The power of powerless speech: The effects of - Alison Fragale

19 Pages·2006·0.47 MB·English

Organizational Behavior and Human Decision Processes 101 (2006) 243–261
www.elsevier.com/locate/obhdp
V
The power of powerless speech: The e ects of speech style
(cid:2)
and task interdependence on status conferral
Alison R. Fragale¤
Department of Organizational Behavior and Strategy, Kenan-Flagler Business School, University of North Carolina – Chapel Hill, CB #3490,
Chapel Hill, NC 27599-3490, USA
Received 15 March 2005
Available online 6 March 2006
Abstract
Two studies examine the eVects of speech styles and task interdependence on status conferral judgments. In both studies, partici-
pants were exposed to an individual who used either a powerful or powerless speech style in a low or high task interdependence
group, and made judgments about the amount of status to confer to the individual. When task interdependence was low, participants
conferred more status to powerful speakers, whereas when interdependence was high, participants conferred more status to power-
less speakers. Furthermore, Study 2 demonstrated that speech styles inXuenced trait inferences about the speaker (agency and com-
munality), but these traits were weighted diVerently in status conferral judgments across groups. These Wndings provide insight into
both the relationship between observed behaviors and status positions and the decision process underlying status conferral judg-
ments.
© 2006 Elsevier Inc. All rights reserved.
Keywords: Status conferral; Speech Styles; Powerful speech; Powerless speech; Task interdependence; Task performance; Agency; Communality
Language is power, in ways more literal than most peo- yes. The popular press is replete with books oVering stra-
ple think. When we speak, we exercise the power of lan- tegic prescriptions for how to use language to climb the
guage to transform reality. corporate ladder. Interested readers can learn “power
words” (GriYn, 1998), master the art of “power talking”
– Julia Penelope, Speaking Freely
(Walther, 2000), and even discover how to “sound like a
Is there a “language of success” (Mindell, 2001)? A leader” (Toogood, 1995). The central tenet unifying
glance at the business section of one’s local bookstore these works is that one’s communication style aVects
suggests that the answer to this question is a deWnitive subsequent status attainment at work and in life: Indi-
viduals who speak assertively are more likely to get
hired, be promoted, and command respect from others
(cid:2) This research was supported, in part, through the State Farm Doc- than individuals who speak in a tentative, uncertain
toral Dissertation in Business Award and the University of North Car-
manner.
olina’s Claude and Eleanor George Fund for Teaching and Learning. I
This claim has received substantial empirical support.
am indebted to Margaret Neale and Larissa Tiedens for their advice
and support throughout this project, and to JeV Edwards for numer- Psychologists and organizational scientists alike have
ous invaluable conversations about this work. I am also grateful to Jer- documented a consistent relationship between speech
ald Greenberg, action editor, three anonymous reviewers, and Peter styles and subsequent status attainment. Even holding
Kim, Jennifer Overbeck, and Maia Young for their detailed and in-
message content constant, stylistic features of a message,
sightful comments on earlier versions of this manuscript.
such as hesitations (e.g., “well,” “um”), tag questions
* Fax: [phone].
E-mail address: [email protected] (e.g., “That’s interesting, don’t you think?”), hedges
0749-5978/$ - see front matter © 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.obhdp.2006.01.004
244 A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261
(e.g.,“kinda,” “sort of”), disclaimers (e.g., “This may be way, Pizzamiglio, & Mount, 1996; Fiske, Cuddy, Glick, &
a bad idea, butƒ”), intensiWers (e.g., “really,” “very”) Xu, 2002; Fiske, Xu, Cuddy, & Glick, 1999) and suggests
and formal addresses (e.g., “yes, sir”), inXuence how a that powerful speech should lead to greater status confer-
speaker is evaluated. Individuals who speak assertively, ral than powerless speech, independent of the context in
by avoiding these tentative speech markers, are judged question. The other perspective, which I term the contin-
by observers as more likely to be hired, promoted, and gent criteria perspective, is based on theories of status
supported by superiors (Gallois, Callan, & Palmer, 1992; characteristics (e.g., Berger, Cohen, & Zelditch, 1972; Ber-
Parton, Siltanen, Hosman, & Langenderfer, 2002; Wiley ger, Conner, & Fisek, 1974) and implies that both power-
& Eskilson, 1985; see also Ng & Bradac, 1993 for a ful and powerless speech can be status enhancing,
review), and are more inXuential (Erickson, Lind, John- depending on the speciWc organizational context. The
son, & O’Barr, 1978) than individuals who include these studies presented in this paper are designed to test the
speech characteristics. These relationships are consid- validity of these two theoretical perspectives.
ered so robust that speech styles have been named
according to their consequences: “Powerless” speech is a
speech style deWned by the presence of the abovemen- How do speech styles aVect status?
tioned linguistic markers, whereas “powerful” speech is
deWned by the absence of these markers (Erickson etal., It is surprising that a behavior as subtle and poten-
1978; see also LakoV, 1975). tially meaningless as one’s speech style should aVect
Although this relationship between powerful speech something as consequential as one’s ability to get pro-
and subsequent status attainment seems valid on its face, moted or be respected. Yet, individuals often make dis-
these Wndings raise an important practical and theoreti- positional inferences about an actor on the basis
cal question that has not been adequately addressed in observable behaviors (e.g., Ross, 1977), and these dispo-
prior research: Is powerless speech truly as powerless as sitional attributions, in turn, aVect how the actor is eval-
it seems? Or, under certain circumstances, can powerless uated. A host of behavioral cues—speech styles (e.g.,
speech actually be more eVective for obtaining promo- Erickson etal., 1978; Parton etal., 2002; Wiley & Eskil-
tions and prestige than powerful speech? The Wrst goal son, 1985), speech rates (e.g., Brown, Strong, & Rencher,
of this paper is to answer this question. SpeciWcally, I 1973), vocal tones (e.g., Ridgeway, 1987), patterns of eye
examine whether one particular group-level factor, the contact (e.g., Washburn & Hakel, 1973), and emotional
level of interdependence in a task group, aVects the rela- expressions (Tiedens, 2001)—have been shown to inXu-
tionship between speech styles and status attainment. In ence status positions because they inXuence perceptions
doing so, I adopt a deWnition of status used in prior of the actor’s personality traits.
research: Status is the extent to which an individual in a Work in the area of person perception has demon-
group is seen as prominent, respected, and inXuential by strated that individuals organize their perceptions of oth-
other group members (Anderson, John, Keltner, & ers around two fundamental trait dimensions (Asch, 1946;
Kring, 2001). This deWnition implies that status positions Bakan, 1966; Carson, 1969; Fiske etal., 2002; Kiesler,
are socially determined; one can only possess as much 1983; Leary, 1957; Pruitt & Rubin, 1986; Wiggins, 1979).
status as others are willing to grant. In this sense, status Bakan (1966) referred to these dimensions as agency, or
is conferred to one individual by another individual (or characteristics associated with self-assertion and mastery
group of individuals). The studies in this paper investi- of one’s environment, such as ambition, dominance, and
gate individuals’ decisions about how much status to independence, and communality, or characteristics associ-
confer to an actor in an organizational context. ated with selXessness and nurturance, such as warmth, sin-
Elucidating the eVects of powerless speech on status cerity, and tolerance. Individuals who use powerful speech
conferral has obvious prescriptive implications for how are rated by observers as more competent, intelligent, and
individuals should communicate to gain status. However, strong (i.e., agentic) than individuals who use powerless
investigating the relationship between speech styles and speech (Carli, 1990; Erickson etal., 1978; Parton etal.,
status conferral has broader theoretical implications as 2002; Siegler & Siegler, 1976; see also Ng & Bradac, 1993
well. The second goal of this paper is to gain some insight for a review). However, there is some evidence to suggest
into the process underlying status conferral judgments. that individuals who use a powerless speech style are
The status beneWts of using powerless speech, if any, viewed as nicer, more likeable and good-natured (i.e.,
depend on how individuals make their status conferral communal) than those who use a powerful speech style
decisions. Two theoretical perspectives oVer conXicting (Lee, 1999; Parton etal., 2002).
views on how status conferral decisions are made, and Even though agency and communality may both be
consequently lead to diVerent predictions about whether viewed as socially desirable traits for individuals to pos-
powerless speech should ever be status enhancing. One sess, evidence to date suggests that only the agency
perspective, which I refer to as the Wxed criteria perspec- dimension is used in making status conferral judgments.
tive, is derived from research on status schemas (e.g., Con- That is, when deciding whether to hire, promote, or
A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261 245
support an actor, individuals use their impressions of the be highly agentic, but not particularly communal,
actor’s intelligence, ambition, and dominance to inform whereas members of low status groups (e.g., housewives,
their judgments, but not their evaluations of the actor’s the elderly) are perceived to be highly communal, but
warmth, sincerity, or agreeableness. For example, indi- not particularly agentic (Conway etal., 1996; Fiske etal.,
viduals who express anger are viewed as more agentic, 1999, 2002).1 Evidence suggests that these status schemas
and consequently are conferred more status, than indi- are widely held and applied broadly, regardless of the
viduals who express sadness, even though sadness particular group or individual in question. For example,
expressers are rated as more communal (Tiedens, 2001). observers apply these schemas to individuals based on
Consistent with these Wndings, Wiley and Eskilson the individuals’ occupations (e.g., doctors vs. nurses;
(1985) found that the Wve traits individuals believe to be Conway etal., 1996), their social categories (e.g., Asians,
most important in determining an individual’s qualiWca- Blacks, welfare recipients, feminists; Fiske etal., 1999,
tions for a leadership position were all characteristics of 2002), and even apply these schemas when making judg-
agency: responsible, intelligent, qualiWed, hardworking, ments about members of Wctional groups (Conway etal.,
and organized. Furthermore, the importance of agency 1996).
as a basis for status conferral judgments has been con- These Wndings suggest that individuals may base sta-
veyed through the names of the trait dimensions them- tus conferral judgments on the extent to which actors
selves: Theories of person perception often refer to the conform to these well-developed high and low status
agency and communality dimensions as status and love, schemas. Individuals perceive high status persons to be
respectively (e.g., Carson, 1969; Kiesler, 1983; Leary, highly agentic and view agency as a high status charac-
1957; Wiggins, 1979). teristic, so when they observe a person that conveys
However, it is possible that attributions of commu- agentic traits they conclude that this person should be
nality may play a role in individuals’ status conferral high status (i.e., they confer status to that individual).
decisions as well. The Wxed and contingent criteria per- Similarly, individuals perceive low status others to be
spectives diVer in their implications about process highly communal and view communality as a lesser val-
underlying status conferral judgments, and hence lead to ued, low-status characteristic, so when they observe an
diVerent predictions about the relative importance of individual that conveys communal traits, they conclude
agency and communality in individuals’ status conferral that the individual should be low status, and confer less
decisions. The Wxed criteria perspective (e.g., Conway status to that individual.
etal., 1996; Fiske etal., 1999, 2002) suggests that agency If individuals rely on these situationally invariant
should be weighted more heavily than communality in schemas when making status conferral decisions, this
status conferral judgments across contexts. In other implies the criteria used in making status judgments
words, the criteria on which status conferral judgments should remain Wxed across situations: Observers should
are based (i.e., agency) should remain Wxed across base status conferral decisions on perceptions of agency,
groups. In contrast, the contingent criteria perspective regardless of the organizational, cultural, or social con-
(e.g., Berger etal., 1972, 1974) implies that status confer- text. Extending this logic to speech styles, powerful
ral decisions should be contingent on the group context: speech should always result in greater status conferred
In some groups or organizations, agency may be to the speaker than powerless speech, since the former
weighted more heavily than communality in status con- conveys a higher level of agency than the latter.
ferral judgments, but the opposite may be true in other
groups. Which of these two perspectives most ade- Contingent criteria perspective
quately accounts for individuals’ status conferral judg- Status characteristics theory (also referred to as
ments should aVect the consequences of powerless expectation states theory, Berger etal., 1972, 1974, or
speech. The status conferral beneWts of powerless speech
depend on whether, and the extent to which, attributions
1 Conway etal. (1996) suggest that the content of these high and low
of communality are weighted in status conferral judg-
status schemas may develop as a result of the correspondence bias
ments. (Gilbert & Malone, 1995), whereby observers make dispositional infer-
ences about an actor based on the actor’s observable behaviors. For
Potential processes underlying status conferral decisions example, high status individuals may talk more and dominate conver-
sations more than low status individuals, not necessarily because of in-
herent diVerences between the two groups, but simply because high
Fixed criteria perspective
status individuals are asked for advice, knowledge, and opinions more
Social psychologists have investigated the schemas, or frequently than low status individuals. Yet, observers may not recog-
mental representations, that individuals hold about nize these situational diVerences, and conclude that high status individ-
members of high and low status groups. Individuals have uals actually are more aggressive, outgoing, intelligent, etc. (i.e., highly
agentic) than low status individuals, but also more rude and argumen-
been shown to hold distinct schemas about high and low
tative (i.e., not very communal). Similarly, observers may interpret the
status others: Members of high status groups (e.g., busi- reticence of low status individuals to reXect politeness and respectful-
ness professionals, rich people) are judged by others to ness (i.e., high communality), but also submissiveness (i.e., low agency).
246 A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261
status generalization, Webster & Driskell, 1978) suggests coordinate, or interact with others to complete their
that status is conferred to an actor based on others’ assigned tasks (Thompson, 1967; Wageman, 1995). The
assessments of the actor’s expected performance on the level of task interdependence in a group can be aVected
task at hand: Individuals are conferred high status when by cultural values of the group, which result in norms
they are expected to perform well on the given task, and about how work should be completed (e.g., Shea &
conferred lower status when they are expected to per- Guzzo, 1989; Wageman, 1995), or by features of the task
form poorly. Proponents of this perspective have actu- itself, which either require or prohibit collective action
ally used this argument to explain why status conferral (Thompson, 1967). Although all task groups or organi-
judgments should be relatively similar across situations zations, by deWnition, necessitate some degree of interde-
(e.g., Ridgeway, 1987; see also Webster & Driskell, 1978): pendence among members, there is substantial variation
Characteristics of agency, such as assertiveness and con- among groups in the level of interdependence required
Wdence, facilitate successful performance in most task for task completion (Wageman, 1995). In general, the
groups, so behaviors that convey a high level of agency, greater the level of task interdependence in a group, the
such as powerful speech, should result in status confer- greater the need for, and expectation of, collective eVort
ral, regardless of the speciWc group in question. and coordination among group members (March &
However, by deWning status in terms of expected per- Simon, 1958; Thompson, 1967).
formance, this theoretical framework suggests that status The level of task interdependence in a group may
conferral decisions may actually be more context-depen- aVect the criteria that individuals use when evaluating
dent than the original researchers assumed. If expectations group members’ performance. When interdependence is
of performance are what individuals rely on when making low, and individuals work primarily independently,
their status conferral decisions, then any trait or attribute characteristics of agency may be viewed as particularly
that is seen as facilitating performance should result in important predictors of a group member’s performance.
status conferral. To the extent that the traits or attributes High agency is associated with the ability to think inde-
associated with successful performance change across pendently, having conWdence in one’s own views, and a
groups or situations, this implies that the criteria for status motivation for achievement, which are all characteristics
conferral decisions should change as well. In other words, that should facilitate an individual’s performance when
behaviors that convey high agency, such as powerful working alone. In comparison, characteristics of com-
speech, should lead to status conferral insituations where munality, which indicate how an individual relates to
agency is viewed as a critical determinant of one’s task and interacts with others, may be viewed as less critical
performance, but behaviors such as powerless speech, that for completing work in low interdependence settings,
convey high communality, should lead to status conferral since interaction with others, by deWnition, is relatively
insituations where communality is indicative of successful low. Thus, in low interdependence groups, individuals
performance. may place greater weight on assessments of a group
member’s agency than on communality when judging
Distinguishing between these perspectives the group member’s expected performance.
Current empirical evidence is not suYcient to deter- In contrast, characteristics of communality may be
mine which of these two perspectives best describes the viewed as more predictive of group members’ perfor-
process underlying individuals’ status conferral deci- mance than characteristics of agency when task interde-
sions, since both the Wxed and contingent perspectives pendence is high. High communality is associated with
can be used to explain the previously observed positive high levels of concern for others, a willingness to be
relationship between perceptions of agency and status cooperative, and a desire for aYliation, which are all
conferral. To distinguish between these perspectives, it is characteristics that should facilitate an individual’s per-
necessary to examine status conferral decisions across formance when working collaboratively. At the same
group contexts. Although both theories predict that time, characteristics of agency may be viewed as less pre-
assessments of agency may serve as the basis for status dictive of performance in highly interdependent groups,
conferral judgments, they diVer in their predictions since the collaborative nature of these groups implies
about whether this relationship generalizes across that members can pool their collective talents and rely
groups. In this paper, I focus on one group-level factor, more on the abilities of the group as a whole rather than
task interdependence, which may provide insight into on the abilities of any one member: An individual that
which of these processes most adequately accounts for exhibits relatively low levels of agentic traits (such as
status conferral judgments. dominance, independence, and competitiveness) may
still be viewed as a star performer in a highly interdepen-
Task interdependence dent group if this individual demonstrates an ability to
relate to and work with others. Furthermore, agency
The level of task interdependence in a group is the may sometimes be viewed as a liability in high interde-
extent to which group members need to collaborate, pendence groups, since these individuals may be
A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261 247
perceived as preferring and pursuing individual accom- in an initial task with an individual that used either a pow-
plishments (e.g., personal recognition and achievements) erful or powerless speech style and then made judgments
at the expense of group accomplishments (e.g., group about the amount of status to confer to their partner in a
performance; Ridgeway, 1982). Together, these forces subsequent task group, which manipulated the level of
imply that perceptions of communality may inXuence task interdependence required. Status conferral was
expectations of successful performance to a greater assessed in two ways in this study. First, participants made
degree than perceptions of agency in high interdepen- evaluations of how much status and respect the individual
dence groups. should have in the task group. Second, prior research had
The Wxed and contingent criteria perspectives make demonstrated that status is often conferred by endowing
diVerent predictions about whether the level of task individuals with formal titles that signify their high status
interdependence in a group should aVect status conferral position (e.g., leader, president, and captain; Tiedens,
decisions. On one hand, a reliance on Wxed criteria 2001). Thus, status conferral was also assessed through
implies that diVerences in task interdependence should participants’ willingness to recommend the individual for
not aVect status conferral judgments. This leads to the the formal position of “group leader.”
following hypotheses derived from the Wxed criteria per-
spective: Method
Hypothesis1. In both high and low interdependence
Participants
groups, speakers using a powerful speech style will be
One-hundred-twenty-four individuals (51 males, 64
conferred more status than speakers using a powerless
females, and 9 unidentiWed) aYliated with a west-coast
speech style.
university participated in this study in exchange for a
Hypothesis2. In both high and low interdependence payment of $15. Participants were recruited from an
groups, individuals will weight assessments of an actor’s electronic mailing list at the university that advertises
agency more than assessments of the actor’s communal- behavioral studies to university members that have
ity when making status conferral judgments. expressed an interest in participating in them.
On the other hand, the contingent criteria perspective
Materials and procedure
predicts that status conferral decisions should change as
Between 6 and 12 individuals participated in each
a function of the level of interdependence in a group, to
experimental session. As participants arrived, they were
the extent that task interdependence aVects the deWni-
seated in one of two laboratory rooms, each at individ-
tion of successful performance. Thus, the contingent cri-
ual study carrels with computer terminals. The experi-
teria perspective leads to the following predictions:
menter informed participants that they would be
Hypothesis3a. In low interdependence groups, speakers participating in two decision-making tasks, each with
using a powerful speech style will be conferred more sta- other participants in the experimental session. The Wrst
tus than speakers using a powerless speech style. exercise was described as a computer-mediated problem-
solving task in which participants would interact with a
Hypothesis3b. In high interdependence groups, speakers
partner via a computer. In actuality, participants did not
using a powerless speech style will be conferred more
interact with a partner; rather, they “conversed” with a
status than speakers using a powerful speech style.
scripted computer program. This exercise was designed
Hypothesis4a. In low interdependence groups, individu- to manipulate the speech style of the “partner.” In the
als will weight assessments of an actor’s agency more powerless condition, the partner used hedges, hesita-
than assessments of the actor’s communality when mak- tions, disclaimers, formal addresses, and tag questions,
ing status conferral judgments. whereas in the powerful condition, the partner did not
use these tentative speech patterns. After completing the
Hypothesis4b. In high interdependence groups, individ-
Wrst task, participants were given instructions for a sec-
uals will weight assessments of an actor’s communality
ond task, described as a 4-person group task, in which
more than assessments of the actor’s agency when mak-
task interdependence was manipulated. In actuality,
ing status conferral judgments.
there was no second team task; this aspect of the cover
story was included to create a context for the status con-
ferral measures of interest in this study, participants’
Study 1 evaluations of their “partner” from the Wrst task.
The objective of Study 1 was to investigate whether the Experimental design. The design of this study was a 2
level of task interdependence in a group would moderate (Task 1 Partner Speech Style: Powerful vs. Powerless)£2
the relationship between speech styles and status confer- (Task 2 Interdependence: High vs. Low) between-subjects
ral. In this experimental paradigm, participants interacted factorial design.
248 A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261
Task 1: The desert survival problem. After completing the partner’s rational regarding the Xashlight, the pro-
the general instructions, the experimenter provided spe- cess of sending and receiving rationales about item rank-
ciWc instructions for the Wrst problem-solving exercise. ings was then repeated for the remaining 11 items.
This exercise was a computer-mediated version of the Although the content of the partner responses was
Desert Survival Problem (DSP; LaVerty & Eady, 1974) held constant across conditions, the style of the
in which participants rank ordered 12 items (e.g., a mir- responses diVered. For example, in the powerful speech
ror, a map, two raincoats) in terms of importance for condition, one statement by the partner read: “The Xash-
survival in a desert.2 First, participants individually light needs to be rated higher. It is the only reliable night
made initial rankings of the 12 items. Participants were signaling device; also, the reXector and lens could be
then asked to engage in a computer-mediated discussion used to start a Wre, which is another way to signal for
about their initial item rankings with another individual, help. Put it higher.” In the powerless speech condition,
described as another participant seated in the other com- the same statement suggesting that the Xashlight should
puter lab. The experimenter informed participants that be rated higher read: “Do you think the Xashlight should
there was a correct answer to the DSP and that they maybe be rated higher? It may be a pretty reliable night
would have an opportunity to revise their initial rank- signaling device. Also, maybe the reXector and lens could
ings after the computer-mediated discussion, so the part- be used to start a Wre, which could possibly be another
ner discussion task would be an opportunity to gather way to signal for help”.3
information that could help participants improve upon After participants sent and received messages about
their initial answers. all 12 items, they were given an opportunity to revise
A web-based computer program (adapted from their initial rankings of these items and then completed a
Shechtman, 2002) was used to facilitate this computer- “Communication Style Questionnaire.” This question-
mediated discussion. The program was designed to cre- naire asked participants to rate aspects of their partner’s
ate the impression that the participant was interacting communication style, including their assertiveness and
with an actual partner. The Wrst screen of the program friendliness. These two items served as manipulation
prompted participants to enter their initial rankings of checks for the speech style manipulation.
the 12 desert survival items and, upon completion, to
click a “send” button to transmit their rankings to their Task 2: 4-person group task. After completing this ques-
partner. After a brief delay, designed to convey the tionnaire, participants were given general instructions
impression that the discussion partner was typing his or about the second task. Participants were told that they
her rankings, the interface displayed both the partici- would be formed into 4-person groups, and that each
pant’s rankings and the partner’s rankings. In actuality, group would have the task of calculating merit increases
the partner’s rankings were a systematic transformation for employees in an organization (referred to as the
of the participant’s rankings such that (a) the participant “performance appraisal task”). This task was selected
and partner disagreed on the rankings for almost all of because it has been used in prior studies of task interde-
the items and (b) the total magnitude of the disagree- pendence (Saavedra, Earley, & Van Dyne, 1993). The
ment was constant across all participants. description of the task manipulated the level of task
The program then prompted participants to type a interdependence. In the low task interdependence condi-
short comment explaining their rationale for their rank- tion, participants read the following instructions (modi-
ing of the Xashlight and to press “send” when they were Wcations for the high interdependence condition appear
ready to transmit the statement to the partner. The part- in parentheses):
ner then responded with an analogous rational about the
In the performance appraisal task, 4-person groups will
Xashlight and a recommendation about how the partici-
rate employees of an organization on several factors and
pant should revise his or her initial ranking of this item.
then recommend merit increases for each employee. To
These computer-generated responses were derived from
reach a merit recommendation for an employee, there
a table based on whether the partner ranking for that
are a series of steps that must be completed. To complete
item was higher or lower than the participant’s, as well
as the partner’s speech condition (powerful or power-
less). To create the impression that a partner was actu-
3 In the studies presented in this paper, speech styles were manipulat-
ally typing responses, all responses from the partner
ed through written, rather than oral, communication. Erickson etal.
were delayed by a half second per word. After receiving
(1978) manipulated presentation style (written vs. oral) in their re-
search on the consequences of powerful and powerless speech. These
researchers found that observers’ inferences about a speaker depended
2 The original Desert Survival Problem required participants to rank on the speech style (powerful vs. powerless), but not on the mode of
15 items. Three items from the original 15-item list, a red and white presentation. Consequently, subsequent research has manipulated
parachute, a pair of sunglasses, and a loaded .45 caliber pistol, were not speech style through either written or oral presentation. For example,
included in this version of the exercise. This modiWcation was made to Parton etal. (2002) used audiotapes, whereas Wiley and Eskilson
shorten the amount of time participants spent on Task 1. (1985) and Hosman (1989) both used written transcripts.
A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261 249
these steps and successfully accomplish the performance Table 1
appraisal task, the 4 members of the rating team must Means, standard deviations, and correlations of dependent measures in
Studies 1 and 2
work independently (collectively). Each member of the
team will be given their own set of employees to rate (a Mean SD 1 2 3
speciWc role to perform), so that each group member will Study 1
complete all (some, but not all) of the steps required to 1. Status conferral 4.42 1.20 —
2. Expected performance 5.34 1.28 .52¤¤¤ —
produce a complete merit recommendation. Given this
division of labor, members of the performance appraisal Study 2
team will be highly independent of (dependent on) each 1. Status conferral 4.18 1.17 —
2. Perceived agency 4.79 .93 .43¤¤¤ —
other. All members need to be able to work on their own
3. Perceived communality 4.35 .92 .34¤¤ ¡.32¤ —
(together) if the task is to be completed successfully.
Note. Study 1 ND111, Study 2 ND56.
To reinforce the manipulations, and to check that ¤ p<.05.
participants attended to them, two questions were pre- ¤¤ p<.01.
¤¤¤ p<.001.
sented below the task instructions: Participants indi-
cated the extent to which group members needed to
collaborate and coordinate with others to complete their participants were excluded because they suspected that
task and the extent to which group members would be their Task 1 conversation partner was actually a com-
required to complete the task independently. Both ques- puter program.4 Data from one additional participant
tions were presented on 7-point scales (1Dnot at all; were excluded because the participant failed to follow
7Da great deal).
instructions. Thus, all reported analyses are based on the
remaining 111 participants.
Partner evaluation questionnaire. The main dependent
measure of interest in this study was participants’ evalu- Manipulation checks
ations of the leadership potential of their partner from Speech style. Participants in the powerful speech condi-
the Wrst task. After receiving the general instructions for tion (coded as 1) rated their partner’s communication
the second task, participants were informed that each of style as signiWcantly more assertive (MD6.3) than par-
the 4-person task groups would have a designated ticipants in the powerless speech condition (coded as ¡1;
leader, and to choose those leaders, the experimenter MD4.6), bD.89, t(109)D7.32, p<.001. Participants in
wanted to solicit each participant’s opinion about the the powerless speech condition rated their partner’s
partner they worked with in the Wrst task. Participants communication style as signiWcantly more friendly
answered three questions about how much status to con- (MD4.8) than participants in the powerful speech con-
fer to their partner in the 4-person group: Participants dition (MD3.1), bD¡.84, t(109)D¡6.50, p<.001.
indicated the extent to which they would recommend
their partner to be the leader in the 4-person group, how Task interdependence. Participants in the low task inter-
much status they thought their partner should have in dependence condition (coded as ¡1) indicated that their
the task group, and how much respect their partner group members would need to work more independently
should receive from other members of the task group. to complete their assigned task (MD6.4) than did partic-
These questions were assessed on 7-point scales (1Dnot ipants in the high task interdependence condition (coded
at all; 7Da great deal) and were averaged to create a as 1; MD3.5), bD¡1.4, t(109)D¡9.73, p<.001. Con-
composite measure of status conferral ((cid:2)D.76). Addi- versely, participants in the high task interdependence
tionally, participants indicated the extent to which they condition indicated that their 4-person group would
thought their partner had the ability to perform well in require a higher level of collaboration and coordination
the group (assessed on a 7-point scale anchored by to complete their assigned task (MD6.3) than did partic-
1Dnot at all and 7Dvery much). This question served as ipants in the low task interdependence condition
a measure of the partner’s expected performance in the (MD2.5), bD1.9, t(109)D13.76, p<.001.
Task 2 group.
After completing this questionnaire, the experimenter
informed participants that there was no second task and
the experiment was Wnished. Participants were fully 4 These suspicious participants were roughly evenly distributed
across all conditions of the study. To determine whether participants
debriefed and paid for their participation.
suspected they were not interacting with an actual partner, I examined
the transcripts of their desert survival discussion (which were recorded
Results by the web-based discussion program) and their questionnaire re-
sponses (which provided a space for participants to justify their rating
of their partner’s leadership potential). Participants were coded as sus-
Descriptive statistics and correlations for the depen-
picious if they noted, in either or both of these places, that they thought
dent measures are reported in Table 1. The data from 12 their partner was not real.
250 A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261
Status conferral judgments
I regressed status conferral on speech style, task interde-
pendence, and the interaction between the two. The results
of this analysis can be seen in Table 2, Regression 2. There
was a marginally signiWcant main eVect of task interdepen-
dence (M D4.6, M D4.2, bD¡.20, t(107)D¡1.85,
low high
pD.067). However, this main eVect was qualiWed by a sig-
niWcant interaction between the partner’s Task 1 speech
style and the level of task interdependence in the Task 2
group (bD¡.28, t(107)D¡2.50, pD.014). As may be seen
in Fig.1, when task interdependence was low, participants
whose partner used a powerful speech style thought that
their partner should be conferred more status in the Task 2 Fig.1. Study 1: Status conferral as a function of partner’s Task 1
group than did participants whose partner used a power- speech style and Task 2 interdependence.
less speech style, t(107)D1.81, pD.073. However, when
task interdependence was high, participants that interacted
with a powerless speaker in Task 1 thought their partner
should be conferred more status in the second task group
than did participants that interacted with a powerful
speaker, t(107)D1.72, pD.089. Thus, Hypotheses 3a and
3b, predicting an interaction between speech style and
interdependence, were supported, but Hypothesis 1, pre-
dicting a main eVect for powerful speech, was not.
Performance expectations
The above Wndings provide support for the contingent
criteria perspective, which suggests that the criteria for sta-
tus judgments in a group should change as the criteria for
Fig.2. Study 1: Performance expectations as a function of partner’s
successful performance changes across groups. To deter-
Task 1 speech style and Task 2 interdependence.
mine whether the partner’s speech style in Task 1 aVected
participants’ expectations of the partner’s performance in
Task 2, I regressed expected performance on speech style, dependence (bD¡.32, t(107)D¡2.65, pD.009). As may be
task interdependence, and the interaction between the two seen in Fig.2, when Task 2 was described as a low interde-
(see Table 2, Regression 6). This analysis revealed only a pendence task, participants that interacted with a powerful
signiWcant interaction between speech style and task inter- speaker in Task 1 thought that their partner would per-
Table 2
Study 1: OLS regressions to predict status conferral and expected performance
Measures Status conferral Expected performance
(1) (2) (3) (4) (5) (6)
Constant 4.42¤¤¤ 4.41¤¤¤ 4.42¤¤¤ 4.42¤¤¤ 5.34¤¤¤ 5.33¤¤¤
Speech style .01 .01 .01 .02 ¡.01 ¡.01
Task interdependence ¡.219 ¡.209 ¡.179 ¡.179 ¡.09 ¡.08
Speech£interdependence ¡.28¤ ¡.13 ¡.13 ¡.32¤¤
Expected performance .46¤¤¤ .46¤¤¤
Performance£interdependence .02
R2 .03 .08 .31 .31 .01 .07
(cid:2)R2 .05¤ .23¤¤¤ .00 .06¤¤
Note. ND111. For speech style, powerful speech was coded 1 and powerless speech was coded ¡1. For task interdependence, low interdependence
was coded ¡1 and high interdependence was coded 1. Continuous independent measures were mean-centered before conducting analyses, and these
mean-centered variables were used in computing the interaction terms between continuous and categorical predictors. Regression models are pre-
sented vertically; numbers across the top of the table in parentheses signify diVerent regression models, and the labels across the top of the table sig-
nify the dependent measure for the regressions in the columns below. Entries in the columns represent unstandardized regression coeYcients.
9 p<.10.
¤ p<.05.
¤¤ p<.01.
¤¤¤ p<.001.
A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261 251
Low TaskInterdependence
Expected
Performance
.31+ .44***
Speech .15 Status Conferral
Style
High TaskInterdependence
Expected
Performance
-.33* .48***
Speech -.11 Status Conferral
Style
Fig.3. Study 1: Simple paths at each level of the moderator variable (task interdependence). Note. ND111. For speech style, powerful speech was
coded 1 and powerless speech was coded ¡1. Figure entries are unstandardized path coeYcients. Paths with underlined coeYcients are signiWcantly
diVerent (p<.05) across levels of the moderator variable. +p<.10, *p<.05, ***p<.001.
form better in the Task 2 group than did participants that Expected performance signiWcantly predicted status con-
interacted with a powerless speaker, t(107)D1.83, pD.069. ferral (bD.46, t(105)D5.78, p<.001) and, comparing
However, when Task 2 was described as a high interdepen- this model to Regression 2, the interaction between
dence task, participants that interacted with a powerless speech style and task interdependence was no longer sig-
speaker thought their partner would perform better on the niWcant when expected performance was added into the
second task than did participants that interacted with a model (bD¡.13, t(105)D1.31, n.s.). To interpret these
powerful speaker, t(107)D1.92, pD.058. eVects, I then conducted a moderated path analysis (see
The contingent criteria perspective also implies that Fig.3) to determine the speciWc form of mediation in
the eVects of an individual’s speech style on status con- question by examining simple paths (or slopes) at each
ferral judgments should be mediated by expectations of level of the moderator variable (Edwards & Lambert,
the individual’s performance in the group: an individ- 2004). The Wrst thing to note is that speech styles exerted
ual’s speech style provides a cue to how the individual signiWcant, yet opposite, eVects on expected performance
will perform in a group, the these performance expecta- across levels of the moderator variable (task interdepen-
tions, in turn, inXuence how much status the individual dence). When task interdependence was low, powerful
should be conferred. Following the procedures recom- speech was associated with higher performance expecta-
mended by Baron and Kenny (1986) and Edwards and tions, whereas when interdependence was high, powerful
Lambert (2004), I regressed status conferral on the inde- speech was associated with lower performance expecta-
pendent variable (speech style), the moderator (task tions. Furthermore, the path from the mediator
interdependence), the interaction between the indepen- (expected performance) to the dependent variable (sta-
dent variable and the moderator, the proposed mediator tus conferral) was statistically signiWcant in both high
(expected performance), and the interaction of the medi- and low interdependence conditions, and the magnitude
ator and the moderator (see Table 2, Regression 4).5 of this path did not diVer across levels of the moderator.
To test for mediation, I then used a bootstrap proce-
dure to test the magnitude of the indirect eVect (the eVect
5 To determine whether the mediator (expected performance) and
of speech style on status conferral through the mediator,
the dependent measure (status conferral) were distinct constructs, I
conducted a two-factor conWrmatory factor analysis in which the fac- expected performance) at each level of the moderator
tors were performance and status conferral, which had one and three variable. Mediation is indicated when the size of an indi-
items, respectively. For performance, the item loading and measure- rect eVect diVers signiWcantly from zero (Shrout & Bol-
ment error variance of the single item were Wxed to reXect the assumed
ger, 2002). I implemented the bootstrap by drawing 1000
reliability of the item (Hayduk, 1987), and for status conferral, the item
random samples with replacement from the full sample
loadings and measurement error variance were freely estimated. These
analyses revealed that, if the reliability of the single performance item (Efron & Tibshirani, 1993; Stine, 1989). The indirect
was assumed to be .65 or higher (a reasonable assumption—see Wa- eVect was computed using each of these bootstrap sam-
nous & Hudy, 2001; Wanous, Reichers, & Hudy, 1997 for discussions ples, and based on these results, I constructed bias-cor-
of the reliability of single-item measures), the performance and status rected conWdence intervals to ascertain whether the
conferral measures achieved discriminant validity, as indicated by a
factor correlation whose 95% conWdence interval excluded unity (Ba- indirect eVect diVered signiWcantly from zero (see MacK-
gozzi & Phillips, 1982). innon, Lockwood, & Williams, 2004). In the low interde-
252 A.R. Fragale / Organizational Behavior and Human Decision Processes 101 (2006) 243–261
pendence condition, the indirect eVect from the original less of the level of interdependence, was not empirically
data set was 0.14 (0.31¤0.44), and the 90% conWdence supported. Thus, this study provides a more conclusive
interval for this eVect excluded zero (0.01, 0.31), indicat- test of the process underlying status judgments than
ing a marginally signiWcant indirect eVect in the low those conducted in prior empirical research.
interdependence condition. In the high interdependence
condition, the indirect eVect from the original data set
was ¡0.16 (¡0.33¤0.48), and the 95% conWdence inter- Study 2
val excluded zero (¡0.37, ¡0.02), indicating a signiWcant
indirect eVect in the high interdependence condition. The objectives of Study 2 were both to replicate the
Collectively, these Wndings support the conclusion that pattern of results obtained in Study 1 and to explore the
expected performance mediated the relationship role of agency and communality in individuals’ status
between speech styles and status conferral in both the conferral decisions. In support of the contingent criteria
low and high interdependence conditions. perspective, Study 1 demonstrated that the status bene-
Wts of powerful versus powerless speech depended on the
Discussion level of task interdependence in a group. This perspec-
tive also predicts these eVects are due to the diVerential
The Wndings of Study 1 extend our current under- weighting of agency and communality in status judg-
standing of status conferral judgments in several ways. ments across groups that vary in task interdependence
First, these Wndings suggest that behaviors that have (Hypotheses 4a and 4b): In low interdependence groups,
been traditionally deemed ineVective for enhancing one’s individuals should weight agency more than communal-
status position may not necessarily be so. When an ity as a basis for status conferral judgments, whereas in
upcoming task necessitated independent work, the high interdependence groups, individuals should weight
results replicated prior research: Participants were more communality more than agency. To explore this possibil-
likely to recommend their partner for a leadership posi- ity, participants in Study 2 were asked to rate an actor
tion in the group, and they thought the partner should on several agentic and communal traits.
receive more status and respect in the group, when the Another objective of Study 2 was to replicate Study
partner used a powerful speech style than when the part- 1’s pattern of results using a more ecologically valid
ner used a powerless speech style. However, this pattern manipulation of task interdependence. In Study 1, par-
reversed when the task necessitated interdependent ticipants were explicitly instructed to use either an inde-
work, and greater status was conferred to the partner pendent or interdependent process in Task 2. In reality,
when the partner used a powerless speech style. To date, however, individuals do not always receive such explicit
little value has been attributed to powerless speech, and instructions about how a task should be completed.
conventional wisdom generally counsels against the use Rather, the increasing prevalence of self-managed work
of such tentative speech patterns. On the surface, charac- teams implies that organizational members are often
teristics of powerless speech, such as hesitations, qualiW- given tasks to complete, and left to their own devices to
ers, and disclaimers, seem to convey only negative determine the appropriate processes (Wageman, 1995).
messages about the speaker, such as the speaker’s lack of In such cases, one important factor that is likely to aVect
certainty (i.e., low agency). However, these negative attri- this process decision is the culture of the group or orga-
butions may be oVset by attributing characteristics of nization. Organizational culture has been conceptualized
communality to the speaker, and consequently behaviors as a form of social control; a way to communicate to
such as powerless speech may be status-enhancing organizational members about the values and norms
insituations where communality is valued for perfor- they are expected to uphold (O’Reilly & Chatman,
mance. Thus, this study suggests that the relationship 1996). An important cultural distinction has been made
between an individual’s verbal and nonverbal cues and between individualistic and collectivistic cultures (e.g.,
their conferred level of status may be more complex than Chatman, Polzer, Barsade, & Neale, 1998; Earley, 1993).
previously thought. One of the most distinguishing characteristics between
Second, this study provides greater insight into the these two types of cultures is the level of task interdepen-
process underlying status conferral decisions. Consistent dence that they foster (Chatman etal., 1998; Cox, Lobel,
with the predictions derived from the contingent criteria & McLeod, 1991): In comparison to individualistic cul-
perspective, the eVects of speech styles on status confer- tures, collectivistic cultures place greater emphasis on
ral judgments depended on the level of task interdepen- collaborative work and collective action (i.e., they have
dence in the group. In further support of this perspective, norms of high task interdependence; Hofstede, Neuijen,
these eVects were mediated by expectations of the part- Ohayv, & Sanders, 1990; Wagner & Moch, 1986). In this
ner’s likely performance. In contrast, the Wxed criteria way, an organization’s culture communicates the level of
perspective, predicting that powerful speech would lead task interdependence that is expected of its members
to greater status conferral than powerless speech regard- through implicit means—without explicitly mandating a