The masculine form and its competing interpretations in French: When linking grammatically masculine role names to female referents is difficult

Abstract

Using a word association paradigm we examined the extent to which readers can overcome the specific interpretation of the grammatical masculine form in French when instructed to embrace its generic meaning. In two experiments participants were to decide whether a person introduced by a kinship term (e.g., aunt) could be part of a group represented by a role name (e.g., musicians). After the completion of the first half of the experiment, participants were explicitly reminded about the generic interpretation and use of the masculine form. Although the reminder resulted in some level of generic interpretation, there were still strong traces of the specific interpretation in the response times, regardless of participants' inhibition capacities (Experiment 1). Adding a supplementary constraint by exposing readers to distractor role names in feminine forms (Experiment 2) did not reveal any different effects. The results indicated that although readers can be motivated to elaboratively activate the generic interpretation of the masculine form, the latter can not completely overrule a more passively activated specific one.

Full text

The masculine form in French..... 1
Running head: THE MASCULINE FORM IN FRENCH
The masculine form and its competing interpretations in French: When linking grammatically
masculine role names to female referents is difficult.
Pascal Gygax
University of Fribourg, Switzerland
Ute Gabriel
Norwegian University of Science and Technology, Trondheim, Norway
Arik Lévy, Eva Pool, Marjorie Grivel, Elena Pedrazzini
University of Fribourg, Switzerland
Address for correspondence:
Pascal Gygax
Département de Psychologie
Université de Fribourg
Rue Faucigny 2
1700 Fribourg, Switzerland
Pascal.Gygax@unifr.ch

The masculine form in French..... 2
Abstract
Using a word association paradigm we examined the extent to which readers can overcome
the specific interpretation of the grammatical masculine form in French when instructed to
embrace its generic meaning. In two experiments participants were to decide whether a person
introduced by a kinship term (e.g. aunt) could be part of a group represented by a role name
(e.g. musicians). After the completion of the first half of the experiment, participants were
explicitly reminded about the generic interpretation and use of the masculine form. Although
the reminder resulted in some level of generic interpretation, there were still strong traces of
the specific interpretation in the response times, regardless of participants’ inhibition
capacities (Experiment 1). Adding a supplementary constraint by exposing readers to
distractor role names in feminine forms (Experiment 2) did not reveal any different effects.
The results indicated that although readers can be motivated to elaboratively activate the
generic interpretation of the masculine form, the latter can not completely overrule a more
passively activated specific one.
<164 words>
Key words: text comprehension, gender representation, masculine bias, inhibition processes

The masculine form in French..... 3
The masculine form and its competing interpretations in French: When linking grammatically
masculine role names to female referents is difficult.
In gender marked languages such as French or German, the grammatical gender of a
noun referring to a person typically matches the sex of the person referred to, such as une
artiste [a female artist] and un libraire [a male librarian]. In addition, the grammatical form is
often explicitly signalled not only by the form of the determiner or the article but also by the
morphological form of the noun. This is most often true for role names, defined as names that
incorporate features used to describe a person or a group of people such as, for example,
musicians or swimmers. In French, you would use une musicienne (feminine form), to refer to
a female musician but un musicien (masculine form) for a male musician, and you would use
une nageuse (feminine form) to refer to a female swimmer, but un nageur (masculine form)
for a male swimmer.
Although as a general rule, in French, the grammatical gender of a human referent
noun as well as its morphological form indicate the sex of the person referred to, this rule is
somehow misleading, as there are instances when it does not apply: When referring to persons
of unknown sex, to persons where the sex of the person is irrelevant or to a group of people of
both sexes, the masculine form is also used and is supposed to be dissociated from its
biological or specific meaning (i.e., referring to men), and interpreted in a generic way (i.e.,
equally referring to men and women). The rule stating that grammatical gender indicates sex
is actually only truly reliable for feminine forms (i.e., feminine form  female referent). For
masculine forms, readers have to rely on additional cues to decide whether the grammatical
information of a human referent noun is relevant for the sex (i.e., masculine form  male
referent) or not (i.e., masculine used as a generic). As a consequence, masculine forms often
generate semantic ambiguity (Irmen & Kurovskaja, 2010). This ambiguity is enhanced by the
possible interpretations of the very notion of generic (Gygax & Gabriel, 2010). Formally, a
group of people of both sexes can mean that there is a majority of men and one or two women,

The masculine form in French..... 4
that there is an equal share of both, but also that there is a majority of women and only one or
two men.
Previous research suggests that the ambiguity of whether a masculine is used in a
specific or generic way is typically resolved to the disadvantage of women (for a review:
Stahlberg, Braun, Irmen, & Sczesny, 2007). However, empirical investigations have not yet
addressed whether masculine forms can at all be understood generically, i.e. whether readers
can be brought to link masculine forms equally to female and male referents.
The present study is intended to fill this gap by exploring how readers embrace the
generic interpretation of the masculine form using a particular word association paradigm.
Gender being a genuine part of readers' mental representations of the text (e.g., Garnham,
Oakhill, & Reynolds, 2002), we were interested in whether readers can easily mentally
represent women and men equally when encountering role names written in the masculine
form when explicitly asked to do so. Although based on a word association paradigm, the
present study tackles processes at stake when readers encounter masculine forms in particular
messages that theoretically equally target persons of both sexes (e.g., a job advertisement).
In terms of a mental representation of a text, readers already have multiple meanings
that may or may not be activated when encountering a role name in the masculine form.
Memory-based approaches would suggest that through a passive bottom-up process (e.g.,
resonance mechanism), textual elements activate all associated meanings (Gerrig & McKoon,
1998) in readers' long-term memory. Only through a subsequent evaluation phase is the final
information chosen and integrated in readers’ mental model of the situation (Cook &
Guéraud, 2005). Explanation-based approaches, on the other hand, would suggest a more
active process, driven by readers' need to maintain a coherent representation of the text
(Graesser, Singer & Tabasso, 1994), whereby readers search the appropriate meaning of the
masculine form in long-term memory. Both approaches assume that at one point in the
integration process, an evaluation has to take place to decide which meaning is referred to in

The masculine form in French..... 5
the text.
A reasonable assumption to make here is that when confronted with the ambiguity
accompanying the masculine form, readers rely on very easily and rapidly activated
information to form a somehow good-enough representation (as in Ferreira, Bailey, &
Ferraro, 2002). As will be discussed later, research on the mental representation of gender
when reading role names has so far mainly identified two types of information as being easily
activated: (a) stereotypical knowledge and (b) the specific meaning of the masculine form (in
gender-marked languages). However, to our knowledge, much less attention has been devoted
to instances where readers are explicitly prompted to rely on other, correct meanings, such as
the generic meaning of masculine form.
On an applied level, clarifying the preconditions and limits for a successful use of the
masculine as a generic becomes important for languages such as French, in which the use of
less ambiguous alternatives, such as balanced forms (i.e., les musiciens et musiciennes), is
not widely implemented (e.g., Sarrasin, Gabriel & Gygax, in press). On a more theoretical
level, identifying the conditions that may favour the activation of the generic meaning over
the specific one may provide us with important insights into the way the masculine form is
represented in memory.
Before describing our research in detail, we briefly review the literature on reader’s
mental representation of gender in role names and the different sources of information they
might rely on.
Constructing a representation of gender: Activations of different sources of information
Past research has mainly focused on the influence of two non-exclusive sources of
information on the mental representation of gender, namely grammatical and stereotypical
information. Most empirical research on the use of the masculine intended as generic in
French (e.g., Colé & Segui, 1994; Gygax & Gabriel, 2008), in German (e.g., Gygax, Gabriel,
Sarrasin, Oakhill & Garnham, 2008; Irmen, 2007; Stahlberg et al., 2007), in Spanish (e.g.,

The masculine form in French..... 6
Flaherty, 2001) and to some extent in Norwegian (e.g., Gabriel, 2008; Gabriel & Gygax,
2008) suggests that when reading a noun referring to persons – such as role names – the use
of the masculine induces male dominant representations of gender, hinting at a very strong
influence of grammatical cues on the mental representation of human referents’ gender. A
reliable influence of stereotypical information when building a representation of gender has
been shown for readers of unmarked languages such as English (e.g., Banaji & Hardin, 1996;
Carreiras, Garnham, Oakhill, & Cain, 1996; Garnham et al., 2002).
Although these studies have documented a certain imbalance in the activation of male
and female representations when reading role names in the masculine form, most of them, if
not all, do not suggest an all-or-none process (i.e., either masculine or feminine). Even when
written in the masculine form, role names have been found to activate some level of female
representation. One wonders, therefore, whether this gender representation imbalance may
fluctuate depending on contextual factors. To our knowledge, only two studies have explicitly
addressed this issue, one investigating textual elements that could affect readers'
representations of gender (Gabriel, Gygax, Sarrasin, Garnham & Oakhill, 2011) and the other
exploring circumstances where a male dominant representation might even be intensified
(Gygax & Gabriel, 2008). Both studies, described further, as well as previously mentioned
studies, are the basis for the specific issue that is tested in this paper, namely the possibility of
establishing or generating a balanced mental representation of gender.
Exploring both mitigation (in German) and reinforcement mechanisms (in French and
to some extent in English), Gabriel et al. (2011), in a partial replication of Gygax et al.’s
experiment (2008), investigated the effect of adding gender marked pronouns on the mental
representation of role names written in the masculine form. In their experiment, they
presented participants with the same pairs of sentences as in Gygax et al. (2008), one sentence
of each pair comprising a role name in the masculine (plural) form (e.g., Les voisins sortirent
de la cafeteria. [The neighbours came out of the cafeteria.]) and one sentence containing

The masculine form in French..... 7
explicit information about the characters’ gender (e.g., A cause du temps nuageux un/e des
femmes/hommes avait un parapluie. [Because of the cloudy weather one of the women/men
had an umbrella]), but also added a pronoun in a small statement associated to the first
sentence (e.g., Les voisins sortirent de la cafeteria. Ils partirent. [The neighbours came out of
the cafeteria. They went away]). Participants had to decide whether the second sentence was a
sensible continuation of the first one (and its associated statement). In English, the congruent
pronoun is gender neutral (they), in French masculine (ils - masculine form used as a generic)
and in German generic but morphologically identical to the singular feminine form (sie).
Adding pronouns did alter participants' mental representations of gender, but only in German.
The male dominance found in previous experiments significantly decreased - yet was still
apparent - both in the female and neutral stereotypicality conditions as a result of using the
pronoun sie.
Based on a paradigm used initially by Banaji and Hardin (1996) and subsequently by
Garnham et al. (2002), Gygax and Gabriel (2008) presented participants with a word
association paradigm in which pairs of words were presented, each composed of a kinship
term in the singular form (e.g., a sister or a brother) and a stereotypically male, female or
neutral role name in the masculine plural form (i.e., musicians). Participants had to decide, as
fast as possible, if the person represented by the kinship term could be part of the group
represented by the role name. In the first part of the experiment, all role names were in the
masculine plural form, whereas in the second part, some role names were also in the feminine
plural form. In line with previous research, in the first part of the experiment, participants
were more likely to respond yes to a pair when it contained a male kinship term, regardless of
the stereotypicality of the role name. In the second part, participants were even more likely to
do so for the same pair (i.e., role name in the masculine form-male kinship), demonstrating an
even more pronounced dominance of the specific interpretation. The authors concluded that
the mere presence of role names written in the feminine form increased an already dominant

The masculine form in French..... 8
activation of the specific interpretation when confronted with the masculine form.
Taken together, these studies document firstly that although there is some malleability
<plasticity? Formeability?> in the interpretation of masculine forms, grammatical
information has a strong influence on readers’ mental representation of gender. Secondly,
readers are constantly confronted with different sources of information (grammatical,
morphological, semantical), at times contradictory, when constructing a representation of
gender. Finally, in gender marked languages, a true activation of the alternative meanings of
the specific interpretation of the masculine form has never been shown, at least never to the
extent of occupying a primary role in readers’ mental representation.
Overcoming the influence of grammatical information
With reference to masculine forms, research suggests that the specific meaning of the
masculine form (i.e., masculine form  male human referent) prevails over the generic one
(i.e. masculine form  refers equally to both, male and female human referent) indicating
that readers strongly rely on grammatical cues and in general might need additional
information to interpret masculine forms as truly generic.
We therefore wanted to explore the effects of reminding readers of the generic
interpretation and consequently motivate them to consider masculine forms as representative
of both women and men. We employed the word association task from Gygax and Gabriel
(2008) and gave participants specific instructions in the middle of the experiment on how to
resolve the ambiguity initiated by the use of the masculine form. We reminded them of the
rule that masculine forms can be used to equally refer to male and female referents and asked
them to keep this in mind when responding.
We were interested in three central ideas. First, we wanted to examine the ease with
which readers would embrace the generic interpretation. Based on the presumption that
linking masculine forms to female referents is more difficult than linking masculine forms to
male referents, we hypothesized that a general increase in readers’ proportion of positive

The masculine form in French..... 9
answers for masculine form/female referent combinations after having received the
instructions might nevertheless be accompanied with differentiated response times, positive
response times still being slower to masculine form/female referent combinations than to
masculine form/male referent combinations, regardless of the instructions.
Second, we examined whether a possible mitigation of the masculine form influence,
even if only partial, would reveal some influence of stereotypical information. When testing
bi-gendered role names in Italian (i.e., non-gendered role names in a gender-marked
language), Cacciari and Padovani (2007, Experiment 2) found that when a pronoun (e.g., lui
[he] or lei [she]) was primed by a matching stereotyped role name (e.g., engineer-he or
housekeeper-she), participants were faster to decide whether the pronoun was masculine or
feminine than when primed by a counter-stereotype role name. Similar results have been
found in Spanish (e.g., Carreiras et al., 1996, fourth experiment) and in German (Irmen,
2007). If instructing the participants does reveal some stereotype effect, it would provide us
with evidence that both sources of information are generally activated when readers encounter
role names written in the masculine forms, and that removing the dominant source may give
way to the less dominant one.
Third and finally, we were interested in the possibility that the processes at hand may
well depend on readers’ inhibition capacities as measured by a Flanker task (Fan, Flombaum,
McCandliss, Raz & Posner, 2002; based on Eriksen & Eriksen, 1974). Individual differences
might appear in the way participants process the specific interpretation of the masculine form
and integrate our instructions to overcome it. In fact, we believe that readers may require
inhibition processes with reference to the grammatical information to consider a generic
meaning as a possible alternative, as “overlooking grammatical information” could be one
way to meet the instructions. Following this line of reasoning, we expected individuals with
higher inhibition capacities to embrace the generic meaning with more ease than those with
lower inhibition capacities.

The masculine form in French..... 10
Experiment 1
Method
Participants. Fourty-nine native French-speaking first and second year psychology
students (42 women and 7 men) from the University of Fribourg (Switzerland) participated in
this experiment to gain course credits. As most of the studies on the generic interpretation of
the masculine form – presented in the introduction – found no gender differences, we were
not interested in this variable and hence did not control for gender balance.
Materials and procedure. As in Gygax and Gabriel (2008), participants were
presented with pairs of terms, each pair composed of a role name in the plural form and a
kinship term in the singular form. The participants’ task was to decide as quickly as possible
whether the person represented by the kinship term (e.g., une soeur [a sister]) could be part of
the group represented by the role name (e.g., musiciens [musicians]). If participants thought,
for example, that a sister could be part of a group of musicians, they pressed the yes button,
but if they thought that a sister could not be part of a group of musicians, they pressed the no
button. Thirty-six role names from Gabriel, Gygax, Sarrasin, Oakhill and Garnham (2008)
were used in this experiment. They were composed of 12 stereotypically female role names,
12 stereotypically male role names and 12 stereotypically neutral (i.e., with no stereotype)
role names. The role names are presented in Table 1. As in Gygax and Gabriel (2008), six
kinship terms were used to create the pairs. These were sister, aunt, mother, brother, uncle,
father. The experiment was divided into two parts, in the first part of the experiment (i.e., Part
I), 18 out of the 36 role names (6 female, 6 male and 6 neutral role names) were used. Each
role name was associated with all kinship terms. In Part I, each role name appeared 6 times
resulting in 108 experimental items. Fifty-four filler items, each composed of an unambiguous
gender role name (e.g., godfathers) and an incongruent kinship term (e.g., a mother) were
added. These filler items were added to ensure that participants would not consistently press

The masculine form in French..... 11
the yes button without properly reading the pairs. For each participant, the pairs were
presented in a random order.
In Part II, the remaining 18 role names were used, and, as in Part I, each role name
was associated with all kinship terms and appeared 6 times resulting in 108 experimental
items. Fifty-four filler items were also added. In total, each participant was presented with
324 items, half in Part I and half in Part II. Two fixed lists were created to ensure that each
role name appeared both in Part I and in Part II across the experiment (i.e., a role name in Part
I in List 1 would appear in Part II in List 2).
When Part I was completed, participants were presented with additional instructions.
They were reminded of the grammatical rule that when referring to persons of unknown sex,
to persons where the sex is irrelevant or to a group of people of both sexes, the masculine
form is used and is supposed to be interpreted in a generic way and were explicitly asked to
keep this in mind when responding to Part II.
After a short break, when both Part I and Part II were completed, participants were
presented with an inhibition Flanker Task (FT) from Fan et al. (2003). In this task, 30 sets of
five black arrows pointing to the left or the right are presented on the screen (i.e., white
background). The arrows are .58 cm large and placed .06 cm from each other, resulting in an
approximate total of 3.27 degree angle of vision, taken that each participant is 30 cm away
from the screen. In each set of arrows, the central one is the target. Participants have to focus
on it and determine as fast as possible its direction (i.e., left or right), while ignoring the other
ones. The other arrows are considered as distractors and can either point in the same direction
(i.e., congruent condition) or in the opposite direction (i.e., incongruent condition) of the
target arrow. A third condition (i.e., neutral condition), in which the distracting arrows are
substituted by simple strokes, is also presented to the participants. In all, each condition is
represented by 10 sets. Before each set appears, a 1-second fixation point appears in the

The masculine form in French..... 12
middle of the screen at the very same location as the target arrow. The 30 items were
presented in a random order.
Apparatus. Participants completed both tasks on a Power Macintosh 4400 using the
PsyScope Software (Cohen, MacWhinney, Flatt, & Provost, 1993). Participants were tested
individually in a quiet laboratory room.
Results and discussion
Before proceeding to the main analyses, we removed incorrect response times of the
Flanker task (2.6% of the data) and replaced correct response times that were 2.5 standard
deviations above or below each participant's mean (3.8% of the correct responses) by their
cut-off values.
For both the proportion of positive responses and the response times analyses, we then
separated participants into high and low inhibition participants by computing response times
to the FT as followed: As in Fan et al. (2003), we subtracted the time it took participants to
respond to neutral items to incongruent ones and performed a median split to divide the
participants into high inhibition participants (M = 21 ms; SD = 55) and low inhibition
participants (M = 133 ms; SD = 141; t(47)=3.62; p<.001)1.
Proportion of positive responses. The mean proportions of positive judgements were
analysed both by-participants (F1) and by-items (F2). In the former (F1), mixed design
ANOVAs were conducted by considering Kinship (Men vs. Women), Part (Part I vs. Part II)
and Stereotype (Female vs. Male vs. Neutral) as within-participant variables and Flanker
(Low vs. High) as a between-participant variable. In the latter (F2), mixed design ANOVAs
were conducted by considering Flanker (Low vs. High), Part (Part I vs. Part II) and Kinship
(Men vs. Women) as within-item variables (i.e., the same role names were presented in each
part, and each role name was followed by male and female continuations) and Stereotype
(Male vs. Female vs. Neutral) as a between-item variable. We only report the main and
interaction effects that were significant (p < .05) for both F1 and F2.

The masculine form in French..... 13
A 2 (Part) by 2 (Flanker) by 3 (Stereotype) by 2 (Kinship) mixed-design ANOVA on
the proportions of positive judgements revealed a main effect of Part (F1 (1, 47) = 53.11, p <
.001; F2 (1, 33) = 274.90, p < .001), the proportion of positive judgements being higher in the
second part (.97) than in the first part (.76) of the experiment, and a main effect of Kinship
(F1 (1, 47) = 41.58, p < .001; F2 (1, 33) = 108.69, p < .001), participants responding more
often yes to pairs including a male kinship (.97) than to pairs including a female kinship (.76).
Both main effects were qualified by a Part by Kinship interaction effect (F1 (1, 47) = 37.72, p
< .001; F2 (1, 33) = 108.12, p < .001), showing an increase in the proportion of positive
answers to pairs including a female kinship, from .58 in Part I to .95 in Part II, but only a
slight increase to pairs including a male kinship, from .95 to .98. There was no other main or
interaction effect including Stereotype or Flanker.
The instructions given to our participants, regardless of their level of inhibition
(capacities), as measured by the Flanker task, and regardless of stereotypicality, resulted in
the participants increasing their positive responses to female kinships-role names pairs,
indicating that they were able to encompass a more generic interpretation of the masculine
form. Note that in Part I, the proportion of positive responses to female kinships-role names
was of .58, indicating that the process by which participants responded, when given no
specific instructions, was not exclusively linked to the specific interpretation of the masculine
form, though still dominant, as suggested by Gygax and Gabriel (2008).
Positive response times. Although in Figure 1 we present raw response times, we
performed two transformations in Parts I and II before analysing the data, mainly to account
for two critical issues First, as role names differed in length in terms of number of letters
composing each role name and as each role name appeared either in Part I or in Part II for
each participant, we transformed the raw response times following Trueswell, Tanenhaus and
Garnsey's (1994) regression method. For each participant, we produced a regression equation
of time (i.e. reading time) against length (i.e. number of characters in the target role name)

The masculine form in French..... 14
using all positive response times. For each participant, a time by role name length regression
was calculated by computing the slope and the intercept of the regression. Residual response
times for each participant were then calculated by subtracting the actual raw response times
from the response times predicted by the regression equation.
We then transformed the data into z-scores per participant and per Part to account for
individual variability as well as the fact that naturally, participants responded faster in the
second part of the experiment, regardless of the materials presented to them (i.e., participants
get habituated to the task and hence gradually respond more rapidly), as shown in Figure 1.
Statistical analyses were then conducted on the z-scores. Again, only response times for
positive judgements were analysed. The low proportion of positive responses in some
conditions (i.e., few positive responses, especially when the kinship term was female) lead to
an imbalanced data set. To accommodate this problem, the data were analysed by fitting
linear mixed-effects models (using the R software by the R Development Core Team, 2010)2,
including both participants and items as random factors (Brysbaert, 2007). This kind of
procedure avoids cumbersome manipulations due to missing data and provides a particularly
adequate alternative to traditional repeated measures ANOVA as it includes both participants
(F1) and items (F2) in the same model which in turn represents a potential solution to the
“language-as-fixed-effect-fallacy” (see Brysbaert, 2007 for an initial presentation of the
controversies related to separate F1 and F2 analyses). In our analyses, experimental factors
(Flanker, Stereotype, Kinship and Part) were treated as fixed effects; participants and role
names were treated as random effects.
Our main interest in this experiment lay in a potential Kinship effect, expressing the
dominance of the specific interpretation of the masculine form, as well as a Part by Kinship
effect, evaluating the impact of our instructions on the above Kinship effect. Two other
models were subsequently tested. One that added Stereotype to inspect the differential effects
due to the role names stereotypicality and one that added Flanker to evaluate whether the

The masculine form in French..... 15
effects that might be found could be explained in terms of inhibition capacities.
Hence, we initially tested a model encompassing Kinship and Part as fixed factors and
participants and role names as random effects. Our first and simplest model included Kinship
as a main effect and Kinship by Part as an interacting effect. We then added Stereotype to the
initial model as a fixed factor, to test a Kinship by Stereotype as well as a Kinship by Part by
Stereotype interaction effect. We assessed the model fitness improvement by means of χ2–
difference based on the models log-likelihoods. The model encompassing Stereotype
improved the model significantly (Δχ2 = 29.78, Δdf = 8, p<.001). Conversely, adding Flanker
to the model by testing a Kinship by Flanker as well as a Kinship by Part by Flanker
(inhibition (capacity) hypothesis) interaction effects did not improve the model at all (Δχ2 =
.33, Δdf = 4, ns).
The model that fit the data best (i.e., with Stereotype) revealed a main effect of
Kinship (F (1, 9128) = 110.78, p <.001), positive responses (see Figure 1) to pairs including
male kinships being faster (.13 standard deviations) than to pairs including female kinships (-
.10 standard deviations), and more importantly, no interaction effect of Part by Kinship (F (2,
9128) = 2.53, ns). The model also revealed a Kinship by Stereotype interaction effect (F (4,
9128) = 3.96, p<.01), responses to pairs including female kinships being only .15 standard
deviations (z-score) slower than pairs including male kinships in the female stereotype
condition compared to .27 slower in the male stereotype condition and .26 in the neutral
stereotype condition. This difference has to be considered in light of a three-way Kinship by
Stereotype by Part (F (4, 9128) = 3.59, p<.01) interaction effect, suggesting that this effect
was greater in Part I (.20 std-difference in the female stereotype condition; .43 in the male
condition and .43 in the neutral condition) than in Part II (.12 std-difference in the female
stereotype condition; .22 in the male condition and .20 in the neutral condition). This latter
interaction suggested that with no particular instructions, the masculine bias found in previous
research was less pronounced in the female stereotype condition than in the others, as

The masculine form in French..... 16
indicated by Figure 1. As one of our primary interest resided in a possible increase of the
influence of stereotypical information as a result of our manipulation, the fact that
stereotypical information seemed pertinent in Part I, and not in Part II, was somehow
unexpected. We believe that this result corroborates the idea that when given no particular
instructions as to how to interpret the masculine form, multiple sources of information are
activated, increasing variance in the information readers use to build a representation of
gender. We will come back to this idea in the General discussion section.
Overall, the data substantiate our first hypothesis, assuming a possible increase in
readers’ proportion of positive answers for masculine form/female referent combinations after
having received the instructions accompanied with differentiated response times, positive
response times still being slower to masculine form/female referent combinations than to
masculine form/male referent combinations, regardless of the given instructions.
If none of the variance could be accounted for significantly by inhibition capacities, as
measured by the Flanker task – contrary to our expectations –, ruling out the influence of
inhibition capacities when dealing with the multiple sources of information associated with
role names might be premature. Indeed, in more applied settings (e.g., job advertisements or
newspaper articles), the masculine form is most often imbedded in texts also using feminine
forms, the latter adding critical constraint as to interpreting the masculine as a generic form
(Gygax & Gabriel, 2008). Such a supplementary constraint might add cognitive workload
(i.e., inhibition capacities might explain some of the variance in the results), and might also
generally impact upon the increase in the proportion of positive responses for masculine
form/female referent combinations, apparent in Experiment 1.
In the following experiment, we tested whether adding role names written in the
feminine form in the second part of the experiment would alter the effects found in
Experiment 1, both in terms of the proportion of positive responses and inhibition capacities.
Experiment 2

The masculine form in French..... 17
Method
Participants. Fourty-six native French-speaking first and second year psychology
students (40 women and 6 men) from the University of Fribourg (Switzerland) participated in
this experiment to gain course credits. None of the participants took part in Experiment 1. As
in Experiment 1 (see comment in the Participants section of Experiment 1), we did not
control for gender balance.
Materials and procedure. The materials and procedure were the same as in
Experiment 1, except for Part II, which, in addition to being preceded by instructions on the
generic interpretation of the masculine form, also included role names written in the feminine
form. In Part I, as in Experiment 1, each of 18 role names appeared 6 times resulting in 108
experimental items. Fifty-four filler items, each composed of an unambiguous gender role
name (e.g., godfathers) and an incongruent kinship term (e.g., a mother) were added. These
filler sentences were added to ensure that participants would not consistently press the yes
button without properly reading the pairs. For each participant, the pairs were presented in a
random order.
In Part II, the remaining 18 role names were used, but this time, each role name
appeared 6 times (i.e., associated with the six kinship terms) in the masculine plural form and
6 times in the feminine plural form. Thus in Part II, each role name appeared in total 12 times
(i.e., instead of only 6 times in Part I). In Part II, there were 216 experimental items. Only
twenty-seven filler items were added, as experimental items written in the feminine form
associated with a male kinship already constituted occasions for participants to respond
negatively.
As in Experiment 1, when Part I was completed, participants were presented with
additional instructions. They were reminded of the grammatical rule stipulating that when
referring to persons of unknown sex, to persons where the sex of the person is irrelevant or to
a group of people of both sexes, the masculine form is used and is supposed to be interpreted

The masculine form in French..... 18
in a generic way, and they were explicitly asked to keep this in mind when responding to Part
II. When both Part I and Part II were completed and after a short break, participants were
presented with the same inhibition Flanker Task (FT) presented in Experiment 1.
Apparatus. Participants completed both tasks on a Power Macintosh 4400 using the
PsyScope Software (Cohen, MacWhinney, Flatt, & Provost, 1993). Participants were tested
individually in a quiet laboratory room.
Results and discussion
All analyses were conducted following the same principles as in Experiment 1. As in
Experiment 1, before proceeding to the main analyses, we removed incorrect responses to the
Flanker task (3% of the data) and replaced correct response times that were 2.5 standard
deviations above or below each participant's mean by their cut-off values (3.18% of the
correct responses). To examine the FT, we then subtracted the time it took participants to
respond to neutral items to incongruent ones and performed a median split to divide the
participants into high inhibition participants (M=37 ms; STD=69) and low inhibition
participants (M=126; STD=49; t(44)=5.07; p<.001).
Proportion of positive responses. A 2 (Part) by 2 (Flanker) by 3 (Stereotype) by 2
(Kinship) mixed-design ANOVA on the proportion of positive judgements to pairs with role
names in the masculine form revealed a main effect of Part (F1 (1, 44) = 12.31, p < .001; F2
(1, 33) = 37.35, p < .001), the proportion of positive judgements being higher in the second
part (.94) than in the first part (.85) of the experiment, a main effect of Kinship (F1 (1, 44) =
16.34, p < .001; F2 (1, 33) = 46.89, p < .001), participants responding more often yes to pairs
including a male kinship (.95) than to pairs including a female kinship (.83), and a main effect
of Stereotype (F1 (2, 88) = 15.27, p < .001; F2 (2, 33) = 3.49, p < .05), participants giving
more positive responses to neutral stereotyped role names (.92) than to female stereotyped
role names (.89) and male stereotyped role names (.87).

The masculine form in French..... 19
Those main effects were qualified by a Part by Kinship interaction effect (F1 (1, 44) =
15.18, p < .001; F2 (1, 33) = 37.74, p < .001), showing again a significant increase in the
proportion of positive answers to pairs including a female kinship from .75 in Part I to .92 in
Part II, but no increase to pairs including a male kinship (.95 in both parts). As shown in
Figure 2, the increase was most apparent in the male stereotyped condition (from .68 to .93).
This larger increase was substantiated by a significant Part by Kinship by Stereotype
interaction effect (F1 (2, 88) = 21.26, p < .001; F2 (2, 33) = 5.56, p < .01). As in Experiment 1
and even though the experimental materials were more constraining, participants managed to
use a generic interpretation of the masculine form when reminded of the grammatical rule.
Positive response times. Before analysing response times of positive judgements to
pairs with role names in the masculine form, we performed the same transformations as in
Experiment 1. In Figure 3, we present raw response times in Parts I and II of the experiment.
As in Experiment 1, we initially tested a simple model including Kinship and Part as fixed
factors and participants and role names as random effects. Our first model included Kinship
as a main effect and Kinship by Part as an interacting effect. We then added Stereotype to the
initial model as a fixed factor, to test a Kinship by Stereotype as well as a Kinship by Part by
Stereotype interaction effects, which improved the model significantly (Δχ2 = 49.54, Δdf = 8,
p<.001). Conversely, when we added Flanker to the model to test a Kinship by Flanker as
well as a Kinship by Part by Flanker (inhibition (capacity) hypothesis) interaction effects, as
in Experiment 1, there was no significant improvement to the model (Δχ2 = 6.60, Δdf = 4,
ns).
The model that fit the data best (i.e., with Stereotype) revealed a main effect of
Kinship (F (1, 8844) = 121.93, p <.001), positive responses to pairs including male kinships
being faster (.11 standard deviations) than to pairs including female kinships (-.14 standard
deviations), and more importantly, no interaction effect of Part by Kinship (F (2, 8844) =
0.16, ns). As in Experiment 1, the model also revealed a Kinship by Stereotype interaction

The masculine form in French..... 20
effect (F (4, 8844) = 8.39, p<.001), responses to pairs including female kinships being only
.12 standard deviations (z-score) slower than pairs including male kinships in the female
stereotype condition compared to .28 slower in the male stereotype condition and .30 in the
neutral stereotype condition. A three-way Kinship by Stereotype by Part (F (4, 8844) = 4.65,
p<.001) interaction effect suggested, as hinted by Figure 3, that this effect was only present in
Part I (.003 std-difference in the female stereotype condition; .41 in the male condition and
.43 in the neutral condition) as in Part II, all conditions were similar (.23 std-difference in the
female stereotype condition; .24 in the male condition and .23 in the neutral condition).
Overall, the additional constraint enforced by role names written in the feminine form did not
alter the results found in Experiment 1. Again, there was an increase in readers’ proportion of
positive answers for masculine form/female referent combinations, and of differentiated
response times, as positive response times were slower for masculine form/female referent
combinations, regardless of the given instructions. Again, and contrary to our expectations,
none of the variance could significantly be accounted for by inhibition capacities, as measured
by the Flanker task. Considering the strong similarities between the results of Experiments 1
and 2, we deemed it unnecessary to run an analysis to compare the two experiments.
General discussion
In this paper, we present two experiments that examined (a) whether motivating
readers to consider the masuline form as representative of both women and men could
mitigate the male dominance found in previous studies (e.g., Gygax & Gabriel, 2008), (b)
whether such a mitigation would reveal stereotypicality effects rarely found in studies on
gender marked languages and (c) whether such a mitigation depended on readers’ general
inhibition capacities, as measured by the Flanker task. Overall, we expected that although
readers may be able to improve their performance when given our particular instructions,
response times should still signal the activation of the specific meaning of the masculine
form.

The masculine form in French..... 21
In both experiments, when explicitly reminded of the possible generic interpretation of
the masculine form, participants did increase their proportion of positive responses when they
had to decide whether a woman could be part of a group represented by a role name written in
the masculine plural form (i.e., sister-musicians). However, response times showed that it
took them longer to do so, with or without the instructions. Our results therefore support the
idea that linking masculine forms to female referents is more difficult than linking masculine
forms to male referents, regardless of what readers are instructed to do. Although our
experiments were conducted in French, in light of previous research on the topic, we expect
very similar results in other gender-marked languages.
Both experiments also indicated some influence of stereotypicality. In the first part of
each experiment (i.e., without specific instructions), and only in the first part, participants’
male dominant representation was not as pronounced in the female stereotypical condition as
in the other stereotypical conditions. Stereotypicality was central in this paper, but we
assumed that stereotypicality effects would only be apparent in the second part of the
experiments. In fact, we hypothesised that by lifting the influence of the masculine-as-specific
form, our results would mirror those of Cacciari and Padovani (2007, Experiment 2) who
found some signals of stereotype effects for both male and female stereotyped role names4. In
both our experiments, stereotypicality effects in response times were associated only to
female stereotyped role names and vanished in the second part of each experiment. We
believe that this bound-to-Part I stereotypicality effect, together with a strong male bias
imputed to the masculine form, suggests the idea that when given no particular instructions,
readers are faced with different sources of information – most likely activated in parallel –
when dealing with the ambiguity of the masculine form. Both the means and standard
deviations in Part I highlight the more difficult task that readers have when no clear
instructions are given as to how to interpret the masculine form. When given clear
instructions, readers may well override some of the processes initially activated, especially

The masculine form in French..... 22
those only sparsially so, such as stereotypes.
We also expected, in both experiments, inhibition capacities to be influential. We
hypothesised that, as overlooking grammatical information could be one way to meet our
instructions, participants would require particular inhibition capacities to do so. Executive
control, which capacities we measured using the Flanker task, might thus be required. Our
results indicate no influential effect of inhibition capacities. If null effects are often quite
difficult and delicate to interpret (see note #3 for a short discussion on the different strategies
that we adopted to analyse our inhibition scores), we would like to propose a tentative
explanation to account for the lack of inhibition capacities influence. Although we introduced
the paper by assuming that to adopt a generic interpretation of the masculine form, the
specific one must be inhibited, one could argue that this is not quite true. Recall that the
masculine generic interpretation is supposed to be activated in three cases: when referring to
persons of unknown sex, to persons where the sex (of the person) is irrelevant or to a group of
people of both sexes. In all three cases, the likelihood of a man or men being represented is
actually quite high. In fact, the grammatical rule stipulates that one man suffices for a group
to be written in the masculine form. Therefore, a generic interpretation could be considered as
one that is not different from the specific interpretation per se, but more as one that includes
the specific one and also embraces an additional female representation. As such, a generic
interpretation should be considered a cumulative representation which may not require
inhibition processes, but instead additional activation. Investigating working memory
capacities might therefore be interesting to access readers' difficulties in embracing such a
representation, as well as focusing on some measure of activation capacities (although these
often seem to be referred to as, or at least associated with, inhibition capacities). In sum, any
explanation of the mechanisms involved in the resolution of the semantic ambiguity
introduced by the use of the masculine form has to account for the semantic association, or
even overlap, between the two meanings. For this, one might also want to consider a task

The masculine form in French..... 23
involving semantic activation such as the Stroop task (e.g., Brown & Besner, 2001; Klopfer,
1996). Altogether, we believe that future work should investigate more closely the cognitive
factors – not just societal factors (e.g., sexist attitudes as in Gabriel et al., 2011) –, that
contribute to differences in the way readers process the masculine form as well as the way
they switch from one interpretation to another.
With reference to our paradigm, we should keep in mind that although the proportion
of positive responses does suggest that readers managed to adopt a generic interpretation of
the masculine form, it only constitutes an off-line indication of what might be. It might even
only mirror participants' attentional capacity to detect the signals that activate the rule that
was given to them just before the second part of the experiment, without them truly
embracing a balanced representation associated with positive responses. Response times, on
the other hand, may more accurately reflect the content of participants' mental representation
constructed (Keenan, Potts, Golding, & Jennings, 1990) when reading the role names in the
masculine form. If this is the case, our results indicate two different processes : (a) readers, by
default, maintain a rather superficial processing of the masculine form (i.e., by default, they
go for the good-enough representation), as hinted by the response times, and (b) the generic
interpretation is being activated only when a more elaborate processing is motivated, forcing
readers to reassess their mental model and change their initial representation. Essentially, our
data suggest that the specific meaning, as the default value, is activated through a passive
process, whereas the generic, more elaborate one, needs active processing. The specific
meaning of the masculine form typically fits a good-enough representation, but when readers
are motivated to build a representation based on another meaning, they need to go beyond this
representation. The term beyond is crucial – and we purposely do not use the term replace
here –, as our response times do indicate traces of this initial good-enough representation.
Now that we have shown differential processes for both meanings, more data is needed to
examine the conditions favoring them.

The masculine form in French..... 24
Thus, future work might also consider more implicit ways (as you would expect in
everyday life) to motivate readers to interpret the masculine form as generic, explicit
instructions as the ones reported in this paper potentially triggering counter-productive
reactance processes.
Still, our results have one main implication. Giving specific instructions for an
intended generic meaning of the masculine form may not result in truly balanced mental
representations of gender. People, being reluctant to replace the masculine form with gender-
fair forms (e.g., les musiciennes et musiciens [the female and male musicians]) yet still keen
to target both men and women (e.g., in job advertisements), often chose to explain at the
beginning of a text (or in a footnote, as discussed by Rothmund & Scheele, 2004) that the
masculine form is “aimed at both men and women”. In light of the results of our experiments,
we believe that this is not the most adequate solution. If our proportion of positive answers
suggests that it might, our response time data suggest that it might not, resulting in a possible
hampering of women's visibility in society (e.g., Braun, 1996; Bussmann, 1995; Peyer &
Wyss, 1998).

The masculine form in French..... 25
References
Baayen, R. H. (2008). Analyzing Linguistic Data. A Practical Introduction to Statistics Using
R. Cambridge University Press.
Banaji, M. R., & Hardin, C. D. (1996). Automatic stereotyping. Psychological Science, 7,
136-141.
Braun, F. (1996). Das grosse I und seine Schwestern - eine kritische Bewertung [The capital I
and its sisters – a critical evaluation]. Deutschunterricht, 48, 54-62.
Brown, M., & Besner, D. (2001). On a variant of Stroop’s paradigm: Which cognitions press
your buttons? Memory & Cognition, 29, 903-904.
Brysbaert, M. (2007). “The language-as-fixed-effect fallacy”: Some simple SPSS solutions to
a complex problem (Version 2.0). Royal Holloway, University of London.
Bussmann, H. (1995). Das Genus, die Grammatik und - der Mensch: Geschlechterdifferenz in
der Sprachwissenschaft [Gender differences in linguistics]. In H. Bussmann & R. Hof
(Eds.): Genus. Zur Geschlechterdifferenz in den Kulturwissenschaften (pp. 114-160).
Stuttgart: Körner.
Cacciari, C. & Padovani, R. (2007). Further evidence of gender stereotype priming in
language: Semantic facilitation and inhibition in Italian role nouns. Applied
Psycholinguistics, 28, 277-294.
Carreiras, M., Garnham, A., Oakhill, J., & Cain, K. (1996). The use of stereotypical gender
information in constructing a mental model: Evidence from English and Spanish.
Quarterly Journal Of Experimental Psychology Section A-Human Experimental
Psychology, 49, 639-663.
Cohen , J.D., MacWhinney, B., Flatt, M., & Provost, J. (1993). PsyScope: A new graphic
interactive environment for designing psychology experiments. Behavioral Research
Methods, Instruments & Computers, 25, 257-271.
Colé, P. & Segui, J. (1994). Grammatical incongruence and vocabulary types. Memory &

The masculine form in French..... 26
Cognition, 22, 387-394.
Cook, A. E., & Guéraud, S. (2005). What have we been missing? The role of general world
knowledge in Discourse processing. Discourse Processes, 39, 265-278.
Eriksen, B.A., & Eriksen, C. W. (1974). Effect of noise letters upon the identification of a
target letter in a nonsearch task. Perception & Psychophysics, 16, 143-149.
Fan, J., McCandliss, B.D., Sommer, T., Raz, A. & Posner, M.I. (2002). Testing the efficiency
and independence of attentional networks. Journal of Cognitive Neuroscience, 14,
340-347.
Ferreira, F., Bailey, K. G. D., & Ferraro, V. (2002). Good-enough representations in language
comprehension. Current Directions in Psychological Science, 11, 11-15.
Flaherty, M. (2001). How a language gender system creeps into perception. Journal of Cross-
Cultural Psychology, 32, 18-31.
Gabriel, U. (2008). Language policies and in-group favouritism: The malleability of the
interpretation of generically intended masculine forms. Social Psychology, 39, 103-
107.
Gabriel, U. & Gygax, P. (2008). Can language amendments change gender representation?
The case of Norway. Scandinavian Journal of Psychology, 49, 451-457.
Gabriel U., Gygax, P., Sarrasin, O., Garnham, A., & Oakhill, J. (2008). Au-pairs are rarely
male: Norms on the gender perception of role names across English, French and
German. Behavior Research Methods, 40, 206-212.
Gabriel, U., Gygax, P., Sarrasin, O., Garnham, A., & Oakhill, J. (2011). Pronouns’ impact on
the mental representation of gender in French and German: When beauticians,
musicians and mechanics remain men. Manuscript submitted for publication.
Garnham, A., Oakhill, J., & Reynolds, D. (2002). Are inferences from stereotyped role names
to characters' gender made elaboratively? Memory & Cognition, 30, 439-446.
Gerrig, R., & McKoon, G. (1998). The readiness is all: The functionality of memory-based

The masculine form in French..... 27
text processing. Discourse Processes, 26, 67-86.
Graesser, A. C., Singer, M., & Tabasso, T. (1994). Constructing inferences during narrative
text comprehension. Psychological Review, 101, 371-395.
Gygax, P., & Gabriel, U. (2008). Can a group of musicians be composed of women? Generic
interpretation of French masculine role names in absence and presence of feminine
forms. Swiss Journal of Psychology, 67, 143-151.
Gygax, P., Gabriel, U., Sarrasin, O., Oakhill, J., & Garnham, A. (2008). Generically intended,
but specifically interpreted: When beauticians, musicians and mechanics are all men.
Language and Cognitive Processes, 23, 464-485.
Gygax, P., & Gabriel, U. (2010). Gender representation in language: More than meets the eye.
Manuscript in press in Mishra, R.M. & Srinivasan, N. (eds). Language & Cognition :
State of the Art.
Irmen, L. (2007). What's in a (role) name? Formal and conceptual aspects of comprehending
personal nouns. Journal of Psycholinguistic Research, 36, 431-456.
Irmen, L., & Kurovskaja, J. (2010). On the semantic content of grammatical gender and its
impact on the representation of human referents. Experimental Psychology, 57, 367-
375.
Keenan, J. M., Potts, G. R., Golding, J. M., & Jennings, T. M. (1990). Which elaborative
inferences are drawn during reading? A question of methodologies. In D. A. Balota, F.
d'Arcais & K. Rayner (Eds.), Comprehension processes in reading (pp. 377-403).
Hove: LEA.
Klopfer, D. S. (1996) . Stroop interference and color-word similarity. Psychological Science,
7, 150- 157.
Peyer, A., & Wyss, E. L. (1998). "JazzmusikerInnen - weder Asketen noch Müesli-Fifis".
Feministische Sprachkritik in der Schweiz, ein Überblick [„Jazz-musicians – neither
ascetics nor cereals-picker“ – Feministic language critic in Switzerland, a review].

The masculine form in French..... 28
Germanistische Linguistik, 139-140, 117-154.
Posner, M. I., & Petersen, S. E. (1990). The attention system of the human brain. Annual
Review of Neuroscience, 13, 25-42.
R Development Core Team (2010). R: A Language and Environment for Statistical
Computing. R Foundation for Statistical Computing: Vienna, Austria.
Rothmund, J. & Scheele, B. (2004). Personenbezeichnungsmodelle auf dem Prüfstand.
Lösungsmöglichkeiten für das Genus-Sexus-Problem auf Textebene. [Putting
genderneutral reference terms to the test: Constructive solutions to the problem of
grammatical vs. referential gender on the textual level]. Zeitschrift für Psychologie,
212, 40-54.
Sarrasin, O., Gabriel, U. & Gygax, P. (in press). Sexism and attitudes toward gender-neutral
language: The case of English, French and German. Swiss Journal of Psychology.
Stahlberg, D., Braun, F., Irmen L., & Sczesny, S. (2007). Representation of the sexes in
language. In Fiedler (Ed.), Social communication. Editors: Kruglanski & Forgas. New
York: Psychology Press.
Trueswell, J. C., Tanenhaus, M. K., & Garnsey, S. M. (1994). Semantic influences on parsing:
Use of thematic role information in syntactic ambiguity resolution. Journal of Memory
& Language, 33, 285-318.

The masculine form in French..... 29
Footnotes
1 In the analysis, there were 24 participants in the high inhibition group and 25 in the
low inhibition group. We also tried the analyses first by taking 25 participants in the high
inhibition group and second by extracting the extra participants. All analyses were identical,
hence we present only the one with 24 participants in the high inhibition group and 25 in the
low inhibition group.
2 All analyses were conducted using version 2.10.1 of R for Mac OS X. The models
were tested with the lme4 package. The Chi-squared values for evaluating log-likelihood
differences between models were obtained with the anova() function (Baayen, 2008). And
finally, the p-values, F-values and degrees of freedom estimates were obtained with the
aovlmer.fnc() function.
3 To ensure that the null results associated with the Flanker factor were not due to our
median split, we performed extra analyses adopting different strategies: (a) we only compared
the lower and the higher quartiles, (b) the lower and the higher thirds, and (c) included
Flanker task scores as a covariate. None of these bared any difference to the analyses
presented in this paper. Finally, as a last attempt to examine possible inhibition effects, we
performed correlation analyses between Flanker scores and positive response times to pairs
including a female kinship in the second part of each experiment (overall and per stereotyped
condition). We presupposed that if any, inhibition capacities should have an effect on the
most difficult positive responses to generate. No correlation was significant, or even close to
significant (all p>.10).
4 Note that in Cacciari and Padovani (2007), the stereotype effects were only revealed
when both the prime presentation time and the prime-target interval were prolonged
(Experiment 2). In addition to this, female pronouns were responded to equally fast when
primed by a male stereotyped role name or a neutral stereotyped condition, suggesting only a
partial stereotype effects (i.e., female pronouns were still responded to faster when primed by

The masculine form in French..... 30
female stereotype role names). The explanation of these effects goes beyond this present
paper.

The masculine form in French..... 31
Table 1
French role names chosen from Gabriel et al. (2008) along with the proportion of men
evaluated by each language participant group.

The masculine form in French..... 32
Figure Captions
Figure 1. Raw response times of positive responses in the different conditions in Experiment
1. Note that the slight difference in response times between the three stereotyped conditions
can be attributed to differences in word length, the female stereotyped condition being
composed of role names longer (M=13.75) than the male (10.92) and neutral conditions
(11.08). As explained in the Results section, word length was corrected for in all analyses.
Figure 2. Proportion of positive responses in Experiment 2.
Figure 3. Raw response times of positive responses in the different conditions in Experiment
2. Note that the slight difference in response times between the three stereotyped conditions
can be attributed to differences in word length, the female stereotyped condition being
composed of role names longer (M=13.75) than the male (10.92) and neutral conditions
(11.08). As explained in the Results section, word length was corrected for in all analyses.

The masculine form in French..... 33
Figure 1

The masculine form in French..... 34
Figure 2.

The masculine form in French..... 35
Figure 3.