Altruism and antisocial behavior: independent tendencies, unique personality correlates, distinct etiologies.

Abstract

The relationship between altruism and antisocial behavior has received limited attention because altruism and antisocial behavior tend to be studied and discussed in distinct literatures. Our research bridges these literatures by focusing on three fundamental questions. First, are altruism and antisocial behavior opposite ends of a single dimension, or can they coexist in the same individual? Second, do altruism and antisocial behavior have the same or distinct etiologies? Third, do they stem from the same or from distinct aspects of a person's personality? Our findings indicate that altruism and antisocial behavior are uncorrelated tendencies stemming from different sources. Whereas altruism was linked primarily to shared (i.e., familial) environments, unique (i.e., nonfamilial) environments, and personality traits reflecting positive emotionality, antisocial behavior was linked primarily to genes, unique environments, and personality traits reflecting negative emotionality and a lack of constraint.

Full text

Page 1

PSYCHOLOGICAL SCIENCE

Research Article

VOL. 12, NO. 5, SEPTEMBER 2001 Copyright © 2001 American Psychological Society

397

ALTRUISM AND ANTISOCIAL BEHAVIOR:
Independent Tendencies, Unique Personality Correlates, Distinct Etiologies

Robert F. Krueger, Brian M. Hicks, and Matt McGue
University of Minnesota–Twin Cities


Abstract—
has received limited attention because altruism and antisocial behav-
ior tend to be studied and discussed in distinct literatures. Our re-
search bridges these literatures by focusing on three fundamental
questions. First, are altruism and antisocial behavior opposite ends of
a single dimension, or can they coexist in the same individual? Sec-
ond, do altruism and antisocial behavior have the same or distinct eti-
ologies? Third, do they stem from the same or from distinct aspects of
a person’s personality? Our findings indicate that altruism and anti-
social behavior are uncorrelated tendencies stemming from different
sources. Whereas altruism was linked primarily to shared (i.e., famil-
ial) environments, unique (i.e., nonfamilial) environments, and per-
sonality traits reflecting positive emotionality, antisocial behavior was
linked primarily to genes, unique environments, and personality traits
reflecting negative emotionality and a lack of constraint.

The relationship between altruism and antisocial behavior


Psychology is in the midst of a renaissance of interest in personal-
ity. Owing directly to this renewed interest, Krueger and his colleagues
have pursued systematic research on links between personality traits and
maladaptive social behaviors in population-based samples (e.g., Krue-
ger, 1999; Krueger, Caspi, Moffitt, Silva, & McGee, 1996; Krueger et al.,
1994). Krueger, Caspi, and Moffitt (2000) recently coined the term
“epidemiological personology” to describe this approach.
To date, research in epidemiological personology has focused
solely on distress and dysfunction. Understanding distress and dys-
function is an important goal, as maladaptive social behavior is costly
to society. But by focusing solely on maladaptive social behavior, re-
searchers may be neglecting key elements of psychology’s original
mission. As noted recently by Seligman and Csikszentmihalyi (2000),
“Psychology is not just the study of pathology, weakness, and dam-
age; it is also the study of strength and virtue” (p. 7). We hypothesized
that epidemiological personology could contribute to this effort to re-
juvenate the study of strength and virtue, and hence, we undertook re-
search to extend our paradigm to the study of prosocial behavior.
Specifically, we sought to compare and contrast a specific form of
prosocial behavior, altruism, with its logical counterpart from our
prior investigations, antisocial behavior (see Krueger et al., 1994). We
pursued research directed at answering three fundamental questions
about relations among altruism, antisocial behavior, and personality
traits.
First, are altruism and antisocial behavior opposite ends of a single
dimension, or can altruistic and antisocial tendencies coexist in the
same individual? Altruism is typically conceptualized as a component
of the broader construct of prosocial behavior, and the term “proso-
cial” entered the social science literature to define the opposite of anti-

social behavior (Batson, 1998; Wispe, 1972). This terminology suggests
that altruism and antisocial behavior should define opposite ends of a
single dimension. Nevertheless, this possibility has received limited
empirical attention because research on antisocial behavior and re-
search on prosocial behavior tend to form two distinct literatures.
Antisocial behavior is typically studied by criminologists and psycho-
pathologists, whereas altruism is typically studied by social psycholo-
gists. Although a few studies have bridged these literatures, findings
from these studies are inconsistent, with some suggesting that pro- and
antisocial tendencies are opposite ends of a single trait (e.g., Eron &
Huesmann, 1984; Goma-i-Freixanet, 1995), and others suggesting that
these domains contain substantial amounts of nonoverlapping vari-
ance (e.g., Axelrod, Widiger, Trull, & Corbitt, 1997; Harris, Rushton,
Hampson, & Jackson, 1996; Levenson, 1990; McCord, 1992; Rush-
ton, Fulker, Neale, Nias, & Eysenck, 1986, 1989). Moreover, existing
studies mix measures of behavior with measures of attitudes and repu-
tation, leaving the question of the correlation between altruistic and
antisocial behavior per se unanswered. In the study we report here, we
directly estimated the correlation between altruistic and antisocial be-
havioral tendencies by measuring these propensities using comparably
reliable self-report inventories that inquired specifically about the fre-
quency of altruistic and antisocial behaviors, as opposed to attitudes or
reputation.
Second, are altruism and antisocial behavior linked to personality,
and do they stem from the same or from distinctive aspects of a per-
son’s personality? As recently as 1990, a review of the field of altru-
ism research concluded that “it is undoubtedly futile to search for the
altruistic personality,” citing “inconsistent relationships between per-
sonality characteristics and prosocial behavior” (Piliavin & Charng,
1990, p. 31). Interestingly, before the 1990s, criminologists also dis-
missed personality as an unfruitful area of inquiry (see Stitt & Gia-
copassi, 1992). Fortunately, these viewpoints have undergone marked
change in recent years. Recent reviews of the altruism literature have
noted that theoretical models of altruism that do not take dispositional
factors into account are likely to be incomplete (Batson, 1998; Schro-
der, Penner, Dovidio, & Piliavin, 1995). In addition, empirical re-
search has begun to demonstrate systematic and meaningful links
between personality and antisocial behavior (Elkins, Iacono, Doyle, &
McGue, 1997; Krueger et al., 1994).
A question that remains to be explored, however, is whether a
structural model of personality traits can help to explain the relation
(or lack thereof) between altruism and antisocial behavior. A structural
model of personality is a formal account of patterns of correlations
among basic personality traits in terms of a smaller number of “super-
traits,” or higher-order personality factors. These higher-order factors
provide a “road map” of the major axes of human variation, and a
structural model of personality therefore provides the theoretical basis
needed to understand patterns of relations among various behavioral
propensities. For example, in our research to date, we have employed
the structural model of personality traits developed by Tellegen
(1985). In Tellegen’s system, basic personality traits can be organized

Address correspondence to Robert F. Krueger, Department of Psychology,
University of Minnesota, N414 Elliott Hall, 75 East River Rd., Minneapolis,
MN 55455-0344; e-mail: krueg038@tc.umn.edu.

Page 2

PSYCHOLOGICAL SCIENCE

Altruism and Antisocial Behavior

398

VOL. 12, NO. 5, SEPTEMBER 2001

around three fundamental and relatively independent super-traits: pos-
itive emotionality (a tendency to experience positive emotions linked
to active engagement with social and work environments), negative
emotionality (a tendency to experience unpleasant emotions such as
anxiety and anger), and constraint (a tendency to endorse social norms,
act in a cautious manner, and avoid thrills). These three structures re-
flect the organization of personality not only at the phenotypic level,
but also at the level of both genetic and environmental effects (Krueger,
2000). In addition, we have found this system to have excellent utility
in explaining correlations among maladaptive behavior patterns, with
negative emotionality being linked to maladjustment generally, but
constraint determining the form maladjustment takes. Specifically, if
negative emotionality is paired with low constraint, maladjustment is
expressed outward, in the form of problems such as antisocial behav-
ior (vs. inward as anxiety and depression, in the absence of low con-
straint; Krueger, 1999; Krueger et al., 1996, 2000).
Positive emotionality, in contrast, has shown few robust correla-
tions with maladaptive behavior in our research. Are there no robust
behavioral correlates of positive emotionality, or have we been look-
ing in the wrong places? The research we report here addressed the
latter possibility—that by focusing exclusively on maladaptive behav-
ior, we have neglected the behavioral correlates of positive emotional-
ity. Keeping in mind that positive emotionality, negative emotionality,
and constraint are relatively independent, this hypothesis implies that
adaptive and maladaptive behavioral tendencies might be uncorre-
lated. More specifically, we hypothesized that altruism would be
linked to the positive emotionality superfactor (cf. Graziano & Eisen-
berg, 1997), but that (as in our past research) antisocial behavior
would be linked to high negative emotionality and low constraint. This
pattern of correlations, in turn, suggests that altruism and antisocial
behavior should be uncorrelated, independent tendencies because they
are linked to distinctive, uncorrelated aspects of personality. In the
current research, we were able to test this hypothesis because partici-
pants completed a personality instrument that measures Tellegen’s su-
perfactors (the Multidimensional Personality Questionnaire, MPQ;
Tellegen, 2000), in addition to the aforementioned indices of altruistic
and antisocial behavior.
Our third question was, do altruistic and antisocial behavior have
the same or distinct etiologies? A powerful means of investigating the
etiology of individual differences in behavior involves studying genet-
ically informative samples, such as samples of twins. Nevertheless, in
the behavior genetic literature, antisocial behavior has received far
greater attention than altruistic behavior. Much is now known about
the genetic and environmental roots of antisocial behavior in adult-
hood; recent, well-designed studies reveal a significant genetic con-
tribution to antisocial behavior in adults, as well as a significant
contribution of the nonshared environment (environmental effects
that make people growing up in the same families different; Bock &
Goode, 1996; Carey & Goldman, 1997; DiLalla & Gottesman, 1989;
Gottesman & Goldsmith, 1994; Lyons et al., 1995; Rutter, 1997; van
den Bree, Svikis, & Pickens, 1998). However, much less is known
about the genetic and environmental roots of altruism; we are aware of
only one genetically informative study focused specifically on altruis-
tic behavior in adults (Rushton et al., 1986). Yet there are reasons to
believe that behavior genetic studies may be as informative about al-
truistic behavior as they have been about antisocial behavior and,
moreover, that such studies may reveal altruism to be etiologically dis-
tinct from antisocial behavior. Specifically, little variance in most be-
havioral traits can be traced to shared environments (environmental
effects that make people growing up in the same families similar;
Rowe, 1994). One possible exception to this basic finding, however,
involves positive traits and behaviors (Plomin & Caspi, 1999). Atti-
tudes toward love and romance (Waller & Shaver, 1994), agreeable-
ness (Bergeman et al., 1993; Riemann, Angleitner, & Strelau, 1997),
and extraversion (Beer, Arnold, & Loehlin, 1998) in adults, as well as
positive affect in children (Goldsmith, Buss, & Lemery, 1997), have
shown significant shared environmental variance. On the basis of this
evidence, we hypothesized that altruistic and antisocial behavior
would be etiologically distinct, with altruism showing a notably larger
shared environmental component of variance than antisocial behavior.
To evaluate this possibility, we conducted our research in a sample of
twins.

METHOD
Research Participants

Data for this report were obtained from mailings to male twins
born between 1961 and 1964. These twins make up the youngest co-
hort enrolled in the Minnesota Twin Registry (MTR). MTR members
were ascertained from birth records provided by the Minnesota State
Health Department. Surviving intact twin pairs were located and re-
cruited by mail, and initial mailings included five questions about twin
similarity that were used to determine zygosity (monozygotic, or MZ,
vs. dizygotic, or DZ). This method achieves 95% accuracy when com-
pared with zygosity established via blood samples (Lykken, Bouchard,
McGue, & Tellegen, 1990). Measures for this report and informed
consent were obtained from 170 MZ pairs, 106 DZ pairs, and 121 in-
dividuals whose twin did not participate. These 673 persons were 33
years old on average when they completed our measures. To correct
for the nonindependence of observations stemming from complete twin
pairs, individuals from complete twin pairs were assigned weights of
1/2 in our phenotypic analyses to reflect the number of independent
observations in the analyses, resulting in a weighted
McGue, Bacon, & Lykken, 1993).

N

of 397 (cf.

Measurement of Personality

Participants completed the 198-item version of the MPQ (Tellegen,
2000). The 11 primary MPQ scales have excellent internal-consis-
tency reliabilities (alphas in the MTR range from .77 to .88) and co-
here in a three-factor, higher-order structure consisting of orthogonal
dimensions labeled positive emotionality, negative emotionality, and
constraint (Krueger 2000; Tellegen, 1985). Positive emotionality is
indicated primarily by Well-Being, Social Potency, Social Closeness,
Achievement, and Absorption. Negative emotionality is indicated pri-
marily by Stress Reaction, Alienation, and Aggression. Constraint is
indicated primarily by Control, Harmavoidance, and Traditionalism.

Measurement of Antisocial Behavior

Antisocial behavior was measured using a self-report questionnaire
consisting of items drawn primarily from the Short-Nye Self-Report
Delinquency Items and the Seattle Self-Report Instrument (see Hinde-
lang, Hirschi, & Weis, 1981, Appendices A and B), as well as from the
Clark Self-Report List of Deviant Behavior (Clark & Tifft, 1966).
Self-report indices of antisocial behavior are commonly employed in
criminological research because they are often less biased than indices

Page 3

PSYCHOLOGICAL SCIENCE

Robert F. Krueger, Brian M. Hicks, and Matt McGue

VOL. 12, NO. 5, SEPTEMBER 2001

399

that reflect only crime detected by authorities (such as conviction
records). Nevertheless, such indices also show good convergent valid-
ity with other indices of antisocial behavior, such as the reports of
peers and officials (e.g., Krueger et al., 1994).
Items that ask for the frequency of each antisocial behavior (4-point
scale, from
never
to
very often
) after high school graduation (or age
17) were used for this report. These items fall into four broad content
classes, and span a range of common and rare behaviors and a range of
seriousness. They inquire about theft (e.g., “How often did you take
something not belonging to you worth more than $50?”), illegal be-
havior involving drugs and alcohol (e.g., “How many times did you
sell marijuana or some other illegal drug?”), force (e.g., “How often
did you take part in fights in which a group of your friends was against
another group?”), and miscellaneous vice (e.g., “How many traffic ci-
tations, other than for parking, did you receive?”). The reliability of
our self-reported adult antisocial behavior index was good (
27 items). Because total scores on the index were positively skewed,
these scores were log-transformed prior to being analyzed.





?



?

.89,

Measurement of Altruism

Altruism was measured using a self-report questionnaire consist-
ing of items adapted from the Self-Report Altruism Scale of Rushton,
Chrisjohn, and Fekken (1981), plus additional items added to expand
upon the range and scope of altruistic actions. Specifically, the Self-
Report Altruism Scale inquires primarily about altruistic actions taken
toward strangers, or intended to benefit organizations, and our items
expanded upon these areas, as well as inquiring about actions intended
to benefit friends and acquaintances. Thus, the 45 items on our ques-
tionnaire fell into four broad content classes, inquiring about the
frequency of altruistic behavior (4-point scale, from
ward friends (e.g., “I have done chores or shopping for one or more
close friends who were sick”), acquaintances (e.g., “I have offered to
give a ride to one or more acquaintances even though it meant going
out of my way”), strangers (e.g., “I have stopped to help a stranger
who was having difficulty carrying their belongings”), and organiza-
tions (e.g., “I have done volunteer work for a charity or community
service organization”). The reliability of this scale in our sample was
good (
?

?
.90).

never

to

often

) to-





RESULTS AND DISCUSSION
Altruism and Antisocial Behavior:
Correlated or Independent?

Were altruism and antisocial behavior correlated or independent
tendencies in our data? The Pearson product-moment correlation be-
tween our altruism and antisocial behavior scales was
correlation suggests that altruism and antisocial behavior were largely
independent in our data. Nevertheless, two forms of error can reduce
the magnitude of observed correlations, potentially leading to under-
estimates of actual correlations: acquiescence (a systematic tendency
to respond “true” to diverse items, regardless of item content) and ran-
dom measurement error. Hence, we pursued analyses to ensure that
these artifacts were not responsible for the low magnitude of the corre-
lation we observed.
First, we used Tellegen’s (2000) TRIN (true-response inconsis-
tency) scale to measure acquiescence, and we examined the correla-
tion between altruism and antisocial behavior, controlling for TRIN.

?

.08, n.s. This

TRIN consists of 23 pairs of MPQ items that are opposite in content,
and is scored as the sum of responses to these items, with no items re-
flected (i.e., all items scored true). Hence, high scores on TRIN repre-
sent the tendency to answer “true” regardless of item content. The
partial correlation between altruism and antisocial behavior control-
ling for TRIN was
?
.08, n.s. Thus, in our data, the nonsignificant cor-
relation between altruism and antisocial behavior was not due to an
acquiescent response set.
Second, we estimated the latent correlation between altruism and
antisocial behavior, taking into account random measurement error, by
using a confirmatory factor analytic model. Specifically, for this anal-
ysis, we divided the items in our altruism and antisocial behavior in-
ventories rationally into facets, or groups of items, based on their
manifest content. The antisocial behavior items were divided into four
groups, theft (6 items), drug-alcohol (3 items), force (6 items), and
vice (12 items). As these facets were positively skewed, we treated the
theft, drug-alcohol, and force variables as ordinal and log-transformed
vice prior to estimating correlations among the variables. The altruism
items were also divided into four groups, friends (7 items), acquain-
tances (9 items), strangers (16 items), and organizations (13 items),
which were treated as continuous because they showed no marked
skew.
Product-moment (for pairs of continuous variables), polychoric
(for pairs of ordinal variables), and polyserial (for pairings between
ordinal and continuous variables) correlations were computed among
the variables using the computer program PRELIS 2.3 (Jöreskog &
Sörbom, 1996b), and may be seen in Table 1. Using the computer pro-
gram LISREL 8.3 (Jöreskog & Sörbom, 1996a), we fit a weighted
least-squares confirmatory factor analytic model to these correlations.
In this model, the altruism facets were allowed to load freely on a la-
tent Altruism factor, the antisocial facets were allowed to load freely
on a latent Antisocial factor, the correlation between these latent fac-
tors was also free to vary, and identification was achieved by fixing the
variances of the latent factors at 1.0. This model fit the data suffi-
ciently,
?
(19,
N

?
397)
?
37.65,
p
seen in Table 1, all variables showed notable standardized loadings on
their target factors; each of these loadings was significant at
The latent correlation between antisocial behavior and altruism in this
model was estimated at
?
.12, which was not significant (
In sum, a straightforward zero-order correlational approach, a correla-
tional approach controlling for acquiescence, and a more elaborate
confirmatory factor analytic approach all converged on the conclusion
that altruism and antisocial behavior were largely independent behav-
ioral tendencies in our data.




2











?

.01,

RMSEA



?

.05. As can be

p



?

.05.



z



?



?

1.37).

Biometry of Altruism and Antisocial Behavior

Given evidence that altruism and antisocial behavior were indepen-
dent tendencies in our data, we sought to determine if the variance in
these tendencies might be linked to different sources. Specifically,
data from twins allows resolution of the sources of variance in behav-
ioral phenotypes into three categories: additive genetic variance (la-
beled A); shared, or “common,” environmental variance (labeled C);
and nonshared environmental variance (labeled E). Our goal was to
find the maximum-likelihood model that best explained the observed
data while invoking as few unknown parameters as possible (i.e., the
model that achieved the smallest chi-square while allowing the largest
number of degrees of freedom). We therefore evaluated model fit us-
ing Akaike’s Information Criterion (AIC; Akaike, 1987), calculated as

Page 4

PSYCHOLOGICAL SCIENCE

Altruism and Antisocial Behavior

400

VOL. 12, NO. 5, SEPTEMBER 2001

?
We fit four models to the altruism and antisocial variables using the
computer program MX (Neale, 1997): a model that included all three
components of variance (additive genes, shared environments, and
nonshared environments, referred to as an ACE model), a model that
dropped C (AE), a model that dropped A (CE), and a model that
dropped both A and C (E).
Results from fitting these models may be seen in Table 2. For anti-
social behavior (MZ cross-twin correlation
.28), the best-fitting model was the AE model, whereas for altruism
(MZ correlation
?
.38; DZ correlation
was the CE model. In the AE model of antisocial behavior, the stan-
dardized maximum-likelihood estimate of A was 52% (95% likeli-
hood-based confidence interval: 41%–62%), and the standardized
maximum-likelihood estimate of E was 48% (38%–59%). In the CE
model of altruism, C was estimated at 35% (25%–45%) and E was es-
timated at 65% (55%–75%). To summarize, individual differences in
antisocial behavior were linked to genes and nonshared environments,

2



?

(2

?



df

), seeking the model with the largest negative AIC value.

?

.51; DZ correlation

?




?

.31), the best-fitting model
whereas individual differences in altruism were linked to shared and
nonshared environments.

Linking Altruism and Antisocial Behavior to
Personality Traits

In our data, altruism and antisocial behavior were largely uncorre-
lated tendencies with distinct etiologies; were they also linked with
distinct personality traits? Correlations between (a) the MPQ scales
and superfactors and (b) altruism and antisocial behavior are presented
in Table 3. The table shows a consistent pattern of relations, in which
positive emotionality and its component scales are correlated with al-
truism, but other personality traits—negative emotionality and a lack
of constraint—are correlated with antisocial behavior.
To ensure that these correlations were not inflated because of con-
tent overlap between our altruism or antisocial behavior measure and
the MPQ, we examined each item on the altruism and antisocial be-
havior measures for possible overlap with the MPQ items. We found
six MPQ items with content that partially overlapped with the antiso-
cial behavior measure; no MPQ items overlapped with the altruism
measure. Each of these six MPQ items inquires about physically vio-
lent actions (e.g., “Sometimes I just like to hit someone”), and all six
items are scored on the Aggression subscale. Hence, we reconstructed
this subscale, eliminating these six items, and recomputed the correla-
tion between aggression and antisocial behavior. The recomputed cor-
relation was .37,
p

?
.01 (vs. the original correlation of .39). In sum,
content overlap could not explain the correlations we found linking
personality to altruism and antisocial behavior.
Finally, to assess the overall magnitude of relations between per-
sonality and altruism and antisocial behavior, we estimated ordinary
least-squares multiple regressions in which all 11 MPQ scales were
entered simultaneously as predictors of altruism and antisocial behav-
ior. As shown in the last row of Table 3, both multiple correlations
were sizable (.50 and .47, respectively). In addition, examination of
the standardized regression coefficients (betas) from these regressions
suggested that at the level of unique effects of the MPQ scales, Social
Potency, Social Closeness, Absorption, and a lack of Aggression were
associated with altruism, whereas Aggression, a lack of Control, and a
lack of Harmavoidance were associated with antisocial behavior. Thus,





Table 1.
antisocial behavior

Correlations among facets of altruism and antisocial behavior and factor loadings for a two-factor model of altruism and

Facet
Correlations
Loadings
Altruism
Antisocial
behavior12345678
1. Altruism-strangers
2. Altruism-acquaintances
3. Altruism-friends
4. Altruism-organizations
5. Antisocial-force
6. Antisocial-drug
7. Antisocial-theft
8. Antisocial-vice
1.00
.63
.64
.61
.00
?
.12
?
.04
?
.09
.85
.80
.81
.69
—
—
—
—
—
—
—
—
.77
.65
.79
.70
1.00
.67
.46
.12
?
.03
.03
.04
1.00
.51
.09
.02
.06
.05
1.00
?
.13
?
.17
?
.17
?
.15


1.00
.37
.48
.61






1.00
.45
.48




1.00
.61




1.00

Note.
marked with “—” refer to loadings fixed at zero in the confirmatory factor model.

Numbers listed under “Correlations” refer to facets, the names of which are given on the left side of the table. In the “Loadings” columns, cells

Table 2.
altruism and antisocial behavior

Results from four biometric models fit to twin data on

Variable
Model
ACE (3)AE (4)CE (4)E (5)
Altruism
?
AIC
Antisocial behavior
?
AIC


2

1.52
4.48
3.15
4.85
1.99
6.01
38.88
28.88

?


?


?



2

1.77
4.23
1.79
6.21
8.63
0.63
60.93
50.93

?


?


Note.
of freedom associated with each model. A
C
?
shared environmental variance; E
variance; AIC
?
Akaike’s Information Criterion.

Numbers in parentheses after the model labels show the degrees

?

additive genetic variance;
nonshared environmental



?

Page 5

PSYCHOLOGICAL SCIENCE

Robert F. Krueger, Brian M. Hicks, and Matt McGue

VOL. 12, NO. 5, SEPTEMBER 2001

401

both altruism and antisocial behavior were strongly and meaningfully
associated with distinctive personality traits.

SUMMARY AND CONCLUSION

Our analyses indicate that altruism and antisocial behavior are in-
dependent tendencies with unique personality correlates and distinct
etiologies. Nevertheless, some limitations must be borne in mind when
considering these findings. First, our findings are limited to men; future
research must determine the generality of these findings across gen-
ders. Also, although our sample was drawn from a broad population
(male twins born in Minnesota in the early 1960s), this population is
not as broad as, for example, the U.S. population. It would be valuable
to examine the generality of our findings in other samples, drawn from
other broad and well-defined populations.
Second, our findings are limited to self-report. Analyses employing
other indices of altruistic and antisocial behavior (e.g., records of tax
deductions resulting from charitable contributions, arrest records)
could provide a very informative extension of the work presented
here. Third, our data are contemporaneous rather than predictive. In
future work, it would be valuable to determine the ways in which per-
sonality traits and altruistic and antisocial actions interact and influ-
ence each other across time and across other developmental periods.
For example, would our findings generalize to adolescence, when anti-
social behavior is at its peak (Moffitt, 1993), or to later adulthood,
when generativity (prosocial behavior intended to benefit the next
generation) becomes important (McAdams & de St. Aubin, 1998)?
Fourth, our findings are limited to a specific, behavioral operation-
alization of the altruism construct, which may be one facet of a broader
prosocial orientation encompassing affective and cognitive propensities
in addition to behavioral ones. Along these lines, Penner, Fritzsche,
Craiger, and Freifeld (1995) gathered measures predictive of prosocial
tendencies and found two correlated facets underlying these measures.
The first facet, “other-oriented empathy,” is indicated more by thoughts
and feelings of concern for others’ welfare, whereas the second facet,
“helpfulness,” is indicated more by reports of past prosocial actions.
As our measure of altruism is closer to the second of these two facets,
future research could extend the efforts reported here by studying per-
sonality, antisocial behavior, and both of the facets found by Penner et
al. in a genetically informative sample.
Despite these limitations, our findings answer some fundamental
questions about altruism and antisocial behavior, and they emphasize
the relevance of both genetically informative studies and structural
models of personality in understanding the psychological bases and
origins of behavioral tendencies. Altruism and antisocial behavior were
independent tendencies in our data. This finding, although potentially
counterintuitive from a terminological perspective (i.e., the tendency to
use the term “prosocial behavior” as an antonym for “antisocial behav-
ior”; Batson, 1998; Wispe, 1972), makes sense in light of our additional
analyses. Specifically, altruism and antisocial behavior were etiologi-
cally distinct; altruism in our sample arose from environmental factors
(both shared and nonshared), but antisocial behavior arose from genes
and nonshared environments. Although behavior genetic designs have
produced a large database attesting to the relative unimportance of
shared family environments in shaping many adult phenotypes (Rowe,
1994), altruism—and perhaps prosocial and positive behaviors more
generally—may provide an important exception. Perhaps altruistic be-
havior in adulthood shows greater shared environmental influence
than antisocial behavior because altruistic behaviors stem more from
cultural experiences that are shared by members of the same families
and that directly promote altruistic behavior. Church affiliation, for ex-
ample, is substantially influenced by the shared environment (D’Ono-
frio, Eaves, Murrelle, Maes, & Spilka, 1999), and might contribute
directly to frequency of altruistic behavior.
Although further research is needed to replicate and extend our
findings to additional positive phenotypes, our finding of distinct etiol-
ogies for altruistic and antisocial behavior could have broad implica-
tions. If the sources of desirable and undesirable behaviors are distinct,
strategies designed to diminish undesirable behavior need not result in
the promotion of desirable behavior, and vice versa. Moreover, re-
searchers’ tendency to focus on undesirable behavior appears to result
in an incomplete picture of human functioning; desirable and undesir-
able qualities can coexist in the same persons, and conceptual models
and assessment approaches need to be broadened to reflect this fact
(Seligman & Csikszentmihalyi, 2000).
The finding of a distinction between the personological sources
of altruism (positive emotionality) and antisocial behavior (uncon-
strained negative emotionality) also helps to explain the independence
of these tendencies. This finding explains why we have encountered
only sporadic relations between positive emotionality and behavior in
our work to date (Krueger et al., 2000); it is not that positive emo-
tionality lacks behavioral correlates, but instead that these correlates
entail adaptive social behaviors, and we have been studying only mal-
adaptive behaviors. Note also that specific personality traits had only
modest validity as predictors of behavioral tendencies (cf. Ahadi &
Diener, 1989), but in aggregate, personality was substantially related

Table 3.
Personality Questionnaire (MPQ) scales and superfactors and
altruism and antisocial behavior

Indices of association between Multidimensional

MPQ
Behavior
Altruism
Antisocial
behavior

r

?

r

?

Scales
Well-Being
Social Potency
Achievement
Social Closeness
Absorption
Stress Reaction
Alienation
Aggression
Control
Harmavoidance
Traditionalism
Superfactors
Positive Emotionality
Negative Emotionality
Constraint
Multiple correlation
.30*
.30*
.25*
.29*
.23*
.10
?
.12
?
.21*
.08
?
.09
.10
.05
.17*
.11
.18*
.20*
.02
.00
?
.21*
.00
?
.06
.07
.03
.11
.04
.03
.16*
.15*
.22*
.39*
.26*
?
.24*
?
.06
.08
.00
.03
.03
.03
.02
.04
.31*
?
.15*
?
.15*
.00

?


?








?













.44*
.10
.07
.50*
.11
.28*
.27*
.47*

?


?


Note.
squares regressions predicting altruism and antisocial behavior from all
11 MPQ scales.
*
p

?
.01.

Multiple correlations and beta weights are from ordinary least-