Implicit Voice Theories: Taken-for-Granted Rules of Self-Censorship at Work

Abstract

In four studies, we examine implicit voice theories - taken-for-granted beliefs about when and why speaking up at work is risky or inappropriate. In Study 1, interview data from a large corporation suggest that fine-grained implicit theories underlie reluctance to voice even pro-organizational suggestions. Study 2 survey data address the generalizability of the implicit theories identified in Study 1. Studies 3 and 4 develop survey measures for five such theories, establishing the measures' discriminant validity and incremental predictive validity for workplace silence. Collectively, our results indicate that implicit voice theories are widely held and significantly augment explanation of workplace silence.

Full text

IMPLICIT VOICE THEORIES:
TAKEN-FOR-GRANTED RULES OF SELF-CENSORSHIP
AT WORK
JAMES R. DETERT
Cornell University
AMY C. EDMONDSON
Harvard University
In four studies, we examine implicit voice theories—taken-for-granted beliefs about
when and why speaking up at work is risky or inappropriate. In Study 1, interview
data from a large corporation suggest that fine-grained implicit theories underlie
reluctance to voice even pro-organizational suggestions. Study 2 survey data address
the generalizability of the implicit theories identified in Study 1. Studies 3 and 4
develop survey measures for five such theories, establishing the measures’ discrimi-
nant validity and incremental predictive validity for workplace silence. Collectively,
our results indicate that implicit voice theories are widely held and significantly
augment explanation of workplace silence.
Upward communication is vital to the success of
contemporary organizations. By speaking up to
those who occupy positions that are hierarchically
higher than their own, employees can help stem
illegal and immoral behavior, address mistreatment
or injustice, and bring problems and opportunities
for improvement to the attention of those who can
authorize action. Employees of all types and levels
confront problems and formulate ideas when car-
rying out day-to-day activities in organizations; this
is the nature of work in a dynamic environment.
Yet, even when they believe they have something
useful to say, people often choose silence over
voice (speaking up [e.g., Milliken, Morrison, &
Hewlin, 2003; Ryan & Oestrich, 1998]). Reluctance
to voice substantive and relevant ideas and ques-
tions at work is widespread and frequently attrib-
uted to employee concerns about personal conse-
quences (e.g., Ashford, Rothbard, Piderit, & Dutton,
1998; Edmondson, 2003; Milliken et al., 2003; Pin-
der & Harlos, 2001; Withey & Cooper, 1989).
Whether seen as primarily rational and calculative
or as fear-driven and spontaneous, the belief that
voice is risky has been described as a general ex-
pectation that speaking up will have undesired out-
comes, such as harm to one’s reputation or image,
reduced self-esteem or emotional well-being, or
negative work evaluations and reduced opportuni-
ties for promotion (e.g., Ashford et al., 1998; Mil-
liken et al., 2003).
Two major approaches to the study of speaking
up in organizations exist. The first, which has dom-
inated the literature, is to predict the occurrence of
this prosocial, potentially risky behavior. This re-
search has amassed considerable evidence about
the individual difference, leader behavior, and or-
ganizational context antecedents of voice (e.g.,
Ashford et al., 1998; Detert & Burris, 2007; LePine &
Van Dyne, 2001; Van Dyne & LePine, 1998). Implic-
itly or explicitly, many of these studies present
psychological safety as a mediator between ante-
cedent variables and voice behavior (e.g., Ashford
et al., 1998; Miceli & Near, 1992). Detert and Burris
(2007), for example, showed that employee percep-
tions of psychological safety mediated relation-
ships between positive managerial behaviors and
employee voice. A second approach, thus far pur-
sued primarily through qualitative accounts (e.g.,
Harlos, 2001; Milliken et al., 2003; Ryan & Oestrich,
1998), is to focus directly on the phenomenon of
reluctance to speak up. As elaborated below, we be-
lieve both approaches are valuable and necessary be-
cause they reflect phenomenological, not just seman-
tic, differences. We propose that further systematic
focus on why employees don’t speak up opens the
possibility for new theoretical explanations of speak-
ing up in the workplace, for two basic reasons.
We thank Alan Johnson, Hanna Rodriquez-Farrar,
Alex Romney, and Alyssa Goldman for research assis-
tance and Ethan Burris, Adam Grant, Nathan Pettit, Sean
Martin, Dave Harrison, and research audiences at Carne-
gie Mellon, Case Western Reserve, Duke, Emory, Har-
vard, INSEAD, Pittsburgh, Queens, and Wharton for
valuable feedback and suggestions.
Editor’s Note: The manuscript for this article was ac-
cepted during Duane Ireland’s term as editor.
娀Academy of Management Journal
2011, Vol. 54, No. 3, 461–488.
461
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express
written permission. Users may print, download or email articles for individual use only.

First, overly broad conceptions of both voice and
silence, together with the assumption that what is
understood about one term fully applies to the
other, can restrict advances in theory and practice
(Tangirala & Ramanujam, 2008; Van Dyne, Ang, &
Botero, 2003). To illustrate, an employee’s not pro-
viding input because of not having ideas is a dif-
ferent phenomenon from withholding ideas from
bosses because of fear of the consequences of ex-
pressing these ideas. And an individual might
speak frequently yet withhold information or opin-
ions about important problems from bosses because
he/she believes these ideas would not be well re-
ceived. Thus, withholding relevant ideas for self-
protective reasons (self-censorship) can occur even
while voice (of another kind) is being offered.
These distinctions are important because it is the
lack of timely input—from those who have infor-
mation they believe is worth contributing, to those
with the power to act—that especially hampers or-
ganizational learning (Edmondson, 2002, 2003).
Our research thus investigates silence, defined as
“the withholding of ideas, suggestions, or concerns
about people, products, or processes that might
have been communicated verbally to someone in-
side the organization with the perceived authority
to act” (Kish-Gephart, Detert, Trevino, & Edmond-
son, 2009: 166–167).
Second, and the central focus of this article, is
the possibility that people remain silent at work
because of socially acquired beliefs, or implicit the-
ories, about what makes voice risky in social hier-
archies. If so, understanding of employee silence
might be improved by focusing explicitly on these
taken-for-granted beliefs, rather than assuming they
merely reflect personality characteristics, current
boss behaviors, or features of an organizational con-
text. This perspective—that implicit theories inde-
pendently drive a variety of social behaviors—has
been demonstrated extensively in the social psy-
chology literature (e.g., Chiu, Hong, & Dweck,
1997), but only alluded to in previous explanations
of voice in the workplace (Milliken et al., 2003).
The present research thus examines implicit voice
theories as a subtle and insidious cause of em-
ployee silence. In particular, we focus on self-pro-
tective implicit voice theories: knowledge struc-
tures that individuals use to avoid trouble that
could arise from speaking up to authorities.
After a brief introduction to the idea of implicit
voice theories, we present four studies representing
a progression from inductive to deductive analysis
that together address three general research ques-
tions: (1) Are there common implicit theories about
why upward voice is risky or inappropriate? (2)
Can these implicit voice theories be efficiently and
validly measured? And (3) Are implicit voice the-
ories related to workplace silence after other theo-
retically relevant individual differences and organ-
izational influences are controlled for?
We address the first question with an exploratory
interview study (Study 1) in which we examine the
reasons for self-protective silence from the point of
view of would-be speakers in one large multina-
tional corporation. Rather than deriving categories
of beliefs from a priori theoretical expectations, we
deepened and expanded our understanding of why
people remain silent by searching for implicit the-
ories that might be revealed in what we term latent
voice episodes: specific instances in which a
would-be speaker believes the possibility exists to
speak up to someone with positional power in a face-
to-face context about something of importance.
In Study 2, we examine the generalizability of the
implicit theories identified in Study 1. Our goal in
this study was not to develop an exhaustive taxon-
omy of all self-protective implicit voice theories
(hereafter, simply “implicit voice theories”) but
rather to confirm that those identified in Study 1
are not idiosyncratic to a single organization and are
common enough to merit subsequent investigation.
In Studies 3 and 4, we address the second and
third research questions by developing and refining
survey measures of five implicit voice theories and
linking them to correlates and consequences. In
Study 3, we operationalize and examine the psy-
chometric properties of five implicit voice theory
measures, including their relationships with an ar-
ray of relevant individual and organizational con-
structs. In Study 4, we use a three-wave survey
design to test the incremental validity of these the-
ories by examining whether they predict silence in
analyses controlling for a wide range of individual
differences and contextual factors. Through a series
of robustness checks, we show that implicit voice
theories can be understood as antecedents of si-
lence in their own right, rather than as mediators of
typically studied contextual factors or as anteced-
ents attenuated or amplified by their interaction
with contextual factors. Collectively, the results
from this hybrid research project studying samples
of hundreds of adults with diverse work experi-
ences indicate both the prevalence of implicit voice
theories and their unique contribution to explana-
tions of workplace silence.
IMPLICIT VOICE THEORIES
The Nature of Implicit Theories
We propose that largely taken-for-granted beliefs
about the risk or inappropriateness of speaking up
462 JuneAcademy of Management Journal

in hierarchical organizations exist and constitute a
type of implicit theory used for everyday “sense
making” (Levy, Chiu, & Hong, 2006). Implicit the-
ories are schema-like knowledge structures that in-
dividuals use to effortlessly process current stimu-
lus cues and choose responses (Ross, 1989: 342);
they have alternatively been called “naı¨ve,” “lay,”
and “commonsense” theories (Heider, 1958; Kelly,
1955) and are similar to “cognitive scripts” (Abel-
son, 1976), “primary frameworks” (Goffman, 1974),
and “logics of action” (Bacharach, Bamberger, &
McKinney, 2000). The term “implicit theory” is
particularly apt, however, because it clarifies that
this type of belief structure contains not just an
organized representation of stimuli, but also as-
sumptions about cause and effect (Anderson &
Lindsay, 1998). That is, implicit theories, like sci-
entific ones, allow an individual to make a priori
predictions (Levy et al., 2006). For example, an
implicit theory that it is unsafe to speak up in
public settings at work contains an “if-then” as-
sumption that speaking up will lead to negative
consequences in a specific type of situation. The
assumptions embedded in implicit theories need
not be accurate—especially as applied to each spe-
cific instance in which they guide behavior—to
serve useful functions for individuals, such as pro-
viding a sense of psychological control, helping
maintain relationships, and protecting oneself or
one’s group (Levy et al., 2006).
Implicit theories allow individuals to quickly
and relatively effortlessly orient themselves and
decide on action by comparing new stimuli with
previously encountered stimuli stored in mental
structures (Chiu et al., 1997). Such theories are
implicit in that they often operate below conscious-
ness, generating in a top-down, automatic fashion
many behavioral responses typically (but incor-
rectly) described as resulting from an intentional,
deliberative process (Bargh & Ferguson, 2000; Goll-
witzer & Brandstatter, 1997; Hertel & Kerr, 2001).
Given their automaticity, implicit theories are usu-
ally poorly articulated (Chiu et al., 1997; Levy et al.,
2006), and people seldom discuss them without
prompting or provocation (Ross, 1989). As such,
people tend not to be aware of the tremendous
impact implicit theories have on their behavior,
even though they know whether they agree or dis-
agree with the theories when they are stated and
can articulate simple versions of the theories they
hold when prompted for explanations of their be-
havior (Bacharach et al., 2000; Levy, Stroesser, &
Dweck, 1998).
Implicit theories develop during an individual’s
life, through both direct experience and vicarious
learning (Abelson, 1976; Anderson & Lindsay,
1998) in repeated situations that are “sufficiently
involving to stimulate the construction of a gener-
alized event representation of the sequence of
events that take place” (Wyer, 2004: 288). For in-
stance, individuals develop implicit theories of
leadership effectiveness via an array of experiences
in which perceived leader actions co-occur with
outcomes apparently attributable to these acts
(Epitropaki & Martin, 2004; Lord & Maher, 1991).
Once formed, implicit theories tend to endure be-
cause people seek out anecdotal verification rather
than falsification of beliefs that serve self-protec-
tive or self-enhancing aims (Furnham, 1988; Levy
et al., 2006; Wyer, 2004). Further, implicit theories
tend to survive empirical disconfirmation because
of certain cognitive and motivational processes
(Anderson & Lindsay, 1998). For example, the ten-
dency to more readily recall confirming instances
bolsters existing implicit theories, particularly
those that predict negative outcomes for the self
(e.g., social, emotional, or bodily harm). This is
partly explained by the bias whereby “bad” is
stronger than “good” in memory (Baumeister, Brat-
slavsky, Finkenauer, & Vohs, 2001) and the surviv-
al-based human fear response that tends toward
“false positives” (Ohman, 2000).
Self-Protective Implicit Voice Theories
As Anderson and Lindsay noted, countless small
decisions are made every day in organizations,
many without conscious awareness on the part of
decision makers “that they are making judgments
or using a naı¨ve theory to guide their judgments
and behaviors” (1998: 23–24). Among the routine
organizational decisions implicit theories are likely
to drive is whether to speak up in specific in-
stances, which is why scholars have called focus-
ing on implicit theories a crucial next step for un-
derstanding decisions to remain silent (Milliken et
al., 2003). Although implicit voice theories can
serve many different motives, such as protecting
others’ feelings or avoiding wasting time or effort
(Van Dyne et al., 2003), we focus here on self-
protective implicit voice theories, those that link
upward voice about certain issues or in certain
types of situations to risk. This focus allows us to
address several of the dominant themes and re-
maining theoretical and empirical questions about
workplace silence.
Studying implicit theories that connect cues em-
bedded in particular situations to silence answers
recent calls for the development of finer-grained
theories of communication behavior that include
factors such as targets, topics, and venues (Grant &
Ashford, 2008). An implicit theory perspective sug-
2011 463Detert and Edmondson

gests that silence can stem from automatically
evoked beliefs rather than from a conscious, calcu-
lative conclusion that “speaking up is risky,” a
view that is consistent with the spontaneous nature
of most routine decisions about speaking up.
1
To
illustrate, individuals might withhold voice draw-
ing on a specific implicit theory that speaking up in
the presence of one’s boss’s boss puts the boss in a
bad position and so should be avoided (Milliken et
al., 2003), irrespective of who the boss is or what
her/his previous behavior has been. This withhold-
ing differs from only withholding voice about prob-
lems from specific leaders whose prior behavior
has created an impression that offering improve-
ment-oriented voice is risky. In short, an implicit
theories perspective recognizes behavior is driven
by automatic, specific judgments (whether con-
scious or not) that link particular situational cues to
extant knowledge structures. Summary judgments
made consciously (e.g., “It’s not safe”) may provide
parsimonious explanations for a pattern of behav-
ior, but they do not capture the nature of motivated
human cognitions that give rise to such summaries.
Uncovering implicit voice theories leads to an
expanded view of the basic causes of workplace
silence and to a greater appreciation of why it is so
difficult to eliminate subordinate reticence to chal-
lenge authority. Prior qualitative studies examining
silence have shown employees attributing a lack of
safety for voice to proximate, external, and presum-
ably verifiable causes. For example, employees of-
ten explain their fear of speaking up by pointing to
harsh or unwelcoming leader behavior (Milliken et
al., 2003), unfavorable company or industry condi-
tions (Dutton et al., 1997), or prior experiences at
work suggesting that speaking up leads to negative
social or career consequences (Dutton et al., 1997;
Milliken et al., 2003). Consider, however, that
speaking up at work is a specific case of speaking
up to authority figures, a domain with which indi-
viduals have past experience in multiple institu-
tional settings starting early in life (Kish-Gephart et
al., 2009). As Milgram (1974) noted, children de-
velop beliefs about upward communication from
routine interactions with parents in which parents
convey not only specific content (e.g., “Go to bed
now”), but also more general expectations about
how to respond to authority (e.g., “You must do as
I say” or “Do not question me”). It therefore stands
to reason that people hold implicit theories that
govern behavior around authorities and that these
theories are stronger predictors of self-protective
silence than other attitudes or beliefs.
In fact, implicit voice theories may be particu-
larly potent predictors of silence toward authorities
because they are among the numerous specific
schemata individuals develop to guide behavior
in “authority ranking” situations (social relations
marked by clear power differences among members
[Fiske, 1991]). In nearly all cultures, by the age of
three children spontaneously demonstrate under-
standing of authority ranking structures, suggesting
that humans are endowed with content-specialized
psychological structures for generating this under-
standing (Cosmides & Tooby, 1992; Fiske, 1991).
Stated another way, recognition of social hierar-
chies and fear of offending those with higher stand-
ing than oneself is part of humans’ evolutionary
heritage (Milgram, 1974; Plutchik, 2003). Although
offending higher-ranking others today rarely pres-
ents physical danger, humans may nonetheless
have developed a specific “motivational system de-
signed to regulate willingness to take competitive
risks in dominance-relevant contexts” (Ermer, Cos-
mides, & Tooby, 2008: 107). Safety now refers pri-
marily to social and material rather than physical
matters (Ermer et al., 2008; Pinker, 1997). Nonethe-
less, the development and internalization of beliefs
about how to safely interact with others who have
greater power (“higher-power others”) may explain
why quantitative studies have revealed modest re-
lationships between measures of current leader be-
havior (e.g., managerial openness) (Ashford et al.,
1998; Detert & Burris, 2007; Dutton, Ashford, Law-
rence, & Miner-Rubino, 2002) and employee beliefs
that it is safe to speak up.
In suggesting that implicit voice theories may be
related to human evolution, we do not presuppose
that all individuals in all cultures equally endorse
or are equally driven by such theories (Buss, 2009;
Tooby & Cosmides, 1990). Beliefs about challeng-
ing higher-ranking others likely require specific ex-
periences to activate, shape, and reinforce them
(LeDoux, 1996; Tooby & Cosmides, 1990). Because
each individual experiences authority through a
unique combination of parents, teachers, coaches,
and religious or other institutional leaders, one
should not expect individuals to have developed
1
Research in psychology has shown that conscious
processing and reasoning about the costs and benefits of
speaking up in a specific instance likely only character-
ize situations in which an employee has a long time to
reflect on a choice and is highly motivated to do such
processing (Kish-Gephart et al., 2009). Such conscious
processing and reasoning may, for instance, occur when
employees “sell” strategic issues (Dutton, Ashford,
O’Neill, Hayes, & Wierba, 1997) and “blow the whistle”
on illegality or immorality (Miceli & Near, 1992), but they
are less likely to occur when employees make on-the-
spot work decisions to speak up or remain silent about
problems or improvement opportunities.
464 JuneAcademy of Management Journal

identical beliefs and behavioral strategies for navi-
gating hierarchical relationships (Buss, 2009).
These arguments parallel findings for other do-
main-specific mental structures. For example, hu-
mans appear to be endowed with specialized men-
tal apparatus for detecting cheaters (Cosmides &
Tooby, 1992), but individuals vary widely in their
actual ability to correctly detect cheating (Ekman,
O’Sullivan, & Frank, 1999). Overall, therefore, hu-
mans have evolved to have elaborate mental struc-
tures around authority, but the content and
strength of beliefs are likely to vary dramatically on
the basis of individual experience.
Implicit Voice Theories and Employee Silence
An implicit theories perspective on voice sug-
gests that ordinary life experiences in hierarchical
social institutions shape beliefs about where,
when, and about what speaking up is risky. That
one often cannot remember where or when these
“fundamental social frameworks” (Goffman, 1974)
were learned does not mean they are not imported
into one’s current work environment, wherein they
serve as a basis for automatic evaluation of stimuli,
rather than being formed anew in each setting
(Bacharach et al., 2000). Stated another way, when
employees encounter a new boss, their minds are
not a tabula rasa in regard to beliefs about speaking
up, just as adults’ minds are not a tabula rasa in
regard to beliefs about leadership (Lord & Maher,
1991), politeness (Brown & Levinson, 1987), and
other standards for general social behavior, when
they enter new contexts. We thus propose that
implicit voice theories can help explain work-
place silence as a direct causal factor. We do not
expect these theories to mediate relationships be-
tween typically studied antecedents and silence
because these antecedents (such as current leader
behavior and other aspects of current work con-
text) are likely only modest contributors to belief
systems about behavior around higher-power
others that form and solidify over a long time
period.
In sum, our argument is that individuals are
likely to arrive at work with a set of implicit voice
theories, constructed through past direct and vicar-
ious learning, or socialization, in hierarchical insti-
tutions. Micro aspects of specific contexts in which
an individual might speak up—such as the content
of desired communication, the venue, and the
number and positions of others present—work as
stimulus cues that are matched to the individual’s
“toolbox” of discrete, specific implicit theories
(Hong, Morris, Chiu, & Benet-Martinez, 2000).
Once invoked, implicit voice theories dictate an
appropriate response; they represent prior forma-
tion of beliefs “about what leads to what” (Bandura,
1986: 183). Implicit voice theories thus can dictate
remaining defensively silent so as to avoid personal
harm (Van Dyne et al., 2003). To the individual,
silence appears to work (that is, it keeps him/her
safe), further reinforcing the implicit theories.
Thus, without somehow engaging in behavior that
directly and specifically challenges and contradicts
them, it is unlikely that individuals will revise, set
aside, or develop new implicit theories related to
speaking up (Bacharach et al., 2000). We thus argue
that it is not easy to counteract or attenuate the
impact of implicit voice theories.
STUDY 1: GROUNDED DISCOVERY OF
IMPLICIT VOICE THEORIES
We undertook a field study in a large, complex
corporation to investigate employees’ beliefs about
the potential risks of speaking up about issues they
believed were important to their organization’s ef-
fectiveness. Our inductive approach was designed
to capture snapshots of reality as it exists for actors
in a social system, allowing them to help define the
conceptual territory rather than limiting our in-
quiry to existing categories and frameworks.
Methods
Context overview. We conducted extensive in-
terviews at a leading high-technology corporation,
referred to here by the pseudonym HiCo. Employ-
ees engaged in strategy, research and development,
manufacturing, marketing, and selling of products
and services in a highly technical and fast-changing
industry context. HiCo presented many advantages
for this research, starting with the central impor-
tance of the effective use of knowledge for success
in its industry. Recognizing this, senior executives
had been dismayed when a company-wide survey
revealed that approximately 50 percent of all HiCo
respondents did not feel comfortable speaking up
at work. These results triggered the creation of an
internal task force and an invitation to the first
author to join this group as an academic researcher
who would design and conduct interviews investi-
gating the individual and contextual factors behind
voice and silence at HiCo.
Interview sample. To maximize, as much as pos-
sible within a single company, the likelihood of
identifying diverse causal factors and employees’
beliefs about speaking up, we selected interviewees
from ten units, two in each of five divisions: R&D,
manufacturing, U.S. sales, international marketing
and sales, and corporate finance. We chose leaders
2011 465Detert and Edmondson

at the top of each focal unit and used a random
selection of direct reports from the top to the bot-
tom of a reporting structure within a given unit,
stratifying by gender and tenure only to ensure a
diverse mix of interviewees. Our interviews
spanned the hierarchy at all ten sites, ranging from
senior managers one or two levels below the divi-
sion president level to the lowest-ranking employ-
ees, and included executives, managers, engineers,
researchers, sales and marketing professionals, fi-
nancial analysts, and frontline operators. The aver-
age age of interviewees was 41 years (s.d. ⫽8.6);
their average organizational tenure was 11 years
(s.d. ⫽7.3); and 56 percent were male. We con-
ducted 190 interviews, 5 in the smallest unit of just
16 members and more than 20 each in units with
several hundred employees.
Data collection. The first author spent between
one and three days interviewing at each of the ten
sites, conducting a total of 75 of the 190 interviews.
Two additional skilled interviewers pilot-tested the
protocol with the first author at HiCo and then
conducted the remaining 115 interviews. The inter-
views, which lasted between 30 and 90 minutes
each, elicited detailed behavioral examples of
situations in which the interviewees either felt
particularly able/unable or willing/unwilling to
speak up. To minimize the imposition of poten-
tial interviewer biases on informants, each inter-
view began by showing the interviewee his/her
unit’s and division’s average scores on the “safe
to speak up” and “challenge traditional ways”
items from the company’s employee survey and
asking, “What do you make of these results for
your unit?” The interviewers then asked more
structured interview questions designed to gen-
erate descriptions of specific voice episodes. The
intent of the interviews was to generate concrete
examples of situations in which individuals
wanted to speak up and either did or did not, and
to elicit descriptions of their thoughts and feel-
ings about the situations. By pushing for specific
examples and clarifying that respondents’ per-
sonal experiences or beliefs could be highly di-
vergent from those of the majority in their unit,
we generated a diverse mix of positively and
negatively valenced descriptions from respon-
dents in each unit.
Tape-recording and transcribing the interviews
generated over 3,000 pages of text describing a
wide variety of general beliefs and specific latent
voice episodes. Approximately 72 percent of the
specific episodes pertained to issues directly re-
lated to organizational performance or improve-
ment. In these episodes, people had offered or
withheld suggestions for improving the quality, ef-
ficiency, or customer experience of a particular
product or service, or for improving a marketing or
research strategy. The remaining 28 percent of the
episodes pertained to situations perceived by the
informant as unjust or otherwise affecting personal
or coworker well-being. No episodes involved legal
boundaries crossed or matters perceived as requir-
ing reporting outside the organization.
Data analysis. We engaged in a multistep induc-
tive process (Miles & Huberman, 1994) to identify
and hone our understanding of the affect-laden be-
liefs that lead to silence. Using qualitative data
management software, we identified and placed in
one electronic folder over 500 text passages, rang-
ing from a few lines to several paragraphs, in which
informants used cognitive and emotional terms
(e.g., “afraid,” “scared,” “anxious,” “tentative,”
“paranoid”) to describe their beliefs about the
safety of speaking up. These analytic units were
systematically analyzed for evidence of common
explanations, not just of why it felt unsafe to speak
up in the specific episode described, but also of
why the episode represented, for the informant, a
more general type of situation in which it is unsafe
or a bad idea to speak up. Our focus on episodes
allowed us to detect situational cues that give rise
to silence, along with the stable factors examined in
prior research (e.g., individual differences and
manager behaviors). In short, we analyzed the data
to include attention to episode-specific features
that might reveal implicit theories about speaking
up. In keeping with prior implicit theory research,
we anticipated that features of a situation may trig-
ger certain stored beliefs about where, when, and to
whom speaking up is unsafe.
Through repeated review and discussion of the
data and the emerging themes, we identified and
named five implicit voice theories that occurred
multiple times in the data. These theories differ as
to the specifics of when, where, and why, yet all are
self-protective in their core assumption that speak-
ing up to authority figures at work is risky, and the
theories are all therefore likely to lead to defensive
withholding of upward input (Van Dyne et al.,
2003). As a check on our developing understand-
ing, we confirmed that an independent coder could
identify examples of each implicit voice theory in
the data. We also checked with our contacts inside
the company as to the face validity of our develop-
ing understanding (Lincoln & Guba, 1985). This
iterative journey between data and theoretical cat-
egorization involved extensive debate among the
authors.
466 JuneAcademy of Management Journal

Findings
Implicit voice theories. We labeled the first im-
plicit theory that emerged “presumed target identi-
fication.” It is a taken-for-granted belief in manag-
ers’ identification with the status quo, which leads
to the causal assumption that those higher in an
organizational hierarchy hear suggestions as per-
sonal criticism. Specifically, we found that many
would-be speakers viewed speaking up as unsafe
because they took for granted that the target of
voice was likely to be personally identified with, or
to feel ownership of, the aspect of the organization
in question. For example, a HiCo research scientist
considered recommending a new procedure to save
resources risky because he worried that the target
(his boss) would view the suggestion as an accusa-
tion that the current method (the one the boss was
presumed to embrace) was wasteful. As a project
manager put it, “People get so attached to their
projects; their projects become part of their self-
image and self-worth so it’s not acceptable to ques-
tion them.” In this way, numerous informants as-
sumed that, rather than being seen as helpful, their
comments about products, processes, or strategy
would threaten, offend, or invoke defensiveness in
authority figures. This was the case even when the
intended comments were not inherently negative—
that is, not related to problems or mistakes but
rather to ideas for improvement.
The second recurring implicit theory is a per-
ceived need to have solid data, polished ideas, or
complete solutions before it is safe to speak. Simi-
larly to NASA members observed by Vaughan
(1996), informants reported a fear of speaking up
without proof that something was wrong or subop-
timal; others reported a need to inoculate them-
selves against negative reactions from authorities
through extensive preparation and data gathering
prior to speaking up. The following statement by a
HiCo manager illustrates the second implicit the-
ory: “There’s that feeling, that belief, that you
should go in [to forums where voice is possible]
extremely prepared. That you should cover every
area, not go out on a limb, don’t reach or overex-
tend any conclusions without substantiation.” Oth-
ers spoke of the perceived need to look good by
being able to “answer questions perfectly,” avoid-
ing “ad lib comments,” and “not crying about
something unless you can offer a constructive so-
lution.” This implicit voice theory is reminiscent of
the issue-selling hypothesis that employees are ret-
icent to speak up without having clear solutions
(Dutton & Ashford, 1993).
The third and fourth implicit theories we identi-
fied involve beliefs about speaking up to bosses in
the presence of others. As Dutton and Ashford
(1993) noted, public voice, and the resulting visi-
bility of an issue so raised, may increase the likeli-
hood that a target takes action, but it also appears to
be associated with a belief that the speaker suffers
consequences. This implicit theory, “Don’t bypass
the boss upward,” refers to the belief that speaking
up in ways that (even inadvertently) challenge,
question, contradict, or expose one’s boss in front
of his/her superiors will be seen as disloyal and
unacceptable (Milliken et al., 2003). As informants
noted, speaking up to those above one’s immediate
boss feels risky: “[He] would maybe hold that
grudge against me, because I disagreed with him”
(manufacturing operator) or because “going over
my manager’s head” would “completely alter our
relationship,” which could “limit my career”
(sales representative). As a manufacturing man-
ager noted, this implicit voice theory hampers
learning because higher-level leaders do not hear
the unvarnished truth from distal subordinates,
especially when the intermediate managers are in
the room:
The reason people won’t speak up to me is not fear
of me, I think, but the fact that their boss was in the
room ....IfI getthem one-on-one I get some great
learnings. But where you get concern is fear of re-
prisal, that somebody will come back and say, “How
dare you go and talk to [the plant manager] without
me getting a chance to have my side of the story.
Don’t you ever talk to him again without me know-
ing.” And bang, that shuts down all the communi-
cation right there.
The fourth implicit theory identified, “Don’t em-
barrass the boss in public,” represents the belief
that bosses dislike hearing bad news, or being chal-
lenged, in front of others in a group without
advance, private notice. Driven by this implicit the-
ory, informants often reported “just dropping it”
altogether, or choosing to wait for a private oppor-
tunity. Both choices can be detrimental to organi-
zational learning, because some improvement-
oriented voice content occurs to people in public
contexts and loses its usefulness if held for later.
For instance, if a decision is being made in a
meeting, waiting for privacy or anonymity to pro-
vide an alternative or express a concern may
mean that voice is offered too late to be imple-
mented. Further, private discussions preclude
the opportunity to utilize group brainstorming to
generate solutions.
As has prior work (e.g., Milliken et al., 2003), we
identified “negative career consequences for speak-
ing up” as a fifth implicit voice theory. This theory
connects challenging the status quo to general, neg-
2011 467Detert and Edmondson

ative career repercussions caused by managerial
retaliation. As one manufacturing line operator ex-
plained, “What good is it going to do me to stand up
and have a legitimate question or maybe challenge
them about something? Nothing but put me lower
in the basement.” Others reported more specific
retribution fears, such as poor formal evaluations.
For example, one manager said, “If I disagree, they
would maybe hold that grudge against me—like our
end-of-year review, they might be nit-picky.” Oth-
ers viewed speaking up to bosses as risky for their
career trajectories. A salesperson said, “My man-
ager determines my destiny at this company, there-
fore I dare not challenge him and what he’s telling
me to do. So, in a sense, it’s not safe to speak up.”
Although the specific work consequences infor-
mants feared varied widely, the general form of this
implicit voice theory was consistent: “If I challenge
authority, retaliation in the form of career conse-
quences may arise.”
Theory or fact? The basis of implicit voice the-
ories. Next, we explored the data for evidence of
the sources of informants’ implicit theories about
speaking up. These analyses suggested that infor-
mants could only sometimes identify specific ex-
periences underpinning an implicit theory. For ex-
ample, some informants pointed to bosses reacting
to prior upward voice with emotional outbursts of
anger, demeaning or derogatory comments, or other
unpredictable or frightening behaviors as the basis
of their general beliefs about when, where, or to
whom to speak up. Not surprisingly, fear-laden ex-
periences such as these lead to a focus on self-
preservation (Rachman, 1990), which people pur-
sue in part by developing cognitive theories about
specific stimuli linked probabilistically to risk and
harm. Although it is not surprising that direct ex-
periences with stimuli like a “bulldozing” or “re-
venge-seeking” boss would give rise to a belief
about the dangers of speaking up, in our data,
such reports were relatively rare. Instead, as the
following examples illustrate, interviewees fre-
quently lacked evidence to support their fears of
speaking up, even when asked directly for such
evidence:
Interviewer: What would lead to your fear that per-
haps you’d lose your job or that whatever you say is
going to be used against you? Are there any exam-
ples of this? and Has anything every happened for
asking hard questions?
Financial analyst: I’ve never seen it happen.
International marketing manager: Everyone knows
that we never fire anybody.
Sales representative: Not that I know, but we’re al-
ways afraid that that can happen. Why? I don’t
know. And I feel the same way. And I cannot ex-
plain why.
Manufacturing operator: I’ve never seen anything
happen as a result....It’s not like a week later all of
a sudden you were gone.
Moreover, some respondents’ reports revealed
implicit theories that inhibit voice despite experi-
ences directly contradicting those theories. For ex-
ample, an international sales manager noted that
his beliefs persisted despite “management, in a
way, always stimulating that you should speak up,
that they should hear your voice.” A research asso-
ciate who, driven by the “don’t embarrass the boss
in public” implicit theory, reported staying silent
about a mistake her boss made during a meeting
also noted that when she later told the boss in
private he responded, “Oh, you should have said so
at the meeting.”
In this way, the data suggest that implicit voice
theories do not merely reflect what people have
learned in their current role and environment but
that these theories endure despite contradictory ev-
idence in the present, as do other implicit theories
(Anderson & Lindsay, 1998). For example, the data
reveal that implicit theories about voice can stem
from general beliefs rather than from actual expe-
riences at work. Said one sales representative, ex-
plaining his reluctance to speak up, “It’s just cor-
porate America in general.” Remarks from two
salespeople illustrate that informants sometimes
recognized, when probed, that their beliefs about
the risks of speaking up were not based on actual
HiCo experience: “I think it’s more societal than
[HiCo]” and “I don’t think it has to do with [HiCo]
specifically, but with people’s perception that
speaking up is viewed in a lot of instances as rock-
ing the boat.” Similarly, we found that the implicit
voice theories identified in these data were often
not grounded in employee recall of specific in-
stances that led to the general beliefs. For instance,
when asked if specific experiences had led to his
belief that he’d be “cutting his boss’s throat” if he
challenged his manager in front of more senior
managers, an employee responded, “No, I just think
it’s kind of a business thing that you don’t disagree
with your manager to his boss.” Although Study 1
respondents may have referenced “corporate Amer-
ica” or “business” as the basis and boundary for
their beliefs, their statements reflect the tendency
of people throughout the world, across all institu-
tional domains, to recognize the need for special
care in dealing with higher-ups (Fiske, 1991).
468 JuneAcademy of Management Journal

STUDY 2: ASSESSING THE BROADER
EXISTENCE OF IMPLICIT VOICE THEORIES
To ensure that our findings were not idiosyn-
cratic to a single organization or an artifact of our
interview approach and specific questions, we con-
ducted a second study using a different method to
uncover implicit voice theories with a sample of
informants from a broad range of organizational
backgrounds. We captured answers to an open-
ended survey question about the conditions under
which speaking up at work might be risky and
coded written responses for evidence of the theo-
ries identified in Study 1 as well as for evidence of
additional common implicit voice theories.
Methods
The sample. A simple survey was given to 185
students enrolled in executive education courses
offered by a large university in the northeastern
United States. These respondents had a mean age of
37.9 years; 79 percent were male, and 44 percent
had been raised outside North America. Their work
experiences and organizational levels spanned
multiple functions and hierarchical levels in a
wide array of public and private organizations.
Data collection and analyses. We developed a
survey requesting written responses to a general,
open-ended question with the following instruc-
tions: “Thinking about the what, to whom, where,
and when of speaking up (or anything else that
comes to mind about speaking up) to managers in
work organizations, please state below–in your
own words–any beliefs you have about what, in
general, makes speaking up to those with more
power feel somewhat or very risky, dangerous, or
inappropriate.” The prompt “Speaking up feels
risky or inappropriate when:” was then presented,
followed by several bullet points indicating respon-
dents could write as many distinct reasons as they
wanted.
Two research associates were instructed in the
general idea of an implicit voice theory and, using
working definitions and examples of each implicit
theory from Study 1, taught to identify evidence of
the five theories in the new data. The research
associates were also trained to search for and re-
cord evidence of any additional general beliefs that
might comprise implicit voice theories, as well as
evidence of contextual factors linked to assess-
ments of risk in respondent statements.
Findings
Implicit voice theories identified. The two inde-
pendent coders identified the five implicit voice
theories discovered in Study 1 frequently (aver-
age ⫽29.2 per theory) and with reasonably high
consistency (Cohen’s kappa ⫽.81) in the Study 2
data. We readily adjudicated the few disagreements
between the coders, noticing that most disagree-
ments stemmed from one coder’s overly liberal ap-
plication of the codes. The most frequent implicit
theory in these data was the “need for solid data or
solutions” (n⫽48), suggesting that this belief is not
merely a reflection of the high-tech, science-driven
context of Study 1. For example, some Study 2
respondents saw speaking up as risky or inappro-
priate when “we don’t have enough points to prove
and sustain the idea,” an “idea is not fully baked,”
or when “there is not a solution in hand.” Table 1
illustrates the five common implicit voice theories
identified in the two studies.
Open coding identified two other findings worth
mentioning. First, approximately 40 percent of the
comments referenced current context factors,
rather than general implicit theories, as the basis
for assessing voice as risky. For example, respon-
dents pointed to specific boss behaviors, aspects of
their organization’s culture, “hot situations” (e.g.,
someone’s bad mood), and a feeling of being on
“shaky ground” (e.g., firm is performing poorly,
their boss is new) as reasons for a calculated assess-
ment of risk. Second, these data suggest several
additional implicit voice theories, but none of these
other potential theories were identified as fre-
quently as the five self-protective ones also found
in Study 1. The bottom part of Table 1 lists these
additional beliefs and shows that each was identi-
fied fewer than ten times in the Study 2 data (that
is, less than one-third as frequently on average as
the five theories from Study 1). These additional
beliefs reflect several primary motives: a desire to
protect oneself, a desire to protect others or one’s
organization, and a desire to protect oneself from
social embarrassment or ostracism (Milliken et al.,
2003; Van Dyne et al., 2003). Finally, both coders
also noted about 60 instances that suggest people
may hold an array of additional implicit theories
about message content. For example, respondents
noted that it feels risky or inappropriate to speak up
about race/diversity issues, performance or com-
pensation decisions, overall strategy as set by se-
nior managers, their boss’s own behavior or man-
agement style, and issues already discussed.
Summary. The Study 2 data suggest that the five
implicit voice theories uncovered in Study 1 cap-
ture common beliefs about speaking up in hierar-
chical organizations, not just beliefs held at HiCo
(the Study 1 context). The data also suggest that
other implicit voice theories exist (relating to self-
protection and other motives, such as protecting
2011 469Detert and Edmondson

others), although these were not as frequently men-
tioned as were the theories also identified in Study
1. Thus, we took assessing whether common im-
plicit voice theories can be measured and linked to
workplace silence as our next goal (versus devel-
oping an exhaustive taxonomy of these theories).
We proceeded to Study 3, in which we developed
survey measures to operationalize the five implicit
theories identified most frequently in Studies 1 and 2.
STUDY 3: DEVELOPMENT OF IMPLICIT VOICE
THEORY MEASURES
Our goal in Study 3, to evaluate the psychometric
properties of proposed scales for five implicit voice
theories, included assessing statistical support for
our argument that implicit voice theories are dis-
tinct (i.e., show discriminant validity) from the in-
dividual difference and contextual factors com-
monly discussed in extant literature.
TABLE 1
Implicit Voice Theories Identified in Studies 1 and 2
Implicit Voice Theories Exemplary Quotations from Informants
Number of
Times
Identified,
Study 2 Only
Found in Studies1&2
Presumed target
identification
“Bosses may feel personal ownership in the tasks I am suggesting are problematic.” 16
“The boss may have created the processes and may be offended or attached to them.”
“If speaking up involves critiquing a process, routine, or belief which is important to the
boss, they may hold a grudge.”
Need solid data or
solutions (to speak up)
“I think that presenting an under-developed, under-researched idea is never a good idea.” 48
“You are questioning their ideas and had better have proof to back up your statements.”
“You lack the necessary data to justify your position.”
Don’t bypass the boss
upward
“If there is a higher-level individual present it is more risky because you would be afraid
that your direct boss would feel as if you were going over their head. [My boss] may
perceive that I am pointing out issues that she should have identified.”
28
“My boss would see [speaking up to his boss] as undermining and insubordinate.”
“When you go around your boss and move up the chain there is the possibility that it
will have dramatic negative effects. By breaking the chain of command it appears that
you do not respect who you report to, even if you only feel that it was an important
issue that was being ignored.”
Don’t embarrass the boss
in public
“Managers hate to be put on the spot in front of others. It is best to brief them first one-
on-one so the boss doesn’t look bad in front of the group.”
25
“You should pass it by the boss in private first, so you don’t cut his legs out from under
him.”
“[Speaking up] is done in front of a group, it is the first time they are hearing the suggestion,
and it is possibly counter or critical to [the manager’s] objectives, beliefs, team.”
Negative career
consequences of voice
“To stop a project or say it is worthless would be a career ender.” 28
“Speaking up may leave a bad impression and impact future promotions.”
“Speaking up may reduce the number of opportunities to be involved in interesting
projects.”
Other potential theories identified in Study 2 only
Don’t speak up if you’re not an expert 7
Don’t speak up when caught off guard 6
Don’t speak up when doing so will harm others 6
Don’t speak up if it makes you look like a show-off, not a team player 6
Don’t speak up in front of clients/customers 3
Don’t speak up when you are too new 3
Don’t speak up when you are the youngest 3
470 JuneAcademy of Management Journal

Methods
Data collection and sample. To reduce respon-
dent fatigue and concerns about common method
bias (Podsakoff, MacKenzie, Lee, & Podsakoff,
2003), we collected survey responses to new im-
plicit theory items and items tapping several poten-
tial correlates on a first survey and then obtained
responses to items assessing other theoretically rel-
evant constructs on a second survey given approx-
imately one week later. We collected survey data
from two sources, both of which (as in Study 2)
provided access to adults with a broad range of
work experiences in hierarchical settings. We used
a community-based online subject pool associated
with one university in the northeastern United
States. Of the 231 adults invited to participate (on
the basis of eligibility criteria including at least two
years of full-time work experience), 157 provided
usable data for survey 1, and 117 provided usable
data for both surveys, giving us an overall net re-
sponse rate of 52 percent of those initially invited
to participate. We also administered surveys 1 and
2 to 110 MBA students at a second northeastern
university. Of the 108 respondents who provided
usable data for survey 1, 66 respondents also pro-
vided usable data for survey 2, giving us a net
response rate of 60 percent.
Of the 265 respondents providing data for the
implicit voice theory measures and other con-
structs on survey 1, 59 percent were female, and 66
percent were between the ages of 22 and 35. Ap-
proximately 85 percent of the sample had spent the
majority of their childhood in North America; 10
percent were raised in Asia. In this sample, 48
percent of the members were currently employed,
and 51 percent had been managers of at least one
employee in their most recent job. Compared to the
MBA respondents, respondents from the commu-
nity pool were, on average, older, and they were
more likely to be currently employed, U.S.-raised,
and female (all p⬍.01). These differences merely
represent increased variance on those dimensions
and were therefore not considered a reason to sep-
arate the samples for assessments of the implicit
voice theory scales or their correlations with indi-
vidual differences. However, because the relation-
ship between the theories and factors in an indi-
vidual’s current environment (e.g., boss behaviors)
might be expected to differ for those currently em-
ployed and those currently unemployed, we report
correlations and discriminant validity evidence for
the implicit voice theories and contextual factors
using only the 94 currently employed respondents
from the community sample.
Measures of implicit voice theories. Use of self-
report survey measures is the most common tech-
nique for assessing implicit theories, and prior re-
search indicates that when presented with
straightforward statements of an implicit theory,
people are able to “reflect and endorse their views
accordingly” (Levy et al., 2006: 5), even if the be-
liefs are not normally made explicit. For example,
implicit person theories are routinely assessed on
surveys with basic statements reflecting beliefs
about the malleability of personality (e.g., Chiu et
al., 1997; Levy et al., 1998; Plaks, Grant, & Dweck,
2005), intelligence, creativity, wisdom (Sternberg,
1985), and morality (Chiu, Dweck, Tong, & Fu,
1997). Our goal was to develop implicit voice
theory measures that were understandable and
appropriate for adults in a wide variety of indus-
try and organizational settings and that had ade-
quate content validity without taxing respon-
dents’ energy or goodwill (Brown, Trevino, &
Harrison, 2005).
We therefore used working definitions and exam-
ples from Studies 1 and 2 to iteratively develop,
discuss, and refine survey items until we had set-
tled on 28 survey items (5 or 6 per theory) deemed
to represent the conceptual domain of our five im-
plicit theories. Throughout the process, we at-
tempted to ensure that our measures had reason-
able initial “substantive validity” (Anderson &
Gerbing, 1991) by constructing items that drew di-
rectly from the language used by working infor-
mants, revising proposed items on the basis of feed-
back from a consultant working in this domain,
and making further revisions based on feedback
from executive MBA students. Because implicit
voice theories are general beliefs imported into
specific situations (e.g., interactions with specific
bosses), the questions treat “speaking upward”
and “to the boss” in hypothetical rather than
concrete terms. Respondents were given a general
instruction to answer “some specific questions
about upward communication in work organiza-
tions” rather than told to think about speaking up
to a particular target (e.g., their current boss).
Questions were answered on a five-point agree-
ment scale.
Measures of potential correlates. The ultimate
value of implicit voice theory measures rests not on
their own reliability and internal validity, but on
whether they are distinct from and have incremen-
tal validity over other predictors. In Study 3, we
therefore began establishing discriminant validity
by collecting data on a number of individual and
organizational attributes. We included several
variables commonly used in prior voice research
(e.g., LePine & Van Dyne, 1998). We coded re-
2011 471Detert and Edmondson

spondents’ gender (1 ⫽“female”), current em-
ployment status (1 ⫽“currently employed”), and
managerial status/experience (1 ⫽“manager”)
using dummy variables. Respondents’ age was
assessed using seven categories, ranging from “21
or less” to “56 or older.” Respondents also indi-
cated the type of industry/company in which
they had the most years of work experience using
one of seven categories provided (e.g., “manufac-
turing,” “consulting”).
We also sought to establish that the implicit
voice theory measures were distinct from disposi-
tional characteristics of respondents linked to voice
in prior research. For example, LePine and Van
Dyne (2001) found assertiveness and vulnerability,
subcomponents for extraversion and neuroticism,
respectively, to be the “Big 5” personality markers
most highly correlated with voice. Similarly, pro-
active personality has been identified as a positive
correlate of voice behavior in prior research (Detert
& Burris, 2007). Endorsement of implicit voice the-
ories may simply be a reflection of respondents’
overall vulnerability or, conversely, may be com-
pletely unnecessary responses to certain stimuli for
those very high in assertiveness and proactive per-
sonality. Assertiveness and vulnerability were
therefore each measured on survey 1 using 10 items
from the International Personality Item Pool (2009).
Sample items for assertiveness were “I take charge”
and “I seek to influence others”; sample items for
vulnerability were “I get overwhelmed by emo-
tions” and “I can’t make up my mind”. We mea-
sured proactive personality on survey 2 using 4
items from Bateman and Crant’s (1993) 17-item
scale. The 4 items, including, “If I see something I
don’t like, I fix it” and “I am always looking for
better ways to do things,” have been used as a
reliable short version of the scale in voice research
(Detert & Burris, 2007).
Because implicit voice theories involve leaders
as targets and because leaders have power to im-
pose material consequences on employees (Magee
& Galinsky, 2008), it was important to determine
whether the theories stemmed from the current be-
haviors of respondents’ bosses. For example, if
these theories have deeper roots than learning from
a current boss’s behavior, then respondents’ level
of agreement with them should be distinct from
their ratings of their current bosses’ behaviors, even
behaviors that are highly “considerate” (Stogdill,
1963) and “intellectually stimulating” (Podsakoff,
MacKenzie, Moorman, & Fetter, 1990). Similarly, if
implicit voice theories represent more than learn-
ing from one’s current context, then endorsements
of them should also be distinguishable from ratings
of a negative current leader behavior such as abu-
sive supervision (Burris, Detert, & Chiaburu, 2008).
We also reasoned that other leader behaviors less
directly linked to voice in prior research, such as
“initiating structure” (leader behaviors involving
clear delineation of standards and expected proce-
dures [Stogdill, 1963]), should be distinct from the
implicit voice theories. Even if these theories are
largely a reflection of current leader behaviors, ini-
tiating structure could minimize the development
of theories if it leads subordinates to see a leader as
even-handed; or, conversely, it could lead to higher
levels of implicit voice theories if subordinates per-
ceive initiation of structure as leader disinterest in
being challenged.
We evaluated the above discriminant validity
questions using four extant measures of perceived
leader behavior. First, in survey 1 we used items
from the Leader Behavior Description Question-
naire (LBDQ-12; Stogdill, 1963) to assess consider-
ation and initiating structure. The 6 items used to
measure consideration include “S/he is friendly
and approachable” and “puts suggestions made by
the group into action”; the five items used to mea-
sure initiating structure include “asks that group
members follow standard rules and regulations”
and “lets group members know what is expected of
them.” Second, we used Podsakoff and colleagues’
(1990) 4-item measure to assess intellectual stimu-
lation (sample items: “S/he challenges me to think
about old problems in new ways” and “asks ques-
tions that prompt me to think”) in survey 2. Third,
we measured abusive supervision in survey 2, ask-
ing respondents 6 questions from Tepper’s (2000)
15-item scale. Four of the 6 items chosen have
been shown to form a reliable scale that is nega-
tively correlated with subordinates’ psychologi-
cal safety and use of voice to their current leader
(Burris et al., 2008). Sample items include “S/he
puts me down in front of others” and “S/he tells
me my thoughts or feelings are stupid.” These
four leader-behavior constructs were all rated on
a five-point frequency scale ranging from “never”
to “always.”
Finally, if individuals develop implicit theories
prior to joining a specific work organization, then
the theories measured here should not be strongly
related to signals about the risks of voice in a cur-
rent work environment. For example, measures of
the theories should be distinct from measures of
organizational centralization, even though hierar-
chical, centralized decision making has been iden-
tified as a feature of organizations in which silence
is pervasive (Milliken et al., 2003). To investigate
this relationship, we included the hierarchy of au-
thority component of Hage and Aiken’s (1969)
472 JuneAcademy of Management Journal

measure of organizational centralization in survey
2. This five-item measure was assessed on a four-
point scale ranging from “definitely false” to “def-
initely true” with items such as “There can be little
action here until a supervisor approves a decision”
and “Even small matters have to be referred to
someone higher up for an answer.”
Data Analysis
The first question addressed by Study 3 was
whether our measures of the five implicit voice
theories have adequate statistical properties. We
computed and evaluated item and scale descrip-
tive statistics and reliabilities, as well as results
from multiple factor analysis models. To do this,
we randomly split the 265 survey 1 responses,
analyzed the first half extensively, and then ran a
single confirmatory model on the hold-out
sample.
With the first half of the sample, we fit the 28
items to a five-factor model in LISREL, using max-
imum-likelihood estimation and allowing for factor
correlation (with direct oblimin rotation). After ex-
amination of item means, reliability information,
and factor loadings (e.g., item loadings ⬎.50 on the
hypothesized factor), we eliminated 8 items (1 or 2
from each implicit theory measure) and estimated a
second five-factor model using the remaining 20
items (see Appendix A for the items). We then
estimated several four-, three-, and one-factor mod-
els to compare against the fit of the hypothesized
five-factor model. For example, we compared the
fit of the hypothesized model to that of a four-factor
model in which the items for the “don’t bypass the
boss upward” and “don’t embarrass the boss in
public” implicit theories were loaded onto a single
factor. When convinced that the five-factor model
was the most appropriate one for the first half of the
data, we then estimated its fit using the hold-out
sample (n⫽132).
To examine whether our implicit voice theory
measures had discriminant validity in relation to
other theoretically relevant variables, we computed
scales for all constructs (the five implicit voice
theories and all other constructs in the nomological
network) by averaging all items tapping a construct
and estimating reliabilities for each scale. We then
computed bivariate correlations among all the vari-
ables and performed additional confirmatory factor
analyses (CFAs). For the demographic characteris-
tics assessed as categorical variables, we assessed
relationships with the implicit voice theories via
analyses of variance (ANOVAs) with post hoc
contrasts.
Findings
Table 2 presents the means, standard deviations,
and intercorrelations of the five implicit voice the-
ories and all other variables.
Implicit voice theory scales. As expected, the
implicit voice theory measures are positively and
moderately intercorrelated. The means are near the
scale midpoint of 3.0, and standard deviations
(0.73 to 0.90) indicate a significant range of agree-
ment with each implicit theory. The estimated re-
liabilities for the implicit voice theory scales (see
the diagonal of Table 2) are acceptable (Nunnally &
Bernstein, 1994).
The hypothesized five-factor structure for the 20
retained implicit voice theory items fit the ran-
domly selected first half of the data well (RMSEA ⫽
.05, CFI ⫽.97, NNFI ⫽.96). Further, the five-factor
model fit the data better (in terms of fit indexes and
significant differences in chi-square statistics per
degrees of freedom) than all more parsimonious
models estimated. For example, the hypothesized
five-factor model corresponding to our implicit
voice theories fit the data better than a four-factor
model combining the items for the “don’t bypass
the boss upward” and “don’t embarrass the boss in
public” implicit theories (⌬
␹
2
⫽49.1, ⌬df ⫽4, p⬍
.01) and a model with a single factor (⌬
␹
2
⫽289.6,
⌬df ⫽10, p⬍.01). The hypothesized five-factor
model also fit the data better than a four-factor
model in which the correlation between the two
most highly correlated implicit theory latent factors
(“don’t bypass upward” and “don’t embarrass the
boss in public”) was fixed at 1.0 (⌬
␹
2
⫽56.8, ⌬df ⫽
1, p⬍.01) (Bagozzi & Phillips, 1982). We con-
firmed the fit of the hypothesized five-factor model
using the unexplored half of the data. All fit in-
dexes were again above acceptable thresholds (e.g.,
RMSEA ⫽.08, CFI ⫽.94, NNFI ⫽.93).
Implicit voice theories and relevant correlates.
We next considered the discriminant validity of the
implicit voice theories from the set of individual
differences and contextual factors described above.
As shown in Table 2, respondents’ demographic
characteristics were largely unrelated to the im-
plicit voice theories. Respondents who were cur-
rently employed reported higher agreement with
the implicit theory that one needs solid data or
solutions when speaking up and with the implicit
theory that speaking up has negative career conse-
quences. The managerial status of respondents was
not related to any theory, suggesting that becoming
a manager does not make one less likely, on aver-
age, to hold the kinds of implicit theories that
thwart speaking up. Gender was also uncorrelated
with any of the implicit voice theories. Further,
2011 473Detert and Edmondson

TABLE 2
Descriptive Statistics and Correlations for Study 3 Variables
a
Variable Mean s.d. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Individual implicit
voice theories
1. Presumed target
identification
2.90 0.76 (.78)
2. Need solid data or
solutions
2.80 0.90 .43** (.78)
3. Don’t bypass boss
upward
2.89 0.76 .36** .33** (.76)
4. Don’t embarrass
boss in public
3.22 0.77 .35** .40** .51** (.77)
5. Negative career
consequences
2.67 0.73 .49** .42** .36** .35** (.76)
Individual difference
variables
6. Gender 0.60 0.49 .03 .08 ⫺.03 .07 .02
7. Current
employment
status
0.51 0.50 .12 .24** .07 .07 .24** .18*
8. Managerial status/
experience
0.51 0.50 ⫺.11 .03 ⫺.10 .02 ⫺.11 .04 .07
9. Assertiveness 3.76 0.62 ⫺.32** ⫺.19* ⫺.23** ⫺.26** ⫺.28** ⫺.02 ⫺.02 .27** (.91)
10. Vulnerability 2.08 0.56 .26** .04 .06 .04 .11 .11 ⫺.13 ⫺.23** ⫺.42** (.89)
11. Proactive
personality
4.20 0.53 ⫺.11 .03 ⫺.04 .07 ⫺.15* .09 .00 .15* .45** ⫺.23** (.77)
Contextual variables
12. Consideration
(leader)
b
3.62 0.78 ⫺.26* ⫺.06 ⫺.06 ⫺.02 ⫺.12 .13 .07 .13 ⫺.21* .18 (.89)
13. Initiating structure
(leader)
3.94 0.66 ⫺.03 ⫺.05 ⫺.02 .00 ⫺.03 .23* .24* .06 ⫺.21* .06 .37** (.76)
14. Intellectual
stimulation
(leader)
3.16 0.85 ⫺.24* ⫺.05 ⫺.30** .04 ⫺.22* .15 .34** .29** ⫺.25* .33** .42** .27** (.89)
15. Abusive
supervision
(leader)
1.50 0.74 .17 .07 .11 .10 .26* .00 ⫺.03 .02 .05 ⫺.04 ⫺.62** ⫺.29** ⫺.43** (.91)
16. Organizational
centralization
2.20 0.61 .38** .06 .18 .19 .27** .12 ⫺.19 ⫺.25* .37** ⫺.22* ⫺.27** ⫺.06 ⫺.42** .38** (.87)
a
Variables 1–10 and 12–13 were measured at time 1; variables 11 and 14–16 were measured approximately one week later. Where appropriate, alpha reliabilities are shown on the
diagonal in parentheses. Empty cells indicate use of a restricted sample for correlations involving contextual variables.
b
n⫽183 for correlations involving variables 1–11, but n⫽94 for correlations involving contextual variables 12–16.
*p⬍.05
** p⬍.01

ANOVA results (not shown in Table 2) for both the
age and the industry categories indicated no overall
differences for any of the implicit theories; post hoc
univariate comparisons also did not reveal clear
differences in the implicit theories for any age or
industry grouping.
The bivariate correlations between the implicit
voice theories and the individual dispositions
examined are generally modest, and only about
half are significant. The most consistent pattern
emerges for assertiveness, which is significantly,
negatively related with all five implicit voice the-
ories (though at a modest level, an average rof
⫺.26). Using the complete data from both sam-
ples (n⫽183), we ran a CFA in which the hy-
pothesized best-fitting model had eight factors:
five implicit voice theories, assertiveness, vul-
nerability, and proactive personality. CFA results
suggested an acceptable fit for the model discrim-
inating the theories from the personality traits
(RMSEA ⫽.06, CFI ⫽.92, NNFI ⫽.92) and a
poorer model fit for all alternative models com-
bining any implicit voice theory with any person-
ality dimension. For example, a seven-factor
model combining items for the personality di-
mension (assertiveness) and the theory (career
consequences) whose independent factors were
most highly correlated (
␾
⫽–.42) in the eight-
factor model fit the data significantly less well
(⌬
␹
2
⫽173.8, ⌬df ⫽7, p⬍.01).
Turning to the contextual factors, we found that
the four measures of perceived leader behavior gen-
erally had correlations with the implicit voice the-
ories in the expected directions (e.g., negative for
consideration, positive for abusive supervision),
but overall these relationships were quite weak and
inconsistent. For example, none of the four extant
perceived leader behavior measures was signifi-
cantly correlated with all five of the implicit theo-
ries, and only one correlation between a leader
behavior and a voice theory reached .30. Intellec-
tual stimulation showed the most consistent pat-
tern, being significantly, negatively related to three
of the five theories, but at an average value of only
⫺.15. Abusive supervision and consideration were
each correlated (positively and negatively, respec-
tively) with only one of the five implicit voice
theories. Finally, centralization was significantly,
positively correlated with two of the five implicit
voice theories (average r⫽.22). To avoid potential
recall/hindsight bias, we ran CFA models on the
implicit voice theory and work context factors us-
ing only the currently employed Study 3 respon-
dents. Because a model with all ten factors (five
theories, four leader behaviors, and centralization)
required estimation of too many parameters rela-
tive to the sample size to produce reliable esti-
mates, we fitted several smaller models to confirm
the five theories were distinct from any of the con-
text factors. No model that combines any contex-
tual factor with any of the implicit voice theories
resulted in a better fit to the data. For example, a
six-factor model with five implicit theory factors
and one factor for organizational centralization fit
the data well (RMSEA ⫽.04, CFI ⫽.95, NNFI ⫽
.95), significantly better than a five-factor model
that combined centralization with its highest cor-
related implicit voice theory, presumed target iden-
tification (⌬
␹
2
⫽108.9, ⌬df ⫽5, p⬍.01).
Summary
The results of Study 3 indicate that five induc-
tively derived implicit voice theories can be mea-
sured with survey scales that show evidence of
reliability and validity. Results also show that the
implicit theories are distinct from theoretically rel-
evant individual differences and contextual factors.
The next question, which we addressed in Study 4,
was whether these implicit voice theories also have
incremental validity—that is, whether they predict
workplace silence in analyses controlling for other
individual and contextual explanations.
STUDY 4: EXAMINING THE INCREMENTAL
VALIDITY OF IMPLICIT VOICE THEORIES
We used a three-wave survey process with a new
sample of working adults to examine the incremen-
tal validity of the implicit voice theories as predic-
tors of silence. We measured an array of theoreti-
cally relevant individual difference and contextual
correlates of silence at time 1, the implicit voice
theories at time 2, and silence about problems and
ideas at time 3.
As already noted, prior theory and research on
voice and the smaller body of research on silence
have described individual demographic, disposi-
tional, and attitudinal factors, leadership behav-
iors, and other contextual factors as primary influ-
ences on voice or silence. The arguments and
initial evidence offered thus far in this article, how-
ever, suggest that implicit voice theories that are
only partially, if at all, rooted in current experience
drive silence in a current organization, directed
toward current authority figures. For example, em-
ployees may remain silent in meetings with “skip-
level” leaders (those in positions one or more hier-
archical steps higher than the employees’
immediate bosses) because they developed a belief
long ago that their boss would see speaking up as
disloyal. Or employees may have internalized gen-
2011 475Detert and Edmondson

eral social norms about “embarrassing others in
public” and therefore refrain from speaking up in
such settings, irrespective of their natural ten-
dencies toward speaking up or current contextual
cues from a leader about its acceptability. We
therefore tested the following hypothesis about
the independent influence of the theories of in-
terest here:
Hypothesis 1. Implicit voice theories predict
workplace silence in analyses controlling for
other individual and contextual influences on
silence.
As described in our introduction, and illustrated
throughout, the nature of implicit theories suggests
that the direct, independent effect hypothesized
above is its own main story, not a simplification of
a moderation or mediation story involving general
workplace influences. Thus, we conducted a series
of robustness checks to rule out alternative expla-
nations, including the idea that we missed more
complicated relationships between the theories
and other variables. We also examined whether our
hypothesis receives support using an alternative
dependent variable (voice) that is less directly
the behavioral manifestation of the implicit voice
theories.
Methods
Data collection and sample. To reduce concerns
about respondent fatigue, common method bias,
and uncertain direction of causality, we collected
the Study 4 data via three surveys completed ap-
proximately one to three weeks apart (Ostroff,
Kinicki, & Clark, 2002). The measures of all indi-
vidual difference and contextual variables were
placed on survey 1, the implicit voice theory mea-
sures on survey 2, and the dependent variable—a
measure of silence—on survey 3. In addition to
staggering surveys to reduce “memory traces” (Har-
rison & McLaughlin, 1993), we used spacing within
the surveys and different scale types (e.g., agree-
ment versus frequency anchors [Podsakoff et al.,
2003]) to facilitate construct differentiation.
Data were collected from students enrolled in
an executive MBA program based in the north-
eastern United States. Of 123 possible respon-
dents, 116 completed all three surveys. However,
because an accurate assessment of the hypothesis
depended on respondents’ being currently em-
ployed and having a current boss, we excluded
responses from 22 people who were unemployed
or self-employed (e.g., company founders). The
average age reported by the 94 in the final sample
was 34.5 (s.d. ⫽5.0); 17 percent were female, 68
percent reported being a manager of one or more
employees, 43 percent reported significant work
experience outside the U.S. or Canada, and 50
percent reported being born and raised outside of
North America (with 33 percent, Asia was the
dominant other origin).
Measures
Implicit voice theories. We assessed the five im-
plicit voice theories using the same 20 items used
in Study 3 (see Appendix A), again rated using a
five-point agreement scale. Because our goal in
Study 4 was to examine the predictive utility of
implicit voice theories as an explanatory category
distinct from personality or contextual factors, not
to assess which particular theories are most related
to silence in this particular sample, we combined
the 20 items into a single higher-order latent con-
struct (self-protective implicit voice theories) using
confirmatory factor analysis in LISREL. In short,
use of a higher-order construct is both parsimoni-
ous and consistent with the recommendation that
independent and dependent variables (in this
case, a general measure of silence) should be
assessed at the same level of aggregation (Ajzen,
1991). We thus computed the value for each re-
spondent on this higher-order implicit voice the-
ory construct by weighting each of the 20 theory
item responses by its loading on the first-order
factor and then further weighting the five first-
order theory factors by their loadings on the sec-
ond-order theory factor. Estimated reliability for
this measure is .93.
Individual difference control variables. We
used evidence from Study 3 as well as from prior
theory and research to select individual differ-
ence and contextual control variables. Gender
(1 ⫽“female”) and age (in years) were assessed as
basic demographic controls that could poten-
tially affect silence (LePine & Van Dyne, 1998).
Because the assertiveness dimension of extraver-
sion and the vulnerability dimension of neuroti-
cism were the highest Big 5 correlates of voice in
prior research (LePine & VanDyne, 2001) and
could likewise be reasons for silence, we again
measured respondents’ assertiveness and vulner-
ability as described in Study 3. We also measured
proactive personality, another dispositional pre-
dictor of speaking up (Detert & Burris, 2007), as
in Study 3. Additionally, because dispositional
affect may be theoretically related to staying si-
lent at work (Tangirala & Ramanujam, 2008) and
is a recommended means of controlling for po-
tential same-source biases (Cropanzano, James, &
Konovsky, 1993), we included measures of re-
476 JuneAcademy of Management Journal

spondents’ positive affect and negative affect us-
ing Watson, Clark, and Tellegen’s (1988) 20-item
PANAS scale, with instructions to report the ex-
tent each adjective was felt “on average,” using
anchors ranging from “very slightly or not at all”
(1) to “extremely” (5).
We also controlled for respondents’ managerial
status (coded 1 for a manager of one or more sub-
ordinates, 0 otherwise), reasoning that formal posi-
tion might alter both beliefs and behaviors regard-
ing speaking up. Finally, because level of
satisfaction with one’s job has been examined as a
reason to speak up more or less (e.g., Withey &
Cooper, 1989), we assessed overall job satisfaction
on a five-point scale using three items adapted from
Hackman and Oldham’s Job Diagnostic Survey
(1975) (sample item: “Generally speaking, I am very
satisfied with my current job”; “strongly disagree”
[1] to “strongly agree” [5]).
Current context control variables. We con-
trolled for four aspects of respondents’ current con-
text that might independently influence workplace
silence. First, drawing from the literature (e.g., Mil-
liken et al., 2003), we assessed respondents’ view of
their organization’s centralization (using the same
measure as in Study 3). Second, we measured re-
spondents’ ratings of the psychological safety in
their work units, because of its clear connection to
willingness to take interpersonal risks, including
speaking up (Detert & Burris, 2007; Edmondson,
2003). We used Edmondson’s (1999) scale, with
items rated from “highly inaccurate” (1) to “highly
accurate” (7) (sample items: “Members of my unit
are able to bring up problems and tough issues”; “It
is safe to take a risk in my unit”). Third, we as-
sessed leader openness on a scale ranging from
“never,” 1, to “always,” 5, using a four-item scale
used by Detert and Burris (2007) (sample items: My
manager is . . . “willing to make changes,” “inter-
ested in my ideas”). Though leader intellectual
stimulation (as used in Study 3) and leader open-
ness are conceptually similar, we chose to use the
latter here, given its more direct theoretical and
empirical link to speaking up beliefs and behavior
in prior work (e.g., Ashford et al., 1998; Detert &
Burris, 2007). Finally, we again measured per-
ceived abusive behaviors of a current boss, using
the same leader abusiveness measure as in Study 3,
because this variable has been linked to reduced
voice (Burris et al., 2008).
Silence. In response to prior arguments that si-
lence should be theorized and measured as distinct
from voice (Van Dyne et al., 2003) and to serve our
specific focus on beliefs leading to self-censor-
ing—or not speaking up—we did not wish to use
extant voice measures as a dependent variable.
Similarly, newly developed measures of silence
were not appropriate for this research, because mo-
tives for silence are embedded in them. For exam-
ple, Parker, Bindl, Van Dyne, and Wong’s (2009)
measure of “defensive silence” includes attribu-
tions about withholding ideas, opinions, or infor-
mation because the respondent does not want to
“damage [his/her] reputation” and “be seen as
difficult or rude.” Use of such measures here
would thus have involved using an explanation
for silence (implicit voice theories) to predict an
outcome measure with its own already embedded
explanation.
We therefore developed our own five-item mea-
sure of silence. The scale is target-specific, in that
respondents were asked to answer questions about
“withholding ideas, opinions, and/or information
from your current boss.” We relied on self-reports
because, unlike observable behaviors (including
voice), instances of intentional withholding of in-
put are most clearly knowable by the self (Tangirala
& Ramanujam, 2008). The items for this measure,
which were rated on a frequency scale (1 ⫽“never”
to 5 ⫽“always”), were “I withhold ideas from my
boss for changing inefficient work policies,” “I
keep ideas for developing new products or services
to myself,” “I do not speak up about difficulties
caused by the way managers and subordinates in-
teract, “I keep quiet in group meetings about prob-
lems with daily routines that hamper perfor-
mance,” and “I withhold thoughts about improving
customers’/stakeholders’ experiences with us.” Es-
timated reliability for this measure was .74. To
confirm that this scale was negatively related to,
but not the mere opposite of, voice, we also col-
lected a self-report measure of voice using Van
Dyne and LePine’s (1998) six-item voice scale. In
keeping with expectation, the correlation between
the silence and voice measures was ⫺.55, indicat-
ing only about 30 percent overlap in self-reported
voice and our new measure of silence that reflected
withholding input.
Findings
Preliminary analyses. Table 3 presents the
means, standard deviations, correlations, and esti-
mated reliabilities (on the diagonal) of the Study 4
variables. In general, variable correlations are as
expected (e.g., leader openness and leader abuse
are negatively correlated; the implicit voice theory
latent variable and silence are positively corre-
lated), and none are so high as to warrant concerns
about multicollinearity.
2011 477Detert and Edmondson

TABLE 3
Descriptive Statistics and Correlations for Study 4 Variables
a,b
Variable Mean s.d. 1 234567891011121314151617
1. Gender 0.17 0.38
2. Age 34.50 5.03 ⫺.11
3. Assertiveness 3.94 0.48 ⫺.05 .08 (.80)
c
4. Vulnerability 1.87 0.47 .11 ⫺.19 ⫺.23* (.79)
5. Proactive
personality
4.22 0.54 .06 ⫺.12 .19 ⫺.29** (.76)
6. Positive affect 3.84 0.48 .13 .05 .34** ⫺.33** .32** (.81)
7. Negative affect 1.69 0.50 .10 ⫺.34** ⫺.05 .53** .03 ⫺.10 (.84)
8. Managerial status 0.68 0.47 .13 .08 .13 .06 .17 ⫺.11 .12
9. Job satisfaction 3.19 0.89 .00 .03 ⫺.03 ⫺.11 .06 .31** ⫺.26** ⫺.04 (.83)
10. Organization
centralization
2.28 0.69 .17 ⫺.17 ⫺.22* .06 .14 .00 .18 ⫺.05 ⫺.38** (.85)
11. Work unit
psychological
safety
5.08 0.98 ⫺.11 .27** .22* ⫺.14 .01 .06 ⫺.36** .08 .24* ⫺.48** (.66)
12. Leader openness 3.64 0.76 ⫺.05 .10 .24* ⫺.27** .05 .16 ⫺.22* .17 .27** ⫺.52** .46** (.88)
13. Leader
abusiveness
1.63 0.71 ⫺.01 ⫺.04 .08 .15 .09 ⫺.01 .36** .08 ⫺.28** .44** ⫺.35** ⫺.61** (.84)
14. IVT-specific
leader behaviors
2.02 0.75 .13 ⫺.01 ⫺.03 .11 ⫺.07 .05 .21* ⫺.08 ⫺.22* .52** ⫺.37** ⫺.60** .75** (.88)
15. Self-protective
IVTs
24.46 4.94 .06 .09 ⫺.14 ⫺.05 ⫺.02 .14 ⫺.12 ⫺.11 .01 .29** ⫺.19 ⫺.01 .05 .22* (.93)
16. Voice 3.93 0.54 ⫺.06 .10 .22* ⫺.01 .05 ⫺.01 ⫺.09 .10 .06 ⫺.16 .29** .17 ⫺.01 ⫺.14 ⫺.30** (.81)
17. Silence 1.95 0.58 ⫺.06 ⫺.02 ⫺.27** .04 .03 ⫺.01 .11 ⫺.12 ⫺.01 .25* ⫺.27** ⫺.22* .15 .24* .44** ⫺.55** (.74)
a
“IVT” is “implicit voice theory.”n⫽94. Variables 1–14 were measured at time 1, variable 15 at time 2, and variables 16–17 at time 3.
b
Variables 3–7 and 9–17 were measured on five-point scales, with these exceptions: 10, four-point scale, and 11, seven-point scale.
c
Values on the diagonal in parentheses are scale reliabilities; reliability for the second-order self-protective implicit voice theory construct was calculated as 1/(1⫹1/x), where xis
the sum of five
␥
2
/psi terms, estimated using CFA in LISREL. For an illustration of this technique, see www.ssicentral.com/lisrel/techdocs/validity.pdf.
*p⬍.05
** p⬍.01

Given the moderate correlations among the five
specific implicit voice theories (in Study 3 and in
this study), we expected and found the fit indexes
for a CFA model loading the 20 voice theory items
onto five first-order factors and then onto a higher-
order self-protective voice theory latent factor to be
acceptable but not high (e.g., RMSEA ⫽.08). Also
as expected, fit indexes for this higher-order latent
construct model were lower than those for a model
with only five first-order implicit voice theory con-
structs (e.g., RMSEA ⫽.06), but much better than
those for a model that simply bypassed the five
first-order constructs and loaded all 20 items di-
rectly onto a single latent factor (e.g., RMSEA ⫽.14,
⌬
␹
2
⫽130.9, ⌬df ⫽5, p⬍.01). Additional models
also confirmed the independence of the implicit
voice theory, voice, and silence constructs (i.e., any
model that combined items for these three factors
into two or fewer factors produced significantly
poorer model fit indexes and chi-square statis-
tics). For example, a three-factor model distin-
guishing the implicit voice theories, voice, and
silence fit significantly better than a two-factor
model combining silence and voice (⌬
␹
2
⫽17.6,
⌬df ⫽2, p⬍.01).
Hypothesis test. We examined our hypothesis
using hierarchical ordinary least squares regression
analysis. We first entered all control variables,
which collectively explained 19 percent of the vari-
ance in silence; Table 4 (model 1) presents results.
In a second model, we added the implicit voice
theory second-order latent factor. This factor is sig-
nificantly related to silence (
␤
⫽0.39, p⬍.01) and
explains an additional 12 percent of the variance in
TABLE 4
Study 4: Implicit Voice Theories as Predictors of Silence
a
Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7
Individual differences
Gender ⫺0.13
a
⫺0.12 ⫺0.11 ⫺0.13 ⫺0.10 0.01 0.00
Age 0.06 0.01 ⫺0.01 0.01 ⫺0.02 0.06 0.09
Assertiveness ⫺0.23 ⫺0.18 ⫺0.20 ⫺0.18 ⫺0.18 0.19 0.15
Vulnerability ⫺0.09 ⫺0.10 ⫺0.13 ⫺0.09 ⫺0.10 0.13 0.14
Proactive personality 0.06 0.08 0.07 0.10 0.06 0.05 0.05
Positive affect 0.02 ⫺0.02 0.00 ⫺0.03 ⫺0.00 ⫺0.11 ⫺0.08
Negative affect 0.12 0.17 0.16 0.16 0.18 ⫺0.10 ⫺0.13
Managerial status ⫺0.08 ⫺0.05 ⫺0.03 ⫺0.04 ⫺0.05 0.00 ⫺0.02
Job satisfaction 0.16 0.12 0.09 0.11 0.14 0.03 0.06
Contextual variables
Organizational centralization 0.19 0.04 0.48 0.02 0.94 0.06 0.15
Work unit psychological safety ⫺0.14 ⫺0.08 ⫺0.47 ⫺0.08 ⫺0.61 0.19 0.15
Leader openness 0.04 ⫺0.10 0.30 ⫺0.09 0.27 0.09 0.18
Leader abusiveness 0.11 0.03 ⫺0.03 ⫺0.04 0.79 0.25 0.26
IVT-specific leader behaviors 0.09 ⫺1.40 ⫺0.22 ⫺0.15
Implicit voice theories
Self-protective IVT latent factor 0.39 0.62 0.38** 0.31 ⫺0.27*
Context by IVT interactions
Organizational centralization ⫻IVT ⫺0.61 ⫺1.20
Climate for psychological safety ⫻IVT 0.52 0.77
Leader openness ⫻IVT ⫺0.54 ⫺0.51
Leader abusiveness ⫻IVT 0.08 ⫺0.94
IVT-specific leader behaviors ⫻IVT 1.81
R
2
0.19 0.30 0.31 0.31 0.34 0.14 0.20
⌬R
2b
0.12 0.01 0.03 0.06
⌬F12.87** 0.25 0.71 5.2*
a
Standardized regression coefficients are reported. “IVT” is “implicit voice theory.” n⫽94.
b
For the model 2 change in R
2
(⌬R
2
), the baseline model is model 1; for model 3, the baseline is model 2; for model 5, the baseline is
model 4; and for model 7, baseline is model 6.
*p⬍.05
** p⬍.01
2011 479Detert and Edmondson

silence beyond that explained by the controls
(⌬F⫽12.87, p⬍.01).
2
This result offers clear
support for our hypothesis.
Robustness Checks
Context moderation. To examine whether the
main effect of implicit voice theories on silence
masked a more complex pattern of amplified or
attenuated impact driven by contextual features,
we examined the results of a model with four in-
teraction terms, each testing a context variable
commonly studied in the voice literature. As
shown in model 3 of Table 4, these results provide
no evidence of an interaction between aspects of
respondents’ current organizational, team, or lead-
ership context and implicit voice theories. Al-
though our sample size limited power to detect
such effects, none of the interaction terms ap-
proached significance, and adding the four interac-
tion terms together did not appreciably improve the
explanatory power of the model (⌬F⫽0.25, n.s.).
Mediation. To examine whether implicit voice
theories mediate between the contextual factors
studied and silence, we used two statistical ap-
proaches. First, we tested whether the theories me-
diate relationships between team psychological
safety, organizational centralization, leader open-
ness, and leader abuse and silence using the Baron
and Kenny (1986) four-step method. In all cases,
the first two conditions were not both met; thus,
this approach did not reveal evidence of mediation.
Second, where appropriate, we conducted Sobel
(1982) tests to examine distal mediation (MacKin-
non, Lockwood, Hoffman, West, & Sheets, 2002)
and again found no statistical evidence to suggest
that the implicit voice theories studied here were
mediating an indirect relationship between the
contextual influences and silence.
We also conducted mediation tests to consider
the possibility that mediation ran in the opposite
direction, as suggested by implicit leadership the-
ory (Lord & Maher, 1991) and demonstrated in re-
cent psychological research (Critcher & Dunning,
2009). Here, too, we found no evidence that im-
plicit voice theories influenced how current con-
textual factors were perceived, precluding as plau-
sible a model in which perceptions of context
mediate between implicit voice theories and
silence.
Implicit voice theory–specific leader behav-
iors. As an additional check on the robustness of
our results, we ran a set of regressions adding an-
other leader behavior variable to models 2 and 3.
Because we reasoned that leader openness and
leader abusiveness, as general constructs, might not
be specific enough to affect silence through im-
plicit voice theories or to moderate the impact they
have on speaking up to a leader, we developed and
included in the regressions a measure of leader
behaviors that directly related to the theories. To
illustrate, an employee might hold implicit theories
about not embarrassing his/her boss in public be-
cause of witnessing the boss reacting angrily when
others brought new issues up in front of coworkers.
In this case, the manager’s specific behavior, not an
implicit theory, could be said to have led to em-
ployee silence. Lacking extant measures of these
implicit voice theory–specific leader behaviors, we
developed ten items (two per implicit theory) to
assess whether our respondents’ current bosses had
engaged in specific behaviors related to the theo-
ries. Respondents were asked to rate each of the ten
statements on a seven-point scale ranging from
“highly inaccurate” to “highly accurate” as a de-
scription of their current direct boss. Instructions
for these questions stated this: “We are interested
here in understanding whether you have actually
experienced the following behaviors and outcomes,
not your beliefs or general impressions about
them.” To illustrate, items included in the implicit
voice theory–specific leader behavior scale captur-
ing experiences related to presumed target identi-
fication, need for data or solutions, and don’t by-
pass the boss upward, respectively, were: “My boss
gets upset when people point out problems with
work routines that s/he has spent time developing
or supporting,” “People in my unit are told not to
bring up problems unless they can present clear
solutions,” and “My boss uses words and actions
indicating that pointing out things needing im-
provement to those higher in the organization is a
sign of disloyalty to her/him.” Estimated reliability
for this ten-item scale was .88.
We first retested our hypothesis by adding the
implicit voice theory–specific leader behaviors
variable to the model (model 2) that included all
the other control variables and the implicit voice
theories. As shown in model 4 in Table 4, the
implicit voice theory latent factor remained a sig-
nificant predictor (
␤
⫽0.38, p⬍.01) of silence, and
the new leader behavior control variable was not a
significant predictor of silence. Further, the trivial
2
Whereas the implicit voice theory latent factor re-
sults in a significant ⌬Fand increase in adjusted R
2
when
added to a model with all controls, the reverse is not true:
when the latent variable is entered alone in step 1, there
is no significant increase in variance explained by adding
the 13 control variables (⌬F⫽0.97, n.s.; ⌬-adjusted-R
2
⫽–.003).
480 JuneAcademy of Management Journal

change in R
2
(⬃1%) between models 2 and 4 sug-
gests that although the general implicit theories
explain significant variance in silence (about 12
percent in both models), implicit voice theory–re-
lated behaviors by current bosses do not. Our re-
sults are thus robust to the inclusion of a variable
assessing current bosses’ implicit voice theory–spe-
cific behaviors.
We also reexamined the case for moderation and
mediation involving the theories. As shown in
model 5, the five terms interacting implicit voice
theories and the respective context variables do
not, collectively, produce any increase in the ex-
planatory power of the model predicting silence
(⌬F⫽0.71, n.s.). There is also no evidence for
mediation using this new leader behavior variable.
Predicting voice. As a final robustness check, we
addressed the concern that the relationship be-
tween implicit voice theories and silence may re-
flect the self-censoring nature of both the beliefs
and the associated behavior by considering the in-
cremental validity of the theories latent factor on
voice. Recall, as argued earlier and supported by
the correlation of ⫺.55 between our measure of
silence and Van Dyne and LePine’s measure of
voice, that silence and voice are not direct oppo-
sites; nonetheless, confidence in our findings and
our hypothesis can be strengthened by showing
that the incremental predictive validity of the the-
ories on silence is replicated for the voice measure.
We thus added the implicit voice theory latent
factor (see model 7) to a controls-only model
(model 6) and found a 6 percent increase in the
amount of voice explained (⌬F⫽5.2, p⬍.05).
DISCUSSION
This article advances research on employee si-
lence by investigating implicit theories of voice
operating in organizational hierarchies. In four
studies employing multiple methods to counterbal-
ance and supplement each other’s contributions,
using data from several hundred adults with varied
work experiences, we identified, measured, and
examined relationships with theoretically relevant
correlates for five implicit theories about speaking
up. Our results provide initial evidence that im-
plicit voice theories help explain employee silence,
adding substantially to prior explanations that have
focused on dispositional or demographic character-
istics of employees and on organizational context.
Together, the studies provide a platform for future
research on implicit theories as an influence on
workplace voice and silence.
Theoretical Contributions
Our focus on self-protective implicit voice theo-
ries advances understanding of reluctance to speak
up at work. The traditional view is that contextual
factors, such as the behavior of an angry boss or the
memory of prior experiences (e.g., what happened
last time one spoke up to person X about topic Y),
give rise to a general conclusion that voice is unsafe
(e.g., Milliken et al., 2003). Our research extends
this view by explaining that people hold implicit
theories relating certain situational cues to negative
consequences that are such that specific features of
the content, context, and target in a latent voice
episode can trigger an implicit theory, thereby lead-
ing to silence. This explanation complements the-
ory maintaining that general features of an employ-
ee’s work environment make voice more or less
risky. The implicit voice theories we identified are
not random but, rather, consistent with what is
known about reluctance to challenge authority
(Milgram, 1974), fear of being ostracized (Williams,
2001), and the tendency to prevent embarrassment
by avoiding face-threatening remarks (Brown &
Levinson, 1987; Morand, 1996). Moreover, our
findings show that sometimes unwillingness to
speak up is not experienced as intense, discrete fear
but rather as a sense of inappropriateness; voice
seems risky because it seems wrong or out of place.
We also argued that because knowledge struc-
tures about undesired consequences of speaking up
are formed throughout a life rather than merely
through recent experience (Kish-Gephart et al.,
2009), we expected implicit theories to be triggered
by relevant stimulus cues and applied in a cogni-
tively automatic, top-down fashion. The evidence
presented in Study 4—that implicit voice theories
show an independent main effect on silence, rather
than mediation of (or moderation by) local condi-
tions—provides initial support for this view. These
findings imply that models that explain silence
with leader behavior and other contextual factors
are incomplete in important ways. In these leader-
centric views, which have dominated the voice lit-
erature, employees choose silence because leader
behavior has caused them to conclude that it is
unsafe or futile to speak up (e.g., Burris et al., 2008;
Nembhard & Edmondson, 2006). In this study,
however, the perceived behaviors of actual leaders
were only modestly correlated with implicit theo-
ries about speaking up to leaders. This finding is
consistent with issue selling research, in which
quantitative assessments of current management
behavior (e.g., top management openness) have
been largely unrelated to the perceived “image
risks” of speaking up, even though qualitative stud-
2011 481Detert and Edmondson

ies have shown that informants attribute the risks
to leader behavior (Ashford et al., 1998). We thus
offer a complementary “follower-centric” perspec-
tive (Shamir, 2007).
In follower-centric models, it is what employees
believe about leaders in general (based on a lifetime
of prototype building) that drives behavior toward
leaders (often in an automatic fashion), not objec-
tive characteristics or behaviors of actual leaders
(Epitropaki & Martin, 2004; Lord & Maher, 1991). In
essence, this perspective suggests that follower be-
liefs are vital to explanation of outcomes usually
considered leadership effects. For research on
voice and silence, a follower-centric perspective
suggests that individuals bring implicit voice theo-
ries to work and automatically use them as filters
for attending to and interpreting current stimuli.
When features of a latent voice episode trigger a
specific belief that it is unwise to speak up about
this issue,orinthat venue, then presumptions
about leader reactions can be more influential on
behavior than a leader’s actual behavior or desires.
In short, employees’ silence can be thought of as
influenced as much by their own cognitive frame-
works as caused by current bosses’ behaviors or
other organizational factors.
We do not intend to suggest that actual boss
behavior and direct experiences at work do not
influence employee silence. Clearly, a domineering
or abusive boss, or a recent experience of being
reprimanded for speaking up, will influence the
behavior of employees (Burris et al., 2008). (Indeed,
although not significant in the regression models
containing many variables, several leader behavior
and other contextual factors showed significant bi-
variate correlations with employee silence.) Ac-
cording to a follower-centric perspective, employee
silence does not require bosses acting in intimidat-
ing ways; implicit theories about bosses are suffi-
cient, just as face saving and many other social
behaviors are largely driven by internalized knowl-
edge structures (Bacharach et al., 2000; Goffman,
1974). Although bosses figure prominently in most
of the implicit voice theories we uncovered, “the
boss” can be a hypothetical or archetypical author-
ity, constructed over many years and experiences,
rather than an accurate reflection of an employee’s
current manager or managers. This observation
helps to explain why many managers are surprised
to learn that people are afraid to speak up to them.
If managers assume that others’ fear has to be a
reaction to them, or to something they have done,
they will be puzzled when others are reluctant to
speak up despite what they accurately see as a lack
of intimidating behavior. Ironically, the beliefs of a
boss may be the opposite of the beliefs contained in
an implicit voice theory. For example, contradict-
ing the “presumed target identification” implicit
theory, a manager may genuinely wish for honest
input about projects and be unaware that others
withhold that input because they tacitly assume
that to offer it could be taken as criticism of the
manager.
Limitations and Future Research Directions
Despite our use of multiple methods, study de-
signs, and analyses to support our core theoretical
arguments, this research has limitations, and much
work remains to be done to understand the nature,
antecedents, and consequences of implicit voice
theories. Our studies rely on self-reports of the spe-
cifics in actual latent voice episodes, for reactions
to an open-ended question, and for the measures of
implicit voice theories. And Studies 3 and 4 used
explicit measures (the most common method for
assessing implicit theories) of the theories of inter-
est. Although we took steps to alleviate common
method concerns (including measurement over
time periods, use of positive and negative affect as
controls, and factor analysis to establish discrimi-
nation among constructs) (Podsakoff et al., 2003), it
is possible that respondents were reporting “more
than they can know” about their beliefs via this
approach (Nisbett & Wilson, 1977). The consistent
emergence of similar implicit theories in analyses
using multiple methods and samples, together with
the results in Studies 3 and 4 showing discriminant
validity from potential correlates and criterion va-
lidity with voice and silence measured at a later
time, strengthen our confidence in the validity of
our approach and results; however, further re-
search is required to test and extend the implicit
theories we identified. This should include re-
search establishing the predictive validity of im-
plicit voice theories with additional controls not
examined here (e.g., conscientiousness, agreeable-
ness, self-esteem [LePine & Van Dyne, 2001]) and
research comparing how survey measures relate to
probes assessing implicit voice theories with more
implicit methods (e.g., association tests or primes).
Because our primary theoretical and empirical
goal was to establish the importance of the concept
of implicit voice theories in explaining silence, we
developed measures and examined relationships
for the five theories most commonly identified in
Studies 1 and 2. Despite the size and breadth of
these samples, other implicit voice theories un-
doubtedly can be identified and perhaps even fit
into a parsimonious taxonomy based on function or
motive. We focused on five self-protective theories
but, as noted in Study 2, we recognize that others
482 JuneAcademy of Management Journal

exist in this category and that implicit voice theo-
ries serving other purposes also exist; such pur-
poses include protecting oneself or others from so-
cial ostracism or feelings of impotence. Systematic
study across national cultures—especially cultures
differing in “power distance” (House, Hanges, Javi-
dian, Dorfman, & Gupta, 2004)—may prove par-
ticularly fruitful for comprehensive identification
of implicit theories governing silence. Future work
could also be done to refine and further test the
survey measures developed for the five focal im-
plicit voice theories of this study. Although several
statistics (e.g., item variances, item-to-total scale
correlations) suggested that our items were not
overly complex, future research might use even
simpler, more direct language to appropriately as-
sess these and other implicit voice theories.
Our goals in Study 4 were conducive to using a
single higher-order latent variable to represent the
five theories. Additional analyses showing that the
five theories entered alone (after all control vari-
ables) each significantly predicted silence strength-
ened our conclusions, as did the finding that the
five implicit voice theories, entered as a set of first-
order variables, predict an increment in silence
(⌬R
2
⫽15.5%) similar to the single higher-order
latent variable. However, future research should
investigate specific implicit voice theories for pre-
dicting specific voice-related criterion variables.
Attention to how implicit voice theories develop
and shape behavior over time will also advance
understanding of this important phenomenon. The
mixed and modest correlations between the theo-
ries studied here and an array of individual differ-
ences and contextual factors, together with the pre-
liminary lack of evidence that the former mediate
the latter’s influence on silence, suggest a need for
future research on the causes of implicit voice the-
ories. This might include identification of salient
experiences in individual work histories, not just
examining current leadership chains (Detert &
Trevino, 2010). In some cases, a single distant in-
cident or life event may stamp into memory long-
enduring beliefs about speaking up (Tooby & Cos-
mides, 1990). For example, one interviewee vividly
recalled an incident 12 years and three managers
ago as the basis for his belief that speaking up is
risky (cited in Kish-Gephart et al. [2009]). Research
might also consider how nonwork experiences that
shape orientation to authority in hierarchical set-
tings (such as childhood and adolescent experi-
ences in families, schools, and other institutions)
endure in the form of implicit voice theories. Pa-
rental style (authoritative, authoritarian, or permis-
sive [Baumrind, 1967]) and perceptions of parents’
work experiences (Barling, Dupre, & Hepburn,
1998), for example, may affect the strength with
which implicit voice theories are held in adult-
hood. In short, if people are born with a basic
grammar for orienting themselves toward authori-
ties, research must seek to identify the types of
experiences that produce varied life stories from
similar starting points (Fiske, 1991).
To uncover the sources of implicit voice theories,
and to distinguish which are shared features of
human socialization and which, if any, are specific
to individual, organizational, and national experi-
ences, research should be conducted on large sam-
ples of individuals from selected cohorts, organiza-
tions, and countries. Such research, especially if
longitudinal, will help uncover the social dynam-
ics through which implicit theories are formed and
reinforced, leading, for example, to particularly
strong “climates of silence” in some organizations
(Morrison & Milliken, 2000). In sum, scholars need
to better understand the factors that shape the de-
velopment, use, reinforcement, and overcoming of
implicit voice theories (Milliken et al., 2003). This
understanding may require researchers to expand
their theoretical (e.g., draw from evolutionary psy-
chology, anthropology, child and adult develop-
ment) and empirical (e.g., use life history narra-
tives, subliminal priming approaches, or PET or
fMRI scans to measure brain activity) scope.
This research did not find evidence for modera-
tion of the impact of implicit voice theories on
silence, despite its use of constructs often used in
voice research: psychological safety, centralized
decision making, and leader openness and abusive-
ness. During the review process for this article, we
collected additional data (from 90 percent of our
Study 4 sample) to explore (post hoc) the potential
moderating role of perceived alternative employ-
ment possibilities. We tested whether access to al-
ternative employment lowered the implicit theory
effects on silence, but the results suggested instead
that better alternative employment is related to
lower silence only for those holding the implicit
voice theories at low levels. Specifically, we found
a significant interaction effect (on silence) for the
interaction of alternative employment times self-
protective implicit voice theories, and a plot of the
interaction revealed that among respondents with
high perceived access to alternative employment,
silence was lower only for those reporting low
agreement with the implicit voice theories (
␤
⫽
0.08, s.e. ⫽.02, p⬍.01). (For respondents with low
perceived access to alternative employment, si-
lence did not change with their level of agreement
with the implicit voice theories [
␤
⫽0.02, s.e. ⫽
.02, p⫽.34, n.s.]). In short, we did not find evi-
dence that strongly held implicit voice theories are
2011 483Detert and Edmondson

attenuated by contextual factors, such as access to
alternative employment, that might indicate less
need for their use.
These results, though initially counterintuitive,
are consistent with the view of implicit voice the-
ories proposed in this article. Specifically, we ar-
gue that implicit voice theories develop from the
hard-wired motive of self-protection and are thus
taken-for-granted, biased toward false positives,
and rarely tested against evidence; therefore, as-
pects of a person’s current context (such as an open
boss or a decentralized organizational structure)
rarely grab sufficient cognitive attention to override
self-protective implicit theories. In this way, the
implicit theories people hold drive behavior when
basic stimulus cues trigger those theories (Bargh,
1997; Ehrlinger & Dunning, 2003). For example, if a
person believes that speaking up in public is risky,
silence will occur during public latent voice epi-
sodes even if the current context includes a good
boss, a supportive culture, or little other basis for
concern about job security or employability.
We do not intend to suggest that implicit voice
theories cannot be moderated or overcome. In the
chaos of competing stimulus cues, “it is the action
schema with the strongest activation level that tri-
umphs in this battle for cognitive supremacy and
guide’s one’s behavior” (Macrae & Johnston, 1998:
404). Thus, for example, schema-driven helping
will lead to picking up another’s dropped pen,
unless the brain’s “supervisory attention system”
(Norman & Shallice, 1986) shifts focus to another
stimulus in the situation—such as messy leaking
ink (Macrae & Johnson, 1998)—and overrides the
initial automatic tendency. Similarly, implicit
voice theories, despite an evolutionary basis, may
be overridden in voice episodes in which suffi-
ciently strong competing environmental stimuli
lead the “prioritizing algorithms” in an individu-
al’s brain to activate alternative emotions, sche-
mata, and action tendencies (Cosmides & Tooby,
2000). For example, situations involving behavior
that is strongly offensive, illegal, or physically dan-
gerous may evoke sufficient anger to override the
withdrawal tendencies associated with implicit
voice theories. As Kish-Gephart and her colleagues
noted, “Higher intensity empathetic anger experi-
enced when a close colleague is belittled, blatantly
discriminated against, or unfairly blamed for a
costly mistake may trigger an automatic response
that over-rides fear’s silencing effects” (2002: 182).
Such appears to be the case for individuals who
blow the whistle despite the fear of retaliation (He-
nik, 2008). Future research may include experi-
ments to identify real-time cues that may lead to
voice despite high levels of one or more implicit
voice theories. Such research would allow con-
trolled study of the necessary conditions for estab-
lishing episodic attenuation or amplification of im-
plicit theories.
Alternatively, field research could investigate or-
ganizational contexts (naturally occurring or ma-
nipulated in quasi-experiments) where norms or
practices counteract the tendency toward silence
associated with specific implicit voice theories. For
example, researchers might study the impact of
managerial statements targeted to contradict com-
mon theories, such as, “I want to hear about con-
cerns even if you don’t yet have solutions, because
we can work on solutions together” or “I want you
to speak up honestly at all times when you have a
problem or idea, no matter who is in the room,
including my boss.”
Conclusion
Implicit voice theories present subtle barriers to
organizational learning. Even when managers do
not behave in ways that actively stifle voice, im-
plicit voice theories can block valuable knowledge
from being shared. Enabling individuals to speak
up to those in power is inherently challenging,
given the presumed risk-reward asymmetry that
favors silence. In short, voice’s benefits are primar-
ily collective (organizational), but individuals bear
voice’s costs. The implicit voice theories discov-
ered in this research add to this simple observation,
exacerbating the asymmetry and increasing the pull
of silence. Viewed this way, our findings can be
seen as reason for further pessimism about the ideal
of an organization that elicits and puts to use the
ideas and intelligence of all its members. Managers
appear saddled not only by their own actual behav-
iors inhibiting voice but also by subordinate beliefs
about managers. However, although managers may
have little choice about whether employees bring
implicit voice theories to work, they do have dis-
cretion about whether to proactively manage them
or to let them be silent killers of voice. Proactive
management would involve acknowledging that
people hold beliefs that foster silence, explicitly
stating theories that contradict those beliefs, and
continuously working to dispel them by modeling,
inviting, and rewarding speaking up.
REFERENCES
Abelson, R. P. 1976. Script processing in attitude forma-
tion and decision making. In J. S. Carroll & J. W.
Payne (Eds.), Cognition and social behavior: 33–46.
Hillsdale, NJ: Erlbaum.
Ajzen, I. 1991. The theory of planned behavior. Organi-
484 JuneAcademy of Management Journal

zational Behavior and Human Decision Processes,
50: 179–211.
Anderson, J. C., & Gerbing, D. W. 1991. Predicting the
performance of measures in a confirmatory factor
analysis with a pretest assessment of their substan-
tive validities. Journal of Applied Psychology, 76:
732–740.
Anderson, C. A., & Lindsay, J. L. 1998. The development,
perseverance, and change of naı¨ve theories. Social
Cognition, 16: 8–30.
Ashford, S. J., Rothbard, N. P., Piderit, S. K., & Dutton,
J. E. 1998. Out on a limb: The role of context and
impression management in selling gender-equity is-
sues. Administrative Science Quarterly, 43: 23–57.
Bacharach, S. B., Bamberger, P., & McKinney, V. 2000.
Boundary management tactics and logics of action:
The case of peer-support providers. Administrative
Science Quarterly, 45: 704–736.
Bagozzi, R. P., & Phillips, L. W. 1982. Representing and
testing organizational theories: A holistic construal.
Administrative Science Quarterly, 27: 459–489.
Bandura, A. 1986. Social foundations of thought and
action: A social cognitive theory. Englewood Cliffs,
NJ: Prentice Hall.
Bargh, J. A. 1997. The automaticity of everyday life. In
R. S. Wyer (Ed.), Advances in social cognition, vol.
10: 1–61. Mahwah, NJ: Erlbaum.
Bargh, J. A., & Ferguson, M. J. 2000. Beyond behaviorism:
On the automaticity of higher mental processes. Psy-
chological Bulletin, 126: 925–945.
Barling, J., Dupre, K. E., & Hepburn, C. G. 1998. Effect of
parents’ job insecurity on children’s work beliefs
and attitudes. Journal of Applied Psychology, 83:
112–118.
Baron, R. M., & Kenny, D. A. 1986. The moderator-medi-
ator variable distinction in social psychological re-
search: Conceptual, strategic, and statistical consid-
erations. Journal of Personality and Social
Psychology, 51: 1173–1182.
Bateman, T. S., & Crant, J. M. 1993. The proactive com-
ponent of organizational behavior: A measure and
correlates. Journal of Organizational Behavior, 14:
103–118.
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., &
Vohs, K. D. 2001. Bad is stronger than good. Review
of General Psychology, 5: 323–337.
Baumrind, D. 1967. Child care practices anteceding three
patterns of preschool behavior. Genetic Psychology
Monographs, 75: 43–88.
Brown, M. E., Trevino, L. K., & Harrison, D. A. 2005.
Ethical leadership: A social learning theory perspec-
tive for construct development. Organizational Be-
havior and Human Decision Processes, 97: 117–
134.
Brown, P., & Levinson, S. C. 1987. Politeness: Some
universals in language usage. Cambridge, U.K.:
Cambridge University Press.
Burris, E. R., Detert, J. R., & Chiaburu, D. S. 2008. Quitting
before leaving: The mediating effects of psychologi-
cal attachment and detachment on voice. Journal of
Applied Psychology, 93: 912–922.
Buss, D. M. 2009. How can evolutionary psychology suc-
cessfully explain personality and individual differ-
ences? Perspectives on Psychological Science, 4:
359–366.
Chiu, C., Dweck, C. S., Tong, J. Y., & Fu, J. H. 1997
Implicit theories and conceptions of morality. Jour-
nal of Personality and Social Psychology, 73: 923–
94.
Chiu, C., Hong, Y., & Dweck, C. S. 1997. Lay disposition-
ism and implicit theories of personality. Journal of
Personality and Social Psychology, 73: 19–3.
Cosmides, L., & Tooby, J. 1992. Cognitive adaptations for
social exchange. In J. H. Barkow, L. Cosmides, & J.
Tooby (Eds.), The adapted mind: 163–228. Oxford,
U.K.: Oxford University Press.
Cosmides, L., & Tooby, J. 2000. Evolutionary psychology
and the emotions. In M. Lewis & J. M. Haviland-
Jones (Eds.), Handbook of emotions (2nd ed.): 91–
115. New York: Guilford.
Critcher, C. R., & Dunning, D. 2009. How chronic self-
views influence (and mislead) self-assessments of
task performance: Self-views shape bottom-up expe-
riences with the task. Journal of Personality and
Social Psychology, 97: 931–945.
Cropanzano, R., James, K., & Konovsky, M. A. 1993.
Dispositional affectivity as a predictor of work atti-
tudes and job performance. Journal of Organiza-
tional Behavior, 14: 595–606.
Detert, J. R., & Burris, E. R. 2007. Leadership behavior
and employee voice: Is the door really open? Acad-
emy of Management Journal, 50: 869–884.
Detert, J. R., & Trevino, L. K. 2010. Speaking up to higher
ups: How supervisors and skip-level leaders influ-
ence employee voice. Organization Science, 21:
249–270.
Dutton, J. E., & Ashford, S. J. 1993. Selling issues to top
management. Academy of Management Review,
18: 397–428.
Dutton, J. E., Ashford, S. J., Lawrence, K. A., & Miner-
Rubino, K. 2002. Red light, green light: Making sense
of the organizational context for issue selling. Orga-
nization Science, 13: 355–369.
Dutton, J. E., Ashford, S. J., O’Neill, R. M., Hayes, E., &
Wierba, E. E. 1997. Reading the wind: How middle
managers assess the context for selling issues to top
managers. Strategic Management Journal, 18: 407–
425.
Edmondson, A. C. 1999. Psychological safety and learn-
ing behavior in work teams. Administrative Science
Quarterly, 44: 350–383.
2011 485Detert and Edmondson

Edmondson, A. C. 2002. The local and variegated nature
of learning in organizations: A group-level perspec-
tive. Organization Science, 13: 128–146.
Edmondson, A. C. 2003. Speaking up in the operating
room: How team leaders promote learning in inter-
disciplinary action teams. Journal of Management
Studies, 40: 1419–1452.
Ehrlinger, J., & Dunning, D. 2003. How chronic self-views
influence (and potentially mislead) estimates of per-
formance. Journal of Personality and Social Psy-
chology, 84: 5–17.
Ekman, P., O’Sullivan, M., & Frank, M. G. 1999. A few
can catch a liar. Psychological Science, 10: 363–
366.
Epitropaki, O., & Martin, R. 2004. Implicit leadership
theories in applied settings: Factor structure, gener-
alizability, and stability over time. Journal of Ap-
plied Psychology, 89: 293–31.
Ermer, E., Cosmides, L., & Tooby, J. 2008. Relative status
regulates risky decision making about resources in
men: Evidence for the co-evolution of motivation
and cognition. Evolution and Human Behavior, 29:
106–118.
Fiske, A. P. 1991. Structures of social life. New York:
Free Press.
Furnham, A. 1988. Lay theories: Everyday understand-
ing of problems in the social sciences. Oxford, U.K.:
Pergamon Press.
Goffman, E. 1974. Frame analysis. Boston: Northeastern
University Press.
Gollwitzer, P. M., & Brandstatter, V. 1997. Implementa-
tion intentions and effective goal pursuit. Journal of
Personality and Social Psychology, 73: 186–199.
Grant, A. M., & Ashford, S. J. 2008. The dynamics of
proactivity at work. In B. M. Staw & A. Brief (Eds.),
Research in organizational behavior, vol. 28: 3–34.
Greenwich, CT: JAI.
Hackman, J. R., & Oldham, G. R. 1975. Development of
the Job Diagnostic Survey. Journal of Applied Psy-
chology, 60: 159–170.
Hage, J., & Aiken, M. 1969. Routine technology, social
structure, and organizational goals. Administrative
Science Quarterly, 14: 366–376.
Harlos, K. P. 2001. When organizational voice systems
fail. Journal of Applied Behavioral Science, 37:
324–342.
Harrison, D. A., & McLaughlin, M. E. 1993. Cognitive
processes in self-report responses: Tests of item con-
text effects in work attitude measures. Journal of
Applied Psychology, 78: 129–140.
Heider, F. 1958. The psychology of interpersonal rela-
tions. New York: Wiley.
Henik, E. 2008. Mad as hell or scared stiff? The effects of
value conflict and emotions on potential whistle-
blowers. Journal of Business Ethics, 80: 111–119.
Hertel, G., & Kerr, N. L. 2001. Priming in-group favorit-
ism: The impact of normative scripts in the minimal
group paradigm. Journal of Experimental Social
Psychology, 37: 316–324.
Hong, Y., Morris, M. W., Chiu, C., & Benet-Martinez, V.
200. Multicultural minds: A dynamic constructivist
approach to culture and cognition. American Psy-
chologist, 55: 709–772.
House, R. J., Hanges, P. J., Javidian, M., Dorfman, P. W., &
Gupta, V. 2004. Culture, leadership, and organiza-
tions: The GLOBE study of 62 societies. Thousand
Oaks, CA: Sage.
International Personality Item Pool. 2009. A scientific
collaboratory for the development of advanced
measures of personality traits and other individual
differences. http://ipip.ori.org/. Accessed April 7,
2009.
Kelly, G. A. 1955. The psychology of personal con-
structs. New York: Norton.
Kish-Gephart, J., Detert, J. R., Trevino, L. K., & Edmond-
son, A. C. 2009. Silenced by fear: Psychological,
social, and evolutionary drivers of voice behavior at
work. In B. M. Staw & A. P. Brief (Eds.), Research in
organizational behavior, vol. 29: 163–193. Green-
wich, CT: JAI.
LeDoux, J. E. 1996. The emotional brain. New York:
Simon & Schuster.
LePine, J. A., & Van Dyne, L. 1998. Predicting voice
behavior in work groups. Journal of Applied Psy-
chology, 83: 853–868.
LePine, J. A., & Van Dyne, L. 2001. Voice and cooperative
behavior as contrasting forms of contextual perfor-
mance: Evidence of differential relationships with
Big Five personality characteristics and cognitive
ability. Journal of Applied Psychology, 86: 326–
336.
Levy, S. R., Chiu, C., & Hong, Y. 2006. Lay theories and
intergroup relations. Group Processes and Inter-
group Relations, 9: 5–24.
Levy, S. R., Stroesser, S. J., & Dweck, C. S. 1998. Stereo-
type formation and endorsement: The role of im-
plicit theories. Journal of Personality and Social
Psychology, 74: 1421–1436.
Lincoln, Y. S., & Guba, E. G. 1985. Naturalistic inquiry.
Newbury Park, CA: Sage.
Lord, R. G., & Maher, K. J. 1991. Leadership and infor-
mation processing. Boston: Unwin/Hyman.
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M.,
West, S. G., & Sheets, V. 2002. A comparison of
methods to test mediation and other intervening
variable effects. Psychological Methods, 7: 83–104.
Macrae, C. N., & Johnston, L. 1998. Help, I need some-
body: Automatic action and inaction. Social Cogni-
tion, 16: 400–417.
Magee, J. C., & Galinsky, A. D. 2008. Social hierarchy:
486 JuneAcademy of Management Journal

The self-reinforcing nature of power and status. In
J. P. Walsh & A. P. Brief (Eds.), Academy of Man-
agement annals, vol. 2: 351–398. Essex, U.K.:
Routledge.
Miceli, M. P., & Near, J. P. 1992. Blowing the whistle:
The organizational and legal implications for com-
panies and employees. New York: Lexington.
Miles, M. B., & Huberman, A. M. 1994. Qualitative data
analysis (2nd ed.). Beverly Hills, CA: Sage.
Milgram, S. 1974. Obedience to authority. New York:
Harper & Row.
Milliken, F. J., Morrison, E. W., & Hewlin, P. F. 2003. An
exploratory study of employee silence: Issues that
employees don’t communicate upward and why.
Journal of Management Studies, 40: 1453–1476.
Morand, D. A. 1996. Dominance, deference and egalitar-
ianism in organizational interaction. Organization
Science, 7: 544–556.
Morrison, E. W., & Milliken, F. J. 2000. Organizational
silence: A barrier to change and development in a
pluralistic world. Academy of Management Re-
view, 25: 706–725.
Nembhard, I. M., & Edmondson, A. C. 2006. Making it
safe: The effects of leader inclusiveness and profes-
sional status on psychological safety and improve-
ment efforts in health care teams. Journal of Orga-
nizational Behavior, 27: 941–966.
Nisbett, R. E., & Wilson, T. D. 1977. Telling more than we
can know: Verbal reports on mental processes. Psy-
chological Review, 84: 231–259.
Norman, D. A., & Shallice, T. 1986. Attention to action:
Willed and automatic control of behavior. In R. J.
Davidson, G. E. Schwartz, & D. Shapiro (Eds.), Con-
sciousness and self-regulation, vol. 4: 1–18. New
York: Plenum.
Nunnally, J. C., & Bernstein, I. H. 1994. Psychometric
theory (3rd ed.). New York: McGraw-Hill.
Ohman, A. 2000. Fear and anxiety: Evolutionary, cogni-
tive, and clinical perspectives. In M. Lewis & J. M.
Haviland-Jones (Eds.), Handbook of emotions (2nd
ed.): 573–593. New York: Guilford.
Ostroff, C., Kinicki, A. J., & Clark, M. A. 2002. Substan-
tive and operational issues of response bias across
levels of analysis: An example of climate-satisfaction
relationships. Journal of Applied Psychology, 87:
355–368.
Parker, S., Bindl, U., Van Dyne, L., & Wong, S. 2009.
Measuring motives for silence. Paper presented at
the annual meeting of the Academy of Management.
Chicago.
Pinder, C. C., & Harlos, K. P. 2001. Employee silence:
Quiescence and acquiescence as responses to per-
ceived injustice. In K. M. Rowland & G. R. Ferris
(Eds.), Research in personnel and human resources
management, vol. 20: 331–369. New York: JAI.
Pinker, S. 1997. How the mind works. New York:
Norton.
Plaks, J. E., Grant, H., & Dweck, C. S. 2005. Violations of
implicit theories and the sense of prediction and
control: Implications for motivated person percep-
tion. Journal of Personality and Social Psychology,
88: 245–262.
Plutchik, R. 2003. Emotions and life: Perspectives from
psychology, biology, and evolution. Washington,
DC: American Psychological Association.
Podsakoff, P., MacKenzie, S., Lee, J., & Podsakoff, N.
2003. Common method biases in behavioral re-
search: A critical review of the literature and recom-
mended remedies. Journal of Applied Psychology,
88: 879–903.
Podsakoff, P., MacKenzie, S., Moorman, R., & Fetter, R.
1990. Transformational leader behaviors and their
effects on followers’ trust in leader, satisfaction, and
organizational citizenship behaviors. Leadership
Quarterly, 1: 107–142.
Rachman, S. 1990. Fear and courage. Gordonsville, VA:
Freeman.
Ross, M. 1989. Relation of implicit theories to the con-
struction of personal histories. Psychological Re-
view, 96: 341–357.
Ryan, K. D., & Oestreich, D. K. 1998. Driving fear out of
the workplace (2nd ed.). San Francisco: Jossey-Bass.
Shamir, B. 2007. From passive recipients to active co-
producers: Followers’ roles in the leadership pro-
cess. In B. Shamir et al. (Eds.), Follower-centered
perspectives on leadership: ix-xxxix. Greenwich,
CT: Information Age.
Sobel, M. E. 1982. Asymptotic confidence intervals for
indirect effects in structural equation models. In S.
Leinhardt (Ed.), Sociological methodology 1982:
290–312. Washington, DC: American Sociological
Association.
Sternberg, R. J. 1985. Implicit theories of intelligence,
creativity, and wisdom. Journal of Personality and
Social Psychology, 49: 607–627.
Stogdill, R. M. 1963. Manual for the Leader Behavior
Description Questionnaire—Form XII. Colum-
bus: Bureau of Business Research, Ohio State Uni-
versity.
Tangirala, S., & Ramanujam, R. 2008. Employee silence
on critical work issues: The cross level effects of
procedural justice climate. Personnel Psychology,
61: 37–68.
Tepper, B. J. 2000. Consequences of abusive supervision.
Academy of Management Journal, 43: 178–190.
Tooby, J., & Cosmides, L. 1990. On the universality of
human nature and the uniqueness of the individual:
The role of genetics and adaptation. Journal of Per-
sonality, 58: 17–67.
Van Dyne, L., Ang, S., & Botero, I. 2003. Conceptualizing
2011 487Detert and Edmondson

employee silence and voice as multidimensional
constructs. Journal of Management Studies, 40:
1359–1392.
Van Dyne, L., & LePine, J. A. 1998. Helping and voice
extra-role behaviors: Evidence of construct and pre-
dictive validity. Academy of Management Journal,
41: 108–119.
Vaughan, D. 1996. The Challenger launch decision:
Risky technology, culture, and deviance at NASA.
Chicago: University of Chicago Press.
Watson, D., Clark, L. A., & Tellegen, A. 1988. Develop-
ment and validation of brief measures of positive
and negative affect: The PANAS scale. Journal of
Personality and Social Psychology, 54: 1063–
1070.
Williams, K. D. 2001. Ostracism: The power of silence.
New York: Guilford.
Withey, M. J., & Cooper, W. H. 1989. Predicting exit,
voice, loyalty, and neglect. Administrative Science
Quarterly, 34: 521–539.
Wyer, R. S. 2004. Social comprehension and judgment:
The role of situation models, narratives, and im-
plicit theories. Mahwah, NJ: Erlbaum.
APPENDIX A
Studies 3 and 4: Implicit Voice Theory
Construct Items
Presumed Target Identification
Someone who helps create a process or routine is
likely to be offended when others suggest changes.
It’s risky to challenge existing processes because it
may be seen as questioning the wisdom of the individu-
als who established or support them.
Speaking up to suggest a better way of doing some-
thing is likely to offend the person(s) currently in charge
of the process or product you’re speaking about.
It is not good to question the way things are done
because those who have developed the routines are likely
to take it personally.
Need Solid Data or Solutions (to Speak Up)
Presenting underdeveloped, under-researched ideas to
your group is never a good idea.
To look good when speaking up with an idea or sug-
gestion you have to be able to answer every question you
get asked.
Saying “I don’t know” or “I’m not sure” when being
questioned about some aspect of a new idea you’re pre-
senting puts you in a bad position.
Unless you have clear solutions, you shouldn’t speak
up about problems.
Don’t Bypass the Boss Upward
When you speak up about problems or areas for im-
provement to your boss in front of people who are even
higher in the organization, you make your boss look bad.
Loyalty to your boss means you don’t speak up about
problems in front of his or her boss.
Pointing out possibilities for improvement in front of
other managers calls attention to the fact that my boss
didn’t identify these possibilities him/herself.
Questions that you’re not sure if your boss can answer
should not be asked in front of your boss’s boss.
Don’t Embarrass the Boss in Public
Pointing out problems or inefficiencies in front of oth-
ers is likely to embarrass the boss.
It is not a good idea to make your manager look bad in
front of the group by speaking up without telling him/her
in advance.
You should always pass your ideas by the boss in
private first, before you speak up publicly at work.
It is important to give your boss time to prepare to
discuss a problem or suggestion you have prior to bring-
ing it up in front of a group.
Negative Career Consequences of Voice
If you want advancement opportunities in today’s
world, you have to be careful about pointing out needs
for improvement to those in charge.
You are more likely to be rewarded in organizational
life by “going along quietly” than by speaking up about
ways the organization can improve.
Pointing out problems, errors, or inefficiencies might
very well result in lowered job evaluations.
Speaking up at work about possible improvements sets
you up for retribution by those above you who felt threat-
ened by your comments.
James R. Detert (jdetert@cornell.edu) is an assistant pro-
fessor of management at the Johnson Graduate School of
Management at Cornell University. His current research
interests include voice and silence in organizations, lead-
ership influences on voice and ethical decision making,
and cognitive moral disengagement as a predictor of un-
ethical behavior. He received his Ph.D. in organizational
behavior from Harvard University.
Amy C. Edmondson (aedmondson@hbs.edu) is the No-
vartis Professor of Leadership and Management at Har-
vard Business School, a chair established to enable the
study of human interactions that lead to the creation of
successful business enterprises for the betterment of so-
ciety. Edmondson’s research focuses on psychological
safety, team learning, and innovation in organizations.
She received her Ph.D. in organizational behavior from
Harvard University.
488 JuneAcademy of Management Journal