Conference PaperPDF Available

An Exploration of the Effectiveness of Infographics in Contrast to Text Documents for Visualizing Census Data: What Works?


Abstract and Figures

The U.S. Census Bureau conducted a usability research study using 3 published Census infographics on different topics containing statistics produced from Census survey data. The study used a mixed-factorial design with repeated measures, comparing user performance (accuracy of information recall and accuracy of searching for information) and satisfaction after using either the 3 infographics, or 3 text documents containing the same information. 55 participants were randomly assigned to either the Infographics condition or the Documents condition, accordingly. Results revealed significant advantage in accuracy for participants with any amount of college education, and a significant interaction between the stimulus type (infographics compared to documents) and the condition (searching versus recall) with infographics being related to increased performance in the searching task and not the recall tasks. However, there were no significant memory-related advantages for infographics above documents, and no interaction between education and stimulus type.
Content may be subject to copyright.
An Exploration of the Effectiveness of Infographics in
Contrast to Text Documents for Visualizing Census Data:
What Works?
Marylisa Gareau, Rebecca Keegan, Lin Wang
U.S. Census Bureau, Suitland, MD, United States of America
{marylisa.gareau, rebecca.keegan,}
Abstract. The U.S. Census Bureau conducted a usability research study using 3
published Census infographics on different topics containing statistics produced
from Census survey data. The study used a mixed-factorial design with repeat-
ed measures, comparing user performance (accuracy of information recall and
accuracy of searching for information) and satisfaction after using either the 3
infographics, or 3 text documents containing the same information. 55 partici-
pants were randomly assigned to either the Infographics condition or the Doc-
uments condition, accordingly. Results revealed significant advantage in accu-
racy for participants with any amount of college education, and a significant in-
teraction between the stimulus type (infographics compared to documents) and
the condition (searching versus recall) with infographics being related to in-
creased performance in the searching task and not the recall tasks. However,
there were no significant memory-related advantages for infographics above
documents, and no interaction between education and stimulus type.
Keywords: Usability, Infographics, Memory
1 Introduction
Infographics represent an emerging medium of information visualization and statisti-
cal communication. In an era of rapidly increasing availability of information, in-
fographics can attenuate the effects of information inundation by presenting infor-
mation in a way that is quickly and easily digestible, as well as engaging and attrac-
tive. Some studies show that the human visual system can process and assign mean-
ing to imagery in less than a second, which can be much faster than reading through a
text explanation [1, 2]. Other research suggests that for online content, people read
less than 28% of the words presented to them [3]. Thus, increasing imagery and using
words sparingly, as in an infographic, seem like obvious choices.
At the Census Bureau, infographics are created with the intent to convey data about a
topic or concept in a way that is appealing and easily interpreted by a wide, public
adfa, p. 1, 2011.
© Springer-Verlag Berlin Heidelberg 2011
audience, especially those who may not have a formal background in statistics, or
high level of education. Dowse, R. & Ehlers, M. (2005) found that medicine labels
containing text and visual representations were interpretable by 95% of subjects as
opposed to labels with text alone for which the interpretability rate was much lower
(70% of subjects) [4].
There is a lack of published research pertaining to the effectiveness of infographics as
an information visualization medium for complex statistics, and infographics can vary
wildly in terms of composition, imagery, length, density, technical terms, use of tradi-
tional graphs, and content.
Currently, while Census data is available to the public, accessing these statistics re-
quires querying complex data tables. If statistics on more than one aspect of a partic-
ular topic are desired, this can require multiple queries. This can be difficult and
time-consuming for a member of the general public, who may have no knowledge of
the structure of Census data. Census infographics allow for some comparison of mul-
tiple statistics on a particular topic to be presented without the need for querying or
understanding the structure of Census data. Unlike data tables, infographics also
allow visualization of the data (e.g. graphs) and imagery that may be engaging and
related to the topic (iconic imagery). However, no research has yet been done to as-
sess whether these presentations are effective and understandable for members of the
general public.
The objective of the present research is to begin examining what makes an infograph-
ic effective at communicating statistical data and appealing to readers by measuring
memorability, searchability, time required to read the infographic, understandability
of language, use of imagery and graphs, user satisfaction data, and eye-tracking anal-
yses. Results will be used to inform the development of future Census infographics
and further research in the area of information visualization for a public audience.
2 Methods
A usability research study was conducted using 3 published infographics from the
U.S. Census Bureau on different topics containing statistics produced from Census
survey data. The study use a 2x3x3 mixed factorial design, comparing user perfor-
mance and satisfaction after using either the 3 infographics, or 3 text documents con-
taining the same information.
Between-subjects variables are “Condition” (2 levels, Infographics and Text Docu-
ments) and “Education” (3 levels, Highschool, Some College, Bachelor’s Degree).
Within-subjects (Repeated Measures) variables are “Stimulus” and “Response Condi-
tion.” Stimulus has three levels: Memorial Day, Home Improvements, Child Care.
Response Condition has three levels: Free-Recall, Multiple-Choice, Search.
Dependent Variables: Accuracy scores for each combination of the Within-Subjects
factors (e.g. “Memorial Day Free-Recall Score” is one dependent variable). Accuracy
was measured on a scale of 0 to 10 and was determined based on a pre-established
rubric. Ranges of acceptable values for Free-Recall responses were established prior
to the study, and partial credit was awarded in multi-part questions.
2.1 Participants
A total of 55 participants were randomly assigned to either the Infographics condition
or the Documents condition, accordingly. Participants were recruited from a pool of
volunteers maintained by the Human Factors and Usability Group at the U.S. Census
Bureau. Participants were recruited based on the following criteria:
Participant is 18 years of age or older.
Participant has completed at least 3 years of high school.
Participant has completed no more than a Bachelor’s degree.
Participant speaks English.
Sessions lasted one hour and participants were given a $40 dollar honorarium. Partic-
ipant characteristics appear in Table 1 below.
Table 1. Participant Characteristics, n=55.
Total Participants
Age (in years)
Mean (SD)
42.5 (16.5)
18 - 74
High school diploma or less
Some college
Bachelor’s degree
2.2 Procedure
Participants were randomly assigned to either the infographics group (ni=28), or the
text documents group (nt=27).
Three Census infographics and their corresponding text-documents were chosen
based on length of the infographic, iconic value of imagery and density of infor-
mation. Iconic value of imagery refers to the extent to which the imagery in the info-
graphic was abstract (such as differently sized boxes representing varying statistical
values) or iconic (imagery relates directly to the topic, e.g. drawings of people to rep-
resent quantities of people). The infographics chosen were of the following topics:
Memorial Day (short length, low iconic value, high density), Home Improvements
(short length, high iconic value, low density), and Child Care (longer length, moder-
ate iconic value, moderate density).
Each participant was verbally introduced to the purpose of the study, signed a consent
form, and was calibrated for eye-tracking.
For the text condition, the publicly available blind-accessible document version of
each infographic was used. The text-only documents contained the same data and
text descriptions of the data as their corresponding infographics, but contained no
imagery or visualizations. An example of a Census infographic section with its asso-
ciated text-document section appears in Figures 1 and 2, respectively.
Fig. 1. Section from the Memorial Day infographic
Fig. 2. Section from the Memorial Day infographic blind-accessible text-document
To examine possible effects of infographics on the memorability of the information,
each participant was presented with a stimulus (either in infographic or text form) and
instructed to read it. Then they responded to 10 factual questions about the content of
the stimulus, first in open-format response (Free Recall condition), then the same 10
questions in multiple-choice format (Multiple Choice condition). This was repeated
for each of 3 stimuli.
To examine searchability, each stimulus was then presented again simultaneously
with the same 10 questions from the prior two conditions, but the participant was
instructed to search the stimuli for the correct answers (Search condition). Finally,
each participant was asked debriefing questions related to their satisfaction with dif-
ferent aspects of the stimuli.
2.3 Hypotheses
It was predicted that infographics would be associated with higher accuracy scores for
both memorability (as measured by Free-Recall and Multiple-Choice accuracy scores)
and searchability (as measured by accuracy scores in the Search condition) as com-
pared with the text documents. Because participants were allowed as much time as
they needed to examine the stimuli, it was predicted that the length of time spent ex-
amining the stimuli would be associated with higher accuracy scores.
Because Census infographics report statistical data, it was predicted that higher edu-
cation level would be positively associated with overall accuracy scores. However,
because of the presence of visualizations and imagery, it was predicted that in-
fographics would be associated with higher accuracy scores for those with less educa-
tion than the text documents, and that this effect would not be present for those with
higher education level.
2.4 Apparatus
Tobii x120 Monitor
Internet Explorer 10
3 Results
3.1 Time Spent on Stimuli
Contrary to predictions, time spent on stimuli was not found to be significantly corre-
lated with accuracy scores (r=.04, p=.52).
3.2 Accuracy
A 2x3x3 mixed ANOVA was performed to test the main effects and interactions of
Condition, Education, Response Condition, and Stimulus on the accuracy scores for
the recall question sets.
Mauchly’s Test of Sphericity was not significant for the within-subject factor “Stimu-
lus,” but was significant for the within-subjects factor “Response Condition”
(X2(2)=21.10, p<.001), and for the interaction between Stimulus and Response condi-
tion (X2(9)=29.03, p=.001). Therefore, the Greenhouse-Geisser correction was used
to interpret significance for F values related to these factors (ε=.73, ε=.77, respective-
ANOVA results revealed a main effect of Stimulus (topic) (F(2, 96)=38.23, p<.001).
Furthermore these differences were significant, as indicated by repeated contrasts
(F(1,48)=37.53, p<.001, F(1,48)=9.65, p=.003). Regardless of experimental group,
Response Condition or Education, accuracy scores for Child Care stimuli were the
highest (M=6.25), followed by Home Improvements (M=5.61), followed by Memorial
Day (M=4.50). In general, the Child Care topic appeared to be the easiest to interpret,
despite it being the longest infographic/document (about 3 times as long as the other
two). Therefore, it was inferred that length of the infographic is not necessarily a
negative factor in the composition of an infographic.
There was a main effect for Response Condition (F(1.47, 70.50)=215.73, p<.001) as
expected. This main effect indicates the overall difference in scores across groups
and education level, with Search having the highest average score (M=7.24), followed
by Multiple-Choice (M=5.37), followed by Free-Recall (M=3.75). Contrasts show
that the differences between these means was significant (F(1,48)=200.22, p<.001,
F(1,48)=82.63, p<.001). Overall, Free-Recall was the most difficult condition, Mul-
tiple-Choice (a form of cued recall) was easier, and finding the answers by searching
the stimulus was easiest, as expected.
There was also a main effect for the between-subjects factor of Education
(F(2,48)=10.36, p<.001), with education level of Bachelor’s degree being highest
(M=6.10), followed by Some College (M=5.92), followed by High school or less
(M=4.34). The Bonferroni post hoc test revealed that the difference in average scores
was significant between Bachelor’s and High school (p<.001), and between Some
College and High school or less (p=.001), but it was not significant between Bache-
lor’s degree and Some College. This overall effect shows the influence of even some
college-level education on the interpretation of the stimuli, because the difference
exists regardless of which stimulus was being viewed, which response condition was
employed, or which type of stimulus (infographic or text document) was being
viewed. Although this effect does not imply causality (i.e. college-level education
cause these participants to perform better), it may be indicative of the difficulty level
of the material being presented.
However, there was no significant interaction, as predicted, between Education and
Condition. The information being presented in infographic form did not offer a dis-
tinct advantage to those of a lower education level in comparison to text documents.
This could be due to the material being difficult enough to understand that imagery
could not benefit the less-educated participants, or perhaps the imagery and visualiza-
tions that were chosen were not particularly effective.
There was a significant interaction between the between-subjects factor Condition
(Infographics or Documents groups) and the repeated-measures factor Response Con-
dition (Free-Recall, Multiple-Choice, Search) (F(2,96)=4.76, p=0.02). Contrasts
show that the interaction was significant between the Multiple-Choice and Search
response conditions (F(1,48)=6.153, p=.017). This interaction shows that Census
infographics did not aid in accuracy for either of the memory conditions, but that they
did improve accuracy for finding information compared to the text documents, re-
gardless of the stimulus topic or the participant education level. The effect, although
significant, represented an advantage of around 10%, or 1 out of 10.
3.3 Eyetracking and Satisfaction
Eyetracking and item-by-item analysis of the recall questions for each stimulus, to-
gether with participant commentary during debriefing led to the identification of di-
mensions of each infographic that were both facilitative for finding information and
detrimental to this task. Major areas of adjustment included size of font, spacing, use
of plain language, and use of iconic imagery. Infographics with adequate spacing,
font size, simple language, and iconic imagery were favored by participants and fa-
cilitated higher accuracy in recall and search task items, and those lacking in these
dimensions were more difficult for participants to interpret. These results are not
reported here.
4 Limitations
This study was partially exploratory in nature, and thus is subject to several limita-
tions, including limitations in design and scope.
4.1 Scope
Perhaps the most obvious limitation of this study is that although we would like to
address the potential benefits of infographics per se, we restricted our investigation to
infographics published by the Census Bureau. Such infographics must necessarily
conform to particular guidelines and are not representative of infographics in general,
such as others that may be produced by other sources or concerning non-statistical
4.2 Design
Participants in this study were asked the same set of 10 “factual” questions three sepa-
rate times, and participants saw each stimulus twice. The second viewing of the stim-
ulus, along with prior knowledge of the questions that would be asked could have
influenced the accuracy scores for the “Search” condition. However it should be
noted that despite having the stimuli present simultaneously with the recall questions
in the “Search” condition (like an open-book test), few participants were able to
achieve a full 10/10 score (8 for Memorial Day, 6 for Home Improvements, 12 for
Child Care, out of 55 participants).
Another design limitation arose from usability. We intended to address regarding how
long participants would spend on the infographics if given no specific direction. Be-
cause we chose to allow participants to spend as much time as they needed to view
the infographics and documents, this aspect of the study was not specifically con-
trolled (i.e. we did not enforce a time limit). However, because education level may
affect the amount of time needed to read the stimuli, and because we wanted to see
possible differences in the behavior and performance of participants of different edu-
cation levels, time was not enforced. In the end, the length of time spent reading the
stimuli was not related to any of the measured factors, including education level, and
is likely attributable to individual differences.
Text documents and infographics were not compared within subjects, so it cannot be
assessed whether an infographic would be more effective than a text document at an
individual level. Learning styles, reading ability, and spatial ability could all play a
role in the interpretation of infographics and text documents.
Finally, the choice to use the blind-accessible documents as a control condition for the
infographics was not truly representative of the differences that may be encountered
when a person tries to extract statistics from Census data tables and compare them. A
more true-to-experience control condition might have been to present participants
with the data tables and require them to compare the data on their own. Providing the
data in the form of blind-accessible text documents allowed us to examine the effect
of the presence of imagery/visualizations, but this may have given the text condition
an advantage over real-life data consumption processes.
5 Discussion
The present study did not provide overwhelming evidence that infographics in and of
themselves increase memory for statistical information, but the results may suggest
that they make information easier to locate when compared to finding the same in-
formation in text documents. This may be due to the imagery being used as a form of
landmark or cue. While the results on accuracy scores may not be compelling on
their own, other usability data was gathered which led to recommendations to the
teams responsible for producing infographics. Thus, the present study combined of
experimental and qualitative data to provide recommendations for the design of future
Census infographics. We list only a few instances of these recommendations.
Some imagery which is very abstract was found to be difficult to interpret and actual-
ly become distracting and confusing to participants. The Memorial Day infographic
was related to less accuracy for the infographics group than for the text-documents
group, which was opposite of the data trend for the other two infographics. The Me-
morial Day infographic used abstract imagery (i.e. differently-sized boxes represent-
ing proportions of soldiers who served versus soldiers who died in each war (Figure
3)) whereas the Child Care and Home Improvements infographics used iconic image-
ry and traditional graphs (e.g. a drawing of a house containing the data for each room
(Figure 4); bar graphs (Figure 5), respectively). This accuracy data was supported by
participants’ verbal feedback that the “boxes” from the Memorial Day infographic
were not understandable (only 3 of the 28 participants in the infographics group could
identify the meaning of the boxes), so a recommendation to use more iconic imagery
to represent data in future infographics was made.
Fig. 3. Abstract “Boxes” from Memorial Day infographic
Fig. 4. Iconic “House Drawing” from Home Improvements infographic
Fig. 5. Traditional bar graphs in Home Improvement infographic
Wording was found through debriefing questions to be an aspect of the infographics
that could have been hindering performance and influencing the effectiveness of the
imagery. Jargon words and complex sentence structure made some graphs difficult
for participants to interpret. Despite the participants saying that they liked the image-
ry and layout of the data in the infographics, they pointed out areas in which the
wording hindered their understanding of the graphs. Two small graphs had titles with
wording that confused participants.
For example, the title of one graph in the Child Care infographic and the text in the
blue bubble next to the graph were confusing to participants due to the lan-
guage/structure (Figure 6). Participants had trouble interpreting these statements and
answering the recall questionnaire question even when they had the infographic in
front of them (and could therefore look directly at the data). The wording, however,
made the data difficult to interpret for some participants. The repeated use of “em-
ployed” and “not employed” as modifiers for the parents within the same sentence
made the sentences difficult to read. In the case of infographics, clarity in wording
may be just as important as the imagery chosen to represent the data.
Fig. 6. Fathers section of the Child Care infographic, with difficult wording
Interestingly, although Census infographics were created with the intention to convey
statistical data to an audience with no statistical background, and to a broad range of
education levels, results showed that those with a high school education or less re-
sponded much less accurately than those with any amount of college education. Fur-
thermore, the infographics did not improve performance for the less educated partici-
pants to a greater extent than the college-educated participants (as compared to the
text documents). To be consumable by a wider audience than just the college-
educated, infographics may benefit from user testing research and reorganization.
Accompanying the data with imagery may improve their appeal, but information clar-
ity via reduction of jargon and simplification of wording may improve the under-
standability of the data.
Possible future studies could include investigations of infographics from sources and
topics external to the Census Bureau, comparison of performance of infographics
versus text-documents within subjects to explore individuals factors, comparison of
infographics to text accompanied by varying levels of iconic imagery and traditional
graphs, to address some of the topics that were not investigated in the present study.
6 Bibliography
1. Thorpe, S., Fize, D., & Marlot, C. (1996). Speed of processing in the human visual system.
nature, 381(6582), 520-522.
2. Holcomb, P., & Grainger, J. (2006). On the time course of visual word recognition: An
event-related potential investigation using masked repetition priming. Cognitive Neurosci-
ence, Journal of, 18(10), 1631-1643.
3. Nielsen, J. (2008). How Little Do Users Read? Retrieved from
4. Dowse, R., & Ehlers, M. (2005). Medicine labels incorporating pictograms: do they influ-
ence understanding and adherence?. Patient education and counseling, 58(1), 63-70.
... Al Hosni (2016) examined the effectiveness of infographics in the classroom and found infographics increased students' comprehension of material and information recall. However, some studies suggested infographics are not beneficial for information recall (Damman et al., 2018;Gareau et al., 2015). Therefore, consumer information recall from infographics about risky agricultural topics must be studied in order to determine infographics future use as an agricultural communication tool. ...
... Many consumers already have strong opinions about GM science (Funk et al., 2015;National Academy of Science, 2016) and may be unable to process information contradictory to their already held belief. The findings are similar to Damman et al. (2018) and Gareau et al. (2015) who found infographics did not improve consumers' information recall; however, they contradict a large body of literature suggesting infographics improve information recall (Al Hosni, 2016;Alrwele, 2017;Bateman et al., 2010;Pjesivac et al., 2017;Yildirim, 2016). Findings from the current research suggest the risk associated with GM foods may outweigh the impact of infographics as a learning tool. ...
... As a consequence, the outcome of the achievements of the current study is in accordance with the results of previous studies, which indicated that using infographics affects student' achievements positively (Al Hosni, 2016;Alotiabi, 2016;Davis & Quinn, 2014;Dur, 2014;Gallicano, Ekachai, & Freberg, 2014;Fowler, 2015;Gareau, Keegan, & Wang, 2015;Islamoglu et al., 2015;Karre, 2015;Rezaei & Sayadian, 2015;Saurbier, 2014;Yildirim, 2016;Young & Ruediger, 2016). ...
... Besides, in some studies conducted by Alias and Hafir (2009) and Rassiah, Chidambaram, and Sihombing (2011), it was indicated that due to the lack of facilities, educational development skills, and poor network, a traditional learning approach still needs to be used; however, learners need changes in the educational environment and materials. Moreover, as recent studies demonstrated, learners who were exposed to study through infographics obtained different results than studying through traditional materials (Al Hosni, 2016;Gallicano et al., 2014;Gareau et al., 2015;Islamoglu et Furthermore, the majority of studies showed that not only students but also faculty were positively affected by the flipped classroom approach, since it increased the enrollment rate in comparison to the traditional classroom, enhanced the self-confidence of students and also decreased the burden on faculty in the long-term (Ugwoke et al., 2018;Moranski & Kim, 2016;Nouri, 2016;Freeman et al., 2014). ...
Today, technologies are utilized extensively to support the process of learning according to learner needs. Hence, the advances in technology and ideology have prompted new directions in the second language (L2) instructional practices. This study is carried out in regard to the usage of infographics in a flipped classroom learning environment called “Flipped Classroom Instructional Infographics”. The goal of the research is to assess the perceptions and evaluate the achievements of ESL students in learning English language through flipped classroom instructional infographics. The study is designed as a case study and uses a mixed research method (quantitative and qualitative) in order to obtain the result in a more consistent manner. The data are gathered from 130 students by means of a questionnaire and a focus group interview is conducted to explore the “students' perceptions of infographics”, “students' perceptions of the FCII learning environment” and “students' academic achievements level”. The findings of the study indicate that students' motivations in the experimental group compared to the control group are more triggered by the engaging and comprehensive nature of flipped classroom instructional infographics, meaning that they could absorb the concepts easier, memorize the information faster, and become more confident in the educational process.
... To our knowledge, previous research in the area of census infographic design for government communication is quite limited. There has been a usability research on the effectiveness of census infographics, emanating from the U.S. Census Bureau [4]. Thus, this preliminary study aimed to contribute towards the advancement of the area of the study by applying a visual content analysis to examine census infographics in the different context, i.e. in a Thai setting. ...
This paper reports an approach to visual content analysis and findings of the initial phase of developing the coding sheet for analyzing government infographics in Thailand. It sought to examine story topics and visual displays appeared in census infographics at the outset. All 108 infographics from the National Statistical Office (NSO) website were chosen to be studied. There were two expert coders responsible for a pilot coding. The first expert was designated to analyze the contents, and another coder was assigned to identify narrative visualization used in the census infographics. Overall, the preliminary findings reveal that the NSO focused on the creation of static infographics in which their contents, together with the static visuals, represented census data in a single page. Although the NSO static infographics were the most common format found for its visual communication, further studies are required to examine different kinds of infographics generated by other Thai government agencies dealing with varied information other than census data.
Full-text available
How long does it take for the human visual system to process a complex natural image? Subjectively, recognition of familiar objects and scenes appears to be virtually instantaneous, but measuring this processing time experimentally has proved difficult. Behavioural measures such as reaction times can be used, but these include not only visual processing but also the time required for response execution. However, event-related potentials (ERPs) can sometimes reveal signs of neural processing well before the motor output. Here we use a go/no-go categorization task in which subjects have to decide whether a previously unseen photograph, flashed on for just 20 ms, contains an animal. ERP analysis revealed a frontal negativity specific to no-go trials that develops roughly 150 ms after stimulus onset. We conclude that the visual processing needed to perform this highly demanding task can be achieved in under 150 ms.
Full-text available
The present study used event-related potentials (ERPs) to examine the time course of visual word recognition using a masked repetition priming paradigm. Participants monitored target words for occasional animal names, and ERPs were recorded to nonanimal critical items that were full repetitions, partial repetitions, or unrelated to the immediately preceding masked prime word. The results showed a strong modulation of the N400 and three earlier ERP components (P150, N250, and the P325) that we propose reflect sequential overlapping steps in the processing of printed words.
The objective was to determine the influence of medicine labels incorporating pictograms on the understanding of instructions and on adherence. Eighty-seven Xhosa participants attending an outpatient clinic who had been prescribed a short course of antibiotics were randomly allocated to either a control group (41 participants given text-only labels), or an experimental group (46 participants given text + pictogram labels). All participants had a maximum of 10 years of formal schooling. Follow-up home visits were conducted after 3-5 days to assess understanding of instructions and to evaluate adherence. A high adherence of greater than 90% was found for 54% of the experimental group, compared with only 2% of the control group. Average percentages for understanding in the control and experimental groups were 70 and 95%, respectively, and average adherence was 72 and 90%, respectively. The presence of pictograms was found to contribute positively to both understanding of instructions and adherence.
How Little Do Users Read?
  • J Nielsen