Tappigraphy: continuous ambulatory assessment and analysis of in-situ map app use behaviour

Abstract

While map apps on smartphones are abundant, their everyday usage is still an open empirical research question. With tappigraphy – the quantification of smartphone touchscreen interactions – we aimed to capture continuous data stream of behavioural human-map app usage patterns. The current study introduces a first tappigraphy analysis of the distribution of touchscreen interactions on map apps in 211 remotely observed smartphone users, accumulating a total of 42 days of tap data. We detail the requirements, setup, and data collection to understand how much, when, for how long, and how people use mobile map apps in their daily lives. Supporting prior research, we find that on average map apps are only sparsely used, compared to other apps. The longitudinal fluctuations in map use are not random and are partly governed by general daily and weekly human behaviour cycles. Smartphone session duration including map app use can be clearly distinguished from sessions without any map apps used, indicating a distinct temporal behavioural footprint surrounding map use. With the transfer of the tappigraphy approach to a mobile map app use context, we see a promising avenue to provide research communities interested in the underlying behavioural mechanisms of map use a continuous, in-situ momentary assessment method.

Full text

ZurichOpenRepositoryandArchiveUniversityofZurichUniversityLibraryStrickhofstrasse39CH-8057Zurichwww.zora.uzh.chYear:2022Tappigraphy:continuousambulatoryassessmentandanalysisofin-situmapappusebehaviourReichenbacher,Tumasch;Aliakbarian,Meysam;Ghosh,Arko;Fabrikant,SaraIAbstract:Whilemapappsonsmartphonesareabundant,theireverydayusageisstillanopenempiricalresearchquestion.Withtappigraphy–thequanticationofsmartphonetouchscreeninteractions–weaimedtocapturecontinuousdatastreamofbehaviouralhuman-mapappusagepatterns.Thecurrentstudyintroducesarsttappigraphyanalysisofthedistributionoftouchscreeninteractionsonmapappsin211remotelyobservedsmartphoneusers,accumulatingatotalof42daysoftapdata.Wedetailtherequirements,setup,anddatacollectiontounderstandhowmuch,when,forhowlong,andhowpeopleusemobilemapappsintheirdailylives.Supportingpriorresearch,wendthatonaveragemapappsareonlysparselyused,comparedtootherapps.Thelongitudinaluctuationsinmapusearenotrandomandarepartlygovernedbygeneraldailyandweeklyhumanbehaviourcycles.Smartphonesessiondurationincludingmapappusecanbeclearlydistinguishedfromsessionswithoutanymapappsused,indicatingadistincttemporalbehaviouralfootprintsurroundingmapuse.Withthetransferofthetappigraphyapproachtoamobilemapappusecontext,weseeapromisingavenuetoprovideresearchcommunitiesinterestedintheunderlyingbehaviouralmechanismsofmapuseacontinuous,in-situmomentaryassessmentmethod.DOI:https://doi.org/10.1080/17489725.2022.2105410PostedattheZurichOpenRepositoryandArchive,UniversityofZurichZORAURL:https://doi.org/10.5167/uzh-219669JournalArticlePublishedVersion
ThefollowingworkislicensedunderaCreativeCommons:Attribution-NonCommercial-NoDerivatives4.0International(CCBY-NC-ND4.0)License.Originallypublishedat:Reichenbacher,Tumasch;Aliakbarian,Meysam;Ghosh,Arko;Fabrikant,SaraI(2022).Tappigraphy:continuousambulatoryassessmentandanalysisofin-situmapappusebehaviour.JournalofLocationBasedServices:1-27.DOI:https://doi.org/10.1080/17489725.2022.2105410

Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=tlbs20
Journal of Location Based Services
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tlbs20
Tappigraphy: continuous ambulatory assessment
and analysis of in-situ map app use behaviour
Tumasch Reichenbacher, Meysam Aliakbarian, Arko Ghosh & Sara I.
Fabrikant
To cite this article: Tumasch Reichenbacher, Meysam Aliakbarian, Arko Ghosh & Sara I.
Fabrikant (2022): Tappigraphy: continuous ambulatory assessment and analysis of in-situ map app
use behaviour, Journal of Location Based Services, DOI: 10.1080/17489725.2022.2105410
To link to this article: https://doi.org/10.1080/17489725.2022.2105410
© 2022 The Author(s). Published by Informa
UK Limited, trading as Taylor & Francis
Group.
Published online: 26 Jul 2022.
Submit your article to this journal
View related articles
View Crossmark data

Tappigraphy: continuous ambulatory assessment and
analysis of in-situ map app use behaviour
Tumasch Reichenbacher
a
, Meysam Aliakbarian
a
, Arko Ghosh
b
and Sara I. Fabrikant
a
a
Department of Geography, University of Zurich, Zurich, Switzerland;
b
Institute of Psychology, Leiden
University, Leiden, The Netherlands
ABSTRACT
While map apps on smartphones are abundant, their everyday
usage is still an open empirical research question. With tappi-
graphy – the quantication of smartphone touchscreen inter-
actions – we aimed to capture continuous data stream of
behavioural human-map app usage patterns. The current
study introduces a rst tappigraphy analysis of the distribution
of touchscreen interactions on map apps in 211 remotely
observed smartphone users, accumulating a total of 42 days of
tap data. We detail the requirements, setup, and data collection
to understand how much, when, for how long, and how people
use mobile map apps in their daily lives. Supporting prior
research, we nd that on average map apps are only sparsely
used, compared to other apps. The longitudinal uctuations in
map use are not random and are partly governed by general
daily and weekly human behaviour cycles. Smartphone session
duration including map app use can be clearly distinguished
from sessions without any map apps used, indicating a distinct
temporal behavioural footprint surrounding map use. With the
transfer of the tappigraphy approach to a mobile map app use
context, we see a promising avenue to provide research com-
munities interested in the underlying behavioural mechanisms
of map use a continuous, in-situ momentary assessment
method.
ARTICLE HISTORY
Received 13 August 2021
Accepted 20 July 2022
KEYWORDS
Mobile map apps; mobile
map app use; tappigraphy
1. Introduction
Maps play an important role in humans’ everyday activities. Map use
encompasses the acquisition of spatial knowledge from maps, sense-
making of the environment, development of a mental representation of
space, and nally the acquisition of spatial knowledge about geographic
features and their relations, including space-time events and processes at
various scales (Kimerling et al. 2016; MacEachren 2004). Map use involves
the reading, analysis, and interpretation of space-time processes (Kimerling
et al. 2016). During the last two decades, map use has shifted from static
CONTACT Tumasch Reichenbacher tumasch.reichenbacher@geo.uzh.ch
JOURNAL OF LOCATION BASED SERVICES
https://doi.org/10.1080/17489725.2022.2105410
© 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives
License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduc-
tion in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.

paper maps to interactive mobile map apps. Mobile maps can be dened as
any cartographic product or application explicitly designed for display and
use on a portable and movable computing device (Ricker and Roth 2018),
such as a smartphone or tablet (Muehlenhaus 2013). With recent technolo-
gical advancements in mobile communication and positioning technology,
smartphone hardware, and online geographic databases, geographic infor-
mation is increasingly consumed in everyday activities often on the go in
the form of so-called map apps. The map app category is, however, loosely
dened, and can include dedicated map apps (e.g. Google Maps, Apple
Maps, maps.me, etc.), travel apps with a map interface, tracking apps, and
basically any location-based services (LBS) relying on map interfaces. Touch-
based user interfaces found on smart, assistive devices, in particular, have
changed the way in which maps are consumed and being used and oer
new ways of studying map use. Still, ambulatory in-situ assessment meth-
ods aimed at collecting a continuous stream of map use data to study
mobile map use behaviour ‘in the wild’ are still rare (Riegelsberger and
Nakhimovsky 2008; Savino et al. 2021). Instead, aggregated map app down-
load statistics and self-reports still serve as a rst proxy for mobile map use.
Averaged download statistics do not oer ne-grained insights on map use
behaviour itself, and self-reports on digital media use are notoriously inac-
curate (Parry et al. 2021). Hence, basic research questions from a location-
based and geographic information science perspective still remain open:
How much do people use mobile map apps in everyday life? When and for
how long do people use mobile map apps? How do people use mobile map
apps? To answer these basic and pertinent questions would allow cartogra-
phers, geographic information scientists, or location-based services (LBS)
developers to better understand mobile map use behaviour ‘in the wild’,
and this, in turn, would support the user-centred design and development
of future human and context-dependent map apps for an intended target
audience (Huang et al. 2018; Thrash et al. 2019). However, the rst chal-
lenge would be to obtain continuously collected, in-situ map use data
which is typically not openly available yet for the research community.
Possible reasons for the scarcity of ecologically valid ne-grained mobile map
use data and research may relate to the signicant eort required to run large-
scale empirical studies, requiring substantial monetary, temporal, technical, and
human resources; in addition, aside from the diculty of nding an adequately
sized and willing participant pool, other reasons include the complex technical
and experimental setups, controlling for potentially confounding variables
occurring ‘in the wild’, handling additional privacy concerns of participants,
and the obtrusiveness of observational procedures, including draining batteries
and push notications.
2T. REICHENBACHER ET AL.

In light of steadily progressing digital transformation, ecologically validated
empirical evidence is indispensable for informing and improving the design for
useful and usable mobile map apps, location-based services, and smart naviga-
tion systems that support the mobility of a broad set of individuals in various
rapidly changing use contexts in their everyday lives.
In sum, despite the ubiquity of mobile devices, we still know very little about
ne-grained typical mobile map app use with smartphones. We thus introduce
tappigraphy to the LBS and GIScience elds, as a complementary method for
continuous, unobtrusive collection and analysis of ‘natural’, ecologically valid
smartphone touch data as a proxy for everyday in-situ map use behaviour. The
main point of this article is thus to explore how far one can push tappigraphy as
the sole method to infer map app user behaviour, without the need of any other
typical human behaviour tracking methods. With the current empirical map app
study, we propose to bridge micro-level behaviour analysis known from cogni-
tive neurosciences and psychology with macro-level eld studies typically con-
ducted in GIScience.
It is important to note here that tappigraphy is dierent from classical,
typically small-scale user studies out in the world, including usability studies
or controlled lab experiments that are commonly used in cartography and LBS
studies (see Table 1). Tappigraphy follows instead the approach of typically
remote, in-situ ambulatory assessment and ecological momentary assessment
(EMA) of a large quantity of users in their everyday settings. Fahrenberg et al.
(2007) dene ambulatory assessment as ‘the use of (mainly) electronic devices
and computer-assisted methods of data collection suitable for use in the eld to
collect self-report data, behaviour observation data, psychometric behaviour
measures, and physiological data in unrestrained daily life settings’ (p.207).
Similarly, EMA involves ‘repeated sampling of subjects’ current behaviours and
experiences in real time, in subjects’ natural environments’ (Shiman, Stone,
and Huord 2008, 1).
Hence, tappigraphy is dened here as the remote, inobtrusive, and almost
continuous registering and quantication of smartphone touchscreen events in
people’s everyday use situations. It records every single tap of a user on the
smartphone screen with its timestamp. It has historically been developed and
Table 1. Comparison of classic user tracking methods in the field compared with tappigraphy.
Experiment design Classic user tracking methods Tappigraphy
Sample size 30 – 50 >200
Ecological validity Low High
Experimental control High Low
Direct interaction with participants Yes No
Experiment goals and tasks Determined Not determined
Obtrusiveness Method-dependent Low
Observation duration Discretely, set minutes to hours Continuously days to months
Data sampling Milliseconds to minutes Milliseconds
Ease of use/running costs High Low
Data Highly variable Touches on smartphone display
JOURNAL OF LOCATION BASED SERVICES 3

applied in the domains of cognitive and behavioural neuroscience, and thus
data is collected on the temporal scale of milliseconds. Tappigraphy has an
emerging role in quantifying hidden human health variables such as sleep
patterns, cognitive processing speeds, and human disease activities (Balerna
and Ghosh 2018; Borger, Huber, and Ghosh 2019; Duckrow et al. 2021; Huber
and Ghosh 2021). Studying the frequency and speed of taps over longer time
periods during humans’ natural usage situations allows, for instance, for the
tapping behaviour to be related to various cognitive processes. Certainly, the
LBS and GIScience community is very familiar with commonly known empirical
methods to digitally observe mobile map use, such as human-map display
interaction logging, automatic map screen recording, mobile eye tracking,
think aloud protocols, digital surveys, etc. Unlike tappigraphy, classic usability
tracking methods typically require either at some point of the empirical data
collection campaign direct interventions or contact with study participants by
researchers, or these are still typically run in a controlled empirical study setting
outside or (even remote online) indoors. Tappigraphy, on the other hand,
provides in its purest form the capacity to unobtrusively capture, ecologically
valid, everyday, ne-grained ambulatory human-system interface interaction
in situ (i.e. tap events are continuously recorded in milliseconds), over a long
period of time (i.e. weeks and months, etc.), and for a large number of totally
anonymous participants (i.e. hundreds of users can be studied in parallel) with-
out any intervention or direct contact by researchers, and all this at considerably
low eort and running costs (see Table 1).
The reason for using tappigraphy in the current study is to evaluate this type
of EMA method for unobtrusively studying the complex human- and context-
dependent process of daily map use including map reading, map analysis, map
interpretation, space-time decision-making and spatial behaviour of many users
in the ‘wild’.
We thus set out to nd dierent types of map use behaviours captured only
by map app touches with respect to individual and user group dierences of
participants, geographic context, movement modalities, and purposes of use
that are totally unknown.
We hypothesise that map use during longer periods of immobility (e.g. at
home, sitting and/or standing for a long time in public transport, during
work, or in a café, etc.) will show signicantly dierent map app touch
patterns from map app touch patterns during self-propelled locomotion,
because these will be interrupted by only a few and short periods of immo-
bility. When exploring new and unknown environments on the map, or while
planning a route, this most likely will result in longer map use sessions with
many map app tap interactions. This will include map app taps to manipulate
the map view, as well as to zooming panning touches, or when searching for
information on the smartphone. In contrast, while en route, map use should
manifest as a series of short and less frequent map app touch interactions, for
4T. REICHENBACHER ET AL.

example, when users try to self-localisation on a GPS-enabled smartphone.
With the availability of ne-grained map use data captured by map app
touches, we thus expect to identify map use behaviour patterns that can
be linked to specic map purposes and use situations. We already know that
people dier in their spatial abilities, background, and training (Hegarty and
Waller 2005). We thus expect not only to be able to infer intra-individual
dierences in map use behaviour, as described above, but also group dier-
ences, even though we have no background information about tracked
participants, except for their tapping behaviour. We aim to detect clusters
of individuals with similar map use behaviours based on their map app
touches, as detailed in Section 3.
With this current study, we therefore aim to transfer the tappigraphy
method from cognitive neuroscience to GIScience, cartography, and LBS as
a complementary unobtrusive and ambulatory empirical user monitoring
approach to better understand how people use map apps for everyday
mobility tasks on their own smartphones with least experimental control. In
doing so, on the one hand, we aim to provide a rst step to be able to upscale
ndings from low-level map app use behaviour obtained from small-scale
controlled user studies to aggregated and publicly map app use statistics
currently available. On the other hand, tappigraphy oers the opportunity to
eventually downscale map app touches collected from a large uncontrolled
user sample to neural correlates of human map use behaviour at individual
level (Fabrikant 2022).
In summary, we seek to 1) introduce tappigraphy borrowed from cognitive
neuroscience into the research context of GIScience, in particular for empirical
user-centred research in LBS and cartography, 2) detail the in-situ tappigraphy
data collection and analysis approach in the context of mobility, and 3) demon-
strate by means of examples how tappigraphy data could be analysed for
getting at individual map use patterns that ultimately might help improve large-
scale urban mobility.
2. Related work
We review related work with respect to large-scale (aggregated) download
statistics and studies of general app use, followed by small-scale studies on
map app use from GIScience, cartography, and human-computer interaction.
2.1. General app download statistics and app use studies
Publicly available data on map app use is non-existent for the open academic
research community. As map apps are considered a market share of the
mobile app business, most data on mobile map app use are based on market
surveys and app download statistics from the major mobile operating systems
JOURNAL OF LOCATION BASED SERVICES 5

providers, i.e. the Google Playstore and the Apple App Store. These data
provide a macroscopic approximation of map app use in comparison to
other app categories, such as gaming, social media, communication, or enter-
tainment. For instance, a survey showed that in 2018 62% of people in Japan
used maps on their smartphones (Sugimoto et al. 2021). Such statistics also
reveal that map app use is infrequent. While Google Maps is ranked sixth with
mobile apps users in the UK in 2019, with about 50% penetration of the total
adult mobile app users, the total app minutes spent on Google Maps is only
5% of the time spent on YouTube, and less than 10% of time spent on
Facebook (UKOM 2019).
In a large-scale survey of mobile app usage Böhmer et al. (2011) measured
the app usage of 4125 users with a specially developed app called AppSensor.
Their results showed an average smartphone use of about 59 minutes per day
and an average app session time of 71 seconds. For the app category travel that
includes Google Maps or Waze, they found an average usage time of 44.72
seconds. Furthermore, total app usage was lowest at 5 am and peaks at 6 pm.
However, at 5 am, relative app usage for the category ‘travel’ was highest, at
2.6%. Similar results were reported by Falaki et al. (2010) in a study with 255
users. However, they found a range of use time per day between 30 and 500
minutes with users uniformly distributed within this range. Average session
times were found to be between 20 and 250 seconds. Interaction sessions
with maps were the longest.
A similar data collection approach was used for another survey of app
usage (Do, Blom, and Gatica-Perez 2011). In addition to app usage logs for
77 smartphone users, the location when using an app was recorded too.
Results for the category of maps showed the lowest numbers compared to
SMS, Voice Call, or Web related apps. Regarding location, maps were used most
on holiday, when relaxing, at restaurants, and to a lesser degree, during
transport.
In an even smaller study, Carrascal and Church (2015) observed 18 partici-
pants for 2 weeks in-situ with an app logger. While they found an average
mobile usage of over four and a half hours per day, they registered an average
duration of sessions in the category Travel&Local (including map apps) of 111.6
seconds, which was half of the duration for Entertainment, but in a similar range
as Social Networking. For app launches, Travel&Local accounted for 1.92%, while
Social Networking accounted for 17.98%, Browser&Search for 1.3%, and SMS/
Texting for 11.04%, respectively.
In another study (Banovic et al. 2014) the authors categorised the smart-
phone use of ten participants over a period of 18–36 days into three dierent
interaction behaviour groups, based on duration and type. They dened very
short interactions on the locked or home screen as glance sessions. Review
6T. REICHENBACHER ET AL.

sessions encompass relatively short interactions periods with one or more apps.
Finally, engage sessions occur when users engage for longer interaction with
apps. Engage session were found to be longer than 60 seconds.
In a large-scale study, Church, Cherubini, and Oliver (2014) collected data on
participants’ information needs and the addressing of these needs in-situ with
a snippet-based diary technique using SMS and MMS data. Their data set
included everyday information needs data for 108 users. Of all information
needs reported by participants, 5890 (61.5%) were satised. Of the 5890 satis-
ed information needs, 106 were addressed by online maps on the internet
(1.8%). A total of 55 information needs (0.8%) were addressed by GPS/map
routing services when a clear destination had to be reached.
In a recent study on navigation app use for pedestrians, Fonseca et al. (2021)
collected responses from 1438 people recruited in the cities of Bologna (Italy)
and Porto (Portugal). While 92% of the respondents reported using smart-
phones intensively, only 42% stated that they used map apps, mainly to nd
locations and to obtain the shortest routes between two locations. Google Maps
was reported to be the most used map app. As the main reason for not using
map apps more extensively, study participants stated a lack of need. As the
study authors suggest, this might indeed be an expected response, as most
people navigate daily within a well-known familiar environment, and an
increased need for a map app is only relevant for navigation and waynding
in a novel and unknown environment, for example, when tourists explore a new
holiday location. Another nding of the study was the individual factors that
explain variations in navigation app use. In particular, the study revealed age to
be an important factor. For the age group of 24 years old or younger, reported
map app use was 50%, compared to only 25% for those aged 65 and older. As
suggested earlier, if map app use is related to background knowledge of
a traversed environment, it would make sense that older respondents might
have a lesser need to use a map, as they might have been exposed longer, and
thus have accumulated more local knowledge of their residential environment
and local neighbourhoods than younger adults.
2.2. Small-Scale mobile map use studies in GIScience, cartography, and
human-computer interaction
The few studies that have already been conducted are not typically dedicated to
specic mobile map app use analysis, but rather study map use as a by-product,
and either focus on navigation and waynding processes, or on the usability
and human-computer interaction with mobile map displays. Specically,
numerous eld studies have already been conducted that investigate waynd-
ing in situ (Brügger, Richter, and Fabrikant 2016, 2019; Delikostidis and van
Elzakker 2011; Huang, Schmidt, and Gartner 2012). Commonly used data collec-
tion methods for such studies are questionnaires, interviews, eye-tracking, and
JOURNAL OF LOCATION BASED SERVICES 7

video recordings and interaction logging. One of the few map use studies in the
eld focusing solely on human-computer interaction issues using a map-based
web browser is described by Riegelsberger and Nakhimovsky (2008). Other
studies focus on understanding the interaction of users with maps by recording
touches on smartphones or touch screens. For example, to evaluate the usability
of smartphone screen interfaces for elderly people in the context of classic
human-computer interaction (HCI) research, Kobayashi et al. (2011) monitored
the display interactions of 20 participants with a classic HCI Wizard-of-Oz pro-
totype, including participants’ tapping, pinching, or dragging behaviour. Colley
and Häkkilä (2014) tested the performance of a novel interaction concept based
on the distinction of dierent ngers used, and in three user studies with small
user samples of 37, 13, and 25 participants, respectively. These authors showed
that multi-nger interaction was perceived to be faster and more valuable to
users. Another study looking into dierences between ‘digital natives’ and
‘digital immigrants’ (e.g. adults) investigated users’ performance in spatial
tasks on touch screens when interacting with 3D environments. Response
time and gestures were recorded accordingly (Herman and Stachoň 2018).
While these studies also investigated human display interactions using touchsc-
reens, they mainly focused on typical HCI usability issues, based on predened
user tasks in a controlled lab experiment setting. As typical in HCI usability
studies, small samples (<100 participants) were being monitored in the above
reviewed studies.
Perhaps the most comprehensive study on mobile map app use that comes
closest to the sample sizes that tappigraphy is designed for was conducted by
Savino et al. (2021). With their developed wrapper app MapRecorder Savino et al.
(2021) continuously monitored participants’ interactions with the Google Maps
app on their own smartphones (e.g. coordinates of map screen touchpoints,
type of touches, keyboard inputs, changes of map zoom level or map display
centre, etc.). MapRecorder requires participants to instal this wrapper app on
their smartphones, and to be willing to use it over the native Google Maps app.
As in previously reviewed studies above, only a small sample of 28 participants
familiar with their environment in total used MapRecorder for four consecutive
weeks. While the number of participants was relatively small compared to our
tappigraphy study, this is justied by their specic aim to focus on usability
issues and compare residents and tourists in a given location. The average
number of Google Maps sessions per participants in the ‘local user’ sample
was only 15, and their Google Maps app session had a duration of 65 seconds on
average. The second sample comprised 60 tourists who used the MapRecorder
app for one single day. The average number of sessions per tourist was 19, and
their map app sessions lasted on average 52 seconds. The authors identied
four typical Google Maps app use types: map-view manipulation, directions,
place, and search. The proportion of map use types for local participants was
67.5% for map-view manipulation, 21.1% for direction, 8.2% for place, and 3.2%
8T. REICHENBACHER ET AL.

for search, which had great similarity with the tourist sample. Local Google Map
users had a typical app use type sequence of map-view manipulation – search –
place – direction. This sequence and the large proportion of map-view manip-
ulations suggests that for local Google Map app users, a central characteristic of
Google Maps use is of exploratory spatial search behaviour, such as zooming
and panning the map app. Additionally, logging text input and thus revealing
actual search terms oers potentially deeper insights into the purpose and
context of a de-facto standard map app use. This, however, also raises critical
privacy issues, especially if the map app is of commercial nature, and aimed for
economic interests. This hinders the scalability of such a targeted wrapper
approach for empirically studying a large and diverse user sample, and limits
open science research.
Next, we present the design, deployment, and analysis of a rst tappigraphy
study with more than 100 users, focusing on smartphone touch interactions
related to map apps, and conceptualised as a steppingstone for studying map
use behaviour.
3. Map app tappigraphy study design and deployment
Unlike the above reviewed app user studies, we take a quite dierent user
monitoring approach. Specically, we do not invite participants to a research
lab to be given a pre-loaded phone to take home to use. We also do not ask
participants to perform specic tasks either at home, in a controlled lab envir-
onment, or remotely online. Our approach lets participants naturally live their
everyday lives in their own settings, using their own smartphones whenever and
for whatever they wish, while we continuously record their smartphone tapping
behaviour. We do not know who the participants are, why they participate,
where they live, or any other personal information, except that they allow us to
track their smartphone touches with their informed consent. This very natur-
alistic study setting comes with great benets of high ecological validity and
large user sample recruitment, but at the cost of information parsimony, thus no
further knowledge about the users, and limited experimental control.
The map app data showcased below in our rst case study of map app use is
a secondary use example of smartphone tappigraphy data (Balerna and Ghosh
2018; Borger, Huber, and Ghosh 2019; Ghosh, Pster, and Cook 2017; Huber and
Ghosh 2021). As mentioned earlier, tappigraphy relates to the inobtrusive,
almost continuous, in-situ recordings of smartphone touchscreen events with-
out any direct interactions with the user. In a classic tappigraphy campaign,
a smartphone user is tracked for weeks and month, and their taps initiated for
various reasons on the smartphone screen, in dierent periods of times, and
within many smartphone sessions are continuously recorded at high temporal
resolution (i.e. in the range of milliseconds) without any direct interactions or
observation by experimenters. In contrast to lab experiments or eld studies,
JOURNAL OF LOCATION BASED SERVICES 9

tappigraphy is a form of ambulatory assessment, i.e. a device-supported record-
ing of human behaviour in everyday activities. Of course, tappigraphy can
complement existing well-known monitoring methods being used to study
human-map interactions, such as interaction logging, video tracking, eye-
tracking, or similar.
As tappigraphy data are continuously recorded over a long campaign period,
the combined data collection from the University of Zurich and Leiden
University was frozen in August 2018 for this study. Specically collected
tappigraphy data for this case study ranges from 2014 to 2018.
3.1. Participants
Participants were recruited via on-campus yers and promotional emails at the
University of Zurich and at Leiden University. Subjects who were not right-
handed, healthy, or with any permanent hand injuries were eliminated (self-
declared). All 211 participants included for analysis provided informed consent
to the anonymously stored and shared smartphone data collection. The studies
were approved by the Ethical Committee of the Institute of Psychology, and by
the Canton of Zurich, enforcing the Swiss Human Experimentation Act. The rst
participant included in this analysis was recruited in January 2014 (from one of
the authors). The raw data were gathered by the Cognition in the Digital
Environment Laboratory (CODELAB), at Leiden University. For representative
demographic information from a largely overlapping database at Leiden
University (see Huber and Ghosh 2021). As the use of maps may be indicative
of additional location services, to preserve user privacy any associated demo-
graphic information was eliminated prior to data sharing.
3.2. Materials
We applied the TapCounter app, provided by QuantActions in Lausanne,
Switzerland, to collect tapping data. This app requires a smartphone with
a touchscreen running on the Android operating system. This was an inclusion
criterion for participation to which participants had to consent. The market
share for Android operating systems from 2014 to 2018 was on average 58%,
and for iOS 39%.
1
Once installed, the TapCounter app runs in the background
and continuously records the name and timestamps of all touchscreen interac-
tions on the app in the foreground of an unlocked smartphone. The recordings
have an approximate error margin of 5 milliseconds. Only user taps to unlock
and lock a smartphone, and the sequence of touches on apps in the foreground,
that is, immediately used by the phone user, are recorded. This means that
touches on apps such as Facebook, Twitter, or Google Maps are recorded, but
no other data or information content is registered to assure the privacy of the
10 T. REICHENBACHER ET AL.

user. Moreover, as individual app labels might allow identifying characteristics
of individuals and possibly infringe their privacy, the individual app names were
not used in the analysis.
3.3. Procedure
After providing anonymously informed consent through a website to be again
anonymously observed, study participants were asked to download and instal
the TapCounter app on their own Android smartphones. Participants were
instructed to not share their smartphone with others during the campaign
and to run the app for at least two consecutive weeks. All recorded data was
assigned a unique user code, encrypted, and streamed to cloud storage, from
where the data was later accessed through cloud-based services provided by
the QuantActions platform.
3.4. Data and processing
The raw tapping timestamps were pre-processed and stored in MATLAB les
using the parser extractTaps (QuantActions Ltd. Lausanne, Switzerland). For this
pre-processing step, all taps were grouped into two app categories based on
the type of app, i.e. map apps and all other apps. We categorised taps on map
apps including Google Maps, CityMaps2Go, Maps.me, Citymapper, MMApp, and
MapMyRun. The second category encompasses those we considered non-map
apps. These two categories were coded with 1 for map apps, and 0 for other
apps, respectively. The collected smartphone touchscreen timestamps were
then extracted from the MATLAB les and loaded into a Postgres database
2
for further analysis. All tapping analyses were run in Jupyter Notebook,
3
using
Python version 3.8.8.
Below, we show an extract of the raw tappigraphy data. The rst column is
a sequentially increasing unique ID to identify each tap, and the second column
contains a participant ID (293), followed by a timestamp in seconds. The next
column stores the tap type, i.e. 0 for unlocking, 10 for locking a phone, and 1 for
any other tap on the smartphone within an unlocking and locking session. The
last column holds the app category code, i.e. 1 for map apps and 0 for all other
apps.
54476530,293,1521299792.768,0,0
26210380,293,1521299793.521,1,0
26210381,293,1521299797.427,1,0
55875211,293,1521299799.827,10,0
54476531,293,1521303003.902,0,0
26210382,293,1521303006.439,1,0
26210383,293,1521303007.082,1,0
26210384,293,1521303007.791,1,0
JOURNAL OF LOCATION BASED SERVICES 11

26,210,385,293,1,521,303,010.233,1,0
26210386,293,1521303011.603,1,0
26210387,293,1521303012.282,1,0
26210388,293,1521303016.718,1,0
26210389,293,1521303050.46,1,0
26210390,293,1521303063.754,1,0
26210391,293,1521303075.009,1,0
26210392,293,1521303081.619,1,0
26210393,293,1521303082.869,1,0
26210394,293,1521303092.675,1,0
26210395,293,1521303121.852,1,0
26210396,293,1521303146.705,1,0
26210397,293,1521303161.935,1,0
26210398,293,1521303204.943,1,0
26210399,293,1521303241.82,1,0
55875212,293,1521303302.458,10,0
Once participants activated the TapCounter app on their smartphone, all taps
were automatically recorded for the campaign period until the moment when
users deactivated the app again (Figure 1). The tappigraphy data is structured
by three dierent types of events: start (0) and stop (10) events that correspond
to unlocking and locking the smartphone, dening a phone session, and all
other tap events on apps in the foreground (1). A phone session may contain
many, one or no taps at all, and they can be of varying durations. The unevenly
spaced small dots in Figure 1 represent a sequence of display touches, i.e. the
continuous sequence of taps by a user.
3.5. Results
Returning to the ne-grained map app use questions posed in the introduction:
How much do people use mobile map apps in everyday life? When and for how
long do people use mobile map apps? How do people use mobile map apps? we
structure the answers to those questions in below results section.
Figure 1. Structure of tappigraphy data: the black dots symbolise individual taps on the display
within phone sessions.
12 T. REICHENBACHER ET AL.

3.6. How much do people use mobile map apps?
For analysing how often people use mobile maps we only included 170 parti-
cipants (80.6%) out of the total of 211 participants who tapped at least once on
a map app during their monitoring period. Map taps were dened as smart-
phone touches on an app categorised as ‘map app’ used in the foreground. The
total number of map taps for our sample was 256,397 (0.7%), compared to
36,241,971 taps (99.3%) on other (non-map) apps. A total of 41 participants
(20%) never used a map app during the data collection campaign, which, on
average, lasted 55 consecutive days. The number of map taps per participant
ranged from 1 to 21,433, on average 1,508 taps, with a standard deviation of
2,647 taps. However, as the recording period varied considerably between
participants (7–430 days), we assumed the larger number of map taps for
some of the participants to be the result of longer total use time (Figure 2, left).
To compensate for this eect, we normalised the raw map tap counts by the
total use period duration for each participant. The histogram (Figure 2, right
panel) shows the normalised map counts as the number of map taps per total
use time. Note that we excluded 11 outliers with more than four map taps from
the histogram to increase legibility. Almost 50% of the participants tapped
fewer than 0.5 times per hour on a map app over the entire recording period
(Figure 2, right panel).
Overall, the total numbers of map taps were very low. We thus interpret this
as map app use is rather scarce in monitored participants’ everyday life. On
average, map taps accounted for only 0.89% of all taps. In other words, non-map
app taps far outnumbered map taps and support the infrequency of map app
use for our studied population. The overall extent of map app use revealed
insights into everyday mobile map use behaviour, particularly the scarcity of
map app use, compared to other apps used on smartphones.
Figure 2. The number of map taps across participants greatly varies due to different lengths of
campaign periods (left panel); most participants had very few map taps; on average map apps
overall were used only once (right panel).
JOURNAL OF LOCATION BASED SERVICES 13

3.7. When and for how long do people use mobile map apps?
Next, we wished to investigate temporal patterns of mobile map use and how
the frequency of map use is distributed over participants’ hours of their day, and
days of the week. The average number of map taps over a whole day showed
a clear day-night pattern (Figure 3). From 7 am the average map tap numbers
steadily increase with a rst peak of about 100 taps around 1 pm (highlighted in
red). The highest average number (113 map taps) was observed around 5 pm
(highlighted in red). After that, the numbers of map taps dropped slightly before
they became signicantly lower from 10 pm onwards. Between 2 and 6 am they
fell below 10 map taps.
Extending our analysis to a whole week, we can observe a weekly use pattern
(Figure 4). The aggregated number of map taps showed a clear day-night
pattern over the week with noticeable peaks around midday and in the early
evening. The pattern appeared very uniform, and weekdays and weekend days
did not substantially dier in the number of map taps. Looking more closely at
the aggregated numbers of map taps over the course of a week, we can detect
the highest numbers of map taps occurring on Thursday evening; Friday after-
noon, and early evening; Saturday around noon; and on Sunday afternoon
(Figure 4). On Saturday, the number of map taps was again lower and more in
the range of the Monday to Wednesday pattern. Moreover, the peaks around
midday and early evening were not as distinct compared to the pattern during
the rest of the week.
Dierent map use frequencies over time suggest that there may also be
dierent modes of map use, e.g. a longer series of map taps when looking up
a location and planning the route to that location, versus one short map tap to
conrm the current position. Aside from the question of how much and when
people use map apps, we further extend our analysis of map app touches to
explore how map apps are used by studying the sequences of map taps within
phone sessions (see Figure 1).
Figure 3. The map taps followed a day (light grey) night (dark grey) usage pattern and showed
clear peaks around lunch (13h) and before dinner time (18h) (red bars).
14 T. REICHENBACHER ET AL.

Figure 4. The weekly pattern showed the highest map tap numbers per hour (dark blue cells)
mainly in the afternoon and evening hours of the weekend (Thursday to Sunday).
JOURNAL OF LOCATION BASED SERVICES 15

3.8. How do people use mobile map apps?
As mentioned earlier, a phone session was dened for our study as the time
span between unlocking (start) and locking (end) of the smartphone screen. We
found a total of 453,452 such phone sessions in our collected dataset.
Analysing map taps at the level of individual phone sessions allowed us to
study map tapping behaviour in more detail. Our intention was twofold. First, as
we can determine from the daily and weekly pattern of map taps (Figures 3 and
4), there were longer periods with no touchscreen events at all, for example
during the night when participants are sleeping. Tap sequences thus seem more
likely to happen within brief and distinct periods of times, during the day/week.
We also contend that most map use tasks require more than one single map app
tap, and thus the analysis of map use behaviour should lend itself to analysing
multiple taps and characteristics of such tap sequences at various levels of
granularity, including a single phone session.
The histogram of map phone session duration shows a similarly skewed
distribution (Figure 5, left panel) as the overall number of recorded map taps
(Figure 2, right panel). Note that we removed 680 outliers with phone session
duration of 20 minutes and longer from the histogram to increase legibility.
From the remaining 9830 phone sessions, about 40% had a duration of equal to
or less than 60 seconds. Most phone sessions with map taps had a duration
between 0 and 1200 seconds, i.e. approximately 20 minutes. If we compare
these numbers with the duration of the phone sessions containing taps that
were not related to map apps, we observe a similar distribution, although with
442,942 non-map app tap sessions, the magnitude was signicantly larger.
To dierentiate between possible use modes of map apps (e.g. map use for
planning, map use while navigating), we needed to consider the frequency of
taps within map phone sessions. The histogram of map taps per map phone
session showed that almost 50% of phone sessions with map taps had fewer
Figure 5. The majority of the 9830 map tap phone sessions had a duration of fewer than two
minutes (left panel); about half of the 10,001 map tap phone sessions had no more than 10 taps
(right panel).
16 T. REICHENBACHER ET AL.

than 10 taps per session (Figure 5, right panel). Please note that we removed 399
outliers with more than 100 map taps per phone session for a better legibility of
the histogram. The frequency of taps in phone sessions with other app taps was
higher, on average 52.89 (SD 78.08), and with a maximum of 434 app taps. It is
noteworthy that the average number of other taps per phone session was
almost three times higher than the number of map taps. If we study the
cumulative distribution function (CDF) for the number of taps in map app
sessions compared to other app sessions, we notice some dierences between
the two groups, particularly for numbers of taps per phone session below 1000
(Figure 6).
Despite these dierences, the data for the two groups appeared to have
a similar distribution. A Kolmogorov-Smirnov test (D: 0.21; p = 0) showed that
the two samples did not come from a population with the same continuous
distribution. For the number of taps per phone session below 1000, the cumu-
lative probability for map taps was higher than for other, non-map taps.
We also plotted the CDF of the duration of map sessions against other app
sessions (Figure 7). Note that the right plot in Figure 7 depicts the same data as
in Figure 7 left panel, but on a log scale.
A Kolmogorov-Smirnov test (D: 0.162; p < 0.01) revealed that the two samples
were not based on the same distribution. In phone sessions up to 30 minutes,
the cumulative probability for a non-map phone session was higher than for
a map app phone session.
We also analysed the frequency and distribution of taps within map app
sessions to get a better understanding of dierent types of map app uses.
First, we grouped participants according to their session duration and the
frequency of taps within phone sessions to achieve a discrimination
Figure 6. The probability for low number of taps and map taps per phone session is significantly
different.
JOURNAL OF LOCATION BASED SERVICES 17

between heavy and light users. For our sample, the average phone session
duration was 5 minutes and 51 seconds. The average number of map taps
was 24.4. To exemplify these dierent user types, we next look at two
distinct participants.
The touch data of participant 13, for example, had an average phone
session duration of 2 minutes and 16 seconds, with on average 20.35 map
taps. At the other end, participant 87 had an average phone session
duration of 12 minutes, and on average 18.75 map taps. Figure 8 shows
the distribution of map taps within all app sessions. For participant 13
(Figure 8, left panel) there were many map taps at around 5% of the app
session, and the maximum number of map taps is reached between 15%
and 20%. The density then decreased, reaching the minimum at around
70%. The number of map taps increased again between 70% and 90%, and
after 90% of an app session, there were fewer map taps in an app session.
Figure 8 (right panel) shows a dierent distribution of map taps within app
sessions for participant 87. This participant had many map taps at the
beginning and towards the end of an app session.
Figure 9 shows the ten app sessions with most map taps of participant 13
(left) and participant 87 (right). Each row represents one app session from 0%
(rst tap on a map app) to 100% (last tap on a map app) and every stroke
represents a single map tap within the app session. The app session lengths are
thus normalised to an interval between 0% and 100% of the app session length.
For participant 13 (Figure 9, left panel) we can observe accumulations and
clustering of map taps in app sessions 1, 5, 6, 7, and 8, particularly at the
beginning of an app session, followed by fewer or no map taps, but then
again, a cluster of outbursts at the end. In app sessions 2, 3, 4, 9, and 10, the
map taps of participant 13 were very homogeneously distributed over the
whole app session length and there was almost no clustering of map taps.
App sessions 1 and 6 were special, since map taps were exclusively clustered in
Figure 7. CDF for the duration of phone sessions with map app taps and other app (left panel)
and in log scale (right panel).
18 T. REICHENBACHER ET AL.

the rst half of the session, while the second half of the app session showed no
map taps. There happened only one map tap for session 1, and three map taps
for session 6 again at the very end of the app sessions.
Figure 9 (right panel) shows a fairly homogenous distribution of the top ten
app sessions with the most map taps for participant 87. The map taps were
distributed in almost regular intervals over the map app session. Exceptions
were app sessions 4 and 6. App session 4 showed a concentration of map taps at
the beginning of the app session. Thereafter there was a short period of no map
taps followed by a small clustering of a few map taps after 20% of the app
session duration. Then, there were no map taps until the very end of the app
Figure 8. Kernel density estimation of map taps of all app sessions of participant 13 (left panel)
and participant 87 (right panel).
Figure 9. Distribution of map taps within the top 10 app sessions for participant 13 (left panel)
and participant 87 (right panel). App session lengths are normalised to an interval of 0 to 100%
of app session length and each tally mark represents a map tap within the respective app
session.
JOURNAL OF LOCATION BASED SERVICES 19

session. App session 6 showed almost the opposite pattern. There was a single
map tap at the beginning and then a gap until 80% of the app session length,
followed by a few map taps towards the end of the map app session.
4. Discussion
The goal of our long-term empirical research programme is to study how often,
how much and in what ways people use mobile maps in their everyday lives.
With the science of tappigraphy – the quantication of smartphone touchscreen
interactions – we aimed to capture of a continuous data stream of behavioural
human-map app patterns remotely, underlying daily life.
Our tappigraphy results based on map taps of 170 participants recorded
during their everyday activities were in line with reported, aggregated publicly
available app download statistics, and conrm, indeed, that map use of
observed participants was rather sparse compared to the use of other apps on
their smartphones, such as their social media consumption, chat, or gaming app
uses.
On the one hand, we nd that the proportion of our observed map taps over
all taps was on average less than 1%. We may interpret this as map apps to be
used very infrequently, and thus acknowledge the signicantly greater popu-
larity of non-map app use. On the other hand, one might also consider that
observed non-map apps, for example, relating to communication or gaming
might simply require far more taps during use. This can include typing for
sending a text message, commenting on a social media post, or controlling
the course of a game with keys or ngers. Given that Fonseca et al. (2021) report,
57.9% of their sample were non-map app users, and of the 42.1% map app
users, only 25.1% used the map apps daily our map app numbers probably also
support relatively sparse map use. This is in line with Böhmer et al. (2011), for
instance, who report an average smartphone use time of 59 minutes per day,
and average app use duration of Google Maps and Waze of 45 seconds, which
corresponds to a proportion of 1.27% of total use app use time, on average.
Infrequent map use statistics could also be a result of a narrow categorisation of
map apps, for example, only apps that contain the Android store app label
‘map’. Although travel support apps, real-estate apps, navigation support apps,
and tourism apps might heavily rely on maps, they might not be directly
labelled or categorised as a ‘map’ app. The more detailed labelling of map
apps monitored on participants’ smart phone will have to be considered for
future tappigraphy studies.
Aside from the overall small number of map taps observed, we did nd
a meaningful temporal distribution over the course of the day and the week.
For the daily map app use, we nd a map tap peak around 5 pm, thus at a time
of the day when most people end their work. It is not yet clear why the time of
the evening commute might be so dierent from the morning commuting
20 T. REICHENBACHER ET AL.

period when considering map taps. Possibly, the evening commute leads to
more potentially novel destination options. That is, other than the well-known
work-to-home route for users, perhaps allowing for less planned leisure activ-
ities into locales that might be less known or infrequently used. Interestingly,
general smartphone-based proxy measures of cognitive performance such as
tapping speed or app-locating speed also peaks at around 5 pm (Huber and
Ghosh 2021). This might suggest that it is also a good time for performing
spatial tasks with map apps.
While this general app use peak around 5 pm is also suggested by Böhmer
et al. (2011) and is also evident in our data, their study also found a peak of app
use in the category travel at 5 am, probably indicating the beginning of the
morning commuting period (Huber and Ghosh 2021). As mentioned earlier, it
would be useful in a future tappigraphy study to disentangle apps that are not
labelled as map apps, but potentially also include maps such as travel apps, etc.
to get deeper insights on map app use throughout the day.
For the weekly cycle, we observed a systematic, periodic pattern of map taps
mostly outside of the work week, suggesting map apps are predominantly used
for leisure activities, similarly to the 5 pm peak in the day analysis. Map taps peak
on Thursday evening, Friday afternoon and evening, and Sunday noon. These
kinds of leisure activities are likely to require more spatial planning and way-
nding than routine activities during the workday and weekend. As for the daily
frequency pattern, the morning commute did not show a high frequency of
map use, while the late afternoon and evening hours showed higher map use
frequencies. Perhaps people use dierent apps (e.g. communication, news,
games, social media) when commuting in the morning, compared to evening
leisure planning when they are leaving work.
We nd that phone sessions with map taps were on average more than 50%
longer than those containing other app taps. In contrast, Böhmer et al. (2011)
found that the average app use time in the travel category which often
included maps or a map interface to travel content, was 50% shorter than
the average app use duration. This dierence could be explained by a dierent
conceptualisation of what a use session is, and points to the diculty of
comparing results across published user studies. Böhmer et al. (2011), for
instance, dened a session as an app use session, that is, as the duration
between starting and closing one single app. In contrast, our phone session
concept, bound by unlocking and locking the smartphone screen, could
include the consecutive use of several apps, and therefore, overall, a phone
session may, by denition, take longer. Regarding the frequency of taps per
phone session then, we observed that sessions with no map taps at all
contained, on average, almost three times more taps than phone sessions
with map taps. We interpret this that phone sessions with higher numbers of
taps more likely fall into the category of social media apps, chat/communica-
tion apps, or gaming apps, as they usually include larger portions of typing for
JOURNAL OF LOCATION BASED SERVICES 21

text entries, which in turn comprise numerous short taps. Our ndings sug-
gested that map app sessions might take longer in time, but typical map app
tasks can be achieved with fewer taps overall such as leaving a map app
running for continuous self-localisation while navigating.
The distribution of map taps within app sessions (Figures 9 and 10) revealed
distinct tapping behaviours for two participants, suggesting dierent map app
use types. While participant 13 showed clusters of map taps over a session,
participant 87’s map tap distribution was quite homogenous with equal inter-
vals between taps. A possible interpretation of this distinct distribution could be
that participant 13 changed the extent and the scale of the map to search for
and identify locations on the map resulting in more clustered map taps. On the
other hand, the behaviour of participant 87 could reect another type of
common map app use, where a navigator regularly checks their current position
on the map display during waynding to ensure being on the right track during
navigation.
Although our empirical results were based on a relatively large sample, both
with respect to participant numbers and observation length, compared to similar
user studies in the wild (i.e. Banovic et al. 2014; Carrascal and Church 2015; Do,
Blom, and Gatica-Perez 2011), there are also limitations to our rst tappigraphy
study. First, the TapCounter app was available for the Android operating system
only. As Apple Maps had at the time of our study an estimated overall market
share of 10%, we might have missed out on a relevant share of potential users,
and possibly a relevant user pool for tappigraphy analyses on map apps. The data
source used here came from recruitment drives performed at university campuses
and was thus dominated by student populations. Dierent recruitment strategies
Figure 10. Kernel density estimation of taps from a participant in geographic space (left panel);
taps along a recorded GPS trajectory around home location (right panel).
22 T. REICHENBACHER ET AL.

could be considered in future to include a broader and more inclusive study
population. What is more, the data collection period from 2014 to 2018 might not
reect the latest developments in technology. Technological advancements, such
as larger screen sizes of smartphones, the roll-out of 5 G cellular networks,
improvement of Wi-Fi accessibility and the arrival of new map apps on the market
including voice-assisted map interfaces (e.g. Siri and Alexa) or AI technology could
possibly inuence map use behaviour in the future and serves as a motivating
source for future tappigraphy studies. Moreover, at this stage of the project we
had predominantly analysed recorded taps as events within a phone session (see
Figure 2). That is, the way we dened the phone sessions may not only have an
inuence on our reported results, but also limit comparison with other similar
studies found in the literature. One could also aggregate individual taps rst to
individual app sessions within a phone session, as shown in Figure 10. By introdu-
cing dierent data analysis granularities, dierent map use behaviour patterns
might emerge, which would allow the study of contextual embedding of map app
use in a sequence of other app uses within a phone session. It would be interest-
ing to further explore what kind of apps precede or follow map apps uses.
Another limitation of the tappigraphy method is that it will not distinctively
capture map use behaviour that includes no haptic interactions. For example, if
a user is using a mobile map for navigation, trac status information or self-
localisation by only gazing at the smartphone screen, it will likely not be detected
if tappigraphy is used as the only method. Finally, the next obvious step by
researchers interested in the spatio-temporal context of map app use would be
to capture and analyse taps in geographic space and over time to answer the key
geographic question: why there and why then? Answers to these questions would
support a deeper understanding of map use patterns in the context of human
mobility.
To move in this direction, we are in the process of leveraging a powerful
combination of tappigraphy with other data channels including map app use
location. In the context of a novel geo-tappigraphy approach, we are currently
collecting map use behaviour data with a modied version of the TapCounter
app that also records spatial data, including the smartphone’s geographic
location, and the acceleration parameters of monitored smartphones.
4
In
doing so, we wish to extend current, mostly temporal, analyses of map use
behaviour to spatially dependent map use behaviour. This will allow us to gain
further insights of where, when, and in what kinds of situations and environ-
mental contexts mobile maps are used, thus linking map taps directly to geo-
graphic space, and to mobility analysis. Figure 10 shows as an example the plot
of a participant’s taps in geographic space as a kernel density estimation (left
panel) and the taps around the home location along a movement trajectory
from GPS locations (right panel).
JOURNAL OF LOCATION BASED SERVICES 23

5. Summary and future work
Taking advantage of tappigraphy borrowed from cognitive neuroscience, we
set out to study mobile map app use in our everyday life. By using a tappigraphy
dataset gathered from a large Android user base, we unobtrusively observed
the frequency, timing, and mode of ambulatory map app use in-situ, with
participants' own smartphones and in participants’ everyday environment. We
identied clear dierences between the use of map apps compared to other
apps. Our results conrm a scarcity of map app use compared to all other apps,
as suggested by aggregated app download statistics and related prior research.
Going beyond aggregated download statistics, our approach allowed us to
uncover a distinct diurnal map app use pattern, that closely followed
people’s day-night rhythms. Also, we observed peaks of map use in the early
evening hours and at weekends, typically leisure time periods. Moreover, we
found that tapping behaviour during phone sessions with map apps was sig-
nicantly dierent from non-map app phone sessions, both with respect to their
duration and the frequency of taps within the sessions.
We wish to further extend the analysis of typical map use behaviour to
specic use contexts and purposes, for example, for planning leisure activities,
for spatial searches of environmental features, during navigation and waynd-
ing in familiar environments, or for exploring unfamiliar environments. We also
hope to recruit a broad mix of mobile map users in the future to study potential
group dierences (e.g. gender and age), and individual dierences (e.g. spatial
abilities and attitude) in everyday mobile map use. And nally, we intend to
align and combine the tappigraphy method with other, traditional user analysis
methods from HCI and cartography (e.g. interaction logging and usability
analysis) to capture a wider range of map use interactions, including non-
tapping interactions and voice interfaces such as Siri or Alexa.
Notes
1. https://www.statista.com/statistics/640039/market-share-mobile-operating-systems-
netherlands/.
2. https://www.postgresql.org.
3. https://jupyter.org.
4. https://www.geo.uzh.ch/microsite/mapontap.
Acknowledgements
Tumasch Reichenbacher would like to thank Jan Weber for the contribution of Figures 8 and
9. Arko Ghosh would like to acknowledge intramural funding from Leiden University and the
resources made available due to grants from Holcim Stiftung, Velux Stiftung (No. 1283), and
the Society in Science Branco Weiss Fellowship. Arko Ghosh would also like to acknowledge
the students from the Applied Cognitive Psychology Master’s programme and Myriam
24 T. REICHENBACHER ET AL.

Balerna for their help in data collection. Sara Fabrikant wishes to further acknowledge
generous funding by the European Research Council (ERC) Advanced Grant GeoViSense,
No. 740426.
Disclosure statement
Arko Ghosh is a co-founder of QuantActions Ltd, Lausanne, Switzerland. The company
focuses on converting smartphone taps to mental health indicators. For this research, soft-
ware and data collection services from QuantActions were used to monitor the smartphone
activity of participants.
Funding
The work was supported by the European Research Council (ERC) Advanced Grant [740426];
Society in Science Branco Weiss Fellowship; Holcim Stiftung zur Förderung der
Wissenschaftlichen Fortbildung; Velux Foundation [1283].
ORCID
Tumasch Reichenbacher http://orcid.org/0000-0002-7833-185X
Arko Ghosh http://orcid.org/0000-0002-2924-5146
Sara I. Fabrikant http://orcid.org/0000-0003-1263-8792
References
Balerna, M, and A Ghosh. 2018. “The Details of Past Actions on a Smartphone Touchscreen are
Reected by Intrinsic Sensorimotor Dynamics.” Npj Digital Medicine 1 (1): 4. doi:https://doi.
org/10.1038/s41746-017-0011-3
.
Banovic, N, C Brant, J Manko, and A K. Dey. 2014. “ProactiveTasks: The Short of Mobile Device
Use Sessions.” 10.1145/2628363.2628380
Böhmer, M, B Hecht, J Schöning, A Krüger, and G Bauer. 2011. “Falling Asleep with Angry
Birds, Facebook and Kindle: A Large Scale Study on Mobile Application Usage.” 10.1145/
2037373.2037383
Borger, J N., R Huber, and A Ghosh. 2019. “Capturing Sleep–wake Cycles by Using Day-To-Day
Smartphone Touchscreen Interactions.” Npj Digital Medicine 2 (1): 73. doi:10.1038/s41746-
019-0147-4.
Brügger, A, K-F. Richter, and S Fabrikant. 2016. “Walk and Learn: An Empirical Framework for
Assessing Spatial Knowledge Acquisition During Mobile Map Use.” International
Conference on GIScience Short Paper Proceedings. 10.21433/b3113hc8k3js
Brügger, A, K-F. Richter, and S I. Fabrikant. 2019. “How Does Navigation System Behavior
Inuence Human Behavior?” Cognitive Research: Principles and Implications 4 (1). doi:10.
1186/s41235-019-0156-5.
Carrascal, J P., and K Church. 2015. An in-Situ Study of Mobile App & Mobile Search Interactions.
doi:10.1145/2702123.2702486.
Church, K, M Cherubini, and N Oliver. 2014. “A Large-Scale Study of Daily Information Needs
Captured in situ.” ACM Transactions on Computer-Human Interaction 21 (2): 1–46. doi:
https://doi.org/10.1145/2552193 .
JOURNAL OF LOCATION BASED SERVICES 25

Colley, A, and J Häkkilä. 2014. “Exploring Finger Specic Touch Screen Interaction for Mobile
Phone User Interfaces.” Proceedings of the 26th Australian Computer-Human Interaction
Conference on Designing Futures: The Future of Design, 539–548. 10.1145/2686612.
2686699
Delikostidis, I, and C van Elzakker. 2011. “Usability Testing of a Prototype Mobile Navigation
Interface for Pedestrians.” ICC 2011: 25th International Cartographic Conference, Paris,
France.
Do, T M. T., J Blom, and D Gatica-Perez. 2011. Smartphone Usage in the Wild: A Large-Scale
Analysis of Applications and Context. doi:10.1145/2070481.2070550.
Duckrow, R B., E Ceolini, H P. Zaveri, C Brooks, and A Ghosh. 2021. “Articial Neural Network
Trained on Smartphone Behavior Can Trace Epileptiform Activity in Epilepsy.” IScience
24 (6): 102538. doi:10.1016/j.isci.2021.102538.
Fabrikant, S. I. 2022. Neuro-adaptive LBS: Towards human- and context-adaptive mobile
geographic information displays (mGIDs) to support spatial learning for pedestrian
navigation. In 17th International Conference on Location Based Services (LBS 2022), edited
by J. Krisp, L. Meng, Munich.
Fahrenberg, J, M Myrtek, K Pawlik, and M Perrez. 2007. “Ambulatory Assessment—monitoring
Behavior in Daily Life Settings.” European Journal of Psychological Assessment 23 (4):
206–213. doi:10.1027/1015-5759.23.4.206.
Falaki, H, R Mahajan, S Kandula, D Lymberopoulos, R Govindan, and D Estrin. 2010. “Diversity
in Smartphone Usage.” 10.1145/1814433.1814453
Fonseca, F, E Conticelli, G Papageorgiou, P Ribeiro, M Jabbari, S Tondelli, and R Ramos. 2021.
“Use and Perceptions of Pedestrian Navigation Apps: Findings from Bologna and Porto.”
ISPRS International Journal of Geo-Information 10 (7): 446. doi:10.3390/ijgi10070446.
Ghosh, A, J-P. Pster, and M Cook. 2017. “Optimised Information Gathering in Smartphone
Users.” ArXiv:1701.02796 [Physics.Soc-Ph]. https://arxiv.org/abs/1701.02796
Hegarty, M, and D A. Waller. 2005. “Individual Dierences in Spatial Abilities.” In The
Cambridge Handbook of Visuospatial Thinking, edited by P. Shah (edited by) & A. Miyake,
121–169. Cambridge University Press. doi:10.1017/CBO9780511610448.005.
Herman, L, and Z Stachoň. 2018. “Controlling 3D Geovisualizations on Touch Screen – the
Role of Users Age and Gestures Intuitiveness.” Conference: 7th International Conference on
Cartography and GIS, Bulgaria: Sozopol.
Huang, H, M Schmidt, and G Gartner. 2012. “Spatial Knowledge Acquisition with Mobile Maps,
Augmented Reality and Voice in the Context of GPS-Based Pedestrian Navigation: Results
from a Field Test.” Cartography and Geographic Information Science 39 (2): 107–116. doi:10.
1559/15230406392107.
Huang, H, G Gartner, J Krisp, M Raubal, and N Van de Weghe. 2018. “Location Based Services:
Ongoing Evolution and Research Agenda.” Journal of Location Based Services 12 (2): 63–93.
doi:10.1080/17489725.2018.1508763.
Huber, R, and A Ghosh. 2021. “Large Cognitive Fluctuations Surrounding Sleep in Daily
Living.” IScience 24 (3): 102159. doi:10.1016/j.isci.2021.102159.
Kimerling, A J., A R. Buckley, P C. Muehrcke, and J O. Muehrcke. 2016. Map Use: Reading,
Analysis, Interpretation. 8th ed. Redlands, CA: Esri Press.
Kobayashi, M, A Hiyama, T Miura, C Asakawa, M Hirose, and T Ifukube. ”Elderly User Evaluation
of Mobile Touchscreen Interactions.” In Human-Computer Interaction – INTERACT 2011,
edited by P. Campos, N. Graham, J. Jorge, N. Nunes, P. Palanque and M. Winckler
P. Campos, N. Graham, J. Jorge, N. Nunes, P. Palanque and M. Winckler 83–99. Berlin
Heidelberg: Springer. 2011.
MacEachren, A M. 2004. How Maps Work Representation, Visualization, and Design. New York,
NY: Guilford Press.
26 T. REICHENBACHER ET AL.

Muehlenhaus, I. 2013. Web Cartography: Map Design for Interactive and Mobile Devices. Boca
Raton, FL: CRC Press.
Parry, D A., B I. Davidson, C J. R. Sewall, J T. Fisher, H Mieczkowski, and D S. Quintana. 2021.
“A Systematic Review and Meta-Analysis of Discrepancies Between Logged and
Self-Reported Digital Media Use.” Nature Human Behaviour 5 (11): 1535–1547. doi:https://
doi.org/10.1038/s41562-021-01117-5
.
Ricker, B, and R Roth. 2018. “Mobile Maps and Responsive Design.” Geographic Information
Science & Technology Body of Knowledge 2018 2018 (Q2). doi:10.22224/gistbok/2018.2.5.
Riegelsberger, J, and Y Nakhimovsky. 2008. Seeing the Bigger Picture: A Multi-Method Field Trial
of Google Maps for Mobile. doi:10.1145/1358628.1358655.
Savino, G-L., M Sturdee, S Rundé, C Lohmeier, B Hecht, C Prandi, N J. Nunes, and J Schöning.
2021. “MapRecorder: Analysing Real-World Usage of Mobile Map Applications.” Behaviour
& Information Technology 40 (7): 646–662. doi:10.1080/0144929X.2020.1714733.
Shiman, S, A A. Stone, and M R. Huord. 2008. “Ecological Momentary Assessment.” Annual
Review of Clinical Psychology 4 (1): 1–32. doi:10.1146/annurev.clinpsy.3.022806.091415.
Sugimoto, M, T Kusumi, N Nagata, and T Ishikawa. 2021. “Online Mobile Map Eect: How
Smartphone Map Use Impairs Spatial Memory.” Spatial Cognition & Computation 1–23.
doi:10.1080/13875868.2021.1969401.
Thrash, T, S Lanini-Maggi, S I. Fabrikant, S Bertel, A Brügger, S Credé, C T. Do, et al. 2019. “The
Future of Geographic Information Displays from Giscience, Cartographic, and Cognitive
Science Perspectives (Vision Paper).” 142, 19:1-19:11. 10.4230/LIPIcs.COSIT.2019.19
UKOM. 2019. UK Digital Market Overview. London, UK: UKOM.
JOURNAL OF LOCATION BASED SERVICES 27