Thick(er) Coding in ATLAS.ti: Visualizing the Data
B. Jane Scales
In this brief paper, I continue the discussion started in Creating the Links: An Exploration of Element Relations Within ATLAS.ti, which examined some of Bob Dylan’s lyrics. My goal for that first paper was to illustrate the various types of links one can create in ATLAS.ti: simple linking, which connects a quotation to a code; and code-to-code linking. “Creating the Links” illustrated the primary cycle of coding – the initial work completed before proceeding to secondary coding for more in-depth or thematic codes.
In Creating the Links, I concluded that my initial coding was not rigorous enough to support a second cycle of coding. I wanted to examine the meaning behind Dylan’s use of “dream” – what he meant when he used the term, what part of speech Dylan favored, as well as identify usage patterns within the lyrics. Before I could do that, however, I needed to review the project and organize my primary cycle codes better.
Organizing the Primary Cycle Codes
Many resources on qualitative research discuss coding in terms of “primary” or “first cycle” and “second cycle.” For example, Saldaña explains, “First Cycle methods are those processes that happen during the initial coding of data…” He then continues to list a set code types typical to the first or primary cycle, several of which are relevant to the Dylan Dream project.
For example, Saldaña describes the “Elemental Method” which includes structural and descriptive coding – both of which are present in the Dylan Dream project. I used structural coding early in the project to demarcate the larger segments of data. Any set of lyrics which contained the word dream was coded. For example, the song “Idiot Wind” includes only one use of the word “dreamed,” yet I created a quotation that included the nearly 60 lines of the song, and coded it “Idiot Wind.” In the end, I had coded a total of 69 song lyrics, which formed the “Dream Lyric Family.” I also used elemental coding to label the album titles from which each song lyric appeared.
Descriptive coding, also called “topic coding,” assigns labels to passages to summarize them in some way. Examples of descriptive coding can be seen in both the part of speech codes, and form of the word dream codes assigned to a line or two of relevant lyrics. See Figure 2.
This process of fine-tuning my primary cycle codes resulted in four code families:
1) Albums – The titles of 25 albums which contain lyrics with the word “dream.”
2) Dream Lyrics – Sixty-nine song titles linked to the text of their lyrics.
3) Dream Uses – Seven variants of the word “dream,” used by Dylan.
4) Parts of Speech – Three codes: noun, verb, and adjective.
Transition between Primary and Secondary Cycle Coding
Satisfied that my primary cycle codes were a good foundation to support further probing of the lyrics, I began the next phase of coding. My intent was to assign an Oxford English Dictionary (OED) definition to each instance of “dream” in Dylan’s lyrics. Coding these definitions into the project facilitated another level of analysis.
Secondary cycle coding, according to Tracy and Saldana, entails organizing the primary cycle codes and developing “focused codes, which adds, as Saldana writes, ”…(a) categorical, thematic, conceptual,…or theoretical organization” to the primary cycle codes.” Assigning one of fourteen OED definition codes, or an “ambiguous” code brought a level of conceptual organization to the project. For Richards, this exercise of “analytic” coding was one that came “…from interpretation and reflection on meaning…” that is, “…key to opening up meaning and creating conceptual categories.”
Coding the Oxford English Dictionary Definitions
Using definitions from the OED, I created fourteen definitional codes – seven nouns and eleven verbs. See Figures 3 and 4.
I expanded quotation sizes beyond just the “dream word” (Figure 5) to include the context in which Dylan used the term (Figure 6). I coded sixty-nine quotations with one of fourteen OED definitions, or, if the meaning of the text was not clear, with an ‘ambiguous’ code. Each quotation usually consisted of two lines of lyrics, enough to lend meaning to the dream word.
The Purpose of ‘Exhaustive Coding’
When researchers code data, they identify chunks of text or information which can be examined and written about. In a Grounded Theory approach, the more complete or exhaustive the coding, the better one is able to discover relationships between those information chunks. One can see evidence of thicker coding in ATLAS.ti by using its discovery, query, and visualization tools.
In its current state, the Dylan dream project, coding may not necessarily be described as “exhaustive,” however; the coding is thicker and more developed than its initial state, thereby allowing a closer reading of meaning and patterns within the text. The overlapping and layering of codes visible within Figures 5 and 6 facilitates this, making code relations more apparent.
Analyzing the Code Use and Relationships within the Project
It is beyond the scope of this short summary paper to explore the nuances of dreams in Dylan’s lyrics. However, I will outline some of the methods one may use to begin the analysis.
Groundedness and density are code qualities that show the researcher how often they are used, and how they relate to one another and to other elements within an ATLAS.ti project. In the Dylan project, the part-of-speech code “Noun” is the most grounded, as it’s linked to 62 quotations – more than any other code. The Noun code also has one of the higher densities, as it is linked to five forms of the word dream. See Figure 7 for a network view of this code’s density.
By comparing the groundedness and density of other codes in the project, I see that Dylan overwhelmingly used the word dream to express a set of “ideals or aspirations” within his lyrics, because that definitional code is more grounded (27) than any other of the OED definitions.
Using ATLAS.ti’s query tool, I am able to discover how many more times Dylan used the plural “dreams” as a noun (thirty-six times) versus using the word as a verb (1). Using ATLAS.ti, the researcher can execute granular queries on data. See Figure 8.
Figure 9 shows yet another way of visualizing code relationships – with a co-occurrence table. The c-coefficient (a number between 0 and 1) shows the relationship strength between the three parts-of-speech codes and the dream words. The higher the c-coefficient, the greater the relationship strength.
Corpus-linguistic Look at Dylan and Dreams
In this short review, I used a content analysis, corpus-linguistic approach to examine how Dylan uses dream imagery, the word “dream” or its variants within his lyrics. I had not previously worked with corpus texts, or methodology as described in Corpus Analysis and Linguistic Theory by Meyer. For that reason, my analysis was more intuitively conducted, based on an interest of Dylan, but also drawing from passive knowledge about semantics, and grammatical constructions.
Only after I had completed much of the work on the Dylan Dream project did I look at books and articles that describe how corpus linguistic scholars work with literary texts. A few studies of this sort have been written on Dylan – a short corpus linguistic analysis by Daniel Schmidtke published as a blog entry, and an article appearing in the journal Oral Tradition entitled, “A Semantic and Syntactic Journey Through the Dylan Corpus.”
In summary, developing a clear system of coding and carefully linked to the appropriate quotations in the PD was time consuming, but ultimately rewarding. Thicker coding of quotations allowed a closer look at the themes and patterns within the PD. By first applying elemental and topical coding, I was able to easily delve into a more thematic analysis and look at the meaning behind the text. ATLAS.ti made this much easier by providing a multitude of ways to visualize these relationships.
Meyer, Charles F. (2002). Corpus Analysis and Linguistic Theory (2002). Cambridge University Press: Port Chester, NY.
Khalifa, Jean-Charles. (2007). “A Semantic and Syntactic Journey Through the Dylan Corpus.” Oral Tradition. 21(1). pp. 162-174.
Richards, Marilyn G. and Janice M. Morse. (2012). README FIRST for a User’s Guide to Qualitative Methods. Sage: Thousand Oaks.
Saldaña, Johnny. (2013). The Coding Manual for Qualitative Researchers. Sage: Thousand Oaks. pp. 58-244.
Mishler, E. (1979). “Meaning in Context: Is There Any Other Kind?” Harvard Educational Review. 49(1): 1-19.
Konopásek, Zdeněk. (2008). “Making Thinking Visible with Atlas.ti: Computer Assisted Qualitative Analysis as Textual Practices.” Forum: Qualitative Social Research. 9(2).
About the Author
B. Jane Scales is the Reference Team Leader, and E-Projects Librarian at the Washington State University Libraries. She holds a bachelor’s degree in Russian Language from Indiana University, a master’s in German Language and Literature from Ohio State University, and a master’s in information science (MLIS) from the University of Kentucky. Her research focus includes information literacy, online learning theories, and academic reference services.