Conference PaperPDF Available

Towards Language Independent Detection of Dyslexia with a Web-based Game

Authors:
Conference Paper

Towards Language Independent Detection of Dyslexia with a Web-based Game

Abstract and Figures

Detecting dyslexia is important because early intervention is key to avoid the negative effects of dyslexia such as school failure. Most of the current approaches to detect dyslexia require expensive personnel (i.e. psychologists) or special hardware (i.e. eye trackers or MRI machines). Also, most of the methods can only be used when children are learning how to read but not before, necessarily delaying needed early intervention. In this work, we present a study with 178 participants speaking different languages (Spanish, German, English, and Catalan) with and without dyslexia using a web-based game built with musical and visual elements that are language independent. The study reveals eighth game measures with significant differences for Spanish children with and without dyslexia, which could be used in future work as a basis for language independent detection. A web- based application like this could have a major impact on children all over the world by easily screening them and suggest the help they need.
Content may be subject to copyright.
Towards Language Independent Detection of Dyslexia
with a Web-based Game
Maria Rauschenberger
WSSC
Universitat Pompeu Fabra
maria.rauschenberger@upf.edu
Ricardo Baeza-Yates
WSSC
Universitat Pompeu Fabra
rbaeza@acm.org
ABSTRACT
Detecting dyslexia is important because early intervention is
key to avoid the negative effects of dyslexia such as school
failure. Most of the current approaches to detect dyslexia
require expensive personnel (i.e. psychologists) or special
hardware (i.e. eye trackers or MRI machines). Also, most
of the methods can only be used when children are learning
how to read but not before, necessarily delaying needed early
intervention. In this work, we present a study with 178
participants speaking different languages (Spanish, German,
English, and Catalan) with and without dyslexia using a
web-based game built with musical and visual elements that
are language independent. The study reveals eighth game
measures with significant differences for Spanish children
with and without dyslexia, which could be used in future
work as a basis for language independent detection. A web-
based application like this could have a major impact on
children all over the world by easily screening them and
suggest the help they need.
CCS Concepts
Human-centered computing Empirical studies in
accessibility; Accessibility design and evaluation meth-
ods; Software and its engineering
Interactive games;
Keywords
Dyslexia; Detection; Pre-Readers; Serious Games; Web-based
Assessment; Universal Screening; Language-Independent;Visual;
Musical; Information Processing; Gamification
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 not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page. Copyrights for components of this work owned by others than the
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specific permission
and/or a fee. Request permissions from permissions@acm.org.
W4A ’18, April 23–25, 2018, Lyon, France
© 2018 Copyright held by the owner/author(s). Publication rights licensed to ACM.
ISBN 978-1-4503-5651-0/18/04. .. $15.00
DOI: https://doi.org/10.1145/3192714.3192816
Luz Rello
HCI Institute
Carnegie Mellon University
luzrello@cs.cmu.edu
Jeffrey P. Bigham
HCI and LT Institutes
Carnegie Mellon University
jbigham@cs.cmu.edu
1. INTRODUCTION
The American Psychiatric Organization defines dyslexia as
a specific learning disorder which affects around 5% to 15% of
the world population [1]. Those affected by dyslexia usually
have difficulties in reading and writing, independently from
the mother tongue. Dyslexia does not affect the general intel-
ligence of a person. Hence, people with dyslexia understand
the meaning of the words but do not always know how to spell
or read a word. Often this results in bad grades at school and
frustration for students and parents over many years (40% to
60% of children with dyslexia have psychological difficulties
[32]). These are common indicators for detecting a person
with dyslexia. Other indicators for detecting dyslexia relate
to linguistic skills, e.g., prosodic or phonological awareness
[9], differences in reading and spelling error rates of people
with and without dyslexia [3, 31], and differences in game
measures derived from phonological awareness, letter recog-
nition or word recognition [28]. These and other language
related indicators have been used in various software for
screening, prediction or detection of dyslexia (we use this
three terms indistinguishably). Examples of software for
detecting dyslexia in English are
Lexercise Screener
[15],
Nessy
[19] and
Dytective
for both English [29] and Spanish
[28].
All these reading and spelling applications are language
dependent. This means on one hand that the content of
the application needs to be adapted for every new language
which is time and resource consuming. On the other hand,
only people who already have language acquisition can be
tested.
Children with dyslexia can learn the spelling of words or
decode words for reading. But they need more time, as well
as special and intense treatment. For example, two years
instead of one for learning how to spell phonetically accurate
words [32]. Hence, to give children with dyslexia more time
to practice, avoid frustration and the possibility to succeed,
early detection is needed.
Detecting dyslexia in children before they learn to read
and write is difficult because the indicators above all use
manifestations of reading and writing. This means that
children can be detected only after they begin to learn to
read and write. This puts students with dyslexia behind.
Therefore, new ways of detecting the risk of having dyslexia
are needed for pre-readers. Prior studies show approaches to
predict future language acquisition of pre-readers e.g., from
Figure 1: Participants playing the visual part (left)
and the musical part (right) of the Game MusVis.
Photos included with the adults’ permission.
newborn with brain recordings [16], from infants with rapid
auditory cues [2], and from kindergarten children with the
perception of visual-spatial attention [5].
Previous research has related speech perception difficulties
to auditory processing, phonological awareness and literacy
skills [30, 34]. Phonological deficits of dyslexia have been
linked to basic auditory processing [10]. The auditory percep-
tion of children with dyslexia has been proven to be related
to the sound structure [12] as well as to the auditory working
memory [17]. None of these require reading ability, and may
be useful in detecting dyslexia.
Related research suggests that reading impairments are
due to the visual-spatial attention and poor coding instead
of phonological difficulties [36]. Apart from that, visual
discrimination and search efficiency are used as predictors for
future reading acquisitions [5]. This prior work motivated us
to design our game content with musical and visual elements
to create a language independent environment to analyze
the differences in the game measures between children with
and without dyslexia. For the content design of the musical
and visual elements prior knowledge of language acquisition,
phonological awareness, letter naming or letter recognition
is not needed.
To create the musical elements, we used acoustic param-
eters in the musical part of our game MusVis. To create
the visual elements, we designed different visual represen-
tations similar to visual features of annotated error words
from people with dyslexia [23, 27] and designed the game as
a simple search task which does not require language acqui-
sition. Additionally, the participants need to store chunks
of information in their short-term memory for both parts of
the game.
Next, we present the first results of the game measures col-
lected from children with and without dyslexia while playing
the game MusVis (see Figure 1) as well as the game content.
With a pre-study like this where participants already diag-
nosed with dyslexia are participating, we reduce the time
to find the indicators and increase the chances of making a
promising approach before smaller children participate in a
long-term study. In this study, we do find game measure-
ments which we can use as indicators to distinguish readers
with and without dyslexia after playing our game.
2. RELATED WORK
Dyslexia is a specific reading disorder, which is probably
caused by the phonological skills deficiencies associated with
phonological coding deficits [35]. A person with dyslexia
has visual and auditory difficulties that cause problems in
reading and writing. It does not affect how intelligent a
person with dyslexia is [1]. We focus in related work on
digital approaches (games) to predict, screen or detect pre-
readers or content unrelated to the knowledge of phonological
awareness or letter naming as well as on the perception of
sound and visual cues.
Audenaeren et al. present a tool called DIESEL-X [7]
which intends to predict the possibility of a child having
dyslexia. The tool includes three mini-games to measure
dyslexia related indicators (e.g., ‘letter knowledge, FM detec-
tion, end-phoneme recognition’ [7]). An example task is that
the child is asked to find a certain letter in the game. The
focus of the development was the gameplay and motivation
for a player. To the best of our knowledge, the validation of
the prediction model is not published yet.
Another screening computer-based game is AGTB 5–12
for children at the age of five to twelve [33]. The game has
twelve tasks in total and every task takes seven minutes
(total game duration is 87 minutes). The tasks focus on
the phonological working memory processing, the central
working memory, and the visual-spatial working memory.
In a similar vein, the Bielefelder Screening provides nine
tasks for children at the last year of kindergarten [33]. In
20 to 25 minutes the children do different types of tasks
on phonological perception, phonological working memory
processing, long-term memory, and visual attention. The
results are the categorization of risk groups for dyslexia. The
published accuracy of the prediction process has not been
found.
Gaggi et al. [6] publish their preliminary results with a
sample size of 24 participants (last year of kindergarten)
using six different visual and/or auditory games. The game
performance was compared to detect a child but the games
were originally developed to train different skills of a child
with dyslexia. There is no information about the total dura-
tion of the game.
2.1 Sound Perception and Dyslexia
Different theories and empirical results motivated us to
use modifications of acoustical parameters as game content.
For example, the rapid auditory processing deficit hypothesis
assumes individuals with dyslexia have problems processing
short auditory cues. Another theory, claims the dynamic
change of the acoustical parameters cause the difficulties [10].
Since the phonological grammar of music [22] is similar to
the prosodic structure of language, music i.e., a combination
of acoustical parameters, can be used to imitate these features.
Studies showed a significant difference in the perception of
readers with dyslexia on the syllable stress compared to the
control group at the age of 9 [8].
For example, the rise time of a sound could imitate stress
levels on syllables. Additionally, findings suggest a relation
between rise time perception and the prosodic and phonolog-
ical development [12].
Even newborns respond automatically to the complex
task of perceiving music [37] and show differences in the
perception of sensitivity to native versus non-native rhythmic
stress [9] by the age of 5 months or to the phonemic length
by 6 months [16].
Because of the similarities of music and language different
acoustic parameters of sound have been explored and proven
significance in the perception of children at the age of 8
to 13 years with and without dyslexia [12] e.g., rise time,
short duration (100ms), intensity, and rhythm. Also, the
perception of pitch and its patterns relates to reading
skills which are one difficulty people with dyslexia have [30,
37].Furthermore, a lab study showed different behavior of
infants (m = 7.5 months) using complex sound frequencies
to predict child’s language skills at the age of three [2].
A recent study found evidence that dyslexia-associated
genes are related to the encoding of sounds in the auditory
brainstem [18].However, there are musicians with dyslexia
which scored better on auditory perception tests than the
general population [17]. At the same time, these participants
score worse on tests of auditory working memory, i.e., the
ability to keep a sound in mind for seconds. This observation
is in line with the results on perceptions for short duration
sounds [12] and the findings on the prosodic similarity
effects of participants with dyslexia [9]. One connection
between the difficulties in the perception of language and
music seems to be the short-term memory and the recall
of information chunks [9]. Since people with dyslexia have
short-term memory difficulties [14, 21] questions like “Which
sound did you hear first” or “Which sound is pitched higher?
would determine the groups [12].
Huss et al. [12] already showed that significant performance
differences can be found using musical metrical structure
between people with and without dyslexia between the age
of 8 and 13 years old in a controlled setting.
2.2 Visual Perception and Dyslexia
Apart from the auditory perception results already men-
tioned, previous research suggests that the cause of reading
impairments could be partly due to the visuo-spatial (also
called visual-spatial [5]) attention and poor visual coding
instead of the auditory difficulties [36]. These would mean
that the difficulties people with dyslexia have in reading
and writing are due to a poor decoding of visual cues, e.g.,
letter recognition, especially, for error cases where a person
has a good phonological awareness but difficulties in reading
non-words.
Further, findings also provide evidence that the cause of
dyslexia might be due to a more basic cross-modal letter-to-
speech sound integration deficit and the pre-reading visual
parietal-attention [5]. They are able to predict reading acqui-
sition in preschoolers with the visual-spatial attention. An
example of a visual-spatial attention task is a search task
(searching for symbols) which shows significant differences in
the error rate for poor readers in first grade [5].
The analysis of error words from children with dyslexia
shows that the wrong and correct letters in errors words are
visually similar as well as through different languages, e.g.,
English, Spanish [27] or German [23]. The annotated error
and correct letters show similarities in different visual features
called mirror letter (e.g., < n > < u >) or fuzzy letter (e.g.,
< s > and < z >). The letters have also similarities in
the vertical (e.g., < m >) and horizontal symmetries (e.g.,
< e >) through the visual features [27].
However, to predict pre-readers which have no knowledge
of language acquisition, phonological awareness or letter
recognition is still a challenge. Our game MusVis aims to
distinguish between readers with and without dyslexia with
the derived measurements of our game. This is the first step
towards a prediction of dyslexia with measurements derived
from a game for readers and pre-readers.
The related work focused on using one evidence, e.g., local
visual search on one game. We are combining findings from
previous literature, which are known to cause troubles for
children with dyslexia as explained before, to create a game
environment to find solid differences for predicting dyslexia
in the future. At the same time, the game should be fun and
not too difficult. We expect people with dyslexia to make
more mistakes and take more time than the control group.
This is a language independent approach which has the
potential of detecting pre-readers by just translating the
instructions of the game because it is independent from
knowledge of, e.g., word recognition, letter recognition or
phonological awareness.
3. GAME DESIGN
The game MusVis aims to measure differences in how
children with and without dyslexia react on musical and
visual cues. For that reason, we designed a musical (see
Figure 2) and a visual part (see Figure 4) of the game MusVis
with features extracted from the literature. Because of the
different perceptions involved (sound and visual), the game
design of each part is different but both games trigger the
short-term memory.
As is well known, children have more difficulties to pay
attention over a longer period of time. Therefore, the two
games have four stages each and eight rounds that need less
than 10 minutes to play. We used game mechanics, e.g.,
rewards (points, score), feedback (instant feedback, progress
bar, visible status) or challenges (time limit) and game com-
ponents, e.g., story for the game design. The content design,
user-interface, interaction and implementation for the musi-
cal and visual part of the game are described in the following
subsections.
3.1 Content Design for Musical Elements
The musical part is adapted from the already existing
visual game Memory.
1
This game was chosen because it is
a well-known child game and could be easily transformed
to musical elements. The musical elements were already
evaluated with a five user study to discover usability problems
that could influence the prediction approach [25].
The participant is asked to find all similar musical elements
instead of the same pictures and no time pressure is given. To
avoid a random match in the first two clicks, the participants
always listen to the first two different sounds of the round.
The last two cards and clicks are always correct because these
are the remaining cards of the memory game. The musical
element is played once when clicked.
This part has four stages which are counter-balanced with
Latin Squares [4]. Each stage is assigned to one acoustic
parameter of sound, i.e., frequency, length, rise time, rhythm
and three musical elements are created, for example, with
different rhythms. Each stage has two rounds with, first two,
and then three, musical elements that must be matched. The
sounds arrangement for every round are in random order.
1
An example of a visual memory game can be found on
https://goo.gl/vhWmYs.
Features General
Complex vs.
simplex Pitch Sound
duration Rise
time Rhythm Shot-term
memory PSE* CAPS**
Literature
Overy [21] x x x
Huss et al.[12] x x x x
Goswami et al.[9] x x x x x
Yuskaitis et al. [37] x x
Johnson [14] x
Stage
frequency x x x x x x
length x x x x
rise time x x x x x x
rhythm x x x x x x
Table 1: Mapping of the evidence from literature to distinguish a person of dyslexia, the features and
general assumptions and the stages of the musical part of the game MusVis. *phonological similarity effect;
**correlation acoustic parameters speech
Figure 2: Example of the musical part from the
game MusVis for the first two clicks on two sound
cards (left) and then a pair of equal sounds is found
(right). The participant is asked to find two equal
musical elements by clicking on sound cards in a row.
The musical elements are generated with a simple sinus
tone using the free software Audacity. The exact parameters
of each musical element are described in the preliminary
usability study [25] and the Musical Elements are available
at GitHub [24]. The musical elements generated for the
game MusVis are designed with the knowledge of previous
literature. We present the mapping of the literature in Table
1 which provide evidence to distinguish a person of dyslexia
to our designed stages for the musical part of the game
MusVis and the following is a short summary.
Same for all musical elements:
Each acoustic stage
has three musical elements (we use MP3 for sound files).
Only one acoustic parameter is changing within a stage.
Stage frequency:
The frequencies used are in the audi-
tory perception range of a person starting from 440 Hz. We
combine the simple tone with a relatively short duration of
0.350s. Each musical element of this stage differs by 0.25 of
a semitone (440 Hz to 452.8929 Hz to 446.3998 Hz). We use
for the first round of two sound pairs the 440 Hz and 446
Figure 3: Waveform for the order of intervals for one
musical element of the stage Rise Time. The exam-
ple starts with a 0.025s fade-in interval and then a
0.250s interval followed by a 0.250s fade-in interval.
Hz musical elements.
Stage length:
Each musical element of this stage has a
different duration (0.350s, 0.437s, 0.525s), i.e., tone length.
The differences between the length of each musical element
follow the suggested short duration (100msec) from Huss et
al. [12]. We use for the first round of two sound pairs the
0.350s and 0.525s musical elements.
Stage rise time:
At this stage, each musical element is
designed with either a short fade in of 0.025s or a fade in of
0.250s or a fade out of 0.250s. We use for the first round of
two sound pairs the 0.025s and 0.250s fade ins.
Stage rhythm:
At this stage, the musical elements are
designed with two intervals of rise time equal to 0.250s fade in
and one interval equal to 0.025s fade in, in a different order.
The order of fade in for each musical element is changed
according to the limit of possibilities (see example in Figure
3). We always use for the first round the two sound pairs
with the order of rise time interval 0.025s, 0.250s, 0.250s
and musical element with the rise time order reversed.
To keep the game duration short, we only include very
promising and easy to deploy acoustic parameters. Param-
eters like intensity, even though, showed significant differ-
ences in controlled environment studies could not easily be
controlled in our online remote study, as different personal
computers and headphones might produce different volume
levels due to hardware or software diversity.
3.2 Content Design for Visual Elements
The visual part of the game is similar to the interaction
of Whac-A-Mole. We adapted the interaction design and
content for this purpose, as shown in Figure 4. For the visual
game, we design elements that have the potential of making
Figure 4: Example of the visual part of the game
MusVis with the priming of the target element sym-
bol (left) and then the nine-squared design including
the distractors for each symbol (right).
more elements with similar features and represent horizontal
and vertical symmetries which are known to be difficult for
a person with dyslexia [23, 27, 36].
At the beginning, participants see the target visual element
(see Figure 4, left) for three seconds. They are asked to
remember this visual element. After that, the participants
are presented with a setting where the target visual element
and distractors are displayed (see Figure 4). Within 15
seconds the participants try to click as often on the target
visual element as possible. The arrangement of target and
distractor elements is randomly changed after every click.
The visual part has 4 stages which are counter-balanced
with Latin Squares [4]. Each stage is assigned to one visual
type (
symbol, z, rectangle, face
) and four visual elements
for each stage are presented. One visual element is the target
which the participants need to find and click (see Figure
5, top). The other three visual elements are distractors for
the participants. Each stage has two rounds (in total the
number of rounds is 8) with first a 4-squared and then a
9-squared design (see Figure 4, right). The target and all
three distractors are displayed in the 4-squared design. In
the 9-squared design, the target is displayed twice as well as
the distractor two and three. Only distractor one is displayed
three times. The stages from Figure 5 are summarized next.
Stage symbol:
This stage uses two lines connected in
an angle of less than 30
as the target visual element and
creates a vertical symmetry. The distractor one is mirrored
while the distractor two and three are rotated by 90
and
-90 .
Stage z:
The target visual element for this stage is created
with two lines parallel to each other connected with a diagonal
line. The diagonal line is drawn from the top right line end
to the down left line end. This creates vertical and horizontal
symmetry of the visual element. This representation looks
very similar to the letter z but we do not use the phonological
awareness of the letter, i.e. the participants do not need to
know that this is also an existing letter of the Latin alphabet.
The distractor one is mirrored while the distractor two and
three are rotated by 90 and -90 .
Stage rectangle:
This stage is the shape of a square and
a right-angled triangle. These shapes have by design vertical
and horizontal symmetries which we use to create a complex
Figure 5: Overview of the designed visual elements.
The figure shows the target element (top) and dis-
tractor elements (below) for the four different stages
(z, symbol, rectangle, face) of the visual part of the
game MusVis.
target. The outline shape is the square and two triangles
are placed within the square. The 90
corner of one triangle
is placed in the up-right corner of the square and the other
triangle in the below-left corner of the square. This creates
a visual element with different ways to perceive similarities
within the element. The distractors are rotated by 90
, 180
and 270 .
Stage face:
The target visual element has three visual
cues combined (two symmetric dots placed horizontally and
an outline around them). The outline is first a straight
horizontal line under the two dots and connects the ending
with a bow around the two dots. The whole target element
is symmetrical on the vertical line. The target is rotated
180
for the first and third distractor. Additionally, the two
dots are slightly staggered up and down for the second and
third distractor.
3.3 User-Interface and Implementation
To support the readability for parents and supervisors
we used a large font size (minimum 18 points) [26]. The
interactive elements (cards to be clicked within the game)
are large enough to be clicked easily. The presentation of
interactive elements (sound cards/squares) are the same
within each game and do not differ in color or shape to avoid
differences in the perception.
Both games are implemented as a web-application using
JavaScript, jQuery, CSS, HTML5 and a backend with a PHP
server and a MySQL database. One reason for this is access
simplicity for remote online-studies. Another reason is the
advantage of adapting the application for different devices
in future research studies.
4. EXPERIMENTAL STUDY
We conducted an independent within-subject design study
with 178 participants and included only participants which
are either diagnosed with dyslexia (n = 67) or without
dyslexia (n = 111). Every participant played all the game
rounds in their mother language with the same game content.
Only the study and game instructions (audio and text) are
translated into the mother tongue. We recruited Spanish par-
ticipants diagnosed with dyslexia mostly over public social
media calls from the non-profit organization ChangeDyslexia
(https://changedyslexia.org/). German participants diag-
nosed with dyslexia have been mostly recruited over social
media calls in support groups. The control group was re-
cruited with the collaboration of two Spanish schools and
two German schools.
4.1 Procedure
The communication with the participants was mostly via
email. The web-application was played at home or in the
school with one researcher (authors of the paper) present or
always available through digital communication.
First, the parents or supervisor filled out the demographic
questionnaire for gaining background information, e.g., age
of the participant, dyslexia diagnosis (yes/no/maybe) and
the mother tongue.
This was followed by explaining instructions for the user
study to the parents or supervisor, e.g., turn up the volume,
use headphones, play without interruptions or explain and
help your child only with the instructions of the games.
Then a short video story for the musical part was played.
After that every participant played first the musical and
then the visual part of MusVis (see Figure 1). Finally,
the parents answered two feedback questions and left their
contact details.
Each input method (computer vs. tablet) needs to be
analyzed separately. We decided to use a laptop or desktop
computer for two reasons: (1) From prior game evaluation [6,
28] we know that readers are able to interact with the device
and (2) these devices are still more available than tablets [13].
4.2 Participants
The analysis data includes only the data from participants
who played all 16 rounds of the game MusVis with a com-
puter. Dropouts happened mostly because participants used
a different browser (e.g., Internet Explorer instead of Google
Chrome) or a different device (tablet instead of a computer).
To have a more accurate analysis we made sure we know
the status of a participant (diagnosed or not) and excluded
participants which reported they might have dyslexia. Only
participants that showed no indication of dyslexia or with
an official diagnosis, by a medical doctor or equivalent, were
included. Thirteen participants were suspected of dyslexia
and therefore, taken out of the analysis.
We report separately the results for the Spanish partic-
ipants (n = 108), German participants (n = 57) and an
analysis with all languages for the language independent
variables where we added English (n = 6), and Catalan (n =
7). For the dependent variables (DV) which show indications
of the same tendency of results and are therefore considered
as language independent .
For the analysis with all languages (n = 178), we considered
for the dyslexia group, 67 participants which were diagnosed
of dyslexia (33 female, 34 male). Their ages ranged from 7
to 12 years (m = 9.8, sd = 1.4). For the control group, we
considered 111 participants (67 female, 44 male). Their ages
ranged from 7 to 12 years (m = 10.5, sd = 1.5).
For the Spanish participants, we considered 41 participants
diagnosed of dyslexia (23 female, 18 male). Their ages ranged
from 7 to 12 years (m = 9.5, sd = 1.1). For the control group,
we took into account 67 participants (42 female, 25 male).
Their ages ranged from 7 to 12 years (m = 10.0, sd = 1.2).
For the German participants, we considered 17 participants
diagnosed with dyslexia (5 female, 12 male). Their ages
ranged from 7 to 12 years (m = 10.7, sd = 1.4). For the
control group, we had 40 participants (21 female, 19 male).
Their ages ranged from 7 to 12 years (m = 11.4, sd = 1.4).
4.3 Dependent Measures
The dependent variables we collected from the user interac-
tion with the web-based game MusVis were for both parts of
the game time intervals of clicks and total number of clicks.
For the musical part, we additionally collected the duration
of each round and average click time (we calculated the
average click time by dividing the duration with the total
number of clicks). Because of the gameplay, the first three
cards need less rethinking to find where the sound is the same.
Therefore, we consider for the musical part the first 4 clicks
as one interval, 4th click interval, to measure the duration.
Every click interval after that can be used as well and we
choose the 6th click interval, which even exists when the
participant finishes the game in the shortest click sequence
possible.
For the visual part we collected time to the first click,
number of hits or correct answers, number of misses or non-
correct answers, efficiency (we calculated the efficiency by
dividing the time of the last click by hits) and accuracy (we
calculated the accuracy as hits divided by the total number
of clicks).
5. RESULTS
In order to find out whether we have new indicators to pre-
dict people with dyslexia after playing MusVis, we analyzed
the dependent variables for our independent within-sub ject
study for the three groups: Spanish, German, all languages.
We applied first the Shapiro-Wilk test. All variables (n =
54) were not normally distributed and we applied, therefore,
the independent Wilcoxon Test. All analyses were conducted
with a Bonferroni correction to avoid type I errors (2.4e-3).
We present the results for Spanish in Table 2, for German in
Table 3, and for all languages in Table 4.
The DVs are categorized for Spanish and German accord-
ing to the tendency that participants with dyslexia compared
to the control group had within each language (see Table 5).
An example of a language independent variable is the DV hits
because the dyslexia group has in German and Spanish signif-
icantly less correct clicks (Spanish 5.7; German 5.6) than the
control group (Spanish 6.6; German 6.3). The DV duration is
an example of the opposite trend because the dyslexia group
for Spanish takes significantly more time while the German
participants with dyslexia take less time compared to their
language control group. Only if the tendency was similar,
the DV were included in the overview of all languages (Table
4). We consider the variables in Table 5 as a first step to
provide evidence towards a language independent detection.
We use the effect size to estimate the likely size of the
effect in the population. The effect size (r) for the Wilcoxon
Test [4] is calculated as
z
r =
N
where z is the z-score and N is the number of observations.
We use the effect size in Tables 2 and 3, only for the
significant results. First, we report the results for the musical
part and then for the visual part of the game.
Total number of clicks (music)
is not language inde-
pendent. The tendency of results is opposite between partici-
Dependent variables
Spanish Control
mean sd Dyslexia
mean sd Wilcoxon
W p-value z effect size
Musical
Total clicks 11.0 5.5 11.3 6.0 86231 0.63 -0.48 0.05
4th click interval 1.6s 0.7s 2.0s 1.3s 63658 7e-12 -6.80 0.66
6th click interval 1.5s 0.8s 1.7s 1.2s 76762 2e-3 -3.13 0.30
Duration 27.5s 17.1s 34.3s 27.0s 72316 1e-5 -4.38 0.42
Average click time 2.5s 0.8s 3.0s 1.2s 59028 2e-16 -8.11 0.78
Visual
Total clicks 8.0 3.3 6.7 2.7 110000 3e-10 6.25 0.60
Time to first click 2.3s 1.4s 2.7s 1.8s 75566 5e-4 -3.47 0.33
Hits 6.6 2.9 5.7 3.0 105670 5e-7 5.02 0.48
Misses 1.3 3.1 1.0 1.8 86340 0.62 -0.50 0.05
Accuracy 0.60 0.50 0.57 0.50 90432 0.43 0.83 0.08
Efficiency 2.8s 2.6s 3.1s 2.8s 73301 4e-5 -4.10 0.39
Table 2: Overview of all reported dependent variables for the musical and visual part of the game MusVis
for Spanish only (n = 108).
Dependent variables
German Control
mean sd Dyslexia
mean sd Wilcoxon
W p-value z effect size
Musical
Total clicks 10.8 5.4 10.6 4.4 20880 0.49 -0.70 0.09
4th click interval 1.8s 0.8s 2.0s 1.2s 19218 0.05 -2.00 0.26
6th click interval 1.7s 0.7s 1.6s 0.7s 21580 0.89 -0.14 0.02
Duration 28.5s 16.9s 27.9s 13.0s 20542 0.34 -0.95 0.13
Average click time 2.6s 0.8s 2.6s 0.5 19.708 0.11 -1.59 0.21
Visual
Total clicks 7.2 3.1 6.8 2.7 23887 0.10 1.67 0.22
Time to first click 2.4s 1.5s 2.5s 1.1s 19314 0.06 -1.90 0.25
Hits 6.3 2.8 5.6 2.6 24675 0.02 2.28 0.30
Misses 0.9 2.1 1.2 2.3 20718 0.36 -0.92 0.12
Accuracy 0.60 0.49 0.59 0.49 22084 0.83 0.30 0.04
Efficiency 2.8s 2.3s 3.2s 2.9s 19357 0.06 -1.87 0.25
Table 3: Overview of all results reported dependent variables for the musical and visual part of the game
MusVis for German only (n = 57).
pants with dyslexia (Spanish m = 11.3 & German m = 10.6)
compared to participants without dyslexia (Spanish m = 11.0
& German m = 10.8). This means that German participants
with dyslexia click less compared to the German control
group and Spanish participants with dyslexia click more com-
pared to the Spanish control group. The total number of
clicks did not reveal significant differences on total clicks for
Spanish (W = 86231, p = 0.63, r = 0.05) or German (W =
20880, p = 0.49, r = 0.09). The effect size for Spanish and
German is nearly zero, so is considered as it has no effect [4].
Click time interval (music)
is not language indepen-
dent over all click intervals and we, therefore, do not report
any click intervals for all languages. Hence, participants with
dyslexia (Spanish 4th click interval m = 2.0s & German 4th
click interval m = 2.0s & Spanish 6th click interval m = 1.7s)
take more time before they make the next click than the
control group (Spanish 4th click interval m = 1.6s & German
4th click interval m = 1.8s & 6th click interval m = 1.5s).
But German participants with dyslexia (6th click interval
m = 1.6s) take less time before they make the next click
than the German control group (m = 1.7s). The 4th time
interval (W = 63658, p = 7e-12, r = 0.66) as well as the
6th click interval (W = 76762, p = 2e 3, r = 0.30) is
significant for Spanish but not for German (W = 21580, p =
0.89, r = 0.02). The effect size for 4th click time interval
Spanish is considered as large where the effect size for 6th
click time interval for Spanish and the 4th click time interval
for German is considered as medium [4]. We report only the
fourth and sixth click interval for the musical game since
the first three intervals do not show, as expected due to the
game design, any significant differences between the groups.
Duration (music)
is not language independent. Hence,
Spanish participants with dyslexia (m = 34.3s) take more
time to find all pairs and finish the round than the Spanish
control group (m = 27.5s). But German participants with
dyslexia (m = 27.9s) take less time before they find all pairs
than the German control group (m = 28.5s). The duration is
significant for Spanish (W = 72316, p = 1e 5, r = 0.42) but
not for German (W = 20542, p = 0.34, r = 0.13). The effect
size for Spanish is considered as medium and for German as
small [4].
Average click time (music)
is not language indepen-
dent. Because Spanish participants with dyslexia (m = 3.0s)
take in average more time for a click than the Spanish con-
trol group (m = 2.5s). But German participants with and
without dyslexia take in average the same time (m = 2.6s).
Spanish participants with dyslexia significantly spend more
time for each click (W = 59028, p = 2e-16), r = 0.78) where
Dependent variables Control Dyslexia Wilcoxon
Visual mean sd mean sd W p-value
Total clicks 7.6 3.2 6.8 2.7 276120 3e-7
Time to first click 2.4s 1.5s 2.6s 1.6s 210850 3e-4
Hits 6.5 2.9 5.8 2.9 272180 4e-6
Accuracy 0.60 0.49 0.58 0.49 240780 0.66
Efficiency 2.8s 2.5s 3.1s 2.7s 209740 2e-4
Table 4: Overview of all language independent results for the visual part of the game (n = 178).
we cannot measure a difference for German (W = 19708, p =
0.11), r = 0.21). The effect size for Spanish is considered as
large and for German as small [4].
Total number of clicks (visual)
is language indepen-
dent. Participants with dyslexia (m = 6.7) significantly
clicked less times than participants without dyslexia (m =
8.0) for Spanish (W = 110000, p = 3e-10, r = 0.60). The
effect size for Spanish is considered as large [4]. The German
participants have the same trend for the control group (m =
7.2) compared with the group of participants with dyslexia
(m = 6.8, W = 23887, p = 0.10, r = 0.22). The effect size
for German is considered as small [4]. Because the trend is
the same, we provide the analysis for the total number of
clicks which confirm the significant difference (W = 276120,
p = 3e 7).
Time to the first click (visual)
is language independent.
This means that participants with dyslexia (Spanish m =
2.6s) and German (m = 2.5s) take more time before they
make the first click than the control group (Spanish m =
2.3s & German m = 2.4s). The time to the first click is
significant for Spanish (W = 89450, p = 1e-3, r = 0.30) but
not for German (W = 19314, p = 0.06, r = 0.25). The effect
size for Spanish and German is considered as medium [4].
Because the trend is the same for both languages even though
it is only significant for Spanish, we provide the analysis for
the time to the first click with a significant difference for all
languages (W = 210850, p = 3e 4).
Hits
is language independent. Hence, participants with
dyslexia (Spanish m = 5.7s) & German (m = 5.6s) have less
hits than the control group (Spanish m = 6.6s & German
m = 6.3s). The hits is significant for Spanish (W = 105670,
p = 5e-7, r = 0.48) and for German (W = 24675, p = 0.02,
r = 0.30). The effect size for Spanish and for German is
considered medium [4]. Because the trend is the same for
both languages, we provide the analysis for the hits with a
significant difference for all languages, W = 272180, p = 4e-6.
Misses
is not language independent. Hence, Spanish
participants with dyslexia (m = 1.0) make less mistakes
than the Spanish control group (m = 1.3). But German
participants with dyslexia (m = 1.2) make more mistakes
than the German control group (m = 0.9). Misses has no
significant difference for Spanish (W = 86340, p = 0.62, r =
0.05) or German (W = 20718, p = 0.36, r = 0.12). The
effect size is considered for both languages small [4].
Accuracy
is language independent. There were no dif-
ferences for participants with dyslexia (Spanish m = 0.57
& German m = 0.59) and the control group (Spanish m =
0.60 & German m = 0.60) in accuracy. The accuracy is not
significant different for Spanish (W = 90432, p = 0.43, r =
0.08) or German (W = 22084, p = 0.83, r = 0.04). Because
the trend is the same for both languages, we provide the
analysis for the accuracy with no significant difference for all
languages (W = 240780, p = 0.66).
Efficiency
is language independent. Hence, participants
with dyslexia (Spanish m = 3.1s & German m = 3.2s) take
more time for a hit than the control group (Spanish m =
2.8s & German m = 2.8s). The efficiency is significant for
Spanish (W = 73301, p = 4e 5, r = 0.39) but not for
German (W = 19357, p = 0.06, r = 0.25). The effect size
for both languages is considered as medium [4]. Because
the trend is the same for both languages, we provide the
analysis for the efficiency with a significant difference for all
languages (W = 209740, p = 2e 4).
Children and parents provided positive (n = 44) and
negative (n = 7) feedback about the gameplay or content.
Translated positive example quote from a boy (8 years) who
participated in a school: This was so cool! It was the best
day at school ever; from the web feedback input field of a girl
(12 years): it was fun and not boring! ; or a boy (10 years): I
love this game. The positive feedback was provided by all age
groups. Translated negative example quotes from a girl (12
years): not exciting more boring or a boy (12 years): game
started to fast.
6. DISCUSSION
The measurement data taken from the game MusVis show
that Spanish participants with dyslexia behave differently
than their control group. Differences can be reported for
the musical game for: 4th click interval, 6th click interval,
duration, and average click time. For the visual part the
following measurements can be reported as indicators: total
clicks, time to the first click, hits, and efficiency. Besides,
similar tendencies can be reported for the variables of the
visual part: total clicks, time to the first click, hits, accuracy,
and efficiency (see Table 5). We can show with our results
over all languages that the effect for each measurement is
confirmed even if we cannot draw strong conclusions about
our sample size on the comparison of German vs. Spanish
speaking participants. Spanish has 8 significant indicators
and we expected to reproduce the same amount of significant
indicators with more German participants.
In general, all participants found the game easy to under-
stand and only children at the age of 12 complained about
missing challenges. The amount of positive feedback and
engagement of all age groups let us conclude that the
game
mechanics and components
applied are also positive to
perceive the MusVis as a game and not as a test.
Dyslexia is known to be present
across different lan-
guages and cultures
[1]. The assumption that the ten-
dency for the indicators are similar over all languages cannot
be proven for all indicators in our study, e.g. German par-
ticipants with dyslexia start faster to click (music) than
the Spanish participants compared to their language control
group. We can exclude external factors like different appli-
cations or different study set up as a possible influence on
opposite tendency. According to the results we may have
Musical Language Independent
Total clicks
4th click interval
6th click interval
Duration
Average click time
No
No
No
No
No
Visual Language Independent
Total clicks
Time to first click
Hits
Misses
Accuracy
Efficiency
Yes
Yes
Yes
No
Yes
Yes
Table 5: Overview of all dependent variables show-
ing the language independent results between the
German and Spanish groups.
to assume that not all indicators for dyslexia are language
independent and have cultural dependencies. To confirm this
assumption we will need an equal larger number of partici-
pants for both language groups (Spanish and German).
The variables time to first click (visual & music) and total
number of clicks (visual & music) provide apart from the
cultural society also dependencies of the
game content and
game design
. Otherwise, we could not explain the trend dif-
ference between the musical and visual part for total number
of clicks, i.e., total clicks for visual is significantly different
than for music. Additionally, the analyses of the musical part
of the game present two limitations: (1) participants could
select a correct pair by chance, and (2) participants could
click through the game board without listening to the sounds.
Children with dyslexia are detected by their slower reading
or
error rate
as explained in the introduction [3, 31]. There-
fore, we designed our game with content that is known to be
difficult to differentiate for children with dyslexia to measure
the errors and the duration. Nevertheless, from previous
literature we knew that children with dyslexia do not make
more mistakes in games than the control group [28]. We
can confirm that misses did not reveal significant differences
for German or Spanish either. It might be possible that we
cannot compare errors in reading and writing with errors in
this type of games. Then, we cannot explain (yet) why the
Spanish control group made more mistakes than the Spanish
group with dyslexia.
Spanish children without dyslexia take significantly more
time to find all pairs and finish the musical game. Children
without dyslexia take more time before they click the first
time (visual) for all languages. The reason for that might be,
the time they need to
process the given information
in
the auditory processing [34] or recall the information from
the short-term memory [9] for auditory and visual. However,
participants with dyslexia from the German group are nearly
as fast as the control group in finding all pairs (music) which
might be due to the cultural differences.
The musical and visual elements are designed on purpose
to be more difficult to process for people with dyslexia than
without. Therefore, children with dyslexia are expected to
need more time (duration) which might be due to a
less
distinctive encoding of prosody
[9] and is line with
indicator of slower reading. Considering that children with
dyslexia need more time to process information we observe
this behavior as well for our indicators. For example, partic-
ipants with dyslexia from the Spanish group take more time
on the 4th click interval and also on the average click time
compared to the control group. Both results are significant
and with a large effect size of 0.7 and 0.8, we can estimate
what effect would be also in the whole population [4].
A person with dyslexia has difficulties to read and write
independently of the mother tongue which also appears when
learning a second language [11, 20]. The analysis of errors
from children with dyslexia show similar error categories for
Spanish, English [27], and German [23] which show similar-
ities of the perception between the languages. Our results
suggest that we can measure a significant difference on four
indicators for the visual game with the same tendency be-
tween Spanish, German, English, and Catalan. These means
that a person with dyslexia might perceive our visual game
content similar independent of the mother tongue. Further
research needs to be done to confirm the results but this
first pilot study shows strong evidence that it will be pos-
sible to measure dyslexia using our content, approach and
game design with the same language independent content
for different languages.
7. CONCLUSIONS AND FUTURE WORK
We presented a game with musical and visual elements that
collects measures that show differences between people with
and without dyslexia for different languages. With our user
study, we found eighth significant indicators for Spanish and
four significant indicators which are language independent
to distinguish a person with and without dyslexia. This is
preliminary evidence for a prediction approach for dyslexia
that would work across different languages.
The next step is to test how language dependent this
approach for different language is and compare tablets (touch)
and desktop computer (mouse) input. We already started to
pilot studies in Catalan and English to be able to compare
the different languages and study whether the game content
is truly language independent. For future work, we are
planning to conduct a large-scale study to be able to apply
a machine learning model to predict dyslexia based on the
eight significant indicators for Spanish.
Additionally, we will carry out a longitudinal study with
pre-readers to be able to predict children before they acquire
reading and writing skills. This would provide children with
dyslexia, more time to practice and compensate their difficul-
ties before starting school. Long term we plan to offer a game,
MusVis, with the musical and visual indicators that could
be language independent, applicable for pre-readers and
easily accessed from any given location with a tablet, laptop
or desktop computer. This will leverage the opportunities of
children with dyslexia being able to predict their difficulties
when they are pre-readers, in time to effectively intervene.
8. ACKNOWLEDGMENTS
This paper and content were partially supported by the
ICT PhD program of Pompeu Fabra through a travel grant
and funded by the fem:talent Scholarship from the Ap-
plied University of Emden/Leer as well as by the Deutschen
Lesepreis 2017 from the Stiftung Lesen and the Commerzbank-
Stiftung. We deeply thank M. Jes´us Blanque and R. No´e
opez, School Colegio Hijas de San Jos´e, Zaragoza; S.
Tena, A. Carrasco y E. endez, innovation team of Colegio
Leonardo da Vinci, Madrid; ChangeDyslexia, Barcelona; in
Spain, and F. Hansch, Grundschule Ofenerdiek, Oldenburg;
L. Klaus, Peter-Ustinov-Schule, Eckernf¨
orde; H. Marquardt,
Gorch-Fock-Schule, Eckernf¨
orde; A. Tammena and W. Volk-
mann, Johannes-Althusius-Gymnasium, Emden; M. Batke
and J. Thomaschewski, Hochschule Emden/Leer, Emden; and
P. St¨We also umpel, AncoraMentis, Rheine, in Germany.
thank all parents and children for playing MusVis. Finally,
thanks to H. Witzel for his advice during the development
of the visual part and to M. Blanca, and M. Herrera on the
translation for the Spanish version.
9. REFERENCES
[1] American Psychiatric Association. Diagnostic and
Statistical Manual of Mental Disorders. American
Psychiatric Association, May 2013.
[2] Benasich et al. Infant discrimination of rapid auditory
cues predicts later language impairment. Behavioural
Brain Research, 136(1):31–49, 2002.
[3] Coleman et al. A comparison of spelling performance
across young adults with and without dyslexia.
Assessment for Effective Intervention, 34(2):94–105, 2008.
[4] Field et al. How to design and report experiments. SAGE
Publications, 2003.
[5]
Franceschini et al. A causal link between visual spatial
attention and reading acquisition. Current Biology,
22(9):814–819, may 2012.
[6] Gaggi et al. Serious games for early identification of
developmental dyslexia. Comput. Entertain. Computers
in Entertainment, 15(4), 2017.
[7] Geurts et al. DIESEL-X: A game-based tool for early
risk detection of dyslexia in preschoolers. In Describing
and Studying Domain-Specific Serious Games, pages
93–114. Springer International Publishing, 2015.
[8]
Goswami et al. Impaired perception of syllable stress in
children with dyslexia: A longitudinal study.
Journal of
Memory and Language, 69(1):1–17, 2013.
[9]
Goswami et al. Prosodic similarity effects in short-term
memory in developmental dyslexia. Dyslexia, 2016.
[10] H¨al¨
am¨ainen et al. Basic auditory processing deficits in
dyslexia. Journal of Learning Disabilities, 46(5):413–427,
2013.
[11]
Helland et al. Dyslexia in English as a second language.
Dyslexia, 11(1):41–60, feb 2005.
[12] Huss et al. Music, rhythm, rise time perception and
developmental dyslexia: Perception of musical meter
predicts reading and phonology. Cortex, 47(6):674–689,
jun 2011.
[13] IDC Worldwide. Shipment forecast of tablets, laptops
and desktop PCs worldwide from 2010 to 2019 (in
million units). Statista, 2020:2020, 2016.
[14] D. J. Johnson. Persistent auditory disorders in young
dyslexic adults. Bulletin of the Orton Society,
30(1):268–276, jan 1980.
[15] Lexercise. Dyslexia test - Online from Lexercise.
http://www.lexercise.com/tests/dyslexia-test, 2016.
[Online; accessed 18-September-2017].
[16] Lyytinen et al. Dyslexia - Early identification and
prevention. Current Developmental Disorders Reports,
2(4):330–338, dec 2015.
[17]
M¨
annel et al. Phonological abilities in literacy-impaired
children: Brain potentials reveal deficient phoneme
discrimination, but intact prosodic processing.
Developmental Cognitive Neuroscience, 23:14–25, 2016.
[18] Neef et al. Dyslexia risk gene relates to representation
of sound in the auditory brainstem. Developmental
Cognitive Neuroscience, 24(February):63–71, 2017.
[19] Nessy. Dyslexia screening - Nessy UK.
https://www.nessy.com/uk/product/dyslexia- screening/,
2011. [Online; accessed 18-September-2017].
[20] J. Nijakowska. Dyslexia in the foreign language classroom.
Multilingual Matters, 2010.
[21] K. Overy. Dyslexia, temporal processing and music:
The potential of music as an early learning aid for
dyslexic children. Psychology of Music, 28(2):218–229,
oct 2000.
[22] R. F. Port. Meter and speech. Journal of Phonetics,
31:599–611, 2003.
[23] Rauschenberger et al. A language resource of German
errors written by children with dyslexia. In
LREC 2016
,
Paris, France, may 2016.
[24] Rauschenberger et al. Supplement: DysMusic musical
llements: Towards the prediction of dyslexia by a
web-based game with musical elements, 2017.
https://doi.org/10.5281/zenodo.809783.
[25] Rauschenberger et al. Towards the prediction of
dyslexia by a web-based game with musical elements.
In W4A’17, pages 4–7, 2017.
[26] Rello et al. Good fonts for dyslexia. Proceedings of the
15th International ACM SIGACCESS Conference on
Computers and Accessibility, page 14, 2013.
[27] Rello et al. A resource of errors written in Spanish by
people with dyslexia and its linguistic, phonetic and
visual analysis.
Language Resources and Evaluation
, pages
1–30, 2016.
[28]
Rello et al. Dytective: Diagnosing risk of dyslexia with
a game. In Proc. Pervasive Health’16, Cancun, Mexico,
2016.
[29] Rello et al. Screening dyslexia for english using hci
measures and machine learning. In Proc. Digital
Health’18, Lyon, France, 2018.
[30]
Rolka et al. A systematic review of music and dyslexia.
Arts in Psychotherapy, 46:24–32, 2015.
[31] Schulte-K¨
orne et al. Familial aggregation of spelling
disability. Journal of Child Psychology and Psychiatry,
37(7):817–822, oct 1996.
[32] G. Schulte-K¨
orne. Diagnostik und therapie der
lese-Rechtscheib-St¨
orung: The prevention, diagnosis,
and treatment of dyslexia. Deutsches ¨
Arzteblatt
international, 107(41), 2010.
[33] Steinbrink et al. Lese-Rechtschreibst¨
orung. Springer
Berlin Heidelberg, 2014.
[34]
P. Tallal. Improving language and literacy is a matter of
time. Nature reviews. Neuroscience, 5(9):721–728, 2004.
[35] Vellutino et al. Specific reading disability (dyslexia):
What have we learned in the past four decades?
Journal
of Child Psychology and Psychiatry, 45(1):2–40, jan 2004.
[36] Vidyasagar et al. Dyslexia: a deficit in visuo-spatial
attention, not in phonological processing. Trends in
Cognitive Sciences, 14(2):57–63, 2010.
[37] Yuskaitis et al. Neural mechanisms underlying musical
pitch perception and clinical applications including
developmental dyslexia. Current neurology and
neuroscience reports, 15(8):51, 2015.
... Historically, the rates of spelling mistakes and reading errors have been the most common way to detect persons with dyslexia, using the popular paper and pencil assessments in different languages (Cuetos et al., 2002(Cuetos et al., , 2007Fawcett and Nicolson, 2004;Grund et al., 2004). Therefore, we compare our game measures and found in our pilot study (n 178) four significant game measurements for Spanish, German, and English as well as eight significant game measurements for Spanish (Rauschenberger et al., 2018b), e.g., total clicks or time to first click. ...
... We combined, for example, the deficits of children with dyslexia in auditory working memory with the results on the short duration of sounds (Huss et al., 2011) while taking the precaution of not measuring hearing ability (Fastl and Zwicker, 2007). Each stage is assigned to one acoustic parameter like frequency or rhythm which is designed with the knowledge of the analysis from previous literature (Rauschenberger et al., 2018b. ...
... The auditory cues are generated with a simple sinus tone using the free software Audacity 3 . The exact parameters of each auditory cue are already published (Rauschenberger et al., 2018b) and the auditory cues are available at GitHub (Rauschenberger et al., 2017a) 4 . Each stage has two rounds, with first two and then three auditory cues that must be assigned by choosing the same sound (see Figure 2). ...
Article
Full-text available
Children with dyslexia have difficulties learning how to read and write. They are often diagnosed after they fail school even if dyslexia is not related to general intelligence. Early screening of dyslexia can prevent the negative side effects of late detection and enables early intervention. In this context, we present an approach for universal screening of dyslexia using machine learning models with data gathered from a web-based language-independent game. We designed the game content taking into consideration the analysis of mistakes of people with dyslexia in different languages and other parameters related to dyslexia like auditory perception as well as visual perception. We did a user study with 313 children (116 with dyslexia) and train predictive machine learning models with the collected data. Our method yields an accuracy of 0.74 for German and 0.69 for Spanish as well as a F1-score of 0.75 for German and 0.75 for Spanish, using Random Forests and Extra Trees, respectively. We also present the collected user data, game content design, potential new auditory input, and knowledge about the design approach for future research to explore universal screening of dyslexia. Universal screening with language-independent content can be used for the screening of pre-readers who do not have any language skills, facilitating a potential early intervention.
... Detection of risks done earlier could minimize the undesirable outcomes for children development. A number of researchers used computer programs or games with young children for screening purpose [1] [2] [3] [4] [5]. Games offer mechanisms that help lessening anxiety of children while playing the games instead of feeling that they are being tested. ...
... Games offer mechanisms that help lessening anxiety of children while playing the games instead of feeling that they are being tested. Recently, several serious games were developed based on dyslexia features on short-term memory, visual perception, or audio perception [1][2] [3][4] [5]. Among them, MusVis and GC are the games with published accuracy results [2] [3]. ...
... Recently, several serious games were developed based on dyslexia features on short-term memory, visual perception, or audio perception [1][2] [3][4] [5]. Among them, MusVis and GC are the games with published accuracy results [2] [3]. ...
Conference Paper
Full-text available
Dyslexia is the disorder affecting reading and learning language acquisition. Children with dyslexia cannot spell words or acquire reading. Screening is a necessary process to distinguish dyslexia children early to treat and help them appropriately. If detection can be done since they are in pre-readers age, it can lessen the undesirable outcomes for children such as lower educational attainment, negative feelings toward learning, and loss of low self-confidence and self-respect. Many researchers start using computer games for screening purpose. Game design is a significant part to achieve the goal. One of the challenges in creating a game for children is designing the game that maintains children’s attention until they finish the game. In addition, the level of difficulties must be suitable for the participants’ skill. Thus, this paper presents a pilot study to develop a picture rotation game. The game was designed mainly to study behavior of children with dyslexia while overlooking the game element principles. This study found that game playing behavior of the two groups of children was different. The difficulty levels of the game were found to have influences on the gameplay behaviors of both groups. The findings are used as the guidelines to improve the design of the game to better assess the risk of dyslexia children.
... Among native English speakers, 10% of Australians suffer from dyslexia while the rate is much higher ( 20%) for other nationals including the citizens of Canada and the UK. A similar trend is present among other language users [1,2]. Developmental Dyslexia (i.e., dyslexia which is genetic, present from birth, and develops over time) is generally observed among the younger school-going population who performs poorly in schools and often faces difficulties during their studies. ...
... Recently, machine learning techniques have become popular in predicting Dyslexia from the data collected through online tests, such as reading and writing exercises, online games, tracking eye movement, or collecting EEG scans and MRI data while the participants engage in reading or writing tasks [5]. Among these techniques, online gamified tests have recently gained popularity among researchers as they are cost-effective, easier to conduct, and can cover a wide participant base [2,4,6]. In this case, participants engage in online games while information about their performance is collected and later analyzed for detecting dyslexia. ...
Article
Developmental Dyslexia is a learning disorder often discovered in school-aged children who face difficulties while reading or spelling words even though they may have average or above-average levels of intelligence. This ultimately results in anger, frustration, low self-esteem, and other negative feelings. Early detection of Dyslexia can be highly beneficial for dyslexic children as their learning needs can be properly addressed. Researchers have used several testing techniques for early discovery where the data is collected from reading and writing tests, online games, Magnetic reasoning imaging (MRI) and Electroencephalography (EEG) scans, picture and video recording. Several Machine learning techniques have also been used in this regard recently. However, existing works did not focus on the problem of the imbalanced dataset where the percentage of dyslexic participants is much higher compared to non-dyslexic participants, which is expected to be the case for pre-screening among a random population. This paper addresses the imbalanced dataset obtained from dyslexia pre-screening tests and proposes an oversampling and ensemble-based machine learning technique for the detection of Dyslexia. Simulation results show that the proposed approach improves the detection accuracy of the minority class, i.e., dyslexic patients from 80.61% to 83.52%.
... Another popular approach is to create a custom game based on theories of cognition or other previous research. For example, several papers in our review created games to detect dyslexia based on theories of visual-spatial attention [55,89,96,100]. As another example, McKanna et al [32] created 21 Tally, described as "blackjack played in two dimensions," based on theories of divided attention. ...
... Some games are very literature-and hypothesis-driven, whereas others offer less rationale for their design. For example, Rauschenberger et al [100] created MusVis, a game to detect dyslexia using language-independent methods. They based their design choices on the visual and auditory processing abilities of individuals with dyslexia. ...
Article
Full-text available
Background Serious games are now widely used in many contexts, including psychological research and clinical use. One area of growing interest is that of cognitive assessment, which seeks to measure different cognitive functions such as memory, attention, and perception. Measuring these functions at both the population and individual levels can inform research and indicate health issues. Attention is an important function to assess, as an accurate measure of attention can help diagnose many common disorders, such as attention-deficit/hyperactivity disorder and dementia. However, using games to assess attention poses unique problems, as games inherently manipulate attention through elements such as sound effects, graphics, and rewards, and research on adding game elements to assessments (ie, gamification) has shown mixed results. The process for developing cognitive tasks is robust, with high psychometric standards that must be met before these tasks are used for assessment. Although games offer more diverse approaches for assessment, there is no standard for how they should be developed or evaluated. Objective To better understand the field and provide guidance to interdisciplinary researchers, we aim to answer the question: How are digital games used for the cognitive assessment of attention made and measured? Methods We searched several databases for papers that described a digital game used to assess attention that could be deployed remotely without specialized hardware. We used Rayyan, a systematic review software, to screen the records before conducting a systematic review. Results The initial database search returned 49,365 papers. Our screening process resulted in a total of 74 papers that used a digital game to measure cognitive functions related to attention. Across the studies in our review, we found three approaches to making assessment games: gamifying cognitive tasks, creating custom games based on theories of cognition, and exploring potential assessment properties of commercial games. With regard to measuring the assessment properties of these games (eg, how accurately they assess attention), we found three approaches: comparison to a traditional cognitive task, comparison to a clinical diagnosis, and comparison to knowledge of cognition; however, most studies in our review did not evaluate the game’s properties (eg, if participants enjoyed the game). Conclusions Our review provides an overview of how games used for the assessment of attention are developed and evaluated. We further identified three barriers to advancing the field: reliance on assumptions, lack of evaluation, and lack of integration and standardization. We then recommend the best practices to address these barriers. Our review can act as a resource to help guide the field toward more standardized approaches and rigorous evaluation required for the widespread adoption of assessment games.
... The aim of this already-collected data was to distinguish between children with and without dyslexia. The interactive system was designed with the HCD approach [6] and is well described in research papers [16][17][18][19]. ...
Conference Paper
Full-text available
In many areas, only small data sets are available and big data does not play a significant role, e.g., in Human-Centered Design research. In the context of machine learning analysis, results of small data sets can be biased due to single variables or missing values. Nevertheless, reliable and interpretable results are essential for determining further actions, such as, e.g., treatments in a health-related use case. In this paper, we explore machine learning clustering algorithms on the basis of a small, health-related (variance) data set about early dyslexia screening. Therefore, we selected three different clustering algorithms from different clustering methods: K-Means, HAC and DBSCAN. In our case, K-Means and HAC showed promising results, while DBSCAN did not deliver distinct results. Based on our experiences, we provide first proposals on how to handle small data set clustering and describe situations in which using Human-Centered Design methods can increase interpretability of machine learning clustering results. Our work represents a starting point for discussing the topic of clustering small data sets.
... Nonetheless, the app was language-dependent, and children should have been able to read and write. Another app [6] for screening dyslexia in children aged 7 to 12 years was also developed to solve the language dependency. There are also findings in the SpLD literature for dyscalculia. ...
Article
Full-text available
Background Specific learning difficulties (SpLD) include several disorders such as dyslexia, dyscalculia, and dysgraphia, and the children with these SpLD receive special education. However, the studies and the educational material so far focus mainly on one specific disorder. Objective This study’s primary goal is to develop comprehensive training material for different types of SpLD, with five serious games addressing different aspects of the SpLD. The second focus is measuring the impact of adaptive difficulty level adjustment in the children’s and their educators’ usability and technology acceptance perception. Receiving feedback from the children and their educators, and refining the games according to their suggestions have also been essential in this two-phase study. Methods A total of 10 SpLD educators and 23 children with different types of SpLD tested the prototypes of the five serious games (ie, Word game, Memory game, Category game, Space game, and Math game), gave detailed feedback, answered the System Usability Scale and Technology Acceptance Model (TAM) questionnaires, and applied think-aloud protocols during game play. Results The games’ standard and adaptive versions were analyzed in terms of average playtime and the number of false answers. Detailed analyses of the interviews, with word clouds and player performances, were also provided. The TAM questionnaires’ average and mean values and box plots of each data acquisition session for the children and the educators were also reported via System Usability Scale and TAM questionnaires. The TAM results of the educators had an average of 8.41 (SD 0.87) out of 10 in the first interview and an average of 8.71 (SD 0.64) out of 10 in the second interview. The children had an average of 9.07 (SD 0.56) out of 10 in the first interview. Conclusions Both the educators and the children with SpLD enjoyed playing the games, gave positive feedback, and suggested new ways for improvement. The results showed that these games provide thorough training material for different types of SpLD with personalized and tailored difficulty systems. The final version of the proposed games will become a part of the special education centers’ supplementary curriculum and training materials, making new enhancements and improvements possible in the future.
... Medical Therapy [1, 38-41, 65, 66, 83, 93, 114, 120, 129, 161, 166, 171, 172] Diagnosis [13,47,80,109,129,138,140,149,155,177] Training [77-79, 124, 141, 175, 184] Social Collaboration [14,17,23,26,28,69,116,127,131,134,143,162] Education [22,29,37,50,68,84,85,90,97,115,119] Communication [24,65] Sports [50,75] Work Skills [106,170] Art & Public [47,76] Self-Guided Free Play [108,142,[144][145][146]183] Some papers span more than one purpose, sometimes even across models of disability. One such instance is the work by Craven et al. which is intended for public, artful engagement as well as envisioned within diagnostic contexts [47]. ...
Article
Play presents a popular pastime for all humans, though not all humans play alike. Subsequently, Human–Computer Interaction Games research is increasingly concerned with the development of games that serve neurodivergent ¹ players. In a critical review of 66 publications informed by Disability Studies and Self-Determination Theory, we analyse which populations , research methods, kinds of play and overall purpose goals existing games address. We find that games are largely developed for children, in a top-down approach. They tend to focus on educational and medical settings and are driven by factors extrinsic to neurodivergent interests. Existing work predominantly follows a medical model of disability, which fails to support self-determination of neurodivergent players and marginalises their opportunities for immersion. Our contribution comprises a large-scale investigation into a budding area of research gaining traction with the intent to capture a status quo and identify opportunities for future work attending to differences without articulating them as deficit.
Chapter
While early identification of children with dyslexia is crucial, it is difficult to assess literacy risks of these children early on before they learn to read. In this study, we expand early work on Dot-to-Dot (DtD), a non-linguistic visual-motor mechanism aimed at facilitating the detection of the potential reading difficulties of children at pre-reading age. To investigate the effectiveness of this approach on touchscreen devices, we conducted a user study with 33 children and examined their task performance logs as well as language test results. Our findings suggest that there is a significant correlation among DtD task and series of language tests. We conclude the work by suggesting different ways in which DtD could be seamlessly embedded into everyday mobile use cases.
Conference Paper
Full-text available
More than 10% of the population has dyslexia, and most are di- agnosed only after they fail in school. This work seeks to change this through early detection via machine learning models that pre- dict dyslexia by observing how people interact with a linguistic computer-based game.We designed items of the game taking into account (i) the empirical linguistic analysis of the errors that peo- ple with dyslexia make, and (ii) specific cognitive skills related to dyslexia: Language Skills, Working Memory, Executive Functions, and Perceptual Processes. . Using measures derived from the game, we conducted an experiment with 267 children and adults in order to train a statistical model that predicts readers with and without dyslexia using measures derived from the game. The model was trained and evaluated in a 10-fold cross experiment, reaching 84.62% accuracy using the most informative features. CCS
Conference Paper
Full-text available
Current tools for screening dyslexia use linguistic elements, since most dyslexia manifestations are related to difficulties in reading and writing. These tools can only be used with children that have already acquired some reading skills and; sometimes, this detection comes too late to apply proper remediation. In this paper, we propose a method and present DysMusic, a prototype which aims to predict risk of having dyslexia before acquiring reading skills. The prototype was designed with the help of five children and five parents who tested the game using the think aloud protocol and being observed while playing. The advantages of DysMusic are that the approach is language independent and could be used with younger children, i.e., pre-readers.
Article
Full-text available
Dyslexia is a reading disorder with strong associations with KIAA0319 and DCDC2. Both genes play a functional role in spike time precision of neurons. Strikingly, poor readers show an imprecise encoding of fast transients of speech in the auditory brainstem. Whether dyslexia risk genes are related to the quality of sound encoding in the auditory brainstem remains to be investigated. Here, we quantified the response consistency of speech-evoked brainstem responses to the acoustically presented syllable [da] in 159 genotyped, literate and preliterate children. When controlling for age, sex, familial risk and intelligence, partial correlation analyses associated a higher dyslexia risk loading with KIAA0319 with noisier responses. In contrast, a higher risk loading with DCDC2 was associated with a trend towards more stable responses. These results suggest that unstable representation of sound, and thus, reduced neural discrimination ability of stop consonants, occurred in genotypes carrying a higher amount of KIAA0319 risk alleles. Current data provide the first evidence that the dyslexia-associated gene KIAA0319 can alter brainstem responses and impair phoneme processing in the auditory brainstem. This brain-gene relationship provides insight into the complex relationships between phenotype and genotype thereby improving the understanding of the dyslexia-inherent complex multifactorial condition.