ArticlePDF Available

Surface-Level Diversity and Decision-Making in Groups: When Does Deep-Level Similarity Help?


Abstract and Figures

We examined how surface-level diversity (based on race) and deep-level similarities influenced three-person decision-making groups on a hidden-profile task. Surface-level homogeneous groups perceived their information to be less unique and spent less time on the task than surface-level diverse groups. When the groups were given the opportunity to learn about their deep-level similarities prior to the task, group members felt more similar to one another and reported greater perceived attraction, but this was more true for surface-level homogeneous than surface-level diverse groups. Surface-level homogeneous groups performed slightly better after discovering deep-level similarities, but discovering deep-level similarities was not helpful for surface-level diverse groups, who otherwise outperformed surface-level homogeneous groups. We discuss the implications of this research for managing diversity in the workplace.
Content may be subject to copyright.
Group Processes &
Intergroup Relations
2006 Vol 9(4) 467–482
Surface-Level Diversity
and Decision-Making in
Groups: When Does
Deep-Level Similarity
Katherine W. Phillips
Northwestern University
Gregory B. Northcraft
University of Illinois, Urbana-Champaign
Margaret A. Neale
Stanford University
We examined how surface-level diversity (based on race) and deep-level similarities influenced
three-person decision-making groups on a hidden-profile task. Surface-level homogeneous
groups perceived their information to be less unique and spent less time on the task than
surface-level diverse groups. When the groups were given the opportunity to learn about their
deep-level similarities prior to the task, group members felt more similar to one another and
reported greater perceived attraction, but this was more true for surface-level homogeneous
than surface-level diverse groups. Surface-level homogeneous groups performed slightly better
after discovering deep-level similarities, but discovering deep-level similarities was not helpful
for surface-level diverse groups, who otherwise outperformed surface-level homogeneous
groups. We discuss the implications of this research for managing diversity in the workplace.
keywords diversity, information sharing task, similarity-attraction, social
PRIMARY reason organizations use groups is
to garner the benefits of the unique knowledge
and information that group members might
bring to the table (e.g. Schneider & Northcraft,
1999). For nearly twenty years the sharing and
integration of unique information in small
group discussions has been the subject of much
experimental (e.g. Stasser & Stewart, 1992;
Stasser & Titus, 1985, 1987; Stasser, Vaughan, &
Stewart, 2000; Stewart & Stasser, 1995; Winquist
& Larson, 1998; Wittenbaum, 2000) and some
Copyright © 2006 SAGE Publications
(London, Thousand Oaks, CA and New Delhi)
9:4; 467–482; DOI: 10.1177/1368430206067557
Author’s note
Address correspondence to Katherine W.
Phillips, Kellogg School of Management,
Northwestern University, 2001 Sheridan Road,
Evanston, IL 60208-2001, USA
peer-00571629, version 1 - 1 Mar 2011
Author manuscript, published in "Group Processes & Intergroup Relations 9, 4 (2006) 467-482"
DOI : 10.1177/1368430206067557
field research (e.g. Larson, Christensen,
Abbott, & Franz, 1996, 1998). Reflecting the
reality that all individuals bring a unique con-
stellation of perspectives and experiences to
small group discussions, this research examines
hidden profile decision situations in which sub-
optimal decisions are likely to be made if unique
information is not shared and integrated into
the group discussion (Stasser & Titus, 1985;
for reviews see Wittenbaum, Hollingshead, &
Botero, 2004; Wittenbaum & Stasser, 1996). In
such situations, organizations and teams that
can create environments where members are
willing to share and discuss unique information
may gain considerable competitive advantage.
In this article, we seek to understand more
about how surface-level (i.e. race/ethnicity)
and deep-level (i.e. experiences, preferences,
and values) diversity affects the ability of groups
to benefit from their unique information. We
move beyond the typical social categorization
perspective on diversity and highlight a by-
product of the social categorization process—
assumptions of in-group similarity—which has
been overlooked by many researchers in this
tradition (cf. Williams & O’Reilly, 1998). We
extend the argument that surface-level diversity
triggers expectations that informational differ-
ences may be present in groups, and legitimates
the expression of unique information (Phillips,
2003; Phillips & Loyd, 2006; Van Knippenberg,
De Dreu, & Homan, 2004; Van Knippenberg &
Haslam, 2003). Moreover, by highlighting
deep-level similarities in an effort to increase
levels of attraction and diminish social
categorization effects, we argue that managers
may undermine the benefits of having surface-
level diversity present in groups that must share
unique information for effective performance.
We provide some empirical evidence, while
integrating research on collective information
sampling in groups with that on the effects of
group diversity.
Collective information sampling in
Research on information sharing in groups has
found that sharing and integrating unique (i.e.
known by a single member) as opposed to
commonly held (i.e. known to all members)
information into group decisions is easier said
than done (for reviews see Stasser, 1999; Witten-
baum & Stasser, 1996; Wittenbaum et al., 2004).
One reason why unique information is men-
tioned and repeated less than commonly held
information is because group members gener-
ally assume that the information they possess is
the same as that possessed by others (unless
contrary information is available) (Stasser,
Stewart, & Wittenbaum, 1995). The assumption
is that there is no unique information, and that
unmentioned information is information that
other group members have deemed not of
sufficient importance to discuss. When unique
information does arise in groups, individuals
are likely to assume that because the infor-
mation is not widely held among the group
members, it is less important than commonly
held information, and therefore may fail to
repeat that unique information during dis-
cussion. Moreover, people may feel uncomfort-
able expressing and focusing on unique
information, because it is often inconsistent
with their perceived expectations that their
information should be similar to that of other
group members (Gruenfeld, Mannix, Williams,
& Neale, 1996; Phillips, Mannix, Neale, &
Gruenfeld, 2004). Sharing unique information
also leads to a lack of social validation from
others, causing individuals to feel less accepted
than when they share commonly held infor-
mation (Wittenbaum & Bowman, 2004).
Despite the information sharing barriers in
groups, the discussion of unique information
has been shown to increase when group
members have greater reason to believe that
unique information is going to be present. For
instance, when expertise is labeled, or if people
are explicitly forewarned that unique infor-
mation is present, groups are better able to
share and integrate the unique information
into the discussion (Franz & Larson, 2002;
Schittekatte & Van Hiel, 1996; Stasser et al.,
1995; Stewart & Stasser, 1995). To this end,
Postmes, Spears and Cihangir (2001) have
shown that unique information is more likely to
be shared when groups have developed a norm
Group Processes & Intergroup Relations 9(4)
peer-00571629, version 1 - 1 Mar 2011
of criticality instead of consensus. In such
critical norm groups, the sharing of unique
information is consistent with the group norms
and the same overvaluing of shared infor-
mation is less likely to occur. Thus, for groups
to use their unique information effectively, the
sharing of such information has to be perceived
as a legitimate part of the groups’ norms and
identity ( Jetten, Postmes, & McAuliffe, 2002;
Postmes, Haslam, & Swaab, 2005; Van Knippen-
berg et al., 2004).
We posit that surface-level diversity may also
serve this legitimation purpose in groups by
making it more acceptable to discuss and value
unique information that must be shared for
effective performance. Recent research has dis-
tinguished between diversity in surface-level
characteristics, which are immediately salient in
groups (like race and gender), versus diversity
in deep-level characteristics (like attitudes,
opinions, information, and values), which
become known only over time through verbal
and nonverbal communication (Harrison,
Price, & Bell, 1998; Harrison, Price, Gavin, &
Florey, 2002; Jackson, May, & Whitney, 1995).
We expect surface-level differences to serve as a
signal to group members that unique infor-
mation may be present, leading them to be
more aware of and willing to share unique
information with the group. This argument is
consistent with recent research on composition
beliefs, which has shown that individuals
believe diverse groups are likely to outperform
homogeneous ones when unique ideas are
needed, whereas homogeneous groups are
likely to outperform diverse ones when com-
monality of ideas is needed (van Knippenberg
& Haslam, 2003). Also see the work on the
mechanical and organic solidarity discussed by
Postmes et al. (2005) supporting the notion
that multiple sources of identity simultaneously
exist in groups.
Our argument is based on the fact that a
primary consequence of categorization pro-
cesses is that people assume that they hold
more similar deep-level perspectives with indi-
viduals who share their surface-level character-
istics than with people who do not, on topics
both relevant and irrelevant to the salient
surface-level distinction (e.g. Allen & Wilder,
1975, 1979; Chen & Kenrick, 2002; Diehl, 1988;
Holtz & Miller, 1985; Phillips, 2003; Phillips &
Loyd, 2006; Tajfel, 1969; Wilder, 1984). For
instance, Allen and Wilder (1979) divided
students into two groups, allegedly on the basis
of their preferences for oil paintings, and found
greater assumed deep-level similarity between
self and similar others than between self and dis-
similar others on opinions about art (category
relevant) and politics (category irrelevant).
Recent research by Phillips and Loyd (2006)
found this same pattern of assumed deep-level
similarity in decision-making settings when
examining the relationship between salient
task-relevant (e.g. functional background) (also
see Phillips, 2003) and irrelevant (e.g. campus
geography) social categories and task opinions.
Especially at the beginning of a group’s exist-
ence, when surface-level characteristics are
most salient (Harrison et al., 1998, 2002), indi-
viduals are likely to use the presence of these
characteristics to predict who shares deep-level
perspectives with whom. In surface-level homo-
geneous groups, group members are likely to
assume that they all possess the same infor-
mation about the task, whereas in surface-level
diverse groups members are likely to expect
there to be differences in information (Antonio
et al., 2004; Phillips, 2003; Phillips & Loyd,
2006; Phillips et al., 2004). Thus, surface-level
diversity triggers expectations that deep-level
diversity will be present in groups, and serves to
legitimize the surfacing of unique information.
Significantly, this legitimation of unique infor-
mation may apply not only to those who are
(surface-level) ‘different’ in the group, but also
to group members who are similar to most
others. Phillips and Loyd (2006) found that dis-
senting members of the social majority voiced
themselves more persistently and confidently
when there was diversity present than when
there was not. They concluded that the mere
presence of diversity encouraged those dissent-
ing group members to voice their disparate
perspective when they might otherwise have
remained silent and conformed to the opinion
of their in-group (e.g. Abrams, Wetherell,
Cochrane, Hogg, & Turner, 1990; Asch, 1952).
Phillips et al. diversity and performance
peer-00571629, version 1 - 1 Mar 2011
For these reasons, surface-level homoge-
neous groups should be less aware of the
unique information they possess. As such, over
the course of the group discussion they will be
less likely to discuss unique information about
the task than will surface-level diverse groups.
This will result in the surface-level homogenous
groups spending less time discussing the task
than their diverse counterparts. Moreover,
spending less time discussing the task will
further hinder the discovery of unique infor-
mation (Larson et al., 1996, 1998) leading to a
confirmation of the group members’ expec-
tations that they all have the same information.
In contrast, in groups where members possess
unique information, surface-level diverse
groups are more likely to discover and discuss
unique information than surface-level hom-
ogeneous groups. Surface-level diverse groups
assume that unique information is more likely
to be there, and the presence of informational
differences will be more consistent with their
expectations. Likewise, if group members are
aware that they might possess unique infor-
mation, they should be inclined to spend more
time discussing the task in an effort to discover
and integrate that information. Thus we
hypothesize that,
Hypothesis 1: Surface-level homogeneous groups
will be less aware of their unique information, and
will spend less time discussing the task than will
surface-level diverse groups.
Highlighting deep-level similarities
Although surface-level diversity may be ben-
eficial to teams or work groups that must share
unique information for effective performance,
diversity researchers have often found that
diversity has a negative impact on communi-
cation and cohesion, and promotes high levels
of detrimental group conflict (Ely & Thomas,
2001; Jackson, Joshi & Erhardt, 2003; Jehn,
Northcraft, & Neale, 1999; Pelled, Eisenhardt,
& Xin, 1999; for extensive reviews see also
Milliken & Martins, 1996; Williams & O’Reilly,
1998). Over the past ten years, diversity
researchers have focused on how to minimize
the detrimental effects of social categorization
on workgroups, with some suggesting that
increasing the level of perceived deep-level
similarity among group members should help
them feel more socially validated and accepted
by the other members of the group (e.g.
Gaertner et al., 2000). Some social categoriz-
ation researchers have advocated this perspec-
tive, suggesting that interventions designed to
minimize the salience of social categories and
instead ‘de-categorize’ or ‘re-categorize’ group
members by highlighting the similarities that
exist across seemingly different individuals can
be beneficial to group functioning (e.g.
Gaertner et al., 2000; Northcraft & Martin,
1982). For instance, Northcraft and Martin
(1982) argued that, ‘. . . the liking, acceptance,
and perceived competence of tokens and solos
can be enhanced by making salient their
similarities to majority group members in
background, attitudes, and interests’ (p. 114).
Further, in a study of corporate outside directors
of Fortune/Forbes 500 companies, Westphal
and Milton (2000) found that minority board
members (categorized on the basis of their
functional background, industry background,
education, race, or gender) were more influen-
tial on their focal boards when they had direct
or indirect social ties, often through their
common experiences with (focal board)
majority members on other corporate boards.
This perspective is built on the well-
established body of findings that similarity
attracts (Byrne, 1971). Individuals generally are
more attracted to and feel more comfortable
interacting with others whom they perceive to
be similar. For both surface-level homogeneous
and surface-level diverse groups, an interven-
tion designed to help group members discover
their deep-level similarities should lead to
greater feelings of attraction. Learning that one
shares deep-level similarities with a fellow group
member should also promote recategorization,
increasing the likelihood that out-group
members (i.e. those who have surface-level dis-
similarities from other group members) will
actually be seen as part of the in-group (e.g.
Gaertner, Mann, Murrell, & Dovidio, 1989;
Kramer & Brewer, 1984). The potential
negative effects of social categorization may,
Group Processes & Intergroup Relations 9(4)
peer-00571629, version 1 - 1 Mar 2011
subsequently, be reduced. As a result, we suggest
Hypothesis 2A: Members of groups who learn
about deep-level similarities should perceive greater
attraction from other group members than those
who do not learn about deep-level similarities.
However, recent research has suggested that
increasing this perceived similarity and attrac-
tion among the group members may come at a
cost to the group’s ability to benefit from the
surface-level differences we have discussed here
(Hornsey & Jetten, 2004; Jetten et al., 2002;
Postmes et al., 2001; van Knippenberg et al.,
2004). For instance, Postmes et al. (2002)
found that a focus on agreement and common-
alities created norms of consensus that in turn
undermined the sharing of unique information
in groups. The effectiveness of recategorization
or promoting the perception of others as
similar to oneself as a means to diminish the
detrimental effects of surface-level diversity has
also been called into question by researchers of
self-verification (e.g. Polzer, Milton, & Swann,
2002; Swann, Milton, & Polzer, 2000). These
researchers argue that promoting the percep-
tion (or recategorization) of all group members
as similar, may also discourage individuals from
thinking and acting in ways associated with their
unique category memberships (Gaertner et al.,
1989). Yet, it is precisely these unique ways of
thinking and acting that constitute the potential
positive contribution of a diverse workgroup
(Polzer et al., 2002, p. 297). Thus, highlighting
deep-level similarities (especially in surface-level
diverse groups), while leading to greater feelings
of attraction toward the other group members,
may undermine the groups’ awareness of and
willingness to embrace unique information.
In sum, the process of highlighting deep-
level similarities may undermine the signaling
effect of surface-level diversity that legitimates
expressing and discussing unique information
by all group members. In other words, learning
about deep-level similarities in surface-level
diverse groups may interfere with the legitima-
tion of difference and disagreement that
surface-level diversity promotes. For example, if
a work group who thinks they are diverse based
on the surface-level characteristic of race finds
that they all share the same attitudes, feelings,
and experiences about the organization, they
may be reluctant to disagree with each other
going forward with the task. They may feel that
they really are not all that different from each
other after all. Thus, for surface-level diverse
groups, although the realization of deep-level
similarities may increase attraction toward the
group, it may simultaneously increase pressures
to conform to the group and undermine the
discussion of unique information (Abrams
et al., 1990).
For surface-level homogeneous groups high-
lighting similarities will also increase attraction,
and may further interfere with the sharing of
unique information since doing so poses a
threat to feelings of acceptance and validation
(Wittenbaum & Bowman, 2004; Wittenbaum,
Hubbell, & Zuckerman, 1999). Thus, we believe
that for surface-level homogeneous groups
there will be somewhat of a ‘floor’ effect—the
lack of surface-level differences will hinder the
expectation of informational differences and
the discussion of unique information, and then
the highlighting of deep-level similarities will
further hinder this process. Thus, we hypothe-
size that:
Hypothesis 2B: Highlighting deep-level similarities
will lead to less awareness of unique information
and less discussion time.
Hypothesis 2C: The effect of highlighting deep-level
similarities on awareness of unique information
and discussion time will be more pronounced for
surface-level diverse groups than for surface-level
homogeneous groups.
In terms of group performance, highlighting
deep-level similarities should be detrimental
because it undermines the legitimacy of dis-
cussing needed unique information. Although
attraction may increase in groups as a result of
learning about deep-level similarities, these
deep-level similarities will be inconsistent with
any expectation of unique information being
present and thus are likely to hurt group per-
formance, especially for surface-level diverse
groups (e.g. Watson, Kumar, & Michaelsen,
1993). As such, we would argue that:
Phillips et al. diversity and performance
peer-00571629, version 1 - 1 Mar 2011
Hypothesis 3A: Surface-level diverse groups will
outperform surface-level homogeneous groups.
Hypothesis 3B: Groups that highlight deep-level
similarities will perform worse than groups that do
Hypothesis 3C: The performance of surface-level
diverse groups that highlight deep-level similarities
will be more negatively affected than the perform-
ance of surface-level homogeneous groups that
highlight deep-level similarities.
To test the hypotheses, groups received an
intervention asking them to discover deep-level
similarities among group members. Exploring
the effects of this intervention should allow for
a better understanding of how different groups
(i.e. those that are surface-level diverse or homo-
geneous) are affected by finding that they have
deep-level similarities prior to engaging in a task
where unique information must be shared.
Participants and overview
A total of 216 undergraduate business students
at a midwestern business school participated in
this research. The median age of the partici-
pants was 21 years, and approximately 42% of
the sample was female. The students partici-
pated in a class exercise designed to provide
insight into group decision-making. Partici-
pants were given extra course credit for their
participation, and a few of the best performing
groups on the prediscussion task were randomly
selected and entered into a raffle to receive cer-
tificates for free meals at local restaurants.
Participants first made individual assessments
about who they believed committed a murder
(see Stasser & Stewart, 1992 for materials)
based on the investigating detective’s reports,
and then discussed the case in three-person
groups. Groups were either all male (42) or all
female (30); the gender composition of the
group is included as a covariate in all analyses
presented. Eighty-two percent of the partici-
pants were white, 14% Asian, and the remain-
der African American or Hispanic. We used a 2
(surface-level homogeneous vs. surface-level
diverse) 2 (deep-level similarities highlighted
vs. control) between-subjects design and all
analyses were conducted at the group level.
There were a total of 31 surface-level homoge-
neous (i.e. three Caucasian group members)
and 41 surface-level diverse groups (i.e. two
Caucasian and one Asian, African American, or
Hispanic). Forty groups were in the deep-level
similarities condition and 32 were in the
control condition. Participants were thoroughly
debriefed about the purpose of the study after
Every participant was given a packet of evidence
from an apparent homicide investigation. The
evidence consisted of interviews and a variety of
supporting materials, including a list of
suspects, a map, a personal note, and a newspa-
per article. All of these materials were adapted
from Stasser and Stewart’s (1992) study. Within
each group, every member received the list of
suspects, transcripts of initial interviews with
each of the four key suspects, the newspaper
article, and maps of the crime scene and sur-
rounding area. The materials contained 42
clues in all, 12 of which were critical for solving
the case. All participants received 30 commonly
shared clues; 12 critical clues were distributed
among the three group members such that
each group member held some unique infor-
mation pertinent to identifying the guilty
suspect. These clues were embedded in follow-
up interviews with the key suspects, and inter-
views with some additional witnesses. In all of
the groups, a hidden profile existed because
the best solution was more likely to be found if
the unique information represented by the 12
unshared critical clues was shared.
When participants arrived at the laboratory or
in the classroom, they were randomly assigned
to three-person groups based on their visible
racial characteristics, with the constraint that all
three members had to be of the same gender.
In some groups the three group members
appeared to all be Caucasian (surface-level
homogeneous groups), and in other groups
two of the members appeared Caucasian and
Group Processes & Intergroup Relations 9(4)
peer-00571629, version 1 - 1 Mar 2011
one was of a different race (Asian, African
American, or Hispanic). At the end of partici-
pation, group members provided their self-
identified race on the post-task questionnaire,
and the experimenters’ group assignments were
validated against the information provided by
the students.
During the study, participants were first given
approximately 20 minutes to read and review
materials in order to determine which of four
suspects committed a murder (materials drawn
from Stasser & Stewart, 1992 and also used by
Gruenfeld et al., 1996; Liljenquist, Galinsky, &
Kray, 2004; Phillips et al., 2004). Individuals
were instructed to take notes on the case while
reading the materials, because they would not
be allowed to keep the case materials during
the group discussion portion of the exercise.
Participants made an individual assessment of
who they believed was more likely to have com-
mitted the murder. They were asked to indicate
how confident they were that each of the four
suspects did or did not commit the murder.
Subjects were also asked to provide a brief,
written justification for their decision.
Participants were then gathered into their
assigned three-person groups, and were
instructed that they had 5 minutes to complete
a short exercise before beginning group dis-
cussion. Groups were separated so that they
could not overhear the deliberations of other
groups. Half of the groups were randomly
assigned to identify their similarities with the
following instructions:
Working together with the other two members of
your group, you have 5 minutes to discover as many
things as possible that the three of you have in
common. You may include anything that you have
in common: friends, experiences, hobbies, books or
movies that you liked, places to which you have all
traveled, places where you all have lived or visited—
anything that the three of you have in common. Put
as many items on your list as possible.
The control groups were given the following
Working by yourself, you have 5 minutes to list as
many US states and their capitals as you can
remember. Put as many states and their capitals on
your list as possible.
Control group members did not work together
on the state capitals task in order to prevent the
inadvertent sharing of similarity information
while working together on the task. (A follow-
up data collection revealed that working
together to identify similarities increased group
members’ perceptions of similarity while
working together on the state capitals task did
In both conditions, participants were
informed that the group during that session
that generated the longest list (of similarities or
state capitals) could receive a prize (certificates
for free meals at a local restaurant).
After completing the 5-minute task, group
members individually completed a 20-item
survey titled ‘Personal Assessment Inventory’,
by circling the number that best indicated how
much they agreed with each of the statements
on a scale from 1 = strongly disagree to 7 =
strongly agree. 54 groups completed the survey.
Embedded in the items were questions assess-
ing how similar group members felt to one
another and how attracted they thought the
group would be to them (adapted from Eagly,
Once the short exercise and questionnaire
were completed, participants were instructed to
come to a group decision regarding which
suspect they believed was most likely to have
committed the murder. They were informed
that they would have to report their group
decision in exactly the same way that they had
reported their individual decisions. Groups
were given up to 35 minutes to complete their
discussions, which were timed by the experi-
menter. Finally, upon completion of the group
discussion, participants filled out an individual
post-task questionnaire where they recorded
the group decision in terms of how confident
they were that each of the four possible suspects
committed the murder. All participants then
answered several questions assessing their
group’s task performance and management of
information, as well as reporting their individ-
ual demographic characteristics.
Group task performance was operationalized
as how confident the group was that the correct
(guilty) suspect committed the murder. For
instance, groups could report that they were
Phillips et al. diversity and performance
peer-00571629, version 1 - 1 Mar 2011
0% to 100% certain that the correct suspect
committed the murder. Group members
responded individually, and individual
responses were aggregated to the group level as
in most groups the individual responses were
exactly the same (one way analysis of variance
(ANOVA) for group membership (F(71, 143) =
9.35, intraclass correlation coefficient (ICC(1))
= .74). To illustrate that our results are indeed
a reflection of group accuracy on this task, and
not just higher levels of overall confidence in
the groups’ assessments, we also examined how
confident groups were that each of the other
three (innocent) suspects committed the
Dependent variables
To measure group members’ awareness about
whether they possessed unique information,
group members were asked to respond to the
following item: ‘The information in my packet
was the same as that in others’ packets’, with 1
labeled ‘not at all the same’ and 7 labeled
‘exactly the same’ (Gruenfeld et al., 1996;
Phillips et al., 2004). Group discussion time was
measured in minutes and seconds with an
average of 20.11 minutes (SD = 6.67). Likewise,
perceptions of similarity were measured on the
post-manipulation questionnaire with the
following item ‘I feel similar to the other
members of my group’. How attracted group
members thought others were to them was
captured by four items measured on a 7-point
Likert scale; ‘I think I will like being a member
of this group’, ‘I think the others in the group
will like me’, ‘I am fun to be with’, and ‘I am
easy to like’ (= .87). The questions were asked
from the perspective of the participant because
the internal feelings of the individual, even if
others in the group were actually more or less
attracted, were thought to be the critical assess-
ment leading to how socially validated group
members would feel during the group dis-
To examine the number and content of the
similarities listed by groups, two independent
raters who were blind to the hypotheses first
counted the number of similarities listed by
each group. They then categorized the deep-
level similarities listed into two categories:
characteristics that were more obvious because
of the nature of the groups and participants
being used, versus less obvious characteristics
that took more discussion and interaction to
Obvious characteristics were things that were
clear to all of the group members and poten-
tially similar among many participants in the
study, such as ‘all members had been in (a
major building on campus)’, that all students
must enter for class registration and other
activities. Other similarities characterized as
obvious included items such as group members
were all wearing shoes; group members all
have hair, etc. Less obvious characteristics were
described as those that could not be discerned
by the group members from simply looking at
the group members or knowing that they were
students at the university, i.e. similarities that
could not have been a matter of common
knowledge. These less obvious similarities
between group members included personal
likes and dislikes (such as all members liked
certain foods or certain restaurants), similari-
ties in places that individuals had visited, and
similarities in nonobvious demographic charac-
teristics (such as the number of siblings). Thus,
group members had to be explicit in divulging
nonobvious information about themselves to
discover hidden interpersonal similarities.
The two raters were trained together on
about 10% of the groups (eight groups). The
162 similarities listed in those eight groups were
coded in the same category 93.2% of the time.
This gave us confidence to allow the two coders
to proceed independently with coding the rest
of the groups. One coder independently coded
39 of the 72 groups and the other coder did the
25 remaining groups.
To test our hypotheses regarding the impact of
highlighting deep-level similarities in surface-
level homogeneous and diverse groups, we
used a 2 (surface-level homogeneous or
surface-level diverse) 2 (deep-level similarity
intervention or control) analysis of covariance
Group Processes & Intergroup Relations 9(4)
peer-00571629, version 1 - 1 Mar 2011
(ANCOVA) approach to analyze the data.
Gender composition of the group (male or
female) was a covariate in the analyses.
All analyses were conducted at the group
level. We tested the validity of aggregating the
dependent variables of perceived uniqueness of
information, feelings of similarity, and per-
ceived attraction to the group level in two steps
(see Bliese, 2000). First, we ran a one-way
analysis of variance with group membership as
the independent variable to ensure that the
variance between teams was greater than the
variance within teams. In all cases, the ANOVAs
were significant (F(71, 143) = 2.15, p < .001;
F(53, 108) = 1.50, p < .04; F(53, 108) = 1.57,
p < .03). We then calculated ICC(1) values for
each of the dependent variables to test how
much variability in individual responses was due
to group membership (Bliese, 2000; Klein &
Kozlowski, 2000). The ICC(1) values were 0.28,
0.15, and 0.16 respectively, suggesting that
there was significant interdependence in indi-
vidual responses and that group membership
accounted for a reasonable proportion of
variance in individual responses on these three
dependent variables (Bliese, 2000; James,
Manipulation checks
The composition of the groups was validated
against the self-reported race of each of the
participants. This revealed that there were two
groups initially categorized by the experimenter
as racially homogeneous that self-identified as
racially diverse. These groups were moved to the
appropriate category before analysis.
To check on the outcomes of the similarity-
manipulation, we first examined the number of
similarities listed by surface-level homogeneous
and diverse groups. We found no significant
differences in the number of similarities listed
by the two types of groups (M = 18.84, SD =
7.98). Likewise, there were no significant differ-
ences in the types of similarities listed by the
groups with an average of 59% (SD = 18%)
being categorized as nonobvious by the coders.
These results suggest that surface-level homoge-
neous and surface-level diverse groups naturally
generated approximately the same number and
type of similarities during the similarity inter-
We also examined how similar group
members felt to one another right after com-
pleting the manipulation, but before beginning
discussion of the task. As expected, group
members felt more similar to one another after
highlighting deep level similarities (M = 4.71,
SD = 0.80) than in the control conditions (M =
4.49, SD = 0.76) (F(1, 49) = 3.92, p = .05).
Notably, there was a marginally significant inter-
action effect (F(1, 49) = 2.86, p = .097), suggest-
ing that deep-level similarities increased
feelings of similarity more for surface-level
homogeneous groups than for surface-level
diverse groups (see Table 2 for means).
Hypotheses tests
In Hypothesis 1, we predicted that surface-level
homogeneous groups would report discovering
less unique information and would discuss the
task less than would surface-level diverse
groups. We conducted a multivariate analysis of
variance analysis and found support for our
hypothesis (F(2, 63) = 3.14, p < .05). The follow-
up univariate tests revealed that surface-level
homogeneous groups reported that the infor-
mation in their packets was more similar (M =
3.31, SD = 1.18) than did surface-level diverse
groups (M = 2.78, SD = 0.80) (F(1, 67) = 4.39,
p < .05). Likewise, we found that surface-level
homogeneous groups (M = 17.90, SD = 7.54)
spent less time discussing the task than surface-
level diverse groups (M = 21.72, SD = 5.52) (F(1,
64) = 5.29, p < .03). Table 1, which includes the
correlations among all of the variables reported
here, shows that discussion time and awareness
of unique information were significantly corre-
lated at r = –.46, p < .01. Controlling for the
effects of discussion time on the awareness of
unique information, the effect of surface-level
group composition on awareness of unique
information was no longer significant (F < 1.0,
p > .30), suggesting that discussion time
mediated the effects of surface-level composi-
tion on the discovery of unique information.
Hypothesis 2A predicted that learning about
nontask relevant deep-level similarities would
increase perceived attraction among the group.
Phillips et al. diversity and performance
peer-00571629, version 1 - 1 Mar 2011
Analysis revealed a marginally significant main
effect for the similarity manipulation (F(1, 49)
= 3.00, p < .10); groups exposed to the similarity
intervention reported greater feelings of attrac-
tion (M = 5.29, SD = 0.54) than those in the
control condition (M = 5.15, SD = 0.54). This
main effect was accompanied by a significant
and unexpected interaction effect (F(1, 49) =
5.38, p < .03). While surface-level homogeneous
groups were more attracted to one another
after the deep-level similarity manipulation, this
was not the case for surface-level diverse groups
(see Table 2 for means). Neither Hypotheses 2B
(that there would be a main effect of the deep-
level similarity manipulation on awareness of
unique information and discussion time) nor
Hypothesis 2C (that the impact of surface-level
composition would interact with the similarity
intervention to affect awareness of unique
information and discussion time) was sup-
ported (Fs < 1, ps = ns).
Group performance was analyzed using an
ANCOVA analysis with two covariates—gender
composition and the average individual confi-
dence about the a priori correct suspect prior
to group discussion. Although the means were
in the expected direction (M = 65.60, SD =
26.76 and M = 56.77, SD = 31.91 for surface-
level diverse and homogeneous groups respec-
tively), the main effects for surface-level
diversity and the deep-level similarity manipu-
lation were not supported (Fs < 1, ps = ns).
Further analysis did reveal a significant inter-
action between surface-level diversity and the
deep-level similarity intervention (F(1, 66) =
6.29, p < .02). As predicted in Hypothesis 3c, we
found that surface-level diverse groups per-
formed worse after being exposed to the deep-
level similarity manipulation (M = 61.00, SD =
28.69) than when they were not (M = 72.79, SD
= 22.41) (F(1, 37) = 4.20, p < .05). In contrast
and contrary to our expectations, surface-level
homogeneous groups seemed to perform
better after experiencing the similarity inter-
vention (M = 67.89, SD = 31.68) than when they
did not (M = 46.35, SD = 29.33), although this
difference did not reach significance (F(1, 27)
= 2.34, p = .138). (Both individual decisions and
gender composition were controlled for in
these follow-up analyses.) Additional analyses
on how confident groups were that each of the
three incorrect suspects committed the murder
revealed no significant effects. This rules out
the possibility that some conditions just made
groups more confident than others did.
Finally, Table 1 includes correlations of all
variables reported in the results. Of note, the
correlations show that group composition was
related to how long groups discussed the task
(r = –.28, p < .02), such that surface-level homo-
geneous groups discussed the task for a shorter
period of time; and group discussion time was
positively related to group performance—the
group’s confidence that the correct suspect
committed the murder (r = .26, p < .04).
Moreover, group discussion time, awareness of
Group Processes & Intergroup Relations 9(4)
Table 1. Means, standard deviations, and correlations for study variables (N = 72)
Mean (SD)1234567
1. Surface-level composition 0.43 (0.50) 1
2. Deep-level similarity intervention 0.56 (0.50) –.13 1
3. Awareness of unique information 3.01 (1.01) 0.26* –.21† 1
4. Time discussing task 20.11 (6.67) –.28* 0.11 –.46** 1
5. Feelings of similarity 4.62 (0.78) 0.17 0.14 0.02 –.29* 1
6. Perceived attraction 5.45 (.54) 0.02 0.13 –.06 –.02 0.53** 1
7. Group performance 61.81 (29.20) –.15 0.07 –.36** 0.26* –.11 –.05 1
p < .10; *p < .05; **p < .01.
Notes: N = 69 for time discussing task and N = 54 for the perceptions of similarity attraction.
Surface-level composition is coded 1 = homogeneous, 0 = diverse; deep-level similarity intervention is coded
1 = yes, 0 = control; and awareness of unique information is coded 1 = information is not the same (i.e.
unique) to 7 = information is the same.
peer-00571629, version 1 - 1 Mar 2011
unique information, and group performance
were all significantly correlated (rs > .26, ps <
.05), meaning that groups performed better
when they spent more time in discussion and,
in turn, became more aware of the differences
in information present among group members.
This article reports how different types of diver-
sity—surface- and deep-level—influence group
decision-making and performance. Under-
standing the impact of diversity on decision-
making groups is often complicated by the
different sources of diversity that can exist and
the assumed correlation among these sources.
In this study, we moved beyond the typical
social categorization perspective on diversity by
highlighting a by-product of the social
categorization process—assumptions of in-
group similarity—which has been overlooked
by researchers in this tradition. We argue that
surface-level diversity triggers expectations that
informational differences may be present in
groups, making it more expected and legiti-
mate for group members to raise and discuss
unique information that may be critical for
group performance. Moreover, by highlighting
deep-level similarities in an effort to increase
levels of attraction and diminish social
categorization effects, managers may under-
mine the benefits of having surface-level diver-
sity present in groups that must share unique
information for effective performance.
While our results revealed a substantial
amount of support for this perspective, there
were some surprises. First, in support of this
perspective we found that surface-level diverse
groups perceived their information to be less
similar, and spent more time discussing the task
than did surface-level homogeneous groups.
Moreover, the more time group members spent
discussing the task, the more likely they were to
perceive the information as different. These
results were strong, expected, and independent
of whether groups learned about deep-level
As argued and shown by past research,
learning about deep-level similarities had a
positive impact on how similar and how
attracted group members felt toward one
another (Byrne, 1971), but mostly for the
surface-level homogeneous groups. The simi-
larity manipulation did not reliably increase
perceived attraction for the surface-level
diverse groups, even though our manipulation
check revealed that the number and type of
similarities generated by the two types of groups
did not differ. Moreover, we did not find
evidence that the deep-level similarities had a
differential impact on surface-level homoge-
neous and diverse groups’ awareness of unique
information or discussion time. There was only
the significant interaction on group perform-
ance showing that surface-level diverse groups
performed worse after being exposed to the
deep-level similarity manipulation. In contrast,
surface-level homogeneous groups seemed to
be helped by the intervention, although not
significantly so. Overall, the general arguments
made in this article were supported with some
exceptions. A closer look at the deep-level simi-
larity manipulation may lend some insight into
future research opportunities.
Phillips et al. diversity and performance
Table 2. Means (standard deviations) for feelings of similarity and attraction (N = 54 groups)
Surface-level Surface-level
homogeneous/ diverse/
Surface-level Deep-level Deep-level
homogeneous/ similarity Surface-level similarity
Control intervention diverse/Control intervention
Feelings of similarity 4.41 (0.88)
5.14 (0.69)
4.60 (0.60) 4.44 (0.75)
Perceived attraction 5.00 (0.54)
5.56 (0.49)
5.34 (0.51) 5.12 (0.51)
Note: Means sharing subscripts within each row differ from one another significantly at p < .05.
peer-00571629, version 1 - 1 Mar 2011
Group members were instructed to spend
five minutes discovering as many things as
possible that all three of the group members
had in common. This meant that in the process
of finding these similarities, groups were also
likely to discover similarities that bonded only
subsets of the members, and differences where
none of the group members were similar.
Without a measure of these partially shared sim-
ilarities or the differences that surfaced, it is dif-
ficult to determine exactly how they might have
affected our results here. However, given that
all groups had the same amount of time for the
exercise and that there were no differences in
the number of similarities discovered, it is likely
that there were no systematic differences across
the conditions on the number of differences
found either. Future research should disentan-
gle the effects of highlighting similarities from
the effects of discovering differences, and the
implications of doing this in light of a group’s
surface-level composition.
Sharing unique information is difficult in
groups, because sharing the information often
means that one has to disagree with the rest of
the group about what is important for making
the group decision. Providing the context that
allows groups to benefit fully from the unique
information and perspectives that each group
member holds is a difficult task. This research
suggests that surface-level diversity might be
beneficial for groups because of the legitimacy
it lends to sharing unique perspectives. For
surface-level homogeneous groups the obvious
similarities among group members hindered
the recognition of unique information and led
them to perform worse. In fact, the surface-
level homogeneous groups in the control con-
dition performed worse than any others in our
sample. For these groups, sharing unique infor-
mation may be particularly difficult because of
the social risks involved (e.g. Gruenfeld et al.,
1996; Wittenbaum et al., 1999).
Surface-level homogeneous groups, where
group members are concerned about being
accepted by their fellow in-group members,
suffer more from conformity pressures that
prevent them from sharing unique information
and opinions than do surface-level diverse
groups (e.g. Abrams et al., 1990; Janis, 1982;
Phillips & Loyd, 2006). Postmes et al. (2001)
suggest that one way to overcome the reluc-
tance to share unique information is to develop
norms of criticality instead of consensus. This
may be even more important for surface-level
homogeneous than for surface-level diverse
groups. Moreover, homogeneity without feelings
of attraction and acceptance may be detrimen-
tal to groups when sharing unique information
is crucial for performance (Gruenfeld et al.,
1996; Jehn & Shah, 1997). There is a fine line
that surface-level homogeneous groups need to
walk. Feeling too similar to one another may
undermine the amount of time group members
spend on the task, as suggested by the correla-
tion found here between feelings of similarity
and discussion time, but not feeling similar
enough may lead group members to feel
insecure and have concern about sharing
unique information at all.
Future research should attempt to disentan-
gle the effects of similarity and familiarity on
work teams. For instance, does familiarity have
the same effect on surface-level homogeneous
as it does on diverse groups? How familiar do
people need to be with one another before the
potential benefits of familiarity outweigh the
potential downsides of similarity/diversity?
From the study conducted here, it seems that
increasing familiarity (e.g. via 5 minutes of dis-
cussing similarities) can only slightly overcome
the conformity pressures that come along with
surface-level homogeneity. However, it may take
considerably longer for surface-level diverse
groups to benefit from familiarity (Watson et
al., 1993). The amount of diversity present in
our groups, and the small group size, may also
be limitations that should be considered in
future research. Would the diverse groups have
performed even better if they had had greater
surface-level diversity? Would the results be the
same if a different type of surface-level diversity
were used?
In conclusion, as organizations attempt to
cope with the changing demography of the
work force there is a natural tendency to believe
that what enhances the performance of surface-
level homogeneous groups may also enhance
Group Processes & Intergroup Relations 9(4)
peer-00571629, version 1 - 1 Mar 2011
the performance of surface-level diverse
groups. The current research suggests that
enhancing the performance of workgroups is
more complex than that. Our research found
that attempting to diminish the salience of
potentially disruptive categorical boundaries by
asking members of racially diverse groups to
focus on their similarities was detrimental for
group performance. Diversity can be beneficial
for groups, not merely because individuals
belonging to different subgroups are likely to
have access to differing information, but also
because the presence of salient differences may
legitimate the introduction and consideration
of unique information in the group’s decision-
making process. Both the positive and negative
effects of surface-level and deep-level diversity
in the workplace are yet to be fully understood.
This work is another step toward understanding
the effects of diversity on groups where sharing
unique information is crucial to performance.
1. We conducted a follow-up study with 12
three-person racially homogeneous groups that
either worked together to identify US states and
capitals or worked together to identify similarities.
Results revealed that identifying similarities does
indeed lead individuals to feel more similar to
one another than just working together recalling
US states and capitals (F(1, 10) = 10.48, p < .01).
Moreover, a comparison of feelings of similarity
by the groups that worked together to identify US
states and capitals (M = 4.71, SD = .56) and those
that worked separately when they identified states
and capitals (from the initial study) (M = 4.41, SD
= .88) revealed that there was no significant
difference between the two types of groups
(t(18) = .821, p = .423). Theoretically, we do not
believe that just working together (without
discussing similarities) could generate our same
pattern of data. The highlighting of similarities in
surface-level diverse groups causes a particular
threat to expectations that just working together
would not. For instance, in the work by
Gruenfeld et al. (1996) familiarity was found to
lead to better performance when unique
information needed to be shared. This suggests
that both the surface-level homogeneous and
diverse groups should have improved their
performance after working together to highlight
deep-level similarities if simply working together
is all that is important. As this was not the case we
believe that our current manipulation of
highlighting similarities does indeed add a
unique contribution to people’s experience,
beyond just working together.
2. Bliese (2000, p. 361) reports that values between
.05 and .30 should be expected in most applied
field settings.
The authors are indebted to the graduate students
who assisted in the data collection efforts. Also, we
thank KTAG for financial support and the members
of the Social Interactions Lab at Northwestern
University for their helpful comments. We thank
Elson Huang and Tanya Canak for coding help. In
addition, I would like to thank the audiences at
several colloquiums who offered their insights on
this work.
Abrams, D. M., Wetherell, M., Cochrane, S., Hogg,
M. A., & Turner, J. C. (1990). Knowing what to
think by knowing who you are: Self-categorization
and the nature of norm formation, conformity
and group polarization. British Journal of Social
Psychology, 29, 97–119.
Allen, V. L., & Wilder, D. A. (1975). Categorization,
beliefs similarity, and intergroup discrimination.
Journal of Personality and Social Psychology, 32,
Allen, V. L., & Wilder, D. A. (1979). Group
categorization and attribution of belief similarity.
Small Group Behavior, 10, 73–80.
Antonio, A. L., Chang, M. J., Hakuta, K., Kenny,
D. A., Levin, S., & Milem, J. F. (2004). Effects of
racial diversity on complex thinking in college
students. Psychological Science, 15, 507–510.
Asch, S. E. (1952). Social psychology. Englewood Cliffs,
NJ: Prentice-Hall.
Bliese, P. D. (2000). Within-group agreement,
non-independence, and reliability: Implication for
data aggregation and analysis. In K. J. Klein &
S. W. J. Kozlowski (Eds.), Multilevel theory, research,
and methods in organizations: Foundations, extensions,
and new directions (pp. 349–381). San Francisco:
Byrne, D. (1971). The attraction paradigm. New York:
Academic Press.
Phillips et al. diversity and performance
peer-00571629, version 1 - 1 Mar 2011
Chen, F. F., & Kenrick, D. T. (2002). Repulsion or
attraction? Group membership and assumed
attitude similarity. Journal of Personality and Social
Psychology, 83, 111–125.
Diehl, M. (1988). Social identity and minimal
groups: The effects of interpersonal and
intergroup attitudinal similarity on intergroup
discrimination. British Journal of Social Psychology,
27, 289–300.
Eagly, A. H. (1967). Involvement as a determinant of
response to favorable and unfavorable
information. Journal of Personality and Social
Psychology, 7, 15.
Ely, R., & Thomas, D. (2001). Cultural diversity at
work: The effects of diversity perspectives on work
group processes and outcomes. Administrative
Science Quarterly, 46, 229–273.
Franz, T. M., & Larson, J. R., Jr. (2002). The impact
of experts on information sharing during group
discussion. Small Group Research, 33, 383–411.
Gaertner, S., L., Dovidio, J. F., Banker, B. S.,
Houlette, M., Johnson, K. M., & McGlynn, E. A.
(2000). Reducing intergroup conflict: From
superordinate goals to decategorization,
recategorization, and mutual differentiation.
Group Dynamics, 4, 98–114.
Gaertner, S. L., Mann, J., Murrell, A., & Dovidio, J.
(1989). Reducing intergroup bias: The benefits of
recategorization. Journal of Personality and Social
Psychology, 57, 239–249.
Gruenfeld, D. H., Mannix, E. A., Williams, K. Y., &
Neale, M. A. (1996). Group composition and
decision making: How member familiarity and
information distribution affect process and
performance. Organizational Behavior and Human
Decision Processes, 67, 1–15.
Harrison, D. A., Price, K. H., & Bell, M. P. (1998).
Beyond relational demography: Time and effects
of surface- and deep-level diversity on work group
cohesion. Academy of Management Journal, 41,
Harrison, D. A., Price, K. H., Gavin, J. A., & Florey,
A. T. (2002). Time, teams, and task performance:
Changing effects of surface- and deep-level
diversity on group functioning. Academy of
Management Journal, 45, 1029–1045.
Holtz, R., & Miller, N. (1985). Assumed similarity
and opinion certainty. Journal of Personality and
Social Psychology, 48, 890–898.
Hornsey, M. J., & Jetten, J. (2004). The individual
within the group: Balancing the need to belong
with the need to be different. Personality and Social
Psychology Review, 8, 248–264.
Jackson, S. E., Joshi, A., & Erhardt, N. L. (2003).
Recent research on team and organizational
diversity: SWOT analysis and implications. Journal
of Management, 29, 801–830.
Jackson, S. E., May, K. E., & Whitney, K. (1995).
Understanding the dynamics of diversity in
decision-making teams. In R. A. Guzzo & E. Salas
(Eds.), Team decision-making effectiveness in
organizations (pp. 204–261). San Francisco:
James, L. R. (1982). Aggregation bias in estimates of
perceptual agreement. Journal of Applied Psychology,
67, 219–229.
Janis, I. L. (1982). Victims of groupthink (2nd ed).
Boston: Houghton Mifflin.
Jehn, K., Northcraft, G., & Neale, M. (1999). Why
differences make a difference: A field study of
diversity, conflict, and performance in work
groups. Administrative Science Quarterly, 44,
Jehn, K. A., & Shah, P. P. (1997). Interpersonal
relationships and task performance: An
examination of mediation processes in friendship
and acquaintance groups. Journal of Personality and
Social Psychology, 72, 775–790.
Jetten, J., Postmes, T., & McAuliffe, B. J. (2002).
‘We’re all individuals’: Group norms of
individualism and collectivism, levels of
identification and identity threat. European Journal
of Social Psychology, 32, 189–207.
Klein, K. J., & Kozlowski, S. W. J. (2000). From micro
to meso: Critical steps in conceptualizing and
conducting multilevel research. Organizational
Research Methods, 3, 211–236.
Kramer, R., & Brewer, M. B. (1984). Effects of group
identity on resource use in a simulated commons
dilemma. Journal of Personality and Social Psychology,
46, 1044–1057.
Larson, J. R., Jr., Christensen, C., Abbott, A. S., &
Franz, T. M. (1996). Diagnosing groups: Charting
the flow of information in medical decision-making
teams. Journal of Personality and Social Psychology, 71,
Larson, J. R., Jr., Christensen, C., Abbott, A. S., &
Franz, T. M. (1998). Diagnosing groups: The
pooling, management, and impact of shared and
unshared case information in team-based medical
decision-making. Journal of Personality and Social
Psychology, 75, 93–108.
Liljenquist, K. A., Galinsky, A. D., & Kray, L. J.
(2004). Exploring the rabbit hole of possibilities
by myself or with my group: The benefits and
liabilities of activating counterfactual mind-sets for
information sharing and group coordination.
Journal of Behavioral Decision Making, 17, 263–279.
Group Processes & Intergroup Relations 9(4)
peer-00571629, version 1 - 1 Mar 2011
Milliken, F. J., & Martins, L. L. (1996). Searching for
common threads: Understanding the multiple
effects of diversity in organizational groups.
Academy of Management Review, 21, 402–433.
Northcraft, G. B., & Martin, J. (1982). Double
jeopardy: Resistance to affirmative action from
potential beneficiaries. In B. Gutek (Ed.), Sex role
stereotyping and affirmative action policy
(pp. 81–130). Los Angeles, CA: Institute of
Industrial Relations, UCLA.
Pelled, L. H., Eisenhardt, K. M., & Xin, K. R. (1999).
Exploring the black box: An analysis of work
group diversity, conflict, and performance.
Administrative Science Quarterly, 44, 1–28.
Phillips, K. W. (2003). The effects of categorically
based expectations on minority influence: The
importance of congruence. Personality and Social
Psychology Bulletin, 29, 3–13.
Phillips, K. W., & Loyd, D. L. (2006). When surface
and deep-level diversity collide: The effects on
dissenting group members. Organizational Behavior
and Human Decision Processes, 99, 143–160.
Phillips, K. W., Mannix, E., Neale, M., &
Gruenfeld, D. (2004). Diverse groups and
information sharing: The effects of congruent ties.
Journal of Experimental Social Psychology, 40,
Polzer, J. T., Milton, L. P., & Swann, W. B. Jr. (2002).
Capitalizing on diversity: Interpersonal
congruence in small work groups. Administrative
Science Quarterly, 47, 296–324.
Postmes, T., Haslam, A., & Swaab, R. I. (2005). Social
influence in small groups: An interactive model of
social identity formation. European Review of Social
Psychology, 16, 1–42.
Postmes, T., Spears, R., & Cihangir, S. (2001).
Quality of decision making and group norms.
Journal of Personality and Social Psychology, 80,
Schittekatte, M., & Van Hiel, A. (1996). Effects of
partially shared information and awareness of
unshared information on information sampling.
Small Group Research, 27, 431–449.
Schneider, S. K., & Northcraft, G. B. (1999). Three
social dilemmas of workforce diversity in
organizations: A social identity perspective.
Human Relations, 52, 1445–1467.
Stasser, G. (1999). The uncertain role of unshared
information in collective choice. In L. L.
Thompson, J. M. Levine, & D. M. Messick (Eds.),
Shared cognition in organizations: The management of
knowledge (pp. 49–69). Hillsdale, NJ: Erlbaum.
Stasser, G., & Stewart, D. (1992). Discovery of hidden
profiles by decision-making groups: Solving a
problem versus making a judgment. Journal of
Personality and Social Psychology, 63, 426–434.
Stasser, G., Stewart, D. D., & Wittenbaum, G. M.
(1995). Expert roles and information exchange
during discussion: The importance of knowing
who knows what. Journal of Experimental Social
Psychology, 31, 244–265.
Stasser, G., & Titus, W. (1985). Pooling of unshared
information in group decision making: Biased
information sampling during discussion. Journal of
Personality and Social Psychology, 48, 1467–1478.
Stasser, G., & Titus, W. (1987). Effects of information
load and percentage of shared information on the
dissemination of unshared information during
group discussion. Journal of Personality and Social
Psychology, 53, 81–93.
Stasser, G., Vaughan, S. I., & Stewart, D. D. (2000).
Pooling unshared information: The benefits of
knowing how access to information is distributed
among group members. Organizational Behavior
and Human Decision Processes, 82, 102–116.
Stewart, D. D., & Stasser, G. (1995). Expert role
assignment and information sampling during
collective recall and decision-making. Journal of
Personality and Social Psychology, 69, 619–628.
Swann, W. B., Jr., Milton, L. P., & Polzer, J. T. (2000).
Should we create a niche or fall in line? Identity
negotiation and small group effectiveness. Journal
of Personality and Social Psychology, 79, 238–250.
Tajfel, H. (1969). Cognitive aspects of prejudice.
Journal of Social Issues, 25, 79–97.
Van Knippenberg, D., De Dreu, C. K. W., & Homan,
A. C. (2004). Work group diversity and group
performance: An integrative model and research
agenda. Journal of Applied Psychology, 89,
Van Knippenberg, D., & Haslam, S. A. (2003).
Realizing the diversity dividend: Exploring the
subtle interplay between identity, ideology and
reality. In S. A. Haslam, D. van Knippenberg,
M. Platow, & N. Ellemers (Eds.), Social identity at
work: Developing theory for organizational practice
(pp. 61–77). New York: Taylor & Francis.
Watson, W., Kumar, K., & Michaelsen, L. (1993).
Cultural diversity’s impact on interaction process
and performance: Comparing homogeneous and
diverse task groups. Academy of Management Journal,
36, 590–602.
Westphal, J. D., & Milton, L. P. (2000). How
experience and network ties affect the influence
of demographic minorities on corporate boards.
Administrative Science Quarterly, 45, 366–398.
Wilder, D. A. (1984). Predictions of belief
homogeneity and similarity following social
Phillips et al. diversity and performance
peer-00571629, version 1 - 1 Mar 2011
categorization. British Journal of Social Psychology,
23, 323–333.
Williams, K., & O’Reilly, C. (1998). Demography and
diversity in organizations: A review of 40 years of
research. In B. M. Staw & R. Sutton (Eds.),
Research in organizational behavior (Vol. 21,
pp. 77–140). Greenwich, CT: JAI Press.
Winquist, J. R., & Larson, J. R. (1998). Information
pooling: When it impacts group decision making.
Journal of Personality and Social Psychology, 74,
Wittenbaum, G. M. (2000). The bias toward
discussing shared information: Why are high status
members immune? Communication Research, 27,
Wittenbaum, G. M., & Bowman, J. M. (2004). A
social validation explanation for mutual
enhancement. Journal of Experimental Social
Psychology, 40, 169–184.
Wittenbaum, G. M., Hollingshead, A. B., & Botero,
I. C. (2004). From cooperative to motivated
information sharing in groups: Moving beyond the
hidden profile paradigm. Communication
Monographs, 71, 286–310.
Wittenbaum, G. M., Hubbell, A. P., & Zuckerman, C.
(1999). Mutual enhancement: Toward an
understanding of the collective preference for
shared information. Journal of Personality and Social
Psychology, 77, 967–978.
Wittenbaum, G. M., & Stasser, G. (1996).
Management of information in small groups. In
J. L. Nye & A. M. Brower (Eds.), What’s social about
social cognition? Research on socially shared cognition
in small groups (pp. 3–28). Thousand Oaks, CA:
Paper received 8 November 2004; revised version accepted
28 February 2006.
Biographical notes
KATHERINE W. PHILLIPS is associate professor of
management and organizations at the Kellogg
School of Management at Northwestern
University. Her research focuses on diversity in
decision making groups, information sharing,
minority influence, identity boundary
management and the effects of status in groups.
She received her PhD from the Graduate School
of Business at Stanford University.
GREGORY B. NORTHCRAFT is the Harry J. Gray
Professor of Executive Leadership in the
department of business administration, and
Institute of Labor and Industrial Relations, at the
University of Illinois. His major research interests
include conflict management and negotiation,
collaboration in teams, managerial decision
making, and employee motivation and job design,
particularly in high-technology manufacturing
settings. He received his PhD in social psychology
at Stanford University.
MARGARET A. NEALE is the John G. McCoy-Banc
One Corporation Professor of Organizations and
Dispute Resolution at the Graduate School of
Business at Stanford University. Professor Neale’s
major research interests include bargaining and
negotiation, distributed work groups, and team
composition, learning, and performance. She
received her PhD in business administration from
the University of Texas.
Group Processes & Intergroup Relations 9(4)
peer-00571629, version 1 - 1 Mar 2011
... Similarity between individuals positively impacts interactions, cohesion, performance, likeness, and perceived competence (Guéguen, Martin, & Meineri, 2011;Harrison, Price, & Bell, 1998;Phillips, Northcraft, Neale, & relations, 2006;Singh et al., 2015). Similarity can be examined from two levels: surfacelevel similarity and deep-level similarity (Harrison et al., 1998;Kacmar, Harris, Carlson, & Zivnuska, 2009;Phillips et al., 2006). Surface-level similarity, also known as demographic similarity, refers to the similarity based on the salient characteristics (i.e., gender and ethnicity) between individuals. ...
... Surface-level similarity, also known as demographic similarity, refers to the similarity based on the salient characteristics (i.e., gender and ethnicity) between individuals. Deep-level similarity, or attitudinal similarity, refers to the shared beliefs, attitudes, and opinions among individuals (Harrison et al., 1998;Phillips et al., 2006). ...
Conference Paper
Full-text available
With the rise and integration of AI technologies within organizations, our understanding of the impact of this technology on individuals remains limited. Although the IS use literature provides important guidance for organization to increase employees' willingness to work with new technology, the utilitarian view of prior IS use research limits its application considering the new evolving social interaction between humans and AI agents. We contribute to the IS use literature by implementing a social view to understand the impact of AI agents on an individual's perception and behavior. By focusing on the main design dimensions of AI agents, we propose a framework that utilizes social psychology theories to explain the impact of those design dimensions on individuals. Specifically, we build on Similarity Attraction Theory to propose an AI similarity-continuance model that aims to explain how similarity with AI agents influence individuals' IT identity and intention to continue working with it. Through an online brainstorming experiment, we found that similarity with AI agents indeed has a positive impact on IT identity and on the intention to continue working with the AI agent.
... Related to this is the idea of surface-level diversity, i.e. that which can be seen (race/gender) and deeplevel diversity, i.e. similarities in beliefs, values etc. Phillips et al. (2006) posits that surface-level diversity is beneficial as it visually signals to others that that individual will have unique information and therefore is of value in the workplace. As our participants remained anonymous, there is no way that such factors were at play in our study revealing that the impact of diversity can be influential with the group without the need for visual cues or identification of group similarity or difference. ...
Diverse groups have been shown to result in higher levels of innovation and collaborative decision making. For the UK to meet our net zero goals in an innovative and timely manner while also representing the diverse needs and aspirations of the population requires adequate diversity within the policymaking sphere. The aim of this project is to use a serious game to understand and assess the impact group diversity has on collective decision-making, set within the context of sustainable development.
... Beyond the determination of guilt or innocence, several studies have demonstrated positive outcomes relating to superior group performance by diverse groups, such as increased creativity, flexibility, information sharing, and idea generation, both in general and in legal contexts (Peter-Hagene, 2019;Phillips et al., 2004Phillips et al., , 2006Sommers et al., 2008). Researchers have examined the observable influence of jury racial diversity and have found significant effects across an array of group performance metrics (ForsterLee et al., 1994;Hinsz & Indahl, 1995). ...
Full-text available
The primary goal of this research was to determine whether the racial composition of a jury impacts the outcome and deliberation in a civil retail discrimination lawsuit. We presented a retail discrimination trial video to 30 separate mock juries. Of the 30 juries, 15 juries had 2 Black jurors, while the remaining 15 had no Black jurors (i.e., only White or White and Latinx participants). After watching the video, each mock jury was given 1 h to deliberate as if they were jurors deciding an actual case. Contrary to previous research, juries with Black jurors were no more likely to deliberate longer or consider more case facts. However, they were more likely to find for the plaintiff and award higher compensation. In addition, content analysis of the deliberation racial discourse revealed that some jurors espoused colorblind racial attitudes, minimizing the significance of race in this case and in society in general, and accusing the plaintiff of playing the “race card.” Other jurors encouraged understanding of the plaintiff and espoused non‐colorblind racial attitudes, recognizing the importance and damaging role of racism in American society. Implications for civil trials, jury selection, and racial discourse are discussed.
Full-text available
Team member differences can be found in various characteristics and be seen as both perks and perils. But what makes one group focus on certain dimensions and differences’ positive implications, while another collective notices other aspects and sees trouble ahead? We address this question in the context of interorganizational teams’ first stages, when impressions are limited and valuations must be made promptly. Our findings from in-depth interviews offer a sensemaking perspective on perceived otherness and explicate when and why differences are interpreted as helping or hindering collaborative practices. Moreover, we illuminate how coorientation and representation dynamics shape otherness perceptions and valuations.
Increasing demographic diversity is undoubtedly important and can aid in debiasing decision makers. Yet, the promises of demographic diversity are not always realized due to social integration problems. We consider why and for whom differences combined with homogeneity make a difference for groups in terms of integratively complex thinking and ideological decision making. Although research has shown that decision makers often rely on political biases, that work has not addressed when and why decision-making groups are able to overcome these biases—a pervasive concern in today’s politically polarized social milieu. Drawing on the common in-group identity model and research on integrative complexity, we theorize that demographic diversity ultimately yields less ideological decision making because it prompts integrative complexity; however, demographic diversity only accrues this benefit in the presence of ideological homogeneity. We also reason that the relationship between integrative complexity and reduced ideological decision making emerges for more conservative (versus more liberal) groups. We find support for our expectations using a natural experiment of judges on the U.S. Courts of Appeals. Supplemental analyses indicate that working within a demographically diverse and ideologically homogeneous group also positively predicts integrative complexity in future decision-making groups. Finally, we find that demographic and ideological diversity can substitute for one another, but no additional integrative complexity benefits accrue when both are present. We discuss implications of this research in light of the ongoing conversation about the value of diversity and today’s polarized political climate. Supplemental Material: The online appendix is available at .
In both academic and policy contexts, two of the most widely researched and discussed diversity ideologies are multiculturalism (i.e., acknowledgment and celebration of group differences) and colorblindness (which can involve focusing on group similarities and characteristics of individuals instead of differences). However, both diversity ideologies have potential drawbacks, and their implications for intergroup contexts beyond race and ethnicity are not well understood. Given that the United States is becoming increasingly diverse in terms of religion, gender identity, and sexual orientation as well as race and ethnicity, we propose a “multiple forms” approach to diversity ideologies, particularly multiculturalism. We suggest that explicitly emphasizing the importance of many types of diversity may be beneficial for dominant and non‐dominant group members and for institutions and organizations more broadly. In this article, we present an overview of the “multiple forms” approach and what it would entail, review empirical evidence supporting its potential effectiveness, discuss the psychological and legal advantages and challenges involved in implementing such an approach, and offer concrete policy recommendations for doing so.
In this book, we explore the workplace experiences, opportunities, and challenges as it relates to the heterogeneity of persons with disabilities. In this chapter, we outline the scope of the text and importantly, the interrelated nature of diversity, discrimination, inclusion and equality in the English-speaking Caribbean, specifically in relation to persons with disabilities, within the context of employment. Many of the islands of the Commonwealth Caribbean have signalled their commitment to inclusivity, diversity and the reduction of disparate treatment for persons with disabilities, by becoming signatories to the United Nations (UN) Convention on the Rights of Persons with Disabilities (CRPD), which entered into force in 2008 and aims to “promote, protect and ensure the human rights, dignity and fundamental freedoms of all persons with disabilities”. Moreover, some Anglo-Caribbean islands have enacted legislation which (congruent with the directives of the CRPD) prohibit discrimination on the basis of disabilities, within the context of education, employment and the provision of goods and services.KeywordsPersons with disabilitiesCommonwealth CaribbeanEqualityDiversityInclusionEmploymentModels of disabilities
Introduction Students in higher education courses need to practice vital groupwork skills that are increasingly relevant in today's collaborative workplace, which often occurs through small group coursework. Statement of the Problem The process of forming small groups with students who are committed to the topic and goals is frequently problematic. Literature Review We review the literature on the advantages and disadvantages of common group formation strategies, such as self-assembly and professor assignment. Teaching Implications We discuss the use of a unique method of forming small groups in a university setting through a fun, interactive activity based on the classic “speed-dating” exercise. Conclusion We recommend that educators consider the use of a speed-interviewing group formation strategy to allow groups to form based on similar topic interests and work styles.
Scholars have extensively documented social class differences in students’ relationships with educational institutions through their interactions with authority figures and the unequal institutional advantages these interactions yield. However, little is known about whether or how social class also shapes students’ peer interactions in ways that produce these inequalities. Using a qualitative case study of an elite engineering school in which I draw on participant observation and interviews with 88 undergraduates and six administrators, I argue that social class context—a proxy for social class—shapes the peer help-seeking and help-giving (collaborative) strategies students use, which can create inequalities in the institutional advantages they secure in the form of academic help, support, and learning opportunities. Focusing specifically on the social class context of students’ high schools, I find that compared to their less-privileged counterparts, privileged students—who came from class-advantaged high school contexts where they became familiar with collaboration and upper-middle-class cultural signals—more easily collaborated with their college classmates and displayed signals that communicated they were “good” collaborators. The findings highlight new mechanisms through which inequalities are reproduced in educational institutions and make theoretical contributions to research on cultural capital, inequality, and education.
Full-text available
The interaction processes of culturally homogeneous and culturally diverse groups were studied for 17 weeks. Initially, homogeneous groups scored higher on both process and performance effectiveness. Over time, both homogeneous and heterogeneous groups showed improvement on process and performance, and between-group differences converged. By week 17, there were no differences in process or overall performance, but the heterogeneous groups scored higher on two task performance measures. Implications for management and future research are given.
Full-text available
The interaction process and performance of culturally homogeneous and culturally diverse groups were studied for 17 weeks. Initially, homogeneous groups scored higher on both process and performance effectiveness. Over time, both types of group showed improvement on process and performance, and the between-group differences converged. By week 17, there were no differences in process or overall performance, but the heterogeneous groups scored higher on two task measures, Implications for management and future research are given.
This study examines how the influence of directors who are demographic minorities on corporate boards is contingent on the prior experience of board members and the larger social structural context in which demographic differences are embedded. We assess the effects of minority status according to functional background, industry background, education, race, and gender for a large sample of corporate outside directors at Fortune/Forbes 500 companies. The results show that (1) the prior experience of minority directors in a minority role on other boards can enhance their ability to exert influence on the focal board, while the prior experience of minority directors in a majority role can reduce their influence; (2) the prior experience of majority directors in a minority role on other boards can enhance the influence of minority directors on the focal board, and (3) minority directors are more influential if they have direct or indirect social network ties to majority directors through common memberships on other boards. Results suggest that demographic minorities can avoid out-group biases that would otherwise minimize their influence when they have prior experience on other boards or social network ties to other directors that enable them to create the perception of similarity with the majority.
We examine interpersonal congruence, the degree to which group members see others in the group as others see themselves, as a moderator of the relationship between diversity and group effectiveness. A longitudinal study of 83 work groups revealed that diversity tended to improve creative task performance in groups with high interpersonal congruence, whereas diversity undermined the performance of groups with low interpersonal congruence. This interaction effect also emerged on measures of social integration, group identification, and relationship conflict. By eliciting self-verifying appraisals, members of some groups achieved enough interpersonal congruence during their first ten minutes of interaction to benefit their group outcomes four months later. In contrast to theories of social categorization, the interpersonal congruence approach suggests that group members can achieve harmonious and effective work processes by expressing rather than suppressing the characteristics that make them unique.
We examined the impact of surface-level (demographic) and deep-level (attitudinal) diversity on group social integration. As hypothesized, the length of time group members worked together weakened the effects of surface-level diversity and strengthened the effects of deep-level diversity as group members had the opportunity to engage in meaningful interactions.