Theoretical background: Sensitive questioning in the courtroom and beyond

Social desirability is a widely recognized obstacle in social surveys (Krumpal Reference Krumpal2013), medical history taking (Heritage Reference Heritage, Freed and Ehrlich2010), clinical counseling (Peräkylä Reference Peräkylä1995; Tracey Reference Tracey2016), and other interactional sources of self-reported data (Whitehead Reference Whitehead2009). This problem may arise from fear of material consequences (Rasinski etal. Reference Rasinski, Willis, Baldwin, Yeh and Lee1999), but a more pervasive concern is the desire for approval (Krumpal Reference Krumpal2013; Yan Reference Yan2021). Interviews are necessarily social encounters as well as research instruments (Antaki & Widdicombe Reference Antaki and Widdicombe1998; Maynard, Houtkoop-Steenstra, Schaeffer, & van der Zouwen Reference Maynard, Houtkoop-Steenstra, Schaeffer and Zouwen2002; Potter & Hepburn Reference Potter and Hepburn2005; Drew, Raymond, & Weinberg Reference Drew, Raymond and Weingerg2006), and as such a locus for self-presentation and impression management (Goffman Reference Goffman1959). This gives rise to the ‘dark number problem’ of underreporting for stigmatized attitudes and behaviors, and conversely the overreporting of what is regarded as mainstream or normative, data quality distortions that can be self-reinforcing over time (Noelle-Neumann Reference Noelle-Neumann1993; Kuran Reference Kuran1997).

Researchers have sought methods for minimizing such distortions. Taking stock of these efforts, Schaeffer & Presser (Reference Schaeffer and Presser2003) note they have been more successful in identifying techniques for managing the overall interview situation (e.g. assuring confidentiality, response verification; see Fowler Reference Fowler1995; Schaeffer & Dykema Reference Schaeffer and Dykema2020; Krumpal Reference Krumpal and Maggino2024 for overviews) than for designing questions themselves. But some studies (Holtgraves, Eck, & Lasky Reference Holtgraves, Eck and Lasky1997; Yan Reference Yan2021) have begun to explore face-saving question frames and response options as efforts toward de-toxifying sensitive questions.

For the case of voir dire, social pressures favoring fairness over bias are readily evident in the vastly different ways that claims of fairness as opposed to bias get expressed. The following episode vividly illustrates this contrast. When the attorney suggests that the candidate juror appears to empathize with the defendant (lines 1–2), the juror’s acknowledgment of bias (line 3) is treated as dispreferred (cf.Davidson Reference Davidson, Atkinson and Heritage1984; Pomerantz Reference Pomerantz, Atkinson and Heritage1984). The admission is substantially delayed by “Well”, a pause of silence, and an extended preface. The admission is also qualified by its preface, marked as both subjective (“in my mind”) and uncertain (“I guess”). And it is followed by a no-fault account (lines 5–6) portraying his partiality as driven by circumstance and as normal for “anyone” in his situation. In short, the juror’s empathetic feelings are put forward with manifest caution and are cast as involuntary.

By contrast in the very next exchange, when asked if he can set his feelings aside at trial (lines 9–11), his vow of fairness (lines 13–16) is delivered straightforwardly and expressed in unqualified and steadfast terms.

While it retains a one-word preface (“Oh” precedes “yes”), this strengthens rather than mitigates the response by suggesting that it is obvious (Heritage Reference Heritage1998). And his subsequent extreme-case account (Pomerantz Reference Pomerantz1986; Edwards Reference Edwards2000) portrays fairness not as an involuntary response to circumstance but rather as a trans-situational principle that he “live[s] by” and applies to “everyone and everything”.

The social inhibition against admitting bias may be counterbalanced by a contrary incentive to claim bias as a way of escaping the burden of jury service. Hypothetically, this could motivate some jurors in the opposite direction, while also encouraging judges to avoid giving an ‘easy out’ to jurors ostensibly seeking one. Legal consultants nonetheless recognize social desirability as the predominant problem in voir dire and advise ‘normalizing bias’ to create a context conducive to candor (Broda-Bahm Reference Broda-Bahm2013; Hamilton & Zephyrhawke Reference Hamilton and Zephyrhawke2015; see also Fowler Reference Fowler1995:35–36). Efforts of this sort can be seen in the present data, both in the framing of particular questions and in more general introductory remarks portraying personal biases as commonplace and a natural part of life.

Beyond explicit normalizing comments, what else do court officials do to manage the sensitivities of voir dire questioning? Despite the centrality of voir dire to jury selection and practical interest among attorneys and consultants, empirical research on actual examination conduct is remarkably sparse. Most work has utilized experimental methods (Middendorf & Luginbuhl Reference Middendorf and Luginbuhl1995; Otis, Greathouse, Kennard, & Kovera Reference Otis, Greathouse, Kennard and Kovera2014), mock trials (Gongola Reference Gongola2019), and post-hoc interviews (Broeder Reference Broeder1965; Johnson & Haney Reference Johnson and Haney1994) to examine contextual factors exogenous to examination conduct itself. Meanwhile direct observational research, prominent in studies of trial examinations and dispute resolution (Ehrlich Reference Ehrlich, Wodak, Johnstone and Kerswill2010; Komter Reference Komter, Sidnell and Stivers2013; Conley, O’Barr, & Riner Reference Conley, O’Barr and Riner2019; Drew & de Almeida Reference Drew, Ferraz de Almeida, Coulthard, May and Sousa-Silva2021), remains underdeveloped for voir dire, although some studies have explored the distribution of certain broad action categories (e.g. questioning vs. commenting, open vs. closed questions; Johnson & Haney Reference Johnson and Haney1994, Grosso & O’Brien Reference Grosso and O’Brien2019). Until these initiatives can be developed, much will remain unknown about how the interaction order of questioning mediates a key process central to jury selection and the implementation of justice ideals.

Resources for the study of voir dire can be found in conversation analytic research, where question design is a well-established field of study. Most relevant is work on polar question-answer sequences (Raymond Reference Raymond2003; Heritage Reference Heritage, Freed and Ehrlich2010), which has yielded insight into how specific question-design formats can invite or ‘prefer’ either yes- or no-type responses (Heritage Reference Heritage, Freed and Ehrlich2010), and relatedly how they may suggest a stance toward the probability of affirmation (Heritage & Raymond Reference Heritage and Raymond2021; Raymond & Heritage Reference Raymond and Heritage2021) or its anticipated strength (cf. Couper-Kuhlen, Thompson, & Fox Reference Couper-Kuhlen, Sandra, Barbara, Bolden, Heritage and Sorjonen2023; see Hayano Reference Hayano, Sidnell and Stivers2013 and Bolden, Heritage, & Sorjonen Reference Bolden, Heritage, Sorjonen, Bolden, Heritage and Sorjonen2023 for overviews). Such language-based stances and response preferences may be mobilized in ways that downplay whatever may be socially undesirable about the subject of inquiry, such as asking about abnormal medical conditions (Heritage Reference Heritage, Freed and Ehrlich2010) or student truancies (Pomerantz Reference Pomerantz and Gene2004). Accordingly, in the courtroom context their study can illuminate how court officials orient to and work to address the sensitivities of asking candidate jurors about bias.

This body of work provides analytic resources for the present study, which examines specimens of voir dire questioning in their natural habitat of pre-trial jury selection. We anatomize the overall activity structure of voir dire, and then focus on the subset of questions explicitly targeting juror fairness/bias. We identify key tendencies and preferences in question design, including how bias is referenced, explained, and projected as likely (or unlikely). For each of these we track variations across expanded sequences, and demonstrate that these patterns are accountable as efforts to balance two overarching social norms: (i) showing due diligence in the task of rooting out bias, while also (ii) avoiding offense to prospective jurors. We conclude by considering broader implications for question design in jury selection and other investigative contexts.

Database and methodology

The primary database for this study (also utilized in Fox & Clayman Reference Fox and Clayman2024) consists of four criminal trials drawn from Arizona state courts, and one civil trial from a federal district court in northern California. The infrequency of voir dire recordings necessitated a focus on these courts which, for different reasons, recorded some trials’ voir dire examinations. Arizona state courts permitted trial recordings for a television documentary on the jury system (CBS Reports: Enter the Jury Room, released April 16, 1997). Video recordings and transcripts are archived and publicly available at various law libraries, and both were obtained for the present study. The northern California federal court trial was recorded in 2016 for a pilot program evaluating the effect of cameras in district courtrooms. Videos were not available at the time of the present study, but transcripts were obtained from the Westlaw legal research platform. Further information about the pilot program and associated data can be found at the US Courts website.Footnote ³

Case selection from these sources continued until saturation was reached. The resulting database consists of criminal trials for armed robbery, aggravated assault, and drug possession and sale (the latter including both an initial trial and a retrial), and a civil trial for a boating collision lawsuit. All are low-profile and routine cases of their type. Except for the retrial, each case was overseen by a different judge. A small number of questions from other Arizona court cases were also examined qualitatively, but these do not figure in the quantitative analysis.

Our primary analysis concerns exchanges where the juror’s fairness or bias is explicitly thematized as the target of inquiry (e.g. “Would you have any difficulty serving as a fair and impartial juror in this case?”). A comprehensive review of the database yielded sixty sequences of this kind. Of these, thirty-five are single question-answer sequences, and twenty-five sequences are expanded with follow-up questions, for a total of N = 107 fairness/bias questions.

Our methods began with context-sensitive case analysis in the conversation analytic tradition. In this phase, each question was examined for its design features, sequential placement, and subsequent uptake. Patterns emerging from the case analyses then suggested features to be coded systematically across the databaseFootnote ⁴ and subjected to statistical measures of frequency and association. This combination of methods—sequential case analysis plus quantification—has become commonplace in applications of conversation analysis to medicine, media, and other institutional areas, as well as longitudinal and cross-cultural comparative projects (see Clayman & Heritage Reference Clayman and Heritage2021; Robinson Reference Robinson, Robinson, Clift, Kobin and Raymond2024; Helmer, Deppermann, & Stivers Reference Helmer, Deppermann and Stivers2025). The presentation of findings broadly mirrors this two-stage approach, with each dimension of questioning first exemplified with sequential case analyses, then examined for frequency and distribution across questioning episodes.

Overview and activity structure

In our data the pool of prospective jurors is co-present throughout voir dire, and the judge plays a predominant role in launching the process, summarizing the case, asking the majority of questions (cf. Johnson & Haney Reference Johnson and Haney1994), and announcing juror selection decisions. Attorneys are invited to ask questions later, usually of jurors already questioned by the judge.

The examination process itself typically unfolds as a series of episodes, each comprised of three ordered phases as sketched in Figure 1.

Figure 1. Voir dire activity structure.

In phase 1, problem identification, the judge (or an attorney) identifies some circumstance as a possible biasing factor. These include being acquainted with the parties involved, having parallel life experiences or deep-seated ethical concerns, or other circumstances that might predispose the juror to disregard trial evidence or favor one side in the case.

In phase 2, juror self-disclosure sequence, the judge/attorney then asks via a show of hands if the circumstance identified in phase 1 applies to any candidate jurors, and selects one to disclose their personal situation. The disclosure may be completed within a single exchange, or expanded with additional questions pursuing clarification or further detail.

In phase 3, juror self-assessment sequence, shifting from the just-disclosed circumstance to the upcoming trial, the judge/attorney asks the candidate to reflect on their capacity for fairness or bias if selected for the jury. The self-assessment may be completed within a single exchange, or expanded with further questions geared to resolving ambiguity or securing a more definitive self-assessment (cf. Houtkoop-Steenstra Reference Houtkoop-Steenstra1987; Lindström Reference Lindström, Raymand, Heritage and Lerner2017).

The last two phases are repeated for other jurors indicating that the same biasing factor may apply to them; and all three phases are repeated to explore other factors. When attorneys return to a juror previously questioned by the judge, the phases may be reduced. The recursive three-phase process sketched above is thus an ideal type, describing in broad strokes what transpires during voir dire until a full complement of jurors and alternates has been selected.

Juror self-assessment question design: Targeting bias over fairness

The remaining analysis zeroes in on phase 3 and the design of juror self-assessment questions. These questions follow the disclosure of some potentially biasing circumstance, and they fall within a new sequence that shifts to future conduct at trial. They have a number of characteristic design features illustrated in excerpts (3) through (6) below. They are commonly formed as polar questions establishing the relevance of a yes- or no-type response (Raymond Reference Raymond2003). And they deploy grammatical forms that lean in favor of a yes-type answer (Heritage Reference Heritage, Freed and Ehrlich2010), most often straight interrogatives (e.g. “Is there…” in excerpt (6) below) or positive declaratives (e.g. “You could be…” in excerpt (3); see also excerpts (4) and (5)) seeking confirmation of the juror’s anticipated performance. Both of these forms advance a proposition that asserts some state of affairs rather than negating it (cf. “You couldn’t be…”), and are in that way positively polarized and hence built for a yes-type response. In general, then, juror self-assessment questions are linguistically designed for a positive affirmation of whatever proposition about future conduct is being put forward.

Notwithstanding these commonalities, the questions vary in other ways, with the most basic distinction centering on the nature of the conduct in question. Questions may reference and thematize either potential fairness, or alternatively potential bias. Those targeting fairness, exemplified in the next two excerpts, invite the recipient to assess their capacity for impartiality (“You could be fair and impartial” in excerpt (3), line 1) or evidence-based conduct (“you would accept and examine all of facts” in excerpt (4), lines 2–4) at trial. In this context an affirmative answer promises fairness if selected for the jury.

Since fairness is, as noted earlier, culturally normative and desirable, these questions’ grammatical preference is congruent with and reinforcing of the questions’ content-based social preference (Heritage Reference Heritage, Freed and Ehrlich2010). These jurors are, in effect, being grammatically invited to affirm a socially desirable ideal, and the subsequent affirmative responses are indeed fully aligned with both preferences. Both responses are also relatively prompt and straightforward. Although they are subjectivized (see the initial response “I believe so” in excerpt (3), line 3; and the subsequent elaboration “I think I could” in excerpt (4), line 6), subjectivization of this kind is a characteristic feature of courtroom talk (Szczyrbak Reference Szczyrbak2013, Reference Szczyrbak2020) and a recurrent method of exhibiting ‘carefulness’ or ‘circumspection’ in this formal investigative setting (see also Edwards & Potter Reference Edwards and Potter2017).

By contrast, questions targeting bias, as in the next two excerpts, invite the recipient to assess their potential for favoritism or partiality. For these questions, an affirmative answer (e.g. excerpt (5)) entails an admission of some potentially disqualifying form of bias.

In bias-targeted questions, the grammatical and social preferences are cross-cutting (Heritage Reference Heritage, Freed and Ehrlich2010), with the question’s conducive grammar favoring affirmation of something that is socially undesirable. In effect, the juror is being invited to admit to a socially noxious predisposition toward favoritism. As might be expected, responses here cannot easily align with both preferences and are more frequently non-conforming (Raymond Reference Raymond2003) and relatively elaborate. In excerpt (5) the admission of bias (“yes I do”) is grammatically aligned but socially dispreferred. It is also substantially delayed and elaborately framed as ‘reluctant’ through a prefatory clause (“I have to answer honestly”) that includes a modal verb of obligation and an explicit claim to sincerity (recurrent in dispreferred responses; Edwards & Fasulo Reference Edwards and Fasulo2006). Conversely, in excerpt (6) the vow of fairness is socially preferred but grammatically disaligned; it is also qualified as probabilistic, and subsequently dialed back by noting circumstances under which they would “probably have some bias”.

The simple binary distinction between fairness- and bias-targeted questions glosses over substantial variation within each category, but this rough distinction nonetheless matters for examiners who discriminate between them in a way that favors questions targeting bias over questions targeting fairness. One general indication of this preference is that the question types are not equally frequent (Table 1),Footnote ⁵ with bias-targeted questions significantly outnumbering fairness-targeted questions by almost a 3:1 ratio.

Table 1. Questions targeting fairness versus bias.

χ² = 24.04, p < .0001***

Beyond this aggregate tendency, a preference for bias-targeted over fairness-targeted questions is further evident in the emergent construction of individual cases, when questions are revised or subject to self-repair (Schegloff, Jefferson, & Sacks Reference Schegloff, Jefferson and Sacks1977). In each of the next two cases, an apparent fairness-targeted question is aborted prior to completion (see (7), line 1 and (8), line 2) and then replaced with an alternative question targeting bias (see (7), line 3 and (8), line 3).

In the first case, the self-repair is ‘over-exposed’ (Bolden et al. Reference Bolden, Hepburn, Potter, Zhan, Wei, Park, Shirokov, Chun, Kurlenkova, Licciardello, Caldwell, Mandelbaum and Mikesell2022) through an explicit comment (“let me ask it a different way”) that retroactively acknowledges the normative breach-in-progress and projects its rectification. The second case is less explicit, but both cases show the judge actively choosing between alternative question-design formats, launching a positively valenced interrogative frame projecting a question about fairness, but then abandoning this in favor of a bias-targeted alternative.

All of this suggests that judges and attorneys are on the lookout for bias, building their questions in such a way as to probe for its possible existence. This orientation toward what is undesirable stands in stark contrast to the predominantly benign orientation found in ordinary conversation (Maynard Reference Maynard2003; Raymond & Heritage Reference Raymond and Heritage2021) and some institutional contexts (Pomerantz Reference Pomerantz and Gene2004; Heritage Reference Heritage, Freed and Ehrlich2010). In medicine for instance (Heritage Reference Heritage, Freed and Ehrlich2010), clinicians’ history-taking questions are routinely geared toward biomedically desirable responses (e.g. indicating health rather than illness, with exceptions for particular patients). The contrasting orientation toward problematicity in voir dire is compatible with the courtroom context in a variety of ways. It fits the immediate sequential context where the juror has just disclosed some potentially biasing circumstance (in phase 2); the subsequent question designedly targets this just-revealed circumstance and probes its consequentiality at trial. Furthermore, targeting bias pushes back against the social desirability pressure to answer ‘correctly’ with a vow of fairness. To frame the question otherwise (cf. “So you could be fair at trial?”), anticipating ideal conduct just when this has been cast into doubt, could be seen as ignoring or perhaps whitewashing a potential problem. Accordingly, targeting bias displays a semblance of ‘due diligence’ in working to uncover elusive indications of favoritism or prejudice.

Probing for bias is the default orientation of judges and attorneys, but this orientation is also context-sensitive and may be modified depending on how the juror responds. In expanded self-assessment sequences, if the juror is initially optimistic regarding their capacity for fairness, the follow-up question may register this by targeting fairness instead. For instance, here the attorney first asks if prior difficulties with public transit might have engendered “some animosity or bias” against transit employees (lines 1–4). But when the juror rejects that view and claims “no specific feelings” toward any transit employees (lines 5–9), the next question shifts the target from bias to fairness (lines 10–11); and the question’s “so” preface (line 10) frames it as building upon the juror’s prior optimistic response (Bolden Reference Bolden2006).

The trajectory evident in this excerpt reflects a recurrent pattern in expanded self-assessment sequences. Table 2 disaggregates the distribution of fairness/bias questions by their position within expanded episodes of questioning. While bias-targeted questions predominate across the board, this tendency is strongest in initial questions and declines later in subsequent questions to the same juror. In the sequence-initiating position, bias-targeted questions comprise more than 80% of the cases, and some of the less-frequent fairness-targeted questions in that position (just 17% of cases) have other compositional features suggesting skepticism about the capacity for fairness. In later questions, after jurors have expressed optimism about their own capacity for fairness, this asymmetry diminishes (the bias-to-fairness ratio drops from ≈5:1 to < 2:1 as the proportion of fairness questions more than doubles) to an extent that is statistically significant. All of this is consistent with court officials being on the lookout for bias by default, but reducing that orientation in subsequent questions contingent on the juror’s response.

Table 2. Fairness/bias questions by sequential position.

χ² = 5.75, p < .05*

Mitigating bias questions: Referencing bias indirectly

Bias-targeted questions, while predominant, have a number of convergent design features geared to mitigating or detoxifying the inquiry. These design patterns reflect court officials’ orientation to the socially stigmatized nature of bias, and they are accountable as efforts to manage the inhibitions that such stigma may entail.

Consider, first, how the concept of bias is referenced. It may be formulated directly as bias (e.g. excerpts (5), (6)) or with synonymous terms such as partiality, sympathy/animosity toward one side, etc. (e.g. excerpt (1), line 1; excerpt (9), line 2).

Alternatively, bias may be referenced indirectly,Footnote ⁶ with two such indirect forms recurrent in the data and relevant to mitigation. In the litotes version, bias is formulated indirectly by problematizing its opposite; that is, characterizing fairness as elusive, hampered, or only partially achieved. The disclosed circumstance may be said to “affect your ability to be fair” (much like excerpts (10), (11)), or to “make you less than fair and impartial” (excerpt (12)).

Litotes formulations, by virtue of their roundabout or circuitous construction (e.g. difficulty being fair instead of biased), are a widely used resource for cautiously broaching sensitive matters (Bergmann Reference Bergmann, Drew and Heritage1992; Lutfey & Maynard Reference Lutfey and Maynard1998).

An alternative form of indirectness, arguably even less direct than the litotes version, avoids the language of fairness/bias altogether in favor of a more allusive formulation. The question may refer in a general way to difficulty “[serving] as a juror” (excerpt (13)) or difficulty “decid[ing] this case” in (14).

Such formulations allude to bias while steering clear of the terminological framework of fairness/bias altogether.

Here again, these distinctions are not merely academic. Judges and attorneys discriminate between alternative formulations in a variety of ways that favor indirect over direct references to bias. This is evident, first, in the aggregate frequency of these alternatives (Table 3), with questioners choosing indirect bias reference forms over three times more frequently than direct forms.

Table 3. Direct vs. indirect bias questions.

χ² = 24.01, p < .0001***

This evident preference for indirect references to bias is, much like the preference for bias-targeted questions more generally, a default orientation that is context-sensitive and subject to modification contingent on the juror’s response. Just as optimistic responses can prompt fairness-targeted follow-up questions (as seen in excerpt (9) above), uncertain or explicitly pessimistic responses can prompt a shift to questions referencing bias in a more direct and straightforward way (lines 13-15).

Here the initial question targets bias while referring to it allusively (“would… affect your ability to be a juror in this case”, lines 4–5). But after the juror’s pessimistically inflected response (“well possibly”, line 6), the follow-up query formulates bias in a direct and even blunt manner (“You believe you might be prejudiced in this case”, lines 13–14).

The trajectories evident in excerpts (9) and (15) above are recurrent in expanded sequences (Table 4). The tendency to target bias indirectly is strongest in sequence-initiating questions where indirect reference forms outnumber direct forms by more than a 5:1 ratio. In subsequent questions to the same juror, indirect forms decline in frequency while both direct bias forms and fairness forms increase in frequency. Questions thus become more polarized over the course of expanded sequences, shifting toward the direct targeting of either bias or fairness contingent on the juror’s response to the initial question.Footnote ⁷ It is noteworthy, however, that this shift toward polarization is modest in the aggregate, with indirect bias questions remaining more frequent than direct bias forms (by a 2:1 ratio) even in follow-up questions, hence after the juror has expressed optimism or pessimism about their capacity to serve.

Table 4. Fairness/bias questions by sequential position.

χ² = 8.29, p < .05*

Mitigating bias questions: Diffusing responsibility for bias

A second locus of variation among bias-targeted questions concerns how the juror’s responsibility for bias may be underscored or diffused through the linguistic encoding of agency. Lexico-grammatical practices for representing agency are a universal feature of human languages (Duranti Reference Duranti and Duranti2004). While such practices do not provide a direct ‘window into the mind’ of participants (Edwards & Potter Reference Edwards and Potter1993; Ochs, Gonzales, & Jacoby Reference Ochs, Gonzalez, Jacoby, Ochs, Emanuel and Sandra1996), they are public discursive acts that evidently matter for observers’ perceptions and inferences (Fausey, Long, Inamori, Boroditsky Reference Fausey, Long, Inamori and Boroditsky2010; McGlone, Bell, Zaitchik, & McGlynn Reference McGlone, Bell, Zaitchik and McGlynn2013; Bell, McGlone, & Dragojevic Reference Bell, McGlone and Dragojevic2014; Chen, McGlone, & Bell Reference Chen, Matthew and Bell2015) and for language users themselves (Budwig Reference Budwig1990; Duranti Reference Duranti1990; Bohner Reference Bohner2010; Conley Reference Conley2013, Reference Conley2015). In the voir dire context, the relevant distinction is between questions that feature (i) the juror as opposed to (ii) the circumstance as the subject of the sentence or clause and, by implication, the main driving force in the process projected as fair or biased.

The juror is spotlighted as the sentential subject and primary agent in both of the following questions, the first targeting fairness (“You could be fair and impartial…”) and the second targeting bias (“Are you going to give their evidence less weight…”).

By contrast, in the next examples the circumstantial factor is spotlighted as the sentential subject and primary agent, while the juror is cast as the object acted upon by circumstance. This construction is used in questions targeting bias directly (excerpt (18)) and through more indirect litotes in (19) and allusive formulations in (20).

In each case, the circumstance is in the subject position and is thus treated as the causal factor that will “bias you” in (18) or “affect your ability” to be fair or to serve (see (19), (20)). Responsibility for bias is thereby diffused in a way that subordinates the juror’s role relative to the determinative circumstances.

Not surprisingly, almost all fairness-targeted questions have the juror as subject and primary agent (e.g. excerpt (16), also (3) and (4)). Responsibility for rendering a fair judgment is routinely attributed to the juror. By contrast, bias-targeted questions exhibit more variability in the construction of agency, with an evident preference for spotlighting the circumstance as the driving factor behind bias. As with other question design preferences documented above, this preference for the circumstance as subject is evident in both single cases and aggregate patterns.

At the single-case level are instances involving self-repair. Here the question is launched with the juror as subject/agent (“Are you…”, line 1), but that construction is aborted before completion and then replaced (line 2) with an alternative version spotlighting the circumstance (“anything about that experience”) as impacting the juror’s “ability to be fair”.

Here the judge can be seen switching between alternate constructions in a manner that ultimately works to deflect responsibility for bias away from the juror (cf. excerpt (10) above).

There is also aggregate distributional evidence for the preference evident in this case (see Table 5). Among bias-targeted questions, constructions spotlighting the circumstance as subject/agent are twice as frequent as those spotlighting the juror. So unlike fairness-targeted questions that overwhelmingly credit the juror for impartiality, bias-targeted questions tend to deflect responsibility away from the juror and toward the circumstance.

Table 5. Bias-targeted questions with juror vs. circumstance as subject.

χ² = 11.85, p < .001***

This is, moreover, a default orientation in questioning that is transformable contingent on the juror’s response. In the next exchange (seen earlier in excerpt (15) above), the judge’s initial bias-targeted question spotlights the circumstance as subject/agent (“Anything about that experience…”, lines 1–5). But after the juror expresses pessimism about their capacity for fairness (lines 6–11), the follow-up question invites re-confirmation of what is treated as an admission while shifting the spotlight from circumstance to juror (“You believe that you might be prejudiced…”, lines 13–15).

The follow-up question retains a reference to circumstance, but this is relegated to a subordinate clause (“because of your prior experience”); the juror is now the sentential subject.

The pattern evident in the preceding excerpt—where the subject/agent spotlight shifts from the circumstance to the juror over the course of an expanded sequence—turns out to be a recurrent feature of expanded sequences targeting bias (see Table 6). Initial questions overwhelmingly spotlight the circumstance over the juror as subject/agent by almost a 7:1 ratio. But later questions tend to spotlight the juror more than the circumstance, albeit to a lesser degree (< 2:1 ratio).

Table 6. Bias-targeted questions with juror vs. circumstance as subject, by sequential position.

χ² = 18.87, p < .0001***

Mitigating bias questions: Projecting bias as minimal or unlikely

A third locus of variation concerns whether the question ‘anticipates’ a response from the juror proclaiming either fairness or bias. As noted earlier, fairness/bias questions are grammatically designed for yes, through straight interrogatives or positive declaratives, in an overwhelming proportion of cases (≈95%). Notwithstanding the near-universality of yes-preferring question formats, there is a residue of variation in the inclusion (or not) of certain lexical items—some/something or any/anything—that have been shown to modulate this preference. In field experiments varying only this one lexeme in polar questions (Heritage, Robinson, Elliott, Beckett, & Wilkes Reference Heritage, Robinson, Elliott, Beckett and Wilkes2007), some and any have been shown to be positively and negatively polarized, respectively, as evidenced by their impact on the polarity of responses. In subsequent naturalistic research focusing on how speakers deploy these terms in context, Heritage & Raymond (Reference Heritage and Raymond2021) conclude that they encode a stance anticipating the likelihood of affirmation, with some casting affirmation as likely and any as unlikely. Other researchers (Couper-Kuhlen et al. Reference Couper-Kuhlen, Sandra, Barbara, Bolden, Heritage and Sorjonen2023) offer a somewhat different account of any, arguing that its import, while context-dependent, can in some contexts work to lower the bar for affirmation.

These analyses of any, although differing in specifics and suggesting different impacts on responses, are not incompatible at the level of stance and implicature. A practice that ‘lowers the bar for affirmation’ may also carry the implication that affirmation is ‘unlikely’, on the basis that the question’s target phenomenon is presumptively scarce so as to require ‘lowering the bar’ to capture it. Furthermore, both interpretations have an a priori plausibility in the voir dire context, being broadly congruent with the dual institutional norms sketched at the beginning of this article. ‘Lowering the bar’ for reportable bias dovetails with the norm of showing due diligence, since it can be understood as working to encourage bias disclosures even of the smallest or most negligible sort. Conversely, casting bias as ‘unlikely’ dovetails with the norm of avoiding offense to prospective jurors.

Actual patterns of use in voir dire reinforces the view that any/anything is normatively evocative in this context and a mitigating resource. Consider, first, that the terms are used primarily within bias-targeted questions, with any/anything being far more frequent than some/something and appearing most often in sequence-initiating questions. For instance, here an initial bias-targeted exchange is launched with any.

And in the following expanded sequence, any/anything is used twice but only in the first bias-targeted question, while some is used the follow-up question (arrowed). Following the juror’s disclosure of a noteworthy personal relationship (her husband is a deputy sheriff), the judge’s initial question (lines 1–4) targets bias allusively and is twice modulated with any/anything suggesting that bias is apt to be minimal or unlikely.

By contrast, after the juror admits to uncertainty about her capacity for fairness (lines 6–9, 11), the judge’s follow-up question (lines 12–13) takes a more direct form and switches from “any” to “some” (“D’you have some reservation then about…”). This inclusion of “then” immediately after the shift marks it as an inference (Haselow Reference Haselow2011) from the juror’s prior admission. In short, the judge shifts from an indirect question that is also relatively ‘optimistic’ about the juror’s capacity to serve, to a more direct and ‘pessimistic’ form after the juror shows reluctance to promise fairness.

Distributional evidence demonstrates that the preceding cases are characteristic of bias-targeted questions generally (Table 7). The inclusion of either some or any is treated as optional, with more than a third of the questions having neither term. When they are used, any is far more frequent than some by almost a 5:1 ratio, with fully half of all bias-targeted questions mitigated in this way. By treating the existence of bias as minimal or unlikely, such questions are inflected with a modicum of ‘optimism’ about the recipient’s capacity to serve.

Table 7. Some vs. any in bias-targeted questions.

χ² = 19.79, p < .0001***

Disaggregating by sequential position (Table 8), any is used disproportionately in initial bias-targeted questions and is much less frequent in later questions. Some exhibits the inverse distribution in being clustered in later questions, although it is not very common even in that position. This overall pattern is indicative of a default impetus to probe for bias while casting it as minimal or unlikely, suggesting initial ‘optimism’ about the juror’s capacity to serve that diminishes and may yield to relative ‘pessimism’ following jurors’ subsequent admissions.

Table 8. Some vs. any in bias-targeted questions, by sequential position.

χ² = 10.8236, p < .01**

Earlier it was proposed (based on Heritage & Raymond Reference Heritage and Raymond2021; Couper-Kuhlen et al. Reference Couper-Kuhlen, Sandra, Barbara, Bolden, Heritage and Sorjonen2023) that the minimizing stance encoded by any (‘lowering the bar’ for reportable bias while anticipating that it is scant or unlikely) is intertwined with the institutional norms of due diligence and offense avoidance. Carrying this argument one step further, this normative functionality may be underscored and accentuated when the lexeme is used multiple times in elaborated questions inviting a multidimensional and fine-grained search for bias.

The two excerpts examined previously illustrate the contrast between relatively broad (excerpt (23)) versus fine-grained (excerpt (24)) bias queries. In excerpt (23), any is used once in a single referring expression pointing toward undifferentiated reservations about serving (“any misgivings about sitting on a jury”). By contrast in excerpt (24) any/anything is used twice within referring expressions that are more narrow and distinctly targeted. The first use (line 1) points toward aspects of her spousal relationship that could be problematic (“is there anything that you may have learned about your husband’s experience or job”); the second use (lines 3–4) points toward various ways this might affect her in court (“that may impact in any fashion your ability to sit as a juror…”). Unlike the single broad use in excerpt (23), these distinctly targeted uses invite a more meticulous search for possible sources and manifestations of bias.

A similar compound and fine-grained query appears in the next excerpt. Here again there are two instances of any/anything, the first referencing the juror’s past experience (first arrow), and the second various ways this might affect her in court (last arrow). But the judge here goes one step further, spelling out three distinct aspects of her experience that could be problematic (lines 2–3: witnessing a similar crime, testifying before a grand jury, and testifying at trial).

Accordingly, this juror is being invited to inspect several different elements of her experience, as well as consider the varying ways they could impair her judgement. This calls for a fine-grained and intensive search, which may implicate the minimal or non-obvious nature of bias as well as a lowered bar for what should be deemed worthy of reporting.