1. Visual distance of map symbols
evaluation of map readability with eye-tracking
Alžběta Brychtová
This presentation is co-financed by the
European Social Fund and the state
budget of the Czech Republic

2. Visual distance
 Jan T. Bjørke, Norway (1996):
„It is necessary to maintain sufficient visual distance
between map symbols to make them distinguishable.“
 visual distance
1. Euclidean distance between symbols
 influenced by the real spatial location
of mapped objects, topology, generalization
and map purpose
2. Rate of difference between symbols appearance
 experiences and ability of map makers
to design easily distinguishable map symbols
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

3. Visual distance
 definition
“Visual distance of map symbols is exactly determined
numerical value describing the degree of variation of
visual variables of compared map symbols.”
 variation of visual variables = change of information
transmitted by a map
 easily distinguishable change of visual variable = easy to
read the information
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

4. Visual distance
 necessity to emphasize sufficient difference of visual
variables
 Jacques Bertin's visual variables (7)
 size
 position
 shape
 orientation
 color hue
 color value
 texture
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

5. Research task
 detect influence of color distance between two map
elements on the readability of the map
 assumption:
 increasing color distance will have positive impact on map
readability
 experimental stimuli were designed to reflect changes in
color value (color hue is currently in progress)
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

6. Color distance
 Visual distance definition: “Visual distance of map symbols is exactly determined numerical value
describing the degree of variation of visual variables of compared map symbols.”
 The International Commission on Illumination (CIE) defines the color
distance as Euclidean distance of two colors in the CIELuv color
space
 In this case study the distance were computed as a dot product of
two RGB vectors in the RGB color space:

First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

7. Experimental design
 Eye-tracking experiment was performed
 statistical analyses of eye-tracking metrics
 Lab setup:
 SMI RED 250 eye-tracker
 120 Hz sampling rate
 0.4°accuracy and 0.03°spatial resolution
 gaze data classification by dispersion threshold algorithm (ID-T)
dispersion threshold = 50 px, duration threshold = 80 ms
 SMI BeGaze
 R Project
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

8. Experimental design - stimuli
 15 simple map stimuli varying in color distance of map labeling
and background
 20%, 40%, 60%, 80% and 98% color distance
 8, 11 ad 14 pt size of labels
 reduction of the number of independent variables to a
minimum
 participants were asked to find a concrete
administrative unit by its name
 avoid the effect of geographical knowledge
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

9.  20% 40% 60% 80% 98%
8 pt
11 pt
14 pt
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

10. Experimental design - process
 within subject design – all participant tested under the same
condition
 randomization of trials – prevention of the learning effect
15
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

11. Experimental design - respondents
 53 volunteers – students of Palacký University
 data from 3 respondents with the tracking ratio less than
90% wasn’t taken into account
 50 respondents
 20-25 years
 30 cartographers + 20 non-cartographers
 30 men + 20 women
 data were collected within bachelor thesis of Veronika
Obadálková
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

12. Monitored metrics
 fixation count
 more overall fixations indicates less efficient searching
 average duration of fixation
 longer fixation duration indicates difficulty in extracting
information, or the object is more engaging in some way
 scanpath
 longer scanpath (the length of gaze trajectory over the
stimulus) indicates less efficient searching
 time to answer
 reflects the success during searching the information
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

13. Results
 Shapiro-Wilk test of normality
average fixation
fixation count scanpath length time to answer
duration
p-value 2.2e-16 3.129e-16 2.2e-16 2.2e-16
 on the significance level α = 0.05 no one measured eye-tracking metric comes from normal distribution
 Mann-Whitney test for median comparison
 different perception between groups of cartographers
and non-cartographers
average fixation
fixation count scanpath length time to answer
duration
p-value 0,09238 0,988 0,7801 0,2094
 On the significance level α = 0.05 no differences between two groups of respondents in measured metrics were proven
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

14. Results
 different perception between groups men and women
 Mann-Whitney test for median comparison
average fixation
fixation count scanpath lenght time to answer
duration
p-value 0.008283 3.875e-09 0.02236 0.6384
 On the significance level α = 0.05 the significant result was proven for fixation count, average
fixation duration and scanpath length
 fixation count F<M
 average fixation duration F>M
 scanpath length F<M
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

15.  Kruskal-Wallis ANOVA for mean rank comparison
 different perception of maps with varying color distance
 no categorization
average fixation
fixation count scanpath lenght time to answer
duration
p-value 0,009817 0,9073 0,005274 0,0012
 On the significance level α = 0.05 the significant result was proven for fixation count (H=
13.3192, DF = 4, N=50, P= 0,009817), scanpath length (H= 14.7391, DF = 4, N=50, P=
0,005274) and time to answer metric (H= 17.9129, DF = 4, N=50, P= 0,009817)
 the mean ranks of these metrics are significantly different among maps with different colour-
distance between map labeling and background.
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

16. Results
 post-hoc Kruskal-Wallis tests
 differences of perception between pairs of map of
concrete color distance
fixation count scanpath length time to answer
color observed critical differen color observed critical differen color observed critical differen
distance difference difference ce distance difference difference ce distance difference difference ce
20%-40% 39.125000 68.99673 FALSE 20%-40% 48.36111 68.99673 FALSE 20%-40% 51.24306 68.99673 FALSE
20%-60% 31.665398 68.87767 FALSE 20%-60% 20.10352 68.87767 FALSE 20%-60% 31.85699 68.87767 FALSE
20%-80% 82.903329 68.87767 TRUE 20%-80% 79.71386 68.87767 TRUE 20%-80% 95.19837 68.87767 TRUE
20%-98% 63.527778 68.99673 FALSE 20%-98% 69.28472 68.99673 TRUE 20%-98% 72.83681 68.99673 TRUE
40%-60% 7.459602 68.87767 FALSE 40%-60% 28.25759 68.87767 FALSE 40%-60% 19.38606 68.87767 FALSE
40%-80% 43.778329 68.87767 FALSE 40%-80% 31.35275 68.87767 FALSE 40%-80% 43.95532 68.87767 FALSE
40%-98% 24.402778 68.99673 FALSE 40%-98% 20.92361 68.99673 FALSE 40%-98% 21.59375 68.99673 FALSE
60%-80% 51.237931 68.75840 FALSE 60%-80% 59.61034 68.75840 FALSE 60%-80% 63.34138 68.75840 FALSE
60%-98% 31.862380 68.87767 FALSE 60%-98% 49.18120 68.87767 FALSE 60%-98% 40.97981 68.87767 FALSE
80%-98% 19.375551 68.87767 FALSE 80%-98% 10.42914 68.87767 FALSE 80%-98% 22.36157 68.87767 FALSE
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

17. First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc

18. Conclusions
 highest values of all analyzed metrics were observed on the
map with the minimal color distance (20%), which means that
respondents had difficulties in extracting information from
these maps of low color distance;
 increasing color distance leads to decreasing count of
fixations, which can mean the higher color distance the more
successful information mining;
 similar statement can be done for scanpath length and time
to answer, except the local maximum of measured metrics for
maps with % color distance;
 color distance has evident influence on map readability, but
its improvement can be observed only between stimuli with
high differences of the color distance.
First InDOG Doctoral Conference, 29th October - 1st November 2012, Olomouc