Visualhttp://vcj.sagepub.

com/ Communication

The use of eye-tracking and retrospective interviews to study teenagers' exposure to online advertising
Kerstin Gidlf, Nils Holmberg and Helena Sandberg Visual Communication 2012 11: 329 DOI: 10.1177/1470357212446412 The online version of this article can be found at: http://vcj.sagepub.com/content/11/3/329

Published by:
http://www.sagepublications.com

Additional services and information for Visual Communication can be found at: Email Alerts: http://vcj.sagepub.com/cgi/alerts Subscriptions: http://vcj.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - Jul 30, 2012 What is This?

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

2012

446412

VCJ11310.1177/1470357212446412Gidlf et al.: The use of eye tracking and retrospective interviewsVisual Communication

visual communication

ARTICLE

The use of eye-tracking and retrospective interviews to study teenagers exposure to online advertising

KERSTIN GIDLF, NILS HOLMBERG AND HELENA SANDBERG Lund University, Sweden

ABSTRACT

This study investigated (1) potential exposure, (2) actual exposure, and (3) perceived exposure to online advertising in Swedish 15-year-olds. Eye movements of these teenagers were measured while they surfed on the internet for 15 minutes. The results show that the teenagers were potentially exposed to 132 advertisements during this time. The actual exposure was 10 per cent of all potential advertisements. A mixed effect model analysis indicates that position and size of advertisements are important factors influencing the teenagers visual attention to advertising, whereas subject gender did not have any significant effect. A retrospective interview based on previously recorded eye-tracking data revealed that there was a substantial difference between the teenagers actual and perceived exposure to advertisements, and that they were mainly unaware of their actual exposure. The retrospective interviews also showed that the subjects had difficulties in identifying the advertisements and advertised product, as well as the advertiser.
KEYWORDS

advertising exposure eye-tracking internet retrospective interviews teenager

INTRODUCTION

Many people in Sweden have recently voiced a need to regulate internet advertising aimed at children in order to protect them in their role as consumers. Before arguing for internet regulation, we need a deeper understanding of childrens and teenagers potential exposure to internet advertisements: the total amount and type of advertisements in the media context; actual exposure (the
SAGE Publications (Los Angeles, London, New Delhi, Singapore and Washington DC: http://vcj.sagepub.com) Copyright The Author(s), 2012. Reprints and permissions: http://www.sagepub.co.uk/journalspermissions.nav/ Vol 11(3): 329345 DOI 10.1177/1470357212446412
Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

true amount of visual attention to advertisements); and perceived exposure (the individuals idea and comprehension of exposure to advertisements). The study reported here was designed to fill this knowledge gap. As far as we know, there are few published studies on childrens exposure to and visual interaction with internet advertising (Neeley, 2007; Sandberg et al., 2011). In our previous studies (e.g. Ekstrm and Sandberg, 2010; Sandberg and Ekstrm, 2008), younger children (up to age 12) were described as a particularly attractive consumer group for advertisers, but at the same time they are a group in need of special consideration and care when framing and designing commercial messages. In the current study, we have chosen to focus on teenagers, who as a group are less often identified as a priority in media and advertising research (Livingstone and Helsper, 2006), but who use new information and communication technology extensively, particularly the internet. On this basis, the population studied here is limited to boys and girls aged 1416, grade 9 in the Swedish school system. The study focuses on mapping out the advertisement landscape on the websites visited by the teenagers, as well as measuring the teenagers exposure to and perceptions of advertisements on the internet. It is an observational and explorative study not driven by hypotheses but with the intention of formulating hypotheses for future research. The research questions are: RQ1: How many and what kind of advertisements are Swedish teenagers exposed to on their favourite websites (potential exposure content analysis of webpages)? RQ2: What is the amount of visual attention that teenagers give to internet advertisements (actual exposure eye tracking)? RQ3: How do graphic and layout properties of advertisements affect the amount of teenagers visual attention (actual exposure eye tracking)? RQ4: If teenagers look at the advertisements, how do they understand the exposure (perceived exposure retrospective interviews)?
BACKGROUND

Childrens media use has changed dramatically over the last decade, with the increased use of computer games, mobile phones and especially the internet. The digitization of media use or the move from traditional media to digital media has been referred to, in some instances, as a paradigm shift. In Sweden, children (aged 914) typically spend an average of 87 minutes daily on the internet, while the older youth group (aged 1524) spend approximately 167 minutes a day online (NORDICOM-Sweden, 2011). The internet has allowed an expanded reach of advertising by incorporating new ways of targeting young audiences (Neeley, 2007). Studies suggest that 98 per cent of childrens websites permit advertising, and two-thirds of the sites made for children rely on advertising for their primary revenue (Moore, 2006).

330

Visual Communication 11(3)

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

As advertising effects are largely based on visual stimuli, visual exposure (potential and actual) to a communication effort is a necessary but not sufficient condition for an effect to occur. The individual must pay attention to the message, cognitively process the content and then act on the basis of the evaluation of the message. However, in comparison to adults, less developed prefrontal control in children may make them more prone to exposure due to less developed ability to inhibit involuntary eye movements (Fukushima et al., 2000; Kramer et al., 2005). The current body of advertising research is immense and complex, and there is still no consensus on the influence that advertising has on children. Advertising effect studies vary greatly with respect to paradigms, perspectives and methodology. Three main dimensions are usually involved when studying the impact of advertising on children: (1) cognitive, (2) affective, and (3) behavioural effects (Valkenburg, 2000). Cognitive effect studies concentrate on childrens ability to differentiate commercials from other media content and their capacity for understanding the purpose of advertisements. Affective studies focus on the appeal of commercials and the trust they generate in children, while studies of behavioural effects examine the extent to which children are persuaded, usually measured by their preferences for a product or their requests to buy particular products (Valkenburg, 2000). In all three cases, advertising impact is defined as a product (outcome) measurement and usually measured by the use of survey techniques and interviews. However, impact may also be defined as a process measurement, answering questions such as: How does the individual visually interact with advertising? What happens during exposure? Eye-tracking (ET) is increasingly used in a number of research fields (Holmqvist et al., 2011). So far, however, researchers in media and communication studies have rarely used ET, although it is probably the most suitable method for studying advertising impact as a process. A positive relationship has been found between the number of fixations on an advertisement and its recall (Maughan et al., 2007). The same researchers also found that advertisements referred to as liked received more fixations. However, just because an area has been fixated on does not mean it has been remembered (Drze and Hussherr, 2003), but there is a high correlation between the amount of visual attention (dwell time) and memory recall (Wedel and Pieters, 2000). In a study of adolescents attention to responsibility messages in alcohol advertisements, Thomsen and Fulton (2007) found that only a relatively small percentage of the adolescents who fixated on the responsibility message at least once could recall its general concept. The recall of a web advertisement is associated with differences in navigation style. A person aimlessly browsing a website performs better in a recall and recognition test than a person searching for specific information (Pagendarm and Schaumburg, 2001). Web advertisements have also been found to hinder visual search (Burke et al., 2005).

Gidlf et al.: The use of eye tracking and retrospective interviews

331

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

Web advertisements are often disliked and therefore ignored. They also affect the perceived workload. Banners with flashing text, in particular, are perceived as frustrating and mentally demanding. In line with this, young peoples attitudes towards modern marketing such as digital advertising (online, email advertising, mobile phone-based SMS and MMS advertising) are more negative than their attitudes towards advertising in traditional media channels, e.g. print, cinema, outdoor advertising and television (Ekstrm and Sandberg, 2010). Web surfers seem to purposefully avoid looking at web advertisements (Drze and Hussherr, 2003). It might be assumed that more experienced users or younger people who grew up with the medium would be better at avoiding web advertisements, but Drze and Hussherr (2003) did not find any differences in the number of web advertisements looked at by young and older surfers or by experts and novices. In our study, we use ET to measure teenagers visual interaction with online advertising.
METHODS

In this section, we present the use of our two main methods ET and retrospective interviews and details of our data collection and analysis procedures. How does ET work? Subjects eyes are illuminated with infrared light whilst being filmed in order to localize the centre of the pupil and the reflection in the cornea. This measurement can be used to calculate the eyes movements because the relationship between the pupil and the cornea reflection changes depending on where the subject looks (Holmqvist et al., 2011). These recordings make it possible to identify periods of time when the eye is still (known as fixations) and the movements between these fixations (known as saccades). It is only during fixations that our eyes can gather information from a scene; during the saccades, we are blind. A fixation can last for about 80 milliseconds (ms) up to a few seconds (sec.). It only takes about 80 ms to read a word and understand its meaning (Rayner, 1998). The duration of a typical fixation is approximately 200300 ms. The movements of the eyes are very important because high visual resolution is limited to only a small part of the scene, about 2 of the visual field, although studies have shown that experts tend to have a larger functional visual field (or perceptual span) than novices and a more task-efficient selection of fixation points (Reingold et al., 2001).

Sample
The study was conducted in a city in southern Sweden. Schools were identified through the local school office and chosen to participate on the basis of municipality data. Two urban schools, out of a total of nine local secondary schools, were strategically selected to take part in the study. The purpose of

332

Visual Communication 11(3)

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

the sampling procedure was to secure variation among the subjects in terms of social background, ethnicity, media use and interests. One class of ninthgraders was selected from each of the two schools (A and B), making a total of 44 students. Parents of two of the students would not allow them to participate, and three dropped out because of illness or because they forgot their appointed times for the test and could not be rescheduled. Altogether there was a loss of five students from the total population of 44, so the study included 39 teenagers, 21 from school A, and 18 from school B. There were 20 girls and 19 boys. According to the classroom survey data compiled a couple of weeks before the ET study, all the teenagers used the internet. The vast majority of the teenagers (98%) had a computer with internet access at home; 43 per cent had computers of their own, and 69 per cent had internet access in their bedrooms; 77 per cent of the teenagers used the internet on a daily basis. More than a third of them used the internet 12 hours on weekdays; a third used it 23 hours on weekdays, and 16 per cent between 3 and 6 hours on weekdays. At weekends, their internet use increased overall, with 63 per cent using the internet more than 3 hours on any given Saturday or Sunday. In the survey questionnaire, we also asked the teenagers to list their favourite websites (not more than 10) to map out their everyday environment in cyberspace. One third of the respondents listed a maximum of three sites, 17 per cent reported as many as seven sites, and 14 per cent used the opportunity to fill in 10 sites. On average, each person indicated five websites. In total, 98 unique sites were reported as the teenagers most visited or preferred websites. There were web addresses to social forums (e.g. LunarStorm, eBuddy), media sites (papers and television networks/channels), search engines (e.g. Google), services (e.g. websites for local transportation, Ticnet, Amazon) and sites reflecting the teenagers personal interests (e.g. sports, music, fashion, celebrity gossip, environmental issues, theology, etc.).

Data collection and procedure

To facilitate teenagers participation, all data collection took place in the schools during school hours. The researchers left their university campus and set up a temporary laboratory incorporating their technical equipment in the schools. The ET setup, instructions for the subjects and experiment procedure for the upcoming data collection were tested on individuals from the designated target group in a pilot study that took place four months ahead of the data collection period; adjustments were made based on the outcomes of this pilot. The aim of the study was to capture the teenagers natural online behaviour. Thus we wanted as few restrictions as possible during the trials in order to increase the ecological validity of the study. The study consisted of two parts: the ET and the retrospective interview recording, conducted in two rooms adjacent to the classrooms.

Gidlf et al.: The use of eye tracking and retrospective interviews

333

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

Each teenager was welcomed into the room where the eye movement recording was taking place and asked to sit in front of the experiment computer. Once in place, the teenager read a leaflet with some background information and the conditions for taking part in the study. The chair and computer were adjusted so that the teenager could sit comfortably. A calibration of the equipment was completed before the eye movement recording began. Test instructions were then presented on the screen, and thereafter the teenager was allowed to ask questions about possible ambiguities before starting.

Instructions
Each teenager was given 15 minutes to surf the internet without restraints or particular tasks. To assist, a list of URLs was presented, which had been developed from the classroom survey data. A selection of 21 of the 98 websites made up the list for the ET test (the list of preset websites is presented in Sandberg et al., 2011). If the teenagers did not like any of the stipulated websites, they had the option of seeking any website they wanted through the Google search engine, which was included in the list. The software used did not permit subjects to enter a URL in the address bar of the web browser. While they were surfing, the teenagers were left alone in the room in order to create a relaxed unsupervised situation and to encourage the most natural online behaviour possible. When 15 minutes had passed, they were asked to complete their activities on the internet, and a short questionnaire on their activities was completed by the investigator. Questions were asked to estimate the consistency of the surfing activity, e.g. Have you visited sites you have never been on before? and Would you say that your activities reflect your everyday web activities or diverge in any way? They were then asked to go into the other room, where a retrospective interview was conducted.

Recording of eye-tracking data

The eye-tracker used in this study to measure potential and actual exposure was a remote SMI iView X 2.5 RED system, which means that it measured the eye movements by filming them from a distance and therefore required no physical contact with the person, which helped to improve ecological validity. This meant that the subject could be at ease and move relatively freely during the data collection. The eye-tracker is attached to a panel below the computer screen and is hard for an untrained eye to detect. The eye-tracking system is a video-based pupil and corneal reflex system, recording data at 60Hz. The screen resolution was set at 1680 1050, and the average viewing distance between screen and the participants was 700 mm. The experimental equipment consisted of a computer, which was used for the teenagers internet surfing as well as the recording of their eye movements, and Experiment Suite 360 from SensoMotoric Instruments, which was used for presentation of stimuli (instructions, list of URLs and browser),

334

Visual Communication 11(3)

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

measurement of eye movements and the presentation of eye movements during the retrospective interview.

Retrospective interviews
In order to assess the teenagers perceived exposure, we set up retrospective structured interviews based on the eye-movement data. As a way of helping them to verbalize their surfing experience, we used a structured interview method (not a cued one) asking concrete questions about the online advertisements. The purpose of the retrospective interview was to reconstruct the trials, making the teenager recall the process, choices and decision-making involved in their online activities. Another important purpose was to explore their awareness of the exposure as well as the reasons behind the attention they had paid to some advertisements but not to others. During these interviews, we also asked the teenagers to identify and actually point out the advertisements on the website. In the retrospective interview, the subject sat on a chair in front of a computer screen on which the personal eye-tracking data were presented. The retrospective interview began with questions about how the teenager had experienced the ET and their recall of the sites visited. The interviewer had a guide to follow to ensure that all important aspects were covered and that a between-subjects comparison could be made of the questions asked at each interview. The interview was filmed for later transcription and analysis. The setup of the video equipment was such that the teenager remained anonymous and the computer screen where the eye movements were presented was in focus. During the retrospective interview, one researcher conducted the interview and another was responsible for the technology (video recording and uploading of the subjects eye-movement data). Each interview lasted a little less than an hour.

Analysis of eye-tracking and coding

The ET data, corresponding to 9.86 hours of recording, were analysed with the software BeGaze from SMI. The web content was captured (recorded) at the time of the ET. In order to analyse the amount and type of advertising content that the teenagers looked at, the advertisements on the internet during the sessions first had to be identified as areas of interest (AOI), and coded into product categories. Internet advertising in this study was restricted to paid advertising space on the websites, such as banners (top or bottom), advertisements placed in boxes in the left or right margins, ad links and floating advertisements that usually appear in the middle of the website and follow the viewer scrolling down the page. The statistics of the compiled eye-movement data were then exported from BeGaze into a database for further analysis. Identified advertisements were coded into one of 11 categories: food and drink, ad links, gambling, technology, entertainment, lifestyle and recreation, banking, education and employment, PSA/NGO, diet/weight

Gidlf et al.: The use of eye tracking and retrospective interviews

335

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

loss and other (for further discussion of category descriptions, see Sandberg et al., 2011). Advertisements considered to be self-promotion for the particular website in use were not included in the analysis.
EMPIRICAL EXAMPLES

The recording session had been set to 15 minutes, but the actual average session time ended at 15.57 minutes. Each recording session was divided into smaller units, or trials. Each time a subject clicked on a link and a website was downloaded to the subjects web browser, a new trial was set up. A trial is thus roughly equivalent to a request. The fixations were calculated with the SMi event detection algorithm designed for low speed sampling rate. According to the manufacturer of the ET system, the average precision error should be less than one visual degree. It is interesting to note that the average duration of fixation for the teenagers in this study is almost 407 ms, which is more than typical mean fixation durations during reading (225 ms), visual search (275 ms) and scene perception (330 ms) (Rayner, 1998). Long fixation times are often interpreted as an indication of visual information of high density, which requires longer fixation times for the viewer to process the content. This result suggests that the internet is a visually very demanding medium for teenagers aged 1416 (grade 9 in the Swedish school system), even though they are familiar with the internet environment and use it several hours per day. Advertising in this environment of high information density might require more concentration to identify and subsequently decode than advertisements on less information-rich media channels. In Figure 1, a scan path is presented, showing a teenagers eye movements during a trial. The round spots in the picture represent eye fixations, i.e. when the eyes are quite still, information is taken in and being processed; the larger the spot, the longer the fixation. The lines between the spots represent the saccades, the very fast eye movements between the fixations when the eye is practically blind. The results from the saccades, when the eye moves across the screen, show that the average saccade duration is 32 ms, which is quite normal (Rayner, 1998) and was expected. An interesting finding is the average saccade amplitude of nearly two visual degrees. This demonstrates that the average distance the eye moves during the internet surfing is about 3 cm on the computer screen. These larger saccades suggest a scanned and exploratory reading, where subjects actively integrate visual information from different places on the page.

Potential exposure RQ1: How many and what kind of advertisements are Swedish teenagers exposed to on their favourite websites?

336

Visual Communication 11(3)

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

Figure 1 A scanpath visualization generated from eye-tracking data. The picture shows the gaze behaviour of one teenager in a single trial, in this case a web page from http:// www.aftonbladet.se/. The red lines correspond to the eye-movements during the first second of the trial. Reproduced with the kind permission of Aftonbladet.

The teenagers visited a total of 121 unique websites and 935 unique URLs (pages), and we achieved a total of 3042 requests (page downloads) across all subjects in the study; 5161 advertisement instances were identified and assigned to one of 11 categories (Sandberg et al., 2011). The largest advertising category was ad links (33.8%), then technology (14.39%), gambling (11.15%) and lifestyle and recreation (10.51%). The total number of ads in this material (5161) indicates the total number of potential advertisement exposures in the study. However, not all these advertisements were in fact viewed by the teenagers, and therefore the potential exposures do not correspond to the actual or absolute advertising exposures. Over an entire session, the average number of possible or potential exposures of advertisements measured 132 (SD = 130.01), and for each trial there was an average of 4.65 advertisements on a web page.

Actual exposure RQ2: What is the amount of visual attention that teenagers direct to the internet advertisements?
The number of actual exposures during a surfing session was no more than 13.69 advertisements (SD = 11.61). This result means that the teenagers

Gidlf et al.: The use of eye tracking and retrospective interviews

337

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

fixated on, and visually paid attention to, about 10 per cent of the potential number of advertising exposures. The total effective (absolute) exposure time for these advertisements, over the entire session, was on average 14.5 seconds per session (SD = 15.2). The average time for an advertisement look was 0.5 seconds, or slightly over 1 fixation of average duration (406.88 ms), which is more than it takes to decode a very simple message such as a word, or to get the gist of a scene. On average, each teenager completed 76.89 requests (downloads) during a session, and each request lasted on average 18.45 seconds. The minimum time a person spent on any particular website was 0.14 seconds. The maximum time spent on a site was 279.29 seconds, or 4 minutes and 39 seconds. These figures clearly demonstrate the range in time for which a certain website is displayed, and therefore it also tells us something about how long advertisements are displayed on the website. Eye-tracking data also help us to address questions about the efficiency or impact of advertising. Some of the advertisement categories occurred frequently in the material but still got a low total dwell time, which could be interpreted as a low impact for those categories. On the other hand, few advertisement instances and longer total dwell time could be interpreted as high impact. Thus, one way to operationalize advertisement impact is to measure the relative dwell time on advertisements by dividing the aggregated total dwell time of a specific advertisement category with the total number of advertisements in the same category. Our data suggest that food and drink is the advertisement category that gives rise to the highest impact of the online advertisement categories in this study (Sandberg et al., 2011). RQ3: How do graphic and layout properties of advertisements affect the amount of teenagers visual attention? In order to study the effects of advertisement size, position and subject gender on actual exposure measured as dwell time (ms), a data set was prepared where observations of advertisement objects receiving zero dwell time were removed from the data. This resulted in a data set with 520 cases, where each case represented an exposure to an advertisement. Advertisement size was measured in square pixels, and advertisement position was coded as five categorical values: top, left, middle, right and bottom. The top category corresponded to banner advertisements spanning the entire width of the top of the web page. The left, middle and right categories corresponded to the three major content columns of the web page, where the middle column is mainly used for displaying editorial material (text, pictures), and the right column represents the prototypical advertising column. Observations coded with the bottom category were subsequently removed because of the relative scarcity of this category (approximately 1%).

338

Visual Communication 11(3)

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

Since the distribution of dwell times on areas of interest typically has a strong positive skew, with the majority of advertisement objects getting very low values, the dwell times were log transformed. After this transformation, the data showed an improved normal distribution. Further, the data were tested for outliers, using box plots. In the inferential statistics analysis, a linear mixed effects model was applied to the data (the lme4 package in the statistics software R), using log transformed dwell time as the dependent variable. This model treated advertisement size, position and subject gender as fixed factors, whereas subject was treated as a random factor with observation/trial as a nested within-subjects variable. The advertisement position category right was judged to be the most prototypical advertisement position, so this position was treated as a baseline, against which the effects of the other factors were measured. The results of the analysis indicated that, compared to the right position baseline, middle and left position had a significant effect on dwell time (t = 1.59 and t = 4.26, respectively). Advertisement size also had a positive effect on dwell time relative to baseline (t = 3.54), whereas subject gender had no significant effect. An interpretation of these results would be that atypical advertisement positions (i.e. advertisements in the left column, or advertisements interleaved with editorial material in the middle column) have a positive effect on advertisement dwell time compared to the more default advertisement position in the far right column of the web page. Advertisement size also has a positive effect on dwell time, but this effect could be partly attributed to the fact that ads occupying a larger portion of the screen have a larger generic probability of being fixated. To investigate further the effect of advertisement size factor, this probability would have to be compensated for (cf. Simola, 2011).

Perceived exposure RQ4: If the children look at the advertisements, how do they perceive the exposure?
The ET results presented above suggest that 90 per cent of the advertisements are not heeded by the teenagers. The retrospective interviews revealed that the teenagers were fairly aware of where advertising is most often placed on the websites. At the same time, however, they seemed to be unable to distinguish advertising content from non-advertising content. In several cases, they also found it difficult to identify the source of the advertisement and the product being promoted. The following quotations will show how this came out in the interviews. Example 1 (boy, school A) I: So, what is this then? R: Is that also an advertisement?

Gidlf et al.: The use of eye tracking and retrospective interviews

339

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

I: Yeah indoor golf course Do you know what they are advertising here? R: No. I: No? How do you know then that it is an advertisement? R: Well, its like those pretty big banners at the side that has nothing to do with the rest, kind of I: Mm, because here you have no apparent advertiser on this one. Is it something you usually look for, the advertiser or? R: No, not really. I: Is it the placement you go for? R: They usually are at the same place on all sites. Most of the teenagers experienced the advertising online as a nuisance. They were irritated, annoyed, and some were also upset about it for various reasons. There was also some scepticism and distrust of online advertising. This can be an important reason why the teenagers deliberately avoid advertising as much as they can. In the interviews, the teenagers also described strategies they used to avoid the advertisements online, e.g. they consistently look away, deliberately try to stay alert and avoid being seduced by concentrating on the task, and they hold their hands up, blocking the screen to avoid seeing the advertisements. Others declared that they immediately click to a new link or simply close down the advertisements that are obtrusive. A total of 76 per cent of the respondents reported in the survey that they avoid advertising on the internet (Ekstrm and Sandberg, 2010). It is clear that teenagers have a huge resistance to internet advertising. Example 2 (boy, school B) I: Do you remember looking at it? (referring to an advertisement on the screen) R: No, not that much. I: No? R: I dont pay attention to them. Its like its not worth it. I: Not worth it? R: Its often like these crappy things. Although the teenagers are somewhat aware of advertising placement and although they have negative attitudes about it, the results of the eye-movement recording evidenced that they actually look at advertising to a certain extent. The amount of online advertising is very large in relation to how much the teenagers actually note (about 10%). The retrospective interviews showed that the teenagers are not fully aware of the advertisements they actually view. Several subjects had no recall of looking at a particular advertisement shortly after exposure.

340

Visual Communication 11(3)

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

Example 3 (girl, school A) I: We can see that you have looked, for quite a long time, at that guy up here (pointing to an advertisement on the screen). R: Oh, thats strange, I really dont recall looking there I: You dont? R: No. I: And not even now afterwards you remember looking at it? R: No, well maybe, vaguely, but its not like I have Well, I have only looked at it, like, briefly.
RELEVANCE

In this study, we investigated the processual as well as the product aspects of online advertising effects, operationalized as potential and actual exposure. Teenagers perceived exposure, which is a cognitive effect, was investigated by interviewing the individuals shortly after exposure. By measuring teenagers eye movements as they surfed the internet, we obtained a good comprehension of their subjective experiences of different websites. We measured to what extent they were looking at advertising, as well as the type of advertising they paid attention to, the time spent on advertising in relation to other media content, the number of advertisements, as well as their category and position in relation to the teenagers visual attention, etc. All these measurements are of high relevance when discussing the medias role and the power of advertising to influence our ideas, emotions and behaviours. The retrospective protocols revealed that there was a substantial difference between the childrens actual and perceived exposure to advertisements. First, the children proved to be uncertain about what web material contained commercial messages. Second, they perceived the advertisements as disturbing, not interesting, and not to be trusted. Third, the retrospective protocols revealed that the children were not aware of their actual exposure, as recorded by the ET equipment. The retrospective interviews were thus important for obtaining a deeper understanding of teenagers perceptions and visual processing of online advertising. Eye-measurements are needless to say important but they cannot give us the more nuanced and personalized knowledge of internet advertising processing. We therefore recommend a combination of the two methods.
LIMITATIONS

During the ET, some technical problems unfortunately occurred on websites containing online games. The program used for the recording could not handle all of the information in these games and therefore stopped on a few occasions. We then asked the teenagers not to play these games during the test; they could visit the sites but were asked not to start the games. Despite this, a few did not comply, and their recordings had to be started again, with a
Gidlf et al.: The use of eye tracking and retrospective interviews

341

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

little risk of data loss each time. Advertising in online games was therefore not measured in the study. Our aim in this study was to measure the teenagers eye movements while they performed their everyday activities on the web. Initially, we wanted to instruct the teenagers to pretend that they were at home and should carry out whatever they usually do on the web during the test. They were also informed that their personal codes would not be saved by the equipment. However, the test situation per se is such an unnatural setting for the children that such an instruction would seem ridiculous to them; consequently, the teenagers were encouraged only to surf unreservedly during the test and do whatever they normally do on the web. To compensate for this little remission, some questions were asked in connection with the test to estimate whether the surfing behaviour and websites visited were representative of their natural activities and web environment. As many as 82 per cent of the teenagers responded that the sites they had visited during the test were sites they already knew of and had visited before. Fewer than a fifth of the teenagers visited one site that was completely new to them. Half of the participants confirmed that, during the test, they visited the same websites they usually did at home. Those who said they would visit other websites if they were at home mentioned specific game sites and personal blogs as the only alternatives. It is therefore our assessment that the teenagers surfing activities during the tests are similar to their usual internet habits and thus the advertising they were exposed to during the test is not unlike their everyday exposure to online advertisements.
DISCUSSION

Our results suggest that the teenagers have developed strategies to efficiently avoid internet advertisements. This was indicated by both the retrospective interviews about perceived exposure and the ET measurement of actual exposure. The large majority of the online advertisements received no attention from the teenagers. The ET study indicates that during the 15-minute session of free surfing, the average teenager faced 132 potential exposures to advertisements. The number of actual exposures was slightly fewer than 14. The teenagers were exposed to approximately 10 per cent of the online advertisements, with an effective exposure time of 14.5 seconds on average. Compared to the time spent on non-advertising content, this represents only 1.6 per cent of the total time for the session (15 minutes), which strongly suggests that teenagers use avoidance strategies. The use of avoidance strategies could be explained by the teenagers internet surfing expertise and their use of parafoveal vision. The classroom survey data show that the teenagers are highly skilled internet users, spending several hours on the internet each day. With that amount of internet usage, it is likely that they have developed a mental representation of the structure of

342

Visual Communication 11(3)

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

websites and know where to expect advertisements to appear in this structure (Drze and Hussherr, 2003). This is also suggested by the retrospective interviews. Such expertise in a specific activity is also known to increase the viewers functional visual field, thus the ability to make use of parafoveal vision (Reingold et al., 2001). The extended parafoveal vision and the salient visual features of advertisements would enable the teenagers to recognize advertisements without looking at them directly. The underlying cognitive and visual processes of avoidance strategies should be the subject of future research. It is difficult to say whether this amount of actual exposure is substantial or not. The figures for actual exposure might be considered high, in view of the high information density on the internet. Coming across in such an information-rich media is not easy, as the teenagers negative attitudes towards advertising in general, and online SMS ads in particular (Ekstrm and Sandberg, 2010), make it increasingly difficult for advertisers to connect with the assumed target group. The average attention paid to an advertisement was estimated at 0.5 seconds. This is considered to be a fairly long time in ET research, since it takes less than 0.1 seconds to take in and process basic information (e.g. a word), which is the first crucial step in any persuasion process. Taking this into account, the probability of the teenagers being influenced by online advertising to which they pay attention is high. Precisely the fact that they have paid visual attention to advertisements during a certain amount of time, measured and demonstrated in our study, confirms that the teenagers have been influenced in some way. Our research design does not allow us to make any conclusions about the sort of influence. The current study points out several important directions for future research. In order to obtain the necessary knowledge for European policy guidelines on internet advertising targeting children, we need to construct models that accurately predict actual visual exposure and advertising impact. These models need to take into account: (1) developmental and demographic differences such as age and gender; (2) individual differences in digital literacy, such as web surfing expertise and the ability to carry out goal-directed surfing; and (3) the visual saliency of low-level visual features of internet advertisements (e.g. colour, contrast, animation). This research endeavour is now being pursued through research grants from the Swedish Research Council and the Crafoord Foundation.
ACKNOWLEDGEMENTS

Thanks to the eye-tracking group at the Humanities Lab for valuable comments on methodology and data analysis, as well as suggestions on how to improve the paper. Special thanks to Dr Joost van de Weijer, methodologist at the Humanities Lab, for help with statistical analysis and interpretation of results. The study was funded by the Nordic Council of Ministers, and commissioned by the Swedish Consumers Association.

Gidlf et al.: The use of eye tracking and retrospective interviews

343

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

REFERENCES

Burke, M. et al. (2005) High-Cost Banner Blindness: Ads Increase Perceived Workload, Hinder Visual Search, and Are Forgotten, ACM Transactions on ComputerHuman Interaction 12(4): 42345. Drze, X. and Hussherr, F.-X. (2003) Internet Advertising: Is Anybody Watching?, Journal of Interactive Marketing 17(4): 823. Ekstrm, L. and Sandberg, H. (2010) Advertising Does Not Work on Me, Youth, Marketing and the Internet (Nord 2010, 502). Copenhagen: Nordic Council of Ministers. Fukushima J, Hatta, T. and Fukushima, K. (2000) Development of Voluntary Control of Saccadic Eye Movements: I. Age-Related Changes in Normal Children, Brain and Development 22(3): 17380. Holmqvist, K. et al. (2011) Eye Tracking: A Comprehensive Guide to Methods and Measures. Oxford: Oxford University Press. Kramer, A.F., Gonzalez de Sather, J.C.M. and Cassavaugh, N. (2005) Development of Attentional and Oculomotor Control, Developmental Psychology 41(5): 76072. Livingstone, S. and Helsper, E.J. (2006) Does Advertising Literacy Mediate the Effects of Advertising on Children? A Critical Examination of Two Linked Research Literatures in Relation to Obesity and Food Choice, Journal of Communication 56(3): 56084. Maughan, L. Gutnikov, G. and Stevens, S. (2007) Like More, Look More. Look More, Like More: The Evidence from Eye-Tracking, Journal of Brand Management 14(4): 33542. Moore, E.S. (2006) Its Childs Play: Advergaming and the Online Marketing of Food to Children. Menlo Park, CA: Kaiser Family Foundation. URL (consulted Feb. 2010): http://www.kff.org Neeley, S. (2007) Internet Advertising and Children, in D.W. Schumann and A. Thorson (eds) Internet Advertising: Theory and Research, pp. 34362. New York: Psychology Press. NORDICOM-Sweden (2011) Mediebarometer 2010, Medienotiser (1). Gothenburg: Gothenburg University. Pagendarm, M. and Schaumburg, H. (2001) Why Are Users Banner Blind?, Journal of Digital Information 2(1). Rayner, K. (1998) Eye Movements in Reading and Information Processing: 20 Years of Research, Psychological Bulletin 124(3): 372422. Reingold, E.M. et al. (2001) Visual Span in Expert Chess Players: Evidence from Eye Movements, Psychological Science 12(1): 4855. Sandberg, H. and Ekstrm, L. (2008) Princess Muffins and Chocolate Eggs: An Analysis of Food Advertisements Directed to Children. Nord: 2008, 003. Copenhagen: Nordic Council of Ministers.

344

Visual Communication 11(3)

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013

Sandberg, H., Gidlf, K. and Holmberg, N. (2011) Childrens Exposure to and Perceptions of Online Advertising, International Journal of Communication 5: 2150 Simola, J. (2011) Investigating Online Reading with Eye Tracking and EEG (Studies in Psychology). Helsinki: University of Helsinki Press. Thomsen, S.R. and Fulton, K. (2007) Adolescents Attention to Responsibility Messages in Magazine Alcohol Advertisements: An Eye-Tracking Approach, Journal of Adolescent Health 41(1): 2734. Valkenburg, P.M. (2000) Media and Youth Consumerism, Journal of Adolescent Health 27(S): 526. Wedel, M. and Pieters, R. (2000) Eye Tracking for Visual Marketing, Foundations and Trends in Marketing 1(4): 231320.
BIOGRAPHICAL NOTES

KERSTIN GIDLF is a PhD student in cognitive science at Lund University, Sweden. Her research interests are in consumer decision making and visual processes in natural environments. Address: Cognitive Science Department, Filosofiska Institutionen, Kungshuset, Lundagrd, 22222 Lund, Sweden. [email: Kerstin.Gidlof@lucs.lu.se] NILS HOLMBERG is a PhD student at the Department of Communication and Media at Lund University, with a background in cognitive science and informatics. His current research interests are in eye-tracking, visual attention, web interaction, advertising literacy and web development. Address: Department of Communication and Media, Lund University, PO Box 201, 221 00 Lund, Sweden. [email: nils.holmberg@kom.lu.se] HELENA SANDBERG is an Associate Professor, PhD, in the Department of Communication and Media, Lund University, Sweden. Her award-winning dissertation Media and Obesity: An Analysis of Weight (2004) gained much scholarly and media attention. Her work has since focused on media and health, consumption, advertising and children. In collaboration with the Humanities Laboratory, Sandberg runs the project on childrens exposure to online advertising, funded by the Crafoord Foundation and the Swedish Research Council. Address: as Nils Holmberg. [email: Helena.Sandberg@kom.lu.se]

Gidlf et al.: The use of eye tracking and retrospective interviews

345

Downloaded from vcj.sagepub.com at National School of Political on January 9, 2013