On Racial Diversity and Group Decision Making: Identifying Multiple Effects of Racial Composition on Jury Deliberations

INTERPERSONAL RELATIONS AND GROUP PROCESSES
On Racial Diversity and Group Decision Making: Identifying Multiple
Effects of Racial Composition on Jury Deliberations
Samuel R. Sommers
Tufts University
This research examines the multiple effects of racial diversity on group decision making. Partici-
pants deliberated on the trial of a Black defendant as members of racially homogeneous or
heterogeneous mock juries. Half of the groups were exposed to pretrial jury selection questions
about racism and half were not. Deliberation analyses supported the prediction that diverse groups
would exchange a wider range of information than all-White groups. This finding was not wholly
attributable to the performance of Black participants, as Whites cited more case facts, made fewer
errors, and were more amenable to discussion of racism when in diverse versus all-White groups.
Even before discussion, Whites in diverse groups were more lenient toward the Black defendant,
demonstrating that the effects of diversity do not occur solely through information exchange. The
influence of jury selection questions extended previous findings that blatant racial issues at trial
increase leniency toward a Black defendant.
Keywords: racial diversity, group composition, decision making, jury deliberations, jury selection
When any large and identifiable segment of the community is
with similar issues across a variety of domains. Much contempo-
excluded from jury service, the effect is to remove from the jury room
rary discourse on racial diversity has focused on competing ide-
qualities of human nature and varieties of human experience, the
ologies and value systems (Norton, Sommers, Apfelbaum, Pura, &
range of which is unknown and perhaps unknowable. It is not neces-
Ariely, in press; Plaut, 2002; Richeson & Nussbaum, 2004). In
sary to assume that the excluded group will consistently vote as a class
debate over affirmative action, for example, one camp has cham-
in order to conclude, as we do, that its exclusion deprives the jury of
pioned colorblindness as a moral imperative and decries as racism
a perspective on human events that may have unsuspected importance
in any case that may be presented.”
any consideration based on race (Connerly, 2000; Gratz v.
—Justice Thurgood Marshall, Peters v. Kiff
Bollinger, 2003, petitioners’ arguments). On the other hand, pro-
ponents of affirmative action have emphasized the need to remedy
The quotation above comes from a ruling in which the U.S.
historical inequities and suggest that racial representativeness in
Supreme Court addressed the controversial issue of racial diversity
the classroom and workplace constitutes a compelling societal
and juries. Even though Thurgood Marshall’s words are over 30
interest (Crosby, 2004; University of California Regents v. Bakke,
years old, they resonate today in a society that continues to wrestle
1978). Such conflicts in ideology are polarizing and difficult to
resolve. As Justice Marshall’s comments imply, a more promising
means of evaluating racial diversity is to consider its observable
influence on group performance.
I would like to recognize Phoebe Ellsworth for her continued collabo-
The present article heeds this suggestion and examines in a
ration and invaluable contributions to this line of research and Nalini
jury context the influence of racial composition on group de-
Ambady, Amy Bradfield Douglass, Monique Fleming, Richard Gonzalez,
Samuel Gross, Norbert Kerr, Keith Maddox, and Michael Norton for their
cision making. The mock jury paradigm provides a realistic and
helpful comments on the interpretation of this study and drafts of this
engaging means for examining group decision making, as par-
article. This data collection would not have been possible without the
ticipants are forced to work together to evaluate ambiguous
interest and generous cooperation of judges Melinda Morris and Donald
information and reconcile their interpretations of it. In this
Shelton, as well as the assistance of Yvonne Boyd and Suzette Rygiel-
setting, a premium is placed on a group’s fact-finding abilities
Abella, all of the Washtenaw County, Michigan courthouse. This article is
and final decision, not on its morale, rendering the jury an ideal
based in part on a doctoral dissertation from the University of Michigan,
vehicle for examining variables that affect group decision mak-
which was supported by a predoctoral fellowship from the Rackham
ing. An additional benefit of the jury context is the potential of
Graduate School.
such research to produce findings of practical as well as theo-
Correspondence concerning this article should be addressed to Samuel
R. Sommers, Department of Psychology, Tufts University, 490 Boston
retical importance. Decades after Peters v. Kiff (1972), contro-
Avenue, Medford, MA 02155. E-mail: sam.sommers@tufts.edu
versy still abounds regarding the role of race in the legal
Journal of Personality and Social Psychology, 2006, Vol. 90, No. 4, 597– 612
Copyright 2006 by the American Psychological Association 0022-3514/06/$12.00
DOI: 10.1037/0022-3514.90.4.597
597

---

598
SOMMERS
system. Among the general public and media, unpopular jury
regarding this first proposition, Moreland et al. (1996) have sug-
verdicts are frequently attributed to racial composition, and
gested that racial diversity may be even more likely than other
intuitions regarding juror race are often treated as facts needing
types of heterogeneity to create interpersonal conflict and threaten
no corroboration (see Cowan & Fairchild, 1997; Reynolds,
morale. Consistent with this conclusion, McLeod, Lobel, and Cox
1996). To date, little research has tested these assumptions or
(1996) reported a study in which White participants enjoyed a
examined the more basic psychological processes through
brainstorming task more when their group was racially homoge-
which diversity affects group decision making. The present
neous, even though diverse groups generated more creative and
research does just that, and though it examines decision making
feasible ideas. In another study—the only experiment published to
in the context of a jury evaluating the trial of a Black defendant,
date that directly examines the cognitive processes through which
this article focuses on the more general relationship between
racial diversity influences group discussion—Whites demonstrated
racial diversity and processes of group decision making.
more complex thinking when assigned to a diverse group than
when assigned to an all-White group (Antonio et al., 2004). Spe-
Effects of Diversity
cifically, White participants who discussed a controversial social
issue (either international child labor practices or capital punish-
It is well documented that a group’s composition can affect its
ment) in a group with a Black confederate wrote postdiscussion
dynamics and performance, but the exact nature of diversity’s
essays that were coded as higher in integrative complexity than did
impact remains the subject of debate (see Mannix & Neale, 2005).
White participants in homogeneous groups. The conclusion that
The most frequently mentioned negative outcome of diversity—
racial diversity has potential performance benefits is further bol-
broadly defined—is interpersonal conflict (see De Dreu & Wein-
stered by the fact that on more than one occasion, social psychol-
gart, 2003). More specifically, various types of heterogeneity can
ogists have attempted to convince trial courts of the positive
reduce the quantity and quality of group communication
relationship between racial heterogeneity and long-term outcomes
(Maznevski, 1994; Zenger & Lawrence, 1989) as well as predict
such as intellectual engagement, academic motivation, and devel-
decreases in group cohesion and morale, outcomes that in turn lead
opment of social skills (Dovidio, 2002; Gurin, 1999).
members to seek alternative groups or to simply drop out (Jackson,
1992; McCain, O’Reilly, & Pfeifer, 1983; O’Reilly, Caldwell, &
Racial Diversity and Group Decision Making
Barnett, 1989). The potential negative impact of diversity is not
limited to morale but can also be seen in a group’s actual perfor-
The objective of the present research is to examine the effects of
mance (Ancona & Caldwell, 1992; DeBiasio, 1986; Mullen &
racial diversity in the specific domain of group decision making.
Copper, 1994). However, as Moreland, Levine, and Wingert
The relationship between group composition and performance in
(1996) have pointed out, deleterious effects for performance are
general is clearly complicated, but from a strictly decision-making
most likely under certain circumstances, such as on a simple task,
perspective, both sides of the debate regarding diversity effects are
when the decision requires convergent thinking or when hetero-
compatible with the hypothesis that groups often benefit from
geneity also leads to variability in group members’ abilities.
racial heterogeneity. The extent to which racial diversity facilitates
Other theorists have noted that cohesion and morale do not
information exchange and problem solving certainly indicates ad-
ensure good group performance (e.g., Janis, 1982; Kameda &
vantages for heterogeneous groups, especially for complex deci-
Sugimori, 1993), and research demonstrating the advantages of
sions. But even interpersonal conflict— often mentioned as the
diversity typically has focused on performance benefits. The oft-
principal negative result of diversity—may be useful when a
cited benefits of diversity include increases in group creativity,
group’s primary goal is not boosting morale but rather good and
information sharing, flexibility, and thoughtfulness (Hoffman &
thorough decision making. Consider, for example, the jury, for
Maier, 1961; Nemeth, 1995; Phillips, Mannix, Neale, & Gruen-
which positive affect and group cohesion are less important than
feld, 2004; Triandis, Hall, & Ewen, 1965). These outcomes are
fact-finding ability and a willingness to consider the entire range of
particularly likely when a task is complex, requires divergent
a community’s viewpoints (Ellsworth, 1989; Johnson v. Louisi-
thinking, or requires interaction with nongroup members (Levine
ana, 1972, dissenting opinion; Wilkenfeld, 2004). In this setting,
& Moreland, 1998; Moreland et al., 1996). Research has also
as in other contexts in which turnover is not a primary concern (or
suggested that many of the threats to morale posed by diversity
possibility), racial heterogeneity is likely to have positive effects
weaken or disappear over time as group members learn to work
on decision-making processes and outcomes.
with one another and even become proud of their heterogeneity
But published comparisons of the decision making of racially
(Allmendinger & Hackman, 1995; Jehn, Northcraft, & Neale,
homogeneous and heterogeneous groups are scarce, and our un-
1999; Watson, Kumar, & Michaelsen, 1993). In other words, to the
derstanding of the processes through which racial composition
extent that a group can weather the initial conflict that diversity
affects decision making is limited. This is the case even in the legal
sometimes creates— or even use that conflict to its advantage—
domain, in which these questions carry great practical importance.
diversity often has observable benefits for group performance and
Recent analyses of actual trials have supported the proposition that
problem solving.
decision making varies by jury racial composition, as the greater
Despite the extensive literature on group composition, very few
the proportion of Whites on a jury, the harsher that jury tends to be
experiments have examined the specific case of racial diversity by
toward non-White defendants (Bowers, Steiner, & Sandys, 2001;
comparing all-White and racially mixed groups. It seems reason-
Daudistel, Hosch, Holmes, & Graves, 1999). However, these anal-
able to predict that as with other forms of diversity, the downsides
yses are correlational and fail to reveal the processes through
of racial heterogeneity typically involve interpersonal conflict,
which group composition is influential. A study by Kerr, Hymes,
whereas its benefits most often relate to performance. Indeed,
Anderson, and Weathers (1995) has demonstrated that the mere

---

RACIAL DIVERSITY AND DECISION MAKING
599
expectation of participating in a racially diverse jury can be influ-
Such an analysis also has clear practical implications for the
ential, though this study did not examine actual deliberation ef-
legal system’s consideration of jury diversity. The bases for re-
fects. In a rare experimental investigation of jury racial composi-
quirements of jury representativeness are not principally perfor-
tion and deliberations, Bernard (1979) showed 10 college student
mance related; they include Constitutional concerns regarding the
juries a trial video with either a White or a Black defendant. The
rights of defendants and jurors (e.g., Batson v. Kentucky, 1986) as
only jury to reach a unanimous guilty verdict was also the only
well as a desire to maintain the perceived legitimacy of the system
all-White jury to view the trial of a Black defendant. Research with
(Wilkenfeld, 2004). However, as the opening quotation of this
Latino and Indian participants has also found all-White juries to be
article reveals, another potential justification is that diverse juries
more conviction prone than diverse juries in judging a minority
are often better decision makers than homogeneous ones. Support
defendant (Chadee, 1996; Perez, Hosch, Ponder, & Trejo, 1993).
for this proposition usually comes in the form of information
Again, though, even these laboratory studies shed little light on the
exchange explanations such as Thurgood Marshall’s: Diverse ju-
processes through which racial composition affects deliberations,
ries enjoy wider ranging discussions because White and Black
relying instead on a demographic argument that composition de-
jurors bring different experiences and perspectives to the jury
termines a jury’s predeliberation vote split and, thus, its final
room (Peters v. Kiff, 1972). But what of other performance ad-
verdict (see Kalven & Zeisel, 1966).
vantages to jury diversity? Hans and Vidmar (1982) raised one
Whereas few studies have examined the processes through
such possibility decades ago, writing that the presence of minority
which racial diversity is influential, more general research on
group jurors “may inhibit majority group members from express-
group decision making provides theoretical background for the
ing prejudice, especially if the defendant is from the same group as
present investigation. In fact, much basic psychological theory on
the minority group jurors” (p. 42). In sum, in the legal context and
group decision making has been derived from empirical studies set
more generally, it is an oversimplification to conclude that the
in the jury context (e.g., Davis, Stasser, Spitzer, & Holt, 1976;
effects of racial composition on decision making can be wholly
Stasser, Kerr, & Bray, 1982; Tindale & Davis, 1983; Tindale,
attributed to differential information conveyed by White and Black
Davis, Vollrath, Nagao, & Hinsz, 1990). Most notably, social
group members. This assertion is theoretically and practically
decision scheme theory (Davis, 1973) quantified the ways in which
important, as well as consistent with seminal models of group
individuals’ preferences are combined and reconciled in the pro-
decision making and minority influence (Davis, 1973; Moscovici,
duction of a group decision. Subsequent extensions of this theory
1980). However, to date no empirical studies have examined it
have considered transitions from one preference configuration to
directly.
another, as well as changes in individuals’ subjective certainty
over time (Kerr, 1981; Stasser & Davis, 1981). Models such as
Race and Legal Judgments
these typically do not address the specific effects of racial com-
position, but they can be extrapolated to predict that to the extent
When psycholegal researchers have focused on race, they have
that a group’s racial composition affects the distribution of indi-
typically examined its influence on the judgments of individual
vidual members’ preferences, so, too, will composition affect the
jurors. A brief review of this literature provides a broader context
group’s ultimate decision.
for the present investigation of race and the decision making of
Indeed, a consistent outcome of these, as well as of less quan-
juries. Research examining the influence of a defendant’s race on
titative models of group decision making is that a group’s majority
individual juror judgments has produced inconsistent results that
tends to carry the day. Moreover, the larger this majority, the
are difficult to reconcile, in large part because these studies are
greater its impact on the group. The most frequent explanation for
idiosyncratic and typically not grounded in any particular theory
this “majority rules” tendency focuses on information exchange:
(for exceptions, see Bodenhausen & Lichtenstein, 1987; Witten-
Not only does a sizable majority translate into more votes for one
brink, Gist, & Hilton, 1997). Some studies have found that a
outcome alternative, but it also leads the group to devote more of
defendant’s race has no reliable effect (Mazzella & Feingold,
its discussion to information that favors this outcome. However, in
1994; McGuire & Bermant, 1977; Nickerson, Mayo, & Smith,
his initial theoretical formulation, Davis (1973) suggested that
1986); others have indicated that jurors are harsher in their judg-
additional explanations for the majority rules tendency are neces-
ments of out-group defendants (DeSantis & Kayson, 1997; Hymes,
sary as well, in that many group decisions—including those made
Leinart, Rowe, & Rogers, 1993; Klein & Creech, 1982), and still
by juries— exhibit “a considerably more complex social process at
others have suggested that jurors are biased in favor of out-group
work. The best-fitting model suggested a mixture of majority,
defendants (McGowen & King, 1982; Poulson, 1990).
conformity, and other effects to be involved” (p. 123). This prop-
In light of these inconsistencies, recent research has attempted to
osition is consistent with the conclusions of other psychologists
place the literature in a more theoretical framework. In a series of
who have also cited both informational and noninformational
studies, Sommers and Ellsworth (2000, 2001) found support for
explanations for the influence of a group’s composition (Levine &
the hypothesis that a defendant’s race is relatively unlikely to
Moreland, 1998; Moscovici, 1980; Moscovici & Lage, 1976).
influence White jurors when a trial’s content is blatantly racial,
Therefore, it would seem that despite its intuitive and theoretical
such as when the crime itself is racially charged or when attorneys
appeal, an exclusive focus on information exchange is not suffi-
inject race-related arguments into the proceedings. Drawing on
cient for explaining the influence of a group’s composition on its
theories that portray modern racism as aversive and subtle (e.g.,
decision making. The present research tests this assertion by iden-
Gaertner & Dovidio, 1986; Kinder & Sears, 1981), Sommers and
tifying multiple effects of racial diversity, thereby expanding our
Ellsworth suggested that race-relevant trial content makes salient
conceptual understanding of the ways in which group composition
many White jurors’ concerns about avoiding prejudice or the
influences decision making.
appearance thereof. However, absent racially charged trial content,

---

600
SOMMERS
when they are presumably less concerned about racism, White
of heterogeneity not subject to the idiosyncrasies of the special
jurors are harsher in their judgments of a Black than a White
case of a minority of one.
defendant (Sommers & Ellsworth, 2000, 2001).
Other studies have provided indirect support for this hypothesis.
Identifying Multiple Effects of Racial Diversity
Fein, Morgan, Norton, and Sommers (1997) found that an inflam-
matory editorial about a Black defendant led White mock jurors to
This investigation was expected to identify multiple simulta-
render harsher verdicts in a subsequent trial, but when it was
neous processes through which racial diversity affects group de-
insinuated that the newspaper’s depiction was racially motivated,
cision making. At the most basic demographic level, racial com-
its biasing effects were eliminated. Pfeifer and Ogloff (1991)
position was predicted to influence groups’ predeliberation vote
found that judicial instructions to reach a verdict “free from
distribution. Consistent with previous findings that Black jurors
sympathy or prejudice” had a similar effect, eliminating Whites’
are more lenient than White jurors toward Black defendants
tendency to judge a Black defendant more harshly than a White
(Skolnick & Shaw, 1997; Sommers & Ellsworth, 2000; Ug-
defendant. Indeed, recent research has begun to reconcile incon-
wuegbu, 1979), the presence of Blacks on a jury was expected to
sistent findings (see also Johnson, Whitestone, Jackson, & Gatto,
translate into fewer guilty votes before deliberations.
1995; Sweeney & Haney, 1992), converging on the hypothesis that
Of greater interest were the psychological processes through
activating White jurors’ concerns about prejudice attenuates the
which diversity exerts its influence. To the extent that researchers
influence of a defendant’s race on judgments. It is worth noting,
have considered the processes through which a group’s composi-
however, that the majority of mock juror investigations—like
tion affects its decision making, they have usually focused on the
examinations of group decision making more generally— continue
prediction that demographic diversity leads to informational diver-
to rely on predominantly White college student samples, and the
sity. That is, the traditional information exchange account for
handful of studies that includes between-races comparisons have
diversity effects is that heterogeneity enables a group to consider
yielded inconsistent findings (cf. Abwender & Hough, 2001; Foley
a wider range of perspectives and information (e.g., Hans &
& Chamblin, 1982; Skolnick & Shaw, 1997; Ugwuegbu, 1979).
Vidmar, 1982; Hoffman & Maier, 1961; Jehn et al., 1999). With
regard to racial diversity in particular, the basis for this prediction
is that Black group members will make different contributions to
The Present Research
a group than Whites (Peters v. Kiff, 1972). An example in the legal
context would be the expectation that racially diverse juries will
The chief objective of the present research was to utilize a mock
more frequently discuss institutional racism or racial profiling
jury paradigm to examine the processes through which racial
because Black jurors are more likely to have personal experiences
diversity influences group decision making. Participants were
with these issues. Despite the appeal of this information exchange
shown the trial of a Black defendant, and the decision making of
prediction, no direct experimental tests of it have been conducted
racially heterogeneous and homogeneous 6-person mock juries
with regard to racial diversity. In fact, many in the legal system
was compared.1 Observing groups of a wide variety of racial
regard hypotheses regarding jury composition as untestable
compositions would have been informative, but in the present
(Marder, 2002). In the present study, the informational content of
study heterogeneity was operationalized as 4 White and 2 Black
all-White and racially diverse jury deliberations was compared. It
jurors (with homogeneous groups consisting of 6 White jurors).
was hypothesized that heterogeneous groups would indeed ex-
These operationalizations were chosen for theoretical as well as
change a wider range of experiences, viewpoints, and interpreta-
practical reasons. Most participants were recruited from an actual
tions than would homogeneous groups.
jury pool in the middle of jury duty. A contingency of this
The present study also considers a less-traditional take on di-
arrangement with the court was that participant recruitment could
versity and information exchange, one that includes the possibility
not affect the demographics of the remaining jury pool. Therefore,
that White group members behave differently depending on a
the proportion of Black participants used in the study was as high
group’s racial composition. The proposition that the effects of
as was permissible. Operationalizing heterogeneity with only one
racial heterogeneity are, at least in part, attributable to the perfor-
Black juror would have been more practically convenient, yet
mance of Whites is consistent with previous findings that demon-
theoretically problematic. A minority of one has unique psycho-
strate the influence of race on motivation, judgment, and cognition.
logical properties when it comes to conformity and social influ-
Consider, for example, the conclusions of Sommers and Ellsworth
ence (Allen, 1975; Asch, 1956); a juror who perceives a lack of
(2000, 2001), which suggest that situational variables that activate
social support or potential allies may be particularly likely to
remain quiet and succumb to group pressure (Ellsworth & Mauro,
1998). Minorities of one are also less likely to exert the consistent
1 The decision to use 6- rather than 12-person juries was based largely on
pressure necessary to have an influence on a majority (Maass &
practical considerations. Most participants in the study were citizens ful-
Clark, 1984; Moscovici & Lage, 1976). Therefore, creating juries
filling their jury duty, and only a limited number of these individuals were
with 4 White and 2 Black jurors held constant the study’s defini-
available for research purposes. Using 6-person groups also facilitated
transcription and coding of the videotaped deliberations. Previous research
tion of heterogeneity and enabled a situation in which a group’s
has identified few process differences between 6- and 12-person juries
composition could conceivably be expected to affect its decision
(Kerr & MacCoun, 1985), and the reduced representativeness of smaller
making. That is, the diversity in this study was meaningful and not
juries was not a concern because group composition was controlled. The
“token,” as a two-person minority would be expected to be able to
present research therefore follows a long line of jury simulations that have
resist majority pressure and even exert its own influence in some
used 6-person mock juries (e.g., Davis et al., 1976; Kerr, 1981; Tindale et
situations. This was chosen as a relatively generalizable definition
al., 1990).

---

RACIAL DIVERSITY AND DECISION MAKING
601
Whites’ concerns about avoiding prejudice also affect their private
Situational Considerations
judgments about a trial, rendering them more lenient toward a
Black defendant. If, as Hans and Vidmar (1982) have suggested,
Previous investigations of race and legal decision making have
membership in a diverse group reminds Whites of their motiva-
illustrated the importance of situational variables. In particular, the
tion to avoid prejudice, then White participants in the present study
presence of blatantly racial issues at trial has been found to
may evaluate and weight the trial evidence differently depending
moderate the influence of race on jurors’ judgments (Sommers &
on the composition of their group. Accordingly, Whites’ informa-
Ellsworth, 2000, 2001). In terms of the present investigation, a
tional contributions to deliberations should vary by group
related question is whether the effects of racial diversity vary
composition.
depending on the racial content of a case. Accordingly, groups
Previous work on race and attitude change suggests another
were randomly assigned to receive either a race-relevant pretrial
related prediction, namely that membership in a diverse group
jury selection questionnaire—which included questions about in-
affects Whites’ information-processing style. Several studies
stitutional racism in the legal system and provided a racially
have found that Whites’ desire to guard against prejudice— or
charged context for the subsequent trial— or a race-neutral ques-
serve as “watchdogs” for bias—leads them to process informa-
tionnaire. This manipulation not only served a conceptual purpose
tion more systematically when it is about a Black target (Sar-
but also addressed the important practical consideration of how to
gent & Bradfield, 2004) or conveyed by a Black source (Petty,
address the previous finding that juror racial bias is most likely to
Fleming, & White, 1999; White & Harkins, 1994). It is not
occur in run-of-the-mill trials without blatantly racial issues. One
much of a leap to propose that a group’s racial composition can
possible remedy is to attempt to counteract potential racial bias
have a similar effect on its members (see Antonio et al., 2004).
through jury selection questioning, or voir dire.
In fact, anticipating a meaningful, potentially race-relevant
Though judges and legal professionals view voir dire as an
discussion with a diverse group is likely a very immediate and
opportunity to identify jurors who are likely to demonstrate bias
salient reminder of one’s motivation to avoid prejudice (see
(see Hans & Vidmar, 1982), prospective jurors often conceal
Vorauer, Hunter, Main, & Roy, 2000). As Sargent and Brad-
socially undesirable tendencies (Seltzer, Venuiti, & Lopes,
field (2004) concluded, “Whites’ attempts to act as watchdogs
1991) and erroneously report that they would not be influenced
may be activated by a variety of situations involving stigma-
by potential biases (Kerr, Kramer, Carroll, & Alfini, 1991;
tized group members” (p. 1003). In sum, though most psychol-
Nisbett & Wilson, 1977). But even if its potential to diagnose
ogists and legal scholars have assumed that the influence of
bias is limited, race-related voir dire may still be useful. The
racial diversity results from the contributions of minority group
experience of answering questions about racism or one’s own
members, the present study offers the novel hypothesis that
racial attitudes could serve as yet another way in which racial
Whites also bear responsibility for the informational effects of
issues and concerns about prejudice are made salient for jurors.
racial composition. Specifically, White participants were ex-
This would be consistent with the idea that voir dire is more
pected to contribute different and potentially more accurate
than just a method for identifying unsuitable jurors and is also
information to deliberations when in racially heterogeneous as
an opportunity to socialize citizens regarding their new role as
opposed to homogeneous groups.
officers of the court (Balch, Giffiths, Hall, & Winfree, 1976). As
Finally, yet another possibility is that some of the effects of
such, the present voir dire manipulation assessed the possibility that
racial diversity occur completely outside the scope of informa-
priming race-related concerns moderates the effects of racial diversity
tion exchange. This is also a hypothesis that has received scant
on group decision making and also tested the practical hypothesis that
attention from psychologists and the legal community. One
pretrial questions about racism color jurors’ subsequent trial judg-
ments. On the basis of previous research, the race-relevant voir dire
exception is Kerr et al.’s (1995) experiment that demonstrated
was expected to lead to more lenient judgments of the Black defen-
the influence of perceived group composition on the judgments
dant than the race-neutral version.
of its ostensible members. Antonio et al. (2004) reported a
comparable, albeit marginally significant finding that member-
ship in a group with a Black confederate led Whites to engage
Method
in more complex thinking about a discussion topic even before
that discussion began. Results such as these cannot be explained
Design
by information exchange, as they emerged in the absence of
discussion. In the present study, participants’ private trial judg-
This study was designed as an engaging and realistic courtroom simu-
ments were assessed before deliberations began (but after they
lation, in which a jury-eligible community sample watched a video trial
were aware of the composition of their group). To the extent
summary and then deliberated as 6-person mock juries. Both independent
that membership in a racially diverse group affects not only the
variables in the 2
2 design—jury racial composition and the nature of
information White participants convey during the discussion
voir dire questions—were manipulated at the group level. Racial compo-
sition was varied so that half of the juries were all-White and half were
but also how they think about and privately evaluate the case—
diverse; diverse juries deliberated as groups of 4 White and 2 Black mock
specifically, rendering them less punitive toward the defendant
jurors. Half of the juries received a race-neutral pretrial voir dire question-
and more thorough in their information processing—Whites
naire with items about general legal attitudes and experiences. The other
were expected to be less likely to vote to convict the Black
half received a race-relevant version of this questionnaire, including items
defendant when in diverse versus all-White groups, even before
assessing participants’ racial attitudes and their beliefs regarding the influ-
deliberations begin.
ence of race in the criminal justice system.

---

602
SOMMERS
Participants
impartiality. Participants were then given one of the two versions of the
voir dire questionnaire.
Several steps were taken to ensure that the ecological validity of this
The first items assessed demographic information. Subsequent questions
simulation exceeded that of most investigations of legal and group decision
asked about participants’ legal experiences, including whether they had
making, beginning with participant recruitment. A total of 200 participants
ever testified at trial, retained an attorney, or been the victim of a crime.
were recruited in one of two ways. The majority (121, or 60.5%) partici-
The race-neutral version finished with four open-ended questions about
pated during jury duty at a county courthouse in Washtenaw County,
perceptions of violent crimes. For example, one of these questions read,
Michigan. Additional participants were recruited through newspaper ad-
“Do you have any beliefs or experiences that might prevent you from
vertisements in the same area.2
judging a defendant fairly in a case of violent crime?” These were designed
Courthouse sessions were run with the cooperation of local judges and
to be comparable to the final items in the race-relevant condition with
jury-pool administrators. Each week during the course of the study, a panel
regard to question length and type of response required, though none of
of 25 prospective jurors was randomly selected by computer for possible
these items referred to race. The final four open-ended items of the
participation. This occurred at the beginning of jury duty, before individ-
race-relevant version were as follows:
uals experienced voir dire for any of the cases on that week’s docket. After
moving from the jury assembly room to an empty administrative office in
“This trial involves an African American defendant and White victims.
another area of the courthouse, the experimenter described to these indi-
How might this affect your reactions to the trial?,”
viduals their opportunity to earn $10 per hour in a court-sponsored project
on jury decision making. Prospective participants were told that the study
“Do you have any biases or prejudices that might prevent you from
entailed viewing the video summary of a sexual assault case and deliber-
judging an African American defendant fairly?,”
ating on it. Over 90% expressed interest in participating, and from this
group a smaller subset of participants was selected at random and assigned
“In your opinion, how does the race of a suspect affect the treatment s/he
to mock juries (subject to the constraint that half of the groups were racially
receives from police?,”
homogeneous and half heterogeneous). Remaining individuals were re-
turned to the jury assembly room.
“In your opinion, how does the race of a defendant influence the
Participants recruited through newspaper advertisements were given the
treatment s/he receives in the legal system as a whole?”
same information about the opportunity to earn $10 per hour as a volunteer
These questions were designed to force participants to think about their
for a court-sponsored project on jury decision making. From this pool of
own racial attitudes and the general role of race in the criminal justice
willing participants, mock juries were randomly created, subject to the
system.
constraints of racial composition and schedule availability. These experi-
On completing voir dire, participants were shown a 30-min Court TV
mental sessions were run in a university seminar room selected for its
(Courtroom Television Network, 1995) video summary of the trial of a
physical similarity to the room used at the courthouse.
Black defendant charged with sexual assault. The trial video included
Because any juror who decided to withdraw from participation in the
highlights from opening arguments, the testimony of seven prosecution and
middle of the session would have led to the loss of data for an entire jury,
three defense witnesses, and closing arguments. The prosecution presented
efforts were made to select an additional alternate juror whenever possible.
Therefore, for 26 of the 29 juries, an additional White participant was
testimony from the two victims, neither of whom could identify the
included in the group that watched the trial video (i.e., a 5th White juror in
assailant’s face, though one was able to describe a scar similar to one on
heterogeneous groups and a 7th White juror in the homogeneous groups).
the defendant’s torso. The core of the prosecution’s case was forensic
As no participants ever withdrew from the study, the alternate was always
analysis of crime scene semen and hair, which were consistent with the
dismissed before deliberations began. The assignment of a White juror to
defendant’s, but not definitive matches. The defense focused on the lack of
the position of alternate was done at random and was unbeknown to that
eyewitness evidence and the idiosyncratic methods used by the laboratory
participant or the rest of the group. Therefore, alternate jurors are included
that conducted the DNA analysis.
in predeliberation but not deliberation analyses.
At the end of the video, the experimenter read aloud jury instructions
All 200 participants were jury-eligible citizens. Comparable demograph-
from the State of Michigan (Michigan Criminal Jury Instructions, 1989,
ics were maintained across the courthouse and newspaper samples, with no
2001). These instructions included the legal elements of criminal sexual
significant differences in participant gender, age, or education level. Over-
conduct, general guidelines for deliberations, and a vague definition of
all, 121 participants were women (60.5%) and 79 participants were men
reasonable doubt. Participants then completed a two-item predeliberation
(39.5%); 170 participants were White (85%) and 30 participants were
questionnaire. First, they were asked to indicate their verdict preference of
Black (15%). Participant age ranged from 18 to 78 years, with a mean of
either guilty or not guilty. Second, they were asked to quantify the likeli-
39 years. The demographics of the nonalternate sample of 174 participants
hood that the defendant committed the crime using a scale of 0%–100%.
who deliberated on the trial were comparable, as 106 deliberating partic-
After collecting the predeliberation questionnaires, the experimenter
ipants were women (61%) and 68 (39%) were men, with an average age of
once again reminded participants of their objective to reach a unanimous
40 years. Removal of 26 White alternates from the sample brought the final
verdict. Other than the instruction to begin by electing a foreperson, no
racial demographics to 83% White and 17% Black.
specific deliberation procedures were suggested. In most experimental
sessions, an alternate White participant was then excused from the remain-
der of the study. The experimenter turned on a video camera mounted on
Procedure
a tripod, pointed it at the jury table, and left participants to deliberate in
private. On occasion, groups sent their foreperson out of the room with
Once a mock jury was created, participants were seated at a rectangular
questions. In these instances, the experimenter provided no additional
table and assigned juror identification numbers. Participants were able to
information about the case. If the foreperson announced that the group
see one another and therefore were aware of the composition of their group
could not reach a unanimous verdict, the experimenter reminded partici-
from the outset of the session. In initial instructions, the experimenter
pants of a jury’s duty to exhaust all avenues in pursuit of unanimity. If the
emphasized that participants should interact and make judgments as if they
were jurors judging an actual case. The experimenter outlined the process
of voir dire, explaining that before a trial prospective jurors are asked
2 As discussed in more detail in the Results section, none of the key
questions about their personal experiences and attitudes to determine their
findings were moderated by participant recruitment method.

---

RACIAL DIVERSITY AND DECISION MAKING
603
foreperson announced that the group had reached a verdict, the experi-
menter reentered the room and turned off the camera; otherwise, deliber-
ations were stopped after 60 min had elapsed.
Results
Sample Comparison
Potential differences between participants from the courthouse
and those recruited via newspaper were examined, and the validity
of treating all 29 mock juries as one sample was confirmed.
Location did not affect any of the predeliberation measures on its
own, nor did it moderate any of the findings reported below. The
only location effect to emerge was that courthouse groups men-
tioned more facts about the defendant’s scar during deliberations
(M
4.33, SD
0.69) than did groups recruited through news-
paper advertisement (M
3.73, SD
0.91), t(27)
2.05, p
.05. This difference has little implication for the present hypoth-
eses, and given that more than 30 analyses were used to test for
potential sample effects and interactions, it may very well reflect
a simple Type I error.
Predeliberation Judgments
Participants’ predeliberation ratings were collected privately
before they interacted. These data therefore satisfy the indepen-
Figure 1.
Predeliberation verdict preferences by group racial composi-
dence assumption of standard participant-level analysis. Dichoto-
tion, voir dire condition, and participant race.
mous predeliberation verdict preference was analyzed using logis-
tic regression. Across the entire sample, 41.0% of participants
voted guilty before deliberations began. Racial composition had a
age response on this measure was 63.9 (SD
26.8).3 The main
significant effect on verdict preference, as 30.7% of participants in
effect for racial composition was significant, as participants in
diverse groups voted guilty compared with 50.5% of participants
diverse groups gave lower estimates (M
58.5, SD
27.6) than
in all-White groups, Wald (1)
9.50, p
.01. The effect of voir
participants in all-White groups (M
69.3, SD
25.0), F(1,
dire on verdict preference was also consistent with predictions, as
196)
8.57, p
.01. Consistent with predictions, participants
only 34.4% of participants who received a race-relevant voir dire
who received a race-relevant voir dire gave lower estimates of
voted guilty compared with 47.1% of participants in the race-
guilt (M
60.0, SD
26.2) than did participants in the race-
neutral condition, Wald (1)
4.07, p
.04. No significant
neutral condition (M
67.9, SD
27.0), F(1, 196)
4.52, p
interaction was found between the independent variables of racial
.04. Once again, no significant interaction emerged, F(1, 196)
1.
composition and type of voir dire, Wald (1)
1.
Participant race was added to the analysis by replacing jury
Chi-square analyses were used for planned comparisons by
racial composition with a new variable with three levels: Blacks in
participant race. Consistent with previous research, Black partici-
diverse groups, Whites in diverse groups, and Whites in all-White
pants (23.3%) were less likely to vote guilty than White partici-
groups. Planned comparisons were conducted using contrasts with
pants (44.1%),
2(1, N
200)
4.50, p
.03. Breaking this
Welch’s separate variance t tests to account for the different size of
result down by jury racial composition, the 23.3% guilty vote rate
these three comparison groups; this resulted in noninteger degrees
among Blacks did not differ significantly from the 33.8% rate of
of freedom for most comparisons. Overall, Black participants gave
Whites in diverse groups,
2(1, N
101)
1.08, p
.30.
lower guilt estimates (M
53.9, SD
32.4) than did Whites (M
However, the difference between Black participants and Whites in
65.9, SD
25.3), though this difference only approached statis-
all-White groups (50.5% guilty votes) was significant, 2(1, N
tical significance, t(35.6)
1.81, p
.08. The difference between
129)
7.33, p
.01. Analysis also revealed that Whites in diverse
Black participants’ responses and those of Whites in diverse
groups were significantly less likely to vote guilty than Whites in
groups (M
61.2, SD
25.2) was not significant, t(44.5)
1.09,
all-White groups, 2(1, N
170)
5.23, p
.02. This result is
p
.28; the difference between the responses of Black participants
particularly noteworthy, as it demonstrates that racial composition
and those of Whites in all-White groups (M
69.3, SD
25.0)
affected White participants before any information exchange oc-
was significant, t(40.0)
2.39, p
.02. Consistent with verdict
curred. No interaction was found between voir dire and participant
race, Wald (1)
1 (Figure 1 presents descriptives for this variable
3 That fewer than half of participants voted to convict the defendant even
by racial composition, voir dire condition, and participant race).
though the average probability of commission estimate was greater than
Similar patterns were found for the second predeliberation mea-
60% suggests that participants took seriously the concept of reasonable
sure, ratings of the likelihood that the defendant committed the
doubt. Debriefing conversations after the study, as well as the earnestness
crime on a scale of 0%–100%. This continuous variable was
of group discussions during deliberations, further confirmed that partici-
analyzed using analysis of variance (ANOVA). Overall, the aver-
pants found the experiment to be a realistic and engaging jury simulation.

---

604
SOMMERS
preference data, Whites in diverse groups gave significantly lower
controlling for this variable, which, as expected, emerged as a
guilt estimates than did Whites in all-White groups, t(150.4)
negative predictor of deliberation length (
–.32). Consistent
2.09, p
.04. That Whites in diverse groups were once again less
with information exchange predictions, this new analysis indicated
convinced of the Black defendant’s guilt than Whites in all-White
a significant difference in deliberation length by racial composi-
groups supports the conclusion that the influence of racial com-
tion, F(1, 24)
5.37, p
.03. Closer examination of this effect is
position is not limited to processes of information exchange. Type
detailed below. No difference was found between the deliberation
of voir dire did not interact with participant race to affect guilt
length of groups in the race-relevant voir dire condition (untrans-
estimates, F(2, 194)
1.
formed M
43.5 min, SD
19.0) and the race-neutral condition
(M
45.7 min, SD
19.0), nor did voir dire condition interact
Deliberation Outcomes
with racial composition to affect deliberation length, both Fs(1,
Groups began deliberations by electing a foreperson. This pro-
25)
1.
cess varied widely: Strategies included looking for someone with
jury experience, explicitly choosing the only man in the group,
Deliberation Content
soliciting volunteers, and turning to the person who spoke first.
Out of the 15 diverse groups, 5 selected a Black foreperson. This
The processes through which the present manipulations affected
33% rate was identical to the proportion of Black participants in
group decision making were assessed by content coding of the
these groups. The selection process was not as balanced regarding
videotaped deliberations. Three coders simultaneously coded five
gender. Even though 61% of the sample was female, the majority
juries, and these overlapping judgments were used to ensure and
of forepersons (52%) were male, a difference that was consistent
quantify reliability. Each coder then scored eight juries individu-
with previous findings (Ellsworth, 1989), though not statistically
ally. Coders were blind to voir dire condition, though racial com-
significant, 2(1, N
29)
1.97, p
.22.
Given that only 40% of predeliberation votes across the
position was apparent as they watched each tape. Coders were,
sample were for guilty, it was not surprising that only one
however, naive to the present hypotheses regarding racial compo-
group—an all-White jury in the race-relevant voir dire condi-
sition. Pairwise kappas for the three coders were computed across
tion— ultimately reached a unanimous guilty verdict. This lack
all variables reported below. To provide a more rigorous test of
of variability in verdicts prevented meaningful statistical com-
reliability, these kappas were based on the coding of each partic-
parisons by condition. Out of the 29 groups, 16 (55%) reached
ipant, not group-level coding. Pairwise values ranged from .79 to
a unanimous not guilty verdict and 12 (41%) were not unani-
.90, surpassing the conventional cutoff for satisfactory reliability
mous after 1 hr (see Table 1 for breakdown of group outcomes
of .70 (Stangor, 1998).
by condition).
For deliberation content examined at the group level (e.g.,
Analysis of deliberation length was more informative. Because
number of case facts discussed by each group), standard two-way
the distribution was negatively skewed, data were squared to
ANOVA were used. For participant-level analyses (e.g., novel
improve normality and equality of variance. Deliberations of di-
case facts raised by participant race), participants were treated as
verse groups (untransformed M
50.7 min, SD
16.8) were
replicates nested within juries, and planned comparisons were
longer than those of all-White groups (M
38.5 min, SD
19.1),
conducted using contrast weights. In other words, for each depen-
though this difference only approached statistical significance,
dent measure two scores were computed for each diverse group:
F(1, 25)
3.67, p
.07. However, there was greater variance in
one for the average response of the 4 White participants and one
predeliberation verdict preferences in all-White groups, as these
for the average response of the 2 Black participants. One average
juries were more evenly divided between guilty and not guilty
score for each dependent measure was also computed for the 6
verdicts than were diverse groups. Ordinarily, this sharper division
Whites in each all-White group. Comparisons between the average
of preferences would predict longer deliberations and more diffi-
scores of Whites and Blacks in diverse groups were performed
culty reaching unanimity for all-White groups (see Stasser, Kerr,
using paired-samples t tests, whereas comparisons between scores
& Davis, 1989). Therefore, each group’s distance from a 50/50
of Whites or Blacks in diverse groups with scores of Whites in
predeliberation vote split was calculated by converting vote split to
a proportion and taking the positive difference between this pro-
all-White groups were calculated using Welch’s separate variance
portion and .50. The analysis of deliberation length was then run
independent-sample t tests. For both group- and participant-level
analyses, this yielded a maximum sample size of 29, the total
number of groups in the study.
Breadth of factual content.
No reliable main effects or inter-
Table 1
actions were found for the influence of voir dire on any measure of
Group Outcomes by Condition
deliberation content. However, racial composition had multiple
effects on deliberations (see Table 2 for a summary of group-level
Race-relevant voir dire
Race-neutral voir dire
analyses; see Table 3 for participant-level analyses). First, the
Diverse
All-White
Diverse
All-White
breadth of information discussed by each group was examined to
Outcome
group
group
group
group
determine whether longer deliberations indicated more substantive
Acquittals
4
5
4
4
discussions. Coders used the trial video to create a checklist of 46
Hung juries
3
1
4
3
major case facts. These facts fell into the more specific categories
Convictions
0
1
0
0
of assault details, assailant’s scar, scientific evidence, and legal

---

RACIAL DIVERSITY AND DECISION MAKING
605
Table 2
significant difference emerged in spite of the fact that all-White
Group-Level Analyses of Deliberation Content
jury deliberations were shorter and included fewer factual state-
ments to begin with.
Diverse
All-White
Nonfactual content.
The analyses above examine discussion
Measure
group
group
of factual information. However, information exchange predic-
Deliberation length, in min
50.67
38.49
tions often do not focus on facts but rather on the expectation that
a
b
No. of case facts discussed
30.48
25.93
a
b
diverse groups will be exposed to more opinions and perspectives
No. of factual inaccuracies
4.14
7.28
a
b
than homogeneous groups. These nonfactual aspects of delibera-
No. of uncorrected inaccurate statements
1.36
2.49
a
b
tions were examined in several ways. First, coders noted partici-
Amount of “missing” evidence cited
1.87
1.07
No. of race-related issues raised
3.79
2.07
pants’ statements about evidence they wished had been presented
a
b
No. of mentions of racism
1.35
0.93
at trial. For example, several participants cited the lack of finger-
% of time mention of racism met with
print evidence as important; others wondered why a child witness
objection
22%
100%
a
b
to the assault did not testify. Analysis indicated that diverse groups
Note.
Values with different subscript letters differ significantly at p
discussed more examples of “missing” evidence (M
1.87, SD
.05; n
15 diverse groups and 14 all-White groups.
1.46) than did all-White groups (M
1.07, SD
0.47), though
this effect only approached significance, F(1, 25)
3.78, p
.06.
Once again, at the participant level, this difference was driven by
issues.4 Consistent with information exchange predictions, diverse
the performance of Whites, who brought up more items of “miss-
groups discussed more case facts (M
30.5, SD
4.87) than
ing” evidence when in diverse groups (M
0.37, SD
0.35) than
all-White groups (M
25.9, SD
6.08), F(1, 25)
4.68, p
.04.
in all-White groups (M
0.18, SD
0.08), t(20.5)
2.02, p
Coders recorded who was the first participant in the group to
.05. A more intuitive way to interpret these fractional means is that
raise each of these facts, and these participant-level data were
1 of every 2.7 White participants in diverse groups cited an item of
analyzed as detailed above. Analyses indicated that the effect of
“missing” evidence, a rate twice as frequent as the 1 of 5.6 rate
racial composition on deliberation breadth was not attributable to
among all-White groups. The average for Black participants (M
the performance of Black participants, who raised only slightly
0.21, SD
0.25; 1 of every 4.8 participants) did not differ
more novel facts on average (M
4.70, SD
1.93) than did
significantly from either group of White participants.
Whites in all-White groups (M
4.32, SD
1.01), t(21.5)
1.
Mention of race.
Novel statements related to race were also
Rather, the main effect for racial composition captured differences
coded. Topics included the relationship between race and eyewit-
in White participants’ performance across condition. Even without
ness accuracy, genetic bases of race, and the role of race in police
controlling for predeliberation verdict preference distribution, the
investigations. The use of race as a simple descriptor (e.g., “the
average of 4.32 (SD
1.01) case facts raised by participants in
African American detective said that”) was not included in this
all-White groups was significantly less than the average of 5.27
(SD
1.29) raised by Whites in diverse groups, t(26.3)
2.21,
analysis, nor were responses to other participants’ race-related
p
.04.
comments. Diverse groups were found to discuss more race-related
Accuracy of factual content.
Breadth of information exchange
topics (M
3.79, SD
1.93) than all-White groups (M
2.07,
in a group’s decision making is important, but so too is the
SD
1.86), F(1, 25)
5.52, p
.03. This effect was largely
accuracy of that information. Accordingly, deliberations were
attributable to Black participants, who raised more novel race-
coded for inaccurate statements. Coding was confined to inaccu-
related issues (M
0.80, SD
0.68) than Whites in diverse
racies related to the case; unrelated factual errors (e.g., “the pop-
groups (M
0.55, SD
0.41) or all-White groups (M
0.35,
ulation of the U.S. is 600 million”) were not included. When
SD
0.31), though only the comparison with the latter group was
multiple participants were inaccurate regarding the same fact, all
significant, t(19.9)
2.35, p
.03.
of these errors were counted.5 However, the same inaccuracy
Specific mention of racism—related to this case in particular or
voiced repeatedly by the same participant was only counted once.
at a societal level—was less frequent. At a group level, diverse
Analysis indicated that the deliberations of diverse groups in-
groups mentioned racism on more occasions (M
1.35, SD
cluded fewer inaccurate statements (M
4.14, SD
2.71) than
1.59) than did all-White groups (M
0.93, SD
1.64), but the
those of all-White groups (M
7.28, SD
2.64), F(1, 24)
9.66,
effect was not signi