whether a change in the confidence of the juror's self-assessment would have produced a different result.

We expect, however, that expressed self-confidence is likely to directly affect rulings on challenges for cause. Research in several disciplines converges to link confident forms of expression with impressions of competence. Goffman (1959) tied a “meek, apologetic” manner with lower status in an interaction and lower expectations from an observer, suggesting, for example, that hesitations in speech signal inappropriate levels of concern with an interaction. Sociologists working in the tradition of expectation states theory (e.g., Berger et al. 1986; Fisek et al. 2005) cite confidence as part of a “behavioral pattern” that raises expectations about the quality of future performance. According to both legal anthropologists and psychologists, in legal settings a confident style of testimony builds witness credibility (Conley et al. 1978; Lindsay et al. 1989). Those who confidently tell a story are less likely to be perceived as liars than those who pause and seem less assured of what they are saying (Hartwig et al. 2004; Vrij 2000). Although there are exceptions (as when the confident person seems too “cocky” or does not know his or her own limitations; e.g., Tenney et al. 2007), and although, as we discuss in more detail later, there are reasons to doubt the validity of confidence as a cue to fairness, confidence nevertheless “will almost always be an enabling factor for effective self-presentations” (DePaulo 1992:218).

The most direct way to demonstrate whether confident speech changes judges' rulings on juror bias is to use an experimental design and assess reactions to the same juror who responds in more or less confident ways to the same question. We predicted excuses for cause to be more likely when jurors use more confident words and phrasing (e.g., “I would be fair”) than when they use words that are more equivocal or uncertain (e.g., “I am pretty sure I would be fair”). This working definition of confidence follows O'Barr (1982), who says that the absence of hedges or qualifiers to a phrase (sometimes also termed “epistemic modals”; Aries 1996) constitutes a more “powerful” (i.e., confident) form of speech.

Current Study

We drew on appellate cases and instances from our own court observations to construct vignettes to manipulate; in each, the prospective juror had a problematic biography but also reported in more or less confident terms that he or she could be fair. We operationalized juror confidence through the phrasing of a single response that a prospective juror gave at the end of each vignette. We then asked legal professionals to rate the hypothetical juror's ability to be fair and impartial, and we asked respondents to indicate whether the “average judge” would excuse the person if an attorney challenged the juror for cause. Although all study participants read all of the vignettes, the particular subset of scenarios containing either confident or less-confident responses was randomly varied across participants.

Our design relied on input from prosecutors and defense attorneys (Studies 1 and 2) as well as judges (Study 2) to report likely judicial decisions and to assess likely juror bias. We used multiple informants to triangulate results and thereby to bolster the validity of the responses about what judges typically do. Lawyers are good informants about judicial behavior because they have seen the behavior of a variety of judges in a number of different trial contexts. Study 1 examined an experienced group of prosecutors and public defenders from a major urban area, a region we call “Lockland” (those who assisted us in recruiting this and all of our samples asked that we not identify the county or state where they work). We surveyed both prosecutors and defense attorneys to account for any bias associated with the particular side an attorney represents. This was particularly important because several of the vignette case facts—all of which are based on actual court cases— came from the appellate literature and necessarily reflected a potential pro-prosecution bias (i.e., those appealing convictions argued that the judge had not removed a juror who may have been biased against the defense).

In Study 2, we submitted the same questions and vignettes to a sample of sitting judges from a different state. As these judges differed from Lockland attorneys in terms of both role and geography, we also sampled a second, smaller group of prosecutors and public defenders from a large urban area in the same state as the judge sample, an area we call “Hartland.” The combination of these samples allowed us to examine whether perceptions of judicial behavior are consistent across types of informants and across different locations that may have somewhat different legal cultures.²

A final set of informants played a different role in this research. We gave samples of laypeople (i.e., prospective jurors) the same vignettes and asked them to estimate potential juror impartiality and to indicate in each case how a judge should rule (Study 3). This lay response told us whether the decision-making behavior in our samples of legal professionals comports with the sensibility of nonprofessionals, as well as how laypeople assess others' abilities to remain impartial.

Study 1

Method

Participants

We distributed questionnaires to 282 public defenders and 297 prosecutors in a large urban county. The questionnaire included a memorandum from a senior supervisor supporting the project, as well as a return envelope. Participants returned questionnaires in the envelopes to their immediate supervisors, and these local supervisors returned them to a central office where we collected them. Anonymity of participation was emphasized in the cover memo, as well as in our introductory page, which described the study as an attempt “to gain a better understanding of the conditions under which challenges for cause are granted.” We received 80 responses from public defenders (response rate=28%) but had to omit two questionnaires because of high levels of missing data, for a sample of 78 defense attorneys (median number of cases tried=20). One hundred and seven prosecutors returned the questionnaire (36% response rate).³ One person was omitted because the participant reported no trial experience, and another had substantial missing data, resulting in a total sample of 105 prosecutors, who reported having previously tried a median of 40 cases.

Questionnaires

All questionnaires contained a single introductory page with instructions, eight one-page vignettes describing a juror's background and verbal exchanges with a judge during voir dire, and a final page on which attorneys reported their trial experience. Participants were presented first with two control vignettes and then with six experimental vignettes; in each experimental vignette the juror gave either a firm or equivocal assessment of his or her ability to be fair.

Control vignettes. The control vignettes provided a check on participants' abilities to identify jurors who were either highly unlikely or highly likely to be excused for cause. The first described an armed robbery case. For the public defender sample, a juror reported a remote connection with the county's elected district attorney because they attended the same church. She stated, however, that the church was very large and she had never spoken with this person; she did not know the assistant district attorney who was actually prosecuting the case. At no point did this juror express any hesitation about her ability to be fair and impartial. We anticipated that few, if any, of the attorneys would believe that a judge would excuse a juror for cause in this situation. For prosecutors, the vignette was identical except that the juror attended the same church as the chief public defender for the county (rather than the elected district attorney). In this way, both attorney samples considered a juror whose association was with the boss of the opposing attorney in the case.

The second control vignette described a drug trafficking case. The juror in question was a police officer, and three police officers were scheduled to testify during the trial, one of whom was the juror's direct supervisor. The juror reported being familiar with the case through work and said he had strong opinions about it. He told the judge that he did not think he could be truly fair. We expected the average judge to excuse this juror for cause. The control vignettes were always the first two vignettes in the packet (we refer to them herein as Vignettes A and B).

Experimental vignettes. Each of the remaining vignettes had two versions that varied according to the confidence with which jurors expressed their abilities to be fair and impartial. The vignettes were:

• • Vignette C: In a capital murder trial, a juror has children the same age as the victims, and these children know the victims through school.

• • Vignette D: In a murder trial following a prison riot, a prospective juror is a former deputy sheriff who currently works for the State Board of Pardons and Paroles. He knew the slain guard casually and worked with him the Friday before the riot.

• • Vignette E: In a case in which a defendant is accused of sexually molesting his step-daughter, a prospective juror states that her daughter had been sexually molested by a family member.

• • Vignette F: In a drunk-driving case, a prospective juror reports that his niece was killed by a drunk driver and he has since joined Mothers Against Drunk Driving (MADD).

• • Vignette G: In an involuntary manslaughter case involving a highway accident resulting in multiple fatalities, a prospective juror reports knowing the wife of one of the victims through his job at a local bank, where he handles investments for the woman. The victim's wife will testify in this criminal case, and the juror reports being aware that the wife has filed a separate civil suit for wrongful death against the defendant.

• • Vignette H: In a rape case, a prospective juror reveals that she was raped 40 years ago. When asked about the experience, she becomes emotional, saying that she wished she would have killed her attacker. She does, however, believe she could be fair and perhaps even more so because she's “been through it” and believes her experience would help her in being able to tell if someone were speaking truthfully or not.

To control for possible order effects, we created eight different sets of experimental vignettes, assigned randomly across participants; order did not alter the results we report.

After presenting the above basic information, all vignettes contained descriptions of the follow-up exchange between the juror and the judge, as the judge attempted to clarify whether the juror could hear the case with an open mind, follow instructions, and distinguish any personal circumstances from the evidence presented in the case. Except for the final question from the judge and the juror's final response, these reflected the exchanges from the actual cases. For example, one version of Vignette C read as follows:

A defendant is charged with capital murder, rape and attempted murder in the stabbing of two children, a brother and a sister. The young girl was raped and killed; the brother received multiple stab wounds but survived the attack. A prospective juror has two children the same ages as the victims, and his children attend the same school as the one attended by the victims. The juror's son has gym class with the surviving victim and sees the scars from his stab wounds regularly. The juror says that he has discussed the case with his children, who have been fearful since the incident.

During voir dire:

• • The juror says has “some ideas” regarding the defendant's guilt, but he has not made up his mind.

• • When asked if he could listen to the case with an open mind, the juror says “Yes.”

• • The judge asks if the juror believes that he will concern himself with what his children might think about the verdict. The juror replies, “No, I don't think so.”

• • Finally the judge asks if the juror is able to decide the case on the facts and evidence and not on anything else; the juror responds, “Yes.”

The juror's last response was the sole difference between the two experimental versions of the vignette. For the above vignette, the equivocal version had the juror saying, “I would try” instead of “Yes.”

We created two versions of the questionnaire and randomly assigned respondents to receive either Version One or Version Two. Each version contained three firm responses and three equivocal ones. The equivocal statements reflect the type of responses we have observed in jury selections and the type of responses appellate courts have reviewed (see, e.g., People v. Torpey 1984 [“I think I can”]; People v. Johnson 2000 [“I would try”]). In Version One, Vignettes C, D, and G ended in firm promises, and Vignettes E, F, and H ended in equivocal responses. In Version Two, Vignettes C, D, and G ended in equivocal responses, while Vignettes E, F, and H closed with firm promises to be fair. Table 1 depicts this design and shows the precise language used in the jurors' responses.

Table 1. Design of Study

^a Indicates confident responses.

^b In Studies 2 and 3, this response was, “I would be fair.”

After they read each vignette, participants responded to the following two questions: (1) “How likely is it that this juror will be fair and impartial in this case?” (1=“very unlikely” and 7=“very likely”); and (2) “If one of the attorneys asked the judge in the case to excuse this juror for cause, what do you think the average judge would do?” with two response options: “Excuse the juror” or “Deny the attorney's request.”

Analytic Strategy

To generate descriptive data, we collapsed across experimental version and examined for each vignette how often, on average, respondents said a judge would excuse the juror in response to a challenge for cause. An overall “excuse score” reflected the number of jurors each participant said a judge would excuse (range: 0 to 6). We also examined the mean estimate from the 1 to 7 rating of how biased the juror appeared to the attorney-participants.

To evaluate the effect of our experimental manipulation, we created two index scores, each based on the design outlined in Table 1. Score 1 represented the number of jurors excused in the first set of vignettes: C, D, and G; Score 2 represented the number of jurors excused in Vignettes E, F, and H.⁴ Thus each participant had both a Score 1 and a Score 2; however, depending upon whether a participant received Version One or Version Two of the questionnaire, Score 1 reflected the “firm” vignettes and Score 2 the “equivocal” ones, or vice-versa.

If juror certainty influenced judgments, then when one set of vignettes all had firm responses (i.e., Vignettes C, D, and G in Version One), the number of excuses expected to be granted for this set (i.e., Score 1) would be lower than when the same vignettes were equivocal (i.e., Vignettes C, D, and G in Version Two). Likewise, Score 2 for participants reading Version One involved vignettes with equivocal language (i.e., Vignettes E, F, and H). The predicted result was more excuses for cause granted and a higher score than Score 2 for Version Two, which had firm responses in Vignettes E, F, and H.

A repeated-measures ANOVA examined the extent to which the pattern of scores was influenced by equivocation, while also accounting for the fact that each person produced both a Score 1 and a Score 2, rendering the two indexes non-independent. If our predicted relationship was confirmed, there would be a significant interaction between the difference in the two scores and the experimental version (Version One versus Version Two); that is, the magnitude of the difference between Score 1 and Score 2 should depend upon whether someone responded to Version One or Version Two. Several patterns of means could generate a significant interaction in this analysis. If our hypothesis was correct, Score 1 minus Score 2 for Version One (firm minus equivocal) would be negative, while the difference between the same scores for Version Two would be positive. To test this, separate analyses of Score 1 and for Score 2 assessed whether the level of equivocation expressed produced the predicted changes in the number of excuses for cause.

Results

Descriptive Results

Both attorney groups responded to the two control vignettes as predicted. For the control vignette involving low apparent bias, only 4 percent of public defenders and 4 percent of prosecutors said that the judge would excuse the juror in Vignette A (who went to the same church as the boss of the opposing lawyer and said she could be fair). By contrast, on the high-apparent-bias vignette, 97 percent of public defenders and 99 percent of the prosecutors said the judge would dismiss the police officer in Vignette B (whose supervisor would be testifying and who said he could not be fair). A t-test for bias ratings indicated that both attorney groups rated the juror in Vignette A as more likely to be fair than the juror in Vignette B (p's<0.0001); however, as might be expected, public defenders also rated both of the prospective jurors as less likely to be fair than did prosecutors: Vignette A, M=3.91 (SD=1.51) for public defenders, and M=4.91 (SD=1.43) for prosecutors (p<0.0001); for Vignette B, M=1.13 (SD=0.63) for public defenders, and M=1.79 (SD=1.06) for prosecutors (p<0.001).

Table 2 presents descriptive results for the remaining six experimental vignettes. Consistent with our expectation that jurors who raised red flags but maintained some ability to be fair might not be viewed consistently by all judges, estimates of what the average judge would do exhibited far less consensus in the experimental vignettes than in the control vignettes. Collapsing across experimental condition, the percentage of public defenders who predicted an excuse for cause ranged from a low of 38 percent (Vignette H, a rape case) to a high of 68 percent (Vignette D, a prison riot). For prosecutors the range was 43 percent (Vignette H) to 75 percent (Vignette F, a DUI case). Both public defenders and prosecutors expected that, on average, judges would grant fewer than four challenges, with the mean only slightly higher for prosecutors (M=3.65) than for public defenders (M=3.12, t=1.84, p<0.07). This convergence in the overall number of expected excuses for cause occurred despite the fact that, as was true for the control vignettes, public defenders had significantly lower confidence in all of these jurors' abilities to be fair and impartial when compared to prosecutors; all juror bias ratings differed at p<0.0001 across the two attorney groups.

Table 2. Lockland Attorney Sample: Percent Indicating Juror Would Be Excused for Cause and Perceived Bias, Overall and by Experimental Condition

^* Means reported reflect the average number of cause challenges participants said would be granted across all six vignettes (range: 0 to 6).

Differences Due to Juror Confidence

Our research design produced a total of 12 ratings, i.e., those for six equivocal vignettes and six confident vignettes. However, as described, in this type of “within-subjects” design, all participants read three vignettes of each type. We report the statistical analysis appropriate for this design below. However, ahead of this, we first examined whether the general pattern of responses fit with our expectations. We therefore separately counted the number of cause excuses given for the six vignettes with equivocal language and for the six vignettes with firm language. The pattern supported our hypothesis that judges would be expected to grant more challenges for cause when jurors used equivocal language than when jurors used confident language. On average, public defenders expected 3.73 excuses for cause when language was equivocal and 2.96 when the language was firm. The parallel averages for prosecutors were 4.04 and 3.51. (To generate these means, we combined the Score 1/firm mean from one experimental group with the Score 2/firm mean from the other, weighted by the sample size for each. We did the same for Score 1/equivocal and Score 2/equivocal. A direct test of significance was not possible with means constructed across experimental groups in this manner.)

The formal test of our hypothesis examined the differences in means for each experimental condition; these means appear in Table 3. A test for an interaction between the differences in Score 1 and Score 2 and experimental version resulted in significant results for both public defenders, F(1, 76)=7.86, p<0.01 and prosecutors, F(1, 103)=6.83, p<0.05.

Table 3. Lockland Attorneys: Mean Numbers of Challenges Granted (and Standard Deviation) by Vignette Set and Equivocation Condition

Range: 0 to 3. Higher scores=greater number of cause challenges encouraged.

^a These means differ from one another at p<0.01.

^b These means differ from one another at p<0.10.

Follow-up tests revealed support for our hypothesis on one set of vignettes but not the other. The experimental manipulation of language did not significantly alter Score 1 (produced by vignettes C, D, and G) for public defenders, F(1, 76)=0.03, p<0.91 (M=1.88 versus M=1.84; effect size measure, Cohen's d=0.04). Prosecutors showed evidence for a small effect (M=1.88 versus M=2.15; Cohen's d=0.25), but this was also not statistically significant, F(1, 103)=1.67, p<0.20. For the vignettes producing Score 2 (E, F, and H), the effects for language were in the predicted direction and highly significant in the public defender sample. As Table 3 shows, when these vignettes ended with a firm response, respondents were less likely to say that the judge would grant a challenge for cause than when these same vignettes ended with an equivocal response, F(1, 76)=11.98, p<0.001 (public defenders; d=0.78). Precisely the same pattern occurred among prosecutors, but the difference in means did not reach conventional levels of significance, F(1, 103)=2.89, p<0.10, (prosecutors; d=0.33).

Although attorneys expected judicial decisions to be affected by language, their own perceptions of juror bias were mostly unrelated to the juror's confidence. In only one instance (Vignette C, when the juror's children knew the victims) was the equivocating juror perceived as less able to be fair (M=3.20, SD=1.47) than the confident juror (M=3.82, SD=1.60), and this effect was limited to prosecutors (t=2.02, p<0.05; d=0.40).

Discussion

In situations in which a prospective juror's background and experiences suggested bias, but the individual claimed some capacity to be fair and impartial, attorneys on both sides of the aisle responded to subtle changes in language—a distinction between, for example, “I would try” versus “Yes”—by altering their predictions about how a judge would rule on a cause challenge. There was a significant effect for language in the second set of the vignettes (i.e., those producing Score 2) among public defenders, with prosecutors showing the same, albeit a slightly weaker, result. For both prosecutors and public defenders, the jurors' language had little effect on whether attorneys personally judged the jurors capable of fairness. This consistency between the two attorney groups is remarkable given the differences in each group's ratings of the jurors' capacities for fairness and impartiality. Public defenders demonstrated “floor” effects on most of these ratings, whereas prosecutors were reliably more optimistic about whether the jurors could be fair and impartial.

We considered several explanations for why language differences predicted the responses for one group of vignettes (E, F, and H, which produced Score 2) but not the other (C, D, and G, which produced Score 1). For example, we wondered if attorneys perceived the level of bias differently across the two sets of vignettes in a way that undermined our hypothesis. However, mean ratings of bias showed no association with language equivocation for either attorney group.⁵ The two sets differed from one another in terms of case facts—for example, the set producing Score 1 all concerned instances of formal or informal associations with the victims, whereas the Score 2 set concerned a personal experience with the crime alleged in the case. Two of the three vignettes in the latter set involved sex crime cases (child molestation and rape). Perhaps attorneys believed that judges have more difficulty adjudicating questions of bias in instances of personal experience with a crime, or in sex crimes in particular, and therefore would rely more heavily on the quality of the jurors' self-reports in these cases. Certainly the result points to the need for replication.

Study 1 also omits an important source. We relied on the reports of attorneys, albeit both prosecutors and public defenders, to describe judicial behavior. In Study 2, we tested the responses of a group of sitting trial judges to these same vignettes. We also recruited a smaller group of attorneys (public defenders and prosecutors) from an urban area (Hartland) in this same state.

Study 2

Method

Participants

Judges. We distributed 283 questionnaires to sitting trial court judges in a single state. All these individuals had presided over at least one criminal jury trial in the past year. With the help of the administrative office of the courts for the state, the presiding judge sent a recruitment e-mail message, which vouched for the study and provided two ways to respond to the questionnaire. The participant could go to a secure Web site and provide answers online (responses were then e-mailed to the first author). Alternatively, the judges could print out an attached .pdf copy of the questionnaire, fill it out, and send it via regular mail to the first author. The judges were contacted twice to request participation. A total of 77 judges responded (27 percent), half via Web responses and half through sending in the printed copy. Response rates were comparable across the two experimental conditions. The sample was highly experienced, reporting that they had overseen a median of 150 criminal jury trials.

Attorneys. Unlike the judges in this state, and the centralized units in Lockland, both prosecutors and public defenders in Hartland were scattered across multiple offices, creating a substantial challenge in generating a sampling frame. To obtain a group of participants, we were again aided by a representative of an administrative office of the courts who, at a regional meeting, invited supervisory heads of public defender offices and district attorney offices to allow us to contact them about the study. Four offices agreed to be contacted: one district attorney office from a suburban part of the area, and three public defender offices representing urban practices. We gave these attorneys the same options we provided to judges (i.e., participation through either a secure Web site or a .pdf attachment), although some supervisors preferred instead to distribute hard copies of the survey directly to their attorneys' mailboxes; we provided return envelopes.

We sent out 515 questionnaires, 75 as hard copies to be distributed and the remainder as e-mail solicitations. We received 54 responses, 17 from the prosecutors (34 percent response rate) and 37 from public defenders (response rate=8 percent). After eliminating one prosecutor who reported no trial experience, we had a total sample of 53 (median cases tried=35 and 25 for prosecutors and public defenders, respectively). Responses were largely balanced across experimental conditions for the prosecutors, but slightly fewer public defenders returned Version One (n=16) than Version Two (n=21).

Questionnaires/Study Design

The stimulus material, study design, and analytic strategy were all identical to those used for the Lockland attorney sample, with two exceptions. First, as the order of presentation did not affect the Lockland results, we used a single order for all participants. This reflected the order we have used for this article, i.e., Vignettes A through H. Second, we noticed that the firm response the juror gave in Vignettes D and F (“I would listen to both sides”) did not quite parallel the equivocal response (“I'm pretty sure I could be fair”). To ensure that this was not introducing noise into results, we made the firm response more direct and parallel: “I would be fair.” As with the Lockland sample, in the version the prosecutors received, Vignette A stated that the juror attended the same church as the public defender's boss, whereas public defenders read about a juror who attended the same church as the prosecutor's boss (as did judges).

Results

Descriptive Results

All groups reacted as predicted to the two control vignettes. All judges said that the average judge would deny a challenge for cause for the juror in Vignette A, and 100 percent of judges said the average judge would excuse the police officer in Vignette B. Judges were also highly confident that the juror in Vignette A was free of bias (mean bias rating=6.39, SD=0.80); at the other extreme, the juror in Vignette B produced a mean of 1.12 (SD=0.33). The comparable numbers for the attorney groups were as follows: 100 percent of the prosecutors and 95 percent of the public defenders said a judge would deny a cause challenge in the church vignette, and 100 percent of both groups agreed on a cause challenge being granted for Vignette B. Prosecutors were more confident about the abilities of the juror in Vignette A than were public defenders (mean bias rating=5.13, SD=1.15, and 3.70, SD=1.27, respectively; p<0.001). The attorney groups rated the juror in Vignette B comparably (mean bias=1.00, SD=0, for prosecutors, and 1.27, SD=1.17, for public defenders; p<0.44).

Table 4 presents descriptive results for the experimental vignettes. The table highlights a clear difference in perceptions between the judges and the attorneys, especially between the judges and the public defenders. In most instances, this sample of judges showed little variability in decision making across the six vignettes. Given six possible challenges to be granted, judge-participants said the average judge would sustain an average of 5.42 (SD=0.92). Those saying the judge would excuse ranged from a low of 76 percent (Vignette G, the juror who invests for the survivor of the alleged victim) to 99 percent (Vignette D, the prison riot case).

Table 4. State Judges and Hartland Attorneys: Percent Indicating Juror Would Be Excused for Cause and Bias Ratings

The Hartland public defenders' ratings were closer to those of the Lockland attorney groups reported in Study 1, with 3.89 (SD=1.66) excuses expected overall; across vignettes, between 51 and 78 percent of public defenders expected a cause challenge to be granted in the scenarios we described. Hartland prosecutors' estimates were higher, with a mean number of expected excuses of 4.69 (SD=1.01; range in predictions: 44 to 94 percent). Consistent with the control vignette, and consistent with what we observed for the Lockland attorneys, Hartland prosecutors not only expected more cause excuses from the judges but also were personally more confident about the jurors' abilities to be fair and impartial than were the public defenders (all p-values for differences across attorney types<0.05), although, as with the judges, all ratings of perceived impartiality clustered well below the midpoint of the scale.

Differences Due to Juror Confidence

Although the three groups perceived the jurors' level of bias differently, and although judges were far more confident that the “average judge” would dismiss this set of jurors, the responses of all three groups indicated that language contributed to judges' decision making. For both attorneys and judges, highly significant interaction terms supported our prediction: F(1, 75)=12.02, p<0.001 (judges) and F(1, 51)=27.16, p<0.0001 for attorneys (attorney groups were combined here due to the small number of prosecutors; the interaction was also significant when the groups were analyzed separately). Means by condition appear in Table 5. For judges, when the set of vignettes associated with Score 1 (C, D, and G) ended with an equivocal response, excuses for cause were more likely than when these same vignettes ended with a firm response, F (1, 75)=6.38, p<0.05 (d=0.58). Differences in Score 2 across the firm and equivocating versions were in the predicted direction but did not reach conventional levels of significance, F(1, 75)=2.55, p<0.12 (d=0.36). The Hartland attorney groups exhibited significant effects in the predicted direction for both vignette sets. For Score 1, F(1, 51)=12.68, p<0.001, d=0.98; for Score 2 (Vignettes E, F, and H), F(1, 51)=6.22, p<0.05, d=0.69.

Table 5. Mean numbers of Challenges Expected to be Granted by Vignette Set and Equivocation Condition

^a These two means differ from one another at p<0.05.

^b These two means differ from one another at p<0.001.

Although the judges reported on the likely excuse pattern for the average judge, as with other respondents, they reported on their own perceptions of the impartiality of these jurors. These ratings provided a window into the perceptions judges had of these jurors' abilities, and whether language differences affected those perceptions. On Vignettes C, E, and G, those jurors described as providing a firm response (M=2.63 [C], 2.38 [E], and 3.50 [G]) were perceived as significantly more likely to be able to be fair than those who equivocated (M=2.00 [C], 1.89 [E], 2.72 [G]; all p-values<0.05; d=0.54, 0.47, and 0.52, respectively). By contrast, confidence did not typically predict attorneys' perceptions of whether the jurors could be impartial, confirming the results we observed for Lockland. The sole exception was for Vignette C, in which public defenders had higher ratings of the juror's impartiality when that juror was confident (M=1.81, SD=0.75) than equivocal (M=1.33, SD=0.48), t=2.36 (p<0.05; d=0.78).

Discussion

Results from five groups of legal professionals in two geographical regions converge on the same result: jurors who are confident about their abilities are less likely to be removed for cause than are jurors who hesitate when declaring fairness. In the set of vignettes represented by Score 1—all of which involved some association with the victims—judges expected excuses for cause more often when the juror gave an equivocating response than when the same person responded more confidently. This effect emerged despite the fact that judges believed nearly all the jurors would be excused. Further, confidence predicted some of the judges' own impressions of the jurors' abilities to be fair. In response to three vignettes, judges' ratings of juror impartiality were higher in the confident condition than in the equivocation condition. By contrast, with a single exception, juror confidence was unrelated to Hartland attorneys' ratings of how fair and impartial the juror could be—the same result we observed among Lockland attorneys. However, small changes in language affected Hartland attorneys' expectations about how judges make decisions. Confidence predicted judges' likely rulings for both sets of vignettes, thus replicating both the Score 1 result among the judges and the Score 2 effect observed in the Lockland sample.

The results of Study 2 demonstrate why it is critical to have multiple groups report on what happens during jury selection. Had we given this survey only to judges, our hypothesis about the relationship between confidence and cause challenge decisions would have been supported, at least for one set of vignettes (Score 1). However, we would have come away with a starkly different view of how judges behave during jury selection more generally. According to the judge sample, the vignettes we created were not difficult to decide; the judges expected an average of 5.42 out of six possible excuses. Contrast this result with those from the Lockland samples, who estimated the average number of challenges granted across experimental vignettes as 3.12 (public defenders) and 3.65 (prosecutors). We could attribute this to a difference in region; however, the Hartland public defenders had results that were far more consistent with Lockland attorneys than they were with their home state judges—expecting 3.89 out of a possible six excuses for cause. (The Hartland prosecutors were more certain that nearly all the jurors would be excused; however, their sample size of 16 was very small and restricted to a single office within Hartland.) Generally speaking, unlike judges, attorney groups gave little indication that jurors with problematic backgrounds would automatically be dismissed as biased.

We suspect that some of the judges' reports were likely affected by this study's particular methodology, which presented respondents with hypothetical scenarios. In an actual voir dire, judges must make decisions about a real person who claims to be fair, and their decisions about these jurors have consequences for other jurors who must replace those excused. Thus our participant judges may well have found it easier than they would in an actual voir dire to say that a hypothetical juror should be excused in the interest of fairness. Given that observational studies suggest that judges often deny cause challenges in cases like the ones we presented (Balch et al. 1976; Rose 2003; Shuy 1995; see also Zalman & Tsoudis 2005), and given the attorney results in this study, we suspect that our judges may have been mistaken in claiming that these jurors would generally be dismissed for cause. More likely, the decision to excuse will depend on a perception that the juror can set aside potential bias. Perceptions of juror bias differed across professional roles, particularly between public defenders and prosecutors—an issue we return to in more detail following Study 3. However, the consensus of the results obtained from all groups was that judges' estimations of likely bias are influenced by the subtle ways that jurors describe their current state of mind.

Viewed one way, our findings suggest that attorneys may view judicial decision making in a cynical manner: the attorneys largely believe the jurors are unlikely to be fair, and they are not persuaded by the jurors' self-reports of their abilities, but they also do not necessarily believe that judges will dismiss these jurors, especially if the jurors are confident-sounding in their self-report. However, it is also possible that judges and attorneys simply disagree about whether juror confidence conveys something important about the juror's state of mind and ability to be fair. If judges see confidence as diagnostic of ability, then altering cause challenge decisions in response to more- or less-confident language is not a cynical way of retaining “obviously” biased jurors who have uttered “magic words.” Instead, judges may simply see value in using confident self-reports as a cue for an ability and willingness to be fair, whereas attorneys do not. We have some evidence for this possibility: judges were more likely than attorneys to alter their ratings of how fair the jurors seemed as a result of changes in juror confidence, an effect that occurred in half the vignettes.

The different conclusions judges and attorneys draw about confident responses necessarily raises a question about how common it is to assume that a relationship exists between confident speech and an ability to remain neutral. As with perceptions of witnesses in a courtroom (Conley et al. 1978; O'Barr 1982), do people who confidently describe their state of mind (“I can be fair”) appear to most people to be more trustworthy than those who are less direct (“I'll try to be fair”)? Or do red flags appear to be more diagnostic of ability than what people say about themselves? In Study 3 we look at this issue through the eyes of prospective jurors, a constituency in the jury selection process that has not been deeply socialized into legal norms and practices. If confidence reflects—accurately or not—that a given individual is more capable of impartiality, then confidence should predict how community members view the individuals in the vignettes and whether they think the judge should dismiss such people for cause.

Study 3

Method

Participants

The participants in Study 3 came from three large urban courts in two states; for ease, we refer to these simply as Counties 1, 2, and 3. Counties 2 and 3 are located in the state where we recruited our judge and Hartland attorney samples; County 1 is in a different state from all the other samples. We summarize the sociodemographic profiles of both the counties and the respondents in Table 6.

Table 6. Characteristics of Jury Participants and Their Counties

Note: County profiles are based on census data. For purposes of preserving the identity of the regions, the county-level information reflects approximations of actual values (rounded to the nearest value of 5 or 10).

In County 1, the court administrator, before telling the jurors in his courthouse that they would be excused shortly, invited the waiting jurors to participate in the survey. Data were collected across several weeks, on days when the administrator could work recruitment into his schedule. No record was maintained on the rate of participation. In the other two courthouses, the authors personally recruited waiting jurors, with the assistance of the jury clerks. We happened to undertake collection on a day when several trials had been scheduled, but due to unexpected circumstances, no trials commenced. The result was that many jurors were waiting around with little or nothing to do. Nearly every juror offered a survey at these courthouses agreed to participate. We omitted from analysis those questionnaires with high levels of missing data (n=7 from County 1, and n=11 for other areas) or instances when respondents told us they did not speak English well and did not understand the questions (n=3).

Questionnaires

Before answering the questions, jurors received a brief description of an excuse for cause and what happens during jury selection (e.g., “The judge asks potential jurors a series of questions. Based on the answers, the judge may excuse a juror from serving on that particular case if the answers indicate that the juror would not be able to be fair in that case.”). For County 1, this description appeared on the introductory page to the questionnaire. For Counties 2 and 3, we read this description orally to jurors during our brief presentation to them about what the study would involve. The remainder of the survey instrument was identical to those given to the legal professionals, except that we asked jurors whether they thought the judge should excuse the juror for cause, and we also asked jurors to provide background information about their sex, age, and prior jury service. (These latter variables did not correlate with fairness ratings for the vignettes or with an overall tendency to say the judge should excuse a juror and are therefore not discussed further.)

Results

Descriptive Results

As was true for all the other samples, the respondents were nearly unanimous in believing that the person described in Vignette B (police officer works for one of the witnesses) should be excused: 94 percent for County 1 and 97 percent in County 2. County 3 offered a significantly lower figure (89 percent) than the other two groups (χ²=5.71, p<0.05), but this was the highest rate of any vignettes from this county. Unexpectedly, these community samples differed markedly from the legal professionals in their responses to Vignette A, designed to represent an “obvious” instance of no bias. Vignette A also produced strong differences across counties. Before discussing those results, we first describe results for the experimental vignettes, which reflected our variable of interest.

Mean bias ratings did not differ by county, with the exception of Vignette C (juror's children know victims), F(2, 333)=3.07, p<0.05. Post hoc contrast tests showed that those in County 3 were somewhat more likely than those in the other two areas to say the juror in the vignette could be fair: M for County 3=3.34 (SD=2.17) versus M for County 1=2.71 (SD=1.81, p<0.02) and M for County 2=2.88 (SD=1.68, p<0.08). Although perceptions of bias in the experimental vignettes were consistent across counties, evaluations of what judges should do were not. As Table 7 indicates, jurors in County 3 were less inclined than those elsewhere to say the judge should excuse the jurors in question. They recommended a total of 3.91 (SD=1.66) excuses for cause out of a possible six, compared to an average of 4.66 (SD=1.46) for the other two areas (4.61 for County 1 and 4.70 for County 2), t=4.01, p<0.0001. Nevertheless, as Table 7 makes clear, a majority of all respondents favored dismissal of the prospective juror in each of the vignettes.

Table 7. Laypeople's Ratings of Experimental Vignettes, by County

Differences Due to Juror Confidence

The confidence with which the jurors in the vignettes expressed themselves had no effect. None of the respondents' mean ratings of impartiality differed by experimental version. Unlike any of the attorney or judge samples, the repeated-measures analysis revealed no association between confidence and the difference in Score 1 and Score 2 and experimental version, interaction F(1, 341)=0.30 (p<0.59).

Juror Attends Same Church as County Prosecutor

Vignette A, which concerned a remote connection to the prosecuting attorney's boss through church, produced results at odds with the patterns in Studies 1 and 2. Recall that all samples of legal professionals were nearly unanimous in saying that judges would not excuse a juror in these circumstances. These results are reprised in Table 8, which contrasts them with those for our laypeople, many of whom believed that the juror in this situation should be excused for cause. Further, unlike the pattern observed in all other vignettes, in which respondents in County 3 recommended fewer excuses for cause, a majority (56 percent) of this group thought the judge should excuse the juror in question, which was significantly higher than estimates from the other two areas (p for both χ² tests<0.01). Table 8 also shows that the estimate of bias for County 3 was in line with those of the public defenders, and these groups' ratings differed from Counties 2 and 3 and the prosecuting attorneys. The ratings of ability to be fair were significantly lower for all groups than the high rating given by the judge sample.

Table 8. Responses to Vignette A: The Juror Attends the Same Church as an Attorney's Boss

Notes: Jurors responded on whether the judge should excuse; attorneys on whether the judge was likely to excuse. All groups except the prosecutors read about the elected district attorney as the church member. Prosecutors read about the chief of the public defender's office. Means with different superscripts differ at p<0.05.

Discussion

Like other informants, the prospective jurors typically suspected that the jurors we described would have a hard time being fair, and in most instances, strong majorities said the judge should excuse the person. These perceptions were unrelated to how confidently the person spoke. Thus to the extent that judges attend to juror confidence to resolve questions about whom to excuse for cause—an effect all samples of legal professionals evidenced—this reaction does not occur because people's actual open-mindedness is self-evidently reflected through the certainty of their words.

Why might judges see value in confidence as a cue for fairness, while others do not? We view the inconsistency between judges and others as likely reflecting the distinct responsibilities and concerns that judges have during jury selection. First, judges naturally consider how their rulings will look to an appellate court. To the extent that appellate judges, like their trial court counterparts, find a confident assertion to be more convincing than a less-confident one, then judges who deny cause challenges on confident jurors and allow them for jurors who equivocate will maintain a more successful appellate record.⁶

In addition, judges have the responsibility for the decision itself and for its effect on other prospective jurors. For reasons of accountability, compared to attorneys, judges may look at all pieces of evidence about a juror with an exacting eye (see, e.g., Tetlock & Kim 1987). There is indeed good reason for judges to feel a good deal of uncertainty about decisions. No data quantify the likelihood that someone with a given background characteristic (e.g., a crime victim) would be unable or unwilling to maintain objectivity in a case.⁷ Even if such data existed, they might not apply to the particular juror or in a particular case. Faced with a need to rule in the face of uncertainty, judges may view an imperfect cue as far preferable to no cue.

Finally, judges have a distinctive social relationship with jurors during voir dire. During jury selection, judges ask jurors to give their opinion about how they will perform as jurors. When a juror confidently asserts an ability and willingness to do his or her duty, a judge's excuse for cause acts as a clear rejection, saying, in essence, “I do not believe you,” or at least, “I know better than you.” In theory, the awkwardness of overruling such a juror should be negligible. A judge who concludes that a juror is most likely not capable of fairness could emphasize the background facts about a juror and minimize a confident self-report of fairness. Indeed, when reading hypothetical scenarios, our judge sample predicted numerous cause excuses no matter what the juror said. However, in practical terms and in real voir dire interviews, judges who sustain challenges for cause make a public declaration about a person's fitness to serve, and they do so as the official arbiters of bias. Equivocation from a juror may relieve some of the potential social awkwardness, since a judge's ruling may appear simply to affirm the juror's own seeming lack of certainty.

Implications for Law and Social Science Scholarship

Although a number of factors may explain why judges perceive prospective jurors' abilities differently than do others, the finding itself contributes new data to literatures in several fields of law and social science, especially social psychology, sociology, and sociolinguistics. Prior work has shown that small changes to language can dramatically alter impressions of people (Conley et al. 1978) and of events (e.g., Loftus 1979). We here link language differences to changes in judges' behavior and, to some extent, to their impression of potential jurors. Our data indicate that confidence acts as a simplifying heuristic. However, we also show that use of this heuristic depends upon social role—that is, upon being a judge—a finding that is consistent with previous suggestions that the value of heuristics in decision making may depend upon social characteristics of the perceiver (e.g., Douglas 1986; Heimer 1988; Tetlock 2002).

Our findings point to a number of additional areas of research to pursue. Trial judges have the opportunity to observe not only what a juror says but how he or she says it, and appellate courts will often cite so-called demeanor evidence in reasoning why they should give great deference to a judge's cause challenge rulings (for a recent example, see Uttecht v. Brown 2007). Although we gave our vignette survey to a wide variety of respondents, and although our results were consistent with the findings obtained in observational studies of actual cases, additional studies should consider presenting richer detail about prospective jurors. A study might focus on smaller groups of participants and present (and systematically vary) a stimulus that more realistically reflects what judges see and hear. In one prior study, Kerr and his colleagues (1991) presented juror information using videotape. In future work, researchers can use this approach to more fully examine the range of behavior and speech patterns that lead some individuals to appear more capable of fairness than others.

Our study also leaves open the question of whether and how language effects depend upon the social category of the person speaking—that is, whether and how confidence interacts with some other indicators of status to shape impressions of credibility and competence. Judgments about whether someone can be fair, given how they speak, may vary by the class, race, or gender of the person speaking. Legal anthropologists and other scholars have noted that some types of people, including those with less education and lower occupational status, are more prone to “powerless language” than are their higher-status counterparts (Conley & O'Barr 2005; see also, e.g., Aries 1996; Lakoff 1975). Additional work might investigate whether confident language leads to different types of inferences when used by, for example, an older white male who appears to be middle-class as compared to a young African American woman who appears to be of more limited means. Research in status characteristics theory suggests that people do not interpret status cues (e.g., behaving assertively) the same way for different status categories (e.g., being female versus male; see Ridgeway 2001). As we noted in our results sections, the effects for language sometimes emerged for one set of vignettes and not for the other, and it may be that some status-related factor—e.g., Score 1 was calculated on a set of all-male characters—contributed to variability in impressions across groups. The present study is the first to show, in a nonobservational, unambiguous way, that confidence is one factor judges attend to in cause challenge decisions. Further research should assess the extent to which the effects observed here would be magnified or reduced when other characteristics of prospective jurors are varied.

Further work should also consider whether confident-sounding people are, in fact, more capable of open-mindedness than those who express equivocation—that is, whether judges give confidence too much weight or whether attorneys and laypeople give it too little. Such a study would examine how equivocating versus confident people behave during decision making and deliberations. According to Kerr and colleagues (1991), juror self-assessments are independent of final verdict preferences. In that experiment, researchers exposed participants (sampled from jury pools) to different forms of pretrial publicity and then interviewed them about whether they could be fair and impartial as jurors in the case. Researchers classified answers as indicating either that the person definitely could not be fair, possibly could be fair, or definitely could be fair. Participants then watched a simulated trial and indicated their own verdicts. Further, the interviews with these mock jurors were videotaped, and a random sample of tapes was sent to judges and attorneys to evaluate. Although judges were far more likely to say that those in the “definitely could not be fair” group should be excused for cause, the proportion of guilty and not guilty verdicts was largely unrelated to the jurors' self-assessment of their ability to be fair.

The Kerr and colleagues study (1991) involved a simulation and a highly abbreviated voir dire, and this may have contributed to the absence of a relationship between answers and views of the case. However, there are a number of theoretical reasons to expect a weak or nonexistent relationship between confidence and neutrality. Meta-analysis finds only a small correlation between an eyewitness's confidence and the accuracy of his or her identification, a relationship that often disappears altogether when confidence is not assessed immediately at the time of the initial eyewitness identification (Sporer et al. 1995). Because confidence can reflect a myriad of factors, including nervousness over self-presentation, confident speech does not consistently separate truth-tellers from liars (Ekman 1985; Vrij 2000). Indeed, according to some recent work, some forms of equivocal speech (e.g., more frequent use of “I” and other first-person pronouns) predict greater truthfulness (Newman et al. 2003). In addition, a lack of confidence may well reflect a manner of speaking publicly and a person's social relationship with an audience (e.g., Conley & O'Barr 2005), rather than someone's true state of mind (see also Jones 1987). This is significant because although social status may predict several other issues in jury deliberations, such as who talks and for how long (see, e.g., Strodtbeck et al. 1957), we know of no evidence to suggest that status, per se, predicts an ability to be impartial—especially in a social context, such as voir dire, that makes the quality of any self-assessment questionable.

Implications for the Legal System

In comparison to other aspects of jury selection, the excuse for cause has been vastly understudied. At the same time, serious controversies exist regarding the proper distribution of decision making power during jury selection. Scholars who call for the elimination of the peremptory challenge (e.g., Hoffman 1997; Marder 1995) are, implicitly or explicitly, assuming that jury selection outcomes would be improved if judges had sole control over who served and who did not. Our findings suggest they may be unduly optimistic.

First, our data indicate that jurors perceived by judges and nonjudges as likely to be biased will quite routinely survive the cause challenge process. With some variations, the laypeople we sampled, especially those from Counties 1 and 2, thought that judges should remove most of the jurors in the vignettes, particularly those who had a personal or familial experience with sex crimes (i.e., Vignettes E and H, involving molestation and rape cases). However, attorneys' estimates of the likelihood of an excuse for cause showed strong variability, and Vignettes E and H typically had some of the lowest rate of excuses for cause in the attorney samples. We checked ratings across experimental condition for these vignettes. When prospective jurors in Vignettes E and H were confident, only 31 percent of Lockland public defenders said the judge would excuse the juror in Vignette E; this figure was 22 percent for Vignette H. For Lockland prosecutors, estimates were 45 and 33 percent for the confident versions of the two vignettes, respectively, which was similar to Hartland public defenders (43 and 38 percent). (The small sample of Hartland prosecutors was outliers, with 75 and 86 percent saying the confident jurors in Vignettes E and H would be excused.)

At this point we cannot say whether confident-sounding people are or are not capable of being fair. In addition, limited response rates mean that the above values may not be precise estimates for the populations of practicing attorneys in those areas. However, both existing observational studies and the patterns in the present work clearly indicate that jurors whom many different types of people perceive as compromised are not, in many cases, excused for cause. Such situations can raise questions about the perceived integrity of trials if there is no mechanism, apart from dismissals for cause, to remove such jurors.⁸ Some studies have shown that such people are removed only through a peremptory challenge (Rose 2003). Thus by allowing attorneys to dismiss confident individuals who are not dismissed for cause, the peremptory challenge offers an important “safety valve” for parties (for our vignettes, the defense in particular).

Further, the reasons we have suggested for why judges attend to the quality of a person's self-report are highly relevant for this debate. Some who call for the elimination of the peremptory (e.g., Marder 1995) address the issue of a safety valve by suggesting that judges should simply become more generous in allowing excuses for cause. However, an increased willingness to sustain challenges for cause should be expected only if judges now make calculations about dismissals with the peremptory in mind—that is, if they deny cause challenges leveled at confident jurors because they know the attorney can still eliminate those of greatest concern. If instead judges either overuse confidence as a cue to impartiality or if their attention to confidence reflects the social implications of “vetoing” a dutiful citizen, then we should be less optimistic that judges would become more active in removing potentially biased prospective jurors if the peremptory challenge was eliminated. These differences in opinions and differences in role responsibilities would remain. Thus despite the many problems with attorneys' decisions documented previously (e.g., Baldus et al. 2001; Finkelstein & Levin 1997; Rose 1999; Sommers & Norton 2007; Zeisel & Diamond 1978), we believe that eliminating attorney discretion in jury selection will generate concerns about the appearance of fairness in jury selection that are worth considering alongside the existing concerns about the potential unfairness of the peremptory.

The Social Construction of Bias

We close by highlighting an unexpected, but revealing, pattern of findings regarding perceptions of bias across different groups. We designed Vignette A as a control case, in which a prospective juror reports a remote connection to the boss of one of the attorneys through church, but this juror also minimizes the connection and promises to be fair. The vignette produced stark differences in perceptions of bias by different groups. Although nearly all legal professionals responded as we expected (i.e., said the juror would not be excused), substantial numbers of laypersons thought the judge should excuse this juror. Even within the layperson samples, we found strong county-level differences in reactions. More than half the respondents in County 3 recommended excusing the juror, a far higher rate than nearby County 2 (22 percent) or County 1 (34 percent). The County 3 result did not reflect any general expectation that judges remove all suspect jurors: respondents in this area recommended fewer excuses for cause overall than did jurors elsewhere.

Even among the legal professionals, who all agreed on what the judge would do, ratings of the prospective juror's ability to be fair and impartial differed sharply by occupation type (judge, prosecutor, public defender). Both sets of public defenders (in Lockland and Hartland) were more skeptical about this juror's fairness than were both sets of prosecutors (see Table 8), a difference that was evident in the ratings of all vignettes. However, as we have noted, when this difference occurred for the same-church vignette, it could not be attributable to a presumed adversarial advantage in retaining the juror, since the scenario was altered across attorney type to control for this potential effect.

What might account for the patterns we observed? Because the findings were not anticipated, our explanations are necessarily only suggestive. Differences across the various sets of legal professionals could reflect differences in values and worldviews that are created by or selected for in different occupational or other roles (see generally Kohn & Schooler 1983; Pratto et al. 1997).⁹ Among the layperson samples, County 3 was notably distinguished by having the greatest percentage of African Americans compared to the other areas, per census data. (Although we did not gather racial data on the respondents, we observed the demographic differences in the two samples in which we personally collected data and saw far more African Americans in County 3 than in County 2.) Compared to other groups, religion has more salience in the identity of African Americans (Taylor et al. 2003; Ellison 1993), and African Americans give and receive more social support in their religious communities (Krause 2002; Pargament 1983). The greater identity and felt-closeness in religious communities may lead African Americans to assume that church-related connections that seem “remote” to others actually reflect a strong and shared allegiance.

Whatever the explanation, these unexpected group differences suggest that, as in other situations of trust, perceptions of neutrality and fairness reflect the qualities of the target being judged, qualities of the person doing the assessing, and features of the setting that call for a given person to be fair (see, e.g., Hardin 2001). Jury selection is a context in which the question of whom to trust arises routinely, but the need for an impartial decision maker is present in many domains of life. However, as yet, no general theory accounts for how people make decisions about impartiality. Prior research in psychology demonstrates some consistent biases in how we perceive the fairness of others (Cohen et al. 1988; Hastorf & Cantril 1954; MacCoun 1998; Vallone et al. 1985), and sociological research indicates that regulations on conflict of interest differ markedly across professional groups (Shapiro 2003; see also Tuchman 1972). Thus it is clear that people's methods of demarcating the boundaries between tolerable versus impermissible levels of bias in others have a systematic component. Precisely how people draw those boundaries, and what social factors contribute to this construction, merit further investigation.