What Breaks a Leader : The Curvilinear Relation Between Assertiveness and Leadership

PERSONALITY PROCESSES AND INDIVIDUAL DIFFERENCES
What Breaks a Leader: The Curvilinear Relation Between Assertiveness
and Leadership
Daniel R. Ames and Francis J. Flynn
Columbia University
The authors propose that individual differences in assertiveness play a critical role in perceptions about
leaders. In contrast to prior work that focused on linear effects, the authors argue that individuals seen
either as markedly low in assertiveness or as high in assertiveness are generally appraised as less effective
leaders. Moreover, the authors claim that observers’ perceptions of leaders as having too much or too
little assertiveness are widespread. The authors linked the curvilinear effects of assertiveness to under-
lying tradeoffs between social outcomes (a high level of assertiveness worsens relationships) and
instrumental outcomes (a low level of assertiveness limits goal achievement). In 3 studies, the authors
used qualitative and quantitative approaches and found support for their account. The results suggest that
assertiveness (and other constructs with nonlinear effects) might have been overlooked in research that
has been focused on identifying what makes a leader rather than on identifying what breaks a leader.
Keywords: assertiveness, leadership, interpersonal relations, individual differences, curvilinear effects
The study of lives and personalities has long been concerned
but that the nature of this link has proven elusive for researchers,
with questions of which types of people emerge as effective
in part, because their focus has been on what makes leaders rather
leaders and why (for recent reviews, see Hogan & Kaiser, 2005;
than on what breaks them. Most researchers conducting leadership
Judge, Bono, Ilies, & Gerhardt, 2002). In this wide-ranging liter-
studies have investigated positive, linear determinants and have
ature, a pattern of seemingly contradictory results revolves around
attempted to specify which personality characteristics are present
assertiveness, which is a person’s tendency to actively defend,
in attributions of successful leadership. Far fewer studies have
pursue, and speak out for his or her own interests. Some scholars
identified attributes associated with ineffective leadership. This
have found that leadership emergence and effectiveness are posi-
make-versus-break distinction would not mean much if leadership
tively related to high-assertiveness constructs, such as dominance,
perceptions were symmetrical—that is, if the concerns that appear
aggressiveness, and nondeference (e.g., Bass, 1990; Gough, 1990;
in everyday descriptions of leader weaknesses were simply the
Hills, 1984; Lord, De Vader, & Alliger, 1986). However, leader-
opposite of those characteristics associated with leader strengths.
ship has also been positively linked to low-assertiveness con-
In the present article, we suggest that the concerns that dominate
structs, such as self-sacrifice, cooperativeness, and consideration
perceived weaknesses are not the mirror image of strengths and
(e.g., Bass, 1990; De Cremer & van Knippenberg, 2004; Guilford,
that this difference can clarify the role of an overlooked compo-
1952; Judge, Piccolo, & Ilies, 2004; van Knippenberg & van
nent of leadership: assertiveness.
Knippenberg, 2005). Given these disparate effects, a reasonable
We suspect that the perceived shortcomings of leaders may
observer may suspect that the overarching link between leadership
often revolve around chronically low levels of assertiveness or
and assertiveness is not meaningful, is extremely situation specific,
chronically high levels of assertiveness. High levels of assertive-
or, perhaps, is unknowable. But is there an integrated story that can
ness may bring instrumental rewards and short-term goal achieve-
reconcile these past results and shed new light on who is seen as
ment but can be costly when relationships fray or fail to take root.
an effective leader and why? Or, to put it more generally: How
In contrast, low levels of assertiveness may bring social benefits
does assertiveness matter to leadership, if it matters at all?
but can undermine goal achievement. Thus, increasing levels of
We believe that individual differences in assertiveness matter
assertiveness may often entail a trade-off between social costs and
greatly to observers’ perceptions of leaders and potential leaders
instrumental benefits— between getting along and getting one’s
way. However, we do not think these trade-offs offset one another
Daniel R. Ames and Francis J. Flynn, Management Division, Columbia
or cancel each other out. Below some level of assertiveness,
Business School, Columbia University.
instrumental costs loom large, and leaders may primarily be seen
We thank Cameron Anderson for comments on the article. We also
as ineffective. Above some level of assertiveness, social costs
thank Jenn Hu, Dora Kanellopoulos, Julia Ostrov, and Shelley Trone for
loom large, and leaders may primarily be seen as antagonistic.
help with coding.
Accordingly, we predicted a curvilinear relation between asser-
Correspondence concerning this article should be addressed to Daniel R.
Ames, Columbia Business School, Columbia University, 707 Uris Hall,
tiveness and overall leadership perceptions, such that above and
3022 Broadway, New York, NY, 10027. E-mail: da358@columbia.edu
below certain levels, leaders tend to be seen as less effective.
Journal of Personality and Social Psychology, 2007, Vol. 92, No. 2, 307–324
Copyright 2007 by the American Psychological Association 0022-3514/07/$12.00
DOI: 10.1037/0022-3514.92.2.307
307

---

308
AMES AND FLYNN
Our approach and results make several potential contributions to
be viewed negatively because their behavior seems hostile and
the literatures of personality and leadership. Specifically, we call
offensive (i.e., they cannot get along); those who lack assertiveness
attention to assertiveness as a critical component of leadership
may be seen as weak leaders because they fail to take charge in
effectiveness. More generally, this work suggests that the impact
situations that require initiative and conviction (i.e., they cannot
of individual differences on leadership—and perhaps on basic
get their way). Although many potential leaders may be seen as
interpersonal relations—may be underestimated to the extent that
having too little or too much assertiveness, we did not expect this
nonlinear associations have been overlooked (cf. Simonton, 1995).
to be the case with most other commonly discussed attributes of
leaders. Other attributes, such as intelligence, charisma, or consci-
Assertiveness: A Behavioral Spectrum
entiousness, would likely be viewed as weaknesses only when
targets possessed them in limited amounts (but see Simonton,
Assertiveness is viewed as a dimension describing people’s
1985).
tendency to speak up for, defend, and act in the interest of them-
Although we anticipated that over- and underassertiveness
selves and their own values, preferences, and goals1 (cf. Costa &
would frequently be cited as weaknesses, we did not expect that
McCrae, 1992; Wilson & Gallois, 1993). Assertive behaviors can
references to moderate assertiveness would similarly dominate
be both proactive (e.g., vocalizing needs) and reactive (e.g., de-
discussions of leadership strengths. Part of the reason for this
fending against imposition), both verbal (e.g., articulating clear
asymmetry may be that assertiveness is not as salient at moderate
demands) and nonverbal (e.g., displaying annoyance), and both
levels. Instead, in the right range, assertiveness may fade into the
local or immediate (e.g., a face-to-face disagreement) and diffuse
background, allowing other attributes with positive, linear rela-
or prolonged (e.g., influence tactics over time).
tions with leadership to become more salient. In causal terms, a
Some scholars have portrayed certain behaviors as occupying
moderate level of assertiveness may be a background condition: a
various points on an assertiveness continuum ranging from pas-
necessary but insufficient cause of perceived leadership. Like salt
sivity and submissiveness to aggressiveness and hostility (e.g.,
in a sauce, too much overwhelms the dish; too little is similarly
Wilson & Gallois, 1993). As we explored it in Study 1, everyday
distracting; but just the right amount allows the other flavors to
use of the term appears to reflect this view, with assertiveness
dominate our experience. Just as food is rarely praised for being
labels attached to a wide range of behavioral levels. Thus, low
perfectly salted, leaders may somewhat infrequently be praised for
assertiveness may refer to showing unwarranted deference, high
being perfectly assertive.
assertiveness may refer to belligerently pursuing goals, and mod-
erate assertiveness may refer to defending against imposition and
Instrumental and Social Consequences of Assertiveness
actively making legitimate claims. We adopt this dimensional view
in the present article.
Highly assertive people often get their way, at least in terms of
short-run instrumental benefits. In organizations, assertive people
tend to be seen by others as more powerful than passive employ-
Assertiveness and Perceptions of Leaders
ees, and they tend to adopt more structurally advantageous posi-
In impression formation, people attend closely to information
tions in social networks. In the domain of interpersonal exchanges,
about assertiveness, showing relatively high levels of observer–
assertive behavior, such as an extreme opening offer in a negoti-
observer and observer–target agreement (John & Robins, 1993;
ation and a reluctance to make concessions, can dramatically
Paunonen, 1989; Schmidt Mast, Hall, Murphy, & Colvin, 2003).
increase instrumental outcomes (e.g., De Dreu, Weingart, &
One related dimension, extraversion, has been identified by nu-
Kwon, 2000; Galinsky & Mussweiler, 2001).
merous scholars as the most observable personality trait (e.g.,
Although a high level of assertiveness may entail instrumental
Kenny, 1994). Another aspect of assertiveness, competitiveness,
benefits, it often carries social costs. Assertive people tend to be
often plays a key role in impression formation. Several researchers
(e.g., De Bruin & Van Lange, 1999; Wojciszke, Bazinska, &
1
Jaworksi, 1998) have found that perceivers are more strongly
It is worth recognizing the correspondence between assertiveness and
constructs featured in models of personality, such as extraversion in the Big
drawn to information about a target’s competitive motives than to
Five personality dimension models (e.g., Costa & McCrae, 1992; John &
his or her intelligence. Work on leadership perceptions has like-
Srivastava, 1999) and agency or communion in circumplex models (e.g.,
wise underscored the importance of assertiveness. For instance,
Fournier & Moskowitz, 2000). Although overlaps exist, we believe the
Gough (1990) noted that perceived leadership was associated with
construct of assertiveness, as we use it, is not synonymous or redundant
ratings on items such as active and assertive (see also Judge et al.,
with extraversion, agency, or communion. In work on circumplex models,
2002). In work on explanations for leadership failures, “problems
Wiggins and Broughton (1985) identified an assertiveness dimension—
with interpersonal relationships” (Leslie & Van Velsor, 1996, p.
spanning from the need for aggression and competitiveness at one end to
14), including being overly assertive and being passive and with-
deference, abasement, and mild-mannered behavior on the other end—that
drawn, have emerged as a central theme.
is rotated somewhat counterclockwise of the vertical dimension (the dom-
Given this work linking assertiveness to basic impressions and
inance or agency dimension). We believe this assertiveness dimension,
reflecting aspects of both agency and communion, corresponds most
leadership perceptions and failures, we predicted that coworker
closely to how our raters and informants used the term assertiveness. As for
comments about the weaknesses of potential leaders would fre-
extraversion, there are important aspects of that trait that are not part of
quently highlight assertiveness. One reason we expected it to be so
assertiveness (e.g., positive affect) and vice versa (e.g., nonverbal displays
prevalent in reports of weaknesses is that negative consequences
of disagreement). We measured and examined both constructs in Study 3
occur in both directions—there are, in effect, two ways to get
to clarify whether the effects of assertiveness are distinct from those of
assertiveness wrong. Those who are seen as highly assertive may
extraversion.

---

WHAT BREAKS A LEADER
309
seen as less likeable and less friendly than unassertive people, even
A
when assertive behavior is considered effective, justified, and
Social
Instrumental
appropriate (e.g., Kelly et al., 1982; Kern, 1982). Although the
addition of extra consideration (Woolfolk & Dever, 1979) or
empathy (Zollo, Heimberg, & Becker, 1985) to assertiveness ap-
pears to diminish social costs, even kinder and gentler versions of
assertiveness are seen as leading to worse impressions than are low
Outcomes
levels of assertiveness. Highly assertive people may damage their
relationships and reputations because they are more willing to
Assertiveness
engage in conflict and to use defensive and/or unconstructive
tactics with others (e.g., Graziano, Jensen-Campbell, & Hair,
1996; Kelley & Stahelski, 1970; Kipnis, Schmidt, & Wilkinson,
1980).
In short, we predicted that assertiveness would be positively
linked with instrumental outcomes and negatively linked with
social ones. This notion of social and instrumental outcomes is
B
related to previous leadership research that distinguished between
consideration, or social-emotional behaviors, and structure, or
initiative-taking behaviors (e.g., House, Filley, & Kerr, 1971;
Benefits
Judge et al., 2004). So-called high– high models portray effective
leaders as displaying both consideration and structure. Our view
shares some of these intuitions (i.e., effective leaders are often
neither solely people oriented nor solely outcome driven), yet our
Costs
distinction is focused at the level of outcomes rather than focused
on separate classes of behaviors. Nonetheless, we believe that
distinct social and instrumental dynamics are central in explaining
perceptions of leaders; we turn now to how these dynamics may be
integrated in these perceptions.
The Curvilinear Effect of Assertiveness
C
It could be that the social and instrumental consequences of
Instrumentally
Socially
assertiveness simply cancel each other out, offsetting each other so
impotent
insufferable
that all levels of assertiveness are somehow seen as equally good
in different ways. We believe that this is unlikely. Instead, we
Leadership
expected that the drawbacks of extremely low levels of assertive-
perceptions
ness or extremely high levels of assertiveness would have a dis-
proportionate impact on observers’ evaluations of leaders. The
Assertiveness
rationale for this prediction is rooted in the idea that bad is stronger
than good in impression formation (Baumeister, Bratslavsky,
Figure 1.
Underlying social and instrumental effects aggregate to a
Finkenauer, & Vohs, 2001). Perceivers tend to show behavior that
curvilinear relation between assertiveness and leadership. A: Assertiveness
has been labeled a negativity bias (Rozin & Royzman, 2001),
is positively linked to instrumental outcomes and negatively linked to
social outcomes. B: Perceivers weigh costs more heavily than benefits;
meaning that they pay more attention to the negative aspects of
below a certain point, perceivers attend more to instrumental costs than to
other people and events than to their positive attributes. Recent
social benefits; above a certain point, perceivers attend more to social costs
work has suggested that this effect extends to subordinates’ per-
than to instrumental benefits. C: These main effects aggregate to a curvi-
ceptions of leaders (Amabile, Schatzel, Moneta, & Kramer, 2004).
linear effect for overall perceptions of leadership.
This tendency to focus on negative information suggests that the
costs of extremely low levels or extremely high levels of asser-
tiveness may often outweigh the benefits in the eyes of observers
the most effective behavior. Thus, the relation between assertive-
(see Figure 1). Below a certain range of assertiveness, instrumental
ness and general leadership effectiveness would be curvilinear, as
costs outweigh social benefits, so an individual low in assertive-
portrayed in Figure 1. We hasten to note that we are not suggesting
ness will be seen primarily as instrumentally impotent rather than
that moderately assertive behavior is always an ideal response; we
as relationally successful. Above a certain range of assertiveness,
return to the questions of moderation and flexibility in the final
social costs outweigh instrumental benefits, so an individual with
discussion.
high levels of assertiveness will be seen primarily as socially
insufferable rather than as instrumentally effective. Taken to-
Predictions and Plan of Study
gether, these points suggest that some middle range of assertive-
ness—in which there are neither chronic and glaring social costs
Our account of assertiveness and leadership led to three sets of
nor chronic and glaring instrumental ones—will often be seen as
predictions. The first concerned prevalence: Assertiveness would

---

310
AMES AND FLYNN
be a common theme in coworker comments about colleagues’
from several former coworkers. Participants identified their re-
leadership weaknesses, and these comments would refer to both
spondents and contacted them directly with a standard set of
over- and underassertiveness. In contrast, we expected other
instructions. Respondents were asked, via an anonymous online
widely studied attributes (e.g., intelligence, charisma, and consci-
survey, to provide comments about the former colleague’s
entiousness) to be mentioned less frequently and to be described as
strengths and weaknesses as a leader. For the comment focusing on
weaknesses more exclusively in terms of a colleague’s not having
strengths, the instructions read, “We’d like to hear your views
enough. In comments about strengths, however, we expected these
about this person’s strengths as a colleague and as a leader. Please
other attributes to be more prevalent. Our second prediction fo-
write a few brief thoughts below.” For the comment focusing on
cused on the link between assertiveness and perceived leadership.
weaknesses, the instructions read, “Consider areas where you think
We predicted that assertiveness would have a curvilinear effect:
this person could improve as a colleague and leader. What do you
Leaders seen as chronically low in assertiveness or high in asser-
wish they would do differently . . . what do you wish they would
tiveness would be judged more negatively than those who were
change? Please be honest and constructive.” To minimize con-
seen as moderately assertive. We expected this curvilinear effect to
trived or meaningless responses, we informed raters that the com-
remain after extraversion was controlled for.
ments were optional: “These comments are important, but if noth-
Our third set of predictions addressed underlying effects. We
ing constructive comes to mind, click below to continue.”
expected that assertiveness would be positively associated with
On average, participants gathered 4.00 (SD
1.28) responses
instrumental outcomes but negatively associated with social out-
from work colleagues, although not all colleagues gave responses
comes. Moreover, we expected these outcomes to mediate the
for both strength and weakness questions. We asked raters to
effects of assertiveness, although mediation would be shaped by
clarify how well they knew the person they were rating, on a
negativity effects. At high levels of assertiveness, social outcomes
4-point scale ranging from 1 (not well at all) to 4 (extremely well).
would account for the effects of assertiveness on leadership; at low
The average rating for familiarity was 2.89 (SD
0.75).
levels, instrumental outcomes would mediate.
Participants’ former coworkers were also asked to complete the
We tested these hypotheses in three studies. In Study 1, we
Thomas–Kilmann Conflict Mode Instrument (TKI; Thomas &
examined coworkers’ qualitative comments about colleagues’
Kilmann, 2002), which categorizes a person’s orientation toward
leadership strengths and weaknesses. In Study 2, we gathered
resolving conflict. The five orientations—accommodating, avoid-
coworker ratings of colleagues, testing for a curvilinear relation
ing, collaborating, competing, and compromising—are located by
between assertiveness and leadership. In Study 3, we examined
crossing two dimensions: an individual’s emphasis on satisfying
subordinates’ perceptions of leaders, testing again for curvilinear
his or her own concerns and an individual’s emphasis on satisfying
effects as well as for underlying effects and mediation.
the concerns of the other party. The TKI contains 30 paired
statements that describe two of the five conflict orientation modes.
Study 1
For each pair, respondents were asked to select the statement that
more aptly described the target. Possible scores on each of the five
In Study 1, we collected and analyzed anonymous comments on
conflict orientations ranged from 0 to 12. In the present research,
leadership strengths and weaknesses for potential leaders (master
we focused on the competing orientation because this mode best
of business administration [MBA] students). Comments were pro-
reflected the dimension of assertiveness as work colleagues ap-
vided by former work colleagues who had typically worked with
peared to use it in their comments. Thomas and Kilmann (2002)
the students for 2 to 5 years. We conducted quantitative text
described this orientation as a power-oriented mode that is asser-
analyses on the comments to assess the prevalence of words
tive and uncooperative; statements consistent with this mode refer
related to dimensions such as assertiveness, charisma, and intelli-
to being firm in pursuing goals and pressing to get one’s own
gence. We also conducted qualitative coding to assess the meaning
points made.
of the comments. We predicted that assertiveness would emerge as
Coding.
We conducted a content analysis of the qualitative
the most prevalent theme (both in shortage and overabundance) in
comments provided by each participant’s raters. The comments
weakness comments but that it would be less prevalent in strength
were read by two independent coders who were unaware of the
comments. Work colleagues also completed a rating measure of
study hypotheses. Coders were given definitions of the four con-
assertiveness regarding the targets, enabling us to validate this
structs that our review of the leadership literature and our pilot
measure against our coding of the comments, in anticipation of
work suggested were critical: intelligence, conscientiousness, cha-
using the measure in Study 2.
risma, and assertiveness. The definitions for these attributes were
refined with the coders during the pilot work and are shown in
Table 1.
Method
For each comment, the coders provided two ratings for each of
Sample.
Our sample of potential leaders consisted of 168
the constructs: a prevalence rating and a direction rating. The
people enrolled in a full-time MBA program located on the East
prevalence rating represented the extent to which the construct was
Coast. The participants were 42 (25.0%) women and 126 (75.0%)
mentioned in the comment and was recorded on a 3-point scale
men. The participants were identified as 63.7% Caucasian, 22.6%
(0
no mention, 1
some mention, 2
clear/strong mention).
Asian or Pacific Islander, 2.4% African or African American, and
Ratings of prevalence were not mutually exclusive for the trait
1.8% Latino or Hispanic; 9.5% declined to specify their ethnicity.
constructs: Some of the comments contained no mention of any of
Their mean age was 28.36 years (SD
2.70).
the constructs, whereas other comments touched on more than one.
Procedure.
As part of an organizational behavior course, par-
Sample comments with a prevalence rating of 2 for each construct
ticipants collected qualitative feedback on their leadership skills
are listed in Table 1.

---

WHAT BREAKS A LEADER
311
Table 1
Definitions and Examples for Coding Attributes in Coworker Comments About Leadership Strengths and Weaknesses
Attribute
Coding definition
Strength comment examples
Weakness comment examples
Intelligence
Using thought or creativity to solve
“Very practical and often insightful
“[X’s] intelligence is usually a positive,
interpersonal or work problems;
thinking, extremely logical approach to
but there are times when he needs to
being quick and/or thoughtful in
problems, quick thinker.”
slow down to consider group thinking.”
assessing situations and finding
“[X] is excellent at identifying a problem’s
“I feel [X] may be too creative a problem
solutions
issue and at formulating a possible
solver at times.”
solution.”
Conscientiousness
Exhibiting dedication and steadfast
“Always pulls his own weight. . . . Hard
“[X] tends to be lazy and irresponsible.”
willingness to complete work;
worker.”
ability to complete work in
“She is responsible, dependable and always
“[X] needs to be more diligent in making
efficient, timely, meticulous way
up to the task.”
sure her efforts meet team quality
concerns rather than just getting the job
done.”
Charisma
Ability to motivate others to become
“Bright, charming, enthusiasm that is
“Expand upon your charisma.”
enthusiastic about following an
persuasive and infectious, engaging,
agenda; energizing people and
radiates good character.”
creating enthusiasm
“[X] brings an open, positive, energetic
“Be even more of a charismatic,
attitude and has an infectious ambition
attention-catching leader with some
which inspires those around him.”
star-like, ‘showman’ behavior.”
Assertiveness
Persistence in displaying and
“[X] is assertive with his view points. In
“Being hardnosed and blunt can be an
defending one’s ideas and
addition, [X] takes initiative in
efficient means of getting things done,
interests in an unwavering manner
approaching work and is likely to take
but can bruise people’s feelings in the
without ambivalence; not being
the lead.”
process.”
intimidated by others; speaking up
“[X] is a real take-charge guy. He is willing
“[X] could be stronger on his point of
confidently
to take the initiative and handle tasks
views. He sometimes is willing to
without being asked.”
sacrifice his proposal to maintain a
relationship and avoid tension.”
The direction rating captured whether a comment referred to low
Results
levels of the focal construct (0) or high levels of the focal construct
(1). Thus, a strength comment suggesting someone was “very
Quantitative results.
Our initial set of results focused on quan-
reliable” would be coded as a clear mention of conscientiousness
titative text analysis. There were 493 strength comments (20,129
(prevalence
2) and as a reference to high levels of conscien-
total words; 2,784 unique words; average comment length of 40.8
tiousness (direction
1), whereas a weakness comment suggest-
words) and 426 weakness comments (15,796 words; 2,569 unique
ing someone needed to “be more assertive” would be coded as a
words; average length of 37.1 words).
clear mention of assertiveness (prevalence
2) and as a reference
Our text analysis focused on adjectives in coworker comments
to low levels of assertiveness (direction
0). Direction ratings
(e.g., “X is extremely bright, and one of the most charming people
were given only for comments that had a prevalence rating of 1 or 2.
I have had the opportunity to work with,” “X could be less rigid,”
The levels of agreement and reliability between the coders
“Sometimes X’s presentation of her thoughts comes out somewhat
exceeded general norms for acceptability. Across the attributes
aggressive”). Several of the adjectives that emerged were ambig-
(intelligence, assertiveness, etc.), the coders had 87.0% agreement
uous, with potential meanings that did not indicate something
in the prevalence ratings for strength comments and 93.0% agree-
substantive about strengths or weaknesses. After a close review of
ment in the prevalence ratings for weakness comments. These
adjectives in context, we set aside adjectives that did not consis-
yielded Cohen’s kappa values of .73 and .80, respectively, in line
tently refer to leadership qualities, including good, great, strong,
with norms for reliability (Landis & Koch, 1977). The direction
well, positive, negative, different, hard, high, easy, big, certain,
ratings also showed high levels of agreement: 99.2% for strength
difficult, and level.
comments and 95.0% for weakness comments, yielding kappas of
As seen in Table 2, there was moderate overlap in the leading
.91 and .88, respectively. The percentage agreement and kappa
adjectives for strength and weakness comments. In the comments,
levels for the individual attributes were likewise above acceptable
4 of the 10 most prevalent adjectives were common to both
levels. In the case of assertiveness, agreement for prevalence
strengths and weaknesses (able, focused, effective, constructive),
ratings in strength comments was 86.2% (
.71), whereas
and 14 of the top 30 were common to both strengths and weak-
agreement for direction ratings was 96.2% (
.85); for weakness
nesses. Assertive emerged as the most frequently used adjective in
comments, agreement for prevalence was 89.8% (
.80),
weakness comments, well ahead of focused; for word frequency,
whereas agreement for direction was 95.7 (
.91). After com-
2(1, N
31,592)
4.60, p
.05; for case frequency, 2(1, N
pleting their independent ratings, coders reconciled their re-
852)
6.78, p
.05. Other assertiveness-related terms emerged
sponses. This final set of ratings was used in the analyses that
among the most frequently used adjectives in weakness comments,
follow.
including aggressive, confident, direct, and proactive.

---

312
AMES AND FLYNN
Table 2
Frequently Occurring Adjectives in Comments About Leadership Strengths and Weaknesses, Study 1
Strength comments
Weakness comments
Word
Case
Word
Case
Rank
Adjective
frequency
frequency
Rank
Adjective
frequency
frequency
1
Able
3.48
11.36
1
Assertive
2.03
7.28
2
Willing
2.63
9.94
2
Focused
1.08
3.29
3
Focused
2.24
8.32
3
Able
1.01
3.05
4
Effective
1.79
6.49
4
Sure
1.01
3.52
5
Analytical
1.74
6.90
5
Effective
0.82
2.35
6
Thoughtful
1.54
5.88
6
Aggressive
0.70
2.35
7
Intelligent
1.34
5.48
7
Constructive
0.70
2.58
8
Motivated
1.09
4.26
8
Firm
0.63
1.88
9
Considerate
1.04
4.26
9
Confident
0.57
1.88
10
Constructive
1.04
4.06
10
Involved
0.57
2.11
11
Creative
1.04
4.26
11
Quick
0.57
1.88
12
Logical
1.04
4.26
12
Direct
0.51
1.88
13
Professional
0.99
4.06
13
Personal
0.44
1.41
14
Involved
0.89
3.65
14
Professional
0.44
1.64
15
Bright
0.84
3.45
15
Smart
0.44
1.64
16
Understanding
0.84
2.64
16
Willing
0.44
1.64
17
Dedicated
0.79
3.25
17
Considerate
0.38
1.41
18
Driven
0.79
3.25
18
Critical
0.38
1.41
19
Clear
0.75
3.04
19
Afraid
0.32
1.17
20
Fair
0.75
3.04
20
Comfortable
0.32
1.17
21
Quick
0.75
3.04
21
Emotional
0.32
1.17
22
Firm
0.70
2.84
22
Honest
0.32
1.17
23
Friendly
0.70
2.84
23
Impatient
0.32
1.17
24
Personal
0.70
2.84
24
Proactive
0.32
1.17
25
Smart
0.70
2.64
25
Reasonable
0.32
1.17
26
Confident
0.65
2.64
26
Selfish
0.32
1.17
27
Direct
0.65
2.64
27
Arrogant
0.25
0.94
28
Fun
0.65
2.43
28
Collaborative
0.25
0.94
29
Capable
0.60
2.23
29
Diplomatic
0.25
0.94
30
Efficient
0.60
2.43
30
Frustrated
0.25
0.94
Assertive
0.40
1.42
Competitive
0.19
0.70
Aggressive
0.15
0.61
Competitive
0.05
0.20
Note.
Adjectives in bold are common to both strength and weakness top 30 lists. Word frequency is count per 1,000 words for 20,129 strength words
and 15,796 weakness words. Case frequency is percentage of comments in which the adjective appears in 493 strength comments and 426 weakness
comments.
Although assertive and aggressive—perhaps the most central
weakness comments addressed these issues and allowed us to
adjectives for our purposes—appeared among the 10 most preva-
further test our predictions.
lent weakness adjectives, they were not among the 10 most prev-
As expected, the coding revealed that assertiveness was a sub-
alent, or even the 30 most prevalent, strength adjectives. As shown
stantially more prevalent theme in weakness comments than were
in Table 3, various combinations of assertiveness-related adjec-
themes of conscientiousness, intelligence, and charisma (see Table
tives were significantly more prevalent in weakness comments. As
4). On the prevalence coding scale, weakness comments were
expected, other attributes were reflected more frequently in
rated on average at 1.06 (SD
0.99) for assertiveness. This was
strength comments. Adjectives related to intelligence, conscien-
significantly higher than were the ratings for conscientiousness,
tiousness, and charisma were significantly more prevalent in
t(414)
9.55, p
.001; intelligence, t(415)
20.80, p
.001;
strength comments (see Table 3).
and charisma, t(415)
21.40, p
.001. Indeed, the average
Qualitative coding results.
Although these quantitative text
prevalence score for assertiveness was significantly higher than the
analyses lend support to our account, they cannot resolve some
combined prevalence scores for the remaining traits (M
0.41,
questions about the contextual meaning of the words in the
SD
0.79), t(415)
9.05, p
.001. Over half of the weakness
strength and weakness comments (e.g., whether the use of the
comments featured a clear mention of assertiveness (coded as 2).
adjective assertive in a weakness comment entailed a discussion of
A far smaller share of weakness comments featured a clear men-
overassertiveness, underassertiveness, or something else entirely).
tion of the other attributes.
Further, although we believe the adjectives reflected a good share
We predicted that weakness comments about assertiveness
of the meaning of the comments, part of raters’ sentiments were
would be more likely to refer to excessive levels (overassertive-
not captured in adjectives. The qualitative coding of strength and
ness) than would weakness comments about the other attributes.

---

WHAT BREAKS A LEADER
313
Table 3
Prevalence of Adjectives and Adjective Groups in Comments About Leadership Strengths and
Weaknesses, Study 1
Strength comments
Weakness comments
Comparison
Word
Case
Word
Case
Adjective
frequency
frequency
frequency
frequency
2(1, N
35,925)
p
Assertive
0.4
1.4
2.0
7.3
21.10
.01
Aggressive
0.1
0.6
0.7
2.4
6.81
.01
Assertive, aggressive,
competitive, active,
proactive, confident, direct
2.2
5.5
4.5
8.5
14.79
.01
Intelligent
1.3
5.5
0.3
0.9
12.15
.01
Intelligent, smart, logical,
analytical, creative, bright
6.7
17.0
0.9
2.1
70.55
.01
Focused, effective, efficient,
conscientious, reliable,
diligent
6.0
14.6
1.9
3.3
35.13
.01
Energetic, charismatic,
enthusiastic
0.9
2.5
0.2
0.4
7.52
.01
Note.
Word frequency is count per 1,000 words for 20,129 strength words and 15,796 weakness words. Case
frequency is percentage of comments in which the adjective appears in 493 strength comments and 426 weakness
comments. Chi-square and p values are comparisons of word frequency for strength comments versus weakness
comments.
As shown in Table 4, the mean direction coding for weakness
898)
29.24, p
.01. The reverse was true for the other
comments referring to assertiveness was 0.48 (SD
0.50), mean-
attributes, which showed greater prevalence in strength com-
ing that 48.0% of weakness comments referred to some form of
ments: conscientiousness, t(401)
10.57, p
.001; intelli-
overassertiveness, whereas 52.0% referred to some form of under-
gence, t(403)
15.40, p
.001; and charisma, t(404)
4.67,
assertiveness. This was significantly higher than the mean direc-
p
.001.
tion coding for conscientiousness, t(225)
10.10, p
.001;
In sum, the qualitative coding results converged with the quan-
intelligence, t(225)
6.97, p
.001; and charisma, t(225)
titative text analysis to support our account: Assertiveness (in both
14.48, p
.001.
excess and absence) was by far the foremost theme in coworkers’
We also predicted that assertiveness, unlike other attributes,
comments about leadership weaknesses. Assertiveness was less
would be more prevalent in comments about weaknesses than in
prevalent in strength comments than in weakness comments; the
comments about strengths. As shown in Table 4, this appeared
reverse was true for conscientiousness, intelligence, and charisma.
to be the case: For prevalence coding, t(404)
5.28, p
.001.
Comment and rating measures of assertiveness.
We also
Although over half of the weakness comments featured a clear
tested whether assertiveness, as measured by the TKI construct of
mention of assertiveness (coded as 2), only a third of strength
competitiveness, converged with the coding of assertiveness in
comments featured a clear mention of assertiveness, 2(1, N
weakness comments. A high correlation would suggest that both
Table 4
Prevalence and Direction Codings for Attributes in Coworker Comments About Leadership Strengths and Weaknesses, Study 1
Strength comments
Weakness comments
Prevalence
Direction
Clear
Prevalence
Direction
Clear
mentions
mentions
Attribute
M
SD
M
SD
(%)
M
SD
M
SD
(%)
Assertiveness
0.69
0.95
0.95
0.23
33.8
1.06
0.99
0.48
0.50
51.7
Conscientiousness
1.08
1.00
1.00
0.00
53.5
0.38
0.77
0.14
0.35
18.1
Intelligence
0.76
0.97
0.99
0.07
37.8
0.02
0.18
0.25
0.50
0.7
Charisma
0.11
0.45
1.00
0
5.4
0.01
0.11
0
0
0.2
Note.
Clear mentions are the percentage of comments coded as 2 on prevalence for the given attribute.

---

314
AMES AND FLYNN
measures reflected coworkers’ opinions about a target’s level of
participants were 100 (25.8%) women and 288 (74.2%) men. They
assertiveness; a low correlation could mean that one or both of
identified themselves as Caucasian (61.1%), Asian or Pacific Is-
these measures did not gauge such impressions.
lander (16.0%), Latino or Hispanic (8.2%), or African or African
To quantify references to assertiveness in weakness comments,
American (4.4%); 10.3% declined to specify their ethnicity. Their
we computed a measure of overassertiveness by multiplying the
mean age was 28.8 years (SD
2.70).
prevalence coding for assertiveness in each comment with a mod-
Procedure.
As in Study 1, participants collected feedback on
ified direction coding for each comment ( 1 for low levels,
1 for
their leadership behavior from several former coworkers as part of
high levels). Thus, a comment with a partial reference to a target
an organizational behavior course. Participants identified their
not being assertive enough would score a
1 (1 Prevalence
1
respondents and contacted them directly with a standard set of
Direction
1), whereas a comment with a clear reference to a
instructions. Respondents were asked, via an online survey, to
target being too assertive would score a 2 (2 Prevalence
1
provide ratings referring to four domains of leadership: motivation,
Direction
2).
social influence, managing conflict, and working in teams. Al-
We conducted the analyses at both the rating and the target
though not necessarily a comprehensive catalog of leadership
(averaging across ratings) levels. In both cases, the TKI assertive-
skills, these domains are central to many descriptions of what
ness rating measure was positively correlated with the overasser-
effective leaders achieve (motivating and influencing others, deal-
tiveness comment measure. At the rating level, r(416)
.44, p
ing with conflicts, and leading teams). For each domain, on 7-point
.001. To gauge reliability in collapsing across raters to the target
scales that range from 1 (never) to 7 (always), respondents rated
level, we computed the Intraclass Correlation Coefficient 2 mea-
how strongly five statements (e.g., social influence: “S/he is able
sure (Bartko, 1976), which yielded values of .61 for the TKI
to direct and steer meetings in his/her favor”) characterized the
measure and .47 for the overassertiveness measure, suggesting
participant’s behavior at work. The items are noted in the Appen-
modest reliability. At the target level, the TKI and comment
dix.
measures were correlated at r(149)
.62, p
.001.
The survey also included the TKI, which assessed coworkers’
Thus, targets described as being overassertive in open-ended
perceptions of the participant’s level of assertiveness (see Study 1
weakness comments were likely to receive higher numerical scores
for a description). Finally, raters were asked to rate how well they
in the TKI assertiveness measure. The substantial correlations
knew the target, on a 4-point scale that range from 1 (not well at
suggest that both measures reflect underlying perceptions of as-
all) to 4 (extremely well). On average, participants had 3.87 (SD
sertiveness. We used the TKI measure to gauge coworker percep-
1.40) former colleagues as raters. The average score on the famil-
tions of assertiveness in a separate sample, in Study 2.
iarity measure was 3.19 (SD
0.68). The survey took about
10 –15 min to complete.
Discussion
Results
Study 1 confirmed our expectations about the prevalence of
assertiveness in comments about leadership weaknesses. With both
Constructs and reliability.
The reliability (alpha) coefficients
quantitative text analysis and qualitative coding, we found that
for the four 5-item leadership scales were .69 (motivating), .64
although assertiveness was not a dominant theme in strength
(influence), .68 (conflict), and .71 (teams). The alpha for the
comments, it was substantially more common than other themes
overall leadership scale, including items from all four subscales,
(including conscientiousness, intelligence, and charisma) in weak-
was .89. To gauge reliability across raters for each participant, we
ness comments. Moreover, although these other attributes were
computed intraclass correlation coefficients (ICC[2]; Bartko,
described almost exclusively in terms of “not enough” in weakness
1976) for each construct, which yielded .60 for assertiveness, .34
comments, references to assertiveness were varied, with nearly
for motivating, .37 for influence, .41 for conflict, .44 for teams,
half of them referring to overassertiveness and the remainder
and .44 for leadership. This suggests modest within-subject reli-
referring to underassertiveness.
ability. We interpreted this as suggesting that different raters
observed different samples of the participants’ behaviors and/or
Study 2
saw them in different contexts, capacities, or roles. Nonetheless,
we expected our predicted effects to emerge whether average
Study 1 showed that, in qualitative comments about leadership
views of an individual participant (separating participants consis-
weaknesses, assertiveness was a prevalent theme and was split
tently seen as low or high in assertiveness) were considered or
between references to “too much” and “too little.” Consistent with
results at the rater level (separating raters according to their per-
this pattern, we expected that quantitative ratings of a colleagues’
ception of a participant as low in assertiveness or high in asser-
leadership and assertiveness would show a curvilinear relation. In
tiveness) were considered. As a result, we conducted analyses at
Study 2, MBA students gathered ratings of their assertiveness and
multiple levels, including both the rating level and the target level.
their leadership from former work colleagues, which allowed us to
We also pursued multilevel modeling.
test for the expected effects.
Curvilinear effects.
To test our prediction that assertiveness
would have a curvilinear link with perceptions of leadership, we
used regression analyses with ratings of assertiveness to predict
Method
leadership measures. Our models featured both linear and squared
Sample.
Our sample of potential leaders in Study 2 consisted
terms for assertiveness. A significant negative coefficient for this
of 388 people enrolled in a full-time MBA program located on the
squared measure would be consistent with the expected inverted-U
East Coast (none of whom had participated in Study 1). The
curvilinear effect.

---

WHAT BREAKS A LEADER
315
Table 5 shows the results of the regression models at the target
p
.001. The linear assertiveness term had a significant and
level, including both linear-term-only results and linear-plus-
positive effect (B
.0568), t(1102)
2.79, p
.01. We ran
squared-term results. The curves from the full models are plotted
analogous multilevel models for each of the four behavioral do-
in Figure 2. As expected, the predicted curvilinear effects emerged
mains; in each one, the squared assertiveness term had the pre-
in each of the four domains as well as in the aggregate leadership
dicted significant and negative effect.
measure (an average of the four behavioral domains). In all five
The significance of the squared term in our results indicates that
cases, the squared term had a significant negative coefficient, with
the effect of assertiveness may be curvilinear but does not establish
ps at or below .01. Similar results emerged in analyses at the rating
whether it is symmetrical—that is, whether leadership perceptions
level, as shown in Table 5. Again, the squared terms were signif-
are reduced both above and below certain levels of assertiveness
icantly negative for each construct.
(see Figure 2 for fitted curves). To examine this, we first con-
Given the nested nature of the data, we also pursued multilevel
ducted a tertiary split of assertiveness at the rating level. The
modeling, using the PROC MIXED routine in SAS (Singer, 1998).
aggregate leadership measure was significantly lower for those in
Our first analysis featured a mixed model that could be used to
the highest third of assertiveness compared with those in the
predict aggregate leadership ratings and that featured two fixed
middle third (5.54 vs. 5.70), t(1019)
3.60, p
.001, although
effects (assertiveness and the squared value of assertiveness) and a
it was not significantly lower for those in the lowest third (5.72 vs.
random effect, which allowed the intercept to vary at the target
5.70), t(914)
0.70, ns. We suspected that the negative effects of
level. As in our regression analyses, and consistent with our
assertiveness might be stronger at more extreme high and low
predictions, we found that the squared assertiveness term had a
levels. Accordingly, we compared the extremes (ratings of 0 for
significant and negative effect (B
.0066), t(1102)
4.24,
low assertiveness and 12 for high assertiveness) with less extreme
Table 5
Results of Multiple Regression Models Predicting Leadership Constructs With Assertiveness at
the Target Level, Study 2
Assertiveness
Assertiveness2
Leadership domain
Model R2
t
p
t
p
Target level
Motivation
Linear term only
0.17
3.33a
.01
.03
Linear and square terms
0.42
1.81b
0.07
0.60
2.59b
.01
.05
Influence
Linear term only
0.23
4.55a
.01
.05
Linear and square terms
0.67
2.97b
.01
0.92
4.06b
.01
.09
Conflict
Linear term only
0.27
5.46a
.01
.07
Linear and square terms
0.57
2.52b
.01
0.85
3.79b
.01
.11
Teams
Linear term only
0.35
7.38a
.01
.12
Linear and square terms
0.32
1.43b
.15
0.68
3.10b
.01
.15
Leadership (aggregate)
Linear term only
0.30
6.08a
.01
.09
Linear and square terms
0.57
2.55b
.01
0.88
3.97b
.01
.12
Rating level
Motivation
Linear term only
0.18
7.02c
.01
.03
Linear and square terms
0.10
0.97d
.33
0.29
2.84d
.01
.04
Influence
Linear term only
0.02
0.61c
.54
.00
Linear and square terms
0.44
4.35d
.01
0.44
4.34d
.01
.01
Conflict
Linear term only
0.14
5.44c
.01
.02
Linear and square terms
0.37
3.71d
.01
0.53
5.27d
.01
.04
Teams
Linear term only
0.18
7.03c
.01
.03
Linear and square terms
0.06
0.58d
.56
0.25
2.44d
.02
.04
Leadership (aggregate)
Linear term only
0.15
5.68c
.01
.02
Linear and square terms
0.30
2.97d
.01
0.46
4.57d
.01
.04
Note.
At the target level, N
388. At the rating level, N
1,501.
a df
386.
b df
385.
c df
1500.
d df
1499.

---

316
AMES AND FLYNN
1
0
t
i
ng
Overall
r
a
)
re
Motivation
p
o
c -1
shi
Influence
-s
er
(Z
Conflict
Teams
Lead
-2
-3
-3
-2
-1
0
1
2
3
Assertiveness (Z-score)
Figure 2.
Fitted curvilinear relations between perceptions of assertiveness and leadership at the target level
from Study 2 are shown.
ratings (ratings of 1–11). Again, the leadership measure was sig-
therefore, was viewed more negatively). To address these issues,
nificantly lower for those at high levels of assertiveness compared
we sought a more wide-ranging sample of managers in Study 3.
with those at middle levels of assertiveness (5.22 vs. 5.67),
t(1448)
5.97, p
.001; this pattern of significance extended to
all of the underlying constructs (motivation, influence, conflict,
Study 3
and teams). The difference was directional although not significant
Study 2 revealed the curvilinear effects of assertiveness on leader-
when those at low levels of assertiveness were compared with
ship. In Study 3, we sought to replicate and extend these results in two
those at middle levels of assertiveness (5.56 vs. 5.67), t(1410)
major ways. First, in Study 3, we examined the curvilinear effects in
1.18, ns. Examining the underlying constructs, we found signifi-
a broader and older sample of target leaders, which featured over 200
cant differences for influence (5.06 vs. 5.34), t(1419)
2.20, p
informants who provided comments and ratings on their most recent
.03, and conflict (5.19 vs. 5.49), t(1414)
2.24, p
.03, although
manager or leader (with a mean estimated target age of nearly 40
not for motivation or teams.
years). We also sought evidence that, compared with moderate levels
of assertiveness, both high and low levels were associated with
Discussion
significantly lower leadership evaluations. Second, Study 3 was de-
signed to explore the processes underlying the curvilinear effect. As
In Study 2, we analyzed coworker ratings of colleagues’ lead-
noted in the introduction, we expected assertiveness to be positively
ership at the rater level, at the target level, and with multilevel
linked with instrumental outcomes and negatively linked with social
models. In all three approaches, assertiveness had a curvilinear
outcomes. We also made a negativity prediction. Assuming that costs
relation with leadership behavior in several domains, including
exert greater weight in perceptions, we expected that instrumental
managing teams, dealing with conflict, and influencing and moti-
outcomes would be the primary determinant of leadership at low
vating others. Our results show that, compared with moderate
levels of assertiveness, whereas social outcomes would be the primary
levels of assertiveness, high levels of assertiveness were associated
determinant of leadership at high levels of assertiveness. Thus, we
with significantly lower ratings of leadership by informants. The
expected different patterns of mediation, depending on level of asser-
comparison of low levels with moderate levels yielded more mixed
tiveness.
results. Although those at very low levels of assertiveness were
Study 3 had several other features worth noting. One concerns
rated significantly worse at influencing others and managing con-
our measures: Study 2 focused on leadership behavior measures;
flict, the differences were not significant for motivating others or
whereas Study 3 captured overall evaluations of leaders’ current
managing teams. Does this cast doubt on the costliness of low
effectiveness, as well as perceptions of future leadership success.
levels of assertiveness? We suspect that our data might have
We also gathered qualitative comments about leader strengths and
featured a restriction of range, given that our targets were MBA
weaknesses and coded them in an effort to replicate our results
students in a competitive program. It seems plausible that our
from Study 1. Finally, we sought to test whether the effects of
sample’s distribution of assertiveness might have been somewhat
individual differences in assertiveness on leadership are accounted
higher than that of the population at large. Further, compared with
for in part or in whole by extraversion. Numerous studies have
the population of working managers, our participants (averaging
identified a connection between extraversion and leadership (see
about 29 years in age) might have been in comparatively lower
Judge et al., 2002 for a meta-analysis); we wanted to clarify
status positions for which low assertiveness was more normative
whether the effects of assertiveness are distinct from any effects of
and role congruent (i.e., high assertiveness was incongruent and,
extraversion.

---