You are on page 1of 26

Journal of Applied Psychology © 2013 American Psychological Association

2014, Vol. 99, No. 1, 87–112 0021-9010/14/$12.00 DOI: 10.1037/a0034284

Speaking More Broadly: An Examination of the Nature, Antecedents, and


Consequences of an Expanded Set of Employee Voice Behaviors

Timothy D. Maynes Philip M. Podsakoff


University at Buffalo, The State University of New York University of Florida

Scholarly interest in employee voice behavior has increased dramatically over the past 15 years.
Although this research has produced valuable knowledge, it has focused almost exclusively on voice as
a positively intended challenge to the status quo, even though some scholars have argued that it need not
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

challenge the status quo or be well intentioned. Thus, in this paper, we create an expanded view of voice;
This document is copyrighted by the American Psychological Association or one of its allied publishers.

one that extends beyond voice as a positively intended challenge to the status quo to include voice that
supports how things are being done in organizations as well as voice that may not be well intentioned.
We construct a framework based on this expanded view that identifies 4 different types of voice behavior
(supportive, constructive, defensive, and destructive). We then develop and validate survey measures for
each of these. Evidence from 5 studies across 4 samples provides strong support for our new measures
in that (a) a 4-factor confirmatory factor analysis model fit the data significantly better than 1-, 2-, or
3-factor models; (b) the voice measures converged with and yet remained distinct from conceptually
related comparison constructs; (c) personality predictors exhibited unique patterns of relationships with
the different types of voice; (d) variations in actual voice behaviors had a direct causal impact on
responses to the survey items; and (e) each type of voice significantly impacted important outcomes for
voicing employees (e.g., likelihood of relying on a voicing employee’s opinions and evaluations of a
voicing employee’s overall performance). Implications of our findings are discussed.

Keywords: types of employee voice behavior, scale development, construct validation, performance
evaluations

More than four decades have passed since Hirschman (1970) (1998). These authors clarified the definition of voice and devel-
first noted the importance of behaviors wherein employees speak oped an operational measure of it. They defined voice as the
up in order to change things rather than live with an ineffective or expression of constructive challenge intended to improve how
inefficient status quo. He called this behavior employee voice and things are done (Van Dyne & LePine, 1998) and identified change-
argued that it serves the vital function of helping organizations oriented suggestions and recommendations for improvements as
adapt to the ever changing business environment. Initially, re- examples of voice. Furthermore, they argued that voice has a
search in this area had very little impact on the field of manage- positive impact on organizational functioning because it identifies
ment. For example, in the 25 years between 1970 and 1994, fewer new or better ways of doing things, directs management’s attention
than 15 articles on voice behavior appeared in management jour- to critical issues that need to be addressed, and corrects problems
nals. However, in the years since 1994 the number of published with existing work practices or procedures (LePine & Van Dyne,
articles has increased more than 10 times to over 200 manuscripts. 1998).
Scholarly interest in voice behavior was slow to materialize be- Although the rapid growth in this area over the past 15 years has
cause early researchers were hindered by a lack of conceptual substantially increased our knowledge of voice behavior, the focus
clarity regarding what voice is (Gorden, 1988), resulting in poor of this research has been overly narrow, concentrating almost
measurement (Rusbult, Farrell, Rogers, & Mainous, 1988) and exclusively on voice as a positively intended challenge to the
inconsistent empirical findings (Withey & Cooper, 1989). status quo (Van Dyne & LePine, 1998). Thus, in this paper, we
The dramatic increase in voice research observed after 1994 is introduce a new, more expansive view of voice behavior. We
likely the result of research conducted by Van Dyne and LePine believe that such a view is warranted for several reasons. First, an
expanded view of voice is supported from a conceptual perspec-
tive, given that some scholars have argued that voice need not
This article was published Online First September 16, 2013. challenge how things are done in organizations or be well inten-
Timothy D. Maynes, Department of Organization and Human Re- tioned (Gorden, 1988; Hirschman, 1970; Van Dyne, Ang, &
sources, University at Buffalo, The State University of New York; Philip
Botero, 2003). For instance, Van Dyne et al. (2003) discussed a
M. Podsakoff, Department of Marketing, Warrington College of Business
type of voice that supports the status quo, and Gorden (1988)
Administration, University of Florida.
Correspondence concerning this article should be addressed to Timothy argued that some voice behaviors are destructive in nature. Unfor-
D. Maynes, Department of Organization and Human Resources, Jacobs tunately, the narrow focus of past research may have precluded
Management Center, University at Buffalo, The State University of New investigations into other types of voice. As evidence of this fact,
York, Buffalo, NY 14260-4000. E-mail: tdmaynes@buffalo.edu although more than a dozen voice behaviors have been discussed
87
88 MAYNES AND PODSAKOFF

in the literature (Gorden, 1988; Liang, Farh, & Farh, 2012; Mor- ing the construct validity of our voice measures using the
rison, 2011; Van Dyne et al., 2003; Van Dyne & LePine, 1998), a nomological network can be augmented with a more direct
validated survey measure exists for only one of them (Van Dyne & technique based on video validation procedures. Finally, we
LePine, 1998). For this reason, scholars have called for the devel- examine some of the personality predictors of voice, as well as
opment of measures that capture other types of voice behavior the mediating mechanisms connecting voice to important out-
(Morrison, 2011; Van Dyne et al., 2003). comes for voicing employees.
Second, many of the other types of voice discussed in the
literature have conceptual problems. This fact has created confu- Expanding the Conceptualization of Voice Behavior
sion regarding the nature of voice. For instance, Gorden (1988)
discussed a “passive” form of voice (characterized by such behav- We followed three steps in expanding the conceptualization of
iors as nonverbal support and calculative silence), which is clearly voice behavior. First, we examined the various definitions of voice
inconsistent with the current view that voice is an active, verbal in the literature in order to determine the boundaries of the domain.
behavior. We address these conceptual issues by developing an On the basis of this review, we developed an overarching defini-
tion of voice that clarified what voice is and what it is not. Next,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

overarching definition of voice behavior that should clarify what


we constructed a framework based on the foundational elements of
This document is copyrighted by the American Psychological Association or one of its allied publishers.

voice is and what it is not.


Third, as the research reported in this paper demonstrates, (a) voice and the overarching definition produced in the first step that
other types of voice behavior actually occur in organizations; (b) identifies four types of voice as well as the attributes that distin-
these behaviors are distinct from each other; and (c) these behav- guish these behaviors from each other. Finally, we labeled and
iors have important consequences for the employees who enact defined the specific forms of voice.
them and for organizations as a whole. Thus, expanding the
domain of voice is warranted from a practical as well as a theo- Employee Voice Behavior
retical perspective. In addition to Hirschman’s (1970) seminal work, the research
Finally, scholars have recently suggested that the existing mea- of other scholars (Gorden, 1988; Liang et al., 2012; Van Dyne
sure of voice may need to be refined (Detert & Burris, 2007; Grant, et al., 2003; Van Dyne & LePine, 1998) has informed the field’s
Parker, & Collins, 2009). Van Dyne and LePine (1998) anticipated conception of employee voice behavior. The conceptual defi-
this need in the original publication of their measure, observing nitions, key attributes, and limitations of several previous con-
that “another task for future research is refinement of the . . . voice ceptualizations of voice behavior are summarized in Table 1.
scales” (p. 118). The need to refine a construct’s measure is not An examination of this table suggests that there are several core
uncommon. Schwab (1980) noted that differences between the attributes of voice behavior. On the basis of these attributes, we
conceptual meaning of a construct and its operational measure- define voice behavior as an individual’s voluntary and open
ment result from a natural process in which changes are made to communication directed toward individuals within the organi-
the construct’s conceptualization or to its measures, thus resulting zation that is focused on influencing the context of the work
in either contamination or deficiency. Schwab recommended pe- environment (Gorden, 1988; Hirschman, 1970; Morrison, 2011;
riodically reevaluating the correspondence between theoretical Van Dyne et al., 2003; Van Dyne, Cummings, & Parks, 1995;
constructs and their measures as a remedy for this issue. Van Dyne & LePine, 1998). Furthermore, voice behaviors share
Thus, our primary objective in this research is to develop a several important attributes: (a) they are exhibited by individual
new, more expansive voice behavior framework; one that in- employees; (b) they are not silent (Van Dyne et al., 2003),
cludes behaviors that either challenge or support the status quo anonymous (Hirschman, 1970), or neutral (Hirschman, 1970);
as well as behaviors that possess either positive or negative (c) they conspicuously stake out an employee’s position relative
attributes. After constructing this framework, we define each of to the status quo; and (d) because others in the organization may
the behaviors identified within it, develop valid measures for disagree with the voicing employee’s position, voicing may
these behaviors, and examine some of their antecedents and damage interpersonal relationships at work (LePine & Van
consequences. We conclude with a discussion regarding direc- Dyne, 1998; Van Dyne & LePine, 1998).
tions for future research. Although voice is often verbally expressed, it is not limited to
This paper makes several important contributions to the verbal behavior (Hirschman, 1970); it includes such actions as
literature. First, the conceptual framework and definitions of the sending e-mails and writing memos (Withey & Cooper, 1989). On
different forms of voice provided in this paper should expand the other hand, not all expressive behavior is voice (Van Dyne et
the voice behavior domain to include important behaviors (be- al., 2003). To be considered voice, the expression must be (a)
haviors that are of both practical and theoretical value) that openly communicated, (b) organizationally relevant, (c) focused
have been relatively ignored up to this point. Second, this work on influencing the work environment, and (d) received by someone
should help to clarify the conceptual nature of voice behavior, inside the organization. For this reason, providing improvement-
which will provide a clear framework for voice researchers as oriented suggestions to a manager is an example of voice, whereas
they move forward in the future. Third, the development and notifying regulatory agencies of misconduct in the organization or
validation of scales for several different types of voice should placing anonymous notes in suggestion boxes is not.
provide the necessary tools with which to test our theories of
voice behavior and thereby facilitate research on the types of
Organizing Framework for Voice Behaviors
voice that have yet to be examined empirically. Fourth, follow-
ing the recent recommendations of Podsakoff, Podsakoff, A further examination of Table 1 indicates that despite the
MacKenzie, and Klinger (2013), we demonstrate how establish- progress that has been made in the voice literature, there are still
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 89

Table 1
Summary of Prior Voice Behavior Conceptualizations

Study Conceptualization of voice behavior Key attributes Limitations

Hirschman (1970) Voice behavior Broad definition leaves room for many
“Voice is . . . [an attempt] to change, Challenges the status quo forms of voice.
rather than to escape from, an No validated measures.
objectionable state of affairs.” (p. 30)
“Voice is . . . [a] ‘messy’ concept Conspicuous/visible
because it can be graduated, all the Direct
way from faint grumbling to violent
protest; it implies articulation of one’s
critical opinions rather than a private,
‘secret’ vote in the anonymity of a
supermarket; and finally, it is direct
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

and straightforward rather than


This document is copyrighted by the American Psychological Association or one of its allied publishers.

roundabout.” (p. 16)


“Voice is like exit in that it can be Critical
overdone: The discontented . . . Prohibitive or promotive
members could become so harassing focus
that their protests would . . . hinder
rather than help.” (p. 31)
Gorden (1988) Active/constructive voice
Expressions that actively reaffirm or Promotes improvement No validated measures of these
correct the activities of the behaviors.
organization. Definitions confound norms and
Behaviors include making suggestions, Preserves or challenges the behavior.
bolsterism (supporting the status quo), status quo Preserving/challenging the status quo
and principled dissent. (p. 285) confounded in one type of voice.
Active/destructive voice Passive dimension is inconsistent with
Expressions that are actively Oriented toward doing Hirschman’s (1970) notion of voice
antagonistic toward the organization harm being “active.”
and its activities. Several behaviors identified as voice
Behaviors include verbal aggression, Challenges the status quo do not involve the sharing of
bad-mouthing, and antagonistic exit. information/opinions with others.
(p. 285)
Passive/constructive voice
Behaviors indicating acceptance of and Preserves the status quo
submission to organizational
activities.
Behaviors include nonverbal support,
compliance, and cooperation. (p. 258)
Passive/destructive voice
Behaviors that indicate resigned Preserves the status quo
conformance to the activities of the
organization.
Behaviors include apathy, calculated
silence, and withdrawal. (p. 285)
Van Dyne & LePine Voice behavior
(1998) “promotive behavior that emphasizes the Promotes improvement Scholars have observed that the
expression of constructive challenge Challenges the status quo scale is contaminated (Detert &
intended to improve rather than Positive intentions Burris, 2007; Grant et al., 2009).
merely criticize” (p. 109)
Van Dyne et al. (2003) Prosocial voice
“non-required behavior that emphasizes Active, discretionary No validated measures of these
the expression of change-oriented behavior behaviors.
comments.” (p. 1370) Promotes improvement Definitions confound motive and
The expression of “work related ideas, Challenges the status quo behavior.
information, or opinions based on Altruistic motive
cooperative motives. Thus, this . . .
behavior is intentional, proactive, and
other-oriented.” (p. 1371)
Behaviors include suggestions for
change and recommendations for
improvements.
(table continues)
90 MAYNES AND PODSAKOFF

Table 1 (continued)

Study Conceptualization of voice behavior Key attributes Limitations

Defensive voice
The expression of “work-related ideas, Self-protective
information or opinions—based on
fear—with the goal of protecting the
self.” (p. 1372)
Behaviors include expressing ideas that Motivated by fear
shift attention elsewhere based on
fear and proposing ideas that focus on
others to protect the self.
Acquiescent voice
The “expression of work-related ideas, Preserves/promotes the
information, or opinions—based on status quo
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

feelings of resignation. Acquiescent


This document is copyrighted by the American Psychological Association or one of its allied publishers.

voice is disengaged behavior that is


based on feeling unable to make a
difference. Thus, it results in
expressions of agreement and support
based on low self-efficacy to affect
any meaningful change.” (p. 1373)
Behaviors include expressions of Motivated by resignation
support and agreement.
Liang et al. (2012) Promotive voice
“employees’ expression of new ideas or Challenges the status quo Does not capture the negative
suggestions for improving the overall connotation associated with the term
functioning of their work unit or prohibitive.
organization. . . . Such voice is
‘promotive’ in the sense that it is
focused on a future ideal state.”
(p. 74)
Behaviors include proposing ways to Promotes change
change the status quo and offering
suggestions for improvement.
Prohibitive voice
“Prohibitive voice describes employees’ Prevents harm
expressions of concern about work
practices, incidents, or employee
behavior that are harmful to their
organization.” (p. 75)
Behaviors include expressing concern Preserves or challenges the
over current work practices that are status quo
problematic and expressions focused
on preventing problematic initiatives
from being implemented.

a number of limitations that should be addressed. First, although definitional meanings associated with the term prohibitive are
scholars tend to agree that there is more than one type of voice, quite negative.
many of the forms of voice discussed in the literature do not have Given these issues, one of our objectives was to develop a
validated measures. As a result, our empirical knowledge of the voice new voice behavior framework; one that both expands the
domain has been restricted. Second, several of the voice models re- domain and clarifies what types of behaviors should be consid-
ported in the literature have definitional issues that have made it ered to be voice. In constructing this framework, we drew on
difficult to investigate these behaviors. For instance, Gorden’s several of the voice models summarized in Table 1. As shown
(1988) definitions conflate norms with voice behaviors, and in in this table, scholars have discussed two different dimensions
some cases his definitions are overly broad and include more than upon which voice behaviors may vary. The first dimension
one type of voice. Finally, the vast majority of empirical work on distinguishes voice behaviors that preserve the status quo (i.e.,
voice has focused on behaviors with positive attributes (e.g., reaffirm, sustain, keep things the same) from those that chal-
improvement oriented, intended to benefit the organization, altru- lenge it (i.e., correct, question, confront). This dimension is
istically motivated), even though several forms of voice possessing rooted in Hirschman’s (1970) description of voice as the artic-
negative attributes (e.g., hinders rather than helps, antagonistic ulation of critical opinions; Gorden’s (1988) observation that
toward the organization) have been identified. This is especially voice behaviors can either “reaffirm” or “correct” the activities
apparent when one considers the voice behaviors that are prohib- of the organization; and Van Dyne et al.’s (2003) distinction
itive in nature. These behaviors have primarily been defined as between voice behaviors that express support versus those that
consisting of positive attributes, even though the connotations and constructively challenge.
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 91

The second dimension distinguishes promotive voice behaviors (d) challenge/ prohibitive quadrant destructive voice. In our view,
(i.e., those that are aimed at advocating, encouraging, or endors- voice is not an overarching construct made up of multiple subdi-
ing) from prohibitive voice behaviors (i.e., those that are aimed at mensions. Instead, voice is a domain consisting of multiple cate-
stopping, blocking, or hindering). This dimension is found in gories of behaviors (Rusbult et al., 1988; Van Dyne et al., 2003;
Hirschman’s (1970) view that voice behavior may be focused Withey & Cooper, 1989). Each category is a separate construct,
either on promoting change or on hindering it; Gorden’s (1988) and the behaviors within each category are reflective of the con-
distinction between voice behavior focused on promoting change struct representing that category. Thus, the forms of voice are
versus voice that is focused on tearing down the organization; and reflective, not formative, constructs (Bollen, 1989).
the prohibitive/promotive dimension from Liang et al.’s (2012) Finally, our conceptualization of voice relies primarily on the
recent work. However, unlike Liang et al. (2012), whose concep- perspective of observers of voice events rather than on the per-
tualization views prohibitive voice as a constructive behavior that spective of voicing employees, because (a) observers are likely to
prevents deterioration in work practices, we conceptualize prohib- be aware of voice events given that these events are highly visible;
itive voice as being more consistent with Gorden’s (1988) sugges- (b) the majority of recent research on voice behavior has examined
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

tion that voice messages can either be constructively or destruc- voice from this perspective (Burris, Detert, & Chiaburu, 2008;
This document is copyrighted by the American Psychological Association or one of its allied publishers.

tively focused. Although this perspective is a departure from Liang LePine & Van Dyne, 1998; Liang et al., 2012; Tangirala &
et al. (2012) and may be viewed as somewhat controversial, we Ramanujam, 2008; Van Dyne & LePine, 1998; Venkataramani &
believe that it is consistent with prior theory (Gorden, 1988; Tangirala, 2010); and (c) self-reports of voice may result in biased
Hagedoorn, Van Yperen, Van de Vliert, & Buunk, 1999; Kassing, responses, because some voice behaviors may be perceived as
1998, 2011) and with the definitional meaning of the term prohib- socially desirable/undesirable (Nederhof, 1985).
itive. It is also consistent with the fact that people don’t always
voice their opinions in a constructive manner.
Definitions for Each Type of Voice
Positioning the preservation/challenge dimension opposite the
promotive/prohibitive dimension in a two by two matrix divides Supportive voice. Supportive voice represents the preserva-
the voice domain into four quadrants (summarized in Table 2). We tion/promotion quadrant in our model. To be supportive means to
labeled the behaviors in the (a) preservation/promotion quadrant sustain (i.e., preserve) by defending or endorsing (i.e., promoting;
supportive voice; (b) challenge/promotion quadrant constructive Borman & Motowidlo, 1993). We define supportive voice as the
voice; (c) preservation/prohibitive quadrant defensive voice; and voluntary expression of support for worthwhile work-related pol-

Table 2
Organizing Framework for Employee Voice Behaviors

Dimension Preservation Challenge

Promotive Supportive voice Constructive voice


Definition: Supportive voice is the voluntary expression of Definition: Constructive voice is the voluntary
support for worthwhile work-related policies, programs, expression of ideas, information, or opinions
objectives, procedures, etc., or speaking out in defense of focused on effecting organizationally
these same things when they are being unfairly criticized. functional change to the work context.
Representative behaviors Representative behaviors
Expressing support for organizational procedures or Suggesting improvements to standard operating
objectives procedures
Verbally defending organizational policies that other Proposing ideas for new or more effective
employees are criticizing work methods
Related constructs Related constructs
Acquiescent voice (Van Dyne et al., 2003) OCB voice (Van Dyne & LePine, 1998)
Active/constructive voice (Gorden, 1988) Prosocial voice (Van Dyne et al., 2003)
Passive/constructive voice (Gorden, 1988) Promotive voice (Liang et al., 2012)
Loyalty (Graham, 1991) Prohibitive voice (Liang et al., 2012)

Prohibitive Defensive voice Destructive voice


Definition: Defensive voice is the voluntary expression of Definition: Destructive voice is the voluntary
opposition to changing an organization’s policies, procedures, expression of hurtful, critical, or debasing
programs, practices, etc., even when the proposed changes opinions regarding work policies, practices,
have merit or making changes is necessary. procedures, etc.
Representative behaviors Representative behaviors
Vocally opposing changes to work policies, even though Bad-mouthing the organization’s policies or
the changes are necessary objectives
Speaking out against changing work policies, even when Making overly critical comments about how
the changes have merit things are done in the organization
Related constructs Related constructs
Defensive voice (Van Dyne et al., 2003) Active/destructive voice (Gorden, 1988)
Resistance to change (Oreg, 2003) Poor sportsmanship (Organ, 1988)
Note. OCB ⫽ organizational citizenship behavior.
92 MAYNES AND PODSAKOFF

icies, programs, objectives, procedures, and so on, or speaking out in order to prevent them from occurring, whereas resistance to
in defense of these same things when they are being unfairly change is more passive, given its focus on avoiding changes
criticized. Representative behaviors include expressing support for that have already been enacted.
valuable work practices or defending valid organizational policies Destructive voice. The final form of voice represents the
that coworkers are criticizing. Supportive voice is similar to ac- challenge/prohibitive quadrant. These behaviors are referred to
quiescent voice (Van Dyne et al., 2003), given that they both as destructive voice. To be destructive is to be hurtful or critical
emphasize the expression of support for work-related practices. (i.e., challenge) or to bring down or end (i.e., prohibit; Gorden,
However, whereas acquiescent voice must be motivated by a 1988; Warren, 2003). We define destructive voice as the vol-
feeling that one cannot make a difference, supportive voice is not untary expression of hurtful, critical, or debasing opinions
subject to the same restriction. Supportive voice is also similar to regarding work policies, practices, procedures, and so on. Rep-
organizational loyalty (i.e., identification with the organization resentative behaviors include bad-mouthing organizational pol-
and its leaders; Graham, 1991) and loyal boosterism (i.e., promo- icies, making disparaging comments regarding work-related
tion of the organizational image to outsiders; Moorman & Blakely, programs, and harshly criticizing the organization’s work meth-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1995). However, supportive voice differs from these constructs, ods. Destructive voice is similar to active/destructive voice
This document is copyrighted by the American Psychological Association or one of its allied publishers.

given that supportive voice must be the communication of infor- (Gorden, 1988), given that they both emphasize harsh criticism
mation to organizational insiders. of the status quo. However, active/destructive voice differs,
Constructive voice. This form of voice represents the chal- given that it includes expressions that are not critical of the
lenge/promotion quadrant. To be constructive is to promote im- work context (e.g., ingratiation and duplicity). Destructive
provement (i.e., to challenge the status quo; Gorden, 1988). We voice is also similar to the OCB construct of a lack of sports-
define constructive voice as the voluntary expression of ideas, manship, which includes such behaviors as “complaining, petty
information, or opinions focused on effecting organizationally grievances, [and] railing against real or imagined slights”
functional change to the work context. Constructive voice expres- (Organ, 1988, p. 11). Despite the similarities, destructive voice
sions may include suggestions of (a) new or improved ways of differs from poor sportsmanship in three important ways: (a)
doing things; (b) ways to fix problems with existing work methods, poor sportsmanship is not limited to the expression of informa-
procedures, and so on; or (c) solutions to problems that have been tion; (b) poor sportsmanship behaviors include expressions that
previously identified. Constructive voice is similar to active/con- are not limited to work practices, policies, or methods; and (c)
structive voice (Gorden, 1988), organizational citizenship behav- poor sportsmanship behaviors do not include disparaging ex-
ior (OCB) voice (the form of voice identified by Van Dyne & pressions.
LePine, 1998), and prosocial voice (Van Dyne et al., 2003), given
that each of these emphasizes improvements to the working envi-
Instrument Development
ronment. However, constructive voice also differs from these
constructs. OCB and active/constructive voice are broader, given After constructing our conceptual framework, we developed
that they include nonexpressive behaviors such as keeping in- and validated survey measures for each voice behavior in our
formed about work-related issues and participative decision mak- model. We did so in five stages based on the recommendations
ing. In contrast, prosocial voice is narrower, given that it must be of several organizational scholars (Churchill, 1979; MacKenzie,
motivated by a desire to benefit others. Podsakoff & Podsakoff, 2011; Nunnally & Bernstein, 1994). In
Defensive voice. Behaviors that reflect the preservation/ the first stage, survey items were generated from the theoretical
prohibitive quadrant are called defensive voice. In this case, to definitions, a review of the literature, and qualitative interviews
be defensive is to shield (i.e., prohibit) the status quo from with working professionals. Following this, the content validity
potential changes (i.e., preserve; Ashforth & Lee, 1990). We of the items was assessed. Next, we conducted a preliminary
define defensive voice as the voluntary expression of opposition test of the convergent, discriminant, and nomological validity of the
to changing an organization’s policies, procedures, programs, scales. In Stage 4, the causal connection between the voice
and practices, even when the proposed changes have merit or behaviors and ratings on the voice scales was assessed (i.e., the
making changes is necessary. Representative behaviors include scale’s veridical validity). In the final stage, the criterion-
stubbornly arguing against changing work methods or vocally related validity of the voice scales was evaluated.
opposing changes to standard operating procedures. Van Dyne
et al.’s (2003) three-factor model also contained a behavior that
Stage 1: Item Generation and Refinement
they labeled defensive voice. Although their conception of this
construct is similar to ours (i.e., both are protective actions The objective of the first stage was to generate a sufficient
intended to reduce a perceived threat), these constructs differ in number of items to adequately tap the conceptual domain of each
two important ways. First, in our model defensive voice is form of voice (Kerlinger, 1973). Items were generated in two
focused on shielding the status quo, whereas in their model the ways. First, 44 items were generated based on a review of previous
behavior is aimed at protecting the voicing employee. Second, research and theory regarding behaviors that are similar to voice
their conception of defensive voice is broader, given that it (e.g., taking charge, issue selling, loyalty, defensive behavior).
includes expressions of support for the organization as well as Second, 38 unique behaviors were identified via interviews with
protective expressions. Defensive voice is also similar to Oreg’s 39 associates working in professional service firms (e.g., manage-
(2003) resistance to change construct, given that both empha- ment consulting, advisory services, financial services). On aver-
size preserving the status quo. However, defensive voice is age, the interviewees were about 35 years old (M ⫽ 34.79, SD ⫽
more active in that defensive voicers speak out against changes 6.28). They were primarily male (87.2%) and Caucasian (94.8%)
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 93

and had just under 10 years of work experience (M ⫽ 9.55, SD ⫽ wording of the items. Evidence of the content validity of each item
5.90). Respondents were given the conceptual definition and ex- relative to the overall dimensions of voice (i.e., preservation/
amples of one of the four forms of voice, and they were then asked challenge and prohibitive/promotive) was then obtained by asking
to describe an incident of someone engaging in the behavior. study participants to rate each survey item twice; once regarding
Between the two procedures, a total of 82 items were generated. the extent to which the items describe behaviors that tend to be
After we eliminated redundant or overlapping items, 68 unique prohibitive (blocks, hinders, hampers, abolishes, or stamps out) or
items remained. promotive (advances, advocates, encourages, or endorses), and
once regarding the extent to which the behaviors tend to preserve
Stage 2: Assessment of Item Content Validity (sustain, support, or keep things the same) or challenge the status
quo (criticize, change, or tear down). These tasks were performed
Content validity evidence was obtained in two separate studies. sequentially and were counterbalanced such that half of the par-
In the first, the extent to which each survey item is consistent with ticipants rated the preservation/challenge dimension first, followed
the definitions of the overall dimensions of voice (i.e., preserva- by the prohibitive/promotive dimension, and the other half per-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

tion/ challenge and promotive/prohibitive) was assessed. Although formed these tasks in reverse order. Items were rated on a 7-point
This document is copyrighted by the American Psychological Association or one of its allied publishers.

an assessment of this type differs from traditional content validity scale ranging either from ⫺3 (Preserve) to 3 (Challenge) or from
assessments, it is valuable because it provides evidence regarding ⫺3 (Prohibitive) to 3 (Promotive).
the validity of our overarching framework of the voice domain. In Results and Discussion. Aggregated results from this assess-
the second study, in which we used a more traditional approach to ment (reported in the first four columns of Table 3) indicate that
content validity (cf. Haynes, Richard, & Kubany, 1995; Nete- the supportive voice items are preservation oriented (M ⫽ ⫺2.77,
meyer, Bearden & Sharma, 2003), the extent to which the survey SD ⫽ 0.70) and promotive in nature (M ⫽ 2.56, SD ⫽ 0.62); the
items are consistent with the definitions of the four types of voice constructive voice items are challenge oriented (M ⫽ 2.70, SD ⫽
behavior was assessed. 0.76) and promotive in nature (M ⫽ 2.76, SD ⫽ 0.47); the
defensive voice items are preservation oriented (M ⫽ ⫺2.90,
Study 1: Content Validity Relative to the Overall SD ⫽ 1.45) and prohibitive in nature (M ⫽ ⫺2.66, SD ⫽ 0.63);
Dimensions of Voice and the destructive voice items are challenge oriented (M ⫽ 2.65,
SD ⫽ 1.00) and prohibitive in nature (M ⫽ ⫺2.72, SD ⫽ 0.51).
Participants. Thirty undergraduate students majoring in busi-
These findings are consistent with our conceptual framework of
ness from a large university in the United States participated in this
voice behavior.
study. On average, participants were 20 years of age (M ⫽ 20,
SD ⫽ .74). They were primarily male (53.3%) and Caucasian Study 2: Content Validity Relative to
(83.3%) and had just over 4 months of full-time work experience
the Definitions of Voice
(M ⫽ 4.63, SD ⫽ 9.23). Students completed this exercise as part
of a requirement to earn course credit. Although some scholars Participants. Forty-seven undergraduate students majoring in
have cautioned against the use of undergraduate students for business at a large university in the United States participated in
research purposes (Anderson & Gerbing, 1991), it is appropriate in this study. On average, participants were 20 years of age (M ⫽ 20,
this case because participants were asked to evaluate survey items SD ⫽ 0.55). They were primarily male (74.4%) and Caucasian
based on definitions (which is similar to the activities of most (91.5%) and had just over 4 months of full-time work experience
undergraduate courses), not to assess whether the items tapped (M ⫽ 4.38, SD ⫽ 8.95). Students completed this exercise as part
actual behaviors in organizations. of a requirement to earn course credit.
Procedure. We first refined the list of survey items from 68 to Procedure. Evidence of item content validity relative to the
48 (supportive voice ⫽ 12, constructive voice ⫽ 13, defensive definitions of voice was obtained by asking study participants to
voice ⫽ 11, and destructive voice ⫽ 12) based on two preliminary rate the extent to which each survey item is captured by the
studies in which we sought input from participants regarding the conceptual definition of each form of voice (Hinkin & Tracey,

Table 3
Mean Ratings of Survey Items Obtained From Content Validity Assessment (Stage 2)

Study 1 Study 2
Preserve (⫺)/ Prohibit (⫺)/
Challenge (⫹) Promote (⫹) Mean ratings Standard deviation of ratings
Type M SD M SD SV CV DFV DSV SV CV DFV DSV

Supportive ⫺2.77 0.70 2.56 0.62 6.81 1.15 1.08 1.05 0.84 0.66 0.48 0.31
Constructive 2.70 0.76 2.76 0.47 1.13 6.88 1.00 1.02 0.48 0.54 0.02 0.16
Defensive ⫺2.90 1.45 ⫺2.66 0.63 1.36 1.13 6.37 1.68 1.20 0.60 1.70 1.71
Destructive 2.65 1.00 ⫺2.72 0.51 1.03 1.01 1.34 6.72 0.13 0.06 1.06 0.92
Note. In Study 1, items were rated on a 7-point scale from –3 (Preserve) to 3 (Challenge) for the preservation versus challenge dimension, and from –3
(Prohibit) to 3 (Promote) for the prohibitive versus promotive dimension. In Study 2, items were rated on a 7-point scale from 1 (not at all) to 7 (completely)
for SV, CV, DFV, and DSV. Italicized means indicate average ratings that were significantly higher than any other mean in the same row. SD ⫽ standard
deviation; SV ⫽ supportive voice; CV ⫽ constructive voice; DFV ⫽ defensive voice; DSV ⫽ destructive voice.
94 MAYNES AND PODSAKOFF

1999; MacKenzie et al., 2011). Thus, four ratings were obtained behavior focused on personality predictors (LePine & Van Dyne,
for each item (1 item ⫻ 4 behaviors).1 Response choices ranged 2001); and (b) prior research indicates that there is a connection
from 1 (the survey item is not captured by the conceptual definition between personality and discretionary behaviors such as voice
at all) to 7 (the survey item is completely captured by the concep- (Crant, Kim, & Wang, 2011; Ilies, Fulmer, Spitzmuller, & John-
tual definition). Content validity was then assessed with a one-way son, 2009; Nikolaou, Vakola, & Bourantas, 2008; Organ & Ryan,
repeated measures analysis of variance (ANOVA) procedure, 1995). We conducted the analysis with structural equation model-
wherein mean ratings on the intended form of voice were com- ing procedures. Because the analysis is preliminary, we do not
pared to mean ratings on the other forms of voice. This was develop formal hypotheses. However, we do expect to see certain
followed up with a series of planned comparisons, which assessed patterns of relationships emerge (in some cases differentially)
whether an item’s mean score was significantly higher on the across the forms of voice. These expectations are described below.
proposed form of voice compared to the other forms. Given that agreeable people tend to be empathetic, cooperative,
Results and Discussion. Summary results, reported in Table and friendly and that they place a high value on social harmony in
3, indicate that the supportive voice (SV) items were captured by
the workplace (Barrick & Mount, 1991; LePine & Van Dyne,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the definition of supportive voice but not the other types of voice
2001), they are unlikely to defend the organization when others are
This document is copyrighted by the American Psychological Association or one of its allied publishers.

(mean ratings for SV ⫽ 6.81, CV ⫽ 1.15, DFV ⫽ 1.08, DSV ⫽


being critical (supportive voice), challenge the status quo by mak-
1.05); the constructive voice (CV) items were captured by the
ing recommendations for change (constructive voice), speak out
definition of constructive voice but not the other types (mean
against changes that are being made (defensive voice), or speak out
ratings for SV ⫽ 1.13, CV ⫽ 6.88, DFV ⫽ 1.00, DSV ⫽ 1.02); the
defensive voice (DFV) items were captured by the definition of against the status quo (destructive voice) because each of these
defensive voice but not the other types (mean ratings for SV ⫽ behaviors may create social discord by offending those who are
1.36, CV ⫽ 1.13, DFV ⫽ 6.37, DSV ⫽ 1.68); and the destructive opposed to the views of the voicer. Thus, agreeableness should be
voice (DSV) items were captured by the definition of destructive negatively related to each type of voice in our model.
voice but not the other types (mean ratings for SV ⫽ 1.03, CV ⫽ Conscientiousness is associated with the traits of being depend-
1.01, DFV ⫽ 1.34, DSV ⫽ 6.72). On an individual level, each item able, responsible, planful, and achievement oriented (Barrick &
received its highest rating on the intended form of voice, and this Mount, 1991). Thus, conscientious employees are likely to speak
rating was always significantly higher than the other ratings (all p up in support of the organization when others are being critical
values ⬍ .05). Taken together, this analysis suggests that the items because they may view it as the responsible thing to do. Further-
possess adequate content validity. more, they would likely speak up with recommendations for im-
provement, if doing so would help them achieve the results they
Stage 3: Preliminary Assessment of Convergent, desired. Thus, conscientiousness should be positively related to the
promotive forms of voice in our model (i.e., supportive and con-
Discriminant and Nomological Validity
structive voice) but not related to the prohibitive forms of voice
The objective in this stage was to conduct a preliminary assessment (i.e., defensive and destructive).
of the convergent, discriminant, and nomological validity of the voice Emotional stability is the trait of being calm and even tempered
scales. This assessment was conducted in two separate steps. In the (Barrick & Mount, 1991). In contrast, emotionally unstable people
first step, we examined the convergent and discriminant validity of the tend to be anxious, angry, worried, and insecure (Barrick &
voice measures at the item- and construct-level using confirmatory Mount, 1991). Empirical evidence indicates that emotionally un-
factor analysis procedures in LISREL 8.8. Two confirmatory factor stable people are uncooperative (LePine & Van Dyne, 2001); tend
analyses (CFA) were examined in this stage. In the first, a four-factor to have low quality interactions with coworkers (LePine & Van
model that included only the items measuring the types of voice was Dyne, 2001); often express negative attitudes about aspects of the
analyzed. This was done in order to assess the convergent/discrimi- work environment (Judge & Bono, 2001); and respond poorly to
nant validity among the voice scales at the item and construct level. In environmental stressors such as changes to work processes or
the second, an eight-factor model containing the measures of voice
procedures (Judge, Thoresen, Pucik, & Welbourne, 1999). Given
and the measures of a comparison construct for each type of voice was
that defensive voice is the expression of a negative attitude about
examined in order to assess the convergent/discriminant validity of
changing work conditions and that destructive voice is the expres-
the voice scales relative to the measures of the comparison constructs.
sion of a negative attitude about how things are currently being
Loyalty was selected as the comparison construct for supportive voice,
done in the organization, emotionally unstable people are likely to
because these two behaviors share an emphasis on supporting elements of
the organization and should therefore covary. OCB voice served as the engage in these types of voice. In contrast, given that supportive
comparison for constructive voice, given that these two forms of voice voice is the expression of a positive attitude about current work
focus on promoting constructive changes. Resistance to change served as practices and that a potential outcome of constructive voice is a
the comparison for defensive voice, because both of these behaviors focus
on preventing change. Finally, poor sportsmanship was selected as the 1
Ratings of item-level content validity relative to the definitions of each
comparison for destructive voice, given that both of these behaviors focus form of voice were obtained by presenting participants with a matrix,
almost exclusively on negative aspects of the organization. wherein the rows of the first column contained the survey items, and the
In the second step, we conducted a preliminary assessment of definitions of supportive, constructive, defensive, and destructive voice
were listed at the tops of columns 2 through 5. A final column titled
the nomological validity of the voice scales by examining dispo- “Other” was provided to capture items that were not consistent with any of
sitional antecedents of voice. Predictors were selected from the Big the forms of voice. A 7-point rating scale was provided at the intersection
Five personality framework because (a) early research on voice of each row and column.
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 95

change to current work practices, emotional instability is unlikely Measures. One goal in this study was to create short-form mea-
to be related to these types of voice. sures by reducing the length of our scales. Thus, although all 48 voice
Extraversion is associated with the traits of sociability/dominance, items from our previous content validity assessment were included in
gregariousness, and talkativeness (Barrick & Mount, 1991). Extra- this study, only five items for each voice scale were retained in the
verts tend to be expressive and assertive in their interpersonal inter- final models. Consistent with the recommendations of MacKenzie et
actions. Thus, they are likely to speak up and express their opinions, al. (2011), scales were trimmed by first conducting a CFA and
especially if doing so helps them achieve their goals. However, looking for items with low/nonsignificant factor loadings, strong/
extraverts also tend to be socially adept; they are unlikely to express significant error covariances, or strong/significant cross-loadings. All
negative views or opinions at work because doing so may arouse of the voice items passed these tests. Thus, the final five items for each
negative feelings in others, creating resentment. Furthermore, prior scale were selected based on (a) the content validity ratings; (b) the
research indicates that extraverts tend to be happy (Pavot, Diener, & ratings on the overall dimensions of voice (e.g., prohibitive/promo-
Fujita, 1990) and to experience high levels of positive emotion (Die- tive); and (c) an assessment regarding whether the items captured,
ner, Suh, Lucas, & Smith, 1999). Thus, extraverts are unlikely to hold collectively, the entire domain of the construct.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

negative views of the work environment. As a consequence, when Reliabilities for these measures were .89 for supportive voice; .95
This document is copyrighted by the American Psychological Association or one of its allied publishers.

extraverts voice their opinion it is likely to be in a positive manner. for constructive voice; .92 for defensive voice; and .93 for destructive
Because promotive voice behaviors are the expression of positive voice. Loyalty (␣ ⫽ .82) was measured with five items from Van
attitudes about opportunities for improvement or about current orga- Dyne, Graham, and Dienesch’s (1994) scale; OCB voice (␣ ⫽ .89)
nizational procedures, extraversion should be positively related to was measured with six items from Van Dyne and LePine’s (1998)
these types of voice (i.e., supportive/constructive). Consistent with scale; resistance to change (␣ ⫽ .85) was measured with five items
this line of reasoning, extraversion is not expected to be related to from Oreg’s (2003) scale; and poor sportsmanship (␣ ⫽ .90) was
either defensive or destructive voice behavior. measured with four items from Podsakoff and MacKenzie’s (1994)
Openness to experience is characterized by active imagination, scale. Traits from the Big Five personality framework were assessed
intellectual curiosity, and broad-mindedness (Barrick & Mount, with three items from each of Saucier’s (1994) trait adjective mea-
1991). People who score high on this dimension are willing to sures. Cooperative, rude (reverse coded), and unsympathetic (reverse
consider diverse perspectives; are likely to seek opportunities to learn coded) were used to assess agreeableness (␣ ⫽ .89); efficient, orga-
new things and to have new experiences; and are very tolerant of nized, and practical were used to assess conscientiousness (␣ ⫽ .88);
diversity (McCrae, 1996). In contrast, those who are closed to expe- moody, temperamental, and touchy (all reverse coded) were used
rience prefer stable routines over new experiences and have a very to assess emotional stability (␣ ⫽ .90); bashful (reverse coded), bold,
narrow range of interests (Whitbourne, 1986). Given that constructive and energetic were used to assess extraversion (␣ ⫽ .82); and cre-
voice initiates change and defensive voice diminishes change, open- ative, imaginative, and unintellectual (reverse coded) were used to
ness to experience should be positively related to constructive voice assess openness to experience (␣ ⫽ .79).
but negatively related to defensive voice.
Results and Discussion
Method Assessment of convergent and discriminant validity. The
convergent/discriminant validity of the voice scales was evaluated
Sample. The questionnaire was administered to a sample of with two CFAs constructed in LISREL 8.8. The first was a
executive MBA students taking classes at a large university in the four-factor model. Only the items measuring the types of voice
United States (210 completed surveys were returned for a response were included. Results, reported in Table 4, indicate high levels of
rate of 81%). On average, respondents were 39 years of age (M ⫽ convergent validity. Each factor loading is strong and significant
38.81 years). They were primarily male (90%) and Caucasian (83.3%) (all p values ⬍ .01), and the average variance extracted (AVE)
and had approximately 8 years of full-time work experience (M ⫽ value for each type of voice is well above Fornell and Larcker’s
7.92). Furthermore, 51% had been required to evaluate the perfor- (1981) suggested cutoff of .50. Results also indicate high levels of
mance of subordinates as part of their job; they had this responsibility discriminant validity. As indicated in the first column of Table 5,
for just over three years on average. Participants were paid $20. the hypothesized four-factor model fit the data significantly better
Procedure. Respondents were instructed to think about a co- (all p values ⬍ .01) than a measurement model in which all of the
worker that they work with frequently in their job. They then rated supportive, constructive, defensive, and destructive voice items
this coworker on the items measuring the types of voice, loyalty, OCB loaded on a single factor (⌬␹2 ⫽ 818.45, df ⫽ 6); a two-factor model
voice, resistance to change, poor sportsmanship, and variables from in which the promotive types of voice loaded on one factor and the
the Big Five personality framework on a 7-point scale (1 ⫽ strongly prohibitive types loaded on another (⌬␹2 ⫽ 980.71, df ⫽ 5); and a
disagree, 7 ⫽ strongly agree). Although personality is often assessed two-factor model in which the preservation-oriented types of voice
with self-reports, we relied on observer ratings because it has been loaded on one factor and the challenge-oriented types loaded on
argued (Goldberg, 1993) that they capture meaningful aspects of another (⌬␹2 ⫽ 1,432.28, df ⫽ 5). Finally, goodness-of-fit indices for
personality not captured in a self-report. Meta-analytic evidence sup- the four-factor model met or exceeded Hu and Bentler’s (1999)
ports this view in that other-reports of personality explain significant suggested cutoff of .95 for the comparative fit index (CFI ⫽ .96), .09
incremental variance over self-reports in important outcomes (Oh, for the standardized root mean residual (SRMR ⫽ .06), and .95 for the
Wang, & Mount, 2011). Thus, we chose to rely on observer ratings of incremental fit index (IFI ⫽ .96).
personality. Respondents also answered demographic questions about The second CFA was constructed as an eight-factor model
themselves and the focal coworker. wherein the measures of voice and the selected comparison con-
96 MAYNES AND PODSAKOFF

Table 4
Summary of Confirmatory Factor Analysis Results (Stage 3)

Constructs/items Factor loadings

Supportive voice (average variance extracted) (.62)


This employee
1. Defends organizational programs that are worthwhile when others unfairly criticize the programs. .79
2. Expresses support for productive work procedures when others express uncalled for criticisms of the procedures. .80
3. Speaks up in support of organizational policies that have merit when others raise unjustified concerns about the policies. .78
4. Defends useful organizational policies when other employees unfairly criticize the policies. .84
5. Defends effective work methods when others express invalid criticisms of the methods. .72
Constructive voice (average variance extracted) (.77)
This employee
1. Frequently makes suggestions about how to do things in new or more effective ways at work. .88
2. Often suggests changes to work projects in order to make them better. .88
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

3. Often speaks up with recommendations about how to fix work-related problems. .89
This document is copyrighted by the American Psychological Association or one of its allied publishers.

4. Frequently makes suggestions about how to improve work methods or practices. .87
5. Regularly proposes ideas for new or more effective work methods. .87
Defensive voice (average variance extracted) (.71)
This employee
1. Stubbornly argues against changing work methods, even when the proposed changes have merit. .83
2. Speaks out against changing work policies, even when making changes would be for the best. .79
3. Vocally opposes changing how things are done, even when changing is inevitable. .84
4. Rigidly argues against changing work procedures, even when implementing the changes makes sense. .91
5. Vocally argues against changing work practices, even when making the changes is necessary. .84
Destructive voice (average variance extracted) (.72)
This employee
1. Often bad-mouths the organization’s policies or objectives. .82
2. Often makes insulting comments about work-related programs or initiatives. .83
3. Frequently makes overly critical comments regarding how things are done in the organization. .91
4. Often makes overly critical comments about the organization’s work practices or methods. .92
5. Harshly criticizes the organization’s policies, even though the criticism is unfounded. .75
Note. All factor loadings are completely standardized lambda loadings and are significant at p ⬍ .01. The chi-square test had 164 degrees of freedom.
␹2 ⫽ 392.19 (p ⬍ .01); comparative fit index ⫽ .96; standardized root mean residual ⫽ .06; incremental fit index ⫽ .96.

structs were modeled as separate factors. The means, standard p ⬎ .05), constructive (␥ ⫽ ⫺.07, p ⬎ .05), or destructive voice
deviations, reliabilities, AVE values, and correlations are reported (␥ ⫽ ⫺.15, p ⬎ .05). Thus, people perceived as more agreeable
in Appendix A. As indicated in the table in this appendix, the types were also perceived to express lower levels of defensive voice than
of voice are significantly correlated with their comparison con- people who were perceived to be less agreeable. Contrary to our
structs (i.e., correlations ranged from .46 to .76), which provides expectations, conscientiousness was not significantly related to
evidence of convergent validity. As indicated in the second column either supportive (␥ ⫽ .09, p ⬎ .05) or constructive voice (␥ ⫽ .08,
of Table 5, the measures also demonstrate a sufficient level of p ⬎ .05). However, as expected, emotional stability was negatively
discriminant validity, given that the hypothesized eight-factor related to defensive (␥ ⫽ ⫺.27, p ⬍ .01) and destructive voice
model fit the data significantly better than a single-factor model or
a four-factor model in which the items from each type of voice and
its comparison construct loaded on the same factor. Finally, the 2
In order to provide a more robust test of the discriminant validity of the
CFI (.97), SRMR (.07), and IFI (.97) values for the eight-factor new voice measures from their comparison constructs, we analyzed a series of
model all met or exceeded standard cutoff levels (Hu & Bentler, two-factor CFA models. In each of these, the focal type of voice and its
1999). These results provide substantial support for the conver- comparison construct were included in the model. Results of these analyses
gent/discriminant validity of the items and overall measures of indicated that there was a substantial amount of support for the distinctiveness
of the voice measures in that (a) all of the goodness-of-fit indices for the
voice.2 hypothesized two-factor solutions met or exceeded the cutoffs suggested by
Assessment of nomological validity. The second objective in Hu and Bentler (1999); (b) all of the two-factor CFA models fit the data better
this study was to conduct a preliminary assessment of the nomological than one-factor CFA models, which collapsed the measures of the focal voice
validity of the voice scales using the personality variables as predic- construct with the measures of the comparison construct (e.g., all chi-square
tors of the voice behaviors. Relationships were examined with struc- differences between the hypothesized two-factor CFA models and the one-
factor models were greater than the critical value of 3.84, and all p values
tural equation modeling procedures in LISREL 8.8. A model with were ⬍ .01); (c) all of the items loaded significantly on their intended factors;
nine latent variables was specified (covariances among the personality (d) consistent with Fornell and Larcker’s (1981) test of convergent validity, all
variables and among the measures of voice were estimated). The five of the AVEs met or exceeded the suggested level of .50; and (e) consistent with
personality variables served as predictors, and the four types of voice Fornell and Larcker’s (1981) test for discriminant validity, the average vari-
ance extracted from the items by the hypothesized constructs was greater than
served as outcomes. Results are summarized in Table 6. the squared correlations between these constructs in every case. Thus, these
As indicated in the table, agreeableness is negatively related to tests provide strong evidence in support of the discriminant validity of the new
defensive voice (␥ ⫽ ⫺.30, p ⬍ .05) but not supportive (␥ ⫽ ⫺.23, voice measures.
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 97

Table 5
Summary of Model Fit Information for All CFAs Examined

Stage 3 Stage 4 Stage 5


Model 4-factor 8-factor SV CV DFV DSV 13-factor

Sample size 210 210 76 76 75 74 222


Hypothesized model
␹2 (df) 392.19 (164) 1,354.03 (712) 431.29 (265) 478.49 (289) 365.84 (265) 340.12 (242) 1,357.57 (849)
CFI .96 .97 .94 .93 .94 .97 .98
SRMR .06 .07 .08 .10 .08 .06 .07
IFI .96 .97 .94 .93 .94 .97 .98
4-factor model
␹2 (df) 2,378.33 (734) 712.06 (269) 537.48 (293) 443.35 (269) 415.01 (246)
CFI .92 .83 .91 .89 .94
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

SRMR .11 .15 .11 .11 .10


This document is copyrighted by the American Psychological Association or one of its allied publishers.

IFI .92 .83 .91 .89 .94


3-factor model
␹2 (df) 4,446.79 (924)
CFI .87
SRMR .12
IFI .87
2-factor model Promotive/prohibitive Mediators/Outcomes
␹2 (df) 1,372.90 (169) 4,734.08 (926)
CFI .81 .86
SRMR .16 .11
IFI .81 .86
2-factor model Preservation/challenge Predictors/Mediators
␹2 (df) 1,824.47 (169) 4,651.41 (926)
CFI .73 .86
SRMR .22 .11
IFI .73 .86
1-factor model
␹2 (df) 1,210.64 (170) 4,291.76 (740) 1,276.34 (275) 1,065.40 (299) 1,152.79 (275) 1,180.59 (252) 5,052.04 (927)
CFI .67 .83 .61 .71 .45 .67 .85
SRMR .19 .16 .25 .20 .22 .23 .11
IFI .67 .83 .62 .72 .46 .67 .85
Note. CFA ⫽ confirmatory factor analysis; SV ⫽ supportive voice; CV ⫽ constructive voice; DFV ⫽ defensive voice; DSV ⫽ destructive voice; df ⫽
degrees of freedom; CFI ⫽ comparative fit index; SRMR ⫽ standardized root mean residual; IFI ⫽ incremental fit index.

(␥ ⫽ ⫺.45, p ⬍ .01) and was not related to supportive (␥ ⫽ .17, be extraverts may try to preserve the status quo of the organization
p ⬎ .05) or constructive voice (␥ ⫽ .10, p ⬎ .05). These findings by speaking up in support of what the organization is currently
suggest that people perceived as emotionally stable are also per- doing. Finally, openness to experience was positively related to
ceived to exhibit lower levels of both forms of prohibitive voice perceptions of constructive (␤ ⫽ .43, p ⬍ .01) and defensive voice
behaviors. We also found that extraversion was positively related (␤ ⫽ ⫺.33, p ⬍ .01), as expected. Although not anticipated,
to supportive voice (␥ ⫽ .29, p ⬍ .05) but not constructive voice openness to experience was also positively related to supportive
(␤ ⫽ .11, p ⬎ .05). This finding suggests that people perceived to voice (␤ ⫽ .36, p ⬍ .01). These findings suggest that people who

Table 6
Summary of Direct Effects of Personality Predictors on Voice Measures for Nomological Validity Study (Stage 3)

Supportive voice Constructive voice Defensive voice Destructive voice


Expected Expected Expected Expected
Predictor relationship ␤ relationship ␤ relationship ␤ relationship ␤

Agreeableness ⫺ ⫺.23 ⫺ ⫺.07 ⫺ ⫺.30 ⫺ ⫺.15
Conscientiousness ⫹ .09 ⫹ .08 0 .15 0 .24
Emotional stability 0 .17 0 .10 ⫺ ⫺.27ⴱⴱ ⫺ ⫺.45ⴱⴱ
Extraversion ⫹ .29ⴱ ⫹ .11 0 .20 0 ⫺.11
Openness to experience NA .36ⴱⴱ ⫹ .43ⴱⴱ ⫺ ⫺.33ⴱⴱ NA ⫺.06
Model R2 .26 .25 .32 .24
Note. All estimates in this table are from the completely standardized solution. Expected relationships are denoted as follows: A plus sign equals expected
positive relationship; a minus sign equals expected negative relationship; 0 means expected nonrelationship; NA means that no expectation was articulated.
The chi-square test for this model had 524 degrees of freedom. ␹2 ⫽ 954.57 (p ⬍ .01); CFI ⫽ .96; SRMR ⫽ .06; IFI ⫽ .96.

p ⬍ .05. ⴱⴱ p ⬍ .01.
98 MAYNES AND PODSAKOFF

are perceived to be open to new experiences are likely to speak up Taken together, the results discussed above provide support for
in support of the organization and to make constructive sugges- many (67%) of the expected relationships between the personality
tions for change but are unlikely to verbally fight against changes variables and our voice measures. However, several relationships
that they do not want to see enacted. that were expected to be significant were not. For instance, al-
These results provide preliminary, albeit mixed, support for the though it was anticipated that agreeableness would negatively
nomological validity of the voice measures. Consistent with our impact each form of voice, the only significant relationship was
expectations, agreeableness was negatively related to defensive with defensive voice. In addition, the anticipated positive relation-
voice. However, contrary to our expectations, this personality ships between conscientiousness and the promotive forms of voice
variable was not related to the other forms of voice behavior. This were not observed. One possible explanation for these nonsignif-
may be the case because defensive voice contains an element of icant findings may be that the personality measures were obtained
stubbornness (not shared by the other voice behaviors) in that from observers rather than from the voicers themselves. This may
defensive voicers speak out against suggested changes even when have attenuated some relationships, because observers may not
others generally agree that making the changes is a good idea. This understand a voicing employee’s motives for speaking up. Thus,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

may lead to a perception that the voicing employee is obstinate and we recommend that future research examining relationships be-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

disagreeable. tween personality and voice collect personality measures from the
Conscientiousness was not related to any of the forms of voice. person expressing voice, rather than from an observer, to see if this
This was consistent with our expectations for defensive and de- makes a difference.
structive voice. However, we were surprised that conscientious-
ness was not related to either supportive or constructive voice, Stage 4: Assessment of Veridical Validity
because LePine and Van Dyne (1998) reported a significant pos-
itive relationship between conscientiousness and voice behavior, Most scholars agree that the primary objective of construct valida-
and other research has demonstrated a relationship between con- tion is to assess whether an instrument measures what it is intended to
scientiousness and other proactive behaviors (Organ & Ryan, measure (cf. Borsboom, Mellenbergh, & van Heerden, 2004). Unfor-
1995). One explanation for the lack of support may be that con- tunately, traditional methods of evaluating the construct validity of
survey measures do not directly examine this assertion, relying in-
scientious people don’t view speaking up in support of the orga-
stead on indirect, correlational evidence (e.g., convergent, discrimi-
nization or with improvement oriented suggestions as one of their
nant, and nomological validity). Borsboom et al. (2004, p. 1061)
responsibilities.
noted the irony of this and argued that “a test is valid for measuring
Emotional stability was negatively related to defensive and
an attribute if and only if (a) the attribute [behavior] exists and (b)
destructive voice but not to supportive or constructive voice. Thus,
variations in the attribute causally produce variations in the outcomes
as expected, emotional stability was a stronger predictor of the
of the measurement procedure.” Given that validity is a causal con-
prohibitive than the promotive types of voice. Given that emotion-
cept, construct validation procedures should include a test that meets
ally unstable people ruminate on the negative attributes of their
the conditions necessary to infer causality if possible. That is, the test
work environment (Barrick & Mount, 1991), it is not surprising
should demonstrate that (a) the measure covaries with the attribute it
that they express negative views regarding potential changes or
is intended to measure; (b) variation in the measure is preceded by
current work practices that are bothersome. variation in the attribute; and (c) alternative causes of variation in the
As expected, extraversion was not related to defensive or de- measure are controlled. Thus, in this phase of validation, we examine
structive voice but was positively related to supportive voice. the causal connection between the voice constructs and their mea-
Given that extraverts tend to focus on the positive aspects of their sures, a type of validity referred to as veridical validity (MacKenzie et
work environment (Barrick & Mount, 1991) and that they are al., 2011).
generally comfortable sharing their opinions, these findings were not
surprising. However, contrary to our expectations and prior research,
extraversion failed to predict constructive voice. This result may be Hypotheses
connected to the fact that prior studies of voice used a measure that Despite the merits of testing veridical validity, no studies that
captured a broader spectrum of proactive behaviors. Regardless, these we are aware of have conducted such an investigation. One reason
results suggest that perceived extraversion is a stronger predictor of for this may be that a method for doing so has only recently been
supportive voice than of the other types of voice. reported in the literature (Podsakoff et al., 2013). Podsakoff et al.
Finally, consistent with our expectations, openness to experi- noted that such an assessment can be conducted for behavioral
ence was positively related to constructive voice and negatively constructs by developing films of the behaviors at high and low
related to defensive voice. This underscores the importance of levels, showing the films to study participants, and then having the
change to constructive voice and change prevention to defensive participants rate the behaviors shown using the scale items. As-
voice, given that intellectual curiosity and finding enjoyment in suming that respondents actively attend to the level of the behav-
new experiences are key traits of this personality type (Barrick & iors portrayed in the videos, encode these behaviors in memory,
Mount, 1991). Although unanticipated, openness to experience and recall this information when they generate a response to each
also predicted supportive voice. When other employees criticize scale item, there should be a strong relationship between the level
the organization, those who are open to experience may speak up of a manipulation and responses on the scale (Tourangeau, Rips, &
and present the alternative point of view because they tend to be Rasinski, 2000). When this procedure is used, a scale is considered
broad minded and are willing to consider information from mul- to be valid if there is a high degree of correspondence between the
tiple perspectives (Barrick & Mount, 1991). level of the manipulation a study participant is exposed to and the
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 99

rating on the scale that the participant provides (Podsakoff et al., one type of voice, she or he responded to items for all the forms of
2013). Consistent with these arguments, we hypothesize: voice and the chosen comparison construct. This design allowed
for tests of veridical and discriminant validity of the voice mea-
Hypothesis 1: Variation in the levels of (a) supportive, (b) sures. Responses were on a 7 point scale (1 ⫽ strongly disagree,
constructive, (c) defensive, and (d) destructive voice behav- 7 ⫽ strongly agree). All experimental materials were presented to
iors produces variation in the measures of these constructs. participants via computer. The total time of the videos was about
This procedure can also be used to assess the degree to which 25 minutes. Veridical validity was evaluated using structural equa-
survey measures of other constructs capture the focal construct that tion modeling procedures.
was manipulated, which provides evidence of the discriminant valid- Measures. The voice items reported in Table 4 were used to
ity of the measures. As was done in Stage 3, the measures of a assess the types of voice. Reliabilities for these measures were .98,
comparison construct for each type of voice were included in the .85, .86, and .91 for supportive voice; .86, .96, .92, and .91 for
questionnaire administered to study participants (i.e., loyalty for sup- constructive voice; .87, .91, .96, and .95 for defensive voice; and
portive voice, OCB voice for constructive voice, resistance to change .88, .81, .89, and .97 for destructive voice. Loyalty, OCB voice,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

for defensive voice, and poor sportsmanship for destructive voice). resistance to change, and poor sportsmanship were all measured
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Given the above arguments regarding how respondents formulate with the same scales we used in Study 3. Reliabilities for these
responses to survey items, when a respondent is presented with survey measures were .93 for loyalty, .89 for OCB voice, .68 for resis-
items for which there is no matching information to retrieve from mem- tance to change, and .80 for poor sportsmanship.
ory, there should be very little correspondence between the level of the
behavior and responses on the scale items. Therefore, we hypothesize: Results
Hypothesis 2: Variation in the levels of (a) supportive, (b) Means, standard deviations, and correlations among the vari-
constructive, (c) defensive, and (d) destructive voice behav- ables in each of the four studies are reported in the table in
iors produce more variation in their corresponding measures Appendix C. Results of a confirmatory factor analysis are reported
than they do in the measures of conceptually related constructs in Table 5. As indicated in columns 3– 6 of this table, the measures
or the other voice behavior measures. demonstrate a sufficient level of discriminant validity, given that
the hypothesized five-factor models fit the data significantly better
than single-factor models or four-factor models in which the items
Method
from each type of voice and the comparison construct loaded on
Participants. Three hundred and one undergraduate business the same factor. To test our hypotheses, we constructed a six-
students from a large U.S. university participated in this study. On variable structural equation model for each type of voice. In each
average, participants were 21 years of age (M ⫽ 21.43 years). of these SEM models, the voice manipulation was modeled as a
They were primarily male (58%) and Caucasian (83.4%) and had manifest independent variable, whereas the survey measures of
approximately 6 months of full-time work experience (M ⫽ 6.29). voice and the appropriate comparison construct were modeled as
All participants completed the experiment for credit in a course latent dependent variables. Paths between the independent variable
related to their major. and each dependent variable were estimated, and the dependent
Procedure. Following the recommendations of Podsakoff et variables were allowed to covary. Generally speaking, fit indices
al. (2013), we used an experimental procedure to assess veridical (reported at the bottom of Figure 1) indicated a relatively good fit
validity. This required five main steps. The first three of these of the models to the data.
steps, which included (a) developing the scripts, (b) validating the The analyses are summarized in Figure 1 (panels A–D). As
scripts with subject matter experts (SMEs) and conducting manip- indicated in the figure, the manipulation of supportive voice had a
ulation checks with students, and (c) filming the videos with strong and significant impact on ratings of supportive voice (␥ ⫽
professional actors and film crews, are described in detail in .96, p ⬍ .01; R2 ⫽ .92); the manipulation of constructive voice
Appendix B. had a strong and significant impact on ratings of constructive
In the fourth step, the films were used as stimuli in four voice (␥ ⫽ .96, p ⬍ .01; R2 ⫽ .92); the manipulation of
experimental studies (one for each type of voice). Each study defensive voice had a strong and significant impact on ratings of
employed a between-subjects design with two conditions (high vs. defensive voice (␥ ⫽ .93, p ⬍ .01; R2 ⫽ .87); and the manipulation
low level of voice). Participants first watched a short instructional of destructive voice had a strong and significant impact on ratings
video in which they were told that they would be playing the role of destructive voice (␥ ⫽ .94, p ⬍ .01; R2 ⫽ .88). Thus, Hypoth-
of a recently hired employee going through a management training eses 1a, 1b, 1c, and 1d were supported.
program, and that they would be required to accurately identify the To test whether the manipulations produced more variation in
behaviors of employees who had been videotaped while interact- their corresponding measures than in the measures of the concep-
ing in the workplace. Participants then viewed a video depicting a tually related constructs, or the other voice measures, we calcu-
focal employee engaging in one of the four forms of voice at either lated the significance of the difference between the focal path (i.e.,
a high or a low level. After this, they were presented with the manipulation to its intended measure) and the other paths in the
measures of voice and related constructs (i.e., loyalty, OCB voice, model using t tests. In every model, the path estimate from the
resistance to change, and poor sportsmanship) and were asked to manipulation to its intended measure was significantly greater than
indicate their level of agreement regarding whether the voicing those of all of the other paths in the model (all p values ⬍ .01).
employee had engaged in the behaviors described by the measures. This was even the case for the difference between the path esti-
Thus, although each participant viewed a manipulation for only mates for constructive voice and OCB voice in the constructive
100 MAYNES AND PODSAKOFF
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Figure 1. Summary of direct effects of video manipulations on measures of voice and related constructs (Stage
4). For each voice manipulation, 0 ⫽ low level of voice and 1 ⫽ high level of voice. All coefficients reported
in this figure are completely standardized. SV ⫽ supportive voice; CV ⫽ constructive voice; DFV ⫽ defensive
voice; DSV ⫽ destructive voice; CFI ⫽ comparative fit index; SRMR ⫽ standardized root mean residual; IFI ⫽
incremental fit index. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01.

voice model (␥constructive ⫺ ␥OCB voice ⫽ .06, p ⬍ .01), even observer’s intentions to rely on the voicer’s opinion and evalua-
though these standardized effect sizes appear to be quite similar in tions of the voicing employee’s overall performance. Intention to
size. Thus, Hypotheses 2a, 2b, 2c, and 2d were all supported. rely on the voicer’s opinion was selected because this variable
represents what is probably the most desirable proximal outcome
Discussion for a voicing employee. Evaluation of a voicer’s overall perfor-
The results of this study provide strong evidence of the veridical mance was selected because prior research indicates that voicing
validity of the voice measures. In each model, the voice manipu- may result in positive evaluations under some conditions but
lation accounted for between 87% and 92% of the variance in its negative evaluations under other conditions (Milliken, Morrison,
intended measure. This evidence is important because it demon- & Hewlin, 2003; Seibert, Kraimer, & Crant, 2001). The mediators
strates that the manipulation of each of the voice constructs had a chosen were (a) perceptions that the voicer is concerned about the
direct, causal effect on the variability in its corresponding scale. In organization and (b) perceptions that the behavior will have a
addition, the variability observed in the voice scales cannot be positive/negative impact on organizational functioning. Although
attributed to common method biases (Podsakoff, MacKenzie, Lee, these mediators have not been examined in prior voice research,
& Podsakoff, 2003), given that the voice behaviors were manip- we elected to examine them in this study because they have been
ulated independently of participants’ ratings on the scales. shown to mediate other OCB– outcome relationships (Allen &
Beyond this, the fact that the path estimate between each voice Rush, 1998; Johnson, Erez, Kiker, & Motowidlo, 2002) and be-
manipulation and its intended measure was significantly greater cause there were good theoretical reasons to believe that they
than the other paths in each model provides evidence that the voice would mediate the relationships between the voice behaviors and
scales possess a high degree of discriminant validity. As further the outcome variables in our study.
evidence of this distinctiveness, Figure 1 indicates that the manip-
ulation of the focal construct accounted for substantially more The Influence of Voice on Perceived Organizational
variance in its intended measure than in the measures of related Concern and Positive/Negative Impact
constructs. This was even the case in the constructive voice model Supportive voicers voluntarily speak up and support, endorse, or
even though the manipulation accounted for considerable variance defend organizational policies, procedures, and work practices. Ob-
in both its intended measure (92%) and the measure of OCB voice servers are likely to rely on the opinion of a supportive voicer and
(81%), which indicates that the behaviors assessed by these mea- evaluate them favorably for several reasons. First, observers may
sures are at least somewhat different. Thus, the measures of the view this type of voice as a signal of an employee’s level of concern
four types of voice appear to possess a high degree of veridical for the organization, given that supportive voicers defend the organi-
validity and are relatively distinct from each other and from zation when others are being critical. Second, observers may believe
measures of related constructs. that supportive voice behaviors positively impact organizational func-
tioning because these behaviors preserve worthwhile work practices,
Stage 5: Assessment of Criterion-Related Validity
thereby reducing confusion over how tasks are to be performed and
The objective in this stage was to examine the criterion-related allowing supervisory staff to allocate less time to overseeing these
validity of the voice measures. The final outcomes examined were functions. Finally, expressions of support for organizational objec-
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 101

tives may also help leaders obtain buy-in from other subordinates, defensive employees have been shown to (a) experience less trust
making it easier to reach consensus on important issues (Galvin, from supervisors (Ashforth & Lee, 1990), (b) consistently under-
Balkundi, & Waldman, 2010). These arguments are consistent with achieve because they fail to learn from prior mistakes (Baumeister
prior research indicating that workplace behaviors that are supportive & Scher, 1988), (c) fail to solve problems (Argyris, 1985), and (d)
of the organization are positively related to perceptions that the experience significant performance anxiety as they anticipate fail-
employee is highly committed to the organization (Allen & Rush, ure (Baumeister & Scher, 1988). Thus, we hypothesize:
1998) and to ratings of overall performance (Whiting, Podsakoff, &
Pierce, 2008). Thus, we hypothesize: Hypothesis 5: Defensive voice is (a) negatively related to
perceptions that the voicer is concerned about the organization
Hypothesis 3: Supportive voice is positively related to per- but (b) positively related to perceptions that the behavior will
ceptions that (a) the voicer is concerned about the organization have a negative impact on organizational functioning.
and that (b) the behavior will have a positive impact on
Destructive voice is the voluntary expression of overly critical or
organizational functioning.
disparaging comments regarding an organization’s work practices,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Constructive voice is the voluntary expression of ideas regard- methods, and programs. There are several reasons why destructive
This document is copyrighted by the American Psychological Association or one of its allied publishers.

ing new or improved work methods, procedures, and programs. voice should negatively impact performance ratings and reduce the
Observers are likely to rely on a constructive voicer’s opinion and likelihood that observers would rely on the voicing employee’s opin-
to evaluate his or her performance favorably for several reasons. ion. First, observers may interpret overly critical or disparaging com-
First, observers may view constructive voice as evidence of the ments regarding organizational practices as evidence of the voicer’s
voicer’s commitment to and concern for the organization, given lack of concern for the organization, given that these behaviors clearly
that the voicing employee expended extra effort in solving an communicate contempt. Second, destructive voice may increase in-
organizational problem that was not required of him or her. Sec- terpersonal tensions between coworkers and thereby create discord
ond, given that the topic of a constructive voice message is about within a given work unit. Furthermore, destructive comments may
improving an organization’s work procedures, methods, and so on, undermine the morale of coworkers, requiring organizational leaders
observers may believe that this type of behavior will actually lead to expend substantial effort in reenergizing the workforce. Finally, it
to improvements in organizational functioning. Finally, because is also possible for managers to feel threatened by overly critical
constructive voice may improve the performance of a given work comments and respond with hostility, especially if the comments are
unit, the manager of the unit may seek to encourage this type of interpreted as a personal attack (Fincham, 1992). Empirical evidence
behavior by rewarding the voicing employee with higher perfor- indicates that destructive behaviors result in a variety of negative
mance appraisals. Consistent with these expectations, prior re- outcomes for the employee who enacts the behavior. For example,
search indicates that constructive workplace behaviors are posi- poor sportsmanship has been linked to negative ratings of employee
tively related to perceptions of organizational commitment (Allen performance (Podsakoff & MacKenzie, 1994), and complaining has
& Rush, 1998) and perceptions that the behavior will positively been related to deteriorating satisfaction with and commitment to
impact organizational performance (Whiting, Maynes, Podsakoff interpersonal relationships (Fincham, 1992). Furthermore, to the ex-
& Podsakoff, 2012). Thus, we hypothesize: tent that employees become branded as complainers, they run the risk
of being shunned by organizational leaders (Graen & Uhl-Bien,
Hypothesis 4: Constructive voice is positively related to per- 1995), excluded from informal groups (Kowalski, 1996), and left out
ceptions that (a) the voicer is concerned about the organization of social interactions (Kowalski, 1996). Thus, we hypothesize:
and that (b) the behavior will have a positive impact on
organizational functioning. Hypothesis 6: Destructive voice is (a) negatively related to
perceptions that the voicer is concerned about the organization
Defensive voicers speak out in opposition to changing how but (b) positively related to perceptions that the behavior will
things are done in the organization, even when change is inevitable have a negative impact on organizational functioning.
or implementing changes makes sense. Observers are unlikely to
rely on a defensive voicer’s opinion when making important The Mediating Effects of Perceived
organizational decisions and are likely to evaluate the voicing
Organizational Concern
employee’s performance negatively for several reasons. First,
given that defensive voicers speak out against making changes Organizational concern is defined as an employee’s desire to dem-
even when the changes are in the best interests of the organization, onstrate pride in the organization and to see the company be success-
observers are likely to perceive that the voicing employee is more ful (Rioux & Penner, 2001). A rater’s perception that a voicing
concerned about the preservation of his or her own self-interests employee’s behavior is motivated by concern (or lack of concern) for
than the organization’s success (Allen & Rush, 1998; Johnson et the organization should impact his or her ratings of the voicer’s
al., 2002). Second, observers may believe that defensive voice will overall performance and intent to rely on the voicer’s opinions,
negatively impact organizational functioning because it prolongs because raters are likely to believe that good employees are commit-
problem resolution and delays the implementation of needed ted to the organization and want to see it do well (Rioux & Penner,
changes. Finally, defensive voice behaviors may force a manager 2001). Attribution theory provides an explanation of how this effect
to work harder and for a longer period of time in order to align the operates. According to this theory, people have a fundamental need to
work of subordinates with organizational objectives, leading the assign causal explanations to the behavior of others (Heider, 1958).
manager to evaluate the voicing employee’s performance nega- When the behavior is at the discretion of the actor, people tend to
tively. Generally consistent with these expectations, chronically perceive that the cause is located within the actor. Consistent with this
102 MAYNES AND PODSAKOFF

proposition, prior research on discretionary workplace behavior indi- rate of 98%). Respondents were on average 30 years of age (M ⫽
cates that people are likely to attribute the behavior to an actor’s 30.4 years) and were primarily male (79%) and Caucasian (79%).
motive and that the assigned motive has a significant impact on how They had approximately 8 years of work experience (M ⫽ 7.8) and
the actor is perceived (Allen & Rush, 1998; Johnson et al., 2002). In had been required to evaluate the performance of subordinates as
general, raters tend to punish selfishly motivated behaviors and re- part of their job for nearly three years (2.64). Furthermore, the
ward behaviors motivated by a concern for others. Thus, based on the subordinate they chose to rate had reported to them for approxi-
reasoning presented here and in the previous sections, perceptions that mately 2.5 years (M ⫽ 2.58).
a voicer’s behavior is motivated by concern (or lack thereof) for the Procedure. Survey respondents were instructed to think about
organization should mediate the impact of supportive, constructive, the subordinate they interact with most frequently in their job.
defensive, and destructive voice on ratings of overall performance and They were then asked to rate this person on the items measuring
the intent to rely on the voicer’s opinions. Therefore, we hypothesize: the different types of voice, mediators, and outcomes. Respondents
also answered demographic questions about themselves and the
Hypothesis 7: The impact of (a) supportive, (b) constructive, focal coworker. In addition, because all of our measures were
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

(c) defensive, and (d) destructive voice on intent to rely on the obtained from the same source, we accounted for the possibility of
This document is copyrighted by the American Psychological Association or one of its allied publishers.

voicer’s opinions is mediated by a rater’s perception that the common method bias by measuring response styles and then
voicer is concerned about the organization. controlling for these in our analyses (Baumgartner & Steenkamp,
2001; Podsakoff, MacKenzie, & Podsakoff, 2012; Weijters, Schil-
Hypothesis 8: The impact of (a) supportive, (b) constructive,
lewaert, & Geuens, 2008).3
(c) defensive, and (d) destructive voice on performance eval-
Measures. All scales were assessed with a 7-point Likert
uations is mediated by a rater’s perception that the voicer is
response format (1 ⫽ Strongly Disagree to 7 ⫽ Strongly Agree).
concerned about the organization.
The measures of voice reported in Table 4 were used for this study.
Reliabilities for these measures were .92 for supportive, .95 for
The Mediating Effects of Perceived constructive, .94 for defensive, and .95 for destructive voice,
Positive/Negative Impact respectively. Perceived concern for the organization was mea-
sured with three items adapted from Rioux and Penner’s (2001)
Rater conceptions of employee performance are likely to in-
scale (␣ ⫽ .82). Perceived positive impact was measured with
clude notions regarding the impact of employee actions on orga-
three items, and perceived negative impact was measured with two
nizational functioning (Ajzen, 1991). Furthermore, intentions to
items. These items were developed based on the work of Gorden
rely on the opinion of another may depend on the perceived
(1988; positive impact ␣ ⫽ .95; negative impact ␣ ⫽ .93). Reli-
outcome of following the advice received. Thus, perceptions that a
ance on the opinion of the voicer was assessed with three items
voice behavior will positively (negatively) impact organizational
adapted from the work of Mayer and Davis (1999; ␣ ⫽ .90).
functioning should influence ratings of a voicing employee’s over-
Overall performance was assessed with three items reported by
all performance and the intent of an observer to rely on the voicer’s
MacKenzie, Podsakoff, and Fetter (1991; ␣ ⫽ .95). Finally, ac-
opinion. It is also plausible that those who stand to benefit from
quiescent, disacquiescent, extreme, and midpoint responding were
behaviors that are perceived to have a positive impact on the
measured with 15 heterogeneous items based on the work of
organization (managers, supervisors, coworkers, etc.) would be
Weijters et al. (2008).
likely to rely on the voice message and reinforce the behavior by
rewarding the voicing employee with higher performance evalua-
tions (Van Dyne & LePine, 1998; Whiting et al., 2012). In con- Results
trast, those who would be adversely impacted by behaviors that are Measurement validity. A confirmatory factor analysis using
perceived to have a negative impact on the organization may give LISREL 8.8 was conducted to examine the measurement validity
lower performance evaluations in an effort to discourage the of all questionnaire items. Results of this analysis (reported in the
behavior’s occurrence in the future. Considering these arguments
and those made previously, we hypothesize:
3
This method builds on the assumption that people tend to respond to
Hypothesis 9: The impact of (a) supportive, (b) constructive, survey questions in a consistent manner (i.e., response style). The most
common styles include acquiescence (tendency to agree with items), di-
(c) defensive, and (d) destructive voice on intent to rely on the
sacquiescence (tendency to disagree with items), extreme responding (ten-
voicer’s opinions is mediated by a rater’s perception that the dency to utilize ends of the rating scaling), and midpoint responding
behavior will positively/negatively impact the organization. (tendency to endorse the middle scale category; Weijters, Geuens, &
Schillewaert, 2010). Measures of these response styles are composed of
Hypothesis 10: The impact of (a) supportive, (b) constructive, items that do not form a “meaningful nomological network.” Instead, items
(c) defensive, and (d) destructive voice on performance eval- are randomly selected from a sample of heterogeneous items. These items
are included in a survey as buffer items between items of substantive
uations is mediated by a rater’s perception that the behavior
interest, and then the response styles are calculated as follows: Acquies-
will positively/negatively impact the organization. cence ⫽ [f(5) ⫻ 1 ⫹ f(6) ⫻ 2 ⫹ f(7) ⫻ 3]/k; Disacquiescence ⫽ [f(1) ⫻
3 ⫹ f(2) ⫻ 2 ⫹ f(3) ⫻ 1]/k; Extreme ⫽ [f(1) ⫹ f(7)]/k; and Midpoint ⫽
f(4)/k; where k is equal to the number of items and f(o) refers to the
Method frequency of response option o. In this study, these response styles were
measured with 15 heterogeneous items based on the work of Weijters et al.
Sample. The questionnaire was administered to a sample of (2008). The items were then subdivided into three groups of five, and the
executive MBA students taking classes at a large university in the response style calculations were performed on each of these groups,
United States (222 completed surveys were returned for a response resulting in three indicators for each response style.
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 103

last column of Table 5) indicate that the hypothesized 13-factor from the model in which they were. Although the pattern of
model (i.e., 4 types of voice, 4 response styles, 3 mediators, and 2 results between these two models is quite similar, we will focus
final outcomes) fit the data very well. This model fit the data our attention on the results that account for method bias because
significantly better than a single-factor model in which all items they are more conservative. In this model, the following direct
loaded on one construct; a two-factor model in which the items for effects were observed. Supportive voice significantly affected
the mediators and final outcomes loaded on one construct and the perceptions that the voicer is concerned about the organization
items for the predictors (i.e., voice behaviors and response styles) (␥ ⫽ .58, p ⬍ .01) and perceptions that the behavior positively
loaded on a separate construct; a two-factor model in which the impacts organizational functioning (␥ ⫽ .28, p ⬍ .01); con-
items for the predictors and mediators loaded on one factor and the structive voice significantly affected perceived positive impact
items for the final outcomes loaded on a separate factor; and a (␥ ⫽ .26, p ⬍ .01) but not perceived organizational concern
three-factor model wherein the items for the final outcomes loaded (p ⬎ .05); defensive voice positively affected perceptions of
on one factor, the items for the mediators loaded on another factor, negative impact (␥ ⫽ .48, p ⬍ .01) but not perceptions of
and the predictor items loaded on a third factor. In addition, all of
organizational concern (p ⬎ .05); and destructive voice nega-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the variables passed Fornell and Larcker’s (1981) test for discrim-
tively impacted perceived organizational concern (␥ ⫽ ⫺.14,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

inant validity.
p ⬍ .05) and positively affected perceived negative impact (␥ ⫽
Test of hypothesized model. A 13-factor structural equation
.26, p ⬍ .01). Thus, Hypotheses 3a, 3b, 4b, 5b, 6a, and 6b were
model was constructed in LISREL 8.8 to examine the direct and
supported, but Hypotheses 4a and 5a were not.
indirect effects of the voice measures on the mediators and out-
comes. Means, standard deviations, and correlations are reported Although not hypothesized, constructive voice had significant
in Appendix D. All paths between the voice measures and medi- effects on perceived negative impact (␥ ⫽ ⫺.21, p ⬍ .01) and
ators and outcomes were estimated. Paths between the response evaluations of the voicer’s overall performance (␥ ⫽ .13, p ⬍ .05),
styles and outcome variables were also estimated. With the excep- and defensive voice significantly affected perceived positive im-
tion of the response style variables, all nonsignificant paths were pact (␥ ⫽ ⫺.21, p ⬍ .01).
removed from the final model. Fit statistics indicate that this model Finally, perceived organizational concern did not have a
fit the data very well (df ⫽ 878; ␹2 ⫽ 2,382.44, p ⬍ .01; CFI ⫽ significant impact on intentions to rely on the voicer’s opinion
.95; SRMR ⫽ .07; IFI ⫽ .95). (␤ ⫽ .16, p ⬎ .05), but it did significantly affect overall
Results from this model are reported in Figure 2 (indirect performance evaluations (␤ ⫽ .28, p ⬍ .01); perceived positive
effects are reported in Appendix E). In the figure, there are two impact significantly impacted intentions to rely on the voicer’s
estimates next to each path (estimates are completely standard- opinion (␤ ⫽ .32, p ⬍ .01) and overall performance evaluations
ized). The first estimate is from the model in which response (␤ ⫽ .18, p ⬍ .05); and perceived negative impact had a
styles were not taken into account, and the second estimate is negative effect on both intentions to rely on the voicer’s opinion

Figure 2. Final model of voice outcomes. All estimates reported in this figure are from the completely
standardized solution. Estimates before the forward slash are from a model in which response styles were not
included as control variables, whereas estimates after the slash are from a model in which response styles were
included. Although correlations among predictors; among mediators; and between the outcomes were estimated,
they were not reported in the figure for the sake of clarity. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01.
104 MAYNES AND PODSAKOFF

(␤ ⫽ ⫺.25, p ⬍ .01) and overall performance evaluations (␤ ⫽ perceived organizational concern, which suggests that speaking up
⫺.22, p ⬍ .01). in support or defense of such things as work procedures signals to
When controlling for method biases, supportive voice impacted others in the workplace that the voicer is motivated by a concern
intentions to rely on the voicer’s opinion indirectly (indirect for the organization’s welfare. In contrast, the primary indirect
effect ⫽ .18, p ⬍ .01) via perceived positive impact (indirect path for constructive voice was through the positive impact others
effect ⫽ .09, p ⬍ .01) but not perceived organizational concern perceive that it will have on organizational functioning, which
(indirect effect ⫽ .09, p ⬍ .08). Thus, Hypothesis 9a was sup- suggests that observers value constructive voice because it poten-
ported but Hypothesis 7a was not. Supportive voice also impacted tially benefits the organization. Constructive voice also had a
performance evaluations indirectly (indirect effect ⫽ .21, p ⬍ .01) significant direct effect on overall performance evaluations,
through perceptions of organizational concern (indirect effect ⫽ whereas none of the other forms of voice did, suggesting that there
.16, p ⬍ .01) and perceived positive impact (indirect effect ⫽ .05, p ⬍ is something unique about constructive voice that is not captured
.05), supporting Hypotheses 8a and 10a (indirect effects are sum- by the mediators included in our study. One attribute that may have
marized in Appendix E). led to this finding is that only constructive voice requires voicers
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Constructive voice indirectly affected intentions to rely on the to proactively think about issues in advance (e.g., come up with
This document is copyrighted by the American Psychological Association or one of its allied publishers.

voicer’s opinion (indirect effect ⫽ .13, p ⬍ .01) via perceived new or innovative ways of doing things).
positive impact (indirect effect ⫽ .08, p ⬍ .01) and perceived Defensive voice exhibited its strongest indirect effects through
negative impact (indirect effect ⫽ .05, p ⬍ .05) but not per- perceived negative impact, suggesting that observers don’t like
ceived organizational concern. Constructive voice also indirectly defensive voice because they believe that the behavior itself may
impacted overall performance evaluations (indirect effect ⫽ .10, harm the organization. Finally, as with defensive voice, the stron-
p ⬍ .10) through perceived positive impact (indirect effect ⫽ .05, gest indirect effects for destructive voice operated through per-
p ⬍ .05) and perceived negative impact (indirect effect ⫽ .05, p ⬍ ceived negative impact. However, unlike defensive voice, destruc-
.05) but not perceptions of organizational concern. Thus, Hypoth- tive voice also exhibited indirect effects through perceived
eses 9b and 10b were supported, but Hypotheses 7b and 8b were organizational concern, suggesting that destructive voice signals a
not. The effects for perceived negative impact were not hypothe- lack of concern for the organization. A final point worth noting is
sized. that because we measured common response styles and accounted
Defensive voice exhibited a significant indirect influence on for these in our statistical analyses, concern regarding the impact
participants’ intent to rely on the voicer’s opinions (indirect of same source biases on our findings should be substantially
effect ⫽ ⫺.19, p ⬍ .01) through perceived negative impact (indi- reduced.
rect effect ⫽ ⫺.12, p ⬍ .01) and perceived positive impact
(indirect effect ⫽ ⫺.07, p ⬍ .05) but not through perceptions of
General Discussion
organizational concern. Defensive voice also had a significant
indirect effect on evaluations of overall performance (indirect The purpose of the present research was to (a) clarify the
effect ⫽ ⫺.15, p ⬍ .01) via perceived negative impact (indirect effect ⫽ boundaries of the employee voice behavior domain; (b) develop a
⫺.11, p ⬍ .01) and perceived positive impact (indirect effect ⫽ new, more expansive voice behavior framework; (c) report on the
⫺.04, p ⬍ .05) but not perceived organizational concern. Thus, construct validation of new measures of the voice behaviors iden-
Hypotheses 9c and 10c were supported, whereas Hypotheses 7c tified by this framework; and (d) examine the antecedents and
and 8c were not. The effects for perceived positive impact were not consequences of these behaviors. Our results suggest that em-
hypothesized. ployee voice behaviors can be conceptualized along two intersect-
Finally, destructive voice impacted intentions to rely on the ing dimensions (i.e., preservation vs. challenge orientation and
voicer’s opinion indirectly (indirect effect ⫽ ⫺.09, p ⬍ .01) via promotive vs. prohibitive), resulting in four different types of
perceived negative impact (indirect effect ⫽ .07, p ⬍ .01) but voice: supportive, constructive, defensive, and destructive voice.
not perceived organizational concern (indirect effect ⫽ .02, p ⬎ Even though our framework may be seen by some as a departure
.05). Destructive voice also impacted performance evaluations from recent conceptualizations of voice that have focused primar-
indirectly (indirect effect ⫽ ⫺.10, p ⬍ .01) via perceived ily on the positive attributes of this behavior (Liang et al., 2012;
negative impact (indirect effect ⫽ .06, p ⬍ .01) but not per- Van Dyne & LePine, 1998), we found strong support for our
ceived organizational concern (indirect effect ⫽ .04, p ⬍ .09). four-factor model in that (a) participants in our content validity
Thus, Hypotheses 9d and 10d were supported, but Hypotheses study easily categorized the voice items in a manner that was
7d and 8d were not. consistent with the overall dimensions of our framework (i.e.,
preservation/challenge and promotive/prohibitive); (b) the confir-
matory factor analyses (summarized in Table 5) we conducted
Discussion
supported the four-factor measurement model, and the latent con-
A summary of the support we found for our hypotheses is structs accounted for more variance in their corresponding items
provided in Table 7. As indicated in the table in this appendix, 15 than these constructs shared with the other forms of voice; and (c)
of our 24 hypotheses were supported. These results provide addi- the forms of voice exhibited differential relationships with the
tional evidence regarding the empirical usefulness of the voice personality predictors and with the outcomes. Although some may
behaviors outlined in our model, given that the measures ac- view our framework as being somewhat controversial because it
counted for significant, unique variance in the outcome variables incorporates negative attributes of voice, we believe that our
and that they exhibited different patterns of effects. For instance, findings build on and extend the original theoretical work of
supportive voice exhibited its strongest indirect effects through Hirschman (1970) and the more recent contributions of Gorden
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 105

Table 7
Summary of Support for Hypothesized Relationships in Study 5 (Outcomes of Voice)

Organizational Positive Negative


Hypothesis concern impact impact

Direct effects
H3 Supportive voice is positively related to (a) perceived Y Y —
organizational concern and (b) perceived positive impact.
H4 Constructive voice is positively related to (a) perceived N Y —
organizational concern and (b) perceived positive impact.
H5 Defensive voice is (a) negatively related to perceived N — Y
organizational concern but (b) positively related to perceived
negative impact.
H6 Destructive voice is (a) negatively related to perceived Y — Y
organizational concern but (b) positively related to perceived
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

negative impact.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Indirect effects
SV CV DFV DSV
H7 Perceived organizational concern mediates the impact of (a) Nⴱ N N Nⴱ
supportive, (b) constructive, (c) defensive, and (d) destructive
voice on intent to rely on a voicer’s opinion.
H8 Perceived organizational concern mediates the impact of (a) Y N N N
supportive, (b) constructive, (c) defensive, and (d) destructive
voice on evaluations of a voicer’s performance.
H9 Perceived positive impact mediates the impact of (a) supportive Y Y Y Y
and (b) constructive voice on intent to rely on a voicer’s
opinion; perceived negative impact mediates the effect of (c)
defensive and (d) destructive voice.
H10 Perceived positive impact mediates the impact of (a) supportive Y Y Y Y
and (b) constructive voice on evaluations of a voicer’s
performance; perceived negative impact mediates the effect of
(c) defensive and (d) destructive voice.
Note. The asterisks next to the Ns for Hypothesis 7 indicate that the relationships were significant when not controlling for method variance. SV ⫽
supportive voice; CV ⫽ constructive voice; DFV ⫽ defensive voice; DSV ⫽ destructive voice; Y ⫽ supported; N ⫽ not supported. A dash indicates not
hypothesized.

(1988) and Van Dyne et al. (2003), in that they suggest that voice Implications for Voice Research
behavior is more complex than has been recognized in much of the
prior empirical research reported in the literature (Rusbult et al., Our findings have several important implications for future
1988; Withey & Cooper, 1989). research. First, prior researchers (Detert & Burris, 2007; Grant et
Aside from the support for the four-factor conceptualization of al., 2009; Van Dyne & LePine, 1998) have acknowledged the lack
voice, the studies reported in this paper highlight several strengths of consensus regarding what voice behavior is. Much of this
of the voice measures. First, the measures possess adequate dis- problem has stemmed from the existence of multiple, incomplete
criminant validity, given that they are empirically distinguishable conceptualizations of voice behavior in the literature. Thus, the
from each other and from a number of conceptually related con- conceptual framework and definitions of the different forms of
structs. Second, the new measures exhibit an acceptable level of voice provided in this paper should help to clarify the domain for
nomological and criterion-related validity, because they are gen- voice researchers in the future.
erally related to the measures of the antecedents and consequences Second, although scholars have contemplated the existence of
that they should be related to theoretically. Finally, the voice various types of voice (Gorden, 1988; Rusbult et al., 1988; Van
measures possess a high degree of veridical validity, given that Dyne et al., 2003; Withey & Cooper, 1989), a validated survey
they were shown to be causally connected to actual voice behav- measure exists for only one of these; unfortunately, even this
iors. This evidence is particularly powerful because it is not subject measure has come under some recent criticism (Detert & Burris,
to the limitations of correlational data or to issues related to 2007; Grant et al., 2009). Furthermore, the history of voice re-
common method variance. Furthermore, findings from this stage search is one of poor measurement and inconsistent empirical
also provide very strong evidence regarding the discriminant
findings (Rusbult et al., 1988; Withey & Cooper, 1989). Thus, the
validity of our measures, because they captured more variance
development and validation of scales for the four types of voice
in the actual behaviors they were intended to capture than did
should give researchers greater confidence when examining
measures of related constructs or measures of the other forms of
questions related to voice behavior and should facilitate re-
voice. Thus, taken together, the evidence provided in this paper
lends a significant amount of support to the validity of the voice search on the types of voice that have been relatively ignored in
scales. past research.
106 MAYNES AND PODSAKOFF

Finally, there has been some debate in the literature regarding the sizes. Therefore, future research should replicate these findings in
impact of constructive voice behaviors on important outcomes for a more homogenous sample in order to determine whether our
voicing employees. Some scholars argue that because constructive observed effect sizes were suppressed.
voice is oriented toward organizationally functional change, decision Another limitation results from the fact that all of the measures
makers in the organization (e.g., managers) will value constructive in Stages 3 and 5 were gathered from the same source. Thus,
voice and reward this type of behavior by granting higher perfor- common method biases (Podsakoff et al., 2003) may have been
mance evaluations, recommending voicing employees for promo- partly responsible for the observed relationships between the voice
tions, and so on (Whiting et al., 2008; Whiting et al., 2012). However, scales and their comparison constructs and the personality vari-
given the challenging nature of constructive voice, other scholars have ables (Stage 3), and between the voice scales and the outcomes for
argued that it may damage interpersonal relationships and result in voicing employees (Stage 5). However, it is important to note that
negative outcomes for employees choosing to engage in this type of these biases could not have accounted for the effects of our
behavior (Milliken et al., 2003; Seibert et al., 2001; Van Dyne et al., manipulations of the voice constructs on their respective measures
1995). Given the issues related to the measurement of voice in Stage 4, because the independent variable (video manipulation
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

noted earlier in our paper, it is possible that the inconsistent of the voice behaviors) was not obtained from the same source as
This document is copyrighted by the American Psychological Association or one of its allied publishers.

empirical findings and differing conceptual views related to the dependent variable (participants’ ratings of measures of the
individual outcomes for voicing employees may be a function focal construct). In addition, method biases also have difficulty
of the type of voice enacted. Consistent with this view, we accounting for the majority of the findings of Study 5, since these
found that promotive types of voice result in positive outcomes findings were largely the same even after we controlled for re-
for the voicing employee, whereas prohibitive forms resulted in sponse style biases. Thus, although additional research will be
negative outcomes. Thus, our understanding of how voicing necessary to assess the generalizability of our findings, the pre-
impacts the voicer should be enhanced by examining the effects liminary results suggest that method biases cannot account for all
of the various types of voice included in our study. of our findings.
A final limitation is that our measures focused exclusively on
the perspective of observers of voice. Although we believe this
Implications for Practitioners
approach is consistent with the majority of prior research in the
This research should be informative to practitioners for several voice domain, Burris, Detert, and Romney (2013) recently
reasons. First, the fact that we found a relationship between the noted the importance of examining voice from the perspective
different forms of voice and performance appraisals and observer of others in the workplace. Indeed, results of the Burris et al.
intentions to rely on the voicer’s opinion should prove useful to (2013) paper suggest that ratings of voice provided by a voicing
employees who are thinking about speaking up, given that the pro- employee may differ from those provided by an observer and
motive forms of voice were much more likely than the prohibitive that both the causes and effects of such differences are impor-
forms to result in positive outcomes for the voicer. Furthermore, when tant areas that warrant further study. Thus, although we focused
observers of voice events perceive that the behavior is motivated by on the perspective of observers of voice behavior, future re-
a concern for the welfare of the organization or that the behavior will search should not be limited to examining voice solely from this
ultimately have a positive impact on organizational functioning, the perspective.
observers tend to reward the voicing employee with higher perfor-
mance evaluations. Thus, we recommend that, before speaking up,
voicers think about the signal that their communication sends to those
Conclusion
who will be receiving the voice message. Scholarly interest in employee voice behavior has grown
Second, given the potential benefits of voice and its voluntary, dramatically over the past few decades. Unfortunately, the
conspicuous, and potentially damaging nature, understanding why voice literature suffers from several limitations, not the least of
employees speak up is important from a practical perspective. which is a need to better understand the conceptual nature of the
Thus, the information related to the personality predictors of voice voice construct and to develop valid measures of the different
reported in this paper should be useful to practicing managers. types of voice. Thus, the purpose of the research reported in this
Unfortunately, beyond the work examining the antecedents of paper has been to gain a deeper understanding of the different
constructive voice (Burris et al., 2008; Detert & Burris, 2007; forms of employee voice behavior and to develop valid mea-
LePine & Van Dyne, 2001; Tangirala & Ramanujam, 2008; Van sures of them. Although additional research is clearly needed to
Dyne & LePine, 1998), the other forms of voice have received establish the generalizability of our findings, we believe that the
very little attention in this area. Thus, we recommend that future studies reported in this paper provide preliminary support for
research address this issue. the validity our new voice behavior framework and for the
measures that we developed and that they provide evidence
Limitations regarding some of the antecedents and outcomes of the various
forms of voice behavior.
As is true of all studies, those reported in this paper have some
limitations. One limitation arises from the heterogeneity (e.g.,
industry, age, work experience) of the executive MBA samples References
used in Stages 3 and 5. Although the breadth of these samples Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior
decreases generalizability concerns, it may also have increased the and Human Decision Processes, 50, 179 –211. doi:10.1016/0749-
amount of error-related variance; resulting in decreased effect 5978(91)90020-T
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 107

Allen, T. D., & Rush, M. C. (1998). The effects of organizational citizen- Gorden, W. I. (1988). Range of employee voice. Employee Responsibilities
ship behavior on performance judgments: A field study and a laboratory and Rights Journal, 1, 283–299. doi:10.1007/BF01556937
experiment. Journal of Applied Psychology, 83, 247–260. doi:10.1037/ Graen, G. B., & Uhl-Bien, M. (1995). Relationship-based approach to
0021-9010.83.2.247 leadership: Development of leader–member exchange (LMX) theory of
Anderson, J. C., & Gerbing, D. W. (1991). Predicting the performance of leadership over 25 years: Applying a multi-level multi-domain approach.
measures in a confirmatory factor analysis with a pretest assessment of Leadership Quarterly, 6, 19 –247. doi:10.1016/1048-9843(95)90036-5
their substantive validities. Journal of Applied Psychology, 76, 732–740. Graham, J. (1991). An essay on organizational citizenship behavior. Em-
doi:10.1037/0021-9010.76.5.732 ployee Responsibilities and Rights Journal, 4, 249 –270. doi:10.1007/
Argyris, C. (1985). Strategy, change, and defensive routines. Boston, MA: BF01385031
Pitman. Grant, A., Parker, S., & Collins, C. (2009). Getting credit for proactive
Ashforth, B., & Lee, R. (1990). Defensive behavior in organizations: A behavior: Supervisor reactions depend on what you value and how you
preliminary model. Human Relations, 43, 621– 648. doi:10.1177/ feel. Personnel Psychology, 62, 31–55. doi:10.1111/j.1744-6570.2008
001872679004300702 .01128.x
Barrick, M. R., & Mount, M. K. (1991). The Big Five personality dimen- Hagedoorn, M., Van Yperen, N. W., Van de Vliert, E., & Buunk, B. P.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

sions and job performance: A meta-analysis. Personnel Psychology, 44, (1999). Employees’ reactions to problematic events: A circumplex struc-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

1–26. doi:10.1111/j.1744-6570.1991.tb00688.x ture of five categories of responses, and the role of job satisfaction.
Baumeister, R. F., & Scher, S. J. (1988). Self-defeating behavior patterns Journal of Organizational Behavior, 20, 309 –321. doi:10.1002/
among normal individuals: Review and analysis of common self- (SICI)1099-1379(199905)20:3⬍309::AID-JOB895⬎3.0.CO;2-P
destructive tendencies. Psychological Bulletin, 104, 3–22. doi:10.1037/ Haynes, S. N., Richard, D. C. S., & Kubany, E. S. (1995). Content validity
0033-2909.104.1.3 in psychological assessment: A functional approach to concepts and
Baumgartner, H., & Steenkamp, J. B. E. M. (2001). Response styles in methods. Psychological Assessment, 7, 238 –247. doi:10.1037/1040-
marketing research: A cross-national investigation. Journal of Market- 3590.7.3.238
ing Research, 38, 143–156. doi:10.1509/jmkr.38.2.143.18840 Heider, E. (1958). The psychology of interpersonal relations. New York,
Bollen, K. A. (1989). Structural equations with latent variables. New NY: Wiley. doi:10.1037/10628-000
York, NY: Wiley. Hinkin, T., & Tracey, J. (1999). An analysis of variance approach to
Borman, W. C., & Motowidlo, S. J. (1993). Expanding the criterion
content validation. Organizational Research Methods, 2, 175–186. doi:
domain to include elements of contextual performance. In N. Schmitt &
10.1177/109442819922004
W. C. Borman (Eds.), Personnel selection in organizations (pp. 71–98).
Hirschman, A. O. (1970). Exit, voice, and loyalty: Responses to decline in
San Francisco, CA: Jossey-Bass.
firms, organizations, and states. Cambridge, MA: Harvard University
Borsboom, D., Mellenbergh, G. J., & van Heerden, J. (2004). The concept
Press.
of validity. Psychological Review, 111, 1061–1071. doi:10.1037/0033-
Hu, L., & Bentler, P. (1999). Cutoff criteria for fit indexes in cova-
295X.111.4.1061
riance structure analysis: Conventional criteria versus new alter-
Burris, E. R., Detert, J. R., & Chiaburu, D. S. (2008). Quitting before
natives. Structural Equation Modeling, 6, 1–55. doi:10.1080/
leaving: The mediating effects of psychological attachment and detach-
10705519909540118
ment on voice. Journal of Applied Psychology, 93, 912–922. doi:
Ilies, R., Fulmer, I. S., Spitzmuller, M., & Johnson, M. D. (2009).
10.1037/0021-9010.93.4.912
Personality and citizenship behavior: The mediating role of job
Burris, E. R., Detert, J. R., & Romney, A. C. (2013). Speaking up vs. being
satisfaction. Journal of Applied Psychology, 94, 945–959. doi:10.1037/
heard: The disagreement around and outcomes of employee voice.
a0013329
Organization Science, 24, 22–38. doi:10.1287/orsc.1110.0732
Churchill, G. A. (1979). A paradigm for developing better measures of Johnson, D. E., Erez, A., Kiker, D. S., & Motowidlo, S. J. (2002). Liking
marketing constructs. Journal of Marketing Research, 16, 64 –73. doi: and attributions of motives as mediators of the relationships between
10.2307/3150876 individuals’ reputations, helpful behaviors, and raters’ reward decisions.
Crant, J. M., Kim, T. Y., & Wang, J. (2011). Dispositional antecedents of Journal of Applied Psychology, 87, 808 – 815. doi:10.1037/0021-9010
demonstration and usefulness of voice behavior. Journal of Business and .87.4.808
Psychology, 26, 285–297. doi:10.1007/s10869-010-9197-y Judge, T. A., & Bono, J. E. (2001). Relationship of core self-evaluations
Detert, J. R., & Burris, E. R. (2007). Leadership behavior and employee traits—self-esteem, generalized self-efficacy, locus of control, and emo-
voice: Is the door really open? Academy of Management Journal, 50, tional stability—with job satisfaction and job performance: A meta-
869 – 884. doi:10.5465/AMJ.2007.26279183 analysis. Journal of Applied Psychology, 86, 80 –92. doi:10.1037/0021-
Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective 9010.86.1.80
well-being: Three decades of progress. Psychological Bulletin, 125, Judge, T. A., Thoresen, C. J., Pucik, V., & Welbourne, T. M. (1999).
276 –302. doi:10.1037/0033-2909.125.2.276 Managerial coping with organizational change: A dispositional perspec-
Fincham, F. D. (1992). The account episode in close relationships. In M. L. tive. Journal of Applied Psychology, 84, 107–122. doi:10.1037/0021-
McLaughlin, M. Cody, & S. J. Read (Eds.), Explaining one’s self to 9010.84.1.107
others: Reason-giving in a social context (pp. 167–182). Hillsdale, NJ: Kassing, J. W. (1998). Development and validation of the Organizational
Erlbaum. Dissent Scale. Management Communication Quarterly, 12, 183–229.
Fornell, C., & Larcker, D. (1981). Evaluating structural equation models doi:10.1177/0893318998122002
with unobservable variables and measurement error. Journal of Market- Kassing, J. W. (2011). Dissent in organizations. Cambridge, United King-
ing Research, 18, 39 –50. doi:10.2307/3151312 dom: Polity Press.
Galvin, B. M., Balkundi, P., & Waldman, D. A. (2010). Spreading the word: The Kerlinger, F. N. (1973). Foundations of behavioral research (2nd ed.).
role of surrogates in charismatic leadership processes. Academy of Management New York, NY: Holt McDougal.
Review, 35, 477–494. doi:10.5465/AMR.2010.51142542 Kowalski, R. M. (1996). Complaints and complaining: Functions, anteced-
Goldberg, L. R. (1993). The structure of phenotypic personality traits. ents, and consequences. Psychological Bulletin, 119, 179 –196. doi:
American Psychologist, 48, 26 –34. doi:10.1037/0003-066X.48.1.26 10.1037/0033-2909.119.2.179
108 MAYNES AND PODSAKOFF

LePine, J. A., & Van Dyne, L. (1998). Predicting voice behavior in work video techniques to supplement traditional construct validation proce-
groups. Journal of Applied Psychology, 83, 853– 868. doi:10.1037/0021- dures. Journal of Applied Psychology, 98, 99 –113. doi:10.1037/
9010.83.6.853 a0029570
LePine, J. A., & Van Dyne, L. (2001). Voice and cooperative behavior as Podsakoff, P. M., & MacKenzie, S. (1994). Organizational citizenship
contrasting forms of contextual performance: Evidence of differential behaviors and sales unit effectiveness. Journal of Marketing Research,
relationships with Big Five personality characteristics and cognitive 31, 351–363. doi:10.2307/3152222
ability. Journal of Applied Psychology, 86, 326–336. doi:10.1037/0021- Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003).
9010.86.2.326 Common method biases in behavioral research: A critical review of the
Liang, J., Farh, C., & Farh, L. (2012). A two-wave examination of the literature and recommended remedies. Journal of Applied Psychology,
psychological antecedents of voice behavior. Academy of Management 88, 879 –903. doi:10.1037/0021-9010.88.5.879
Journal, 55, 71–92. doi:10.5465/amj.2010.0176 Podsakoff, P. M., MacKenzie, S. B., & Podsakoff, N. P. (2012). Sources of
MacKenzie, S. B., Podsakoff, P. M., & Fetter, R. (1991). Organizational method bias in social science research and recommendations on how to
citizenship behavior and objective productivity as determinants of man- control it. Annual Review of Psychology, 63, 539 –569. doi:10.1146/
agerial evaluations of salespersons’ performance. Organizational Behav- annurev-psych-120710-100452
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ior and Human Decision Processes, 50, 123–150. doi:10.1016/0749- Rioux, S. M., & Penner, L. A. (2001). The causes of organizational
This document is copyrighted by the American Psychological Association or one of its allied publishers.

5978(91)90037-T citizenships behavior: A motivational analysis. Journal of Applied Psy-


MacKenzie, S. B., Podsakoff, P. M., & Podsakoff, N. P. (2011). Construct chology, 86, 1306 –1314. doi:10.1037/0021-9010.86.6.1306
measurement and validation procedures in MIS and behavioral research: Rusbult, C. E., Farrell, D., Rogers, G., & Mainous, A. G. (1988). Impact of
Integrating new and existing techniques. MIS Quarterly, 35, 293–334. exchange variables on exit, voice, loyalty, and neglect: An integrative
Mayer, R. C., & Davis, J. H. (1999). The effect of the performance model of responses to declining job status satisfaction. Academy of
appraisal system on trust for management: A field quasi-experiment. Management Journal, 31, 599 – 627. doi:10.2307/256461
Journal of Applied Psychology, 84, 123–136. doi:10.1037/0021-9010.84 Saucier, G. (1994). Mini-markers: A brief version of Goldberg unipolar
.1.123 Big-Five markers. Journal of Personality Assessment, 63, 506 –516.
McCrae, R. R. (1996). Social consequences of experiential openness. doi:10.1207/s15327752jpa6303_8
Psychological Bulletin, 120, 323–337. doi:10.1037/0033-2909.120.3 Schwab, D. P. (1980). Construct validity in organizational behavior. In
.323
B. M. Staw & L. L. Cummings (Eds.), Research in organizational
Milliken, F. J., Morrison, E. W., & Hewlin, P. F. (2003). An exploratory
behavior (Vol. 2, pp. 3– 43). Greenwich, CT: JAI Press.
study of employee silence: Issues that employees don’t communicate
Seibert, S. E., Kraimer, M. L., & Crant, J. M. (2001). What do proactive
upward and why. Journal of Management Studies, 40, 1453–1476.
people do? A longitudinal model linking proactive personality and
doi:10.1111/1467-6486.00387
career success. Personnel Psychology, 54, 845– 874. doi:10.1111/j.1744-
Moorman, R. H., & Blakely, G. L. (1995). Individualism– collectivism as
6570.2001.tb00234.x
an individual difference predictor of organizational citizenship behavior.
Tangirala, S., & Ramanujam, R. (2008). Exploring nonlinearity in em-
Journal of Organizational Behavior, 16, 127–142. doi:10.1002/job
ployee voice: The effects of personal control and organizational identi-
.4030160204
fication. Academy of Management Journal, 51, 1189 –1203. doi:
Morrison, E. W. (2011). Employee voice behavior: Integration and direc-
10.5465/AMJ.2008.35732719
tions for future research. Academy of Management Annals, 5, 373– 412.
Tourangeau, R., Rips, L. J., & Rasinski, K. (2000). The psychology of
doi:10.1080/19416520.2011.574506
survey response. Cambridge, England: Cambridge University Press.
Nederhof, A. J. (1985). Methods of coping with social desirability bias: A
doi:10.1017/CBO9780511819322
review. European Journal of Social Psychology, 15, 263–280. doi:
10.1002/ejsp.2420150303 Van Dyne, L., Ang, S., & Botero, I. (2003). Conceptualizing employee
Netemeyer, R. G., Bearden, W. O., & Sharma, S. (2003). Scaling proce- silence and employee voice as multidimensional constructs. Journal of
dures: Issues and applications. Thousand Oaks, CA: Sage. Management Studies, 40, 1359 –1392. doi:10.1111/1467-6486.00384
Nikolaou, I., Vakola, M., & Bourantas, D. (2008). Who speaks up at work? Van Dyne, L., Cummings, L. L., & Parks, J. M. (1995). Extra-role
Dispositional influences on employees’ voice behavior. Personnel Re- behaviors: In pursuit of construct and definitional clarity. Research in
view, 37, 666 – 679. doi:10.1108/00483480810906892 Organization Behavior, 17, 215–285.
Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). Van Dyne, L., Graham, J. W., & Dienesch, R. M. (1994). Organizational
New York, NY: McGraw-Hill. citizenship behavior: Construct redefinition, measurement, and valida-
Oh, I. S., Wang, G., & Mount, M. K. (2011). Validity of observer ratings tion. Academy of Management Journal, 37, 765– 802. doi:10.2307/
of the five-factor model of personality traits: A meta-analysis. Journal of 256600
Applied Psychology, 96, 762–773. doi:10.1037/a0021832 Van Dyne, L., & LePine, J. A. (1998). Helping and voice extra-role
Oreg, S. (2003). Resistance to change: Developing an individual differ- behaviors: Evidence of construct and predictive validity. Academy of
ences measure. Journal of Applied Psychology, 88, 680 – 693. doi: Management Journal, 41, 108 –119. doi:10.2307/256902
10.1037/0021-9010.88.4.680 Venkataramani, V., & Tangirala, S. (2010). When and why do central
Organ, D. W. (1988). Organizational citizenship behavior: The good employees speak up? An examination of mediating and moderating
soldier syndrome. Lexington, MA: Lexington Books. variables. Journal of Applied Psychology, 95, 582–591. doi:10.1037/
Organ, D. W., & Ryan, K. (1995). A meta-analytic review of attitudinal a0018315
and dispositional predictors of organizational citizenship behavior. Per- Warren, D. (2003). Constructive and destructive deviance in organizations.
sonnel Psychology, 48, 775– 802. doi:10.1111/j.1744-6570.1995 Academy of Management Review, 28, 622– 632.
.tb01781.x Weijters, B., Geuens, M., & Schillewaert, N. (2010). The stability of
Pavot, W., Diener, E., & Fujita, F. (1990). Extraversion and happiness. individual response styles. Psychological Methods, 15, 96 –110. doi:
Personality and Individual Differences, 11, 1299 –1306. doi:10.1016/ 10.1037/a0018721
0191-8869(90)90157-M Weijters, B., Schillewaert, N., & Geuens, M. (2008). Assessing response
Podsakoff, N. P., Podsakoff, P. M., MacKenzie, S. B., & Klinger, R. L. styles across modes of data collection. Journal of the Academy of
(2013). Are we really measuring what we say we’re measuring? Using Marketing Science, 36, 409 – 422. doi:10.1007/s11747-007-0077-6
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 109

Whitbourne, S. K. (1986). Openness to experience, identity flexibility, and Whiting, S. W., Podsakoff, P. M., & Pierce, J. R. (2008). Effects of task
life change in adults. Journal of Personality and Social Psychology, 50, performance, helping, voice, and organizational loyalty on performance
163–168. doi:10.1037/0022-3514.50.1.163 appraisal ratings. Journal of Applied Psychology, 93, 125–139. doi:
Whiting, S. W., Maynes, T. D., Podsakoff, N. P., & Podsakoff, P. M. 10.1037/0021-9010.93.1.125
(2012). Effects of message, source, and context on evaluations of em- Withey, M. J., & Cooper, W. H. (1989). Predicting exit, voice, loyalty, and
ployee voice behavior. Journal of Applied Psychology, 97, 159 –182. neglect. Administrative Science Quarterly, 34, 521–539. doi:10.2307/
doi:10.1037/a0024871 2393565

Appendix A
Variable Means, Standard Deviations, Intercorrelations, and Internal Consistency Reliabilities (Stage 3)
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Variable M SD AVE 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Supportive voice (SV) 5.03 1.21 .62 (.89)


2. Constructive voice (CV) 4.75 1.57 .77 .57 (.95)
3. Defensive voice (DFV) 2.72 1.36 .71 ⫺.27 ⫺.32 (.92)
4. Destructive voice (DSV) 2.86 1.61 .72 ⫺.36 ⫺.09 .33 (.93)
5. Loyalty (LY) 4.93 1.25 .57 .46 .34 ⫺.14 ⫺.52 (.82)
6. OCB voice (OCBV) 4.78 1.34 .58 .58 .76 ⫺.24 ⫺.17 .47 (.89)
7. Resistance to change (RC) 3.68 1.35 .55 ⫺.42 ⫺.56 .48 .24 ⫺.24 ⫺.56 (.85)
8. Poor sportsmanship (PS) 4.88 1.63 .70 ⫺.40 ⫺.22 .42 .66 ⫺.50 ⫺.29 .40 (.90)
9. Agreeableness (AG) 4.44 1.78 .72 .06 .10 ⫺.17 ⫺.16 .10 .08 ⫺.08 ⫺.19 (.89)
10. Conscientiousness (CO) 4.41 1.69 .58 .20 .17 ⫺.01 ⫺.02 .16 .14 ⫺.06 ⫺.12 .59 (.88)
11. Emotional stability (ES) 3.55 1.76 .60 .14 .15 ⫺.41 ⫺.40 ⫺.34 ⫺.20 .28 .66 ⫺.31 ⫺.09 (.90)
12. Extraversion (EX) 4.44 1.71 .60 .02 .15 .12 .01 .13 .21 ⫺.07 ⫺.05 .46 .53 .04 (.82)
13. Openness to experience (OE) 4.93 1.31 .63 .44 .56 ⫺.36 ⫺.21 .40 .54 ⫺.62 ⫺.39 .10 .11 ⫺.31 .09 (.79)
Note. Cronbach alpha reliabilities are reported in parentheses on the diagonal. Correlations with an absolute value greater than or equal to.14 are
significant at the p ⬍ .05 level. AVE ⫽ average variance extracted; OCB ⫽ organizational citizenship behavior.

(Appendices continue)
110 MAYNES AND PODSAKOFF

Appendix B
Summary of Procedures Used in Developing and Validating Video Scripts

The following is a description of first three steps used to validate four ways. First, participants selected the intended voice category
the scripts for the veridical validity study (Stage 4) 89% of the time. Second, the average ratings across all four of the
voice behaviors for the high level scripts was high (M ⫽ 6.36), and
Step 1 the average ratings across all of the behaviors for the low level
scripts was low (M ⫽ 1.92). Third, ANOVA comparisons of the
Scripts for the experimental stimuli (videos of workplace inter- ratings on the high versus low scripts for each type of voice
actions) were developed and validated. Scripts were based on the indicated that they were significantly different from each other in
conceptual definitions of the four voice behaviors. Each script each case (all Fs ⬎ 22.84; dfs ranging from [1, 12] to [1, 14],
depicted workplace interactions in a fictitious auditing firm depending on sample size; all ps ⬍ .01). Finally, results of
(Brooks, Wilson, & Kline), wherein a focal employee engaged in
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ANOVA comparisons within the manipulation levels (i.e., high or


either a high or a low level of one form of voice. The employees
This document is copyrighted by the American Psychological Association or one of its allied publishers.

low) indicated that the high level scripts were equally high, F (3,
described in the scripts were members of a financial statement 27) ⫽ 0.64, p ⬎ .05, and that the low level scripts were equally
audit team responsible for the completion of the year-end audit of low, F (3, 27) ⫽ .18, p ⬎ .05, across the four types of voice. Thus,
a fictitious client (Capital Bank). The audit team was composed of the high level scripts depicted high level voice behaviors, the low
a manager (Luke), the team member exhibiting voice (Sarah), and level scripts depicted low level voice behaviors, the high and low
two coworkers (Josh and Ryan). (The scripts are available from the level scripts depicted significantly different levels of behaviors,
first author upon request.) and the manipulation levels were consistent across the types of
voice.
Step 2
We examined the validity of the scripts. This was done in two Step 3
stages. First, subject matter experts (SME) analyzed the scripts to
ensure that the manipulations were content valid. Second, we We filmed professional actors playing the roles described in the
conducted manipulation checks of the scripts with a sample of 62 scripts. The manager role was played by a male actor in his early
undergraduate students majoring in business. Each participant 30s, and the voicing employee was played by a female actor in her
reviewed one of the eight video scripts. They were then provided late 20s. The final two team members were played by male actors
with five categories of behavior (i.e., supportive, constructive, in their mid-to-late 20s. For the purposes of filming, the actors
defensive, destructive, or other) and were asked to categorize the were trained to limit their facial expressions, voice variation, and
behavior of the voicing employee (Sarah). After they selected the expressed emotions so as to avoid introducing confounding effects.
category, participants were asked to rate the level of the behavior The filming was conducted by a professional film crew using
on a 7 point scale (1 ⫽ Very Low Level, 7 ⫽ Very High Level). commercial grade video equipment to ensure quality. A total of
Results of this study support the validity of the manipulations in eight films were created (high/low for each type of voice).

(Appendices continue)
EXPANDED SET OF EMPLOYEE VOICE BEHAVIORS 111

Appendix C
Variable Means, Standard Deviations, and Intercorrelations for Veridical Validity Study (Stage 4)

Supportive voice model Constructive voice model


Variable M SD 1 2 3 4 5 6 Variable M SD 1 2 3 4 5 6

1. Manipulation — — — 1. Manipulation — — —
2. Supportive voice 3.98 2.44 .96ⴱⴱ (.98) 2. Supportive voice 2.11 1.07 .22 (.85)
3. Constructive voice 1.52 0.81 .47ⴱⴱ .46ⴱⴱ (.86) 3. Constructive voice 3.98 2.34 .93ⴱⴱ .28ⴱ (.96)
4. Defensive voice 1.41 0.67 ⫺.21 ⫺.19 .13 (.87) 4. Defensive voice 1.82 1.16 ⫺.17 .24ⴱ ⫺.14 (.91)
5. Destructive voice 1.36 0.71 ⫺.24ⴱ ⫺.24ⴱ .03 .71ⴱⴱ (.88) 5. Destructive voice 1.84 1.01 ⫺.17 .32 ⴱⴱ
⫺.17 .14 (.81)

6. Loyalty 3.57 1.17 .17 .21 .27 ⫺.06 ⫺.06 (.93) 6. OCB voice 4.22 1.67 .80ⴱⴱ .25 ⴱ
.80 ⴱⴱ
.07 ⫺.15 (.89)

Defensive voice model Destructive voice model


This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

M SD 1 2 3 4 5 6 M SD 1 2 3 4 5 6
This document is copyrighted by the American Psychological Association or one of its allied publishers.

1. Manipulation — — — 1. Manipulation — — —
2. Supportive voice 1.96 0.88 ⫺.29ⴱ (.86) 2. Supportive voice 1.66 0.93 ⫺.32ⴱⴱ (.91)
3. Constructive voice 1.51 0.83 .06 .24ⴱ (.92) 3. Constructive voice 1.37 0.62 .29ⴱ .33ⴱⴱ (.91)
4. Defensive voice 4.01 2.04 .90ⴱⴱ ⫺.18 .09 (.96) 4. Defensive voice 2.75 1.66 .62ⴱⴱ ⫺.15 .22 (.95)
5. Destructive voice 1.92 1.05 .15 .13 .08 .22 (.89) 5. Destructive voice 3.82 2.17 .93ⴱⴱ ⫺.29ⴱ .18 .60ⴱⴱ (.97)
6. Resistance to change 6.23 0.82 .23 ⫺.05 ⫺.24ⴱ .23ⴱ ⫺.02 (.68) 6. Poor sportsmanship 4.07 1.35 .44ⴱⴱ ⫺.43ⴱⴱ ⫺.03 .28ⴱ .45ⴱⴱ (.80)

Note. For each voice manipulation, 0 ⫽ low level of voice and 1 ⫽ high level of voice. Cronbach alpha reliabilities are reported in parentheses on the
diagonal.

p ⬍ .05. ⴱⴱ p ⬍ .01.

Appendix D
Variable Means, Standard Deviations, and Intercorrelations for Criterion-Related Validity Study (Stage 5)

Variable M SD 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Supportive voice (SV) 4.91 1.37 (.92)


2. Constructive voice (CV) 5.01 1.53 .55 (.95)
3. Defensive voice (DFV) 2.88 1.47 ⫺.38 ⫺.34 (.94)
4. Destructive voice (DSV) 3.14 1.69 ⫺.52 ⫺.16 .48 (.95)
5. Acquiescence response style (ARS) 0.95 0.37 .37 .28 ⫺.23 ⫺.27 (.48)
6. Disacquiescence response style (DRS) 0.55 0.35 ⫺.23 ⫺.27 .26 .28 ⫺.34 (.60)
7. Extreme response style (ERS) 0.21 0.16 .16 .01 .01 ⫺.03 .45 .46 (.57)
8. Midpoint response style (MRS) 0.21 0.15 ⫺.02 ⫺.02 ⫺.01 ⫺.07 ⫺.42 ⫺.46 ⫺.32 (.60)
9. Perceived organizational concern (POC) 5.41 1.35 .66 .37 ⫺.40 ⫺.50 .35 ⫺.27 .10 .01 (.82)
10. Perceived positive impact (PPI) 4.79 1.29 .57 .55 ⫺.48 ⫺.40 .36 ⫺.29 .07 ⫺.04 .56 (.95)
11. Perceived negative impact (PNI) 2.68 1.37 ⫺.45 ⫺.40 .65 .50 ⫺.26 .19 ⫺.05 .02 ⫺.54 ⫺.58 (.93)
12. Intent to rely on voicer’s opinion (Rely) 5.41 1.55 .53 .49 ⫺.45 ⫺.35 .42 ⫺.35 .09 ⫺.01 .56 .66 ⫺.59 (.90)
13. Evaluation of overall performance (Eval) 5.61 1.48 .58 .54 ⫺.46 ⫺.40 .39 ⫺.36 .01 .01 .60 .65 ⫺.60 .77 (.95)
Note. Cronbach alpha reliabilities are reported in parentheses on the diagonal. Correlations with an absolute value greater than or equal to .16 are
significant at the p ⬍ .05 level.

(Appendices continue)
112 MAYNES AND PODSAKOFF

Appendix E
Summary of Direct, Indirect, and Total Effects of the Final Model for Study 5 With and Without Accounting for
Common Method Variance

Intent to rely on voicer’s opinion Overall performance evaluations


Predictor Direct Indirect Total Direct Indirect Total

Model 1 (without response styles)


Supportive voice .32ⴱⴱ .32ⴱⴱ .31ⴱⴱ .31ⴱⴱ
Constructive voice .17ⴱⴱ .17ⴱⴱ .15ⴱⴱ .12ⴱⴱ .27ⴱⴱ
Defensive voice ⫺.22ⴱⴱ ⫺.22ⴱⴱ ⫺.17ⴱⴱ ⫺.17ⴱⴱ
Destructive voice ⫺.11ⴱⴱ ⫺.11ⴱⴱ ⫺.12ⴱⴱ ⫺.12ⴱⴱ
Organizational concern .26ⴱⴱ .26ⴱⴱ .33ⴱⴱ .33ⴱⴱ
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Positive impact .40ⴱⴱ .40ⴱⴱ .25ⴱⴱ .25ⴱⴱ


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Negative impact ⫺.24ⴱⴱ ⫺.24ⴱⴱ ⫺.23ⴱⴱ ⫺.23ⴱⴱ


Model R2 .60 .61
Model 2 (controlling for response styles)
Supportive voice .18ⴱⴱ .18ⴱⴱ .21ⴱⴱ .21ⴱⴱ
Constructive voice .13ⴱⴱ .13ⴱⴱ .13ⴱ .10ⴱⴱ .23ⴱⴱ
Defensive voice ⫺.19ⴱⴱ ⫺.19ⴱⴱ ⫺.15ⴱⴱ ⫺.15ⴱⴱ
Destructive voice ⫺.09ⴱⴱ ⫺.09ⴱⴱ ⫺.10ⴱⴱ ⫺.10ⴱⴱ
Organizational concern .16 .16 .28ⴱⴱ .28ⴱⴱ
Positive impact .32ⴱⴱ .32ⴱⴱ .18ⴱ .18ⴱ
Negative impact .25ⴱⴱ .25ⴱⴱ ⫺.22ⴱⴱ ⫺.22ⴱⴱ
Model R2 .65 .65

Model fit statistics, chi-square values, and degrees of freedom


Sample size df ␹2 CFI SRMR IFI
SEM without response styles 222 470 864.47 .98 .06 .98
SEM controlling for response styles 222 878 2,382.44 .95 .07 .95
Note. df ⫽ degrees of freedom; CFI ⫽ comparative fit index; SRMR ⫽ standardized root mean residual; IFI ⫽ incremental fit index; SEM ⫽ structural
equation modeling.

p ⬍ .05. ⴱⴱ p ⬍ .01.

Received February 22, 2012


Revision received June 25, 2013
Accepted July 31, 2013 䡲

You might also like