20
Homepage aesthetics: The search for preference factors and the challenges of subjectivity Muzeyyen Pandir a , John Knight b, * a Department of Sociology, Lancaster University, Lancaster LA1 4YD, UK b User-Lab, UCE Birmingham Institute of Art and Design, Gosta Green, Birmingham B4 7DX, UK Received 8 February 2006; received in revised form 22 March 2006; accepted 24 March 2006 Available online 5 June 2006 Abstract While many studies have considered the usability of website homepages, subjective issues such as preference have been under explored. This paper describes a pilot study that investi- gates subjects’ preferences for different homepages. The study applies Berlyne’s theory of experimental aesthetics to website homepages. This theory suggests that there is an inverted-U shape relationship between preference for a stimulus and its complexity. Twelve subjects evaluated 12 homepages. The study used a ranking method to measure subjects’ pref- erences and the relationships between complexity, pleasure and interestingness. In addition, verbal reports were collected. No support was found for an inverted-U shape relationship and the findings indicate that complexity is not a predictor of pleasure. However, the results uncovered a number of subjective factors that underlie preference. These factors include indi- vidual differences in taste and lifestyle all of which are highly personal factors that change and develop over time. In addition, the findings suggest a link between interestingness and curios- ity. Lastly, the findings show an agreement on the judgements of complexity, and disagree- ment on aesthetic preferences. In conclusion, the paper points out the challenges faced in researching preference because of its highly subjective character. Ó 2006 Elsevier B.V. All rights reserved. Keywords: Aesthetic pleasure; Experimental aesthetics; Preference factors; Website preferences; Subjec- tivity 0953-5438/$ - see front matter Ó 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.intcom.2006.03.007 * Corresponding author. E-mail addresses: [email protected] (M. Pandir), [email protected] (J. Knight). www.elsevier.com/locate/intcom Interacting with Computers 18 (2006) 1351–1370

Homepage aesthetics: The search for preference factors and the challenges of subjectivity

Embed Size (px)

Citation preview

Page 1: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

www.elsevier.com/locate/intcom

Interacting with Computers 18 (2006) 1351–1370

Homepage aesthetics: The search for preferencefactors and the challenges of subjectivity

Muzeyyen Pandir a, John Knight b,*

a Department of Sociology, Lancaster University, Lancaster LA1 4YD, UKb User-Lab, UCE Birmingham Institute of Art and Design, Gosta Green, Birmingham B4 7DX, UK

Received 8 February 2006; received in revised form 22 March 2006; accepted 24 March 2006Available online 5 June 2006

Abstract

While many studies have considered the usability of website homepages, subjective issuessuch as preference have been under explored. This paper describes a pilot study that investi-gates subjects’ preferences for different homepages. The study applies Berlyne’s theory ofexperimental aesthetics to website homepages. This theory suggests that there is aninverted-U shape relationship between preference for a stimulus and its complexity. Twelvesubjects evaluated 12 homepages. The study used a ranking method to measure subjects’ pref-erences and the relationships between complexity, pleasure and interestingness. In addition,verbal reports were collected. No support was found for an inverted-U shape relationshipand the findings indicate that complexity is not a predictor of pleasure. However, the resultsuncovered a number of subjective factors that underlie preference. These factors include indi-vidual differences in taste and lifestyle all of which are highly personal factors that change anddevelop over time. In addition, the findings suggest a link between interestingness and curios-ity. Lastly, the findings show an agreement on the judgements of complexity, and disagree-ment on aesthetic preferences. In conclusion, the paper points out the challenges faced inresearching preference because of its highly subjective character.� 2006 Elsevier B.V. All rights reserved.

Keywords: Aesthetic pleasure; Experimental aesthetics; Preference factors; Website preferences; Subjec-tivity

0953-5438/$ - see front matter � 2006 Elsevier B.V. All rights reserved.

doi:10.1016/j.intcom.2006.03.007

* Corresponding author.E-mail addresses: [email protected] (M. Pandir), [email protected] (J. Knight).

Page 2: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1352 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

1. Introduction

Recent research in HCI has shifted the quality of the user experience awayfrom objective qualities (e.g. performance) to more subjective issues of usersatisfaction and enjoyment. These factors include emotion (Picard, 1997),pleasure (Jordan, 2000), fun (Blythe et al., 2003), engagability (Knight andPandir, 2005), ambiguity (Gaver and Beaver, 2003) and aesthetics (Tilghman,2004). The rationale for this shift is based on a growing awareness of theimportance of subjective factors in influencing preference in choosing productsand services.

Websites need to be usable but they must also engage users. In particular, web-site homepages give users their first impression of the site and thus are critical insetting up expectations and even influencing whether users return. Furthermore,given the millions of sites available to users, website owners are keen to understandthe ways to attract users, keep them on their site and provide an experience thatthey will want to repeat. A key way of achieving this is through effective homepagedesign. In ‘Homepage Usability: 50 Websites deconstructed’ Nielsen and Tahir(2001:1) go as far as to suggest that ‘Homepages are the most valuable real estatein the world’.

Studies in interface aesthetics are scarce and the few studies carried out alsohave limitations (Park et al., 2004). Some (Shenkman and Jonsson, 2000; Tractin-sky et al., 2000) suggest that visual appearance is important in users’ preferencefor websites. Thus, beautiful websites are preferred and are even perceived asbeing more usable. According to Tractinsky et al. (2000) ‘What is beautiful isusable’. How long this effect lasts is unknown and studies of e-commerce sites(e.g. Lohse, 1998) have to date not identified aesthetics as increasing conversionrates or sales.

The dearth of research leaves many questions unanswered. These questionsinclude how aesthetics affect interaction over time and how aesthetics interact withother qualities of use. The authors have a keen interest in trying to answer thesequestions and in order to begin to unpick the role of aesthetics in interaction, theauthors looked at identifying different approaches to the subject. From this research,a pilot study was carried out.

This pilot research aims to explore the aesthetic preferences of websites. The studyapplies Berlyne’s aesthetic theory (1974). This suggests that aesthetic pleasure is con-nected to variations in arousal potential. According to Berlyne, arousal potential islinked to the potency of collative properties such as complexity, novelty and ambi-guity, etc. Berlyne proposed a curvilinear relationship between pleasure and thesevariables.

In order to explore the correlation between these variables in websites, theauthors conducted a pilot experiment. Twelve subjects ranked 12 websitehomepages in terms of their complexity, pleasingness and interestingness.The participants also described the reasons for their choices as part of verbalreports. The experiment uncovered judgements on the visual appearance of thehomepages.

Page 3: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1353

2. Objective, subjective and holistic approaches to aesthetics

2.1. Objective definitions of aesthetics

Aesthetics has been dominated by intellectual speculation rather than empiricalinvestigation. Much of this speculation has centred on understanding and definingbeauty. In the West, beauty has been associated with the pleasure experiencedthrough interacting with works of art. Traditionally, beauty was thought to be anintrinsic property of an object and one of the central debates in aesthetics has beenwhether beauty is in the object or the subject.

The term ‘aesthetica’, deriving from the Greek term ‘aisthesis’ for sensory percep-tion, was coined by Baumgarten with the aim of forming a philosophical discipline(Shusterman, 2000:263-4). But it was Aristotle’s aesthetic judgements that dominat-ed Western thought and they were based on introspection. Aristotle proposed thatbeauty resided in the object and its properties such as order, symmetry and definite-ness (Hanfling, 1992). Hogarth, as a painter and engraver, searched for the relation-ship between beauty and regularity in his book Analysis of Beauty (Hanfling, 1992).He produced illustrations in which he manipulated objective qualities such as thestraightness or curvature of lines. He found that straight lines were the least satisfac-tory lines and argued that perfect regularity was not satisfying. For him, the curvedlines were more beautiful and he concluded that beauty was predicated by variety.Burke (Hanfling, 1992) also considered the importance of variety and described abeautiful object as small and smooth with a certain kind of variety, delicacy and col-oration. Arnheim also suggested that order without some complexity was monoto-nous (Lavie and Tractinsky, 2004).

2.2. Subjective definitions of aesthetics

The idea that beauty resides merely in the object’s characteristics (e.g. size and col-our) has been challenged by subjectivists. According to simple subjectivism, aestheticjudgement is determined by the pleasure or displeasure that perception of the objectarouses in any spectator (Ward, 1992). Subjectivism admits the possibility of differ-ences in judgements despite the objective properties of the object; one subject mayfind an object beautiful while another may find it ugly.

As a sophisticated subjectivist, Hume argued that aesthetics was subjective butgeneralisable. Sophisticated subjectivism argues that, ‘within a given area of the arts,some people are possessed of a high degree of discrimination and that there is anacceptable measure of agreement among them’ (Ward, 1992:244). Here, a highdegree of discrimination refers to high degree of sensibility. ‘On this account, anyaesthetic rules of composition that may exist will simply be empirical generalizations,based on the discovery that features of a recognizably similar kind have been foundto please discriminating spectators in a variety of different objects’ (Ward, 1992:245).

In his essay ‘Of the Standard of Taste’, Hume (1997) proposed a commonality ofaesthetic tastes, since people share common emotions and experiences. From thesecommonalities, he suggested a ‘standard of taste’ that unites people and concluded,

Page 4: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1354 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

‘the general principles of taste are uniform in human nature’ (Hume (1997):361).This is why:

‘The same Homer, who pleased at Athens and Rome 2000 years ago, is still

admired at Paris and at London. All the changes of climate, government, reli-gion, and language, have not been able to obscure his glory’ (Hume(1997):1997:354).

2.3. Holistic definitions of aesthetics

Some have argued that aesthetics combines elements that are holistic and combineobjective (e.g. colour) and subjective (e.g. preference) factors as well as physiologicaland psychological aspects of experience. Bell’s (1997:15) aesthetic judgements werebased on personal experience, particularly, ‘personal experiences of a particular emo-tion’ that echoes Dewey’s (1958) notion of the aesthetic as experience. Bell claimedthat works of visual art provoke a particular kind of emotion, an ‘aesthetic emotion’.He suggested that if common qualities of the objects that provoke these emotionscan be discovered, then it is possible to uncover the essential qualities in works ofart. He called this quality ‘significant form’, a combination of lines and colours thatmove people aesthetically and produce aesthetic emotion (Bell, 1997).

Bell’s suggestion of a causal relationship between significant form and aestheticemotion makes a distinction between aesthetic and nonaesthetic qualities. Sibley(1959) classifies nonaesthetic qualities as objective qualities such as lines, shapes orcolour, which can be perceived by anyone. Aesthetic qualities, however, are evalua-tive judgements based on several factors including sensibility, education and taste.Petit (cited in Tilghman, 2004) suggests that one may closely examine a paintingand may still not see elegance. This may be due to a lack of aesthetic sensitivityor knowledge or because of variations in taste. For this reason, Tilghman (2004) not-ed that judgements should be understood in a ‘context that includes the people, theircharacters, and histories, involved in the conversation in which the judgement ismade’.

Despite the differences between objective, subjective and holistic definitions of aes-thetics, research into the subject has centred on introspective and philosophicalaccounts. Shusterman (2000), drawing on Dewey (1958), suggested aesthetics shouldbe grounded in felt experience. Dewey and Shusterman characterised aesthetics as atemporal multimodal interaction between viewer, object and maker. Furthermore,by emphasising the holism of aesthetic experience, they paved the way for an aesthet-ic inquiry based on both philosophical introspection and empirical research into feltexperience.

3. Experimental aesthetics

Experimental aesthetics is a scientific approach to understanding aesthetic experi-ences. Experimenters, including Berlyne (1974), adapted the methods used by Fech-

Page 5: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1355

ner (1876). The most popular of these is the ‘method of choice’. Here, subjectschoose the most pleasing object from a selection presented to them by researchers.In addition, Fechner described the ‘method of production’, whereby subjects changethe characteristics of an object to make it more pleasing. Finally, Fechner suggestedthe ‘method of use’, where subjects examine a work of art in the whole on the basisthat it contains generic aesthetic elements.

Experiments in this field test the effects of isolated elements (e.g. triangles, cir-cles) or aesthetic and artistic qualities (e.g. complexity, novelty) of an object onhuman subjects. Dutton (2001) noted that empirical aesthetics sought ‘the exis-tence of universal aesthetic values’ from the empirical data collected from subjec-tive judgements. Incidently, Fechner’s experiments, with aesthetic stimuliinvestigations, were crucial to early developments of methods and knowledge ofcognitive phenomena. Daniel Berlyne became the leading exponent of experimen-tal aesthetics in the second half of the 20th century. Berlyne’s (1971) aestheticstheory proposed a Wundt-curve function (Bryman and Cramer, 1990) linkingpreference for any stimulus with the fluctuations in arousal. The theory suggestedthat:

‘. . .aesthetic patterns produce their hedonic effects by acting on arousal. . . posi-tive hedonic values can come about in either of two ways, namely through amoderate increase in arousal. . . or through a decrease in arousal when arousalreached an uncomfortably high level. . .’ (Berlyne, 1974:8)

.The theory proposed that moderate increments in the arousal potential of stimu-

lus were pleasurable. In contrast, sharp rises in arousal were experienced as beingunpleasant and punishing (Berlyne, 1974:9). Berlyne considered collative propertiesas the most important factors in arousal potential and thus for the aesthetic pleasure(Berlyne, 1971). These properties define the structural properties of the stimulus,such as complexity, ambiguity, novelty and familiarity. Berlyne called these proper-ties collative variables because the viewer perceives them through collating informa-tion from different sources (Crozier, 1994:62).

For Berlyne (1960), complexity was partly objective and partly subjective. It wasobjective because the physical properties of an object are the same for all subjects.It was subjective because its relative complexity could vary from subject to subject.Nevertheless, his approach to complexity positioned it as an objective property. Inmany experiments, Berlyne would create patterns himself and define their relativecomplexity according to the number of elements within them. This has led to somecriticism in determining the level of complexity at the expense of subjectivity (Cro-zier, 1994).

In using artificially created patterns Berlyne was also open to criticisms of valid-ity. For example, Munsinger and Kessen (1964); and Martindale et al. (1990) report-ed that semantic factors (meaningfulness) rather than collative variables were themost important determinants of aesthetic preference. Normore (1974) denied the solefunctioning of collative variables and suggested an interaction between collativevariables and other characteristics of a stimulus, including meaningfulness. Kaplanet al. (1972) suggested the role that content played in preference. They conducted

Page 6: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1356 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

experiments with photographs of rural and urban landscapes and found that, regard-less of complexity, the photographs of rural settings were preferred to urban ones.However, when each type of stimuli were analysed separately there was a slight posi-tive linear relationship between preference and complexity.

Other studies have found relations between preference, objective qualities and sub-jective factors. Eysenck (1942) linked preference factors to personality characteristicsand claimed that extraverts liked simplicity and introverts liked complexity. Barronand Welsh’s (1952) findings indicated that subjects who preferred simplicity were con-servative, organised and conventional compared to those who preferred complexity,who tended to be pessimistic, hostile, depreciative and admittedly creative. Eysenckand Castle (1970) found that artists preferred simplicity and non-artists complexity.Others claimed that different aesthetic responses for the same stimuli were related todemographic factors such as age, gender, education, life experiences, religious back-ground (Lee, 1998) and other social differences such as lifestyles (Ritterfeld, 2002).

Despite the criticism that Berlyne’s work has received, his many experiments sup-ported his theory of complexity. Furthermore, most criticisms of Berlyne have beenbased on theoretical disputes rather than confounding empirical findings. Lastly,growing interest in aesthetics in HCI suggested an empirical approach to under-standing subjective use qualities was most appropriate. In this light, the authorssought to test Berlyne’s theory through a pilot study on Website homepages.

4. Experiment

The authors aimed to investigate two issues. Firstly, how Berlyne’s theory relatedto website homepages. Secondly, the authors wanted to test whether an experimentalaesthetics approach was useful and relevant to design research. Rather than provingBerlyne right or wrong, the authors were interested in unpicking the aesthetic dimen-sion of design and providing practical guidance. Given that both the stimuli (home-pages) and method (cardsort ranking) differed from Berlyne’s approach, the authorsexpected different results from Berlyne’s. Unlike the abstracted nature of the stimuli(e.g. shapes and colours) used by Berlyne, the experiment used ‘naturally’ foundstimuli (e.g. pages from the World Wide Web. In terms of experimental design,the authors aimed to avoid the methodological criticisms of Berlyne’s work.

4.1. Participants

Twelve participants, five female and seven male, ranged in age from 24 to 54, wererecruited for the experiment. They were all graduates either in art and design, scienceand engineering or the humanities.

4.2. Stimuli

Homepages were chosen as stimuli for two reasons. Firstly, because homepagesare users’ entry point to websites. As such, it would seem plausible that aesthetics

Page 7: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1357

is a more important factor in homepages, in comparison to other web pages. Second-ly, unlike abstracted patterns, homepages are ‘real’ stimuli containing nonaestheticand aesthetic elements.

Twelve website homepages were included in the experiment. Each homepage wasprinted as a colour screenshot of equal size on A4 white paper and assigned a letterfrom A to L for identification. Paper screenshots were used so that subjects couldview all 12 stimuli at the same time and thus make a comparison. The websites werechosen from Macromedia’s ‘site of the day’ web page (www.macromedia.com/cfusion/showcase/) and are shown in Fig. 1. The names of the websites were keptsecret to avoid any influence.

Fig. 1. Twelve websites used as stimuli.

Page 8: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1358 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

4.3. Procedure

Each participant was assigned a time to take part in the experiment and all of themtook part in the experiment individually. Subjects were instructed to carry out a num-ber of tasks by a researcher using a script. The first task entailed participants complet-ing a paper questionnaire that collected demographic data. Then the screen shots weregiven to subjects face down and the subjects were asked to spread them out on a bigtable. They were asked to rank one set of homepage in descending order according totheir complexity. Having ordered the websites, they were asked to reflect on theirchoices and to write down their reasons for ranking the highest and lowest choice. Thisprocedure was repeated with subjects ranking the homepages in terms of interesting-ness and pleasingness and again giving and writing down verbal reports of theirchoices. These writings were used as the basis of verbal reports. The stimuli were ran-domised for each subject and the order of ranking the variables (complexity, pleasure,interestingness) were also randomised to avoid familiarisation.

4.4. Method of analysis

Each subject’s ranking order and verbal reports were collected. The rankingorders were then graded. Grading involved scoring each homepage for each subjectbetween one and 12. Homepages ranked the highest were given the highest score of12, while the websites ranked lowest were given the lowest score of 1. The gradeswere tabulated and the total for each website was calculated. Thus, each of the 12websites had a score of its complexity, pleasure and interestingness.

Statistical analysis was then carried out on the scores. Since the data obtainedfrom the experiment is ordinal (ranked), not normally distributed and related, non-parametric tests were used.

Although our initial aim was to explore the correlation between the variables ofthe experiment, we also became interested in seeing the agreement among the judg-ments. We realized that, finding out the most liked website was not enough, withoutknowing the degree of the agreement for it. Therefore, Kendall’s coefficient of con-cordance (W) was calculated to find out the degree of agreement among the judg-ments of pleasantness, complexity and interestingness of the websites. The resultof Kendall’s W changes between 0 for no agreement, to 1 for a perfect agreement.

After calculating the agreement, we also wanted to see, whether there was differ-ence among the websites; in other words, whether the websites were coming from thesame population or not. This was particularly important for complexity rankingssince we wanted to have stimuli varying in complexity degrees, so that we couldsee the relationship between different levels of complexity and pleasure. For this rea-son, the Friedman’s test, a nonparametric alternative to the analysis of variance, wasemployed to see if the ranking totals for each variable were different from each other.

Kendall’s W and Friedman’s test were applied and interpreted together because,although they have a close relationship, nevertheless, they reveal different facets ofthe data. For example, by applying the two tests, we expand the interpretation ofthe results. For example, Kendall’s W revealed a low agreement on pleasurable

Page 9: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1359

websites and in conjunction with Friedman’s test showed that the websites wereobserved to be different from each other. In this way, these two tests corroboratethe analysis of each other.

Correlations between pleasure and complexity, and also interestingness and com-plexity were then calculated. This was important for this research since the ‘measuresof correlation indicate both the strength and the direction of the relationship between apair of variables’ (Bryman and Cramer, 1990:162–3). Because the research is concernedwith the measures of rank correlation, we had to use the Spearman’s rank correlationcoefficient (rho), rather than the more common correlation test Pearson’s r (Brymanand Cramer, 1990:173). The interpretation of the result of Spearman’s rho changesbetween �1 for exactly negative relation, and + 1 for a perfect positive relation.

Finally, no formal analysis of the verbal reports was undertaken. Verbal reportsincluded single words and short narratives. The authors looked for similarity anddifference in the words used to rationalise choices. In addition, the authors soughtto find evidence of personal values and interests in choices in order to identify otherfactors affecting preferences.

4.5. Rationale for the method

Previous experiments (Berlyne, 1974) projected stimuli one at a time and subjectsevaluated each one using Osgood’s 7-point rating scales. This method was not usedin this study because it limits the ability to make comparisons. In contrast, the studyemployed a ranking method. This meant that the subjects were able to view all 12stimuli on a table in front of them and could compare them. The rankings of web-sites revealed not only the most liked or the most complex website, but also thejudgements made relatively to the others. By collecting verbal reports, the aim wasto find out other factors influencing aesthetic responses. This qualitative aspect ofthis experiment differentiates it from traditional quantitative approaches.

Another difference between this experiment and previous studies is how the colla-tive variables were measured. Berlyne determined the complexity of the stimuli him-self, whereas the current experiment asked participants to rank the homepages fromthe most complex to the simplest. Before the experiment, when selecting the stimuli,the authors included homepages of various degrees of complexity. We agree withSmets (1973) that this is necessary to be able to investigate the nature of the relation-ship between the variables. Therefore, the selection of the stimuli was made througha subjective scaling of the authors. However, their complexity levels were decided byparticipants in the process of ranking.

5. Results

5.1. Agreement among judgements

In order to calculate the agreement among participants’ judgements, Kendall’s W

was calculated. Complexity judgements were measured and the results showed a

Page 10: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1360 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

statistically significant high agreement (0.816, df = 11, p < 0.01). This shows thatsubjects agree on what is complex and what is not. The Friedman test was employedto see whether the websites were coming from the same population or not. TheFriedman test result revealed significant differences between each website’s rank totalfor complexity (chi-square = 107.76, df = 11, p < 0.01). This means that there was anagreement on what is complex and that homepages differ in their levels of complex-ity. Complexity can relate to the quantity of visual elements as well as the individualcomplexity of each element. In either case, the results suggest that complexity is anobject quality and that there is some subjective agreement of the level of complexity.So, with the Kendall test result, we see the agreement on the complexity levels of thewebsites. With the Friedman test result, we see that, for the subjects, the complexitylevel of each website differed from each other. Table 1 shows that subjects agreedwebsite E was the most complex.

In terms of pleasure, Kendall’s W test showed that the agreement on pleasurablewebsites was significantly low (0.193, df = 11, p < 0.01). Accordingly, the Friedmantest revealed that pleasure rankings for each website were significantly different (chi-square = 25.474, df = 11, p < 0.01). This means that none of the subjects had thesame preference for all 12 websites.

The agreement on interestingness was low but not statistically significant (0.110,df = 11, p > 0.05). The rankings of interestingness were also different but not signif-icant (chi-square = 14.474, df = 11, p > 0.05). Because there was no agreement onpleasure and interestingness, it appears from the means (see Tables 2 and 3) thatnone of the homepages were found particularly pleasing, displeasing, interestingor uninteresting. The mean score of pleasure and interestingness was 6.5. The rangeof scores, however, did not deviate from their mean. For example, when we look atthe means of the scores that the websites got from rankings, we see that the mostpleasurable website (I) had a mean score of 8.25, which is not much more thanthe total average. The mean score of the least interesting homepage (C), was 4.17and this is also not far below the total average. However, the most complex

Table 1Mean complexity rankings of the websites, the most complex at the top

Ranks Websites Mean SD Minimum Maximum

1 E 12.00 0.00 12.00 12.002 C 10.33 0.78 9.00 11.003 L 9.00 1.86 5.00 11.004 J 8.92 2.19 5.00 11.005 A 6.92 1.73 4.00 9.006 B 6.50 2.24 3.00 10.007 G 6.42 1.73 4.00 9.008 I 6.17 2.25 3.00 10.009 D 4.67 1.61 3.00 7.00

10 H 4.08 0.90 3.00 6.0011 F 1.83 0.39 1.00 2.0012 K 1.17 0.39 1.00 2.00Total 6.50 3.26

Page 11: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

Table 2Mean pleasure rankings of the websites, the most pleasurable at the top

Ranks Websites Mean SD Minimum Maximum

1 I 8.25 2.30 4.00 11.002 G 8.00 2.59 4.00 12.003 H 7.92 3.16 2.00 12.004 D 7.67 3.28 2.00 12.005 K 7.58 4.98 1.00 12.006 L 7.33 3.82 2.00 12.007 F 7.08 3.85 1.00 12.008 B 6.25 2.05 2.00 8.009 J 4.67 2.64 1.00 9.0010 C 4.67 3.45 1.00 12.0011 A 4.58 2.54 1.00 11.0012 E 4.00 3.05 1.00 9.00Total 6.50 1.58

Table 3Mean interestingness rankings of the websites, the most interesting at the top

Ranks Websites Mean SD Minimum Maximum

1 K 8.25 4.29 1.00 12.002 D 7.75 2.67 4.00 12.003 H 7.50 3.51 3.00 12.004 F 7.50 3.97 1.00 12.005 G 7.17 3.22 2.00 12.006 J 6.58 3.12 2.00 12.007 I 6.42 3.70 1.00 12.008 B 6.00 2.00 1.00 9.009 L 5.75 3.52 2.00 12.0010 E 5.75 4.33 1.00 12.0011 A 5.17 3.16 1.00 10.0012 C 4.17 2.55 1.00 9.00Total 6.50 1.19

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1361

homepage (E) was chosen by all of the subjects and had the highest mean score of 12,which is far above the total average (6.50).

In conclusion, none of the homepages had particularly high or low scores of plea-sure and interestingness and the Friedman variance of mean scores stayed at lowdegrees in comparison to complexity scores.

5.2. Correlations between the variables

In order to determine the correlation between variables, Spearman’s correlationanalysis was calculated. Between pleasure and complexity, there was a statisticallysignificant negative correlation (rho = �0.644, n = 12, p < 0.05). Fig. 2 shows therelationship between the mean ranked scores for pleasure and complexity. As the fig-ure shows, there is no inverted-U shape relationship between the two. Table 4 shows

Page 12: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

Fig. 2. Curves relating mean pleasure ratings with increasing complexity.

Table 4Contingency table presenting the observed frequencies of variables in terms of percentages

Complexity

Simple (%) Medium (%) High (%)

Pleasure 38.8 34.7 26.5Interestingness 39.7 31.7 28.5

1362 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

the distribution of pleasure scores according to the levels of complexity. By lookingat the distribution of complexity scores on the histogram, the scores between 0–4.99,5–8.99 and 9–12 were categorised as simple, medium and highly complex, respective-ly. Fig. 2 and Table 4 show that high scores of pleasure correlated with low andmedium levels of complexity. This indicates that simple and moderately complexwebsites were liked more. The most complex websites received the lowest scores ofpleasure. There is an outlier in the results, which is seen in Fig. 2. This was interpret-ed as an extreme sample and was left out from the relationship.

The correlation between complexity and interestingness was significantly high andnegative (rho = �0.867, n = 12, p < 0.01). Therefore, it appears that interestingnessdecreased as complexity increased. Fig. 3 shows this negative correlation. Table 4also shows the difference between the distributions of the scores between complexityand interestingness. The homepages with low degrees of complexity were found moreinteresting. Mid-ranged complex websites had middling degrees of interest whilehighly complex websites were chosen as the least interesting.

5.3. Verbal reports

Verbal reports from the subjects provided an indication of the factors influencingthe ranking process. According to these, the terms used to describe the properties ofthe most complex website were overpowering, confusing, intense, daunting and unor-dered. On the other hand, the descriptors of the least complex website were simple,

Page 13: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

Fig. 3. Curves relating mean interestingness ratings with increasing complexity.

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1363

unified and clean. The terms used for the most pleasing websites were balanced, clear,minimalist, stylish, intriguing and restrained. The ones used to describe the least pleas-ing websites were stressful, aggressive, complex, heavy, boring, incomplete and over-loaded. The qualities of the most interesting website were described as appealing,original, intriguing and minimal. Finally, the subjects described their choice of the leastinteresting website as unsurprising, too established, expected, usual and confusing.

6. Discussion

6.1. Agreement on complexity and variability in tastes

The high agreement on judgements indicates that complexity was perceived objec-tively. Most probably, complexity is related to the structural and nonaesthetic quali-ties, which are objectively perceived properties of the stimuli. This finding confirmsBerlyne’s definition of objective complexity. This finding also supports Hekkert andvan Wieringen (1996) who found that in general people have similar understandingsof complexity.

The low levels of agreement on pleasure and interestingness judgements, however,support the argument that preference is highly subjective and poses the question towhat degree can preference be generalised, if at all. The variation in subjects’ judge-ments indicates that there are significant differences in people’s tastes. Indeed, as Kors-meyer (2001) suggested, because there is more variation in taste, aesthetic judgements(e.g. pleasure, interestingness) diverge more than judgements of nonaesthetic qualities(e.g. complexity).

The findings regarding the variations in taste echo other studies (Jacobsen, 2004).Certain features of the stimuli, which were perceived as contributing to the home-pages’s beauty by one participant, were found displeasing by another. It was seen

Page 14: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1364 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

in the verbal reports that the reason for this was that subjects had different personalinterests and motivations and perceived the stimuli under these different influences.One of the homepages included a landscape and appealed to a subject who had aninterest in outdoor pursuits. A website that was perceived to have a distinctive1970s aesthetic was preferred by a subject who was found (from the verbal reports)to have a preference for retro. The homepages with a minimalist design appealed topeople who preferred simplicity. As a result, there was no agreement observed on thejudgements of pleasure, thus, it is impossible, at least within the frames of this study,to make generalisations and examine the aesthetic qualities of pleasing websites bylooking at their common qualities.

Our experiment was not designed to investigate individual differences as a prefer-ence factor. However, findings from the verbal reports provide us with informationabout the subjects’ personal interests and tastes. These findings point to differences ininterests and tastes and recall the findings of Ritterfeld (2002), who linked disagree-ment on aesthetic responses to social differences such as lifestyles. This indicates thatmore research focusing on the relationship between demographic factors, individualdifferences, taste and preference is needed. The difficulty of unpicking the personalfactors that impinge on preference is in turn complicated by the fact that it can oftenbe non-verbal. Indeed, clothes, possessions and objects can often express our tastesand preferences in a way that words can never do. This shift away from the con-sciously linguistic to the multisensory unconscious again, makes study difficult in thisarea.

While the question of taste is beyond the scope of this research, by looking at ourfindings, the authors believe that it determines people’s preference rather than objec-tive properties. For Kant (1952), aesthetic judgements (thus taste) are determined bythe complex interaction of feelings, viewers and objects. Since we should think ofviewers in terms of their individual characteristics (Lee, 1998), we suggest that tasteis formed over a period of time and long before the subject gets in contact with theobject. In addition, taste is often conditioned externally through peers and society.This means that the viewer brings as much to the object as the designer and that,although the designer may be able to influence the viewer, subjectivity and socialnorms limit how much preference can be manipulated. However, it is commonsensethat although people have different sensibilities there is a finite degree of difference,otherwise there would not be the discernable changes in fashion and preference forcertain goods; the challenge is to find how research can interpret these changes andwhether predictions can be made.

6.2. Correlations between variables

Overall, the correlations between the variables do not support Berlyne’s inverted-U theory. The relationship observed between the variables echoes Smets work, whichbacks the linear relationship between arousal and stimuli and concludes that ‘arousalmust be seen as a linear function of stimulus complexity’ (1973:22).

However, it must be stated that our findings partly do support Berlyne’s findings.For example, the most pleasurable websites had moderate levels of complexity and

Page 15: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1365

the least liked websites were the ones with high degrees of complexity. In contrast toBerlyne’s findings, it was observed that low complexity (simplicity) was also pre-ferred. The observed correlation between complexity and pleasure shows that com-plexity is not a strong factor in preference. The findings also show a negativecorrelation between complexity and interestingness. This is in contrast to Berlyne’sfindings (1971, 1974). However, Berlyne also suggested that people lose interest inhighly complex stimuli (Crozier, 1994:73), which chimes with the results since theleast interesting websites were highly complex websites.

Berlyne (1971) noted the positive relationship between pleasure and interesting-ness. In this experiment, nearly half of the subjects mentioned that they thought ‘in-terestingness’ was close to attractiveness and thus pleasingness. Therefore, it is notsurprising that pleasure and interestingness have similar relationships withcomplexity.

The authors were expecting that the use of a ranking method for measuringdegrees of complexity would yield different results from Berlyne’s. And, indeed,we observed different results. We concluded that the reason for this was not relatedto the differences in determining complexity levels. Berlyne made judgements himselfabout the relative complexity of stimuli. Although this approach is open to chargesof bias, the current experiment suggests that complexity is indeed an objective qual-ity. Thus, Berlyne’s method cannot be a problem. It would seem that the differentcorrelations found in the experiment were due to other subjective factors that are dis-cussed in relation to the verbal reports.

Smets (1973) pointed out that even if the investigators have the same conceptionof complexity, there might still be variations in the results concerning the relation-ship between complexity and preference. For example, if we accept Berlyne’s modelas being valid and use stimuli with comparatively low levels of complexity, then therewould be a monotonically increasing relationship, which would be representing thefirst half of the inverted-U relationship. In such cases, the ranking method, used inthis experiment may avoid such problems, since all the stimuli, even if they all havelow degrees of complexity, are ranked according to their relative degrees of simplicityand complexity.

The experiment investigated Berlyne’s collative variables on the basis of theirrobustness in his experimentation in the context of interface aesthetics. This a prioricategorisation has both positive and negative affects on the results. On the one hand,there is artificiality in using external categories (e.g. complexity) to describe phenom-ena that may be different from the subjects. On the other hand, it is perhaps difficultto verbalise aesthetic experience, which some argue (McCarthy and Wright, 2004)can be pre-linguistic, without pre-categorisation. Further research is necessary toinvestigate whether or not categories help in understanding aesthetic experiencesand whether these are best generated by researchers or elicited from subjects.

6.3. Verbal reports and observed factors of preferences

Verbal reports for pleasing websites recall Gestalt theory. As well as balance,which is a recognised quality in the theory for aesthetic pleasure (Berlyne,

Page 16: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1366 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

1960:241), orderly and clear designs were found favourable. This concurs with otherstudies (e.g. Chang et al., 2002; Fisher and Smith-Gratto, 1998, 1999; Parizotto-Ribeiro and Hammond, 2004) that suggest particular ‘Gestalt laws’ for creatingmore pleasing interfaces. Consistent with these studies is our findings that the sim-plest website homepages was also the most interesting website and the most complexwas the least pleasing. In this sense, there is a high degree of similarity between thewords used to describe the least pleasing website (stressful, complex, heavy and over-loaded) and the most complex website (intense, overpowering).

The verbal reports also indicated the important role that content plays in prefer-ence. Unlike the visual stimuli used by Berlyne, homepages contain not only shapesbut texts, photography and interactive controls. In their verbal reports, subjectsoften justified their preferences in terms of their personal interests. Homepagesincluding information relating to their interests were preferred to other sites. Thisfinding points to the important role of content in preference and in this sense it sup-ports the findings of Kaplan et al. (1972), who found that particular content (e.g.photographs of rural or urban places) was preferred regardless of its complexity.

There are also results supporting the findings of Martindale et al. (1990), whoclaimed that semantic factors are more important than collative variables in aestheticpreference. Some subjects explained that they did not like some of the homepagesbecause they did not understand what they meant. Similarly, it was also observed thatan unfavourable message might have a negative effect on pleasure. For example, onehomepage featuring a ring was found visually appealing. However, because its messagewas perceived as conveying weddings by one participant, it was considered ‘boring’. Asmuch as supporting semantic factors, this finding also supports the proposed linkbetween preference and individual differences. Weddings may be boring to some peo-ple, but can be of great interest to others. Given the difficulty of predicting people’sinterpretation and meaning of content, it is perhaps more pragmatic to focus on dis-crete groups of individuals rather than attempting to understand and design for thegeneral.

Another factor that came to light in the verbal reports was curiosity. Some subjectsmentioned that a lack of information made them curious and want to explore the sitemore. Berlyne (1971) defined curiosity as a form of motivation to learn more about asource of uncertainty, with the goal of gaining knowledge and reducing uncertainty.From this definition, one should not think of uncertainty as the sole factor of curiosity.This is because, people may not be curious about everything they do not have sufficientknowledge of. There needs to be a curiosity factor, other than uncertainty, to motivatepeople to explore. According to our experiment, this factor is interestingness. From thefindings we see that the most interesting homepage is the one that is ranked as the sim-plest homepage that contains the least information. This finding concurs with Berlyne’ssuggestion that patterns containing information, which cannot be absorbed immedi-ately but swiftly through perceptual and intellectual efforts, are judged interesting(1974:179). The relationship between curiosity and ambiguity is unclear. The verbalreports suggest that curiosity occurs when a tantalising trail of information is offeredand users want to find out more. Ambiguity, on the other hand suggests duality anduncertainty. From a usability perspective, ambiguity may not be as compelling in this

Page 17: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1367

context. Rather than the gradual revealing of meaning, Gaver et al. (2003:236), sug-gests that ambiguity leads to ‘multiple interpretations’.

Given that curiosity decreases with familiarity, it is unknown whether this influ-ences repeated exposure to stimuli. In this context, it is perhaps necessary forresearch in HCI aesthetics to consider how preference, predicated by curiosity chang-es with time and interaction.

Although the results highlight different factors such as content, semantics andcuriosity, these are all linked to subjectivity. Content featuring personal interests,websites with favourable messages and understandable meanings represent nothingother than different dimensions of individual differences externalised in the formof webpages. It may be that rather than looking for preference factors in the object,research needs to account for the different factors of subjectivity within the subject.

6.4. Some thoughts on the method and the challenges in experimental aesthetics

Because the initial aim of the experiment was to test the effect of formal qualities onpreference, the authors used a ranking method that would minimise subjective factors.Eysenck and Castle (1970) said that the ranking method eliminates individual differ-ences and had a statistical advantage. In addition, the ranking method is commonlyused in HCI research in the form of card-sorting and is easy and inexpensive to carryout.

During the experiment, the authors observed no signs of difficulty or boredomfrom the subjects. It is very likely that this was due to the small size of the stimuli(n = 12). However, it was observed that, while some subjects performed the rankingquickly in just a few minutes, others took their time, observed all the stimuli intense-ly, and deliberated over their choices. In some cases, subjects changed the rankingsduring the experiment. This shows that, some of the subjects did the ranking quickly,whereas some judged the stimuli carefully after close examination.

After reading the introduction about how to perform the experiment, the authorsdid not intervene with the subjects about their way of deciding the preference. It wasthought that this could affect their decisions. Therefore, it is not known whether thesubjects, who did the ranking with their first impressions, would change their choicesif they were told to consider the stimuli a second time. We believe that this behaviourof deciding on the preferences is also linked to individual characteristics. Some peo-ple know what they like and make their decisions quickly. Some like to take theirtime and make a well-judged decision.

The ranking method is also useful when the nature of the relationships betweenthe variables is being explored. For this research, we needed to include stimuli vary-ing in complexity. Although the authors tried to provide this variety subjectivelywhile choosing the stimuli, the method nevertheless provides this variety as it ranksthe complexity of the stimuli. Even if the stimuli do not differ from each other sig-nificantly, still, rankings provide three different values—low (simple), medium, high(complex). For this reason, the ranking method provides an opportunity, for study-ing Berlyne’s model in particular, to observe the functioning of low, medium andhigh complexity as a preference factor.

Page 18: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1368 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

However, the biggest challenge that the authors faced during this research was theinfluence of individual differences that were observed in the preference judgements.The ranking method is useful as it eliminates individual differences. However, guess-ing that the preference factors could be subjective, we decided to include verbalreports to the experiment, to see the other factors at work. Finally, the verbal reportsprovided us with necessary explanations for subjects’ responses. This shows the needto tailor the experimental design in order to include individual differences, althoughit also brings some challenges. In conclusion, the methods seem to be robust and use-ful for investigation aesthetic issues in interface design.

Due to the size of this pilot research, it may be unwise to attempt to generalise thefindings. It is planned to repeat the experiment with a greater number of subjects andto explore the role of the other factors. Moreover, allowing an interaction betweenthe subjects and the homepages would produce more complex but more reliableresults. It is plausible that while traditional usability studies can be carried out withrelatively few subjects, research into preference requires substantially more data toidentify trends, if indeed there are useful and discernable trends to discover. Perhapsa more pragmatic approach is to see how this type of research can help designersmake good decisions. Rather than looking for universal laws of aesthetics, theresearch methods used in this experiment suggest that they could be useful in solvingdesign trade-offs within a discrete user population.

7. Conclusion

The study started with the assumption that, if we could find the common qualitiesin the most liked objects, then we could identify the components of beauty. Wedecided to start from Berlyne’s proposal, to see whether complexity was really a fac-tor of preference.

Having seen that Berlyne’s theory did not work exactly, we tried to search for otherfactors in the verbal reports the subjects provided. There was no agreement on themost liked stimuli and there was no particular common factor for preferences either.But most of the descriptions pointed to the importance of individual differences,personal interests and subjectivity.

The existence of subjective factors resulted in preferences for different websitesand this caused the disagreement on what is pleasing. Therefore, for this experiment,it has not been possible to make generalisations on the aesthetic qualities that pro-duce pleasure for everyone. There was a strong agreement on what is complex. How-ever, complexity was not observed as a predictor for preference. Rather, semantics(as suggested by Martindale et al., 1990) and content (as suggested by Kaplanet al., 1972) appeared to be the factors acting on pleasure. And curiosity wasobserved as a factor for influencing interestingness.

The study considers these factors as different dimensions of subjectivity. Thus, thestudy emphasises the influence of individual differences, personal interests and tasteon preference judgements. Consequently, this study suggests the need for an integrat-ed investigation of subjectivity, aesthetics and interaction.

Page 19: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370 1369

References

Barron, F., Welsh, S., 1952. Artistic perception as a possible factor in personality style: its measurement bya figure preference test. The Journal of Psychology 33, 199–203.

Bell, C., 1997. The aesthetic hypothesis. In: Feagin, S., Maynard, P. (Eds.), Aesthetics. Oxford UniversityPress, Oxford, pp. 15–23.

Berlyne, D.E., 1960. Conflict, Arousal and Curiosity. McGraw-Hill Book Company, London.Berlyne, D.E., 1971. Aesthetics and Psychobiology. Meredith Corporation, New York.Berlyne, D.E., 1974. Studies in the New Experimental Aesthetics—steps toward an Objective Psychology

of Aesthetic Appreciation. Hemisphere Publishing, Washington.Blythe, M.A., Overbeeke, K., Monk, A.F., Wright, P.C., 2003. Funology: From Usability to Enjoyment—

Human–Computer Interaction Series 3. Kluwer, Dordrecht, The Netherlands.Bryman, A., Cramer, D., 1990. Qualitative Data Analysis for Social Scientists. Routledge, London.Chang, D., Dooley, L., Tuovinen, J.E., 2002. Gestalt theory in visual screen design: A new look at an old subject.

Proceedings of the seventh World Conference on Computers in Education, vol. 8. Denmark, pp. 5–12.Crozier, R., 1994. Manufactured Pleasures: Psychological Responses to Design. Manchester University

Press, Manchester.Dewey, J., 1958. Art as Experience. Capricorn Books, New York.Dutton, D., 2001. Aesthetic universals. In: Gaut, B., Lopes, D.M. (Eds.), The Routledge Companion to

Aesthetics. Routledge, London, pp. 203–215.Eysenck, H.J., 1942. The empirical study of good gestalt: a new approach. Psychological Review 48, 344–364.Eysenck, H.J., Castle, M., 1970. Training in art as a factor in the determination of preference judgements

for polygons. British Journal of Psychology 61 (1), 65–81.Fechner, G., 1876. Vorschule der Asthetik. Liepzig. Breithof and Hartel.Fisher, M., Smith-Gratto, K., 1998. Gestalt theory: a foundation for instructional screen design. Journal

of Educational Technology Systems 27 (4), 361–371.Gaver, W., Beaver, J., 2003. Ambiguity as a resource for design. In: Bellotti, V., Erickson, T. (Eds.),

Proceedings of CHI 2003. Ft. Lauderdale, Florida. ACM Press, New York www.equator.ac.uk/var/uploads/2002-gaver-0.pdf.

Hanfling, O., 1992. Aesthetic qualities. In: Hanfling, O. (Ed.), Philosophical Aesthetics: An Introduction.Blackwell Publishers, Oxford, pp. 41–73.

Hekkert, P., van Wieringen, P.C.W., 1996. Beauty in the eye of expert and nonexpert beholders: a study inthe appraisal of art. American Journal of Psychology 109 (3), 389–407.

Hume, D., 1997. Of the standard of taste. In: Feagin, S., Maynard, P. (Eds.), Aesthetics. OxfordUniversity Press, Oxford, pp. 350–364.

Jacobsen, T., 2004. Individual and group modelling of aesthetic judgment strategies. British Journal ofPsychology 95 (1), 41–56.

Jordan, P.W., 2000. Designing Pleasurable Products. Taylor & Francis, London.Kant, I., 1952. The Critique of Judgement. (J.C. Meredith, Trans.). Oxford University Press, Oxford.Kaplan, S., Kaplan, R., Wendt, S., 1972. Rated preference and complexity for natural and urban visual

material. Perception and Psychophysics 12, 354–356.Knight, J., Pandir M., Harte, D., 2005. Assessing the Engagability of Online Courses: The case of

Professional Studies Curriculum. In: Iasıs, P., Nunes, M.B., Palma Dos Reis, A. (Eds.), Proceedings ofIADIS Virtual Multi Conference on Computer Science and Information Systems. Portugal.

Korsmeyer, C., 2001. Taste. In: Gaut, B., Lopes, D.M. (Eds.), The Routledge Companion to Aesthetics.Routledge, London, pp. 193–202.

Lavie, T., Tractinsky, N., 2004. Assessing dimensions of perceived visual aesthetics of web sites.International Journal of Human–Computer Studies 60 (3), 269–298.

Lee, K., 1998. The measurement evaluation of emotion through understanding emotion and sensibility.Journal of Emotion and Sensibility 1 (1), 113–122.

Lohse, G.L., Spiller, P., 1998. Quantifying the effect of user interface design features on cyberstore trafficand sales, in: Karat, C.M., Lund, A., Coutaz J., Karat, J. (Eds.), Proceedings of CHI98: SIGCHIconference on Human factors in computing systems, Los Angeles.

Page 20: Homepage aesthetics: The search for preference factors and the challenges of subjectivity

1370 M. Pandir, J. Knight / Interacting with Computers 18 (2006) 1351–1370

Macromedia Website, Site of the Day. Available Online: http://www.macromedia.com/cfusion/showcase/[Accessed: 28 July 2004].

Martindale, C., Moore, K., Borkum, J., 1990. Aesthetic preference: Anomalous findings for Berlyne’spsychobiological theory. American Journal of Psychology 103 (1), 53–80.

McCarthy, J., Wright, P., 2004. Technology as experience. MIT, Cambridge.Munsinger, H.L., Kessen, W., 1964. Uncertainty, structure and preference. Psychological Monographs 78

(9), 1–24 (Whole No. 586).Nielsen, J., Tahir, M., 2001. Homepage Usability; 50 Websites Deconstructed. New Riders Press.Normore, L.F., 1974. Verbal responses to visual sequences varying in uncertainty level. In: Berlyne, D.E.

(Ed.), Studies in the New Experimental Aesthetics—Steps Toward an Objective Psychology ofAesthetic Appreciation. Hemisphere Publishing, Washington, pp. 109–120.

Parizotto-Ribeiro, R., Hammond, N., 2004. What is aesthetics anyway? Investigating the use of the designprinciples, Proceedings of the NordiCHI 2004 Workshop on Aesthetic Approaches to Human-Computer Interaction, Tampere, Finland, pp. 37–40.

Park, S., Choi, D., Kim, J., 2004. Critical factors for the aesthetic fidelity of web pages: empirical studieswith professional web designers and users. Interacting with Computers 16, 351–376.

Picard, R., 1997. Affective Computing. MIT Press, Cambridge, MA.Ritterfeld, U., 2002. Social heuristics in interior design preferences. Journal of Environmental Psychology

22 (4), 369–386.Shenkman, B.O., Jonsson, F., 2000. Aesthetics and preferences of web pages. Behaviour & Information

Technology 19 (5), 367–377.Shusterman, R., 2000. Pragmatist aesthetics— living beauty, rethinking art, second ed. Rowman &

Littlefield Publishers, Oxford.Sibley, F., 1959. Aesthetic concepts. International Philosophical Review 68, 421–450.Smets, G., 1973. Aesthetic judgment and arousal: an experimental contribution to psycho-aesthetics.

Leuven University Press, Belgium.Tilghman, B.R., 2004. Reflections on aesthetic judgement. British Journal of Aesthetics 44 (3), 248–260.Tractinsky, N., Katz, A.S., Ikar, D., 2000. What is beautiful is usable. Interacting with Computers 13,

127–145.Ward, A., 1992. Simple and sophisticated subjectivism. In: Cooper, D. (Ed.), Blackwell Companions to

Philosophy—A Companion To Aesthetics. Blackwell, Oxford, pp. 244–245.