ANALYSING EYE MOVEMENT PATTERNS TO IMPROVE MAP DESIGN
Kristien Oomsa, Philippe De Maeyera, Veerle Fackb
a
Ghent University, Department of Geography
Krijgslaan 281, S8, B-9000, Ghent, Belgium
b
Ghent University, Department of Applied Mathematics and Computer Science
Krijgslaan 281, S9, B-9000, Ghent, Belgium
{Kristien.Ooms, Philippe.DeMaeyer, Veerle.Fack}@UGent.be
KEY WORDS: map design, eye tracking, qualitative analyses
ABSTRACT:
Recently, the use of eye tracking systems has been introduced in the field of cartography and GIS to support the evaluation of the quality
of maps towards the user. The quantitative eye movement metrics are related to for example the duration or the number of the fixations
which are subsequently (statistically) compared to detect significant differences in map designs or between different user groups. Hence,
besides these standard eye movement metrics, other - more spatial - measurements and visual interpretations of the data are more suitable
to investigate how users process, store and retrieve information from a (dynamic and/or) interactive map. This information is crucial to
get insights in how users construct their cognitive map: e.g. is there a general search pattern on a map and which elements influence this
search pattern, how do users orient a map, what is the influence of for example a pan operation. These insights are in turn crucial to be
able to construct more effective maps towards the user, since the visualisation of the information on the map can be keyed to the user his
cognitive processes. The study focuses on a qualitative and visual approach of the eye movement data resulting from a user study in
which 14 participants were tested while working on 20 different dynamic and interactive demo-maps. Since maps are essentially spatial
objects, the analysis of these eye movement data is directed towards the locations of the fixations, the visual representation of the
scanpaths, clustering and aggregation of the scanpaths. The results from this study show interesting patterns in the search strategies of
users on dynamic and interactive maps.
INTRODUCTION
The goal of this research is to improve the quality of maps
towards the user which is related to the design of the map. But
how to define ‘a good map design’? The user has to be able to
interpret its content in the first place correctly, but also efficiently.
If a user can interpret the content of a map correctly, but only after
studying the map for a long time, it cannot be considered to have a
good design or a good quality towards the user. How users
interpret the content of a map, store this information internally and
retrieve it later on is related to the structure of their cognitive or
mental map (Montello, 2002; Slocum et al., 2001; Downs and
Stea, 1977). Therefore, to be able to improve the quality of the
map towards the user, insights in the user’s cognitive processes
while working with maps are necessary. Consequently, a suitable
method is needed to get in touch with the user’s cognitive map
during a user study. But maps have a spatial dimension which
cannot be neglected when analysing the results of a user study,
since it is inherently connected with how users interpret the map’s
content.
From the long list of techniques (see Rubin and Chisnell, 2008
and Nielsen, 1993 for an overview), the eye tracking method is
considered to be the most suitable method to conduct the user
study since the close link between a user’s eye movements and his
cognitive processes has been proven multiple times in the past
(Duchowski, 2007, Jacob and Karn, 2003, Poole and Ball, 2006,
Rayner, 1998). Furthermore, already in the 1970s the feasibility of
the tracking technique to study map use was demonstrated
(Montello, 2002). Recently, the eye tracking method has also been
used to study the design of maps and their usability: their
symbology (Brodersen et al., 2001), map animations (Fabrikant et
al., 2008), design of the map interface (Coltekin et al., 2009).
But the software packages accompanying the eye tracking devices
are not fully suitable to study the spatial dimension of the
movement data which is essential to understand how users
orientate the map and interpret the map’s content. Eye movement
data is essentially not that different from other movement data
such as GPS-tracks: a list of (x,y)-positions at a certain timestamp
t. The software package The Visual Analytics Toolkit (also known
as CommonGIS) was selected to visualize and analyze the
scanpaths since its suitability to summarize the eye movement
data is already demonstrated in the work of Fabrikant et al.
(2008). This package is developed by Andrienko G. and
Andrienko A. and its functionalities are described in a number of
articles (e.g. Andrienko et al., 2007, Andrienko and Andrienko,
2010). In the next section, the design of the study is described,
followed by the results, a discussion and a conclusion.
A special joint symposium of ISPRS Technical Commission IV & AutoCarto
in conjunction with
ASPRS/CaGIS 2010 Fall Specialty Conference
November 15-19, 2010 Orlando, Florida
STUDY DESIGN
Apparatus & Participants
The tests were conducted in the Eye Tracking laboratory of the
Department of Experimental Psychology, Ghent University. This
laboratory is equipped with an Eye Link 1000 device from SR
Research (Mississauga, Ontario, Canada) and sample a person’s
POR (Point of Regard) at a rate of 1,000 Hz (or once every ms).
The movements from one eye only are recorded during the tests.
The recorded eye movements of 14 subjects were analyzed all of
which were students and most of them studied courses at the
Faculty of Psychology and Educational Sciences, Ghent
University.
Stimuli & Tasks
Twenty demo-maps were presented to each user in a random
order. Each demo-map in the experiment had a simple
background – equal on all maps – with point objects (symbolising
cities) and associated name labels. On the right, a list with five
names was visible. The participants were asked to locate these
five names on the map and push a button each time they found a
name. By using this task, the user has to perform a visual search
on the map. After 50 seconds the map image was translated
horizontally over a fixed difference, simulating a pan operation
lasting one second. The list with the five names had also changed
during this translation: three new names were displayed and two
which were already in the former list (but on a different location).
Again the user had to locate these five names in the map and push
a button when they found one. The eye movements of the
participants were recorded during these twenty trials. The
combination of the time measurements form the button actions
and the location where the user was looking (derived from the eye
movement recordings) allows identifying if and which label was
found. The visualization of the eye movements gives insights in
how users search on the map and if any patterns can be detected
in this. An example of a demo-map is depicted in Figure 1: the
initial view on top, a view during the translation on the left and
the final view on the right.
This task corresponds to an operation which is actually executed
rather often by users on dynamic and interactive maps: the user is
trying to locate the position of an interest area. To be able to do
this, the user has to orientate the map and subsequently scan its
content to discover the position of a certain symbol, such as a
label. Patterns in these scanpaths give insights in how users
interpret and process the content of the map while trying to
retrieve information from it and which elements have an influence
on these scanpaths. This information in turn allows keying the
visualization parameters of maps (such as the position of labels)
to the actual cognitive or mental map of the user. Difficulties in
the interpretation of the map and thus in locating the labels
indicate usability problems which would show in the visualized
scanpaths of the subjects, through for example very long and
chaotic scanpaths.
Figure 1. Example of a demo map with on top the initial view,
bottom left a view during the pan-operation and bottom right the final view
A special joint symposium of ISPRS Technical Commission IV & AutoCarto
in conjunction with
ASPRS/CaGIS 2010 Fall Specialty Conference
November 15-19, 2010 Orlando, Florida
RESULTS & DISCUSSION
As illustrated in Figure 2, the massive amount of eye movement
data quickly leads to overcrowded visualizations from which no
meaningful conclusions can be drawn. Even when only
visualising the eye movements of one person, no patterns can be
detected in the movement data due to the massive amount of data
gathered with the eye tracking study. The Visual Analytics
Toolkit includes a wide range of possibilities to select, aggregate,
summarize and visualize the movement data. These analysis and
visualization tools are crucial to detect patterns in the eye
movements which can provide insights in how users orientate a
map.
Figure 2. Visualization of the eye movements from all users (left)
or even from only one user cannot be used to detect patterns
One interesting selection technique available in The Visual
Analytics Toolkit is the time function which allows defining time
intervals; only the movement data which occurred in this interval
is visualized. This is an ideal tool to investigate how the scanpaths
of the users evolve over time which provides crucial insights in
how users orientate the map information and consequently
construct their cognitive map. Below in Figure 3, a time series of
[0 - 10]
the scanpaths of all users during all trials is depicted. The first
picture depicts the start-situation and every subsequent picture
shows the scanpaths in a next interval of 500ms. Picture eight in
Figure 3 shows the eye movements in the time interval
[49,500ms-50,000ms], that is the time interval right before the
simulated pan-operation.
[0 - 500]
[500 – 1,000]
[2,000 – 2,500]
[2,500 –3,000]
[1,000-1,500]
...
[1,500 – 2,000]
A special joint symposium of ISPRS Technical Commission IV & AutoCarto
in conjunction with
ASPRS/CaGIS 2010 Fall Specialty Conference
November 15-19, 2010 Orlando, Florida
[49,500 – 50,000]
[50,000 – 50,500]
[50,500 –51,000]
[51,000-51,500]
[51,500 – 52,000]
[52,000 – 52,500]
[52,500 –53,000]
[53,000-53,500]
Figure 3. Time series of the user’s scanpaths with subsequent intervals of 500ms
The first row in Figure 3 shows that all users start looking at the
map near its centre and are subsequently drawn to the list with
five names which is read during a certain amount of time. Next,
the users start looking for the five names in the top part of the
map. Since the eye movements of all participants are visualized, a
general pattern in the search behaviour is visible. After 1.5
seconds more users start scanning the lower part of the map, but
still a large part is of the eye movements is situated in its upper
part. Only after 2500ms, the search pattern of the users is more
homogeneously spread over the entire map area, which continues
during the remaining intervals. The interval right before the start
of the simulated pan-operation also shows this homogeneous
distribution of the scanpaths. During the first part of the simulated
pan-operation (interval [50,000ms-50,500ms]) this homogeneous
distribution continues. This is different during the second part
[50,500ms-51,00ms]: the users already start looking at the newly
displayed list with names. At this moment, the list with the names
has already changed, but the map is not yet completely in its final
position. However, most users already start reading this newly
displayed list before the end of the simulation. After the
simulation, most users are still reading the list with names and
some of them already started searching for the names. In this case,
the scanpaths are directed more diagonal across the map, but are
still mainly situated in the upper part. Another element which can
be noticed here is that a larger subset of the scanpaths goes
directly to the left part of the map, which was already visible
before the simulated interaction. This indicates that the users
remember this part of the map and use this information to locate
the names in the list in a more efficient way. Two seconds after
the end of the simulation, the pattern of the scanpaths starts to
look more disperse which also continues until the end of the tests.
how users orientate it and interpret its content. Visualising and
analysing the eye movements of the users with the Visual
Analytic Toolkit allows detecting patterns in the search behaviour
of the users while performing a visual search in the map. The
preference of nearly all users in the study to start their visual
search in the upper part of the map, both in the initial view and
after the interaction, is such a pattern which could be detected by
visualising the eye movements in subsequent small time intervals.
Another patterns detected using this visual analytic method is the
more diagonal search strategy which is present after the
interaction. This information is important to obtain insights in
how the user’s cognitive map is constructed and how information
is subsequently extracted from it. Keying the designs of maps to
the user’s mental map ensure the construction of more effective
maps towards the user. This is especially important when dealing
with dynamic and interactive maps which are found in a wide
range of applications on the Internet today.
REFERENCES
Andrienko, G., Andrienko, N. & Wrobel, S. (2007) Visual
analytics tools for analysis of movement data. SIGKDD
explorations, 9, 38-46.
Andrienko, N. & Andrienko, G. (2010) Spatial Generalization and
Aggregation of Massive Movement Data. IEEE Transactions on
visualization and computer graphics.
Brodersen, L., Andersen, J. H. K. & Weber, S. (2001) Applying
the eye-movement tracking for the study of map perception and
map design. Copenhagen, Denmark, National Survey and
Cadastre.
CONCLUSION
To be able to improve the quality of maps, and thus their design,
towards the user, it is crucial to get insights in the user’s cognitive
process while working with the map. The spatial dimension of the
maps cannot be neglected in this, since it is closely linked with
Coltekin, A., Heil, B., Garlandini, S. & Fabrikant, S. I. (2009)
Evaluating the effectiveness of interactive map interface designs:
a case study integrating usability metrics with eye-movement
analysis. Cartography and Geographic Information Science, 36,
5-17.
A special joint symposium of ISPRS Technical Commission IV & AutoCarto
in conjunction with
ASPRS/CaGIS 2010 Fall Specialty Conference
November 15-19, 2010 Orlando, Florida
Downs, R. M. & Stea, D. (1977) Maps in Minds. Reflection on
cognitive mapping, New York, Harper & Row.
Duchowski, A. T. (2007) Eye movement analysis. Eye tracking
methodology - Theory and practice. Springer.
Fabrikant, S. I., Rebich-Hespanha, S., Andrienko, N., Andrienko,
G. & Montello, D. R. (2008) Novel method to measure inference
affordance in static small-multiple map displays representing
dynamic processes. Cartographic Journal, 45, 201-215.
Jacob, R. & Karn, K. (2003) Eye tracking in human-computer
interaction and usability research: Ready to deliver the promises.
In Radach, R., Hyona, J. & Deubel, H. (Eds.) The Mind's Eye:
Cognitive and Applied Aspects of Eye Movement Research.
Amsterdan, Elsevier.
Montello, D. R. (2002) Cognitive map-design research in the
twentieth century: theoretical and empirical approaches.
Cartography and Geographic Information Science, 29, 283-304.
Nielsen, J. (1993) Usability Engineering, San Francisco, Morgan
Kaufmann.
Poole, A. & Ball, L. J. (2006) Eye tracking in human computer
interaction and usability research: current status and future
prospects. IN GHAOUI, C. (Ed.) Encyclopedia of Human
Computer Interaction. Idea Group.
Rayner, K. (1998) Eye movement in reading and information
processing: 20 years of research. Psychological Bulletin, 124,
372-422.
Rubin, J. & Chisnell, D. (2008) Handbook of Usability Testing.
How to Plan, Design and Conduct Effective Tests, Indianapolis,
Wiley Publishing.
Slocum, T. A., Blok, C., Jiang, B., Koussoulakou, A., Montello,
D. R., Fuhrman, S. & Hedley, N. R. (2001) Cognitive and
usability issues in geovisualisation. Cartography and Geographic
Information Science, 28, 61-75.
A special joint symposium of ISPRS Technical Commission IV & AutoCarto
in conjunction with
ASPRS/CaGIS 2010 Fall Specialty Conference
November 15-19, 2010 Orlando, Florida