You are on page 1of 10

Participatory Design of Touch Gestures for Informational

Search on a Tablet Device


Tsele Rakubutu Helene Gelderblom Jason Cohen
University of the Witwatersrand University of Pretoria University of the Witwatersrand
Dept of Information Systems Department of Informatics Dept of Information Systems
Johannesburg, South Africa Pretoria Johannesburg, South Africa
27 (0)11 717-8164 27 (0)12 4203352 27 (0)11 717-8164
TseleRakubutu@gmail.com helene.gelderblom@up.ac.za Jason.Cohen@wits.ac.za

ABSTRACT as MS Windows [6].


This study set out to answer the question: what would a gesture Leveraging off pointer-based interactions in designing gesture sets
set for conducting an informational search on a multi-touch tablet could be justified [9]. The key motivation for doing so is that
web browser look like if designed in collaboration with users with users would be able to apply their existing knowledge of pointer-
limited touch screen experience? In addressing this question, we based interactions when confronted with touch screen interaction.
developed such a user-defined gesture set, based on gestures An opposing view is that moving away from the classic WIMP
elicited from participants with little or no experience with touch elements (windows, icons, menus, pointer) would improve the
screen devices. Focusing on 24 specific search-related browser quality of touch screen interactions [4]. According to this latter
functions, 20 participants were asked, in a lab setting, to view, multi-touch devices can enable exceptional freedom in
experiment with and then suggest a gesture for each of the gestural input while the use of pointer-based, single finger, one-
functions. Data collection included video and audio recordings, point tap or drag interactions, analogous to mouse clicks and
think-aloud data, informal interviews and self-evaluation of drags, negates the benefits that multi-touch technology offers.
proposed gestures. A combination of qualitative and quantitative
analysis revealed the following: one-handed gestures are preferred These conflicting views reflect an uncertainty about users’
over two-handed gestures; users who have mostly been exposed to preference between pointer-based interactions in touch screen
mouse interaction prefer gestures that correspond to point-and- applications and novel interactions that are divorced from
click actions; completion of a task should be possible with more traditional point-and-click interactions. The answer to the question
than one gesture; and complex tasks should allow varying of which gestures would be most intuitive to users in different
combinations of gestures). application contexts may not have a simple answer.
An intuitive gesture is one where the gesture afforded by the
Categories and Subject Descriptors interface design is aligned to how users expect to provide input
H.5.2 [Information interfaces and presentation]: [4]. Various factors, including the levels of previous exposure to
User interfaces – interaction styles, user-centered design. different interaction mechanisms, might influence what users
regard as intuitive. Therefore, understanding users and their
General Terms expectations is critical in designing appropriate touch screen
Design, Experimentation, Human Factors. gesture sets.
Given the particular interest of this study in web searching on
Keywords multi-touch tablets, we asked the following question: What would
Multi-touch, Tablet, Web browsing, Participatory design, Gesture an intuitive user-defined gesture set for conducting an
design, Referents. informational search on a multi-touch tablet web browser look
like? In addressing this question, we developed a user-defined
1. INTRODUCTION gesture set based on gestures elicited from participants with little
Many touch screen applications have their own specific gesture or no experience with touch screen devices.
sets that are used to interact with digital objects and provide input Next, we briefly discuss the related research. We then describe the
[4]. Such a gesture set is used throughout the application or mechanisms used to engage individual users in the process of
interface and is also called a gesture vocabulary [8]. Many gesture reasoning, reflecting and articulating to construct the proposed
sets use varying combinations of one finger tap interactions, gesture set. We describe our data collection and analysis methods
which suggests that they have simply been adapted from the and present our findings in the form of a gesture set together with
pointer-based paradigm used in popular operating systems, such general guidelines for the design of intuitive gestures.

Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are 2. RELATED WORK
not made or distributed for profit or commercial advantage and that Previous research on gesture elicitation has predominantly
copies bear this notice and the full citation on the first page. Copyrights focused on tabletop devices. Some of the key studies are discussed
for components of this work owned by others than ACM must be honored. case-by-case below.
Abstracting with credit is permitted. To copy otherwise, or republish, to
post on servers or to redistribute to lists, requires prior specific Nielsen, Störring, Moeslund and Granum [8] put forward a
permission and/or a fee. Request permissions from: procedure for developing an intuitive and ergonomic gesture set.
Permissions@acm.org.

SAICSIT2014, September 29 - October 01 2014, Centurion, South Africa


Copyright 2014 ACM 978-1-4503-3246-0/14/09…$15.00
http://dx.doi.org/10.1145/2664591.2664594 n

276
This procedure guides participants through the the process of Kane, Wobbrock, and Ladner [3] conducted two user studies that
creating the gesture set, as well as testing it The study then compared how blind people and sighted people use touch screen
provides an example of how to apply the procedure through an gestures. Their first study showed that blind people have different
investigation of participants’ interactions with virtual 3D shapes. gesture preferences than sighted people and in a subsequent study
The method used in our study was informed by this procedure. found significant differences in the speed, size, and shape of
gestures performed by blind people versus those performed by
Wobbrock, Aung, Rothrock and Myers [10] analysed participants’ sighted people.
unistroke gestures (these are gestures where the participant does
not lift his or her finger until the gesture is complete) on a It is worth noting that in the research surveyed above, only
touchpad to indicate letters of the alphabet. They found that their Mauney et al. [5] and Kane et al. [3] addressed handheld
approach for developing gestures resulted in gestures that first- touchscreen devices, with the rest focusing on tabletop devices. It
time users were able to guess with considerable accuracy. Their should be further noted that the application contexts that have
procedure also informed our study. been investigated have not specifically addressed the use of search
engines or web browsers for conducting any form of web search –
Epps, Lichman and Wu [1] investigated gestures used for general despite this being a common activity. These gaps in literature
operating system (Windows) tasks and identified users’ provided justification for our study.
preferences for performing common Windows tasks using
gestures. They found, for example, that users prefer to use their
index finger and illustrated the need to reduce or remove the 3. DATA COLLECTION
Window-centric elements in screens presented to participants Users should play a central, participative role in gesture design,
during experiments. because gestures designed autonomously by system designers are
not always reflective of what end-users would prefer [7, 11].
Frisch, Heydekorn and Dachselt [2] created a user-defined gesture There are various ways in which users may participate in gesture
set (comprising of one-handed and two-handed gestures, as well design. The approach employed in this study is based on the prior
as those involving hands and pen together) for editing node-link work of Nielsen et al. [8], Wobbrock et al. [10], Wobbrock et al.
diagrams on multi-touch, pen-enabled tabletops. They share their [11], Frisch et al. [2] and Micire et al. [6] who elicited gestures
procedure for creating this gesture set, which involved the use of from participants through a series of experimental tasks. The
think-aloud data and requesting participants to provide ratings, approach taken is summarised below.
using Likert scales, for each gesture they propose.
Wobbrock, Morris and Wilson [11] studied users’ interactions
3.1 Referents for Relevant Functions
The first step in developing an intuitive gesture set for
with 2D shapes (with no specific application context). They
informational web searching was to identify the functions for
present an approach to designing gestures that relies on eliciting
which gestures would be needed [2]. An analysis of 80 web
gestures from non-technical users through the use of referents.
browsers, comprising of both tablet and desktop web browsers,
Their findings indicated, among other things, that users rarely care
yielded the 24 functions which were used in the study, and for
about the number of fingers they use when gesturing, that one
which referents were created. These referents are short video clips
hand is preferred to two, and that desktop computing strongly
(created using Mozilla Firefox, Windows Internet Explorer and
influences users’ mental models. This study greatly influenced
Google Chrome) illustrating the effects of the function i.e. the
ours in terms of methodology, hypotheses and data analysis.
output or effect that would be displayed when the function is
Micire, Desai, Courtemanche, Tsui and Yanco [6] conducted an activated. The 24 functions are listed in Table 1.
experiment to determine the gestures that people would naturally
use for controlling robot teams. They presented guidelines for 3.2 Participants
designing gesture sets, which we adopted in our experiments. Twenty participants, who were mainly university staff, served as
They also provided a taxonomy of user-defined gestures, which volunteers. Half were academics appointed at different levels
provides the following classifications for gestures: Selection (from junior lecturers to full professors). The remainder consisted
gestures (used for selecting individual or multiple robots), of research assistants, administrative staff and a lab manager. One
Position gestures (indicate desired change in location of the robot participant was a business analyst. They represented a variety of
or object), Rotation gestures (express the desired rotation of disciplines: computer science, information systems, linguistics,
robots, objects or map), Viewpoint gestures (used to change the sociology and business management. The inclusion criteria that
participant’s view by either moving the virtual world or changing were used to recruit participants were:
the position of the camera) and User Interface Elements (which  The required application domain knowledge, i.e. familiarity
include commonly used interface elements like buttons, menus, with conducting an informational search on a desktop or
virtual keyboards and handwriting recognition). laptop web browser using the functionality detailed above in
Table 1.
Mauney, Howarth, Wirtanen and Capra [5] analysed gestures for
 Little or no experience with any form of touch screen device,
28 actions commonly done on handheld devices across
including touch screen phones or tablets.
participants in nine different countries in order to identify cultural
Participants’ ages ranged from 27 to 65, 6 participants were male
similarities and differences in the use of gestures. Their findings
and 14 were female, and 17 were right-handed and 3 were left-
influenced the hypotheses of our study e.g. they found that
handed. The average age was 40. Eight had never used a touch
experience with gesture-enabled devices influenced the gestures
screen device before, while 12 had very limited exposure to touch
that participants created and that participants still relied on menus
screen interaction.
when gesturing.

277
Table 1. Web browsing functions that support an During gesture elicitation, the researcher sat next to the participant
informational search and remained present throughout the session (see Figure 1). All
No Function No Function explanations and conversations that happened between the
researcher and the participant were recorded.
1 Go to previous 13 Paste section of text
webpage into search engine
2 Go to next 14 Open link on the
webpage webpage in new tab
3 Input search 15 Open a new browsing
query tab
4 Stop loading 16 Close a browsing tab
webpage
content
5 Refresh / reload 17 View record of
webpage visited pages
content
6 Pan down the 18 Delete entry from
webpage record of visited
webpages Figure 1 Experiment setting
7 Pan up the 19 Save current 3.4 Gesture Elicitation
webpage webpage address The referents were presented to participants on an iPad and the
within the browser participants were requested to propose gestures that they felt
for viewing at a later matched the referents – each one in turn. To ensure complete
stage understanding of the referents, the facilitator explained each
(“bookmarking”) referent to the participants prior to showing it to them. He
carefully explained the event(s) in the referent while showing it to
8 Pan across the 20 Call up saved them for the first time. If required, he showed it to them several
webpage webpage address times, addressing any questions the participants had, until they
9 Make webpage 21 Save image on indicated that they understood it clearly. The video referents of
content larger webpage to memory the web browser ensured that minimal elements relating to
(“zoom in”) Windows or the Macintosh were displayed; this was to avoid bias
[11].
10 Make webpage 22 Save current
content smaller webpage to memory When the participants were ready to offer a gesture or gestures, a
thin glass sheet was placed over the iPad. The referent video was
(“zoom out”)
then played on the iPad, and the participants performed the
11 Search for text 23 Request help gesture(s) for the event(s) in the video, as if interacting with the
on the webpage iPad. So, although the users’ gestures did not actually cause the
changes on the screen (because of the glass sheet that covered the
12 Copy a section 24 Close web browser iPad screen), the set-up very closely mimicked such real
of text on the interaction. The users performed the gestures on the iPad as if
webpage they were controlling the changes on the screen. Wobbrock et al.
[11] explain that with such an approach, the lack of feedback is
beneficial as it creates a “monologue” where all participants’ input
3.3 Experiment Setting is always acceptable, and therefore enables participants to exhibit
The experiments were conducted in a HCI laboratory equipped their natural, unrevised and uninhibited behaviour.
with two built-in, high-resolution, overhead cameras. In addition Participants were allowed to use one or more fingers, as well as
to these cameras, two standalone video cameras were positioned at one or both hands, any part or shape of their finger(s) and/or
the table where the participants’ gestures were elicited. All four hand(s) in any way they wanted, as long as they expressed the
cameras recorded the participants unobtrusively and were gesture which they felt would best matched the function shown in
positioned to optimally capture finger, hand and arm movements the referent [6]. Participants were requested to think aloud while
from different angles [2, 6, 8, 11]. providing their gestures [2, 6, 11]. These think-aloud data
A separate microphone was used to record the audio think-aloud provided verbal explanations and reasoning with regard to
data during the experiments. Noldus and Camtasia Studio participants’ gestures.
software were used to combine and synchronise the different As with Wobbrock et al. [10], if participants were unsatisfied with
video recordings and the audio data into a single video. the gesture they had proposed, they were allowed to redo it, but
once they had proceeded to the next gesture they were not

278
permitted to go back and change it. Where participants felt that an system provides feedback) or continuous (requires on-going
additional widget, such as a menu or special button, was required recognition by the system and feedback is provided while the
on the interface for a particular activity, they were allowed to gestures is being performed).
assume it was present – provided they explained in detail what the
Following this classification, the gestures were assigned to their
widget was, the form they expected it to take and how they
respective functions. This process is explained in the subsections
expected it to work [6, 11].
below.
3.5 Gesture Ratings
In addition to the gestures and think-aloud data, participants were 4.2 Assignment of Gestures to Functions
also requested to provide ratings, using Likert scales, for each Assignment of gestures to functions was based on agreement
gesture they proposed. This was done immediately after between the different participants; that is, the gesture that had
performing that gesture [2, 11]. These gesture ratings represented been proposed the most number of times by different participants
the participants’ perception of the quality of their gestures in for a specific referent would be assigned to that referent [10, 11].
terms of suitability and ease of performance and were elicited It is important to note that the other gestures that were proposed
using two Likert scales. for that referent, i.e. those that were proposed less frequently,
were not discarded, but were also assigned to the referent
The first Likert scale, which elicited ratings for the suitability of provided they did not conflict with other gestures [2, 4, 6, 11].
the proposed gesture for the referent displayed, read, “The gesture Three types of conflicts existed; conflicts within the participants’
I picked is a good match for its intended purpose” [11]. The own gesture sets, internal conflicts and external conflicts. These
second scale, which elicited ratings for the ease with which the are explained below.
proposed gesture could be performed, read, “The gesture I picked
is easy to perform” [11]. Five point Likert scales were employed, 4.2.1 Conflicts within Participants Gesture Sets
both of which elicited ordinal responses from 1 = strongly As with Wobbrock et al. [10] participants were required to resolve
disagree to 5 = strongly agree. Obtaining the subjective ratings is conflicts among their own gesture sets, i.e. participants were
important as functions with low gesture ratings are functions that instructed not to propose the same gesture for different referents
users would likely struggle with – either in terms of their mapping where the interactive contexts was the same. For example, a
(reflected through the “gesture goodness” rating) or in terms of participant would not be permitted to use a single finger drag from
how easy it is, physically, to perform them (reflected through the the left edge of the screen towards the middle of the screen for the
“gesture ease” rating), or both. “Go to next page” referent and use the same gesture for the “Go to
the previous page” referent as the interactive contexts for these
4. DATA ANALYSIS referents are identical. The participant would, however, be
permitted to use that gesture when highlighting an entry from a
4.1 Classification of Gestures menu list that appears on the left edge of the screen.
The video data was evaluated in order to extract the gestures that
the participants used in their interactions. This was done by 4.2.2 Internal and External Conflicts
extracting the gestures that each participant performed for each of An internal conflict occurred when a participant’s gesture for a
the referents. These gestures were then classified using the certain referent conflicted with another participant’s gesture for
Taxonomy of Surface Gestures [11], in corroboration with that same referent. External conflicts, on the other hand, occurred
categorisations used by [8] and [4]. The gestures were classified when different participants used the same gesture, within identical
according to 4 dimensions: interactive contexts, for different referents.
 Gesture Form: this relates to the fingers or parts of hand To resolve any one of these types of conflicts between two
used, whether the gesture was static (the finger(s) or hand is gestures, the gesture that had been proposed the most number of
in a fixed position over time) or dynamic (the finger(s) or times by different participants for that referent was assigned to
hand change positions over time) and whether it was a simple that referent and the one that was in the minority was excluded
(gesture comprises a single finger or hand action) or complex from the final gesture set. While the conflicting gesture was
gesture (gesture comprises two or more gestures, performed excluded, other gestures that did not have conflicts were still
sequentially or in parallel). considered for inclusion in the final gesture set for that referent
 Gesture Nature: this refers to whether the gesture is physical even if only proposed by a minority of participants. This was to
(physically acts on objects), metaphorical (acts on, with, or allow for alternatives in performing that function.
like something else), symbolic (visually depicts a symbol) or
arbitrary (arbitrarily mapped to referent).
4.2.3 Agreement Scores
Once all the conflicts had been resolved, the agreement scores
 Gesture Binding: this refers to whether the gesture is object- were calculated. As previously mentioned, assignment of gestures
centric (performed with respect to a specific object), world- to functions was based on agreement between the different
dependent (takes the real world context into consideration), participants; agreement scores are single numbers that serve as a
world-independent (can generally be performed anywhere on means of quantifying to what extent there was agreement between
the interface) or has mixed dependencies (is world- participants for a gesture [11].
independent in one respect but world-dependent or object-
The computation of an agreement score was done by taking all
centric in another).
gestures that had been proposed for a specific referent, and based
 Gesture Flow: this refers to whether the gesture is discrete on their classifications, grouping identical gestures together.
(performed in its entirety and once completely performed the These groups, which represent the frequency with which the

279
different gestures were proposed by different people, were then ratings (based on the scale of “The gesture I picked is a good
used to compute the agreement score A using the following match for its intended purpose”) and “gesture ease” ratings (based
formula [11]: on the scale of “The gesture I picked is easy to perform”) were
calculated [11]. These data provided crucial complementary data
to the actual elicited gestures and provided an opportunity to
evaluate the extent to which participants adjudged their gestures to
be suitable for the referents they had proposed them for, as well as
how easy they considered them to be to perform.
The gestures with low gesture ratings are functions that users
would likely struggle with – either in terms of their suitability for
where r is a referent in the set of all referents R, Pr is the set of the function they were proposed for (reflected through the
proposed gestures for referent r, and Pi is a subset of identical “gesture goodness” rating) or in terms of how easy they are,
gestures from Pr. Using the formula to calculate agreement scores physically, to perform (reflected through the “gesture ease”
is important because the greater the agreement between different rating), or both. No gestures had to be removed from the gesture
participants in the gestures used for a referent, the more likely it is set as a result of low ratings. The lowest rating across all
that other novice users would also agree or subscribe to using participants for both Likert scales was 3, with the averages for the
those gestures. Those functions with low agreement scores, are the “gesture goodness” and “gesture ease” ratings being 4.494 and
functions that users would likely struggle with their gestures. 4.527 respectively.

4.3 Think-Aloud Data 5. FINDINGS


The physical form of a gesture, i.e. the gesture as it is physically
performed, may not always reveal the nature of the gesture. The 5.1 The Gesture Set
think-aloud data, which was participants’ narrations, explaining A coherent gesture set for conducting an informational search on a
their thought processes for the gestures they were proposing while multi-touch tablet web browser, based on gestures elicited from
they were simulating the gestures, were used in revealing the participants, was successfully created. Table 2 presents the 24
nature of these gestures. It assisted in revealing the participants’ search functions and the gesture(s) found to be most suitable for
mental models within the context of the referent, assisting us in those functions. It should be noted that the gestures presented
the classification of the gesture as previously described. below are only a subset of the total gesture set. Note that the
buttons that appear on the interfaces in the images below The full
4.4 Analysing the Gesture Ratings set can be requested from the authors.
Based on the subjective Likert scale ratings provided by each
participant for each gesture proposed, average “gesture goodness”

Table 2. Web browsing functions and their gestures

Function Gesture(s) Function Gesture(s)

1. Go to 13. Paste
previous section of text
webpage into search
engine

The user taps on the “Back” button.

2. Go to next 14. Open link


webpage on the
webpage in
new tab

The user taps on the “Forward” button. The user drags the link to the “New Tab
Button”.

280
3. Input search 15. Open a
query new browsing
tab

The user calls up a keyboard by tapping


once on the Google textbox. The user taps once on the “New Tab
Button”.

4. Stop 16. Close a


loading browsing tab
webpage
content

The user taps once on a “Stop” button. The user taps once on the “Close Tab”
cross.
5. Refresh / 17. View
reload record of
webpage visited pages
content

The user taps once on a “Refresh / The user taps once on a “History Button”
Reload” button.

6. Pan down 18. Delete


the webpage entry from
record of
visited
webpages

The user drags their finger downwards The user by taps once on the item, and
along the right edge of the screen. taps once on a “Delete Button”.

281
7. Pan up the 19. Save
webpage current
webpage
address within
the browser
for viewing at
a later stage

The user taps once on the “Favourites


The user drags their finger upwards
Button” to call up a list of webpages that
along the right edge of the screen.
have been saved and taps once on the
“Add Button”.
8. Pan across 20. Call up
the webpage saved
webpage
address

The user taps once on the “Favourites


The user drags their finger sideways Button” to call up a list of webpages that
(left to right) along the bottom edge of have been saved and taps once on the
the screen. page to be opened.

9. Make 21. Save


webpage image on
content larger webpage to
(“zoom in”) memory

The user draws a box around the image


The user makes the webpage content by dragging the finger in a rectangle
larger by dragging his / her thumb and around the image, and taps once on the
index finger apart. “Save Button”.

10. Make 22. Save


webpage current
content webpage to
smaller memory
(“zoom out”)

The user makes the webpage content The user calls up the main menu by
smaller by dragging his / her thumb tapping on a menu button, and taps once
and index finger together. on the “Save webpage” option.

282
11. Search for 23. Request
text on the help
webpage

The user calls up a textbox by tapping The user taps once on a “Help” button.
on a “search button”.
12. Copy a 24. Close web
section of text browser
on the
webpage

The user taps once on a “Close Button”


The user presses with a single finger located at the top right corner of the
where the text begins, and drags the screen.
finger over the text.

This gesture set revealed the following:  Novel gestures for tablet web browsing that do not
correspond to any of the ways in which desktop web
Co-existence of One-Handed and Two-Handed Gestures
browsing is performed, may not be intuitive to adult users
It was found that users prefer one-handed gestures to two-handed who had mostly been exposed to mouse based interaction.
gestures. Of the 478 gestures that were recorded, 459 (96.025%)
 Should novel gestures be developed for tablet web browsing,
were one-handed, and 19 (3.975%) were two-handed. This
alternative gestures that are desktop computing adaptations,
implies that:
including those that require menu access, should be made
 The use of two-handed gestures should be limited. available to users.

 If two-handed gestures are developed for a specific function, A key observation is that users drew upon their experiences of
an alternative one-handed gesture should be made available. desktop computing despite the exclusion of graphical elements
relating to Windows or Macintosh operating systems (these
It is also worth noting that during the experiments, participants included menus, title bars and buttons such as Back and Refresh).
did not actually hold the iPad in their hands – it was placed on a For example, 14 out of 20 participants tapped on an imaginary
table, facing participants at an incline. During normal use a tablet “X” in the top right-hand corner of the screen for closing the web
would be held in one hand, while the other hand gestures. Under browsers despite the fact that none of the referents had the
those circumstances it would be difficult to use two-handed Minimise, Maximise and Close buttons typically found on the top
gestures. Our participants had both hands available for gesturing right corner of windows.
and still opted to use one-handed gestures.
Another observation was that participants generally dragged their
The only notable exceptions were the referents for resizing (i.e. fingers from the top to the bottom to scroll down, and from the
zooming in and out). bottom to the top to scroll up. On most tablets, the opposite is true
Inclusion of Desktop Computing Elements i.e. users drag their fingers from the top to the bottom to scroll up
and from the bottom to the top to scroll down. The fact that the
We found that participants relied heavily on desktop computing as Windows operating system generally employs arrow keys or
a point of reference for their gestures. The single finger tap, scroll bars explains why it is that most participants (13 out of 20)
analogous to the mouse click, was the most used gesture and in this study dragged or flicked downwards on the screen to scroll
accounted for 50.727% of all the gestures used. The single finger down the webpage, and vice versa.
drag, analogous to the mouse drags, was second in line, and
accounted for 24.128% of all gestures performed. Furthermore, The complexity introduced by adopting a desktop computing
the majority of participants incorporated the use of menus (main adaptation in this situation, and allowing users to drag downwards
menus and context menus) into their gestures. The following in order to scroll down, means that users that have become
implications arise: accustomed to dragging/flick upwards to scroll down will find
this method of scrolling unintuitive. One recommendation in this

283
situation would be to allow users to decide which outcomes search on a multi-touch tablet web browser, it is not inconceivable
should be produced when dragging downwards or upwards by that designers may employ some of the gestures presented in this
allowing users to configure what function should be mapped to study in other application contexts. Functions such as “searching
which gesture for text” or “request help” are not limited to web browsing
application context.
A major disadvantage of desktop computing adaptations is that
they may not make the best use of multi-touch technology, In addition to the user-defined gesture set, this study provided
especially given the predominant use of single finger taps. The various insights with respect to gesture design for multi-touch
advantage of this approach, however, is that users can move tablet browsing. These enhance the understanding or appreciation
between their desktops, which they are familiar with, and their of what users, specifically adult users who have had limited or no
tablet devices without having to learn or memorise new ways with prior experience with touch screen devices, consider to be
which to interact with the tablet web browser. intuitive. This understanding will enable designers to design
gestures that are aligned to the naïve or natural expectations of
The use of desktop computing adaptations should, however, not their users, and would therefore result in satisfactory or enjoyable
overshadow the fact that several participants produced novel experiences for these users – potentially attracting more users to
gestures, such as dragging a finger repeatedly over an item to their applications. The contributions of this study will therefore
delete it, as if “rubbing it out” (Referent 18), dragging a browsing prove beneficial to those designing or developing multi-touch
tab “off the screen” to close it (Referent 16) and tracing a tablet interfaces.
question mark on the screen to activate the “Help” function
(Referent 23), to mention a few. These were not excluded from There are various implications for future research. An obvious
the gestures set, but incorporated to co-exist with gestures that are next step would be to conduct a usability study on the user-
desktop computing adaptations. defined gesture set, using completely different participant groups,
to determine the extent to which the gestures presented in this
5.2 Gesture Re-use study are usable in practice.
It was found that participants will attempt using the same gesture It is recommended that similar studies be conducted for tablet
to accomplish different tasks (in different interactive contexts) devices, but focusing on other application contexts, such as word
and re-use the same gestures in different combinations based on processing, note taking, viewing videos, playing music and
their judgment of how tasks should be decomposed. This gaming. With tablet devices becoming more popular within the
reinforces the existing knowledge that users differ in how they go public domain, it becomes increasingly important to ensure that
about completing various tasks. For designers and developers it usability requirements for the various activities that tablet devices
means that gesture re-use is essential and should be maximised, support, are well understood by those in practice.
while obviously ensuring that no conflicts occur. Our data
revealed five gestures that should receive special consideration in
terms of re-use. They are: the one finger tap, the one finger drag,
8. REFERENCES
the one finger flick, the one finger press and the one finger [1] Epps, J., Lichman, S., and Wu, M. (2006). A Study of Hand
double-tap. These accounted for 86.065% of all gestures provided Shape Use in Tabletop Gesture Interaction. In Proceedings of
by participant across all referents. the 2006 Annual Conference on Human Factors in
Computing Systems – CHI ’06, Montreal, Canada, 22 – 27
April.
6. LIMITATIONS [2] Frisch, M., Heydekorn, J., and Dachselt, R. (2009).
The depth of our data collection and consequently the amount of
Investigating Multi-touch and Pen Gestures for Diagram
time spent with each participant (and also the time required for
Editing on Interactive Surfaces. In Proceedings of the ACM
analysis of each participant’s data) restricted the number of
International Conference on Interactive Tabletops and
participants to 20. The sample size proved large enough to inform
Surfaces – ITS ‘09, Calgary, AB, Canada, 23 – 25
an intuitive gesture set, but we could not analyse the gestures in
November.
relation to the demographics of the participants, e.g. identifying
whether there were any patterns in the gestures provided by male [3] Kane, S. K., Wobbrock, J. O., and Ladner, R. E. (2011).
participants in comparison to female participants, or between Usable Gestures for Blind People: Understanding Preference
gestures proposed by left-handed versus right-handed people. and Performance. In Proceedings of the 2011 Annual
Conference on Human Factors in Computing Systems – CHI
The study addressed 24 specific functions. This means that while ’11, Vancouver, Canada, 7 – 12 May.
the user-defined gesture set addressed the key functions for
conducting an information search, it does not cover the full [4] Lao, S., Heng, X., Zhang, G., Ling, Y., and Wang, P. (2009).
complement of web browsing functions. A Gestural Interaction Design Model for Multi-touch
Displays. In Proceedings of the 23rd British HCI Group
7. CONCLUSIONS AND FUTURE WORK Annual Conference on People and Computers: Celebrating
The study has achieved its aim of creating a user-defined gesture People and Technology – BCS-HCI ’09, Cambridge, United
set for conducting an informational search on a multi-touch tablet Kingdom, 1 – 5 September.
web browser based on gestures elicited from participants. This [5] Mauney, D., Howarth, J., Wirtanen, A., and Capra. M.
user-defined gesture set may be used as presented here by (2010). Cultural Similarities and Differences in User-
designers or developers in new versions of web browsing software Defined Gestures for Touchscreen User Interfaces. In
for multi-touch tablet devices. Designers may also decide only to Proceedings of the 28th ACM CHI Conference on Human
use certain gestures, or to modify the proposed gestures as they Factors in Computing Systems, Atlanta, USA, 10 – 15 April.
deem fit for their applications. Furthermore, while these gestures
have been created specifically for conducting an informational

284
[6] Micire, M., Desai, M., Courtemanche, A., Tsui K. M., and [9] Wang, F., Cao, X., Ren, X., and Irani, P. (2009). Detecting
Yanco, H. A. (2009). Analysis of Natural Gestures for and Leveraging Finger Orientation for Interaction with
Controlling Robot Teams on Multi-touch Tabletop Surfaces. Direct-Touch Surfaces. In Proceedings of the 22nd annual
In Proceedings of the ACM International Conference on ACM symposium on User Interface Software and
Interactive Tabletops and Surfaces – ITS ‘09, Calgary, AB, Technology – UIST ’09, British Columbia, Canada, 4–7
Canada, 23 – 25 November. October.
[7] Morris, M. R., Wobbrock, J. O., and Wilson, A. D. (2010). [10] Wobbrock, J.O., Aung, H.H., Rothrock, B., and Myers, B.A.
Understanding Users' Preferences for Surface Gestures. In (2005). Maximizing the guessability of symbolic input. . In
Proceedings of Graphics Interface 2010 – GI '10, Toronto, Proceedings of the 2005 Annual Conference on Human
Ontario, Canada, 31 May – 2 June. Factors in Computing Systems–CHI '05, Portland, Oregon,
[8] Nielsen, M., Störring, M., Moeslund, T.B., and Granum, E. 2–7 April.
(2003). A procedure for developing intuitive and ergonomic [11] Wobbrock, J. O., Morris, M. R., and Wilson, A. D. (2009).
gesture interfaces for HCI. In Proceedings of the 5th User-defined gestures for surface computing. In Proceedings
International Gesture Workshop 2003, Genova, Italy, 15–17 of the 27th international conference on human factors in
April. computing systems – CHI 2009, Boston, Massachusetts,
USA, 4 – 9 April.

285

You might also like