The Aesthetic Appeal of Fractals:
A Boost for Advertising Effectiveness?
Cora Graf
The Aesthetic Appeal of Fractals:
A Boost for Advertising Effectiveness?
Master Thesis
University of Groningen
Faculty of Business and Economics
MSc Marketing Management
11
thJanuary 2016
First Supervisor: dr.Yannick Joye
Second
Supervisor: Carmen M. Donato
Cora Graf
Schuitendiep 25 1/A
III
Abstract
Fractals have been noted to successfully describe and quantify the complex visual character of natural scenes. Besides their describing character, fractals are also known for their high aesthetic appeal. By using an online survey, this experimental study investigates the effect of different levels of fractal dimension in visual advertising on consumer responses in the form of willingness to buy and approach behavior. Participants are presented with a visual adver-tisement, which displays a fractal element with either a high, intermediate or low fractal di-mension as well as a hedonic or utilitarian product. The results show that both the level of fractal dimension and the product type do not have an effect on consumers’ willingness to buy or approach behavior toward the promoted product. However, it has been found that consumers’ aesthetic liking of the advertisement has a significant positive influence on their willingness to buy, which demonstrates the relevance of aesthetic design in visual advertis-ing.
Preface
The completion of this thesis also represents the end of my Master program of Marketing Management at the University of Groningen. The last year of spent in Groningen absolutely flew by, and I can definitely say that getting back to University, after gaining two years of practical work experience, has been a blast.
When applying for a thesis topic, I have, frankly speaking, never heard of fractals before. But after peeking into this option, I was sure that I wanted to research this topic. The process of writing this thesis allowed me to dive into the fascinating world of fractals. It has been challenging and complex at times but looking back it definitely provided a lot of fun and excitement, too.
I would like to take the opportunity to thank dr. Yannick Joye for his guidance and valuable support throughout the process of writing this thesis. He has been a great supervisor by pro-viding constructive feedback and motivating me to continuously improve my work.
Furthermore, I would like to thank my family for their motivating words and valuable sug-gestions. Last but not least, a big thank you goes to Janina for accompanying me along the road and making the year in Groningen an amazing time, and to Eddie, for encouraging and supporting me during the entire period.
11th January 2016
V
Table of Contents
List of Figures ...VII List of Tables... VIII List of Abbreviations... IX
1. Introduction ... 1
2. Literature Review... 5
2.1 Human Visual System and Perception ... 5
2.2 Aesthetic Experience ... 6
2.3 Fractals ... 10
2.4 The Effects of Fractal-like Image Properties on Aesthetic Liking... 13
2.5 The Effects of Aesthetic Liking of an Advertisement on WTB... 15
2.6 The Effects of Aesthetic Liking of an Advertisement on Approach Behavior ... 16
2.6 Sensitivity to Aesthetics ... 17
2.7 Product Type: Hedonic vs. Utilitarian... 17
2.8 Hypotheses and Conceptual Model ... 18
3. Methodology ... 22 3.1 Research Design ... 22 3.2 Participants ... 22 3.3 Materials ... 23 3.3.1 Stimuli ... 23 3.4 Measurements... 25
3.5 Procedure of the Survey ... 28
4. Results ... 30
4.1 Descriptive Statistics and Preliminary Analysis... 30
4.1.1 Normality Tests ... 32
4.2 Testing Hypotheses ... 32
4.2.1 Effect of Fractals and Product Type on WTB ... 32
4.2.2 Effect of Fractals and Product Type on Approach Behavior ... 34
4.2.3 Effect of Fractals and Product Type on Aesthetic Liking ... 35
4.2.4 Effect of Aesthetic Liking on WTB ... 36
4.2.5 Effect of Aesthetic Liking on Approach Behavior... 37
4.2.6 Mediating Effect of Aesthetic Liking... 38
4.2.7 Moderating Effect of Aesthetic Sensitivity ... 39
5. Discussion... 40
5.1 Conclusion... 40
5.3 Limitations and Future Research... 43
References ... 46
Appendix I: Scales... 54
VII
List of Figures
Figure 1: Photographs of Trees at Different Levels of Magnification (taken from
Taylor, Newell, Spehar & Clifford, 2005a). ... 1 Figure 2: Pictures of a Mathematical Fractal, a Natural Fractal and a Fractal in Artwork. .... 11 Figure 3: A Comparison of Patterns with Different D Values (taken from
Taylor et al., 2005a). ... 12 Figure 4: Visualization of Box-Counting Method (taken from Gardi, 2014). ... 12 Figure 5: Conceptual Model... 21 Figure 6: Six Manipulated Advertisements with Flat Screen TVs and Lamps as well
List of Tables
Table 1: 2 by 3 Research Design with the Factors Fractal Dimension and Product Type... 22
Table 2: Pre-Test: Mean Hedonic Values of Selected Products ... 24
Table 3: Condition Distribution, Results of ANOVA and Pearson Chi-squared Test ... 31
Table 4: Normality Tests for Dependent Variables, Mediator and Moderator ... 32
Table 5: Means of WTB and Results of Post Hoc Test: Pairwise Comparisons for WTB ... 33
Table 6: Means of Approach Behavior and Results of Post Hoc Test: Pairwise Comparison for Approach Behavior ... 35
Table 7: Means of Aesthetic Liking and Results of Post Hoc Test: Pairwise Comparisons for Aesthetic Liking ... 36
Table 8: Moderating Effect of Aesthetic Sensitivity on the Effect of Fractals on Aesthetic Liking ... 39
IX
List of Abbreviation
α Scale-invariance
BAS Behavioral Activation System BIF Behavior Identification Form df Degree(s) of freedom
e.g. Exempli gratia
1. Introduction
“Beauty is a manifestation of secret natural laws,
which otherwise would have been hidden from us forever.” - Johann Wolfgang von Goethe
Imagine standing in front of a tree and while looking through the lens of a camera, you closely observe all the single elements of the tree, such as the leaves, twigs and branches. Through zooming in and out of the individual elements, it can be seen that the complexity of the whole scene seems to derive from the fact that all single parts of the tree resemble the entire picture of the tree (see Figure 1). This describes the essential character of fractal pat-terns, which is referred to as self-similarity (Voss, 1988). A fractal represents a mathematical construct (Joye, 2006), which describes complex patterns that recur on increasingly finer scales (Taylor et al., 2007). Only after Mandelbrot introduced the concept of fractals in 1977 (Spehar, Clifford, Newell & Taylor, 2003), it has become possible to describe and quantify the visual complexity of many natural patterns (Taylor, Spehar, Van Donkelaar & Hagerhall, 2011). Although we are surrounded by fractals on a daily basis through the natural environ-ment, most people are not aware of the existence of fractal patterns and their describing character of nature.
Figure 1: Photographs of Trees at Different Levels of Magnification (taken from Taylor, Newell, Spehar & Clifford 2005a).
A significant characteristic of fractals is their pleasing aesthetic value for individuals (Spehar & Talyor, 2013). Since aesthetics are closely linked to marketing strategies, new product development and also retailing (Lavie & Tractinsky, 2004), fractals might represent an
Mathematics and Culture III
54
Fig. 1.
Trees are an
example of a natural
fractal object. Although
the patterns observed at
different magnifications do n’t r ep ea t e xa ctl y,
analysis shows them to
have the same statistical
qualities (photograph
2 interesting construct for marketers in general. Particularly, aesthetic perception is crucial in the field of advertising. It has been researched that aesthetic liking of an advertisement has a positive influence on a recipient’s attitude toward ads (Stammerjohan, Wood, Chang & Thorson, 2005), which in turn are a key driver in a consumer’s decision making process and for the degree of willingness to buy (MacKenzie & Lutz, 1989; Spears & Singh, 2004). Due to the fact that many companies spend large budgets on advertising campaigns that are not effective (Abraham & Lodish, 1990), there is a high relevance of creating ads that provide better results in stimulating sales. Additionally, with increasing competition and an amplified advertising clutter, it becomes exceedingly difficult for a print ad to attract the attention of consumers and to effectively influence their purchase decision making (Fennis & Stroebe, 2010). Hence, it is important that marketers on the one hand create advertisements that cap-ture the attention of consumers (Pieters, Wedel & Batra, 2010) and on the other hand take further steps to optimize the aesthetic appeal of print ads, which eventually lead to a positive attitude toward the ad (Stammerjohan et al., 2005) and a higher willingness to buy (Spears & Singh, 2004). Insights on fractals and their aesthetic appeal might also be relevant for mar-keters due to the potential of aesthetic stimuli to evoke positive emotions in individuals, which in turn serve as a basis for the motivation to approach and seek out to the stimuli (Desmet & Hekkert, 2007). If an advertisement is able to trigger approach behavior, recipi-ents might go to the store for a live inspection of the promoted product, which could eventu-ally lead to a purchase.
Further investigations of the relationship between fractal patterns and their aesthetic perception are required in order to understand how the visual character of fractals affects individuals and to derive effective strategies for the use of fractal-like image properties in advertising. Due to the fact that the topic of fractals has not yet been researched from a downright marketing perspective, new insights and subsequent guidelines for marketers on how to optimize aesthetic perception of ads through the strategic application of fractal pat-terns can potentially be gained. Since aesthetic liking of advertising is a crucial element in a consumer’s decision making process, the results might have a large leverage for practical use as well as positive impact on advertising effectiveness and product sales once applied in real business context.
Do different levels of fractal dimensions (low, intermediate, high) in visual advertisements differentially influence willingness to buy and approach behavior toward the promoted product?
Many studies have examined the relationship between nature and human responses in regard to aesthetic liking. It has been revealed that natural environments trigger and affect the aesthetic preferences of individuals in a universal way (Joye, 2006). Based on the fact that fractals are closely related to nature, it is of large interest how fractals can be related to aesthetic responses. Several scholars have made this connection and showed that images with a fractal dimension, similar to the one of natural environments, are more aesthetically pleasing (Spehar & Taylor, 2013). Thus, aesthetic liking is used as a mediator to test the re-lationship between fractal-like image statistics, and willingness to buy as well as approach behavior.
Continuing the focus on aesthetics, not only the effect of aesthetics in advertisements in general needs to be taken into consideration but also the type of product that is depicted in the ad ought to be differentiated. It is suggested that people usually place a higher emphasis on aesthetics in the evaluation of hedonic as opposed to utilitarian products (Hoyer & Stock-burger-Sauer, 2012). Therefore, it is researched whether the effects of aesthetic liking of an ad, caused by the ad’s fractal properties, on willingness to buy are different for hedonic, compared to utilitarian products. In a similar way it is argued that an aesthetic advertisement for a hedonic product triggers a higher level of approach behavior toward the good as op-posed to an ad for a utilitarian product. Hence, the present research also investigates whether the product type, promoted in the print advertisement, plays a role in the relations between the fractal dimensions of the ad and aesthetic liking, willingness to buy as well as approach behavior.
Furthermore, aesthetic sensitivity is a factor that is unique for every individual. When individuals have a more positive attitude toward aesthetics, they are more touched by an aes-thetic stimulus, thus their aesaes-thetic liking is enhanced (Diessner, Solom, Frost, Parsons & Davidson, 2008). Therefore, the effect a specific fractal dimension in the ad has on the aes-thetic preference may vary between individuals. Hence, the aesaes-thetic sensitivity of individu-als, i.e. the extent of importance a person attaches to visual aesthetics in response to stimuli (Bloch, Brunel & Arnold, 2003), is assessed, in order to find out whether this construct mod-erates the effect of fractals on aesthetic liking.
2. Literature Review
2.1 Human Visual System and Perception
Vision can be defined as “(…) the process of acquiring knowledge about environmental ob-jects and events by extracting information from the light the obob-jects emit or reflect“ (Hyvärinen, Hurri & Hoyer, 2009, p. 2). Although this process is characterized as a complex computational task, people are not aware of performing it since the underlying processes in the brain take place automatically and without any apparent effort (Hyvärinen, Hurri & Hoyer, 2009). A way to examine these underlying processes is the assessment of the percep-tual process, which is defined as a sequence of processes that determine an individual’s re-sponse to visual stimuli.
The perceptual process is characterized by four categories: 1) Stimulus, i.e. to what individuals pay attention to in their visual environment; 2) electricity, represents the creation of electrical signals by the receptors, which are then transmitted to the brain; 3) experience and action, refer to the perception, recognition and reaction to the stimuli; and 4) knowledge, represents the information that individuals have stored and which they add to the perceptual situation (Goldstein, 2013). Especially the second category in the perceptual process, i.e. electricity, which describes how our brain processes the vast amount of incoming informa-tion, is a primary issue in neuroscience (Olshausen & Field, 2004). The underlying processes of the electricity category are examined in more detail in the following because they serve as an explanation of how humans process visual information. This is relevant for the course of this thesis in order to provide further understanding of the visual brain.
(Ol-6 shausen & Field, 2004) because capacity in the resources of the nervous system is expanded, by decreasing the neural activity attributed to visual stimuli (Redies, 2007).
It is suggested that the human visual system is not static, but that adaptations and de-velopments take place over time and are influenced by the following three concepts: 1) indi-vidual tasks that need to be performed by humans; 2) limited resource capacity of the nerv-ous system; 3) living environment of the individual. Focusing on the role of the environ-ment, Simoncelli and Olshausen (2001) state that stimuli, which are encountered and experi-encedmore frequently, are also processed in a more efficient way by the perceptual system, compared to the stimuli, which are less common. Since, for example, natural environments frequently surround humans, the human visual system has adapted respectively. This means that neuronal activities respond to natural scenes in a specific way and hence result in effi-cient processing of natural images (Graham & Meng, 2011; Sekuler & Bennett, 2001; Si-moncelli & Olshausen, 2001). Besides the adaptation of the visual system through develop-mental processes, it is argued that evolutionary processes also have had an influencing effect (Simoncelli & Olshausen, 2001). The human visual brain has evolved in a natural environ-ment and therefore, as a result, it has become adapted to optimally process natural scene sta-tistics (Redies, 2007).
2.2 Aesthetic Experience
Although aesthetic perception is one of the oldest topics researched in psychology and many attempts have been made to define aesthetic preference, no general definition ex-ists (Palmer et al., 2013). Even though the concept of aesthetics is diverse and can be as-sessed from different perspectives, in the field of visual aesthetics, a possible definition is given by Redies (2015, p. 2), who defines aesthetic experience as “intense feeling of pleas-ure”, which is elicited by the visual encounter with a stimulus. According to this definition, the subjective experience of the beholder in response to the visual characteristics of an object underlies aesthetic preference. Furthermore, the character of aesthetics is twofold. On the one side, it needs to explain aesthetic value, i.e. which kind of objects are aesthetically pleas-ing. On the other hand, it is supposed to explain judgments by specifying visual properties related to aesthetic perception (Armstrong & Detweiler-Bedell, 2008).
Fluency as a Driver of Aesthetics
8 be used for optimizing stimulus presentation schedule and visual elements of marketing communications in order to increase their effectiveness (Janiszewski & Meyvis, 2001).
To assess the underlying determinates of aesthetic liking in further detail, the rela-tionship between visual design properties of stimuli and aesthetic responses is examined and followed by an assessment of the effect of repetition on aesthetic liking.
Visual Properties of Stimuli
Symmetry and visual complexity are two design principles, which receive a high level of attention in research (Creusen, Veryzer & Schoormans, 2010) and are important determi-nants of aesthetic liking (Jacobsen & Höfel, 2002; Reber et al., 2004).
Symmetry
Symmetric patterns are defined by the following two properties: “(…) all the virtual lines between symmetry points are parallel and have midpoints that are collinear” (van der Vloed, Csathó & van der Helm, 2005, p. 76). Symmetry in visual stimuli has been found to posi-tively affect aesthetic preferences (Palmer et al., 2013). Since a higher redundancy is related to symmetry, people have to extract less visual information from the presented stimulus, which results in more positive responses and higher aesthetic liking. The explanation for the positive relation between symmetry and aesthetic liking is that symmetrical patterns are eas-ier to process, thus eliciting a higher processing fluency and higher preference (Palmer et al., 2013; Reber et al., 2004). A further explanation of aesthetic liking for symmetric patterns can be presented from an evolutionary perspective, which suggests that people have an in-nate preference for symmetry (Etcoff, 1999). It is argued that symmetry increases the prefer-ence and attractiveness of human faces because it signals mate quality, in form of biological fitness and health (Rhodes, 2006).
Visual Complexity
Visual complexity can be defined based on the number and diversity of elements, which are present in an image or contained in a stimulus (Marin & Leder, 2013). In the literature on visual complexity the findings regarding the relation between complexity and aesthetic pref-erence are mixed. From the perspective of perceptual fluency, it is known that the relation-ship between processing fluency and aesthetic preference has a linear character. Visual com-plexity is argued to negatively affect processing ease and hence leads to lower aesthetic lik-ing (Reber et al. 2004). Challenglik-ing the theory of processlik-ing fluency, Armstrong and Detweiler-Bedell (2008, p. 305) state that people have an enhanced aesthetic liking for stim-uli with high levels compared to low levels of complexity. They reason that the fast recogni-tion of simple objects can only be mildly pleasing, “whereas sensing the prospect of success-fully representing a complex object can be exhilarating”. Meaning that processing fluency can only serve as an explanation for positive affect connected to simple objects, but cannot account for the enhanced pleasure associated with more complex stimuli (Armstrong & Detweiler-Bedell, 2008; Joye, Steg, Ünal & Pals, 2015). There is further empirical evidence that shows an increased preference for more complex stimuli (Landwehr, Labroo & Herrmann, 2011). In contrast, however, a number of studies suggest that an intermediate level of visual complexity is preferred (Leder, Belke, Oeberst & Augustin, 2004; Palmer et al., 2013). These findings are in line with Berlyne’s (1974) model of arousal that proposes an inverted U-shaped relationship between complexity and liking (Cox & Cox, 2002; Lavie & Tractinsky, 2014).
10 Repetition
Repetition in regard to a stimulus is twofold: On the one hand repetition is related to re-peated encounters with an object, and on the other hand it refers to the internal visual repeti-tion of an image.
Assessing the history of a perceiver’s experience with a certain stimulus, it has been shown that mere repeated exposure to a stimulus enhances the positive affect toward this stimulus (Zajonc, 1968). Thus, a higher exposure frequency can amplify peoples’ aesthetic judgments (Palmer et al., 2013). Understanding the underlying causes between the relation of exposure and positive evaluations, the processing fluency/attribution model provides a possible explanation. Through exposure to a stimulus, a representation of that stimulus is created in memory, which in turn serves as a processing facilitation when the stimulus is reencountered at a later time (Janiszewski & Meyvis, 2001). However, this might not apply for all types of stimuli. In their study of preference and exposure, Cox and Cox (2002) find that repeated exposure enhances the liking of complex stimuli, while for more simple de-signs repeated exposure leads to declining preferences. In the literature numerous researchers have focused on examining how repetition influences advertising effectiveness, however there is no distinct finding. The most prominent theory argues that low levels of repetition first increase the effectiveness of an ad message, while a further increase of repetition leads to a decrease of effectiveness (Campbell & Keller, 2003).
Internal visual repetition is a concept that describes repeating the same shapes within one image. Joye et al. (2015) empirically studied the effects of visual complexity, which is derived from internal repetition, on perceptual fluency. They show that individuals can more easily process complex stimuli when high levels of internal repetition characterize these stimuli. Since evidence in this field is rare, further empirical research is needed in order to gain more insights into the effects of internal repetition on aesthetic liking.
2.3 Fractals
Hayn-Leichsenring, Langner, Wiese, & Redies, 2015). This self-similarity is the essential property of fractal patterns (Joye, 2006; Voss, 1988), which were first introduced by Mandelbrot in 1977. The word “fractal” is derived from the Latin word “frangere”, which can be translated with “to break” (Mandelbrot, 1983). With the discovery of fractals, it has become possible to mathematically quantify the complex and fractured characteristics of natural elements (Spe-har & Taylor, 2013). Hence, fractals are often described as fingerprints of nature (Forsythe, Nadal, Sheehy, Cela-Conde & Sawey, 2011).
A specific form of self-similarity is scale invariance, which implies that as one zooms in or out of a fractal pattern, the detail and spatial frequency presented in the image remain constant, i.e. self-similar at different levels of resolution (Olshausen & Field, 2000; Redies, 2007). However, fractal structures are not only present in nature. As presented in Figure 2, a differentiation between three categories of fractal patterns can be made: Natural fractals (e.g. coastlines), mathematical fractals (e.g. computer simulations of mountains) and human frac-tals (e.g. present in artwork) (Spehar & Taylor, 2013).
Figure 2: Pictures of a Mathematical Fractal, a Natural Fractal and a Fractal in Art-work. (a) A fragment from a Mandelbrot Set (taken from Wikipedia, 2013), (b) Pic-ture of a Romanesco Cauliflower (taken from Wikipedia, 2008), (c) Painting Number 32 from Jason Pollock (1950).
12
scribed by familiar integer values – for a smooth line (containing no fractal
struc-ture)
D has a value of 1, whilst for a completely filled area (again containing no
fractal structure) its value is 2. However, the repeating structure of a fractal
pat-tern causes the line to begin to occupy area.
D then lies between 1 and 2 and, as the
complexity and richness of the repeating structure increases, its value moves
closer to 2. Figure 4 demonstrates how a fractal pattern’s
D value has a profound
effect on its visual appearance. For fractals described by a low
D value close to one
(left), the patterns observed at different magnifications repeat in a way that builds
a very smooth, sparse shape. However, for fractals described by a
D value closer to
two the repeating patterns build a shape full of intricate, detailed structure
(right).
The research by Rogowitz and Voss indicates that people perceive imaginary
objects (such as human figures, faces, animals etc.) in fractal patterns
character-ized by low
D values [6]. For fractal patterns with increasingly high D values this
perception falls off markedly. This result caused Rogowitz and Voss to speculate
that the ink blots used to induce projective imagery in psychology tests of the
1920s were fractal patterns described by low
D values. Indeed, their subsequent
Fractals: A Resonance between Art and Nature57
Fig. 3. Ink blot patterns created by R.P. Taylor using the technique employed
by Rorschach when generating his ten original patterns
Fig. 4. A comparison of patterns with different D values:
1 (left), 1.1, 1.6, 1.9 and 2 (right)
D=1
(non-fractal) D=1.1 D=1.6 D=1.9 (non-fractal)D=2
one and two (Taylor et al., 2007). With increasing richness and complexity of the fractal structures, the value of D moves closer to two (Taylor & Sprott, 2008). This is exemplified in Figure 3. In regard to the fractal dimension D in nature, it has been found that natural pat-terns are most frequently characterized by mid-range D values around 1.3 (Aks & Sprott, 1996; Taylor, Spehar, Van Donkelaar & Hagerhall, 2011).
Figure 3: A Comparison of Patterns with Different D Values (taken from Taylor et al., 2005a).
For measuring fractal-like image properties in form of the fractal dimension D, the box-counting method is well established (Mureika & Taylor, 2013; Taylor et al., 2007). This is provided through the ability to measure the fractal dimension of not entirely self-similar im-ages (Forsythe et al., 2011). The technique examines the fractal dimension D by using a network of identically sized boxes to cover the surface of an image. The method counts the number of boxes, which intersect with part of the image. The procedure is repeated with con-tinuously smaller boxes within the network, which represents the examination at a higher spatial frequency. The fractal dimension D is then determined by measuring the proportion of boxes filled by the pattern and the proportion of squares that are empty (Mather, 2013; Mureika & Taylor, 2013; Spehar & Taylor, 2013; Taylor, 2003). A visualization of the de-scribed box-counting method is presented in Figure 4.
Figure 4: Visualization of Box-Counting Method(taken from Gardi, 2014).
statis-tics are extracted, i.e. the slope of the power/amplitude spectrum of an image is measured (Párraga, Troscianko & Tolhurst, 2000; Redies et al., 2015). The slope of the power/amplitude spectrum quantifies the relative contribution of different spatial frequencies or Fourier elements to the image as a whole. In other words, the slope value is a characteri-zation of the relative proportion of low spatial frequencies (i.e. rough detail) and high spatial frequencies (i.e. fine detail) in an image (Graham & Field 2008; Graham & Redies, 2010; Redies et al., 2015). It has been found that 1/fα (or equivalent f -α) describes the dependency between frequency and amplitude, where f is the frequency and exponent α quantifies the scale-invariance of an image (Graham & Field 2008; Mather, 2014; Spehar & Taylor, 2013). Several authors state that a Fourier slope of -2 characterizes the spatial frequency distribu-tion of natural scenes, implying that nature possesses fractal-like properties in form of scale-invariance (Koch, Denzler & Redies, 2010; Menzel et al., 2015).
The relationship between exponent α and parameter D, can be defined as an inverted correlation, i.e. the higher the D value, the lower α. Hence, image patterns with high com-plexity and detail can be characterized by a high D and a low α value. The main difference between the two constructs is that the D value quantifies the “(…) scale-invariance of the fractal boundaries lines“, whereas α measures the “(…) scale-invariance of the grayscale fractal surface (…)“ (Spehar & Taylor, 2013, p. 3).
2.4 The Effects of Fractal-like Image Properties on Aesthetic Liking
14 Connecting these findings to the fact that the fractal dimension of natural environ-ments can most commonly be described by a D value around 1.3 (Aks & Sprott, 1996; Tay-lor et al., 2011), research claims that the visual preference for mid-value fractal dimensions is linked to a “continuous visual exposure” to nature (Spehar & Taylor, 2013, p. 2). In a similar notion, it is also suggested in the literature, that images characterized by a Fourier slope of -2, as found in natural scenes, are aesthetically preferred (Menzel et al., 2015).
Two explanations exist for this preference of individuals for natural scene statistics. First, it is suggested that people have an evolved preference and affinity for nature, which is referred to as ‘biophilia hypothesis’ (Kellert & Wilson, 1993). Second, it is argued that the human visual system has developmentally adapted to the processing of visual patterns in the natural environment (Cheun & Wells, 2004; Redies, 2007), as explained in Chapter 2.1 in the context of visual processing. Therefore, fractal patterns can be processed with a higher fluency, leading to enhanced aesthetic liking for fractal stimuli, corresponding to the fractal dimensions of natural scenes in the range around 1.3 (Párraga et al., 2000; Redies, 2007; Spehar & Taylor, 2013) and a Fourier slope of -2 (Koch et al., 2010; Menzel et al., 2015).
Empirical findings from Joye et al. (2015) support that perceptual fluency accounts for the preference of fractals. They reason that the processing fluency theory might not only has explanatory power in regard to aesthetic liking of simpler non-fractal structures, but also regarding complex fractal stimuli. It is known that complex fractal patterns are characterized by a high visual redundancy (Joye & Van den Berg, 2011). Visual complexity theory states that images with a high redundancy are less complex (Donderi, 2006), leading to facilitated processing (Joye & Van den Berg, 2011) and hence higher aesthetic preference (Joye et al., 2015). Forsythe and Sheehy (2011) contribute to the discussion about the relation between fractals and aesthetics by stating that fractals capture complexity as well as order and strive a balance between the two, which makes them aesthetically pleasing.
the effectiveness of visual stimuli. Consequently, it is advisable to examine the effect of fractals in the field of marketing.
2.5 The Effects of Aesthetic Liking of an Advertisement on WTB
Print advertising is a specific form of a designed image. Through the use of visually pleasing stimuli (Phillips & McQuarrie, 2004), advertising first aims at attracting consumers’ atten-tion (Braun et al., 2013) and in the following step intends to change their responses, so that as a result the promoted product gets purchased (Fennis & Stroebe, 2010). Applying a hier-archy-of-effects model the intermediate steps between advertising exposure and the resulting consumer responses can be described. More precisely, advertising takes consumers through a sequence of cognitive, affective and conative stages while forming purchase intentions (Fen-nis & Stroebe, 2010; Vakratsas & Ambler, 1999), which refer to a consumer’s plan to buy a certain product or brand (Spears & Singh, 2004). Purchase intentions are related to willing-ness to buy (WTB), which refers to the likelihood that the consumer has an intention to buy a product (Dodds, Monroe & Grewal, 1991).
16
2.6 The Effects of Aesthetic Liking of an Advertisement on Approach
Behavior
In the literature it is suggested that aesthetic experience can be seen as a specific form of appraisal (Desmet & Hekkert, 2007). Appraisal theories assume that emotions are the out-come of evaluative processes. Individuals continuously evaluate events that they encounter and if an event is identified as relevant, it is assessed in more detail. The examination ulti-mately results in corresponding emotional responses (Lazarus, 1991). Hence, the encounter with a stimulus can trigger both positive and negative emotions, such as pleasure or dis-pleasure. Furthermore, it is assumed that individuals are motivated to seek stimuli that pro-vide pleasure and avoid those that propro-vide displeasure (Desmet & Hekkert, 2007).
In general such behavior is either referred to as approach motivation or avoidance motivation (Arnold & Reynolds, 2012). Approach motivation is related to the drive toward an object, whereas withdrawal motivation refers to an urge to move away from an object (Gable & Harmon-Jones, 2008). Hence, when a stimulus is evaluated as aesthetically pleas-ing, it encourages approach behavior within an individual (Desmet & Hekkert, 2007). In a research of Ulrich (1983), studying aesthetic responses of individuals to natural environ-ments, it is stated that visual stimuli can elicit aesthetic preferences and increase arousal within the beholder. Furthermore, these reactions to the visual object lead to a behavior change in form of approach behavior, such as further exploring and seeking out for the stimulus (Mehrabian & Russell, 1974). In another study from Gable and Harmon-Jones (2008), participants report a higher approach motivation toward pictures, which they find more pleasing and desirable.
2.6 Sensitivity to Aesthetics
Besides having diverse aesthetic preferences toward stimuli, research has found that people also differ in the amount of importance they place on aesthetics in their formation of pur-chase intentions (Bloch, Brunel & Arnold, 2003). This variability can be attributed to a per-son’s individual sensitivity to aesthetics (Reinecke & Gajos, 2014). Aesthetic sensitivity of a person can be defined as the extent of importance a person attaches to visual aesthetics in response to stimuli (Bloch et al., 2003). It is suggested that a person who is highly aestheti-cally sensitive, perceives a stimulus with higher aesthetic value as more pleasing compared to a stimulus with lower aesthetic value (Child, 1964). Therefore, for people with a high aes-thetical sensitivity, visual aesthetics are in general of higher importance (Bloch et al., 2003) and they are more touched by an aesthetic stimuli, thus their aesthetic liking is enhanced (Diessner et al., 2008).
This individual connection of a person to aesthetics is an essential factor when de-termining his/her responsiveness to advertisements containing aesthetic images. When an individual is more sensitive to aesthetics, he/she might have an enhanced experience of the positive effects of aesthetically pleasing images in general and might be more responsive compared to people with a lower sensitivity to aesthetics. In order to increase advertising effectiveness, these insights could, for example, allow marketers to tailor advertisings cam-paigns to the individual preference of consumer segments regarding aesthetics.
2.7 Product Type: Hedonic vs. Utilitarian
18 as hedonic or utilitarian also depends on the classification and importance of its attributes to the consumer (Park & Moon, 2003). For example, a car can be bought for its functional pur-pose of transportation (utilitarian) from one person, whereas for others the feeling and pleas-ure it provides when driving is of higher importance (hedonic).
The distinction between hedonic and utilitarian product classifications also affects a consumer’s decision making process. It has been found that consumers decide upon hedonic consumption based on their expectation of what will be most pleasurable for a large amount of time (Alba & Williams, 2012). Moreover, since aesthetics have a strong association with hedonic properties, aesthetics play an important role in consumers’ decision making process in regard to hedonic products (Holbrook & Moore, 1981; Hoyer & Stockburger-Sauer, 2012). In contrast, for utilitarian goods, consumers’ expertise and knowledge of a product-class are crucial for forming purchase decisions (Hoyer & Stockburger-Sauer, 2012). Al-though a lack of empirical results exists that examines the effect of aesthetics in the decision making process regarding ads of hedonic or utilitarian products, it is commonly suggested that the two product types differ in the way they affect a consumer’s decision making proc-ess (Dhar & Wertenbroch, 2000; Hoyer & Stockburger-Sauer, 2012; Voss, Spangenberg & Grohmann, 2003). Hence, in order to allow for a higher quality of results in this research, it is important to make a distinction between hedonic and utilitarian products.
2.8 Hypotheses and Conceptual Model
The discussed literature provides the basis for the present research and the established con-ceptual model with its corresponding hypotheses. This paper investigates the effect of frac-tal-like image characteristics in print advertising on aesthetic liking, willingness to buy and approach behavior. Furthermore, the differences between hedonic and utilitarian products featured in the ad and its corresponding effects on aesthetic liking are analyzed under the consideration of the moderating effect of individual sensitivity to aesthetics.
Since the introduction of the concept of fractals a large amount of research has been done by relating fractal properties to visual processing, art, music (Mureika, 2005) and also architecture (Joye, 2007). However, in the field of fractals in relation to marketing, evidence is very rare. Therefore, this research aims to examine whether fractal-like image properties in print advertisements lead to more positive consumer responses in regard to WTB and ap-proach behavior. Hence, the following hypotheses are formulated:
H1b: Fractal-like image properties have a positive effect on approach behavior for both he-donic and utilitarian products.
It is generally acknowledged that people derive aesthetic pleasure from nature (Joye et al., 2015) and furthermore that fractals have a high aesthetic appeal (Spehar & Taylor, 2013). Moreover, it is known that natural fractal patterns cluster within an intermediate frac-tal dimension D of 1.3-1.5 (Spehar & Taylor, 2013). Hence, it is suggested that with chang-ing visual complexity related to a higher or lower level of D as opposed to the intermediate range of 1.3-1.5, aesthetic liking might go down. Therefore, it is hypothesized that:
H2: The aesthetic liking of a print advertisement is higher when the fractal-like image prop-erties in the ad correspond to the mid-range fractal dimension of natural scenes, as opposed to lower or higher fractal dimensions.
Hedonic products are defined as goods with a high experimental as well as affective value. Utilitarian products, on the other hand, have a primary functional character (Dhar & Wertenbroch 2000). It is suggested that aesthetics have a strong association with hedonic properties and are therefore more pronounced for hedonic as opposed to utilitarian products (Hoyer & Stockburger-Sauer, 2012). Hence, it is assumed that an advertisement with inter-mediate fractal dimension leads to a higher level of aesthetic liking of the ad, when promot-ing hedonic products as opposed to utilitarian products.
H3: Aesthetic liking of a print advertisement with intermediate fractal dimension is higher for hedonic compared to utilitarian products.
Print advertising aims at changing consumer buying behavior in favorable ways through the use of visually pleasing images (Phillips & McQuarrie, 2004). Attitude toward the ad has been identified as a key driver of consumer behavior (Spears & Singh, 2004), which can be positively influenced by a high aesthetic value of an advertisement (Stammer-johan et al., 2005). Hence, if advertisements are aesthetically liked and elicit positive affec-tive responses from recipients, the willingness to buy the promoted products increases, lead-ing to the followlead-ing hypothesis:
H4a: Aesthetic liking of a print advertisement has a positive effect on WTB.
20 (Gable & Harmon-Jones, 2008). It is assumed that individuals have a higher approach moti-vation toward stimuli that provide pleasure and avoid those that provide displeasure. Hence, when a stimulus is evaluated as aesthetically pleasing, it encourages approach behavior within an individual (Desmet & Hekkert, 2007). Based on these insights from the literature, it is hypothesized that:
H4b: Aesthetic liking of a print advertisement has a positive effect on approach behavior. Based on the characteristic attributes of the hedonic and utilitarian product classifica-tions, it is argued that aesthetics play a crucial role in a consumer’s decision making process in regard to hedonic products. The converse argument applies for utilitarian products (Hoyer & Stockburger-Sauer, 2012). In a similar way it can be argued that aesthetics are more re-lated to hedonic values (Hoyer & Stockburger-Sauer, 2012) and therefore trigger a higher level of approach behavior toward hedonic as opposed to utilitarian products. Therefore, the following hypotheses are formulated:
H5a: The effect of aesthetic liking of a print advertising on WTB is higher for hedonic com-pared to utilitarian products.
H5b: The effect of aesthetic liking of a print advertising on approach behavior is higher for hedonic compared to utilitarian products.
As discussed in the literature, fractals have been acknowledged for their high aes-thetic value (Spehar & Taylor, 2013). However, evidence in this field is rare and it is not clear how fractal-like image properties in print advertisements affect aesthetic liking, which in turn has an effect on WTB and approach behavior. To gain insights into these relation-ships, it is hypothesized that the positive effect of fractals on WTB and approach behavior is mediated by aesthetic liking.
H6a: The aesthetic liking of a print advertisement, caused by fractal-like image properties, has a positive mediating effect on WTB.
H6b: The aesthetic liking of a print advertisement, caused by fractal-like image properties, has a positive mediating effect on approach behavior.
in-sights, it is suggested that liking of an aesthetic stimuli, i.e. an ad with intermediate fractal dimension, is higher for aesthetically sensitive individuals. Hence, the following hypothesis is derived:
H7: The higher the aesthetic sensitivity of an individual, the more positive is the relationship between advertisements with intermediate fractal dimensions and aesthetic liking.
Figure 5: Conceptual Model
22
3. Methodology
In the previous chapter the relevant literature for this research has been reviewed, hypotheses derived and summarized in a conceptual model. In order to test the hypotheses included in the conceptual model, empirical research has been conducted. The present chapter specifies the methodology of the research by addressing the following: research design, participants, materials of the study, measurement of the different variables and procedure of the question-naire.
3.1 Research Design
An online survey via the research platform Qualtrics was chosen as instrument for data ac-quisition because it is relatively cost-efficient and allows reaching high coverage. Since this research aims at measuring the effects of two product types (hedonic vs. utilitarian product) as well as three different levels of fractal dimension (levels of fractal dimension: high, in-termediate, low) as factors in an experiment, there are two independent variables with two and three levels. Hence, this research made use of a 2 x 3 design. In the experiment, manipu-lation was achieved through altogether six different advertisements that participants were asked to look at. Furthermore, a between-subject design was used in the experiment, mean-ing that each respondent was assigned to a different experimental condition, i.e. a specific combination of one “level of fractal dimension" and one “product type” condition. This has the advantage that sequence effects, which could lead to awareness bias of respondents after being exposed to one experimental condition, can be avoided (Aronson, Wilson & Brewer, 1998). Random assignment of participants to experimental conditions was used to make sure that the results of the research are not influenced by preexisting differences between partici-pants, but are only caused by the manipulation of the variables.
Table 1: 2 by 3 Research Design with the Factors Fractal Dimension and Product Type
3.2 Participants
The online questionnaire was distributed in Germany and the Netherlands. Potential partici-pants received a link to the study through the social media platform Facebook or via email.
Low Fractal Dimension Intermediate Fractal Dimension High Fractal Dimension
Hedonic Product Condition 1 Condition 2 Condition 3
The study was online for 8 days and was accessible via computer and Smartphone. In total 194 respondents participated in the survey. Forty-one participants (21.13%) had to be ex-cluded, as they dropped out of the questionnaire after seeing the introduction. Hence, in total 153 participants conducted the survey until the end and their answers are used in the analy-sis. Of these 153 respondents, 68 (44.4%) were male and 85 (55.6%) were female. The age among the participants ranged from 17 to 63 years (M=29.1, SD = 8.27).
3.3 Materials
3.3.1 Stimuli
Selection of Products
As discussed in the literature review a distinction can be made between hedonic and utilitar-ian products. Hedonic products can provide pleasure, fun and excitement to the consumer (Hirschman & Holbrook 1982), whereas utilitarian products are non-sensory and the func-tional character of the product is most important for consumers (Batra & Ahtola, 1991). This research aims at controlling for these differences between the two products types. Therefore, the experiment includes one advertisement that depicts a product of hedonic nature and the other one of utilitarian. In order to make sure that from the finally chosen products that are depicted in the advertisements, one is generally seen as ‘high hedonic/low utilitarian’ and the other product as ‘low hedonic/high utilitarian’, a pre-test was conducted. A pre-selection of six products was rated on one dimension (hedonic value). The choice of products for the pre-test is based on the following criteria: a) the product can be depicted in such a way that it is easily recognizable for the participants what type of product it is; b) the product can be pre-sented in a pleasant way in connection with fractal patterns, so the ad produces a harmonious overall picture; c) the products are all durable goods, so they are more comparable.
24
Table 2: Pre-Test: Mean Hedonic Value of Selected Products
Based on the results of the pre-test, a lamp is chosen as the utilitarian product since it has the lowest mean hedonic value and a flat screen TV is chosen as the hedonic product because it has the highest mean hedonic value. In order to assess whether a lamp and a flat screen TV do not deviate too much from each other in attractiveness, another pre-test was conducted. The second pre-test was conducted with German and Dutch respondents (four female and four male, from different age categories). Respondents were asked to rate their aesthetic lik-ing of a range of eight images of different lamps and eight images of various flat screen TVs on a 7-point Likert scale ranging from 1 (“I don’t like it”) to 7 (“I like it very much”) (Landwehr, Wentzel & Herrmann, 2013, p. 13). Based on the results an image of a lamp and a flat screen were chosen, which closely resemble each other based on their attractiveness. These two products were used in the advertisement stimuli of this study.
Manipulated Advertisements
The stimuli that were used within this study are six self-made print advertisement (see Fig-ure 6), each of them showing a specific product image and a fractal element. The fractals were designed with an online tool (Disseldorp, 2013), which allows to select a specific shape and increase the number of iterations or internal repetition of this shape. Although the dis-tinct fractal dimensions of the fractal elements were not measured, it is suggested that through increasing the level of internal repetition of the image, also the fractal dimension of the image increases. Hence, in the following an ad with a relatively low number of iterations is referred to as low fractal dimension and with increasing iterations as intermediate and high fractal dimension. The same size of the fractal image is used in all of the six print advertise-ments. Although the pictures do not have a typical advertisement character, they are referred to as print advertisements throughout the course of this paper. Within the survey they were referred to as pictures. The ads are in black and white and no wording is depicted, so partici-pants are neither influenced by color nor text.
Product Mean Hedonic Value
Figure 6: Six Manipulated Advertisements with Flat Screen TVs and Lamps as well as Fractal Elements with Low, Intermediate and High Number of Iterations.
3.4 Measurements
Aesthetic Liking
26 Willingness to Buy
After being presented with the advertisement, the participants’ willingness to buy the pro-moted product was measured. This was done by using the scale of Grewal, Monroe and Krishnan (1998, p. 51), which includes three statements that participants had to rate on a 7-point Likert Scale, ranging from 1 (very low) to 7 (very high). Depending on the condition (hedonic vs. utilitarian product), in which the participant was in, he/she received statements regarding the advertised lamp or flat screen TV: “If I were going to buy a lamp/flat screen TV, the probability of buying this model is”, “The probability that I would consider buying this lamp/flat screen TV is”, “The likelihood that I would purchase this lamp/flat screen TV is”. The WTB scale shows a strong internal consistency with a Cronbach’s Alpha of .943. Approach Behavior
The level of approach behavior toward the promoted product was measured by using the behavioral activation system (BAS) scale. In a study of Schmeichel, Harmon-Jones and Harmon-Jones (2010) it is argued that this scale was originally designed as a trait measure. However, since currently no state measure of approach behavior exists, the authors suggest that the scale can be used to as a dependent measure.
Only the “BAS Drive” part of the scale was used because it consists of items that measure the pursuit of desired goals and hence fits best within the scope of this study. Par-ticipants were asked to rate their agreement with the following four items in regard to the advertised lamp/flat screen TV on a 7-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree): “When I want something, I usually go all-out to get it”, “I go out of my way to get things I want”, “If I see a chance to get something I want, I move on it right away”, “When I go after something I use a ‘no holds barred’ approach” (Carver & White, 1994, p. 323). The BAS scale shows a strong internal consistency with a Cronbach’s Alpha of .797.
Sensitivity to Aesthetics
percep-tual/cognitive (“I notice beauty”); 2) physiological arousal (“I feel a lump in my throat” or “an expansion in my chest”); 3) conscious emotion (“I feel emotional, it moves me”) and 4) transcendent or spiritual (“I feel a spiritual experience, a sense of oneness, or being united with the universe”) (p. 309). Participants had to rate 4 statements on a 7-point Likert scale, ranging from 1 (very unlike me) to 7 (very much like me). All statements of the EBS scale, which were used in the experiment, are shown in Appendix A. The EBS scale shows a strong internal consistency with a Cronbach’s Alpha of .762.
Action Identification
Although not discussed in the literature review of this thesis, a wide literature on construal and consumer behavior exists. Hence, for explanatory purposes this research also tests for individual preferences for action identification1. This was measured by using the Behavior
Identification Form (BIF), which has been introduced by Vallacher and Wegner (1989). In-dividuals were presented with different behaviors and for each behavior, participants could choose between two alternatives that describe the same behavior in different ways, i.e. either low or high level action identification. Low levels refer to the details or means of the action, whereas high levels frame the action in terms of its end. For example, “eating” was pre-sented as one action behavior for which “chewing and swallowing” was prepre-sented as low level and “getting nutrition” as high level action (Fujita, Henderson, Eng, Trope & Liber-man, 2006, p. 280). The complete list of items is presented in Appendix B.
Control Variables
The questionnaire included various control variables. The following demographic variables were incorporated in the survey through questions that describe the participants in the re-search and give insights into the representativeness of the sample: 1) “What is your gender?” This is a nominal variable with the answer possibilities (1) Male and (2) Female. 2) “What is your age?” This is an open question, resulting in interval data. 3) “What is your highest level of education?” This is a nominal variable with an 8-point scale. Answer options are: Basic education, Mbo 2/3 (secondary school), Mbo 4 (A levels), Havo/Vwo (vocational training), Bachelor, Master/Diploma, PhD or Other.
An additional control variable was included in the study, which assesses the general attitude of participants toward print advertising. It is suggested in the literature to measure
1 Action Identification is not discussed in the results chapter because no significant effects have been
28 the general attitude to advertising because it affects how participants react to advertisements. Hence, the scale of Mehta (2000, p. 69) was used. Six statements were included in the study, which assess the general attitude of the participants toward advertising. For example, “I like to look at advertising”. Ratings were made on a 7-point Likert Scale ranging from 1 (strongly disagree) to 7 (strongly agree). The complete list of statements can be found in Appendix C. The scale to measure the general attitude toward advertising has a Cronbach’s Alpha of .725. Deleting the question “On average, brands that are advertised are better in quality than brands that are not advertised” would improve the Cronbach’s Alpha to .757 but since this represents only a minor improvement of the internal consistency, it is decided to proceed with the original scale.
Moreover, the questionnaire also included a manipulation check regarding the donic and utilitarian value of the promoted products. Participants were asked to rate the he-donic value of the lamp and flat screen TV on the same five attributes on a 5-point Likert scale, which have also been used in the pre-test (Voss et al., 2003). The scale, which was used to measure the hedonic product value, has a Cronbach’s Alpha of .802. Deleting the item “Enjoyable” would improve the Cronbach’s Alpha to .858. Since this presents a rather large improvement of .56, this item is taken out of the further analysis with this scale.
3.5 Procedure of the Survey
30
4. Results
In the present chapter the data gathered through the quantitative research are described and analyzed. First, the descriptive statistics of the data set are presented in order to provide in-sights into the demographics of the respondents. Furthermore, it is tested to what extent de-mographical differences are present within the different groups of participants assigned to the experimental conditions. Subsequently, the normality of the data is tested. In the third and most important part of this chapter, ANOVA, a correlation analysis as well as mediation and moderation analyses are used to validate the hypotheses.
4.1 Descriptive Statistics and Preliminary Analysis
Prior to the data set description and analysis, the data had to be edited. At first, all partici-pants’ answers were reviewed with regard to precision and accuracy (Malhotra, 2010). In total 194 participants have started the survey, of which 153 filled in the survey until the end. This implies a dropout rate of 21.13%. All unfinished surveys were removed from the data set because the participants dropped out after seeing the introduction and hence did not pro-vide any insights for this research. Furthermore, outlier analyses on the dependent variables, WTB and approach behavior as well as on the mediating variable, aesthetic liking, were conducted. The box plot for the variable approach behavior (Appendix D) detected five out-liers. Based on the rule of thumb (Mean ± 3SD), the results of the outlier detection calcula-tions show that the outliers are still in range of the acceptable tolerance and hence no further participants had to be deleted.
In this research each of the 153 respondents was randomly assigned to one of the six different experimental conditions. The overview of this distribution of the participants across the conditions is presented in Table 3a. Due to the deletion of respondents who did not fill in the survey until the end, the distribution of participants to the conditions is slightly uneven. To reveal systematic differences between the participants, assigned to the six experimental conditions, with regard to the demographic variables gender, age and education, an ANOVA and a Pearson’s Chi-squared test were conducted. The metric variable age was tested by a one-way ANOVA, while education and gender as non-metric variables were reviewed by Pearson’s Chi-squared test. The results of the ANOVA and the Pearson’s Chi-squared test are presented in Table 3b and show that the null hypothesis (there are no significant differ-ences) cannot be rejected (p > .05). Hence, the six experimental conditions did not differ significantly with regard to the demographic variables, age, gender and education.
Table 3: Condition Distribution, Results of ANOVA and Pearson Chi-squared Test Before proceeding with the analysis of the data, a manipulation check was conducted in or-der to assess whether participants confirm the two products, which were used in the present study, as respectively hedonic or utilitarian. The control variable hedonic product value shows that participants rated the hedonic product with a higher hedonic value (M = 3.03, SD = 0.81) than the utilitarian product (M = 2.27, SD = 0.93). In order to test whether significant differences exist between the two independent sample means, i.e. hedonic product value for hedonic and utilitarian product type, a t-test was conducted. The t-test for an independent sample was chosen because the product type was manipulated between subjects. The results show that significant differences exist in the mean values of the hedonic product value be-tween the two product types, hedonic and utilitarian (t(151) = 5.35, p = .000). Thus
32 ing that the selected products can be seen as a representation for either the hedonic or the utilitarian product type.
4.1.1 Normality Tests
Normality tests were used in order to gain insights into the distribution of the collected data. This was done by consulting the values of skewness and kurtosis for the dependent variables as well as the moderating and mediating variables. In theory, the skewness statistic indicates whether the data are normally and symmetrical distributed, whereas the kurtosis statistic depicts how peaked or flat the data is. For skewness values between -0.5 and 0.5 and for the kurtosis statistics between -1.96 and 1.96, the data shows a normal distribution. However, the closer the value is to zero, the more normally distributed the data is (Field, 2009). Table 4 provides an overview of the normality tests for the variables of the conceptual model. Most variables are normally distributed; only the variables WTB and aesthetic liking are slightly negatively skewed, which is appropriate and within the margin to continue with the analysis.
Table 4: Normality Tests for Dependent Variables, Mediator and Moderator
4.2 Testing Hypotheses
4.2.1 Effect of Fractals and Product Type on WTB
Examining the mean values of WTB for the independent variables, fractal dimension and product type, presented in Table 5, it can be seen that for both the hedonic product and the utilitarian product a high level of fractal dimension leads to the highest level of WTB.
A two-way ANOVA has been used to examine group differences in the mean values of WTB across the experimental conditions fractal dimension and product type. Moreover, it was tested whether the interaction between the two independent variables (fractal dimension and product type) has had an effect on WTB. Fractal dimension (low = 0, intermediate = 1, high = 2) and product type (hedonic = 0, utilitarian = 1) served as the independent variables and WTB as dependent variable. The results of the two-way ANOVA (Appendix E) show
that the main effect of fractal dimension F(2, 147) = 0.64, p = .529 and the main effect of product type F(1, 147) = 0.12, p = .729 on WTB are statistically not significant. Further-more, the interaction effect between fractal dimension and product type on WTB is not sig-nificant F(2, 147) = 0.01, p = .993. Hence, hypothesis 1a, which suggests that fractal-like image properties have a positive effect on WTB for both a hedonic and a utilitarian product, is not confirmed.
Since no significant main effect of fractal dimension and product type on WTB had been found, a post hoc test with pairwise comparisons for the three levels of fractal dimen-sion and the two different product types has been conducted. The results are presented in Table 5 and show that the six experimental conditions do not significantly differ from each other in regard to WTB.
Table 5: Means of WTB and Results of Post Hoc Test: Pairwise Comparisons for WTB
Means of WTB
Fractal Dimension Product Type Mean Standard Deviation
Low Hedonic 2.99 1.36 Utilitarian 2.87 1.61 Intermediate Hedonic 3.16 1.49 Utilitarian 3.07 1.29 High Hedonic 3.28 1.45 Utilitarian 3.23 1.43
Pairwise Comparisons for WTB
Fractal Dimension Fractal Dimension Mean Difference Sig.
Low Intermediate -.182 .520 High -.325 .263 Intermediate Low .182 .520 High -.143 .625 High Low .325 .263 Intermediate .143 .625
Product Type Product Type Mean Difference Sig.
Hedonic Utilitarian .081 .729
Utilitarian Hedonic -.081 .729
Means of Aesthetic Liking
Fractal Dimension Product Type Mean Standard Deviation
Low Hedonic 3.62 1.19 Utilitarian 3.49 1.39 Intermediate Hedonic 3.75 1.11 Utilitarian 3.76 1.27 High Hedonic 3.96 1.34 Utilitarian 3.39 1.14
Pairwise Comparisons for Aesthetic Liking
Fractal Dimension Fractal Dimension Mean Difference Sig.
Low Intermediate -.201 .410 High -.120 .630 Intermediate Low .201 .410 High .081 .749 High Low .120 .630 Intermediate -.081 .749
Product Type Product Type Mean Difference Sig.
Hedonic Utlitarian .229 .261
Utilitarian Hedonic -.229 .261
Means of Approach Behavior
Fractal Dimension Product Type Mean Standard Deviation
Low Hedonic 4.54 1.30 Utilitarian 4.81 1.30 Intermediate Hedonic 4.31 1.09 Utilitarian 4.64 1.02 High Hedonic 4.66 0.78 Utilitarian 4.69 1.27
Pairwise Comparisons for Approach Behavior
Fractal Dimension Fractal Dimension Mean Difference Sig.
Low Intermediate .201 .370 High -.001 .997 Intermediate Low -.201 .370 High -.202 .385 High Low .001 .997 Intermediate .202 .385
Product Type Product Type Mean Difference Sig.
Hedonic Utilitarian -.211 .260
34 4.2.2 Effect of Fractals and Product Type on Approach Behavior
Analyzing the mean values of approach behavior for the independent variables, fractal di-mension and product type, presented in Table 6, it can be seen that for a hedonic product a high level of fractal dimension leads to the highest level of approach behavior and for a utili-tarian it is a low level of fractal dimension.
A two-way ANOVA has been used to examine group differences in the mean values of approach behavior across the experimental conditions fractal dimension and product type. Moreover, it was tested whether the interaction between the two independent variables (frac-tal dimension and product type) has had an effect on approach behavior. Frac(frac-tal dimension (low = 0, intermediate = 1, high = 2) and product type (hedonic = 0, utilitarian = 1) served as the independent variables and approach behavior as the dependent variable. The results of the two-way ANOVA (Appendix F) show that the main effect of fractal dimension F(2, 147) = 0.53, p = .592 and product type F(1, 147) = 1.28, p = .260 on approach behavior are not significant. Furthermore, the interaction effect between the two independent variables on approach behavior is not significant F(2, 147) = 0.21, p = .811. Thus, hypothesis 1b, which proposes that fractal-like image properties have a positive effect on approach behavior for both a hedonic and a utilitarian product, is rejected.
