Running head: FOOD STIMULI AND GENDER

UNIVERSITÄT HOHENHEIM
INSTITUT FÜR ERNÄHRUNGSMEDIZIN

Food Stimuli and Gender
Master thesis
Master of Science in Clinical Nutrition
Presented by
Sara Teresa Salazar Winter

First Supervisor:

Prof. Dr. Nanette Ströbele-Benschop

Second Supervisor:

Prof. Dr. Dr. Anja Bosy-Westphal

Submission date:

27.02.2017

Due date:

09.03.2017

FOOD STIMULI AND GENDER

I

Abstract
The literature consistently reports that women have a healthier food choice pattern than men.
Food choices have significant health implications, and research can help to determine
possible gender differences in food choice in order to better support necessary dietary
changes. Therefore, gender attentional bias toward high-calorie and low-calorie foods and
sweet and non-sweet foods was measured using eye-tracking technology. It was examined
whether pre-existing preferences, attitudes, and food intake history correlated with the eyetracking measures. In this study, men displayed a higher attentional bias toward high-calorie
foods than women. Food preferences and dietary intake demonstrated a strong correlation
with eye-tracking data in men; men who preferred high-calorie foods paid less attention to
low-calorie foods and men who preferred low-calorie food paid less attention to high calorie
foods. The results also suggest that more attention was paid to low-calorie foods when
participants had stronger attitudes toward a healthy diet. No significant gender bias was
present between sweet and non-sweet foods. However, the level of processing of low-calorie
stimuli seemed to influence the perception of food stimuli.

Keywords
Eye-tracking, Food Stimuli, Gender Bias, Sex Differences

FOOD STIMULI AND GENDER

II

Table of Contents
Abstract ................................................................................................................................... I
Table of Contents ................................................................................................................ II
List of Figures ...................................................................................................................... IV
List of Tables ........................................................................................................................ V
1

Introduction .................................................................................................................... 1

2

Theoretical Framework .................................................................................................. 3
2.1 Decision Theory and Human Behavior ....................................................................... 3
2.1.1 Food Choice and Human Behavior ....................................................................... 6
2.1.2 Eating Behavior and Gender ................................................................................. 8
2.1.3 Stimuli and Behavioral Response ....................................................................... 11
2.2 Eye Movements, Attention, and Preference .............................................................. 13
2.2.1 The Human Eye ................................................................................................... 13
2.2.2 Eye movements ................................................................................................... 14
2.2.3 Attention and preferences.................................................................................... 15
2.3 Healthy and Unhealthy .............................................................................................. 17

3

Experimental Design .................................................................................................... 20
3.1 Stimuli ........................................................................................................................ 20
3.2 Data collection ........................................................................................................... 25
3.2.1 Eye Tracking.................................................................................................... 25
3.2.2 Questionnaires ................................................................................................. 26
3.3 Subjects ...................................................................................................................... 29
3.4 Experimental Procedure ............................................................................................. 30
3.5 Data Analysis ............................................................................................................. 32
3.5.1 Areas of Interest: Defining and Position ............................................................. 32
3.5.2 Coding Procedure and Validation ....................................................................... 33
3.5.3 Data Preparation and Data Analysis.................................................................... 33

4

Results .......................................................................................................................... 37
4.1 Participants................................................................................................................. 37
4.2 Eye-tracking Data ...................................................................................................... 38
4.2.1 Analysis of High and Low-Calorie Groups......................................................... 38
4.2.2 Analysis of Food Groups..................................................................................... 41

FOOD STIMULI AND GENDER

III

4.3 Questionnaires ........................................................................................................... 46
4.3.1

Analysis of Attitudes Toward Healthy Eating Questionnaire ....................... 46

4.3.2 Analysis of Dutch Eating Behavior Questionnaire ............................................. 52
4.3.3 Correlations Between the FLS, the FFQ, and ET Data ....................................... 56
5.

Discussion .................................................................................................................... 59
5.1 Main Findings ............................................................................................................ 60
5.2 Methodological Considerations ................................................................................. 65
5.2.1 Stimuli and Areas of Interest ............................................................................... 65
5.2.2 Presentation Software .......................................................................................... 65
5.2.3 Task Instructions ................................................................................................. 66
5.4 Further Research ........................................................................................................ 67

References ........................................................................................................................... 68
Appendices ............................................................................................................................. I
A Index of Abbreviations ...................................................................................................I
B Flyer .............................................................................................................................. II
C Consent Information and Participant Information ....................................................... III
D Details of Stimuli ....................................................................................................... VII
E Questionnaires .............................................................................................................. IX
F Encoding Protocol ..................................................................................................... XXI
G Inter Coder Reliability ........................................................................................... XXIII
H Eye-tracking Recordings Exclusion ....................................................................... XXIV
Affirmation .................................................................................................................... XXVI
Aknowledgment ...........................................................................................................XXVII

FOOD STIMULI AND GENDER

IV

List of Figures
Figure 1: Weighing Desirability in an Intertemporal Choice Setting.................................... 4
Figure 2: Food Choice Factors and Relationships ................................................................. 7
Figure 3: “Sex Differences in the Size of Various Human Brain Regions” .......................... 8
Figure 4: Stimulus Response Model for Vision .................................................................. 12
Figure 5: The Human Eye ................................................................................................... 14
Figure 6: Caloric Density .................................................................................................... 18
Figure 7: Trial Sample ......................................................................................................... 22
Figure 8: AOI around Food Stimuli .................................................................................... 32
Figure 9: Boxplots of Low- and High-calorie ET Parameters Divided by Gender ............. 39

FOOD STIMULI AND GENDER

V

List of Tables
Table 1: Major Determinants of Food Choice ....................................................................... 6
Table 2: Comparison of Food Stimuli Categories ............................................................... 21
Table 3: Food Stimuli used for the experiment ................................................................... 23
Table 4: Inclusion and Exclusion Criteria ........................................................................... 29
Table 5: Sample Description Differentiated by Gender ...................................................... 37
Table 6: Attention Bias for Men and Women by High- and Low-calorie Group ............... 40
Table 7: Attention Bias for Men and Women by Food Group ............................................ 44
Table 8: Reliability of the EGE Scales ................................................................................ 46
Table 9 Correlations of Eye-tracking Parameters and Survey Data .................................... 47
Table 10: Correlations of Eye-tracking Parameters and Questionnaire Items in Men ........ 49
Table 11: Correlations of Eye-tracking Parameters and Questionnaire for Women ........... 51
Table 12: Reliability of FEV Scales .................................................................................... 52
Table 13: FEV Scales by Gender ........................................................................................ 52
Table 14: Correlations Between FEV and ET Variables for Men ....................................... 54
Table 15: Correlations Between FEV and ET Variables for Women ................................. 55
Table 16: Correlations from ET Data with FLS and FFQ for Men ..................................... 56
Table 17: Correlations from ET Data with FFQ and FLS for Women................................ 57

FOOD STIMULI AND GENDER

1

1

Introduction

It has been consistently reported that women have a healthier food choice pattern
than men (Garriguet, 2009; Liebman et al., 2006; Westenhoefer, 2005). Women generally
consume vegetables and fruits at higher rates and fat and salt at lower rates; women also
indicate greater health beliefs and health consciousness (Fagerli & Wandel, 1999; Leblanc,
Begin, Corneau, Dodin, & Lemieux, 2015; Wardle et al., 2004; Westenhoefer, 2005). On
the other hand, it has been reported that men have a higher consumption of calories and
foods with higher sugar, fat, and salt content (Bugge & Lavik, 2012; Rolls, Fedoroff, &
Guthrie, 1991).
Motivational variables behind gender differences in food choice have been attributed
to health consciousness and weight control (Wardle et al., 2004; Westenhoefer, 2005).
Eating behavior may also be affected by gender roles in reproduction. As mothers, women
have to maintain an adequate food supply for the survival of not only themselves but also
their offspring (Logue, 2014). Women’s historical role of being habitually responsible for
preparing and providing food for the family still appears to have an effect today (Meiselman,
2009).
The aim of this study is to examine differences between men and women’s food
perceptions to better identify motivational variables of food choice. Food choices have
significant health implications, and more knowledge in this area could provide better support
in necessary dietary changes.
In this study, attentional bias to high- versus low-calorie and sweet versus non-sweet
food stimuli was measured with eye-tracking technology. Food choice is greatly influenced
by personal preferences. Therefore, the eye-tracking parameters were compared with
questionnaires to determine whether pre-existing preferences, attitudes, and food intake

FOOD STIMULI AND GENDER

2

could explain gender differences that have been observed in previous studies of food intake
and behavior (Bugge & Lavik, 2012; Fagerli & Wandel, 1999; Leblanc et al., 2015; Wardle
et al., 2004; Westenhoefer, 2005).
The next chapter offers a literature review of the theory of human decision making,
food choice, and eye movements, providing a theoretical framework for the study. The third
chapter provides a description of the experiment design, and the fourth chapter offers the
results and discussion of the present study.

FOOD STIMULI AND GENDER

2

3

Theoretical Framework

2.1 Decision Theory and Human Behavior
Human behavior is addressed by different sciences like economics, sociology, and
psychology. Economics could be described as a science that predicts behavior. However,
economics concentrates mainly on observed actions and infers preferences, tastes, and
values through observed behaviors (Schröder, 2003). For a better understanding of behavior,
this research concentrates on not only behavior but also why certain behaviors take place.
This psychology is of great importance in investigating the motivational variables behind
actions (Schröder, 2003). It is interesting to examine what foods people buy and consume
and the reasons for their choices.
Food advertising always promises utility or gains or loss prevention to consumers.
This form of advertisement supports the traditional economic theory postulating that humans
choose according to their values, and the decisions people make are always dependent
maximizing utility or profit. Thus, if a person must choose between two options that will
both provide utility, he or she will choose the more profitable option (Simon, 1959). This
theory also claims that decisions are made rationality. This theory, and the rational actor
model has been important in not only economics but also the biological modeling of animal
behavior (Gintis, 2007; L. Real & Caraco, 1986; L. A. Real, 1991). The beliefs, preferences,
and constraints (BPC) model is an integrative approach used to unify the behavioral
sciences. (Gintis, 2007).
When the outcome of a decision does not take place at the same time when the
decision is made it is known as intertemporal choice (Frederick, Loewenstein, &
O'donoghue, 2002). If benefits are immediate, individuals choose wisely but often make
poor choices when the greater benefit is to be acquired in the long run. This situation occcurs

FOOD STIMULI AND GENDER

4

when people make poor food choices because pleasure in the moment outweighs health
benefits in the future (Gintis, 2009). For example, a person might have pleasures and health
on his or her utility curve preferences. The preferences may or may not coincide with healthy
choices. If this person places a higher value on health because of health beliefs or
knowledge, he or she would be able to make a choice that may be incompatible with his or

health

Pleasure

health

Time

Pleasure

Value

Value

her preferences and pleasure in exchange for a long-term reward.

Time
Choice

Figure 1: Weighing Desirability in an Intertemporal Choice Setting

The pleasure experienced when food is consumed is not constant and could be
dependent on time of day and the context in which it is consumed (Schröder, 2003).
Emotions also play a major role in the decision-making process. A somatic-marker
hypothesis proposes that emotions are represented as body states and are needed for
reasoning generating positive or negative somatic states. Rewards generate positive somatic

FOOD STIMULI AND GENDER

5

states and punishment generates negative somatic states. Thus, these somatic states give
signals to inhibit or enable a specific response (Bechara, Damasio, & Damasio, 2000;
Spence, 1995). Evoked emotions have been used to predict food choices, along with food
liking scales (Dalenberg et al., 2014).
It has been suggested that the desirability of each available choice is computed by
neurons, supporting the previously described economic theory. The mechanism through
which the decision is made is as simple as performing the more desirable action. It has been
proposed that neural architecture might compute and portray in a physiological way, as
utility theory proposes (Glimcher, Dorris, & Bayer, 2005). The algorithms through which
these circuits calculate the economic variables for which physiologically expected utilities
are consequent are an area of intense research (Glimcher et al., 2005).
It should be recognized that maximization of utility in economic theory is not
necessarily consciously achieved (Gintis, 2007; Jaynes, 2003). In the past several years,
research has pointed out that many decisions are made without conscious awareness and
even emotions can be unconscious (Bargh, 2002; Berridge & Winkielman, 2003; Bornstein
& Pittman, 1992; Damasio & Sutherland, 1994; Winkielman & Berridge, 2004).
Dhar and Gorlin (2013) propose “that certain choice effects arise mainly from
intuitive processing and require little deliberation, whereas others can be attributed
primarily to deliberate thought and effortful comparisons among options” (Dhar & Gorlin,
2013). This intuitive processing is an unconscious and automatic process that is more prone
to be activated by visual and sensory attributes generating an intuitive preference (Dhar &
Gorlin, 2013).

FOOD STIMULI AND GENDER

6

Eye-tracking technology is one of the methods that open the door to investigate what
people unconsciously fixate on (Gofman, Moskowitz, Fyrbjork, Moskowitz, & Mets, 2009)
and the relationship with preferences and food choices.

2.1.1 Food Choice and Human Behavior
A study performed in the European Union found gender differences. Women
reported that they were more likely to choose food because of “quality or freshness, price,
trying to eat healthy, and family preferences”(Lennernäs et al., 1997). Men were more likely
to choose food because of habit and taste (Lennernäs et al., 1997).
Table 1: Major Determinants of Food Choice

Economics

Psychology
Major Determinants
of Food Choice

Sociology

Biology

•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•

Income
Time
Skills
Availability
Knowledge
Personality
Attitudes
Beliefs
Emotions
Mood
Memory and learning
Conditioning
Norms
Lifestyle
Status
Social class
Hunger and appetite
Palatability
Taste
Genetics

Broadly based on (Schröder, 2003)

Food choices are influenced by many aspects and are very complex, as many factors
also interrelate (see Table 1). After hunger, personal preferences and taste are probably the

FOOD STIMULI AND GENDER

7

main reasons for food choices. However, these choices can be influenced by social aspects,
gender, education, and advertising (Schröder, 2003).
External factors such as temperature, visual cues, and smell also affect hunger and
satiety (Kraly & Blass, 1976; Logue, 2014) and are largely involved in food choice. Imram
(1999) nicely describes the relationship of visual cues and preferences as he states: “First
taste is almost always with the eye” (Imram, 1999). In Section 2.1.3, an overview of the
process involved between visual stimuli and behavioral response is provided.
Wądołowska et al. (2008) have summarized an analysis of Poles’ food choices and
the interaction between food preferences, food intake and sociodemographic features in the
following figure.

Figure 2: Food Choice Factors and Relationships (Wądołowska et al., 2008)

This figure illustrates how food preferences are influenced by many factors that
finally lead to food choices and dietary intake.
As previously mentioned, gender differences have been reported in relation to food
choices. In the following section, some known gender differences in eating behaviors are
described.

FOOD STIMULI AND GENDER

8

2.1.2 Eating Behavior and Gender
Gender differences related to food behavior are well documented, and the underlying
neurobiology is an area of investigation (Del Parigi et al., 2002). The size and morphology
of some areas of the human brain have presented sex differences; hormones produced by the
gonads at the development stage influence behavior patterns to a great extent. These
hormones cause mostly irreversible sex differences in central nervous system function
(MacLusky & Naftolin, 1981). Some gender differences and hemispheric asymmetries were
reported (Gur et al., 1995). These observations support the assumption that emotional and
cognitive gender differences could have a biological foundation (Del Parigi et al., 2002).
Gender-based anatomical differences of the brain should not be underestimated, as all lobes
of the brain display gender differences that should not be attributed solely to hormones
(Cahill, 2006).

Figure 3: “Sex Differences in the Size of Various Human Brain Regions”

(Cahill, 2006)

FOOD STIMULI AND GENDER

9

Gender differences in responses to visual food stimuli have been reported in some
neuroimaging studies. An activation of energy intake regulation areas of the brain has been
detected in both men and women. Neuronal responses to food stimuli measured by fMRIs
found a difference in women compared to men. Women showed stronger parietal (attention)
and prefrontal (cognitive processing) responses to food stimuli and seemed to be more
sensitive to food intake measured by post-meal satiety ratings. (Cornier, Salzberg, Endly,
Bessesen, & Tregellas, 2010; Uher, Treasure, Heining, Brammer, & Campbell, 2006)
A study presenting food stimuli of different tastes and measuring brain activity in
hungry and sated state showed that women had a lower variation in brain activation going
from hunger to satiation than mane and higher brain activation at a satiated or hungry state
(Haase, Green, & Murphy, 2011). Another study presented high-calorie and low-calorie
food stimuli compared to non-food stimuli. This study also showed gender differences in
brain activation in a satiated state (Frank et al., 2010)
Based on the reported gender differences in the response to food stimuli, the first
hypothesis for the present study is as follows:
H1: Attention to food stimuli differs between men and women.
A stronger preference for hot and spicy food has been found in men; a correlation
has been found between blood testosterone levels and hot food (Bègue, Bricout, Boudesseul,
Shankland, & Duke, 2015). A study on comfort food preferences found that: “Males
preferred warm, hearty, meal-related comfort foods (such as steak, casseroles, and soup),
while females instead preferred comfort foods that were more snack related (such as
chocolate and ice cream)” (Wansink, Cheney, & Chan, 2003). Other studies on dietary
intake support this gender difference (Bugge & Lavik, 2012; Hiroyasu, Chigusa, Hiroyuki,
Takesumi, & Akiko, 2005).

FOOD STIMULI AND GENDER

10

Based on these findings, two hypotheses are proposed and are as follows:
H1a: Men show an increased attentional bias for savory food cues compared
to women.
H1b: Women show an increased attentional bias toward high-density sweet
foods compared to men.
A review of the literature illustrates notable differences in food choices between men
and women (Garriguet, 2009; Liebman et al., 2006; Westenhoefer, 2005). Women generally
consume more vegetables and fruits and less fat and salt; women have indicated greater
health beliefs and health consciousness (Fagerli & Wandel, 1999; Leblanc, Begin, Corneau,
Dodin, & Lemieux, 2015; Wardle et al., 2004; Westenhoefer, 2005). It has been described
that, in general, women have a healthier diet than men (Garriguet, 2009; Liebman et al.,
2006) and that men consume more foods with higher sugar, fat, and salt content (Bugge &
Lavik, 2012). Research has also shown that men and women’s eating styles are different
(Rolls, Fedoroff, & Guthrie, 1991). Men consume more calories than women and eat at
higher speeds, taking more food at a time (Rolls et al., 1991).
Based on reports that women show greater consumption of fruit and vegetables, the
following hypothesis was derived:
H2: Women show an increased attentional bias toward fruit and vegetables
compared to men.
It is assumed that women are usually more concerned about weight control and have
stronger health beliefs (Wardle et al., 2004; Westenhoefer, 2005). Women are generally
more displeased with their weight and shape than men (Rolls et al., 1991). It has been
reported that weight control and health beliefs as motivational factors could make up as
much as 50% of gender differences in eating behavior (Wardle et al., 2004; Westenhoefer,

FOOD STIMULI AND GENDER

11

2005). Even young children have shown concern about weight control affecting their food
choices (Westenhoefer, 2002). However, studies indicate that gender differences in food
choices and intake appear for the first time in adolescence (Rolls et al., 1991). A Czech study
on adolescents showed a gender difference in food preferences. Boys had less healthy food
preferences than girls. Nevertheless, comparing preferences with actual consumption,
preferences were reflected only in part (Fiala, Bienertová-Vasšku, Brazdová, Švancara, &
Kukla, 2015).
Dietary restraint is defined as “intentional efforts to achieve or maintain a desired
weight through reduced caloric intake” (Stice, Ozer, & Kees, 1997). A correlation has been
found between dietary restraint and lower body mass index (BMI), suggesting that dietary
restraint could contribute to successful weight management (Rideout & Barr, 2009). Dietary
restraint has also been negatively correlated with energy and fat intake (De Castro, 1995;
Lindroos et al., 1997). Dietary restraint is not necessarily equivalent to dieting. However,
more research is needed to clarify this distinction (Rideout & Barr, 2009). Women have
presented higher scores in dietary restraint than men (Provencher, Drapeau, Tremblay,
Després, & Lemieux, 2003).
Other factors that could explain gender differences in consumption and eating
behaviors are health beliefs and attitudes (Emanuel, McCully, Gallagher, & Updegraff,
2012) and nutritional knowledge (Baker & Wardle, 2003).

2.1.3 Stimuli and Behavioral Response
The previous section mentioned external factors that affect hunger. A stimulusresponse model is proposed that includes different steps before a behavioral response takes
place. The first step in the stimulus-response model is an external stimulus such as a visual

FOOD STIMULI AND GENDER

12

cue. The stimulus is recognized by a sense organ such as the eyes and processed, giving rise
to the formation of sensations. The sensations generate a perception that leads to a behavioral
response (see Figure 2) (Schröder, 2003). In some cases, an environmental stimulus may be
incomplete; some information can be completed by memory if necessary. This process can
also occur in an inverted way when no external factors are present. Individual can experience
hunger that leads them to think about a particular food, and this situation could generate a
visual perception of food (Schröder, 2003) without a visual stimulus affecting the eyes.

Stimulus
(proximal**

Attention

Memory

Sensation

Mental
processing

Stimulus
(distal*)

Perception

Response

*Object
**Image of object in the eye
Figure 4: Stimulus Response Model for Vision (Schrö der, 2003)

Most stimulus perception occurs out of awareness, and an entire decision-making
process could take place unconsciously (Fitzsimons et al., 2002). Before a visual stimulus
can be processed, it must go through the visual pathway. In the next section, general insight
about the visual pathway is provided.

FOOD STIMULI AND GENDER

13

2.2 Eye Movements, Attention, and Preference
A brief description of the human eye and its movements is offered for a better
understanding of the eye-tracking technology used in the present study.

2.2.1 The Human Eye
The human eyeball is divided into three layers:
1. An outer layer called the fibrous tunic consists of the sclera (the “white” part)
and cornea (transparent and curved).
2.

A middle layer named the vascular tunic is composed of the ciliary body,
choroid, and iris. The iris is a colored circle with a hole in the center, known as
the pupil, where light enters the eyeball (see figure 4).

3. The inner layer known as the retina gives the starting point of the visual path and
is divided into the pigmented and neural layers. The neural layer consists of three
layers: the photoreceptor, bipolar cell, and ganglion cell layers (Tortora &
Derrickson, 2010).

FOOD STIMULI AND GENDER

14

Figure 5: The Human Eye (Clker-Free-Vector-Images, 2016)

The photoreceptors located on the neural layer of the retina are specialized cells
called cones and rods. These two kinds of photoreceptors are responsible of the beginning
of the process of converting light into nerve impulses. Rods are accountable for shades of
gray and cones give rise to highly acute color vision. Cones are more concentrated in the
fovea centralis that is located in the middle of the macula lutea (also known as yellow spot)
and is the area of the highest sharpness of vision. To view an image it must be placed on the
fovea. This positioning of an image on the fovea leads to eye and head movements (Tortora
& Derrickson, 2010).
2.2.2 Eye movements
Three pairs of muscles are responsible for vertical, horizontal, and torsional eye
movements. The orientation of the eye decides the direction of the “look” (attention). This
orientation of the eyes is decided by the brain, which directs muscles toward relevant
locations. Most common types of eye movements are the following (Holmqvist et al., 2011).

FOOD STIMULI AND GENDER

•

15

Fixation is a temporary cease of major eye movements, though micromovements still occur . Fixations are the most reported measure used in eye
tracking and are usually seen as “attention.”

•

Saccade is movement of the eye from one fixation to another; it is the fastest
movement that the body can produce. Saccades are also an often-reported
measure in eye tracking.

•

Glissade is a post-saccadic movement.

•

Smooth pursuit occurs when the eye follows an object in motion.

2.2.3 Attention and preferences
Attention, or where people look is “…normally defined as a selectivity in the
perception” (Orquin & Mueller Loose, 2013). Attention chooses which part of the visual
input is put into the fovea (Tortora & Derrickson, 2010). Certain cases may occur in which
a visual scene is familiar and the stimulus is stored in the memory. In such cases, the subject
can choose whether to view the stimulus or not; this choice happens even before the stimulus
is perceived. Thus, attention is not passive acquisition of information. It is proposed that the
eye movements and attention moves may be driven by similar internal mechanisms
(Rizzolatti, Riggio, Dascola, & Umiltá, 1987).
Eye movements “shape decisions by gatekeeping information in the decision
process” (Orquin & Mueller Loose, 2013). Evidence suggests that eye movements cannot
only account for image features or sensory information such as luminance and contraste
(Treue, 2003). Eye movements can also account for cognitive factors or attentional
influences, such as knowledge, intentions, and expectations (Henderson & Ferreira, 2004;
Treue, 2003). In the literature, top-down and button-up mechanisms are generally seen as

FOOD STIMULI AND GENDER

16

complementary, and the interaction between these two mechanisms has been discussed
(Einhäuser, Kruse, Hoffmann, & König, 2006). Attention may also play an active a role in
preference formation and be used as a link to conscious choice (Shimojo, Simion, Shimojo,
& Scheier, 2003). This effect has been validated in decision trials by dynamically controlling
gaze duration (Glaholt, Wu, & Reingold, 2010).
Eye movement measurement, such as eye-tracking technology, is seen as a direct
method to measure visual attention and attentional bias related to food stimuli. Eye
movements are recorded as participants are exposed to visual stimuli (Doolan, Breslin,
Hanna, & Gallagher, 2015; Orquin & Mueller Loose, 2013). One main measurement is the
duration of total fixation (the total time a participant fixates on an area of interest) (Tobii
AB Technology, 2015). This measure has been seen as being influenced by decision-making
goals and preference formation (Orquin & Mueller Loose, 2013; Shimojo et al., 2003; van
der Laan, Hooge, de Ridder, Viergever, & Smeets, 2015). Though some evidence suggests
that fixation has an effect on the process of decision making, this correlation should be made
with care (Krajbich, Armel, & Rangel, 2010). Eye tracking has been proven to be successful
in studying how consumers perceive the quality of foods. The eye-tracking technique also
promises more insights into human behavior, allowing the study of underlying components
of attention and decision making (Mitterer-Daltoé, Queiroz, Fiszman, & Varela, 2014).
In the literature, two mechanisms are described as being responsible for stimulus
attention: \ automatic and controlled processes. The automatic process is a fast and
unconscious process. The controlled process is slower than the automatic process and not
necessarily conscious (Leven & Leven, 1991). Some separate these two mechanisms, also
seen as initial and maintained attention, in terms of time. Some propose that the first 80-100
ms are initial attention, but others see initial attention as time up to 500 ms (Doolan et al.,

FOOD STIMULI AND GENDER

17

2015; Leven & Leven, 1991). No consensus is found in the literature concerning time for
initial attention and maintained attention (Doolan et al., 2015), and some argue that it is
impossible to distinguish these two mechanisms (Leven & Leven, 1991).
The present study used eye-tracking technology to measure eye movements and
fixations in timing when stimulus attention is still controlled by an automatic process and
not driven as much by a conscious process. The goal was to measure intuitive eye
movements. As previously mentioned, it has been reported that women have a healthier food
choice pattern than men (Garriguet, 2009; Liebman et al., 2006; Westenhoefer, 2005). In
this study the goal was to investigate whether this healthier food choice pattern could be
reflected as attentional bias. For this reason a comparison between healthy and unhealthy
food stimuli was chosen.
2.3 Healthy and Unhealthy
It is very difficult to state that a certain food is healthy or not and find guidelines to
define whether a certain food is healthy. Nevertheless, fast food is thought to be unhealthy,
as it has been linked with obesity; it has been proposed that foods with high sugar, salt, and
fat content are unhealthy (Bugge & Lavik, 2012; James, 1990). Due to the nutritional profile,
these mentioned foods can be categorized as foods with high energy density.
Energy density is usually measured in calories per gram and is defined as the quantity
of energy per food weight. Research has pointed out that lowering the intake of energy
density could be useful to sustain satiety and lower energy consumption (Rolls, 2009). It is
proposed that foods with high energy density constitute a challenge to the systems that
control human appetite with circumstances for which these systems were not intended
(Prentice & Jebb, 2003), promoting weight gain and obesity. This physiological failure can
lead to “passive over-consumption” (Prentice & Jebb, 2003). Prentice and Jebb (2003) state

FOOD STIMULI AND GENDER

18

that humans exhibit a “weak innate ability to recognize foods with a high energy density and
to appropriately down-regulate the bulk of food eaten to maintain energy homeostasis”
(Prentice & Jebb, 2003). In an experiment, the main variable that influenced meal size was
not level of hunger but the nutrient content of the assortment of foods consumed (Lawton,
Burley, Wales, & Blundell, 1993). It is suggested that individuals have a tendency eat the
same weight of food regardless of the diet composition (Bell, Roe, & Rolls, 2003; Rolls,
2009). Supposedly stretch receptors located throughout the stomach induce a sensation of
satiety (Starkebaum & Gerber, 2004).

Figure 6: Caloric Density (Hever, 2012)

It is important to note that even though high-density foods appear to be less healthy
as those with a lower caloric density profile, consumption of these foods in moderate
amounts and as part of a balanced diet would not constitute a problem in most cases. This
information suggests that high-density foods themselves are not the problem; eating habits
play a major role. Nevertheless, the challenge presented by these high calorie foods should
not be underestimated, since they could be undermining humans’ appetite control systems.
In the current study, the categorization of foods as healthy or unhealthy did seem to
be the most appropriate, as no clear guidelines were found in the literature. Therefore, the
visual food stimuli were categorized according to caloric density. Food stimuli were

FOOD STIMULI AND GENDER

19

categorized as high and low calorie, without disregarding the fact that low-calorie foods are
mostly referred to as healthier choices (Westenhoefer, 2005).

FOOD STIMULI AND GENDER

3

20

Experimental Design

3.1 Stimuli
A total of 60 food images were used as visual stimuli in this study. These images
were selected from the FoodPics and FoodCast research image databases that offer images
with associated normative data (Blechert, Meule, Busch, & Ohla, 2014; Foroni, Pergola,
Argiris, & Rumiati, 2013). These databases have been created for eating behavior research
(Blechert et al., 2014; Foroni et al., 2013) (Also see www.food-pics.sbg.ac.at and
https://foodcast.sissa.it/neuroscience/).
To study whether an attentional gender bias exists concerning high- and low-calorie
foods (H2), the main criteria for image selection was caloric density (kcal/100g). The
selected pictures were divided into two main groups. Thirty pictures were categorized as
high calorie food pictures and 30 were categorized as low calorie food pictures based on
calorie density. Other image properties were also considered in order to avoid attentional
bias between high-calorie and low-calorie groups due to these image properties. As color
can also influence visual behavior (Jantathai, Danner, Joechl, & Dürrschmid, 2013), one
goal was to have a variety of colors in both image groups. Other factors taken into account
were complexity, brightness, contrast, valence, craving, and arousal. These picture attributes
were offered by the databases. For this experiment, the given values were tested statistically
between the two image groups to avoid a significant difference between both. The
complexity was quantified by computing the proportion of outline-related pixels within the
image using the Canny edge detection algorithm, and arousal was rated using a visual analog
scale (Blechert et al., 2014). The main picture attribute computations are presented in Table
2; for more details on other attributes, refer to Appendix D.

FOOD STIMULI AND GENDER

21

Table 2: Comparison of Food Stimuli Categories

High Calorie (n)
Variable

Sweet

Kcal/100g

Not Sweet

Low Calorie (n)
Ready to
Eat

30
15

Not Ready
to Eat

Arousal

15

30
15

9

82,00 n.s.

105,00 n.s.

15

29
12

p

416,00 n.s.

30

21
Complexity

Z

0,00 p <
.001
59,00 p < .05

30

21

Mann-Whitney
U Test

9

57,00 n.s.
302,00 n.s.

27

86,00 n.s.

15
20

9

71,00 n.s.

Note: The numbers in each category are n = number of pictures.

Many of the images of low calorie foods had little or no processing. For example,
some were not peeled, washed, or cooked. The research team felt that it was improper to
compare low-calorie foods that were not ready to eat with high-calorie ready-to-eat foods.
Amount of processing has not been taken into account in previous studies (Blechert, Klackl,
Miedl, & Wilhelm, 2016; Meule, Kübler, & Blechert, 2013); for this reason it seemed
interesting to investigate whether the amount of processing of low-calorie foods could have
an effect. The low-calorie category was divided into ready-to-eat and not ready-to-eat
categories. The foods categorized as ready to eat did not require additional processing to eat
and included pre-cut or pre-peeled fruits and vegetables. In the not ready-to-eat category,
the foods chosen required processing in order to be eaten. For a description of the food
stimuli chosen, refer to Table 3 below. To test whether a gendered attentional bias existed

FOOD STIMULI AND GENDER

22

concerning sweet and savory foods (H1a and H1b), the high-calorie category was subdivided
into two groups: sweet and non-sweet.

Figure 7: Trial Sample

The selected pictures were presented in pairs and placed to the right and left near the
center of the screen because of the known increased attention at the center (Leven & Leven,
1991). Placement of high- and low-calorie pictures on right and left was randomized to avoid
special location bias (Glaholt et al., 2010).
The sequence of the food stimuli was randomized for each participant. A black
fixation cross was shown for 1000 ms before each picture pair to draw attention to the
middle, to avoid any initial fixation on either one of the subsequent images and provide a
better overview of the calibration of glasses during the eye-tracking presentation
(Castellanos et al., 2009). Each picture pair was shown for 3000 ms (Meule et al., 2013) (see
Figure 5).

FOOD STIMULI AND GENDER

23

Table 3: Food Stimuli used for the experiment

High Calorie Not Sweet:

High Calorie Sweet:

Hamburger with bacon
Cheese and cold meat platter
French fries
Chips
Pizza
Doner kebab
Pasta bake
Salami sausage
Fish sticks
Wiener schnitzel
Fried sausage with roll
Camembert
Lasagna
Nachos
Sausages
Fast Food Menu:
French fries, hamburger, soft drink, ice cream
Low Calorie Ready:
Watermelon
Salad plate
Raspberries
Cucumber and carrot
Oranges
Blueberries
Radishes
Tomatoes
Fruit salad
Beans and carrots, cooked
Grapes, white
Cherries
Pear
Yellow bell pepper, sliced
Brussels sprouts
Mixed vegetables
Peaches
Strawberries
Pieces of fruit
Tomato and mozzarella skewer
Apricot

Ice cream with chocolate beans
Donut with chocolate sprinkles
Chocolate popsicles
Muffins
Raspberry cake
Bar of chocolate with nuts
Sundae (peach)
Cake
Chocolate muffin
Gummy candy (gold bears)
Mini chocolate cake bar
Mini chocolate marshmallows
Cookies filled with chocolate
Colored chocolate beans
Waffle

Low Calorie Not Ready:
Paprika peppers
Cucumbers
Red cabbage
Corn (on a cob)
Banana
Orange
Carrots
Pineapple
Lettuce

The timing chosen for the trials was short in order to favor an attentional bias based
on an uncounsous process of intuitive choice due to pre-existing preferences, (Dhar &
Gorlin, 2013). Measuring errors are common in software timing, and it is important to
choose suitable software (Garaizar, Vadillo, López-de-Ipiña, & Matute, 2014). PsychoPy

FOOD STIMULI AND GENDER

24

software was selected for this experiment. The builder interface was used for this experiment
to facilitate easier usage (Peirce, 2008).

FOOD STIMULI AND GENDER

25

3.2 Data collection
The experiment was conducted in a laboratory setting at Hohenheim University in
the Department of Nutrition Psychology and lasted approximately 30 minutes for each
participant.
3.2.1 Eye Tracking
Tobii Pro Glasses 2 were used to capture subjects’ eye movements and fixation
points. Tobii Pro Glasses 2 are a wireless eye tracker device with a live view software
function. The eye tracker has four cameras, two for each eye. A gaze sampling of 50 Hz was
utilized; this is a sampling interval of 20 ms. This particular eye tracker uses pupil corneal
reflection technology, which creates a reflection of light on the cornea. The eye moves, but
the reflection remains stationary, and eye movements can be recorded. For this purpose,
Tobii Pro Glasses 2 have a set of illuminators that project infrared light and create
reflections. This eye tracker can choose between bright pupil (pupil appears brighter than
the iris) and dark pupil (pupil appears black or darker than the iris) tracking; depending on
the position of the light source one or the other is used (Holmqvist et al., 2011; Tobii AB
Technology, 2015).
A calibration of each participant is required, since the fovea in not in the same
position in every person and is not necessarily at the center. A one-point calibration and
three-point validation were performed on each participant before starting each recording
using the calibration card and instructions provided by Tobii AB Technology (Holmqvist et
al., 2011; Tobii AB Technology, 2015). Through the three-point validation of the
calibration, a maximum of 1.2 grade accuracy error was admitted, which was adequate for
this experiment due to the large size of the areas of interest.

FOOD STIMULI AND GENDER

26

3.2.2 Questionnaires
Two questionnaires were developed from different questionnaires. One was
distributed at the beginning of the experiment, and the second was distributed after the food
picture presentation. Descriptions of the questionnaires and the order in which they were
completed are listed below; the questions can be found in Appendix E.
Questionnaire I (completed before the food picture presentation):
•

Demographic questions included age, gender, weight, height, affiliation to
the university, and study course.

•

Participants were asked to indicate how hungry they were and their appetite
at the moment on the 10 cm analog Hunger and Appetite Scales (0 = not
hungry / no appetite 10 = very hungry / large appetite). Participants were also
required to specify the time of their last meal. For the analysis, answers were
measured on the scales with a ruler.

•

The authorized German translation of the Dutch Eating Behavior
Questionnaire (DEBQ) (Fragebogen zum Ernährungsverhalten or FEV I)
was used for assessment of eating behavior. This questionnaire contains 33
questions from which 13 on emotional eating, 10 are on restrained eating,
and 10 on external eating behavior (Grunert, 1989; van Strien, Frijters,
Bergers, & Defares, 1986).
In this master’s thesis, the DEBQ questionnaire was analyzed to investigate
whether previous reports (Provencher et al., 2003) on gender differences in
restrictiveness were also present in our study sample. The sample of our study
was composed only of students and university staff, which is not equal to the
sample used to validate the DEBQ questionnaire. An instrument for

FOOD STIMULI AND GENDER

27

measuring a latent variable is generally seen as reliable if the Cronbach's α is
greater than 0.7 (Breakwell, Hammond, & Fife-Schaw, 2006). The four latent
variables of the Cronbach's α were calculated, with each of the corresponding
items having a uniform procedure, as with the attitudes toward healthy eating
questionnaire (Einstellungen zu Gesunde Ernährung or EGE).
Questionnaire II (completed after the eye-tracking presentation):
•

The above-described Hunger and Appetite Scales were included again.

•

A German questionnaire on attitudes toward healthy eating (Einstellungen zu
Gesunde Ernährung or EGE) was included with four-point categorical
answer options ranging from “does not apply” to “applies” (Diehl, 1980).
The German EGE questionnaire consisted of 30 items that measured four
latent variables: attitudes on the effectiveness of a healthy diet, appreciation
of a healthy diet, practicing a healthy diet, and consumption of a healthy,
low-fat diet. The questionnaire was analyzed for correlations between the
variables from the questionnaire with the eye-tracking data. An instrument
for measuring a latent variable is generally seen as reliable if the Cronbach's
α is greater than 0.7 (Breakwell, Hammond, & Fife-Schaw, 2006). To
examine the reliability (internal consistency) of the questionnaire, the four
latent variables of the Cronbach's α were calculated with each of the
corresponding items.

•

A shortened form of the German Food Frequency Questionnaire (FFQ) was
adapted for the food pictures in the present study (Haftenberger et al., 2010;
Truthmann, Mensink, & Richter, 2011). This questionnaire was developed to
assess the food consumption of adults in Germany between 2008-2011

FOOD STIMULI AND GENDER

28

(Gößwald, Lange, Kamtsiuris, & Kurth, 2012). The questionnaire assesses
consumption over the past four weeks by asking frequency with five-point
categorical answer options (never, 1-3 times per month, 1-3 times per week,
4-6 times per week, at least once every day). In the present thesis, this
questionnaire was used to identify whether self-reported consumption
correlated correlation with the eye-tracking data.
•

A Food Liking Scale (FLS) was included with five-point categorical answer
options ranging between “I do not like it at all” to “I like it very much”
(Peryam, 1998). This questionnaire was adapted to the food picture
categories in order to investigate whether foods with an attentional bias were
also reported as pre-existing preferences.

FOOD STIMULI AND GENDER

29

3.3 Subjects
Seventy-three students and staff ages 18-35 were recruited from the Universities of
Hohenheim and Stuttgart to participate in the experiment. The recruitment was completed
by placing poster advertisements in different locations in both universities, such as libraries
and scientific institutes. Flyers were distributed at the end of lectures, on cafeteria tables,
and to individuals. Participation in a prize draw of five €20 gift certificates was offered (see
Appendix B for flyer and poster details). Individuals made an appointment through a Doodle
link and were contacted via e-mail to ensure that the inclusion criteria were met and none of
the exclusion criteria were met (see Table 4).
Table 4: Inclusion and Exclusion Criteria

Inclusion Criteria:
Age 18-35
Good German language
Student or scientific worker

Exclusion Criteria:
Eye illnesses
Corrective glasses
Photosensitive epilepsy

Though participants with eye illness or corrective glasses were excluded, participants
with contact lenses were allowed to take part in the experiment.

FOOD STIMULI AND GENDER

30

3.4 Experimental Procedure
A brief description of the experimental procedure is presented to provide readers
with an overview of how the eye-tracking experiment was accomplished. The experiment
consisted of a single session. Preparation for each experimental session included checking
the projector settings and running the stimuli presentation once before the participant
arrived. Participants were asked to abstain from eating for three hours before visiting the
laboratory to avoid a potential effect of satiation and facilitate a better reaction time in a
hungry state (Doolan et al., 2015; Nijs, Muris, Euser, & Franken, 2010).
Only one participant at a time was allowed in the laboratory for the eye-tracking
presentation. After a participant was welcomed, he or she was asked to turn off his or her
mobile device, read the participant information, and read and sign the consent information
(see Appendix C). Subsequently, he or she was asked to fill out Questionnaire I. For a
detailed description of the questionnaires, please refer to the Data Collection section and
Appendix E. The Tobii Pro 2 glasses were adjusted to the participant’s facial features, and
a calibration of the glasses was performed at a distance of 1 m from the screen using the
calibration target card provided by Tobii Technology. The accuracy error admitted by the
calibration was of less than 0.6 grades. The initial calibration was then validated by a threepoint verification admitting less than 1.2 grades of accuracy error.
After successful calibration, participants were shown the randomized food stimulus
picture presentation without any prior instructions to avoid any induced bias in the viewing
behavior. It is important to evade priming that could influence certain unconscious behaviors
(Aarts, Custers, & Marien, 2008).
After the presentation, participants finished the experiment by completing
Questionnaire II. See the description of the questionnaire in the Data Collection section and

FOOD STIMULI AND GENDER

31

Appendix E. A piece of fruit and a pen with the university logo were offered to thank
participants. The experimenter was always present in the laboratory; a second person was
monitoring in the technical room and was responsible for starting the stimuli presentation
and securing the data.

FOOD STIMULI AND GENDER

32

3.5 Data Analysis
3.5.1 Areas of Interest: Defining and Position
The accuracy of the eye tracker is the average difference between gaze position and
the position recorded by the eye tracker (Holmqvist et al., 2011). For this experiment,
accuracy was verified by calibrating before each recording with the calibration card provided
by Tobii Technology. Due to the accuracy error of the eye-tracking glasses, drawing areas
of interests (AOIs) too close to the edges of the image would allow a loss of fixation and a
margin drawn too large would allow fixation counts, which are outside of the stimuli in
reality. Consequently, the accuracy of the eye tracker is important when deciding the size of
an AOI (Holmqvist et al., 2011).

Figure 8: AOI around Food Stimuli

Areas of interests (AOIs) for this experiment were defined with a margin of 0.9
degrees for all food stimuli (see Figure 7). The margin was defined based on the calibration
error admitted, with 0.9 being the mean between 0.6 and 1.2 degrees (see Chapter 3.4). Thus,
AOIs were drawn at the edges of each food without leaving any margin. Each AOI was
carefully drawn once for each food stimuli and then used for all participants via the copy
and paste function of the Tobii Pro Glasses Analyzer Software, assuring identical AOIs for
all participants.

FOOD STIMULI AND GENDER

33

3.5.2 Coding Procedure and Validation
The eye-tracking video-based recordings must be encoded or mapped into snapshots
to obtain the coordinates of the eye movements necessary for data analysis.
Snapshots
Snapshots were taken of each slide of each randomized presentation using a Canon
PowerShot G9 X camera. A total of 2,263 snapshots were taken, all with the same manual
settings: ISO-160, exposure 1/60, focal distance 10 mm, and no flash.
Encoding
The encoding procedure was performed using the Tobii Pro Glasses Analyzer
Version1.36.1430 (x64) Software. The auto-mapping tool of the Tobii Pro Glasses Analyzer
Software was used when possible; when this function did not work, the points were encoded
manually.
Four people were assigned to this task. To ensure a high reliability between all
coders, three training sessions were conducted, and an encoding protocol was implemented
(see Appendix F). The four people performed a total of three reliability tests of the ET
recording until a match of greater than 90 % was achieved with deviations of less than 1%
(see Appendix G).
3.5.3 Data Preparation and Data Analysis
Once the coding of the raw eye-tracking data was accomplished, a velocity-threshold
identification (IV-T) filter was applied to all projects before the data was exported. A filter
is important in order to classify the fixations, “If the velocity of the eye movement is below
a certain threshold, the samples are classified as part of a fixation, and if the velocity is above
the threshold is classified as a saccade” (Tobii AB Technology, 2015). The I-VT filter

FOOD STIMULI AND GENDER

34

settings were adjusted based on previous studies to suit this study. These settings were also
verified with Tobii AB Technology. The IV-T filter settings used were as follows:
•

30o Velocity Threshold (Nijs et al., 2010)

•

100 ms for fixations (Doolan, Breslin, Hanna, Murphy, & Gallagher, 2014;
Mogg, Bradley, Field, & De Houwer, 2003; Nijs et al., 2010; Popien, Frayn,
von Ranson, & Sears, 2015; Werthmann et al., 2011)

•

20 ms window length

•

Noise reduction: off

•

Gap fill-in: off

•

Merge adjacent fixations: off

Four eye-tracking measurements were chosen for statistical analysis:
•

Total visit duration (TVD) is “the total time each participant has visited each
AOI on all Media, including averages, medians, sums, and the share of total
time spent in each AOI out of all AOIs” (Tobii AB Technology, 2015).

•

Visit count (VC) is “the number of visits within each AOI on all Media,
including averages, medians, and the percentage of participants that fixated
within each AOI at least once” (Tobii AB Technology, 2015).

•

Total fixation duration (TFD) is “the total time each participant has fixated
each AOI on all Media, including averages, median, sums, and the share of
total time spent in each AOI out of all AOIs” (Tobii AB Technology, 2015).

•

Fixation count (FC) is “the number of fixation within each AOI on all Media,
including averages, medians, sums, total number of fixations within the
Analysis Set, and the total Analysis Set and Recording Durations” (Tobii AB
Technology, 2015).

FOOD STIMULI AND GENDER

35

The eye-tracking data sheets output provided by the Tobii Pro Glasses Analyzer
Software was placed together on a matrix with the questionnaire data for each participant.
The matrix was created using Microsoft Excel Software 2016. This matrix contained all the
data collected in the experiment. All statistical analyisis was computed using the IBM SPSS
Statistics 23 Software.
Inferential analysis and significance tests were used to investigate whether the
observed gender differences were statistically significant. Since the study included two
independent samples of men and women, two tests could have been used to compare the two
groups, either the t-test for independent samples or the nonparametric Mann-Whitney U test.
For the t-test, a normal distribution and homogeneity of variances are required in both
groups. Nevertheless, in the case of missing homogeneity of variances the t-test could be
used if a Welch correction is applied. If a normal distribution is not provided, the MannWhitney U test should be used, since for the Mann-Whitney U test the distribution of the
data is not a prerequisite.
A Shapiro-Wilk test was used to test the high and low calorie groups for the normal
distribution of each of the indices among men and women. The results are in the lines p SW
Men and p SW Women in Appendix I. The prerequisite of homogeneity of variance was
examined using the Levene test. A normal distribution and homogeneity of variance were
confirmed in all cases. For this reason, a two-sample t-test was always applied without any
Welch correction.
In the analysis of the food groups, the above-described t-test or Mann-Whitney U
test was used to test for statistical significance. In most cases, the p-value was above the
usual significance level of 0.05. Therefore most variables displayed normal distribution. The
p-value was above 0.05 for three variables (HCS_TVD, HCS_TVD, AND HCNS_FC); they

FOOD STIMULI AND GENDER

36

did not have a normal distribution. For these three variables, the non-parametric MannWhitney U test was used.
In the Levene test, none of the eye-tracking variables had a p-value of p = ˂ 0.05;
this result demonstrated homogeneity of variance. No Welch correction was applied when
the t-test was calculated. The values marked with an “A” are z-values; these are cases in
which a Mann-Whitney U test was performed.

FOOD STIMULI AND GENDER

37

4

Results

In this chapter, the participants are briefly described, and the results of the eyetracking data are presented. Statistical tests were conducted, as explained in Section 3.5.3,
to test the gender differences hypothesis established in Section 2.1.2. In Section 4.3, a
correlation analysis of the questionnaires and eye-tracking data is described. An analysis of
variables that may account for gender differences is also provided.
4.1 Participants
Of the 73 subjects who participated in the experiment, the data of only 50 was usable,
mainly because of the poor quality of the data. See Appendix H for detailed exclusion
criteria. Of the 50 eye-tracking recordings, 18 single trials had to be excluded from the
analysis due to AOI overlap and poor quality of data (see Appendix H).
Of the 50 participants included in the analysis, 24 subjects were male and 26 were
female. A sample description can be seen in Table 5 below.
Table 5: Sample Description Differentiated by Gender

Age in Years
BMI kg/cm²
Hours of Fasting

Men n = 24
Mean
SD
24
3
24.77
3.28
4.53
3.54

Note: n= number of participants

Gender
Women n = 26
Mean
SD
22
3
21.95
2.35
4.59
3.13

Total
Mean
23
23.31
4.57

SD
3
3.14
3.29

FOOD STIMULI AND GENDER

38

4.2 Eye-tracking Data
Four eye-tracking parameters were investigated for each food group: total visit
duration (TVD), visit count (VC), total fixation duration (TFD), and fixation count (FC).

4.2.1 Analysis of High and Low-Calorie Groups
For a comparison of gender differences in relation to calorie density, as stated in
hypotheses 1 and 2, the food stimuli presented in this study were divided into high- and lowcalorie groups (see Chapter 2.3). Furthermore, average indices were calculated; each of the
two food groups and the four eye-tracking variables values (TVD, VC, TFD, and FC) were
averaged. A total of eight indices resulted for each combination of the food groups and eye
tracking (ET) variables. The descriptive statistics for each of the indices were calculated
separately for men and women (mean, standard deviation, minimum, and maximum).
In the low-calorie food group, women had a significantly higher mean value (p <
0.05) than men for the ET variables TVD, TFD, and FC. In the high-calorie food group, men
had a significantly higher mean value (p < 0.05) for all of the ET variables. Boxplots of all
parameters are presented in Figure 7. In these boxplots, a gender-specific gaze bias can be
appreciated for all ET variables except for VC in the low-calorie food group.

FOOD STIMULI AND GENDER

Low Calorie

39

High Calorie

Figure 9: Boxplots of Low- and High-calorie ET Parameters Divided by Gender
Total fixation duration and total visit duration are measured in seconds; visit count and fixation
count are measured in counts within the areas of interest.

FOOD STIMULI AND GENDER

40

The results of the tests for group comparisons are illustrated in Table 6, which
presents the significance values. Gender differences were significant for all ET variables
except for VC and FC in the low-calorie group.

Table 6: Attention Bias for Men and Women by High- and Low-calorie Group

Stimuli
Low
Calorie

ET Variable
Total Visit
Duration
Visit Count
Total Fixation
Duration

High
Calorie

Fixation Count
Total Visit
Duration
Visit Count
Total Fixation
Duration
Fixation Count

Gender
M
W
M
W
M
W
M
W
M
W
M
W
M
W
M
W

N
24
26
24
26
24
26
24
26
24
26
24
26
24
26
24
26

M
1.13
1.28
1.59
1.56
1.06
1.19
2.83
3.10
1.21
1.08
1.71
1.53
1.14
1.02
3.17
2.86

SD
0.21
0.16
0.22
0.21
0.19
0.15
0.57
0.42
0.21
0.16
0.25
0.24
0.20
0.14
0.50
0.51

Min
0.58
1.01
1.24
1.18
0.57
0.95
1.67
2.16
0.74
0.65
1.17
1.17
0.72
0.63
2.10
1.63

Max
1.54
1.74
1.99
1.94
1.46
1.61
4.18
3.94
1.69
1.33
2.18
1.87
1.55
1.26
4.40
3.77

t

p

-2.70 0.01
0.46

0.65

-2.58 0.01
-1.93 0.06
2.41

0.02

2.60

0.01

2.54

0.01

2.17

0.03

Note: N = 50; total fixation duration and total visit duration are measured in seconds; visit count and fixation count are
measured in counts within the areas of interest; TVD = total visit duration, VC = visit count, TFD = total fixation duration,
FC = fixation count

The total fixation duration and total visit duration had significant gender difference
in this experiment with the low-calorie stimuli, and all eye-tracking variables had significant
differences with the high-calorie stimuli.

FOOD STIMULI AND GENDER

41

4.2.2 Analysis of Food Groups
The food stimuli were subdivided into four food groups to test H1a and H1b and
observe whether the level of processing had an influence on the low-calorie stimuli. The
low-calorie food images were subdivided into not ready to eat (lcnr) and ready to eat (lcr).
The high-calorie images were subdivided into sweet (hcs) and not sweet (hcns) (see Section
3.1). Furthermore, the four eye-tracking variables (TVD, VC, TFD, and FC) were
investigated for each food group (lenr, ler, hcs and hcns). To execute this examination, a
similar procedure to that described above was conducted. The average value indices were
calculated for each of the four food groups and the four ET variable values TVD, VC, TFD,
and FC . A total of 16 indices resulted for each combination of groups and ET variables.
Boxplots of the high-calorie foods are presented, dividing the food stimuli between
sweet and not sweet and taking gender into account (see Figure 9). A gender-specific bias
can be observed in that the ET variables illustrate the increased attention of men to sweet
and not sweet high-calorie foods.

FOOD STIMULI AND GENDER

High Calorie Sweet

42

High Calorie Not Sweet

Figure 9: Boxplots of High Calorie Groups; ET Parameters Divided by Gender
Total fixation duration and total visit duration are measured in seconds; visit count and fixation count are
measured in counts within the areas of interest.

The lcnr group comprised of nine food stimuli was created to observe whether the
level of processing could have an impact on the ET variables (see Section 3.1). Boxplots of
the low-calorie significant parameters are presented in Figure 10. A gender difference was

FOOD STIMULI AND GENDER

43

present only in the ready to eat group. This finding suggests that level of processing has an
influence on perception.

Low Calorie Ready to Eat

Low Calorie Not Ready to

Figure 10: Boxplots of Low Calorie Groups; ET Parameters Divided by Gender
Total fixation duration and total visit duration are measured in seconds.

From the boxplots for all groups (Figures 9 and 10) it is clear that women displayed
increased attention to low-calorie foods while men tended to prefer high-calorie foods. Tests
for statistical significance can be viewed in the following table.

FOOD STIMULI AND GENDER

44

Table 7: Attention Bias for Men and Women by Food Group

Stimuli
Low
Calorie
Not Ready
to Eat
(lncr)

ET Variable
Total Visit
Duration
Visit Count
Total Fixation
Duration
Fixation Count

Low
Calorie
Ready to
Eat
(lcr)

Total Visit
Duration
Visit Count

High
Calorie
Sweet
(hcs)

Total Visit
Duration

Total Fixation
Duration
Fixation Count

Visit Count
Total Fixation
Duration
Fixation Count

High
Total Visit
Calorie Not Duration
Sweet
(hcns)
Visit Count
Total Fixation
Duration
Fixation Count

t (z)

p

Gender
M
W
M
W
M
W
M
W
M
W
M
W

M
1.09
1.19
1.60
1.56
1.03
1.11
2.71
2.98
1.18
1.37
1.57
1.55

SD
0.26
0.20
0.27
0.23
0.24
0.19
0.61
0.47
0.21
0.19
0.27
0.26

Min
0.31
0.89
1.20
1.10
0.31
0.81
1.40
2.10
0.83
1.04
1.00
1.17

Max
1.61
1.65
2.20
2.10
1.51
1.57
4.30
3.80
1.62
1.95
2.20
2.00

M
W
M
W
M
W
M

1.10
1.28
2.95
3.22
1.21
1.07
1.68

0.18
0.18
0.59
0.54
0.24
0.19
0.24

0.82
0.94
1.89
2.22
0.66
0.47
1.20

W
M
W
M
W
M
W
M

1.51
1.15
1.01
2.97
2.73
1.21
1.10
1.75

0.25
0.23
0.17
0.53
0.54
0.22
0.19
0.30

1.00
0.65
0.46
1.87
1.40
0.71
0.64
1.07

W
M
W
M
W

1.55
1.13
1.02
3.38
2.99

0.26
0.20
0.16
0.59
0.57

1.13
0.65
0.61
2.33
1.86

1.46
-3.37 0.00
1.80
4.06
-1.69 0.10
4.39
1.67
2.27A 0.02
1.50
2.14
2.46 0.02
1.93
1.58
2.45A 0.01
1.34
3.80
1.56 0.12
3.73
1.71
2.03 0.05
1.44
2.33
2.41 0.02
2.00
1.51
2.06 0.04
1.31
5.27 2.03A
0.04
4.00

-1.43 0.16
0.54

0.59

-1.31 0.19
-1.77 0.08
-3.44 0.00
0.25

0.80

Note: N = 50, male (n = 24), women (c); total fixation duration and total visit duration are measured in seconds; visit
count and fixation count are measured in counts within the areas of interest; in column t(z), t-values and z-values are
shown; z-values are marked with an A