Evolving Co-occurrence Tools
Author of this page: Ray Maietta
Affiliation: ResearchTalk Inc.
Date written: 30th October 2005
From forthcoming Maietta, Raymond C. (2006). State of the Art: Integrating Software with Qualitative Analysis. In Applying Qualitative and MixedMethods in Aging and Public Health Research. Leslie Curry, Renee Shield and Terrie Wetle, (Eds.) Washington, DC: American Public Health Association and the Gerontological Society of America.
Ray Maietta supports users of qualitative software: working from New York and Los Angeles USA. http://www.researchtalk.com/
Code combination tools (ATLAS.ti’s query tool, Text Retrieval in MAXqda, and Searches in NVIVO) built from Boolean and Proximity operators can lead users to assume that themes exist in places where more than one code is applied to the same section of text. For example, we thought that people who completed journal interviews in a study of life satisfaction were “becoming their parents.” To explore this notion, we used a code combination tool to find instances when we coded to both the parents code and the self growth code. However, the only instance of this code combination was from a woman who related going to yoga classes with her mom. We then decided to step back to ask the question in a simpler way. Dropping the assumption that we could learn about people becoming their parents only where we coded these two main categories to the same text, we retrieved both the self growth and parents codes and read each segment coded to parents or self growth in the order they appeared in the data document.
The first document we read perfectly characterized our expectation. A young woman related that when she lived at home, she couldn’t wait to go to college to escape her father who she described as a military drill sergeant. Everything had to be done on time and kept in an orderly manner. This segment was coded to parents. In a later section of the document we coded to self growth, she wrote how proud she was to listen to a speech announcing that she won an award for being a dorm leader. She described how the speech characterized her as good at keeping order and making sure things happen on time. When we then read segments coded to our categories of interest in sequence, we could see how the nature of the text about the two categories was similar. The participant didn’t acknowledge the similarity as would be required if the same text was about both ideas. Instead, she gave us information about each component part of this theme, and we put the pieces together by organizing the information we were reading in a new way.
In Computer Programs for Qualitative Data Analysis (1995) Weitzman and Miles discuss search tools available in packages they classify as ‘theory building packages.’ Searching for combinations of codes is described as a powerful feature of these programs. Tools like the Query Tool of ATLAS.ti, the Search Procedures in ETHNOGRAPH, Text Retrieval in MAXqda and the Search Tool in NVIVO 2.0 quickly locate all instances where more than one code was applied to the same text or in the same document in a pattern designated by the user. You can find every time Code A comes before Code B or Code A is inside Code B, when they are coded to the same text or coded in overlapping ways and more. These powerful tools are a strong argument for qualitative software. Tasks like these searches would be difficult, and in some cases impossible, to replicate off-screen.
While powerful, these tools nonetheless assume that the researcher knows what code combinations to put into a search before he or she performs it. While we may know many combinations for any project, surprise combinations are equally common.
Tools in MAXqda, ATLAS.ti and NVIVO 2.0 let researchers see a matrix of code combinations that provide access to the associated text. The code relations browser from MAXqda is pictured here. Codes in the project (an entire code list or selection of codes designated by the user) are displayed in both columns and rows of a table. Boxes of different color and size indicate the popularity of different combinations of codes. Large red boxes indicate the most popular combinations and small blue boxes indicate less popular combinations. Double clicking on a particular box brings all instances that fit that combination to the retrieved segments window of MAXqda. The combination of education and work is highlighted in the lower right of the window below.
Figure 1. Code relation browser in MAXqda
The motivation behind tools like the code relations browser is to provide direct access to anticipated and unanticipated code combinations in your project. It is also possible to see a list, or in some programs, like MAXqda and NVIVO, a matrix diagram, of all codes applied within the body of data documents.