Hierarchy in the Library: Egalitarian Dynamics in Victorian Novels

Evolutionary Psychology
www.epjournal.net – 2008. 6(4): 715-738
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Original Article
Hierarchy in the Library: Egalitarian Dynamics in Victorian Novels
John A. Johnson, Department of Psychology, Pennsylvania State University, DuBois, PA, USA. Email:
j5j@psu.edu (Corresponding author)
Joseph Carroll, Department of English, University of Missouri, St. Louis, MO, USA.
Jonathan Gottschall, Department of English, Washington & Jefferson College, Washington, PA, USA.
Daniel Kruger, Prevention Research Center, University of Michigan, Ann Arbor, MI, USA.
Abstract: The current research investigated the psychological differences between
protagonists and antagonists in literature and the impact of these differences on readers. It
was hypothesized that protagonists would embody cooperative motives and behaviors that
are valued by egalitarian hunter-gatherers groups, whereas antagonists would demonstrate
status-seeking and dominance behaviors that are stigmatized in such groups. This
hypothesis was tested with an online questionnaire listing characters from 201 canonical
British novels of the longer nineteenth century. 519 respondents generated 1470 protocols
on 435 characters. Respondents identified the characters as protagonists, antagonists, or
minor characters, judged the characters’ motives according to human life history theory,
rated the characters’ traits according to the five-factor model of personality, and specified
their own emotional responses to the characters on categories adapted from Ekman’s seven
basic emotions. As expected, antagonists are motivated almost exclusively by the desire for
social dominance, their personality traits correspond to this motive, and they elicit strongly
negative emotional responses from readers. Protagonists are oriented to cooperative and
affiliative behavior and elicit positive emotional responses from readers. Novels therefore
apparently enable readers to participate vicariously in an egalitarian social dynamic like
that found in hunter-gatherer societies. We infer that agonistic structure in novels simulates
social behaviors that fulfill an adaptive social function and perhaps stimulates impulses
toward these behaviors in real life.
Keywords: egalitarian groups, literature, social dominance, stigmatization.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Introduction

---

Egalitarian dynamics
Since the early 1990s, literary scholars have been assimilating the insights of
evolutionary psychology and envisioning radical changes in the conceptual foundations of
literary study. The “literary Darwinists” have produced numerous essays in literary theory
and criticism (Carroll 2004, 2005, 2007, 2008a, 2008b; Gottschall, 2008b; Gottschall and
Wilson, 2005). Until recently, though, most literary Darwinists have remained within the
methodological boundaries of traditional humanistic scholarship. Their work has been
speculative, discursive, and rhetorical. They have drawn on empirical research but have
not, for the most part, adopted empirical methods (Gottschall, 2005, 2008a.) The study
described in this article integrates literary Darwinism with empirical methodology.
Drawing on research in evolutionary psychology and related fields, we (a) deploy a model
of human nature—of motives, emotions, and features of personality, (b) use that model to
analyze a specific body of literary texts and the responses of readers to those texts, and (c)
produce data—information that can be quantified and can serve to test specific hypotheses
about those texts.
The Methods section gives a synoptic account of the model of human nature used to
derive the categories for this study. (A more extensive, discursive exposition of the model
can be found in J. Carroll, 2008a.) The literary texts in the present study are canonical
British novels of the longer nineteenth century (Jane Austen to E. M. Forster). The focal
point of the study is “agonistic” structure: the organization of characters into protagonists,
antagonists, and minor characters. The terms “protagonist” and “antagonist” are part of the
common parlance of literary discussion, and critics commonly distinguish between major
and minor characters. Such categories are key features in the organization of characters in
plots; they enter into emotional responses and are presupposed in generic structures, like
romantic comedy and tragedy, that are heavily inflected by emotional responses. In
inferring moral and ideological values in novels and plays, readers depend crucially on
recognizing protagonists and antagonists. And yet, very little literary theory focuses
primarily on the concept of agonistic structure, and that concept has never been tested for
empirical validity on any large scale.
The very existence of agonistic structure is a topic about which speculative opinion
could easily differ, and about which speculative arguments could go on endlessly and
inconclusively. Are characters actually divided in both authors’ intentions and readers’
responses into protagonists, antagonists, and minor characters? From one perspective, such
suppositions could be deprecated as naïve and misconceived, moralistic and simple-minded.
A theorist adopting this perspective might argue that characters in novels (just like real
human beings) possess both egocentric and cooperative dispositions, that they are too
complex to be neatly categorized into good guys and bad guys. Some such idea is at work
behind all contrasts between "serious" fiction, which depicts morally complex
characters, and "melodramas," which depict morally polarized good and bad characters.
(For a canonical instance, see Leavis, 1973, p. 19.) Or the theorist might argue that values
are context-dependent, so that what counts as good and bad alters with circumstances, with
varying cultural norms, and with differences in personal identity. (These latter
contentions, on the relativity of values, are consistent with "reader-response" theories and
with the forms of "cultural relativism" that have bulked so large in literary studies over the
past few decades. See Fish, 1980.)
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -716-

---

Egalitarian dynamics
From another perspective, the categories that make up agonistic structure could be
deprecated as so obvious, so self-evident, that they need no confirmation. From this second
perspective, a research design oriented to substantiating the existence of “good guys” who
are liked by readers and “bad guys” who are disliked by readers could not fail to produce
positive results, and would thus be trivial.
By themselves, the claims originating from either the perspective that it is too
simple-minded to think that agonistic categories exist or the perspective that agonistic
categories obviously exist might seem plausible enough. But because they contradict each
other, they cannot both be true. If it can plausibly be claimed that the idea of agonistic
structure is naïve and misguided, it cannot be the case that the idea of agonistic structure is
so obvious as not to need support or argument. The current research was designed to settle
the matter empirically, by testing for the existence of meaningful agonistic structure. If the
data confirmed that agonistic structure existed, evidence bearing on its function could be
gathered.
The central hypotheses were that agonistic structure exists and that it provides a
medium for exercising evolved dispositions for forming cooperative social groups. Within
the past decade or so, a wide range of evolutionists in diverse disciplines have made cogent
arguments that human social evolution has been driven in part by competition between
human groups. That competition is the basis for the evolution of cooperative dispositions—
dispositions in which impulses of personal domination are subordinated, however
imperfectly, to the collective endeavor of the social group. Suppressing or muting the sense
of competition within a social group enhances the sense of group solidarity and organizes
the group psychologically for cooperative endeavor (Alexander, 1979, pp. 220-35, 1987,
pp. 77-81, 233-84, 1989; Axelrod and Hamilton, 1981; Bingham, 1999; Boehm, 1999;
Cummins, 2005; Darwin, 1981, vol. 1, pp. 70-106; Deacon, 1997; Eibl-Eibesfeldt, 1998;
Flinn, Geary, and Ward, 2005; Geary, 2005, pp. 136-39, 142-44, 247-48; Kenrick, Maner,
and Li, 2005; Krebs, 2005; Kurzban and Neuberg, 2005; Premack and Premack, 1995;
Ridley, 1996; Richerson and Boyd, 1998, 2001, 2005; Salter, 2007; Schaller, Park, and
Kenrick, 2007; Smith, 2007, pp. 129-46; Sober and Wilson, 1998, pp. 159-95, 329-37;
Turchin, 2006; D.S. Wilson, 2006, 2007a, b, c; Wilson and Wilson, 2007).
The real-world pervasiveness of subordinating impulses toward personal domination
in order to form cooperative groups led to the prediction that, in novels, protagonists would
form communities of cooperative endeavor and that antagonists would exemplify
dominance behavior. If this hypothesis proved correct, the ethos reflected in the agonistic
structure of the novels would replicate the egalitarian ethos of hunter-gatherers, who
stigmatize and suppress status-seeking in potentially dominant individuals (Boehm, 1999).
If dispositions for suppressing dominance fulfill an adaptive social function, and if agonistic
structure in the novels reflects and reinforces dispositions for suppressing dominance, the
current research would lend support to the hypothesis that literature fulfills an adaptive
social function (Boyd, 2005; Carroll, 2005, 2007, 2008a, 2008b; Dissanayake, 2000;
Salmon and Symons, 2001 Scalise-Sugiyama, 2005; Tooby and Cosmides, 2001).
The ability of novels to serve an adaptive social function depends on readers
responding to characters in novels in much the same way, emotionally, as they respond to
people in everyday life. They like or dislike them, admire them or despise them, fear them,
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -717-

---

Egalitarian dynamics
feel sorry for them, or are amused by them. In writing fabricated accounts of human
behavior, novelists select and organize their material for the purpose of generating such
responses, and readers willingly cooperate with this purpose. They participate vicariously in
the experiences depicted and form personal opinions about the qualities of the characters.
Authors and readers thus collaborate in producing a simulated experience of emotionally
responsive evaluative judgment. If agonistic structure in the novels reflects the evolved
dispositions for forming cooperative social groups, the novels would provide a medium of
shared imaginative experience through which authors and readers affirm and reinforce
cooperative dispositions on a large cultural scale. (On literature as a form of “simulation,”
see Oatley, 1999, 2002; Tan, 2000, pp. 126-27. On the emotionally responsive character of
the reader’s experience, see J. Carroll, 2004, pp. 114-16, 126-27; N. Carroll, 1997; Feagin,
1997; Hogan, 2003; McEwan, 2005; Matravers, 1997; Oatley and Gholamain, 1997;
Özyürek and Trabasso, 1997; Storey, 1996, pp. 8-15; Tan, 2000; Van Peer, 1997. On the
parallel responses to “real” and “fictive” people, see Bower and Morrow, 1990; Grabes,
2004.)
The current research therefore goes far beyond testing the simple propositions that
good guys and bad guys exist and that readers like good guys more than bad guys. The
research first tests, quantitatively, for the existence of protagonists, antagonists, and minor
characters by measuring the degree of agreement in coding characters into the agonistic
categories. If there is no such thing as agonistic structure, there would be no basis for
systematically coding characters into these categories, so each coder would be using
idiosyncratic and arbitrary rules for the coding task, resulting in zero agreement across
coders. High levels of agreement would support the hypothesis that agonistic structure
exists.
Furthermore, by assessing the ways in which protagonists and antagonists differ, the
research was designed to discover why readers experience different emotional responses
toward protagonists and antagonists. The prediction was that differences in emotional
reaction could be explained by protagonists demonstrating higher degrees of cooperation
and antagonists, social dominance.
Materials and Methods
Participants
Potential research participants were identified by scanning lists of faculty in
hundreds of English departments worldwide and selecting specialists in nineteenth-century
British literature, especially scholars specializing in the novel. Invitations were also sent to
multiple electronic mailing lists dedicated to the discussion of Victorian literature or
specific authors or groups of authors used in the study.
All participation was anonymous, but those who accepted the invitation provided
the following identifying information on the on-line questionnaire used to collect data: sex,
age, level of education, how they had heard about the study, how recently they had read the
novel they were coding, and why they had read it. On the basis of this information,
identification strings were produced to calculate the total number of individual respondents
and segregate them into demographic categories. The data set contains a total of 519 unique
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -718-

---

Egalitarian dynamics
identification strings. Out of 519 unique coders, 178 (34%) were male and 341 (66%)
female. The youngest coder was 15; the oldest was 83, and the mean age was about 40. The
standard deviation for the age of coders was about 15 years. The majority of the
respondents thus ranged between 25 and 55 years of age. 81% of the respondents had a
bachelor’s degree or higher; 58% had advanced degrees; and 32% had doctorates. 52% of
the respondents had read the novel within the past year, and 85% within the past five years.
60% read the novel for their own enjoyment, 20% for a class they were taking, and 19% for
a class they were teaching.

Online questionnaire

A copy of the questionnaire used in the study can be accessed at the following
URL: http://www-personal.umich.edu/~kruger/carroll-survey.html. (The form is no longer
active and will not be used to collect data.) After providing identifying information,
research participants were directed to a list of roughly 2,000 characters from 201 canonical
British novels of the nineteenth century. Participants coded various attributes of the
characters of their choice and their emotional responses to the characters. Coded attributes
reported on in this study include (1) agonistic role assignment, (2) motives, and (3)
personality. (Other attributes such as mate preferences were coded but not used in the
present study.) Each of these attributes and the dimensions of emotional reaction are
described in more detail below.

Agonistic Structure
For each character selected, respondents assigned the character to one of the
following agonistic roles: (1) protagonist, (2) friend or associate of a protagonist, (3)
antagonist, or (4) friend or associate of an antagonist. Alternatively, respondents could
check “other” and thus decline to assign characters to agonistic roles. If respondents
differed on role assignment, the character was assigned to the role for which the majority of
respondents voted. In case of a tie, the character was not assigned to a role, but scores for
that character were still included in statistical correlations among categories of analysis
such as motives and personality factors.

The four agonistic roles form four sets of characters. We ourselves identified each
character as either male or female. Further dividing these four character sets into male and
female sets formed a total of eight character sets: male protagonists, female protagonists,
male associates of protagonists, female associates of protagonists, male antagonists, female
antagonists, male associates of antagonists, and female associates of antagonists.

For purposes of statistical analysis, the eight sets of characters were conceptualized
in terms of three underlying dimensions: Sex, Salience, and Valence. Under Sex, characters
are classified as “male” or “female.” Under Salience, characters are classified as “major” or
“minor.” Major characters are protagonists and antagonists. Minor characters are the
friends and associates of protagonists and antagonists. Under Valence, characters are
classified as “good” or “bad.” Good characters are protagonists and their friends and
associates. Bad characters are antagonists and their friends and associates. The designations
“good” and “bad” are a matter of convenience and simply follow popular usage. The use of
the terms has no pre-emptive moral content, but as it happens, in practice, the distribution
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -719-

---

Egalitarian dynamics
of characters into the good and bad sets is heavily inflected with morally relevant character
traits.

The classification of characters along the three dimensions of agonistic structure
allowed for a series of 2x2x2 multivariate analyses of variance in which sex, salience, and
valence were conceived as quasi-independent variables, while the other responses were
used as dependent variables. We predicted that characters identified as “good” would have
attributed to them, on average, the features associated with communitarian endeavor and
positive emotional reactions. Characters identified as “bad” would have attributed to them,
on average, the characteristics associated with personal dominance and negative emotional
reactions. We predicted further an interaction with salience: that good major characters
(protagonists) would most completely realize the approbatory tendencies in reader response
and that bad major characters (antagonists) would most completely realize the aversive
tendencies.

Motives
In devising a set of categories for motives, we sought to take account of the features
of human life history that have been preserved from our mammalian and primate lineage
(A. Buss, 1997; Lancaster and Kaplan, 2007; Low, 2000; Silk, 2007); the specifically
human reproductive characteristics that involve long-term pair-bonding, differing male-
female mate-selection strategies, paternal investment, and the existence of extended kin
networks (Bjorklund and Pellegrini, 2002; D.M. Buss, 2000, 2003; Deacon, 1997; Flinn
and Ward, 2005; Geary, 1998, 2005; Geary and Flinn, 2001; Kruger, Fisher, and Jobling,
2003; Salmon and Symons, 2001; Schmitt, 2005); evolved human dispositions for forming
coalitions, dominance hierarchies, and in-groups and out-groups (Alexander, 1987; Boehm,
1999; Cummins, 2005; Eibl-Eibesfeldt, 1998; Flinn, Geary, and Ward, 2005; Kurland and
Gaulin, 2005; Kurzban and Neuberg, 2005; Premack and Premack, 1995; Salter, 2007; D.S.
Wilson, 2006, 2007a,c); and the peculiarly human dispositions for acquiring and producing
culture (Baumeister, 2005; Carroll, 2004, 2008a, 2008b; Dissanayake, 2000; Hill, 2007;
Richerson and Boyd, 2005; Sober and Wilson, 1998; Sterelny, 2003; Tomasello, Carpenter,
Call, Behne, and Moll 2005; Tooby and Cosmides, 2001; E.O. Wilson, 1998).
Out of these features, we produced the following list of 12 motives: (1) Survival
(fending off imminent physical danger or privation); (2) Finding a short-term romantic
partner; (3) Finding or keeping a spouse; (4) Gaining or keeping wealth; (5) Gaining or
keeping power; (6) Gaining or keeping prestige; (7) Obtaining education or culture; (8)
Making friends and forming alliances; (9) Nurturing/fostering offspring or aiding other kin;
(10) Aiding non-kin; (11) Building, creating, or discovering something; and (12)
Performing routine tasks to gain a livelihood.
Respondents were asked to rate each character on each motive, on a scale of 1 to 5,
with 1 being “unimportant” and 5 “very important.” We predicted that protagonists would
be generally affiliative in their motives—concerned with helping kin and making friends—
and we predicted that antagonists would be chiefly concerned with acquiring wealth,
power, and prestige. We predicted that protagonists would on average be much more
concerned than antagonists or minor characters with acquiring education and cultural
knowledge. Taking into account both adaptively conditioned sex differences and the sex-
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -720-

---

Egalitarian dynamics
differenced social roles in the period of these novels, we predicted that female characters
would be more interested in marriage and family than the male characters. We predicted
also that male characters would be more oriented to activity in the public domain.
Because the total number of variables in the study was large, the data set was
simplified by reducing the number of variables through factor analyses. To that end, a
principal components factor analysis was conducted on averaged ratings of the 12 motives.
No predictions were made on the precise number of factors that would emerge nor the
loadings of every variable on the factors, but we did expect the motives related to personal
gain (wealth, power, prestige) to mark a factor, motives related to prosociality (making
friends, helping non-kin) to mark a separate factor, and motives related to romance and
family to mark additional factors.

Personality Factors
The predominant conceptualization of personality at the present time is the five-
factor or “big five” model of personality (Costa and McCrae, 1997; Figueredo et al., 2005;
Nettle, 2006, 2007.) Many instruments of the five factors are available. To keep the online
questionnaire as short as possible, thereby encouraging participation in the study, the Ten
Item Personality Inventory (TIPI; Gosling, Rentfrow, and Swann, 2003) was chosen.
Gosling et al. have documented that the TIPI possesses adequate psychometric reliability
and validity. Under the lead-in phrase, “I see this character as,” respondents scored each
character on each of ten attributes (two for each factor). Ratings were on a seven-point
scale ranging from “disagree strongly” to “agree strongly.” We predicted that protagonists
and their friends would on average score higher on the personality factor Agreeableness, a
measure of warmth and affiliation. We also predicted that protagonists would score higher
than antagonists and minor characters on the personality factor Openness to Experience, a
measure of intellectual vivacity.

Emotional Responses
In building emotional responses into our research design, we sought to identify
emotions that are universal and that are thus likely to be grounded in universal, evolved
features of human psychology. We started with a core set of seven terms from Ekman’s
(2003) list of basic emotions and adapted those terms for the purpose of registering graded
responses specifically to persons or characters. Four of the seven terms were used
unaltered: anger, disgust, contempt, and sadness. Fear was divided into two distinct items:
fear of a character, and fear for a character. To adapt the terms “joy” and “enjoyment,” to
make them idiomatically appropriate as a response to a person and also to have it register
some distinct qualitative differences, we chose two terms, “liking” and “admiration.”
“Surprise,” like “joy,” seems more appropriate as a descriptor for a response to a situation
than as a descriptor for a response to a person or character. Consequently, we did not use
the word “surprise” by itself. Instead, we used “amusement,” which combines the idea of
surprise with an idea of positively valenced emotionality. We included one further term in
our list of possible emotional responses: indifference. A number of researchers have
included a term such as “interest” to indicate general attentiveness, the otherwise
undifferentiated sense that something matters, that it is important and worthy of attention
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -721-

---

Egalitarian dynamics
(Plutchik, 2003). Respondents gave a score on each of the ten emotions, on a scale from 1
to 5, with 1 signifying “not at all” and 5 “very strong.”

We predicted that protagonists would receive high scores on the positive emotional
responses “liking” and “admiration” and that antagonists would receive high scores on the
negative emotions “anger,” “disgust,” “contempt,” and “fear-of” the character. We
predicted that good major characters (protagonists) would most completely realize the
approbatory tendencies in reader response and that bad major characters (antagonists)
would most completely realize the aversive tendencies. We also predicted that good
characters would score higher on “sadness” and “fear-for” the character than bad
characters. We predicted that major characters would score lower on “indifference” than
minor characters.

Again, to simplify analyses by reducing the total number of variables, a principal
components factor analysis was conducted on the averaged emotional response ratings.
Previous factor analyses of emotions and mood terms usually locate a factor of positive
emotions and a factor of negative emotions. We predicted that such factors would obtain in
our data.
Results
Reliability Estimates
Coefficient alpha estimates of inter-coder reliability were computed for a sample of
characters that were coded by two or more respondents. As expected, measurement
reliability increased as the number of judges increased. Consider the following list of
characters with the number of coders and corresponding Cronbach alphas: Adam Bede (2,
0.73); Weena [no surname] (3, 0.83); Augusta Elton (7, 0.94); Elizabeth Bennett (81, 0.99).
Several observations can be drawn about these findings. First, the reliability coefficients are
remarkably high, indicating that the respondents took the task seriously and provided high
quality data. Reliability coefficients from as few as two coders are above .70, clearly in a
psychometrically acceptable range. Although reliabilities for characters based upon one
coder cannot be computed, it is not unreasonable to assume that these coders also took their
task seriously. Finally, these high coefficient alpha coefficients justify averaging the
responses for characters who were judged by two or more respondents (almost half of the
characters in the study).
Out of the total of 435 characters who were coded, 53 characters (12%) were not
included in character sets—22 characters who were singly coded and not assigned a role
(respondents checked “other” or “I do not remember”), two characters who had two
codings each and were not assigned a role, and 29 characters who tied in role assignments.
The remaining 382 characters (88% of 435) were assigned to character sets. To calculate
the level of consensus in assigning characters to roles, the total number of respondents over
the whole range of characters who agree with the majority in assigning a character to a role
was divided by the total number of respondents. If missing values (“other” and “I do not
remember”) are retained, the average consensus rating for all 206 multiply coded characters
is 81%. If missing values are eliminated, the average consensus rating for all 206 multiply
coded characters is about 87%. This high degree of consensus affirms the existence of
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -722-

---

Egalitarian dynamics
agonistic structure.
To assess whether male and female respondents vary in their responses to male and
female characters, a set of 117 characters in which each character was rated by at least one
male and one female coder was analyzed. In cases in which there were more than one male
or female coder, the scores from all respondents of either sex were averaged. A 2x2x2x2
analysis of variance with four variables: Coder Sex, Character Sex, Valence, and Salience
(male and female respondents; male and female characters; good and bad characters, and
major and minor characters) was conducted. The result was unequivocal. Across all of the
dependent variables, the sex of respondents simply did not matter. Coder sex had neither
main effects nor interactive effects with the other independent variables.

Specific Categories of Analysis
Motives
Factor analysis of the 12 motives produced five motive factors. See Table 1.
Wealth, power, and prestige all have strong positive loadings on Social Dominance, and
helping non-kin has a moderate negative loading. Constructive Effort is defined by strong
loadings from the two cultural motives, seeking education or culture, and creating,
discovering, or building something, and it also has substantial loadings on two pro-social or
affiliative motives: making friends and alliances and helping non-kin. Romance is a mating
motive, with a chief loading from short-term mating and long-term mating. A secondary
loading on this factor for acquiring wealth reflects a sex-specific female mate-selection
preference. Subsistence combines two motives—survival, and performing routine tasks to
gain a livelihood. Nurture is defined primarily by a significant positive loading for
nurturing/fostering offspring or other kin and a negative loading from short-term mating.
Helping non-kin also loads moderately on this factor, bringing affiliative kin-related
behavior into association with generally affiliative social behavior.
Factor scores with a mean of zero and standard deviation of one were saved and
used as dependent variables in subsequent multivariate analyses of variance employing
Sex, Valence, and Salience as factors. In cases where the multivariate results were
statistically significant, follow-up F-tests identified significant effects on each dependent
variable separately. In cases of statistical interaction, univariate post-hoc comparisons with
Bonferroni adjustment were conducted within levels of a factor to identify simple main
effects.
The Wilks Lambda multivariate test of overall differences among groups was
statistically significant for Sex (F5,365 = 3.80, p = 0.002), Valence (F5,365 = 20.21, p <
0.001), Salience (F5,365 = 5.95, p < 0.001), and the Valence * Salience interaction (F5,365 =
3.91, p = 0.002).
The follow-up test for Social Dominance showed a Valence * Salience interaction
effect (F1,369 = 13.82, p < 0.001). Post-hoc comparisons showed that antagonists score
significantly higher on Social Dominance than protagonists (0.88 vs. -0.15, F1,372 = 55.12, p
< 0.001); and antagonists higher than bad minor characters (0.88 vs. -0.21, F1,372 = 20.50, p
< 0.001). (The mean scores reported here and elsewhere are estimated marginal means.)
For Constructive Effort, significant main effects for Valence and Salience were
found. Post-hoc comparisons showed that good characters score significantly higher on
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -723-

---

Egalitarian dynamics
Constructive Effort than bad characters (0.26 vs. -0.67, F1,369 = 47.78, p < 0.001); and
major characters higher than minor characters (-0.03 vs. -0.37, F1,369 = 6.21, p = 0.013).

Table 1. Factor Analysis of Motives-Rotated Component Matrix

Social Constructive
Motives Dominance Effort Romance Subsistence Nurture

Survival
0.12
-0.05
-0.01
0.80
-0.02
Short-Term Mating
0.07
0.13
0.63
-0.07
-0.56
Long-Term Mating
0.02
0.05
0.83
0.00
0.16
Wealth
0.70
-0.27
0.38
0.22
0.00
Power
0.89
0.01
-0.08
0.00
-0.16
Prestige
0.89
0.13
0.02
-0.05
0.02
Education or Culture
0.08
0.77
0.18
0.00
-0.02
Friends and Alliances
0.12
0.62
0.28
0.01
0.26
Helping Kina
-0.06
0.11
0.12
-0.02
0.82
Helping Non-Kin
-0.34
0.56
-0.06
0.14
0.41
Creating, Discovering b
-0.10
0.73
-0.28
0.13
-0.10
Routine Work c
-0.08
0.20
-0.01
0.76
0.05
Notes: Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser
Normalization. Rotation converged in 7 iterations. The five factors accounted for 69% of the total
variance: Social Dominance 21.5%; Constructive Effort 17.3%; Romance 11.3%; Subsistence
9.7%; Nurture 9.1%. Loadings > ± 0.3 in bold font.
a The whole phrase in the questionnaire was “Nurturing/fostering offspring or aiding other kin.”
b The whole phrase in the questionnaire was “Building, creating, or discovering something.”

Thus, a primary hypothesis of the study was confirmed: good characters differ from
Evolutionary Psychology – ISSN 1474-7049 – Volume 6(4). 2008. -724-

---