• No results found

6.3 Methodology

6.3.1 Measuring Aesthetic Preference:

6.3.3. Measuring connectedness to nature

The following section outlines the research questions, rationale and hypotheses to be examined within this thesis in an attempt to add to the currently limited field of research exploring aesthetics responses to fractal patterns. This chapter will also specify the methods adopted in the thesis and the rationale behind these methodological choices. The methodology section will first explore the different types of established methods for measurement of aesthetic judgments/responses, secondly will explore the stimulus used within the study, the chapter will then examine the different methods to analysing complexity and fractal dimension. Finally measurements of ‘connectedness to nature’ will be discussed as a way of taking the research in an applied direction.

6.1 Rationale Summaries

Study One

Fractal Dimensions and Visual Complexity: An interrelated concept?

Visual complexity and fractal dimension have been considered distinct fields of perceptual stimulus, however this study aims to explore the relationships between the two related concepts. Study one of this thesis explores the relationship between the fractal stimuli and their associated fractal dimension (FD) developed for use in this thesis to measures of computational visual complexity obtained by analysing the fractal stimulus using the GIF ratio compression technique.

Study Two

Cross-cultural comparisons between UK and Egypt samples: Rating Scale Method

.

The aim of the second study used UK and Egyptian samples in a bid to explore cross-cultural preference for fractal complexity in addition to recreate Souief & Eysenck’s 1971 study exploring differences in complexity across the two cultures. Souief & Eysenck’s previously found that British people (with no art training) preferred complex figures but Egyptian people (with no art training) preferred the simple images. As visual complexity is significantly related to Fractal Dimension (FD) this study aims to test this hypothesis with new and improved methods of quantifying complexity. Eysenck and Souief (1971) did not believe that their data supported large aesthetic differences between cultures but instead believed that the findings point towards a universal preference and to unpick these findings further. The second study within this thesis considers the response of UK and Egyptian

aim is to explore in greater detail the impact of country/culture on visual preferences for fractal patterns.

Study Three

Validating the mid-range hypothesis for fractal preference

This study aimed to re-test an established theory of fractal preference, the mid- range hypothesis established by Taylor et al (2001). Taylor and colleagues found evidence that preferences for fractal patterns consistently fall within the mid-range of the fractal continuum (D1.3-1.5). With lower preferences being shown for the images at the higher (D1.7-1.9) or the lower end (D1.1-1.2) of the fractal continuum. To allow comparisons to be made and validate the current established thinking, study three also introduces two further models to understanding preference for fractal complexity including a linear model of preference (with a directional relationship) as well as an equalised-mid model (systematic grouping of fractal dimension instead of lower end weighing in Taylor’s model) to explore how well each model fit the preference data. The study adopts an online design, allowing participants from different countries and cultures to complete the study. This study aims to test if the mid-range preference hypothesis is stable across a wider international and cross-cultural sample adding support from Taylor and colleagues conclusions. Within the field there was a great need for the study, as so far the samples within the field of fractal aesthetics have been limited to WEIRD samples (Henrich, Heine & Norenzayan, 2010) meaning that the majority of data collection is done on Western, Educated, Industrialised, Rich and Democratic populations- it is the view of this author that if assertions are to be made about the universality of preferences, it is important to explore this from a cross-cultural and more varied sample.

Study Four

Optimal Fractal Preference; Stability across culture and within sub- cultural visual environments

Following from the cross-cultural differences found in study three. Study 4 aimed to explore if sub-cultural factors could be a powerful predictor of differences in preferences found in previous literature (see chapter 4). This study aims to explore not only a greater controlled cross-cultural sample but also explore sub-cultural differences looking at the differences between urban, rural and suburban classifications of the visual environment. Previous literature has found differences between aesthetic judgments of those living in Urban and Rural backgrounds, in addition the Mere-exposure hypothesis would suggest that the environment in which we live influences preferences. It is therefore hypothesised that the classification of a person’s environment can change preferences for peak level of fractal complexity. As those living in rural environments viewers are exposed to a high number of fractal and complex natural patterns it is proposed highest preferences will be reported for high FD/complex images. Alternatively those living in urban environments are exposed to mainly Euclidean and man-made shapes opposed to natural and commonly fractal patterns, therefore it is proposed preferences for higher complexity will be lower than the rural group.

Study Five

Connectedness of Nature & Environmental Classification

This study aimed to explore if our aesthetic responses to fractal patterns is related to how connected we feel to nature. Results of previous studies within this thesis suggest that individuals living in rural environments demonstrated higher preference for higher complexity/fractal patterns than those living in urban environments. Previous literature exploring landscape and aesthetics has shown that the environment in which we spend time and see regularly governs our

preferences. It is proposed that this difference in aesthetic judgment may result in differing opinion in how connected we feel towards nature. The study aims to explore if preference towards complex fractal patterns based on visual experience goes further than purely aesthetics response and instead has additional impact beyond aesthetics such as how connected, and as a result how likely we are to protect the natural environment.

Study Six

The relationship between Lifespan, Culture & Gender as predictors to Fractal Preference

The study aimed to explore the strength of the individual differences Age, Continent and Gender on preference for fractal patterns. Each was found as significant predictor model of preference in the previous studies with this thesis. Study 6 examines a combination of the entire data and one additional small set of ‘elderly’ participants to test the reliability of the age effects across a wider sample. Previous landscape research suggests that younger people have higher preference for busy and complex environments where as elderly people show less preference for ‘wild’ nature. Does this mean less preference for fractal patterns? The wider sample of participants allows more reliable contrasts between continents. This study aims to further test the complexity and mid-range models of fractal preference explored throughout previous studies within this thesis.

6.2 Hypotheses Table

Table 6.1- Thesis Hypothesis Table

Study

One

Fractal Dimension a component of Visual Complexity?

It is hypothesised that the fractal stimulus images used within the thesis will correlate significantly to GIF compression ratio scores; a computational measure of visual complexity. If confirmed this finding would suggest that fractal dimension can be considered as a related component or sub-component of visual complexity.

Study

Two

Cross-cultural Difference in Fractal Preference?

Mirroring the samples of Souief & Eysenck’s 1971 study exploring the cross-cultural stability of aesthetic preference with UK and Egyptian participants, this study hypothesises that responses for fractal patterns will demonstrate cross-cultural differences for non-art training participants. The study also hypothesises support the mid-range hypothesis with highest scores being awarded to images that lie within the D range of 1.3-1.5.

Study

Three

Re-testing the Mid-Range Hypothesis in Fractal Preference

 It is hypothesised that the overall frequency patterns of preference would display inverted-U shaped function, with heightened preference at the mid-range (D1.3-1.5).

There are three different models of aesthetic patterns explored in this study and as such three different experimental hypotheses:

 It is hypothesised that the variables Country, Age and Gender would significantly predict the mid-range model of preference

more so than the null model.

 It is hypothesised that the variables Country, Age and Gender would significantly predict linear the Complexity model of preference more so than the null model.

 It is hypothesised that the variables Country, Age and Gender would significantly predict Equalized Mid model of preference more so than the null model.

Study

Four

Cross & Sub-Cultural Differences in Fractal Preference

 It is hypothesised that the overall frequency patterns of preference would display inverted-U shaped function, with heightened preference at the mid-range (D1.3-1.5).

There are three different models of aesthetic patterns explored in this study and as such three different experimental hypotheses:

 It is hypothesised that the variables Country, Environment, Age and Gender would significantly predict the mid-range model of preference more so than the null model.

 It is hypothesised that the variables Country, Environment, Age and Gender would significantly predict linear the Complexity model of preference more so than the null model.

 It is hypothesised that the variables Country, Environment, Age and Gender would significantly predict the Equalized Mid model of preference more so than the null model.

Study

Five

Environment, Fractal Complexity and Connectedness to Nature

 It is hypothesised that the overall frequency patterns of preference would display inverted-U shaped function, with heightened preference at the mid-range (D1.3-1.5).

There are two different models of aesthetic patterns explored in this study and as such two different experimental hypotheses:

 It is hypothesised that the variables Connectedness-to-Nature Score, Environment, Age and Gender would significantly predict the mid-range model of preference more so than the null model.

 It is hypothesised that the variables Connectedness-to-Nature Score,, Environment, Age and Gender would significantly predict the Complexity model of preference more so than the null model.

Study

Six

Lifespan, Continent & Gender- predictors of fractal preference?

The final study combines all 2A-FC design data from this thesis with the addition of a sample of older participants responses.

 It is hypothesised that the overall frequency patterns of preference would display inverted-U shaped function, with heightened preference at the mid-range (D1.3-1.5).

There are two different models of aesthetic patterns explored in this study and as such two different experimental hypotheses:

 It is hypothesised that the variables Continent, Age and Gender would significantly predict the mid-range model of preference more so than the null model.

 It is hypothesised that the variables Continent, Age and Gender would significantly predict the Complexity model of preference more so than the null model.

6.3 Methodology

6.3.1 Measuring Aesthetic Preference:

Within the field, there are different established methods of measuring aesthetic judgment. There is vast variety within the field between the ways that researchers try to tap into aesthetic judgments of their participants. While this variety has continued to develop and grow the field often researchers do not outline clearly their rationale behind methodological choices (Augustin et al., 2012; Faerber et al., 2011). In an attempt to avoid this pit-fall, the following section will explore some the different methods used to elicit data about aesthetic judgment and justify the choices made within this thesis. Palmer, Schloss & Sammartino (2013) reviewed the current states of aesthetics and human preference, this paper provides a thorough summary of the methodological issues when measuring aesthetic responses. The section will use their outline as a structure from which to further explore the methodological choices available and used in the field.

Ratings:

Scales such as the likert scale (discrete) or line-mark rating (continual) methods is perhaps the most common way of eliciting aesthetic responses. The method allows researchers to show participants a series or single image allows collection of individual ratings for each based on a large sample. This method benefits from being able to collect data for a large number of images from a large number of participants in a short period of time, it is also a relatively simple task that can be altered to fit with the specific design, for example the vocabulary used when collecting scores is variable to include ‘liking’ ‘beauty’ ‘preference’ or behavioural choices such as ‘how likely would you be to visit this place’, ‘how likely would you be to buy this product’. This versatility means likert scales are widely used across various discipliners and research fields therefore results gathered this way can be comparable to others of similar design. Despite its wide use and versatility, problems can occur with consistency in scoring when using the rating method, particular at the start of trials. It has been suggested to over come

this issue, that a full range of stimulus should be shown to the participant ahead of rating therefore allowing participants to anchor responses in preferences ahead of the trials (Palmer et al, 2013). Other potential issues with this approach include the variance with scores between participants, there are trends of ratings with some choosing extreme ends of the scales and others being more modest with their scores, or clustering around the mid-points of the scale. We cannot truly conclude that the extreme scores show extreme preference responses more so than the modest responses. It must be acknowledged that choices made may be indicative of context or individual differences approach and personality differences in which the ratings are made (Ogden & Lo, 2011).

Scored from 0 to 10, how much do you like the above picture? (0 meaning extremely dislike, 10 meaning extreme like)

Ranking:

Rank ordering methods commonly involve a participant being given a set of stimulus to order from most to least preferred. The average rank given to each stimulus across the study is calculated and used as a measure of overall preference. The task is simple and something that is commonly experienced in daily life decision-making. Researchers believe that rank ordering offering a more reliable and valid measure than rating individual stimulus alone, and this is especially marked when pairwise ranking is used between 2 choices (Hochberg & Rabinovitch, 2000) as seen in the 2A-FC design discussed below.

While this method offers good and robust methods of collecting data, with the stimulus used consisting of 81 images, allowing participants to rank order these images for preference would be a difficult, complex and time consuming method of collecting preference data. Therefore ranking was not considered usable within this thesis.

Please order these images 1st, 2nd & 3rd in order of preferences. (1st = most liked, 3rd

= least liked)

Figure 6.2- Example of ’order’ aesthetic methodology with fractal stimulus.

2A-FC- Forced-choice:

From ranking a series of images, the forced choice method allowed the same process with smaller numbers of stimuli. Commonly the pairwise or two-alternate forced-choice method is used to unpick aesthetic responses. This method mirrors many behaviours in everyday life decisions in which we make preference choices

for a variety of different situations including which station to have on the radio or which piece of art to hang on our wall. Given this task is a commonality in daily experience, it is a relatively simple and understandable task for participants. The 2A-FC method was first used by Gustav Fechner (1860) during the first recorded empirical study of aesthetics; during this study participants were asked to choose from 2 version of Holbein’s ‘Madonna’. This method has been found to be particularly beneficial if the images are not overtly beautiful, therefore ‘beauty’ or ‘preference’ ratings or scores are unlikely to be accurate as they would be if using artistic or realistic photographic stimulus as used in a large range of studies. An alternate-choice design allows exploration of aesthetic preference and threshold specific information. Using a forced choice method allows regression models analysis, which can provide predictive or probability statistics for the likelihood of an aesthetic choice to be made.

The method could be critiqued for its inability to offer the magnitude of preference for the stimulus. If participants are making choices between two images, using this method it cannot be verified that a participant’s choice based on the stimulus being aesthetically pleasing rather than choices being based on strong/moderate dislike for the image not chosen. Despite the limitations, the findings offer one of the most controlled and suitable methods, and as such will be used (in conjunction with 1 ratings study) within the current thesis.

Which image do you like most? Tick/click/mark the one you like the most.

Production method:

The production method is a lesser-used method for exploring aesthetic responses. The method involves participants changing parameters in the image, whether that is the colours, shape or content of an image to explore individual’s ideal aesthetic worldview. The method is limited in participants artistic abilities and confidence, people are sometimes asked to draw or create something that appeals to them, their artistic talents or methods may not produce a piece that is aesthetically pleasing to them or others. Stimulus manipulation is a technique that has been used to develop the production method in aesthetic research and in recent years Chris McManus (UCL) have began using this method to crop photographs (McManus et al, 2011) or alter the proportions of Piet Mondrian’s (See Figure 6.4) painting to meet ‘optimal’ aesthetic experiences for the viewer. This adaption addresses many of the previous issues faced with the method and is enabled with new developments in technology. The current thesis uses a set of pre-defined fractal images controlled for FD, given the development of this took place ahead of collection, the stimulus do not currently allow this method to take place, therefore the production method was not included within this thesis as a method.

Physiological & Neuroaesthetics Measures:

The above methods have been useful in providing a wealth of evidence exploring aesthetic response to a variety of stimulus however all self-reported measures are limited with human judgment bias. Studying self-report behavioural methods which are open to error therefore, with ever increasing developments in technology, researchers have began to use other techniques to measure human responses to stimulus while avoiding the potential bias or errors from self-report measures. There are a variety of ways to infer preference using physiological measures; Galvanic skin response, Heart rate, Eye movements, EEG and fMRI are just a selection. These measures allow us to explore further than behavioural judgments and infer the potential physiological responses to a variety of stimuli.