Data forms (focus on words):
- handwritten or typed field notes;
- audio or video recordings of interviews or other events in the field setting;
- documents or other print/digital artifacts;
- photographs can be part of the data corpus (best analyzed through memoing).
The basic, raw data (scribbled field notes, recordings) must be processed before they are available for analysis. Field notes must be converted into expanded write-ups. Raw field notes may contain private abbreviations. They are also sketchy. A formal write-up usually will add back some of the missing content because the raw field notes stimulate the field-worker to remember things that happened at the time that are not in the notes. A write-up is an intelligible product for anyone. It can be read, edited for accuracy, commented on, coded, and analyzed .
First cycle coding-->second cycle or pattern codes-->the process of deriving even more general themes through jottings and analytic memoing-->assertion and proposition development.
Codes are labels that assign symbolic meaning to the descriptive or inferential information complied during a study. Codes usually are attached to data "chunks" of varying size and can take the form of a straightforward, descriptive label or a more evocative and complex one. Codes are primarily, but not exclusively, used to retrieve and categorize similar data chunks so the researcher can quickly find, pull out, and cluster the segments relating to a particular research question, hypothesis, construct, or theme. Clustering and the display of condensed chunks then set the stage for further analysis and drawing conclusions.
The conceptual frameworks and research questions are the best defense against overload of information. Codes are prompts or triggers for deeper reflection on the data's meanings. Coding is thus a data condensation task that enables you to retrieve the most meaningful material, to assemble chunks of data that go together, and to further condense the bulk into readily analyzable units.
Codes are first assigned to data chunks to detect reoccurring patterns. From these patterns, similar codes are clustered together to create a smaller number of categories or pattern codes. The interrelationships of the categories with each other then are constructed to develop higher level analytic meanings for assertion, proposition, hypothesis, and/or theory development.
There are 3 elemental methods that serve as foundation approaches to coding:
- Descriptive: A descriptive code assigns labels to data summarize in a word or short phrase--most often a noun--the basic topic of a passage of qualitative data. These eventually provide an inventory of topics for indexing and categorizing, which is especially helpful for ethnographies and studies with a wide variety of data forms. More appropriate for social environments than social action.
- In Vivo: uses words or short phrases from the participant's own language in the data record as codes. It is appropriate for virtually all qualitative studies but particularly for beginning qualitative researchers learning how to code data, and studies that prioritize and honor the participant's voice. In Vivo codes are placed in quotation marks to differentiate them from researcher-generated codes.
- Process coding: This coding method uses gerunds ("-ing" words) exclusively to connote observable and conceptual action in the data. Process also imply actions interwinded with the dynamics of time, such as things that emerge, change, occur in particular sequences, or become strategically implemented. Process coding is appropriate for virtually all qualitative studies, but particularly for grounded theory research that extracts participant action/interaction and consequences.
3 affective methods that tap into the more subjective experiences:
- emotion coding: appropriate for studies that explore intrapersonal and interpersonal participant experiences and actions. A participant himself or herself may sometimes label the emotion, and thus, it should be in Vivo coded in quotation marks.
- values coding: data that reflect a participants' values, attitudes, and beliefs, representing his or her perspectives or worldview. A value (V) is the importance we attribute to ourselves, another person, thing, or idea. An attitude (A) is the way we think and feel about oneself, another person, thing, or idea. A belief (B) is part of a system that includes values and attitudes, plus personal knowledge, experiences, opinions, prejudices, morals, and other interpretive perceptions of the social world. Values coding is appropriate for studies that explore cultural values, identity, intrapersonal and interpersonal participant experiences and action in case studies, appreciative inquiry, oral history, and critical ethnography.
- evaluation coding: applies primarily nonquantitative codes onto qualitative data that assign judgments about the merit, worth, or significance of programs or policy. It is appropriate for policy, critical, action, organizational, and evaluation studies, particularly across multiple cases and extended periods of time. A+ symbol before a code tags it as a positive evaluation. Second-order codes that follow primary code and a colon are called subcodes.
One literary and language method, dramaturgical coding, explores human action and interaction through strategic analysis of people's motives.
Dramaturgical coding: This method applies the terms and conventions of character, play script, and production analysis onto qualitative data. For character, these terms include items such as participant objectives (OBJ), conflicts (CON), tacitcs (TAC), attitudes (ATT), emotions (EMO), and subtexts (SUB). Dramaturgical coding is appropriate for exploring intrapersonal and interpersonal participant experiences and actions in case studies, power relationships, and the processes of human motives and agency.
3 exploratory methods, make preliminary or global coding assignments, based on what the researcher deductively assumes may be present in the data before they are analyzed.
- Holistic coding: applies a single code to a large unit of data in the corpus, rather than line-by-line, to capture a sense of the overall contents and the possible unit of analysis before a more detailed coding or categorization process through First or Second cycle methods. Holistic coding is most applicable when the researcher has a general idea as what to investigate in the data.
- Provisional coding: this approach begins with a "start list" of researcher generated codes, based on what preparatory investigation suggests might appear in the data before they are collected and analyzed. This method is appropriate for qualitative studies that build on or corroborate previous research and investigations.
- Hypothesis coding: the application of a researcher-generated, predetermined list of codes onto qualitative data specifically to assess a research-generated hypothesis. The codes are developed from a theory/prediction about what will be found in the data codes are developed from a theory/prediction about what will be found in the data before they have been collected or analyzed. This method is appropriate for hypothesis testing, content analysis, and analytic induction of the qualitative data set, particularly the search for rules, causes, and explanations in the data.
2 procedural methods utilize specific rather than open-ended ways of coding data:
- Protocol coding: according to a pre-established, recommended, standardized, or prescribed system. Protocol coding is appropriate for qualitative studies in disciplines with previously developed and field-tested coding systems.
- Causation coding: extracts attributions or casual beliefs from participant data bout not just how but why particular outcomes came about. The analyst searches for combinations of antecedent and mediating variables that lead toward certain pathways and attempts to map a three-part process as a CODE 1 > CODE 2 > CODE 3 sequence. Causation coding is appropriate for discerning motives, belief systems, worldviews, processes, recent histories, inter-relationships, and the complexity of influences and affects on human actions and phenomena. The + symbol refers to a combination of variables that are mentioned by the participant as connected parts of the causation sequence; the > symbol means "leads to".
4 grammatical methods play a role in the mechanic of coding:
- attribute coding: the notation of basic descriptive information such as the fieldwork setting, participant characteristics or demographics, data format, and other variables of interest for qualitative and some applications of quantitative analysis. This method is particularly appropriate for those with multiple participants and sites, cross-case studies, and studies with a wide variety of data forms.
- magnitude coding: magnitudes consist of supplemental alphanumeric or symbolic codes or subcodes applied to existing coded data a category to indicate their intensity, frequency, direction, presence, or evaluative content. They are most appropriate for mixed methods and qualitative studies in education, social science, and health care disciplines that also support quantitative measures as evidence of outcomes.
- Subcoding: A second-order tag assigned after a primary code to detail or enrich the entry. The method is particularly appropriate for ethnography and content analyses, studies with multiple participants and sites, and studies with a wide variety of data forms.
- Simultaneous coding: this is the application of 2 or more different codes to a single qualitative datum, or the overlapped occurrence of two or more codes applied to sequential units of qualitative data. The method is appropriate when the data's content suggests multiple meanings that necessitate and justify more than one code.
- deductive coding: developing a provisional 'start list' of codes prior to fieldwork (conceptual framework, research questions, hypotheses, problem areas, or key variables that researchers bring to the study).
- inductive coding: codes emerge progressively during data collection; are better grounded empirically and are especially satisfying to the researcher who has uncovered an important local factor.
Whether codes are created and revised early or late is basically less important than whether they have some conceptual and structural unity. Codes should relate to one another in coherent, study-important ways; they should be part of a unified structure.
An operative coding scheme is not a catalog of disjointed descriptors but rather a conceptual web, including larger meanings and their constitutive characteristics. CAQDAS is especially helpful in displaying the structure of coding schemes, either in hierarchical form or in a network.