Content uploaded by Mimi V. Chapman
Author content
All content in this area was uploaded by Mimi V. Chapman on Mar 02, 2018
Content may be subject to copyright.
Article
What is a picture worth?
A primer for coding
and interpreting
photographic data
Mimi V Chapman and Shiyou Wu
School of Social Work, University of North Carolina at Chapel
Hill, USA
Meihua Zhu
Department of Social Work, East China University of Science and
Technology, China
Abstract
Society is becoming increasingly image based. As individuals regularly record moments
both mundane and momentous, images potentially lose or gain power to communicate
important information. Social work scholars have argued that social work should
incorporate images into both interventions (Chapman and Hall, 2016; Chapman
et al., 2014) and research (Marshal et al., 2009). A recent review provides an overview
of visual methodologies in social work (Clark and Morriss, 2015). The most popular
means of doing this has been through the incorporation of Photovoice (Wang and
Burris, 1997) into the social work research repertoire. Yet, in Photovoice, although
images are central, text remains the unit of analysis. This paper aims to augment the
existing literature in social work by focusing on ways in which images can be data in and
of themselves and how image-based data interact with text-based data. We will
begin with ethical considerations, proceed to step-by-step instructions for coding
and analyzing image-based data in ATLAS.ti, and finally discuss interpretation. A case
example drawing on a visually based project originally conducted with in-country
Chinese migrant mothers will illustrate the outlined methods.
Keywords
Visual methods, photographic coding, qualitative research, Photovoice, ATLAS.ti
Qualitative Social Work
0(00) 1–15
!The Author(s) 2016
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1473325016650513
qsw.sagepub.com
Corresponding author:
Shiyou Wu, School of Social Work, University of North Carolina at Chapel Hill, NC 27599-3550, USA.
Email: shiyouwu@live.unc.edu
by guest on June 3, 2016qsw.sagepub.comDownloaded from
As social work researchers are becoming more interested in visual methods as a
data collection mechanism, guidance in analyzing image-based data is required.
Using images as data is currently uncommon in social work research. Although,
participatory research methods such as Photovoice (Wang and Burris, 1997)
have gained traction, in Photovoice, participant-generated images work as
photo-elicitation devices in which participant images begin a dialogue with
fellow group members and researchers (Sandelowski, 2000). The data most often
analyzed in Photovoice are the transcribed text participants provide as they discuss
their photographs in preparation for advocacy experiences such as community
forums or photo exhibitions. Yet, participant-generated photographs contain
additional undiscussed visual data that are often ignored by researchers and par-
ticipants but may contain valuable information that may augment or enhance the
text-based findings.
Although videotaped interactions are often coded for content in the helping
professions (Dunn et al., 2011), photographs as data sources largely have been
ignored. Scholars, art historians, and marketing professionals have long recognized
the power of images to communicate overt and covert messages (Davis, 1992).
Indeed, still images provide a powerful means for quickly giving voice to complex
experiences by allowing individuals to tell their stories in their own way. In
addition, they allow for reflective processing that can lead to rich and insightful
individual and group conversation (Kross et al., 2005). The photographer, whether
professional or amateur, researcher, or participant, chooses particular moments to
depict over others making both the content contained in photographs, as well as
what is absent, important components for understanding the photographer’s mean-
ing (Berger, 1969). When photographs are coded, patterns included in and excluded
from an individual’s or group’s series of images may become evident. These
patterns can be reflected back to participants or other key informants as a form
of member checking or photo-elicitation. Yet within social work, methods for
coding and analyzing photographs have not typically been a part of either quali-
tative or quantitative research training. This paper begins to fill this gap by exam-
ining ethical considerations when using images as data, providing step-by-step
guidance for coding visual data using ATLAS.ti, and considering ways in which
we might understand particular functions in ATLAS.ti to aid in interpretation and
hypothesis generation.
We will use a case example based on data obtained through a Photovoice study
based in Shanghai, China in which we asked migrant mothers to describe their
parenting experiences using participant-created photographs and group discussion
(Chapman et al., 2013). This exploratory project aimed to understand migrant
mothers’ life experiences, particularly parenting experiences and adjustment to
daily life in the host city. The goal was to take this information to decision-
makers who had the power to assist these mothers in creating needed changes in
their migrant village. For complete information on this project, see Chapman et al.
(2013). In total, 13 mothers participated in taking and discussing photographs, and
advocated for change through a forum with community leaders. The audio-taped
2Qualitative Social Work 0(00)
by guest on June 3, 2016qsw.sagepub.comDownloaded from
discussion was transcribed and translated from Mandarin to English for analysis.
Our initial goal, which was accomplished, was to follow the classic Photovoice
procedure (Wang and Burris, 1997) in which group discussion transcripts were
analyzed with participants in order to create a community forum to share findings
with key stakeholders and advocate for change. During the project, a research
team member who shared a similar migration history to the participants noticed
elements in the pictures that had not come up in the discussion. Further, he sug-
gested other elements that might have been depicted but were not. In addition,
information in the pictures supported the text-based findings in interesting
ways. Our team opted to code the photographs themselves in addition to the
text. Data obtained through this project will be used to provide examples through-
out the paper. We will begin with laying out ethical considerations when using
photographs as data and move to a step-by-step guide to coding and analyzing
images.
Ethical considerations
In the current climate self-made images are ubiquitous. People of all ages and
backgrounds are regularly posting images to social media sites and platforms,
thus, asking participants to take and share images, may be considered low-risk
research. However, an image, particularly one taken for research purposes, may
represent deeply personal dimensions of participants’ identities or circumstances.
Because photography is closely related to visual perception (Solso, 2003), even a
photograph taken without much thought may represent and communicate more
than what a participant might consciously choose to reveal. At the most basic level,
an image may communicate personal data such as location, age, or gender in ways
that threaten confidentiality or protection from deductive disclosure. Further,
taking a picture and discussing it, may produce emotional responses that partici-
pants do not anticipate when they agree to participate in image-based research.
Therefore, when working with participant-generated photographs, researchers
must consider research ethics as vigilantly as they would when collecting highly
sensitive survey or interview data.
Two informed consent elements warrant detailed attention: confidentiality and
potential distress from participation. In the community-based participatory
research context, images are created for advocacy purposes (Lorenz and Kolb,
2009). Participants take photographs knowing that they will engage with others
in their community to discuss their chosen photographs. Yet, photographs
obtained as data are often published in academic journals and, in recent years,
those articles are widely accessible online. Participants may or may not be coau-
thors on such articles. But, small sample sizes and data contained in images may
produce potential deductive disclosures that participants do not anticipate.
For example, one mother in our Photovoice study took a photograph of her
husband working. In the picture, he is wearing a cap, and the picture is taken
from an angle showing his face only in shadow. At first glance, the photograph
Chapman et al. 3
by guest on June 3, 2016qsw.sagepub.comDownloaded from
appears deidentified. A closer look reveals that the cap has a logo on it that indi-
cates his employer. If enough information is given in the article to contextualize
how and where the research was done, how many participants were involved, or
other elements common to a typical sample description, the person photographed
and the person who created the photograph may be able to be identified. The
participant may have been aware and willing to participate in a community
forum, or even to have her photographs published in a journal, but may not rec-
ognize that small elements in the photographs may compromise privacy in unex-
pected ways. Researchers then bear the added responsibility of thinking through
these many possibilities both at the point of obtaining informed consent and of
dissemination. Photographs may need to be edited—in the above example, we
removed the logo and blurred the face to prevent inadvertent disclosure. When a
photograph is particularly compelling, yet will compromise confidentiality in ways
the participant may not have recognized or anticipated, a researcher must refrain
from using it or return to the participant for specific permission. Even when con-
sent has been obtained, the researcher may be in the position to recognize new risks
in certain dissemination venues and should either return to their participants for
further risk/benefit discussion or refrain from dissemination. Modifications to
Institutional Review Boards (IRB) protocols may be necessary in such situations.
Discomfort related to participating in the research also has a particular flavor in
visual research, particularly when researchers code images. As posited earlier in this
manuscript, images may contain data over and above what participants actually
describe when talking about an image they have created. That data may be present
in the image or it may be informed by what is missing. Both present and missing
elements may represent a conscious or an unconscious choice to depict or neglect
that data. Yet, when researchers code images, they may look for elements not
previously mentioned by participants. For example, migrant mothers in our project
provided no pictures of elder family members even though filial piety is a central
value throughout China. When we asked them about this after attempting to code
pictures for elders, many mothers at the table began to cry and voiced deep distress
in not being able to physically care for their parents on a day-to-day basis. Through
asking to discuss photographic elements about which participants did not initiate
discussion, researchers may be opening up conversations that are painful and pro-
duce emotional consequences that participants did not anticipate when they con-
sented. Of course, these conversations can be rich and meaningful, particularly for
social work researchers seeking to understand needs and intervention points, but
participants should be made aware of these possibilities and resources should be
present to help participants with any emotional distress they may encounter by
being a part of the research.
Coding pictures step by step
Several qualitative data analysis programs allow for coding visual data, linking
coded photographs to text, and analyzing relationships between coded data.
4Qualitative Social Work 0(00)
by guest on June 3, 2016qsw.sagepub.comDownloaded from
This article will focus on analytic steps within ATLAS.ti because it is widely used
for qualitative data analysis and supports the analysis of visual data.
By way of overview, we suggest five steps, depicted in Figure 1, for analyzing
photographic data: data organization, code creation, coding, finding relationships
or patterns, and interpretation.
Although data, both text- and image-based, can be imported to ATLAS.ti in
multiple formats (Archer, 2012; Friese, 2013, 2014; Muhr, 1997) to analyze images,
each image should be separated from text. If text and images are stored together in
one file, the photographs cannot be coded partially although the text can. For
instance, in our analysis, we initially created a file that included each mother’s
discussed photographs and the transcribed text that accompanied those images.
This arrangement was useful for analyzing text but we could not analyze the photo-
graphs. Instead, we saved each photograph to be considered for analysis in *.jpg
format and treated them as independent primary documents (P-Docs on the main
toolbar; see Friese (2014: 3–4) for more details about the ATLAS.ti toolbar). We also
saved each mother’s transcript as an independent Microsoft Word document file.
Step 1: Data organization
The P-Docs button on the left of the main ATLAS.ti toolbar accesses a ‘‘Primary
Doc Manager’’ window. Choosing Documents, then, New, and then Add Document
creates a new project. To access an existing project, known as a Hermeneutic Unit
(HU) file in ATLAS.ti, go to Project on the main menu, choose Open, and then
select the desired ATLAS.ti file (e.g. mother2.hpr6, where the ‘‘hpr6’’ is ATLAS.ti
HU’s filename extension). All related documents, code lists, memos, and other
ATLAS.ti files saved for that HU will be available to edit. This process mirrors
the process for text files; in this case, photographs are uploaded instead of text
documents.
Step 2: Code creation
Coding can begin once all primary documents are uploaded. Just as with text, a
priori codes based on previous research, conversations with participants, or theory
can be created prior to the start of coding. Free codes also can be added to images
throughout the coding process. The process is the same.
2. Code
creation
- Free codes
- Codes list
5. Interpretation
- Code frequencies
- Relationship to text
- Member checking
- Examining
relationships between
image and text
3. Coding
photographs
- Coding by list
- Open coding
1. Data
organization
- Photographs (as
separate file)
- Text (transcripts;
optional)
4. Finding
relationships
- Network views
- Codes-primary
documents table
- Code co-occurrence
table
- Diagrams
Figure 1. Steps in photographic data analysis.
Chapman et al. 5
by guest on June 3, 2016qsw.sagepub.comDownloaded from
A priori codes. In our Photovoice project, one focus was the migrant mother’s
parenting concerns. Knowing from the literature that education was a highly sali-
ent issue in Chinese culture, we created an a priori code called ‘‘education.’’
Pictures that depicted parents engaged in helping children with homework or
driving their children to school were coded as ‘‘education.’’
To create an a priori code list, we chose Codes (at the middle of main
menu), then Create Free Code(s) and type the desired code name in the
pop-up window. The down arrow can be used to create another new code. Once
all the a priori codes names were entered, the Codes window displayed an a priori
code list.
After we began to code using our a priori codes, we found that most pictures of
fathers showed them involved with their children’s education. Thus, we added an
additional code called ‘‘fathers in education’’ to our code list. The process of
creating free codes is the same as for a priori codes. It is done after coding has
begun instead of in advance.
Step 3: Coding photographs
To code our photographs we selected a primary document (e.g. a photograph) from
the ‘‘P-Docs’’ document list and selected the elements within the photograph to
code by dragging the mouse over particular parts of the picture. To use a priori
codes we moved the mouse over the selected part of the photograph, right clicked,
and then chose first Coding, and then Select Codes(s) From List, and then double
clicked the desired code name.
When we wanted to add an open code called ‘‘smile,’’ we moved the mouse over
the selected image area and chose Coding. Then we chose the option Enter Code
Name(s) and typed the code name being added. Figure 2 visually demonstrates this
process of creating an open code and then using that code. After we added a code
called ‘‘smile’’ during the open coding process, we were able to apply the code to
other photographs. Once a code is created, either a priori or during the coding
process, accessing that code again is done in the same way. Auto-coding in which
the program searches for a particular word within the texts is not applicable to
photographic data.
Step 4: Finding relationships
Creating code families (CFs) can help organize codes and help researchers think
about relationships and preliminary themes. Table 1 presents CF creation using
nine photographs taken by one mother. These photographs contain 16 codes which
we grouped into five CFs. The nine photos in this example contain six codes that
refer to people, two that refer to places, three codes that refer to how time is spent,
two that reference changing status, and three that address other issues or special
topics. Table 1 presents the progression of codes from individual codes found in
pictures, to groupings, to naming CFs.
6Qualitative Social Work 0(00)
by guest on June 3, 2016qsw.sagepub.comDownloaded from
Figure 2. Process of selecting codes from a code list.
Table 1. Process of generating code families.
P-Docs !Codes (A–Z) !Grouping codes !Codes families
M2-P1.jpg 1. Children 1. Children
1. People
M2-P2.jpg 2. Daughter 2. Daughter
M2-P3.jpg 3. Education 6. Husband
M2-P4.jpg 4. Family portrait 8. Mother
M2-P5.jpg 5. Home 9. Others
M2-P6.jpg 6. Husband 14. Son
M2-P7.jpg 7. Leisure time 5. Home 2. Places
M2-P8.jpg 8. Mother 16. Working place
M2-P9.jpg 9. Others 3. Education
3. Time spent
10. Parenting 7. Leisure time
11. Poor condition 10. Parenting
12. Shanghai symbol 15. TV 4. Change symbols
13. Smile 12. Shanghai symbol
14. Son 4. Family portrait
5. Special issue
15. TV 11. Poor condition
16. Working place 13. Smile
Chapman et al. 7
by guest on June 3, 2016qsw.sagepub.comDownloaded from
Figure 3 demonstrates the way in which the codes in one photograph contribute
to CF creation. Consider photograph 7 that contains six coded elements. Each
element is referred to as a quotation, even though it is image and not text. In
photograph 7, the primary document or P-Docs, we coded quotation 1 as daughter
(code 1), quotation 2 as husband (code 2), quotation 3 as education (code 3), quota-
tion 4 as home (code 4), quotation 5 as smile (code 5), quotation 6 as poor condition
(code 6), and quotation 7 as parenting (code 7). Then we grouped code 1 and 2 as
one CF named ‘‘people (CF1)’’ and grouped code 3 and code 7 as another CF
called ‘‘time spent (CF2).’’ Creating CFs may prompt researchers to look at ways
in which codes fit together either across all of the visual data or among subgroups
of images.
In ATLAS.ti, there are five ways to examine relationships among and between
codes. The first is the network view of primary documents and codes which displays
all the codes within a particular photograph. A second option is the network view of
codes and quotations which allows for review of all the quotations/photographs that
contain a particular code. Third is codes–quotations–documents relationships which
quantifies codes by showing the distributions or percentages of codes in all con-
sidered photographs. A code co-occurrence table depicts the correlations among the
codes and, finally, the diagram function creates visualizations of the relationships
among codes. Figure 4 provides an example in which steps 1–4.1 take us through
these options.
The network view of primary documents and codes provides information on what
codes are present in particular photographs. As depicted in Figure 4, Mother 2’s
second photograph named P3:M2-P2.JPG was chosen as the photograph of inter-
est by choosing P-Docs on the main toolbar. Next we chose Open Network View
that displays a network view window of P3:M2-P2.JPG. Next we right clicked the
mouse on the photo and chose Import Neighbors and Import Codes.
As Figure 4 shows these steps allow all the codes present in this picture to be
visible. If a picture contains many coded elements, the window may be difficult to
CF2: Time Spent
Code 3:
Education
Code 7:
Parenting
Code 1:
Daughter
CF1: People
Primary documents: Photographs of mother 2
Code 2:
Husband
Quotations: Parts of photo 7
Other
Q
uotations…
Figure 3. Codes relationship to codes families.
8Qualitative Social Work 0(00)
by guest on June 3, 2016qsw.sagepub.comDownloaded from
understand. To organize the codes use the step labeled 4.1 in Figure 4, first choos-
ing layout from the main menu of the network view window and then choose
Semantic Layout to create a more organized picture.
Network view of codes and quotations. We used the network view of codes and quota-
tions to find out which primary documents in a HU, in our example Mother 2’s
photographs, contain the code education. Note that quotations can refer to text or
to an element of a photograph. We first chose Codes from the main toolbar and
selected ‘‘education’’ from the code list. We then chose the code name to Open
Network View and then, Import Neighbors followed by Import Quotations. Three of
her nine pictures contain the code ‘‘education.’’
Codes–quotations–documents relationships. Another way to examine how coded elem-
ents are connected is to create a codes-primary documents table. Figure 5 provides
an example. All codes or subgroups of codes can be chosen in order to examine the
relationships between them. We first chose Analysis from the main menu and then
selected Codes-Primary Documents Table. In our example we selected a series of
codes for examination. We selected them in the pop-up window from the Code
Families cell and moved the coded files into the Primary Document Families cell.
Then, we used the ‘‘>>>’’ button to move the selected codes into the analytic data
source cell and clicked Create Report which created a table like that shown in
Figure 5. The table provides frequencies and percentages for the chosen codes or
CFs. The text embedded in Figure 5 gives examples of how these analyses related to
our eventual published findings.
Figure 4. Process to obtain a network view of primary documents and codes.
Chapman et al. 9
by guest on June 3, 2016qsw.sagepub.comDownloaded from
Code co-occurrence table. Acode co-occurrence table displays the relationship
between codes using ratios. Similar to correlation coefficients, higher ratios suggest
a stronger relationship between codes. To begin we chose Analysis from the main
menu and then code co-occurrence table. In the resulting code co-occurrence table
window, we selected the codes for both columns and rows and then specified which
codes to use in the analysis using the ‘‘ 44 ’’ button. These steps produced a code
co-occurrence table like the table shown in Figure 6.
The ratios in the co-occurrence table may suggest the ‘‘shared meaning’’ between
two codes. For example, the co-occurrence rate of code education and code children
is 0.51, which means there are 73 places we coded as education, and 87 places as
children, and the overlap of education and children is 54 (as shown in the left of the
ratio of 0.51 in the cell). Therefore, 0.51 comes from 54 (as the numerator) divided
by the denominator as the sum of 73 and 87 minus the overlap (which is 54). Ratios
closer to 1 suggest a greater relationship between two codes. When codes do not
overlap ‘‘n/a’’ appears in the cell. Although this strategy may suggest relationships
between codes, member checking or supporting text in the transcripts creates more
confidence in such findings.
Diagrams for visualizing relationships. Diagrams like the one shown in Figure 7 also
depict connections among codes. Using the ratios from code co-occurrence table,
Figure 5. Codes-primary documents table results (quotations were cited from Chapman
et al., 2013).
10 Qualitative Social Work 0(00)
by guest on June 3, 2016qsw.sagepub.comDownloaded from
we augmented the program-produced diagram to further visualize important rela-
tionships by using different line styles or color. For example, when the high ratios
between education, children, and home were highlighted using a thicker bold line to
represent the ratios which were higher than 0.5, a triangle emerged that suggested
the deep connection between children, their home, and their education.
Default diagram. We created this diagram by using the Codes button in the main
toolbar and dragging one code onto another code to create a chosen type of rela-
tionship using chosen symbols to specify the type of relationship. Options are ‘‘¼¼:
is associated with ‘‘, ‘‘[]: is part of’’, ‘‘¼>: is cause of’’, ‘‘<>: contradicts’’. The
selected relationship codes are chosen by opening network view, selecting a relation
line, changing the relation, and opening the relation editor. Note that by selecting the
relationship direction as described above, the researcher is simply creating a visual of
hypothesized relationships for which he or she has support through other means.
To create these relationships in the default diagram, we chose Links in the net-
work view window menu; then we chose, Edit Relations, then Code-Code-Relations
which created new code relations by filling out the options in the ‘‘code-
code-relations editor’’ window.
Figure 7. The default and augmented diagram of code relationships.
Figure 6. Code co-occurrence table.
Chapman et al. 11
by guest on June 3, 2016qsw.sagepub.comDownloaded from
Step 5: Interpretation
When looking exclusively at photographs or text, the way in which findings are
interpreted may vary based on what qualitative tradition the researcher espouses.
Photographs, like text, can be analyzed and interpreted using content analysis to
count the number of times particular codes arise, co-occur, in order to identify
particular themes or build theory (Bell, 2001). Alternatively, researchers might use
a more phenomenological approach to describe important ideas that are repre-
sented by depicted image elements. However, when photographs and text are
used together as data sources, interpretation involves considering what is present
and absent in each data type as well as what the participants’ role is in creating
meaning.
Consider Table 2, which contains information from nine photographs taken by
one mother and the transcribed text that is associated with those photographs.
Column two shows which codes were found in each picture. Column three shows
which codes were noted in the transcribed text for each of these photographs.
Column four shows which codes were noted in both the text and the photograph.
Note, that in Column two, ‘‘poor living conditions’’ were coded by researchers in
each photograph. However, in the discussion transcript, the participant only men-
tioned ‘‘poor living conditions’’ in reference to photograph number 5. Returning to
participants with findings is helpful to interpretation. In this example, the mothers
had migrated to Shanghai from the Chinese countryside in which living conditions
are often very poor, a reality that sometimes prompts a family’s migration. To our
eyes, this mother’s living conditions looked dismal and we coded them as such in
every one of her photographs. Yet the mother’s text said very little about her living
conditions. This discrepancy provided an opportunity to think in more nuanced
ways about her experience. To this mother’s eyes, her living conditions may
represent an improvement over conditions in her home province. Alternatively, a
participant may not have given much thought to their surroundings prior to taking
the photographs and seeing their environment through a camera lens may prompt
new thought, reflection, and conversation. Indeed, although these mothers did not
focus on their living conditions as hardships for themselves, they lamented the
paucity of safe, outdoor, play spaces for their children. Through continued con-
versation during the coding and interpretation process, we were able to gain more
insight into what the mothers’ photographs and text together reflect.
What is not seen in photographs can also be telling (Packard, 2008) and is best
considered through dialogue with participants (Drew and Guillimin, 2014). In our
photographs, we noticed no elders, something quite unusual in China where grand-
parents are typically very involved with their grandchildren and filial piety is a
central cultural value. When we asked about the absence of elders in the photo-
graphs, our participants became quite emotional telling us how difficult it was to
live far away from the elders in their family and the great lengths to which they
went to maintain contact. Had we not brought this observation to their attention,
we could have misconstrued our observation assuming a more Western viewpoint
about people wanting to reinvent their lives away from their families of origin.
12 Qualitative Social Work 0(00)
by guest on June 3, 2016qsw.sagepub.comDownloaded from
Conclusion
Analyzing photographs provides another means to help participants share nuanced
perspectives which, together with other methods, will help social workers design
and test interventions and advocate for needed change. In our example, the overall
study was exploratory and our analyses were used to understand the mothers’
experiences and need from their perspective. This process resulted in interven-
tions implemented by social workers working in the mothers’ migrant village.
Table 2. Codes from photographs and texts.
Mother2 From photographs From texts Combined codes
Picture 1 1. Home 2. Poor condition
3. Others 4. Leisure time
5. TV 6. Smile 7. Working
place
1. Children
2. Mother
3. Neighbor
n/a
Picture 2 1. Home 2. Poor condition
3. Education 4. Daughter
5. Smile
1. Daughter
2. Education
– Daughter
– Education
Picture 3 1. Home 2. Poor condition
3. Mother 4. TV
5. Working place
1. Mother
2. Job
– Mother
Picture 4 1. Home 2. Poor condition
3. Daughter 4. Son
5. Smile 6. Leisure time
1. Children
2. Daughter
3. Son
– Daughter
– Son
Picture 5 1. Home 2. Poor condition
3. Son 4. Smile
5. Education
1. Son
2. Education
3. Poor condition
– Son
– Education
– Poor condition
Picture 6 1. Home 2. Poor condition
3. Family portrait
4. Mother 5. Husband
6. Daughter 7. Son
1. Family portrait
2. Daughter 3. Paintings
4. Shanghai symbol
– Family portrait
– Daughter
Picture 7 1. Home 2. Poor condition
3. Education 4. Parenting
5. Husband 6. Daughter
7. Smile
1. Daughter 2. Education
3. Husband 4. Parenting
5. Job
– Daughter
– Education
– Husband
– Parenting
Picture 8 1. Home 2. Poor condition
3. Leisure time 4. Others
5. Daughter 6. Son 7. Smile
8. Shanghai symbol
1. Relative
2. Daughter
3. Son
– Daughter
– Son
Picture 9 1. Home 2. Poor condition
3. Leisure time 4. Mothers
5. Reading
1. Mother
2. Reading
3. Leisure time
4. Neighbor
– Mother
– Reading
– Leisure time
Chapman et al. 13
by guest on June 3, 2016qsw.sagepub.comDownloaded from
Image-based data collection and analysis lends itself to exploratory work because
image creation enfranchises communities by allowing them to depict clearing their
priorities. When researchers are collaboratively engaged with participants, further
data may be extracted from participant-created images as described in this manu-
script. Yet, social work researchers must be careful not to make assumptions about
what they see in participant-generated images. Indeed, the data’s visual nature
means researcher biases and viewpoints can be easily introduced into each analysis
step without conscious awareness. Even more so than in text-based analysis, a
team approach using insider and outsider perspectives and working closely with
participants throughout the analysis mitigates against a researcher’s point of view
coloring the findings.
Photographic data collection has several advantages and benefits. Participants,
regardless of age, literacy level, and from all walks of life can engage with images in
ways perhaps not possible through a standard interview. Through carefully con-
sidered analysis and interpretation, visual analysis of participant-generated images
shows promise as a tool in the qualitative social work research toolkit.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, author-
ship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication
of this article.
References
Archer E (2012) Introduction to Atlas.ti., 4th ed. Pretoria: Unisa.
Bell P (2001) Content analysis of visual images. In: Van Leeuwen T and Jewitt C (eds) The
Handbook of Visual Analysis. London: Sage, pp. 10–13.
Berger J (1969) Understanding a photograph. In: Dyer G (ed.) Understanding a Photograph
John Berger. London: Penguin Press, pp. 17–21.
Chapman MV and Hall WJ (2016) Outcome results from Yo Veo: A visual intervention for
teachers working with immigrant Latino/Latina Students. Research on Social Work
Practice 26(2): 180–188.
Chapman MV, Hall WJ, Colby R, et al. (2014) How images work: An analysis of a visual
intervention used to facilitate a difficult conversation and promote understanding.
Qualitative Social Work 13: 456–476.
Chapman MV, Zhu M and Wu S (2013) Mothers in transition: Using images to understand
the experience of migrant mother in Shanghai. Journal of the Society for Social Work and
Research 4(3): 245–260.
Clark A and Morriss L (2015) The use of visual methodologies in social work research over
the last decade: A narrative review and some questions for the future. Qualitative Social
Work. Epub ahead of print 20 August 2015. DOI: 10.1177/1473325015601205.
14 Qualitative Social Work 0(00)
by guest on June 3, 2016qsw.sagepub.comDownloaded from
Davis JF (1992) The power of images: Creating the myths of our time. Media and Values 57.
Available at: http://www.medialit.org/reading-room/power-images-creating-myths-our-
time (accessed 31 October 2015).
Drew S and Guillimin M (2014) From photographs to findings: Visual meaning-making and
interpretive engagement in the analysis of participant-generated images. Visual Studies
29(1): 54–67.
Dunn MJ, Rodriguez EM, Miller KS, et al. (2011) Direct observation of mother–child
communication in pediatric cancer: Assessment of verbal and non-verbal behavior and
emotion. Journal of Pediatric Psychology 36(5): 565–575.
Friese S (2013) ATLAS.ti 7 User Guide and Reference. Berlin: ATLAS.ti Scientific Software
Development GmbH.
Friese S (2014) Qualitative Data Analysis with ATLAS.ti. Thousand Oaks, CA:
SAGE Publications. Available at: http://zdenek.konopasek.net/archiv/kpa/filez/Atlasti_
workshop_manual_english.pdf.
Kross E, Ozlem A and Mischel W (2005) When asking ‘‘why’’ does not hurt. Distinguishing
rumination from reflective processing of negative emotions. Psychological Science 16:
709–715.
Lorenz LS and Kolb B (2009) Involving the public through participatory visual research
methods. Health Expectations 12: 262–274.
Marshal HS, Craun SW and Theriot MT (2009) The big picture: How social work can
effectively utilize photographs. Social Work 54(4): 317–325.
Muhr T (1997) ATLAS.ti User’s Manual and Reference (Version 4.1). Berlin, Germany:
Scientific Software Development.
Packard J (2008) ‘‘I’m gonna show you what it’s really like out here:’’ The power and
limitation of participatory visual methods. Visual Studies 23(1): 63–77.
Sandelowski M (2000) Combining qualitative and quantitative sampling, data collection,
and analysis techniques in mixed-method studies. Research in Nursing and Health 23:
246–255.
Solso R (2003) The Psychology of Art and the Evolution of the Conscious Brain. Cambridge,
MA: The MIT Press.
Wang C and Burris MA (1997) Photovoice: Concept, methodology, and use for participa-
tory needs assessment. Health Education and Behavior 24(3): 369–387.
Chapman et al. 15
by guest on June 3, 2016qsw.sagepub.comDownloaded from
