Elsevier

Computers in Human Behavior

Volume 30, January 2014, Pages 206-221
Computers in Human Behavior

Research Report
Don’t miss your train! Just follow the computer screen animation: Comprehension processes of animated public information graphics

https://doi.org/10.1016/j.chb.2013.08.010Get rights and content

Highlights

  • Computer graphic animation used to deliver public messages instead of announcement.

  • Animated sequential presentation better understood than static simultaneous format.

  • Animation promptly trigger in traveler’s mind task-appropriate script of events.

  • Eye movements showing animation enhance efficient task-related strategies.

  • Animation provide best condition for segmentation-composition of chain of events.

Abstract

Computer graphic animated information displays have the potential to communicate public information in situations where normal announcement types are ineffective. This study used eye tracking techniques to analyze comprehension mechanism of event-related information on railway traffic disruptions presented via different graphic formats presented on computer screen. 86 participants were asked to understand series of traffic disruption messages delivered via four purely visual formats: Static simultaneous, Static sequential, Animated simultaneous and Animated sequential. Across these four conditions, and contrary to the most common materials used in the studies on animation comprehension, the sequentiality and the animated properties of the entities of the presentation were not confounded. Results revealed the Animated sequential displays were the most effective presentation type. Eye tracking data showed why an animation facilitates comprehension of public information graphics: it enhances processing strategies which provide the best condition for segmenting and composing the causal chain of the events provided in the message.

Introduction

The growth of technology has seen computers become widely used to deliver real time public information messages on platforms ranging from large displays screen to individual mobile devices. Such displays are increasingly common in railway stations, airports, bus terminals and city centers. Information that is useful or even critical to the public can change frequently and suddenly. Today’s global society is highly dependent on various forms of travel and any disruptions to those services can have major consequences. Travelers therefore need access to efficient, effective information sources so that they can respond rapidly and appropriately to such disruptions. Recent major travel disruptions in Europe exposed severe shortcomings in existing ways of presenting information that exacerbated an already chaotic situation.

Unfortunately, the design of important public information to be displayed on electronic screens is rarely based on empirical studies. Rather, the basis for their design tends to be intuition or traditional text-based approaches such as those that have evolved for railway station loudspeaker announcements and information boards. An overarching motivation for the research reported in this paper is to provide information that could guide the design of more effective and efficient railway disruption messages for the National French Railway Organization (SNCF).

The present paper explores the potential of dynamic graphics-based display screens, composed of series of dynamic pictographs, as alternatives to current spoken or written approaches for presenting public information about train traffic disruptions. Such information should be readily available and comprehensible to a broad cross-section of travelers1 including the elderly, those with hearing impediments, and foreign travelers who are not familiar with the local language. Graphic messages could also assist travelers in general who cannot hear the usual loudspeaker announcements properly because of ambient station noise or their distance from loudspeakers. The experimental study presented here investigated the effect of different types of graphic design format on the comprehension of traveler action messages.

The conventional spoken railway announcements typically follow a standard format. In French stations, the texts that are spoken for these announcements are structured in terms of a succession of related events such as: “Your attention please. Contrary to the information that has been displayed, the Regional Train (TER) number 3458 for Paris, will not start from platform B but will start instead from platform D. Please take the stairs to access platform D”. The formulaic nature of these announcement texts gives them much in common with script schemas (Armbruster, 1996, Mandler, 1984). In some railway systems, disruptions can be a part of the everyday context of train travel. Travelers often have to comprehend and respond appropriately to disruption-related information under considerable time pressure. How to convey train announcements non-verbally? Graphic messages, cycling on computer screens distributed around the station could be an effective alternative to traditional spoken or written announcements (Fig. 1).

In railway stations, announcement texts for the most common disruption-related loudspeaker messages are structured as an ordered sequence. Examples of messages include switching of a railway platform, delay of a train, cancellation of a train, reduction of services due to a strike, or consequences of extreme weather (snow, ice, etc.); and a message related to general safety, passing of a nonstop high speed train on the railway near the passengers. Regular train travelers typically possess knowledge structures in long term memory that represent such event-related information as scripts (event schemas) that include various slots. The content of most messages that travelers hear in train stations may also be thought of in terms of categorical slots that can be populated with specific types of information. A typical structure is: (i) a warning (“your attention please”) (ii) the cause of the disruption (“because of the snow”) (iii) the effect of that disruption (“the regional train number1556 for Paris, will be delayed by 15 min”) and (iv) the action that the traveler should take as a result (“for further information please check the central panel situated in the main hall”). Depending of the type of disruption, this series of four event slots may or may not all be filled. For example, in the following message, the causal event is not given: “Your attention please, contrary to the information displayed previously, the Regional Train (TER) number 3458 for Paris, will not start from platform B but will start from platform D, please take the underground pathway”. Each event slot could be defined as an “episode” which applies on different (but limited in numbers) objects (such as a train, a platform, a stair, and city name panels). Regular train travelers in France are often exposed to such messages and so typically develop knowledge structures in long term memory representing relevant event-related information as scripts for various scenarios. In order to convey train announcements, which are (well) known events episodes with their usual objects, series of dynamic visual icons or dynamic pictographs could be developed.

In everyday life, single static pictographs are widely (and efficiently) used in different domains such as traffic and road signs, public areas, human–machine interfaces, industrial areas (for safety purposes), and healthcare centers. Previous studies in this field of “information design” and or “iconic communication” have been focused mostly on taxonomies of pictographs and on the testing of their usability (Bodenreider, 2004, Familant and Detweiler, 1993; Isherwood, McDougall, & Curry, 2007; Rogers, 1989, Yazdani and Barker, 2000, Zwaga et al., 1999). Conveying information with pictographs “relies on pre-established code and convention” (Nakamura & Zeng-Treitler, 2012, p. 535) and implies the meaningful relation between the referent and the pictorial representation is quite transparent, direct or known. In the example of the train disruptions, the meanings of the referents in the context of the railway station are common. Such familiarity could help in understanding the graphic representation of the events and objects of the disruption.

In previous studies on visual pictographic representations, three levels of external pictorial representations of the referent have been identified (see the recent taxonomy by Nakamura and Zeng-Treitler (2012), and the reviews in the book by Zwaga et al. (1999). (i) Direct representations use the visual similarity between a pictograph and its referent (for example, the picture of TGV train to refer to a TGV); (ii) arbitrary representations use social convention (for example the letter P to refer to a parking area, the letter I to refer to an information point); (iii) indirect representations use semantic relation, and semantic association, between a pictograph and its referent (for example, the picture of a fork and a knife to represent the concept of restaurant, or the picture of a clock to represent time and delays, see Fig. 2, first message). Different subcategories of semantic association were recently proposed by Nakamura and Zeng-Treitler (2012) in the domain of healthcare: comparison or contrast, exemplification, semantic narrowing, physical decomposition, temporal decomposition, body language, metaphor and contiguity.

The visual alternatives to spoken disruption messages built in this study were closely in line with recent taxonomies and associated recommendations. However, in previous studies, and in everyday situations as well (i) one graphic representation is usually presented alone in a single picture format (but see exceptions with series of static simultaneous pictographs in Dewar & Arthur, 1999, or in Nakamura & Zeng-Treitler, 2012); (ii) further, pictographs refer mostly to entities and objects (more than 86% of the pictographs used in the study by Nakamura and Zeng-Treitler (2012), cf. page 542) and rarely to dynamic events and actions; (iii) until now, the presentation of the graphic is always (at least mostly) static. Finally, very few studies on pictographic communication were focused on comprehension mechanism.

Because train messages are composed of series of predetermined categorical slots, each message must be presented with a series of several pictures, instead of a single pictograph. In the present study, we used a composition of (dynamic) pictographs to depict complete messages which convey a complex meaning about train disruptions. Further, in the train messages, events usually described via the loudspeaker announcements, have key dynamic dimensions. They do not deliver information about static statement like do a majority of the pictographs printed on physical panels (Nakamura & Zeng-Treitler, 2012). Finally, train messages are structured according to a fixed temporal sequence. How to use visual graphics to present a sequence of dynamic concepts? It could be relevant to follow “the syntax structure” of the loudspeaker well known messages and also to build sequences of graphics in which the relation between the referent and the representation would be as close as possible.

The goal of the two research questions investigated in the present article was to study (i) whether it helps comprehension of visual messages to present the graphics of dynamic events and concepts dynamically rather than statically, and (ii) whether the graphics should be presented sequentially or simultaneously.

Current technology allows graphics and pictographs to be animated rather than merely static. In contrast with static graphics, animations are able to deliver explicit, precise and continuous information about the fine detail of events and process dynamics (Lowe and Boucheix, 2010, Lowe and Schnotz, 2008, Tversky et al., 2002). Most previous research on the comprehension of animated graphics has concerned formal instruction in scientific and technical domains (e.g., Bernay and Bétrancourt, 2009, De Koning and Tabbers, 2011, Höffler and Leutner, 2007, Lowe and Schnotz, 2008, Ploetzner and Lowe, 2012, Scheiter and Eitel, 2010). However, by focusing on the provision of public information in an informal, non-educational context, the present investigation extends the scope of animation research into a rather different but most important area that hitherto has remained essentially unexplored in this way. Its goal was to test the effect of different graphic presentation formats on comprehension of train disruption messages.

In previous literature, one main reason why animated graphics are supposed to be better than their static counterparts is because they present information about the microsteps of dynamic processes (Bétrancourt, 2005, Lowe and Schnotz, 2008, Tversky et al., 2002). The presence of this fine detail should assist viewers to build higher quality mental model of the referent’s dynamics (Boucheix and Lowe, 2010, Hegarty, 2004, Hegarty et al., 2010, Lowe and Boucheix, 2008, Lowe and Boucheix, 2011, Lowe and Boucheix, 2012). Two recent meta-analyses (Bernay and Bétrancourt, 2009, Höffler and Leutner, 2007) reported an overall superiority of animations over static graphics with medium effect sizes (d = .31 and d = .37 respectively). However, other studies (for example, Boucheix and Schneider, 2009, Hegarty, 2004, Kriz and Hegarty, 2007, Mayer et al., 2005) have shown that under some circumstances a series of static pictures (key frames depicting the main steps of the referent process, Spanjers et al., 2010, Wouters et al., 2008) can be as good as an animation for fostering comprehension.

In contrast with previous studies, the present research explores the potential of dynamic visual displays, not to help comprehension in an educational end, but to promptly trigger in the traveler’s mind a task-appropriate script of the relevant events. When travelers are under extreme time pressure to catch their trains, the bottom-up features of the information they receive from an external representation (in this case, visualizations) should perfectly map to top-down aspects of their existing internal representations. By this reasoning, if travelers’ mental scripts for railway disruption situations themselves have key dynamic dimensions, then it could be expected that a dynamic presentation with animations of the events involved would allow a more efficient and effective match to the corresponding internal representations than would a static presentation. Therefore, the advantage that animation could have over static picture, in the case of public information, would not be so much in showing the microsteps of the dynamic process (that could be shown with series of static pictures as well). But, in the train announcements, animation could work as quickly clarifying a dynamic concept (or event), like the train that ‘arrives’ at a platform, a platform that ‘changes’, and travelers having to ‘move’ from one platform to another platform. The dynamic presentation would have a positive effect on comprehension if these dynamic concepts are not well understood from static pictures. A key issue in train announcements consists in understanding what is going on and knowing what to do. Travelers may not recognize so easily the dynamic events when they are displayed as static pictures.

In the context of animation research, much previous experiments comparing animated with static graphics has set a continuous animation of the dynamic process against series of static pictures (for example, Arguel and Jamet, 2009, Boucheix and Schneider, 2009, Bétrancourt, 2005, Hasler et al., 2007, Höffler and Leutner, 2007, Imhof et al., 2011, Imhof et al., 2012, Kim et al., 2007, Kriz and Hegarty, 2007, Lowe and Schnotz, 2008, Lowe et al., 2010, Mayer et al., 2005). With a few exceptions (e.g., Lowe et al., 2010), typically, the series of static pictures is derived from the animation by extracting single frames that show only the main steps of the dynamic content (e.g., key frames). This extraction inevitably reduces the amount of the information available compared with that in the full animation format.

Depending on the study, these sets of static pictures are usually presented to the participant in two possible delivery regimes: sequential or simultaneous. With the sequential regime, two variants can occur (i) replacement in which each of the pictures is delivered to the same spatial location in order, one picture after another – the first frame disappears when the second appears, and so forth. Under such circumstances, the new information may override the predecessor information in working memory. Further, there may be interference with the perception of continuity of movements (Lowe et al., 2010) and increased cognitive demands because the learner must maintain the first picture in working memory while the second is processed (Ayres and Paas, 2007, Lowe, 1999, Paas et al., 2007). (ii) Addition in which the static pictures appear sequentially but are retained (each on its own piece of screen real estate) once they have been presented so that the set of pictures is built up progressively. This approach was used in the present study because it avoids some of the processing challenges associated with replacement.

With the simultaneous regime, all pictures in the set are delivered to adjacent locations on the screen at exactly the same moment. Because the full set of pictures is accessible at once, the learner has the opportunity from the outset to perform direct visual comparisons between different steps of the depicted process. However, simultaneous presentation does not provide the temporal staging effect that is available with sequential presentations. As a result, viewers receive no explicit guidance as to the order in which pictures should be inspected. They could however invoke knowledge of reading direction conventions to navigate through the picture set. However, this format could provoke in an initial searching activity which induces undesirable delays in situation where there is time pressure (as can be the case for rail travelers).

Although previous research has used both sequential and simultaneous presentations for static graphics (see for example, Boucheix and Schneider, 2009, Imhof et al., 2011, Imhof et al., 2012, Kim et al., 2007, Lowe et al., 2010), animations are usually presented in only the sequential modality. In such presentations, each new frame is an alteration of the previous frame, so that in most studies, sequentiality of the presentation is therefore inevitably confounded with animation of the display’s components.

This is an important issue because recent research indicates that sequential and simultaneous presentations are anything but equivalent and could result in significant differences in task performance. The few studies that have been conducted on sequentiality with static visualizations have led to inconsistent results. For example, in the case of static graphics depicting steps in the functioning of a three pulley system, Boucheix and Schneider (2009) found superior comprehension for simultaneous presentation of a five-picture series than for their sequential presentation. More recently, in a fish locomotion pattern classification task from visualizations, Imhof et al., 2011, Imhof et al., 2012 found the similar results, with participants in an animated condition outperforming those using nine static sequential pictures. However, participants in the static simultaneous condition (a row format) had the same performance as participants in the animated condition. In contrast, the study with six graders on the understanding of a bicycle pump by Kim et al. (2007) showed that static simultaneous visualizations led to lower comprehension scores than animated and static-sequential visualizations. Further, in a task where learners were required to arrange kangaroo hopping pictures into the correct order, Lowe et al. (2010) found that a static sequential presentation of the learning material was better than an animated and or a static simultaneous presentation. In the case of animated graphics, in a recent study by Morand and Bétrancourt (2010) a traditional sequential animated presentation of meiosis was compared with a simultaneous animated version in which small animated segments of the process were delivered simultaneously spread across a screen. It was found that transfer task performance of participants given the simultaneous animation was superior to that of those in the sequential condition.

More generally, Lowe et al. (2010) showed that aligning affordances of the graphics with the task requirements could be an important aspect of visualization design. This means that here, outside of the educational animation field, graphic presentation should be aligned on the structure of the traveler’s internal scripts of the train disruptions, in order to map with them. These scripts, as the announcement texts for the most common disruption-related loudspeaker messages, are structured as an ordered (and chronological) sequence. When watching the visual announcements, travelers will use their mental scripts (that are based upon the structure of these loudspeaker messages). Consequently a sequential presentation of pictures would allow a more efficient and effective match to the corresponding internal representations than would a simultaneous presentation.

In the present study, four purely graphic presentation conditions were used to deliver train disruption messages: animated sequential (AnSeq), animated simultaneous (AnSim), static sequential (StaSeq), and static simultaneous (StaSim).

This experiment was a crucial part of a broader project composed of a program of experiments which intended to address separately different issues of the comprehension processes of train disruption messages.

A first issue concerned the intended population: all travelers in train stations, and particularly, travelers with hearing impairment and old persons. Regarding this point, experiments had previously been conducted with small samples of hearing impaired and elderly people (see Boucheix et al., 2010, Paire-Ficout et al., 2013). In those studies, only three presentation conditions were compared; animated sequential, static sequential and static simultaneous. The results, for these two samples, indicated that comprehension was higher with animated than with static presentation. There was also some indication of an advantage for sequential presentation. Another experiment is currently in progress with different categories of hearing impairments.

A second important issue is about the speed of comprehension of the messages in the context of railway stations. Indeed, travelers have sometimes to catch a train under time pressure. This fact raises the issue of the ecological value of the experiment. A future experiment is planned yet in a more “real” train station context (using virtual reality) in which manipulation of time on task and starting point of visual information will be examined.

The present study addressed another crucial issue related to the comprehension processes of every part of the pictographic information within each message. For that reason, in the experiment was used a controlled comprehension task in which participants were given the same information and the same controlled amount of time (thus allowing them to have enough time) to process all the pictographs. The main issue was: is the whole pictographic message understood and how, depending of the experimental conditions? Such investigation was required given the general aim of the study about using animation to make people understand a dynamic situation and how to convey train announcements in a series of visual pictographs.

The goals of the present study were therefore: (i) to make a more refined and comprehensive comparison of the effect of different graphic message conditions formats on comprehension, by teasing out the individual contributions of graphic format (animated versus static) and delivery regime (sequential versus simultaneous); (ii) to investigate underlying cognitive processing reasons for the differences in comprehension of public information displays that employ different graphic formats and display regimes using eye tracking techniques; (iii) to study a larger and less restricted sample of participants; and (iv) to examine the comprehension of each part of the message given that each participant had the opportunity to process every part of the information delivered in the message.

Our prediction was that a sequentially presented animated condition would be of most help to participants in following the course of depicted events and mapping that information to relevant script schemas in long term memory. More specifically, in the case of the train messages, the sequential animation is most consistent with how internal scripts are likely to be organized with respect to spatiotemporal information. It was therefore expected that (i) Hypothesis 1, comprehension performance for the sequential presentation regime would be superior to that for the simultaneous presentation regimes and (ii) Hypothesis 2, comprehension performance for animated format would be superior to that for static format. As a consequence of expected both main effects, the optimal condition would be when the animated format is presented sequentially.

A growing body of research has used eye tracking to study comprehension of, or learning from, multimedia presentations (including animations). Most of these investigations have dealt with text-picture combinations, Canham and Hegarty, 2010, Jarodzka et al., 2010, Meyer et al., 2010, Schmidt-Weigand et al., 2010, Van Gog and Scheiter, 2010. Less research has been conducted on animated graphics presented without accompanying text (Boucheix and Lowe, 2010, Boucheix et al., 2013, De Koning et al., 2010, Lowe and Boucheix, 2011, Scheiter and Eitel, 2010) and to our knowledge almost no experimental studies (Zwaga et al., 1999) has addressed public transport information.

Because the research reported here investigated not only static graphics but also their animated counterparts, our use of eye tracking extended beyond measuring just total viewing durations (dwell time). In addition to total viewing duration in Areas of Interest (AOIs), we used eye tracking to investigate how the different presentation conditions may influence attention direction during visual search of the messages, particularly the order in which the graphic material constituting each message was processed for the various message versions. For this purpose, we determined the number of fixations participants made before attention (first fixation) arrived on the nominated target either a whole frame or an entity within a frame. For this measure, taking only strict raw data into account may result of 20 ms glances on an AOI that would be not meaningful. So, we used a minimal amount of time spent on AOI based on raw data with a fixation definition of 100 ms. This measure was termed the pre-target count and could apply either to a whole picture or to a specific graphic component within a picture.

Different eye movement patterns were expected for each of the four presentation conditions. More specifically, for the animated sequential condition, it was expected that, (i) Hypothesis 3, the pre-target count for each of the four pictures comprising a message would strongly reflect the onset order of the pictures. This is because the dynamic unfolding of the animated sequential condition would exert a powerful influence on viewers’ attention resulting in them closely following the sequence of events. This ‘following’ behavior (c.f. Lowe et al., 2010) was expected to happen not only across the series of four pictures (see Fig. 1) but also within each picture where animated components (train, platform letters, arrows, etc.) could also appear sequentially. For example, in picture two of Fig. 1 (animated sequential version), the train moved before the platform letter was crossed out. Our expectation was that the pre-target count for picture components would reflect such appearance orders.

For the static sequential condition, it was expected that (ii), Hypothesis 4, the pre-target count would also strongly reflect the picture onset order. However, close following behavior was expected only across pictures because within-picture components all appear at the same time. For the static simultaneous condition, it was expected that (iii) Hypothesis 5, the pre-target count for each picture would not reflect the order of the pictures as strongly because their sequencing was only spatial rather than also being temporal. In the absence of any temporal influence on attention direction, there would be less coordination with the events schema both across and within pictures. Such less ordered searching behavior, not coordinated with the script of the events depicted by the pictures, would apply across pictures as well as inside pictures.

For the animated simultaneous condition, it was also expected that (iv), Hypothesis 6, although pre-target counts for within-picture components would reflect the order in which they appeared, there would be much less correspondence between pre-target counts for the whole pictures and their spatial ordering.

Regarding the total viewing time for each picture it was expected that (v), Hypothesis 7, participants would spend more time on the pictures depicting the cause and effect (main events, pictures 2 and 4) of the disruption message than on the pictures showing the warning and the recommended action (pictures 1 and 4). Time spent on the familiar warning (picture 1) would be relatively short (essentially involving recognition only) because it is a widely used, standardized symbol that can be processed holistically. The cause and effect pictures (2 and 3) are specific to the railway context and differ according to the particular situation being represented so viewer processing should take longer. However, once interpreted, they trigger a relevant event schema that includes a likely response. Picture 4 therefore serves a mainly confirmatory function so less processing time is needed.

Section snippets

Participants

A total of 86 undergraduate students (psychology and education areas, 13 males and 73 females) took part in the study for course credit. Participants were all native French speakers with normal vision accuracy, (M = 19, 25 years, SD = 2, 54). Before the experiment, participants completed a multiple choice question on how frequently they caught a train, including four levels: “at least one time a week, at least one time a month, at least one time a year, or never”. These four levels of train use were

Discussion and conclusion

This research investigated the comprehension of computer screen animated public information graphics about train disruption messages in railway stations. Manipulating two presentation factors, sequentiality and animation, four visual presentation conditions, alternatives to traditional spoken announcements, were compared: animated sequential, animated simultaneous, static sequential and static simultaneous. In the approach followed in this paper, animations were used to explore the potential of

Acknowledgments

This research was a part of larger project called “SurDyn” which aimed at the accessibility of public information for everybody in transportation. This research was realized in collaboration with the IFSTTAR (The French institute of science and technology for transport, development and networks) and funded by the PREDIT, which is a department of the French National Research Agency, A.N.R (2009–2013) and also partly founded by a French special interest group for research (GDR-CNRS, 3169,

References (51)

  • B. Imhof et al.

    Learning about locomotion patterns from visualizations: Effects of presentation format and realism

    Computers and Education

    (2011)
  • H. Jarodzka et al.

    In the eyes of the beholder: How experts and novices interpret dynamic stimuli

    Learning and Instruction

    (2010)
  • S. Kriz et al.

    Top-down and bottom-up influences on learning from animations

    International Journal of Human–Computer Studies

    (2007)
  • R.K. Lowe

    Animation and learning: Selective processing of information in dynamic graphics

    Learning and Instruction

    (2003)
  • R.K. Lowe et al.

    Cueing complex animations: Does direction of attention foster learning?

    Learning and Instruction

    (2011)
  • K. Meyer et al.

    Effects of animation’s speed of presentation on perceptual processing and learning

    Learning and Instruction

    (2010)
  • C. Nakamura et al.

    A taxonomy of representation strategies in iconic communication

    International Journal of Human Computer studies

    (2012)
  • R. Ploetzner et al.

    A systematic characterization of expository animations

    Computers in Human Behavior

    (2012)
  • Y. Rogers

    Icons at interface: Their usefulness

    Interacting with Computers

    (1989)
  • F. Schmidt-Weigand et al.

    A closer look at split visual attention in system- and self-paced instruction in multimedia learning

    Learning and Instruction

    (2010)
  • B. Tversky et al.

    Animation: Can it facilitate?

    International Journal of Human–Computer Studies

    (2002)
  • T. Van Gog et al.

    Eye tracking as a tool to study and enhance multimedia learning

    Learning and Instruction

    (2010)
  • B. Armbruster

    Schema theory and the design of content-area textbooks

    Educational Psychologist

    (1996)
  • P. Ayres et al.

    Can the cognitive load approach make instructional animations more effective?

    Applied Cognitive Psychology

    (2007)
  • M. Bétrancourt

    The animation and interactivity principle in multimedia learning

  • Cited by (0)

    View full text