Visual and Textual Analysis Using ATLAS.ti
Author: Ashley C. Feely
ATLAS.ti may be used to analyze text and its supporting images through the creation of carefully designed code prefixes and code families, which are useful for researchers looking to perform combined visual and textual analyses using qualitative content analysis methodologies. This blog will explore how to apply and analyze data using ATLAS.ti codes to apply the principles of grounded theory, utilizing an article and its accompanying images from the popular women’s magazine, Cosmopolitan, as sample data.
Challenges of Grounded Theory
One challenge that is common for researchers using grounded theory as a guiding principle is the development of codes, particularly for analyses that involve more than one type of data. Since grounded theory focuses on theory building from emergent themes (Glaser and Strauss  2011), the creation and revision of codes may be one of the most time-consuming parts of data analysis.
Code prefixes and code families are important and useful analytical tools for theory building, since they not only provide organizational benefits but also the ability to quickly and conveniently view groundedness and density[i] in the Code Manager and in the creation of networks. As theory building progresses, codes sometimes require extensive revision and often need to be resituated in their relationships. Therefore, it is helpful to create networks and keep track of codes’ characteristics by creating and editing memos and code comments.
One important component of theory building is flexibility. Since ATLAS.ti provides shortcuts for accessing memos in the project menu and flexibility in making edits at any time, researchers—particularly those following the principles of grounded theory—should take full advantage of this function to carefully document any and all changes in codes and code families.
The most common application of ATLAS.ti using a content analysis methodology seems to be textual data. Less common but very important are ATLAS.ti’s other notable applications. Whether data takes the form of audio, geographical, video or still image data, ATLAS.ti is a valuable tool in qualitative analysis. Using my own research of Cosmopolitan[ii] as an illustration, this blog details how ATLAS.ti can be a valuable tool in performing combined visual and textual data analyses.
Still image Data
When analyzing still images featured within texts, choosing one of these approaches will fundamentally shape how you go about performing your analysis:
- Disaggregating images from text and importing images into your library as primary documents (hereafter referred to as P-Doc images)
- Allowing images to remain embedded within text and analyzing them within context (hereafter referred to as embedded images)
Each approach has practical and theoretical advantages and disadvantages. Choosing the preferred approach should be determined by your data. Although it may take some time to familiarize yourself with the technical workings of ATLAS.ti, this program is user-friendly and its use becomes more intuitive with practice.
P-Doc Images and Embedded Images in Research
Creating P-Doc images can be particularly beneficial when your data involves multiple images per original text document, like those characterizing newspaper or magazine articles. In such cases, it is critical to construct file names carefully when disaggregating images from text documents, since P-Doc labels correspond with file names.[i] Specifically, I have found it useful to construct P-Doc image file names by signifying their location within the dataset, as to avoid confusion and increase my own efficiency.
Carefully naming P-Doc images also becomes helpful when generating output using the Codes-Primary Document Table, since this careful construction allows users to quickly identify image locations in their selection of P-Docs included in a Codes-Primary Document Table report. Users can view output for P-Doc images or embedded images separately through the creation of these reports by selecting the P-Docs or P-Doc families of interest.
A perceived disadvantage of importing images as P-Docs may be that embedded images and P-Doc images cannot be analyzed within the same ATLAS.ti file. In some cases this concern may be warranted, since files consisting of both embedded images and P-Docs can create some initial difficulties without proper consideration. For example, using the same coding scheme for embedded images and P-Doc images in the same file (i.e., repeating the same codes in two or more places for a repeated image) will double the count in generated output by the Codes Co-Occurrency Table. However, different types of P-Docs can be analyzed separately or jointly within the same file by creating and using filters.
Keeping track of images can be one of the most challenging and cumbersome parts of analyzing data comprised of many images. Although it is possible to keep track of embedded images using networks, P-Doc images are advantageous because they allow users to view image data at a glance.
Networks are useful tools for researchers looking to visually represent code relationships. To view all P-Doc images in a dataset, create a network following these steps:
- From the project menu, choose Networks > New Network View
- Title the network view
- Choose Nodes > Import Nodes
- Select Primary Documents from the drop-down menu
- Choose the P-Doc images of interest
- Click Import
- If desired, resize and move the P-Doc images within the Network View to create a comprehensive visual reference with Codes, Code Families, Memos, Memo Families and so on.
To save this network view, select Network > Save
To create a network view of embedded images, researchers must code the entire image as a single selection. To code an embedded image as one selection:
- Draw a boxed selection around the image using your mouse
- Right-click on the image
- Select Coding > Enter Code Name(s)
- Enter all code names necessary for this embedded image
To access the network for your embedded image, right-click the quotation bar next to the image and select Open Network View. This view shows all codes (or Code Families, Memos, Memo Families, Network Views, Primary Documents, Primary Document Families or Quotations) linked to the embedded image. As of version 7.5.2, the embedded image quotation will show in the network with its ID and name, but the image itself will not show.
It is recommended that you rename the network to reflect the image location within the dataset or to otherwise differentiate it from preceding and successive images. To rename the network, right-click on the image name and select Rename. You may also connect the image with its location and codes by copying and pasting (Copy > Paste Special) the image into the comment box (select Edit Comment to access comment box).
To create a more comprehensive graphical representation of embedded images and their accompanying Codes, Memos, Quotations and so forth, users can merge network views. Within the Network View Manager, choose Nodes > Merge Network View, select the useful networks and choose OK. Another means of creating a comprehensive network view is through hyperlinking. For example, it may be useful to link all embedded images to an article title in the PDF text document. To create a chain:
- Right-click the quotation bar next to the source segment. Select Create Link Source.
- Right-click the quotation bar next to the targeted segment. Select Create Link Target for each of the segments you want to link back to the source.
Right-click the quotation bar next to the link source and select Open Network View to resize or reposition codes and to create a visual representation of the node density and groundedness by selecting Display > Set Colors > Color by Density & Groundedness. Groundedness and density will then be represented by color and shading. To save this network view for reference or future use, select Network > Save As.
Since code prefixes are an important analytical tool when similar themes emerge in image and text data, it is in the creation of networks that carefully naming codes and using code prefixes and families become critical steps in the analysis of embedded images. Since ATLAS.ti sorts codes alphabetically, code-prefixes and code families will make related codes easier to locate and manage in the Code Manager.
Code prefixes should be simple enough to read at a glance, but still provide the detail necessary for differentiating one code from another—you can always edit comments or memos for codes if you need to include more detail.
For organizational purposes, it may be helpful to denote the relationship among codes. To access the Code-to-Code Relations editor, select Networks > Edit Relations > Code-to-Code Relations and insert existing the relationship descriptors.
Both P-Doc images and embedded images possess their own strengths in the analysis of still images. For example, P-Doc images allow the researcher to keep track of all images by creating reference images in network views. Using P-Doc images also means it is not necessary to code the image in a single selection, thus allowing users more flexibility and precision in coding. However, embedded images allow researchers to analyze content within the context of accompanying text, so those interested in interpreting images may find it advisable to leave these images embedded in the text. The approach you choose to visual and textual analysis should be decided by the data.
While it may seem that P-Doc images provide researchers more practical advantages than embedded images, both approaches enable users the flexibility to apply principles of grounded theory by continuously creating and editing codes, code families, and memos throughout analysis. Through the creation of carefully constructed code prefixes and families, analysis of combined visual and textual data becomes more manageable.
About the author
Ashley C. Feely is a graduate teaching and research assistant at the University of Illinois, Urbana-Champaign, where she is pursuing a PhD in Sociology. She holds a bachelor’s degree in Sociology from Saint Mary’s College in Notre Dame, Indiana. Her research focuses on racial, gender and class inequality in contemporary American society.
Friese, Susanne and Thomas G. Ringmayr. 2013. ATLAS.ti 7 User Manual. Berlin: ATLAS.ti Scientific Software Development GmbH.
Glaser, Barney G. and Anselm L. Strauss. ( 2011). The Discovery of Grounded Theory: strategies for qualitative research. New Brunswick, New Jersey: Aldine Transaction.
Talarico, Brittany. 2012. “10 Easy Ways to Save Cash.” Cosmopolitan, March 2. Retrieved February 28, 2013 (http://www.cosmopolitan.com/advice/work-money/easy-ways-to-savecash).
[i]Image P-Docs should be saved as a JPEG, BMP, or TIFF formats before adding them to your library. Images embedded within text should be saved as a PDF, since ATLAS.ti enables users to code any part of PDF files and this type of file does not disrupt or distort the context (Friese and Ringmayr 2013).
[ii] Groundedness refers to code frequency and density indicates the strength of links among codes.
[iii] Sample data derived from Talarico (2012) article, “10 Easy Ways to Save Cash.”