See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/247720472

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

Level Similarity Help?

Article  in  Group Processes & Intergroup Relations · October 2006

DOI: 10.1177/1368430206067557

CITATIONS READS

172 4,868

3 authors, including:

Katherine Williams Phillips Margaret A. Neale

Columbia University Stanford University
77 PUBLICATIONS   5,843 CITATIONS    91 PUBLICATIONS   12,662 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Prenups in Start-ups: How to Write and Frame Contracts Among Founders to Ensure Start-up Success View project

The Intersecting Role of Culture and Gender in Negotiations View project

All content following this page was uploaded by Katherine Williams Phillips on 08 May 2014.

The user has requested enhancement of the downloaded file.

 Author manuscript, published in "Group Processes & Intergroup Relations 9, 4 (2006) 467-482"
DOI : 10.1177/1368430206067557

Group Processes &

G
Intergroup Relations P
2006 Vol 9(4) 467–482
I
R
Surface-Level Diversity
and Decision-Making in
Groups: When Does
Deep-Level Similarity
Help?
peer-00571629, version 1 - 1 Mar 2011

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

categorization

A P R I M A R Y reason organizations use groups is Stewart, 2000; Stewart & Stasser, 1995; Winquist
to garner the benefits of the unique knowledge & Larson, 1998; Wittenbaum, 2000) and some
and information that group members might
bring to the table (e.g. Schneider & Northcraft, Author’s note
1999). For nearly twenty years the sharing and Address correspondence to Katherine W.
integration of unique information in small Phillips, Kellogg School of Management,
group discussions has been the subject of much Northwestern University, 2001 Sheridan Road,
experimental (e.g. Stasser & Stewart, 1992; Evanston, IL 60208-2001, USA
Stasser & Titus, 1985, 1987; Stasser, Vaughan, & [email: [email protected]]

Copyright © 2006 SAGE Publications

(London, Thousand Oaks, CA and New Delhi)
9:4; 467–482; DOI: 10.1177/1368430206067557
www.sagepublications.com
 Group Processes & Intergroup Relations 9(4)

field research (e.g. Larson, Christensen, known by a single member) as opposed to

Abbott, & Franz, 1996, 1998). Reflecting the commonly held (i.e. known to all members)
reality that all individuals bring a unique con- information into group decisions is easier said
stellation of perspectives and experiences to than done (for reviews see Stasser, 1999; Witten-
small group discussions, this research examines baum & Stasser, 1996; Wittenbaum et al., 2004).
hidden profile decision situations in which sub- One reason why unique information is men-
optimal decisions are likely to be made if unique tioned and repeated less than commonly held
information is not shared and integrated into information is because group members gener-
the group discussion (Stasser & Titus, 1985; ally assume that the information they possess is
for reviews see Wittenbaum, Hollingshead, & the same as that possessed by others (unless
Botero, 2004; Wittenbaum & Stasser, 1996). In contrary information is available) (Stasser,
such situations, organizations and teams that Stewart, & Wittenbaum, 1995). The assumption
can create environments where members are is that there is no unique information, and that
willing to share and discuss unique information unmentioned information is information that
may gain considerable competitive advantage. other group members have deemed not of
In this article, we seek to understand more sufficient importance to discuss. When unique
peer-00571629, version 1 - 1 Mar 2011

about how surface-level (i.e. race/ethnicity) information does arise in groups, individuals
and deep-level (i.e. experiences, preferences, are likely to assume that because the infor-
and values) diversity affects the ability of groups mation is not widely held among the group
to benefit from their unique information. We members, it is less important than commonly
move beyond the typical social categorization held information, and therefore may fail to
perspective on diversity and highlight a by- repeat that unique information during dis-
product of the social categorization process— cussion. Moreover, people may feel uncomfort-
assumptions of in-group similarity—which has able expressing and focusing on unique
been overlooked by many researchers in this information, because it is often inconsistent
tradition (cf. Williams & O’Reilly, 1998). We with their perceived expectations that their
extend the argument that surface-level diversity information should be similar to that of other
triggers expectations that informational differ- group members (Gruenfeld, Mannix, Williams,
ences may be present in groups, and legitimates & Neale, 1996; Phillips, Mannix, Neale, &
the expression of unique information (Phillips, Gruenfeld, 2004). Sharing unique information
2003; Phillips & Loyd, 2006; Van Knippenberg, also leads to a lack of social validation from
De Dreu, & Homan, 2004; Van Knippenberg & others, causing individuals to feel less accepted
Haslam, 2003). Moreover, by highlighting than when they share commonly held infor-
deep-level similarities in an effort to increase mation (Wittenbaum & Bowman, 2004).
levels of attraction and diminish social Despite the information sharing barriers in
categorization effects, we argue that managers groups, the discussion of unique information
may undermine the benefits of having surface- has been shown to increase when group
level diversity present in groups that must share members have greater reason to believe that
unique information for effective performance. unique information is going to be present. For
We provide some empirical evidence, while instance, when expertise is labeled, or if people
integrating research on collective information are explicitly forewarned that unique infor-
sampling in groups with that on the effects of mation is present, groups are better able to
group diversity. share and integrate the unique information
into the discussion (Franz & Larson, 2002;
Schittekatte & Van Hiel, 1996; Stasser et al.,
Collective information sampling in
1995; Stewart & Stasser, 1995). To this end,
groups Postmes, Spears and Cihangir (2001) have
Research on information sharing in groups has shown that unique information is more likely to
found that sharing and integrating unique (i.e. be shared when groups have developed a norm

468
 Phillips et al. diversity and performance

of criticality instead of consensus. In such surface-level distinction (e.g. Allen & Wilder,
critical norm groups, the sharing of unique 1975, 1979; Chen & Kenrick, 2002; Diehl, 1988;
information is consistent with the group norms Holtz & Miller, 1985; Phillips, 2003; Phillips &
and the same overvaluing of shared infor- Loyd, 2006; Tajfel, 1969; Wilder, 1984). For
mation is less likely to occur. Thus, for groups instance, Allen and Wilder (1979) divided
to use their unique information effectively, the students into two groups, allegedly on the basis
sharing of such information has to be perceived of their preferences for oil paintings, and found
as a legitimate part of the groups’ norms and greater assumed deep-level similarity between
identity ( Jetten, Postmes, & McAuliffe, 2002; self and similar others than between self and dis-
Postmes, Haslam, & Swaab, 2005; Van Knippen- similar others on opinions about art (category
berg et al., 2004). relevant) and politics (category irrelevant).
We posit that surface-level diversity may also Recent research by Phillips and Loyd (2006)
serve this legitimation purpose in groups by found this same pattern of assumed deep-level
making it more acceptable to discuss and value similarity in decision-making settings when
unique information that must be shared for examining the relationship between salient
effective performance. Recent research has dis- task-relevant (e.g. functional background) (also
peer-00571629, version 1 - 1 Mar 2011

tinguished between diversity in surface-level see Phillips, 2003) and irrelevant (e.g. campus
characteristics, which are immediately salient in geography) social categories and task opinions.
groups (like race and gender), versus diversity Especially at the beginning of a group’s exist-
in deep-level characteristics (like attitudes, ence, when surface-level characteristics are
opinions, information, and values), which most salient (Harrison et al., 1998, 2002), indi-
become known only over time through verbal viduals are likely to use the presence of these
and nonverbal communication (Harrison, characteristics to predict who shares deep-level
Price, & Bell, 1998; Harrison, Price, Gavin, & perspectives with whom. In surface-level homo-
Florey, 2002; Jackson, May, & Whitney, 1995). geneous groups, group members are likely to
We expect surface-level differences to serve as a assume that they all possess the same infor-
signal to group members that unique infor- mation about the task, whereas in surface-level
mation may be present, leading them to be diverse groups members are likely to expect
more aware of and willing to share unique there to be differences in information (Antonio
information with the group. This argument is et al., 2004; Phillips, 2003; Phillips & Loyd,
consistent with recent research on composition 2006; Phillips et al., 2004). Thus, surface-level
beliefs, which has shown that individuals diversity triggers expectations that deep-level
believe diverse groups are likely to outperform diversity will be present in groups, and serves to
homogeneous ones when unique ideas are legitimize the surfacing of unique information.
needed, whereas homogeneous groups are Significantly, this legitimation of unique infor-
likely to outperform diverse ones when com- mation may apply not only to those who are
monality of ideas is needed (van Knippenberg (surface-level) ‘different’ in the group, but also
& Haslam, 2003). Also see the work on the to group members who are similar to most
mechanical and organic solidarity discussed by others. Phillips and Loyd (2006) found that dis-
Postmes et al. (2005) supporting the notion senting members of the social majority voiced
that multiple sources of identity simultaneously themselves more persistently and confidently
exist in groups. when there was diversity present than when
Our argument is based on the fact that a there was not. They concluded that the mere
primary consequence of categorization pro- presence of diversity encouraged those dissent-
cesses is that people assume that they hold ing group members to voice their disparate
more similar deep-level perspectives with indi- perspective when they might otherwise have
viduals who share their surface-level character- remained silent and conformed to the opinion
istics than with people who do not, on topics of their in-group (e.g. Abrams, Wetherell,
both relevant and irrelevant to the salient Cochrane, Hogg, & Turner, 1990; Asch, 1952).

469
 Group Processes & Intergroup Relations 9(4)

For these reasons, surface-level homoge- on workgroups, with some suggesting that
neous groups should be less aware of the increasing the level of perceived deep-level
unique information they possess. As such, over similarity among group members should help
the course of the group discussion they will be them feel more socially validated and accepted
less likely to discuss unique information about by the other members of the group (e.g.
the task than will surface-level diverse groups. Gaertner et al., 2000). Some social categoriz-
This will result in the surface-level homogenous ation researchers have advocated this perspec-
groups spending less time discussing the task tive, suggesting that interventions designed to
than their diverse counterparts. Moreover, minimize the salience of social categories and
spending less time discussing the task will instead ‘de-categorize’ or ‘re-categorize’ group
further hinder the discovery of unique infor- members by highlighting the similarities that
mation (Larson et al., 1996, 1998) leading to a exist across seemingly different individuals can
confirmation of the group members’ expec- be beneficial to group functioning (e.g.
tations that they all have the same information. Gaertner et al., 2000; Northcraft & Martin,
In contrast, in groups where members possess 1982). For instance, Northcraft and Martin
unique information, surface-level diverse (1982) argued that, ‘. . . the liking, acceptance,
peer-00571629, version 1 - 1 Mar 2011

groups are more likely to discover and discuss and perceived competence of tokens and solos
unique information than surface-level hom- can be enhanced by making salient their
ogeneous groups. Surface-level diverse groups similarities to majority group members in
assume that unique information is more likely background, attitudes, and interests’ (p. 114).
to be there, and the presence of informational Further, in a study of corporate outside directors
differences will be more consistent with their of Fortune/Forbes 500 companies, Westphal
expectations. Likewise, if group members are and Milton (2000) found that minority board
aware that they might possess unique infor- members (categorized on the basis of their
mation, they should be inclined to spend more functional background, industry background,
time discussing the task in an effort to discover education, race, or gender) were more influen-
and integrate that information. Thus we tial on their focal boards when they had direct
hypothesize that, or indirect social ties, often through their
common experiences with (focal board)
Hypothesis 1: Surface-level homogeneous groups
will be less aware of their unique information, and majority members on other corporate boards.
will spend less time discussing the task than will This perspective is built on the well-
surface-level diverse groups. established body of findings that similarity
attracts (Byrne, 1971). Individuals generally are
more attracted to and feel more comfortable
Highlighting deep-level similarities interacting with others whom they perceive to
Although surface-level diversity may be ben- be similar. For both surface-level homogeneous
eficial to teams or work groups that must share and surface-level diverse groups, an interven-
unique information for effective performance, tion designed to help group members discover
diversity researchers have often found that their deep-level similarities should lead to
diversity has a negative impact on communi- greater feelings of attraction. Learning that one
cation and cohesion, and promotes high levels shares deep-level similarities with a fellow group
of detrimental group conflict (Ely & Thomas, member should also promote recategorization,
2001; Jackson, Joshi & Erhardt, 2003; Jehn, increasing the likelihood that out-group
Northcraft, & Neale, 1999; Pelled, Eisenhardt, members (i.e. those who have surface-level dis-
& Xin, 1999; for extensive reviews see also similarities from other group members) will
Milliken & Martins, 1996; Williams & O’Reilly, actually be seen as part of the in-group (e.g.
1998). Over the past ten years, diversity Gaertner, Mann, Murrell, & Dovidio, 1989;
researchers have focused on how to minimize Kramer & Brewer, 1984). The potential
the detrimental effects of social categorization negative effects of social categorization may,

470
 Phillips et al. diversity and performance

subsequently, be reduced. As a result, we suggest on the surface-level characteristic of race finds

that: that they all share the same attitudes, feelings,
and experiences about the organization, they
Hypothesis 2A: Members of groups who learn may be reluctant to disagree with each other
about deep-level similarities should perceive greater
attraction from other group members than those
going forward with the task. They may feel that
who do not learn about deep-level similarities. they really are not all that different from each
other after all. Thus, for surface-level diverse
However, recent research has suggested that groups, although the realization of deep-level
increasing this perceived similarity and attrac- similarities may increase attraction toward the
tion among the group members may come at a group, it may simultaneously increase pressures
cost to the group’s ability to benefit from the to conform to the group and undermine the
surface-level differences we have discussed here discussion of unique information (Abrams
(Hornsey & Jetten, 2004; Jetten et al., 2002; et al., 1990).
Postmes et al., 2001; van Knippenberg et al., For surface-level homogeneous groups high-
2004). For instance, Postmes et al. (2002) lighting similarities will also increase attraction,
found that a focus on agreement and common- and may further interfere with the sharing of
peer-00571629, version 1 - 1 Mar 2011

alities created norms of consensus that in turn unique information since doing so poses a
undermined the sharing of unique information threat to feelings of acceptance and validation
in groups. The effectiveness of recategorization (Wittenbaum & Bowman, 2004; Wittenbaum,
or promoting the perception of others as Hubbell, & Zuckerman, 1999). Thus, we believe
similar to oneself as a means to diminish the that for surface-level homogeneous groups
detrimental effects of surface-level diversity has there will be somewhat of a ‘floor’ effect—the
also been called into question by researchers of lack of surface-level differences will hinder the
self-verification (e.g. Polzer, Milton, & Swann, expectation of informational differences and
2002; Swann, Milton, & Polzer, 2000). These the discussion of unique information, and then
researchers argue that promoting the percep- the highlighting of deep-level similarities will
tion (or recategorization) of all group members further hinder this process. Thus, we hypothe-
as similar, may also discourage individuals from size that:
thinking and acting in ways associated with their
unique category memberships (Gaertner et al., Hypothesis 2B: Highlighting deep-level similarities
will lead to less awareness of unique information
1989). Yet, it is precisely these unique ways of and less discussion time.
thinking and acting that constitute the potential
positive contribution of a diverse workgroup Hypothesis 2C: The effect of highlighting deep-level
similarities on awareness of unique information
(Polzer et al., 2002, p. 297). Thus, highlighting
and discussion time will be more pronounced for
deep-level similarities (especially in surface-level surface-level diverse groups than for surface-level
diverse groups), while leading to greater feelings homogeneous groups.
of attraction toward the other group members,
may undermine the groups’ awareness of and In terms of group performance, highlighting
willingness to embrace unique information. deep-level similarities should be detrimental
In sum, the process of highlighting deep- because it undermines the legitimacy of dis-
level similarities may undermine the signaling cussing needed unique information. Although
effect of surface-level diversity that legitimates attraction may increase in groups as a result of
expressing and discussing unique information learning about deep-level similarities, these
by all group members. In other words, learning deep-level similarities will be inconsistent with
about deep-level similarities in surface-level any expectation of unique information being
diverse groups may interfere with the legitima- present and thus are likely to hurt group per-
tion of difference and disagreement that formance, especially for surface-level diverse
surface-level diversity promotes. For example, if groups (e.g. Watson, Kumar, & Michaelsen,
a work group who thinks they are diverse based 1993). As such, we would argue that:

471
 Group Processes & Intergroup Relations 9(4)

Hypothesis 3A: Surface-level diverse groups will vs. control) between-subjects design and all
outperform surface-level homogeneous groups. analyses were conducted at the group level.
Hypothesis 3B: Groups that highlight deep-level There were a total of 31 surface-level homoge-
similarities will perform worse than groups that do neous (i.e. three Caucasian group members)
not. and 41 surface-level diverse groups (i.e. two
Hypothesis 3C: The performance of surface-level Caucasian and one Asian, African American, or
diverse groups that highlight deep-level similarities Hispanic). Forty groups were in the deep-level
will be more negatively affected than the perform- similarities condition and 32 were in the
ance of surface-level homogeneous groups that control condition. Participants were thoroughly
highlight deep-level similarities. debriefed about the purpose of the study after
participation.
To test the hypotheses, groups received an
intervention asking them to discover deep-level Materials
similarities among group members. Exploring Every participant was given a packet of evidence
the effects of this intervention should allow for from an apparent homicide investigation. The
a better understanding of how different groups evidence consisted of interviews and a variety of
peer-00571629, version 1 - 1 Mar 2011

(i.e. those that are surface-level diverse or homo- supporting materials, including a list of
geneous) are affected by finding that they have suspects, a map, a personal note, and a newspa-
deep-level similarities prior to engaging in a task per article. All of these materials were adapted
where unique information must be shared. from Stasser and Stewart’s (1992) study. Within
each group, every member received the list of
suspects, transcripts of initial interviews with
Method each of the four key suspects, the newspaper
Participants and overview article, and maps of the crime scene and sur-
A total of 216 undergraduate business students rounding area. The materials contained 42
at a midwestern business school participated in clues in all, 12 of which were critical for solving
this research. The median age of the partici- the case. All participants received 30 commonly
pants was 21 years, and approximately 42% of shared clues; 12 critical clues were distributed
the sample was female. The students partici- among the three group members such that
pated in a class exercise designed to provide each group member held some unique infor-
insight into group decision-making. Partici- mation pertinent to identifying the guilty
pants were given extra course credit for their suspect. These clues were embedded in follow-
participation, and a few of the best performing up interviews with the key suspects, and inter-
groups on the prediscussion task were randomly views with some additional witnesses. In all of
selected and entered into a raffle to receive cer- the groups, a hidden profile existed because
tificates for free meals at local restaurants. the best solution was more likely to be found if
Participants first made individual assessments the unique information represented by the 12
about who they believed committed a murder unshared critical clues was shared.
(see Stasser & Stewart, 1992 for materials)
based on the investigating detective’s reports, Procedure
and then discussed the case in three-person When participants arrived at the laboratory or
groups. Groups were either all male (42) or all in the classroom, they were randomly assigned
female (30); the gender composition of the to three-person groups based on their visible
group is included as a covariate in all analyses racial characteristics, with the constraint that all
presented. Eighty-two percent of the partici- three members had to be of the same gender.
pants were white, 14% Asian, and the remain- In some groups the three group members
der African American or Hispanic. We used a 2 appeared to all be Caucasian (surface-level
(surface-level homogeneous vs. surface-level homogeneous groups), and in other groups
diverse)
2 (deep-level similarities highlighted two of the members appeared Caucasian and

472
 Phillips et al. diversity and performance

one was of a different race (Asian, African Control group members did not work together
American, or Hispanic). At the end of partici- on the state capitals task in order to prevent the
pation, group members provided their self- inadvertent sharing of similarity information
identified race on the post-task questionnaire, while working together on the task. (A follow-
and the experimenters’ group assignments were up data collection revealed that working
validated against the information provided by together to identify similarities increased group
the students. members’ perceptions of similarity while
During the study, participants were first given working together on the state capitals task did
approximately 20 minutes to read and review not).1 In both conditions, participants were
materials in order to determine which of four informed that the group during that session
suspects committed a murder (materials drawn that generated the longest list (of similarities or
from Stasser & Stewart, 1992 and also used by state capitals) could receive a prize (certificates
Gruenfeld et al., 1996; Liljenquist, Galinsky, & for free meals at a local restaurant).
Kray, 2004; Phillips et al., 2004). Individuals After completing the 5-minute task, group
were instructed to take notes on the case while members individually completed a 20-item
reading the materials, because they would not survey titled ‘Personal Assessment Inventory’,
peer-00571629, version 1 - 1 Mar 2011

be allowed to keep the case materials during by circling the number that best indicated how
the group discussion portion of the exercise. much they agreed with each of the statements
Participants made an individual assessment of on a scale from 1 = strongly disagree to 7 =
who they believed was more likely to have com- strongly agree. 54 groups completed the survey.
mitted the murder. They were asked to indicate Embedded in the items were questions assess-
how confident they were that each of the four ing how similar group members felt to one
suspects did or did not commit the murder. another and how attracted they thought the
Subjects were also asked to provide a brief, group would be to them (adapted from Eagly,
written justification for their decision. 1967).
Participants were then gathered into their Once the short exercise and questionnaire
assigned three-person groups, and were were completed, participants were instructed to
instructed that they had 5 minutes to complete come to a group decision regarding which
a short exercise before beginning group dis- suspect they believed was most likely to have
cussion. Groups were separated so that they committed the murder. They were informed
could not overhear the deliberations of other that they would have to report their group
groups. Half of the groups were randomly decision in exactly the same way that they had
assigned to identify their similarities with the reported their individual decisions. Groups
following instructions: were given up to 35 minutes to complete their
discussions, which were timed by the experi-
Working together with the other two members of
your group, you have 5 minutes to discover as many menter. Finally, upon completion of the group
things as possible that the three of you have in discussion, participants filled out an individual
common. You may include anything that you have post-task questionnaire where they recorded
in common: friends, experiences, hobbies, books or the group decision in terms of how confident
movies that you liked, places to which you have all they were that each of the four possible suspects
traveled, places where you all have lived or visited— committed the murder. All participants then
anything that the three of you have in common. Put
as many items on your list as possible.
answered several questions assessing their
group’s task performance and management of
The control groups were given the following information, as well as reporting their individ-
instructions: ual demographic characteristics.
Group task performance was operationalized
Working by yourself, you have 5 minutes to list as
many US states and their capitals as you can as how confident the group was that the correct
remember. Put as many states and their capitals on (guilty) suspect committed the murder. For
your list as possible. instance, groups could report that they were

473
 Group Processes & Intergroup Relations 9(4)

0% to 100% certain that the correct suspect level similarities listed into two categories:
committed the murder. Group members characteristics that were more obvious because
responded individually, and individual of the nature of the groups and participants
responses were aggregated to the group level as being used, versus less obvious characteristics
in most groups the individual responses were that took more discussion and interaction to
exactly the same (one way analysis of variance learn.
(ANOVA) for group membership (F(71, 143) = Obvious characteristics were things that were
9.35, intraclass correlation coefficient (ICC(1)) clear to all of the group members and poten-
= .74). To illustrate that our results are indeed tially similar among many participants in the
a reflection of group accuracy on this task, and study, such as ‘all members had been in (a
not just higher levels of overall confidence in major building on campus)’, that all students
the groups’ assessments, we also examined how must enter for class registration and other
confident groups were that each of the other activities. Other similarities characterized as
three (innocent) suspects committed the obvious included items such as group members
murder. were all wearing shoes; group members all
have hair, etc. Less obvious characteristics were
peer-00571629, version 1 - 1 Mar 2011

Dependent variables described as those that could not be discerned

To measure group members’ awareness about by the group members from simply looking at
whether they possessed unique information, the group members or knowing that they were
group members were asked to respond to the students at the university, i.e. similarities that
following item: ‘The information in my packet could not have been a matter of common
was the same as that in others’ packets’, with 1 knowledge. These less obvious similarities
labeled ‘not at all the same’ and 7 labeled between group members included personal
‘exactly the same’ (Gruenfeld et al., 1996; likes and dislikes (such as all members liked
Phillips et al., 2004). Group discussion time was certain foods or certain restaurants), similari-
measured in minutes and seconds with an ties in places that individuals had visited, and
average of 20.11 minutes (SD = 6.67). Likewise, similarities in nonobvious demographic charac-
perceptions of similarity were measured on the teristics (such as the number of siblings). Thus,
post-manipulation questionnaire with the group members had to be explicit in divulging
following item ‘I feel similar to the other nonobvious information about themselves to
members of my group’. How attracted group discover hidden interpersonal similarities.
members thought others were to them was The two raters were trained together on
captured by four items measured on a 7-point about 10% of the groups (eight groups). The
Likert scale; ‘I think I will like being a member 162 similarities listed in those eight groups were
of this group’, ‘I think the others in the group coded in the same category 93.2% of the time.
will like me’, ‘I am fun to be with’, and ‘I am This gave us confidence to allow the two coders
easy to like’ ( = .87). The questions were asked to proceed independently with coding the rest
from the perspective of the participant because of the groups. One coder independently coded
the internal feelings of the individual, even if 39 of the 72 groups and the other coder did the
others in the group were actually more or less 25 remaining groups.
attracted, were thought to be the critical assess-
ment leading to how socially validated group
members would feel during the group dis-
Results
cussion. To test our hypotheses regarding the impact of
To examine the number and content of the highlighting deep-level similarities in surface-
similarities listed by groups, two independent level homogeneous and diverse groups, we
raters who were blind to the hypotheses first used a 2 (surface-level homogeneous or
counted the number of similarities listed by surface-level diverse)
2 (deep-level similarity
each group. They then categorized the deep- intervention or control) analysis of covariance

474
 Phillips et al. diversity and performance

(ANCOVA) approach to analyze the data. type of similarities during the similarity inter-
Gender composition of the group (male or vention.
female) was a covariate in the analyses. We also examined how similar group
All analyses were conducted at the group members felt to one another right after com-
level. We tested the validity of aggregating the pleting the manipulation, but before beginning
dependent variables of perceived uniqueness of discussion of the task. As expected, group
information, feelings of similarity, and per- members felt more similar to one another after
ceived attraction to the group level in two steps highlighting deep level similarities (M = 4.71,
(see Bliese, 2000). First, we ran a one-way SD = 0.80) than in the control conditions (M =
analysis of variance with group membership as 4.49, SD = 0.76) (F(1, 49) = 3.92, p = .05).
the independent variable to ensure that the Notably, there was a marginally significant inter-
variance between teams was greater than the action effect (F(1, 49) = 2.86, p = .097), suggest-
variance within teams. In all cases, the ANOVAs ing that deep-level similarities increased
were significant (F(71, 143) = 2.15, p < .001; feelings of similarity more for surface-level
F(53, 108) = 1.50, p < .04; F(53, 108) = 1.57, homogeneous groups than for surface-level
p < .03). We then calculated ICC(1) values for diverse groups (see Table 2 for means).
peer-00571629, version 1 - 1 Mar 2011

each of the dependent variables to test how

much variability in individual responses was due Hypotheses tests
to group membership (Bliese, 2000; Klein & In Hypothesis 1, we predicted that surface-level
Kozlowski, 2000). The ICC(1) values were 0.28, homogeneous groups would report discovering
0.15, and 0.16 respectively, suggesting that less unique information and would discuss the
there was significant interdependence in indi- task less than would surface-level diverse
vidual responses and that group membership groups. We conducted a multivariate analysis of
accounted for a reasonable proportion of variance analysis and found support for our
variance in individual responses on these three hypothesis (F(2, 63) = 3.14, p < .05). The follow-
dependent variables (Bliese, 2000; James, up univariate tests revealed that surface-level
1982).2 homogeneous groups reported that the infor-
mation in their packets was more similar (M =
Manipulation checks 3.31, SD = 1.18) than did surface-level diverse
The composition of the groups was validated groups (M = 2.78, SD = 0.80) (F(1, 67) = 4.39,
against the self-reported race of each of the p < .05). Likewise, we found that surface-level
participants. This revealed that there were two homogeneous groups (M = 17.90, SD = 7.54)
groups initially categorized by the experimenter spent less time discussing the task than surface-
as racially homogeneous that self-identified as level diverse groups (M = 21.72, SD = 5.52) (F(1,
racially diverse. These groups were moved to the 64) = 5.29, p < .03). Table 1, which includes the
appropriate category before analysis. correlations among all of the variables reported
To check on the outcomes of the similarity- here, shows that discussion time and awareness
manipulation, we first examined the number of of unique information were significantly corre-
similarities listed by surface-level homogeneous lated at r = –.46, p < .01. Controlling for the
and diverse groups. We found no significant effects of discussion time on the awareness of
differences in the number of similarities listed unique information, the effect of surface-level
by the two types of groups (M = 18.84, SD = group composition on awareness of unique
7.98). Likewise, there were no significant differ- information was no longer significant (F < 1.0,
ences in the types of similarities listed by the p > .30), suggesting that discussion time
groups with an average of 59% (SD = 18%) mediated the effects of surface-level composi-
being categorized as nonobvious by the coders. tion on the discovery of unique information.
These results suggest that surface-level homoge- Hypothesis 2A predicted that learning about
neous and surface-level diverse groups naturally nontask relevant deep-level similarities would
generated approximately the same number and increase perceived attraction among the group.

475
 Group Processes & Intergroup Relations 9(4)

Table 1. Means, standard deviations, and correlations for study variables (N = 72)

Mean (SD) 1 2 3 4 5 6 7

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

Analysis revealed a marginally significant main action between surface-level diversity and the
effect for the similarity manipulation (F(1, 49) deep-level similarity intervention (F(1, 66) =
= 3.00, p < .10); groups exposed to the similarity 6.29, p < .02). As predicted in Hypothesis 3c, we
intervention reported greater feelings of attrac- found that surface-level diverse groups per-
tion (M = 5.29, SD = 0.54) than those in the formed worse after being exposed to the deep-
control condition (M = 5.15, SD = 0.54). This level similarity manipulation (M = 61.00, SD =
main effect was accompanied by a significant 28.69) than when they were not (M = 72.79, SD
and unexpected interaction effect (F(1, 49) = = 22.41) (F(1, 37) = 4.20, p < .05). In contrast
5.38, p < .03). While surface-level homogeneous and contrary to our expectations, surface-level
groups were more attracted to one another homogeneous groups seemed to perform
after the deep-level similarity manipulation, this better after experiencing the similarity inter-
was not the case for surface-level diverse groups vention (M = 67.89, SD = 31.68) than when they
(see Table 2 for means). Neither Hypotheses 2B did not (M = 46.35, SD = 29.33), although this
(that there would be a main effect of the deep- difference did not reach significance (F(1, 27)
level similarity manipulation on awareness of = 2.34, p = .138). (Both individual decisions and
unique information and discussion time) nor gender composition were controlled for in
Hypothesis 2C (that the impact of surface-level these follow-up analyses.) Additional analyses
composition would interact with the similarity on how confident groups were that each of the
intervention to affect awareness of unique three incorrect suspects committed the murder
information and discussion time) was sup- revealed no significant effects. This rules out
ported (Fs < 1, ps = ns). the possibility that some conditions just made
Group performance was analyzed using an groups more confident than others did.
ANCOVA analysis with two covariates—gender Finally, Table 1 includes correlations of all
composition and the average individual confi- variables reported in the results. Of note, the
dence about the a priori correct suspect prior correlations show that group composition was
to group discussion. Although the means were related to how long groups discussed the task
in the expected direction (M = 65.60, SD = (r = –.28, p < .02), such that surface-level homo-
26.76 and M = 56.77, SD = 31.91 for surface- geneous groups discussed the task for a shorter
level diverse and homogeneous groups respec- period of time; and group discussion time was
tively), the main effects for surface-level positively related to group performance—the
diversity and the deep-level similarity manipu- group’s confidence that the correct suspect
lation were not supported (Fs < 1, ps = ns). committed the murder (r = .26, p < .04).
Further analysis did reveal a significant inter- Moreover, group discussion time, awareness of

476
 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)a 5.14 (0.69)ab 4.60 (0.60) 4.44 (0.75)b
Perceived attraction 5.00 (0.54)a 5.56 (0.49)ab 5.34 (0.51) 5.12 (0.51)b

Note: Means sharing subscripts within each row differ from one another significantly at p < .05.

unique information, and group performance perspective we found that surface-level diverse
were all significantly correlated (r s > .26, ps < groups perceived their information to be less
.05), meaning that groups performed better similar, and spent more time discussing the task
when they spent more time in discussion and, than did surface-level homogeneous groups.
peer-00571629, version 1 - 1 Mar 2011

in turn, became more aware of the differences Moreover, the more time group members spent
in information present among group members. discussing the task, the more likely they were to
perceive the information as different. These
results were strong, expected, and independent
Discussion of whether groups learned about deep-level
This article reports how different types of diver- similarities.
sity—surface- and deep-level—influence group As argued and shown by past research,
decision-making and performance. Under- learning about deep-level similarities had a
standing the impact of diversity on decision- positive impact on how similar and how
making groups is often complicated by the attracted group members felt toward one
different sources of diversity that can exist and another (Byrne, 1971), but mostly for the
the assumed correlation among these sources. surface-level homogeneous groups. The simi-
In this study, we moved beyond the typical larity manipulation did not reliably increase
social categorization perspective on diversity by perceived attraction for the surface-level
highlighting a by-product of the social diverse groups, even though our manipulation
categorization process—assumptions of in- check revealed that the number and type of
group similarity—which has been overlooked similarities generated by the two types of groups
by researchers in this tradition. We argue that did not differ. Moreover, we did not find
surface-level diversity triggers expectations that evidence that the deep-level similarities had a
informational differences may be present in differential impact on surface-level homoge-
groups, making it more expected and legiti- neous and diverse groups’ awareness of unique
mate for group members to raise and discuss information or discussion time. There was only
unique information that may be critical for the significant interaction on group perform-
group performance. Moreover, by highlighting ance showing that surface-level diverse groups
deep-level similarities in an effort to increase performed worse after being exposed to the
levels of attraction and diminish social deep-level similarity manipulation. In contrast,
categorization effects, managers may under- surface-level homogeneous groups seemed to
mine the benefits of having surface-level diver- be helped by the intervention, although not
sity present in groups that must share unique significantly so. Overall, the general arguments
information for effective performance. made in this article were supported with some
While our results revealed a substantial exceptions. A closer look at the deep-level simi-
amount of support for this perspective, there larity manipulation may lend some insight into
were some surprises. First, in support of this future research opportunities.

477
 Group Processes & Intergroup Relations 9(4)

Group members were instructed to spend groups (e.g. Abrams et al., 1990; Janis, 1982;
five minutes discovering as many things as Phillips & Loyd, 2006). Postmes et al. (2001)
possible that all three of the group members suggest that one way to overcome the reluc-
had in common. This meant that in the process tance to share unique information is to develop
of finding these similarities, groups were also norms of criticality instead of consensus. This
likely to discover similarities that bonded only may be even more important for surface-level
subsets of the members, and differences where homogeneous than for surface-level diverse
none of the group members were similar. groups. Moreover, homogeneity without feelings
Without a measure of these partially shared sim- of attraction and acceptance may be detrimen-
ilarities or the differences that surfaced, it is dif- tal to groups when sharing unique information
ficult to determine exactly how they might have is crucial for performance (Gruenfeld et al.,
affected our results here. However, given that 1996; Jehn & Shah, 1997). There is a fine line
all groups had the same amount of time for the that surface-level homogeneous groups need to
exercise and that there were no differences in walk. Feeling too similar to one another may
the number of similarities discovered, it is likely undermine the amount of time group members
that there were no systematic differences across spend on the task, as suggested by the correla-
peer-00571629, version 1 - 1 Mar 2011

the conditions on the number of differences tion found here between feelings of similarity
found either. Future research should disentan- and discussion time, but not feeling similar
gle the effects of highlighting similarities from enough may lead group members to feel
the effects of discovering differences, and the insecure and have concern about sharing
implications of doing this in light of a group’s unique information at all.
surface-level composition. Future research should attempt to disentan-
Sharing unique information is difficult in gle the effects of similarity and familiarity on
groups, because sharing the information often work teams. For instance, does familiarity have
means that one has to disagree with the rest of the same effect on surface-level homogeneous
the group about what is important for making as it does on diverse groups? How familiar do
the group decision. Providing the context that people need to be with one another before the
allows groups to benefit fully from the unique potential benefits of familiarity outweigh the
information and perspectives that each group potential downsides of similarity/diversity?
member holds is a difficult task. This research From the study conducted here, it seems that
suggests that surface-level diversity might be increasing familiarity (e.g. via 5 minutes of dis-
beneficial for groups because of the legitimacy cussing similarities) can only slightly overcome
it lends to sharing unique perspectives. For the conformity pressures that come along with
surface-level homogeneous groups the obvious surface-level homogeneity. However, it may take
similarities among group members hindered considerably longer for surface-level diverse
the recognition of unique information and led groups to benefit from familiarity (Watson et
them to perform worse. In fact, the surface- al., 1993). The amount of diversity present in
level homogeneous groups in the control con- our groups, and the small group size, may also
dition performed worse than any others in our be limitations that should be considered in
sample. For these groups, sharing unique infor- future research. Would the diverse groups have
mation may be particularly difficult because of performed even better if they had had greater
the social risks involved (e.g. Gruenfeld et al., surface-level diversity? Would the results be the
1996; Wittenbaum et al., 1999). same if a different type of surface-level diversity
Surface-level homogeneous groups, where were used?
group members are concerned about being In conclusion, as organizations attempt to
accepted by their fellow in-group members, cope with the changing demography of the
suffer more from conformity pressures that work force there is a natural tendency to believe
prevent them from sharing unique information that what enhances the performance of surface-
and opinions than do surface-level diverse level homogeneous groups may also enhance

478
 Phillips et al. diversity and performance

the performance of surface-level diverse performance after working together to highlight

groups. The current research suggests that deep-level similarities if simply working together
enhancing the performance of workgroups is is all that is important. As this was not the case we
more complex than that. Our research found believe that our current manipulation of
that attempting to diminish the salience of highlighting similarities does indeed add a
unique contribution to people’s experience,
potentially disruptive categorical boundaries by
beyond just working together.
asking members of racially diverse groups to 2. Bliese (2000, p. 361) reports that values between
focus on their similarities was detrimental for .05 and .30 should be expected in most applied
group performance. Diversity can be beneficial field settings.
for groups, not merely because individuals
belonging to different subgroups are likely to
have access to differing information, but also Acknowledgments
because the presence of salient differences may The authors are indebted to the graduate students
legitimate the introduction and consideration who assisted in the data collection efforts. Also, we
of unique information in the group’s decision- thank KTAG for financial support and the members
making process. Both the positive and negative of the Social Interactions Lab at Northwestern
peer-00571629, version 1 - 1 Mar 2011

effects of surface-level and deep-level diversity University for their helpful comments. We thank
in the workplace are yet to be fully understood. Elson Huang and Tanya Canak for coding help. In
addition, I would like to thank the audiences at
This work is another step toward understanding
several colloquiums who offered their insights on
the effects of diversity on groups where sharing this work.
unique information is crucial to performance.

References
Notes Abrams, D. M., Wetherell, M., Cochrane, S., Hogg,
1. We conducted a follow-up study with 12 M. A., & Turner, J. C. (1990). Knowing what to
three-person racially homogeneous groups that think by knowing who you are: Self-categorization
either worked together to identify US states and and the nature of norm formation, conformity
capitals or worked together to identify similarities. and group polarization. British Journal of Social
Results revealed that identifying similarities does Psychology, 29, 97–119.
indeed lead individuals to feel more similar to Allen, V. L., & Wilder, D. A. (1975). Categorization,
one another than just working together recalling beliefs similarity, and intergroup discrimination.
US states and capitals (F(1, 10) = 10.48, p < .01). Journal of Personality and Social Psychology, 32,
Moreover, a comparison of feelings of similarity 971–977.
by the groups that worked together to identify US Allen, V. L., & Wilder, D. A. (1979). Group
states and capitals (M = 4.71, SD = .56) and those categorization and attribution of belief similarity.
that worked separately when they identified states Small Group Behavior, 10, 73–80.
and capitals (from the initial study) (M = 4.41, SD Antonio, A. L., Chang, M. J., Hakuta, K., Kenny,
= .88) revealed that there was no significant D. A., Levin, S., & Milem, J. F. (2004). Effects of
difference between the two types of groups racial diversity on complex thinking in college
(t(18) = .821, p = .423). Theoretically, we do not students. Psychological Science, 15, 507–510.
believe that just working together (without Asch, S. E. (1952). Social psychology. Englewood Cliffs,
discussing similarities) could generate our same NJ: Prentice-Hall.
pattern of data. The highlighting of similarities in Bliese, P. D. (2000). Within-group agreement,
surface-level diverse groups causes a particular non-independence, and reliability: Implication for
threat to expectations that just working together data aggregation and analysis. In K. J. Klein &
would not. For instance, in the work by S. W. J. Kozlowski (Eds.), Multilevel theory, research,
Gruenfeld et al. (1996) familiarity was found to and methods in organizations: Foundations, extensions,
lead to better performance when unique and new directions (pp. 349–381). San Francisco:
information needed to be shared. This suggests Jossey-Bass.
that both the surface-level homogeneous and Byrne, D. (1971). The attraction paradigm. New York:
diverse groups should have improved their Academic Press.

479
 Group Processes & Intergroup Relations 9(4)

Chen, F. F., & Kenrick, D. T. (2002). Repulsion or Recent research on team and organizational
attraction? Group membership and assumed diversity: SWOT analysis and implications. Journal
attitude similarity. Journal of Personality and Social of Management, 29, 801–830.
Psychology, 83, 111–125. Jackson, S. E., May, K. E., & Whitney, K. (1995).
Diehl, M. (1988). Social identity and minimal Understanding the dynamics of diversity in
groups: The effects of interpersonal and decision-making teams. In R. A. Guzzo & E. Salas
intergroup attitudinal similarity on intergroup (Eds.), Team decision-making effectiveness in
discrimination. British Journal of Social Psychology, organizations (pp. 204–261). San Francisco:
27, 289–300. Jossey-Bass.
Eagly, A. H. (1967). Involvement as a determinant of James, L. R. (1982). Aggregation bias in estimates of
response to favorable and unfavorable perceptual agreement. Journal of Applied Psychology,
information. Journal of Personality and Social 67, 219–229.
Psychology, 7, 15. Janis, I. L. (1982). Victims of groupthink (2nd ed).
Ely, R., & Thomas, D. (2001). Cultural diversity at Boston: Houghton Mifflin.
work: The effects of diversity perspectives on work Jehn, K., Northcraft, G., & Neale, M. (1999). Why
group processes and outcomes. Administrative differences make a difference: A field study of
Science Quarterly, 46, 229–273. diversity, conflict, and performance in work
peer-00571629, version 1 - 1 Mar 2011

Franz, T. M., & Larson, J. R., Jr. (2002). The impact groups. Administrative Science Quarterly, 44,
of experts on information sharing during group 741–763.
discussion. Small Group Research, 33, 383–411. Jehn, K. A., & Shah, P. P. (1997). Interpersonal
Gaertner, S., L., Dovidio, J. F., Banker, B. S., relationships and task performance: An
Houlette, M., Johnson, K. M., & McGlynn, E. A. examination of mediation processes in friendship
(2000). Reducing intergroup conflict: From and acquaintance groups. Journal of Personality and
superordinate goals to decategorization, Social Psychology, 72, 775–790.
recategorization, and mutual differentiation. Jetten, J., Postmes, T., & McAuliffe, B. J. (2002).
Group Dynamics, 4, 98–114. ‘We’re all individuals’: Group norms of
Gaertner, S. L., Mann, J., Murrell, A., & Dovidio, J. individualism and collectivism, levels of
(1989). Reducing intergroup bias: The benefits of identification and identity threat. European Journal
recategorization. Journal of Personality and Social of Social Psychology, 32, 189–207.
Psychology, 57, 239–249. Klein, K. J., & Kozlowski, S. W. J. (2000). From micro
Gruenfeld, D. H., Mannix, E. A., Williams, K. Y., & to meso: Critical steps in conceptualizing and
Neale, M. A. (1996). Group composition and conducting multilevel research. Organizational
decision making: How member familiarity and Research Methods, 3, 211–236.
information distribution affect process and Kramer, R., & Brewer, M. B. (1984). Effects of group
performance. Organizational Behavior and Human identity on resource use in a simulated commons
Decision Processes, 67, 1–15. dilemma. Journal of Personality and Social Psychology,
Harrison, D. A., Price, K. H., & Bell, M. P. (1998). 46, 1044–1057.
Beyond relational demography: Time and effects Larson, J. R., Jr., Christensen, C., Abbott, A. S., &
of surface- and deep-level diversity on work group Franz, T. M. (1996). Diagnosing groups: Charting
cohesion. Academy of Management Journal, 41, the flow of information in medical decision-making
96–107. teams. Journal of Personality and Social Psychology, 71,
Harrison, D. A., Price, K. H., Gavin, J. A., & Florey, 315–330.
A. T. (2002). Time, teams, and task performance: Larson, J. R., Jr., Christensen, C., Abbott, A. S., &
Changing effects of surface- and deep-level Franz, T. M. (1998). Diagnosing groups: The
diversity on group functioning. Academy of pooling, management, and impact of shared and
Management Journal, 45, 1029–1045. unshared case information in team-based medical
Holtz, R., & Miller, N. (1985). Assumed similarity decision-making. Journal of Personality and Social
and opinion certainty. Journal of Personality and Psychology, 75, 93–108.
Social Psychology, 48, 890–898. Liljenquist, K. A., Galinsky, A. D., & Kray, L. J.
Hornsey, M. J., & Jetten, J. (2004). The individual (2004). Exploring the rabbit hole of possibilities
within the group: Balancing the need to belong by myself or with my group: The benefits and
with the need to be different. Personality and Social liabilities of activating counterfactual mind-sets for
Psychology Review, 8, 248–264. information sharing and group coordination.
Jackson, S. E., Joshi, A., & Erhardt, N. L. (2003). Journal of Behavioral Decision Making, 17, 263–279.

480
 Phillips et al. diversity and performance

Milliken, F. J., & Martins, L. L. (1996). Searching for problem versus making a judgment. Journal of
common threads: Understanding the multiple Personality and Social Psychology, 63, 426–434.
effects of diversity in organizational groups. Stasser, G., Stewart, D. D., & Wittenbaum, G. M.
Academy of Management Review, 21, 402–433. (1995). Expert roles and information exchange
Northcraft, G. B., & Martin, J. (1982). Double during discussion: The importance of knowing
jeopardy: Resistance to affirmative action from who knows what. Journal of Experimental Social
potential beneficiaries. In B. Gutek (Ed.), Sex role Psychology, 31, 244–265.
stereotyping and affirmative action policy Stasser, G., & Titus, W. (1985). Pooling of unshared
(pp. 81–130). Los Angeles, CA: Institute of information in group decision making: Biased
Industrial Relations, UCLA. information sampling during discussion. Journal of
Pelled, L. H., Eisenhardt, K. M., & Xin, K. R. (1999). Personality and Social Psychology, 48, 1467–1478.
Exploring the black box: An analysis of work Stasser, G., & Titus, W. (1987). Effects of information
group diversity, conflict, and performance. load and percentage of shared information on the
Administrative Science Quarterly, 44, 1–28. dissemination of unshared information during
Phillips, K. W. (2003). The effects of categorically group discussion. Journal of Personality and Social
based expectations on minority influence: The Psychology, 53, 81–93.
importance of congruence. Personality and Social Stasser, G., Vaughan, S. I., & Stewart, D. D. (2000).
peer-00571629, version 1 - 1 Mar 2011

Psychology Bulletin, 29, 3–13. Pooling unshared information: The benefits of

Phillips, K. W., & Loyd, D. L. (2006). When surface knowing how access to information is distributed
and deep-level diversity collide: The effects on among group members. Organizational Behavior
dissenting group members. Organizational Behavior and Human Decision Processes, 82, 102–116.
and Human Decision Processes, 99, 143–160. Stewart, D. D., & Stasser, G. (1995). Expert role
Phillips, K. W., Mannix, E., Neale, M., & assignment and information sampling during
Gruenfeld, D. (2004). Diverse groups and collective recall and decision-making. Journal of
information sharing: The effects of congruent ties. Personality and Social Psychology, 69, 619–628.
Journal of Experimental Social Psychology, 40, Swann, W. B., Jr., Milton, L. P., & Polzer, J. T. (2000).
498–510. Should we create a niche or fall in line? Identity
Polzer, J. T., Milton, L. P., & Swann, W. B. Jr. (2002). negotiation and small group effectiveness. Journal
Capitalizing on diversity: Interpersonal of Personality and Social Psychology, 79, 238–250.
congruence in small work groups. Administrative Tajfel, H. (1969). Cognitive aspects of prejudice.
Science Quarterly, 47, 296–324. Journal of Social Issues, 25, 79–97.
Postmes, T., Haslam, A., & Swaab, R. I. (2005). Social Van Knippenberg, D., De Dreu, C. K. W., & Homan,
influence in small groups: An interactive model of A. C. (2004). Work group diversity and group
social identity formation. European Review of Social performance: An integrative model and research
Psychology, 16, 1–42. agenda. Journal of Applied Psychology, 89,
Postmes, T., Spears, R., & Cihangir, S. (2001). 1008–1022.
Quality of decision making and group norms. Van Knippenberg, D., & Haslam, S. A. (2003).
Journal of Personality and Social Psychology, 80, Realizing the diversity dividend: Exploring the
918–930. subtle interplay between identity, ideology and
Schittekatte, M., & Van Hiel, A. (1996). Effects of reality. In S. A. Haslam, D. van Knippenberg,
partially shared information and awareness of M. Platow, & N. Ellemers (Eds.), Social identity at
unshared information on information sampling. work: Developing theory for organizational practice
Small Group Research, 27, 431–449. (pp. 61–77). New York: Taylor & Francis.
Schneider, S. K., & Northcraft, G. B. (1999). Three Watson, W., Kumar, K., & Michaelsen, L. (1993).
social dilemmas of workforce diversity in Cultural diversity’s impact on interaction process
organizations: A social identity perspective. and performance: Comparing homogeneous and
Human Relations, 52, 1445–1467. diverse task groups. Academy of Management Journal,
Stasser, G. (1999). The uncertain role of unshared 36, 590–602.
information in collective choice. In L. L. Westphal, J. D., & Milton, L. P. (2000). How
Thompson, J. M. Levine, & D. M. Messick (Eds.), experience and network ties affect the influence
Shared cognition in organizations: The management of of demographic minorities on corporate boards.
knowledge (pp. 49–69). Hillsdale, NJ: Erlbaum. Administrative Science Quarterly, 45, 366–398.
Stasser, G., & Stewart, D. (1992). Discovery of hidden Wilder, D. A. (1984). Predictions of belief
profiles by decision-making groups: Solving a homogeneity and similarity following social

481
 Group Processes & Intergroup Relations 9(4)

categorization. British Journal of Social Psychology, Biographical notes

23, 323–333.
K AT H E R I N E W . P H I L L I P S
is associate professor of
Williams, K., & O’Reilly, C. (1998). Demography and
management and organizations at the Kellogg
diversity in organizations: A review of 40 years of
School of Management at Northwestern
research. In B. M. Staw & R. Sutton (Eds.),
University. Her research focuses on diversity in
Research in organizational behavior (Vol. 21,
decision making groups, information sharing,
pp. 77–140). Greenwich, CT: JAI Press.
minority influence, identity boundary
Winquist, J. R., & Larson, J. R. (1998). Information
management and the effects of status in groups.
pooling: When it impacts group decision making.
She received her PhD from the Graduate School
Journal of Personality and Social Psychology, 74,
of Business at Stanford University.
371–377.
Wittenbaum, G. M. (2000). The bias toward
G R E G O R Y B . N O RT H C R A F T
is the Harry J. Gray
discussing shared information: Why are high status
Professor of Executive Leadership in the
members immune? Communication Research, 27,
department of business administration, and
379–401.
Institute of Labor and Industrial Relations, at the
Wittenbaum, G. M., & Bowman, J. M. (2004). A
University of Illinois. His major research interests
social validation explanation for mutual
include conflict management and negotiation,
peer-00571629, version 1 - 1 Mar 2011

enhancement. Journal of Experimental Social

collaboration in teams, managerial decision
Psychology, 40, 169–184.
making, and employee motivation and job design,
Wittenbaum, G. M., Hollingshead, A. B., & Botero,
particularly in high-technology manufacturing
I. C. (2004). From cooperative to motivated
settings. He received his PhD in social psychology
information sharing in groups: Moving beyond the
at Stanford University.
hidden profile paradigm. Communication
Monographs, 71, 286–310.
MARGARET A. NEALE is the John G. McCoy-Banc
Wittenbaum, G. M., Hubbell, A. P., & Zuckerman, C.
One Corporation Professor of Organizations and
(1999). Mutual enhancement: Toward an
Dispute Resolution at the Graduate School of
understanding of the collective preference for
Business at Stanford University. Professor Neale’s
shared information. Journal of Personality and Social
major research interests include bargaining and
Psychology, 77, 967–978.
negotiation, distributed work groups, and team
Wittenbaum, G. M., & Stasser, G. (1996).
composition, learning, and performance. She
Management of information in small groups. In
received her PhD in business administration from
J. L. Nye & A. M. Brower (Eds.), What’s social about
the University of Texas.
social cognition? Research on socially shared cognition
in small groups (pp. 3–28). Thousand Oaks, CA:
Sage.

Paper received 8 November 2004; revised version accepted

28 February 2006.

482

View publication stats