Virtual navigation tested on a mobile app is predictive of real-world wayfinding navigation performance
Virtual navigation tested on a mobile app is predictive of real-world wayfinding navigation performance
Antoine CoutrotID 0 2
Sophie Schmidt 0 2
Lena Coutrot 0 2
Jessica Pittman 0 2
Lynn Hong 0 2
Jan M. Wiener 2
Christoph H o?lscher 2
Ruth C. Dalton 1 2
Michael Hornberger 2
Hugo J. Spiers 0 2
0 Institute of Behavioural Neuroscience, Department of Experimental Psychology, Division of Psychology and Language Sciences, University College London , London , United Kingdom , 2 Laboratoire des Sciences du Num e ?rique de Nantes - CNRS - Universite ? de Nantes, Nantes, France, 3 Institut Jean Nicod, ENS, EHESS, CNRS, Paris, France, 4 Department of Psychology, Ageing and Dementia Research Centre, Bournemouth University , Poole , United Kingdom , 5 ETH Z u ?rich , Swiss Federal Institute of Technology , Zu ? rich , Switzerland
1 Department of Architecture and Built Environment, Northumbria University , Newcastle upon Tyne , United Kingdom , 7 Norwich Medical School, University of East Anglia , Norwich , United Kingdom
2 Editor: Laura Zamarian, Medical University of Innsbruck , AUSTRIA
Virtual reality environments presented on tablets and smartphones have potential to aid the early diagnosis of conditions such as Alzheimer's dementia by quantifying impairments in navigation performance. However, it is unclear whether performance on mobile devices can predict navigation errors in the real world. We compared the performance of 49 participants (25 females, 18-35 years old) at wayfinding and path integration tasks designed in our mobile app 'Sea Hero Quest' with their performance at similar tasks in a real-world environment. We first performed this experiment in the streets of London (UK) and replicated it in Paris (France). In both cities, we found a significant correlation between virtual and realworld wayfinding performance and a male advantage in both environments, although smaller in the real world (Cohen's d in the game = 0.89, in the real world = 0.59). Results in London and Paris were highly similar, and controlling for familiarity with video games did not change the results. The strength of the correlation between real world and virtual environment increased with the difficulty of the virtual wayfinding task, indicating that Sea Hero Quest does not merely capture video gaming skills. The fact that the Sea Hero Quest wayfinding task has real-world ecological validity constitutes a step toward controllable, sensitive, safe, low-cost, and easy to administer digital cognitive assessment of navigation ability.
Funding: Deutsche Telekom supported and funded
this research - https://www.telekom.com/en to HS.
Alzheimer?s Research UK funded the analysis
https://www.alzheimersresearchuk.org/ to HS. The
funders had no role in study design, data collection
Virtual reality (VR) provides a powerful means to study and quantify how humans navigate,
because the properties of a virtual environment can be completely controlled and repeated
across participants. Since the late nineties, it has been a critical tool to understanding how
brain regions support navigation and unveiling the structural and functional neural correlates
and analysis, decision to publish, or preparation of
of spatial navigation [
]. VR tests of spatial cognition have proved more sensitive in
identifying spatial navigation deficits in patient populations compared to more classic visuospatial
?pencil-and-paper? tests like the Mental Rotation Test . VR has the added advantage to be a
less costly and safer alternative to real-world navigation tests, which are time and space
consuming, as well as difficult to administer to a population sometimes less able to walk [
recently, most VR used in research was presented on a desktop display and movement
controlled via a joystick or keyboard. Such an interface presents difficulties for older people, less
exposed to technology than younger participants [
]. However, with the advent of tablet and
smart-phone touch screen mobile devices, older participants have found engaging in VR
tasks much easier and intuitive than with desktop computers [
]. As a consequence, mobile
devices have recently been used in several fields such as neuropsychological assessment ,
stroke rehabilitation [
] and mental health [
]. We recently developed a VR navigation task
for mobile and tablet devices?Sea Hero Quest?with the aim that this may provide an early
diagnostic tool for Alzheimer?s Disease (AD) [
]. For this test to be useful it is important that
it has real-world validity, with errors on the VR task predicting errors in real-world navigation
Past research comparing navigation in real and VR environments has generally found a
good concordance in performance across both environments in the normal population [
], in younger and older age groups , in individuals with brain injury [
], in chronic
stroke patients , and in patients with Mild Cognitive Impairment (MCI) or early AD [
], for reviews see [
]. However, this consistency seems to be modulated by the type of
spatial navigation task, as a previous study showed that performance in real life and virtual
environments were similar for tasks such as landmark recognition or route distance estimate,
but different for pointing to the beginning and endpoint of the route, or drawing a map of the
Most prior studies comparing VR and real-world navigation performance have used
desktop VR or immersive VR to simulate environments, and paper and pencil tests such as line
orientation, road map, or delayed recall when assessing ?real-world navigation behavior?. A few
studies made use of actual navigation tasks but often in a limited spatial range, like the lobby of
a hospital [
]. A notable exception is [
], where the authors tested 978 military college
students on a 6 km orienteering task and replicated many laboratory-based findings, including
gender differences. However, the authors did not test their participants in a VR task and were
thus unable directly compare the two environments.
Numerous studies found a male advantage for navigation in VR tasks [
], but only
a few looked for gender differences in real-world navigational tasks [
]. This led some
authors to suggest that previously reported gender differences in spatial ability may be driven
by familiarity with technology, men being more comfortable with virtual tasks than women
who are sometimes less exposed video games .
Here, for the first time we directly compared in a within-subject design the spatial
navigation performance measured on a mobile device with our Sea Hero Quest virtual tasks, and
in a large-scale real-world environment covering a whole neighborhood of London (Covent
Garden, South of the British Museum) and of Paris (South of the Montparnasse cemetery).
We designed the real-world counterparts of the Sea Hero Quest wayfinding and path
integration tasks, which are known to tap into different cognitive processes [
wayfinding task relies on various skills, including interpretation of a map, planning a multi-stop
route, memory of the route, monitoring progress along the route and updating of route
plan, and transformation of birds-eye perspective to an egocentric perspective needed for
navigation , while the path integration task typically only requires working memory
2 / 15
We hypothesized that performance in the real world in both cities will significantly
correlate with performance in the virtual environment. Based on our original mobile-based results
] and on previous gender differences found in navigational studies in the real world [
we hypothesized that males will perform better than females in both environments. Finally, we
predicted that familiarity with video games will not influence performance in either
environment, in line with a previous study showing no effect of computer experience on spatial
memory errors [
]. In particular, the correlation between the real-world wayfinding performance
and the performance at the first training Sea Hero Quest level?where no spatial ability is
required?should be null. Comparing this study to the original large dataset would enable
testing whether our results hold true not simply in small cohorts but on a population level.
This study has been approved by UCL Ethics Research Committee. The ethics project ID
number is CPB/2013/015. Written consent was obtained from each participant and the data were
analyzed anonymously. Participants were tested on specific levels from Sea Hero Quest [
on a tablet, and then on equivalent tasks in the real world, see Fig 1. Participants were also
asked to answer a few demographic questions. We first ran this experiment in London in
summer/fall 2017. We then replicated it with a different team in Paris in spring 2018. The whole
experiment lasted around three hours.
In London?We tested a total of 30 participants (15 males) but data from 1 participant was
missing due to a technical problem and real-world wayfinding data from 6 participants (3
males and 3 females) were discarded due to GPS recording issues. Subsequent analyses hence
include 23 participants (11 males), aged 18-30 y.o. (M = 21.52, s.d. = 1.81). Participants had
normal or corrected to normal vision and gave their written consent to participate. Path
integration data was not collected for the first 11 participants as this task was not yet implemented.
Participants received 3 class credits or ?20 for their participation.
In Paris?We tested a total of 30 participants (15 males), but real-world wayfinding data
from four participants (1 male and 3 females) were discarded due to GPS recording issues.
Subsequent analyses hence include 26 participants (14 males), aged 18-30 y.o. (M = 23.15,
s.d. = 2.52). Participants had normal or corrected to normal vision and gave their written
consent to participate. Participants received 30 euros for their participation.
We devised a mobile video game designed to measure human spatial navigation ability
through gameplay?Sea Hero Quest (SHQ, www.seaheroquest.com). This video game involves
navigating a boat in a virtual environment (lake or river networks) and has been extensively
described in [
]. It features two main tasks, which have been designed to tackle different
aspects of spatial navigation.
1- Wayfinding. Participants were required to view a map displaying current position and
goal locations to find (Fig 1A and 1B). Participants could study the map without time
restrictions and had to navigate to the goal locations in the order indicated, e.g. goal 1 must be found
first, then goal 2, etc. Goals were buoys with flags marking the goal number. The task is
complete when all goals have been located. If the participant takes more than a set time, an arrow
indicates the direction along the Euclidean line to the goal to aid navigation. On basis of the
data from the mobile video game, we selected a subset of 5 of the total 75 levels in the game
that varied in difficulty. In order to compare the data recorded in this study to the
population3 / 15
Fig 1. Task in real world (bottom row) vs virtual environment (top row). (A-B) Wayfinding task in the video game: participants had to
memorize a map and navigate as fast as possible toward an ordered set of goals. Participants played Sea Hero Quest on a tablet. (C) Path
Integration task in the video game: participants had to navigate in a maze until they find a flare and shoot it back toward their starting
position. (D) Wayfinding task in the real world. Identical as the virtual task, but takes place in the streets of (E) London and (F) Paris. All
other maps are displayed in supporting S1?S4 Figs.
level dataset, we chose three levels of increasing but moderate difficulty appearing quite early
in the game (levels 6, 11 and 16). Indeed, the sample size of the mobile video game logically
dropped rapidly across levels (see [
]). We also included a training level (level 1) and a level
of great difficulty (level43), see S1 Fig. The helping arrow appeared after 80 s in level 1, 70 s in
level 6, 80 s in level 11, 80 s in level 16 and 200 s in level 43.
Performance was quantified with the Euclidean distance travelled in each level (in pixels).
The coordinates of participants? trajectories were sampled at Fs = 2 Hz. We summed the
distance travelled over levels 6 to 43. We did not include level 1 because it did not require any
spatial ability (the goal was visible from the starting point) and was only designed to assess
4 / 15
Fig 2. Spatial ability at a wayfinding task in real world vs virtual environment. (A) Correlation between the distance navigated in the video game and the
normalized distance navigated in the real-world wayfinding task in London (skipped Pearson?s r = 0.46, p = 0.01) and (B) in Paris (skipped Pearson?s
r = 0.57, p = 0.001). Real-world normalized distance is the distance travelled by participants divided by the number of goals they reached. Outliers have been
determined with skipped-correlation. (C) Skipped correlation coefficients between the distance navigated in each video game level and the total normalized
distance navigated in the real world. Video game levels are sorted by increasing difficulty according to Table 1. (D) Gender differences at the wayfinding task
in the video game (right) and in the real world (left). Distances have been standardized (zscore) to allow a direct comparison between environments. Black
dots represent individual data points. Error bars represent standard errors.
participants? ability to learn to control the boat. We only considered level 1 data in Fig 2C to
compare participants? performance in real life with the distance they travelled in each level.
2- Path Integration. During path integration, participants integrate perceived ego motion
while they move to update their position and orientation. It is a more basic navigation
mechanism than wayfinding, which typically only requires working memory processes [
wellestablished tool in the study of path integration is the triangle completion task, where
participants move along the first two sides of a triangular pathway, and then are asked to return to
their starting position, thus completing the triangle [
]. The task we designed here is a
direct implementation of this paradigm. Participants were required to navigate along a river
5 / 15
with bends until they find a flare and shoot it back toward their starting position. Participants
could choose among three directions, as shown in Fig 1C. We selected a subset of 5 levels that
varied in difficulty: level 14 had one bend, level 34 three, level 44 two, level 54 four and level 74
five, see S4A Fig.
Performance was measured with the number of stars obtained by the player. Stars were
awarded based on participant?s choice between 3 proposed directions: 3 stars for the correct
answer (their starting point), 2 stars for the second closest direction, and 1 star for the third
1- Wayfinding. The real-world wayfinding task consisted of 6 wayfinding trials which varied
in difficulty in terms of the number of streets to be navigated, the number of goals and the
relative location of the goals to each other. Each trial consisted of a different starting point and
required exploration through different street networks South of the British Museum in
London (Covent Garden area) and South of the Montparnasse cemetery in Paris. We chose less
busy streets to avoid traffic and made sure the participants were not familiar with them. Before
each trial, participants were shown a map that only indicated the facing direction, the network
of the local streets and the location and the order of the goals (in London see S2 Fig, in Paris
see S3 Fig). Maps were displayed on a tablet (IPad MP24B/A, 9.7 inches). The goals were doors
and gates with distinct features (e.g. specific colour, size, or material). Participants had up to
1 min to memorize the map. Once the minute was up, the map was removed and they were
asked to go locate the goals. During navigation they were provided with colour photographs of
the goal. To calibrate the time limit of each route, we pilot tested 3 participants in London and
2 participants in Paris, not included in the analyses. We chose these time limits to allow for a
few mistakes at a reasonable walking pace. Pilot testing indicated that if participants required
any longer than that these time restraints they were likely guessing and had failed to remember
the goal locations or street layout. To take into account the fact that some participant did not
finish some routes, we divided this distance by the number of goals reached by the participant
plus 1. We added 1 to cope with cases where the participant didn?t reach any goal (this only
happened once). We refer to this as the metric normalized distance, and summed it over routes
1 to 6. If participants reached the limits of the defined region shown in the map they had
studied then they were told by the experimenter that they had reached the edge of the search area
and should turn back. In London, route one: 6 minutes, route two: 6 minutes, route three: 6:30
minutes, route four: 6:30 minutes, route five: 12 minutes, route six: 14 minutes. In Paris, route
one: 5 minutes, route two: 8 minutes, route three: 8 minutes, route four: 9 minutes, route five:
16 minutes, route six: 20 minutes.
The coordinates of participants? trajectories were sampled at Fs = 1 Hz with the
experimenter?s smartphone GPS via the Beeline app. We visually inspected all recorded GPS trajectories
to deal with potential losses of signal. For losses of signal where the participant did not make
any turn, we linearly interpolated between the first and the last missing points. When we
couldn?t reconstruct the trajectory because the participant changed direction during the loss of
signal, we discarded the data (5 trials out of 180 in Paris, 6 out of 180 in London). Performance
was quantified with the Euclidean distance travelled in each route (in meters).
2- Path Integration. The real-world path integration task consisted of 4 path integration
trials which varied in difficulty in terms of the number of turns they featured (1, 2, 3 and 4
turns, see S4B Fig). To avoid familiarity effect, path integration routes were chosen not to
intersect with any wayfinding route. Participants were informed when they were at a starting
point then they were asked to follow the experimenter to an endpoint where they were
6 / 15
Video game levels
1 (1 goal)
6 (3 goals)
11 (3 goals)
16 (3 goals)
43 (4 goals)
instructed to point back toward the starting point. We used a numeric compass to precisely
record the direction. Performance was defined as the inverse of the angle between the direction
pointed toward by the participant and the ground-truth, in degrees. We then summed the
absolute values of the path integration error angles.
Correlation between performance in real world and virtual environments
To visualize participants? raw data in real world and virtual environments we created a video
showing on the left side the trajectories of participants in London?s route 6 and on the right
side the trajectories of the same participants in the 43rd level of Sea Hero Quest (S1 Video).
19.8? ? 4.5
21.7? ? 2.8
19.5? ? 2.4
25.6? ? 4.4
Video game levels
14 (1 turn)
34 (2 turns)
54 (3 turns)
44 (4 turns)
74 (5 turns)
7 / 15
The relationship between the wayfinding performance in the real world and in the video game
is shown Fig 2A (in London) and Fig 2B (in Paris). One can notice a few outliers in the upper
right corner of the scatter plots. Traditional Pearson?s correlation measure is known to be
highly sensitive to outliers, which can severely bias the estimation of the strength of the
association among the bulk of the points . To deal with this, we used skipped-correlation, which
protects against outliers by taking into account the overall structure of the data [
are detected using a projection method, removed, and Pearson?s correlation is computed using
the remaining data. Hence, Pearson?s skipped correlation is a direct reflection of Pearson?s r.
We used an implementation of this algorithm available in a free toolbox [
]. The detected
outliers are tagged in Fig 2A and 2B with black edges, and discarded from further analysis.
95% Confidence Intervals (CI) were computed via bootstrap: pairs of observations were
resampled with replacement and their correlation values obtained and sorted. Values between the
2.5 and 97.5 percentiles yielded the 95% CI. Skipped correlation were significant both in
London (r = 0.46, 95% CI = [0.14, 0.68], p = 0.01) and in Paris (r = 0.57, 95% CI = [0.37, 0.76],
p = 0.001).
To confirm that our virtual task captured participants? wayfinding ability, we checked
whether the strength of the correlation between performance in real world and in the video
game was modulated by the difficulty of the virtual task. Under this hypothesis, the correlation
should be null between real world and level 1 performance, since level 1 is a training level
where no spatial ability is required: the end goal is visible from the starting point. The
correlation should then increase with the difficulty of the level. We broke down the global correlation
score for each Sea Hero Quest level, comparing participants? performance in real life with
the distance they travelled in each level. To increase the sample size, we combined the data
recorded in both London and Paris. In order to take into account the difference in route length
between cities (the task in Paris being slightly longer than the one in London), we calculated
zscores of the performance for each level before combining the data of the two cities. In the
following we work with this cross-city normalized metric, called Standardized Distance. The
skipped correlation coefficient between the Standardized Distance in the real world and in
level 1 is close to 0 (r = 0.06), confirming that this instruction level does not measure spatial
ability. We sorted the game levels by increasing difficulty based on the results of Table 1: level
1, 6, 16, 11 and 43. As shown in Fig 2C, the skipped correlation coefficient increases with level
difficulty, from r = 0.06 in level 1 to r = 0.44 in level 43.
The skipped correlation between real world and video game path integration score in Paris
was not significant, r = -0.23, 95% CI = [-0.51, 0.08], p = 0.11. However, the sign of the
correlation is logical since higher scores mean better performance in the game (number of stars) but
not in the real-world task (error angle). In London, the first 11 participants were not tested on
path integration as this task was not yet designed. The skipped correlation based on the other
19 participants is consistent with Paris data: r = -0.40, 95% CI = [-0.73, 0.07], p = 0.05.
However, correlational analyses with small sample size can lead to strongly biased correlation
estimates and this result should be considered with caution.
For the wayfinding task, we found that in both environments male participants had an
advantage, although smaller in the real world (Fig 2D). In the real world, Cohen?s d = 0.59, 95% CI =
[0.02 1.15], t(47) = -2.08, p = 0.04; in the video game, Cohen?s d = 0.89, 95% CI = [0.35 1.42],
t(56) = -3.43, p = 0.001.
For the path integration task in Paris, we did not find a significant gender difference.
However, the tendency was similar to the wayfinding results, with male having a small advantage in
8 / 15
both environments, smaller in the real world, see S5 Fig. In the real world, Cohen?s d = -0.21,
95% CI = [-0.91 0.49], negative values correspond to a male advantage; in the video game,
Cohen?s d = 0.29, 95% CI = [-0.40 0.99], positive values correspond to a male advantage. The
gender effect size is much smaller for path integration than for wayfinding and is not
significative, as shown by the wide 95% CIs.
Influence of familiarity with video games
We showed that the correlation between performance in real world and in the training level of
the virtual environment was weak, indicating that gaming ability does not predict navigation
ability in the real world (Fig 2C). To further test this claim, we asked participants the average
duration they play video games per week and used this variable along with gender to predict
performance in the real-world and in the virtual environment. On average, females played
video games 2.99 ? 8.38 hours per week and males played 2.95 ? 4.21 hours per week. To
control for the influence of familiarity with video games on real-world and virtual spatial abilities,
we computed a multiple linear regression to predict performance based on gender and on the
time participants spent playing video game (VGT), in hours per week.
With standardized distances recorded in the real-world wayfinding task, gender was a
significant predictor (t(47) = ?2.22, p = 0.03), but not VGT (t(47) = 0.31, p = 0.76). Similarly,
with standardized distances recorded in the virtual wayfinding task, gender was a significant
predictor (t(55) = ?3.44, p = 0.001), but not VGT (t(55) = ?0.99, p = 0.32).
With error angles recorded in the real-world path integration task, gender was not a
significant predictor (t(55) = ?0.65, p = 0.52), nor was VGT (t(55) = 1.09, p = 0.28). Similarly, with
standardized flare accuracy recorded in the virtual path integration task, gender was not a
significant predictor (t(55) = ?0.17, p = 0.87), nor was VGT (t(55) = 0.09, p = 0.92).
Correlation between wayfinding and path integration
The skipped correlation between path integration and wayfinding scores is not significant in
the real world (Paris data): r = -0.06, 95% CI = [-0.34 0.44], nor in the virtual environment: r =
-0.18, 95% CI = [-0.49 0.15]. Higher scores mean better performance in the path integration
task in the game (number of stars) lower score mean better performance in the path
integration task in the real-world task (error angle), and in the wayfinding tasks in both environments
Comparison to the population-level original dataset
To check whether the 60 participants we recruited for this experiment were representative of
the much larger dataset recorded with the mobile version of Sea Hero Quest [
], we plotted
in Fig 3 the performance of the participants at this study (vertical red dotted lines) against the
corresponding distribution of the performance of the French and British Sea Hero Quest
players from the original dataset. Since the number of players per level drops rapidly, we focused
on level 11 to maximise the difficulty / number of players ratio (N = 78,724, see [
] for full
data). Fig 3 clearly shows that the performance of the participants recruited for this study
closely follows the performance distribution of the original dataset. Gender differences
followed the same direction in this study (Cohen?s d = 0.81, 95% CI = [0.28 1.33]) as in the
subsample of the original dataset (Cohen?s d = 0.40, 95% CI = [0.39 0.42], positive values
correspond to a male advantage.
9 / 15
Fig 3. Comparison with the large scale video game dataset. Distance in the Video Game (level 11) of the French and British participants
tested in the original Sea Hero Quest database (N = 78,724, blue histogram), see [
] for full data. The red dotted vertical lines represent the
performance of the participants recorded in the current study.
We report evidence that wayfinding navigation performance on a mobile app-based VR
navigation task (Sea Hero Quest) is significantly correlated with performance in a real-world city
street wayfinding task. We directly compared participants performance at a subset of Sea Hero
Quest wayfinding levels with their performance at an equivalent task in the Covent Garden
area, London. We found a strong correlation between the distance participants travelled in
the video game (in pixels) and in the real-world street network (in meters, measured by a GPS
device). We replicated this result with another set of participants in the Montparnasse area,
Paris. The high similarity of the results in the two cities is a strong indicator of the robustness
of the results presented above. Our findings are consistent with a number of studies that
showed that spatial navigation assessment in a desktop VR [
15, 16, 18, 20, 22, 23, 27
immersive VR [
] environments transferred well to the real world, and extend them to
tablet device presentation and real-world spatial task spanning complex street networks.
However skill assessment don?t always generalize from VR to the real world. For instance reading
10 / 15
skills assessed in a digital device can be partially predicted by participant?s ability to perform
basic computer tasks . The similarity both in term of performance and of gender difference
of this study with the original large scale Sea Hero Quest study [
] suggests that our findings
hold true not simply in small cohorts but on a population level.
We found a significant male advantage in the wayfinding task in both real world and virtual
environments, although weaker in the real world. This difference in effect size between the two
environments couldn?t be explained by males being more familiar with video games, as
suggested in a previous study [
] for three reasons. First, our male and female participants
reported playing video games the same average duration per week (2.95 vs 2.99 hours per
week). Second, when using gender and time playing video games (VGT) as covariates in a
linear model to predict wayfinding performance in the real world (resp. in the virtual
environment), gender came out as a significant predictor, but not VGT. Third, we found a very weak
correlation coefficient (skipped Pearson?s r = 0.06) between the performance at the real-world
wayfinding task and at the first training Sea Hero Quest level, which did not require any spatial
ability (the endpoint being visible from the start). The strength of the correlation increased
with the difficulty of the video game level (up to r = 0.44 in level 43), indicating that Sea Hero
Quest does not merely capture video gaming skills. The discrepancy with [
] might stem
from the difference between tasks, Richardson et al.?s task being closer to the path integration
task than to the wayfinding task discussed in this paragraph. The underlying causes of gender
differences in spatial ability are still debated in the literature and include sex hormones
variation, evolution, differences in self-confidence and anxiety [
]. In a previous paper based on
the global video game dataset, we showed that gender differences in spatial ability measured in
the game correlate with gender differences in the society measured with the Gender Gap Index
(World Economic Forum) [
We did not find a significant correlation between the performance at the real world and the
Sea Hero Quest path integration task. At least three reasons could account for this null result.
First, as mentioned in the introduction, the consistency between spatial navigation ability in
the real world and in a virtual environment task depends on the type of navigational task [
In particular, the aforementioned study reported a poor concordance for a task involving
pointing to the beginning and endpoint of the route, which is quite close to our path
integration task. This hypothesis is consistent with the weak correlations we found between
wayfinding and path integration performances, both in the real world and in the virtual environment:
the two tasks involve different cognitive processes, which don?t generalize similarly from one
environment to the other. The wayfinding task requires quite elaborate processing, while the
path integration only requires working memory processes (see Introduction). Second, this null
result could be caused by the low sensitivity of our virtual path integration task. Indeed, while
in the real-world performance was a continuous variable defined as the inverse of the error
angle, in Sea Hero Quest it could only take three values: one, two, or three stars. This ternary
metric might not be sensitive enough to capture subtle differences in the moderate sample size
used in this study (60 participants), unlike the original Sea Hero Quest study on mobile and
tablet (2.5m participants) [
]. Third, one could argue that path integration in a city is
different from path integration in a controlled virtual environment, as there are environmental
structures (e.g. street grid) and landmarks (e.g. buildings), which may help to judge distances
and directions. This would explain the small difference in mean error angle captured by the
real-world path integration routes between supposedly easy (one turn) and difficult (four
turns) routes, see Table 2.
Altogether, these results constitute a step toward the ability to remotely test people. This is
particularly valuable when certain categories of the population have difficulties in mobility,
like older people. Currently our results focused on young university students, and it will be
11 / 15
useful to extend to a broader population including elderly participants. Spatial ability
assessment provides the potential to act as an early stage diagnostic tool for Alzheimer?s dementia
(AD), because spatial disorientation is one of the earliest symptoms [
there is no standardized test for navigation deficits with AD patients, as diagnostics measures
are still focused on episodic memory deficits, despite their low sensitivity and specificity for
identifying at-risk individuals . Sea Hero Quest wayfinding task having real-world
ecological validity holds future promise for controllable, sensitive, safe, low-cost, and easy to
administer digital cognitive assessment.
S1 Fig. Wayfinding virtual task. Maps of wayfinding Sea Hero Quest levels 1, 6, 11, 16 and
43. Starting position and facing direction are indicated by a pale blue arrow, ordered goals by
red flags. Participants must memorize the map, and then navigate towards the goals in the
right order as quick as possible.
S2 Fig. Wayfinding real-world task in London (UK). Maps of real-world wayfinding routes
(top). Starting position and facing direction are indicated by a yellow arrow, ordered goals by
red dots. Participants must memorize the map, and then walk towards the goals in the right
order as quick as possible. Goals are materialized by remarkable doors (bottom).
S3 Fig. Wayfinding real-world task in Paris (France). Maps of real-world wayfinding routes
(top). Starting position and facing direction are indicated by a green arrow, ordered goals by
yellow dots. Participants must memorize the map, and then walk towards the goals in the right
order as quick as possible. Goals are materialized by remarkable facade (bottom).
S4 Fig. Path integration task in the virtual and real-world environments. A?Maps of Sea
Hero Quest path integration levels 14, 34, 44, 54 and 74. B?Itineraries of the path integration
task in Paris (France). Each color corresponds to a different itinerary.
S5 Fig. Path integration gender effect. Gender differences for the path integration task in the
video game (right) and in the real world (left) in Paris. In the real world, path integration
performance is the opposite of the average error angle. In the video game, path integration
performance is the average number of stars. Both measures have been standardized for comparison.
Black dots represent individual data points. Error bars represent standard errors.
S1 Video. Visualization of the wayfinding task in London and in the video game. Credits to
We thank Emmanuelle Bourigault and Mariam Yusuf for help in collecting pilot data in London.
Conceptualization: Lena Coutrot, Michael Hornberger, Hugo J. Spiers.
12 / 15
Data curation: Antoine Coutrot, Sophie Schmidt, Jessica Pittman, Lynn Hong.
Formal analysis: Antoine Coutrot, Jan M. Wiener.
Funding acquisition: Michael Hornberger, Hugo J. Spiers.
Investigation: Antoine Coutrot, Sophie Schmidt, Lena Coutrot, Jessica Pittman, Lynn Hong.
Methodology: Antoine Coutrot, Sophie Schmidt, Lena Coutrot, Hugo J. Spiers.
Project administration: Antoine Coutrot, Hugo J. Spiers.
Supervision: Antoine Coutrot, Hugo J. Spiers.
Validation: Antoine Coutrot.
Visualization: Antoine Coutrot.
Writing ? original draft: Antoine Coutrot, Hugo J. Spiers.
Writing ? review & editing: Antoine Coutrot, Jan M. Wiener, Christoph Ho?lscher, Ruth C.
Dalton, Michael Hornberger, Hugo J. Spiers.
13 / 15
14 / 15
1. Maguire EA , Burgess N , Donnett JG , Frackowiak RS , Frith CD , O'keefe J. Knowing where and getting there: a human navigation network . Science . 1998 ; 280 ( 5365 ): 921 - 924 . https://doi.org/10.1126/ science.280.5365.921 PMID: 9572740
2. Pine DS , Grun J , Maguire EA , Burgess N , Zarahn E , Koda V , et al. Neurodevelopmental aspects of spatial navigation: a virtual reality fMRI study . Neuroimage . 2002 ; 15 ( 2 ): 396 - 406 . https://doi.org/10.1006/ nimg. 2001 .0988 PMID: 11798274
3. Epstein RA , Patai EZ , Julian JB , Spiers HJ . The cognitive map in humans: spatial navigation and beyond . Nature neuroscience . 2017 ; 20 ( 11 ): 1504 - 1513 . https://doi.org/10.1038/nn.4656 PMID: 29073650
4. Ekstrom AD , Spiers HJ , Bohbot VD , Rosenbaum RS . Human Spatial Navigation . Princeton University Press; 2018 .
5. Mitolo M , Gardini S , Caffarra P , Ronconi L , Venneri A , Pazzaglia F . Relationship between spatial ability, visuospatial working memory and self-assessed spatial orientation ability: a study in older adults . Cognitive Processing . 2015 ; 16 ( 2 ): 165 - 176 . https://doi.org/10.1007/s10339-015 -0647-3 PMID: 25739724
6. Lokka IE , C?o?ltekin A , Wiener J , Fabrikant SI , Ro?cke C. Virtual environments as memory training devices in navigational tasks for older adults . Scientific Reports . 2018 ; 8 ( 10809 ): 1 - 15 .
7. Moffat SD , Zonderman AB , Resnick SM . Age differences in spatial memory in a virtual environment navigation task . Neurobiology of Aging . 2001 ; 22 : 787 - 796 . https://doi.org/10.1016/S0197- 4580 ( 01 ) 00251 - 2 PMID: 11705638
8. Tu S , Wong S , Hodges JR , Irish M , Piguet O , Hornberger M. Lost in spatial translation-A novel tool to objectively assess spatial disorientation in Alzheimer's disease and frontotemporal dementia . Cortex . 2015 ; 67 : 83 - 94 . https://doi.org/10.1016/j.cortex. 2015 . 03 .016 PMID: 25913063
9. Tsai HyS , Shillair R , Cotten SR , Winstead V , Yost E. Getting grandma online: are tablets the answer for increasing digital inclusion for older adults in the US? Educational Gerontology . 2015 ; 41 ( 10 ): 695 - 709 . https://doi.org/10.1080/03601277. 2015 .1048165 PMID: 26877583
10. Ruggeri K , Maguire A ? , Andrews JL , Martin E , Menon S. Are we there yet? Exploring the impact of translating cognitive tests for dementia using mobile technology in an aging population . Frontiers in aging neuroscience . 2016 ; 8 : 21 . https://doi.org/10.3389/fnagi. 2016 .00021 PMID: 27014053
11. Mallet KH , Shamloul RM , Corbett D , Finestone HM , Hatcher S , Lumsden J , et al. Recovernow: Feasibility of a mobile tablet-based rehabilitation intervention to treat post-stroke communication deficits in the acute care setting . PloS one . 2016 ; 11 ( 12 ):e0167950. https://doi.org/10.1371/journal.pone. 0167950 PMID: 28002479
12. Area?n PA, Ly KH , Andersson G . Mobile technology for mental health assessment . Dialogues in clinical neuroscience . 2016 ; 18 ( 2 ): 163 . PMID: 27489456
13. Coutrot A , Silva R , Manley E , de Cothi W , Sami S , Bohbot V , et al. Global determinants of navigation ability . Current Biology . 2018 ; 28 : 1 - 6 . https://doi.org/10.1016/j.cub. 2018 . 06 .009
14. Witmer BG , Bailey JH , Knerr BW , Parsons KC . Virtual spaces and real world places: transfer of route knowledge . International Journal of Human-Computer Studies . 1996 ; 45 ( 4 ): 413 - 428 . https://doi.org/10. 1006/ijhc. 1996 .0060
15. Ruddle RA , Payne SJ , Jones DM . Navigating buildings in ?desk-top? virtual environments: Experimental investigations using extended navigational experience . Journal of Experimental Psychology: Applied . 1997 ; 3 ( 2 ): 143 - 159 .
16. Richardson AE , Montello DR , Hegarty M. Spatial knowledge acquisition from maps and from navigation in real and virtual environments . Memory & cognition . 1999 ; 27 ( 4 ): 741 - 750 . https://doi.org/10.3758/ BF03211566
17. Conroy RA . Spatial navigation in immersive virtual environments . University College London; 2001 .
18. Wolbers T , Hegarty M. What determines our navigational abilities? Trends in Cognitive Sciences . 2010 ; 14 ( 3 ): 138 - 146 . https://doi.org/10.1016/j.tics. 2010 . 01 .001 PMID: 20138795
19. Ventura M , Shute V , Wright TJ , Zhao W. An investigation of the validity of the virtual spatial navigation assessment . Frontiers in psychology. 2013 ; 4 ( 852 ): 1 - 7 .
20. Cushman LA , Stein K , Duffy CJ . Detecting navigational deficits in cognitive aging and Alzheimer disease using virtual reality . Neurology . 2008 ; 71 ( 12 ): 888 - 895 . https://doi.org/10.1212/01.wnl. 0000326262 .67613. fe PMID : 18794491
21. Livingstone SA , Skelton RW . Virtual environment navigation tasks and the assessment of cognitive deficits in individuals with brain injury . Behavioural Brain Research . 2007 ; 185 ( 1 ): 21 - 31 . https://doi.org/10. 1016/j.bbr. 2007 . 07 .015 PMID: 17727970
22. Sorita E , N'Kaoua B , Larrue F , Criquillon J , Simion A , Sauze?on H, et al. Do patients with traumatic brain injury learn a route in the same way in real and virtual environments? Disability and Rehabilitation. 2013 ; 35 ( 16 ): 1371 - 1379 . https://doi.org/10.3109/09638288. 2012 .738761 PMID: 23244232
23. Claessen MH , Visser-Meily JM , de Rooij NK , Postma A , van der Ham IJ . A direct comparison of realworld and virtual navigation performance in chronic stroke patients . Journal of the International Neuropsychological Society . 2016 ; 22 ( 4 ): 467 - 477 . https://doi.org/10.1017/S1355617715001228 PMID: 26689246
24. Coughlan G , Coutrot A , Khondoker M , Minihane AM , Spiers H , Hornberger M. Impact of Sex and APOE Status on Spatial Navigation in Pre-symptomatic Alzheimer's disease . bioRxiv . 2018 ; 287722 .
25. Cogne ? M, Taillade M , N'Kaoua B , Tarruella A , Klinger E , Larrue F , et al. The contribution of virtual reality to the diagnosis of spatial navigation disorders and to the study of the role of navigational aids: A systematic literature review . Annals of Physical and Rehabilitation Medicine . 2017 ; 60 ( 3 ): 164 - 176 . https:// doi.org/10.1016/j.rehab. 2015 . 12 .004 PMID: 27017533
26. Coughlan G , Laczo? J, Hort J , Minihane AM , Hornberger M. Spatial navigation deficits-overlooked cognitive marker for preclinical Alzheimer disease? Nature Reviews Neurology . 2018 . https://doi.org/10. 1038/s41582-018-0031 -x PMID : 29980763
27. van der Ham IJ , Faber AM , Venselaar M , van Kreveld MJ , Lo?ffler M. Ecological validity of virtual environments to assess human navigation ability . Frontiers in Psychology. 2015 ; 6 ( 637 ): 1 - 6 .
28. Malinowski JC , Gillespie WT . Individual differences in performance on a large-scale, real-world wayfinding task . Journal of Environmental Psychology . 2001 ; 21 ( 1 ): 73 - 82 . https://doi.org/10.1006/jevp. 2000 . 0183
29. Moffat SD , Hampson E , Hatzipantelis M. Navigation in a ?Virtual? Maze: Sex Differences and Correlation With Psychometric Measures of Spatial Ability in Humans . Evolution and Human Behavior . 1998 ; 19 ( 2 ): 73 - 87 . https://doi.org/10.1016/S1090- 5138 ( 97 ) 00104 - 9
30. Gro?n G , Wunderlich AP , Spitzer M , Tomczak R . Brain activation during human navigation: gender-different neural networks as substrate of performance . Nature Neuroscience . 2000 ; 3 ( 4 ): 404 - 408 . https:// doi.org/10.1038/73980 PMID: 10725932
31. Astur RS , Tropp J , Sava S , Constable RT , Markus EJ . Sex differences and correlations in a virtual Morris water task, a virtual radial arm maze, and mental rotation . Behavioural brain research . 2004 ; 151 ( 1- 2 ): 103 - 115 . https://doi.org/10.1016/j.bbr. 2003 . 08 .024 PMID: 15084426
32. Goeke C , Kornpetpanee S , Ko?ster M, Ferna?ndez- Revelles AB , Gramann K , Ko?nig P. Cultural background shapes spatial reference frame proclivity . Scientific Reports . 2015 ; 5 ( 11426 ): 1 - 13 .
33. Vashro L , Cashdan E. Spatial cognition, mobility, and reproductive success in northwestern Namibia . Evolution and Human Behavior . 2014 ; 36 ( 2 ): 123 - 129 . https://doi.org/10.1016/j.evolhumbehav. 2014 . 09 .009
34. Richardson AE , Powers ME , Bousquet LG . Video game experience predicts virtual, but not real navigation performance . Computers in Human Behavior . 2011 ; 27 : 552 - 560 . https://doi.org/10.1016/j.chb. 2010 . 10 .003
Wiener JM , Bu?chner SJ, Ho?lscher C. Taxonomy of human wayfinding tasks: A knowledge-based approach . Spatial Cognition & Computation . 2009 ; 9 ( 2 ): 152 - 165 . https://doi.org/10.1080/ 13875860902906496
36. Etienne AS , Jeffery KJ . Path integration in mammals . Hippocampus . 2004 ; 14 ( 2 ): 180 - 192 . https://doi. org/10.1002/hipo.10173 PMID: 15098724
37. Loomis JM , Klatzky RL , Golledge RG , Cicinelli JG , Pellegrino JW , Fry PA . Nonvisual navigation by blind and sighted: assessment of path integration ability . Journal of Experimental Psychology: General . 1993 ; 122 ( 1 ): 73 . https://doi.org/10.1037/ 0096 - 3445 . 122 .1. 73
38. Harris MA , Wolbers T. Ageing effects on path integration and landmark navigation . Hippocampus . 2012 ; 22 ( 8 ): 1770 - 1780 . https://doi.org/10.1002/hipo.22011 PMID: 22431367
40. Wilcox RR , Rousselet GA . A Guide to Robust Statistical Methods in Neuroscience . Current Protocols in Neuroscience. 2018 ; 82 : 8 .42.1- 8 . 42 .30. https://doi.org/10.1002/cpns.41
Wilcox R. Inferences Based on a Skipped Correlation Coefficient . Journal of Applied Statistics . 2004 ; 31 ( 2 ): 131 - 143 . https://doi.org/10.1080/0266476032000148821
41. Pernet CR , Wilcox RR , Rousselet GA . Robust correlation analyses: false positive and power validation using a new open source Matlab toolbox . frontiers in Psychology. 2013 ; 3 : 1 - 18 . https://doi.org/10.3389/ fpsyg. 2012 .00606
42. Starrett MJ , Ekstrom AD . Perspective: Assessing the Flexible Acquisition, Integration, and Deployment of Human Spatial Representations and Information . Frontiers in Human Neuroscience. 2018 ; 12 : 281 . https://doi.org/10.3389/fnhum. 2018 .00281 PMID: 30050422
43. Hahnel C , Goldhammer F , Naumann J , Kro?hne U. Effects of linear reading, basic computer skills, evaluating online information, and navigation on reading digital text . Computers in Human Behavior . 2016 ; 55 : 486 - 500 . https://doi.org/10.1016/j.chb. 2015 . 09 .042
44. Zakzanis KK , Quintin G , Graham SJ , Mraz R . Age and dementia related differences in spatial navigation within an immersive virtual environment . Medical Science Monitor . 2009 ; 15 ( 4 ): 140 - 150 .
45. Serino S , Riva G . Getting lost in Alzheimer's disease: a break in the mental frame syncing . Medical Hypotheses . 2013 ; 80 ( 4 ): 416 - 421 . https://doi.org/10.1016/j.mehy. 2012 . 12 .031 PMID: 23374425
46. Lithfous S , Dufour A , Despre?s O. Spatial navigation in normal aging and the prodromal stage of Alzheimer's disease: insights from imaging and behavioral studies . Ageing Research Reviews . 2013 ; 12 ( 1 ): 201 - 213 . https://doi.org/10.1016/j.arr. 2012 . 04 .007 PMID: 22771718
47. Vl?ek K , Laczo? J. Neural correlates of spatial navigation changes in mild cognitive impairment and Alzheimer's disease . Frontiers in Behavioral Neuroscience . 2014 ; 8 ( 89 ): 1 - 6 .
48. Tu S , Spiers HJ , Hodges JR , Piguet O , Hornberger M. Egocentric versus Allocentric Spatial Memory in Behavioral Variant Frontotemporal Dementia and Alzheimer's Disease . Journal of Alzheimer's Disease . 2017 ; 59 ( 3 ): 883 - 892 . https://doi.org/10.3233/JAD-160592 PMID: 28697554