Content uploaded by Benjamin Bach
Author content
All content in this area was uploaded by Benjamin Bach on Apr 18, 2019
Content may be subject to copyright.
Content uploaded by Benjamin Bach
Author content
All content in this area was uploaded by Benjamin Bach on Feb 26, 2019
Content may be subject to copyright.
Comparing Eectiveness and Engagement of Data
Comics and Infographics
Zezhong Wang1Shunming Wang1,2 Matteo Farinella3Dave Murray-Rust1Nathalie Henry Riche4
Benjamin Bach1
1University of Edinburgh 2Peking University 3Columbia University 4Microsoft Research
zezhong.wang@ed.ac.uk shunming.wang@pku.edu.cn {bbach, d.Murray-Rust}@inf.ed.ac.uk nath@microsoft.com
mf3094@columbia.edu
Figure 1: Two stories were presented in the wild in both Infographic and Comic form, for an empirical observation study
conducted with pedestrians, measuring reading time, interactions (i.e. pointing and talking) and opinions as evidence for
engagement and enjoyment.
ABSTRACT
This paper compares the eectiveness of data comics and
infographics for data-driven storytelling. While infograph-
ics are widely used, comics are increasingly popular for ex-
plaining complex and scientic concepts. However, empirical
evidence comparing the eectiveness and engagement of in-
fographics, comics and illustrated texts is still lacking. We
report on the results of two complementary studies, one in a
controlled setting and one in the wild. Our results suggest
participants largely prefer data comics in terms of enjoyment,
focus, and overall engagement and that comics improve un-
derstanding and recall of information in the stories. Our
ndings help to understand the respective roles of the inves-
tigated formats as well as inform the design of more eective
data comics and infographics.
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 prot or commercial advantage and that
copies bear this notice and the full citation on the rst 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 specic
permission and/or a fee. Request permissions from permissions@acm.org.
CHI 2019, May 4–9, 2019, Glasgow, Scotland Uk
©
2019 Copyright held by the owner/author(s). Publication rights licensed
to ACM.
ACM ISBN 978-1-4503-5970-2/19/05.. .$15.00
https://doi.org/10.1145/3290605.3300483
KEYWORDS
Visualization, Comics, Eectiveness, Engagement
ACM Reference Format:
Zezhong Wang, Shunming Wang, Matteo Farinella, Dave Murray-Rust,
Nathalie Henry Riche, Benjamin Bach. 2019. Comparing Eectiveness and
Engagement of Data Comics and Infographics. In CHI Conference on Hu-
man Factors in Computing Systems Proceedings (CHI 2019), May 4-9, 2019,
Glasgow, Scotland, UK. ACM, New York, NY, USA, Paper 253, 13 pages.
https://doi.org/10.1145/3290605.3300483
1 INTRODUCTION
Data-driven storytelling is concerned with eective commu-
nication around data through visualization [
26
,
38
]. It covers
a multitude of formats to match the diversity of audiences,
messages, contexts, data, and communication media [
44
]:
magazine style, annotated charts, partitioned posters, ow-
chart, comic strips, sideshows and videos. Many of these
formats can be seen as complementary, in that they support
specic media, might be more attractive to specic audiences,
have dierent ways of supporting a specic message, and
involve specic skills and tools for creation.
In this paper, we are interested in data comics [
3
,
54
], a
relatively new and underexplored format. Data comics draw
from the tradition of comics, and combine techniques from
infographics, data visualization, journalism, and other for-
mats of visual explanations. They are based on the notion
of a sequence of panels [
32
] with each panel being a combi-
nation of text and picture to illustrate a particular message,
the author wants to convey. For sequential explanations in
general, many studies have shown benets (e.g., [
13
,
36
],
summarized in [
18
]). Generally, the use of comics in class-
room education and health-related communication appears
to improve the reader’s comprehension and engagement.
While these results may encourage the use of data comics,
generalizing existing study results to data comics is challeng-
ing due to varying content, visual representations, audiences,
and purposes.
As a step towards a better understanding, we provide the
rst structured investigation into the eectiveness of data
comics. We compare data comics with the two closest pop-
ular alternatives: infographics and texts accompanied by
visualizations (illustrated texts). While there are numerous
dierences between these three formats, we focus our com-
parison on i) the degree to which an explicit reading order
is given, and ii) how close the integration of text and pic-
ture is (Section 3). While data comics typically have a highly
structured reading order and text closely integrated pictures,
illustrated texts and infographics exhibit a varying degree in
each of these dimensions. We compared these techniques in
a controlled lab study, where 36 participants read the same 3
stories, each presented in a dierent format (Section 4). We
collected empirical and subjective data on understandability,
recall, preferences, and engagement. To complement this
controlled study, we observed 50 groups of visitors engag-
ing with comics and infographics in an open public space
(Section 6, Figure 1).
Based on a variety of collected data including question-
naires, interviews, think-aloud protocols, explanations, me-
mory-recall, and visitor observation, we found that data
comics are seen as more fun, help readers to stay focused,
and are overall more engaging. Quantitative results suggest
that data comics improve understanding for most cases, and
were generally rated high for enjoyment and engagement. In
addition, subjective feedback collected gives a richer picture
of the respective merits and drawbacks of each format. With
these results, we aim to encourage professional communi-
cators to explore the use of data comics for explanation, as
well as to better inform the design of future data comics and
communicative graphics. Study materials for the Lab study,
in the wild study, questions, some interview quotations and
examples of reading sequence record, can be found online:
http://datacomics.net.
2 BACKGROUND
This section overviews previous studies on illustrated texts,
infographics and other related data storytelling media.
Texts and Pictures
Textual descriptions (printed and spoken words) and graphi-
cal depictions (photography, drawings, paintings, maps and
videos) are the basic forms of representation [
43
,
51
]. In-
formation is remembered better when it is supported with
pictures [
28
], even more so when presented through both
channels at the same time [
6
,
13
,
36
,
48
]. Data visualizations
have been found less memorable than natural scenes [
11
],
yet adding embellishments and unique presentations can
improve memorability [
7
,
11
]. Studies comparing comics and
other visual formats [
27
] with text-only material [
1
] and il-
lustrated texts [
31
] conrm these trends and show increased
memorability. For storytelling in general, the use of sequence
has been found to increase recall, facilitated by information
being split into chunks [
10
]. Psychological studies have inves-
tigated the eectiveness of the panel layouts [
14
,
15
] used in
traditional comics through eye-tracking [
19
]. Other studies
have found that in cases where text is unnecessary, closely
integrating text and picture can distract the reader and hin-
der learning [
12
,
40
], while providing inappropriate graphics
can impede understanding [29].
Comics for Explanation
Comics are becoming increasingly popular for explaining
complicated processes [
18
,
49
]. They oer a set of unique
characteristics for communication and optimal understand-
ing, being highly accessible to a large audience, compatible
with many dierent media, do not require a presenter and
can be read at one’s own pace [
53
]. Comics can blend expla-
nation (e.g. schemata, illustrations, data visualization) with
narration, characters and dialogue. From a structural point
of view, a notion central to comics is the panel, which en-
capsulates a specic message (or information) represented
as an integrated combination of text and picture [
32
]. The
(mostly) linear order of panels creates a sequence of view-
points which together can build a deeper understanding in
the readers’ mind [46].
Comics, and related formats such as juxtaposed annotated
pictures, have consequently been used for explaining a range
of phenomena, blending schematic drawings and illustra-
tions with the narration and characters of traditional comics.
Empirical evidence from using comics as classroom mate-
rial [
21
,
45
,
47
] as well as for health-related communication
[
16
,
17
,
33
,
50
] suggests that comics may be more engaging
and easier to comprehend, especially for non-expert readers.
Recently, comics have also been applied to the communi-
cation of insights from data, resulting in a format called data
comics [
3
,
4
,
54
], where comic strip elements are combined
with data visualization techniques. A growing number of
data comics have been created for networks [
2
,
54
], geog-
raphy [
8
,
34
], personal data and scientic experiments [
4
],
summarized online [
5
]. While research highlighted design
patterns [
4
], authoring tool-support [
25
], and workshop for-
mats [
52
] for creating data comics, several factors prevent a
wider distribution and exploration of data comics: the nov-
elty of the format as well as the lack of a better understanding
of the eectiveness of the format.
In fact, while most of the evidence from previous studies
supports a strong motivation for the use of comics, the eects
were observed in traditional or scientic comics and may
or may not generalize to data comics. In particular, graph-
ics in data comics focus more on abstract data visualization
[
3
], require visualization knowledge and largely deal with
concepts requiring data literacy such as relations, temporal
change, and quantities. For example, Qu and Hullman point
to the importance of consistency in creating sequences of
visualizations [
37
], a factor that very likely inuences the
acceptability and usability of data comics. Given the popular-
ity of infographics and traditional visualisation formats, as
well as the body of expertise built up around their creation
and use, it would be wrong to simply assume there will be
enhanced understanding and engagement with data comics.
Our study aims to support the further investigation, applica-
tion and creation of data comics through empirical results
on their eectiveness. As far as we are aware, ours is the
rst study to address this gap.
Infographics and Illustrated Texts
Two popular alternatives to data comics are infographics and
illustrated texts (“magazine style” [
44
]). Both are text-picture
combinations, diering in how these elements are combined.
Infographics strongly emphasize the visual content, often
using visual embellishments to stimulate attractiveness and
memory [
7
]. They tend to be open-ended, inviting the reader
to explore the content without any specic direction. For
more structured narrative content, e.g., in journalism and
scientic papers, texts illustrated with visualizations have
become the common option: a text for the main narration,
referring to visualizations as needed.
All three formats (comics, infographics, and illustrated
texts) are expressions of dierent combinations of reading
order and text-picture integration. Current studies are not
conclusive and say little about these three formats ability to
support understanding and memory as well as how engaging
each format appears to potential readers.
3 STUDY DESIGN SPACE
Infographics, illustrated texts, and data comics vary in a
range of characteristics and it is not easy to precisely dene
each format. Thus, to inform our baseline techniques for the
study, we focus on two characteristics across these formats:
1Text-Picture-Integration
is the spatial distance between
the verbal message and the visualization [
3
]. In their general
form, comics exhibit a high integration of text and picture;
one piece of text is related exactly to one image, both
supporting the same information within the panel. A low
integration would mean that text and picture are presented
in parallel with occasional implicit or explicit references
such as found in scientic and news articles.
2Reading Guidance
is how strong the reader is guided while
reading the story [
3
,
4
]. While exceptions exist [
4
], comics
provide high guidance through an explicit order of panels.
Low guidance can be found in most infographics where
text and pictures are usually not organized in a linear order
but can be read in a non-linear way.
Figure 2: Design space
We use these dimen-
sions to map out a de-
sign space for visual sto-
rytelling formats for our
study (Figure 2). Comics
are located high on both
dimensions, with tight
text-picture integration and
strong reader guidance.
Infographics provide a less
strong integration of tex-
tual and pictorial content
because the same picture is usually annotated with several
textual information and the specic scope of the text, with
respect to the picture. Infographics provide low reading guid-
ance as the content can usually be read in any order. Illus-
trated Texts are by visualizations, sometimes referenced in
the text. This format provides a low integration of text and
picture, a higher reading guidance than infographics, but
less than comic as it is up to the reader to switch between
text and picture.
To investigate the eectiveness and engagement of these
formats, we opted to perform two complementary studies:
one in laboratory settings and one in the wild.
4 LAB STUDY
Our rst study aimed to compare comics, infographics, and
illustrated text in a controlled lab study. We were gener-
ally interested in how each of the dimensions, laid out in
the previous section, inuences comprehension, recall, and
engagement with the formats. More specic questions con-
cerned how people read each format, which congurations
they found helpful or confusing, how well did they under-
stand the communicated stories, as well as how much of
the content do participants remember after a while. To that
end, we invited participants to read a series of stories in
each of the formats, answer questions related to content un-
derstanding, and report their subjective impressions about
Figure 3: Samples of Alliances in the lab study. IllustratedText (left) ; Infographic (center); Comic (right).
each format. Prior to the study, we carried out a pilot with
three visualization-experienced participants to ensure the
suitability of our material and techniques. The full material
(including all questions and results) is found in the supple-
mentary material.
Stories
During the study, participants read three data stories, a num-
ber we chose with respect to how long we expected partici-
pants to remain attentive and remember story content. Our
main criteria in nding appropriate stories for a controlled
lab study were (i) comfortably presentable on A3 paper and
(ii) requiring little previous knowledge. To extend the varia-
tion among the study examples, we selected dierent types of
common visualizations including multivariate data, dynamic
networks, and geological distribution. To ensure realistic
stories, we selected three existing infographics (see below).
For two of the stories, the labels, text, and data were changed
to prevent users from relying on prior knowledge.
•Energy:Renewables’ Mix in Power Generation in Eu-
rope
talks about the production and distribution of renew-
able energy among European countries. The visualization
was found in the Energy Atlas [
9
]. Bars on each country
showed production in 2011 and 2017 respectively. Texts
describe features of the distribution (Figure 4-left).
•Alliances:Kings’ Alliances and Aggressions in Dul-
lama
(Figure 3) was based on a graph comic on dynamic
networks European Alliances before World War I [
2
]. Dates
were changed and the story was transposed into a cti-
tious medieval kingdom. Country names were replaced
by common names and the background map was altered.
•Economics:Global Interest Rate And Tax Burden
was
taken from Hans Rosling’s TED talk [
41
], featuring a dy-
namic scatterplot of country indicators over time [
20
]
(Figure 4-right). We changed axes labels to Interest rate
and tax burden respectively, implying no specic mean-
ing. The narration was guided by the changes to specic
countries over time.
While we tried to keep stories to an equal level of complex-
ity (e.g., number of panels, number of messages per panel,
Figure 4: Infographics: Energy (left) and Economics (right).
expected commonness of visualizations and diculty of un-
derstanding visualizations), all three participants in the pilot
study gave the same ordering in terms of increasing com-
plexity: Energy,Economics, and Alliances.
Techniques
The techniques in the lab study are the three techniques men-
tioned in Section 3. As boundaries between the formats are
blurred, we settled on the following criteria to create the ma-
terial for our comparison. Fig. 3 shows the three formats for
Alliances. The full material for the other stories is available
in https://datacomics925658343.wordpress.com/study/.
•IllustratedText
: One or two narrative paragraphs
were placed right below the title, followed by the legend
below or on the right. Text and pictures were visually
separated, captions and essential names were added on
the pictures referring to the text description.
•Infographic
: Besides a 1-2 sentence summarization un-
der the title, texts were shown in text boxes, placed close
to the visual information they referred to. Leader lines
sometimes connected text to visual elements. No explicit
visual hint implied a specic reading sequence.
•Comic
: All text and visuals were placed into 16-18 panels
in an unambiguous linear left-right, top-down order. Vi-
suals were adapted to support the text, e.g., focusing on
specic information, highlighting important and remov-
ing unimportant information. We made no other changes
to visual encoding or text.
We ensured everything was readable comfortably from
40cm distance, printed on A3 paper format. Texts were kept
the same across formats, except minor adjustments to the
format. For example, we added “see gure x” for Illustrat-
edText and simplied few longer sentences to t the panels
in Comic. We created all materials through several iterations
to make sure they were readable and understandable and
to lower confusion; at each stage, we discussed in depth
across the authors and asked for feedback from three ex-
ternal comic artists as well as two external colleagues. The
pilot study identied minor problems in understanding and
design and we consequently modied several ambiguous
expressions in the stories and the questionnaires. To provide
a high-resolution and minimize the operational distractions
such as panning and zooming, we printed each story on
one-side A3 paper instead of showing them on a monitor.
Hypotheses
We developed our hypotheses based on the existing literature,
and the two dimensions of our design space.
•H-Accuracy
: We expect participants to have a more cor-
rect understanding of the messages after reading Comic than
Infographic, and least with IllustratedText. We believe
that making a connection between visual and verbal for
each message improves understanding. Comics with mes-
sages embedded in individual panels will be more eective
than combined information in Infographic and separated
text and pictures.
•H-Engaging
: we expect Comic and Infographic to be
rated as more engaging than IllustratedText. This means
that people are more willing to spend time with Comic
and Infographic and believe they are more fun to read.
•H-Memorability
: We expect Comic to increase reten-
tion of information compared to the other formats due to
the separation of information into clear chunks, supported
through tightly integrated texts-and-pictures, presented
in a clear reading order [31].
Data Collection
Seven types of data were collected during the lab study: (1)
error rate from multiple-choice questions to measure under-
standing of each story and format (including four answer
possibilities and “I’m not sure”); (2) story recall scores, coded
by the experimenter and explained in Section 5; (3) reading
paths drawn onto the material by the participants; (4) sub-
jective scores, and (5) qualitative feedback from participants
on all three formats for the same story (shown at the end
of the study) gathered through semi-structured interviews.
Questions included “Compared to the other formats of same
story, how would you judge these formats?” and “Do you have
any preference among these formats and why?”. Subjective
ratings were collected through a questionnaire after partici-
pants saw the three formats of all stories. While engagement
is hard to dene and measure [
30
], we assessed aesthetic
feeling, willingness of spending time to explore, enjoyment,
attention, engagement and preference using 7-point Likert
scales, inspired by existing frameworks [35, 42].
Participants
We recruited 38 participants through mailing lists of a Western-
European university. One subject was eventually denied par-
ticipation due to missing English language prociency and
one subjects results were invalidated as she did not complete
phase 1. Of the remaining 36 participants, 19 were male, ages
ranged between 18 and 35 years, 1 A-level, 9 undergraduates
and 20 graduates, 4 computer science doctoral students, 1
research associate in computer science and 1 lecturer of Lin-
guistics and English Language. Participants’ backgrounds
included Design, Art, Computer Science, Engineering, Lin-
guistics, Philosophy and Psychology. We had 16 Europeans,
15 Asians, 3 North American and 2 African; 34 had lived in
an English speaking country for more than 1 year, the other
2 had scores in English language test equal to B2 level of
Common European Framework of Reference for Languages.
To assess participants’ level of reading ability, comic and
visualization literacy, self rated questions were lled in be-
fore the study. For text reading ability, 13 read illustrated
texts daily and 13 indicated weekly; 11 read scientic text-
only articles daily and 12 read them weekly. For infographics,
20 read them weekly and 7 daily. Familiarity with data vi-
sualizations was rated dierently (
1 7
). For comic
literacy, 8 read comics daily, 7 weekly, 4 monthly, 9 few times
year, 3 yearly and 5 of them have never read comics with
20 of them having rst read comics aged between 6 and 12
years old. 12 participants had experience in drawing comics,
but mostly only a few times and for less than one year. Par-
ticipants reported on their preferred learning method: verbal
(spoken or written) media (6), visual stimuli (8), combined
visual and verbal media (17). Repetition was a popular strat-
egy mentioned by 8 participants. Participants were paid a
compensation of $7.
Procedure
Our study employed a mixed design. For both stories and
techniques we employed a within-design, i.e., every partici-
pant read all stories and all formats. However, as we could
not provide the same story twice to the same participant, we
employed an in-between design for story-format conditions,
i.e., every participant read each story using a dierent format.
We used Latin-square randomization to assign stories to for-
mats. With 36 participants each seeing 3 presentations, each
of the 9 story-format combinations (3 formats
×
3 stories)
was seen by 12 participants. The experiment was divided
into three stages, taking place on dierent days. The rst
part (lasting between 45 and 55min) included the following
steps: (1) Upon arrival, participants lled out a background
questionnaire asking for, e.g., familiarity with comics and
visualization. Participants sat in a quiet room with the in-
structor. Material was printed and placed on the table, one
at a time. (2) Participants read each story in a xed order
(Energy,Alliances,Economics), each time with a dierent
format. Participants were told to read comfortably without
rush but to try to not spend more than 5min reading, and that
they were asked questions about their understanding after
reading. (3) Immediately after reading a story, participants
were asked to re-tell the story with their own words. Read-
ing and retelling was recorded through a video camera, not
including the participant’s face. Then (4), they were given a
questionnaire with 6 multiple choice questions to assess par-
ticipants’ understanding (see next subsection). Eventually, (5)
participants were asked to explain their reading strategy for
each story in the format they had viewed, by drawing arrows
onto the printed material. Participants came back the next
day for their 2nd session, lasting between 20 and 30min each.
For each story, participants were asked to (6) retell the story
with their own words to assess recall. Then, (7) they were
presented with all the materials (Comic,IllustratedText,
Infographic) for each story—including the material they
had not seen during their reading session—and were asked
for their preferences through another questionnaire and then
(8) were asked additional feedback on all three techniques
in a brief semi-structured interview. Eventually, (9) 1 month
after these sessions, we emailed participants the multiple-
choice questionnaire form (4) to further help assess recall.
Participation was rewarded with the possibility of winning
a $25 Amazon voucher.
Understanding estions
In step (4) participants answered 6 questions, measuring their
understanding of the story content. Each question provided
4 answer options plus an “I’m not sure” option. Questions
were naturally dierent for each story, yet for each story
we created one question covering the following information
types:
distribution
(e.g., “What is the geographical distribu-
tion characteristics of renewables production in EU?”),
time
(e.g., “During what time did African countries suer from cri-
sis”),
single fact
(e.g., “The tax burden reduced in which coun-
try due to family planning according to this story?”
outliers
(e.g., “Which King was isolated when the Three King Alliance
was created in 772?”),
comparison
(e.g., “Compared to 1990
to 2000, African countries tend to have (?) interest rate and
(?) tax burden from 2000 to 2015.” higher / lower), and
visual
encoding
(e.g., “Which colour is used to present countries in
Latin America”).
Figure 5: Results for (top) understanding including means
and 98% CIs and (boom): overall eect sizes with 98% CIs.
5 LAB STUDY RESULTS
All 36 participants completed the study, with each session
lasting just below 1h. For each of the 9 story-format com-
binations, we obtained true or false answers to 6 questions
from 12 participants. We collected all video materials and
annotated print outs from each participants, available in
https://datacomics925658343.wordpress.com/study.
Understanding
Understanding was measured as accuracy for the questions
from the questionnaire in step (4). Error per question was
binary, i.e., participants selected the right or a wrong answer.
For each participant we calculated the mean accuracy score
across all the answers per story. Using D’Agostino’s K-
sqared test, we found accuracy scores for two stories to
be not normally distributed. Below and in Figure 5 we report
on mean values, condence intervals (CIs), and eect sizes.
Using Bonferroni correction for multiple comparisons, we
report on 98% CIs for our three comparisons (1
−
0
.
05
/
3).
P-values are indicated for pair-wise comparisons yielding
signicance at the respective 0.02 level or close, using a
Mann-Whitney-U test.
Across Stories
we found Comic (mean=.70) to be more
accurate than Infographic and IllustratedText (p
<
.006).
No dierence was found between IllustratedText and
Infographicwith the same mean accuracy of .57. Eect sizes
between techniques are reported in Figure 5-bottom. Our
eect sizes represent the overall improvement (or decrease)
in understanding averaged for each participant, i.e. for each
participant we calculated dierences between formats and
averaged these values. Dierences between Comic and the
other techniques amount to .13 points in understanding,
implying that comics are on average 23% more accurate than
both IllustratedText and Infographic.
Figure 6: Mean results for understanding after one 1-4 weeks.
Upper numbers indicate original values (as shown in Fig 5),
lower numbers indicate correct results after 1-4 weeks with
change in percent (Comic=blue, IllustratedText=yellow,
Infographic=red).
For
Energy
we found Comic (mean=.79) more accurate
than Infographic (mean=.68). However, leading to the same
mean accuracy than Infographic,IllustratedText was
not found dierent from the other two techniques and showed
much wider CIs. For one question, we could nd a real dif-
ference between techniques; the question asked for spatial
distribution of countries. As this information was highlighted
explicitly in one panel in the comic, we believe it was easier
for participants to understand and remember.
For
Alliances
we could not nd any clear dierences,
while Comic (mean=.52) was still the most accurate on aver-
age but with largely overlapping CIs. Infographic and Illus-
tratedText had similar mean values (.42 and .43). Again,
we found a dierence for one question asking about the
meaning of a specic visual encoding in the visualization
(“What does the black dashed relation represent?”); Comic was
signicantly worse (mean=.25) than the other two formats
(IllustratedText=.66, Infographic=.83). We believe par-
ticipants overlooked this information in the comic as they
might have paid less attention to this particular—not explic-
itly highlighted—detail. We attribute the good performance
of Infographic to the fact that a respective text was placed
close to these lines to explain their meaning.
For
Economics
we found similar results to Energy;Comic
showed a slightly higher accuracy (mean=.79) while Info-
graphic(mean=.61)(p
<
.036) and IllustratedText (mean
=
.62) (p
<
.04) were similar. As with the other stories, we found
a dierence for a question on time (“What is the charac-
teristic of the countries in 1962?”). Here, Comic was more
accurate (mean=.91)(p
<
.015) then any of the other formats
(IllustratedText=.58, Infographic=.33). We conjecture
that the bad performance of Infographic can be related,
again, to the missing temporal visualization, while Comic
showed that information in a single panel. While the explana-
tion should indicate a similar bad performance for Illustrat-
edText, the average here was slightly higher. However, this
particular infographic has been described as very cluttered,
which may explain some of the poor performance.
Figure 7: Examples of users’ traces while reading the same
story (Alliances) in dierent formats. Percentages in
brackets indicate the fraction of users showing comparable
traces. Left: IllustratedText (67%), center: Infographic
(75%), right: Comic (100%).
Recall
After reading the stories on the rst day, participants were re-
quired to retell the story in their own words. We took notes of
each explicitly presented fact they mentioned (11 for Energy,
14 for Economics, and 17 for Alliances). Self discovered
messages, i.e., those not being mentioned in the texts, were
not considered. On the second day, participants were asked
to again retell the story with their own words. Following
the methodology by Bateman et al. [
7
], we coded the dier-
ence between both versions as follows: 3 points for every
correct fact (e.g., correct values), 2 points for remembering
general trends (e.g., increase, type of temporal change), 1
point for vaguely remembering (e.g., mentioning type of in-
formation) and 0 points for not or wrongly recalling. Two of
the authors independently coded all of the recordings, then
discussed until reaching consensus. We found Comic to yield
slightly more precise results on average (35%), compared to
IllustratedText (32%) and Infographic (30%).
After one to four weeks, we sent the questionnaire from
step (4) again to all participants (Figure 6): Overall, Comic
caused participants to remember most on average (55%, down
from 75%), followed by Infographic (43%, down from 57%)
and IllustratedText (41%, also down from 57%). Partici-
pants lost around 1/3 of their performance. However, values
varied across stories, with each format performing best for
one story. In two cases, mean understanding rates did re-
main the same ( IllustratedText for Energy and Comic
for Economics).
Reading Strategy
By asking for the reading sequence of the story, we simu-
lated an "eye tracking" phase (step (9)) in our study. In the
pilot study, we found participants walking us through their
reading order more accurate and informative for our pur-
pose than actual eye-tracking. Figure 7 shows example traces,
drawn by participants, while explaining their reading order.
With Comic, all participants followed the panel order, read-
ing both text and looking at gures in each panel except 1
participant who read the Comic of Energy in a right-left or-
der from the second row. 16 participants (44.4%) jumped back
to previous panels to make comparisons. For Infographic,
Figure 8: Subjective measurement from the lab study in step
(2) of the 2nd session including means and 98% CIs. Distri-
butions of the answers are shown on the left.
most participants (83%) started with title and abstract, the
others started with gures. We found no specic sequence
in which participants read text-boxes in Infographic, i.e.
no prevalence for left-right or top-down order. We found
reading sequence was guided by the layout of text boxes,
e.g., clock-wise or randomly in Economics and Alliances.
Generally, participants were guided by text boxes and only
then looked at the gure to obtain more information. All par-
ticipants checked the legend in Alliances and Economics.
In IllustratedText, most participants (83%) started with
the text, few (17%) started with pictures. 28% checked gures
in the end, while a majority (72%) checked the gures dur-
ing reading (2-3 times), especially when they found specic
values in the text.
Subjective Feedback
Subjective results from the preference questionnaire (step
7) are summarized in Figure 8, following the same conven-
tions and analysis as Figure 5. We found Comic to be highest
rated (averages) on three measures (fun,engaging, and al-
lowing people to stay focused) while scoring slightly less
then Infographic for aesthetics and exploration. While, dif-
ferences between Comic and Infographic are minor and
not signicant, Comic is generally ranked higher as shown
in the respective distribution (small bar charts left side of
Figure 8). IllustratedText was rated generally least across
all measures with huge dierences to the two other tech-
niques for the measures engaging and fun.Infographic was
rated as aesthetically pleasing, with opinions being more
equally distributed for Comic. Asked which format partici-
pants would chose next time, we found similar results: Comic
rated best by 47% (
worst best
), IllustratedText rated
least by 64% (
worst best
) and Infographic in the
middle (
worst best
). In the following, we report on
participants’ feedback during the interview in step (8). Fre-
quencies of issues raised are reported by the numbers in
brackets. Full material and an overview table with the re-
ported merits and drawbacks of each technique can be found
in https://datacomics925658343.wordpress.com/study/.
Comic
—Comics were appreciated for their clear reading
order (33%) and their ability to break down the complexity
into pieces (28%). Reading order was found use support mem-
ory (6%)(“[...] because the important information is repeated all
the time, which helps [...] memorize.”,“Comic makes a story in
your head”) and were found to facilitate understanding: “just
follow the sequence. It is logic and well organized”, especially
for temporal content (39%) (“just because [Alliances] is so
complicated and chronological”). Breaking down complexity
into individual pieces has been found useful for the same
reasons (“You have the option to see the information by steps,
you can easily have your memory when it is happened (sic).
It is like the same way we remember history when I was a
child.”“is quite easier to memorize”. Additionally, participants
liked that comics could group higher-level messages into
rows and potentially pages. Participants also commented on
the ability to quickly overview information and nd/recall
the information they wanted (“If I don’t have time, I’ll go for
data comic”). In fact, during recall, 4 participants used their
hands to air-point to where the respective panel was located.
On the downside of Comic, visual repetitions were dis-
tracting for 2 participants, indicating too much information
to process: “every time I see new pictures (panels) I expect
something new”,“I need to compare to dene what is new”.
There is a tension here, as building a new message is on the
basis of previous messages uses repeated visual elements,
causing redundancy. For Energy, one participant mentioned
“I think it is unnecessary to have that kind of level to break
down, because it is simple enough to understand the image by
just looking at the whole picture and explore it.”. This story
indeed featured a high degree of visual repetition (the map
of Europe) as it was hard to break down the content into
simpler images. Again, other people explicitly preferred a
more open format (“I can only read the comic step by step, it
is hard to nd the part that I am interested in”) and found
that an overview picture was missing. The same participant
was found to jump between panels. Looking at the comic the
rst time, one participant noted “This one can be confusing,
considering there are all types of graph. It seems a little much
at rst glance”.
Infographic
—Infographics were rated highly for explo-
ration (33%) by providing overview and detail at the same
time which helped to make comparisons (6%). Some partic-
ipants (31%) liked the strong connection between text and
picture. (19%) participants found Infographic easier for spa-
tial relations and (17%) mentioned understanding time was
harder. (28%) found a clear reading order lacking.
IllustratedText
—Some participants liked the clean-
ness and familiarity of IllustratedText (11%) and reported
that they would use the text to understand the story, and
could look up information on demand. However, the major-
ity found jumping between text and picture negative (42%) as
they had to create their own connections (14%), complaining
about the high density of the text (31%) (“Along with the high
density of text, and hard time of bridging verbal and visual”).
Summary
Our lab study yielded consistent and conclusive results: comics
helped with understanding while there was no dierence
between the other two formats, thus partially conrming H-
Accuracy. Dierences across questions highlighted specic
aspects of each format. Eventually, all participants found
comics more engaging, thus mostly conrming H-Engaging.
Subjective feedback revealed more precise information about
the respective advantages and drawbacks of each format. We
could not nd evidence to fully support H-Memorability.
While this evidence was collected in a controlled lab set-
ting, the question remains of how comics and infographics
compare in more ecologically valid settings, and how well
each format is able to keep readers engaged who are not
paid and asked to answer questions after reading. For ex-
ample, our comics did not involve characters or elaborate
artistic styles, nor could they feature actual topics and infor-
mation. To complement our initial results, we thus designed
a in-the-wild study.
6 IN-THE-WILD STUDY
The in-the-wild study was carried out during an interna-
tional art festival where visitors varied in age, interest, and
cultural origin, enabling us to study a more diverse set of
participants compared to using standard mailing list recruit-
ing. The study focused on the attraction of each format, how
people engaged with each format and for how long, and their
preferences for consuming information. To that end, we ob-
served and coded people’s behavior as well as conducting
semi-structured interviews with some of the readers.
Techniques
—We only used Infographic and Comic in
this study. Two comics were created from existing infograph-
ics, using the same process as the lab study. Presented topics
were chosen for public interest, but unrelated to any of the
surrounding exhibitions to mitigate potential confounds on
the respective audience attracted by our visualization. To
increase attractiveness of the material for the public space,
we were more free with visual presentation. Comics were
drawn by hand by an experienced comic artist and co-author
of this paper. The basic visual encodings in the infographic
such as colors, height and size for data variations were pre-
served, while also using metaphors and sketch styles from
data comics (Figure 1). We used material on two subjects:
•Hot spots—The Carbon Atlas (CarbonFootprint)
[
39
]
shows global carbon dioxide emission by country, using
bubbles of dierent size to indicate the amount of pro-
duced carbon dioxide, colored by region.
•The Global Water Print (WaterUsage)
[
24
] indicates
the volume of water needed for production and human
services, showing of amount of freshwater available, the
highest consumption, renewable water sources and the
highest usage in food production.
Study Design
—An A0 sized copy of an Infographic
about CarbonFootprint and a Comic about WaterUsage
were placed on a pair of easels in the street, with take-away
copies of the comics and Infographic attached (Figure 1).
After 1h, we exchanged the respective material to show an
Infographic about WaterUsage and a Comic about Car-
bonFootprint. The study was conducted twice on the same
Saturday, once between 11:30am-1:30pm and again from
2pm-4pm. Two study instructors were seated approx. 10
meters distant from the easels, tracking audience behavior.
Visitors who read both formats were approached for an in-
terview after they turned away from the graphics.
Data Collection
—The time visitors spent at each for-
mat was measured, with talking and pointing interactions
counted manually. When we interviewed visitors, we asked
four questions: why did they stop? Did they nish reading?
Was there anything they did not understand? and Which
format they found easiest to understand?
Results
During the 4 hours a total of 43 groups stopped for more
than 10 seconds to view the graphics. Most
visitors
came
in groups of 2 to 4 with a wide age range from adolescent
to elderly. From this, we interviewed 8 groups including 14
people of which 11 were adults, 2 elders and 1 adolescent.
We did not nd any dierences in time spent on either
format or story. CarbonFootprint comic was read 13 times
while Infographic was read 15 times; For WaterUsage , both
the comic and infographic format was read 14 times. Viewers
spend between 8 and 132 seconds reading each story, with
averages ranging between 37 and 48 seconds across formats.
We did not nd a signicant dierence in reading time.
Counting visitors
interactions
with the material through
pointing and discussing each format, we found slightly more
engagement with Infographic (8 groups for Comic vs. 12
groups for Infographic). This could be seen as an indicator
that the more exploratory nature of infographics prompts
people wanting to share and discuss their observations, while
comics focus the reader on understanding the message.
Asked about what attracted them to our graphics in the
rst place, 4 groups replied topic, 1 replied color, 2 replied
illustration in the comic, 1 replied data. Asked about which
format provided
better understanding
of the content, 6
groups preferred Comic, 1 group preferred Infographic and
1 group didn’t nd any dierence. Reasons for preferring
comics the way it visualized, the grouped messages, the
metaphor, and the easy-to-follow layout. Consequently, 12
people reported that Comic were their overall choice, while 1
adult and one adolescent from dierent groups would chose
Infographic (WaterUsage). This supports the evidence
from the Lab Study results, that readers nd comics more
“fun and enjoyable” even if this may not necessarily translate
to longer reading time or improved memorability.
Participants took away slightly more copies of Carbon-
Footprintin comic format compared to infographics format
(20 vs 16). However, this is not the case for WaterUsage, for
which participants took 16 copies of each.
7 DISCUSSION
Main Study Findings
Data Comics improve understanding and engagement
—
In general, data comics led to more correct answers on aver-
age. Data comics have been rated more engaging and more
enjoyable, more easy to stay focused, and received better
overall ratings. The reasons for these results may be ex-
plained by a variety of factors. For example, clear sequencing
increases the readers ability to focus and navigate spatial-
temporal information, while panels help to divide informa-
tion into easily memorable chunks, with rows grouping indi-
vidual messages into higher level messages. From analyzing
specic questions, we found that comics performed better
for some information that was explicitly shown and high-
lighted in panels, such as some temporal events (Alliances)
or distributions (Energy). On the contrary, not highlight-
ing important information in comics, such as visual encod-
ing (Alliances) can lead to participants overlooking details.
These results suggest that increased text-picture integration
and more reading guidance (c.f. the design space in Section
3) can lead to better understanding.
Large text-picture distance impairs understanding
and increases cognitive load
—While illustrated texts were
seen as clean and simple, some participants complained
about the higher cognitive load required by the constant
switching between text and gures. This can explain why
participants preferred formats minimizing that distance.
Infographics foster exploration and overview
—Info-
graphics are well suited to represent spatial content and are
good at delivering both overview and detail. Participants
liked the way they allow for comparison, and were more
likely to want to share their discoveries with other viewers.
We believe techniques from data comics and infographics
can be seamlessly integrated with each other, depending on
data and message.
Designing Data Comics
Our results can be used to discuss and inform the design
of data comic (e.g., [
52
]). Below, we illustrate some of the
complexity of designing good data comics, often requiring
multiple trade-os.
Balancing repetition and highlighting
—while most of
our results point to an increase in understanding with comics,
subjective feedback highlighted potential problems with ex-
cessive repetition and sequencing. Too much (visual) repeti-
tion and redundancy between panels can lead to confusion
as readers struggle to notice the dierences. Possible solu-
tions include explicitly highlighting changes, using a cut-out
pattern [
4
] to emphasize small changes, or combining several
messages into one panel by using elements from infograph-
ics such as annotations. Complicated information could be
explained in a large-picture pattern [
4
] to serve as a mental
map, before individual changes are explained in detail.
Balancing sequence and overview
—Sequences support
temporal and complex causal information, while overviews
support comparison and spatial (non-temporal) information
and help readers to keep a mental map. The lack of overview
has been criticized in data comics, especially if panels show
details of the general visualization (e.g., map, scatter plot).
However, repetition can be distracting, as mentioned above.
A solution could be to carefully pace overview pictures and
to make sure zoomed-in content is understood within the
larger context. Where necessary, larger pictures (especially
for spatial and detailed visualizations) can incorporate ele-
ments from infographics, such as annotations.
Using the layout to structure information
—A comic
layout provides several means to visually structure informa-
tion and the story, using panels, panels inside panels, rows
and potentially pages. Panel size, number and layout [
4
] can
be used to group and relate messages, to pace reading and
attention, as well as to demonstrate importance. A clear page
layout, potentially including overview panels, can support
information look-up and relation during reading.
Reducing visual complexity
—Comics can quickly be-
come visually overwhelming when seen at a glance, as men-
tioned by some participants. While we designed our comics
with this issue in mind, panels full of abstract information
remain a natural source of visual clutter, especially if small.
Possible solutions include creating larger panels (and hence
less panels per page) when panel content gets visually com-
plex. Consistency and repetition of visual information [
37
]
can be another solution to keep the overall visual clutter low,
if the respective changes between panels are highlighted
properly, as mentioned above. Yet, we could not conrm that
visual complexity at a rst glance actually impacts partici-
pants performance negatively.
Limitations and Future work
Type of stories and visualization
—Clearly, the type of
story and visualization presented in the studies may inu-
ence the the reading experience. We chose maps, networks
and scatter plots as representative examples of visualizations.
Future studies need to evaluate whether our ndings hold
for other visualization types including simpler (e.g. bar charts,
line charts) as well as more complicated and less familiar
visualizations such as parallel coordinate plots, matrices or
tree maps. Such types of visualizations require careful expla-
nations to be understood and used in a storytelling context.
We further believe that comics could be successful in achiev-
ing this, given their sequential nature and tight integration
of textual and pictorial information and we see signicant
potential for data comics to explain complex data as well as
contribute to aspects of visualization and data literacy.
Style and design choices
—In creating panels in our data
comics, we made specic design choices. Dierent sequences
and pacing could potentially lead to dierent results [
23
] and
further studies are required, e.g., in order to determine the
appropriate pacing or the amount of redundancy between
text and visualization. Similarly, comics may adopt dierent
drawing styles and visualization strategies. For our lab-study,
we decided on a simple, neutral style that used the same
visuals and colors of the infographics, to allow for a direct
comparison. For the in-the-wild study, we opted for a more
visually elaborate presentation to catch people’s attention
on the street. Hence, Infographic and Comic were slightly
more dierent in their visual appeal (Figure 1). However,
interviews revealed that visitors were attracted mainly by the
presented topics (CarbonFootprint,WaterUsage) rather
than the visuals. It still is possible that introducing more
elaborate drawings, characters and metaphors, will aect
readers’ engagement and attention [18].
Context and Audience
—Our results are naturally lim-
ited by the study context and audience. While the audience
of our in-the-wild study included a wide range of ages, inter-
ests, cultures, and pre-knowledge about visualizations and
the presented topics, people may have been reticent to en-
gage with the material at all, whether due to lack of interest
or external distraction. Feedback and insights into people’s
behaviour may also vary in other contexts (e.g., students
focused on studying with textbook) but our setup appears
to align with the general public’s consumption of infograph-
ics, as echoed by online news article reading behaviors such
as reported in Amanda Cox’ talk at IEEE VIS 2011 about
consumption of New York Times online articles.1
1
http://ieeevis.org/year/2011/keynote/visweek/how-editing-and-design-
changes-news-graphics
Story Formats
—Our design space in Section 3 was cho-
sen to motivate and structure our study. For both axes dif-
ferent solutions are possible and there is no unique measure
that locates a specic solution. In designing our comics, we
made certain choices in panel layout, sizes, style, message
chunking, highlighting, text and picture redundancy, which
might have had an impact on our results. Moreover, in our
rst experiment (Section 4), story and texts were the same
across techniques, in real settings one could be more specic
and add more text in infographics. Eventually, as pointed
out by Bach et al. [
4
] data comics and infographics span a
continuum with terminological and conceptual boundaries
not clearly dened. Given our design space, we can adopt
the same argument of uidity for illustrated texts [
3
]. For
example, text plus pictures can include several pictures, pic-
tures can be linked to places inside the text, infographics
can contain several pictures, can have more or less linearity,
even involve features from comics. Thus, rather than compar-
ing prototype formats, we compared locations in our design
space. Our study aims to provide some clarity about the usage
and impact of each dimension on understanding and engage-
ment. Future designs should take the best of both worlds, i.e.,
using sequential, narrative, and metaphorical elements from
comics, combined with exploratory and image-focused ele-
ments from infographics. Gaining more insights into design
decisions will be the major challenge for future studies.
Finally
, infographics, illustrated texts and comics can be
compared along other dimensions and for dierent contexts:
picture size, type of visualization and information, audience
and etc [
22
]. Eventually, comparison with other formats for
data-driven storytelling (videos, interactives, physicaliza-
tions, etc. [
27
]) can yield more insights in the respective
drawbacks and merits of each format. Better understanding
of the potentials of data comics will lead to better authoring
support and education.
8 CONCLUSION
The new genre of data comics combines many features with
the potential of making data-driven stories accessible and un-
derstandable. In order to verify the comics’ eectiveness on
reader’s understanding, memorability and engagement, we
conducted two experiments comparing data comics with in-
fographics and illustrated texts. Our results are encouraging
for the use of data comics, especially for complex spatio-
temporal data, which are naturally hard to visualize in info-
graphics. Our results also lead to valuable implications for
designing future comics.
ACKNOWLEDGEMENTS
We wish to thank all the participants in this study, including
the visitors in the in-the-wild study, and the colleagues and
reviewers for their feedback.
REFERENCES
[1]
Paul A Aleixo and Krystina Sumner. 2017. Memory for biopsychology
material presented in comic book format. Journal of Graphic Novels
and Comics 8, 1 (2017), 79–88.
[2]
Benjamin Bach, Natalie Kerracher, Kyle Wm Hall, Sheelagh Carpendale,
Jessie Kennedy, and Nathalie Henry Riche. 2016. Telling stories about
dynamic networks with graph comics. In Proceedings of the 2016 CHI
Conference on Human Factors in Computing Systems (CHI). ACM, 3670–
3682.
[3]
Benjamin Bach, Nathalie Henry Riche, Sheelagh Carpendale, and
Hanspeter Pster. 2017. The Emerging Genre of Data Comics. IEEE
computer graphics and applications 38, 3 (2017), 6–13.
[4]
Benjamin Bach, Zezhong Wang, Matteo Farinella, Dave Murray-Rust,
and Nathalie Henry Riche. 2018. Design patterns for data comics. In
Proceedings of ACM SIGCHI Conference on Human Factors in Computing
Systems (CHI). ACM, 38.
[5]
Benjamin Bach, Zezhong Wang, Matteo Farinella Nathalie Henry Riche,
Dave Murray-Rust, Sheelagh Carpendale, and Hanspeter Pster. 2018.
online: retrieved from http://datacomics.net.
[6]
Alan D Baddeley. 1997. Human memory: Theory and practice. Psychol-
ogy Press.
[7]
Scott Bateman, Regan L Mandryk, Carl Gutwin, Aaron Genest, David
McDine, and Christopher Brooks. 2010. Useful junk?: the eects of
visual embellishment on comprehension and memorability of charts. In
Proceedings of the SIGCHI Conference on Human Factors in Computing
Systems (CHI). ACM, 2573–2582.
[8]
Arnd Bernaerts. 2006. Booklet on Naval War changes Climate. (2006).
[9]
Rebecca Bertram and Radostina Primova. 2018. Energy Atlas 2018: Fig-
ures and Facts about Renewables in Europe. Heinrich Böll Foundation.
[10]
John B Black and Gordon H Bower. 1979. Episodes as chunks in
narrative memory. Journal of verbal learning and verbal behavior 18, 3
(1979), 309–318.
[11]
Michelle A Borkin, Azalea A Vo, Zoya Bylinskii, Phillip Isola, Shashank
Sunkavalli, Aude Oliva, and Hanspeter Pster. 2013. What makes
a visualization memorable? IEEE Transactions on Visualization and
Computer Graphics 19, 12 (2013), 2306–2315.
[12]
Paul Chandler and John Sweller. 1991. Cognitive load theory and the
format of instruction. Cognition and instruction 8, 4 (1991), 293–332.
[13]
James M Clark and Allan Paivio. 1991. Dual coding theory and educa-
tion. Educational psychology review 3, 3 (1991), 149–210.
[14]
Neil Cohn. 2013. The Visual Language of Comics: Introduction to the
Structure and Cognition of Sequential Images. A&C Black.
[15]
Neil Cohn. 2014. The architecture of visual narrative comprehension:
the interaction of narrative structure and page layout in understanding
comics. Frontiers in Psychology 5 (2014). https://doi.org/10.3389/fpsyg.
2014.00680
[16]
C. Delp and J. Jones. 1996. Communicating information to patients:
the use of cartoon illustrations to improve comprehension of instruc-
tions. Academic Emergency Medicine: Ocial Journal of the Society for
Academic Emergency Medicine 3, 3 (March 1996), 264–270.
[17]
Judy Diamond, Julia McQuillan, Amy N. Spiegel, Patricia Wonch Hill,
Rebecca Smith, John West, and Charles Wood. 2016. Viruses, Vaccines
and the Public. Museums & Social Issues 11, 1 (Jan. 2016), 9–16. https:
//doi.org/10.1080/15596893.2016.1131099
[18]
Matteo Farinella. 2018. The potential of comics in science communica-
tion. Journal of Science Communication 17, 01 (2018), Y01–1.
[19]
Tom Foulsham, Dean Wybrow, and Neil Cohn. 2016. Reading without
words: Eye movements in the comprehension of comic strips. Applied
Cognitive Psychology 30, 4 (2016), 566–579.
[20]
Ola Rosling Hans Rosling and Anna Rosling RÃűnnlund. [n. d.]. online:
https://www.gapminder.org, [last accessed: 17 Aug 2018].
[21]
Jay Hosler and KB Boomer. 2011. Are comic books an eective way to
engage nonmajors in learning and appreciating science? CBEâĂŤLife
Sciences Education 10, 3 (2011), 309–317.
[22]
Jessica Hullman and Nick Diakopoulos. 2011. Visualization rhetoric:
Framing eects in narrative visualization. IEEE transactions on visual-
ization and computer graphics 17, 12 (2011), 2231–2240.
[23]
Jessica Hullman, Steven Drucker, Nathalie Henry Riche, Bongshin
Lee, Danyel Fisher, and Eytan Adar. 2013. A deeper understanding of
sequence in narrative visualization. IEEE Transactions on visualization
and computer graphics 19, 12 (2013), 2406–2415.
[24]
US Infrastructure. 2010. online: retrieved from http://aquadoc.typepad.
com/waterwired/water_quantity/page/121/. Original website not avail-
able anymore..
[25]
Nam Wook Kim, Nathalie Henry Riche, Benjamin Bach, Guanpeng A
Xu, Matthew Brehmer, Ken Hinckley, Michel Pahud, Haijun Xia,
Michael McGun, and Hanspeter Pster. 2019. DataToon: Draw-
ing Dynamic Network Comics With Pen + Touch Interaction. In Proc.
of ACM Conference of Human Factors in Computing Systems (CHI).
[26]
Robert Kosara and Jock Mackinlay. 2013. Storytelling: The next step
for visualization. Computer 46, 5 (2013), 44–50.
[27]
Stephanie A. Kraft, Melissa Constantine, David Magnus, Kathryn M.
Porter, Sandra Soo-Jin Lee, Michael Green, Nancy E. Kass, Benjamin S.
Wilfond, and Mildred K. Cho. 2016. A randomized study of multimedia
informational aids for research on medical practices: Implications for
informed consent. Clinical Trials 14, 1 (Sept. 2016), 94–102. https:
//doi.org/10.1177/1740774516669352
[28]
W Howard Levie and Richard Lentz. 1982. Eects of text illustrations:
A review of research. ECTJ 30, 4 (1982), 195–232.
[29]
Huifen Lin and Tsuiping Chen. 2007. Reading authentic EFL text
using visualization and advance organizers in a multimedia learning
environment. (2007).
[30]
Narges Mahyar, Sung-Hee Kim, and Bum Chul Kwon. 2015. Towards a
taxonomy for evaluating user engagement in information visualization.
In Workshop on Personal Visualization: Exploring Everyday Life, Vol. 3.
2.
[31]
Richard E. Mayer and Joan K. Gallini. 1990. When is an illustration
worth ten thousand words? Journal of Educational Psychology 82, 4
(1990), 715–726. https://doi.org/10.1037/0022- 0663.82.4.715
[32]
Scott McCloud. 1993. Understanding comics: The invisible art.
Northampton, Mass (1993).
[33]
Sarah McNicol. 2017. The potential of educational comics as a health
information medium. Health Information & Libraries Journal 34, 1
(2017), 20–31.
[34]
Antoni B. Moore, Mariusz Nowostawski, Christopher Frantz, and
Christina Hulbe. 2018. Comic Strip Narratives in Time Geography.
ISPRS International Journal of Geo-Information 7, 7 (2018). https:
//doi.org/10.3390/ijgi7070245
[35]
Heather L O’Brien and Elaine G Toms. 2010. The development and
evaluation of a survey to measure user engagement. Journal of the
American Society for Information Science and Technology 61, 1 (2010),
50–69.
[36]
Allan Paivio. 1990. Mental representations: A dual coding approach.
Oxford University Press.
[37]
Zening Qu and Jessica Hullman. 2018. Keeping multiple views consis-
tent: Constraints, validations, and exceptions in visualization author-
ing. IEEE Transactions on Visualization and Computer Graphics (TVCG)
24, 1 (2018), 468–477.
[38]
Nathalie Henry Riche, Christophe Hurter, Nicholas Diakopoulos, and
Sheelagh Carpendale. 2018. Data-driven Storytelling. CRC Press.
[39]
Simon Rogers and Mark Mccormick. 2010. online: retrieved
from http://image.guardian.co.uk/sys-les/Guardian/documents/2007/
12/17/CARBON_ATLAS.pdf. Original website not available anymore..
[40]
Gertjan Rop, Anne Schüler, Peter PJL Verkoeijen, Katharina Scheiter,
and Tamara Van Gog. 2018. The eect of layout and pacing on learning
from diagrams with unnecessary text. Applied Cognitive Psychology
(2018).
[41]
Hans Rosling. [n. d.]. Some cool motion sensor stu.
https://www.ted.com/talks/hans_rosling_shows_the_best_stats_
you_ve_ever_seen#t-230391
[42] Bahador Saket, Alex Endert, and John Stasko. 2016. Beyond usability
and performance: a review of user experience-focused evaluations in
visualization. In Proceedings of the Sixth Workshop on Beyond Time and
Errors on Novel Evaluation Methods for Visualization. ACM, 133–142.
[43]
Wolfgang Schnotz and Maria Bannert. 2003. Construction and interfer-
ence in learning from multiple representation. Learning and instruction
13, 2 (2003), 141–156.
[44]
Edward Segel and Jerey Heer. 2010. Narrative visualization: Telling
stories with data. IEEE transactions on visualization and computer
graphics 16, 6 (2010), 1139–1148.
[45]
Jeremy C. Short, Brandon Randolph-Seng, and Aaron F. McKenny.
2013. Graphic Presentation An Empirical Examination of the Graphic
Novel Approach to Communicate Business Concepts. Business Com-
munication Quarterly 76, 3 (Sept. 2013), 273–303. https://doi.org/10.
1177/1080569913482574
[46]
Nick Sousanis. 2015. Unattening. Harvard University Press, Cam-
bridge, Massachusetts.
[47]
Amy N Spiegel, Julia McQuillan, Peter Halpin, Camillia Matuk, and
Judy Diamond. 2013. Engaging teenagers with science through comics.
Research in science education 43, 6 (2013), 2309–2326.
[48]
Nicole Sultanum, Michael Brudno, Daniel Wigdor, and Fanny Chevalier.
2018. More Text Please! Understanding and Supporting the Use of
Visualization for Clinical Text Overview. In Proceedings of the 2018 CHI
Conference on Human Factors in Computing Systems. ACM, 422.
[49]
M Tatalovic. 2009. Science comics as tools for science education and
communication: a brief, exploratory study. Jcom 8, 4 (2009), A02.
[50]
Redda Tekle-Haimanot, Preux Pierre-Marie, Gerard Daniel, Dawit Ki-
bru Worku, Hanna Demissie Belay, and Meron Awraris Gebrewold.
2016. Impact of an educational comic book on epilepsy-related knowl-
edge, awareness, and attitudes among school children in Ethiopia.
Epilepsy & Behavior: E&B 61 (Aug. 2016), 218–223. https://doi.org/10.
1016/j.yebeh.2016.05.002
[51]
Barbara Tversky. 2011. Visualizing thought. Topics in Cognitive Science
3, 3 (July 2011), 499–535. https://doi.org/10.1111/j.1756-8765.2010.
01113.x
[52]
Zezhong Wang, Harvey Dingwall, and Benjamin Bach. 2019. Teaching
Data Visualization and Storytelling with Data Comic Workshops. In
Proc. of ACM Conference of Human Factors in Computing Systems (CHI),
Extended Abstracts.
[53]
Gene Yang. 2008. Graphic Novels in the Classroom. Language Arts
85.3 (2008), 185.
[54]
Zhenpeng Zhao, Rachael Marr, and Niklas Elmqvist. 2015. Data Comics:
Sequential Art for Data-Driven Storytelling. Tech. report (2015).