Lund University Cognitive Studies- LUCS 127 ISSN 1101-8453 © 2005 by authors
1
The role of local design factors for newspaper reading behaviour – an eye-tracking perspective
Kenneth Holmqvist and Constanze Wartenberg
火的用途Lund University Cognitive Science
Kunsghut, Lundagård
S-222 22 Lund Sweden
Kenneth@lucs.lu.
I NTRODUCTION
Newspaper design is a creative art, and the possibilities to vary design are endless. Basically, newspaper designers judge the value of a designed spread by looking at it themlves and imagining how readers could perceive it or n what it feels like. Actual feedback on readers perc
eption of the design is very scarce. The only method widely in u is target group interviews. Interviews can cover large samples from a reader population, but readers have only a very limited lf-knowledge and memory of their own reading (Helander, Landauer and Prabhu 1997). Also, very few general interview questions are asked. It is unlikely that the can reflect all potentially interesting details and process that occur during the half hour that Scandinavian readers on average spend with their newspaper.
Eye-tracking during newspaper reading gives us preci information 50 times per cond on where readers are looking. Thus, eye-tracking methodology provides preci insight into how the individual reader perceives newspaper spreads.
This article describes a study of local design factors such as colour, drop quotes, fact boxes etc. in 34 spreads from 17 Nordic newspapers. The study was initiated by the board of the Society for Newspaper
Design/Scandinavia (SND/S) in the context of preparations for the SND/S 2003 conference. In particular, this study looks at the effect of each factor on scanning order and on dwell time. We try to answer questions like: Do fact boxes attract reader attention? Does colour lead to longer visual dwell times?
Wartenberg and Holmqvist (2004), using the same data, prent results on how well designers predict the effect of the different design factors. For instance: Do designers assume that fact boxes attract attention? This article does not prent any results concerning the role of content or of global design factors such as overall composition or the degree of white space on a spread.
L OCAL DESIGN FACTORS
There are a number of local design factors that the designer could u as tools to emphasi objects – a story or an ad – on a spread. Whether the tools work the way it is assumed is the main question in this study. There have been many hypothes on newspaper design factors: •
Positioning an object at the top of a spread, for instance, could lead to readers eing it earlier.
Lund University Cognitive Studies- LUCS 127 ISSN 1101-8453 © 2005 by authors
2
•
Giving the object a bigger size should make readers e it earlier. (Lundqvist and Holmqvist 2001).
•赛扬g530
Colour in a photo may attract the reader’s eyes (possibly Garcia and Stark 1991, e also Hann 1994 and Jophson 1996).
• The drop quote s are introduced into texts for the very reason to attract attention during scanning. • Fact boxes may be ud for another purpo, but could have the same effect as drop quotes. •
Pictures are known from Garcia and Stark as well as from Hann to attract the earliest fixations on a spread.
•
The axiality of an object has been thought to matter for reading depth – a long vertical text could scare readers off.
Some of the hypothes are very wide-spread assumptions in the design world. Most of them have never been tested before. Colour, pictures, position and size have been addresd in various studies. In the current study, we will prent results concerning all of the above-mentioned local design factors. Global design factors, such as the choice of typeface, the amount of white space in t
he spreads, or the typical “look” of a particular newspaper have not been included as factors in this study.
P REVIOUS RESEARCH
hard的反义词A small number of explorative and experimental eye tracking studies of newspaper reading have been conducted beginning in the late 1980s. Küpper (1989) tested 60 readers of Badisches Tagblatt, asking basic questions such as whether pictures are looked at more than text, whether there are favorable positions on a spread and whether short texts are preferred to long ones. Several questions that later were studied in more detail were assd for the first time in this study; in particular reading order on pages, and the influence of position, pictures and headlines. Unfortunately, data treatment and statistics are rather poorly described, making it difficult to interpret some of the results.
Widman and Polansky (1990) tested 129 readers of the Stockholm newspaper Dagens Nyheter for advertiment reading. The unpublished internal
report shows that 39 % of all ads are en, and the bigger the ad, the more likely that it is en, and remembered. This report also investigates the importance of different positions and contents of news
paper ads, noting that a position slightly to the left of the middle is optimal for the smaller ads. Photographs, pictures and colour in ads are reported to increa fixation frequency on the ad. Widman and Polansky also note that readers very quickly forget that they have actually looked at particular ads. Lundqvist and Holmqvist (2001) tested 14 readers of Dagens Nyheter in a follow-up study to Widman and Polansky, but focussing entirely on the size effect on attention, attitude, and memory of ads. Results show that the variables strongly correlate: A bigger ad size makes the reader e, like and remember the ad better. Dwell time on an ad is thus a strong predictor of how well it is remembered. An unpublished follow-up study (by the same authors) on Svenska Dagbladet and Metro indicates that tabloid and broadsheet formats do not differ in how ads are perceived.
Garcia and Stark (1991) tested 90 readers of three newspapers at three different sites in the US. The editions (prototypes A and B) given to the readers were manipulated with respect to colour – a major motivation behind this study was to e whether colour in itlf attracts reader attention. Participants could read as long as they wished and their eye movements were tracked during the reading ssion. A videotape with newspaper pages and the tracks of the eye movements across the pages was ud to find out whether elements on the page were procesd, read or “read in depth”. The material was considered ‘read’ if the reader’s eyes moved across one or more lines of print from
left to right. When at least one half of any text was read, it was considered ‘read in depth’. ‘Processing’ in Garcia and Stark’s termininology means just looking at an item, probably but not necessarily acquiring information. A major finding in Garcia and Stark’s study was that readers not really read but rather scan newspapers. At certain so-called entry points they stop scanning and start reading the story that the entry point belongs to. Garcia and Stark’s discussion ends by defining newspaper design as the task “to give readers material that is worthy of their scan, that makes them stop scanning and start reading” (ibid. 1991:67). Pictures and graphics were identified as the main entry points, followed by front-page promotion boxes. Readers usually enter the page through the dominant photo and then move to a prominent
Lund University Cognitive Studies- LUCS 127 ISSN 1101-8453 © 2005 by authors
3
headline or another dominant photo.
The entry point concept in Garcia & Stark is a central one. Entry points are entities on a spread that draw initial attention: Pictures, headline, quotes, fact boxes could be entry points. But virtually all elements, anywhere, even editorial text, can rve as entry points into reading. This, the authors arg
ue, is strong evidence against the classical inverted information pyramid. The inverted information pyramid tells us that important information should be given most space, and placed at the top, while gradually less important information should be given less space further down. Instead they argue in favour of a creative design dividing stories into substories and fact boxes with a marked graphical support.一尘不染什么意思
According to Garcia and Stark, headlines, cutlines and briefs are procesd – i.e. looked at – often and in depth. However, only 25% of the text itlf are procesd. On average, text processing is highest in the news ction, lowest in the sports ctions. Readers devote more time to photo groupings when they are in colour. Size increas the attraction to a photo. Unfortunately, in their book, Garcia and Stark do not prent their underlying statistical analysis, which makes it difficult for the reader to interpret Garcia and Stark’s results scientifically.
Hann (1994) studied 12 readers of the Copenhagen newspaper Det Fri Aktuelt. Hann’s unpublished study investigates the order in which objects on spreads were scanned. He found that pictures are en first, then icons and graphics followed by headlines of different sizes and text, with form items being obrved last. Hann’s hierarchical order can be en as a more preci version of Garcia and Stark’s entry points.
Hann also investigated readers’ priorities with respect to the length of articles, their placement and genre (news, features, debate, sport etc.). He measured how many centimetres of text article were read by subjects and calculated a depth index (the amount of text that has actually been read, in % of the whole text length) and a total respon index
(averagedepth index for a number of readers). His results show that only short articles are fully read. The longer the article, the smaller the proportion of it will be read. In this respect, Hann’s results remble the result in Garcia and Stark (1991) that a mere 25 % of all articles are en, and only 12 % are read dee per than half of its length. Hann’s subjects were most engaged in reading at the pages 2-6 and after that, their interest in reading decread with the exception of the last page of the newspaper. Stories on the left-hand side of the spread were en significantly earlier than stories on the right hand side. News, features and debates had approximately the same total respon index whereas sport articles got significantly lower total respon indices. Hann stress the importance of designing the newspaper layout so that it helps the reader to prioriti among all the information on the spread.
Hann’s study did not confirm the Garcia & Stark finding that colour objects attract attention more than do greyscale objects.
Jophsson (1996) studied the colour vs. greyscale effect for photos in 4 manipulated newspaper pages, which 32 subjects were allowed to look at for 10 conds per page. Her findings indicate that the position of the photo on the page is more important than the colour/greyscale factor. Her subjects looked at the top of pages earlier than at the bottom, and the earliest position was the upper left, regardless of colour.
Holmqvist, Holsanova, Barthelsson and Lundqvist (2003) developed a reading filter to study how much reading takes place in newspapers and net papers. When reading on a spread, readers exhibit one of two very distinct behaviours: They either scan or read. Reading is a well-defined movement of the eye from left to right, with approximately one stop at each word and small jumps called saccades between them. It is assumed that readers process the text when they do this.
Lund University Cognitive Studies- LUCS 127 ISSN 1101-8453 © 2005 by authors
4
Figure 1: Reading across text.
Scanning is accompanied by quite different eye-movements: Saccades are much longer and can go in practically any direction. The eye often hits pictures, headlines and drop quotes during scanning. Only a few words can be procesd at each stop. The purpo of scanning appears to be to find interesting entry points; points where deeper reading can continue.
Results show that newspaper readers read 55 % of the time and scan 45 %. Net paper readers read only 44 % of the time, scanning 56 %. All differences were significant. Holmqvist et. al (2003) argue that this indicates that newspapers are a more efficient medium for readers: In newspapers, they can read more and have to scan (arch) less than in net papers. The explanation to this difference is in the large design differences between net and paper: The many small items on the net pages, as well as the cumbersome page turing, force readers to do more scanning
千岛湖游玩攻略T ECHNOLOGY
In the first eye tracking studies of newspaper reading, eye movement data were recorded as videos .
统计英文
妹妹是哥哥的A video recording of a reader going through a newspaper is very illustrative. It is possible to follow the reader’s eye movements online and to e on a very detailed level which elements on the pages interested that particular reader. If however, we want to come to general conclusions on reading behaviour, the video must be analyd frame by frame according a defined number of parameters that can be summarid over a sample of different readers. This is very time consuming. Garcia and Stark (1991) judge the work behind their somewhat larger study to equal a Ph.D. thesis, or four years of full time work.
Absolute data coordinates reflect an exact point of gaze on a specified plane. For newspaper reading, this technology was first ud in recordings of eye-movements when reading tabloid newspapers (Holmqvist et. al 2003). The reason for the limitation to tabloid format was that recordings had to be made by so-called remote equipment, which require a stable spatial configuration of both reader and newspaper as well as the eye. The degree to which the eye-tracker could follow the reader’s head and eye-movements was limited, and such an eye tracker can therefore not cover a broader field of vision than that of a tabloid spread. This technology could not be expanded to studies of broad-sheet.
特工皇妃楚乔传2The solution ud in the current study combines eye-tracking with headtracking and a virtual reality
model of the table and newspaper. Readers wear a helmet equipped with head-tracking and eye-tracking devices which measures both the position and the direction of the head as well as the direction of the eye in the head. The combined head-eye vector hits the newspaper plane at a certain position, the absolute coordinates of this point of gaze on the plane are
recorded.
Figure 2: Scanning across a newspaper. by the eye tracking software.
Lund University Cognitive Studies- LUCS 127 ISSN 1101-8453 © 2005 by authors
5
There is therefore no need to watch the video in slow motion and encode specific parameters frame by frame. Instead, the computer does the decoding bad on the absolute data coordinates reprenting the point of gaze and knowing which spread corresponded to the specified plane when data were recorded.
Not only does this allow us to record for simplified analysis on both tabloid and broadsheet. Some other clear advantages of this new technology are the higher spatial precision, the better temporal resolution and the possibility to compensate for calibration off-ts.
This study is to our knowledge the first eye-tracking study of newspaper reading that utilis eye- and headtracking technology. The time cost for preparations, data recording and analysis was approximately four man-months, to be compared to the estimated four man-years in Garcia and Stark (1991).
P URPOSE
The purpo of the prent study was to examine the effect of local design factors on readers’ visual behaviour.
D ATA COLLECTION
Stimulus material: Chief designers from 17 Nordic newspapers each nt us two spreads from the current issue (Danish & Swedish) or a very recent issue (Norwegian & Finnish). Virtually all spreads were chon from the ctions of the papers that have much editorial material. Very few spreads had larger ads.
The designers divided each spread into at most 16 areas of interest, and predicted 1) in what order the areas were obrved, 2) for how long each area was obrved, and 3) how deep each area was read. Areas of interest were usually texts with accompanying headline and picture. Sometimes an area could be an ad. Sometimes picture and accompanying text were divided into parate areas. Subjects : A total of 26 readers participated, i.e. 6 or 7 readers for each of the four languages.
Equipment : For eye movement recordings, we ud the SMI iView X Headt with Polhemus headtracking. The headtracker knows where the head is, while the eyetracker knows to where the eye is directed. Together, they allowed for freedom of movement – necessary when reading broadsheet newspapers – while at the same time giving us absolute data coordinates in the coordinate system of the newspaper – much simplifying data analysis.
Figure 3: Participant fitted with head-mounted eye-tracking equipment, reading the newspaper
Procedure: Newspapers were laid out on a slightly inclined table in front of the reader. The reader was asked to put on the headt, esntially a bicycle helmet with cameras and a mirror (e figure 3). After a short calibration, they were asked to read the spreads as they would normally read a newspaper. Readers turned pages themlves, and were allowed to go through pages backwards if they wanted to. They read for as long as they wanted.
