EMPIRICAL STUDIES OF THE ARTS, Vol.

29(1) 89-109, 2011

THE READER IN THE TEXT: THE CONSTRUCTION

OF LITERARY CHARACTERS

ALDO NEMESIO
University of Turin

M. CHIARA LEVORATO
LUCIA RONCONI

University of Padua

ABSTRACT

When we read the description of a character, we receive explicit information

and construct a semantic representation of his/her aspect and personality.
This constructive process also involves inferential processes based on the
elaboration of explicit information. In this article we examined the first
introduction of four female characters in Italian novels. When readers
reported information that was not given in the texts, we had evidence of their
constructive processes based on their inferences and world knowledge. In
our study readers tended to concretize traits that were not explicitly stated
in the text, with high values of the Gini index, showing different concreti-
zations in different readers. This tendency was quite pervasive among
readers, regardless of gender, academic education and interest in the text read;
such constructive process seems to be intrinsic to the act of reading itself.

In literary theory, there is general agreement that characters in narrative texts

are constructed by readers by means of inferences, making use of readers’
individual competence and experience (Rimmon-Kenan, 1983, p. 36; van Peer,
1988, p. 9). In a theoretical, but non empirical, book Jouve (1992) claimed that

89

Ó 2011, Baywood Publishing Co., Inc.

doi: 10.2190/EM.29.1.f
http://baywood.com
90 / NEMESIO, LEVORATO AND RONCONI

literary characters are the product of the interaction between text and readers;
for this reason, the critic’s point of view should be replaced by a study of the act
of reading. Culpeper (1996) focused on the issue of how and when readers infer
information about characters, comparing foregrounding and attribution theories.
Several literary theorists and psychologists agree that the information provided
at the beginning of a text is particularly compelling (Gerrig & Allbritton, 1990,
p. 385; Jouve, 1992, p. 52; Stanzel, 1981, p. 9). The constructed character may
vary during the act of reading (Emmott, 1997, p. 181).
In empirical studies, the issue of the construction of characters has been
investigated by few researchers. Gernsbacher, Goldsmith, and Robertson
(1992) and De Vega, Leon, and Diaz (1996) studied readers’ comprehension of
emotions experienced by characters as consequences of narrated events, showing
that readers form an explicit lifelike mental representation of fictional characters’
emotional states, even when there is no explicit information in the text. In both
studies, the passages presented to the participants were “textoids,” that is texts
produced on the occasion of the experiment (Zwaan, 1996). The use of textoids
limits the generalizability of the results to natural texts. Nevertheless, it is pre-
sumable that similar constructive inferential processes also occur in the case of
literary texts; the construction of a mental representation that includes infor-
mation not explicitly provided in the text is part of reading comprehension.
Since the goal of reading is to build up a coherent mental representation of what
the text is about, missing information is retrieved by readers using their previous
world knowledge.
Bortolussi and Dixon (2003) based their work on empirical evidence and
remarked that literary characters are processed much in the same way as real
people. In addition:

the constructive operations of readers are so significant that they render it

difficult, if not impossible, for readers to distinguish between what is actually
said about characters in the text and what they themselves have supplied [. . .]
It is entirely conceivable that some readers impose on the text character
constructions that are discordant with the text (Bortolussi & Dixon, 2003,
p. 153).

In previous work (Nemesio, 2002a, 2002b, 2008; Nemesio, Levorato, &

Ronconi, 2008) which examined what readers remember of the description of
characters presented in a novel, empirical evidence was provided, showing dif-
ferences in readers’ constructions after reading literary passages. On one hand,
the use of literary texts, rather than textoids, is more ecologically valid; on the
other hand, information given in the texts cannot be completely controlled.
Since a certain amount of ambiguity is often present in literary texts, in particular
when emotions and internal states are described, we examined the process
of inference production concerning rather concrete traits, such as physical
appearance, age, evident personality traits, income, and educational level.
 THE CONSTRUCTION OF LITERARY CHARACTERS / 91

The present investigation was aimed at exploring the inferential processes

that accompany the reading of the description of characters, checking what
explicit information is retained and what non-explicit information is supplied by
readers. We expected that readers remember information that was not explicitly
given in the text. We interpret this phenomenon as a result of constructive
processes that operate during the reading of literary texts; readers, in fact, integrate
actual information with previous world knowledge and connect various pieces
of information provided in the text with one another. These inferential processes
allow the construction of a coherent mental representation of what the text is
about (Kintsch, 1998). The current study was aimed at analyzing the nature of
such integrative processes. Readers’ concretizations of four female characters
were explored. Differences in constructions were examined in relation to
various readers’ characteristics: gender (male vs. female), university education
(Humanities vs. Applied Sciences), and degree of interest experienced while
reading. Every time we found information that was not given in the texts, we
had evidence of material coming from readers’ inferences and world knowledge.
We hypothesized that readers’ characteristics might affect the perception of the
characters via their inferences and world knowledge.

METHOD
Participants
Four hundred and six students (193 males, 213 females) participated in the
present study. They attended either faculties of Humanities (N = 223) or of
Applied Sciences (N = 183) at the University of Turin. Their ages ranged from 18
to 34 (M = 21.46; SD = 2.38; 3.4% older than 26). Faculties of Humanities include
several courses on Italian literature that are not taught in faculties of Applied
Sciences (such as engineering, architecture, or medicine).
Material
Four passages introducing female characters in four Italian novels were
used. The characters were: Lucia in Manzoni’s (1840) The Betrothed; Fosca in
Tarchetti’s (1869) Passion; Angiolina in Svevo’s (1898) As a man grows older;
and Carla in Moravia’s (1929) The time of indifference. The four passages and
their translation into English are reported in the Appendix. Since the passages
are real literary texts, they differ in length (Manzoni: 220 words, Tarchetti: 230,
Svevo: 74, Moravia: 59). The four novels were chosen as examples of 18th- and
19th-century Italian narrative literature, often taught in schools and universities.
The four female characters are central in the novels’ plots.
Manzoni’s Lucia is presented when she is meeting her friends after her mother
has just finished dressing her up for her wedding day. She is a “peasant girl”
of “modest beauty.” Various emotions appear in her face: in particular “a great
happiness.” She has “dark young hair.” Details of her wedding apparel are given.
92 / NEMESIO, LEVORATO AND RONCONI

Fosca is presented by Tarchetti as an extraordinarily ugly woman. In addition,

she is very ill. But, on the basis of her description, we know very little for
certain: the text contains several evaluations, to be interpreted according to
readers’ criteria and expectations. Her “ugliness” is mainly due to her illness,
that devastated her face (“all of her horror was in her face”).
Svevo’s Angiolina is described as a very pretty woman. Readers find infor-
mation about various aspects of the character, in particular her hair, eyes, height,
figure, face, and health. She is “a tall, healthy blonde, with big blue eyes
and a supple, graceful body, an expressive face and transparent skin glowing
with health.” Angiolina is presented in the open air, while she walks with an
“elegant parasol.”
Carla is introduced in the beginning of Moravia’s novel. Her description
centers around her clothing, emphasizing her sensuality. She is wearing “a
brown woolen frock.” Her skirt is “so short that the movement she made in
shutting the door was enough to pull it up several inches over the slack wrinkles
her stockings formed about her legs,” but she is “unaware of this.” Her bearing is
awkward: she moves “clumsily and uncertainly forward.” In addition to this
information, allusive rather than descriptive, readers are not told anything about
her physical appearance, psychological traits, economic situation and education.

Procedure

In order to explore readers’ inferences, participants were asked to answer

some questions concerning the four female characters, after reading passages
describing them. The three authors of this experiment agreed on a list of traits
commonly associated with the description of characters in a narrative text. Not
all the traits were explicitly given in each passage. Eight traits were chosen:
four concerning characters’ appearance (dress, hair color, height, and size), two
concerning psychological traits (whether the character was calm or nervous,
boring or brilliant), and two concerning socioeconomic status (economical situ-
ation and education).
During regular classes in university classrooms, each student was given one
of the texts. The participants were told that it was the introduction of a character
in a novel, but were not given any further information, such as title, name of the
author or year of publication. They were told that the test was aimed at studying
reading behavior, but did not aim at evaluating their skills. They were asked to
read at their normal speed. Approximately 100 participants read each passage
(Manzoni: N = 103; Tarchetti: N = 104; Svevo: N = 104; Moravia: N = 95).
After reading, the participants returned the copies of the texts. A questionnaire
followed, asking their interest in the text, namely whether they would like to
continue reading the book: answers were given on a 5-point scale (1 = not at all,
5 = very much). Then, participants were asked to estimate the character’s age.
Questions concerning the eight traits followed, in the form of multiple choice
 THE CONSTRUCTION OF LITERARY CHARACTERS / 93

questions. The three choices presented to the readers the extreme poles and the
medium position of the answer (e.g., tall, short, and medium height; rich, poor, and
average economic condition; etc.), with the exception of hair color (which could
be blond, black, or red), and emotional condition (calm, quite calm, and nervous).
Lastly, the subjects were asked in what season of the year the novel was set.
Since some questions concerned information that was not explicitly stated in
the passages, the subjects were allowed to answer “I do not know” and they
were informed that this answer was perfectly legitimate in some cases. In all the
cases in which the information was not given explicitly in the passages and
the readers gave an answer different from “I do not know,” we had information
on their inferential processes and their construction of the character based on
their world knowledge and expectations.
The participants were also asked if they had already read the passage on
some other occasion. We did not consider the answers coming from subjects who
were already familiar with the text or were not native Italian speakers (114
questionnaires were excluded). The test was anonymous; the participants were
only asked to indicate their age and gender.

RESULTS

Interest in the Passages

The first analysis concerned the degree of interest experienced by the
participants. Tables 1 and 2 show the average interest in the text experienced by
male and female students and by students attending faculties of Humanities and
Applied Sciences respectively.
A 4 × 2 × 2 ANOVA was carried out on the scores on the 5-point scale
concerning the degree of interest experienced as dependent variable and the four

Table 1. Interest Experienced by Male and Female Readers

Male Female

Text M SD M SD

Manzoni (NM = 41; NF = 62) 2.05 0.74 2.66 0.96

Tarchetti (NM = 46; NF = 58) 2.61 0.95 3.24 0.71
Svevo (NM = 57; NF = 47) 2.35 1.01 2.91 0.86
Moravia (NM = 49; NF = 46) 2.65 1.16 3.00 0.94
M (NM = 193; NF = 213) 2.42 1.01 2.95 0.89
Note: NM: number of males; NF: number of females; Maximum score = 5.
94 / NEMESIO, LEVORATO AND RONCONI

Table 2. Interest Experienced by Students of Humanities

and Applied Sciences

Humanities Applied Sciences

Text M SD M SD

Manzoni (NH = 59; NAS = 44) 2.69 0.84 2.05 0.91

Tarchetti (NH = 61; NAS = 43) 3.07 0.79 2.81 0.98
Svevo (NH = 55; NAS = 49) 2.76 0.96 2.43 0.98
Moravia (NH = 48; NAS = 47) 2.90 1.08 2.74 1.07
M (NH = 223; NAS = 183) 2.86 0.91 2.51 1.03
Note: NH: number of students of Humanities; NAS: number of students of Applied
Sciences; Maximum score = 5.

passages, the readers’ gender and their type of studies (Humanities vs Applied
Sciences) as between subject factors. The three main factors yielded significant
effects. The factor Passage, F(3, 390) = 9.08, p = < .001, hp2 = .06, showed that
Tarchetti’s passage was considered as more interesting than Manzoni’s
(Bonferroni post-hoc p < .001) and Svevo’s (Bonferroni post-hoc p = .034)
passages and was as interesting as Moravia’s passage. Since Tarchetti’s and
Moravia’s passages are the longest and the shortest respectively, this result
suggests that the length of the passage did not affect the degree of interest
experienced by the readers. The main effect of the readers’ Gender showed
that women reported to be more interested than men, F(1, 390) = 25.17, p < .001,
hp2 = .06. The factor Type of studies showed that Humanities students were
more interested than Applied Sciences students, F(1, 390) = 4.09, p = .029,
hp2 = .01. No interaction between factors yielded significance.

Characters’ Traits: Explicit Information

Two authors of this article analyzed the four passages, with the aim of identi-
fying the information presented explicitly in the texts. A complete agreement was
found in 95% of cases (Cohen’s kappa: k = 0.93); the two conflicting judgments
were discussed until an agreement was reached. It emerged that Moravia’s short
passage did not contain any explicit information concerning the eight traits,
whereas for the other three texts only dress, hair color, height and size were
explicitly mentioned. Table 3 shows readers’ percentage of correct answers
regarding information explicitly presented in the passages. In most cases readers
chose the correct answer for traits explicitly given in the passage, showing that
the comprehension of what they read was good. Tarchetti’s Fosca is described
 THE CONSTRUCTION OF LITERARY CHARACTERS / 95

Table 3. Explicit Information in the Passages

Percentage of
Text Traits correct answers

Manzoni Dress (elegant) 95%*

Hair color (dark-haired) 90%*

Tarchetti Dress (elegant) 60%–

Hair color (dark-haired) 89%*
Height (tall) 44%–
Size (slim) 93%*

Svevo Dress (elegant) 79%*

Hair color (dark-haired) 94%*
Height (tall) 82%*
Size (slim) 58%–
Note: The asterisk indicates the cases in which the difference between correct and
incorrect or “I do not know” responses was significant; in Manzoni’s passage: Dress (c2 =
83.97, df = 1, p < .05) and Hair color (c2 = 66.88, df = 1, p < .05); in Tarchetti’s passage: Hair
color (c2 = 92.35, df = 1, p < .05) and Size (c2 = 77.89, df = 1, p < .05); in Svevo’s text:
Dress (c2 = 34.62, df = 1, p < .05), Hair color (c2 = 81.39, df = 1, p < .05), and Height
(c2 = 41.89, df = 1, p < .05). The minus symbol indicates the cases in which such a
difference was not significant; in Tarchetti’s passage: Dress (c2 = 3.85, df = 1, p = .05)
and Height (c2 = 1.39, df = 1, p = .24); in Svevo’s text: Size (c2 = 2.46, df = 1, p = .12).

as very ill and ugly, two very negative traits that probably overshadowed the
recollection of other traits such as her being elegant and tall. Svevo’s Angiolina
is described as “alta e forte, ma snella e flessuosa” (“tall and strong, but slim and
supple”). This concentration of adjectives, and in particular the presence of
“tall and strong,” probably confused the recollection of “slim.” It seems plausible
that the errors were not casual, but rather based on inferential processes.
Next we explored whether the correct recollection of explicit information was
related to readers’ characteristics such as gender, education and interest in the
text. As for Tarchetti’s passage, the recollection of Size was more accurate in
women than in men (98% vs 87%; c2 = 5.24, df = 1, p < .05) and in students of
Humanities than in students of Applied Sciences (98% vs 86%; c2 = 6.09, df = 1,
p < .05); the recollection of Dress was more accurate in students who declared
high (4-5) or medium (3) interest in the text, than in those who declared low
(1-2) interest (68%, 63%, and 35%, respectively; c2 = 7.81, df = 2, p < .05). As
for Svevo’s passage, the recollection of Height was more accurate in women
than in men (92% vs 74%; c2 = 5.47, df = 1, p < .05).
In general, readers had a good comprehension of explicitly given infor-
mation, showing that they used it to construct a coherent and faithful picture of
96 / NEMESIO, LEVORATO AND RONCONI

the character. Nevertheless, a tendency seems to emerge suggesting that women,

students of Humanities, and readers more interested in the text produce more
accurate recollections than men, students of Applied Sciences, and less inter-
ested readers.

Characters’ Traits: Non-Explicit Information

The next analysis concerned characters’ traits that were not explicitly presented
in the passages. Before administering the test, participants were told that they
could use their inferences in answering and that they could choose not to answer a
question when they thought that the text did not contain the relevant information.
Therefore, when readers chose an answer related to information that was not
given in the texts, we have evidence of material coming from their inferences
and world knowledge. This conclusion is also legitimated by the results of the
analysis of the answers concerning explicit information which demonstrated that
errors were scarce. Table 4 lists the traits for which the texts did not give any
explicit information and the Gini index (described below) corresponding to the
percentages of readers’ choices.
In order to explore systematic tendencies in readers’ responses, chi-square test
and Gini coefficient were used. Then we explored whether the construction of
non-explicit information was related to readers’ characteristics such as gender,
education, and interest in the text.
In the case of Manzoni’s Lucia, missing answers were very scarce, suggesting
that readers believed they had enough information to choose a specific answer.
She was perceived as being rather slim (c2 = 34.74, df = 2, p < .05) and of average
height (c2 = 190.32, df = 3, p < .05) and sociability (c2 = 81.89, df = 3, p < .05).
The other features attributed to Lucia were poor (c2 = 38.35, df = 2, p < .05) and
uneducated (c2 = 64.18, df = 3, p < .05), while the possibility that she could be
calm (chosen by only 16% of readers) was excluded (c2 = 46.09, df = 3, p < .05).
These were legitimate inferences, given that she is a peasant girl described in
the day of her marriage.
Tarchetti’s Fosca was concretized as calm (c2 = 62.46, df = 3, p < .05), rich
(c = 24.31, df = 3, p < .05), and educated (c2 = 34.85, df = 3, p < .05); most of the
2
readers believed they did not have enough information about her sociability
(missing plus average = 81%).
Svevo’s Angiolina was considered as calm (c2 = 67.23, df = 3, p < .05), not
boring (c2 = 26.62, df = 3, p < .05), and rich (c2 = 63.60, df = 2, p < .05);
the possibility that she might be uneducated was ruled out (1% of the choices;
c2 = 35.46, df = 3, p < .05).
Moravia’s text does not contain any explicit information concerning the
eight traits considered in the current study. Missing or average answers were
the most frequent choice for traits concerning Carla’s appearance, such as
dress (c2 = 41.38, df =3, p < .05), hair color (c2 = 34.73, df = 3, p < .05), and height
 Table 4. Information Not Given Explicitly in the Passages and Readers’ Constructions

Text Traits Gini index Readers’ constructions Missing

Manzoni Height .38 tall: 2% short: 3% average: 83% 12%

Size .72 slim: 52% fat: 0% average: 41% 7%
Emotional condition .85 nervous: 38% calm: 16% quite calm: 44% 2%
Sociability .73 boring: 4% brilliant: 25% average: 61% 10%
Economic situation .72 rich: 6% poor: 54% average: 40% 0%
Education .79 educated: 2% uneducated: 50% average: 38% 10%

Tarchetti Emotional condition .80 nervous: 11% calm: 58% quite calm: 10% 21%
Sociability .86 boring: 6% brilliant: 12% average: 40% 41%
Economic situation .92 rich: 42% poor: 10% average: 29% 19%
Education .89 educated: 42% uneducated: 3% average: 23% 32%

Svevo Emotional condition .78 nervous: 10% calm: 59% quite calm: 22% 9%
Sociability .91 boring: 5% brilliant: 30% average: 39% 26%
Economic situation .62 rich: 70% poor: 0% average: 15% 14%
Education .89 educated: 29% uneducated: 1% average: 30% 40%

Moravia Dress .85 bad: 22% elegant: 20% normal: 52% 6%

Hair color .88 fair: 16% dark: 28% red: 8% 48%
Height .84 tall: 19% short: 0% average: 46% 35%
Size .77 slim: 58% fat: 1% average: 18% 23%
Emotional condition .87 nervous: 51% calm: 23% quite calm: 18% 8%
Sociability .84 boring: 4% brilliant: 13% average: 37% 46%
THE CONSTRUCTION OF LITERARY CHARACTERS

Economic situation .84 rich: 3% poor: 15% average: 47% 35%

Education .83 educated: 10% uneducated: 10% average: 27% 53%
/ 97
98 / NEMESIO, LEVORATO AND RONCONI

(c2 = 10.76, df = 2, p < .05), as well as for traits such as sociability (c2 = 44.83,
df = 3, p < .05), economical situation (c2 = 44.75, df = 3, p < .05), and education
(c2 = 49.80, df = 3, p < .05). On the other hand, readers seemed to be quite inclined
to perceive her as slim and nervous (c2 = 64.96, df = 3, p < .05 and c2 = 37.25,
df = 3, p < .05, respectively).
In order to explore the distribution of readers’ answers concerning traits that
were not explicitly mentioned in the passages, the Gini index was applied to the
proportions with which each answer was chosen. The Gini index ranges from
0 to 1: the higher the value, the stronger the tendency of different participants
to choose different answers. Table 4 reports the Gini index and the percentage
obtained by the different answers. The values of the Gini index computed for
the traits that are not explicitly described in the passages tended to be high. This
result suggests that different readers may produce different concretizations of
the same character on the basis of their experiences, expectations and world
knowledge. While the tendency to produce inferences is general, the specific
content of such inferences may vary among the readers. This tentative conclusion
suggested to us that it was necessary to identify the latent semantic structure of
the characters, based on their features, and interpret it on the basis of the readers’
characteristics considered in the present investigation.

Multiple Correspondence Analysis

With the purpose of identifying the salient traits of each character and trace
the dimensions that underlie these traits, for each text a multiple correspondence
analysis was carried out (Benzécri, 1976). The variables considered are the eight traits
(dress, hair color, height, size, emotional condition, sociability, economical situation,
and education) used to construct the factorial space in which the illustrative variables
are projected: readers’ gender, university education (Humanities vs. Applied
Sciences), and degree of interest experienced (Interest Low referring to a degree of
interest equal to 1 or 2; Interest Medium equal to 3 and Interest High equal to 4 or 5).
Figures 1, 2, 3, and 4 show the results of these analyses: the illustrative variables
referring to the reader’s characteristics are printed in capital letters and the features of
the characters presented in the questionnaire are printed in small letters. The position
in the factorial space is derived from the factorial analysis.
For all the passages the same two factors emerged. The first one correctly
explained a proportion of global inertia (Greenacre, 1984) ranging from 15%
in the case of Fosca (Tarchetti), to 20% in the case of Carla (Moravia). The second
factor explained a proportion of global inertia ranging from 11% for Carla and
Fosca to 13% for Angiolina (Svevo). Each figure represents in a bidimensional
space the two factors that emerged. For all the passages, the first factor (Dim 1)
refers to the tendency either to choose the answer “I do not know” (missing
answer) or to place the character in a medium position as to the various traits
(average answer). This dimension shows that, in the case of traits not explicitly
 THE CONSTRUCTION OF LITERARY CHARACTERS / 99

Figure 1. Multiple correspondence analysis for Manzoni’s

character: Lucia.

mentioned in the text, participants who preferred not to answer “I do not know” made
the inference that the most appropriate answer was the medium one. It might be
hypothesized that, when a piece of information was not given in the text, readers
tended to infer that it was not relevant and therefore unmarked. This inference led
them to choose the average answer. This shows that the process of inference making is
a parsimonious one. When it is not necessary, the readers avoid producing a specific
inference, for instance the determination of a particular trait.
The second factor (Dim 2) refers to the answers that differ both from “I do not
know” and the average option. This factor identifies different structural relation-
ships for the different passages and is affected in different ways by the readers’
characteristics. As for Manzoni’s Lucia, students of Humanities tended to attribute
traits such as poor, uneducated, and nervous, while students of Applied Sciences
tended to perceive the character as calm, with a medium economic and cultural
position, or even brilliant. Gender and interest did not seem to affect the perception
of the character. Tarchetti’s Fosca was perceived as elegant, rich, tall, and edu-
cated by women. Svevo’s Angiolina was considered as educated, rich, brilliant,
and nervous by readers with high interest in the text. The concretization of
Moravia’s Carla differentiates between a positive description of the character
100 / NEMESIO, LEVORATO AND RONCONI

Figure 2. Multiple correspondence analysis for Tarchetti’s

character: Fosca.

through traits such as elegant and calm and a negative characterization with
traits such as poor, inelegant, and nervous. Readers’ characteristics and interest
did not affect the perception of this character: the information provided in
Moravia’s passage was very scarce and readers seem to differentiate on the basis
of individual preferences.
The analyses performed on the non-mentioned traits proved to be useful in
order to understand the processes through which the readers construct a mental
representation of the characters. An elaborative process takes place, where
previous knowledge is integrated with explicit information in order to construct
a coherent representation of the main traits of the character. Age is a relevant
feature for defining a character. Readers’ inferences about character’s age are
analyzed in the following section.

Characters’ Age

The four passages do not indicate the characters’ ages explicitly. Table 5 shows
readers’ inferences on the characters’ age. Manzoni’s Lucia is a young woman
 THE CONSTRUCTION OF LITERARY CHARACTERS / 101

Table 5. Readers’ Inferences on Characters’ Age

Text Min Max M SD

Manzoni (N = 102) 16 30 21.75 3.43

Tarchetti (N = 99) 18 55 34.56 8.40
Svevo (N = 101) 17 37 24.26 4.34
Moravia (N = 95) 15 40 24.61 6.55
Note: Nine outliers were excluded.

on the day of her marriage. Our readers attribute her an average age of 22 years,
ranging from 16 to 30.
Tarchetti’s Fosca is generically presented as youthful (“her still youthful
person”), but not very young (“one could not possibly believe that she had
ever been beautiful, but it was evident that her ugliness was for the most part
the effect of the illness, and that, when a girl, she was perhaps pleasant”). In

Figure 3. Multiple correspondence analysis for Svevo’s

character: Angiolina.
102 / NEMESIO, LEVORATO AND RONCONI

Figure 4. Multiple correspondence analysis for Moravia’s

character: Carla.

another part of the novel we are told that she is 25. But, on the grounds of her
first description, readers attribute her an average age of 35. The character is very
ill and illness is often associated with old age. In a few cases this association
overshadows the text’s explicit information, so much so that 12 readers (11.5%)
think that Fosca is older than 50. On the whole, the concretizations of Fosca’s
age are very divergent, ranging from 18 to 55. Differences of this type produce
very different expectations, that have strong consequences when, in order to
reconstruct the world presented in the text, readers have to choose relevant
information in their world knowledge.
Also Angiolina’s age is not explicitly stated in Svevo’s text. The text, however,
contains several indications of health, strength and beauty, which are usually
associated with youth, e.g., “a tall, healthy blonde, with big blue eyes and a supple,
graceful body, an expressive face and transparent skin glowing with health.”
The readers think that she is between 17 and 37 (average: 24). The divergence
is smaller than the one found for Fosca, but it is still remarkable.
Moravia’s passage tells us very little about Carla. However her apparel
(“a brown woolen frock with a skirt so short”) and her bearing (“she moved
 THE CONSTRUCTION OF LITERARY CHARACTERS / 103

clumsily and uncertainly forward”) are normally associated with youth. Accord-
ing to our readers, her age is between 15 and 40 years (42.1% between 15 and
20; 31.6% between 30 and 40; average: 25). Also these ages lead to different
concretizations.
An ANOVA was carried out on the age attributed to each character with two
factors between subjects: the readers’ gender and type of undergraduate studies
(Humanities vs. Applied Sciences). Main effects and interaction between factors
were not found to be significant. Similarly, no significant interaction emerged
between the degree of interest and the age attributed to the characters. This
result, associated with the fact that a wide range of ages was obtained for every
character, suggests that the perception of this feature was based on idiosyncratic
attributions, rather than systematic inferences. The next analysis investigated
whether the same was true for the answers concerning the season of the year in
which the novel was set.

Season of the Year

The season of the year represents another piece of information that was not
given explicitly in the four passages. Table 6 shows the values for the Gini index.
and the percentages with which each season was chosen. The Gini index can be
used only with categorical variables, as in the case of the eight traits and the
seasons of the year and does not allow for the estimation of standard deviation
(Pace & Salvan, 1998). For Manzoni’s passage, readers indicate spring, pre-
sumably because the dress described seems to be appropriate for this season and
because spring might be considered an appropriate season for a marriage. In the
other cases, the Gini index is high, showing that different concretizations have
been performed by the different readers.
The concretization of the season may be related to the perception of the
character. It has important consequences on the type of material readers acti-
vate in their memory in order to reconstruct the situation presented in the text,
since different seasons entail many differences, like different clothing, hours of
daylight, and types of possible actions.

Table 6. Readers’ Inferences on the Season of the Year

Text Gini index Spring Summer Autumn Winter Missing

Manzoni .67 63% 21% 2% 1% 13%

Tarchetti .91 20% 2% 31% 13% 35%
Svevo .84 47% 25% 17% 0% 11%
Moravia .91 29% 3% 37% 19% 12%
104 / NEMESIO, LEVORATO AND RONCONI

DISCUSSION

Our study provided empirical evidence supporting literary theoreticians’ claims

(Culpeper, 1996; Jouve, 1992; Rimmon-Kenan, 1983, p. 36; van Peer, 1988, p. 9)
that literary characters are constructs, resulting from the interaction between
text and readers who make use of their individual competence and experience.
In our experiment readers tended to concretize traits that were not explicitly
stated in the text with very high values of the Gini index, thus showing different
concretizations in different readers. Although our readers were told that they
could choose not to answer a question if they thought that the text did not contain
any information leading to an answer, they tended to concretize the character
through a constructive process based on the production of inferences, even when
the character was presented with relatively few details, as in the case of Moravia’s
Carla. This tendency seemed to be the effect of the need to build up a coherent and
complete semantic representation of the character. The constructed representation
resulted both from information explicitly given and from inferences, more or less
legitimate, produced by the readers. This tendency was quite pervasive among
readers, regardless of whether they were male or female, students of Humanities
or of Applied Sciences, or whether they experienced high or low interest in the
text read. It seemed to be intrinsic to the act of reading itself.
In the answers concerning the elegance and the height of Tarchetti’s Fosca,
two very negative traits (illness and ugliness) weakened the recollection of
other traits mentioned explicitly, such as her being elegant and tall. This shows
that readers’ working memory was selective, removing information that was
considered irrelevant. When questioned about removed information, readers were
unable to recover it in full and thus were forced to draw inferences that led to
results that were different from the text’s surface.
The introduction of a character may be evaluative rather than descriptive.
For instance, in Tarchetti’s Fosca we find adjectives such as: “brutta” (ugly),
“eccessiva” (excessive), “inconcepibile” (inconceivable), “spaventosa” (fearful),
“bella” (pretty), and “giusta” (harmonious). The meanings of such adjectives
depend on the narrator’s criteria, in order to understand them, we need to know
his/her mentality. When this information is not available, the adjectives remain
undetermined and allow readers to concretize them in a way that fits their
expectations or their desires. Readers fill out the details of a character, and do so
according to their own interests and experience.
When information is given, a tendency seems to emerge that women, students
of Humanities, and readers more interested in the text produce more accurate
recollections than men, students of Applied Sciences, and less interested readers.
Women were more interested in the text they read than men and students of
Humanities were more interested than students of Applied Sciences. The tendency
of female readers to express a higher degree of interest for passages taken from
novels emerged also in a previous study (Nemesio et al., 2008). In that study,
 THE CONSTRUCTION OF LITERARY CHARACTERS / 105

passages describing both male and female characters were examined. We can,
therefore, exclude that women’s higher interest was due to the fact that the
current study concerned the description of female characters.
A multiple correspondence analysis showed that, when a piece of informa-
tion was not given in a text, readers tended to infer that it was not relevant
and therefore unmarked and for this reason they tended to choose the medium
answer. When it is not necessary, readers avoid producing a specific inference.
This inferential parsimony allows the allocation of cognitive resources to the
constructive processes that are necessary for the comprehension of the text (see
van den Broek, 1994). On the basis of the representation provided in Figures 1
to 4, it seems to be plausible that this dimension, which might be called the
“parsimony factor,” was scarcely affected by readers’ individual differences,
suggesting that it is a structural characteristic of the comprehension process.
Further studies should test the tentative explanatory hypothesis that the
“parsimony factor” is common to passages that describe characters, whereas a
second factor, that is consistent with the analyses carried out with the Gini
index, differentiates the passages and is related to readers’ characteristics such
as those taken into consideration in the current studies.

APPENDIX
Texts Used—Italian and English Versions

Manzoni, I promessi sposi (1840)

Lucia usciva in quel momento tutta attillata dalle mani della madre. Le amiche si
rubavano la sposa, e le facevan forza perché si lasciasse vedere; e lei s’andava
schermendo, con quella modestia un po’ guerriera delle contadine, facendosi
scudo alla faccia col gomito, chinandola sul busto, e aggrottando i lunghi e neri
sopraccigli, mentre però la bocca s’apriva al sorriso. I neri e giovanili capelli,
spartiti sopra la fronte, con una bianca e sottile dirizzatura, si ravvolgevan, dietro il
capo, in cerchi moltiplici di trecce, trapassate da lunghi spilli d’argento, che si
dividevano all’intorno, quasi a guisa de’ raggi d’un’aureola, come ancora usano le
contadine nel Milanese. Intorno al collo aveva un vezzo di granati alternati con
bottoni d’oro a filigrana: portava un bel busto di broccato a fiori, con le maniche
separate e allacciate da bei nastri: una corta gonnella di filaticcio di seta, a pieghe
fitte e minute, due calze vermiglie, due pianelle, di seta anch’esse, a ricami. Oltre
a questo, ch’era l’ornamento particolare del giorno delle nozze, Lucia aveva
quello quotidiano d’una modesta bellezza, rilevata allora e accresciuta dalle varie
affezioni che le si dipingevan sul viso: una gioia temperata da un turbamento
leggiero, quel placido accoramento che si mostra di quand’in quando sul volto
delle spose, e, senza scompor la bellezza, le dà un carattere particolare.
106 / NEMESIO, LEVORATO AND RONCONI

At that very moment Lucia’s mother had finished dressing her in all her
finery, and she came out to her friends. Each of them wanted Lucia all to herself;
they tried to force her to let them see her properly; and she was warding them
off with all the somewhat brusque modesty of a peasant girl, shielding her face
with one arm, or ducking it down against her bosom. The long dark line of her
eyebrows was gathered in a frown, but her lips opened in a smile at the same time.
Her dark young hair was divided in front by the narrow white line of her parting;
at the back of her head it was twisted up into a series of concentric rings, secured
by long silver pins arranged in a pattern like the rays of a halo—a fashion still
followed by peasant girls in the territory of Milan. She wore a necklace of alternate
garnets and filigree gold beads; a smart bodice of flowered brocade, and sleeves
laced with coloured ribbons; a shirt skirt of rough silk with many small, fine
pleats; scarlet stockings; and embroidered slippers, also of silk. But besides
the special ornaments that she had put on for her wedding morning, Lucia had
one which she wore every day—a modest beauty, which was thrown into relief
and enhanced by the various emotions which appeared in her face—a great
happiness, qualified by a faint air of confusion, and that calm melancholy which
appears from time to time on the face of a bride, not detracting from her beauty,
but giving it a special character [transl. by Bruce Penman].

Tarchetti, Fosca (1869)

Né tanto era brutta per difetti di natura, per disarmonia di fattezze,—ché anzi
erano in parte regolari,—quanto per una magrezza eccessiva, direi quasi incon-
cepibile a chi non la vide; per la rovina che il dolore fisico e le malattie avevano
prodotto nella sua persona ancora così giovine. Un lieve sforzo d’immaginazione
poteva lasciarne travedere lo scheletro, gli zigomi e le ossa delle tempie avevano
una sporgenza spaventosa, l’esiguità del suo collo formava un contrasto vivissimo
colla grossezza della sua testa, di cui un ricco volume di capelli neri, folti,
lunghissimi, quali non vidi mai in altra donna, aumentava ancora la sproporzione.
Tutta la sua vita era ne’ suoi occhi che erano nerissimi, grandi, velati—occhi d’una
beltà sorprendente. Non era possibile credere che ella avesse mai potuto essere
stata bella, ma era evidente che la sua bruttezza era per la massima parte effetto
della malattia, e che, giovinetta, aveva potuto forse esser piaciuta. La sua persona
era alta e giusta; v’era ancora qualche cosa di quella pieghevolezza, di quella
grazia, di quella flessibilità che hanno le donne di sentimento e di nascita distinta;
i suoi modi erano così naturalmente dolci, così spontaneamente cortesi che
parevano attinti dalla natura più che dall’educazione: vestiva con la massima
eleganza, e veduta un poco da lontano, poteva trarre ancora in inganno. Tutta
la sua orribilità era nel suo viso.
Nor did her ugliness stem so much from some natural defect, a disharmony
among her features (which in fact were somewhat symmetrical), as from an
excessive thinness which I would almost call inconceivable to anyone who had
 THE CONSTRUCTION OF LITERARY CHARACTERS / 107

not seen her—it was the ruin that physical pain and illness inflicted on her still
youthful person. A slight effort of imagination would permit a glimpse of her
skeleton. Her cheekbones and temples protruded fearfully, and her slender neck
formed the most striking contrast with the bulk of her head, whose rich mass
of hair, black, thick, longer than I had ever seen on a woman, further augmented
the disproportion. All her life was concentrated in her eyes, which were jet
black, large, veiled—eyes of a surprising beauty. One could not possibly believe
that she had ever been beautiful, but it was evident that her ugliness was for
the most part the effect of the illness, and that, when a girl, she was perhaps
pleasant. She was tall and stood erect. She also displayed some of the pliancy,
grace, flexibility that distinguishes women of sentiment and gentle birth; her
manners were so naturally sweet, so spontaneously courteous that they seemed
to be drawn more from nature than from education. She dressed with a great deal
of elegance, and seen from a distance, she might even deceive. All of her horror
was in her face [transl. by Lawrence Venuti].

Svevo, Senilità (1898)

Angiolina, una bionda dagli occhi azzurri grandi, alta e forte, ma snella e
flessuosa, il volto illuminato dalla vita, un color giallo di ambra soffuso di rosa da
una bella salute camminava accanto a lui, la testa china da un lato come piegata dal
peso del tanto oro che la fasciava, guardando il suolo ch’ella ad ogni passo toccava
con l’elegante ombrellino come se avesse voluto farne scaturire un commento
alle parole che udiva.
Angiolina walked beside him. She was a tall, healthy blonde, with big blue eyes
and a supple, graceful body, an expressive face and transparent skin glowing
with health. As she walked, she held her head slightly on one side, as if it were
weighed down by the mass of golden hair which was braided round it, and she
kept looking down at the ground which she tapped at each step with her elegant
parasol, as if she hoped there might issue from it some comment on the words
that had just been spoken [transl. by Beryl de Zoete].

Moravia, Gli indifferenti (1929)

Entrò Carla; aveva indossato un vestitino di lanetta marrone con la gonna così
corta, che bastò quel movimento di chiudere l’uscio per fargliela salire di un buon
palmo sopra le pieghe lente che le facevano le calze intorno alle gambe; ma ella
non se ne accorse e si avanzò con precauzione guardando misteriosamente davanti
a sé, dinoccolata e malsicura.
Carla came into the room. She was wearing a brown woolen frock with a skirt
so short that the movement she made in shutting the door was enough to pull it
up several inches over the slack wrinkles her stockings formed about her legs;
but she was unaware of this, and advanced cautiously, peering in front of her as
she moved clumsily and uncertainly forward [transl. by Angus Davidson].
108 / NEMESIO, LEVORATO AND RONCONI

REFERENCES
Benzécri, J. P. (1976). L’analyse des données. Paris: Dunod.
Bortolussi, M., & Dixon, P. (2003). Characters and characterization. In Psychonarratology
(pp. 133-165). Cambridge: Cambridge University Press.
Culpeper, J. (1996). Inferring character from exts: Attribution Theory and Foregrounding
Theory. Poetics, 23(5), 335-361.
De Vega, M., Leon, I., & Diaz, J. M. (1996). The representation of changing emotions
in reading comprehension. Cognition and Emotion, 10, 303-321.
Emmott, C. (1997). Narrative comprehension. A discourse perspective. Oxford: Oxford
University Press.
Gernsbacher, M. A., Goldsmith, H. H., & Robertson, R. R. W. (1992). Do readers mentally
represent characters’ emotional states? Cognition and Emotion, 6(2), 89-111.
Gerrig, R. J., & Allbritton, D. W. (1990). The construction of literary character: A view
from cognitive psychology. Style, 24(3), 380-391.
Greenacre, M. J. (1984). Theory and applications of correspondence analysis. London:
Academic Press.
Jouve, V. (1992). L’effet-personnage dans le roman [The character effect in the novel].
Paris: Presses Universitaires de France.
Kintsch, W. (1998). Comprehension–A paradigm for cognition. Cambridge: Cambridge
University Press.
Manzoni, A. (1840). I promessi sposi. Firenze: La Nuova Italia, 1983 (The betrothed,
trans. by Bruce Penman. London: Penguin, 1972).
Moravia, A. (1929). Gli indifferenti. Milano: Bompiani, 1974 (The time of indifference,
trans. by Angus Davidson. London: Secker & Warburg, 1953).
Nemesio, A. (2002a). The concretization of character in narrative texts. In IAEA 2002.
Proceedings of the 17th Congress of the International Association of Empirical
Aesthetics (pp. 537-540). Osaka: Japan.
Nemesio, A. (2002b). La costruzione del testo. Ricerche empiriche su testi italiani
dell’Otto-Novecento [Text construction. Empirical research on eighteenth and
nineteenth Italian texts]. Turin: Thélème.
Nemesio, A. (2008). I diversi volti del personaggio nel testo narrativo. Lo studio
empirico delle concretizzazioni [The different looks of character in narrative texts.
An empirical study of concretizations]. In C. Lombardi (Ed.), Il personaggio: Figure
della dissolvenza e della permanenza [Character in literature: Patterns of evanescence
and permanence] (pp. 397-403). Alessandria: Edizioni dell’Orso.
Nemesio, A., Levorato, M. C., & Ronconi, L. (2008). The time of reading: The first
perception of characters in narrative texts. In Online Proceedings of the Annual
Conference of the Poetics and Linguistics Association (pp. 1-12). Sheffield: University
of Sheffield. Available from http://www.pala.ac.uk/resources/proceedings/2008/
Pace, L., & Salvan, A. (1998). Introduzione alla statistica—Statistica descrittiva [An
introduction to statistics—Descriptive statistics]. Padova: Cedam.
Rimmon-Kenan, S. (1983). Narrative fiction: Contemporary poetics. London: Routledge.
Stanzel, F. K. (1981). Teller-characters and reflector-characters in narrative theory.
Poetics Today, 2(2), 5-15.
Svevo, I. (1898). Senilità. In Opere (pp. 425-593). Milano: Dall’Oglio, 1962 (As a man
grows older, trans. by Beryl de Zoete. London/New York: Putman, 1932).
 THE CONSTRUCTION OF LITERARY CHARACTERS / 109

Tarchetti, I. U. (1869). Fosca. In E. Ghidetti (Ed.), Tutte le opere (Vol. 2, pp. 235-427).
Bologna: Cappelli, 1967 (Passion, trans. by Lawrence Venuti. San Francisco, CA:
Mercury House, 1994).
van den Broek, P. (1994). Comprehension and memory of narrative texts: Inferences and
coherence. In M. A. Gernsbacker (Ed.), Handbook of psycholinguistics (pp. 539-588).
San Diego, CA: Academic Press.
van Peer, W. (Ed.). (1988). The taming of the text: Explorations in language, literature
and culture. London: Routledge.
Zwaan, R. A. (1996). Toward a model of literary comprehension. In B. K. Britton &
A. C. Graesser (Eds.), Models of understanding text (pp. 241-255). Hillsdale, NJ:
Erlbaum.

Direct reprint requests to:

Aldo Nemesio
Via Tetti Borra 13
10023 Chieri (TO), Italy
e-mail: aldo.nemesio@unito.it