Content uploaded by Simone Morgagni
Author content
All content in this area was uploaded by Simone Morgagni on Apr 21, 2022
Content may be subject to copyright.
XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE
Do you believe in transit schematic maps? Design
influences on route choice.
Elise Grison
Direction de la Recherche, SNCF
La Plaine Saint-Denis, France
elise.grison@sncf.fr
Florian Leprévost
Direction de la Recherche, SNCF
La Plaine Saint-Denis, France
f.leprevost@sncf.fr
Simone Morgagni
Direction de la Recherche, SNCF
La Plaine Saint-Denis, France
simone.morgagni@sncf.fr
Abstract—This study completes previous research developed
by Morgagni and Grison [5] on the impact of design
modifications of the Greater Paris (France) transit map. Over
2000 non-residents of the region were asked to plan routes using
several modified versions of the map to further explore the links
between map design and interpretations made by travelers of
the transit network’s characteristics. This paper reports on how
new design modifications linked to transfer stations influenced
route choices. Results complete previous findings, confirming
the specific impact of design modifications on non-residents’
route choices for transport lines and transfer stations. This
paper strengthens previous findings and provides perspectives
for potential applications.
Keywords—transport, transit maps, diagram, cognitive bias
I. INTRODUCTION
Recent research has demonstrated that when planning
routes using a transit schematic map, travelers are strongly
impacted by the implemented design characteristics [1, 2, 4,
6, 7]. Beyond the readability, complexity and perceptual
biases that could be linked to this kind of operational maps, a
hypothesis is that travelers interpret map design not only as a
diagrammatic representation of the network travel alternatives
and characteristics but as some kind of accurate geographical
representation [12]. Indeed, Raveau et al. [11] found that
topological factors presented on a distorted transit map are
more important than actual topology to travelers’ route choice
decisions. Since then, a few studies have tried to better
understand the effect of map design modification on route
choice. Moreover, a better understanding of the relationship
between transit map design and travelers’ route choice might
be of interest for transport operators. Indeed, it might help
them to develop tools for passenger flow management to
improve global comfort of passenger and transport reliability,
especially in saturated megacities’ transit networks.
II. LITERATURE REVIEW
To contribute mitigating possible bottleneck congestions
in transport networks, Guo, et al. [3] recently proposed a route
choice study in the Washington subway network, between
Metro Center and Pentagon stations. Two routes are possible
here, one without transfer, and a second one with a transfer.
The authors proposed to online participants who did not know
the city to choose a route between these two stations under
various graphical design conditions. All designs were
conceived to make the direct route less advantageous
compared to the original version. Results show that increasing
the length ratio of the direct route by 20% leads up to 6% more
participants to choose the route with one transfer compared to
the initial version. This can go up to 10% if the length ratio
increased to 40%.
With these results in mind, Morgagni & Grison [5]
recruited frequent travellers in Greater Paris to participate in a
similar experiment using the region's transit map. Frequent
travellers were asked to plan routes on a printed full-size
transit map of the regional network. In the southern part of the
RER D line, to cross over the area between Juvisy and
Corbeil-Essonnes stations, two routes are available to
travellers. The first one, the western one, has no transfer and
has 4 stops, and the second one, the eastern one, has a transfer
and 3 stops. To alleviate congestion on the direct western
route, the authors applied the same kind of design
modifications as Guo et al. [3] such as increasing the length
ratio between the two routes by 20% or 40%, resulting in an
elongation of the western route. Results highlight that a 20%
increase in the length led travellers to choose more the eastern
option than with the original design. On the contrary, the
preference towards the eastern route is lower in the 40%
condition than in the 20% one. What can explain this kind of
difference between the two studies?
In a 2008 study, Vertesi [12] asked Londoners to draw a
sketch map of Greater London. Her results suggested that
participants structure their graphical productions by relying
heavily on the city’s transit map, i.e., on a diagrammatic
representation of underground metro lines and stations. More
recently, Prabhakar, Grison, Lhuillier & Morgagni [9]
observed that sketch map drawings of Greater Paris, Greater
London and Greater Berlin region inhabitants were more
correlated to the regional transit schematic maps than the
regional geographical maps while including specific
schematic distortions for each city (compression, expansion,
rotation, etc.). This effect was not observed for trained foreign
participants. The contrast and the possible interactions
between previous knowledge and perceptual visual biases
could thus bring to question of the operational validity of the
results observed by Guo et al. [3], especially regarding their
potential use by transport operators to optimize passenger
flow in transit networks which classically have high
percentages of frequent travellers.
Following the same line of questioning, Xu [10] realised a
follow up study of Guo et al.’s work to explore the influence
of the design modification following travellers’ network
knowledge. Participants were classified into three categories
depending on their supposed familiarity with the transport
network: Washington subway travellers (familiar), residents
of the 8 counties covered by the Washington subway (a bit
familiar), and residents of the 22 counties of the Washington
DC area (unfamiliar). For the unfamiliar participants, the
authors observed the same results as Guo et al. [3] for the
conditions in which direct route sees its ratio increased by
20% or 40%. Indeed, in those condition participants tend to
report they choice toward the undirect route. The same result
is observed for the familiar participants when the direct route
length is increased of 20% or when the undirect route is
shortened. However, interestingly, for familiar participants,
increasing of 40% the length of the direct route did not lead
them to choose the undirect route. The 40% ratio condition
appears as less effective than the 20% ratio conditions for the
familiar participants.
While these studies seemed to confirm the influence of
transit lines’ form on route choice, they also reveal the
importance of considering and assessing the effects of
travellers’ familiarity with the transport network.
III. RESEARCH QUESTION
Taking all together, these results suggest that design
modifications of transit maps can have a real impact on how
travellers interpret information about routes. Thus, if the route
seems longer because the graphic line has been lengthened or
complexified with turns, people will interpret this as a reality.
This will consequently impact the choice of route.
Nevertheless, it appears that effect of modification might not
be the same depending on the traveller’s familiarity with the
network. One hypothesis to explain the differences is that
familiar travellers will be sensitive to light and subtle change
but not to major and more visible modification of the schema.
The research presented in this paper is conducted to
validate this hypothesis, but also to validate the generalisation
of those results.
Thus, a first objective of the present study is to confirm
previous results showing that transit line form modifications
impact differently the route choices of familiar (residents of
the region) and unfamiliar (non-residents) participants.
According to the literature, we hypothesize that non-residents
will be more influenced than residents by more glaring
modifications of transit maps, such as an increased 40% length
ratio between routes.
A second objective is to test design modifications of
another main element of transit maps, the transfer station
symbol [2], on the decision to make transfers at a specific
station. To do so, new designs were made to represent the
previously preferred transfer station as increasingly more
complex. We hypothesized that if participants interpret the
graphical complexity of the transfer station as real, they will
be more likely to make had a transfer in a visually simpler
station.
The presented study draws on results observed in our
previous one [5] on the southern part of the RER D in the
Greater Paris transit map.
IV. METHODOLOGY
A. Participants
2482 participants took part in an online study (50,3%
women; 49.7% men). They were aged from 18 to 65 years (M
= 32.6, SD = 12). They did not reside in France and did not
have any knowledge of the Greater Paris region transit
network. They all were recruited using the online platform
Prolific.
B. Material
Maps
As the experiment took place online on possibly small
screens, maps presented to participants showed only parts of
the Greater Paris transit map published by the Parisian
transport authority, Île-de-France Mobilités. The maps
focused on the specific tested line and could incorporate a
possible new version of RER D line. They were simplified
versions of the actual network map presenting only high-
capacity transport modes (Bus Rapid Transport were
excluded). Seven different parts of the network maps were
selected for the test. Six of them were used as distractors and
to prevent participants from guessing the purpose of the study.
Seven design variations of the southern RER D line map
were produced to test our first hypothesis, following the same
design rules used in our previous study [3, 4, 5, 8] as follows
(see Figure 1):
• Control: a standard adaptation of the map according
to actual Île-de-France Mobilités transit map design,
• Vertical: the eastern option is vertically oriented
(vertical-horizontal effect),
• Directness: the eastern option appears more direct
(directness effect),
• 20 % ratio: augmenting of 20% the length ratio
between the two routes,
• 40 % ratio: augmenting of 40% the different of
length between the two routes,
• Directness + 20 % ratio: combination of directness
and 20% ratio conditions
• Acute angle + 20 % ratio: the western option is
designed with an acute angle.
Fig. 1. The 6 modified versions of the map studied.
Two other variations were designed to test our second
hypothesis and focus on the transfer node of Juvisy station, as
follow (see Figure 2):
• Crossing: we inserted another line between the two
RER D branches:
• Separation: we separated the two branches on the
RER D line by adding a lateralized projection to the
transfer point.
Fig. 2. The 3 versions of the transfer at Juvisy.
Routes
Six routes in the area of interest (south of RER D) and 11
control routes in the other areas of the map were identified. In
the area between Juvisy and Corbeil-Essonnes, the eastern
route is the shortest one (3 stations, 90 mm) and the western
route is the longer one (4 stations, 145 mm). For 1 of the 6
routes of this area, the eastern route needed less transfers than
the western. For 2 other routes, the western route needed less
transfers than the eastern one. Finally, for 3 of the 6 routes, the
number of transfers was equivalent. The following Table I
summarizes all the 6 routes and their conditions.
TABLE I. ROUTES TESTED WITH THE INDICATION OF THE ONE WITH
THE MINIUM NUMER OF TRANSFER.
Routes
Less transfer
Juvisy – Moulin Galant
EAST (0)
Corbeil-Essonnes – Créteil Pompadour
WEST (0)
Essonnes Robinson – Villeneuve Saint Georges
WEST (1)
Mennecy – Maison-Alfort Alfortville
EQUAL (1)
Le Vert de Maisons - Boigneville
EQUAL (1)
Vigneux-sur-Seine - Boutigny
EQUAL (1)
Post experiment questionnaire
To collect information about participants’ socio-
economical and transportation profiles, an online
questionnaire composed of 8 items, about age, sex, profession,
and general use of transportation was created.
C. Procedure
The experiment was hosted by the online platform Gorilla,
and participants were recruited through the Prolific platform.
When starting the experiment, participants were first provided
with the general instructions.
Instructions explained that they will have to plan 19 routes
(Origin – Destination pair, i.e., OD) using parts of Greater
Paris (France) schematic public transit network map. For each
trial (route to plan), the part of the map corresponding to the
OD pair was displayed at the screen. At the bottom left of the
screen, the OD pair was presented as follows: “you have to go
from origin to destination”. To select the route, the participant
had to click on every station the route pass by (they will then
be coloured with a yellow dot); and to indicate if necessary,
using a drop-down menu (at the bottom right) the name of
transfer station(s). When finished the participant was invited
to click on the “next” button to proceed to the following trial
(see Figure 3 for a completed trial where a yellow dot
appeared each time a participant clicked on a station).
Trial order was randomized, and the OD pairs direction
was counterbalanced across participants.
Once participants completed the 19 trials, they were
invited to respond to the additional questionnaire. The
experiment took on average 25 minutes.
Fig. 3. Example of one completed trial.
D. Data analysis
Two hundred and one participants were excluded based on
the poor quality of their responses (did not click on stations,
for example). In this paper we overlook the distractive trials
and focus on the analysis of the 7 trials in the RER D part that
have been implemented to answer our hypotheses.
Coordinates of the dots placed on the schema by
participants were collected to code which route option was
selected by the participant, giving us a binary variable (“0” for
the western option an “1” for the eastern one). The transfer
station name was also recorded. We verified the veracity of
the answer by combining the two variables, chosen path and
corresponding transfer station, if the two did not correspond,
the answer was classified as an error.
An average of 32.4% errors was recorded, which is high
but can be explained by the difficulty to understand the
transfers, and the way it was indicated on the map [4, 8].
Indeed, a considerable number of participants’ responses
indicated a plausible route but missed to indicate that there
was a transfer. The percentage of errors was higher in the two
conditions where the transfer station design was modified,
with a mean 36.5% of errors.
For the following analysis, to focus on design effect on
route choice, the choice was made to use percentage of route
choice without considering the error rate [3, 11].
Pearson Chi2 tests were used to observe general and two-
by-two variations between the control condition and modified
alternatives in frequencies of route choice towards the eastern
and western routes. The same analysis was used for transfer
stations considering only the routes for which a transfer at
Juvisy or Viry-Châtillon was needed.
V. RESULTS
A. Route choice depending on design rules
Table II presents the results in comparison with those
obtained previously [5].
The general chi2 (route choice*map design) is significant
(X2(6, 8354) = 34.4; p < .001), indicating that the design
influenced the proportion of route choice toward eastern or
western option.
Looking into details with the two-by-two comparisons, we
observe significant difference in distribution for all
comparisons, except between the control and directness
conditions. For all significant comparisons, design
modifications led to a higher percentage of choice toward the
eastern route, showing a positive effect of the design
modification.
TABLE II. TWO-BY-TWO COMPARISON
Map
Eastern
choice
present
study
Difference
to control
Chi2
p
value
Control
81.8%
NA
NA
NA
Vertical
85.4%
3.6%
4.5
< .05
Directness
85.1%
3.3%
3.6
= .06
20 % ratio
84.1%
2.3%
1.9
= .17
40 % ratio
87.5%
5.7%
11.7
< .001
Directness
+ 20%
89.5%
7.6%
22.3
< .001
Acute
angle +
20%
88.0%
6.2%
13.8
< .001
B. Effect of design on transfer station choice
Table III presents the percentage of choice toward Juvisy
or Viry-Châtillon stations.
The Chi2 test performed on the 3 maps showed a
significant effect, X2(2, 1892) = 365.6, p <.001. In the control
condition, participants preferred to make their transfer at the
Juvisy station.
TABLE III. PERCENTAGE OF CHOICE TOWARD JUVISY OR VIRY
TRANSFER STATION DEPENDING ON THE MAP.
Map
Juvisy transfer
Viry transfer
Control
78.9%
21.1%
Crossing
36.4%
63.6%
Separation
83.2%
16.8%
According to our hypothesis, the modification applied to
the map in the crossing condition led participants to change
their transfer station toward Viry-Châtillon (X2(1, 1291) =
240.2, p <.001. On the contrary, the modification applied to
the separation map did not produce the expected effect.
Indeed, significantly more participants chose the Juvisy
station with the separation map than with the control map
(X2(1,1306) = 3.9, p = .047).
VI. DISCUSSION
The study presented in this paper reinforces previous
results observed in literature on the impact of transit map
design on route choice.
First, we observe the impacts of transit map modifications
on route choice, for non-residents of the region. Non-residents
seem more impacted by significative and important changes
in the map than previously tested residents and travellers (see
Table IV for comparisons), having previous knowledge of the
transit network characteristics [5, 10]. Indeed, contrary to
what has been observed in our first study on residents [5], a
change of 40% of length on one route showed a significant
effect on route choice (+0,6% for residents vs. +5,7% for non-
residents of eastern choice). This result is consistent with Xu
et al. [10] findings on the Washington network on non-
residents’ choices. Combinations of changing form
(directness, verticality) and length are effective for non-
residents too. Moreover, while the modification resulting of
the combination of directness and lengthening of 20% did not
produce any effect on residents (-0,7% for eastern route) [5]
we do observe a change of route choice of +7.6% for the
eastern route for non-residents. All these results confirm the
hypothesis that choices of people that do not know well the
network are more impacted by the design changes.
TABLE IV. COMPARISON OF RESULTS OF THE STUDIES CONDUCTED IN
PARIS AND DC ON BITH RESIDENTS AND NON RESIDENTS
Map/
Difference
to control
Paris
non
residents
Paris
residents
[5]
DC non-
residents
[3]
DC
periphery
residents
[10]
DC
center
residents
[10]
DC
metro
frequent
user [10]
Control
(81.8%)
(83.8%)
(72.1%)
(71.6%)
(71.6%)
(72.4%)
Vertical
+ 3.6%
+ 1 %
Directness
+ 3.3%
+ 2.1 %
20 % ratio
+ 2.3%
+ 3.4 %
+ 3.1 %
+ 6.7 %
+ 2.4 %
+ 6 %
40 % ratio
+ 5.7%
- 1.4 %
+ 9.5 %
+ 12.1 %
+ 8.8 %
+ 3.9 %
Directness
+ 20%
+ 7.6%
- 0.7 %
Acute
angle +
20%
+ 6.2%
- 2.3 %
+ 5.7%
+ 6 %
+ 5.7%
+ 7.1 %
We also introduced new types of design modifications for
transfer stations, with the goal estimate to what extent it will
be possible to make a transfer in bigger already overcrowded
stations less attractive. As for the previous modifications, the
more subtle condition (separation of the big station into two
parts, one per line) didn’t have a significant effect. However,
the more visible Crossing condition (pulling away the two
lines and putting one across another line) had the
overwhelming effect of reversing the preference for the bigger
station, from almost 80% to 36.4% choice. Note that we do
observe a high percentage of error in both conditions of
transfer node modification. This might be explained by the
participant’s difficulty to understand complex transfer nodes
as we observed that most errors made are due to a
misinterpretation of them. In this context the proposed design
change may have increased their misunderstanding.
Additional work is thus needed to improve the understanding
of transfer node and then validate the effect of their design on
route planning.
To sum up, we reproduced previous results on transit map
design effects, introduced new effective modifications, and
confirmed the difference effect of design on resident and non-
residents. Improving the understanding of the effect of these
modifications should help transport operators use them more
adequately and effectively to improve passenger flow and
comfort.. Note that an additional step might be needed to fill
the gap between this research on static transit maps and their
dynamic application, which will be to study this question on
new planning aid tools such as smartphone apps.
VII. ACKNOWLEDGEMENT
This research is partly founded by a French National
Research Agency grant ANR 18-CE22-0016-01.
REFERENCES
[1] M.J. Roberts, H. Gray, & J. Lesnik,, “Preference versus performance:
Investigating the dissociation between objective measures and
subjective ratings of usability for schematic metro maps and intuitive
theories of design”, in International Journal of Human-Computer
Studies. 98, 109-128, 2017.
[2] Z. Guo, “Mind the map! The impact of transit maps on path choice in
public transit”, in Transportation Research Part A, 45 (7), 625–639,
2011.
[3] Z. Guo, J. Zhao, C. Xhong, P. Mishra & L. Wyman, “Redesigning
subway map to mitigate bottleneck congestion: An experiment in
Washington DC using Mechanical Turk”, in Transportation Research,
Part A, 106 pp. 159-169, 2017.
[4] E. Grison, “Schematic map design: from performance to preference? A
literature review of empiral research.”, 2nd Schematic Mapping
Workshop, 2019.
[5] S. Morgagni, & E. Grison, “East or West ? Map Design and Passenger
Path Decisions on Mass Transit Networks”, 2nd Schematic Mapping
Workshop, 2019.
[6] H. Hochmair, “The influence of map design on route choice from
public transportation maps in urban areas”, in Cartograph. J. 46 (3),
242–256, 2009.
[7] M.J. Roberts, E.J. Newton, F.D. Lagattolla, S. Hughes, & M.C.
Hasler, “Objective versus subjective measures of Paris Metro map
usability: Investigating traditional octolinear versus all-curves
schematics”, International Journal of Human-Computer Studies. 71
(3), 363-386, 2013.
[8] P.B. Lloyd, P. Rodgers, & M.J. Roberts, “Metro map colour-coding:
Effect on usability in route tracing”, in: Diagrammatic Representation
and Inference. Editors: Chapman, P., Stapleton, G., Moktefi, A., Perez-
Kriz, S. and Bellucci, F., . Springer. 411- 428, 2018.
[9] A. Prabhakar, E. Grison, S. Lhuillier, & S. Morgagni, “Transport-made
cities: transit maps as major cognitive frames of metropolitan areas”,
submitted.
[10] J. Xu, “Map Sensitivity vs. Map Dependency: A Case Study of Subway
Maps’ Impact on passenger Route Choices in Washington DC”,
Behavioral Sciences, 7, 72, 2017.
[11] S. Raveau, Z. Guo, J.C.Muñoz, & N.H.M Wilson, “A behavioural
comparison of route choice on metro networks: Time, transfers,
crowding, topology and socio-demographics”, Transportation
Research Part A: Policy and Practice, 66, 185–195, 2014
[12] J. Vertesi, “Mind the gap: the londonLondon underground map and
users’ representations of urban space,” Social Studies of Science, vol.
38, no. 1, pp. 7-33, DOI2008, doi: 10.1177/0306312707084153.
