Agreeableness: Dimension of Personality
or Social Desirability Artifact?
William G. Graziano
Renée M. Tobin
Texas A&M University
ABSTRACT Agreeableness is linked to socially valued traits and prosocial
motives, so self-reported Agreeableness may be distorted by self-favoring
biases (SFB). A three-study multimethod research program explored links
between the Big Five dimension of Agreeableness and SFB from three
perspectives. First, we examined zero-order relations between Agreeableness
and SFB measures (N = 316). Next, we used a round robin design (N = 351)
and the Social Relations Model analyses (Kenny, 1994) to partition
perceptions of Agreeableness into target and perceiver effects. These effects
then were related to SFB concerns, and differential responsiveness to a
manipulation of the social desirability of Agreeableness. Study 3 (N = 312)
examined a manipulation of Agreeableness as a moderator of the relation
William G. Graziano and Renée M. Tobin, Department of Psychology.
We are especially grateful to Delroy L. Paulhus for his advice in designing and
interpreting this program of research. We thank William H. M. Bryant, David Funder,
Elizabeth C. Hair, Lauri A. Jensen-Campbell, Deborah A. Kashy, W. Joel Schneider,
Michele Marie Tomarelli, and Stephen Gammaguy West for their help and expertise
in design, data collection, and analysis. We also thank Linda Albright, Winfred E.
Arthur, Jr., John F. Finch, Lowell A. Gaertner, David Kenny, Thomas Malloy,
Rowland Spence Miller, Les Morey, Cynthia Riccio, Bradley E. Sheese, David E.
Tobin, and Thomas J. Tobin for their suggestions and comments on earlier versions of
this manuscript. We thank members of the Texas A&M Personality and Social
Influence Research Team for their help in completing the data collection. This
research was supported by grant R01 MH50069 from the National Institute of Mental
Health to William G. Graziano. Correspondence can be sent to W. G. Graziano,
Department of Child Development and Family Studies, Room 102, Purdue University,
West Lafayette, IN 47907-1267 (USA), e-mail: [e-mail].
Journal of Personality 70:5, October 2002.
#
Copyright 2002 by Blackwell Publishing, 350 Main Street, Malden, MA 02148,
USA, and 208 Cowley Road, Oxford, OX4 1JF, UK.
696 Graziano & Tobin
between dispositional Agreeableness and interpersonal conflict tactics.
Overall, results indicate that Agreeableness is not easily manipulated nor
distorted by SFB. Results are discussed in terms of Agreeableness as a
substantive system of motives.
Agreeableness: Dimension of Personality or
Social Desirability Artifact?
Social desirability confounds have concerned personality researchers
from the founding of the modern field of personality. Allport and
Odbert (1936) recognized the social desirability problem in their
groundbreaking lexical analysis of trait names. Allport and Odbert
carefully segregated censorial, evaluative words (e.g.,   amiable, 
  Agreeable  ) from   neutral  trait words. These authors suggested
that the evaluative words   should be avoided by psychologists unless
they are prepared to deal with the subject of social judgment. . . 
(p. vii). Given this state of affairs, one might not have expected factor-
analytic studies of personality based on the Allport-Odbert purified list
to uncover a major dimension of   social evaluation.  When Digman
and Takemoto-Chock (1981) reanalyzed data from six major large-
scale studies, the first factor to emerge was labeled   friendly
compliance vs. hostile noncompliance.  This dimension is now
commonly called Agreeableness (Digman, 1996; Graziano &
Eisenberg, 1997; Havill, Besevegis, & Mouroussaki, 1998). Appar-
ently, some substance remained after the lexic cleansing.
Of the dimensions in the five-factor approach, Agreeableness is
arguably the least well understood (e.g., Havill, Besevegis, &
Mouroussaki, 1998; Wiggins & Trapnell, 1997, pp. 748 750). From
a conceptual perspective, Agreeableness seems to be the dimension
concerned with interpersonal relations. Specifically, it is concerned
with individual differences in the motivation to maintain positive
relations with others (e.g., Digman, 1997; Hogan, 1983; Graziano &
Eisenberg, 1997; MacDonald, 1992, 1995; Wiggins & Trapnell, 1997).
From an empirical perspective, Agreeableness seems to be a major
dimension of personality, perhaps even the largest single dimension
(e.g., Goldberg, 1995; Graziano & Eisenberg, 1997; Kohnstamm,
Halverson, Mervielde, & Havill, 1998).
The three major self-report measures of Agreeableness (NEO-PI,
BFI, Goldberg Markers) show substantial convergence (John &
Agreeableness & Social Desirability 697
Srivastava, 1999). Moreover, research with these self-report measures
has accumulated substantial support for predictive validity. For
example, self-reports show considerable agreement with expert, peer,
and spouse ratings (John & Srivastava, 1999). Self-reports also predict
behaviors such as efforts to minimize interpersonal conflict (Graziano,
Jensen-Campbell, & Hair, 1996; Jensen-Campbell & Graziano, 2001),
to maintain intragroup cooperation (Graziano, Hair, & Finch, 1997), and
to control negative emotions in the presence of others (Tobin, Graziano,
Vanman, & Tassinary, 2000).
Hafdahl, Panter, Gramzow, Sedikides, and Insko (2000) found that
Agreeableness was the single most descriptive dimension among the
Big Five for all selves (e.g., Actual, Ideal), but was especially
descriptive of the Ought Self. The importance of Agreeableness as a
dimension of personality is also evident in its close connection with
communion, one of the   Big Two  appearing in circumplex models of
personality (Wiggins & Trapnell, 1997).
There is a second construal of the empirical findings that offers a
somewhat different perspective, but harkens back to Allport and
Odbert. Agreeableness may appear to be a large dimension of
personality because a major artifact contaminates it. What may be
pervasive in Agreeableness assessment is socially desirable responding
(SDR), defined by Paulhus (1991) as   the tendency to give answers
that make the respondent look good  (p. 17). Perhaps more than any of
the other dimensions in the five-factor approach, Agreeableness
involves socially desirable qualities (e.g., Hafdahl et al., 2000; Paulhus,
Bruce, & Trapnell, 1995), so persons motivated to look good in the eyes
of others may be especially likely to describe themselves as being
agreeable. In this construal, the motivation underlying SDR, not
Agreeableness, is the   active ingredient  in predicting and explaining
external criteria like cooperation or the control of negative emotions.
Consistent with this second construal, Paulhus et al. (1995) found
that when participants were asked to present themselves in the best
possible light, Agreeableness and Conscientiousness were among the
top dimensions of the Big Five to show differences with an   honest 
appraisal. Additional support comes from distributional data. If self-
favoring biases differentially affect Agreeableness in comparison with
the other Big Five dimensions, we might expect Agreeableness raw
scores to be more negatively skewed than scores for the other
dimensions, with a relatively smaller number of persons describing
themselves as low in Agreeableness compared to the other dimensions.
698 Graziano & Tobin
In examining the distributional characteristics of a large sample of raw
score responses to Goldberg s (1992) adjective markers of the Big
Five dimensions, Tobin et al. (2000, Study 2, N = 1757) found that the
Agreeableness dimension showed a much larger negative skew
(gamma-1 = 1.53) and more peaked kurtosis (gamma-2 = 4.23)
than any of the other Big Five dimensions. For purposes of
comparison, the mean skew and kurtosis for the other four dimensions
were 0.57(gamma-1) and 1.36 (gamma-2). The same pattern was
replicated in an independent sample (Tobin et al., Study 1, N = 316).
This evidence is consistent with a self-favoring bias interpretation, but
it is not definitive because substantive factors other than self-favoring
biases could influence the distribution of Agreeableness raw scores.1
At the least, these data suggest that Agreeableness is somehow
  different  from the other Big Five dimensions, and self-favoring
biases may be one of the contributors.
Complicating the assessment situation is that SDR is not a single
process (e.g., Gangestad & Snyder, 2000; Lautenschlager, 1986;
Lautenschlager & Flaherty, 1990; Ones, Viswesvaran, & Reiss, 1998;
Strosahl, Linehan, & Chiles, 1984). Paulhus (1984, 1986) developed a
theoretical integration of these diverse findings and initially suggested
that SDR could be separated into two processes. The first is
  impression management,  a purposeful tailoring of responses to
impress an audience. The second is   self-deceptive enhancement,  an
unconscious, self-deceptive, self-favoring bias. Subsequent research
showed that both processes could produce unconscious self-favoring
distortions (Paulhus, in press). At the least, this work suggests that
assessments of SDR must contain measures of both impression
management and self-deception.
1. S. G. West (personal communication) observed that skew and kurtosis differences
for Agreeableness (Ag) could arise from other artifacts besides SDR. It is possible, for
example, that the Ag items are   easy,  in comparison with items from the other
dimensions. It may take fewer recalled acts of helpfulness for a person to rate him/
herself as helpful than it takes acts of diligence for a person to rate himself or herself
as conscientious. If less easy items had been included, then perhaps the skew and
kurtosis would be more nearly normal. That being said, in psychometric terms Ag is a
  sentiment,  not an ability. There is no clear objective anchor for the high or low
ends, as in intelligence testing or word recognition. Ag does not scale intuitively or
easily along an easy-difficult dimension, although the behavioral correlates of Ag
might scale this way. The helpful behavior of rescuing Jews from Nazis represents a
  higher hurdle  for Ag than does helping someone pick up a few dropped books (e.g.,
Campbell, 1963).
Agreeableness & Social Desirability 699
Recent evidence suggests that SDR can be traced still further to at
least two different processes, labeled egoistic bias and moralistic bias
(Paulhus & John, 1998). These self-deceptive biases are an expression
of the two fundamental values of agency and communion, which
compel motives for power and approval, respectively. As a result, two
personality constellations, labeled Alpha and Gamma, coalesce. Alpha
is associated with an egoistic bias toward exaggerating one s social
and intellectual status. This tendency is expressed in overly positive
ratings on Extraversion and dominance. Gamma is associated with a
moralistic bias toward claiming   saint-like  attributes, expressed in
overly positive ratings on Agreeableness, dutifulness, and restraint.
Research on Agreeableness, and corresponding theoretical expan-
sion, will remain underdeveloped until links to social desirability are
addressed explicitly, systematically, and empirically. On one hand,
there are plausible reasons for assuming that Agreeableness is
associated with social evaluation and with an apparent self-favoring
bias toward SDR. If we assume that being   warm  and   kind  are
generally more socially desirable than being   cold  and   unkind,  then
SDR may be especially likely on this dimension in both self-ratings and
in the ratings of others. On the other hand, there may be substantive
individual differences in motives for maintaining positive relations with
others, and even if they are related to the self-favoring biases of social
desirability, they may still parsimoniously explain external, positive
social behaviors like intragroup cooperation, conflict reduction, and
efforts to control emotions noted previously (Cunningham, Wong, &
Barbee, 1994; McCrae & Costa, 1983; Smith, 1998).
Either way, the issue is important. If Agreeableness is contaminated
by various forms of SDR, or worse, if its empirical relations to
external criteria are due more to SDR than to dispositional motives for
maintaining positive social relations, then theoretical descriptions of
Agreeableness need to be adjusted. If contamination is minimal, or if
empirical relations are not altered after appropriate adjustments for
SDR, then further theoretical development of Agreeableness as a
personality dimension is in order.
The three studies presented here use both correlational and
experimental methods as part of a converging, multimethod approach.
We explore links between Agreeableness and SDR, in both self-ratings
(S-data) and ratings by others (R-data). First, we examine the simple,
zero-order correlations among S-data Agreeableness and several
measures of SDR. Second, we examine the relation between S-data
700 Graziano & Tobin
Agreeableness and R-data by observers. Third, we explore the
potential moderating effects of dispositional SDR measures (Paulhus
Technique 3) and of situational manipulations of SDR (Paulhus
Technique 4) on self-reported and other-rated Agreeableness. Finally,
we explore the potential impact of such moderation for predicting the
relations between Agreeableness and the external behavioral criteria of
interpersonal conflict tactics.
STUDY 1
The potential SDR biases in Agreeableness may be seen directly in the
zero-order correlations with various established measures of SDR. To
cast a reasonably wide net, we identified three different forms of SDR
that might be related to Agreeableness: Impression Management (IM),
Self-Deception (SDE), and Self-Monitoring (SM). Specifically, we use
the Balanced Inventory of Desirable Responding (BIDR; Paulhus,
1988) to assess IM and SDE, and the Self-Monitoring scale (Snyder,
1987) to assess SM differences in concerns with strategic self-
presentation. Individuals who score higher on the impression-
management (IM) portion of the BIDR are more likely than their
peers to overstate their performance of desirable behaviors and deny
performing undesirable behaviors. Individuals who score higher on the
self-deception (SDE) portion of the BIDR are more likely than their
peers to see themselves as more competent socially and intellectually
than they are seen by well-informed others. At least conceptually, SM
is similar to IM, in that individuals who score higher on the SM scale
are said to be more concerned with strategically creating positive
social impressions than those individuals who score lower in self-
monitoring. In addition, we disaggregated the full SM scale into three
components (Acting, SM Extraversion, & Other Directedness), which
were derived empirically and used in previous research (Briggs &
Cheek, 1986; Graziano & Bryant, 1998).
The focus of this research was on the link between Agreeableness
and SDR. It is possible, however, that Agreeableness is not the only
Big Five dimension of personality related to SDR (e.g., Paulhus &
John, 1998). To explore this possibility, and to address issues of
convergent and discriminant validity, we collected data using two
different measures of the Big Five dimensions of personality structure,
and correlated them with the various measures of SDR in Study 1.
Agreeableness & Social Desirability 701
METHOD
Research Participants (RPs)
A total of 316 Texas A&M University undergraduates (126 males)
volunteered to participate in exchange for partial fulfillment of an
introductory psychology course requirement. All RPs completed both the
Goldberg (1992) unipolar markers and the Big Five Inventory (BFI; John &
Srivastava, 1999) in addition to the SDR measures. Agreeableness, as
measured by the two instruments, correlated .77. Due to space limitations,
unless stated otherwise, only results for the Goldberg adjective markers are
reported here. Following McCrae and John (1992), the Big Five dimensions
are labeled Extraversion, Agreeableness, Conscientious-ness, Emotional
Stability, and Openness (Goldberg s Intellect) for ease of exposition.
RESULTS AND DISCUSSION
Zero-order correlations among the measures are reported in Table 1.
The overall pattern suggests that Agreeableness is correlated with
several SDR measures, notably the impression-management aspect of
SDR. The IM measure had correlations comparable in strength for
Agreeableness, Conscientiousness, and Emotional Stability. Agree-
ableness is related to a lesser extent to self-deception (.07) and self-
monitoring ( .11). At the same time, two of the other Big Five
dimensions (Emotional Stability and Conscientiousness) show more
significant SDR correlations overall, with similar weak links to self-
monitoring. Furthermore, the self-deception aspect of SDR almost
completely bypasses Agreeableness, appearing again in Emotional
Stability (.43), Openness (.22), Extraversion (.21), and Conscientious-
ness (.17). The pattern of correlations with self-monitoring is different
from the one found for IM and SDE, restricted largely to the Big Five
dimension of Extraversion.
The outcomes for Study 1 suggest that if Agreeableness is related to
SDR, then (a) the IM aspect may be more central than SDE or SM; (b)
among the Big Five dimensions, Agreeableness may not represent a
unique problem of socially desirable responding. Emotional Stability
and Conscientiousness may be even more worthy of special attention
than Agreeableness. Prescriptively, these data suggest that if concerns
about Agreeableness as an artifact are warranted, they should be
focused on the impression-management aspect of SDR, and its
underlying processes.
Table 1
Intercorrelations and Cronbach Alphas of Big Five Dimensions and SDR Measures (Study 1)
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
(N = 316)
1. Extraversion .89
2. Agreeableness .36* .93
3. Conscientiousness .23* .56* .87
4. Emotional Stability .30* .34* .30* .85
5. Openness .46* .55* .49* .35* .88
6. Sex .21* .15* .08 .10 .13* 
7. Impression Management .00 .23* .22* .25* .12* .04 .76
8. Self-Deception .21* .07 .17* .43* .22* .10 .35* .67
9. Self-Monitoring (SM) .17* .11* .14* .07 .09 .10 .39* .16* .74
10. SM Acting .21* .12* .12* .07 .20* .14* .21* .14* .68* .66
11. SM Extraversion .47* .06 .04 .11* .19* .04 .12* .18* .60* .55* .71
12. SM Other Directedness .05 .09 .11 .16* .04 .00 .36* .36* .80* .26* .20* .67
Goldberg (1992) markers provided the data for the five-factor personality measures.
Cronbach alphas appear on the diagonal.
For sex, M = 1, F = 2.
* p < .05.
Agreeableness & Social Desirability 703
Study 1 has several potential limitations. First, the outcomes are
based on traditional verbal self-report measures. How these responses
relate to overt behavior in situ, however, remains an empirical
question. A second, related, limitation is that Study 1 treated RPs as
Leibnitzian windowless monads; that is, individuals were assessed in
social isolation. Some aspects of SDR may be entirely intrapsychic,
but other aspects may be emergent, as a result of interpersonal
interaction. Paulhus (1991) describes impression management as
  self-presentation tailored to an audience  (p. 21), noting that   some
subjects are purposefully tailoring their answer to create the most
positive social image  (p. 21). In emphasizing a tailoring metaphor,
Paulhus implies that impression management is a custom-made local
adjustment and is directed to a particular situation, as opposed to an off-
the-rack approach that is designed to fit more generally (self-
deception?). In a subsequent, more refined account, Paulhus and John
(1998) suggest that certain situational pressures and threats may activate
the SDR motives, which otherwise might be dormant in a conventional
paper-and-pencil assessment. Perceived threats to an individual s
agentic or communal worth may differentially activate alpha and
gamma biases in SDR. Such threats may appear typically during
interpersonal interaction. Furthermore, given the interpersonal nature of
Agreeableness (Graziano & Eisenberg, 1997; Jensen-Campbell &
Graziano, 2001), interpersonal situations may be the better forum for
examining SDR links to Agreeableness than isolated situations. To
overcome some of these limitations, and to gain a better picture of
Agreeableness as a potentially unique artifact of SDR and of
impression management in particular, we designed two further
experimental studies that included manipulations to elicit different
kinds of social desirability motivation.
STUDY 2
METHOD
Research Participants (RPs)
Sample size was determined by power calculations (Cohen, 1987) for zero-
order correlations, assuming power of .90, a two-tailed alpha of .05, and a
medium effect size (q = .30). A total of 348 introductory psychology students
(113 males) participated in exchange for partial fulfillment of a course
requirement.
704 Graziano & Tobin
Procedure
Upon arrival, RPs reported their self-perceptions of Agreeableness and
Extraversion, using the 100 Goldberg (1992) self-report markers. Given the
time-consuming nature of the round-robin procedures, we assessed only A and
E, the two Big Five dimensions most immediately relevant to the manipulation
and to SDR theoretically (Paulhus & John, 1998). To measure SDR, we asked
RPs also to complete the Balanced Inventory of Desirable Responding (BIDR;
Paulhus, 1988) and the 25-item Self-Monitoring scale (Snyder, 1987).
Following the completion of the self-report measures, RPs were randomly
assigned to one of three instructional conditions. Specifically, in the
desirable-to-be-Agreeable instructional condition, RPs were given the
following instructions:
As you probably already know, for a group to be successful, its members must
get along. That is, they must compromise and coordinate their positions, help
each other complete tasks, and be reasonable with one another. When
members of groups cannot work together, groups as a whole often fail.
In contrast, RPs in the undesirable-to-be-Agreeable instructional condition
were told the following:
As you probably already know, for a group to be successful, its members
must pull their fair share of the load. That is, individuals must not just lay
back, let others take control, and make poor decisions. When individuals
are more interested in getting along with each other than in getting the task
done right, groups often fail, and disasters can happen. Psychologists call
this kind of conformity,  Group Think. You may have heard that this is
what happened when the space shuttle exploded. A tragedy resulted
because individuals failed to voice their concerns.
RPs in the control condition were given no specific instructions regarding
Agreeableness; they were simply told that they were to complete a group
decision-making task.
Following the instructional manipulation, triads composed of one
participant from each instructional condition were formed. That is, each
discussion triad was heterogeneous for the instructional manipulation. Each
triad performed a group decision-making task (selecting items for survival
following a plane crash in the mountains). The decision-making activity was
used as an opportunity for each member of the triad to have exposure to a
behavioral sample of the other members of the group. To make salient the
appropriate instruction-consistent behavior, RPs were told that they were
being evaluated in regard to future success by experimenters in the room,
although no behavioral data were recorded. Following the group decision-
making task, RPs were separated and instructed to rate the other members of
their triad on the Agreeableness and Extraversion scales of the Big Five.
Agreeableness & Social Desirability 705
Once everyone was finished with the questionnaires, the experimenter
conducted the funnel debriefing (Aronson & Carlsmith, 1968).
RESULTS
Correlations between SDR and the Big Five dimensions of Agree-
ableness and Extraversion collapsing across experimental conditions
are shown in Table 2.
Hypothesis Testing
To test the hypothesis that ratings of Agreeable-ness varied by
experimental condition, we conducted two one-way ANOVAs with
three levels. There was no evidence of a significant effect for
experimental condition on either self- or other-rated (target) Agree-
ableness (both Fs < 2.00, ns). These results indicate that individuals
did not see themselves, nor did others see them, as behaving
differently by experimental condition. We also examined whether
Table 2
Intercorrelations of and Cronbach Alphas for Extraversion,
Agreeableness, Sex, and SDR Measures (Study 2)
1. 2. 3. 4. 5. 6. 7. 8. 9.
(N = 351)
1. Extraversion .87
2. Agreeableness .07 .84
3. Sex .06 .16* 
4. Impression
Management .07* .35* .19* .68
5. Self-Deception .20* .20* .09 .36* .63
6. Self-Monitoring (SM) .26* .18* .09 .47* .30* .78
7. SM Acting .34* .17* .19* .35* .06 .74* .73
8. SM Extraversion .58* .06 .10 .14* .14* .55* .42* .69
9. SM Other
Directedness .02 .20* .01 .45* .46* .83* .40* .22* .66
Goldberg (1992) markers provided the data for Extraversion and Agreeableness.
Cronbach alphas appear on the diagonal.
For sex, male = 1, female = 2.
For correlations involving sex, n = 333.
* p < .05.
706 Graziano & Tobin
persons higher in SDR (especially IM) were more responsive than
their peers to the experimental manipulations using centered cross-
product regression (Aiken & West, 1991). We found that adding the
cross-product terms to the model did not explain significantly more
variance, DR2 = .01, F (2, 341) = 1.70, ns. None of the effects for any
form of SDR (IM, SD, or SM), or their cross products were significant
(all Fs < 2.00, ns).
When a corresponding set of analyses was run with other-rated Big
Five Extraversion (replacing other-rated Agreeableness) as the
criterion, the Experimental Condition effect was significant,
DR2 = .02, F (2, 343) = 3.02, p < .05. The manipulation of the social
desirability of Agreeableness influenced other-rated Extraversion such
that persons assigned to the Bad condition were rated as significantly
more extraverted (M = 1.46, SD = 1.18) than persons assigned to the
Good condition (M = 1.09, SD = 1.11), h2 = .03, t(229) = 2.44, p < .01.
Apparently, when persons were told it was undesirable to be
Agreeable, they responded by being observably more extraverted,
but not less observably Agreeable. The mean difference was smaller
between Bad and Control (.11) than Good and Control (.26), but
neither experimental group differed significantly from the Control
(M = 1.35, SD = 1.18), h2-values = .00, t-values < 1.68, ns. There was
no evidence that the experimental manipulation influenced self-rating
of Extraversion, h2 = .00, F (2, 342) = 0.07, ns.
Another related way to examine differential responsiveness is
through Self-Criterion Residual Analyses (SCR). Paulhus and John
(1998) presented the SCR, a bias index that was calculated as the
residual variance that remained when a self-report measure was
regressed on the corresponding criterion measure for the same
variable. Paulhus and John give an example that explicitly involves
Agreeableness:   For example, scores on a self-reported Agreeable-
ness scale are regressed on a set of peer ratings of the same items and
the self-reported residual is isolated as a separate variable. Because all
the self-reported variance shared with the peer rating has been
removed, the residual represents only self-report inflation: High
scores indicate overclaiming Agreeableness (and low scores indicate
underclaiming) relative to the peer rated criterion. These self-favoring
bias scores can then be correlated with any other variable in the data
set  (p. 1032).
There was no significant effect for the IM experimental
manipulation cross product, DR2 = .01, F (2, 341) = 1.80, ns, for the
Agreeableness & Social Desirability 707
manipulation main effect, for any of the SDR measures, or for their
cross products (all DR2-values = .00, all Fs < 2.00, ns). Collapsing
across the experimental conditions, the zero-order correlations with
the SCR measure were .03 (Impression Management), .03 (Self-
Deception), and .06 (Self-Monitoring), none of which were significant.
It is possible, however, that these results support an alternative
interpretation of the SCR. That is, SCR may reflect not just self-
favoring biases, but other processes as well (e.g., error in ability to rate
accurately an individual following brief acquaintanceship).
Social Relations Model Analysis
Another way to examine the relations among these variables
involves decomposing self- and other data into perceiver and target
effects using a Social Relations Model Analysis (SRM; Kenny,
1994). These effects then may be related to SDR concern measures.
According to the SRM, the perceiver effect measures an individual s
tendency to see others similarly on a dimension. Perceiver effects
accounted for 41.10% of the variance in ratings of others
Agreeableness. This large positive perceiver effect indicates that
individuals   project  Agreeableness onto others. The link between
self-presentation and perceiver effects examines whether individuals
higher in self-presentation generally see others as highly agreeable.
None of the correlations between IM, SD, SM (and its components)
with perceiver effects in Agreeableness were significant (r-values
ranged from .13 to .02). There was no evidence that individuals
who rated themselves as higher on these SDR concern measures
tended to rate all others as agreeable.
The target effect indicates how much raters agree in rating targets
on a trait. Thus, a person with a large negative target effect for
Agreeableness is seen by others as low in Agreeableness. Target
effects accounted for 15.20% of this variance. Thus, approximately
2.7 times as much variance in ratings of others Agreeableness was a
function of the rater than of the target. The significant target effect,
however, indicated that there was some convergence on which
individuals in the group were seen as high (or low) in Agreeable-
ness. Furthermore, if self-presentation is positively related to target
effects in Agreeableness, then individuals who are more concerned
with self-presentation are generally seen by others as more
Agreeable. A marginally significant correlation was found between
708 Graziano & Tobin
SM and target effects in Agreeableness, r = .20, p < .10. Thus,
there was some evidence that their partners saw individuals high in
SM as less Agreeable. There was no evidence of a similar effect for
IM, r = .11, ns.
Additional findings from the SRM analysis included a significant
self-perceiver effect for Agreeableness, r = .28, p < .01. This finding
indicated that individuals who saw themselves as Agreeable also saw
others as Agreeable. The marginally significant correlation between
perceiver and target effects in Agreeableness indicated that individuals
who saw everyone as Agreeable were also seen by everyone as
Agreeable, r = .30, p < .10. However, the correlation between self-
ratings and target effects in Agreeableness indicated that individuals
who saw themselves as Agreeable were not necessarily seen by others
as Agreeable, r = .10, ns.
A similar SRM analysis (Kenny, 1996) partitioned Big Five
Extraversion ratings into perceiver and target effects. These effects
were then correlated with the three SDR concern measures. Perceiver
effects accounted for 17.60% of the variance in ratings of others
Extraversion. Target effects accounted for 46.10% of this variance.
Thus, approximately 2.62 times as much variance in ratings of
others Extraversion was a function of the target than of the rater.
Consistent with past research (e.g., Albright, Kenny, & Malloy,
1988), these findings suggest that greater consensus is found in the
rating of group members Extraversion than in the rating of group
members Agreeableness.
Of the correlations between IM, SD, SM (and its components) with
SRM perceiver effects in Extraversion, only the Extraversion
component of SM was significant, r = .28, p < .01. Individuals
who rated themselves as higher on this measure tended to rate all
others as more extraverted. A significant correlation also was found
between the Extraversion component of SM and target effects in Big
Five Extraversion, r = .23, p < .01. A marginally significant
correlation was found between SM and target effects of Extraversion,
r = .13, p < .10. Thus, there is some evidence that individuals higher
in SM were seen as more extraverted by their partners. There was no
evidence of similar perceiver or target effects for IM, (all rs < .08,
ns). A significant correlation also was found between perceiver
effects of Big Five Extraversion and self-ratings of Agreeableness,
r = .35, p < 01. Individuals higher in Agreeableness tended to rate
others as more extraverted.
Agreeableness & Social Desirability 709
DISCUSSION
Like Study 1, results from Study 2 suggest that Agreeableness has a
complex relation with various forms of SDR concerns. In self-ratings
(S-data) at the zero-order level, Agreeableness is positively but weakly
related to both IM and SDE. Agreeableness is negatively related to
Snyder s SM measure of impression management. In rating the
Agreeableness of others (R-data), the refined SRM analysis showed
that more variance in ratings of Agreeableness was due to perceivers
than to targets, further undermining the artifact explanation. That is,
when social desirability is used to explain Agreeableness effects, the
usual assumption is that the social desirability is located in the actor,
not in the perceiver of action. It is possible, of course, that social
desirability constrains what some perceivers are willing to say about
others, but conceptually, this would seem to be a different form of
SDR, and none of the SDR concern measures were related to perceiver
effects. Moreover, significant target variance was present, suggesting
some convergence in evaluating peer Agreeableness among raters.
One of the SDR concern measures (SM) was related to target variance
of Agreeableness, but in an inverse and weak link. There was no
evidence that SDR concerns were related to differential responsiveness
to instructional conditions to be or not to be Agreeable. Indeed, it
could be argued that despite the relatively large sample and associated
high statistical power, and despite the debriefing checks on it, our
manipulation of the social desirability of Agreeableness was
ineffective in Study 2. In partial rebuttal, we note that the experimental
manipulations were potent enough to influence other ratings of
Extraversion.
Taken together, Studies 1 and 2 suggest that if Agreeableness is
related systematically to SDR concerns, the link is complex.
Consistent with past research (Graziano et al., 1996), the round-robin
and SRM results from Study 2 suggest that Agreeableness differences
can be detected even among minimally acquainted strangers. Both
Agreeableness and Extraversion seem to be related to how individuals
chronically view themselves and others. Both also appear to be related
to how people are seen by others. The pattern for Agreeableness is not
the same as the pattern for Extraversion, however, suggesting
discriminant validity, especially in ratings by others. We infer, then,
that our evidence so far points away from the artifact explanation for
Agreeableness differences, especially the ones rated by others.
710 Graziano & Tobin
There are, however, at least five potential limitations to Study 2.
First, the apparent stability of Agreeableness in the face of
experimental intervention may be an artifact of a weak manipulation
to make Agreeableness desirable or undesirable. All RPs in Study 2
were asked about the effectiveness of the manipulation as part of our
post-experimental debriefing, but the experimental manipulation still
may not have been effective. During face-to-face debriefing discus-
sions with an experimenter, RPs may have been reluctant to admit
publicly that they did not recognize the manipulation. The effective-
ness of the manipulation for each participant should be checked more
intensively, using explicit, quantitative methods. It is possible that a
more a focused, private manipulation check may offer a more accurate
assessment of the differential perceptions of the manipulations.
Second, RPs were assigned to completely heterogeneous discussion
groups following the experimental manipulations. That is, each
discussion group had a mix of persons who had been told nothing
(control), or that it was desirable or undesirable to be Agreeable. In
such a group mix, the distinctive manipulations may have been
subverted by the more concrete, pressing need to adjust to the other
individuals within the discussion groups.
Third, RPs had little opportunity to become acquainted before
making their assessments of each other. The interaction context was
constrained and perhaps left insufficient opportunity for expression of
individual differences. Had RPs been given the chance to discuss
issues freely, especially issues that might reveal underlying differences
in Agreeableness, then links to SDR might have been clearer.
Fourth, no preintervention assessment of Agreeableness was
collected in Study 2. Following the traditional logic of randomized
experimentation, we assumed that because RPs had been randomly
assigned to conditions, then mean Agreeableness levels were
comparable in all three conditions prior to the intervention. The
assumption, then, was that if no differences were found among
experimental conditions, given an effective manipulation and adequate
statistical power to detect such differences, then Agreeableness was
resistant to interventions. This is a long assumptive chain that may not
be warranted.
A final limitation is that both Studies 1 and 2 concentrated on the
relations among Agreeableness and different kinds of SDR as
predictors, rather than on the predictor-outcome relation. In effect,
Studies 1 and 2 focused on the relations among predictors, with one of
Agreeableness & Social Desirability 711
the predictors, Agreeableness, serving as a kind of outcome variable.
No external behavioral outcome empirically associated with Agree-
ableness (e.g., choice of conflict tactics) was measured in either
Studies 1 or 2. That is, given that Agreeableness has been shown to be
a predictor of interpersonal relationship outcomes in previous studies,
we might attempt to improve prediction if we could (a) identify
alternative SDR correlates of Agreeableness that might be more
plausible as potential predictors (Study 1); or (b) identify situational
forces that might increase or decrease scores on measures of
Agreeableness (Study 2); or (c) show how the external criteria
(outcomes) themselves might be affected by SDR concerns, thereby
altering the Agreeableness-outcome link.
STUDY 3
Study 3 examined the generality of Studies 1 and 2, with an eye to
overcoming their potential limitations. In Study 3, RPs were assigned
to discussion groups. They were given a conflict tactics task that would
be potentially diagnostic of Agreeableness differences (Graziano et al.,
1996; Jensen-Campbell & Graziano, 2001). They also were given
enough discussion time to express these tendencies freely. In terms of
the experimental manipulations, RPs were assigned to one of three
homogeneous discussion groups (Good-to-be-Agreeable, Bad-to-be-
Agreeable, Control). Both pre- and postintervention measures
of Agreeableness were collected to obtain a more precise estimate of
individual change following the manipulations. A focused set of
quantitative manipulation checks for the experimental intervention was
collected from each participant.
METHOD
Research Participants (RPs)
As in the previous studies, sample size was determined by power calculations,
assuming a medium effect size, two-tailed alpha of .05, and power of .90. A
total of 312 RPs (150 males) took part in this study in exchange for partial
fulfillment of a course requirement. RPs were drawn from a population of
more than 2000 students who were assessed previously on the Goldberg
(1992) self-report markers for the five-factor model. From the full distribution
of scores from group testing, only RPs who fell within the top or bottom third
of the Agreeableness dimension distribution participated.
712 Graziano & Tobin
Group Composition
RPs were scheduled, in groups of three (triads), into one of two different
heterogeneous group compositions (LLH or HHL), based on their scores on
Agreeableness (H = top 33%, L = bottom 33%). The purpose of
heterogeneous composition of groups was to increase chances that persons
in each triad would be exposed to samples of behavior characteristic of both
low and high Agreeable individuals.
Procedures
Upon arrival, RPs were given a number badge and assigned to a seat within a
group of three RPs. The number badges helped each member identify each
other member of their triad uniquely. RPs were informed that they would be
reading several stories, independently rating the quality of possible solutions to
the conflict in the stories, discussing these solutions in their groups for
5 minutes, and re-rating the solutions independently. The   stories  were
actually vignettes involving dyadic interpersonal conflict, used in previous re-
search (Graziano et al., 1996). The vignettes presented a set of prototypic
dyadic conflict, and allowed for a diverse range of 11 conflict resolution options.
RPs were told that their job was to tell us how good or bad each choice was
in solving conflicts with others. RPs were informed that there are many ways
to end a conflict and several ways may be equally good or bad in a given
situation. They were told that their answers were confidential. After being
shown how to mark their evaluations of the solutions by drawing a single line
along the continuum, RPs were given a practice story with which to
familiarize themselves with the experimental procedure. (After RPs left the
experimental session, these continua were converted to Likert-type scales
anchored at 1 = Poor Choice and 9 = Excellent Choice by a research assistant.)
Following completion of the practice vignette, groups were randomly
assigned to one of three different social desirability conditions: Bad-to-be-
Agreeable, Good-to-be-Agreeable, and control. The manipulations were
given in both written and spoken formats. First, a written explanation for the
study (actually, the manipulation) was distributed. Then a summary statement
of this document was read aloud to the RPs. In the Bad condition, RPs were
told the following summary statement verbatim:   We ask you to keep in
mind that it is more important to present your individual perspective than it is
to get along with other group members,  whereas RPs in the Good condition
were told,   We ask you to keep in mind that it is more important to cooperate
and coordinate your activity with other group members than it is to present
your individual perspective.  RPs in the Control condition were told,   We
ask you to keep in mind these issues as you deal with the task and the other
group members. 
Agreeableness & Social Desirability 713
After the experimental manipulation, RPs were given the first of two
experimental conflict vignettes, and rated the merits of each of the 11
resolution tactics, in order. The tactics included physical action, threats,
criticism, submission, walking away, and third-party mediation. After pre-
ratings were collected, RPs had 5-minute group discussions about each of the
tactics. They also were encouraged to discuss their own past experiences with
interpersonal conflicts. RPs then re-rated each of the conflict tactics privately
and independently. The procedure for the second vignette was identical to
that used for the first vignette. The order of vignette presentation was
completely counterbalanced across conditions.
RPs then were seated in new triads at separate tables so that no two
members of any previous group were together. RPs confidentially completed
the Goldberg (1992) personality markers (for self-ratings) and the Goldberg
(1992) Agreeableness markers (for the other-ratings). After experimenters
collected the personality measures, RPs completed a seven-item manipulation
check before the experimenter conducted a funnel debriefing (Aronson &
Carlsmith, 1968).
RESULTS
Manipulation Checks
First, the seven-item manipulation checks were analyzed in a single-
factor, three-level (Good-to-be-Agreeable, Bad-to-be-Agreeable, Con-
trol), multivariate analysis of variance (MANOVA), followed by
separate univariate ANOVAs. The multi-variate experimental manip-
ulation effect was significant, h2 = .31, F (14, 584) = 18.56, p < .01.
(Means for all of the manipulation check items are presented in
Table 3.) Three of the seven items focused on the specific impact of the
manipulations, and all three of these were significant in the univariate
ANOVAs, with group mean differences also statistically significant
and in the appropriate direction. The first item,   In this study, I was
encouraged to get along with other members of my discussion group, 
was rated significantly different in the three experimental groups,
h2 = .25, F (2, 297) = 49.30, p < .01. Likewise, the second item,   In this
study I was encouraged to present my individual perspective rather than
to get along with other group members,  was rated significantly
different in the three experimental groups, h2 = .45, F (2, 297) = 119.51,
p < .01. Similarly, the third item,   In this study, it was important to get
along with others in my group,  was rated significantly different in the
three experimental groups, h2 = .21, F (2, 297) = 39.72, p < .01. Results
from these manipulation check items were in the expected direction.
714 Graziano & Tobin
Table 3
Means and Standard Deviations for Manipulation
Check Items (Study 3)
Good Bad Control Overall
M (SD)
(n = 108) (n = 111) (n = 93) (N = 312)
1. In this study, I was 4.10a 2.53b 3.30c 3.30
encouraged to get along (0.98) (1.25) (1.28) (1.34)
with the other members of
my discussion group.
2. In this study, I was 2.49a 4.54b 4.06c 3.68
encouraged to present my (1.27) (0.69) (0.95) (1.34)
individual perspective,
rather than to get along
with other group members.
3. The information I received 2.59a 2.40a 2.32a 2.44
after the practice situation (1.11) (1.04) (1.15) (1.10)
influenced my behavior in
this study.
4. The information I received 2.01ab 1.87a 2.23b 2.03
after the practice situation (0.98) (0.97) (1.10) (1.02)
influenced my ratings of
myself.
5. The information I received 2.57ab 2.23a 2.75b 2.51
after the practice situation (1.21) (1.24) (1.31) (1.27)
influenced my ratings of
the other individuals in my
group.
6. In this study, it was 3.51a 2.14b 2.98c 2.87
important to get along with (1.18) (1.06) (1.23) (1.29)
others in my group.
7. I was able to form an initial 3.65a 3.63a 3.68a 3.65
impression of the (1.12) (1.01) (1.21) (1.11)
individuals in my group
based on our discussions.
Note: Means in the same row that do not share subscripts differ at p < .05.
From the same set of manipulation checks, there was no evidence that
the manipulations selectively influenced perceivers confidence in their
ability to rate other group members, as indicated by the item   I was
Agreeableness & Social Desirability 715
able to form an impression of the individuals in my group based on our
discussions,  h2 = .01, F (2, 297) = 1.74, ns.
Social Relations Analyses of Self and Others
As in Study 2, a Social Relations analysis (Kenny, 1996) was used to
partition Agreeableness ratings into perceiver and target effects. These
effects were then correlated with the remaining Big Five personality
dimensions. According to the SRM variance decomposition, perceiver
effects accounted for 64.80% of the variance in ratings of others
Agreeableness. Target effects accounted for 15.70% of this variance.
Thus, approximately 4.13 times as much variance in ratings of others
Agreeableness was a function of the rater than of the target. As in
Study 2, the significant target effect, however, indicated that there was
some convergence as to which individuals in the group were seen as
high in Agreeableness and which were seen as low in Agreeableness.
Additional findings from the SRM analysis included a significant
self-perceiver effect for Agreeableness, r = .43, p < .01. This finding
indicated that individuals who saw themselves as Agreeable also saw
others as Agreeable. The significant correlation between perceiver and
target effects in Agreeableness indicated that individuals who saw
everyone as Agreeable were also seen by everyone as Agreeable,
r = .34, p < .05. However, the correlation between self-ratings and
target effects in Agreeableness indicated that individuals who saw
themselves as Agreeable were not necessarily seen by others as
Agreeable, r = .18, ns.
Tests of Hypotheses
We tested our hypotheses several ways. First, the mean self-ratings
of Agreeableness were compared across experimental conditions
following the manipulations. The manipu-lation influenced mean
self-rating of Agreeableness across conditions, h2 = .03, F (2, 309) =
3.10, p < .05, but not in the expected direction. RPs, told it was Bad-
to-be-Agreeable (M = 2.51, SD = 1.10), rated themselves signifi-
cantly higher in Agreeableness than did their peers in the Control
condition (M = 2.09, SD = 1.35). RPs ratings in the Good condition
(M = 2.24, SD = 1.25) did not differ from those in either the Control
condition or the Bad condition. A second approach was to compare
the stability of correlations in self-rated Agreeableness, from
716 Graziano & Tobin
prescreening to postmanipulation self-rating. If the experimental
manipulations were successful, then the correlation would be larger
(i.e., more stable) in the control condition than in either of the two
conditions in which Agreeableness was described as good or bad. In
all three conditions the correlations were statistically significant
(overall r [310] = .82), but there was no evidence that the
correlations were different across conditions (the correlations for
Agreeableness were .83, .82, and .82 for Bad, Good, and Control,
respectively, all p-values < .01). The internal consistency reliabilities
of the scores for the 10-item Goldberg (1992) Agreeableness markers
prior to the manipulation and following the manipulation were .95
and .93, respectively. The corresponding reliabilities for Extraversion
scores were .90 (prior) and .90 (following manipulation).
A third approach is more idiographic and social-cognitive and
focuses on the correlation between Agreeableness and the manipulation
checks. That is, if Agreeableness is driven primarily by SDR
motivation, and if people also differ in their perception of the
requirements for being socially desirable in a given situation (as
measured by the manipulation checks), then Agreeableness should be
correlated within cells to the manipulation checks. When it is Good to
be Agreeable, Agreeableness should be correlated positively with the
corresponding check, whereas, when it is Bad to be Agreeable,
Agreeableness should be correlated inversely with the corresponding
check. Agreeableness did not correlate significantly within cells (or
between cells) with the specific manipulation checks (r-values < .10.)
Only one manipulation check item correlated significantly with
Agreeableness (r [310] = .13, p < .05):   In this study, it was important
to get along with others in my group.  Taken together, these outcomes
suggest that (a) the manipulation of the desirability of Agreeableness
was completed successfully, as shown by the manipulation checks, but
there was no evidence that (b) the manipulation systematically
influenced self-ratings of Agreeableness, either as postmanipulation
means or as changes in stability of self-rated Agreeableness over time.
Third, if Agreeableness were more closely associated with
impression management than with self-deception (as found in Studies
1 and 2), then manipulating the social desirability of Agreeableness
would have more influence on ratings of Agreeableness by other group
members than on self-ratings. The experimental manipulation effect
was significant for other-rating of Agreeableness, h2 = .03, F (2,309) =
3.34, p < .05. Post hoc comparisons revealed that persons assigned to
Agreeableness & Social Desirability 717
the Bad condition (M = 2.60, SD = 1.07) were rated by others as
significantly more Agreeable than were persons assigned to the Good
condition (M = 2.32, SD = 0.86), h2 = .02, t (210) = 2.11, p < .05.
Persons assigned to the Control condition (M = 2.29, SD = 0.90) were
seen as less Agreeable than persons assigned to the Bad condition,
h2 = .02, t (202) = 2.22, p < .05. There was no evidence that persons
assigned to the Control conditions were seen as different from those
assigned to the Good condition, h2 = .00, t (192) = 0.25, ns.2
As in Study 2, we conducted a Self-Criterion Residual (SCR)
analysis. The effect for the experimental manipulation was significant,
h2 = .02, F (2, 309) = 3.12, p < .05. Contrary to the bias hypothesis, the
SCR residual scores were significantly higher for the Bad-to-be-
Agreeable condition (M = .18, SD = 1.05) than for the Good condition
(M = .09, SD = .86), h2 = .02, t (211) = 2.07, p < .03. The Control
2. Study 3 was different from Study 2 in one important procedural detail. In Study 3,
participants were assigned to triads, which received a common experimental
manipulation, and within which RPs also conducted peer ratings and round-robin
assessments. (In Study 2, experimental manipulations were independent of triads
because triads were constructed from individuals receiving different experimental
manipulations.) Thus, in Study 3 the ratings of other individuals within triads were
potentially dependent on each other. We used intraclass correlation (ICC) procedures
recommended by Kashy and Kenny (2000) to assess the degree of group dependence
and to adjust significance tests accordingly for its occurrence. Results indicate that in
Study 3, ratings of other members Agreeableness within triads were significantly
related, h2 = .40, F(101,208) = 1.35, p < .05, with an overall ICC of .19. Furthermore,
there was evidence that triadic dependence was differentially related to the
experimental manipulation, h2 = .03, F(2,208) = 2.87, p < .06. Recall that prior to
any manipulation, triads were purposefully constructed to be heterogeneous on
pretested Agreeableness (e.g., HHL), so in the absence of a manipulation effect, ICC
values of self-ratings would be negative. Consistent with this hypothesis, the ICC for
other ratings in the Control condition following the manipulation was .12. The
corresponding postmanipulation ICC values were .16 and .45, for the Good and Bad
conditions respectively. These outcomes show differential homogenization of triads
within conditions, with the Bad condition ratings the most homogenous. Despite the
differential homogenization the evidence for mean level of other-rated Agreeableness
(as well as the SRM analyses) was not consistent with a social desirability bias
interpretation (see p. 693). Nevertheless, to address the possibility of group
dependence bias in tests of hypotheses analyzed at the individual level, we tested
each of our hypotheses at the group level, as recommended by Kashy and Kenny.
Results of the analyses at the group level followed a pattern virtually identical to those
reported at the individual level. Information about the group dependency analyses is
available from the authors on request.
718 Graziano & Tobin
condition (M = .11, SD = .92) differed from the Bad condition,
h2 = .02, t (202) = 2.13, p < .03, but not from the Good condition,
h2 = .00, t (191) = .19, ns.
Interpersonal Conflict
The first ratings of the conflict vignettes (prior to any peer group
discussion) on the 11 conflict resolution tactics were subjected to
confirmatory factor analysis. Three factors were extracted, replicating
the three reported previously in Graziano et al. (1996). They accounted
for 62.77% of the variance. Based on these factor analyses, each RP s
ratings were reduced to three separate aggregated scores for Power
Assertion, Disengagement, and Negotia-tion. These three scores were
then used as dependent variables in subsequent analyses.
First, we report general trends in rating the goodness/efficacy of
conflict tactics, using multivariate analysis of variance. In these
analyses, we used RPs Agreeableness scores from prescreening, not
Agreeableness scores collected during the experimental session. There
was a significant main effect for Strategy, h2 = .49, F (2, 308) = 149.10,
p < .01. Among the three strategies, Negotiation received the highest
rating (M = 6.28, SD = 1.28), with Disengagement (M = 3.71, SD = 1.19)
and Power Assertion (M = 3.21, SD = 1.29) rated significantly lower.
This main effect was qualified by a Strategy Agreeableness
interaction, h2 = .03, F (2, 308) = 5.50, p < .05. Persons high in
Agreeableness gave higher ratings to Negotiation (M = 6.47, SD = 1.26)
and Disengagement (M = 3.91, SD = 1.15) than did persons low in
Agreeableness (M = 6.17, SD = 1.14, and M = 3.62, SD = 1.24, for
Negotiation and Disengagement, respectively). In addition, persons
high in Agreeableness rated Power Assertion as less good (M = 3.04,
SD = 1.34) than did persons low in Agreeableness (M = 3.41, SD = 1.29).
Next, we evaluated the hypothesis that the link between Agree-
ableness and conflict tactics was qualified by the experimental
manipulation. There was one significant interaction involving the
experimental manipulation of the social desirability of Agreeableness:
Time Manipulation Agreeableness, h2 = .01, F (1, 305) = 3.16,
p < .05. Disaggregation of the interaction, using simple slopes analysis
(Aiken & West, 1991), showed that persons high in Agreeableness
changed their ratings of conflict tactics more from Time 1 to Time 2 in
the Control condition, but less in either Good or Bad to be Agreeable
conditions, than did persons low in Agreeableness. That is, within the
Agreeableness & Social Desirability 719
general trend for RPs to change over time toward less positive ratings
of Power Assertion and Disengagement tactics, the high-Agreeable
persons changed more in a Control condition, and less when given any
specific manipulation for the direction of the social desirability of
Agreeableness. None of the other effects involving the experimental
manipulation was significant. Despite the apparent effects of the
manipulation as seen in the manipulation checks and on the evaluation
of conflict tactics, there was no evidence that the link between
dispositional Agreeableness and conflict tactics was qualified by the
experimental manipulation (e.g., the Manipulation Strategy
Agreeableness interaction yielded h2 = .01, F (2, 305) = 1.72, ns).
GENERAL DISCUSSION
Social desirability confounds and self-favoring biases have been a
concern of personality researchers from the founding of the modern
field of personality. Allport and Odbert (1936) provided one of the
first systematic attempts to isolate the dross of SDR from the ore of
personality attributes. Later foundational work in personality assess-
ment focused on systematically separating response styles from
substantive aspects of personality, per se (e.g., Jackson, 1971; Jackson
& Messick, 1962). Thus, there is a precedent for being concerned
about an aspect of personality that contains social evaluative terms like
  kind,    warm,  and   considerate.  At least in theory, there are
reasons to suspect that Agreeableness, one of the largest dimensions of
personality, might be deflated if socially desirable responding could be
partitioned from it. Agreeableness may appear to be a large dimension
of personality because self-favoring biases contaminate it. Perhaps
more than any of the other Big Five, Agreeableness involves socially
desirable qualities (e.g., Hafdahl et al., 2000; Paulhus et al., 1995), so
persons motivated to look good in the eyes of others may be especially
likely to describe themselves as being Agreeable. In this construal, the
motivation underlying self-favoring biases, not Agreeableness, is the
  active ingredient  in predicting and explaining external criteria like
cooperation or the efforts to control negative emotions.
Recent research and theory suggest, however, that what was once
simple is now more complicated.   Socially desirable responding  is
not a single process, but a loosely related federation of processes.
Paulhus (1984, 1986, 1991) provided evidence that at least two
different processes, impression management and self-deception, can
720 Graziano & Tobin
generate self-favoring biases. Furthermore, impression management
itself is probably not a single process or set of processes (Gangestad &
Snyder, 2000). Starting from the assumption that Agreeableness might
contain self-favoring elements, and that self-favoring biases can
appear in at least two (and probably three) forms, we addressed the
empirical link between Agreeableness and self-favoring biases with
the three studies presented here, using converging correlational and
experimental methods that included both self-ratings and observer
ratings of Agreeableness. Consistent with Paulhus and John (1998),
Study 1 found that self-reported Agreeableness was correlated with
SDR measures, but was more closely associated with impression
management than with self-deception. The link to impression
management was stronger for Emotional Stability and for Conscien-
tiousness, suggesting that Agreeableness was not unique among the
Big Five in its association to SDR.
Study 2 involved a   constructive replication  of Study 1. In this
regard, Lykken (1970) argues that   constructive replications,  which
focus on the specific hypothesis, offer stronger validity evidence than
do   operational replications,  which attempt to duplicate an empirical
outcome using the same experimental methods and sampling as the
original study. When a constructive replication does not corroborate an
original hypothesis, it suggests that the original results may be artifacts
of method variance or sample specific aspects of the original study. If
impression management is an important contributor to self-reported
Agreeableness, or if impression management motives influenced the
construction of an Agreeable image generated for the consumption of
others during interpersonal interaction (Paulhus, 1991), then it should
be possible to manipulate situations to elicit impression-management
motives. Study 2 examined the link between impression management
and Agreeableness using experimental manipulations. Replicating
Study 1, Study 2 found that self-reported Agreeableness was related to
self-reported impression management, but found no evidence that
Agreeableness could be manipulated through experimental inductions
or that persons high on Impression Management as a disposition were
differentially sensitive to the experimental manipulations. SRM
analyses showed that Agreeableness differences were tied more
closely to perceiver effects than to target effects, suggesting that
Agreeableness is related to the way people perceive others. (For
related discussions, see Hafdahl et al., 2000; Jensen-Campbell &
Graziano, 2001; Tobin et al., 2000.)
Agreeableness & Social Desirability 721
Study 3 replicated and extended Study 2, showing that Agreeable-
ness differences were tied more closely to perceiver effects than to
target effects. In addition, Study 3 replicated previous research
showing a link between Agreeableness and an external criterion,
conflict tactics (Graziano et al., 1996). Despite evidence that
experimental manipulations were perceived by RPs as intended and
had independent effects on the evaluation of conflict tactics, there was
no evidence that the manipulations altered self-reported Agreeableness
or the link between Agreeableness and conflict tactics. The
experimental manipulation did influence other-rated Agreeableness,
but in a pattern not consistent with the SDR bias hypothesis.
Some important issues and qualifications remain. First, outcomes are
generally consistent with those predicted by Paulhus and John (1998),
but the construct of socially desirable responding, and its components,
need further theoretical development. Impression management in
particular seems to be an elusive construct. Paulhus (1991) describes
impression management as   self-presentation tailored to an audience 
(p. 21), noting later that   some subjects are purposefully tailoring their
answers to create the most positive social image  (p. 21). Given this
definition and the usual desiderata of convergent validity, one might
expect that conceptually similar measures of impression management
would be related to each other and that IM concerns would be related to
sensitivity to situational demands. Yet in both Study 1 and Study 2, the
BIDR measure of impression management correlated negatively with
Self-Monitoring, which is presented as another measure of individual
differences in impression management motivation (Snyder, 1987,
pp. 7 13). Agreeableness correlated positively with IM (r [314] = .26),
but negatively with SM (r [314] = .11). Furthermore, IM and SM had
different patterns of correlation with the Big Five dimensions. When
the desirability of Agreeableness was manipulated experimentally, we
found no evidence that either IM or SM interacted with the
manipulation or that either was related to the self-presentation of
Agreeableness or to any other Big Five dimension. It could be argued
that IM and SM have important conceptual differences (Paulhus, 1984;
Gangestad & Snyder, 2000; Graziano & Bryant, 1998), or that one
measure or the other has psychometric deficits (Briggs & Cheek, 1986;
Finch & West, 1997), but the simplest overall interpretation is that
impression management is a heterogeneous set of processes, and, as
such, is unlikely to offer a single, systematic, plausible alternative
account for more reliable Agreeableness effects.
722 Graziano & Tobin
The self-favoring biases of the sort described by Paulhus and John
(1998) are dynamic and seem more likely to contaminate the
substantive structural dimension of Agreeableness during threats to
an individual s communal standing. If the experimental manipulations
in the present research did not threaten, or served only as recommended
guidelines for action, then self-favoring contamination may have been
minimal (Paulhus et al., 1995). Further research should examine the
influence of Communal and Agentic threat on self-favoring biases.
That being said, this interpretation suggests that previous empirical
findings for Agreeableness not involving threats to the individual s
communal standing were probably not artifacts of SDR.
Second, it could be argued that the present program of research did
not provide sufficient opportunity for the expression or detection of
self-favoring biases. In Studies 2 and 3, RPs had a limited time to
become acquainted, and no serious issues were at stake. Perhaps when
persons are better acquainted (e.g., spouses), or are dealing with matters
of importance, self-favoring biases may be easier to detect. In
particular, the Self-Criterion Residual index (Paulhus & John, 1998)
may require minimal levels of acquaintanceship between perceivers and
targets to provide meaningful measures of self-favoring biases. Future
research should explore Agreeableness and differences associated with
length of acquaintanceship between perceivers and targets (Colvin,
1993; Funder & Colvin, 1997; Kolar, Funder, & Colvin, 1996; Tobin
et al., 2000, Study 2; cf. Jensen-Campbell & Graziano, 2001).
Third, it could be argued that outcomes of the present program
of research are based on null outcomes. In both Studies 2 and 3,
there was no evidence that the experimental manipulations
influenced Agreeableness in self-ratings. Perhaps the manipulations
were too weak to be effective or in their implementation did not
address the core motives underlying SDR. The argument has some
merit, but there are several rebuttals to it. First the results were not
uniformly null. The experimental manipulations were potent enough
in Study 2 to influence other-rated Extraversion, and in Study 3 to
influence an Agreeableness-related behavior (conflict tactics), but
not self-rated Agreeableness, per se, or the Agreeableness-conflict
strategy link itself. Further, the experimental manipulation influ-
enced other-rated Agreeableness, but in a pattern inconsistent with
SDR bias hypothesis.
Our interest lies not in any single set of comparison of means, or
even in any single study, but with the total nomological network
Agreeableness & Social Desirability 723
relevant to the hypothesis. The outcomes for the experimental
manipulations of Agreeableness in Studies 2 and 3 cannot be
attributed to a lack of statistical power, since we obtained statistically
significant effects in some cases. Neither can these outcomes be
attributed to recalcitrant subjects; the manipulation checks and conflict
measures in Study 3 showed that subjects were very responsive. Nor
can they be attributed to lack of attention to individual differences that
might have moderated the experimental manipulations. It is possible,
and even probable, that other kinds of inductions might influence self-
or other-rated Agreeableness, but such an induction would need to be
more extreme than the ones commonly used. Within the range of
ethically acceptable manipulations typically used in laboratory social/
personality and social cognition research, we speculate that Agree-
ableness is probably not easily influenced by situational inductions.
Outcomes of these three studies suggest that if Agreeableness is
contaminated by self-favoring biases, the contamination is limited in
scope, relatively small, and more likely to be associated with
dispositional impression management than self-deception. There was
no evidence that self-reported Agreeableness was influenced by
situational inductions of the desirability of Agreeableness. In the case
of Agreeableness, the overall pattern of evidence suggests that
relations with various forms of self-favoring biases do not threaten
the interpretation of Agreeableness as a substantive personality
dimension related to the motivation to maintain positive relations
with others.
Going far beyond the data, several large issues surround this
research. The topic of socially desirable responding may appear to
some to be superficial, or worse, a digression tangential to the larger
issues facing current personality psychology. In recent years,
personality researchers have uncovered striking regularities in
behavior that seem to be robust across settings and times (e.g.,
Goldberg, 1995). Social desirability artifacts appear to be ephemeral in
the face of so many robust regularities. Why study holes when donuts
are on the table?
Social desirability behaviors may be holes, but the Agreeableness-
related motives and dispositions toward such behavior are donuts,
and delectable ones at that. This research focused on personality and
the motives underlying social desirability because the social
desirability of specific behavior is evanescent, and even behavioral
regularities across settings probably reflect complex adaptations to
724 Graziano & Tobin
the demands of larger sociohistorical conditions. The social
desirability of specific behavior varies greatly across social roles
and historical eras (see Barzun, 2000,   The Reign of Etiquette, 
pp. 285 305). Nevertheless, the motives underlying many specific
Agreeableness-like behaviors of the past appear remarkably familiar
to readers from later eras. Perhaps this familiarity is an illusion
because writers of history often focus on motives, are selective in
their examples, and pick cases that illustrate issues in the current
Zeitgeist. Familiarity with the outcomes of history also biases
interpretations (e.g., Fischhoff, 1975). Alternatively, readers may see
regularities because persons of all eras and social systems differ in
their motivation to maintain positive relationships with others
(Graziano & Eisenberg, 1997). The studies outlined here suggest
that Agreeableness and the motives for maintaining positive
relationships with others are transhistorical regularities and may
appear because they are authentic, enduring aspects of persons. How
differences in the motives associated with Agreeableness express
themselves, and interact with the social contingencies for purposes of
adaptation to social environments, remains an intriguing question for
both history and psychology.
REFERENCES
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting
interactions. Newbury Park, NJ: SAGE Publications.
Albright, L., Kenny, D. A., & Malloy, T. E. (1988). Consensus in personality
judgments at zero-acquaintance relationships. Journal of Personality and Social
Psychology, 55, 387 395.
Allport, G. W., & Odbert, H. S. (1936). Trait names: A psycho-lexical study.
Psychological Monographs, 47 (1, Whole No. 211).
Aronson, E., & Carlsmith, J. M. (1968). Experimentation in social psychology. In
G. Lindzey & E. Aronson (Eds.), Handbook of social psychology: Vol. 2. Research
methods (2nd ed., pp. 1 79). Reading, MA: Addison Wesley.
Barzun, J. (2000). From dawn to decadence: 1500 to the present. 500 years of western
cultural life. New York: Harper Collins.
Block, J. (1995). A contrarian view of the five-factor approach to personality
description. Psychological Bulletin, 117, 187 215.
Briggs, S. R., & Cheek, J. (1986). The role of factor analysis in the development and
evaluation of personality scales. Journal of Personality, 54, 106 148.
Campbell, D. T. (1963). Social attitudes and other acquired behavioral dispositions. In
S. Koch (Ed.), Psychology: A study of a science. Vol. 6: Investigations of man as a
socius (pp. 94 172). New York: McGraw Hill.
Agreeableness & Social Desirability 725
Cohen, J. (1987). Statistical power analyses for the behavioral sciences (rev. ed.).
Hillsdale, NJ: Erlbaum.
Colvin, C. R. (1993).   Judgable  people: Personality, behavior, and competing
explanations. Journal of Personality and Social Psychology, 64, 861 873.
Cunningham, M. R., Wong, D. T., & Barbee, A. P. (1994). Self-presentation dynamics
on overt integrity tests: Experimental studies of the Reid Report. Journal of
Applied Psychology, 79, 643 658.
Digman, J. M. (1996). The curious history of the five-factor model. In J. S.
Wiggins et al. (Eds.), The five-factor model of personality: Theoretical
perspectives (pp. 1 20). New York: Guilford Press.
Digman, J. M. (1997). Higher-order factors of the Big Five. Journal of Personality
and Social Psychology, 73, 1246 1256.
Digman, J. M., & Takemoto-Chock, N. K. (1981). Factors in the natural language of
personality: Re-analysis, comparison, and interpretation of six major studies.
Multivariate Behavioral Research, 16, 149 170.
Finch, J. F., & West, S. G. (1997). The investigation of personality structure:
Statistical models. Journal of Research in Personality, 31, 439 485.
Fischhoff, B. (1975). Hindsight is not equal to foresight: The effect of outcome
knowledge on judgment under uncertainty. Journal of Experimental Psychology:
Human Perception and Performance, 1, 288 299.
Funder, D., & Colvin, C. R. (1997). Congruence of others and self-judgments of
personality. In R. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of personality
psychology (pp. 617 647). San Diego: Academic Press.
Gangestad, S., & Snyder, M. (2000). Self-monitoring: Appraisal and reappraisal.
Psychological Bulletin, 126, 530 555.
Goldberg, L. R. (1992). The development of markers of the Big Five factor structure.
Psychological Assessment, 4, 26 42.
Goldberg, L. R. (1995). What the hell took so long? Donald W. Fiske and the Big-Five
factor structure. In P. E. Shrout, S. T. Fiske et al. (Eds.), Personality research,
methods, and theory: A festschrift honoring Donald W. Fiske (pp. 29 43).
Hillsdale, NJ: Erlbaum.
Graziano, W. G., & Bryant, W. H. (1998). Self-monitoring and the self-attribution of
positive emotions. Journal of Personality and Social Psychology, 74, 250 261.
Graziano, W. G., & Eisenberg, N. (1997). Agreeableness: A dimension of personality.
In R. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of personality psychology
(pp. 795 824). San Diego: Academic Press.
Graziano, W. G., Hair, E. C., & Finch, J. F. (1997). Competitiveness mediates the link
between personality and group performance. Journal of Personality and Social
Psychology, 73, 1394 1408.
Graziano, W. G., Jensen-Campbell, L. A., & Hair, E. C. (1996). Perceiving
interpersonal conflict and reacting to it: The case for Agreeableness. Journal of
Personality and Social Psychology, 70, 820 835.
Hafdahl, A. R., Panter, A. T., Gramzow, R. H., Sedikides, C., & Insko, C. A. (2000).
Free-response self-discrepancies across, among, and within FFM personality
dimensions. Journal of Personality, 68, 111 151.
Havill, V. L., Besevegis, E., & Mouroussaki, S. (1998). Agreeableness as a diachronic
human trait. In G. A. Kohnstram, C. F. Halverson, I. Mervielde, & V. L. Havill (Eds.),
726 Graziano & Tobin
Parental descriptions of child personality: Developmental antecedents of the Big
Five? (pp. 49 64). Mahwah, NJ: Erlbaum.
Hogan, R. T. (1983). A socioanalytic theory of personality. In M. Page (Ed.),
Nebraska Symposium on Motivation: Personality Current theory and research
(pp. 58 89). Lincoln, NE: University of Nebraska Press.
Jackson, D. N. (1971). The dynamics of structured personality tests. Psychological
Review, 78, 229 248.
Jackson, D. N., & Messick, S. J. (1962). Response styles and the assessment of
psychopathology. In S. J. Messick & J. Ross (Eds.), Measurement in personality
and cognition. New York: Holt, Rinehart, & Wilson.
Jensen-Campbell, L. A., & Graziano, W. G. (2001). Agreeableness as a moderator of
interpersonal conflict. Journal of Personality, 69, 323 361.
John, O. P., & Srivastava, S. (1999). The Big Five trait taxonomy: History, mea-
surement, and theoretical perspectives. L. A. Pervin & O. P. John (Eds.), Handbook
of personality: Theory and research (2nd ed., pp. 102 138). New York: Guilford.
Kashy, D. A., & Kenny, D. A. (2000). The analysis of data from dyads and groups. In
H. T. Reis & C. M. Judd (Eds.), Handbook of research methods in social
psychology. New York: Cambridge University Press.
Kenny, D. A. (1994). Interpersonal perception: A social relations analysis. New
York: Guilford.
Kenny, D. A. (1996). The design and analysis of social-interaction research. Annual
Review of Psychology, 47, 59 86.
Kohnstamm, G. A., Halverson, Jr., C. F., Mervielde, I., & Havill, V. L. (1998).
Analyzing parental free description of child personality. In G. A. Kohnstamm,
C. F. Halverson, Jr., I. Mervielde, & V. L. Havill (Eds.), Parental description of
child personality: Developmental antecedents of the Big Five? (pp. 1 19).
Mahwah, NJ: Erlbaum.
Kolar, D. W., Funder, D. C., & Colvin, C. R. (1996). Comparing the accuracy of
personality judgments by the self and knowledgeable others. Journal of
Personality, 64, 311 338.
Lautenschlager, G. J. (1986). Within-subjects measures for the assessment of
individual differences in faking. Educational and Psychological Measurement,
46, 309 316.
Lautenschlager, G. J., & Flaherty, V. L. (1990). Computer administration of questions:
More desirable or more socially desirable? Journal of Applied Psychology, 75,
310 314.
Lykken, D. T. (1970). Statistical significance in psychological research. In D. E.
Morrison & R. E. Henkel (Eds.), The significance test controversy: A reader
(pp. 267 279). Chicago: Aldine.
MacDonald, K. (1992). Warmth as a developmental construct: An evolutionary
analysis. Child Development, 63, 753 773.
MacDonald, K. (1995). Evolution, the five-factor model, and levels of personality.
Journal of Personality, 63, 525 567.
McCrae, R. R., & Costa, P. T. (1983). Social desirability scales: More substance than
style. Journal of Consulting and Clinical Psychology, 51, 882 888.
McCrae, R. R., & John, O. J. (1992). An introduction to the five-factor model and its
applications. Journal of Personality, 60, 175 216.
Agreeableness & Social Desirability 727
Ones, D., Viswesvaran, C., & Reiss, A. D. (1998). Role of social desirability in
personality testing for personnel selection: The red herring. Journal of Applied
Psychology, 81, 660 679.
Paulhus, D. L. (1984). Two component model of socially desirable responding.
Journal of Personality and Social Psychology, 46, 598 609.
Paulhus, D. L. (1986). Self-deception and impression management in test responses.
In A. Angleitner & J. S. Wiggins (Eds.), Personality assessment via ques-
tionnaires: Current issues in theory and measurement (pp. 143 165). New York:
Springer-Verlag.
Paulhus, D. L. (1988). The Balanced Inventory of Desirable Responding: Version 6.
(Manual available from the author at the Department of Psychology, University of
British Columbia, Vancouver, B. C., V6T 1Y7, Canada.)
Paulhus, D. L. (1991). Measurement and control of response bias. In J. P.
Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.), Measures of personality
and social psychological attitudes. Vol. 1: Measures of social psychological
attitudes (pp. 17 59). San Diego, CA: Academic Press.
Paulhus, D. L. (in press). Socially desirable responding: The evolution of a construct.
To appear in H. Braun, D. E. Wiley, & D. N. Jackson (Eds.), Personality and
Intellect, validity and values: Cross-cutting themes. New York: Guilford.
Paulhus, D. L., Bruce, M. N., & Trapnell, P. D. (1995). Effects of self-presentation
strategies on personality profiles and their structure. Personality and Social
Psychology Bulletin, 21, 100 108.
Paulhus, D. L., & John, O. P. (1998). Egoistic and moralistic biases in self-perception:
The interplay of self-deception styles with basic traits and motives. Journal of
Personality, 66, 1025 1060.
Smith, H. L. (1998). The structure and utility of social desirability scales in
psychological research. Dissertation Abstracts International, 58, 5695.
Snyder, M. (1987). Public appearances, private realities: The psychology of self-
monitoring. New York: W. H. Freeman.
Strosahl, K. D., Linehan, M. M., & Chiles, J. A. (1984). Will the real social
desireability please stand up? Hopelessness, depression, social desireability, and
the prediction of suicidal behavior. Journal of Consulting & Clinical Psychology,
52, 449 457.
Tobin, R. M., Graziano, W. G., Vanman, E. J., & Tassinary, L. G. (2000). Personality,
emotional experience, and efforts to control emotions. Journal of Personality and
Social Psychology, 79, 656 669.
Wiggins, J. S., & Trapnell, P. D. (1997). Personality structure: The return of the Big
Five. In R. Hogan, J. Johnson, & S. Briggs (Eds.), Handbook of personality
psychology (pp. 737 765). San Diego, CA: Academic Press.
728