Gender differences in graphic design for the Web
Institute of Informatics
University of Namur
rue Grandgagnage, 21
Institute of Informatics
University of Namur
rue Grandgagnage, 21
Nowadays, diﬀerences between men and women in computer
science have become a concern in the scientiﬁc society. But
few studies focus on possible gender diﬀerences in web de-
sign. We have thus tackled the problem and conducted a
statistical analysis in the ﬁeld. We have selected male and
female sites and analyzed them according to a list of graphic
variables including the number of colours in the site, the type
of these, the type of background, the presence of graphics
and their types etc. We have also questionned male and
female students about their preferences in web design.
Categories and Subject Descriptors
D.2.2 [Design Tools and Techniques]: User Interface
Gender diﬀerences, web design, visualization
Many studies focus on the human being to try to understand
how we behave, think and so on. In today’s society, we have
to cope with a tool which is now part of our daily lives: the
computer. We have thus developed a way of interacting with
the machine. It appeared that men and women don’t behave
in the same way regarding the latter. Scientists then decided
to tackle this problem and to ﬁnd what these diﬀerences
consist of and why they exist. For example, they have found
diﬀerences in CMC (Computer Mediated Communication),
programming styles, the use of the Internet, computer games
and so on.
However, few investigations have been conducted regarding
web design. It thus seemed interesting to get into the topic
in greater depth and to ﬁnd if there are diﬀerences between
men and women in this particular ﬁeld of Human Computer
In the following sections, we will present a few gender stud-
ies in the context of web design. We will especially focus on
the visual aspect of the latter. Then, we will introduce you
to the research we have conducted. This survey consists of
a quantitative analysis among Australian academics. Actu-
ally, we have assessed 15 male and 15 female sites according
to a list of features, among which some are visual. But we
also wanted to question today’s young generation and we
have thus collected students’ opinions about their prefer-
ences in web design.
2. PREVIOUS RESEARCH
Even if gender studies related to web design are not numer-
ous, they examine gender patterns in interface design for
many categories of people, from the kindergarten children’s
preferences to academics’. In , Simon looked at the im-
pact of gender on websites. Out of 160 female and male
students, females indicated they overwhelmingly (84%) pre-
fer sites that are less cluttered, with minimal use of graph-
ics. Females (52%) suggested that sites making use of pull-
down menus are easier to navigate than those with levels
that require them to click through to achieve their objec-
tive. Males, on the other hand, indicated that sites making
extensive use of graphics are clearly their preference (77%).
Although ﬁndings are ambiguous, many investigations have
indicated that there are diﬀerences between gender regard-
ing preferences for colors. A review of color studies done
by Eysenck in early 1940’s notes the following results to the
relationship between gender and color . St.George (1938)
found that blue for men stands out far more than for women.
An even earlier study by Jastrow found men preferred blue
to red and women red to blue. In expressing the prefer-
ences for light versus dark colors, there was no signiﬁcant
diﬀerence between men and women.
In , Arnold and Miller looked at homepages produced by
people in institutional or commercial settings. Given that
it is often suggested it is particularly in such settings that
women ﬁnd it diﬃcult to have their status, authority and
credibility recognised, the authors decided to see how the
“oﬃcial” personal web pages of women and men might diﬀer.
They selected some academics’ websites and found several
diﬀerences. Often amongst the women (though they suspect
in some cases with heavy irony) there was a “feminine” style
of self-image, but they have not found any women’s pages
that use jokey pictures of themselves, as some men do. On
the web, people can “belong” to a group of people, e.g. in
a department or subject grouping, which is dominated by a
house style. Yet even here, gender diﬀerences intrude in the
cyberspace equivalent of “ﬂuﬀy” feminine (such as the use
of a substitute picture e.g. “ﬂowers”) compared to technical
“images” (e.g. a computer) used by men.
In , it was found out that while several male academics
also include family within their online representations, men’s
pages tend to focus more on presenting a self-image to the
viewer. Women’s pages, in contrast, often feature more pic-
tures of family members than themselves, and in many cases
completely exclude their own image.
In , Miller and Mather looked at 35 women’s and 35 men’s
personal homepages. The authors identiﬁed four categories
of self-image on the page:
1. straight: an image which is meant to be a straightfor-
2. joke: a distorted or caricatured or unrepresentative im-
age, e.g. cartoon, baby photo, author just after falling oﬀ
bike into mudhole, author caricatures as frog, etc
3. symbolic: an image which represents a human being,
but not the actual person who posted the page. This is of-
ten a piece of clip art, like a cherub or a generic silhouette
4. none: no images of humans.
The authors of this study counted blurred or pixellated pho-
tos which might be of the author, but were so unclear that
they didn’t really represent an individual but didn’t ﬁnd any
gender diﬀerence. The main diﬀerence was related to the use
of jokey images. Indeed, joke images only featured on men’s
pages (on 4), and symbolic images only on women’s (on 10
pages, the most common form of image on women’s pages).
3. FOCUS ON AUSTRALIAN ACADEMICS
The quantitative part of our survey aimed at determining if
women and men design their websites in a diﬀerent way. We
thus decided to conduct the investigation in the academic
context in order to compare our results with the ﬁndings
from previous literature. The ﬁrst step in the assessment of
the sites was of course the selection of these. Academics’ web
homepages have been therefore selected among Australian
IT departments. To try to reduce the inﬂuence of the cul-
tural factor, the sample didn’t include any Asian professor.
We also tried to make sure that the web homepage had been
designed by the academic him-/herself by checking the au-
thor of the site or by checking it was not the common style
of the department. There was no restriction regarding the
housing place of the site.
According to a ﬁrst insight of the sites, we have formulated
several hypotheses. We can classify them into two cate-
gories. The ﬁrst one consists of the numerical variables and
the second is composed of the binary variables. For the ﬁrst
category, we have:
1. Text in women’s websites is more spaced out than
- Women use more white spaces than men
- Women use more paragraphs per page than men
2. Women tend to use more diﬀerent fonts than men
3. Women tend to use more colours for text and hy-
4. Women’s websites have a more colourful back-
5. Men show more self-photos than women
6. Men show more photos apart from self-photos
7. Women use more graphics than men
For the second category, we assume:
1. Men and women both use classic fonts
2. Men and women don’t use girlish fonts
3. Men and women diﬀer in the type of colours used
for text and hypertext
- Women tend to use more reddish colours
- Men tend to use more blueish colours
- Both use black
- Men tend to use white more often than women
- Men tend to use grey more often than women
4. Men tend to have more technological websites
5. Women and men diﬀer regarding the type of
colours used for their backgrounds
- Women use softer colours
- Men use darker colours
- Women use more reddish colours
- Men use more blueish colours
- Men use black more often
- Men and women do not diﬀer regarding the use of white
- Men use grey more often
6. Men and women diﬀer regarding the type of back-
- There are more women’s websites with a classic background
than men’s sites
- There are more men’s websites with an original background
than women’s sites
7. There are more women using graphic accents
(smileys, emoticons etc) than men
8. Women and men diﬀer regarding the type of self-
- Women and men both show the oﬃcial picture
- Men show more non-oﬃcial pictures of themselves than
- Men show more family pictures with themselves than women
- Men show more pictures of themselves with friends than
- Men and women do not diﬀer regarding pictures of them-
selves with colleagues
- Men and women do not diﬀer regarding pictures of them-
selves with their pets
- Men show more pictures of themselves in their leisure time
- Men show more computer-related pictures with themselves
9. Men and women both show good-quality self-
10. Women and men diﬀer regarding the type of
- Men show more pictures of their families than women
- Men and women do not diﬀer regarding pictures showing
- Women show more pictures of their colleagues than men
- Men show more pictures of their pets than women
- Men show more pictures of their leisure time than women
- Men show more computer-related pictures than women
11. Women and men both show good-quality pic-
12. Men and women diﬀer in the type of graphics
- Women use more basic graphics
- Women use more modern graphics
- Women use trendier graphics
- Women use more artistic graphics
- Men use more comics graphics
- Women and men do not diﬀer regarding the use of computer-
3.2 Statistical analysis
In this paper, we have used diﬀerent kinds of statistical tests.
When collecting the data, histograms were already drawn at
that time and it turned out that most male and female dis-
tributions for the diﬀerent characteristics didn’t have a nor-
mal shape, even after transformations. That is why we have
applied non-parametric statistics. The Mann-Wilcoxon-
Whitney test aimed at ﬁnding a location diﬀerence between
male and female distributions. For the possible location
diﬀerences, we have applied the Hodges-Lehmann estimator
to determine the extent of such a diﬀerence. To test the
equality of variances, we have used the squared rank test for
variance. Finally, we applied the Smirnov test to deter-
mine if it was reasonable to assume that the male sample
and the female one came from identically distributed popu-
lations. We have also run a k-means clustering method
in order to classify the academics.
In our list of variables, some are binary as described in the
hypotheses. To analyze these, we have applied the Fisher’s
test in order to ﬁgure out if our male and female popula-
tions were homogeneous or not. The binomial test has
been used to bring out genderless tendencies among the aca-
demics. The discriminant method allowed us to know
which variables were most discriminant. We have also run
a multiple correspondence analysis in order to ﬁnd variable
patterns. Finally, the segmentation method highlighted pro-
ﬁles among our sample.
Let’s ﬁrst mention that the Smirnov test aiming at stating
if males and females could come from a common population
was never signiﬁcative. That is why we won’t mention this
test in the following results.
The ﬁrst question we raised was to ﬁgure out if women’s
websites were more spaced-out than men’s. In this case, we
looked at the proportion of white spaces and the number of
paragraphs per page. The location test was not signiﬁcative
for none of the features. Regarding the variances, we could
conclude to a variability diﬀerence for the number of para-
graphs per page (with an error of 5%). In general, apart
from this variability diﬀerence, we couldn’t draw any other
Regarding the number of fonts, nor the location test or the
variance test were signiﬁcative. Thus, we couldn’t say if
women use more fonts than men or vice-versa. The same
statement could be made for the number of colours for text
and hypertext. On the face of it, we could have thought that
women’s websites would have a more colourful background,
but again, the tests were not signiﬁcative, preventing us from
According to previous literature, we could think that women
would show fewer self-photos on their websites than men.
But the only result we had is a variability diﬀerence between
men and women regarding this variable (with an error of
5%). Regarding the number of photos, we had a location and
a variability diﬀerence (both with an error of 5%). However,
the Hodges-Lehmann estimator indicated there was only a
diﬀerence of one photo between men’s and women’s websites.
For the number of graphics, there was only a variability
diﬀerence again (with an error of 5%).
When classifying the academics, we obtained two clusters.
The second one was very interesting since it only consisted of
two females. These diﬀer from the other academics (cluster
1) by the number of photos (1083 for female 1), the number
of graphics (2855 for female 1 and 1217 for female 2) and
the number of self-photos (89 for female 2).
The following table will help you understand how the Fisher’s
GraphicAccents yes no total
M0 15 15
F5 10 15
total 5 25 30
As you can see in this ﬁrst table, no male use graphic ac-
cents (smileys, emoticons etc) whereas ﬁve females do. For
the analysis, let’s consider the cell situated at the intersec-
tion between the ﬁrst row and the ﬁrst column, denoted by
C(1,1), being A,C(1,2) being B,C(2,1) being C,C(2,2)
being Dand H0stating both populations are homogeneous.
According to Fisher’s tables, for A + B = 15, C + D =
15 and B = 15, the maximum value for D (above which we
cannot reject H0) is 11. Since our D (the one in the table)
equals 10, we can reject H0with an error of 5 %. Regard-
ing the presence of photos, we have the following table:
Photos yes no total
M9 6 15
F2 13 15
total 11 19 30
According to Fisher’s tables, for A + B = 15, C + D =
15 and A = 9, the maximum value for C (above which we
cannot reject H0) is 3. Since our C (the one in the table)
equals 2, we can reject H0with an error of 5 %. If we
consider all the binary variables, we have found out that
only the binary variables “graphic accents” and “photos”
allowed to reject the homogeneity hypothesis stating that
the male and female populations are homogeneous. That
means males and females have diﬀerent behaviours: women
use more graphic accents and males put more photos on
their websites than women.
Since the Fisher’s test only put two diﬀerences forward, we
have conducted a binomial test to see if there were any gen-
derless diﬀerences of behaviour. This test has given inter-
esting results. A majority of academics don’t have a tech-
nological site, use classic fonts and no girlish fonts. They
use bleuish colours, black but no white or grey for text and
hypertext. They don’t show graphic accents but include
graphics in their websites. Their backgrounds are not black
nor dark or blueish but white in general. Backgrounds are
also classic most of the time and not original.
The discriminant analysis has highlighted the variable “graphic
accents” better separate the men from the women than the
variable “photos”. The canonical discriminant function is
F(x) = −2,041g+ 1,759p−0,305
with gstanding for “graphic accents” and pfor “photos”.
Since the absolute value of the coeﬃcient of the graphic
accents is greater than the one for the photos, it conﬁrms
the fact graphic accents is the best discriminating variable.
The classiﬁcation functions are the following ones:
F0(x) = 2,747g+ 0,528p−1,186
F1(x) = −0,317g+ 3,170p−1,644.
We clearly have a much higher coeﬃcient for the females re-
garding the graphic accents, meaning these use more graphic
accents than the males. But regarding the photos, we have
the contrary. Let’s remember that a male (female) do not
necessarily behave like a male (female). When running the
classiﬁcation functions on our sample, we have found out
that one female had a higher score with F1(x), that is to say
her proﬁle corresponds to a male whereas six males had a
higher score with F0(x), thus behaving like a female.
The segmentation analysis has allowed us to divide the males
and the females of our sample into groups which are as ho-
mogeneous as possible. Actually, the segmentation method
uses another method called “decision-tree learning”. A deci-
sion tree describes a tree structure wherein leaves represent
classiﬁcations and branches represent conjunctions of vari-
ables that lead to those classiﬁcations. A decision tree can
be learned by splitting the source set into subsets based on
an variable value test. This process is repeated on each de-
rived subset in a recursive manner. Splitting is done thanks
to the computation of a distance measure. The recursion is
completed when splitting is either non-feasible, or a singular
classiﬁcation can be applied to each element of the derived
subset. In the context of our research, the set of variables
was composed of the binary variables. We chose the entropy
reduction method as a distance measure . The results
are shown on ﬁgure 1. As you can see, the ﬁrst discriminant
variable is the graphic accents. What we can conclude from
this tree conﬁrms the results from the discriminant analysis
and the Fisher’s test. Indeed, the variable “graphic accents”
best discriminate males and females. For those who do not
use graphic accents (right branch of the tree), the presence
of photos is the most discriminant factor.
We have also conducted a multiple correspondence analysis.
This kind of analysis is part of the factorial analyses group.
The goal of factorial analyses is to summarize and organize
the information into a hierarchy, information which can be
found in a matrix of n rows (the subjects) and p columns
(the variables). The n subjects are described by a cloud of
p variables. The information represented by this cloud is
the dispersion of the npoints. So, computing a summary of
this information means projecting these points into a space
which dimension is below p. The axes of this subspace are
called “factors”. In the context of the multiple correspon-
dence analysis, the p variables are qualitative (modality 0
for its absence, modality 1 for its presence). For our sur-
vey, we focussed on the ﬁrst ten factors since they explain
71% of the dispersion. For each axis, we kept the four most
important variables for the negative part and the four most
important variables for the positive one. Then we projected
the males and females of our sample into factorial plans.
That is how we could conclude that males tend to have an
original background that is not white, with trendy graphics.
Indeed, regarding the factorial plan consisting of axis 3 and
axis 4 (ﬁgure 2), we can see there are only three males
(males are in dark blue) in the bottom part of the chart,
meaning males tend to have an original background that
is not white, with trendy graphics. Women (women are in
light pink) do not seem to have a particular preference
since they can be found in both parts of the plan. Let’s
note that the negative part of axis 3 consists of the follow-
ing variables: the presence of girlish fonts, then the absence
of classic fonts, the presence of non-oﬃcial self-photos and
the absence of blueish colours for text. Regarding the posi-
tive part of the same axis, we have the following order: the
absence of girlish fonts, the presence of classic fonts, the
absence of non-oﬃcial self-photos, the presence of blueish
colours for text. For axis 4, the variables composing the neg-
ative part of the axis are ordered as follows: the absence of
an original background, the absence of trendy graphics, the
presence of a white background, the presence of leisure time
photos. Regarding the positive part of axis 4, we have: the
presence of an original background, the presence of trendy
graphics, the absence of a white background and ﬁnally the
absence of leisure time photos. By considering the whole
sample, that is to say males and females put together with-
out any gender distinction, we could bring out two tenden-
cies. The ﬁrst one represents the less common behaviour
among our academics, which is deﬁned by the fact of having
a website with an original background, girlish fonts, col-
leagues and computer-related self-photos, pets photos and
no blueish colours for text and hypertext. The second ten-
dency represents the most common behaviour among the
professors, deﬁned by the fact of having a website with a
classic background, no colleagues and computer-related self-
photos, classic fonts, blueish colours for text and hypertext
and no pets photos.
4. SURVEY AMONG AUSTRALIAN STU-
In the frame of our research, we have also conducted a survey
among students of a web design class. These students had to
answer ﬁfty questions about their design preferences. Ninety
questionnaires were collected: ﬁfty-ﬁve from male students
and thirty-ﬁve from female students. The ﬁrst questions fo-
cussed on the students’ preferences in web design. Then,
they focussed on the way they would design their own web
homepages. For each question (apart from questions 1 and
2), the students could answer “I strongly disagree”, “I dis-
agree”, “I am neutral” (no preference), “I agree” and “I
Here is the list of questions. 1. Are you a male or a female?
2. What nationality are you?
3. I prefer a page for which I have to scroll down in order
to see all text than a page in which all text is cluttered.
4. I prefer pull-down menus.
5. I prefer menus you have to click through in order to
achieve my goal.
Figure 1: Segmentation result - 1 represents the presence of the variable, 0 the absence
Figure 2: MCA for BN: axis 3 vs axis 4
6. I prefer when a variety of fonts are used for text and
7. I prefer soft colours like pastel colours to dark colours
like dark blue or black.
8. I prefer reddish colours (red, yellow, pink, orange etc) to
blueish colours (blue, green, purple etc).
9. I prefer when many colours are used for text and hyper-
10. I prefer websites in which there are many white spaces
between the elements of the site (images, text etc).
11. I prefer when there are a lot of white spaces inside a
12. I prefer when there are many static images in the site.
13. I prefer when there are many graphic animations on a
14. I prefer trendy graphics to basic graphics.
15. I prefer comics graphics to basic graphics.
16. I prefer computer-related graphics to basic graphics.
17. I prefer jokey graphics to basic graphics.
18. I prefer a site in which the background is colourful.
19. I prefer a site in which the pages do not look similar.
For instance, if there are 10 pages on the site, I prefer when
they don’t have the same background, the same fonts etc.
20. I prefer a background with motifs than a plain back-
If I had to design my own web homepage...
21. I would put a picture of myself.
22. I would put many pictures of myself.
23. I would put jokey pictures of myself.
24. I would put pictures showing myself on the main page.
25. I would insert pictures of my private life (family, pets,
friends, leisure time etc) in my own site.
26. I would insert many pictures of my private life (family,
pets, friends, leisure time etc) in my own site.
27. I would put pictures representing my private life on the
28. If I had to include graphics in my web homepage, I
would try to make these jokey.
We have carried out a discriminant analysis in order to know
which questions separate the best the male and the female
students. The preference for menus you have to click
through in order to achieve your goals was the most dis-
criminant feature between male and female students. Actu-
ally, the discriminant analysis showed the girls have a greater
preference for menus you have to click through than the
boys. In , 52 % of women said they prefer pull-down
menus rather than navigating through the site. The two
other questions which could discriminate male and female
students are content questions and not design questions.
The segmentation analysis didn’t allow to bring out binary
diﬀerences between males and females, that is to say fea-
tures that men would ﬁt and that women wouldn’t or vice
versa. Unfortunately we haven’t enough space to display the
ntire tree. But on the other hand, we highlighted 7 homoge-
neous leaves. It means that the semgentation analysis allows
to bring out some “proﬁles” regarding the students’ prefer-
ences, with some groups consisting of a majority of males
and others of a majority of females. In order to highlight
these, we have used the SODAS (version 2.5) data mining
software . Each proﬁle is represented by a star. For each
Figure 3: Stars for the groups of males: M1 being
the ﬁrst group, M2 the second and M3 the third
Figure 4: Stars for the groups of females: F1 being
the ﬁrst group, F2 the second and F3 the third
star, the axes represent the variables, that is to say the ques-
tions we want to study in greater depth. Each axis has been
assigned a scale from 1 to 3, 1 for “I don’t agree”’, 2 for “I
am neutral and 3 for “I agree”. For each value, histograms
show the proportion of students having chosen the corre-
sponding answer. In this paper, we will only present a few
of these proﬁles according to their general preferences in web
design. On 3, you can see the histograms for three groups of
males. On 4, you can see those for three groups of females.
•Their preferences in web design...
Regarding the males of the ﬁrst group (M1: 11 males,1
female), we know they do not like when many colours are
used for text (Q9) as we could expect from boys (the cliche
being boys do not like to use many colours like girls). They
don’t like spaced-out websites (Q10) either, what we could
expect since this feature is the prerogative of women accord-
ing to . They don’t care about menus you have to click
through (Q5). They don’t agree or they are neutral con-
cerning pages for which you have to scroll down (Q3).
As the males of the ﬁrst group, the males of the second
group (M2: 10 males, 3 females) do not like when many
colours are used for text (Q9). They do not like pages for
which they have to scroll down (Q3), thus going along with
the fact the males of the ﬁrst group do not like spaced-out
websites. Regarding menus you have to click through (Q5),
they do not show any tendency. For spaced-out websites
(Q10), the males having answered do not agree, meaning
they do not like this kind of websites.
The males of the third group (M3: 9 males, 1 female)
are more tolerant than the two previous groups of males
regarding the use of many colours for text (Q9) since they
don’t care about it. Regarding the menus you have to click
through (Q5), they don’t care like the ﬁrst and second group
of males. They tend to be neutral or to agree with pages
you have to scroll down to see all text (Q3). Concerning
spaced-out websites (Q10), they do not show any particular
The ﬁrst group of females (F1: 8 females, 3 males) do
not care about the use of many colours for text (Q9) (like
the males of the third group). We could have thought they
would have answered they prefer when many colours are
used after reading . They don’t care about spaced-out
websites either (Q10). We could have thought they would
prefer spaced-out websites after reading the same paper.
Like the boys, they don’t care about menus you have to
click through (Q5). We could have thought they wouldn’t
like this kind of menus since you have to be better at spatial
skills (like the boys) since they are not pull-down menus.
They do not show any particular tendency for pages you
have to scroll down (Q3).
Like the males of the second group, the second group of
females (F2: 8 females, 3 males) doesn’t care if they have to
scroll down or if they have to read cluttered text (Q3). We
could have thought they would prefer pages to scroll down
after reading . Like the other groups, they don’t care
about menus you have to click through (Q5). For the use of
colours (Q9), they do not show any tendency. They tend to
be neutral or to agree with spaced-out websites (Q10)
The females of the third group (F3: 9 females, 3 males)
do not like when many colours are used for text (Q9) like
the males of the ﬁrst and second groups (and unlike the fe-
males interviewed in ). They do prefer a page for which
they have to scroll down (Q3), like the females of . They
do also prefer menus you have to click through (Q5). They
must be able to situate themselves easily in unfamiliar envi-
ronments then. Regarding spaced-out websites (Q10), they
tend to be neutral or to agree.
•When we compare some groups of males with
some groups of females...
If we focus on the second group of females and the sec-
ond group of males (F2 and M2), we will notice they share
the same characteristic: they are more extrovert than the
other groups of the same gender. Both are neutral regard-
ing menus you have to click through (Q5). However, the
males do not like pages for which they have to scroll down
(Q3) whereas the females are neutral regarding this feature.
The third group of females and the third group of
males (F3 and M3) have diﬀerent answering behaviours re-
garding their preferences.The females don’t like when many
colours are used for text and hypertext (Q9). The males
don’t seem to care about an intensive use of colours. Re-
garding the menus you have to click through (Q5), the girls
prefer these whereas the boys don’t have any particular opin-
ion regarding the topic.
If we compare the males of the third group with the
females of the ﬁrst group (M3 and F1), we will notice
both don’t care about menus you have to click through (Q5).
They don’t mind either about an intensive use of colours for
Now if we focus on the males of the second group and
the females of the third group (M2 and F3), we will no-
tice that both do not like an intensive use of colours for text
(Q9). However, the boys do not like pages for which they
have to scroll down (Q3) whereas the girls do. Regarding
menus you have to click through (Q5), the boys don’t care
about using these but the girls prefer to use this type of
The last comparison focus on the ﬁrst group of males and
the third group of females. Both dislike an intensive use
of colours for text (Q9). However, the boys don’t have
any particular opinion regarding menus you have to click
through (Q5) when the girls prefer these.
5. CONCLUSIONS AND PROSPECTS
In the section devoted to the literature review, we have re-
alized gender diﬀerences in a web design context is an issue
preoccupying many sociologists. In the following sections,
we have been able to conﬁrm some ﬁndings of previous re-
search or to highlight opposite behaviours to those described
by the sociologists. It was not always possible to conclude
for each feature after having run the statistical tests. Con-
cerning the students’ preferences, we have no clear diﬀerence
between boys and girls. But we have been able to distinguish
male and female groups with common preferences. Indeed,
some male students (but not all) do not like when many
colours are used for text and hypertext. Some do not like
spaced-out websites either. All the groups of males do not
care about menus you have to click through. Regarding their
preferences for pages to scroll down, every group of males
has its own preference. Regarding the groups of girls, some
girls are neutral or agree with the use of spaced-out websites.
Some are neutral with menus you click through. Regarding
their preferences for the use of colours for text and hyper-
text and for pages you have to scroll down, we can ﬁnd dif-
ferent preferences among the female groups. Nevertheless,
the diﬀerent analyses we have conducted have allowed us
to identify predominantly common proﬁles and types of dis-
tinct behaviours with feminine-higher or masculine-higher
However, we have to keep in mind our sample is small:
thirty academics divided into ﬁfteen males and ﬁfteen fe-
males. That is the reason why we can’t generalize and ap-
ply our ﬁndings to the whole population of male and female
academics. Therefore, it would be interesting to go further
with this study and try to conduct it on an international
level as well, that is to say in diﬀerent cultures. Thus we
would enhance the scope of this work in order to include the
We could also think of going into the subject in greater depth
by applying our study to academics who do not belong to
IT departments. Indeed, the basic idea in our study was
to choose IT professors in order to have more chance these
design their sites on their own since they had the skills to
do so. On the face of it, designing a site must be less easy
for arts professors. But of course, this should be examined
Another idea would be that each team who would continue
on this task consists of specialists forming an heterogeneous
panel: sociologists, statisticians, psychologists, amateur and
professional web designers etc. Interviews should also be
conducted to examine the results obtained with statistical
methods in greater depth and to explore the reasons why
the person has designed the website in this way and not in
Lastly, we could also think of analyzing the way professional
designers design their websites according to the gender of
the ﬁnal user. So the question would become: “Is there a
speciﬁc way of designing for women and for men?”
 J. Arnold and H. Miller. Gender and web home pages.
In CAL99 Virtuality in Education Conference,
London, UK, March 1999.
(Last visit on December 22nd, 2004).
 A. Collard. Comparaison de deux proportions: une
approche baysienne pratique dans le contexte des tables
deux fois deux avec petite taille d’chantillon. PhD
thesis, Facults universitaires Notre-Dame de la Paix
Namur, Namur, Belgique, 1998. Publi par les Presses
Universitaires de Namur.
 P. Dagnelie. Statistique thorique et applique, volume
Tome 2. De Boeck Universit, Bruxelles, Belgique,
 T. W. et Ronald Wonnacott. Statistique. Economica,
4me dition edition, Paris, France, 1991.
 H. Eysenck. A critical and experimental study of color
preferences. The American Journal of Psychology,
 H.-H.Bock and E.Diday. Analysis of Symbolic Data:
Exploratory Methods for extracting statistical
information from complex data. Springer, Germany,
 M. Hess. A nomad faculty: English studies’ online
representations of work, product and workplace.
(Last visit on December 22nd 2004).
 N. Khouw. The meaning of color for gender.
http://www.colormatters.com/khouw.html (Last visit
on July 31st 2005).
 H. Miller and R. Mather. The presentation of self in
www home pages. In IRISS ’98: Conference Papers,
Bristol, UK, March 1998.
(Last visit on December 22nd, 2004).
 T. Mitchell. Machine learning. McGraw-Hill
International Editions, Singapore, Malaysia, 1997.
 M. Noirhomme-Fraiture. Cours de probabilits et
 E. Pearson and H. Hartley. Biometrika Tables for
Statisticians. Cambridge University Press, Cambridge,
 T. Santner and M. Snell. Small-sample conﬁdence
intervals for p1-p2 and p1/p2 in 2x2 contingency
tables. Journal of the American Statistical
Association, 75(370):386–394, June 1980.
 G. Saporta. Probabilits, analyse des donnes et
statistique. Technip, Paris, France, 1990.
 S. Siegel. Nonparametric statistics for the behavioral
sciences. MacGraw-Hill, 2nd edition edition, New
York, USA, 1988.
 S. Simon. The impact of gender on web sites: An
empirical study. The DATA BASE for Advances in
Information Systems, 32(1):18–37, Winter 2001.
 P. Sprent. Applied non-parametric statistical methods.
Chapman and Hall, London, UK, 1989.
 I. Wagner, A. Birbaumer, and M. Tolar. Gender gap in
computer science: cultural and psychological factors.
In Widening Women’s Work in information and
Communication Technology: conceptual framework
and state of the art (Deliverable N1), pages 67–85.
European Commission, Brussels, Belgium, 2002.