News photography for Facebook: effects of images on the visual behaviour of readers in three simulated newspaper formats

Introduction

In the past ten years, the readership of the paper press has declined while the number of readers of news on the Internet has risen (Tuñez, 2009). The Internet has been the main explanation behind the changes in the frequency of news consumption, primarily because it has encouraged circular communication models, generating a new form of journalism in which the distances between those who make the news and those who consume it, namely journalists and audiences, are blurred (Deuze, 2006). The Web is the platform that is shaping new news and media consumption habits, especially among younger people, who do not see the Internet as a technological advance but instead consider it an everyday, common medium in their lives.

In 2010, the Interactive Advertising Bureau (2010) reported that while the readership of the paper press in Spain had dropped 19%, the readership of the Internet-based press had risen 28%. The last General Media Study in Spain (Asociación..., 2014) performed by the AIMC (Media Communication Research Association) described a substantial presence of news media on the Websites visited the most by Spanish Internet users. It also noted that 52.4% of the respondents stated that they used the Internet to read the news. According to research carried out by Reuters Institute (Newman, 2013), worldwide audiences increasingly want news on any device, in any format and at any time of the day. However, reports reveal that the multi-platform and digital revolution is not taking place at an even pace in all countries. Hence the Spanish cultural context should not be considered to be necessarily identical to other European or American contexts.

What happens in the US does not necessary follow automatically in Europe or elsewhere. Geography, culture, and government policy also play their part, with Germany and France still showing strong allegiance to traditional forms of media. We also see marked differences in ‘participatory cultures’, with very different rates of take up in social media, commenting, and voting across our surveyed countries (Newman, 2013, p.9).

Although desktop computers or notebooks are still the primary devices for accessing digital news, the key underlying trend is the growth in accesses from multiple devices. One-third of Newman’s sample was found to obtain news on at least two devices and 9% used more than three.

In terms of the use and penetration of social networks in Spain, 79% of Internet users use social networks; the largest user group is young females (34% are between the ages of eighteen and thirty). Among the social networks, Facebook is used the most, although Tuenti is the favourite in the fourteen to seventeen year old target group. The main activity on social networks among Spaniards is reading content, while only 14% claim to generate content on a regular basis (Interactive..., 2014).

Newman (2013) observed that there is a significant division between those under thirty-five, who prefer online news, and those aged over forty-five who strongly prefer television news. News consumption for the younger generations has become mediated by social networks such as Twitter and Facebook, either by joining fan-pages or as a result of sharing by friends who are members of the same social network. This relationship between social networks and the online press has grown significantly in recent years. Some studies based on Web data-mining analyse the contents and underlying communication models established by the major media in the use of social networks (Sáez-Trumper, Castillo and Lalmas, 2013). Specifically, by comparing the impact of different press media on Facebook, Cárcamo-Ulloa and Sáez-Trumper (2013) studied the membership of fan-pages of the Chilean media and the effects of likes, comments and sharing in this country’s press. Scifleet, Henninger and Albright (2013) explain how social media can become the top-priority source of communication for readers in the event of catastrophes. Yamkovenko (2013) offers considerations on which is the best time to publish and what kind of content is the best to improve the relationship with Facebook audiences.

This study examines the visual behaviour of people when reading online press posts they receive on their Facebook wall. To do this, we created three news formats: text only, text with a large image and text with a small image. We studied twenty-four volunteers in sessions which were recorded using an eye tracker.

Our hypotheses in this study are the following:

1. News that comes with a larger image will be more attractive to users. This visual attraction will be shown by the speed of the users’ first eye fixation.
2. A news story with an image will generate more interest in readers than news stories with no images. This interest will be manifested in the amount of time they spend on this news story.
3. A news story with an image will generate more interest to read further than news stories with no images. This interest will be manifested by clicking on it.
4. Finally, once the experiment is over, it is expected that the readers will declare that they prefer to get news in media with larger-sized images.

Overall, the results of the study will help us to determine how the presence and size of an image affect Facebook users’ visual behaviour and tracking of news.

Although our study focuses on low level visual perception and this issue is not supposed to have direct relationships with affectivity in social networks, certainly at higher cognitive levels many other sources of representations may be elicited by images, alongside emotional states, both having sensitive effects in meaning construction. These elicited elements, although not assessed, can be assumed to be randomized within the respondents. In the next section, a survey of previous studies is presented, after that the research methodology used is set forth. Finally, results and conclusions are discussed.

Theoretical framework

In the last two decades, the mass media have experienced both the advantages and the disadvantages of the analogical-digital transition. This change process brings technological and workplace adaptations common to the digitalisation of signals and the new platforms on which information is published. Thus, for example, Sánchez and Méndez (2013) report the appearance of social media editors and community managers as crucial agents in this stage of the media.

The changes in professional competences reveal how the Internet has become a space to be urgently colonised by the media and today large newspapers and radio and television stations have Websites and use social networks to compete in attracting Internet users to increase the value of these Websites. On this subject, Fumero (2011) states that the Web is changing rapidly and transforming three basic dimensions of our social nature, namely, information, interaction and communication. Other studies stress the role of social networks in promoting news:

It is revealed that in a brief period of time social media (Facebook and Twitter) have shifted from being systems devoted to forging online communities to playing a major role in cybermedia themselves, even taking on the jobs of promotion, dissemination and information. (Tejedor-Calvo, 2010, p. 617).

For the traditional news media, this technological change implies not only a technological adaptation from paper to bytes, but also a larger number of news stories, the competition of newspapers, radio and television stations on the same screen, and the adaptation to new spaces and communicative tools, such as social networks. Social networks are not aimed at the same target as traditional media, which have historically targeted adult audiences. So far, social networks are largely inhabited by adolescents and young adults, and this has meant that the media has largely adapted to this audience.

In its obsession with attracting adolescents, the media system has become adolescent itself: experimenting with formats, new programmes and messages. It is experimenting with itself. (Callejo-Gallego, 2012, p. 23)

As a reflection of the Internet, and particularly social media, being a part of youth culture, Facebook has adopted the metaphor of a fan to refer to users connecting with institutional, commercial or service supplier pages on its Website.

Viral news spread

The advent of social networks also made media writing more complex. The social network’s power of dissemination poses the challenge of attracting online media users who combine audio-visual-reading content (Cebrián, 2009) and who also evaluate, comment on and share news on social networks like Facebook and Twitter. Once this information is shared, it is expected to generate attention from other readers and to gain users’ loyalty, becoming fans or followers of the media, forming news consumption communities.

On the Internet, the concept of community is increasingly important, since it contextualises and personalises the contents produced there. And conversely, to get a member of a given community to optimally increase their knowledge, the contents they receive must be tagged, perceived as valuable and commented on in accordance with their needs. (Astigarraga, Azpillaga, Fernández and Naberan, 2011, p. 543).

In 2010 Facebook started offering the possibility of making a fan-page and the media started implementing it in their news routines. With this arose the demand for reporters and/or community managers to define the content to be posted on the social network, the right frequency of posts and the visual format to be used to deliver news. When vying for audiences, the news cycle usually means that the larger media outlets start with a certain advantage when positioning themselves on the Internet; that is:

the number of followers on the Web will rise in the future. In other words, their popularity in the 'offline' world helps them to be popular online. (Sáez, 2011, p.73).

However, the means of popularising news on social networks, i.e., posts, likes, comments and sharing on Facebook, and tweets and re-tweets on Twitter, seem more like a word-of-mouth means than the metaphor of the powerful antenna that says all and reaches everywhere. The methodology used by Facebook to rank content is not public. Companies like Facebook or Google never reveal the way they rank content, because this will allow spammers to game the results. Nevertheless, some community managers describe EdgeRank as the algorithm behind the Facebook timeline. EdgeRank is supposed to consider three elements: users’ preferences (based on friend preferences); intrinsic content value (a comment has a higher score than a like); and how recent news is (fresh content has a higher score than older content).

Experts now promise to help advertisers understand the phenomenon of news-propagation in social media (e.g., SocialFlow). Articles in the trade press have speculated for years about Facebook’s Edge Rank algorithm, revealed by Facebook engineers at a development conference in 2010 (Hamilton, 2014, p.635).

Indeed, the intelligent use of irony in a post or the impact of an image may be an interesting source of novelty, thus capturing the attention of followers of this media outlet on the social networks. Cárcamo-Ulloa and Sáez-Trumper (2013) cite the example of how some alternative Chilean media are capable of competing for Facebook audiences and vie with large conglomerates to disseminate news, reaching important levels of viral spread with interactions such as post, like, comments or sharing. In this case, the users become the best promoters of the media.

In a study that exploited a dataset of almost 300,000 posts gathered from roughly 1000 popular Google+ users, Guerini, Staiano and Albanese (2013) explored which of three types of posts was most likely to go viral: posts with static images, dynamic images and only text. They found that posts with images were more viral than those with only text, according to the number of +1s, replies and re-shares. Moreover, they confirmed that fun images and informative ones had a higher probability of being shared, while colourful images or those containing people’s faces obtained more +1s and more replies. Another recent study (Khosla, Das Sarma and Hamid, 2014) used a sample of 2.3 million images from Flickr in order to explore what makes a photograph popular. They investigated two components: the image content (colour, gradients and the set of objects present) and its social context (number of friends and number of photos uploaded by a person). They conclude that the most popular images all had in common cluttered colours and elements such as miniskirt, jersey, bikini, bra, cup, perfume and revolver.

Photographs and multimedia in news media

Guallar (2011) points to the fact that digital newspapers have the possibility of using photographs with fewer limitations than in the printed media, which opens up new perspectives for images in the press. He further states that the media should outline and/or develop appropriate strategies for image use. Studies such as that of Thurman and Lupton (2008) report on the inclusion of multimedia content such as video and storytelling (linear narrative with relatively little text, consisting mostly of photographs) in the press. Masip, Díaz-Noci, Domingo, Micó-Sanz and Salaverría (2010) explain that a high degree of experimentation with multimedia resources in the press can be seen in cyber media.

Images have a long history in the press as a way of capturing attention; often the content of these images outstrips journalistic ethics and morbid curiosity becomes an important factor when trying to attract readers (Torres, 2012). A study conducted by Cárcamo and Marcos (2013), which analysed 301 posts issued during the course of one week in ten different media, determined that there are three predominant formats for delivering information via Facebook posts:

1. Text, link and large image
2. Text, medium-sized image and link
3. Text and link

It should be noted that picture reception using social networks on the Internet is part of visual culture, because,

the vision is (as we say) a cultural construction, that it is learned and cultivated, not simply given by nature; that therefore it might have a history related in some yet to be determined way to the history of arts, technologies, media, and social practices of display and spectatorship (Mitchell, 2002, p. 166).

Some further relevant contributions related to culture were made by Barthes (1980) and Freedberg (1989). Barthes proposed that rhetorical figures could also be observed in pictures, contributing to the construction of meaning; Freedberg stressed the role of images throughout human history and their ability to be socio-culturally provocative.

Visual perception and eye tracking

There is a great deal of literature that reveals the relationship between attention and gaze (Fischer, 1999; Rayner, 1992; Rayner, 1998; Remington, 1980; Shepherd, Findlay and Hockey, 1986). One of the key theories on selective attention comes from research by Lachter, Forster and Ruthurk (2004). Following Broadbent’s selective filter theory (1958), these authors state that attention works as a filter and that it can be directed to one source of stimulation or another.

A recent study by Smit, Neijens and Heath (2013) reaffirms the approach of Greenwald and Leavitt (1984), who distinguish between four sequential levels when reading advertisements in newspapers. These levels outline a macro-process that spans from pre-conscious attention levels to the elaboration of meanings in dialogue with memory.

Depending on how the retina registers images, the human visual system has three levels of perception (Rayner, 1998): foveal vision is the kind that is registered with the most clarity, and the one that comes from things to which we have paid more conscious attention; it is the one we use for reading, driving and all activities in which we require a great deal of clarity. Parafoveal vision is the kind that is registered in the zone around the fovea, and it is less clear than foveal vision. Finally, peripheral vision registers images that reach the furthest area from the fovea; this vision has lower resolution and does not capture features like colour.

When we interact with an interface, we make constant eye movements which are extremely quick (up to 500 movements per second). They occur when we shift the focus of attention and they are called saccades. In turn, when we keep our eyes staring at a given place for several milliseconds, this is called fixation.

Some technology enables us to register the fixations and saccades of the human eye: so-called eye tracking devices. Eye trackers detect, track and record the movements we make with our eyes, the duration of our gaze and the pupil dilation, measurements that we can relate to attention, so they are extremely useful when studying reading, behaviour and interaction between people and visual interfaces. To track the eye movements, modern eye trackers operate remotely, that is, without intensively interfering with people’s eyes. The most common technique today is known as pupil centre and corneal reflection; to apply it, the eye tracker comes with an infrared light and a video camera. When it is activated, the device illuminates the user with two projections of infrared rays that generate a reflection on the retina, specifically on the fovea. A video camera integrated into the eye tracker captures these reflections along with the user’s position, and by digitally processing the image it can ascertain the location of the pupils at a ratio of 50 Hertz or higher, depending on the device. After this, the pupils’ position is mapped with the location of the eyes onscreen and in this way it is possible to ascertain where the person is directing their attention at each moment recorded.

The eye tracker has numerous applications in research, especially in cognitive psychology, since it provides extremely relevant information on visual attention and the identification and visual categorisation of objects, as shown in the studies by Abbot (2006) and Altmann and Kamide (2009). The field of person-computer interaction has also applied this technique intensively, in this case primarily to learn about people’s behaviour when browsing the Web (Nielsen and Pernice, 2010).

For the purposes of this paper, we are interested in image perception. Yarbus (1967) has already shown that image content strongly influences eye movements. Moreover, we know that there is a tendency for humans to fixate on images of faces, and the parts of faces looked at can be identified from gaze paths (Belle, Laeng, Brennen and Øvervoll, 2009). Klami (2010) also investigated which parts of images were relevant for users. Based on this idea, and given that humans have the ability to perceive visual information and we are able to assign a given tag to a region in an image, Walber, Scherp and Staab (2013) had the idea to benefit from this human ability to obtain a better understanding of depicted scenes and use the eye tracking technology to explore the possibility of doing this task in an automated way.

Eye tracking studies on social networks

Although the majority of user behaviour studies are carried out through observations and surveys, recent research shows that eye tracking provides valuable insights into users’ perception of online content. Adnan, Hassan, Addullah and Taslim (2013) conducted a study to analyse users’ behaviour, in terms of their activities performed on social networking sites, by means of eye tracking techniques. Four main measurements were examined: the first point that the user looks at, time spent on areas of interest, main activities and total gazing time. Results indicated that wall posts located at the centre of the screen received most attention, based on the observation of the first place users looked, as well as on the duration of gaze recorded. Findings also showed that the main activity among users was reading friends’ statuses on the wall posts area.

Ozturk and Rizvanoglu (2010) explored the usability of the profile pages on five of the most popular social networks (Facebook, Myspace, Twitter, FriendFeed and Last.fm) through an eye-tracking study with twenty-four university students. Their findings emphasized several ways to improve usability in profile pages: only relevant information should be presented on profile pages and the content block needs to be visually separated, in order to make the content readable and easy to explore and perceive. Results showed that users mostly paid attention to profile pictures and recent activities on the profile page.

Boyd, Nugent, Donnelly, Sterritt and Bond (2012) performed a study aimed at observing people’s behaviour in a re-designed interface of Facebook. They asked twenty participants, divided into young and old, to perform typical tasks such as updating their status, sending a message or uploading a photo, while the sessions were recorded with an eye tracking device. The older participants had more problems with familiarising themselves with the new design, they needed more time to complete the tasks and were concerned about the possibility of errors in tasks, while young users had fun exploring the new design.

Eye tracking studies on online newspapers

The changes in the way news is presented today have come not only in response to media editorial approaches, but also in response to the demands from new readers. These demands originate in changes such as the amount of time available and new ways of interacting with text and images.

There is not a comprehensive bibliography to analyse the visual behaviour of people reading newspapers online, although it is more than two decades since the Poynter Institute conducted the first studies of online media. The literature search, setting apart the Poynter Institute studies, also provides studies focusing on the usability of news interfaces, works comparing the behaviour of readers in online media and paper, advertisement analysis in online newspapers and a study on the selection of news from a search engine. The main findings follow.

The Poynter Institute has sponsored at least three news studies which used the eye tracking device. In Eyes on the News, García and Stark (1991) present valuable information on the use of colour and reading routes in paper newspapers. Nine years later, Lewenstein, Edwards, Tatar and DeViga (2000) made a surprising finding in their study, in which they concluded that users of digital newspapers first read the captions before looking at the pictures, contrary to the pattern in paper newspapers. Some of the explanations of this phenomenon suggest that this is a behaviour inherited from the older version of the Internet, when html allowed the text to appear quickly as the images were loading.

Outing and Ruel (2004) point to the study entitled ‘What we saw when we looked through their eyes', which provided valuable information on reading patterns and defined priority zones of information diagramming on computer screens, as shown in Figure 3.

Among the previous studies investigating foveal behaviour before new devices and media, we should spotlight the study by Mosconi, Porta and Ravarelli (2008). They studied the use of multimedia in online newspapers. In another contribution from the Poynter Institute, Adams, Quinn and Edmonds (2007) compared users’ reading of digital and online newspapers, and Quinn (2012) recently explained how the news is read on tablets and its interactive relationship with page sliding. Recently, Arapakis, Lalmas, Cambazoglu, Marcos and Jose (2013) studied how user behaviour in news portals is also constructed in an association between news and feeling.

Michailidou, Harper and Bechhofe (2008) investigated users’ browsing behaviour in the context of Web accessibility. In this study, nine Web pages were investigated to determine how the page’s visual arrangement is related to users’ browsing patterns. Results showed that salient elements attract users’ attention first and users tend to spend more time on the main content of the page. This study also emphasized that common gaze patterns begin at the salient elements of the page, move to the main content, header, right column and left column of the page and finish at the footer area.

Leckner (2012) examines reading behaviour in readers of printed and online newspapers. The aim is to identify how much reading behaviour is dependent on appearance factors and how this dependence and its magnitude are related to electronic or paper format. The work is based on the review and analysis of empirical studies, primarily those employing eye-tracking technology. Results show that, for text-based elements, size and placement are important guides to salience in both kinds of media. Cantoni, Porta, Ricotti and Zanin (2013) exploited eye tracking technology to study the eye behaviour of fifty users while watching the home pages of different newspapers, each one containing masthead banners with different arrangements and dimensions. The purpose was to investigate whether there some solutions are better than others in capturing the user’s attention, in order to reduce banner blindness (the phenomenon for which users, consciously or not, tend to ignore what looks like a banner) and therefore improve newspapers' banner attractiveness. They noted that the lower banner, located just above the bar that displays a newspaper’s sections, was watched by more participants, sooner and more frequently than the upper and left banners, but for a lesser time. The average fixation duration was longer on the left banner than on the upper and lower ones, thus confirming previous research from the Poynter Institute (Adam, et al., 2007; Garcia and Stark, 1991).

Recently, Rovira, Capdevila and Marcos (2014) presented the results of a study about the search engine Google News, in order to determine the importance that users give to headlines, sources and summaries in search results. By the use of eye tracking, researchers tallied the duration of the gaze from fifty users viewing search results lists, to test which of these elements captured their attention for a longer time. Results indicate that sources receive more attention than summaries, while sources and title showed no significant differences.

Unlike these studies, ours has the particularity of offering an experiment on news formats on the social network Facebook. Today, the media not only compete for audiences on paper and Websites but have also become integrated into social networks in a bid to capture fans or followers. These users are attracted to getting news on a particular event through posts that use text, images and links, and the media chooses to confer different priorities on text and images when composing the format of the news story.

Research method

Participants

A total of twenty-four volunteer users were asked to participate, eleven men and thirteen women between the ages of twenty and forty, with an average age of 27.5 years old. All the participants had graduated from or were currently studying at university and participated in academic activities at Barcelona’s Pompeu Fabra University. There were two inclusion criteria for participating in this test: being a Facebook user and being a regular reader (at least twice a week) of online news.

Material

We used a Tobii 1750 eye tracking device integrated into a 17’’ TFT monitor with a resolution of 1280 x 1024 pixels. The device was gauged for each recording and the light source was always in the same location. The distance between the user and the monitor was fixed at 60 cm (24 inches), and was controlled through the use of a stationary chair. Tobii Studio software was used to get the heat-maps and metrics used for the analysis.

Variables

This study examines four dimensions related to people’s news reading behaviour in three presentation formats on their Facebook wall:

1. Visual attraction, measured through the speed of attention capturing.
2. Interest in the news, measured through attention persistence.
3. Choice of news, measured by means of the clicks performed by participants.
4. Oral declaration of preferred format.

In the experiment, the independent variables were the formats in which the news stories were presented and each format corresponded to the publication style of three newspapers invented for this study.

• Prensa de Hoy, which presents the news stories exclusively in text
• Diario Actualidad, which uses large images
• Informatorio, which has a style that combines text and image proportionately

The dependent variables chosen to respond to the specific objectives of this study were the following.

1. To describe the visual attraction of all three news formats proposed, we used the time to first fixation metric, which measures the time that elapses from when the news content is first displayed on the Facebook wall until the user fixes their attention on each of the three proposed formats for the first time. The lower the time, the greater the format’s capacity for attraction. To neutralise the effect of the position occupied by each format on the page, we alternated the positions on the three walls such that each news story and all formats occupied all the positions the same number of times.
2. To study the interest sparked by a news story in each of the three formats, we recorded the amount of time the users spent reading/looking at each story. The eye tracking metric that furnishes this information is the total fixation duration.
3. To measure which formats were chosen most often by users to get news, we analysed their clicks using the click count metric, which indicates the number of times a news story has been clicked on, and time to first click, which measures the time that elapses from when the content is displayed on the Facebook wall until the user clicks on a result.
4. Finally, to determine users’ preferred format, we orally asked each subject which format they preferred after they had performed the test.

Design

Phase 1. Selection of news stories chosen for the study

We created three general news stories similar to the ones the media might have published on three different topics:

• news story on university students receiving scholarships
• news story on a meteorite that will orbit near the Earth
• news story on the health benefits of the Mediterranean diet.

For each news story, three texts were written, each of which was to be used in a given format. These texts were written to have a very similar style, so that the writing style would not be a factor that would particularly capture readers’ attention. Each news story was also associated with an image, in this case the same image in all three formats, given that the images were the main purpose of the study and changing them would have biased the results. Results correspond to a specific experimental group and could obviously vary in a different cultural context.

Phase 2. Format design

Three fictitious press outlets were created; we invented them because we did not want the users to have prejudices or preferences when choosing or clicking on news stories. Each media outlet always used the same news format to post on Facebook (Fig. 4):

1. Diario Actualidad, with text, a link and an image that covers 75% of the post
2. Informatorio, which divided the space, 50% for the image and 50% for the text
3. Prensa de Hoy, text only.

A Facebook profile and three simulated walls were created, each with a news story presented in the three formats being studied. That is, on the same wall the users saw the news just as it had been published in the three newspapers.

Phase 3. Marking areas on the screen

Briefly, the eye tracker works as follows: when the device begins to record the user’s eye movements, it takes a screen shot. Later, when the recording is being reviewed, the analysis software superimposes this screen shot of the information on the eye movements, yielding heat maps showing the zones that have received the most attention. If there are also plans to get statistical measurements, as in this study, the zones of the pages being analysed must be marked. Tobii Studio software calls them Areas of Interest (AOI). For the purposes of this study, we marked three large areas of interest on each wall, which corresponded to the three news stories, and within each story we marked the area occupied by each image in the two fictitious media sources that used images (Figure 4).

Procedure

The experiment consisted of presenting a Facebook wall with the three news stories by the three fictitious media outlets posted parallel to each other. The order in which the posts appeared on the wall was alternated to ensure that the location of the image on the experimental interfaces was not a source of bias. The study just focuses on Facebook’s desktop interface. Following the Latin square layout, each participant was sequentially shown all three Facebook walls.

Participants read each of the three interfaces presented and clicked on the media source from which they wanted more information. The session lasted approximately seven minutes per user.

The process was recorded using the eye tracking device, which not only tracks foveal vision but also saves a record of the clicks and the amount of time the users stay on each page. At the end of the recording, the users were asked which news format they preferred.

Statistical analysis

For this study we made a quantitative analysis, which includes descriptive statistics and tests to contrast means. For the latter, we used the non-parametric test (Wilcoxon’s T) since data did not show a normal distribution, perhaps because of the size of the sample (twenty-four users). The calculations were made using the SPSS/PC+ statistical package (version 15.0) and the statistical tests were bilateral with type 1 errors variable at 5%.

Results

Before answering the research questions, we reviewed the gaze plot of the twenty-four users on the three walls. Dovetailing with Outing and Ruel (2004), we noted a reading behaviour pattern based on fixations and saccades: the eye movements went top to bottom and left to right, and more effort was concentrated on the texts than on the images. This eye movement map remained the same even when the order of the post formats changed randomly. This map also showed that the side bars to the left and right on Facebook attracted very little attention (Fig. 5).

Visual attraction

To determine the attraction of the news posting formats shown to users, we measured the amount of time each person took to look at each of the news stories, that is, the time that elapsed between when they opened the Webpage until their first eye fixation on each news story was recorded. On the eye tracker, this metric is called time to first fixation. The results of the test, displayed in Table 1, showed that the post that attracted the first fixation the most quickly is Diario Actualidad, that is, the one with the largest image.

When we performed the Wilcoxon signed-rank test, we found significant differences in the time to first fixation in Diario Actualidad and the other two formats. In both cases, the time users took to look at the news story in Diario Actualidad was significantly lower (p=0.001 and p<0.001), as shown in Table 2) [In all cases in the tables below, the statistic is based on negative ranks.]

Based on the figures presented, we can see that the Diario Actualidad format, which contains the largest image, is the one that attracts users’ attention the first, while there were no significant differences between the other two formats.

Interest

Another important element to analyse is the time spent solely on the images (not on the news story in general) between the two formats that use this resource. The figures shown in Table 5 reveal significant differences (p <0.001): users spent more time looking at the larger images.

A comparison of measurements between the texts in both formats is also significant (p=0.007). Even though the texts are identical, the participants spent more time reading the text accompanying the larger image.

Selecting a news story

In the procedure section we explained that each user had to read each wall and choose the format from which they would seek further information on the news story. This choice was made with a simple click, which the eye tracker recorded. This click provides us with two pieces of information:

• >which format each user chose on each wall, and
• how much time elapsed from when the user opened the Webpage until they chose a format.

We also sought some kind of regularity in the time to first click metric, although we found no statistically significant differences. Diario Actualidad got an average time of 8.79 seconds, Informatorio 9.03 seconds and Prensa de Hoy 7.88 seconds.

Our analysis of the clicks received by each format reveals that the news posted by Diario Actualidad, which has the largest images, received the most clicks, while the remaining clicks were fairly evenly distributed between the other two formats: medium-sized image and no image. (Table 8).

The heat map in Figure 6 shows the concentration of the gazes in terms of fixation duration. Red areas are those that have been fixated longer, followed by yellow ones. It is interesting to note that the clicks, which are indicated with the mouse icon, are concentrated on the format with the largest image, a situation that was repeated in the other cases studied, regardless of the position of the formats.

Subjective preference for a format

Participants were asked which format of the news stories they found most pleasant. Table 9 shows users’ responses to this question.

As can be seen in Tables 8 and 9, the question revealed an acceptable degree of consistency between the actions performed by the users during the experiment and the subsequent expression of their preference for one of the three formats regularly offered by news providers.

Summary of results

Table 10 shows that foveal fixation happens earlier when looking at a larger image, as the post with no image is the one that took the longest to attract foveal fixation. Likewise, the length of fixation is more intense depending on the existence and size of an image. That is, photographs are capable of attracting more interest in the news story and participants spend more time looking at it.

Finally, the click count enables us to state that the size of the photograph is influential when choosing a news post. Not only did the subjects mainly choose the news stories which had larger images, they also remained there or spent more time on the format that featured images, and the first foveal fixations happened more quickly in the stories with larger images.

Discussion

If one thinks of photographs as mere decoration or an accessory to the text, we fall into the error of remaining stuck in the paradigm of the old paper-based press. Today, both the technical possibilities and users’ visual culture suggest that we promote messages with visual content. The choice of the size and content of the images is connected with considering images not only as an appealing element but as an important part of the message (Guallar, 2011). As was stated by Mitchell (2002), images, by themselves, convey information and meaning that interact with text content. Consequently, combining an image and text is not just adding two sources of information in two representational formats, rather it implies a reciprocal modulation.

In regard to the hypotheses, we can state that:

A. Both the presence of an image and the size of this image positively influence the capacity to attract users’ sight to news posts on Facebook, since larger images were proven to require less time for viewers to fix their gaze on the news story.
B. Both the presence and size of an image positively influence the readers’ interest in a particular news story, since the duration of the eye fixations was higher when larger images were included in the posts

The design of this study does not enable us to determine whether there is a stable, linear relationship between size, attraction and attention time. However, the results do enable us to consider the two-fold efficiency of posts that use images when posting news stories on Facebook.

Having managed to attract interest in a news story, the next step is hooking the user into interacting with it:

C. Both the presence and size of an image positively influence the number of readers who choose a particular news story for further information, since the news stories with large images received many more clicks.
D. After the experiment, the participants in the study orally stated their preference for the format that has a larger image. Therefore, we can state that the selection process of this format is conscious.

In short, the results of this experiment are coherent with the hypotheses posited at the start of the study, confirming that the presence or absence of images and the size of these images condition the visual behaviour of Facebook users from the start of the perceptive process until the end of their interaction with news stories.

Results of this study are similar to the processes observed by Greenwald and Leavitt (1984), Michailidou, Harper and Bechhofe (2008) and Smit, Neijens and Heath (2013). We have also noticed an early pre-attention phase, which entails an unconscious, non-deliberate exploration of the news options, where images attract users quickly. We also detected a second phase of active focal attention in which users stayed longer to examine the posts that use larger images. What is more, participants also went through a third phase where they took the decision to learn more about the news story via one of the three options offered by the research design. In contrast, our study did not deeply examine the subjects’ phase of elaborating meaning.

Nevertheless, when asked about their preference for the format that is most appealing to them, and by observing that the preference of most of them was formats with larger photographs, we can infer a phase in which the subject forges personal connections with the images, once again concurring with Greenwald and Leavitt (1984) that images facilitate closer ties between the news post and the user’s memory.

Limitations

In order to improve this study’s reliability, it is recommended that future experiments should be carried out to evaluate a larger number of subjects to obtain more precise results in the contrast analyses of the time to first click metric, which did not reveal significant differences, perhaps due to the small number of participants. The absence of significance seems to point to a low effect. A larger sample would be more representative, thereby enhancing the validity of the results, since they can barely be extrapolated beyond the university student population. However, the eye-tracker technology assesses low-level behaviour and brain circuitry, something that may compensate for the small sample, avoiding many sampling issues that could compromise the validity of the test.

Furthermore, the technological device where information is presented might have a considerable influence on the way perception acts. Devices such as phones or tablets represent a significant portion of users’ means to access news on Facebook, where the different hardware – particularly in the case of the screen –implies substantial changes in the information display, as well as in the perception process.

Another aspect that has not been thoroughly dealt with in this article is the analysis of the heat-maps which, at first glance, seem to indicate a significant effort to encode the texts and a lesser effort to view the images. We can assume that perceiving images requires different cognitive and brain resources, being more holistic, while those involved in reading a text must be sequential; this might account for the larger number of foveal fixations on the text.

Future research could build on this to study samples of different cultures, since it might provide some clues about whether the pattern can be generalised or verify that different cultural contexts build the vision (Mitchell, 2002). In a similar manner, repeating the experiment in people from different generations could provide an estimation of the effect of former habits in news inspection and selection.

Conclusions

The main contribution of this study is that it describes the visual behaviour of readers of news stories in a controlled situation by observing different news formats, using an automated procedure that records eye movements and applying a real-time recording protocol of their visual search strategies when reading the news in the online context of social media.

Results suggest that, as stated by Leckner (2012), in order to take effective editorial decisions, news formats that use images of a large size should be chosen. This format has been proven to be what the majority of participants prefer. They click on them and, even more importantly, their attention is captured more quickly and their interest held for a longer period. This seems to be a good strategy for overcoming the constant exploration of alternatives that the media are undertaking (Masip, et al., 2010; Thurman and Lypton, 2008) and it is also important for the transition from paper-based to digital publishing (Interactive..., 2010).

Acknowledgements

Our thanks to the participants in our study and to the anonymous reviewers for their constructive comments on the submitted version of the paper.

About the authors

Luis Cárcamo-Ulloa is Associate Professor of Communication and Education in the Department of Social Communication of the Universidad Austral de Chile (UACh), Chile. He received a Bachelor's Degree in Journalism in Universidad Austral de Chile and a Ph.D. in Perception, communication and time from the Universidad Autónoma de Barcelona, Spain. He is a member of the Kelluwen Research Group (www.kelluwen.cl) at UACh. Dr. Cárcamo-Ulloa can be contacted at: lcarcamo@uach.cl
Mari-Carmen Marcos is Associate Professor of Information and Library Sciences in the Department of Communication of the Universitat Pompeu Fabra, Barcelona. She received a Bachelors Degree and a Ph.D. in Information and Library Science from the Universidad Carlos III de Madrid and the Universidad de Zaragoza (Spain) respectively. She is a member of the Web Research Group at Universitat Pompeu Fabra. Dr. Marcos can be contacted at: mcarmen.marcos@upf.edu
Ramon Cladellas Pros, PhD, is a Lecturer in the Department of Basic, Development and Educational Psychology of the Universitat Autònoma Barcelona, Bellaterra. His research interests include: assessment’s learning, Teacher’s Methodology, Perception and Management Time. Dr. Cladellas can be contacted at ramon.cladellas@uab.es.
Antoni Castelló Tarrida, PhD, is a Senior Lecturer in the Department of Basic, Development and Educational Psychology of the Universitat Autònoma de Barcelona, Bellaterra. His research interests include: cognitive processes and intelligence both in human and artificial intelligences. Dr. Castelló can be contacted at Toni.castello@uab.es